[PATCH] arm64: suspend: Only track cpuidle context for idle tasks

[PATCH] arm64: suspend: Only track cpuidle context for idle tasks

[syzbot]

From: Uladzislau Zhauniarovich <uladzislau.zhauniarovich@gmail.com>

Commit 19235e472798 ("cpuidle, arm64: Fix the ARM64 cpuidle logic") moved
the calls to ct_cpuidle_enter() and ct_cpuidle_exit() into cpu_suspend() so
that arm64 cpuidle could keep RCU watching longer and manage context
tracking internally. However, cpu_suspend() is called not only by cpuidle
via psci_cpu_suspend_enter() (which runs in the context of the idle task),
but also during system suspend via psci_system_suspend_enter() (which runs
in the context of the task initiating suspend, such as a process writing to
/sys/power/state).

When entering system suspend, current is a non-idle task. Calling
ct_cpuidle_enter() invokes ct_kernel_exit() from kernel space (!user) on a
non-idle task. With CONFIG_RCU_EQS_DEBUG enabled, this violates context
tracking invariants and triggers a warning:

WARNING: kernel/context_tracking.c:120 at ct_kernel_exit+0x1b4/0x1f4
kernel/context_tracking.c:118
Call trace:
 ct_kernel_exit+0x1b4/0x1f4 kernel/context_tracking.c:118
 ct_idle_enter+0x20/0x34 kernel/context_tracking.c:299
 ct_cpuidle_enter include/linux/cpuidle.h:134 [inline]
 cpu_suspend+0x264/0x474 arch/arm64/kernel/suspend.c:136
 psci_system_suspend_enter+0x4c/0x80 drivers/firmware/psci/psci.c:544
 suspend_enter kernel/power/suspend.c:468 [inline]
 suspend_devices_and_enter+0x88c/0xd0c kernel/power/suspend.c:533
 enter_state kernel/power/suspend.c:618 [inline]
 pm_suspend+0x498/0x698 kernel/power/suspend.c:644
 state_store+0x14c/0x1fc kernel/power/main.c:819
 kobj_attr_store+0x6c/0x90 lib/kobject.c:842
 sysfs_kf_write+0xe0/0x108 fs/sysfs/file.c:145
 kernfs_fop_write_iter+0x260/0x3b8 fs/kernfs/file.c:345
 vfs_write+0x51c/0xa0c fs/read_write.c:687
 ksys_write+0x12c/0x228 fs/read_write.c:739

A corresponding warning is subsequently triggered in ct_kernel_enter() via
ct_cpuidle_exit() upon resume.

Fix this by checking is_idle_task(current) in cpu_suspend() and only
invoking ct_cpuidle_enter() and ct_cpuidle_exit() when executing in the
context of an idle task.

Fixes: 19235e472798 ("cpuidle, arm64: Fix the ARM64 cpuidle logic")
Assisted-by: Gemini:gemini-3.8-flash Gemini:gemini-3.1-pro-preview syzbot
Reported-by: syzbot+79883955da3a62652a80@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=79883955da3a62652a80
Link: https://syzkaller.appspot.com/ai_job?id=5359e16f-013b-4850-9cbc-ddfb3bac95e5
Signed-off-by: Uladzislau Zhauniarovich <uladzislau.zhauniarovich@gmail.com>

---
diff --git a/arch/arm64/kernel/suspend.c b/arch/arm64/kernel/suspend.c
index c41724a40..cde0b8c19 100644
--- a/arch/arm64/kernel/suspend.c
+++ b/arch/arm64/kernel/suspend.c
@@ -5,6 +5,7 @@
 #include <linux/uaccess.h>
 #include <linux/pgtable.h>
 #include <linux/cpuidle.h>
+#include <linux/sched.h>
 #include <asm/alternative.h>
 #include <asm/cacheflush.h>
 #include <asm/cpufeature.h>
@@ -99,6 +100,7 @@ int cpu_suspend(unsigned long arg, int (*fn)(unsigned long))
 	int ret = 0;
 	unsigned long flags;
 	struct sleep_stack_data state;
+	bool idle = is_idle_task(current);
 
 	/*
 	 * Some portions of CPU state (e.g. PSTATE.{PAN,DIT}) are initialized
@@ -133,7 +135,8 @@ int cpu_suspend(unsigned long arg, int (*fn)(unsigned long))
 	 */
 	pause_graph_tracing();
 
-	ct_cpuidle_enter();
+	if (idle)
+		ct_cpuidle_enter();
 
 	if (__cpu_suspend_enter(&state)) {
 		/* Call the suspend finisher */
@@ -149,9 +152,11 @@ int cpu_suspend(unsigned long arg, int (*fn)(unsigned long))
 		if (!ret)
 			ret = -EOPNOTSUPP;
 
-		ct_cpuidle_exit();
+		if (idle)
+			ct_cpuidle_exit();
 	} else {
-		ct_cpuidle_exit();
+		if (idle)
+			ct_cpuidle_exit();
 		__cpu_suspend_exit();
 	}
 


base-commit: 72d3fcf802c45d00b300f25b848a93c3a2bd7c7e
-- 
See https://goo.gle/syzbot-ai-patches for information about AI-generated patches.
The person who has signed off on the patch is responsible for
addressing comments.
syzbot engineers can be reached at syzkaller@googlegroups.com.