]> git.ipfire.org Git - thirdparty/kernel/stable.git/commitdiff
rv: Add rtapp_sleep monitor
authorNam Cao <namcao@linutronix.de>
Wed, 9 Jul 2025 19:21:21 +0000 (21:21 +0200)
committerSteven Rostedt (Google) <rostedt@goodmis.org>
Wed, 9 Jul 2025 19:27:01 +0000 (15:27 -0400)
Add a monitor for checking that real-time tasks do not go to sleep in a
manner that may cause undesirable latency.

Also change
RV depends on TRACING
to
RV select TRACING
to avoid the following recursive dependency:

 error: recursive dependency detected!
symbol TRACING is selected by PREEMPTIRQ_TRACEPOINTS
symbol PREEMPTIRQ_TRACEPOINTS depends on TRACE_IRQFLAGS
symbol TRACE_IRQFLAGS is selected by RV_MON_SLEEP
symbol RV_MON_SLEEP depends on RV
symbol RV depends on TRACING

Cc: John Ogness <john.ogness@linutronix.de>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/75bc5bcc741d153aa279c95faf778dff35c5c8ad.1752088709.git.namcao@linutronix.de
Reviewed-by: Gabriele Monaco <gmonaco@redhat.com>
Signed-off-by: Nam Cao <namcao@linutronix.de>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
kernel/trace/rv/Kconfig
kernel/trace/rv/Makefile
kernel/trace/rv/monitors/sleep/Kconfig [new file with mode: 0644]
kernel/trace/rv/monitors/sleep/sleep.c [new file with mode: 0644]
kernel/trace/rv/monitors/sleep/sleep.h [new file with mode: 0644]
kernel/trace/rv/monitors/sleep/sleep_trace.h [new file with mode: 0644]
kernel/trace/rv/rv_trace.h
tools/verification/models/rtapp/sleep.ltl [new file with mode: 0644]

index 6f86d8501e87ec54f1e34390154bb3eaaed0bbb8..942d57575e67b32ae830f04d0f10019974e86510 100644 (file)
@@ -20,7 +20,7 @@ config RV_LTL_MONITOR
 
 menuconfig RV
        bool "Runtime Verification"
-       depends on TRACING
+       select TRACING
        help
          Enable the kernel runtime verification infrastructure. RV is a
          lightweight (yet rigorous) method that complements classical
@@ -43,6 +43,7 @@ source "kernel/trace/rv/monitors/snep/Kconfig"
 source "kernel/trace/rv/monitors/sncid/Kconfig"
 source "kernel/trace/rv/monitors/rtapp/Kconfig"
 source "kernel/trace/rv/monitors/pagefault/Kconfig"
+source "kernel/trace/rv/monitors/sleep/Kconfig"
 # Add new monitors here
 
 config RV_REACTORS
index 353ecf939d0e90f6bb7a968ca5d1d48639b9fbe0..13ec2944c66508c1e191ff80b8f35b6d437a621e 100644 (file)
@@ -14,6 +14,7 @@ obj-$(CONFIG_RV_MON_SNEP) += monitors/snep/snep.o
 obj-$(CONFIG_RV_MON_SNCID) += monitors/sncid/sncid.o
 obj-$(CONFIG_RV_MON_RTAPP) += monitors/rtapp/rtapp.o
 obj-$(CONFIG_RV_MON_PAGEFAULT) += monitors/pagefault/pagefault.o
+obj-$(CONFIG_RV_MON_SLEEP) += monitors/sleep/sleep.o
 # Add new monitors here
 obj-$(CONFIG_RV_REACTORS) += rv_reactors.o
 obj-$(CONFIG_RV_REACT_PRINTK) += reactor_printk.o
diff --git a/kernel/trace/rv/monitors/sleep/Kconfig b/kernel/trace/rv/monitors/sleep/Kconfig
new file mode 100644 (file)
index 0000000..6b7a122
--- /dev/null
@@ -0,0 +1,22 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+config RV_MON_SLEEP
+       depends on RV
+       select RV_LTL_MONITOR
+       depends on HAVE_SYSCALL_TRACEPOINTS
+       depends on RV_MON_RTAPP
+       select TRACE_IRQFLAGS
+       default y
+       select LTL_MON_EVENTS_ID
+       bool "sleep monitor"
+       help
+         Monitor that real-time tasks do not sleep in a manner that may
+         cause undesirable latency.
