* [PATCH v2 0/3] introduce watchdog driver for Mellanox systems. @ 2019-02-07 21:10 michaelsh 2019-02-07 21:10 ` [PATCH v2 1/3] platform_data/mlxreg: addittions for mellanox watchdog driver michaelsh ` (2 more replies) 0 siblings, 3 replies; 8+ messages in thread From: michaelsh @ 2019-02-07 21:10 UTC (permalink / raw) To: wim, linux, andy, dvhart Cc: linux-watchdog, platform-driver-x86, vadimp, Michael Shych From: Michael Shych <michaelsh@mellanox.com> This patchset introduces watchdog driver for a various range of Mellanox Ethernet and Infiniband switch systems. Mellanox watchdog device is implemented in programmable logic device. There are 2 types of HW watchdog implementations with diferent capabilities. This mlx-wdt driver supports both HW watchdog implementations. Mellanox system can have 2 watchdogs: main and auxiliary. Main and auxiliary watchdog devices can be enabled together on the same system. There are several actions that can be defined in the watchdog: system reset, start fans on full speed and increase register counter. Michael Shych (3): platform_data/mlxreg: addittions for mellanox watchdog driver. watchdog: mlx-wdt: introduce watchdog driver for Mellanox systems. Documentation/watchdog: Add documentation mlx-wdt driver Documentation/watchdog/mlx-wdt.txt | 48 ++++++ drivers/watchdog/Kconfig | 16 ++ drivers/watchdog/Makefile | 1 + drivers/watchdog/mlx_wdt.c | 323 +++++++++++++++++++++++++++++++++++ include/linux/platform_data/mlxreg.h | 6 + 5 files changed, 394 insertions(+) create mode 100644 Documentation/watchdog/mlx-wdt.txt create mode 100644 drivers/watchdog/mlx_wdt.c -- 2.11.0 ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH v2 1/3] platform_data/mlxreg: addittions for mellanox watchdog driver. 2019-02-07 21:10 [PATCH v2 0/3] introduce watchdog driver for Mellanox systems michaelsh @ 2019-02-07 21:10 ` michaelsh 2019-02-18 16:12 ` Guenter Roeck 2019-02-07 21:10 ` [PATCH v2 2/3] watchdog: mlx-wdt: introduce watchdog driver for Mellanox systems michaelsh 2019-02-07 21:10 ` [PATCH v2 3/3] Documentation/watchdog: Add documentation mlx-wdt driver michaelsh 2 siblings, 1 reply; 8+ messages in thread From: michaelsh @ 2019-02-07 21:10 UTC (permalink / raw) To: wim, linux, andy, dvhart Cc: linux-watchdog, platform-driver-x86, vadimp, Michael Shych From: Michael Shych <michaelsh@mellanox.com> There are two new fields added to mlxreg core structure: features - supported features of device and identity - device identity name. Add new defines for watchdog features. Signed-off-by: Michael Shych <michaelsh@mellanox.com> --- v1->v2 Change feature name MLXREG_CORE_WD_FEATURE_NOSTOP_AFTER_START to MLXREG_CORE_WD_FEATURE_NOWAYOUT --- include/linux/platform_data/mlxreg.h | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/include/linux/platform_data/mlxreg.h b/include/linux/platform_data/mlxreg.h index 19f5cb618c55..aadc80700573 100644 --- a/include/linux/platform_data/mlxreg.h +++ b/include/linux/platform_data/mlxreg.h @@ -35,6 +35,8 @@ #define __LINUX_PLATFORM_DATA_MLXREG_H #define MLXREG_CORE_LABEL_MAX_SIZE 32 +#define MLXREG_CORE_WD_FEATURE_NOWAYOUT BIT(0) +#define MLXREG_CORE_WD_FEATURE_START_AT_BOOT BIT(1) /** * struct mlxreg_hotplug_device - I2C device data: @@ -110,11 +112,15 @@ struct mlxreg_core_item { * @led_data: led private data; * @regmap: register map of parent device; * @counter: number of led instances; + * @features: supported features of device; + * @identity: device identity name; */ struct mlxreg_core_platform_data { struct mlxreg_core_data *data; void *regmap; int counter; + u32 features; + char identity[MLXREG_CORE_LABEL_MAX_SIZE]; }; /** -- 2.11.0 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* Re: [PATCH v2 1/3] platform_data/mlxreg: addittions for mellanox watchdog driver. 2019-02-07 21:10 ` [PATCH v2 1/3] platform_data/mlxreg: addittions for mellanox watchdog driver michaelsh @ 2019-02-18 16:12 ` Guenter Roeck 0 siblings, 0 replies; 8+ messages in thread From: Guenter Roeck @ 2019-02-18 16:12 UTC (permalink / raw) To: michaelsh, wim, andy, dvhart; +Cc: linux-watchdog, platform-driver-x86, vadimp On 2/7/19 1:10 PM, michaelsh@mellanox.com wrote: > From: Michael Shych <michaelsh@mellanox.com> > > There are two new fields added to mlxreg core structure: > features - supported features of device and > identity - device identity name. > Add new defines for watchdog features. > > Signed-off-by: Michael Shych <michaelsh@mellanox.com> > Subject: s/addittions/additions/ Otherwise Reviewed-by: Guenter Roeck <linux@roeck-us.net> > --- > v1->v2 > Change feature name MLXREG_CORE_WD_FEATURE_NOSTOP_AFTER_START > to MLXREG_CORE_WD_FEATURE_NOWAYOUT > --- > include/linux/platform_data/mlxreg.h | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/include/linux/platform_data/mlxreg.h b/include/linux/platform_data/mlxreg.h > index 19f5cb618c55..aadc80700573 100644 > --- a/include/linux/platform_data/mlxreg.h > +++ b/include/linux/platform_data/mlxreg.h > @@ -35,6 +35,8 @@ > #define __LINUX_PLATFORM_DATA_MLXREG_H > > #define MLXREG_CORE_LABEL_MAX_SIZE 32 > +#define MLXREG_CORE_WD_FEATURE_NOWAYOUT BIT(0) > +#define MLXREG_CORE_WD_FEATURE_START_AT_BOOT BIT(1) > > /** > * struct mlxreg_hotplug_device - I2C device data: > @@ -110,11 +112,15 @@ struct mlxreg_core_item { > * @led_data: led private data; > * @regmap: register map of parent device; > * @counter: number of led instances; > + * @features: supported features of device; > + * @identity: device identity name; > */ > struct mlxreg_core_platform_data { > struct mlxreg_core_data *data; > void *regmap; > int counter; > + u32 features; > + char identity[MLXREG_CORE_LABEL_MAX_SIZE]; > }; > > /** > ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH v2 2/3] watchdog: mlx-wdt: introduce watchdog driver for Mellanox systems. 2019-02-07 21:10 [PATCH v2 0/3] introduce watchdog driver for Mellanox systems michaelsh 2019-02-07 21:10 ` [PATCH v2 1/3] platform_data/mlxreg: addittions for mellanox watchdog driver michaelsh @ 2019-02-07 21:10 ` michaelsh 2019-02-18 16:26 ` Guenter Roeck 2019-02-07 21:10 ` [PATCH v2 3/3] Documentation/watchdog: Add documentation mlx-wdt driver michaelsh 2 siblings, 1 reply; 8+ messages in thread From: michaelsh @ 2019-02-07 21:10 UTC (permalink / raw) To: wim, linux, andy, dvhart Cc: linux-watchdog, platform-driver-x86, vadimp, Michael Shych From: Michael Shych <michaelsh@mellanox.com> Introduce watchdog driver for variious rang of Mellanox Ethernet and Infiniband switch systems. Watchdog driver for Mellanox watchdog devices, implemented in programmable logic device. Main and auxiliary watchdog devices can exist on the same system. There are several actions that can be defined in the watchdog: system reset, start fans on full speed and increase a counter. The last 2 actions are performed without a system reset. Actions without reset are provided for auxiliary watchdog devices, which is optional. Access to HW registers is performed through generic regmap interface. There are 2 types of HW watchdog implementations. Type 1: actual HW timeout can be defined as power of 2 msec. e.g. timeout 20 sec will be rounded up to 32768 msec.; maximum timeout period is 32 sec (32768 msec.); get time-left isn't supported Type 2: actual HW timeout is defined in sec. and it's the same as user-defined timeout; maximum timeout is 255 sec; get time-left is supported; Watchdog driver is probed from the common mlx_platform driver. Signed-off-by: Michael Shych <michaelsh@mellanox.com> --- v1->v2: Comments pointed out by Guenter: 1. Remove unneeded lock in access functions. 