| // SPDX-License-Identifier: GPL-2.0-only |
| /* |
| * Intel_SCU 0.2: An Intel SCU IOH Based Watchdog Device |
| * for Intel part #(s): |
| * - AF82MP20 PCH |
| * |
| * Copyright (C) 2009-2010 Intel Corporation. All rights reserved. |
| */ |
| |
| #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
| |
| #include <linux/compiler.h> |
| #include <linux/kernel.h> |
| #include <linux/moduleparam.h> |
| #include <linux/types.h> |
| #include <linux/miscdevice.h> |
| #include <linux/watchdog.h> |
| #include <linux/fs.h> |
| #include <linux/notifier.h> |
| #include <linux/reboot.h> |
| #include <linux/init.h> |
| #include <linux/jiffies.h> |
| #include <linux/uaccess.h> |
| #include <linux/slab.h> |
| #include <linux/io.h> |
| #include <linux/interrupt.h> |
| #include <linux/delay.h> |
| #include <linux/sched.h> |
| #include <linux/signal.h> |
| #include <linux/sfi.h> |
| #include <asm/irq.h> |
| #include <linux/atomic.h> |
| #include <asm/intel_scu_ipc.h> |
| #include <asm/apb_timer.h> |
| #include <asm/intel-mid.h> |
| |
| #include "intel_scu_watchdog.h" |
| |
| /* Bounds number of times we will retry loading time count */ |
| /* This retry is a work around for a silicon bug. */ |
| #define MAX_RETRY 16 |
| |
| #define IPC_SET_WATCHDOG_TIMER 0xF8 |
| |
| static int timer_margin = DEFAULT_SOFT_TO_HARD_MARGIN; |
| module_param(timer_margin, int, 0); |
| MODULE_PARM_DESC(timer_margin, |
| "Watchdog timer margin" |
| "Time between interrupt and resetting the system" |
| "The range is from 1 to 160" |
| "This is the time for all keep alives to arrive"); |
| |
| static int timer_set = DEFAULT_TIME; |
| module_param(timer_set, int, 0); |
| MODULE_PARM_DESC(timer_set, |
| "Default Watchdog timer setting" |
| "Complete cycle time" |
| "The range is from 1 to 170" |
| "This is the time for all keep alives to arrive"); |
| |
| /* After watchdog device is closed, check force_boot. If: |
| * force_boot == 0, then force boot on next watchdog interrupt after close, |
| * force_boot == 1, then force boot immediately when device is closed. |
| */ |
| static int force_boot; |
| module_param(force_boot, int, 0); |
| MODULE_PARM_DESC(force_boot, |
| "A value of 1 means that the driver will reboot" |
| "the system immediately if the /dev/watchdog device is closed" |
| "A value of 0 means that when /dev/watchdog device is closed" |
| "the watchdog timer will be refreshed for one more interval" |
| "of length: timer_set. At the end of this interval, the" |
| "watchdog timer will reset the system." |
| ); |
| |
| /* there is only one device in the system now; this can be made into |
| * an array in the future if we have more than one device */ |
| |
| static struct intel_scu_watchdog_dev watchdog_device; |
| |
| /* Forces restart, if force_reboot is set */ |
| static void watchdog_fire(void) |
| { |
| if (force_boot) { |
| pr_crit("Initiating system reboot\n"); |
| emergency_restart(); |
| pr_crit("Reboot didn't ?????\n"); |
| } |
| |
| else { |
| pr_crit("Immediate Reboot Disabled\n"); |
| pr_crit("System will reset when watchdog timer times out!\n"); |
| } |
| } |
| |
| static int check_timer_margin(int new_margin) |
| { |
| if ((new_margin < MIN_TIME_CYCLE) || |
| (new_margin > MAX_TIME - timer_set)) { |
| pr_debug("value of new_margin %d is out of the range %d to %d\n", |
| new_margin, MIN_TIME_CYCLE, MAX_TIME - timer_set); |
| return -EINVAL; |
| } |
| return 0; |
| } |
| |
| /* |
| * IPC operations |
| */ |
| static int watchdog_set_ipc(int soft_threshold, int threshold) |
| { |
| u32 *ipc_wbuf; |
| u8 cbuf[16] = { '\0' }; |
| int ipc_ret = 0; |
| |
| ipc_wbuf = (u32 *)&cbuf; |
| ipc_wbuf[0] = soft_threshold; |
| ipc_wbuf[1] = threshold; |
| |
| ipc_ret = intel_scu_ipc_command( |
| IPC_SET_WATCHDOG_TIMER, |
| 0, |
| ipc_wbuf, |
| 2, |
| NULL, |
| 0); |
| |
| if (ipc_ret != 0) |
| pr_err("Error setting SCU watchdog timer: %x\n", ipc_ret); |