+
+         If you are developing a real-time system and not entirely sure whether
+         the applications are designed correctly for real-time, you want to say
+         Y here.
+
+         Enabling this monitor may have performance impact (due to select
+         TRACE_IRQFLAGS). Therefore, you probably should say N for
+         production kernel.
diff --git a/kernel/trace/rv/monitors/sleep/sleep.c b/kernel/trace/rv/monitors/sleep/sleep.c
new file mode 100644 (file)
index 0000000..eea447b
--- /dev/null
@@ -0,0 +1,237 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/ftrace.h>
+#include <linux/tracepoint.h>
+#include <linux/init.h>
+#include <linux/irqflags.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/rv.h>
+#include <linux/sched/deadline.h>
+#include <linux/sched/rt.h>
+#include <rv/instrumentation.h>
+
+#define MODULE_NAME "sleep"
+
+#include <trace/events/syscalls.h>
+#include <trace/events/sched.h>
+#include <trace/events/lock.h>
+#include <uapi/linux/futex.h>
+#include <rv_trace.h>
+#include <monitors/rtapp/rtapp.h>
+
+#include "sleep.h"
+#include <rv/ltl_monitor.h>
+
+static void ltl_atoms_fetch(struct task_struct *task, struct ltl_monitor *mon)
+{
+       /*
+        * This includes "actual" real-time tasks and also PI-boosted
+        * tasks. A task being PI-boosted means it is blocking an "actual"
+        * real-task, therefore it should also obey the monitor's rule,
+        * otherwise the "actual" real-task may be delayed.
+        */
+       ltl_atom_set(mon, LTL_RT, rt_or_dl_task(task));
+}
+
+static void ltl_atoms_init(struct task_struct *task, struct ltl_monitor *mon, bool task_creation)
+{
+       ltl_atom_set(mon, LTL_SLEEP, false);
+       ltl_atom_set(mon, LTL_WAKE, false);
+       ltl_atom_set(mon, LTL_ABORT_SLEEP, false);
+       ltl_atom_set(mon, LTL_WOKEN_BY_HARDIRQ, false);
+       ltl_atom_set(mon, LTL_WOKEN_BY_NMI, false);
+       ltl_atom_set(mon, LTL_WOKEN_BY_EQUAL_OR_HIGHER_PRIO, false);
+
+       if (task_creation) {
+               ltl_atom_set(mon, LTL_KTHREAD_SHOULD_STOP, false);
+               ltl_atom_set(mon, LTL_NANOSLEEP_CLOCK_MONOTONIC, false);
+               ltl_atom_set(mon, LTL_NANOSLEEP_CLOCK_TAI, false);
+               ltl_atom_set(mon, LTL_NANOSLEEP_TIMER_ABSTIME, false);
+               ltl_atom_set(mon, LTL_CLOCK_NANOSLEEP, false);
+               ltl_atom_set(mon, LTL_FUTEX_WAIT, false);
+               ltl_atom_set(mon, LTL_FUTEX_LOCK_PI, false);
+               ltl_atom_set(mon, LTL_BLOCK_ON_RT_MUTEX, false);
+       }
+
+       if (task->flags & PF_KTHREAD) {
+               ltl_atom_set(mon, LTL_KERNEL_THREAD, true);
+
+               /* kernel tasks do not do syscall */
+               ltl_atom_set(mon, LTL_FUTEX_WAIT, false);
+               ltl_atom_set(mon, LTL_FUTEX_LOCK_PI, false);
+               ltl_atom_set(mon, LTL_NANOSLEEP_CLOCK_MONOTONIC, false);
+               ltl_atom_set(mon, LTL_NANOSLEEP_CLOCK_TAI, false);
+               ltl_atom_set(mon, LTL_NANOSLEEP_TIMER_ABSTIME, false);
+               ltl_atom_set(mon, LTL_CLOCK_NANOSLEEP, false);
+
+               if (strstarts(task->comm, "migration/"))
+                       ltl_atom_set(mon, LTL_TASK_IS_MIGRATION, true);
+               else
