From: Tero Kristo <t-kristo@ti.com>
To: Guenter Roeck <linux@roeck-us.net>, <wim@linux-watchdog.org>,
<linux-watchdog@vger.kernel.org>
Cc: <linux-kernel@vger.kernel.org>, <jan.kiszka@siemens.com>
Subject: Re: [PATCHv3 3/4] watchdog: rti-wdt: attach to running watchdog during probe
Date: Tue, 14 Jul 2020 09:50:22 +0300 [thread overview]
Message-ID: <2a92083c-f8bd-7a42-25e2-ee6ae354a4de@ti.com> (raw)
In-Reply-To: <0c69734d-b301-7762-7614-0644e2a719ba@roeck-us.net>
On 14/07/2020 08:15, Guenter Roeck wrote:
> On 7/13/20 6:18 AM, Tero Kristo wrote:
>> If the RTI watchdog is running already during probe, the driver must
>> configure itself to match the HW. Window size and timeout is probed from
>> hardware, and the last keepalive ping is adjusted to match it also.
>>
>> Signed-off-by: Tero Kristo <t-kristo@ti.com>
>> ---
>> drivers/watchdog/rti_wdt.c | 111 +++++++++++++++++++++++++++++++++----
>> 1 file changed, 101 insertions(+), 10 deletions(-)
>>
>> diff --git a/drivers/watchdog/rti_wdt.c b/drivers/watchdog/rti_wdt.c
>> index d456dd72d99a..45dfc546e371 100644
>> --- a/drivers/watchdog/rti_wdt.c
>> +++ b/drivers/watchdog/rti_wdt.c
>> @@ -35,7 +35,11 @@
>>
>> #define RTIWWDRX_NMI 0xa
>>
>> -#define RTIWWDSIZE_50P 0x50
>> +#define RTIWWDSIZE_50P 0x50
>> +#define RTIWWDSIZE_25P 0x500
>> +#define RTIWWDSIZE_12P5 0x5000
>> +#define RTIWWDSIZE_6P25 0x50000
>> +#define RTIWWDSIZE_3P125 0x500000
>>
>> #define WDENABLE_KEY 0xa98559da
>>
>> @@ -48,7 +52,7 @@
>>
>> #define DWDST BIT(1)
>>
>> -static int heartbeat;
>> +static int heartbeat = DEFAULT_HEARTBEAT;
>>
>> /*
>> * struct to hold data for each WDT device
>> @@ -79,11 +83,9 @@ static int rti_wdt_start(struct watchdog_device *wdd)
>> * be petted during the open window; not too early or not too late.
>> * The HW configuration options only allow for the open window size
>> * to be 50% or less than that; we obviouly want to configure the open
>> - * window as large as possible so we select the 50% option. To avoid
>> - * any glitches, we accommodate 5% safety margin also, so we setup
>> - * the min_hw_hearbeat at 55% of the timeout period.
>> + * window as large as possible so we select the 50% option.
>> */
>> - wdd->min_hw_heartbeat_ms = 11 * wdd->timeout * 1000 / 20;
>> + wdd->min_hw_heartbeat_ms = 500 * wdd->timeout;
>>
>> /* Generate NMI when wdt expires */
>> writel_relaxed(RTIWWDRX_NMI, wdt->base + RTIWWDRXCTRL);
>> @@ -110,7 +112,48 @@ static int rti_wdt_ping(struct watchdog_device *wdd)
>> return 0;
>> }
>>
>> -static unsigned int rti_wdt_get_timeleft(struct watchdog_device *wdd)
>> +static int rti_wdt_setup_hw_hb(struct watchdog_device *wdd, u32 wsize)
>> +{
>> + /*
>> + * RTI only supports a windowed mode, where the watchdog can only
>> + * be petted during the open window; not too early or not too late.
>> + * The HW configuration options only allow for the open window size
>> + * to be 50% or less than that.
