RE: [EXT] Re: [PATCH 2/2 v4] Watchdog: Add marvell GTI watchdog driver

From: Bharat Bhushan
Date: Tue May 02 2023 - 13:05:32 EST


Please see inline

> -----Original Message-----
> From: Guenter Roeck <groeck7@xxxxxxxxx> On Behalf Of Guenter Roeck
> Sent: Saturday, April 29, 2023 8:44 PM
> To: Bharat Bhushan <bbhushan2@xxxxxxxxxxx>; wim@xxxxxxxxxxxxxxxxxx;
> robh+dt@xxxxxxxxxx; krzysztof.kozlowski+dt@xxxxxxxxxx; linux-
> watchdog@xxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx; linux-
> kernel@xxxxxxxxxxxxxxx; Sunil Kovvuri Goutham <sgoutham@xxxxxxxxxxx>
> Subject: [EXT] Re: [PATCH 2/2 v4] Watchdog: Add marvell GTI watchdog driver
>
> External Email
>
> ----------------------------------------------------------------------
> On 4/27/23 00:14, Bharat Bhushan wrote:
> > This patch add support for Marvell GTI watchdog. Global timer unit
> > (GTI) support hardware watchdog timer. Software programs watchdog
> > timer to generate interrupt on first timeout, second timeout is
> > configured to be ignored and system reboots on third timeout.
> >
> > Signed-off-by: Bharat Bhushan <bbhushan2@xxxxxxxxxxx>
> > ---
> > v4:
> > - Use generic name "GTI (Global Timer)" and removed OcteonTX2 as this driver
> > is applicable for other marvell devices as well.
> >
> > drivers/watchdog/Kconfig | 11 ++
> > drivers/watchdog/Makefile | 1 +
> > drivers/watchdog/marvell_gti_wdt.c | 271
> +++++++++++++++++++++++++++++
> > 3 files changed, 283 insertions(+)
> > create mode 100644 drivers/watchdog/marvell_gti_wdt.c
> >
> > diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig index
> > f0872970daf9..a96e7d7c0ad2 100644
> > --- a/drivers/watchdog/Kconfig
> > +++ b/drivers/watchdog/Kconfig
> > @@ -1779,6 +1779,17 @@ config OCTEON_WDT
> > from the first interrupt, it is then only poked when the
> > device is written.
> >
> > +config MARVELL_GTI_WDT
> > + tristate "Marvell GTI Watchdog driver"
> > + depends on ARCH_THUNDER || COMPILE_TEST
>
> As 0-day reported, COMPILE_TEST does not work if there are asm includes in the
> code.

Will remove asm include dependency in next version

>
> > + help
> > + Marvell GTI hardware supports watchdog timer. First timeout
> > + works as watchdog pretimeout and installed interrupt handler
> > + will be called on first timeout. Hardware can generate interrupt
> > + to SCP on second timeout but it is not enabled, So second
> > + timeout is ignored. If device poke does not happen then system
> > + will reboot on third timeout.
> > +
> > config BCM2835_WDT
> > tristate "Broadcom BCM2835 hardware watchdog"
> > depends on ARCH_BCM2835 || (OF && COMPILE_TEST) diff --git
> > a/drivers/watchdog/Makefile b/drivers/watchdog/Makefile index
> > 9cbf6580f16c..bd425408fcaa 100644
> > --- a/drivers/watchdog/Makefile
> > +++ b/drivers/watchdog/Makefile
> > @@ -98,6 +98,7 @@ obj-$(CONFIG_VISCONTI_WATCHDOG) += visconti_wdt.o
> > obj-$(CONFIG_MSC313E_WATCHDOG) += msc313e_wdt.o
> > obj-$(CONFIG_APPLE_WATCHDOG) += apple_wdt.o
> > obj-$(CONFIG_SUNPLUS_WATCHDOG) += sunplus_wdt.o
> > +obj-$(CONFIG_MARVELL_GTI_WDT) += marvell_gti_wdt.o
> >
> > # X86 (i386 + ia64 + x86_64) Architecture
> > obj-$(CONFIG_ACQUIRE_WDT) += acquirewdt.o diff --git
> > a/drivers/watchdog/marvell_gti_wdt.c
> > b/drivers/watchdog/marvell_gti_wdt.c
> > new file mode 100644
> > index 000000000000..a3f0740dedca
> > --- /dev/null
> > +++ b/drivers/watchdog/marvell_gti_wdt.c
> > @@ -0,0 +1,271 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/* Marvell GTI Watchdog driver
> > + *
> > + * Copyright (C) 2023 Marvell.
> > + */
> > +
> > +#include <linux/module.h>
> > +#include <linux/interrupt.h>
> > +#include <linux/of_platform.h>
> > +#include <linux/platform_device.h>
> > +#include <linux/watchdog.h>
> > +
> > +#include <asm/arch_timer.h>
> > +
> > +/*
> > + * Hardware supports following mode of operation:
> > + * 1) Interrupt Only:
> > + * This will generate the interrupt to arm core whenever timeout happens.
> > + *
> > + * 2) Interrupt + del3t (Interrupt to firmware (SCP processor)).
> > + * This will generate interrupt to arm core on 1st timeout happens
> > + * This will generate interrupt to SCP processor on 2nd timeout happens
> > + *
> > + * 3) Interrupt + Interrupt to SCP processor (called delt3t) + reboot.
> > + * This will generate interrupt to arm core on 1st timeout happens
> > + * Will generate interrupt to SCP processor on 2nd timeout happens,
> > + * if interrupt is configured.
> > + * Reboot on 3rd timeout.
> > + *
> > + * Driver will use hardware in mode-3 above so that system can reboot
> > +in case
> > + * a hardware hang. Also h/w is configured not to generate SCP
> > +interrupt, so
> > + * effectively 2nd timeout is ignored within hardware.
> > + */
> > +
> > +
> > +/* GTI CWD Watchdog Registers */
> > +#define GTI_CWD_GLOBAL_WDOG_IDX 63
>
> It might make sense to explain "63".

