2024-04-25 09:08:16

by Peter Yin

[permalink] [raw]
Subject: [PATCH v7 0/1] drivers: watchdog: revise watchdog bootstatus

Regarding the AST2600 specification, the WDTn Timeout Status Register
(WDT10) has bit 1 reserved. Bit 1 of the status register indicates
on ast2500 if the boot was from the second boot source.
It does not indicate that the most recent reset was triggered by
the watchdog. The code should just be changed to set WDIOF_CARDRESET
if bit 0 of the status register is set. However, this bit can be clear when
watchdog register 0x0c bit1(Reset System after timeout) is enabled.
Thereforce include SCU register to veriy WDIOF_EXTERN1 and WDIOF_CARDRESET
in ast2600 SCU74 or ast2400/ast2500 SCU3C.

Change Log:

v6 -> v7
- To use syscon_regmap_lookup_by_compatibleys to get scu base
- Power on reset is set when triggered by AC or SRSRST.
Thereforce, we clear flag to ensure next boot cause is a real watchdog case.
We use the external reset flag to determine
if it is an external reset or card reset.

v5 -> v6
- Fixed missing WDT_TIMEOUT_STATUS_EVENT.

v4 -> v5
- Revert indentation.

v3 -> v4
- Add error handling for syscon_regmap_lookup_by_phandle and
regmap_read.

v2 -> v3
- Fixed WDIOF_CARDRESET status bit check and added support
for WDIOF_EXTERN1 on ast2500 and ast2600.

v1 -> v2
- Add comment and support WDIOF_CARDRESET in ast2600

v1
- Patch 0001 - Add WDIOF_EXTERN1 bootstatus
---

drivers/watchdog/aspeed_wdt.c | 109 ++++++++++++++++++++++++++++++++--
1 file changed, 103 insertions(+), 6 deletions(-)

--
2.25.1



2024-04-25 09:08:34

by Peter Yin

[permalink] [raw]
Subject: [PATCH v7 1/1] drivers: watchdog: revise watchdog bootstatus

Regarding the AST2600 specification, the WDTn Timeout Status Register
(WDT10) has bit 1 reserved. Bit 1 of the status register indicates
on ast2500 if the boot was from the second boot source.
It does not indicate that the most recent reset was triggered by
the watchdog. The code should just be changed to set WDIOF_CARDRESET
if bit 0 of the status register is set. However, this bit can be clear when
watchdog register 0x0c bit1(Reset System after timeout) is enabled.
Thereforce include SCU register to veriy WDIOF_EXTERN1 and WDIOF_CARDRESET
in ast2600 SCU74 or ast2400/ast2500 SCU3C.

Signed-off-by: Peter Yin <[email protected]>
---
drivers/watchdog/aspeed_wdt.c | 109 ++++++++++++++++++++++++++++++++--
1 file changed, 103 insertions(+), 6 deletions(-)

diff --git a/drivers/watchdog/aspeed_wdt.c b/drivers/watchdog/aspeed_wdt.c
index b4773a6aaf8c..4c58593658bc 100644
--- a/drivers/watchdog/aspeed_wdt.c
+++ b/drivers/watchdog/aspeed_wdt.c
@@ -11,10 +11,12 @@
#include <linux/io.h>
#include <linux/kernel.h>
#include <linux/kstrtox.h>
+#include <linux/mfd/syscon.h>
#include <linux/module.h>
#include <linux/of.h>
#include <linux/of_irq.h>
#include <linux/platform_device.h>
+#include <linux/regmap.h>
#include <linux/watchdog.h>

static bool nowayout = WATCHDOG_NOWAYOUT;
@@ -82,6 +84,16 @@ MODULE_DEVICE_TABLE(of, aspeed_wdt_of_table);
#define WDT_RESET_MASK1 0x1c
#define WDT_RESET_MASK2 0x20