| |
| return ipc_ret; |
| }; |
| |
| /* |
| * Intel_SCU operations |
| */ |
| |
| /* timer interrupt handler */ |
| static irqreturn_t watchdog_timer_interrupt(int irq, void *dev_id) |
| { |
| int int_status; |
| int_status = ioread32(watchdog_device.timer_interrupt_status_addr); |
| |
| pr_debug("irq, int_status: %x\n", int_status); |
| |
| if (int_status != 0) |
| return IRQ_NONE; |
| |
| /* has the timer been started? If not, then this is spurious */ |
| if (watchdog_device.timer_started == 0) { |
| pr_debug("spurious interrupt received\n"); |
| return IRQ_HANDLED; |
| } |
| |
| /* temporarily disable the timer */ |
| iowrite32(0x00000002, watchdog_device.timer_control_addr); |
| |
| /* set the timer to the threshold */ |
| iowrite32(watchdog_device.threshold, |
| watchdog_device.timer_load_count_addr); |
| |
| /* allow the timer to run */ |
| iowrite32(0x00000003, watchdog_device.timer_control_addr); |
| |
| return IRQ_HANDLED; |
| } |
| |
| static int intel_scu_keepalive(void) |
| { |
| |
| /* read eoi register - clears interrupt */ |
| ioread32(watchdog_device.timer_clear_interrupt_addr); |
| |
| /* temporarily disable the timer */ |
| iowrite32(0x00000002, watchdog_device.timer_control_addr); |
| |
| /* set the timer to the soft_threshold */ |
| iowrite32(watchdog_device.soft_threshold, |
| watchdog_device.timer_load_count_addr); |
| |
| /* allow the timer to run */ |
| iowrite32(0x00000003, watchdog_device.timer_control_addr); |
| |
| return 0; |
| } |
| |
| static int intel_scu_stop(void) |
| { |
| iowrite32(0, watchdog_device.timer_control_addr); |
| return 0; |
| } |
| |
| static int intel_scu_set_heartbeat(u32 t) |
| { |
| int ipc_ret; |
| int retry_count; |
| u32 soft_value; |
| u32 hw_value; |
| |
| watchdog_device.timer_set = t; |
| watchdog_device.threshold = |
| timer_margin * watchdog_device.timer_tbl_ptr->freq_hz; |
| watchdog_device.soft_threshold = |
| (watchdog_device.timer_set - timer_margin) |
| * watchdog_device.timer_tbl_ptr->freq_hz; |
| |
| pr_debug("set_heartbeat: timer freq is %d\n", |
| watchdog_device.timer_tbl_ptr->freq_hz); |
| pr_debug("set_heartbeat: timer_set is %x (hex)\n", |
| watchdog_device.timer_set); |
| pr_debug("set_heartbeat: timer_margin is %x (hex)\n", timer_margin); |
| pr_debug("set_heartbeat: threshold is %x (hex)\n", |
| watchdog_device.threshold); |
| pr_debug("set_heartbeat: soft_threshold is %x (hex)\n", |
| watchdog_device.soft_threshold); |
| |
| /* Adjust thresholds by FREQ_ADJUSTMENT factor, to make the */ |
| /* watchdog timing come out right. */ |
| watchdog_device.threshold = |
| watchdog_device.threshold / FREQ_ADJUSTMENT; |
| watchdog_device.soft_threshold = |
| watchdog_device.soft_threshold / FREQ_ADJUSTMENT; |
| |
| /* temporarily disable the timer */ |
| iowrite32(0x00000002, watchdog_device.timer_control_addr); |
| |
| /* send the threshold and soft_threshold via IPC to the processor */ |
| ipc_ret = watchdog_set_ipc(watchdog_device.soft_threshold, |
| watchdog_device.threshold); |
| |
| if (ipc_ret != 0) { |
| /* Make sure the watchdog timer is stopped */ |
| intel_scu_stop(); |
| return ipc_ret; |
| } |
| |
| /* Soft Threshold set loop. Early versions of silicon did */ |
| /* not always set this count correctly. This loop checks */ |
| /* the value and retries if it was not set correctly. */ |
| |
| retry_count = 0; |
| soft_value = watchdog_device.soft_threshold & 0xFFFF0000; |
| do { |
| |
| /* Make sure timer is stopped */ |
| intel_scu_stop(); |
| |
| if (MAX_RETRY < retry_count++) { |
| /* Unable to set timer value */ |
| pr_err("Unable to set timer\n"); |
| return -ENODEV; |
| } |
| |
| /* set the timer to the soft threshold */ |
| iowrite32(watchdog_device.soft_threshold, |
| watchdog_device.timer_load_count_addr); |
| |
| /* read count value before starting timer */ |
