freezer: implement and use kthread_freezable_should_stop()

Writeback and thinkpad_acpi have been using thaw_process() to prevent
deadlock between the freezer and kthread_stop(); unfortunately, this
is inherently racy - nothing prevents freezing from happening between
thaw_process() and kthread_stop().

This patch implements kthread_freezable_should_stop() which enters
refrigerator if necessary but is guaranteed to return if
kthread_stop() is invoked.  Both thaw_process() users are converted to
use the new function.

Note that this deadlock condition exists for many of freezable
kthreads.  They need to be converted to use the new should_stop or
freezable workqueue.

Tested with synthetic test case.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Henrique de Moraes Holschuh <ibm-acpi@hmh.eng.br>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Oleg Nesterov <oleg@redhat.com>

drivers/platform/x86/thinkpad_acpi.c

@@ -2456,8 +2456,9 @@ static int hotkey_kthread(void *data)
 	u32 poll_mask, event_mask;
 	unsigned int si, so;
 	unsigned long t;
-	unsigned int change_detector, must_reset;
+	unsigned int change_detector;
 	unsigned int poll_freq;
+	bool was_frozen;
 
 	mutex_lock(&hotkey_thread_mutex);
 
@@ -2488,14 +2489,14 @@ static int hotkey_kthread(void *data)
 				t = 100;	/* should never happen... */
 		}
 		t = msleep_interruptible(t);
-		if (unlikely(kthread_should_stop()))
+		if (unlikely(kthread_freezable_should_stop(&was_frozen)))
 			break;
-		must_reset = try_to_freeze();
-		if (t > 0 && !must_reset)
+
+		if (t > 0 && !was_frozen)
 			continue;
 
 		mutex_lock(&hotkey_thread_data_mutex);
-		if (must_reset || hotkey_config_change != change_detector) {
+		if (was_frozen || hotkey_config_change != change_detector) {
 			/* forget old state on thaw or config change */
 			si = so;
 			t = 0;
@@ -2528,10 +2529,6 @@ exit:
 static void hotkey_poll_stop_sync(void)
 {
 	if (tpacpi_hotkey_task) {
-		if (frozen(tpacpi_hotkey_task) ||
-		    freezing(tpacpi_hotkey_task))
-			thaw_process(tpacpi_hotkey_task);
-
 		kthread_stop(tpacpi_hotkey_task);
 		tpacpi_hotkey_task = NULL;
 		mutex_lock(&hotkey_thread_mutex);

fs/fs-writeback.c

@@ -947,7 +947,7 @@ int bdi_writeback_thread(void *data)
 
 	trace_writeback_thread_start(bdi);
 
-	while (!kthread_should_stop()) {
+	while (!kthread_freezable_should_stop(NULL)) {
 		/*
 		 * Remove own delayed wake-up timer, since we are already awake
 		 * and we'll take care of the preriodic write-back.
@@ -977,8 +977,6 @@ int bdi_writeback_thread(void *data)
 			 */
 			schedule();
 		}
-
-		try_to_freeze();
 	}
 
 	/* Flush any work that raced with us exiting */

include/linux/freezer.h

@@ -47,7 +47,7 @@ static inline bool should_send_signal(struct task_struct *p)
 /* Takes and releases task alloc lock using task_lock() */
 extern int thaw_process(struct task_struct *p);
 
-extern bool __refrigerator(void);
+extern bool __refrigerator(bool check_kthr_stop);
 extern int freeze_processes(void);
 extern int freeze_kernel_threads(void);
 extern void thaw_processes(void);
@@ -57,7 +57,7 @@ static inline bool try_to_freeze(void)
 	might_sleep();
 	if (likely(!freezing(current)))
 		return false;
-	return __refrigerator();
+	return __refrigerator(false);
 }
 
 extern bool freeze_task(struct task_struct *p, bool sig_only);
@@ -180,7 +180,7 @@ static inline void set_freeze_flag(struct task_struct *p) {}
 static inline void clear_freeze_flag(struct task_struct *p) {}
 static inline int thaw_process(struct task_struct *p) { return 1; }
 
-static inline bool __refrigerator(void) { return false; }
+static inline bool __refrigerator(bool check_kthr_stop) { return false; }
 static inline int freeze_processes(void) { return -ENOSYS; }
 static inline int freeze_kernel_threads(void) { return -ENOSYS; }
 static inline void thaw_processes(void) {}

include/linux/kthread.h

@@ -35,6 +35,7 @@ struct task_struct *kthread_create_on_node(int (*threadfn)(void *data),
 void kthread_bind(struct task_struct *k, unsigned int cpu);
 int kthread_stop(struct task_struct *k);
 int kthread_should_stop(void);
+bool kthread_freezable_should_stop(bool *was_frozen);
 void *kthread_data(struct task_struct *k);
 
 int kthreadd(void *unused);

kernel/freezer.c

@@ -9,6 +9,7 @@
 #include <linux/export.h>
 #include <linux/syscalls.h>
 #include <linux/freezer.h>
+#include <linux/kthread.h>
 
 /*
  * freezing is complete, mark current process as frozen
@@ -23,7 +24,7 @@ static inline void frozen_process(void)
 }
 
 /* Refrigerator is place where frozen processes are stored :-). */
-bool __refrigerator(void)
+bool __refrigerator(bool check_kthr_stop)
 {
 	/* Hmm, should we be allowed to suspend when there are realtime
 	   processes around? */
@@ -50,7 +51,8 @@ bool __refrigerator(void)
 
 	for (;;) {
 		set_current_state(TASK_UNINTERRUPTIBLE);
-		if (!frozen(current))
+		if (!frozen(current) ||
+		    (check_kthr_stop && kthread_should_stop()))
 			break;
 		was_frozen = true;
 		schedule();

kernel/kthread.c

@@ -58,6 +58,31 @@ int kthread_should_stop(void)
 }
 EXPORT_SYMBOL(kthread_should_stop);
 
+/**
+ * kthread_freezable_should_stop - should this freezable kthread return now?
+ * @was_frozen: optional out parameter, indicates whether %current was frozen
+ *
+ * kthread_should_stop() for freezable kthreads, which will enter
+ * refrigerator if necessary.  This function is safe from kthread_stop() /
+ * freezer deadlock and freezable kthreads should use this function instead
+ * of calling try_to_freeze() directly.
+ */
+bool kthread_freezable_should_stop(bool *was_frozen)
+{
+	bool frozen = false;
+
+	might_sleep();
+
+	if (unlikely(freezing(current)))
+		frozen = __refrigerator(true);
+
+	if (was_frozen)
+		*was_frozen = frozen;
+
+	return kthread_should_stop();
+}
+EXPORT_SYMBOL_GPL(kthread_freezable_should_stop);
+
 /**
  * kthread_data - return data value specified on kthread creation
  * @task: kthread task in question

mm/backing-dev.c

@@ -600,15 +600,11 @@ static void bdi_wb_shutdown(struct backing_dev_info *bdi)
 
 	/*
 	 * Finally, kill the kernel thread. We don't need to be RCU
-	 * safe anymore, since the bdi is gone from visibility. Force
-	 * unfreeze of the thread before calling kthread_stop(), otherwise
-	 * it would never exet if it is currently stuck in the refrigerator.
+	 * safe anymore, since the bdi is gone from visibility.
 	 */
-	if (bdi->wb.task) {
-		thaw_process(bdi->wb.task);
+	if (bdi->wb.task)
 		kthread_stop(bdi->wb.task);
 }
-}
 
 /*
  * This bdi is going away now, make sure that no super_blocks point to it