entry: Add support for TIF_NOTIFY_SIGNAL

Add TIF_NOTIFY_SIGNAL handling in the generic entry code, which if set,
will return true if signal_pending() is used in a wait loop. That causes an
exit of the loop so that notify_signal tracehooks can be run. If the wait
loop is currently inside a system call, the system call is restarted once
task_work has been processed.

In preparation for only having arch_do_signal() handle syscall restarts if
_TIF_SIGPENDING isn't set, rename it to arch_do_signal_or_restart(). Pass
in a boolean that tells the architecture specific signal handler if it
should attempt to get a signal, or just process a potential syscall
restart.

For !CONFIG_GENERIC_ENTRY archs, add the TIF_NOTIFY_SIGNAL handling to
get_signal(). This is done to minimize the needed architecture changes to
support this feature.

Signed-off-by: Jens Axboe <[e-mail]>
Signed-off-by: Thomas Gleixner <[e-mail]>
Reviewed-by: Oleg Nesterov <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]

arch/x86/kernel/signal.c

···
  * want to handle. Thus you cannot kill init even with a SIGKILL even by
  * mistake.
  */
-void arch_do_signal(struct pt_regs *regs)
+void arch_do_signal_or_restart(struct pt_regs *regs, bool has_signal)
 {
 	struct ksignal ksig;
 
-	if (get_signal(&ksig)) {
+	if (has_signal && get_signal(&ksig)) {
 		/* Whee! Actually deliver the signal.  */
 		handle_signal(&ksig, regs);
 		return;

include/linux/entry-common.h

···
 # define _TIF_UPROBE			(0)
 #endif
 
+#ifndef _TIF_NOTIFY_SIGNAL
+# define _TIF_NOTIFY_SIGNAL		(0)
+#endif
+
 /*
  * TIF flags handled in syscall_enter_from_usermode()
  */
···
 
 #define EXIT_TO_USER_MODE_WORK						\
 	(_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_UPROBE |		\
-	 _TIF_NEED_RESCHED | _TIF_PATCH_PENDING |			\
+	 _TIF_NEED_RESCHED | _TIF_PATCH_PENDING | _TIF_NOTIFY_SIGNAL |	\
 	 ARCH_EXIT_TO_USER_MODE_WORK)
 
 /**
···
 #endif
 
 /**
- * arch_do_signal -  Architecture specific signal delivery function
+ * arch_do_signal_or_restart -  Architecture specific signal delivery function
  * @regs:	Pointer to currents pt_regs
+ * @has_signal:	actual signal to handle
  *
  * Invoked from exit_to_user_mode_loop().
  */
-void arch_do_signal(struct pt_regs *regs);
+void arch_do_signal_or_restart(struct pt_regs *regs, bool has_signal);
 
 /**
  * arch_syscall_exit_tracehook - Wrapper around tracehook_report_syscall_exit()

include/linux/entry-kvm.h

···
 # define ARCH_XFER_TO_GUEST_MODE_WORK	(0)
 #endif
 
-#define XFER_TO_GUEST_MODE_WORK					\
-	(_TIF_NEED_RESCHED | _TIF_SIGPENDING |			\
+#define XFER_TO_GUEST_MODE_WORK						\
+	(_TIF_NEED_RESCHED | _TIF_SIGPENDING | _TIF_NOTIFY_SIGNAL |	\
 	 _TIF_NOTIFY_RESUME | ARCH_XFER_TO_GUEST_MODE_WORK)
 
 struct kvm_vcpu;

include/linux/sched/signal.h

···
 
 static inline int signal_pending(struct task_struct *p)
 {
+#if defined(TIF_NOTIFY_SIGNAL)
+	/*
+	 * TIF_NOTIFY_SIGNAL isn't really a signal, but it requires the same
+	 * behavior in terms of ensuring that we break out of wait loops
+	 * so that notify signal callbacks can be processed.
+	 */
+	if (unlikely(test_tsk_thread_flag(p, TIF_NOTIFY_SIGNAL)))
+		return 1;
+#endif
 	return task_sigpending(p);
 }
 
