[1/1] Watchdog: OMAP3: fix wrong boot status from wdt reboot
diff mbox

Message ID 1341417241-11563-1-git-send-email-zumeng.chen@windriver.com
State New, archived
Headers show

Commit Message

Zumeng Chen July 4, 2012, 3:54 p.m. UTC
Does the following fix make sense?

WDIOC_GETBOOTSTATUS always return 0 even if the machine
comes from omap-wdt reboot. Because WKUP_MOD is not right
for OMAP3, so give the right addr 0xA00 of PRM_RSTST for
get_reset_sources, which inputs the signal from omap-wdt
reboot, and return 1 when coming from omap-wdt reboot for
WDIOC_GETBOOTSTATUS.

Signed-off-by: Zumeng Chen <zumeng.chen@windriver.com>
---
 arch/arm/mach-omap2/prcm.c  |    4 +++-
 drivers/watchdog/omap_wdt.c |    3 +++
 drivers/watchdog/omap_wdt.h |    3 +++
 3 files changed, 9 insertions(+), 1 deletions(-)

Comments

zumeng.chen July 4, 2012, 3:57 p.m. UTC | #1
To: Shubhrajyoti
On 2012?07?04? 23:54, Zumeng Chen wrote:
> Does the following fix make sense?
>
> WDIOC_GETBOOTSTATUS always return 0 even if the machine
> comes from omap-wdt reboot. Because WKUP_MOD is not right
> for OMAP3, so give the right addr 0xA00 of PRM_RSTST for
> get_reset_sources, which inputs the signal from omap-wdt
> reboot, and return 1 when coming from omap-wdt reboot for
> WDIOC_GETBOOTSTATUS.
>
> Signed-off-by: Zumeng Chen<zumeng.chen@windriver.com>
> ---
>   arch/arm/mach-omap2/prcm.c  |    4 +++-
>   drivers/watchdog/omap_wdt.c |    3 +++
>   drivers/watchdog/omap_wdt.h |    3 +++
>   3 files changed, 9 insertions(+), 1 deletions(-)
>
> diff --git a/arch/arm/mach-omap2/prcm.c b/arch/arm/mach-omap2/prcm.c
> index 480f40a..43f3feb 100644
> --- a/arch/arm/mach-omap2/prcm.c
> +++ b/arch/arm/mach-omap2/prcm.c
> @@ -49,8 +49,10 @@ void __iomem *prcm_mpu_base;
>   u32 omap_prcm_get_reset_sources(void)
>   {
>   	/* XXX This presumably needs modification for 34XX */
> -	if (cpu_is_omap24xx() || cpu_is_omap34xx())
> +	if (cpu_is_omap24xx())
>   		return omap2_prm_read_mod_reg(WKUP_MOD, OMAP2_RM_RSTST)&  0x7f;
> +	if (cpu_is_omap34xx())
> +		return omap2_prm_read_mod_reg(0xA00, OMAP2_RM_RSTST)&  0x7f;
>   	if (cpu_is_omap44xx())
>   		return omap2_prm_read_mod_reg(WKUP_MOD, OMAP4_RM_RSTST)&  0x7f;
>
> diff --git a/drivers/watchdog/omap_wdt.c b/drivers/watchdog/omap_wdt.c
> index 8285d65..ea57078 100644
> --- a/drivers/watchdog/omap_wdt.c
> +++ b/drivers/watchdog/omap_wdt.c
> @@ -234,6 +234,9 @@ static long omap_wdt_ioctl(struct file *file, unsigned int cmd,
>   		if (cpu_is_omap24xx())
>   			return put_user(omap_prcm_get_reset_sources(),
>   					(int __user *)arg);
> +		if (cpu_is_omap34xx())
> +			return put_user(omap_prcm_get_reset_sources()&  0x10>>
> +				OMAP3_PRM_RSTST_BIT, (int __user *)arg);
>   		return put_user(0, (int __user *)arg);
>   	case WDIOC_KEEPALIVE:
>   		pm_runtime_get_sync(wdev->dev);
> diff --git a/drivers/watchdog/omap_wdt.h b/drivers/watchdog/omap_wdt.h
