Message-ID: <41dc5cc8-4ce3-62ee-132f-e8117190b850@microchip.com>
Date:   Wed, 2 Oct 2019 11:07:58 +0000
From:   <Eugen.Hristev@...rochip.com>
To:     <alexandre.belloni@...tlin.com>
CC:     <wim@...ux-watchdog.org>, <linux@...ck-us.net>,
        <robh+dt@...nel.org>, <linux-watchdog@...r.kernel.org>,
        <devicetree@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <linux-arm-kernel@...ts.infradead.org>,
        <Nicolas.Ferre@...rochip.com>
Subject: Re: [PATCH 2/3] watchdog: sam9x60_wdt: introduce sam9x60 watchdog
 timer driver



On 02.10.2019 13:23, Alexandre Belloni wrote:

> Hi,
> 
> On 02/10/2019 07:35:26+0000, Eugen.Hristev@...rochip.com wrote:
>> +static void wdt_write(struct sam9x60_wdt *wdt, u32 field, u32 val)
>> +{
>> +	/*
>> +	 * WDT_CR and WDT_MR must not be modified within three slow clock
>> +	 * periods following a restart of the watchdog performed by a write
>> +	 * access in WDT_CR.
>> +	 */
>> +	while (time_before(jiffies, wdt->last_ping + WDT_DELAY))
>> +		usleep_range(30, 125);
>> +	writel_relaxed(val, wdt->reg_base + field);
>> +	wdt->last_ping = jiffies;
>> +}
>> +
>> +static void wdt_write_nosleep(struct sam9x60_wdt *wdt, u32 field, u32 val)
>> +{
>> +	if (time_before(jiffies, wdt->last_ping + WDT_DELAY))
>> +		usleep_range(123, 250);
> 
> So you have a _nosleep function that does sleep?
> 
>> +	writel_relaxed(val, wdt->reg_base + field);
>> +	wdt->last_ping = jiffies;
>> +}
>> +
>> +static int sam9x60_wdt_start(struct watchdog_device *wdd)
>> +{
>> +	struct sam9x60_wdt *wdt = watchdog_get_drvdata(wdd);
>> +
>> +	wdt->mr &= ~AT91_WDT_WDDIS;
>> +	wdt_write(wdt, AT91_WDT_MR, wdt->mr);
>> +	wdt_write_nosleep(wdt, AT91_WDT_IER, wdt->ir);
> 
> I don't think AT91_WDT_IER needs to be protected, you can probably write
> it directly. Also, you certainly need to do that before starting the
> watchdog to avoid race conditions.
> 
>> +
>> +	return 0;
>> +}
>> +
>> +static int sam9x60_wdt_stop(struct watchdog_device *wdd)
>> +{
>> +	struct sam9x60_wdt *wdt = watchdog_get_drvdata(wdd);
>> +
>> +	wdt->mr |= AT91_WDT_WDDIS;
>> +	wdt_write(wdt, AT91_WDT_MR, wdt->mr);
>> +	wdt_write_nosleep(wdt, AT91_WDT_IDR, wdt->ir);
>> +
> 
> I don't think AT91_WDT_IDR needs to be protected.
> 
>> +	return 0;
>> +}
>> +
>> +static int sam9x60_wdt_ping(struct watchdog_device *wdd)
>> +{
>> +	struct sam9x60_wdt *wdt = watchdog_get_drvdata(wdd);
>> +
>> +	wdt_write(wdt, AT91_WDT_CR, AT91_WDT_KEY | AT91_WDT_WDRSTT);
>> +
>> +	return 0;
>> +}
>> +
>> +static int sam9x60_wdt_set_timeout(struct watchdog_device *wdd,
>> +				   unsigned int timeout)
>> +{
>> +	struct sam9x60_wdt *wdt = watchdog_get_drvdata(wdd);
>> +
>> +	wdt_write(wdt, AT91_WDT_WLR,
>> +		  AT91_WDT_SET_COUNTER(WDT_SEC2TICKS(timeout)));
>> +
> 
> I don't think AT91_WDT_WLR needs to be protected.
> 
>> +	wdd->timeout = timeout;
>> +
>> +	return 0;
>> +}
>> +
>> +static const struct watchdog_info sam9x60_wdt_info = {
>> +	.options = WDIOF_SETTIMEOUT | WDIOF_MAGICCLOSE | WDIOF_KEEPALIVEPING,
>> +	.identity = "Microchip SAM9X60 Watchdog",
>> +};
>> +
>> +static const struct watchdog_ops sam9x60_wdt_ops = {
>> +	.owner = THIS_MODULE,
>> +	.start = sam9x60_wdt_start,
>> +	.stop = sam9x60_wdt_stop,
>> +	.ping = sam9x60_wdt_ping,
>> +	.set_timeout = sam9x60_wdt_set_timeout,
>> +};
>> +
>> +static irqreturn_t sam9x60_wdt_irq_handler(int irq, void *dev_id)
>> +{
>> +	struct sam9x60_wdt *wdt = platform_get_drvdata(dev_id);
>> +
>> +	if (wdt_read(wdt, AT91_WDT_ISR)) {
>> +		pr_crit("Microchip Watchdog Software Reset\n");
>> +		emergency_restart();
>> +		pr_crit("Reboot didn't succeed\n");
>> +	}
> 
> I'm not really convinced by the software restart use case but I guess it
> is to be able to shut down while still flushing data to the storage.
> This would not protect against kernel issues then.

Hi Alexandre,

That's correct. It is to do a software shutdown instead of hard reboot 
by hardware. It has it;s use cases, so I preserved the same level of 
functionality as in sama5d4_wdt

> 
>> +
>> +	return IRQ_HANDLED;
>> +}
>> +
>> +static int of_sam9x60_wdt_init(struct device_node *np, struct sam9x60_wdt *wdt)
>> +{
>> +	const char *tmp;
>> +
>> +	wdt->mr = AT91_WDT_WDDIS;
>> +
>> +	if (!of_property_read_string(np, "atmel,watchdog-type", &tmp) &&
>> +	    !strcmp(tmp, "software"))
>> +		wdt->ir = AT91_WDT_PERINT;
>> +	else
>> +		wdt->mr |= AT91_WDT_PERIODRST;
>> +
>> +	if (of_property_read_bool(np, "atmel,idle-halt"))
>> +		wdt->mr |= AT91_WDT_WDIDLEHLT;
>> +
>> +	if (of_property_read_bool(np, "atmel,dbg-halt"))
>> +		wdt->mr |= AT91_WDT_WDDBGHLT;
>> +
>> +	return 0;
>> +}
>> +
>> +static int sam9x60_wdt_init(struct sam9x60_wdt *wdt)
>> +{
>> +	u32 reg;
>> +	/*
>> +	 * When booting and resuming, the bootloader may have changed the
>> +	 * watchdog configuration.
>> +	 * If the watchdog is already running, we can safely update it.
>> +	 * Else, we have to disable it properly.
>> +	 */
>> +	if (wdt_enabled) {
>> +		wdt_write_nosleep(wdt, AT91_WDT_MR, wdt->mr);
>> +		wdt_write_nosleep(wdt, AT91_WDT_IER, wdt->ir);
>> +		wdt_write(wdt, AT91_WDT_WLR,
>> +			  AT91_WDT_SET_COUNTER(WDT_SEC2TICKS(WDT_DEFAULT_TIMEOUT)));
>> +
>> +	} else {
>> +		reg = wdt_read(wdt, AT91_WDT_MR);
>> +		if (!(reg & AT91_WDT_WDDIS))
>> +			wdt_write_nosleep(wdt, AT91_WDT_MR,
>> +					  reg | AT91_WDT_WDDIS);
>> +	}
>> +	return 0;
>> +}
>> +
>> +static int sam9x60_wdt_probe(struct platform_device *pdev)
>> +{
>> +	struct watchdog_device *wdd;
>> +	struct sam9x60_wdt *wdt;
>> +	struct resource *res;
>> +	void __iomem *regs;
>> +	u32 irq = 0;
>> +	int ret;
>> +
>> +	wdt = devm_kzalloc(&pdev->dev, sizeof(*wdt), GFP_KERNEL);
>> +	if (!wdt)
>> +		return -ENOMEM;
>> +
>> +	wdd = &wdt->wdd;
>> +	wdd->timeout = WDT_DEFAULT_TIMEOUT;
>> +	wdd->info = &sam9x60_wdt_info;
>> +	wdd->ops = &sam9x60_wdt_ops;
>> +	wdd->min_timeout = MIN_WDT_TIMEOUT;
>> +	wdd->max_timeout = MAX_WDT_TIMEOUT;
>> +	wdt->last_ping = jiffies;
>> +
>> +	watchdog_set_drvdata(wdd, wdt);
>> +
>> +	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>> +	regs = devm_ioremap_resource(&pdev->dev, res);
>> +	if (IS_ERR(regs))
>> +		return PTR_ERR(regs);
>> +
>> +	wdt->reg_base = regs;
>> +
>> +	irq = irq_of_parse_and_map(pdev->dev.of_node, 0);
>> +	if (!irq)
>> +		dev_warn(&pdev->dev, "failed to get IRQ from DT\n");
>> +
>> +	ret = of_sam9x60_wdt_init(pdev->dev.of_node, wdt);
>> +	if (ret)
>> +		return ret;
>> +
>> +	if ((wdt->ir & AT91_WDT_PERINT) && irq) {
>> +		ret = devm_request_irq(&pdev->dev, irq, sam9x60_wdt_irq_handler,
>> +				       IRQF_SHARED | IRQF_IRQPOLL |
>> +				       IRQF_NO_SUSPEND, pdev->name, pdev);
>> +		if (ret) {
>> +			dev_err(&pdev->dev,
>> +				"cannot register interrupt handler\n");
>> +			return ret;
>> +		}
>> +	}
>> +
>> +	watchdog_init_timeout(wdd, wdt_timeout, &pdev->dev);
>> +
>> +	ret = sam9x60_wdt_init(wdt);
>> +	if (ret)
>> +		return ret;
>> +
>> +	watchdog_set_nowayout(wdd, nowayout);
>> +
>> +	ret = watchdog_register_device(wdd);
>> +	if (ret) {
>> +		dev_err(&pdev->dev, "failed to register watchdog device\n");
>> +		return ret;
>> +	}
>> +
>> +	platform_set_drvdata(pdev, wdt);
>> +
>> +	dev_info(&pdev->dev, "initialized (timeout = %d sec, nowayout = %d)\n",
>> +		 wdd->timeout, nowayout);
>> +
>> +	return 0;
>> +}
>> +
>> +static int sam9x60_wdt_remove(struct platform_device *pdev)
>> +{
>> +	struct sam9x60_wdt *wdt = platform_get_drvdata(pdev);
>> +
>> +	sam9x60_wdt_stop(&wdt->wdd);
>> +
>> +	watchdog_unregister_device(&wdt->wdd);
>> +
>> +	return 0;
>> +}
>> +
>> +static const struct of_device_id sam9x60_wdt_of_match[] = {
>> +	{ .compatible = "microchip,sam9x60-wdt", },
>> +	{ }
>> +};
>> +MODULE_DEVICE_TABLE(of, sam9x60_wdt_of_match);
>> +
>> +#ifdef CONFIG_PM_SLEEP
> 
> Most of the logic has been copy/pasted from sama5d4_wdt.c and this
> already miss some improvement that have been made between the time you
> copied it and now.

I will fix accordingly. As I said in the commit message, sama5d4_wdt is 
used as a starting point so yes, all the functionality is the same, 
except the actual hardware interaction.

> 
> Are you sure both drivers shouldn't be merged? I feel like this will be a
> maintenance hell if we don't do that now.

It could be merged, but we should do so ?
Could have two compatibles, with platform data, selectable, and with 
different functions, that can be selected.. either this or that.
You think that's a better way to handle this new IP block ?
I would like to avoid having a big driver covering multiple different 
hardware pieces, but that's just my preference. I can rework this into a 
single driver if it's better that way.

Eugen

> 
>> +static int sam9x60_wdt_resume(struct device *dev)
>> +{
>> +	struct sam9x60_wdt *wdt = dev_get_drvdata(dev);
>> +
>> +	/*
>> +	 * FIXME: writing MR also pings the watchdog which may not be desired.
>> +	 * This should only be done when the registers are lost on suspend but
>> +	 * there is no way to get this information right now.
>> +	 */
>> +	sam9x60_wdt_init(wdt);
>> +
>> +	return 0;
>> +}
>> +#endif
>> +
>> +static SIMPLE_DEV_PM_OPS(sam9x60_wdt_pm_ops, NULL,
>> +			 sam9x60_wdt_resume);
>> +
>> +static struct platform_driver sam9x60_wdt_driver = {
>> +	.probe		= sam9x60_wdt_probe,
>> +	.remove		= sam9x60_wdt_remove,
>> +	.driver		= {
>> +		.name	= "sam9x60_wdt",
>> +		.pm	= &sam9x60_wdt_pm_ops,
>> +		.of_match_table = sam9x60_wdt_of_match,
>> +	}
>> +};
>> +module_platform_driver(sam9x60_wdt_driver);
>> +
>> +MODULE_AUTHOR("Eugen Hristev");
>> +MODULE_DESCRIPTION("Microchip SAM9X60 Watchdog Timer driver");
>> +MODULE_LICENSE("GPL v2");
>> -- 
>> 2.7.4
>>
> 

Powered by blists - more mailing lists