+                       ltl_atom_set(mon, LTL_TASK_IS_MIGRATION, false);
+
+               if (strstarts(task->comm, "rcu"))
+                       ltl_atom_set(mon, LTL_TASK_IS_RCU, true);
+               else
+                       ltl_atom_set(mon, LTL_TASK_IS_RCU, false);
+       } else {
+               ltl_atom_set(mon, LTL_KTHREAD_SHOULD_STOP, false);
+               ltl_atom_set(mon, LTL_KERNEL_THREAD, false);
+               ltl_atom_set(mon, LTL_TASK_IS_RCU, false);
+               ltl_atom_set(mon, LTL_TASK_IS_MIGRATION, false);
+       }
+
+}
+
+static void handle_sched_set_state(void *data, struct task_struct *task, int state)
+{
+       if (state & TASK_INTERRUPTIBLE)
+               ltl_atom_pulse(task, LTL_SLEEP, true);
+       else if (state == TASK_RUNNING)
+               ltl_atom_pulse(task, LTL_ABORT_SLEEP, true);
+}
+
+static void handle_sched_wakeup(void *data, struct task_struct *task)
+{
+       ltl_atom_pulse(task, LTL_WAKE, true);
+}
+
+static void handle_sched_waking(void *data, struct task_struct *task)
+{
+       if (this_cpu_read(hardirq_context)) {
+               ltl_atom_pulse(task, LTL_WOKEN_BY_HARDIRQ, true);
+       } else if (in_task()) {
+               if (current->prio <= task->prio)
+                       ltl_atom_pulse(task, LTL_WOKEN_BY_EQUAL_OR_HIGHER_PRIO, true);
+       } else if (in_nmi()) {
+               ltl_atom_pulse(task, LTL_WOKEN_BY_NMI, true);
+       }
+}
+
+static void handle_contention_begin(void *data, void *lock, unsigned int flags)
+{
+       if (flags & LCB_F_RT)
+               ltl_atom_update(current, LTL_BLOCK_ON_RT_MUTEX, true);
+}
+
+static void handle_contention_end(void *data, void *lock, int ret)
+{
+       ltl_atom_update(current, LTL_BLOCK_ON_RT_MUTEX, false);
+}
+
+static void handle_sys_enter(void *data, struct pt_regs *regs, long id)
+{
+       struct ltl_monitor *mon;
+       unsigned long args[6];
+       int op, cmd;
+
+       mon = ltl_get_monitor(current);
+
+       switch (id) {
+       case __NR_clock_nanosleep:
+#ifdef __NR_clock_nanosleep_time64
+       case __NR_clock_nanosleep_time64:
+#endif
+               syscall_get_arguments(current, regs, args);
+               ltl_atom_set(mon, LTL_NANOSLEEP_CLOCK_MONOTONIC, args[0] == CLOCK_MONOTONIC);
+               ltl_atom_set(mon, LTL_NANOSLEEP_CLOCK_TAI, args[0] == CLOCK_TAI);
+               ltl_atom_set(mon, LTL_NANOSLEEP_TIMER_ABSTIME, args[1] == TIMER_ABSTIME);
+               ltl_atom_update(current, LTL_CLOCK_NANOSLEEP, true);
+               break;
+
+       case __NR_futex:
+#ifdef __NR_futex_time64
+       case __NR_futex_time64:
+#endif
+               syscall_get_arguments(current, regs, args);
+               op = args[1];
+               cmd = op & FUTEX_CMD_MASK;
+
+               switch (cmd) {
+               case FUTEX_LOCK_PI:
+               case FUTEX_LOCK_PI2:
+                       ltl_atom_update(current, LTL_FUTEX_LOCK_PI, true);
+                       break;
+               case FUTEX_WAIT:
+               case FUTEX_WAIT_BITSET:
+               case FUTEX_WAIT_REQUEUE_PI:
+                       ltl_atom_update(current, LTL_FUTEX_WAIT, true);
+                       break;
+               }
+               break;
+       }
+}
+