> index 09b774c..d8d5daa 100644
> --- a/drivers/watchdog/omap_wdt.h
> +++ b/drivers/watchdog/omap_wdt.h
> @@ -40,6 +40,9 @@
>   #define OMAP_WATCHDOG_WPS		(0x34)
>   #define OMAP_WATCHDOG_SPR		(0x48)
>
> +/* PRM_RSTST MPU_WD_RST bit */
> +#define OMAP3_PRM_RSTST_BIT		4
> +
>   /* Using the prescaler, the OMAP watchdog could go for many
>    * months before firing.  These limits work without scaling,
>    * with the 60 second default assumed by most tools and docs.
Shubhrajyoti Datta July 5, 2012, 5:39 a.m. UTC | #2
Hello,
On Wednesday 04 July 2012 09:27 PM, zumeng.chen wrote:
> To: Shubhrajyoti
> On 2012?07?04? 23:54, Zumeng Chen wrote:
>> Does the following fix make sense?
yes , thanks for the patch.
IIRC Rajendra had a similar one.
Some comments below.
>>
>> WDIOC_GETBOOTSTATUS always return 0 even if the machine
>> comes from omap-wdt reboot. Because WKUP_MOD is not right
>> for OMAP3,so give the right addr 0xA00 of PRM_RSTST for
>> get_reset_sources, which inputs the signal from omap-wdt
>> reboot, and return 1 when coming from omap-wdt reboot for
>> WDIOC_GETBOOTSTATUS.
>>
>> Signed-off-by: Zumeng Chen<zumeng.chen@windriver.com>
>> ---
>>   arch/arm/mach-omap2/prcm.c  |    4 +++-
>>   drivers/watchdog/omap_wdt.c |    3 +++
>>   drivers/watchdog/omap_wdt.h |    3 +++
>>   3 files changed, 9 insertions(+), 1 deletions(-)
>>
>> diff --git a/arch/arm/mach-omap2/prcm.c b/arch/arm/mach-omap2/prcm.c
>> index 480f40a..43f3feb 100644
>> --- a/arch/arm/mach-omap2/prcm.c
>> +++ b/arch/arm/mach-omap2/prcm.c
>> @@ -49,8 +49,10 @@ void __iomem *prcm_mpu_base;
>>   u32 omap_prcm_get_reset_sources(void)
>>   {
>>       /* XXX This presumably needs modification for 34XX */
>> -    if (cpu_is_omap24xx() || cpu_is_omap34xx())
>> +    if (cpu_is_omap24xx())
>>           return omap2_prm_read_mod_reg(WKUP_MOD, OMAP2_RM_RSTST)& 
>> 0x7f;
>> +    if (cpu_is_omap34xx())
>> +        return omap2_prm_read_mod_reg(0xA00, OMAP2_RM_RSTST)&  0x7f;
>>       if (cpu_is_omap44xx())
>>           return omap2_prm_read_mod_reg(WKUP_MOD, OMAP4_RM_RSTST)& 
>> 0x7f;
>>
>> diff --git a/drivers/watchdog/omap_wdt.c b/drivers/watchdog/omap_wdt.c
>> index 8285d65..ea57078 100644
>> --- a/drivers/watchdog/omap_wdt.c
>> +++ b/drivers/watchdog/omap_wdt.c
>> @@ -234,6 +234,9 @@ static long omap_wdt_ioctl(struct file *file,
>> unsigned int cmd,
>>           if (cpu_is_omap24xx())
>>               return put_user(omap_prcm_get_reset_sources(),
>>                       (int __user *)arg);
>> +        if (cpu_is_omap34xx())
>> +            return put_user(omap_prcm_get_reset_sources()&  0x10>>
>> +                OMAP3_PRM_RSTST_BIT, (int __user *)arg);
Actually instead of returning yes/no.
The correct  expectation is to return WDIOF_* flags as defined in
include/linux/watchdog.h.

(BTW I agree that was something even current code is not
following).Since you are at it may be that’s
something you can consider.