+//AST SCU Register
+#define AST2400_AST2500_SYSTEM_RESET_EVENT 0x3C
+#define AST2400_WATCHDOG_RESET_FLAG BIT(1)
+#define AST2400_RESET_FLAG_CLEAR GENMASK(2, 0)
+#define AST2500_WATCHDOG_RESET_FLAG GENMASK(4, 2)
+#define AST2600_SYSTEM_RESET_EVENT 0x74
+#define POWERON_RESET_FLAG BIT(0)
+#define EXTERN_RESET_FLAG BIT(1)
+#define AST2600_WATCHDOG_RESET_FLAG GENMASK(31, 16)
+
/*
* WDT_RESET_WIDTH controls the characteristics of the external pulse (if
* enabled), specifically:
@@ -310,6 +322,7 @@ static int aspeed_wdt_probe(struct platform_device *pdev)
const struct of_device_id *ofdid;
struct aspeed_wdt *wdt;
struct device_node *np;
+ struct regmap *scu_base;
const char *reset_type;
u32 duration;
u32 status;
@@ -458,15 +471,99 @@ static int aspeed_wdt_probe(struct platform_device *pdev)
writel(duration - 1, wdt->base + WDT_RESET_WIDTH);
}

- status = readl(wdt->base + WDT_TIMEOUT_STATUS);
- if (status & WDT_TIMEOUT_STATUS_BOOT_SECONDARY) {
- wdt->wdd.bootstatus = WDIOF_CARDRESET;
+ /*
+ * Power on reset is set when triggered by AC or SRSRST.
+ * Thereforce, we clear flag to ensure
+ * next boot cause is a real watchdog case.
+ * We use the external reset flag to determine
+ * if it is an external reset or card reset
+ */
+ if (of_device_is_compatible(np, "aspeed,ast2600-wdt")) {
+ scu_base = syscon_regmap_lookup_by_compatible(
+ "aspeed,ast2600-scu");
+ if (IS_ERR(scu_base))
+ return PTR_ERR(scu_base);
+
+ ret = regmap_read(scu_base,
+ AST2600_SYSTEM_RESET_EVENT,
+ &status);
+ if (ret)
+ return ret;
+
+ if ((status & POWERON_RESET_FLAG) == 0 &&
+ status & AST2600_WATCHDOG_RESET_FLAG) {
+ if(status & EXTERN_RESET_FLAG)
+ wdt->wdd.bootstatus = WDIOF_EXTERN1;
+ else
+ wdt->wdd.bootstatus = WDIOF_CARDRESET;
+ }
+ status = AST2600_WATCHDOG_RESET_FLAG |
+ POWERON_RESET_FLAG |
+ EXTERN_RESET_FLAG;
+
+ ret = regmap_write(scu_base,
+ AST2600_SYSTEM_RESET_EVENT,
+ status);
+ } else if (of_device_is_compatible(np, "aspeed,ast2500-wdt")) {
+ scu_base = syscon_regmap_lookup_by_compatible(
+ "aspeed,ast2500-scu");
+ if (IS_ERR(scu_base))
+ return PTR_ERR(scu_base);
+
+ ret = regmap_read(scu_base,
+ AST2400_AST2500_SYSTEM_RESET_EVENT,
+ &status);
+ if (ret)
+ return ret;
+
+ if ((status & POWERON_RESET_FLAG) == 0 &&
+ status & AST2500_WATCHDOG_RESET_FLAG) {
+ if(status & EXTERN_RESET_FLAG)
+ wdt->wdd.bootstatus = WDIOF_EXTERN1;
+ else
+ wdt->wdd.bootstatus = WDIOF_CARDRESET;
+ }
+
+ status = AST2500_WATCHDOG_RESET_FLAG |
+ POWERON_RESET_FLAG |
+ EXTERN_RESET_FLAG;
+
+ ret = regmap_write(scu_base,
+ AST2400_AST2500_SYSTEM_RESET_EVENT,
+ status);