| ioread32(watchdog_device.timer_load_count_addr); |
| |
| /* Start the timer */ |
| iowrite32(0x00000003, watchdog_device.timer_control_addr); |
| |
| /* read the value the time loaded into its count reg */ |
| hw_value = ioread32(watchdog_device.timer_load_count_addr); |
| hw_value = hw_value & 0xFFFF0000; |
| |
| |
| } while (soft_value != hw_value); |
| |
| watchdog_device.timer_started = 1; |
| |
| return 0; |
| } |
| |
| /* |
| * /dev/watchdog handling |
| */ |
| |
| static int intel_scu_open(struct inode *inode, struct file *file) |
| { |
| |
| /* Set flag to indicate that watchdog device is open */ |
| if (test_and_set_bit(0, &watchdog_device.driver_open)) |
| return -EBUSY; |
| |
| /* Check for reopen of driver. Reopens are not allowed */ |
| if (watchdog_device.driver_closed) |
| return -EPERM; |
| |
| return stream_open(inode, file); |
| } |
| |
| static int intel_scu_release(struct inode *inode, struct file *file) |
| { |
| /* |
| * This watchdog should not be closed, after the timer |
| * is started with the WDIPC_SETTIMEOUT ioctl |
| * If force_boot is set watchdog_fire() will cause an |
| * immediate reset. If force_boot is not set, the watchdog |
| * timer is refreshed for one more interval. At the end |
| * of that interval, the watchdog timer will reset the system. |
| */ |
| |
| if (!test_and_clear_bit(0, &watchdog_device.driver_open)) { |
| pr_debug("intel_scu_release, without open\n"); |
| return -ENOTTY; |
| } |
| |
| if (!watchdog_device.timer_started) { |
| /* Just close, since timer has not been started */ |
| pr_debug("closed, without starting timer\n"); |
| return 0; |
| } |
| |
| pr_crit("Unexpected close of /dev/watchdog!\n"); |
| |
| /* Since the timer was started, prevent future reopens */ |
| watchdog_device.driver_closed = 1; |
| |
| /* Refresh the timer for one more interval */ |
| intel_scu_keepalive(); |
| |
| /* Reboot system (if force_boot is set) */ |
| watchdog_fire(); |
| |
| /* We should only reach this point if force_boot is not set */ |
| return 0; |
| } |
| |
| static ssize_t intel_scu_write(struct file *file, |
| char const *data, |
| size_t len, |
| loff_t *ppos) |
| { |
| |
| if (watchdog_device.timer_started) |
| /* Watchdog already started, keep it alive */ |
| intel_scu_keepalive(); |
| else |
| /* Start watchdog with timer value set by init */ |
| intel_scu_set_heartbeat(watchdog_device.timer_set); |
| |
| return len; |
| } |
| |
| static long intel_scu_ioctl(struct file *file, |
| unsigned int cmd, |
| unsigned long arg) |
| { |
| void __user *argp = (void __user *)arg; |
| u32 __user *p = argp; |
| u32 new_margin; |
| |
| |
| static const struct watchdog_info ident = { |
| .options = WDIOF_SETTIMEOUT |
| | WDIOF_KEEPALIVEPING, |
| .firmware_version = 0, /* @todo Get from SCU via |
| ipc_get_scu_fw_version()? */ |
| .identity = "Intel_SCU IOH Watchdog" /* len < 32 */ |
| }; |
| |
| switch (cmd) { |
| case WDIOC_GETSUPPORT: |
| return copy_to_user(argp, |
| &ident, |
| sizeof(ident)) ? -EFAULT : 0; |
| case WDIOC_GETSTATUS: |
| case WDIOC_GETBOOTSTATUS: |
| return put_user(0, p); |
| case WDIOC_KEEPALIVE: |
| intel_scu_keepalive(); |
| |
| return 0; |
| case WDIOC_SETTIMEOUT: |
| if (get_user(new_margin, p)) |
| return -EFAULT; |
| |
| if (check_timer_margin(new_margin)) |
| return -EINVAL; |
| |
| if (intel_scu_set_heartbeat(new_margin)) |
| return -EINVAL; |
| return 0; |
| case WDIOC_GETTIMEOUT: |
| return put_user(watchdog_device.soft_threshold, p); |
| |
| default: |
| return -ENOTTY; |
| } |
| } |
| |
| /* |
| * Notifier for system down |
| */ |
| static int intel_scu_notify_sys(struct notifier_block *this, |
| unsigned long code, |
| void *another_unused) |
| { |
| if (code == SYS_DOWN || code == SYS_HALT) |
| /* Turn off the watchdog timer. */ |
| intel_scu_stop(); |
| return NOTIFY_DONE; |
| } |
| |
| /* |
| * Kernel Interfaces |
| */ |
| static const struct file_operations intel_scu_fops = { |
| .owner = THIS_MODULE, |
| .llseek = no_llseek, |
| .write = intel_scu_write, |
| .unlocked_ioctl = intel_scu_ioctl, |
| .compat_ioctl = compat_ptr_ioctl, |
| .open = intel_scu_open, |
| .release = intel_scu_release, |
| }; |
| |
| static int __init intel_scu_watchdog_init(void) |
| { |
| int ret; |
| u32 __iomem *tmp_addr; |
| |
| /* |
| * We don't really need to check this as the SFI timer get will fail |
| * but if we do so we can exit with a clearer reason and no noise. |
| * |
| * If it isn't an intel MID device then it doesn't have this watchdog |
| */ |
| if (!intel_mid_identify_cpu()) |
| return -ENODEV; |
| |
| /* Check boot parameters to verify that their initial values */ |
| /* are in range. */ |
| /* Check value of timer_set boot parameter */ |
| if ((timer_set < MIN_TIME_CYCLE) || |
| (timer_set > MAX_TIME - MIN_TIME_CYCLE)) { |
| pr_err("value of timer_set %x (hex) is out of range from %x to %x (hex)\n", |
| timer_set, MIN_TIME_CYCLE, MAX_TIME - MIN_TIME_CYCLE); |
| return -EINVAL; |
| } |
| |
| /* Check value of timer_margin boot parameter */ |
| if (check_timer_margin(timer_margin)) |
| return -EINVAL; |
| |
| watchdog_device.timer_tbl_ptr = sfi_get_mtmr(sfi_mtimer_num-1); |
| |
| if (watchdog_device.timer_tbl_ptr == NULL) { |
| pr_debug("timer is not available\n"); |
| return -ENODEV; |
| } |
| /* make sure the timer exists */ |
| if (watchdog_device.timer_tbl_ptr->phys_addr == 0) { |
| pr_debug("timer %d does not have valid physical memory\n", |
| sfi_mtimer_num); |
| return -ENODEV; |
| } |
| |
| if (watchdog_device.timer_tbl_ptr->irq == 0) { |
| pr_debug("timer %d invalid irq\n", sfi_mtimer_num); |
| return -ENODEV; |
| } |
| |
| tmp_addr = ioremap(watchdog_device.timer_tbl_ptr->phys_addr, |
| 20); |
| |
| if (tmp_addr == NULL) { |
| pr_debug("timer unable to ioremap\n"); |
| return -ENOMEM; |
| } |
| |
| watchdog_device.timer_load_count_addr = tmp_addr++; |
| watchdog_device.timer_current_value_addr = tmp_addr++; |
| watchdog_device.timer_control_addr = tmp_addr++; |
| watchdog_device.timer_clear_interrupt_addr = tmp_addr++; |
| watchdog_device.timer_interrupt_status_addr = tmp_addr++; |
| |
| /* Set the default time values in device structure */ |
| |
| watchdog_device.timer_set = timer_set; |
| watchdog_device.threshold = |
| timer_margin * watchdog_device.timer_tbl_ptr->freq_hz; |
| watchdog_device.soft_threshold = |
| (watchdog_device.timer_set - timer_margin) |
| * watchdog_device.timer_tbl_ptr->freq_hz; |
| |
| |
| watchdog_device.intel_scu_notifier.notifier_call = |
| intel_scu_notify_sys; |
| |
| ret = register_reboot_notifier(&watchdog_device.intel_scu_notifier); |
| if (ret) { |
| pr_err("cannot register notifier %d)\n", ret); |
| goto register_reboot_error; |
| } |
| |
| watchdog_device.miscdev.minor = WATCHDOG_MINOR; |
| watchdog_device.miscdev.name = "watchdog"; |
| watchdog_device.miscdev.fops = &intel_scu_fops; |
| |
| ret = misc_register(&watchdog_device.miscdev); |
| if (ret) { |
| pr_err("cannot register miscdev %d err =%d\n", |
| WATCHDOG_MINOR, ret); |
| goto misc_register_error; |
| } |
| |
| ret = request_irq((unsigned int)watchdog_device.timer_tbl_ptr->irq, |
| watchdog_timer_interrupt, |
| IRQF_SHARED, "watchdog", |
| &watchdog_device.timer_load_count_addr); |
| if (ret) { |
| pr_err("error requesting irq %d\n", ret); |
| goto request_irq_error; |
| } |
| /* Make sure timer is disabled before returning */ |
| intel_scu_stop(); |
| return 0; |
| |
| /* error cleanup */ |
| |
| request_irq_error: |
| misc_deregister(&watchdog_device.miscdev); |
| misc_register_error: |
| unregister_reboot_notifier(&watchdog_device.intel_scu_notifier); |
| register_reboot_error: |
| intel_scu_stop(); |
| iounmap(watchdog_device.timer_load_count_addr); |
| return ret; |
| } |
| late_initcall(intel_scu_watchdog_init); |