The watchdog is an essential building block for the usage of Linux in
safety-critical systems because it allows the system to be monitored from
an external element - the watchdog hardware, acting as a safety-monitor.
A user-space application controls the watchdog device via the watchdog
interface. This application, hereafter safety_app, enables the watchdog
and periodically pets the watchdog upon correct completion of the safety
related processing.
If the safety_app, for any reason, stops pinging the watchdog,
the watchdog hardware can set the system in a fail-safe state. For
example, shutting the system down.
Given the importance of the safety_app / watchdog hardware couple,
the interaction between these software pieces also needs some
sort of monitoring. In other words, "who monitors the monitor?"
The safe watchdog (safe_wtd) RV monitor monitors the interaction between
the safety_app and the watchdog device, enforcing the correct sequence of
events that leads the system to a safe state.
Furthermore, the safety_app can monitor the RV monitor by collecting the
events generated by the RV monitor itself via tracing interface. In this way,
closing the monitoring loop with the safety_app.
To reach a safe state, the safe_wtd RV monitor requires the
safety_app to:
- Open the watchdog device
- Start the watchdog
- Set a timeout
- ping at least once
The RV monitor also avoids some undesired actions. For example, to have
other threads to touch the watchdog.
The monitor also has a set of options, enabled via kernel command
line/module options. They are:
- watchdog_id: the device id to monitor (default 0).
- dont_stop: once enabled, do not allow the RV monitor to be stopped
(default off);
- safe_timeout: define a maximum safe value that an user-space
application can set as the watchdog timeout
(default unlimited).
- check_timeout: After every ping, check if the time left in the
watchdog is less than or equal to the last timeout set
for the watchdog. It only works for watchdog devices that
provide the get_timeleft() function (default off).
For further information, please refer to:
Documentation/trace/rv/watchdog-monitor.rst
The monitor specification was developed together with Gabriele Paoloni,
in the context of the Linux Foundation Elisa Project.
Cc: Wim Van Sebroeck <wim@xxxxxxxxxxxxxxxxxx>
Cc: Guenter Roeck <linux@xxxxxxxxxxxx>
Cc: Jonathan Corbet <corbet@xxxxxxx>
Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Will Deacon <will@xxxxxxxxxx>
Cc: Catalin Marinas <catalin.marinas@xxxxxxx>
Cc: Marco Elver <elver@xxxxxxxxxx>
Cc: Dmitry Vyukov <dvyukov@xxxxxxxxxx>
Cc: "Paul E. McKenney" <paulmck@xxxxxxxxxx>
Cc: Shuah Khan <skhan@xxxxxxxxxxxxxxxxxxx>
Cc: Gabriele Paoloni <gpaoloni@xxxxxxxxxx>
Cc: Juri Lelli <juri.lelli@xxxxxxxxxx>
Cc: Clark Williams <williams@xxxxxxxxxx>
Cc: linux-doc@xxxxxxxxxxxxxxx
Cc: linux-kernel@xxxxxxxxxxxxxxx
Cc: linux-trace-devel@xxxxxxxxxxxxxxx
Signed-off-by: Daniel Bristot de Oliveira <bristot@xxxxxxxxxx>
---
include/trace/events/rv.h | 11 +
kernel/trace/rv/Kconfig | 10 +
kernel/trace/rv/Makefile | 1 +
kernel/trace/rv/monitors/safe_wtd/safe_wtd.c | 300 +++++++++++++++++++
kernel/trace/rv/monitors/safe_wtd/safe_wtd.h | 84 ++++++
5 files changed, 406 insertions(+)
create mode 100644 kernel/trace/rv/monitors/safe_wtd/safe_wtd.c
create mode 100644 kernel/trace/rv/monitors/safe_wtd/safe_wtd.h
diff --git a/include/trace/events/rv.h b/include/trace/events/rv.h
index 00f11a8dac3b..895eb3435ed7 100644
--- a/include/trace/events/rv.h
+++ b/include/trace/events/rv.h
@@ -66,6 +66,17 @@ DEFINE_EVENT(error_da_monitor, error_wip,
TP_PROTO(char *state, char *event),
TP_ARGS(state, event));
#endif /* CONFIG_RV_MON_WIP */
+
+#ifdef CONFIG_RV_MON_SAFE_WTD
+DEFINE_EVENT(event_da_monitor, event_safe_wtd,
+ TP_PROTO(char *state, char *event, char *next_state, bool safe),
+ TP_ARGS(state, event, next_state, safe));
+
+DEFINE_EVENT(error_da_monitor, error_safe_wtd,
+ TP_PROTO(char *state, char *event),
+ TP_ARGS(state, event));
+#endif /* CONFIG_RV_MON_SAFE_WTD */
+
#endif /* CONFIG_DA_MON_EVENTS_IMPLICIT */
#ifdef CONFIG_DA_MON_EVENTS_ID
diff --git a/kernel/trace/rv/Kconfig b/kernel/trace/rv/Kconfig
index 21f03fb3101a..b14ae63e792b 100644
--- a/kernel/trace/rv/Kconfig
+++ b/kernel/trace/rv/Kconfig
@@ -45,6 +45,16 @@ config RV_MON_WWNR
illustrates the usage of per-task monitor. The model is
broken on purpose: it serves to test reactors.