···
 static inline void restore_saved_sigmask_unless(bool interrupted)
 {
 	if (interrupted)
-		WARN_ON(!test_thread_flag(TIF_SIGPENDING));
+		WARN_ON(!signal_pending(current));
 	else
 		restore_saved_sigmask();
 }

include/linux/tracehook.h

···
 	blkcg_maybe_throttle_current();
 }
 
+/*
+ * called by exit_to_user_mode_loop() if ti_work & _TIF_NOTIFY_SIGNAL. This
+ * is currently used by TWA_SIGNAL based task_work, which requires breaking
+ * wait loops to ensure that task_work is noticed and run.
+ */
+static inline void tracehook_notify_signal(void)
+{
+#if defined(TIF_NOTIFY_SIGNAL)
+	clear_thread_flag(TIF_NOTIFY_SIGNAL);
+	smp_mb__after_atomic();
+	if (current->task_works)
+		task_work_run();
+#endif
+}
+
+/*
+ * Called when we have work to process from exit_to_user_mode_loop()
+ */
+static inline void set_notify_signal(struct task_struct *task)
+{
+#if defined(TIF_NOTIFY_SIGNAL)
+	if (!test_and_set_tsk_thread_flag(task, TIF_NOTIFY_SIGNAL) &&
+	    !wake_up_state(task, TASK_INTERRUPTIBLE))
+		kick_process(task);
+#endif
+}
+
 #endif	/* <linux/tracehook.h> */

kernel/entry/common.c

···
 }
 
 /* Workaround to allow gradual conversion of architecture code */
-void __weak arch_do_signal(struct pt_regs *regs) { }
+void __weak arch_do_signal_or_restart(struct pt_regs *regs, bool has_signal) { }
+
+static void handle_signal_work(struct pt_regs *regs, unsigned long ti_work)
+{
+	if (ti_work & _TIF_NOTIFY_SIGNAL)
+		tracehook_notify_signal();
+
+	arch_do_signal_or_restart(regs, ti_work & _TIF_SIGPENDING);
+}
 
 static unsigned long exit_to_user_mode_loop(struct pt_regs *regs,
 					    unsigned long ti_work)
···
 		if (ti_work & _TIF_PATCH_PENDING)
 			klp_update_patch_state(current);
 
-		if (ti_work & _TIF_SIGPENDING)
-			arch_do_signal(regs);
+		if (ti_work & (_TIF_SIGPENDING | _TIF_NOTIFY_SIGNAL))
+			handle_signal_work(regs, ti_work);
 
 		if (ti_work & _TIF_NOTIFY_RESUME) {
 			clear_thread_flag(TIF_NOTIFY_RESUME);

kernel/entry/kvm.c

···
 	do {
 		int ret;
 
+		if (ti_work & _TIF_NOTIFY_SIGNAL)
+			tracehook_notify_signal();
+
 		if (ti_work & _TIF_SIGPENDING) {
 			kvm_handle_signal_exit(vcpu);
 			return -EINTR;

kernel/signal.c

···
 	struct signal_struct *signal = current->signal;
 	int signr;
 
+	/*
+	 * For non-generic architectures, check for TIF_NOTIFY_SIGNAL so
+	 * that the arch handlers don't all have to do it. If we get here
+	 * without TIF_SIGPENDING, just exit after running signal work.
+	 */
+#ifdef TIF_NOTIFY_SIGNAL
+	if (!IS_ENABLED(CONFIG_GENERIC_ENTRY)) {
+		if (test_thread_flag(TIF_NOTIFY_SIGNAL))
+			tracehook_notify_signal();
+		if (!task_sigpending(current))
+			return false;
+	}
+#endif
+
 	if (unlikely(uprobe_deny_signal()))
 		return false;