- if (of_device_is_compatible(np, "aspeed,ast2400-wdt") ||
- of_device_is_compatible(np, "aspeed,ast2500-wdt"))
- wdt->wdd.groups = bswitch_groups;
+ wdt->wdd.groups = bswitch_groups;
+ } else {
+ scu_base = syscon_regmap_lookup_by_compatible(
+ "aspeed,ast2400-scu");
+ if (IS_ERR(scu_base))
+ return PTR_ERR(scu_base);
+
+ ret = regmap_read(scu_base,
+ AST2400_AST2500_SYSTEM_RESET_EVENT,
+ &status);
+ if (ret)
+ return ret;
+ /*
+ * Ast2400 external reset can clear watdog dog rest flag, so
+ * only support WDIOF_CARDRESET
+ */
+ if ((status & POWERON_RESET_FLAG) == 0 &&
+ status & AST2400_WATCHDOG_RESET_FLAG)
+ wdt->wdd.bootstatus = WDIOF_CARDRESET;
+
+ status = AST2400_RESET_FLAG_CLEAR;
+
+ ret = regmap_write(scu_base,
+ AST2400_AST2500_SYSTEM_RESET_EVENT,
+ status);
+
+ wdt->wdd.groups = bswitch_groups;
}

+ if (ret)
+ return ret;
+
dev_set_drvdata(dev, wdt);

return devm_watchdog_register_device(dev, &wdt->wdd);
--
2.25.1


2024-04-26 00:42:21

by Andrew Jeffery

[permalink] [raw]
Subject: Re: [PATCH v7 1/1] drivers: watchdog: revise watchdog bootstatus

On Thu, 2024-04-25 at 17:07 +0800, Peter Yin wrote:
> Regarding the AST2600 specification, the WDTn Timeout Status Register
> (WDT10) has bit 1 reserved. Bit 1 of the status register indicates
> on ast2500 if the boot was from the second boot source.
> It does not indicate that the most recent reset was triggered by
> the watchdog. The code should just be changed to set WDIOF_CARDRESET
> if bit 0 of the status register is set. However, this bit can be clear when
> watchdog register 0x0c bit1(Reset System after timeout) is enabled.
> Thereforce include SCU register to veriy WDIOF_EXTERN1 and WDIOF_CARDRESET
> in ast2600 SCU74 or ast2400/ast2500 SCU3C.
>
> Signed-off-by: Peter Yin <[email protected]>
> ---
> drivers/watchdog/aspeed_wdt.c | 109 ++++++++++++++++++++++++++++++++--
> 1 file changed, 103 insertions(+), 6 deletions(-)

After this patch the probe() implementation is ~250loc with a whole
bunch of conditional behaviours based on the SoC version. Maybe it's
time to break it up into version-specific functions that are called
from the probe() implementation?

Andrew

2024-04-26 14:46:16

by Peter Yin

[permalink] [raw]
Subject: Re: [PATCH v7 1/1] drivers: watchdog: revise watchdog bootstatus

I can include reset condition in struct maybe like this

static const struct aspeed_wdt_config ast2600_config = {
ext_pulse_width_mask = 0xfffff,
irq_shift = 0,
irq_mask = GENMASK(31, 10),
compatible = "aspeed,ast2600-scu",
reset_event = AST2600_SYSTEM_RESET_EVENT,
watchdog_reset_flag = AST2600_WATCHDOG_RESET_FLAG,
extern_reset_flag = EXTERN_RESET_FLAG,
reset_flag_clear = AST2600_RESET_FLAG_CLEAR,
};

in probe( ) we just call

scu_base = syscon_regmap_lookup_by_compatible(wdt->cfg->compatible);
if (IS_ERR(scu_base))
return PTR_ERR(scu_base);

ret = regmap_read(scu_base, wdt->cfg->reset_event, &status);
if (ret)
return ret;

if ((status & POWERON_RESET_FLAG) == 0 &&
status & wdt->cfg->watchdog_reset_flag)
wdt->wdd.bootstatus = (status & wdt->cfg->extern_reset_flag) ?
WDIOF_EXTERN1 : WDIOF_CARDRESET;

status = wdt->cfg->watchdog_reset_flag | POWERON_RESET_FLAG |
wdt->cfg->extern_reset_flag;

ret = regmap_write(scu_base, wdt->cfg->reset_event, status);

Does this meet your expectations?