2. Change return int to void in functions with constant 0 return value or with unchecked returned code. 3. Use regmap_update_bits() or regmap_update_bits_base() functions. 4. Remove unneeded dev_info messages. 5. Change set_timeout function. 6. Changes in probe 7. Delete unneeded remove function. --- drivers/watchdog/Kconfig | 16 +++ drivers/watchdog/Makefile | 1 + drivers/watchdog/mlx_wdt.c | 323 +++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 340 insertions(+) create mode 100644 drivers/watchdog/mlx_wdt.c diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig index 57f017d74a97..f1766eb081bb 100644 --- a/drivers/watchdog/Kconfig +++ b/drivers/watchdog/Kconfig @@ -241,6 +241,22 @@ config RAVE_SP_WATCHDOG help Support for the watchdog on RAVE SP device. +config MLX_WDT + tristate "Mellanox Watchdog" + depends on MELLANOX_PLATFORM + select WATCHDOG_CORE + select REGMAP + help + This is the driver for the hardware watchdog on Mellanox systems. + If you are going to use it, say Y here, otherwise N. + This driver can be used together with the watchdog daemon. + It can also watch your kernel to make sure it doesn't freeze, + and if it does, it reboots your system after a certain amount of + time. + + To compile this driver as a module, choose M here: the + module will be called mlx-wdt. + # ALPHA Architecture # ARM Architecture diff --git a/drivers/watchdog/Makefile b/drivers/watchdog/Makefile index a0917ef28e07..941b74185c9c 100644 --- a/drivers/watchdog/Makefile +++ b/drivers/watchdog/Makefile @@ -142,6 +142,7 @@ obj-$(CONFIG_INTEL_MID_WATCHDOG) += intel-mid_wdt.o obj-$(CONFIG_INTEL_MEI_WDT) += mei_wdt.o obj-$(CONFIG_NI903X_WDT) += ni903x_wdt.o obj-$(CONFIG_NIC7018_WDT) += nic7018_wdt.o +obj-$(CONFIG_MLX_WDT) += mlx_wdt.o # M68K Architecture obj-$(CONFIG_M54xx_WATCHDOG) += m54xx_wdt.o diff --git a/drivers/watchdog/mlx_wdt.c b/drivers/watchdog/mlx_wdt.c new file mode 100644 index 000000000000..c8fad72c9137 --- /dev/null +++ b/drivers/watchdog/mlx_wdt.c @@ -0,0 +1,323 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* + * Mellanox watchdog driver + * + * Copyright (C) 2019 Mellanox Technologies + * Copyright (C) 2019 Michael Shych <mshych@mellanox.com> + */ + +#include <linux/bitops.h> +#include <linux/device.h> +#include <linux/errno.h> +#include <linux/log2.h> +#include <linux/module.h> +#include <linux/platform_data/mlxreg.h> +#include <linux/platform_device.h> +#include <linux/regmap.h> +#include <linux/spinlock.h> +#include <linux/types.h> +#include <linux/watchdog.h> + +#define MLXREG_WDT_CLOCK_SCALE 1000 +#define MLXREG_WDT_MAX_TIMEOUT_TYPE1 32 +#define MLXREG_WDT_MAX_TIMEOUT_TYPE2 255 +#define MLXREG_WDT_MIN_TIMEOUT 1 +#define MLXREG_WDT_HW_TIMEOUT_CONVERT(hw_timeout) ((1 << (hw_timeout)) \ + / MLXREG_WDT_CLOCK_SCALE) +#define MLXREG_WDT_OPTIONS_BASE (WDIOF_KEEPALIVEPING | WDIOF_MAGICCLOSE | \ + WDIOF_SETTIMEOUT) + +/** + * enum mlxreg_wdt_type - type of HW watchdog + * + * TYPE1 can be differentiated by different register/mask + * for WD action set and ping. + */ +enum mlxreg_wdt_type { + MLX_WDT_TYPE1, + MLX_WDT_TYPE2, +}; + +/** + * struct mlxreg_wdt - wd private data: + * + * @wdd: watchdog device; + * @device: basic device; + * @pdata: data received from platform driver; + * @regmap: register map of parent device; + * @timeout: defined timeout in sec.; + * @hw_timeout: real timeout set in hw; + * It will be roundup base of 2 in WD type 1, + * in WD type 2 it will be same number of sec as timeout; + * @action_idx: index for direct access to action register; + * @timeout_idx:index for direct access to TO register; + * @tleft_idx: index for direct access to time left register; + * @ping_idx: index for direct access to ping register; + * @reset_idx: index for direct access to reset cause register; + * @wd_type: watchdog HW type; + * @hw_timeout: actual HW timeout; + */ +struct mlxreg_wdt { + struct watchdog_device wdd; + struct mlxreg_core_platform_data *pdata; + void *regmap; + int action_idx; + int timeout_idx; + int tleft_idx; + int ping_idx; + int reset_idx; + enum mlxreg_wdt_type wdt_type; + u8 hw_timeout; +}; + +static enum mlxreg_wdt_type +mlxreg_wdt_check_watchdog_type(struct mlxreg_wdt *wdt, + struct mlxreg_core_platform_data *pdata) +{ + if (pdata->data[wdt->action_idx].reg == + pdata->data[wdt->ping_idx].reg && + pdata->data[wdt->action_idx].mask == + pdata->data[wdt->ping_idx].mask) + return MLX_WDT_TYPE2; + + return MLX_WDT_TYPE1; +} + +static void mlxreg_wdt_check_card_reset(struct mlxreg_wdt *wdt) +{ + struct mlxreg_core_data *reg_data; + u32 regval; + int rc; + + if (wdt->reset_idx == -EINVAL) + return; + + if (!(wdt->wdd.info->options & WDIOF_CARDRESET)) + return; + + reg_data = &wdt->pdata->data[wdt->reset_idx]; + rc = regmap_read(wdt->regmap, reg_data->reg, ®val); + if (!rc) { + if (regval & ~reg_data->mask) { + wdt->wdd.bootstatus = WDIOF_CARDRESET; + dev_info(wdt->wdd.parent, + "watchdog previously reset the CPU\n"); + } + } +} + +static int mlxreg_wdt_start(struct watchdog_device *wdd) +{ + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt->action_idx]; + + return regmap_update_bits(wdt->regmap, reg_data->reg, ~reg_data->mask, + BIT(reg_data->bit)); +} + +static int mlxreg_wdt_stop(struct watchdog_device *wdd) +{ + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt->action_idx]; + + return regmap_update_bits(wdt->regmap, reg_data->reg, ~reg_data->mask, + ~BIT(reg_data->bit)); +} + +static int mlxreg_wdt_ping(struct watchdog_device *wdd) +{ + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt->ping_idx]; + + return regmap_update_bits_base(wdt->regmap, reg_data->reg, + ~reg_data->mask, BIT(reg_data->bit), + NULL, false, true); +} + +static int mlxreg_wdt_set_timeout(struct watchdog_device *wdd, + unsigned int timeout) +{ + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt->timeout_idx]; + u32 regval, set_time; + int rc; + + if (wdt->wdt_type == MLX_WDT_TYPE1) { + rc = regmap_read(wdt->regmap, reg_data->reg, ®val); + if (rc) + goto read_error; + + wdt->hw_timeout = order_base_2(timeout * + MLXREG_WDT_CLOCK_SCALE); + regval = (regval & reg_data->mask) | wdt->hw_timeout; + /* Rowndown to actual closest