/* Copyright (c) 2011-2015, 2020 The Linux Foundation. All rights reserved. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License version 2 and * only version 2 as published by the Free Software Foundation. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. */ #define pr_fmt(fmt) "subsys-restart: %s(): " fmt, __func__ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define BUF_SIZE 30 #define DISABLE_SSR 0x9889deed /* If set to 0x9889deed, call to subsystem_restart_dev() returns immediately */ static uint disable_restart_work; module_param(disable_restart_work, uint, S_IRUGO | S_IWUSR); static int enable_debug; module_param(enable_debug, int, S_IRUGO | S_IWUSR); static struct subsys_device *desc_to_subsys(struct device *d); static void send_notif_to_child(struct subsys_desc *desc, enum subsys_notif_type notif); /* The maximum shutdown timeout is the product of MAX_LOOPS and DELAY_MS. */ #define SHUTDOWN_ACK_MAX_LOOPS 100 #define SHUTDOWN_ACK_DELAY_MS 100 /** * enum p_subsys_state - state of a subsystem (private) * @SUBSYS_NORMAL: subsystem is operating normally * @SUBSYS_CRASHED: subsystem has crashed and hasn't been shutdown * @SUBSYS_RESTARTING: subsystem has been shutdown and is now restarting * * The 'private' side of the subsytem state used to determine where in the * restart process the subsystem is. */ enum p_subsys_state { SUBSYS_NORMAL, SUBSYS_CRASHED, SUBSYS_RESTARTING, }; static const char * const subsys_states[] = { [SUBSYS_OFFLINE] = "OFFLINE", [SUBSYS_ONLINE] = "ONLINE", }; static const char * const restart_levels[] = { [RESET_SOC] = "SYSTEM", [RESET_SUBSYS_COUPLED] = "RELATED", }; /** * struct subsys_tracking - track state of a subsystem or restart order * @p_state: private state of subsystem/order * @state: public state of subsystem/order * @s_lock: protects p_state * @lock: protects subsystem/order callbacks and state * * Tracks the state of a subsystem or a set of subsystems (restart order). * Doing this avoids the need to grab each subsystem's lock and update * each subsystems state when restarting an order. */ struct subsys_tracking { enum p_subsys_state p_state; spinlock_t s_lock; enum subsys_state state; struct mutex lock; }; /** * struct subsys_soc_restart_order - subsystem restart order * @subsystem_list: names of subsystems in this restart order * @count: number of subsystems in order * @track: state tracking and locking * @subsys_ptrs: pointers to subsystems in this restart order */ struct subsys_soc_restart_order { struct device_node **device_ptrs; int count; struct subsys_tracking track; struct subsys_device **subsys_ptrs; struct list_head list; }; struct restart_log { struct timeval time; struct subsys_device *dev; struct list_head list; }; /** * struct subsys_device - subsystem device * @desc: subsystem descriptor * @work: context for subsystem_restart_wq_func() for this device * @ssr_wlock: prevents suspend during subsystem_restart() * @wlname: name of wakeup source * @device_restart_work: work struct for device restart * @track: state tracking and locking * @notify: subsys notify handle * @dev: device * @owner: module that provides @desc * @count: reference count of subsystem_get()/subsystem_put() * @id: ida * @restart_level: restart level (0 - panic, 1 - related, 2 - independent, etc.) * @restart_order: order of other devices this devices restarts with * @crash_count: number of times the device has crashed * @do_ramdump_on_put: ramdump on subsystem_put() if true * @crashed: indicates if subsystem has crashed * @notif_state: current state of subsystem in terms of subsys notifications */ struct subsys_device { struct subsys_desc *desc; struct work_struct work; struct wakeup_source ssr_wlock; char wlname[64]; struct work_struct device_restart_work; struct subsys_tracking track; void *notify; struct device dev; struct module *owner; int count; int id; int restart_level; int crash_count; struct subsys_soc_restart_order *restart_order; bool do_ramdump_on_put; struct cdev char_dev; dev_t dev_no; bool crashed; int notif_state; struct list_head list; }; static struct subsys_device *to_subsys(struct device *d) { return container_of(d, struct subsys_device, dev); } static struct subsys_tracking *subsys_get_track(struct subsys_device *subsys) { struct subsys_soc_restart_order *order = subsys->restart_order; if (order) return &order->track; else return &subsys->track; } static ssize_t name_show(struct device *dev, struct device_attribute *attr, char *buf) { return snprintf(buf, PAGE_SIZE, "%s\n", to_subsys(dev)->desc->name); } static ssize_t state_show(struct device *dev, struct device_attribute *attr, char *buf) { enum subsys_state state = to_subsys(dev)->track.state; return snprintf(buf, PAGE_SIZE, "%s\n", subsys_states[state]); } static ssize_t crash_count_show(struct device *dev, struct device_attribute *attr, char *buf) { return snprintf(buf, PAGE_SIZE, "%d\n", to_subsys(dev)->crash_count); } static ssize_t restart_level_show(struct device *dev, struct device_attribute *attr, char *buf) { int level = to_subsys(dev)->restart_level; return snprintf(buf, PAGE_SIZE, "%s\n", restart_levels[level]); } static ssize_t restart_level_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) { struct