+static void handle_sys_exit(void *data, struct pt_regs *regs, long ret)
+{
+       struct ltl_monitor *mon = ltl_get_monitor(current);
+
+       ltl_atom_set(mon, LTL_FUTEX_LOCK_PI, false);
+       ltl_atom_set(mon, LTL_FUTEX_WAIT, false);
+       ltl_atom_set(mon, LTL_NANOSLEEP_CLOCK_MONOTONIC, false);
+       ltl_atom_set(mon, LTL_NANOSLEEP_CLOCK_TAI, false);
+       ltl_atom_set(mon, LTL_NANOSLEEP_TIMER_ABSTIME, false);
+       ltl_atom_update(current, LTL_CLOCK_NANOSLEEP, false);
+}
+
+static void handle_kthread_stop(void *data, struct task_struct *task)
+{
+       /* FIXME: this could race with other tracepoint handlers */
+       ltl_atom_update(task, LTL_KTHREAD_SHOULD_STOP, true);
+}
+
+static int enable_sleep(void)
+{
+       int retval;
+
+       retval = ltl_monitor_init();
+       if (retval)
+               return retval;
+
+       rv_attach_trace_probe("rtapp_sleep", sched_waking, handle_sched_waking);
+       rv_attach_trace_probe("rtapp_sleep", sched_wakeup, handle_sched_wakeup);
+       rv_attach_trace_probe("rtapp_sleep", sched_set_state_tp, handle_sched_set_state);
+       rv_attach_trace_probe("rtapp_sleep", contention_begin, handle_contention_begin);
+       rv_attach_trace_probe("rtapp_sleep", contention_end, handle_contention_end);
+       rv_attach_trace_probe("rtapp_sleep", sched_kthread_stop, handle_kthread_stop);
+       rv_attach_trace_probe("rtapp_sleep", sys_enter, handle_sys_enter);
+       rv_attach_trace_probe("rtapp_sleep", sys_exit, handle_sys_exit);
+       return 0;
+}
+
+static void disable_sleep(void)
+{
+       rv_detach_trace_probe("rtapp_sleep", sched_waking, handle_sched_waking);
+       rv_detach_trace_probe("rtapp_sleep", sched_wakeup, handle_sched_wakeup);
+       rv_detach_trace_probe("rtapp_sleep", sched_set_state_tp, handle_sched_set_state);
+       rv_detach_trace_probe("rtapp_sleep", contention_begin, handle_contention_begin);
+       rv_detach_trace_probe("rtapp_sleep", contention_end, handle_contention_end);
+       rv_detach_trace_probe("rtapp_sleep", sched_kthread_stop, handle_kthread_stop);
+       rv_detach_trace_probe("rtapp_sleep", sys_enter, handle_sys_enter);
+       rv_detach_trace_probe("rtapp_sleep", sys_exit, handle_sys_exit);
+
+       ltl_monitor_destroy();
+}
+
+static struct rv_monitor rv_sleep = {
+       .name = "sleep",
+       .description = "Monitor that RT tasks do not undesirably sleep",
+       .enable = enable_sleep,
+       .disable = disable_sleep,
+};
+
+static int __init register_sleep(void)
+{
+       return rv_register_monitor(&rv_sleep, &rv_rtapp);
+}
+
+static void __exit unregister_sleep(void)
+{
+       rv_unregister_monitor(&rv_sleep);
+}
+
+module_init(register_sleep);
+module_exit(unregister_sleep);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Nam Cao <namcao@linutronix.de>");
+MODULE_DESCRIPTION("sleep: Monitor that RT tasks do not undesirably sleep");
diff --git a/kernel/trace/rv/monitors/sleep/sleep.h b/kernel/trace/rv/monitors/sleep/sleep.h
new file mode 100644 (file)
index 0000000..2ab46fd
--- /dev/null
@@ -0,0 +1,257 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+/*
+ * C implementation of Buchi automaton, automatically generated by
+ * tools/verification/rvgen from the linear temporal logic specification.