Okay, will add below comment
/* GTI hardware block supports total 64 timers and timer 63 is used for system watchdog */

"IDX" seems confusing, will rename to "GTI_CWD_GLOBAL_WDOG_TIMER"

>
> > +#define GTI_CWD_WDOG (0x8 *
> GTI_CWD_GLOBAL_WDOG_IDX)
> > +#define GTI_CWD_WDOG_MODE_INT_DEL3T_RST 0x3
> > +#define GTI_CWD_WDOG_MODE_MASK 0x3
> > +#define GTI_CWD_WDOG_LEN_SHIFT 4
> > +#define GTI_CWD_WDOG_CNT_SHIFT 20
> > +
> > +/* GTI Per-core Watchdog Interrupt Register */
> > +#define GTI_CWD_INT 0x200
> > +#define GTI_CWD_INT_PENDING_STATUS (1ULL <<
> GTI_CWD_GLOBAL_WDOG_IDX)
>
> Please use BIT_ULL().

Ok

>
> > +
> > +/* GTI Per-core Watchdog Interrupt Enable Clear Register */
> > +#define GTI_CWD_INT_ENA_CLR 0x210
> > +#define GTI_CWD_INT_ENA_CLR_VAL (1ULL <<
> GTI_CWD_GLOBAL_WDOG_IDX)
> > +
> > +/* GTI Per-core Watchdog Interrupt Enable Set Register */
> > +#define GTI_CWD_INT_ENA_SET 0x218
> > +#define GTI_CWD_INT_ENA_SET_VAL (1ULL <<
> GTI_CWD_GLOBAL_WDOG_IDX)
> > +
> > +/* GTI Per-core Watchdog Poke Registers */
> > +#define GTI_CWD_POKE (0x10000 + 0x8 *
> GTI_CWD_GLOBAL_WDOG_IDX)
>
> There should be a comment explaining the "8".

GTI_CWD_POKE is poke register for watchdog counter GTI_CWD_GLOBAL_WDOG_TIMER .
Above calculation is for getting the register address for watchdog timer GTI_CWD_GLOBAL_WDOG_TIMER
Can add comment
/* Each POKE register is 64bit and so GTI_CWD_POKE is register address for watchdog counter GTI_CWD_GLOBAL_WDOG_TIMER */

> Also, there is a define for "0x8 *
> GTI_CWD_GLOBAL_WDOG_IDX" above.
> Why not use it here ?

GTI_CWD_GLOBAL_WDOG_TIMER is used for register address calculation, so will keep then separate to avoid any confusion.

>
> > +#define GTI_CWD_POKE_VAL (1ULL << GTI_CWD_GLOBAL_WDOG_IDX)
>
> I don't think it makes sense to have three defines with the same value.

Yes they are same value, but they are bit field in different h/w register. I want to keep a separate name to avoid any confusion.