>> + */
>> + switch (wsize) {
>> + case RTIWWDSIZE_50P:
>> + /* 50% open window => 50% min heartbeat */
>> + wdd->min_hw_heartbeat_ms = 500 * heartbeat;
>> + break;
>> +
>> + case RTIWWDSIZE_25P:
>> + /* 25% open window => 75% min heartbeat */
>> + wdd->min_hw_heartbeat_ms = 750 * heartbeat;
>> + break;
>> +
>> + case RTIWWDSIZE_12P5:
>> + /* 12.5% open window => 87.5% min heartbeat */
>> + wdd->min_hw_heartbeat_ms = 875 * heartbeat;
>> + break;
>> +
>> + case RTIWWDSIZE_6P25:
>> + /* 6.5% open window => 93.5% min heartbeat */
>> + wdd->min_hw_heartbeat_ms = 935 * heartbeat;
>> + break;
>> +
>> + case RTIWWDSIZE_3P125:
>> + /* 3.125% open window => 96.9% min heartbeat */
>> + wdd->min_hw_heartbeat_ms = 969 * heartbeat;
>> + break;
>> +
>> + default:
>> + return -EINVAL;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static unsigned int rti_wdt_get_timeleft_ms(struct watchdog_device *wdd)
>> {
>> u64 timer_counter;
>> u32 val;
>> @@ -123,11 +166,18 @@ static unsigned int rti_wdt_get_timeleft(struct watchdog_device *wdd)
>>
>> timer_counter = readl_relaxed(wdt->base + RTIDWDCNTR);
>>
>> + timer_counter *= 1000;
>> +
>> do_div(timer_counter, wdt->freq);
>>
>> return timer_counter;
>> }
>>
>> +static unsigned int rti_wdt_get_timeleft(struct watchdog_device *wdd)
>> +{
>> + return rti_wdt_get_timeleft_ms(wdd) / 1000;
>> +}
>> +
>> static const struct watchdog_info rti_wdt_info = {
>> .options = WDIOF_KEEPALIVEPING,
>> .identity = "K3 RTI Watchdog",
>> @@ -148,6 +198,7 @@ static int rti_wdt_probe(struct platform_device *pdev)
>> struct watchdog_device *wdd;
>> struct rti_wdt_device *wdt;
>> struct clk *clk;
>> + u32 last_ping = 0;
>>
>> wdt = devm_kzalloc(dev, sizeof(*wdt), GFP_KERNEL);
>> if (!wdt)
>> @@ -169,6 +220,14 @@ static int rti_wdt_probe(struct platform_device *pdev)
>> return -EINVAL;
>> }
>>
>> + /*
>> + * If watchdog is running at 32k clock, it is not accurate.
>> + * Adjust frequency down in this case so that we don't pet
>> + * the watchdog too often.
>> + */
>> + if (wdt->freq < 32768)
>> + wdt->freq = wdt->freq * 9 / 10;
>> +
>
> So this is now only a problem is the window size was set restrictively
> in the BOS/ROMMON. Meaning it is still a problem. How is this better than
> before ? Why not just always set the window size to something reasonable ?
Re-programming of the watchdog only takes effect at the next ping, then
and only then will it adjust the window size + timeout period.