number of sec. */ + set_time = MLXREG_WDT_HW_TIMEOUT_CONVERT(wdt->hw_timeout); + } else { + wdt->hw_timeout = timeout; + set_time = timeout; + regval = timeout; + } + + wdd->timeout = set_time; + rc = regmap_write(wdt->regmap, reg_data->reg, regval); + +read_error: + if (!rc) { + /* + * Restart watchdog with new timeout period + * if watchdog is already started. + */ + if (watchdog_active(wdd)) { + rc = mlxreg_wdt_stop(wdd); + if (!rc) + rc = mlxreg_wdt_start(wdd); + } + } + + return rc; +} + +static unsigned int mlxreg_wdt_get_timeleft(struct watchdog_device *wdd) +{ + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt->tleft_idx]; + u32 regval; + int rc; + + rc = regmap_read(wdt->regmap, reg_data->reg, ®val); + + return rc == 0 ? regval : 0; +} + +static const struct watchdog_ops mlxreg_wdt_ops_type1 = { + .start = mlxreg_wdt_start, + .stop = mlxreg_wdt_stop, + .ping = mlxreg_wdt_ping, + .set_timeout = mlxreg_wdt_set_timeout, + .owner = THIS_MODULE, +}; + +static const struct watchdog_ops mlxreg_wdt_ops_type2 = { + .start = mlxreg_wdt_start, + .stop = mlxreg_wdt_stop, + .ping = mlxreg_wdt_ping, + .set_timeout = mlxreg_wdt_set_timeout, + .get_timeleft = mlxreg_wdt_get_timeleft, + .owner = THIS_MODULE, +}; + +static const struct watchdog_info mlxreg_wdt_main_info = { + .options = MLXREG_WDT_OPTIONS_BASE + | WDIOF_CARDRESET, + .identity = "mlx-wdt-main", +}; + +static const struct watchdog_info mlxreg_wdt_aux_info = { + .options = MLXREG_WDT_OPTIONS_BASE + | WDIOF_ALARMONLY, + .identity = "mlx-wdt-aux", +}; + +static void mlxreg_wdt_config(struct mlxreg_wdt *wdt, + struct mlxreg_core_platform_data *pdata) +{ + struct mlxreg_core_data *data = pdata->data; + int i; + + wdt->reset_idx = -EINVAL; + for (i = 0; i < pdata->counter; i++, data++) { + if (strnstr(data->label, "action", sizeof(data->label))) + wdt->action_idx = i; + else if (strnstr(data->label, "timeout", sizeof(data->label))) + wdt->timeout_idx = i; + else if (strnstr(data->label, "timeleft", sizeof(data->label))) + wdt->tleft_idx = i; + else if (strnstr(data->label, "ping", sizeof(data->label))) + wdt->ping_idx = i; + else if (strnstr(data->label, "reset", sizeof(data->label))) + wdt->reset_idx = i; + } + + wdt->pdata = pdata; + if (strnstr(pdata->identity, mlxreg_wdt_main_info.identity, + sizeof(mlxreg_wdt_main_info.identity))) + wdt->wdd.info = &mlxreg_wdt_main_info; + else + wdt->wdd.info = &mlxreg_wdt_aux_info; + + wdt->wdt_type = mlxreg_wdt_check_watchdog_type(wdt, pdata); + if (wdt->wdt_type == MLX_WDT_TYPE2) { + wdt->wdd.ops = &mlxreg_wdt_ops_type2; + wdt->wdd.max_timeout = MLXREG_WDT_MAX_TIMEOUT_TYPE2; + } else { + wdt->wdd.ops = &mlxreg_wdt_ops_type1; + wdt->wdd.max_timeout = MLXREG_WDT_MAX_TIMEOUT_TYPE1; + } + wdt->wdd.min_timeout = MLXREG_WDT_MIN_TIMEOUT; +} + +static int mlxreg_wdt_init_timeout(struct mlxreg_wdt *wdt, + struct mlxreg_core_platform_data *pdata) +{ + u32 timeout; + + timeout = pdata->data[wdt->timeout_idx].health_cntr; + return mlxreg_wdt_set_timeout(&wdt->wdd, timeout); +} + +static int mlxreg_wdt_probe(struct platform_device *pdev) +{ + struct mlxreg_core_platform_data *pdata; + struct mlxreg_wdt *wdt; + int rc; + + pdata = dev_get_platdata(&pdev->dev); + if (!pdata) { + dev_err(&pdev->dev, "Failed to get platform data.\n"); + return -EINVAL; + } + wdt = devm_kzalloc(&pdev->dev, sizeof(*wdt), GFP_KERNEL); + if (!wdt) + return -ENOMEM; + + wdt->wdd.parent = &pdev->dev; + wdt->regmap = pdata->regmap; + mlxreg_wdt_config(wdt, pdata); + + if ((pdata->features & MLXREG_CORE_WD_FEATURE_NOWAYOUT)) + watchdog_set_nowayout(&wdt->wdd, WATCHDOG_NOWAYOUT); + watchdog_stop_on_reboot(&wdt->wdd); + watchdog_stop_on_unregister(&wdt->wdd); + watchdog_set_drvdata(&wdt->wdd, wdt); + rc = mlxreg_wdt_init_timeout(wdt, pdata); + if (rc) + goto register_error; + + if ((pdata->features & MLXREG_CORE_WD_FEATURE_START_AT_BOOT)) { + rc = mlxreg_wdt_start(&wdt->wdd); + if (rc) + goto register_error; + } + mlxreg_wdt_check_card_reset(wdt); + rc = devm_watchdog_register_device(&pdev->dev, &wdt->wdd); + +register_error: + if (rc) + dev_err(&pdev->dev, + "Cannot register watchdog device (err=%d)\n", rc); + return rc; +} + +static struct platform_driver mlxreg_wdt_driver = { + .probe = mlxreg_wdt_probe, + .driver = { + .name = "mlx-wdt", + }, +}; + +module_platform_driver(mlxreg_wdt_driver); + +MODULE_AUTHOR("Michael Shych <michaelsh@mellanox.com>"); +MODULE_DESCRIPTION("Mellanox watchdog driver"); +MODULE_LICENSE("GPL"); +MODULE_ALIAS("platform:mlx-wdt"); -- 2.11.0 ^ permalink raw reply related [flat|nested] 8+ messages in thread
* Re: [PATCH v2 2/3] watchdog: mlx-wdt: introduce watchdog driver for Mellanox systems. 2019-02-07 21:10 ` [PATCH v2 2/3] watchdog: mlx-wdt: introduce watchdog driver for Mellanox systems michaelsh @ 2019-02-18 16:26 ` Guenter Roeck 2019-02-19 17:00 ` Michael Shych 0 siblings, 1 reply; 8+ messages in thread From: Guenter Roeck @ 2019-02-18 16:26 UTC (permalink / raw) To: michaelsh, wim, andy, dvhart; +Cc: linux-watchdog, platform-driver-x86, vadimp On 2/7/19 1:10 PM, michaelsh@mellanox.com wrote: > From: Michael Shych <michaelsh@mellanox.com> > > Introduce watchdog driver for variious rang of Mellanox Ethernet and > Infiniband switch systems. > > Watchdog driver for Mellanox watchdog devices, implemented in > programmable logic device. > > Main and auxiliary watchdog devices can exist on the same system. > There are several actions that can be defined in the watchdog: > system reset, start fans on full speed and increase a counter. > The last 2 actions are performed without a system reset. > Actions without reset are provided for auxiliary watchdog devices, > which is optional. > Access to HW registers is performed through generic > regmap interface. > > There are 2 types of HW watchdog implementations. > Type 1: actual HW timeout can be defined as power of 2 msec. > e.g. timeout 20 sec will be rounded up to 32768 msec.; > maximum timeout period is 32 sec (32768 msec.); > get time-left isn't supported > Type 2: actual HW timeout is defined in sec. and it's the same as > user-defined timeout; > maximum timeout is 255 sec; > get time-left is supported; > > Watchdog driver is probed from the common mlx_platform driver. > > Signed-off-by: Michael Shych <michaelsh@mellanox.com> > > --- > v1->v2: > Comments pointed out by Guenter: > 1. Remove unneeded lock in access functions. > 2. Change return int to void in functions with constant 0 return value > or with unchecked returned code. > 3. Use regmap_update_bits() or regmap_update_bits_base() functions. > 4. Remove unneeded dev_info messages. > 5. Change set_timeout function. > 6. Changes in probe > 7. Delete unneeded remove function. > --- > drivers/watchdog/Kconfig | 16 +++ > drivers/watchdog/Makefile | 1 + > drivers/watchdog/mlx_wdt.c | 323 +++++++++++++++++++++++++++++++++++++++++++++ > 3 files changed, 340 insertions(+) > create