>
> > +
> > +struct gti_wdt_priv {
> > + struct watchdog_device wdev;
> > + void __iomem *base;
> > + u64 clock_freq;
> > + int irq;
> > +};
> > +
> > +static irqreturn_t gti_wdt_interrupt(int irq, void *data) {
> > + struct watchdog_device *wdev = data;
> > + struct gti_wdt_priv *priv = watchdog_get_drvdata(wdev);
> > +
> > + /* Clear Interrupt Pending Status */
> > + writeq(GTI_CWD_INT_PENDING_STATUS, priv->base + GTI_CWD_INT);
> > +
> > + watchdog_notify_pretimeout(wdev);
> > +
> > + return IRQ_HANDLED;
> > +}
> > +
> > +static int gti_wdt_ping(struct watchdog_device *wdev) {
> > + struct gti_wdt_priv *priv = watchdog_get_drvdata(wdev);
> > +
> > + writeq(GTI_CWD_POKE_VAL, priv->base + GTI_CWD_POKE);
> > +
> > + return 0;
> > +}
> > +
> > +static int gti_wdt_start(struct watchdog_device *wdev) {
> > + struct gti_wdt_priv *priv = watchdog_get_drvdata(wdev);
> > + u64 regval;
> > +
> > + if (!wdev->pretimeout)
> > + return -EINVAL;
> > +
> > + set_bit(WDOG_HW_RUNNING, &wdev->status);
> > +
> > + /* Clear any pending interrupt */
> > + writeq(GTI_CWD_INT_PENDING_STATUS, priv->base + GTI_CWD_INT);
> > +
> > + /* Enable Interrupt */
> > + writeq(GTI_CWD_INT_ENA_SET_VAL, priv->base +
> GTI_CWD_INT_ENA_SET);
> > +
> > + /* Set (Interrupt + SCP interrupt (DEL3T) + core domain reset) Mode */
> > + regval = readq(priv->base + GTI_CWD_WDOG);
> > + regval |= GTI_CWD_WDOG_MODE_INT_DEL3T_RST;
> > + writeq(regval, priv->base + GTI_CWD_WDOG);
> > +
> > + return 0;
> > +}
> > +
> > +static int gti_wdt_stop(struct watchdog_device *wdev) {
> > + struct gti_wdt_priv *priv = watchdog_get_drvdata(wdev);
> > + u64 regval;
> > +
> > + /* Disable Interrupt */
> > + writeq(GTI_CWD_INT_ENA_CLR_VAL, priv->base +
> GTI_CWD_INT_ENA_CLR);
> > +
> > + /* Set GTI_CWD_WDOG.Mode = 0 to stop the timer */
> > + regval = readq(priv->base + GTI_CWD_WDOG);
> > + regval &= ~GTI_CWD_WDOG_MODE_MASK;
> > + writeq(regval, priv->base + GTI_CWD_WDOG);
> > +
> > + return 0;
> > +}
> > +
> > +static int gti_wdt_settimeout(struct watchdog_device *wdev,
> > + unsigned int timeout)
> > +{
> > + struct gti_wdt_priv *priv = watchdog_get_drvdata(wdev);
> > + u64 timeout_wdog, regval;
> > +
> > + /* Update new timeout */
> > + wdev->timeout = timeout;
> > +
> > + /* Pretimeout is 1/3 of timeout */
> > + wdev->pretimeout = timeout / 3;
> > + if (!wdev->pretimeout)
> > + return -EINVAL;
>
> Since min_timeout is set to 3 seconds, this can not happen, and the check is
> unnecessary.
>
> > +
> > + /* Get clock cycles from timeout in second */
>
> from pretimeout ?

Yes, will change comment.

>
> > + timeout_wdog = wdev->pretimeout * priv->clock_freq;
> > +
> > + /* Watchdog counts in 1024 cycle steps */
> > + timeout_wdog = timeout_wdog >> 10;
> > +
> > + timeout_wdog = (timeout_wdog + 0xff) >> 8;
> > + if (timeout_wdog >= 0x10000)
> > + timeout_wdog = 0xffff;
> > +
> > + /*
> > + * GTI_CWD_WDOG.LEN have only upper 16-bits of 24-bits
> > + * GTI_CWD_WDOG.CNT, need addition shift of 8.
> > + */
> > + regval = readq(priv->base + GTI_CWD_WDOG);
> > + regval &= GTI_CWD_WDOG_MODE_MASK;
> > + regval |= (timeout_wdog << (GTI_CWD_WDOG_CNT_SHIFT + 8)) |
> > + (timeout_wdog << GTI_CWD_WDOG_LEN_SHIFT);
> > + writeq(regval, priv->base + GTI_CWD_WDOG);
> > +
> > + return 0;
> > +}
> > +
> > +static int gti_wdt_set_pretimeout(struct watchdog_device *wdev,
> > + unsigned int timeout)
> > +{
> > + struct gti_wdt_priv *priv = watchdog_get_drvdata(wdev);
> > + struct watchdog_device *wdog_dev = &priv->wdev;
> > +
> > + /* pretimeout should 1/3 of max_timeout */
> > + if ((timeout * 3) <= wdog_dev->max_timeout)
>
> Unnecessary ()
>
> > + return gti_wdt_settimeout(wdev, timeout * 3);
> > +
> > + return -EINVAL;
> > +}
> > +
> > +static const struct watchdog_info gti_wdt_ident = {
> > + .identity = "Marvell GTI watchdog",
> > + .options = WDIOF_SETTIMEOUT | WDIOF_PRETIMEOUT |
> WDIOF_KEEPALIVEPING |
> > + WDIOF_MAGICCLOSE | WDIOF_CARDRESET, };
> > +
> > +static const struct watchdog_ops gti_wdt_ops = {
> > + .owner = THIS_MODULE,
> > + .start = gti_wdt_start,
> > + .stop = gti_wdt_stop,
> > + .ping = gti_wdt_ping,
> > + .set_timeout = gti_wdt_settimeout,
> > + .set_pretimeout = gti_wdt_set_pretimeout, };
> > +
> > +static int gti_wdt_probe(struct platform_device *pdev) {
> > + struct gti_wdt_priv *priv;
> > + struct device *dev = &pdev->dev;
> > + struct watchdog_device *wdog_dev;
> > + int irq;
> > + int err;
> > +
> > + priv = devm_kzalloc(&pdev->dev, sizeof(*priv), GFP_KERNEL);
> > + if (!priv)
> > + return -ENOMEM;
> > +
> > + priv->base = devm_platform_ioremap_resource(pdev, 0);
> > + if (IS_ERR(priv->base))
> > + return dev_err_probe(&pdev->dev, PTR_ERR(priv->base),
> > + "reg property not valid/found\n");
> > +
> > + priv->clock_freq = arch_timer_get_cntfrq();
>
> arch_timer_get_cntfrq() returns u32. Storing it as u64 to save a couple of
> typecasts really doesn't add value.
>
> Note that this limits compilability to arm and arm64.