+ * For further information, see kernel documentation:
+ *   Documentation/trace/rv/linear_temporal_logic.rst
+ */
+
+#include <linux/rv.h>
+
+#define MONITOR_NAME sleep
+
+enum ltl_atom {
+       LTL_ABORT_SLEEP,
+       LTL_BLOCK_ON_RT_MUTEX,
+       LTL_CLOCK_NANOSLEEP,
+       LTL_FUTEX_LOCK_PI,
+       LTL_FUTEX_WAIT,
+       LTL_KERNEL_THREAD,
+       LTL_KTHREAD_SHOULD_STOP,
+       LTL_NANOSLEEP_CLOCK_MONOTONIC,
+       LTL_NANOSLEEP_CLOCK_TAI,
+       LTL_NANOSLEEP_TIMER_ABSTIME,
+       LTL_RT,
+       LTL_SLEEP,
+       LTL_TASK_IS_MIGRATION,
+       LTL_TASK_IS_RCU,
+       LTL_WAKE,
+       LTL_WOKEN_BY_EQUAL_OR_HIGHER_PRIO,
+       LTL_WOKEN_BY_HARDIRQ,
+       LTL_WOKEN_BY_NMI,
+       LTL_NUM_ATOM
+};
+static_assert(LTL_NUM_ATOM <= RV_MAX_LTL_ATOM);
+
+static const char *ltl_atom_str(enum ltl_atom atom)
+{
+       static const char *const names[] = {
+               "ab_sl",
+               "bl_on_rt_mu",
+               "cl_na",
+               "fu_lo_pi",
+               "fu_wa",
+               "ker_th",
+               "kth_sh_st",
+               "na_cl_mo",
+               "na_cl_ta",
+               "na_ti_ab",
+               "rt",
+               "sl",
+               "ta_mi",
+               "ta_rc",
+               "wak",
+               "wo_eq_hi_pr",
+               "wo_ha",
+               "wo_nm",
+       };
+
+       return names[atom];
+}
+
+enum ltl_buchi_state {
+       S0,
+       S1,
+       S2,
+       S3,
+       S4,
+       S5,
+       S6,
+       S7,
+       RV_NUM_BA_STATES
+};
+static_assert(RV_NUM_BA_STATES <= RV_MAX_BA_STATES);
+
+static void ltl_start(struct task_struct *task, struct ltl_monitor *mon)
+{
+       bool task_is_migration = test_bit(LTL_TASK_IS_MIGRATION, mon->atoms);
+       bool task_is_rcu = test_bit(LTL_TASK_IS_RCU, mon->atoms);
+       bool val40 = task_is_rcu || task_is_migration;
+       bool futex_lock_pi = test_bit(LTL_FUTEX_LOCK_PI, mon->atoms);
+       bool val41 = futex_lock_pi || val40;
+       bool block_on_rt_mutex = test_bit(LTL_BLOCK_ON_RT_MUTEX, mon->atoms);
+       bool val5 = block_on_rt_mutex || val41;
+       bool kthread_should_stop = test_bit(LTL_KTHREAD_SHOULD_STOP, mon->atoms);
+       bool abort_sleep = test_bit(LTL_ABORT_SLEEP, mon->atoms);
+       bool val32 = abort_sleep || kthread_should_stop;
+       bool woken_by_nmi = test_bit(LTL_WOKEN_BY_NMI, mon->atoms);
+       bool val33 = woken_by_nmi || val32;
+       bool woken_by_hardirq = test_bit(LTL_WOKEN_BY_HARDIRQ, mon->atoms);
+       bool val34 = woken_by_hardirq || val33;
+       bool woken_by_equal_or_higher_prio = test_bit(LTL_WOKEN_BY_EQUAL_OR_HIGHER_PRIO,
+            mon->atoms);
+       bool val14 = woken_by_equal_or_higher_prio || val34;
+       bool wake = test_bit(LTL_WAKE, mon->atoms);
+       bool val13 = !wake;
+       bool kernel_thread = test_bit(LTL_KERNEL_THREAD, mon->atoms);
+       bool nanosleep_clock_tai = test_bit(LTL_NANOSLEEP_CLOCK_TAI, mon->atoms);
+       bool nanosleep_clock_monotonic = test_bit(LTL_NANOSLEEP_CLOCK_MONOTONIC, mon->atoms);
+       bool val24 = nanosleep_clock_monotonic || nanosleep_clock_tai;
+       bool nanosleep_timer_abstime = test_bit(LTL_NANOSLEEP_TIMER_ABSTIME, mon->atoms);
+       bool val25 = nanosleep_timer_abstime && val24;
+       bool clock_nanosleep = test_bit(LTL_CLOCK_NANOSLEEP, mon->atoms);
+       bool val18 = clock_nanosleep && val25;
+       bool futex_wait = test_bit(LTL_FUTEX_WAIT, mon->atoms);
+       bool val9 = futex_wait || val18;
+       bool val11 = val9 || kernel_thread;
+       bool sleep = test_bit(LTL_SLEEP, mon->atoms);
+       bool val2 = !sleep;
+       bool rt = test_bit(LTL_RT, mon->atoms);
+       bool val1 = !rt;
+       bool val3 = val1 || val2;
+
+       if (val3)
+               __set_bit(S0, mon->states);
+       if (val11 && val13)
+               __set_bit(S1, mon->states);