>
>
>> pm_runtime_enable(dev);
>> ret = pm_runtime_get_sync(dev);
>> if (ret) {
>> @@ -185,11 +244,8 @@ static int rti_wdt_probe(struct platform_device *pdev)
>> wdd->min_timeout = 1;
>> wdd->max_hw_heartbeat_ms = (WDT_PRELOAD_MAX << WDT_PRELOAD_SHIFT) /
>> wdt->freq * 1000;
>> - wdd->timeout = DEFAULT_HEARTBEAT;
>> wdd->parent = dev;
>>
>> - watchdog_init_timeout(wdd, heartbeat, dev);
>> -
>> watchdog_set_drvdata(wdd, wdt);
>> watchdog_set_nowayout(wdd, 1);
>> watchdog_set_restart_priority(wdd, 128);
>> @@ -201,12 +257,47 @@ static int rti_wdt_probe(struct platform_device *pdev)
>> goto err_iomap;
>> }
>>
>> + if (readl(wdt->base + RTIDWDCTRL) == WDENABLE_KEY) {
>> + u32 time_left_ms;
>> + u64 heartbeat_ms;
>> + u32 wsize;
>> +
>> + set_bit(WDOG_HW_RUNNING, &wdd->status);
>> + time_left_ms = rti_wdt_get_timeleft_ms(wdd);
>> + heartbeat_ms = readl(wdt->base + RTIDWDPRLD);
>> + heartbeat_ms <<= WDT_PRELOAD_SHIFT;
>> + heartbeat_ms *= 1000;
>> + heartbeat_ms /= wdt->freq;
>
> Ah yes, the pitfalls of 64-bit divide operations.
Should I convert this to use do_div? With 64bit archs this code is
targeted at it runs just fine.
>
>> + if (heartbeat_ms / 1000 != heartbeat)
>> + dev_warn(dev, "watchdog already running, ignoring heartbeat config!\n");
>> +
>> + heartbeat = heartbeat_ms / 1000;
>> +
>> + wsize = readl(wdt->base + RTIWWDSIZECTRL);
>> + ret = rti_wdt_setup_hw_hb(wdd, wsize);
>> + if (ret) {
>> + dev_err(dev, "bad window size.\n");
>> + goto err_iomap;
>> + }
>> +
>> + last_ping = heartbeat_ms - time_left_ms;
>> + if (time_left_ms > heartbeat_ms) {
>> + dev_warn(dev, "time_left > heartbeat? Assuming last ping just before now.\n");
>> + last_ping = 0;
>> + }
>
> All that complexity makes me wonder if it wouldn't be better to just reprogram
> the watchdog if it is already running. Any reason for not doing that ?
Well, you can re-program it but... It does not take effect until next
window starts, so basically we need to take care of the currently
running window anyways after which re-programming it doesn't make much
sense anymore. And handling the switch from one setup to next would add
extra complexity.
-Tero
>
>> + }
>> +
>> + watchdog_init_timeout(wdd, heartbeat, dev);
>> +
>> ret = watchdog_register_device(wdd);
>> if (ret) {
>> dev_err(dev, "cannot register watchdog device\n");
>> goto err_iomap;
>> }
>>
>> + if (last_ping)
>> + watchdog_set_last_hw_keepalive(wdd, last_ping);
>> +
>> return 0;
>>
>> err_iomap:
>>
>
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
next prev parent reply other threads:[~2020-07-14 6:50 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-07-13 13:18 [PATCHv3 0/4] watchdog: rti: support attach to running timer Tero Kristo
2020-07-13 13:18 ` [PATCHv3 1/4] watchdog: use __watchdog_ping in startup Tero Kristo
2020-07-13 13:18 ` [PATCHv3 2/4] watchdog: add support for adjusting last known HW keepalive time Tero Kristo
2020-07-13 13:18 ` [PATCHv3 3/4] watchdog: rti-wdt: attach to running watchdog during probe Tero Kristo
2020-07-13 20:28 ` kernel test robot
2020-07-13 20:33 ` kernel test robot
2020-07-14 5:15 ` Guenter Roeck
2020-07-14 6:50 ` Tero Kristo [this message]
2020-07-14 7:06 ` Guenter Roeck
2020-07-14 7:13 ` Tero Kristo
2020-07-14 14:42 ` Guenter Roeck
2020-07-13 13:18 ` [PATCHv3 4/4] watchdog: rti-wdt: balance pm runtime enable calls Tero Kristo
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=2a92083c-f8bd-7a42-25e2-ee6ae354a4de@ti.com \
--to=t-kristo@ti.com \
--cc=jan.kiszka@siemens.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-watchdog@vger.kernel.org \
--cc=linux@roeck-us.net \
--cc=wim@linux-watchdog.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®