subsys_device *subsys = to_subsys(dev); const char *p; int i, orig_count = count; p = memchr(buf, '\n', count); if (p) count = p - buf; for (i = 0; i < ARRAY_SIZE(restart_levels); i++) if (!strncasecmp(buf, restart_levels[i], count)) { subsys->restart_level = i; return orig_count; } return -EPERM; } static ssize_t firmware_name_show(struct device *dev, struct device_attribute *attr, char *buf) { return snprintf(buf, PAGE_SIZE, "%s\n", to_subsys(dev)->desc->fw_name); } static ssize_t firmware_name_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) { struct subsys_device *subsys = to_subsys(dev); struct subsys_tracking *track = subsys_get_track(subsys); const char *p; int orig_count = count; p = memchr(buf, '\n', count); if (p) count = p - buf; pr_info("Changing subsys fw_name to %s\n", buf); mutex_lock(&track->lock); strlcpy(subsys->desc->fw_name, buf, min(count + 1, sizeof(subsys->desc->fw_name))); mutex_unlock(&track->lock); return orig_count; } static ssize_t system_debug_show(struct device *dev, struct device_attribute *attr, char *buf) { struct subsys_device *subsys = to_subsys(dev); char p[6] = "set"; if (!subsys->desc->system_debug) strlcpy(p, "reset", sizeof(p)); return snprintf(buf, PAGE_SIZE, "%s\n", p); } static ssize_t system_debug_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) { struct subsys_device *subsys = to_subsys(dev); const char *p; int orig_count = count; p = memchr(buf, '\n', count); if (p) count = p - buf; if (!strncasecmp(buf, "set", count)) subsys->desc->system_debug = true; else if (!strncasecmp(buf, "reset", count)) subsys->desc->system_debug = false; else return -EPERM; return orig_count; } int subsys_get_restart_level(struct subsys_device *dev) { return dev->restart_level; } EXPORT_SYMBOL(subsys_get_restart_level); enum subsys_state subsys_get_state(struct subsys_device *subsys) { enum subsys_state state = subsys->track.state; return state; } EXPORT_SYMBOL(subsys_get_state); static void subsys_set_state(struct subsys_device *subsys, enum subsys_state state) { unsigned long flags; spin_lock_irqsave(&subsys->track.s_lock, flags); if (subsys->track.state != state) { subsys->track.state = state; spin_unlock_irqrestore(&subsys->track.s_lock, flags); sysfs_notify(&subsys->dev.kobj, NULL, "state"); return; } spin_unlock_irqrestore(&subsys->track.s_lock, flags); } /** * subsytem_default_online() - Mark a subsystem as online by default * @dev: subsystem to mark as online * * Marks a subsystem as "online" without increasing the reference count * on the subsystem. This is typically used by subsystems that are already * online when the kernel boots up. */ void subsys_default_online(struct subsys_device *dev) { subsys_set_state(dev, SUBSYS_ONLINE); } EXPORT_SYMBOL(subsys_default_online); static struct device_attribute subsys_attrs[] = { __ATTR_RO(name), __ATTR_RO(state), __ATTR_RO(crash_count), __ATTR(restart_level, 0644, restart_level_show, restart_level_store), __ATTR(firmware_name, 0644, firmware_name_show, firmware_name_store), __ATTR(system_debug, 0644, system_debug_show, system_debug_store), __ATTR_NULL, }; struct bus_type subsys_bus_type = { .name = "msm_subsys", .dev_attrs = subsys_attrs, }; EXPORT_SYMBOL(subsys_bus_type); static DEFINE_IDA(subsys_ida); static int enable_ramdumps; module_param(enable_ramdumps, int, S_IRUGO | S_IWUSR); static int enable_mini_ramdumps; module_param(enable_mini_ramdumps, int, S_IRUGO | S_IWUSR); struct workqueue_struct *ssr_wq; static struct class *char_class; static LIST_HEAD(restart_log_list); static LIST_HEAD(subsys_list); static LIST_HEAD(ssr_order_list); static DEFINE_MUTEX(soc_order_reg_lock); static DEFINE_MUTEX(restart_log_mutex); static DEFINE_MUTEX(subsys_list_lock); static DEFINE_MUTEX(char_device_lock); static DEFINE_MUTEX(ssr_order_mutex); static struct subsys_soc_restart_order * update_restart_order(struct subsys_device *dev) { int i; struct subsys_soc_restart_order *order; struct device_node *device = dev->desc->dev->of_node; mutex_lock(&soc_order_reg_lock); list_for_each_entry(order, &ssr_order_list, list) { for (i = 0; i < order->count; i++) { if (order->device_ptrs[i] == device) { order->subsys_ptrs[i] = dev; goto found; } } } order = NULL; found: mutex_unlock(&soc_order_reg_lock); return order; } static int max_restarts; module_param(max_restarts, int, 0644); static long max_history_time = 3600; module_param(max_history_time, long, 0644); static void do_epoch_check(struct subsys_device *dev) { int n = 0; struct timeval *time_first = NULL, *curr_time; struct restart_log *r_log, *temp; static int max_restarts_check; static long max_history_time_check; mutex_lock(&restart_log_mutex); max_restarts_check = max_restarts; max_history_time_check = max_history_time; /* Check if epoch checking is enabled */ if (!max_restarts_check) goto out; r_log = kmalloc(sizeof(struct restart_log), GFP_KERNEL); if (!r_log) goto out; r_log->dev = dev; do_gettimeofday(&r_log->time); curr_time = &r_log->time; INIT_LIST_HEAD(&r_log->list); list_add_tail(&r_log->list, &restart_log_list); list_for_each_entry_safe(r_log, temp, &restart_log_list, list) { if ((curr_time->tv_sec - r_log->time.tv_sec) > max_history_time_check) { pr_debug("Deleted node with restart_time = %ld\n", r_log->time.tv_sec); list_del(&r_log->list); kfree(r_log); continue; } if (!n) { time_first = &r_log->time; pr_debug("Time_first: %ld\n", time_first->tv_sec); } n++; pr_debug("Restart_time: %ld\n", r_log->time.tv_sec); } if (time_first && n >= max_restarts_check) { if ((curr_time->tv_sec - time_first->tv_sec) < max_history_time_check) panic("Subsystems have crashed %d times in less than " "%ld seconds!", max_restarts_check, max_history_time_check); } out: mutex_unlock(&restart_log_mutex); } static int is_ramdump_enabled(struct subsys_device *dev) { if (dev->desc->ramdump_disable_gpio) return !dev->desc->ramdump_disable; return enable_ramdumps; } static void for_each_subsys_device(struct subsys_device **list, unsigned count, void *data, void (*fn)(struct subsys_device *, void *)) { while (count--) { struct subsys_device *dev = *list++; if (!dev) continue; fn(dev, data); } } static void notify_each_subsys_device(struct subsys_device **list, unsigned count, enum subsys_notif_type notif, void *data) { while (count--) { struct subsys_device *dev = *list++; struct notif_data notif_data; struct platform_device *pdev; if (!dev) continue; pdev = container_of(dev->desc->dev, struct platform_device, dev); dev->notif_state = notif; notif_data.crashed = subsys_get_crash_status(dev); notif_data.enable_ramdump = is_ramdump_enabled(dev); notif_data.enable_mini_ramdumps = enable_mini_ramdumps; notif_data.no_auth = dev->desc->no_auth; notif_data.pdev = pdev; subsys_notif_queue_notification(dev->notify, notif, ¬if_data); } } int wait_for_shutdown_ack(struct subsys_desc *desc) { int count; if (!desc || !desc->shutdown_ack_gpio) return 0; for (count = SHUTDOWN_ACK_MAX_LOOPS; count > 0; count--) { if (gpio_get_value(desc->shutdown_ack_gpio)) return count; msleep(SHUTDOWN_ACK_DELAY_MS); } pr_err("[%s]: Timed out waiting for shutdown ack\n", desc->name); return -ETIMEDOUT; } EXPORT_SYMBOL(wait_for_shutdown_ack); static void subsystem_shutdown(struct subsys_device *dev, void *data) { const char *name = dev->desc->name; pr_info("[%s:%d]: Shutting down %s\n", current->comm, current->pid, name); if (dev->desc->shutdown(dev->desc, true) < 0) panic("subsys-restart: [%s:%d]: Failed to shutdown %s!", current->comm, current->pid, name); dev->crash_count++; subsys_set_state(dev, SUBSYS_OFFLINE); } static void subsystem_ramdump(struct subsys_device *dev, void *data) { const char *name = dev->desc->name; if (dev->desc->ramdump) if (dev->desc->ramdump(is_ramdump_enabled(dev), dev->desc) < 0) pr_warn("%s[%s:%d]: Ramdump failed.\n", name, current->comm, current->pid); dev->do_ramdump_on_put = false; } static void subsystem_free_memory(struct subsys_device *dev, void *data) { if (dev->desc->free_memory) dev->desc->free_memory(dev->desc); } static void subsystem_powerup(struct subsys_device *dev, void *data) { const char *name = dev->desc->name; pr_info("[%s:%d]: Powering up %s\n", current->comm, current->pid, name); if (dev->desc->powerup(dev->desc) < 0) { notify_each_subsys_device(&dev, 1, SUBSYS_POWERUP_FAILURE, NULL); pr_emerg("[%s:%d]: Powerup error: %s!", current->comm, current->pid, name); return; } subsys_set_state(dev, SUBSYS_ONLINE); subsys_set_crash_status(dev, false); } static int __find_subsys(struct device *dev, void *data) { struct subsys_device *subsys = to_subsys(dev); return !strcmp(subsys->desc->name, data); } static struct subsys_device *find_subsys(const char *str) { struct device *dev; if (!str) return NULL; dev = bus_find_device(&subsys_bus_type, NULL, (void *)str, __find_subsys); return dev ? to_subsys(dev) : NULL; } static int subsys_start(struct subsys_device *subsys) { int ret; notify_each_subsys_device(&subsys, 1, SUBSYS_BEFORE_POWERUP, NULL); ret = subsys->desc->powerup(subsys->desc); if (ret) { notify_each_subsys_device(&subsys, 1, SUBSYS_POWERUP_FAILURE, NULL); subsys->desc->shutdown(subsys->desc, false); return ret; } subsys_set_state(subsys, SUBSYS_ONLINE); if (subsys->desc->is_not_loadable) return 0; notify_each_subsys_device(&subsys, 1, SUBSYS_AFTER_POWERUP, NULL); return ret; } static void subsys_stop(struct subsys_device *subsys) { notify_each_subsys_device(&subsys, 1, SUBSYS_BEFORE_SHUTDOWN, NULL); subsys->desc->shutdown(subsys->desc, false); subsys_set_state(subsys, SUBSYS_OFFLINE); notify_each_subsys_device(&subsys, 1, SUBSYS_AFTER_SHUTDOWN, NULL); } void *__subsystem_get(const char *name, const char *fw_name) { struct subsys_device *subsys; struct subsys_device *subsys_d; int ret; void *retval; struct subsys_tracking *track; if (!name) return NULL; subsys = retval = find_subsys(name); if (!subsys) return ERR_PTR(-ENODEV); if (!try_module_get(subsys->owner)) { retval = ERR_PTR(-ENODEV); goto err_module; } /*parent must start before child*/ if (subsys->desc->parent) { void *tmp = subsystem_get(subsys->desc->parent->name); if (IS_ERR_OR_NULL(tmp)) { pr_err("can't get %s subsystem\n", subsys->desc->parent->name); return tmp; } } subsys_d = subsystem_get(subsys->desc->depends_on); if (IS_ERR(subsys_d)) { retval = subsys_d; goto err_depends; } track = subsys_get_track(subsys); mutex_lock(&track->lock); if (!subsys->count) { if (fw_name) { pr_info("Changing subsys fw_name to %s\n", fw_name); strlcpy(subsys->desc->fw_name, fw_name, sizeof(subsys->desc->fw_name)); } ret = subsys_start(subsys); if (ret) { retval = ERR_PTR(ret); goto err_start; } } subsys->count++; mutex_unlock(&track->lock); return retval; err_start: mutex_unlock(&track->lock); subsystem_put(subsys_d); err_depends: module_put(subsys->owner); err_module: put_device(&subsys->dev); return retval; } /** * subsytem_get() - Boot a subsystem * @name: pointer to a string containing the name of the subsystem to boot * * This function returns a pointer if it succeeds. If an error occurs an * ERR_PTR is returned. * * If this feature is disable, the value %NULL will be returned. */ void *subsystem_get(const char *name) { return __subsystem_get(name, NULL); } EXPORT_SYMBOL(subsystem_get); /** * subsystem_get_with_fwname() - Boot a subsystem using the firmware name passed in * @name: pointer to a string containing the name of the subsystem to boot * @fw_name: pointer to a string containing the subsystem firmware image name * * This function returns a pointer if it succeeds. If an error occurs an * ERR_PTR is returned. * * If this feature is disable, the value %NULL will be returned. */ void *subsystem_get_with_fwname(const char *name, const char *fw_name) { return __subsystem_get(name, fw_name); } EXPORT_SYMBOL(subsystem_get_with_fwname); /** * subsystem_put() - Shutdown a subsystem * @peripheral_handle: pointer from a previous call to subsystem_get() * * This doesn't imply that a subsystem is shutdown until all callers of * subsystem_get() have called subsystem_put(). */ void subsystem_put(void *subsystem) { struct subsys_device *subsys_d, *subsys = subsystem; struct subsys_tracking *track; if (IS_ERR_OR_NULL(subsys)) return; track = subsys_get_track(subsys); mutex_lock(&track->lock); if (WARN(!subsys->count, "%s: %s: Reference count mismatch\n", subsys->desc->name, __func__)) goto err_out; if (!--subsys->count) { subsys_stop(subsys); if (subsys->do_ramdump_on_put) subsystem_ramdump(subsys, NULL); subsystem_free_memory(subsys, NULL); } mutex_unlock(&track->lock); subsys_d = find_subsys(subsys->desc->depends_on); if (subsys_d) { subsystem_put(subsys_d); put_device(&subsys_d->dev); } if (subsys->desc->parent) { struct subsys_device *subsys_p = desc_to_subsys(subsys->desc->parent->dev); subsystem_put(subsys_p); } module_put(subsys->owner); put_device(&subsys->dev); return; err_out: mutex_unlock(&track->lock); } EXPORT_SYMBOL(subsystem_put); static void update_crash_status_of_child(struct subsys_desc *desc) { struct subsys_child *sub_child; list_for_each_entry(sub_child, &desc->child, node) { struct subsys_desc *tmp_desc = (struct subsys_desc *)sub_child->handle; struct subsys_device *tmp_dev = desc_to_subsys(tmp_desc->dev); if (tmp_dev && tmp_dev->track.state == SUBSYS_ONLINE) { subsys_set_crash_status(tmp_dev, CRASH_STATUS_ERR_FATAL); tmp_dev->track.p_state = SUBSYS_CRASHED; } } } static void restart_multipd_subsystem(struct subsys_device *dev) { struct subsys_child *sub_child; struct subsys_desc *desc = dev->desc; unsigned long flags; struct subsys_tracking *track = &dev->track; update_crash_status_of_child(desc); /*shut down child's*/ list_for_each_entry(sub_child, &desc->child, node) { struct subsys_desc *tmp_desc = (struct subsys_desc *)sub_child->handle; struct subsys_device *tmp_dev = desc_to_subsys(tmp_desc->dev); if (tmp_dev && tmp_dev->track.state == SUBSYS_ONLINE) { notify_each_subsys_device(&tmp_dev, 1, SUBSYS_BEFORE_SHUTDOWN, NULL); for_each_subsys_device(&tmp_dev, 1, NULL, subsystem_shutdown); notify_each_subsys_device(&tmp_dev, 1, SUBSYS_AFTER_SHUTDOWN, NULL); } } /*shut down parent*/ for_each_subsys_device(&dev, 1, NULL, subsystem_shutdown); /*send ramdump notif to childs as well*/ send_notif_to_child(desc, SUBSYS_RAMDUMP_NOTIFICATION); /*Take crash dump of parent only, it consist's child dump also*/ notify_each_subsys_device(&dev, 1, SUBSYS_RAMDUMP_NOTIFICATION, NULL); spin_lock_irqsave(&track->s_lock, flags); track->p_state = SUBSYS_RESTARTING; spin_unlock_irqrestore(&track->s_lock, flags); for_each_subsys_device(&dev, 1, NULL, subsystem_ramdump); /*power on parent*/ notify_each_subsys_device(&dev, 1, SUBSYS_BEFORE_POWERUP, NULL); for_each_subsys_device(&dev, 1, NULL, subsystem_powerup); if (dev->track.state != SUBSYS_ONLINE) { pr_info("[%s:%s:%d]: Restart sequence for %s failed.\n", __func__, current->comm, current->pid, desc->name); return; } notify_each_subsys_device(&dev, 1, SUBSYS_AFTER_POWERUP, NULL); /*Power on child*/ list_for_each_entry(sub_child, &desc->child, node) { struct subsys_desc *tmp_desc = (struct subsys_desc *)sub_child->handle; struct subsys_device *tmp_dev = desc_to_subsys(tmp_desc->dev); if (tmp_dev && tmp_dev->track.p_state == SUBSYS_CRASHED) { notify_each_subsys_device(&tmp_dev, 1, SUBSYS_BEFORE_POWERUP, NULL); for_each_subsys_device(&tmp_dev, 1, NULL, subsystem_powerup); if (tmp_dev->track.state != SUBSYS_ONLINE) { pr_info("Restart sequence for %s failed.\n", tmp_desc->name); continue; } notify_each_subsys_device(&tmp_dev, 1, SUBSYS_AFTER_POWERUP, NULL); tmp_dev->track.p_state = SUBSYS_NORMAL; } } pr_info("[%s:%s:%d]: Restart sequence for %s completed.