On Fri, Apr 26, 2024 at 8:42 AM Andrew Jeffery
<[email protected]> wrote:
>
> On Thu, 2024-04-25 at 17:07 +0800, Peter Yin wrote:
> > Regarding the AST2600 specification, the WDTn Timeout Status Register
> > (WDT10) has bit 1 reserved. Bit 1 of the status register indicates
> > on ast2500 if the boot was from the second boot source.
> > It does not indicate that the most recent reset was triggered by
> > the watchdog. The code should just be changed to set WDIOF_CARDRESET
> > if bit 0 of the status register is set. However, this bit can be clear when
> > watchdog register 0x0c bit1(Reset System after timeout) is enabled.
> > Thereforce include SCU register to veriy WDIOF_EXTERN1 and WDIOF_CARDRESET
> > in ast2600 SCU74 or ast2400/ast2500 SCU3C.
> >
> > Signed-off-by: Peter Yin <[email protected]>
> > ---
> > drivers/watchdog/aspeed_wdt.c | 109 ++++++++++++++++++++++++++++++++--
> > 1 file changed, 103 insertions(+), 6 deletions(-)
>
> After this patch the probe() implementation is ~250loc with a whole
> bunch of conditional behaviours based on the SoC version. Maybe it's
> time to break it up into version-specific functions that are called
> from the probe() implementation?
>
> Andrew

2024-04-26 16:24:45

by Guenter Roeck

[permalink] [raw]
Subject: Re: [PATCH v7 1/1] drivers: watchdog: revise watchdog bootstatus

On 4/26/24 07:45, Chia Hsing Yin wrote:
> I can include reset condition in struct maybe like this
>
> static const struct aspeed_wdt_config ast2600_config = {
> .ext_pulse_width_mask = 0xfffff,
> .irq_shift = 0,
> .irq_mask = GENMASK(31, 10),
> .compatible = "aspeed,ast2600-scu",
> .reset_event = AST2600_SYSTEM_RESET_EVENT,
> .watchdog_reset_flag = AST2600_WATCHDOG_RESET_FLAG,
> .extern_reset_flag = EXTERN_RESET_FLAG,
> .reset_flag_clear = AST2600_RESET_FLAG_CLEAR,
> };
>
> in probe( ) we just call
>
> scu_base = syscon_regmap_lookup_by_compatible(wdt->cfg->compatible);
> if (IS_ERR(scu_base))
> return PTR_ERR(scu_base);
>
> ret = regmap_read(scu_base, wdt->cfg->reset_event, &status);
> if (ret)
> return ret;
>
> if ((status & POWERON_RESET_FLAG) == 0 &&

If you do that, please use
if (!(status & POWERON_RESET_FLAG) && ...

> status & wdt->cfg->watchdog_reset_flag)
> wdt->wdd.bootstatus = (status & wdt->cfg->extern_reset_flag) ?
> WDIOF_EXTERN1 : WDIOF_CARDRESET;
>
> status = wdt->cfg->watchdog_reset_flag | POWERON_RESET_FLAG |
> wdt->cfg->extern_reset_flag;
>
> ret = regmap_write(scu_base, wdt->cfg->reset_event, status);
>
> Does this meet your expectations?
>
> On Fri, Apr 26, 2024 at 8:42 AM Andrew Jeffery
> <[email protected]> wrote:
>>
>> On Thu, 2024-04-25 at 17:07 +0800, Peter Yin wrote:
>>> Regarding the AST2600 specification, the WDTn Timeout Status Register
>>> (WDT10) has bit 1 reserved. Bit 1 of the status register indicates
>>> on ast2500 if the boot was from the second boot source.
>>> It does not indicate that the most recent reset was triggered by
>>> the watchdog. The code should just be changed to set WDIOF_CARDRESET
>>> if bit 0 of the status register is set. However, this bit can be clear when
>>> watchdog register 0x0c bit1(Reset System after timeout) is enabled.
>>> Thereforce include SCU register to veriy WDIOF_EXTERN1 and WDIOF_CARDRESET
>>> in ast2600 SCU74 or ast2400/ast2500 SCU3C.
>>>
>>> Signed-off-by: Peter Yin <[email protected]>
>>> ---
>>> drivers/watchdog/aspeed_wdt.c | 109 ++++++++++++++++++++++++++++++++--
>>> 1 file changed, 103 insertions(+), 6 deletions(-)
>>
>> After this patch the probe() implementation is ~250loc with a whole
>> bunch of conditional behaviours based on the SoC version. Maybe it's
>> time to break it up into version-specific functions that are called
>> from the probe() implementation?
>>
>> Andrew