>>           return put_user(0, (int __user *)arg);
>>       case WDIOC_KEEPALIVE:
>>           pm_runtime_get_sync(wdev->dev);
>> diff --git a/drivers/watchdog/omap_wdt.h b/drivers/watchdog/omap_wdt.h
>> index 09b774c..d8d5daa 100644
>> --- a/drivers/watchdog/omap_wdt.h
>> +++ b/drivers/watchdog/omap_wdt.h
>> @@ -40,6 +40,9 @@
>>   #define OMAP_WATCHDOG_WPS        (0x34)
>>   #define OMAP_WATCHDOG_SPR        (0x48)
>>
>> +/* PRM_RSTST MPU_WD_RST bit */
>> +#define OMAP3_PRM_RSTST_BIT        4
>> +
>>   /* Using the prescaler, the OMAP watchdog could go for many
>>    * months before firing.  These limits work without scaling,
>>    * with the 60 second default assumed by most tools and docs.
>
Zumeng Chen July 5, 2012, 6 a.m. UTC | #3
? 2012?07?05? 13:39, Shubhrajyoti ??:
> Hello,
> On Wednesday 04 July 2012 09:27 PM, zumeng.chen wrote:
>> To: Shubhrajyoti
>> On 2012?07?04? 23:54, Zumeng Chen wrote:
>>> Does the following fix make sense?
> yes , thanks for the patch.
> IIRC Rajendra had a similar one.
> Some comments below.
>>> WDIOC_GETBOOTSTATUS always return 0 even if the machine
>>> comes from omap-wdt reboot. Because WKUP_MOD is not right
>>> for OMAP3,so give the right addr 0xA00 of PRM_RSTST for
>>> get_reset_sources, which inputs the signal from omap-wdt
>>> reboot, and return 1 when coming from omap-wdt reboot for
>>> WDIOC_GETBOOTSTATUS.
>>>
>>> Signed-off-by: Zumeng Chen<zumeng.chen@windriver.com>
>>> ---
>>>    arch/arm/mach-omap2/prcm.c  |    4 +++-
>>>    drivers/watchdog/omap_wdt.c |    3 +++
>>>    drivers/watchdog/omap_wdt.h |    3 +++
>>>    3 files changed, 9 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/arch/arm/mach-omap2/prcm.c b/arch/arm/mach-omap2/prcm.c
>>> index 480f40a..43f3feb 100644
>>> --- a/arch/arm/mach-omap2/prcm.c
>>> +++ b/arch/arm/mach-omap2/prcm.c
>>> @@ -49,8 +49,10 @@ void __iomem *prcm_mpu_base;
>>>    u32 omap_prcm_get_reset_sources(void)
>>>    {
>>>        /* XXX This presumably needs modification for 34XX */
>>> -    if (cpu_is_omap24xx() || cpu_is_omap34xx())
>>> +    if (cpu_is_omap24xx())
>>>            return omap2_prm_read_mod_reg(WKUP_MOD, OMAP2_RM_RSTST)&
>>> 0x7f;
>>> +    if (cpu_is_omap34xx())
>>> +        return omap2_prm_read_mod_reg(0xA00, OMAP2_RM_RSTST)&   0x7f;
>>>        if (cpu_is_omap44xx())
>>>            return omap2_prm_read_mod_reg(WKUP_MOD, OMAP4_RM_RSTST)&
>>> 0x7f;
>>>
>>> diff --git a/drivers/watchdog/omap_wdt.c b/drivers/watchdog/omap_wdt.c
>>> index 8285d65..ea57078 100644
>>> --- a/drivers/watchdog/omap_wdt.c
>>> +++ b/drivers/watchdog/omap_wdt.c
>>> @@ -234,6 +234,9 @@ static long omap_wdt_ioctl(struct file *file,
>>> unsigned int cmd,
>>>            if (cpu_is_omap24xx())
>>>                return put_user(omap_prcm_get_reset_sources(),
>>>                        (int __user *)arg);
>>> +        if (cpu_is_omap34xx())
>>> +            return put_user(omap_prcm_get_reset_sources()&   0x10>>
>>> +                OMAP3_PRM_RSTST_BIT, (int __user *)arg);
> Actually instead of returning yes/no.
> The correct  expectation is to return WDIOF_* flags as defined in
> include/linux/watchdog.h.
Yes, return WDIOF_CARDRESET is more general, thanks.
>
> (BTW I agree that was something even current code is not
> following).Since you are at it may be that’s
> something you can consider.
Yes, I just find them, I'll try it to get it more following, like 
get_status.