\n", __func__, current->comm, current->pid, desc->name); } static void subsystem_restart_wq_func(struct work_struct *work) { struct subsys_device *dev = container_of(work, struct subsys_device, work); struct subsys_device **list; struct subsys_desc *desc = dev->desc; struct subsys_soc_restart_order *order = dev->restart_order; struct subsys_tracking *track; unsigned count; unsigned long flags; bool is_parent = false; struct subsys_child *sub_child; /* * It's OK to not take the registration lock at this point. * This is because the subsystem list inside the relevant * restart order is not being traversed. */ if (order) { list = order->subsys_ptrs; count = order->count; track = &order->track; } else { list = &dev; count = 1; track = &dev->track; } mutex_lock(&track->lock); do_epoch_check(dev); if (dev->track.state == SUBSYS_OFFLINE) { mutex_unlock(&track->lock); WARN(1, "SSR aborted: %s subsystem not online\n", desc->name); return; } /* * It's necessary to take the registration lock because the subsystem * list in the SoC restart order will be traversed and it shouldn't be * changed until _this_ restart sequence completes. */ mutex_lock(&soc_order_reg_lock); pr_debug("[%s:%d]: Starting restart sequence for %s\n", current->comm, current->pid, desc->name); /*check whether it's a parent*/ list_for_each_entry(sub_child, &desc->child, node) { is_parent = true; break; } if (is_parent) { restart_multipd_subsystem(dev); goto unlock; } notify_each_subsys_device(list, count, SUBSYS_BEFORE_SHUTDOWN, NULL); for_each_subsys_device(list, count, NULL, subsystem_shutdown); notify_each_subsys_device(list, count, SUBSYS_AFTER_SHUTDOWN, NULL); notify_each_subsys_device(list, count, SUBSYS_RAMDUMP_NOTIFICATION, NULL); spin_lock_irqsave(&track->s_lock, flags); track->p_state = SUBSYS_RESTARTING; spin_unlock_irqrestore(&track->s_lock, flags); /* Collect ram dumps for all subsystems in order here */ for_each_subsys_device(list, count, NULL, subsystem_ramdump); for_each_subsys_device(list, count, NULL, subsystem_free_memory); notify_each_subsys_device(list, count, SUBSYS_BEFORE_POWERUP, NULL); for_each_subsys_device(list, count, NULL, subsystem_powerup); if (dev->track.state != SUBSYS_ONLINE) { pr_info("[%s:%d]: Restart sequence for %s failed.\n", current->comm, current->pid, desc->name); } else { notify_each_subsys_device(list, count, SUBSYS_AFTER_POWERUP, NULL); pr_info("[%s:%d]: Restart sequence for %s completed.\n", current->comm, current->pid, desc->name); } unlock: mutex_unlock(&soc_order_reg_lock); mutex_unlock(&track->lock); if (dev->track.state == SUBSYS_ONLINE) { spin_lock_irqsave(&track->s_lock, flags); track->p_state = SUBSYS_NORMAL; __pm_relax(&dev->ssr_wlock); spin_unlock_irqrestore(&track->s_lock, flags); } } static void send_notif_to_child(struct subsys_desc *desc, enum subsys_notif_type notif) { struct subsys_child *sub_child; list_for_each_entry(sub_child, &desc->child, node) { struct subsys_desc *tmp_desc = (struct subsys_desc *)sub_child->handle; struct subsys_device *tmp_dev = desc_to_subsys(tmp_desc->dev); if (tmp_dev) notify_each_subsys_device(&tmp_dev, 1, notif, NULL); } } static void __subsystem_restart_dev(struct subsys_device *dev) { struct subsys_desc *desc = dev->desc; const char *name = dev->desc->name; struct subsys_tracking *track; unsigned long flags; struct subsys_device **list; struct subsys_soc_restart_order *order = dev->restart_order; unsigned int count; pr_debug("Restarting %s [level=%s]!\n", desc->name, restart_levels[dev->restart_level]); track = subsys_get_track(dev); if (order) { list = order->subsys_ptrs; count = order->count; } else { list = &dev; count = 1; } /* * Allow drivers to call subsystem_restart{_dev}() as many times as * they want up until the point where the subsystem is shutdown. */ spin_lock_irqsave(&track->s_lock, flags); if (track->p_state != SUBSYS_CRASHED && dev->track.state == SUBSYS_ONLINE) { if (track->p_state != SUBSYS_RESTARTING) { track->p_state = SUBSYS_CRASHED; /* * Send fatal notification to userpd(s) if rootpd * crashed, irrespective of userpd status. */ send_notif_to_child(desc, SUBSYS_PREPARE_FOR_FATAL_SHUTDOWN); notify_each_subsys_device(list, count, SUBSYS_PREPARE_FOR_FATAL_SHUTDOWN, NULL); __pm_stay_awake(&dev->ssr_wlock); queue_work(ssr_wq, &dev->work); } else { panic("Subsystem %s crashed during SSR!", name); } } else WARN(dev->track.state == SUBSYS_OFFLINE, "SSR aborted: %s subsystem not online\n", name); spin_unlock_irqrestore(&track->s_lock, flags); } static void device_restart_work_hdlr(struct work_struct *work) { struct subsys_device *dev = container_of(work, struct subsys_device, device_restart_work); notify_each_subsys_device(&dev, 1, SUBSYS_SOC_RESET, NULL); /* * Temporary workaround until ramdump userspace application calls * sync() and fclose() on attempting the dump. */ msleep(100); panic("subsys-restart: Resetting the SoC - %s crashed.", dev->desc->name); } int subsystem_restart_dev(struct subsys_device *dev) { const char *name; if (!get_device(&dev->dev)) return -ENODEV; if (!try_module_get(dev->owner)) { put_device(&dev->dev); return -ENODEV; } name = dev->desc->name; /* * If a system reboot/shutdown is underway, ignore subsystem errors. * However, print a message so that we know that a subsystem behaved * unexpectedly here. */ if (system_state == SYSTEM_RESTART || system_state == SYSTEM_POWER_OFF) { pr_err("%s crashed during a system poweroff/shutdown.