mode 100644 drivers/watchdog/mlx_wdt.c > > diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig > index 57f017d74a97..f1766eb081bb 100644 > --- a/drivers/watchdog/Kconfig > +++ b/drivers/watchdog/Kconfig > @@ -241,6 +241,22 @@ config RAVE_SP_WATCHDOG > help > Support for the watchdog on RAVE SP device. > > +config MLX_WDT > + tristate "Mellanox Watchdog" > + depends on MELLANOX_PLATFORM > + select WATCHDOG_CORE > + select REGMAP > + help > + This is the driver for the hardware watchdog on Mellanox systems. > + If you are going to use it, say Y here, otherwise N. > + This driver can be used together with the watchdog daemon. > + It can also watch your kernel to make sure it doesn't freeze, > + and if it does, it reboots your system after a certain amount of > + time. > + > + To compile this driver as a module, choose M here: the > + module will be called mlx-wdt. > + > # ALPHA Architecture > > # ARM Architecture > diff --git a/drivers/watchdog/Makefile b/drivers/watchdog/Makefile > index a0917ef28e07..941b74185c9c 100644 > --- a/drivers/watchdog/Makefile > +++ b/drivers/watchdog/Makefile > @@ -142,6 +142,7 @@ obj-$(CONFIG_INTEL_MID_WATCHDOG) += intel-mid_wdt.o > obj-$(CONFIG_INTEL_MEI_WDT) += mei_wdt.o > obj-$(CONFIG_NI903X_WDT) += ni903x_wdt.o > obj-$(CONFIG_NIC7018_WDT) += nic7018_wdt.o > +obj-$(CONFIG_MLX_WDT) += mlx_wdt.o > > # M68K Architecture > obj-$(CONFIG_M54xx_WATCHDOG) += m54xx_wdt.o > diff --git a/drivers/watchdog/mlx_wdt.c b/drivers/watchdog/mlx_wdt.c > new file mode 100644 > index 000000000000..c8fad72c9137 > --- /dev/null > +++ b/drivers/watchdog/mlx_wdt.c > @@ -0,0 +1,323 @@ > +// SPDX-License-Identifier: GPL-2.0+ > +/* > + * Mellanox watchdog driver > + * > + * Copyright (C) 2019 Mellanox Technologies > + * Copyright (C) 2019 Michael Shych <mshych@mellanox.com> > + */ > + > +#include <linux/bitops.h> > +#include <linux/device.h> > +#include <linux/errno.h> > +#include <linux/log2.h> > +#include <linux/module.h> > +#include <linux/platform_data/mlxreg.h> > +#include <linux/platform_device.h> > +#include <linux/regmap.h> > +#include <linux/spinlock.h> > +#include <linux/types.h> > +#include <linux/watchdog.h> > + > +#define MLXREG_WDT_CLOCK_SCALE 1000 > +#define MLXREG_WDT_MAX_TIMEOUT_TYPE1 32 > +#define MLXREG_WDT_MAX_TIMEOUT_TYPE2 255 > +#define MLXREG_WDT_MIN_TIMEOUT 1 > +#define MLXREG_WDT_HW_TIMEOUT_CONVERT(hw_timeout) ((1 << (hw_timeout)) \ > + / MLXREG_WDT_CLOCK_SCALE) A bit nitpickish: The _CONVERT and hw_ are really unnecessary, and (1 << (x) -> BIT(x). > +#define MLXREG_WDT_OPTIONS_BASE (WDIOF_KEEPALIVEPING | WDIOF_MAGICCLOSE | \ > + WDIOF_SETTIMEOUT) > + > +/** > + * enum mlxreg_wdt_type - type of HW watchdog > + * > + * TYPE1 can be differentiated by different register/mask > + * for WD action set and ping. > + */ > +enum mlxreg_wdt_type { > + MLX_WDT_TYPE1, > + MLX_WDT_TYPE2, > +}; > + > +/** > + * struct mlxreg_wdt - wd private data: > + * > + * @wdd: watchdog device; > + * @device: basic device; > + * @pdata: data received from platform driver; > + * @regmap: register map of parent device; > + * @timeout: defined timeout in sec.; > + * @hw_timeout: real timeout set in hw; > + * It will be roundup base of 2 in WD type 1, > + * in WD type 2 it will be same number of sec as timeout; > + * @action_idx: index for direct access to action register; > + * @timeout_idx:index for direct access to TO register; > + * @tleft_idx: index for direct access to time left register; > + * @ping_idx: index for direct access to ping register; > + * @reset_idx: index for direct access to reset cause register; > + * @wd_type: watchdog HW type; > + * @hw_timeout: actual HW timeout; > + */ > +struct mlxreg_wdt { > + struct watchdog_device wdd; > + struct mlxreg_core_platform_data *pdata; > + void *regmap; > + int action_idx; > + int timeout_idx; > + int tleft_idx; > + int ping_idx; > + int reset_idx; > + enum mlxreg_wdt_type wdt_type; > + u8 hw_timeout; > +}; > + > +static enum mlxreg_wdt_type > +mlxreg_wdt_check_watchdog_type(struct mlxreg_wdt *wdt, > + struct mlxreg_core_platform_data *pdata) > +{ > + if (pdata->data[wdt->action_idx].reg == > + pdata->data[wdt->ping_idx].reg && > + pdata->data[wdt->action_idx].mask == > + pdata->data[wdt->ping_idx].mask) > + return MLX_WDT_TYPE2; > + > + return MLX_WDT_TYPE1; > +} > + > +static void mlxreg_wdt_check_card_reset(struct mlxreg_wdt *wdt) > +{ > + struct mlxreg_core_data *reg_data; > + u32 regval; > + int rc; > + > + if (wdt->reset_idx == -EINVAL) > + return; > + > + if (!(wdt->wdd.info->options & WDIOF_CARDRESET)) > + return; > + > + reg_data = &wdt->pdata->data[wdt->reset_idx]; > + rc = regmap_read(wdt->regmap, reg_data->reg, ®val); > + if (!rc) { > + if (regval & ~reg_data->mask) { > + wdt->wdd.bootstatus = WDIOF_CARDRESET; > + dev_info(wdt->wdd.parent, > + "watchdog previously reset the CPU\n"); > + } > + } > +} > + > +static int mlxreg_wdt_start(struct watchdog_device *wdd) > +{ > + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); > + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt->action_idx]; > + > + return regmap_update_bits(wdt->regmap, reg_data->reg, ~reg_data->mask, > + BIT(reg_data->bit)); > +} > + > +static int mlxreg_wdt_stop(struct watchdog_device *wdd) > +{ > + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); > + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt->action_idx]; > + > + return regmap_update_bits(wdt->regmap, reg_data->reg, ~reg_data->mask, > + ~BIT(reg_data->bit)); > +} > + > +static int mlxreg_wdt_ping(struct watchdog_device *wdd) > +{ > + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); > + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt->ping_idx]; > + > + return regmap_update_bits_base(wdt->regmap, reg_data->reg, > + ~reg_data->mask, BIT(reg_data->bit), > + NULL, false, true); > +} > + > +static int mlxreg_wdt_set_timeout(struct watchdog_device *wdd, > + unsigned int timeout) > +{ > + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); > + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt->timeout_idx]; > + u32 regval, set_time; > + int rc; > + > + if (wdt->wdt_type == MLX_WDT_TYPE1) { > + rc = regmap_read(wdt->regmap, reg_data->reg, ®val); > + if (rc) > + goto read_error; > + > + wdt->hw_timeout = order_base_2(timeout * > + MLXREG_WDT_CLOCK_SCALE); > + regval = (regval & reg_data->mask) | wdt->hw_timeout; > + /* Rowndown to actual closest number of sec. */ > + set_time = MLXREG_WDT_HW_TIMEOUT_CONVERT(wdt->hw_timeout); > + } else { > + wdt->hw_timeout = timeout; > + set_time = timeout; > + regval = timeout; > + } > + > + wdd->timeout = set_time; > + rc = regmap_write(wdt->regmap, reg_data->reg, regval); > + > +read_error: This label target doesn't really make sense: rc is not 0 if it is used. Might as well return immediately above and reduce reader confusion. > + if (!rc) { > + /* > + * Restart watchdog with new timeout period > + * if watchdog is already started. > + */ > + if (watchdog_active(wdd)) { > + rc = mlxreg_wdt_stop(wdd); > + if (!rc) > + rc = mlxreg_wdt_start(wdd); > + } > + } > + > + return rc; > +} > + > +static unsigned int mlxreg_wdt_get_timeleft(struct watchdog_device *wdd) > +{ > + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); > + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt->tleft_idx]; > + u32 regval; > + int rc; > + > + rc = regmap_read(wdt->regmap, reg_data->reg, ®val); > + > + return rc == 0 ? regval : 0; > +} > + > +static const struct watchdog_ops mlxreg_wdt_ops_type1 = { > + .start = mlxreg_wdt_start, > + .stop = mlxreg_wdt_stop, > + .ping = mlxreg_wdt_ping, > + .set_timeout = mlxreg_wdt_set_timeout, > + .owner = THIS_MODULE, > +}; > + > +static const struct watchdog_ops mlxreg_wdt_ops_type2 = { > + .start = mlxreg_wdt_start, > + .stop = mlxreg_wdt_stop, > + .ping = mlxreg_wdt_ping, > + .set_timeout = mlxreg_wdt_set_timeout, > + .get_timeleft = mlxreg_wdt_get_timeleft, > + .owner = THIS_MODULE, > +}; > + > +static const struct watchdog_info mlxreg_wdt_main_info = { > + .options = MLXREG_WDT_OPTIONS_BASE > + | WDIOF_CARDRESET, > + .identity = "mlx-wdt-main", > +}; > + > +static const struct watchdog_info mlxreg_wdt_aux_info = { > + .options = MLXREG_WDT_OPTIONS_BASE > + | WDIOF_ALARMONLY, > + .identity = "mlx-wdt-aux", > +}; > + > +static void mlxreg_wdt_config(struct mlxreg_wdt *wdt, > + struct mlxreg_core_platform_data *pdata) > +{ > + struct mlxreg_core_data *data = pdata->data; > + int i; > + > + wdt->reset_idx = -EINVAL; > + for (i = 0; i < pdata->counter; i++, data++) { > + if (strnstr(data->label, "action", sizeof(data->label))) > + wdt->action_idx = i; > + else if (strnstr(data->label, "timeout", sizeof(data->label))) > + wdt->timeout_idx = i; > + else if (strnstr(data->label, "timeleft", sizeof(data->label))) > + wdt->tleft_idx = i; > + else if (strnstr(data->label, "ping", sizeof(data->label))) > + wdt->ping_idx = i; > + else if (strnstr(data->label, "reset", sizeof(data->label))) > + wdt->reset_idx = i; > + } > + > + wdt->pdata = pdata; > + if (strnstr(pdata->identity, mlxreg_wdt_main_info.identity, > + sizeof(mlxreg_wdt_main_info.identity))) > + wdt->wdd.info = &mlxreg_wdt_main_info; > + else > + wdt->wdd.info = &mlxreg_wdt_aux_info; > + > + wdt->wdt_type = mlxreg_wdt_check_watchdog_type(wdt, pdata); > + if (wdt->wdt_type == MLX_WDT_TYPE2) { > + wdt->wdd.ops = &mlxreg_wdt_ops_type2; > + wdt->wdd.max_timeout = MLXREG_WDT_MAX_TIMEOUT_TYPE2; > + } else { > + wdt->wdd.ops = &mlxreg_wdt_ops_type1; > + wdt->wdd.max_timeout = MLXREG_WDT_MAX_TIMEOUT_TYPE1; > + } > + wdt->wdd.min_timeout = MLXREG_WDT_MIN_TIMEOUT; > +} > + > +static int mlxreg_wdt_init_timeout(struct mlxreg_wdt *wdt, > + struct mlxreg_core_platform_data *pdata) > +{ > + u32 timeout; > + > + timeout = pdata->data[wdt->timeout_idx].health_cntr; > + return mlxreg_wdt_set_timeout(&wdt->wdd, timeout); > +} > + > +static int mlxreg_wdt_probe(struct platform_device *pdev) > +{ > + struct mlxreg_core_platform_data *pdata; > + struct mlxreg_wdt *wdt; > + int rc; > + > + pdata = dev_get_platdata(&pdev->dev); > + if (!pdata) { > + dev_err(&pdev->dev, "Failed to get platform data.\n"); > + return -EINVAL; > + } > + wdt = devm_kzalloc(&pdev->dev, sizeof(*wdt), GFP_KERNEL); > + if (!wdt) > + return -ENOMEM; > + > + wdt->wdd.parent = &pdev->dev; > + wdt->regmap = pdata->regmap; > + mlxreg_wdt_config(wdt, pdata); > + > + if ((pdata->features & MLXREG_CORE_WD_FEATURE_NOWAYOUT)) > + watchdog_set_nowayout(&wdt->wdd, WATCHDOG_NOWAYOUT); This coming from platform data makes me wonder how it is enabled. Is there another config flag, or is it a hardware bit ? Anotehr config flag would not really make sense since it would duplicate WATCHDOG_NOWAYOUT. > + watchdog_stop_on_reboot(&wdt->wdd); > + watchdog_stop_on_unregister(&wdt->wdd); > + watchdog_set_drvdata(&wdt->wdd, wdt); > + rc = mlxreg_wdt_init_timeout(wdt, pdata); > + if (rc) > + goto register_error; > + > + if ((pdata->features & MLXREG_CORE_WD_FEATURE_START_AT_BOOT)) { > + rc = mlxreg_wdt_start(&wdt->wdd); > + if (rc) > + goto register_error; You do have to set HW_RUNNING here to inform the watchdog core that the watchdog is running. > + } > + mlxreg_wdt_check_card_reset(wdt); > + rc = devm_watchdog_register_device(&pdev->dev, &wdt->wdd); > + > +register_error: > + if (rc) > + dev_err(&pdev->dev, > + "Cannot register watchdog device (err=%d)\n", rc); > + return rc; > +} > + > +static struct platform_driver mlxreg_wdt_driver = { > + .probe = mlxreg_wdt_probe, > + .driver = { > + .name = "mlx-wdt", > + }, > +}; > + > +module_platform_driver(mlxreg_wdt_driver); > + > +MODULE_AUTHOR("Michael Shych <michaelsh@mellanox.com>"); > +MODULE_DESCRIPTION("Mellanox watchdog driver"); > +MODULE_LICENSE("GPL"); > +MODULE_ALIAS("platform:mlx-wdt"); > ^ permalink raw reply [flat|nested] 8+ messages in thread
* RE: [PATCH v2 2/3] watchdog: mlx-wdt: introduce watchdog driver for Mellanox systems. 2019-02-18 16:26 ` Guenter Roeck @ 2019-02-19 17:00 ` Michael Shych 2019-02-19 17:46 ` Guenter Roeck 0 siblings, 1 reply; 8+ messages in thread From: Michael Shych @ 2019-02-19 17:00 UTC (permalink / raw) To: Guenter Roeck, wim, andy, dvhart Cc: linux-watchdog, platform-driver-x86, Vadim Pasternak > -----Original Message----- > From: Guenter Roeck [mailto:groeck7@gmail.com] On Behalf Of Guenter > Roeck > Sent: Monday, February 18, 2019 6:26 PM > To: Michael Shych <michaelsh@mellanox.com>; wim@linux-watchdog.org; > andy@infradead.org; dvhart@infradead.org > Cc: linux-watchdog@vger.kernel.org; platform-driver-x86@vger.kernel.org; > Vadim Pasternak <vadimp@mellanox.com> > Subject: Re: [PATCH v2 2/3] watchdog: mlx-wdt: introduce watchdog driver > for Mellanox systems. > > On 2/7/19 1:10 PM, michaelsh@mellanox.com wrote: > > From: Michael Shych <michaelsh@mellanox.com> > > > > Introduce watchdog driver for variious rang of Mellanox Ethernet and > > Infiniband switch systems. > > > > Watchdog driver for Mellanox watchdog devices, implemented in > > programmable logic device. > > > > Main and auxiliary watchdog devices can exist on the same system. > > There are several actions that can be defined in the watchdog: > > system reset, start fans on full speed and increase a counter. > > The last 2 actions are performed without a system reset. > > Actions without reset are provided for auxiliary watchdog devices, > > which is optional. > > Access to HW registers is performed through generic > > regmap interface. > > > > There are 2 types of HW watchdog implementations. > > Type 1: actual HW timeout can be defined as power of 2 msec. > > e.g. timeout 20 sec will be rounded up to 32768 msec.; > > maximum timeout period is 32 sec (32768 msec.); > > get time-left isn't supported > > Type 2: actual HW timeout is defined in sec. and it's the same as > > user-defined timeout; > > maximum timeout is 255 sec; > > get time-left is supported; > > > > Watchdog driver is probed from the common mlx_platform driver. > > > > Signed-off-by: Michael Shych <michaelsh@mellanox.com> > > > > --- > > v1->v2: > > Comments pointed out by Guenter: > > 1. Remove unneeded lock in access functions. > > 2. Change return int to void in functions with constant 0 return value > > or with unchecked returned code. > > 3. Use regmap_update_bits() or regmap_update_bits_base() functions. > > 4. Remove unneeded dev_info messages. > > 5. Change set_timeout function. > > 6. Changes in probe > > 7. Delete unneeded remove function. > > --- > > drivers/watchdog/Kconfig | 16 +++ > > drivers/watchdog/Makefile | 1 + > > drivers/watchdog/mlx_wdt.c | 323 > +++++++++++++++++++++++++++++++++++++++++++++ > > 3 files changed, 340 insertions(+) > > create mode 100644 drivers/watchdog/mlx_wdt.c > > > > diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig > > index 57f017d74a97..f1766eb081bb 100644 > > --- a/drivers/watchdog/Kconfig > > +++ b/drivers/watchdog/Kconfig > > @@ -241,6 +241,22 @@ config RAVE_SP_WATCHDOG > > help > > Support for the watchdog on RAVE SP device. > > > > +config MLX_WDT > > + tristate "Mellanox Watchdog" > > + depends on MELLANOX_PLATFORM > > + select WATCHDOG_CORE > > + select REGMAP > > + help > > + This is the driver for the hardware watchdog on Mellanox systems. > > + If you are going to use it, say Y here, otherwise N. > > + This driver can be used together with the watchdog daemon. > > + It can also watch your kernel to make sure it doesn't freeze, > > + and if it does, it reboots your system after a certain amount of > > + time. > > + > > + To compile this driver as a module, choose M here: the > > + module will be called mlx-wdt. > > + > > # ALPHA Architecture > > > > # ARM Architecture > > diff --git a/drivers/watchdog/Makefile b/drivers/watchdog/Makefile > > index a0917ef28e07..941b74185c9c 100644 > > --- a/drivers/watchdog/Makefile > > +++ b/drivers/watchdog/Makefile > > @@ -142,6 +142,7 @@ obj-$(CONFIG_INTEL_MID_WATCHDOG) += intel- > mid_wdt.o > > obj-$(CONFIG_INTEL_MEI_WDT) += mei_wdt.o > > obj-$(CONFIG_NI903X_WDT) += ni903x_wdt.o > > obj-$(CONFIG_NIC7018_WDT) += nic7018_wdt.o > > +obj-$(CONFIG_MLX_WDT) += mlx_wdt.o > > > > # M68K Architecture > > obj-$(CONFIG_M54xx_WATCHDOG) += m54xx_wdt.o > > diff --git a/drivers/watchdog/mlx_wdt.c b/drivers/watchdog/mlx_wdt.c > > new file mode 100644 > > index 000000000000..c8fad72c9137 > > --- /dev/null > > +++ b/drivers/watchdog/mlx_wdt.c > > @@ -0,0 +1,323 @@ > > +// SPDX-License-Identifier: GPL-2.0+ > > +/* > > + * Mellanox watchdog driver > > + * > > + * Copyright (C) 2019 Mellanox Technologies > > + * Copyright (C) 2019 Michael Shych <mshych@mellanox.com> > > + */ > > + > > +#include <linux/bitops.h> > > +#include <linux/device.h> > > +#include <linux/errno.h> > > +#include <linux/log2.h> > > +#include <linux/module.h> > > +#include <linux/platform_data/mlxreg.h> > > +#include <linux/platform_device.h> > > +#include <linux/regmap.h> > > +#include <linux/spinlock.h> > > +#include <linux/types.h> > > +#include <linux/watchdog.h> > > + > > +#define MLXREG_WDT_CLOCK_SCALE 1000 > > +#define MLXREG_WDT_MAX_TIMEOUT_TYPE1 32 > > +#define MLXREG_WDT_MAX_TIMEOUT_TYPE2 255 > > +#define MLXREG_WDT_MIN_TIMEOUT 1 > > +#define MLXREG_WDT_HW_TIMEOUT_CONVERT(hw_timeout) ((1 << > (hw_timeout)) \ > > + / > MLXREG_WDT_CLOCK_SCALE) > > A bit nitpickish: The _CONVERT and hw_ are really unnecessary, > and (1 << (x) -> BIT(x). Ok. Removed this define as well as hw_timeout field > > > +#define MLXREG_WDT_OPTIONS_BASE (WDIOF_KEEPALIVEPING | > WDIOF_MAGICCLOSE | \ > > + WDIOF_SETTIMEOUT) > > + > > +/** > > + * enum mlxreg_wdt_type - type of HW watchdog > > + * > > + * TYPE1 can be differentiated by different register/mask > > + * for WD action set and ping. > > + */ > > +enum mlxreg_wdt_type { > > + MLX_WDT_TYPE1, > > + MLX_WDT_TYPE2, > > +}; > > + > > +/** > > + * struct mlxreg_wdt - wd private data: > > + * > > + * @wdd: watchdog device; > > + * @device: basic device; > > + * @pdata: data received from platform driver; > > + * @regmap: register map of parent device; > > + * @timeout: defined timeout in sec.; > > + * @hw_timeout: real timeout set in hw; > > + * It will be roundup base of 2 in WD type 1, > > + * in WD type 2 it will be same number of sec as timeout; > > + * @action_idx: index for direct access to action register; > > + * @timeout_idx:index for direct access to TO register; > > + * @tleft_idx: index for direct access to time left register; > > + * @ping_idx: index for direct access to ping register; > > + * @reset_idx: index for direct access to reset cause register; > > + * @wd_type: watchdog HW type; > > + * @hw_timeout: actual HW timeout; > > + */ > > +struct mlxreg_wdt { > > + struct watchdog_device wdd; > > + struct mlxreg_core_platform_data *pdata; > > + void *regmap; > > + int action_idx; > > + int timeout_idx; > > + int tleft_idx; > > + int ping_idx; > > + int reset_idx; > > + enum mlxreg_wdt_type wdt_type; > > + u8 hw_timeout; > > +}; > > + > > +static enum mlxreg_wdt_type > > +mlxreg_wdt_check_watchdog_type(struct mlxreg_wdt *wdt, > > + struct mlxreg_core_platform_data *pdata) > > +{ > > + if (pdata->data[wdt->action_idx].reg == > > + pdata->data[wdt->ping_idx].reg && > > + pdata->data[wdt->action_idx].mask == > > + pdata->data[wdt->ping_idx].mask) > > + return MLX_WDT_TYPE2; > > + > > + return MLX_WDT_TYPE1; > > +} > > + > > +static void mlxreg_wdt_check_card_reset(struct mlxreg_wdt *wdt) > > +{ > > + struct mlxreg_core_data *reg_data; > > + u32 regval; > > + int rc; > > + > > + if (wdt->reset_idx == -EINVAL) > > + return; > > + > > + if (!(wdt->wdd.info->options & WDIOF_CARDRESET)) > > + return; > > + > > + reg_data = &wdt->pdata->data[wdt->reset_idx]; > > + rc = regmap_read(wdt->regmap, reg_data->reg, ®val); > > + if (!rc) { > > + if (regval & ~reg_data->mask) { > > + wdt->wdd.bootstatus = WDIOF_CARDRESET; > > + dev_info(wdt->wdd.parent, > > + "watchdog previously reset the CPU\n"); > > + } > > + } > > +} > > + > > +static int mlxreg_wdt_start(struct watchdog_device *wdd) > > +{ > > + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); > > + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt- > >action_idx]; > > + > > + return regmap_update_bits(wdt->regmap, reg_data->reg, > ~reg_data->mask, > > + BIT(reg_data->bit)); > > +} > > + > > +static int mlxreg_wdt_stop(struct watchdog_device *wdd) > > +{ > > + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); > > + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt- > >action_idx]; > > + > > + return regmap_update_bits(wdt->regmap, reg_data->reg, > ~reg_data->mask, > > + ~BIT(reg_data->bit)); > > +} > > + > > +static int mlxreg_wdt_ping(struct watchdog_device *wdd) > > +{ > > + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); > > + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt- > >ping_idx]; > > + > > + return regmap_update_bits_base(wdt->regmap, reg_data->reg, > > + ~reg_data->mask, BIT(reg_data->bit), > > + NULL, false, true); > > +} > > + > > +static int mlxreg_wdt_set_timeout(struct watchdog_device *wdd, > > + unsigned int timeout) > > +{ > > + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); > > + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt- > >timeout_idx]; > > + u32 regval, set_time; > > + int rc; > > + > > + if (wdt->wdt_type == MLX_WDT_TYPE1) { > > + rc = regmap_read(wdt->regmap, reg_data->reg, ®val); > > + if (rc) > > + goto read_error; > > + > > + wdt->hw_timeout = order_base_2(timeout * > > + MLXREG_WDT_CLOCK_SCALE); > > + regval = (regval & reg_data->mask) | wdt->hw_timeout; > > + /* Rowndown to actual closest number of sec. */ > > + set_time = MLXREG_WDT_HW_TIMEOUT_CONVERT(wdt- > >hw_timeout); > > + } else { > > + wdt->hw_timeout = timeout; > > + set_time = timeout; > > + regval = timeout; > > + } > > + > > + wdd->timeout = set_time; > > + rc = regmap_write(wdt->regmap, reg_data->reg, regval); > > + > > +read_error: > > This label target doesn't really make sense: rc is not 0 if it is used. > Might as well return immediately above and reduce reader confusion. > Ok. Changed. > > + if (!rc) { > > + /* > > + * Restart watchdog with new timeout period > > + * if watchdog is already started. > > + */ > > + if (watchdog_active(wdd)) { > > + rc = mlxreg_wdt_stop(wdd); > > + if (!rc) > > + rc = mlxreg_wdt_start(wdd); > > + } > > + } > > + > > + return rc; > > +} > > + > > +static unsigned int mlxreg_wdt_get_timeleft(struct watchdog_device > *wdd) > > +{ > > + struct mlxreg_wdt *wdt = watchdog_get_drvdata(wdd); > > + struct mlxreg_core_data *reg_data = &wdt->pdata->data[wdt- > >tleft_idx]; > > + u32 regval; > > + int rc; > > + > > + rc = regmap_read(wdt->regmap, reg_data->reg, ®val); > > + > > + return rc == 0 ? regval : 0; > > +} > > + > > +static const struct watchdog_ops mlxreg_wdt_ops_type1 = { > > + .start = mlxreg_wdt_start, > > + .stop = mlxreg_wdt_stop, > > + .ping = mlxreg_wdt_ping, > > + .set_timeout = mlxreg_wdt_set_timeout, > > + .owner = THIS_MODULE, > > +}; > > + > > +static const struct watchdog_ops mlxreg_wdt_ops_type2 = { > > + .start = mlxreg_wdt_start, > > + .stop = mlxreg_wdt_stop, > > + .ping = mlxreg_wdt_ping, > > + .set_timeout = mlxreg_wdt_set_timeout, > > + .get_timeleft = mlxreg_wdt_get_timeleft, > > + .owner = THIS_MODULE, > > +}; > > + > > +static const struct watchdog_info mlxreg_wdt_main_info = { > > + .options = MLXREG_WDT_OPTIONS_BASE > > + | WDIOF_CARDRESET, > > + .identity = "mlx-wdt-main", > > +}; > > + > > +static const struct watchdog_info mlxreg_wdt_aux_info = { > > + .options = MLXREG_WDT_OPTIONS_BASE > > + | WDIOF_ALARMONLY, > > + .identity = "mlx-wdt-aux", > > +}; > > + > > +static void mlxreg_wdt_config(struct mlxreg_wdt *wdt, > > + struct mlxreg_core_platform_data *pdata) > > +{ > > + struct mlxreg_core_data *data = pdata->data; > > + int i; > > + > > + wdt->reset_idx = -EINVAL; > > + for (i = 0; i < pdata->counter; i++, data++) { > > + if (strnstr(data->label, "action", sizeof(data->label))) > > + wdt->action_idx = i; > > + else if (strnstr(data->label, "timeout", sizeof(data->label))) > > + wdt->timeout_idx = i; > > + else if (strnstr(data->label, "timeleft", sizeof(data->label))) > > + wdt->tleft_idx = i; > > + else if (strnstr(data->label, "ping", sizeof(data->label))) > > + wdt->ping_idx = i; > > + else if (strnstr(data->label, "reset", sizeof(data->label))) > > + wdt->reset_idx = i; > > + } > > + > > + wdt->pdata = pdata; > > + if (strnstr(pdata->identity, mlxreg_wdt_main_info.identity, > > + sizeof(mlxreg_wdt_main_info.identity))) > > + wdt->wdd.info = &mlxreg_wdt_main_info; > > + else > > + wdt->wdd.info = &mlxreg_wdt_aux_info; > > + > > + wdt->wdt_type = mlxreg_wdt_check_watchdog_type(wdt, pdata); > > + if (wdt->wdt_type == MLX_WDT_TYPE2) { > > + wdt->wdd.ops = &mlxreg_wdt_ops_type2; > > + wdt->wdd.max_timeout = > MLXREG_WDT_MAX_TIMEOUT_TYPE2; > > + } else { > > + wdt->wdd.ops = &mlxreg_wdt_ops_type1; > > + wdt->wdd.max_timeout = > MLXREG_WDT_MAX_TIMEOUT_TYPE1; > > + } > > + wdt->wdd.min_timeout = MLXREG_WDT_MIN_TIMEOUT; > > +} > > + > > +static int mlxreg_wdt_init_timeout(struct mlxreg_wdt *wdt, > > + struct mlxreg_core_platform_data *pdata) > > +{ > > + u32 timeout; > > + > > + timeout = pdata->data[wdt->timeout_idx].health_cntr; > > + return mlxreg_wdt_set_timeout(&wdt->wdd, timeout); > > +} > > + > > +static int mlxreg_wdt_probe(struct platform_device *pdev) > > +{ > > + struct mlxreg_core_platform_data *pdata; > > + struct mlxreg_wdt *wdt; > > + int rc; > > + > > + pdata = dev_get_platdata(&pdev->dev); > > + if (!pdata) { > > + dev_err(&pdev->dev, "Failed to get platform data.\n"); > > + return -EINVAL; > > + } > > + wdt = devm_kzalloc(&pdev->dev, sizeof(*wdt), GFP_KERNEL); > > + if (!wdt) > > + return -ENOMEM; > > + > > + wdt->wdd.parent = &pdev->dev; > > + wdt->regmap = pdata->regmap; > > + mlxreg_wdt_config(wdt, pdata); > > + > > + if ((pdata->features & MLXREG_CORE_WD_FEATURE_NOWAYOUT)) > > + watchdog_set_nowayout(&wdt->wdd, > WATCHDOG_NOWAYOUT); > > This coming from platform data makes me wonder how it is enabled. > Is there another config flag, or is it a hardware bit ? Anotehr config flag > would not really make sense since it would duplicate > WATCHDOG_NOWAYOUT. Probably I don't understand your concern here. The aim is to disable demon to stop watchdog timer when it was started. Tested and it works as desired: " [ 200.589339] watchdog: watchdog0: nowayout prevents watchdog being stopped! [ 200.597056] watchdog: watchdog0: watchdog did not stop! " MLXREG_CORE_WD_FEATURE_NOWAYOUT doesn't duplicate exactly WATCHDOG_NOWAYOUT. It's passed from parent platform driver in common features bitfield. It can be passed together with START_AT_BOOT bit. > > > + watchdog_stop_on_reboot(&wdt->wdd); > > + watchdog_stop_on_unregister(&wdt->wdd); > > + watchdog_set_drvdata(&wdt->wdd, wdt); > > + rc = mlxreg_wdt_init_timeout(wdt, pdata); > > + if (rc) > > + goto register_error; > > + > > + if ((pdata->features & > MLXREG_CORE_WD_FEATURE_START_AT_BOOT)) { > > + rc = mlxreg_wdt_start(&wdt->wdd); > > + if (rc) > > + goto register_error; > > You do have to set HW_RUNNING here to inform the watchdog core that > the watchdog is running. I understood from watchdog-kernel-api.txt document that this is not required In our case. " * WDOG_HW_RUNNING: Set by the watchdog driver if the hardware watchdog is running. The bit must be set if the watchdog timer hardware can not be stopped. The bit may also be set if the watchdog timer is running after booting, before the watchdog device is opened. If set, the watchdog infrastructure will send keepalives to the watchdog hardware while WDOG_ACTIVE