+       if (val11 && val14)
+               __set_bit(S4, mon->states);
+       if (val5)
+               __set_bit(S5, mon->states);
+}
+
+static void
+ltl_possible_next_states(struct ltl_monitor *mon, unsigned int state, unsigned long *next)
+{
+       bool task_is_migration = test_bit(LTL_TASK_IS_MIGRATION, mon->atoms);
+       bool task_is_rcu = test_bit(LTL_TASK_IS_RCU, mon->atoms);
+       bool val40 = task_is_rcu || task_is_migration;
+       bool futex_lock_pi = test_bit(LTL_FUTEX_LOCK_PI, mon->atoms);
+       bool val41 = futex_lock_pi || val40;
+       bool block_on_rt_mutex = test_bit(LTL_BLOCK_ON_RT_MUTEX, mon->atoms);
+       bool val5 = block_on_rt_mutex || val41;
+       bool kthread_should_stop = test_bit(LTL_KTHREAD_SHOULD_STOP, mon->atoms);
+       bool abort_sleep = test_bit(LTL_ABORT_SLEEP, mon->atoms);
+       bool val32 = abort_sleep || kthread_should_stop;
+       bool woken_by_nmi = test_bit(LTL_WOKEN_BY_NMI, mon->atoms);
+       bool val33 = woken_by_nmi || val32;
+       bool woken_by_hardirq = test_bit(LTL_WOKEN_BY_HARDIRQ, mon->atoms);
+       bool val34 = woken_by_hardirq || val33;
+       bool woken_by_equal_or_higher_prio = test_bit(LTL_WOKEN_BY_EQUAL_OR_HIGHER_PRIO,
+            mon->atoms);
+       bool val14 = woken_by_equal_or_higher_prio || val34;
+       bool wake = test_bit(LTL_WAKE, mon->atoms);
+       bool val13 = !wake;
+       bool kernel_thread = test_bit(LTL_KERNEL_THREAD, mon->atoms);
+       bool nanosleep_clock_tai = test_bit(LTL_NANOSLEEP_CLOCK_TAI, mon->atoms);
+       bool nanosleep_clock_monotonic = test_bit(LTL_NANOSLEEP_CLOCK_MONOTONIC, mon->atoms);
+       bool val24 = nanosleep_clock_monotonic || nanosleep_clock_tai;
+       bool nanosleep_timer_abstime = test_bit(LTL_NANOSLEEP_TIMER_ABSTIME, mon->atoms);
+       bool val25 = nanosleep_timer_abstime && val24;
+       bool clock_nanosleep = test_bit(LTL_CLOCK_NANOSLEEP, mon->atoms);
+       bool val18 = clock_nanosleep && val25;
+       bool futex_wait = test_bit(LTL_FUTEX_WAIT, mon->atoms);
+       bool val9 = futex_wait || val18;
+       bool val11 = val9 || kernel_thread;
+       bool sleep = test_bit(LTL_SLEEP, mon->atoms);
+       bool val2 = !sleep;
+       bool rt = test_bit(LTL_RT, mon->atoms);
+       bool val1 = !rt;
+       bool val3 = val1 || val2;
+
+       switch (state) {
+       case S0:
+               if (val3)
+                       __set_bit(S0, next);
+               if (val11 && val13)
+                       __set_bit(S1, next);
+               if (val11 && val14)
+                       __set_bit(S4, next);
+               if (val5)
+                       __set_bit(S5, next);
+               break;
+       case S1:
+               if (val11 && val13)
+                       __set_bit(S1, next);
+               if (val13 && val3)
+                       __set_bit(S2, next);
+               if (val14 && val3)
+                       __set_bit(S3, next);
+               if (val11 && val14)
+                       __set_bit(S4, next);
+               if (val13 && val5)
+                       __set_bit(S6, next);
+               if (val14 && val5)
+                       __set_bit(S7, next);
+               break;
+       case S2:
+               if (val11 && val13)
+                       __set_bit(S1, next);
+               if (val13 && val3)
+                       __set_bit(S2, next);
+               if (val14 && val3)
+                       __set_bit(S3, next);
+               if (val11 && val14)
+                       __set_bit(S4, next);
+               if (val13 && val5)
+                       __set_bit(S6, next);
+               if (val14 && val5)
+                       __set_bit(S7, next);
+               break;
+       case S3:
+               if (val3)
+                       __set_bit(S0, next);
+               if (val11 && val13)
+                       __set_bit(S1, next);
+               if (val11 && val14)
+                       __set_bit(S4, next);
+               if (val5)
+                       __set_bit(S5, next);
+               break;
+       case S4:
+               if (val3)
+                       __set_bit(S0, next);
+               if (val11 && val13)
+                       __set_bit(S1, next);
+               if (val11 && val14)
+                       __set_bit(S4, next);
+               if (val5)
+                       __set_bit(S5, next);
+               break;
+       case S5:
+               if (val3)
+                       __set_bit(S0, next);
+               if (val11 && val13)
+                       __set_bit(S1, next);
+               if (val11 && val14)
+                       __set_bit(S4, next);
+               if (val5)
+                       __set_bit(S5, next);
+               break;
+       case S6:
+               if (val11 && val13)
+                       __set_bit(S1, next);
+               if (val13 && val3)
+                       __set_bit(S2, next);
+               if (val14 && val3)
+                       __set_bit(S3, next);
+               if (val11 && val14)
+                       __set_bit(S4, next);
+               if (val13 && val5)
+                       __set_bit(S6, next);
+               if (val14 && val5)
+                       __set_bit(S7, next);
+               break;
+       case S7:
+               if (val3)
+                       __set_bit(S0, next);
+               if (val11 && val13)
+                       __set_bit(S1, next);
+               if (val11 && val14)
+                       __set_bit(S4, next);
+               if (val5)
+                       __set_bit(S5, next);
+               break;
+       }
+}
diff --git a/kernel/trace/rv/monitors/sleep/sleep_trace.h b/kernel/trace/rv/monitors/sleep/sleep_trace.h
new file mode 100644 (file)
index 0000000..22eaf31
--- /dev/null
@@ -0,0 +1,14 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+/*
+ * Snippet to be included in rv_trace.h
+ */
+
+#ifdef CONFIG_RV_MON_SLEEP
+DEFINE_EVENT(event_ltl_monitor_id, event_sleep,
+            TP_PROTO(struct task_struct *task, char *states, char *atoms, char *next),
+            TP_ARGS(task, states, atoms, next));
+DEFINE_EVENT(error_ltl_monitor_id, error_sleep,
+            TP_PROTO(struct task_struct *task),
+            TP_ARGS(task));
+#endif /* CONFIG_RV_MON_SLEEP */
index 98eee8ec96e4fc94cd8a4b400f7895d92374f787..b6f3104984664888a4e49c13fa7571888b447006 100644 (file)
@@ -173,6 +173,7 @@ DECLARE_EVENT_CLASS(error_ltl_monitor_id,
        TP_printk("%s[%d]: violation detected", __get_str(comm), __entry->pid)
 );
 #include <monitors/pagefault/pagefault_trace.h>
+#include <monitors/sleep/sleep_trace.h>
 // Add new monitors based on CONFIG_LTL_MON_EVENTS_ID here
 #endif /* CONFIG_LTL_MON_EVENTS_ID */
 #endif /* _TRACE_RV_H */
diff --git a/tools/verification/models/rtapp/sleep.ltl b/tools/verification/models/rtapp/sleep.ltl
new file mode 100644 (file)
index 0000000..6379bbe
--- /dev/null
@@ -0,0 +1,22 @@
+RULE = always ((RT and SLEEP) imply (RT_FRIENDLY_SLEEP or ALLOWLIST))
+
+RT_FRIENDLY_SLEEP = (RT_VALID_SLEEP_REASON or KERNEL_THREAD)
+                and ((not WAKE) until RT_FRIENDLY_WAKE)
+
+RT_VALID_SLEEP_REASON = FUTEX_WAIT
+                     or RT_FRIENDLY_NANOSLEEP
+
+RT_FRIENDLY_NANOSLEEP = CLOCK_NANOSLEEP
+                    and NANOSLEEP_TIMER_ABSTIME
+                    and (NANOSLEEP_CLOCK_MONOTONIC or NANOSLEEP_CLOCK_TAI)
+
+RT_FRIENDLY_WAKE = WOKEN_BY_EQUAL_OR_HIGHER_PRIO
+                or WOKEN_BY_HARDIRQ
+                or WOKEN_BY_NMI
+                or ABORT_SLEEP
+                or KTHREAD_SHOULD_STOP
+
+ALLOWLIST = BLOCK_ON_RT_MUTEX
+         or FUTEX_LOCK_PI
+         or TASK_IS_RCU
+         or TASK_IS_MIGRATION