Regards,
Zumeng
>
>>>            return put_user(0, (int __user *)arg);
>>>        case WDIOC_KEEPALIVE:
>>>            pm_runtime_get_sync(wdev->dev);
>>> diff --git a/drivers/watchdog/omap_wdt.h b/drivers/watchdog/omap_wdt.h
>>> index 09b774c..d8d5daa 100644
>>> --- a/drivers/watchdog/omap_wdt.h
>>> +++ b/drivers/watchdog/omap_wdt.h
>>> @@ -40,6 +40,9 @@
>>>    #define OMAP_WATCHDOG_WPS        (0x34)
>>>    #define OMAP_WATCHDOG_SPR        (0x48)
>>>
>>> +/* PRM_RSTST MPU_WD_RST bit */
>>> +#define OMAP3_PRM_RSTST_BIT        4
>>> +
>>>    /* Using the prescaler, the OMAP watchdog could go for many
>>>     * months before firing.  These limits work without scaling,
>>>     * with the 60 second default assumed by most tools and docs.
Vaibhav Bedia July 5, 2012, 1:05 p.m. UTC | #4
Hello,

On Wed, Jul 04, 2012 at 21:24:01, Zumeng Chen wrote:
> Does the following fix make sense?
> 
> WDIOC_GETBOOTSTATUS always return 0 even if the machine
> comes from omap-wdt reboot. Because WKUP_MOD is not right
> for OMAP3, so give the right addr 0xA00 of PRM_RSTST for
> get_reset_sources, which inputs the signal from omap-wdt
> reboot, and return 1 when coming from omap-wdt reboot for
> WDIOC_GETBOOTSTATUS.
> 
> Signed-off-by: Zumeng Chen <zumeng.chen@windriver.com>
> ---
>  arch/arm/mach-omap2/prcm.c  |    4 +++-
>  drivers/watchdog/omap_wdt.c |    3 +++
>  drivers/watchdog/omap_wdt.h |    3 +++
>  3 files changed, 9 insertions(+), 1 deletions(-)
> 
> diff --git a/arch/arm/mach-omap2/prcm.c b/arch/arm/mach-omap2/prcm.c
> index 480f40a..43f3feb 100644
> --- a/arch/arm/mach-omap2/prcm.c
> +++ b/arch/arm/mach-omap2/prcm.c
> @@ -49,8 +49,10 @@ void __iomem *prcm_mpu_base;
>  u32 omap_prcm_get_reset_sources(void)
>  {
>  	/* XXX This presumably needs modification for 34XX */
> -	if (cpu_is_omap24xx() || cpu_is_omap34xx())
> +	if (cpu_is_omap24xx())
>  		return omap2_prm_read_mod_reg(WKUP_MOD, OMAP2_RM_RSTST) & 0x7f;
> +	if (cpu_is_omap34xx())
> +		return omap2_prm_read_mod_reg(0xA00, OMAP2_RM_RSTST) & 0x7f;
>  	if (cpu_is_omap44xx())
>  		return omap2_prm_read_mod_reg(WKUP_MOD, OMAP4_RM_RSTST) & 0x7f;

Instead of adding more cpu_is_* checks maybe you could switch to a
function pointers based approach here?

>  
> diff --git a/drivers/watchdog/omap_wdt.c b/drivers/watchdog/omap_wdt.c
> index 8285d65..ea57078 100644
> --- a/drivers/watchdog/omap_wdt.c
> +++ b/drivers/watchdog/omap_wdt.c
> @@ -234,6 +234,9 @@ static long omap_wdt_ioctl(struct file *file, unsigned int cmd,
>  		if (cpu_is_omap24xx())
>  			return put_user(omap_prcm_get_reset_sources(),
>  					(int __user *)arg);
> +		if (cpu_is_omap34xx())
> +			return put_user(omap_prcm_get_reset_sources() & 0x10 >>
> +				OMAP3_PRM_RSTST_BIT, (int __user *)arg);
>  		return put_user(0, (int __user *)arg);

Usage of PRCM bit masks in the driver looks wrong. Why not
introduce an API like omap_prcm_check_reset_reason() which
returns true or false based on the reset reason being checked?

In case of WDT, the driver can then pass the right flag to
userspace.

Regards,
Vaibhav B.

>  	case WDIOC_KEEPALIVE:
>  		pm_runtime_get_sync(wdev->dev);
> diff --git a/drivers/watchdog/omap_wdt.h b/drivers/watchdog/omap_wdt.h
> index 09b774c..d8d5daa 100644
> --- a/drivers/watchdog/omap_wdt.h
> +++ b/drivers/watchdog/omap_wdt.h
> @@ -40,6 +40,9 @@
>  #define OMAP_WATCHDOG_WPS		(0x34)
>  #define OMAP_WATCHDOG_SPR		(0x48)
>  
> +/* PRM_RSTST MPU_WD_RST bit */
> +#define OMAP3_PRM_RSTST_BIT		4
> +
>  /* Using the prescaler, the OMAP watchdog could go for many
>   * months before firing.  These limits work without scaling,
>   * with the 60 second default assumed by most tools and docs.
> -- 
> 1.7.5.4
> 
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>
zumeng.chen July 5, 2012, 3:03 p.m. UTC | #5
On 2012?07?05? 21:05, Bedia, Vaibhav wrote:
> Hello,
>
> On Wed, Jul 04, 2012 at 21:24:01, Zumeng Chen wrote:
>> Does the following fix make sense?
>>
>> WDIOC_GETBOOTSTATUS always return 0 even if the machine
>> comes from omap-wdt reboot. Because WKUP_MOD is not right
>> for OMAP3, so give the right addr 0xA00 of PRM_RSTST for
>> get_reset_sources, which inputs the signal from omap-wdt
>> reboot, and return 1 when coming from omap-wdt reboot for
>> WDIOC_GETBOOTSTATUS.
>>
>> Signed-off-by: Zumeng Chen<zumeng.chen@windriver.com>
>> ---
>>   arch/arm/mach-omap2/prcm.c  |    4 +++-
>>   drivers/watchdog/omap_wdt.c |    3 +++
>>   drivers/watchdog/omap_wdt.h |    3 +++
>>   3 files changed, 9 insertions(+), 1 deletions(-)
>>
>> diff --git a/arch/arm/mach-omap2/prcm.c b/arch/arm/mach-omap2/prcm.c
>> index 480f40a..43f3feb 100644
>> --- a/arch/arm/mach-omap2/prcm.c
>> +++ b/arch/arm/mach-omap2/prcm.c
>> @@ -49,8 +49,10 @@ void __iomem *prcm_mpu_base;
>>   u32 omap_prcm_get_reset_sources(void)
>>   {
>>   	/* XXX This presumably needs modification for 34XX */
>> -	if (cpu_is_omap24xx() || cpu_is_omap34xx())
>> +	if (cpu_is_omap24xx())
>>   		return omap2_prm_read_mod_reg(WKUP_MOD, OMAP2_RM_RSTST)&  0x7f;
>> +	if (cpu_is_omap34xx())
>> +		return omap2_prm_read_mod_reg(0xA00, OMAP2_RM_RSTST)&  0x7f;
>>   	if (cpu_is_omap44xx())
>>   		return omap2_prm_read_mod_reg(WKUP_MOD, OMAP4_RM_RSTST)&  0x7f;
> Instead of adding more cpu_is_* checks maybe you could switch to a
> function pointers based approach here?
I don't see any more checks VS before like ( cpu_is_omap24xx() || 
cpu_is_omap34xx())

Actually what we want is just to read a register with different offset
responding to the different SOC.
>>
>> diff --git a/drivers/watchdog/omap_wdt.c b/drivers/watchdog/omap_wdt.c
>> index 8285d65..ea57078 100644
>> --- a/drivers/watchdog/omap_wdt.c
>> +++ b/drivers/watchdog/omap_wdt.c
>> @@ -234,6 +234,9 @@ static long omap_wdt_ioctl(struct file *file, unsigned int cmd,
>>   		if (cpu_is_omap24xx())
>>   			return put_user(omap_prcm_get_reset_sources(),
>>   					(int __user *)arg);
>> +		if (cpu_is_omap34xx())
>> +			return put_user(omap_prcm_get_reset_sources()&  0x10>>
>> +				OMAP3_PRM_RSTST_BIT, (int __user *)arg);
>>   		return put_user(0, (int __user *)arg);
> Usage of PRCM bit masks in the driver looks wrong. Why not
Maybe the not proper definition causes "looks wrong".
It should be MPU_WD_RST_BIT, so you see, it is about
watchdog bit.