Will remove using arch_timer_get_cntfrq() and clock reference in device tree.

>
> > +
> > + wdog_dev = &priv->wdev;
> > + wdog_dev->info = &gti_wdt_ident,
> > + wdog_dev->ops = &gti_wdt_ops,
> > + wdog_dev->parent = dev;
> > + /*
> > + * Watchdog counter is 24 bit where lower 8 bits are zeros, so maximum
> > + * value that can be programmed in hardware is 0xffff00. This counter
> > + * decrements every one microsec.
> > + * Hardware will generate interrupt on first timeout (pretimeout)
> > + * reset the system on 3rd timeout.
> > + * Max watchdog pretimeout can be 16 sec, so max timeout is 48 sec.
> > + * Let's have min pretimeout to 1sec, so min timeout is 3sec.
> > + */
> > + wdog_dev->max_timeout = 48;
>
> The code setting the actual timeout checks for overflow, suggesting that the
> actual maximum timeout depends on the clock speed and may be less than 48
> seconds. If that ever happens, the actual timeout would be set to less than 48
> seconds, which could cause random reboots.

Will use clock for these calculation.

Thanks
-Bharat

>
> > + wdog_dev->min_timeout = 3;
> > + wdog_dev->timeout = 30;
> > + wdog_dev->pretimeout = wdog_dev->timeout / 3;
> > +
> > + priv->irq = irq;
> > + watchdog_set_drvdata(wdog_dev, priv);
> > + platform_set_drvdata(pdev, priv);
> > + gti_wdt_settimeout(wdog_dev, wdog_dev->timeout);
> > + watchdog_stop_on_reboot(wdog_dev);
> > + watchdog_stop_on_unregister(wdog_dev);
> > +
> > + err = devm_watchdog_register_device(dev, wdog_dev);
> > + if (err)
> > + return err;
> > +
> > + irq = platform_get_irq(pdev, 0);
> > + if (irq < 0)
> > + return dev_err_probe(&pdev->dev, irq, "IRQ resource not
> found\n");
> > +
> > + err = devm_request_irq(dev, irq, gti_wdt_interrupt, 0,
> > + pdev->name, &priv->wdev);
> > + if (err)
> > + return dev_err_probe(dev, err, "Failed to register interrupt
> > +handler\n");
> > +
> > + dev_info(dev, "Watchdog enabled (timeout=%d sec)\n", wdog_dev-
> >timeout);
> > + return 0;
> > +}
> > +
> > +static const struct of_device_id gti_wdt_of_match[] = {
> > + { .compatible = "marvell,gti-wdt", },
> > + { },
> > +};
> > +MODULE_DEVICE_TABLE(of, gti_wdt_of_match);
> > +
> > +static struct platform_driver gti_wdt_driver = {
> > + .driver = {
> > + .name = "gti-wdt",
> > + .of_match_table = gti_wdt_of_match,
> > + },
> > + .probe = gti_wdt_probe,
> > +};
> > +module_platform_driver(gti_wdt_driver);
> > +
> > +MODULE_AUTHOR("Bharat Bhushan <bbhushan2@xxxxxxxxxxx>");
> > +MODULE_DESCRIPTION("Marvell GTI watchdog driver");