Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit c992cf65 authored by Arun KS's avatar Arun KS Committed by Kyle Yan
Browse files

soc: qcom: ssr: Add option to relax on ssr failures 



SSR failures are considered fatal and results in system panic.
In certain scenarios system can continue to work even with a
failed subsystem.

Add an option in subsystem descriptor to relax on ssr failures.

Change-Id: I86dcaa615d6443937077880d9a91070d9c22ea1f
Signed-off-by: default avatarArun KS <arunks@codeaurora.org>
[satyap@codeaurora.org: trivial merge conflict resolution]
Signed-off-by: default avatarSatya Durga Srinivasu Prabhala <satyap@codeaurora.org>
Signed-off-by: default avatarAvaneesh Kumar Dwivedi <akdwived@codeaurora.org>
parent f48b62b7

Documentation/devicetree/bindings/pil/subsys-pil-tz.txt

+1 −0
Original line number Diff line number Diff line
@@ -67,6 +67,7 @@ Optional properties: 
- qcom,complete-ramdump: Boolean. If set, complete ramdump i.e. region between start address of

			first segment to end address of last segment will be collected without

			leaving any hole in between.

- qcom,ignore-ssr-failure: Boolean. If set, SSR failures are not considered fatal.



Example:

	qcom,venus@fdce0000 {

drivers/soc/qcom/subsystem_restart.c

+57 −17
Original line number Diff line number Diff line
@@ -481,17 +481,21 @@ static void send_sysmon_notif(struct subsys_device *dev) 
	mutex_unlock(&subsys_list_lock);

}



static void for_each_subsys_device(struct subsys_device **list,

static int for_each_subsys_device(struct subsys_device **list,

		unsigned int count, void *data,

		void (*fn)(struct subsys_device *, void *))

		int (*fn)(struct subsys_device *, void *))

{

	int ret;

	while (count--) {

		struct subsys_device *dev = *list++;



		if (!dev)

			continue;

		fn(dev, data);

		ret = fn(dev, data);

		if (ret)

			return ret;

	}

	return 0;

}



static void notify_each_subsys_device(struct subsys_device **list,
@@ -593,21 +597,31 @@ static int wait_for_err_ready(struct subsys_device *subsys) 
	return 0;

}



static void subsystem_shutdown(struct subsys_device *dev, void *data)

static int subsystem_shutdown(struct subsys_device *dev, void *data)

{

	const char *name = dev->desc->name;

	int ret;



	pr_info("[%s:%d]: Shutting down %s\n",

			current->comm, current->pid, name);

	if (dev->desc->shutdown(dev->desc, true) < 0)

	ret = dev->desc->shutdown(dev->desc, true);

	if (ret < 0) {

		if (!dev->desc->ignore_ssr_failure) {

			panic("subsys-restart: [%s:%d]: Failed to shutdown %s!",

				current->comm, current->pid, name);

		} else {

			pr_err("Shutdown failure on %s\n", name);

			return ret;

		}

	}

	dev->crash_count++;

	subsys_set_state(dev, SUBSYS_OFFLINE);

	disable_all_irqs(dev);



	return 0;

}



static void subsystem_ramdump(struct subsys_device *dev, void *data)

static int subsystem_ramdump(struct subsys_device *dev, void *data)

{

	const char *name = dev->desc->name;
@@ -616,15 +630,17 @@ static void subsystem_ramdump(struct subsys_device *dev, void *data) 
			pr_warn("%s[%s:%d]: Ramdump failed.\n",

				name, current->comm, current->pid);

	dev->do_ramdump_on_put = false;

	return 0;

}



static void subsystem_free_memory(struct subsys_device *dev, void *data)

static int subsystem_free_memory(struct subsys_device *dev, void *data)

{

	if (dev->desc->free_memory)

		dev->desc->free_memory(dev->desc);

	return 0;

}



static void subsystem_powerup(struct subsys_device *dev, void *data)

static int subsystem_powerup(struct subsys_device *dev, void *data)

{

	const char *name = dev->desc->name;

	int ret;
@@ -632,11 +648,17 @@ static void subsystem_powerup(struct subsys_device *dev, void *data) 
	pr_info("[%s:%d]: Powering up %s\n", current->comm, current->pid, name);

	init_completion(&dev->err_ready);



	if (dev->desc->powerup(dev->desc) < 0) {

	ret = dev->desc->powerup(dev->desc);

	if (ret < 0) {

		notify_each_subsys_device(&dev, 1, SUBSYS_POWERUP_FAILURE,

								NULL);

		if (!dev->desc->ignore_ssr_failure) {

			panic("[%s:%d]: Powerup error: %s!",

				current->comm, current->pid, name);

		} else {

			pr_err("Powerup failure on %s\n", name);

			return ret;

		}

	}

	enable_all_irqs(dev);
@@ -644,11 +666,16 @@ static void subsystem_powerup(struct subsys_device *dev, void *data) 
	if (ret) {

		notify_each_subsys_device(&dev, 1, SUBSYS_POWERUP_FAILURE,

								NULL);

		if (!dev->desc->ignore_ssr_failure)

			panic("[%s:%d]: Timed out waiting for error ready: %s!",

				current->comm, current->pid, name);

		else

			return ret;

	}

	subsys_set_state(dev, SUBSYS_ONLINE);

	subsys_set_crash_status(dev, CRASH_STATUS_NO_CRASH);



	return 0;

}



static int __find_subsys(struct device *dev, void *data)
@@ -910,6 +937,7 @@ static void subsystem_restart_wq_func(struct work_struct *work) 
	struct subsys_tracking *track;

	unsigned int count;

	unsigned long flags;

	int ret;



	/*

	 * It's OK to not take the registration lock at this point.
@@ -957,7 +985,9 @@ static void subsystem_restart_wq_func(struct work_struct *work) 
	pr_debug("[%s:%d]: Starting restart sequence for %s\n",

			current->comm, current->pid, desc->name);

	notify_each_subsys_device(list, count, SUBSYS_BEFORE_SHUTDOWN, NULL);

	for_each_subsys_device(list, count, NULL, subsystem_shutdown);

	ret = for_each_subsys_device(list, count, NULL, subsystem_shutdown);

	if (ret)

		goto err;

	notify_each_subsys_device(list, count, SUBSYS_AFTER_SHUTDOWN, NULL);



	notify_each_subsys_device(list, count, SUBSYS_RAMDUMP_NOTIFICATION,
@@ -973,12 +1003,19 @@ static void subsystem_restart_wq_func(struct work_struct *work) 
	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);

	ret = for_each_subsys_device(list, count, NULL, subsystem_powerup);

	if (ret)

		goto err;

	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);



err:

	/* Reset subsys count */

	if (ret)

		dev->count = 0;



	mutex_unlock(&soc_order_reg_lock);

	mutex_unlock(&track->lock);
@@ -1470,6 +1507,9 @@ static int subsys_parse_devicetree(struct subsys_desc *desc) 
			desc->generic_irq = ret;

	}



	desc->ignore_ssr_failure = of_property_read_bool(pdev->dev.of_node,

						"qcom,ignore-ssr-failure");



	order = ssr_parse_restart_orders(desc);

	if (IS_ERR(order)) {

		pr_err("Could not initialize SSR restart order, err = %ld\n",

include/soc/qcom/subsystem_restart.h

+3 −0
Original line number Diff line number Diff line
@@ -56,6 +56,8 @@ struct module; 
 * @sysmon_shutdown_ret: Return value for the call to sysmon_send_shutdown

 * @system_debug: If "set", triggers a device restart when the

 * subsystem's wdog bite handler is invoked.

 * @ignore_ssr_failure: SSR failures are usually fatal and results in panic. If

 * set will ignore failure.

 * @edge: GLINK logical name of the subsystem

 */

struct subsys_desc {
@@ -91,6 +93,7 @@ struct subsys_desc { 
	u32 sysmon_pid;

	int sysmon_shutdown_ret;

	bool system_debug;

	bool ignore_ssr_failure;

	const char *edge;

};