+config RV_MON_SAFE_WTD
+ select DA_MON_EVENTS_IMPLICIT
+ bool "Safety watchdog"
+ help
+ Enable safe_wtd, this monitor observes the interaction
+ between a user-space safety monitor and a watchdog device.
+
+ For futher information see:
+ Documentation/trace/rv/safety-monitor.rst
+
config RV_REACTORS
bool "Runtime verification reactors"
default y if RV
diff --git a/kernel/trace/rv/Makefile b/kernel/trace/rv/Makefile
index 963d14875b45..904db96c7eae 100644
--- a/kernel/trace/rv/Makefile
+++ b/kernel/trace/rv/Makefile
@@ -3,6 +3,7 @@
obj-$(CONFIG_RV) += rv.o
obj-$(CONFIG_RV_MON_WIP) += monitors/wip/wip.o
obj-$(CONFIG_RV_MON_WWNR) += monitors/wwnr/wwnr.o
+obj-$(CONFIG_RV_MON_SAFE_WTD) += monitors/safe_wtd/safe_wtd.o
obj-$(CONFIG_RV_REACTORS) += rv_reactors.o
obj-$(CONFIG_RV_REACT_PRINTK) += reactor_printk.o
obj-$(CONFIG_RV_REACT_PANIC) += reactor_panic.o
diff --git a/kernel/trace/rv/monitors/safe_wtd/safe_wtd.c b/kernel/trace/rv/monitors/safe_wtd/safe_wtd.c
new file mode 100644
index 000000000000..9856e0770d0d
--- /dev/null
+++ b/kernel/trace/rv/monitors/safe_wtd/safe_wtd.c
@@ -0,0 +1,300 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/ftrace.h>
+#include <linux/tracepoint.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/rv.h>
+#include <rv/instrumentation.h>
+#include <rv/da_monitor.h>
+
+#include <linux/watchdog.h>
+#include <linux/moduleparam.h>
+
+#include <trace/events/rv.h>
+#include <trace/events/watchdog.h>
+
+#define MODULE_NAME "safe_wtd"
+
+/*
+ * This is the self-generated part of the monitor. Generally, there is no need
+ * to touch this section.
+ */
+#include "safe_wtd.h"
+
+/*
+ * Declare the deterministic automata monitor.
+ *
+ * The rv monitor reference is needed for the monitor declaration.
+ */
+struct rv_monitor rv_safe_wtd;
+DECLARE_DA_MON_GLOBAL(safe_wtd, char);
+
+/*
+ * custom: safe_timeout is the maximum value a watchdog monitor
+ * can set. This value is registered here to duplicate the information.
+ * In this way, a miss-behaving monitor can be detected.
+ */
+static int safe_timeout = ~0;
+module_param(safe_timeout, int, 0444);
+
+/*
+ * custom: if check_timeout is set, the monitor will check if the time left
+ * in the watchdog is less than or equals to the last safe timeout set by
+ * user-space. This check is done after each ping. In this way, if any
+ * code by-passed the watchdog dev interface setting a higher (so unsafe)
+ * timeout, this monitor will catch the side effect and react.
+ */
+static int last_timeout_set = 0;
+static int check_timeout = 0;
+module_param(check_timeout, int, 0444);
+
+/*
+ * custom: if dont_stop is set the monitor will react if stopped.
+ */
+static int dont_stop = 0;
+module_param(dont_stop, int, 0444);
+
+/*
+ * custom: there are some states that are kept after the watchdog is closed.
+ * For example, the nowayout state.
+ *
+ * Thus, the RV monitor needs to keep track of these states after a start/stop
+ * of the RV monitor itself, and should not reset after each restart - keeping the
+ * know state until the system shutdown.
+ *
+ * If for an unknown reason an RV monitor would like to reset the RV monitor at each
+ * RV monitor start, set it to one.
+ */
+static int reset_on_restart = 0;
+module_param(reset_on_restart, int, 0444);
+
+/*
+ * open_pid takes note of the first thread that opened the watchdog.
+ *
+ * Any other thread that generates an event will cause an "other_threads"
+ * event in the monitor.