is not set." The 2nd and 3rd sentences from this quotation aren't true for our watchdog. Does it required anyway in spite that these cases are false? I added set of WDOG_HW_RUNNING but I'm not sure that it's required. > > > + } > > + mlxreg_wdt_check_card_reset(wdt); > > + rc = devm_watchdog_register_device(&pdev->dev, &wdt->wdd); > > + > > +register_error: > > + if (rc) > > + dev_err(&pdev->dev, > > + "Cannot register watchdog device (err=%d)\n", rc); > > + return rc; > > +} > > + > > +static struct platform_driver mlxreg_wdt_driver = { > > + .probe = mlxreg_wdt_probe, > > + .driver = { > > + .name = "mlx-wdt", > > + }, > > +}; > > + > > +module_platform_driver(mlxreg_wdt_driver); > > + > > +MODULE_AUTHOR("Michael Shych <michaelsh@mellanox.com>"); > > +MODULE_DESCRIPTION("Mellanox watchdog driver"); > > +MODULE_LICENSE("GPL"); > > +MODULE_ALIAS("platform:mlx-wdt"); > > Except that, Vadim forward me your remark about mlxreg_wdt_check_watchdog_type. I removed it and added common field version that will be passed from parent platform driver. Thank you, Michael. ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH v2 2/3] watchdog: mlx-wdt: introduce watchdog driver for Mellanox systems. 2019-02-19 17:00 ` Michael Shych @ 2019-02-19 17:46 ` Guenter Roeck 0 siblings, 0 replies; 8+ messages in thread From: Guenter Roeck @ 2019-02-19 17:46 UTC (permalink / raw) To: Michael Shych, wim, andy, dvhart Cc: linux-watchdog, platform-driver-x86, Vadim Pasternak On 2/19/19 9:00 AM, Michael Shych wrote: >>> + watchdog_stop_on_reboot(&wdt->wdd); >>> + watchdog_stop_on_unregister(&wdt->wdd); >>> + watchdog_set_drvdata(&wdt->wdd, wdt); >>> + rc = mlxreg_wdt_init_timeout(wdt, pdata); >>> + if (rc) >>> + goto register_error; >>> + >>> + if ((pdata->features & >> MLXREG_CORE_WD_FEATURE_START_AT_BOOT)) { >>> + rc = mlxreg_wdt_start(&wdt->wdd); >>> + if (rc) >>> + goto register_error; >> >> You do have to set HW_RUNNING here to inform the watchdog core that >> the watchdog is running. > > I understood from watchdog-kernel-api.txt document that this is not required > In our case. > " * WDOG_HW_RUNNING: Set by the watchdog driver if the hardware watchdog is > running. The bit must be set if the watchdog timer hardware can not be > stopped. The bit may also be set if the watchdog timer is running after > booting, before the watchdog device is opened. If set, the watchdog > infrastructure will send keepalives to the watchdog hardware while > WDOG_ACTIVE is not set." > The 2nd and 3rd sentences from this quotation aren't true for our watchdog. > Does it required anyway in spite that these cases are false? > I added set of WDOG_HW_RUNNING but I'm not sure that it's required. > So you want the watchdog to time out and cause a reboot after the initial watchdog period has expired and the watchdog device has not been opened ? Fine with me, just please add a respective comment into the driver. Thanks, Guenter ^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH v2 3/3] Documentation/watchdog: Add documentation mlx-wdt driver 2019-02-07 21:10 [PATCH v2 0/3] introduce watchdog driver for Mellanox systems michaelsh 2019-02-07 21:10 ` [PATCH v2 1/3] platform_data/mlxreg: addittions for mellanox watchdog driver michaelsh 2019-02-07 21:10 ` [PATCH v2 2/3] watchdog: mlx-wdt: introduce watchdog driver for Mellanox systems michaelsh @ 2019-02-07 21:10 ` michaelsh 2 siblings, 0 replies; 8+ messages in thread From: michaelsh @ 2019-02-07 21:10 UTC (permalink / raw) To: wim, linux, andy, dvhart Cc: linux-watchdog, platform-driver-x86, vadimp, Michael Shych From: Michael Shych <michaelsh@mellanox.com> Add documentation with details of mellanox watchdog driver. Signed-off-by: Michael Shych <michaelsh@mellanox.com> --- Documentation/watchdog/mlx-wdt.txt | 48 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 48 insertions(+) create mode 100644 Documentation/watchdog/mlx-wdt.txt diff --git a/Documentation/watchdog/mlx-wdt.txt b/Documentation/watchdog/mlx-wdt.txt new file mode 100644 index 000000000000..f9e57f9aa530 --- /dev/null +++ b/Documentation/watchdog/mlx-wdt.txt @@ -0,0 +1,48 @@ + Mellanox watchdog drivers + for x86 based system switches + +This driver provides watchdog functionality for various Mellanox +Ethernet and Infiniband switch systems. + +Mellanox watchdog device is implemented in a programmable logic device. + +There are 2 types of HW watchdog implementations. + +Type 1: +Actual HW timeout can be defined as a power of 2 msec. +e.g. timeout 20 sec will be rounded up to 32768 msec. +The maximum timeout period is 32 sec (32768 msec.), +Get time-left isn't supported + +Type 2: +Actual HW timeout is defined in sec. and it's the same as +a user-defined timeout. +Maximum timeout is 255 sec. +Get time-left is supported. + +Type 1 HW watchdog implementation exist in old systems and +all new systems have type 2 HW watchdog. +Two types of HW implementation have also different register map. + +Mellanox system can have 2 watchdogs: main and auxiliary. +Main and auxiliary watchdog devices can be enabled together +on the same system. +There are several actions that can be defined in the watchdog: +system reset, start fans on full speed and increase register counter. +The last 2 actions are performed without a system reset. +Actions without reset are provided for auxiliary watchdog device, +which is optional. + +This mlx-wdt driver supports both HW watchdog implementations. + +Watchdog driver is probed from the common mlx_platform driver. +Mlx_platform driver provides an appropriate set of registers for +Mellanox watchdog device, identity name (mlx-wdt-main or mlx-wdt-aux), +initial timeout, performed action in expiration and configuration flags. +watchdog configuration flags: nostop_after_start i.e. nowayout and +start_at_boot. +The driver reports if watchdog set timeout is rounded in type1 HW watchdog. +The driver checks during initialization if the previous system reset +was done by the watchdog. If yes, it makes a notification about this event. + +Access to HW registers is performed through a generic regmap interface. -- 2.11.0 ^ permalink raw reply related [flat|nested] 8+ messages in thread
end of thread, other threads:[~2019-02-19 17:46 UTC | newest] Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2019-02-07 21:10 [PATCH v2 0/3] introduce watchdog driver for Mellanox systems michaelsh 2019-02-07 21:10 ` [PATCH v2 1/3] platform_data/mlxreg: addittions for mellanox watchdog driver michaelsh 2019-02-18 16:12 ` Guenter Roeck 2019-02-07 21:10 ` [PATCH v2 2/3] watchdog: mlx-wdt: introduce watchdog driver for Mellanox systems michaelsh 2019-02-18 16:26 ` Guenter Roeck 2019-02-19 17:00 ` Michael Shych 2019-02-19 17:46 ` Guenter Roeck 2019-02-07 21:10 ` [PATCH v2 3/3] Documentation/watchdog: Add documentation mlx-wdt driver michaelsh
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).