\n", name); return -EBUSY; } pr_info("Restart sequence requested for %s, restart_level = %s.\n", name, restart_levels[dev->restart_level]); if (disable_restart_work == DISABLE_SSR) { pr_warn("subsys-restart: Ignoring restart request for %s.\n", name); return 0; } switch (dev->restart_level) { case RESET_SUBSYS_COUPLED: __subsystem_restart_dev(dev); break; case RESET_SOC: __pm_stay_awake(&dev->ssr_wlock); schedule_work(&dev->device_restart_work); return 0; default: panic("subsys-restart: Unknown restart level!\n"); break; } module_put(dev->owner); put_device(&dev->dev); return 0; } EXPORT_SYMBOL(subsystem_restart_dev); int subsystem_restart(const char *name) { int ret; struct subsys_device *dev = find_subsys(name); if (!dev) return -ENODEV; ret = subsystem_restart_dev(dev); put_device(&dev->dev); return ret; } EXPORT_SYMBOL(subsystem_restart); int subsystem_crashed(const char *name) { struct subsys_device *dev = find_subsys(name); struct subsys_tracking *track; if (!dev) return -ENODEV; if (!get_device(&dev->dev)) return -ENODEV; track = subsys_get_track(dev); mutex_lock(&track->lock); dev->do_ramdump_on_put = true; /* * TODO: Make this work with multiple consumers where one is calling * subsystem_restart() and another is calling this function. To do * so would require updating private state, etc. */ mutex_unlock(&track->lock); put_device(&dev->dev); return 0; } EXPORT_SYMBOL(subsystem_crashed); void subsys_set_crash_status(struct subsys_device *dev, bool crashed) { dev->crashed = crashed; } bool subsys_get_crash_status(struct subsys_device *dev) { return dev->crashed; } static struct subsys_device *desc_to_subsys(struct device *d) { struct subsys_device *device, *subsys_dev = 0; mutex_lock(&subsys_list_lock); list_for_each_entry(device, &subsys_list, list) if (device->desc->dev == d) subsys_dev = device; mutex_unlock(&subsys_list_lock); return subsys_dev; } void notify_proxy_vote(struct device *device) { struct subsys_device *dev = desc_to_subsys(device); if (dev) notify_each_subsys_device(&dev, 1, SUBSYS_PROXY_VOTE, NULL); } void notify_proxy_unvote(struct device *device) { struct subsys_device *dev = desc_to_subsys(device); if (dev) notify_each_subsys_device(&dev, 1, SUBSYS_PROXY_UNVOTE, NULL); } static int subsys_device_open(struct inode *inode, struct file *file) { struct subsys_device *device, *subsys_dev = 0; void *retval; mutex_lock(&subsys_list_lock); list_for_each_entry(device, &subsys_list, list) if (MINOR(device->dev_no) == iminor(inode)) subsys_dev = device; mutex_unlock(&subsys_list_lock); if (!subsys_dev) return -EINVAL; retval = subsystem_get_with_fwname(subsys_dev->desc->name, subsys_dev->desc->fw_name); if (IS_ERR(retval)) return PTR_ERR(retval); return 0; } static int subsys_device_close(struct inode *inode, struct file *file) { struct subsys_device *device, *subsys_dev = 0; mutex_lock(&subsys_list_lock); list_for_each_entry(device, &subsys_list, list) if (MINOR(device->dev_no) == iminor(inode)) subsys_dev = device; mutex_unlock(&subsys_list_lock); if (!subsys_dev) return -EINVAL; subsystem_put(subsys_dev); return 0; } static const struct file_operations subsys_device_fops = { .owner = THIS_MODULE, .open = subsys_device_open, .release = subsys_device_close, }; static void subsys_device_release(struct device *dev) { struct subsys_device *subsys = to_subsys(dev); wakeup_source_trash(&subsys->ssr_wlock); mutex_destroy(&subsys->track.lock); ida_simple_remove(&subsys_ida, subsys->id); kfree(subsys); } static int subsys_char_device_add(struct subsys_device *subsys_dev) { int ret = 0; static int major, minor; dev_t dev_no; mutex_lock(&char_device_lock); if (!major) { ret = alloc_chrdev_region(&dev_no, 0, 4, "subsys"); if (ret < 0) { pr_err("Failed to alloc subsys_dev region, err %d\n", ret); goto fail; } major = MAJOR(dev_no); minor = MINOR(dev_no); } else dev_no = MKDEV(major, minor); if (!device_create(char_class, subsys_dev->desc->dev, dev_no, NULL, "subsys_%s", subsys_dev->desc->name)) { pr_err("Failed to create subsys_%s device\n", subsys_dev->desc->name); goto fail_unregister_cdev_region; } cdev_init(&subsys_dev->char_dev, &subsys_device_fops); subsys_dev->char_dev.owner = THIS_MODULE; ret = cdev_add(&subsys_dev->char_dev, dev_no, 1); if (ret < 0) goto fail_destroy_device; subsys_dev->dev_no = dev_no; minor++; mutex_unlock(&char_device_lock); return 0; fail_destroy_device: device_destroy(char_class, dev_no); fail_unregister_cdev_region: unregister_chrdev_region(dev_no, 1); fail: mutex_unlock(&char_device_lock); return ret; } static void subsys_char_device_remove(struct subsys_device *subsys_dev) { cdev_del(&subsys_dev->char_dev); device_destroy(char_class, subsys_dev->dev_no); unregister_chrdev_region(subsys_dev->dev_no, 1); } static void subsys_remove_restart_order(struct device_node *device) { struct subsys_soc_restart_order *order; int i; mutex_lock(&ssr_order_mutex); list_for_each_entry(order, &ssr_order_list, list) for (i = 0; i < order->count; i++) if (order->device_ptrs[i] == device) order->subsys_ptrs[i] = NULL; mutex_unlock(&ssr_order_mutex); } static struct subsys_soc_restart_order *ssr_parse_restart_orders(struct subsys_desc * desc) { int i, j, count, num = 0; struct subsys_soc_restart_order *order, *tmp; struct device *dev = desc->dev; struct device_node *ssr_node; uint32_t len; if (!of_get_property(dev->of_node, "qcom,restart-group", &len)) return NULL; count = len/sizeof(uint32_t); order = devm_kzalloc(dev, sizeof(*order), GFP_KERNEL); if (!order) return ERR_PTR(-ENOMEM); order->subsys_ptrs = devm_kzalloc(dev, count * sizeof(struct subsys_device *), GFP_KERNEL); if (!order->subsys_ptrs) return ERR_PTR(-ENOMEM); order->device_ptrs = devm_kzalloc(dev, count * sizeof(struct device_node *), GFP_KERNEL); if (!order->device_ptrs) return ERR_PTR(-ENOMEM); for (i = 0; i < count; i++) { ssr_node = of_parse_phandle(dev->of_node, "qcom,restart-group", i); if (!ssr_node) return ERR_PTR(-ENXIO); of_node_put(ssr_node); pr_info("%s device has been added to %s's restart group\n", ssr_node->name, desc->name); order->device_ptrs[i] = ssr_node; } /* * Check for similar restart groups. If found, return * without adding the new group to the ssr_order_list. */ mutex_lock(&ssr_order_mutex); list_for_each_entry(tmp, &ssr_order_list, list) { for (i = 0; i < count; i++) { for (j = 0; j < count; j++) { if (order->device_ptrs[j] != tmp->device_ptrs[i]) continue; else num++; } } if (num == count && tmp->count == count) goto err; else if (num) { tmp = ERR_PTR(-EINVAL); goto err; } } order->count = count; mutex_init(&order->track.lock); spin_lock_init(&order->track.s_lock); INIT_LIST_HEAD(&order->list); list_add_tail(&order->list, &ssr_order_list); mutex_unlock(&ssr_order_mutex); return order; err: mutex_unlock(&ssr_order_mutex); return tmp; } static int __get_gpio(struct subsys_desc *desc, const char *prop, int *gpio) { struct device_node *dnode = desc->dev->of_node; int ret = -ENOENT; if (of_find_property(dnode, prop, NULL)) { *gpio = of_get_named_gpio(dnode, prop, 0); ret = *gpio < 0 ? *gpio : 0; } return ret; } static int __get_irq(struct subsys_desc *desc, const char *prop, unsigned int *irq, int *gpio) { int ret, gpiol, irql; if (of_property_read_bool(desc->dev->of_node, "qca,extended-intc") || of_property_read_bool(desc->dev->parent->of_node, "qca,extended-intc")) { struct platform_device *pdev = container_of(desc->dev, struct platform_device, dev); irql = platform_get_irq_byname(pdev, prop); if (irql < 0) { pr_err("[%s]: Error getting IRQ \"%s\"\n", desc->name, prop); return irql; } *irq = irql; return 0; } ret = __get_gpio(desc, prop, &gpiol); if (ret) return ret; irql = gpio_to_irq(gpiol); if (irql == -ENOENT) irql = -ENXIO; if (irql < 0) { pr_err("[%s]: Error getting IRQ \"%s\"\n", desc->name, prop); return irql; } else { if (gpio) *gpio = gpiol; *irq = irql; } return 0; } static int subsys_parse_devicetree(struct subsys_desc *desc) { struct subsys_soc_restart_order *order; int ret; struct platform_device *pdev = container_of(desc->dev, struct platform_device, dev); ret = __get_irq(desc, "fatal", &desc->err_fatal_irq, &desc->err_fatal_gpio); if (ret && ret != -ENOENT) return ret; ret = __get_irq(desc, "stop-ack", &desc->stop_ack_irq, NULL); if (ret && ret != -ENOENT) return ret; ret = __get_gpio(desc, "qcom,gpio-force-stop", &desc->force_stop_gpio); if (ret && ret != -ENOENT) return ret; ret = __get_gpio(desc, "qcom,gpio-ramdump-disable", &desc->ramdump_disable_gpio); if (ret && ret != -ENOENT) return ret; ret = __get_gpio(desc, "qcom,gpio-shutdown-ack", &desc->shutdown_ack_gpio); if (ret && ret != -ENOENT) return ret; if (!desc->parent) { if (of_property_read_bool(desc->dev->of_node, "qca,extended-intc")) ret = platform_get_irq_byname(pdev, "wdog"); else /* There is only one irq */ ret = platform_get_irq(pdev, 0); if (ret > 0) desc->wdog_bite_irq = ret; } order = ssr_parse_restart_orders(desc); if (IS_ERR(order)) { pr_err("Could not initialize SSR restart order, err = %ld\n", PTR_ERR(order)); return PTR_ERR(order); } return 0; } static int subsys_setup_irqs(struct subsys_device *subsys) { struct subsys_desc *desc = subsys->desc; int ret; char *err_fatal_int; char *stop_ack_int; if (desc->err_fatal_irq && desc->err_fatal_handler) { if (!desc->parent) err_fatal_int = "err_fatal_interrupt"; else { err_fatal_int = devm_kzalloc(desc->dev, BUF_SIZE, GFP_KERNEL); if (!err_fatal_int) return -ENOMEM; strlcpy(err_fatal_int, desc->name, BUF_SIZE); strlcat(err_fatal_int, "_fatal", BUF_SIZE); } ret = devm_request_threaded_irq(desc->dev, desc->err_fatal_irq, NULL, desc->err_fatal_handler, IRQF_TRIGGER_RISING | IRQF_ONESHOT, err_fatal_int, desc); if (ret < 0) { dev_err(desc->dev, "[%s]: Unable to register error fatal IRQ handler!: %d\n", desc->name, ret); return ret; } } if (desc->stop_ack_irq && desc->stop_ack_handler) { if (!desc->parent) stop_ack_int = "stop_ack_interrupt"; else { stop_ack_int = devm_kzalloc(desc->dev, BUF_SIZE, GFP_KERNEL); if (!stop_ack_int) return -ENOMEM; strlcpy(stop_ack_int, desc->name, BUF_SIZE); strlcat(stop_ack_int, "_stop_ack", BUF_SIZE); } ret = devm_request_threaded_irq(desc->dev, desc->stop_ack_irq, NULL, desc->stop_ack_handler, IRQF_TRIGGER_RISING | IRQF_ONESHOT, stop_ack_int, desc); if (ret < 0) { dev_err(desc->dev, "[%s]: Unable to register stop ack handler!: %d\n", desc->name, ret); return ret; } } if (desc->wdog_bite_irq && desc->wdog_bite_handler && !desc->parent) { ret = devm_request_threaded_irq(desc->dev, desc->wdog_bite_irq, NULL, desc->wdog_bite_handler, IRQF_TRIGGER_RISING | IRQF_ONESHOT, "q6_wdog_interrupt", desc); if (ret < 0) { dev_err(desc->dev, "[%s]: Unable to register wdog bite handler!: %d\n", desc->name, ret); return ret; } } return 0; } static void subsys_free_irqs(struct subsys_device *subsys) { struct subsys_desc *desc = subsys->desc; if (desc->err_fatal_irq && desc->err_fatal_handler) devm_free_irq(desc->dev, desc->err_fatal_irq, desc); if (desc->stop_ack_irq && desc->stop_ack_handler) devm_free_irq(desc->dev, desc->stop_ack_irq, desc); if (desc->wdog_bite_irq && desc->wdog_bite_handler) devm_free_irq(desc->dev, desc->wdog_bite_irq, desc); } void subsys_add_child(struct subsys_desc *desc, struct subsys_child *child) { list_add(&child->node, &desc->child); } EXPORT_SYMBOL(subsys_add_child); void subsys_remove_child(struct subsys_desc *desc, struct subsys_child *child) { list_del(&child->node); } EXPORT_SYMBOL(subsys_remove_child); struct subsys_device *subsys_register(struct subsys_desc *desc) { struct subsys_device *subsys; struct device_node *ofnode = desc->dev->of_node; int ret; subsys = kzalloc(sizeof(*subsys), GFP_KERNEL); if (!subsys) return ERR_PTR(-ENOMEM); subsys->desc = desc; subsys->owner = desc->owner; subsys->dev.parent = desc->dev; subsys->dev.bus = &subsys_bus_type; subsys->dev.release = subsys_device_release; subsys->notif_state = -1; subsys->desc->sysmon_pid = -1; strlcpy(subsys->desc->fw_name, desc->name, sizeof(subsys->desc->fw_name)); subsys->notify = subsys_notif_add_subsys(desc->name); snprintf(subsys->wlname, sizeof(subsys->wlname), "ssr(%s)", desc->name); wakeup_source_init(&subsys->ssr_wlock, subsys->wlname); INIT_WORK(&subsys->work, subsystem_restart_wq_func); INIT_WORK(&subsys->device_restart_work, device_restart_work_hdlr); spin_lock_init(&subsys->track.s_lock); subsys->id = ida_simple_get(&subsys_ida, 0, 0, GFP_KERNEL); if (subsys->id < 0) { wakeup_source_trash(&subsys->ssr_wlock); ret = subsys->id; kfree(subsys); return ERR_PTR(ret); } dev_set_name(&subsys->dev, "subsys%d", subsys->id); mutex_init(&subsys->track.lock); ret = device_register(&subsys->dev); if (ret) { put_device(&subsys->dev); return ERR_PTR(ret); } ret = subsys_char_device_add(subsys); if (ret) { goto err_register; } if (ofnode) { ret = subsys_parse_devicetree(desc); if (ret) goto err_register; subsys->restart_order = update_restart_order(subsys); ret = subsys_setup_irqs(subsys); if (ret < 0) goto err_setup_irqs; if (of_property_read_u32(ofnode, "qcom,ssctl-instance-id", &desc->ssctl_instance_id)) pr_debug("Reading instance-id for %s failed\n", desc->name); subsys->desc->edge = of_get_property(ofnode, "qcom,edge", NULL); if (!subsys->desc->edge) pr_debug("Reading qcom,edge for %s failed\n", desc->name); if (of_property_read_bool(desc->dev->of_node, "qca,auto-restart")) subsys->restart_level = RESET_SUBSYS_COUPLED; } mutex_lock(&subsys_list_lock); INIT_LIST_HEAD(&subsys->list); list_add_tail(&subsys->list, &subsys_list); mutex_unlock(&subsys_list_lock); INIT_LIST_HEAD(&subsys->desc->child); return subsys; err_setup_irqs: if (ofnode) subsys_remove_restart_order(ofnode); err_register: device_unregister(&subsys->dev); return ERR_PTR(ret); } EXPORT_SYMBOL(subsys_register); void subsys_unregister(struct subsys_device *subsys) { struct subsys_device *subsys_dev, *tmp; struct device_node *device = subsys->desc->dev->of_node; if (IS_ERR_OR_NULL(subsys)) return; if (get_device(&subsys->dev)) { mutex_lock(&subsys_list_lock); list_for_each_entry_safe(subsys_dev, tmp, &subsys_list, list) if (subsys_dev == subsys) list_del(&subsys->list); mutex_unlock(&subsys_list_lock); if (device) { subsys_free_irqs(subsys); subsys_remove_restart_order(device); } mutex_lock(&subsys->track.lock); WARN_ON(subsys->count); device_unregister(&subsys->dev); mutex_unlock(&subsys->track.lock); subsys_char_device_remove(subsys); put_device(&subsys->dev); } } EXPORT_SYMBOL(subsys_unregister); static int subsys_panic(struct device *dev, void *data) { struct subsys_device *subsys = to_subsys(dev); if (subsys->desc->crash_shutdown) subsys->desc->crash_shutdown(subsys->desc); return 0; } static int ssr_panic_handler(struct notifier_block *this, unsigned long event, void *ptr) { bus_for_each_dev(&subsys_bus_type, NULL, NULL, subsys_panic); return NOTIFY_DONE; } static struct notifier_block panic_nb = { .notifier_call = ssr_panic_handler, .priority = 100, }; static int __init subsys_restart_init(void) { int ret; ssr_wq = alloc_workqueue("ssr_wq", WQ_CPU_INTENSIVE, 0); BUG_ON(!ssr_wq); ret = bus_register(&subsys_bus_type); if (ret) goto err_bus; char_class = class_create(THIS_MODULE, "subsys"); if (IS_ERR(char_class)) { ret = -ENOMEM; pr_err("Failed to create subsys_dev class\n"); goto err_class; } ret = atomic_notifier_chain_register(&panic_notifier_list, &panic_nb); if (ret) goto err_soc; return 0; err_soc: class_destroy(char_class); err_class: bus_unregister(&subsys_bus_type); err_bus: destroy_workqueue(ssr_wq); return ret; } arch_initcall(subsys_restart_init); MODULE_DESCRIPTION("Subsystem Restart Driver"); MODULE_LICENSE("GPL v2");