+ */
+static int open_pid = 0;
+ cond_react(msg);
+ }
+}
+
+static void handle_set_safe_timeout(void *data, struct watchdog_device *wdd, u64 timeout)
+{
+ char msg[128];
+
+ if (wdd->id != watchdog_id)
+ return;
+
+ if (open_pid && current->pid != open_pid) {
+ da_handle_init_run_event_safe_wtd(other_threads_safe_wtd);
+ return;
+ }
+
+ da_handle_event_safe_wtd(set_safe_timeout_safe_wtd);
+
+ if (timeout > safe_timeout) {
+ snprintf(msg, 128, "set safety timeout is too high: %d", (int) timeout);
+ cond_react(msg);
+ }
+
+ if (check_timeout)
+ last_timeout_set = timeout;
+}
+
+static void handle_start(void *data, struct watchdog_device *wdd)
+{
+ if (wdd->id != watchdog_id)
+ return;
+
+ if (open_pid && current->pid != open_pid) {
+ da_handle_init_run_event_safe_wtd(other_threads_safe_wtd);
+ return;
+ }
+
+ da_handle_event_safe_wtd(start_safe_wtd);
+}
+
+static void handle_stop(void *data, struct watchdog_device *wdd)
+{
+ if (wdd->id != watchdog_id)
+ return;
+
+ if (open_pid && current->pid != open_pid) {
+ da_handle_init_run_event_safe_wtd(other_threads_safe_wtd);
+ return;
+ }
+
+ da_handle_event_safe_wtd(stop_safe_wtd);
+}
+
+static int mon_started = 0;
+
+static int start_safe_wtd_monitor(void)
+{
+ int retval;
+
+ if (!mon_started || reset_on_restart) {
+ retval = da_monitor_init_safe_wtd();
+ if (retval)
+ return retval;
+
+ mon_started = 1;
+ }
+
+ rv_attach_trace_probe("safe_wtd", watchdog_close, handle_close);
+ rv_attach_trace_probe("safe_wtd", watchdog_nowayout, handle_nowayout);
+ rv_attach_trace_probe("safe_wtd", watchdog_open, handle_open);
+ rv_attach_trace_probe("safe_wtd", watchdog_ping, handle_ping);
+ rv_attach_trace_probe("safe_wtd", watchdog_set_timeout, handle_set_safe_timeout);
+ rv_attach_trace_probe("safe_wtd", watchdog_start, handle_start);
+ rv_attach_trace_probe("safe_wtd", watchdog_stop, handle_stop);
+ rv_attach_trace_probe("safe_wtd", watchdog_set_keep_alive, blocked_events_timeout);
+ rv_attach_trace_probe("safe_wtd", watchdog_keep_alive, blocked_events);
+ rv_attach_trace_probe("safe_wtd", watchdog_set_pretimeout, blocked_events_timeout);
+ rv_attach_trace_probe("safe_wtd", watchdog_pretimeout, blocked_events);
+
+ return 0;
+}
+
+static void stop_safe_wtd_monitor(void)
+{
+ if (dont_stop)
+ cond_react("dont_stop safe_wtd is set.");
+
+ rv_safe_wtd.enabled = 0;
+
+ rv_detach_trace_probe("safe_wtd", watchdog_close, handle_close);
+ rv_detach_trace_probe("safe_wtd", watchdog_nowayout, handle_nowayout);
+ rv_detach_trace_probe("safe_wtd", watchdog_open, handle_open);
+ rv_detach_trace_probe("safe_wtd", watchdog_ping, handle_ping);
+ rv_detach_trace_probe("safe_wtd", watchdog_set_timeout, handle_set_safe_timeout);
+ rv_detach_trace_probe("safe_wtd", watchdog_start, handle_start);
+ rv_detach_trace_probe("safe_wtd", watchdog_stop, handle_stop);
+ rv_detach_trace_probe("safe_wtd", watchdog_set_keep_alive, blocked_events_timeout);
+ rv_detach_trace_probe("safe_wtd", watchdog_keep_alive, blocked_events);
+ rv_detach_trace_probe("safe_wtd", watchdog_set_pretimeout, blocked_events_timeout);
+ rv_detach_trace_probe("safe_wtd", watchdog_pretimeout, blocked_events);
+
+ da_monitor_destroy_safe_wtd();
+}
+
+/*
+ * This is the monitor register section.