Anyway, I'll try to send V2 patch with Hubhrajyoti's and your
comments

Regards,
Zumeng
> introduce an API like omap_prcm_check_reset_reason() which
> returns true or false based on the reset reason being checked?
>
> In case of WDT, the driver can then pass the right flag to
> userspace.
>
> Regards,
> Vaibhav B.
>
>>   	case WDIOC_KEEPALIVE:
>>   		pm_runtime_get_sync(wdev->dev);
>> diff --git a/drivers/watchdog/omap_wdt.h b/drivers/watchdog/omap_wdt.h
>> index 09b774c..d8d5daa 100644
>> --- a/drivers/watchdog/omap_wdt.h
>> +++ b/drivers/watchdog/omap_wdt.h
>> @@ -40,6 +40,9 @@
>>   #define OMAP_WATCHDOG_WPS		(0x34)
>>   #define OMAP_WATCHDOG_SPR		(0x48)
>>
>> +/* PRM_RSTST MPU_WD_RST bit */
>> +#define OMAP3_PRM_RSTST_BIT		4
>> +
>>   /* Using the prescaler, the OMAP watchdog could go for many
>>    * months before firing.  These limits work without scaling,
>>    * with the 60 second default assumed by most tools and docs.
>> -- 
>> 1.7.5.4
>>
>>
>> _______________________________________________
>> linux-arm-kernel mailing list
>> linux-arm-kernel@lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>>

Patch
diff mbox

diff --git a/arch/arm/mach-omap2/prcm.c b/arch/arm/mach-omap2/prcm.c
index 480f40a..43f3feb 100644
--- a/arch/arm/mach-omap2/prcm.c
+++ b/arch/arm/mach-omap2/prcm.c
@@ -49,8 +49,10 @@  void __iomem *prcm_mpu_base;
 u32 omap_prcm_get_reset_sources(void)
 {
 	/* XXX This presumably needs modification for 34XX */
-	if (cpu_is_omap24xx() || cpu_is_omap34xx())
+	if (cpu_is_omap24xx())
 		return omap2_prm_read_mod_reg(WKUP_MOD, OMAP2_RM_RSTST) & 0x7f;
+	if (cpu_is_omap34xx())
+		return omap2_prm_read_mod_reg(0xA00, OMAP2_RM_RSTST) & 0x7f;
 	if (cpu_is_omap44xx())
 		return omap2_prm_read_mod_reg(WKUP_MOD, OMAP4_RM_RSTST) & 0x7f;
 
diff --git a/drivers/watchdog/omap_wdt.c b/drivers/watchdog/omap_wdt.c
index 8285d65..ea57078 100644
--- a/drivers/watchdog/omap_wdt.c
+++ b/drivers/watchdog/omap_wdt.c
@@ -234,6 +234,9 @@  static long omap_wdt_ioctl(struct file *file, unsigned int cmd,
 		if (cpu_is_omap24xx())
 			return put_user(omap_prcm_get_reset_sources(),
 					(int __user *)arg);
+		if (cpu_is_omap34xx())
+			return put_user(omap_prcm_get_reset_sources() & 0x10 >>
+				OMAP3_PRM_RSTST_BIT, (int __user *)arg);
 		return put_user(0, (int __user *)arg);
 	case WDIOC_KEEPALIVE:
 		pm_runtime_get_sync(wdev->dev);
diff --git a/drivers/watchdog/omap_wdt.h b/drivers/watchdog/omap_wdt.h
index 09b774c..d8d5daa 100644
--- a/drivers/watchdog/omap_wdt.h
+++ b/drivers/watchdog/omap_wdt.h
@@ -40,6 +40,9 @@ 
 #define OMAP_WATCHDOG_WPS		(0x34)
 #define OMAP_WATCHDOG_SPR		(0x48)
 
+/* PRM_RSTST MPU_WD_RST bit */
+#define OMAP3_PRM_RSTST_BIT		4
+
 /* Using the prescaler, the OMAP watchdog could go for many
  * months before firing.  These limits work without scaling,
  * with the 60 second default assumed by most tools and docs.