+ */
+struct rv_monitor rv_safe_wtd = {
+ .name = "safe_wtd",
+ .description = "A watchdog monitor guarding a safety monitor actions",
+ .start = start_safe_wtd_monitor,
+ .stop = stop_safe_wtd_monitor,
+ .reset = da_monitor_reset_all_safe_wtd,
+ .enabled = 0,
+};
+
+int register_safe_wtd(void)
+{
+ rv_register_monitor(&rv_safe_wtd);
+ return 0;
+}
+
+void unregister_safe_wtd(void)
+{
+ if (rv_safe_wtd.enabled)
+ stop_safe_wtd_monitor();
+
+ rv_unregister_monitor(&rv_safe_wtd);
+}
+
+module_init(register_safe_wtd);
+module_exit(unregister_safe_wtd);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Daniel Bristot de Oliveira <bristot@xxxxxxxxxx>");
+MODULE_DESCRIPTION("Safe watchdog RV monitor");
diff --git a/kernel/trace/rv/monitors/safe_wtd/safe_wtd.h b/kernel/trace/rv/monitors/safe_wtd/safe_wtd.h
new file mode 100644
index 000000000000..835c9d0979f6
--- /dev/null
+++ b/kernel/trace/rv/monitors/safe_wtd/safe_wtd.h
@@ -0,0 +1,84 @@
+enum states_safe_wtd {
+ init_safe_wtd = 0,
+ closed_running_safe_wtd,
+ closed_running_nwo_safe_wtd,
+ nwo_safe_wtd,
+ opened_safe_wtd,
+ opened_nwo_safe_wtd,
+ reopened_safe_wtd,
+ safe_safe_wtd,
+ safe_nwo_safe_wtd,
+ set_safe_wtd,
+ set_nwo_safe_wtd,
+ started_safe_wtd,
+ started_nwo_safe_wtd,
+ stoped_safe_wtd,
+ state_max_safe_wtd
+};
+
+enum events_safe_wtd {
+ close_safe_wtd = 0,
+ nowayout_safe_wtd,
+ open_safe_wtd,
+ other_threads_safe_wtd,
+ ping_safe_wtd,
+ set_safe_timeout_safe_wtd,
+ start_safe_wtd,
+ stop_safe_wtd,
+ event_max_safe_wtd
+};
+
+struct automaton_safe_wtd {
+ char *state_names[state_max_safe_wtd];
+ char *event_names[event_max_safe_wtd];
+ char function[state_max_safe_wtd][event_max_safe_wtd];
+ char initial_state;
+ char final_states[state_max_safe_wtd];
+};
+
+struct automaton_safe_wtd automaton_safe_wtd = {
+ .state_names = {
+ "init",
+ "closed_running",
+ "closed_running_nwo",
+ "nwo",
+ "opened",
+ "opened_nwo",
+ "reopened",
+ "safe",
+ "safe_nwo",
+ "set",
+ "set_nwo",
+ "started",
+ "started_nwo",
+ "stoped"
+ },
+ .event_names = {
+ "close",
+ "nowayout",
+ "open",
+ "other_threads",
+ "ping",
+ "set_safe_timeout",
+ "start",
+ "stop"
+ },
+ .function = {
+ { -1, nwo_safe_wtd, opened_safe_wtd, init_safe_wtd, -1, -1, -1, -1 },
+ { -1, closed_running_nwo_safe_wtd, reopened_safe_wtd, closed_running_safe_wtd, -1, -1, -1, -1 },
+ { -1, closed_running_nwo_safe_wtd, started_nwo_safe_wtd, closed_running_nwo_safe_wtd, -1, -1, -1, -1 },
+ { -1, nwo_safe_wtd, opened_nwo_safe_wtd, nwo_safe_wtd, -1, -1, -1, -1 },
+ { init_safe_wtd, -1, -1, -1, -1, -1, started_safe_wtd, -1 },
+ { nwo_safe_wtd, -1, -1, -1, -1, -1, started_nwo_safe_wtd, -1 },
+ { closed_running_safe_wtd, -1, -1, -1, -1, set_safe_wtd, -1, opened_safe_wtd },
+ { closed_running_safe_wtd, -1, -1, -1, safe_safe_wtd, -1, -1, stoped_safe_wtd },
+ { closed_running_nwo_safe_wtd, -1, -1, -1, safe_nwo_safe_wtd, -1, -1, -1 },
+ { -1, -1, -1, -1, safe_safe_wtd, -1, -1, -1 },
+ { -1, -1, -1, -1, safe_nwo_safe_wtd, -1, -1, -1 },
+ { closed_running_safe_wtd, -1, -1, -1, -1, set_safe_wtd, -1, stoped_safe_wtd },
+ { closed_running_nwo_safe_wtd, -1, -1, -1, -1, set_nwo_safe_wtd, -1, -1 },
+ { init_safe_wtd, -1, -1, -1, -1, -1, -1, -1 },
+ },
+ .initial_state = init_safe_wtd,
+ .final_states = { 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },
