kernel/thread: restore signal mask just before returning to userland

* otherwise the signal to be handled might be blocked. fixes #15193
* also remove automatic syscall restart on _kern_select, to match Linux and
BSDs behavior: this fixes parallel build with newer gnu make, which happens
to use pselect.
* also remove automatic syscall restart on _kern_poll.

from https://man7.org/linux/man-pages/man7/signal.7.html
"The following interfaces are never restarted after being
       interrupted by a signal handler, regardless of the use of
       SA_RESTART; they always fail with the error EINTR when
       interrupted by a signal handler: ...
	select(2), and pselect(2)."
from https://notes.shichao.io/unp/ch6/
"Berkeley-derived kernels never automatically restart select."

Change-Id: I3e9488f60c966b38d427f992f06e6e2217d4adc5
Reviewed-on: https://review.haiku-os.org/c/haiku/+/3636
Reviewed-by: Jérôme Duval <jerome.duval@gmail.com>
Reviewed-by: Axel Dörfler <axeld@pinc-software.de>

headers/private/kernel/thread_types.h

@@ -455,6 +455,10 @@ struct Thread : TeamThreadIteratorEntry<thread_id>, KernelReferenceable {
 		// non-0 after a return from _user_sigsuspend(), containing the inverted
 		// original signal mask, reset in handle_signals(); only accessed by
 		// this thread
+	sigset_t		old_sig_block_mask;
+		// the old sig_block_mask to be restored when returning to userland
+		// when THREAD_FLAGS_OLD_SIGMASK is set
+
 	ucontext_t*		user_signal_context;	// only accessed by this thread
 	addr_t			signal_stack_base;		// only accessed by this thread
 	size_t			signal_stack_size;		// only accessed by this thread
@@ -844,5 +848,7 @@ using BKernel::ProcessGroupList;
 #define	THREAD_FLAGS_COMPAT_MODE			0x2000
 	// the thread runs a compatibility mode (for instance IA32 on x86_64).
 #endif
+#define	THREAD_FLAGS_OLD_SIGMASK			0x4000
+	// the thread has an old sigmask to be restored
 
 #endif	/* _KERNEL_THREAD_TYPES_H */

src/system/kernel/thread.cpp

@@ -1938,6 +1938,16 @@ dump_thread_list(int argc, char **argv)
 }
 
 
+static void
+update_thread_sigmask_on_exit(Thread* thread)
+{
+	if ((thread->flags & THREAD_FLAGS_OLD_SIGMASK) != 0) {
+		thread->flags &= ~THREAD_FLAGS_OLD_SIGMASK;
+		sigprocmask(SIG_SETMASK, &thread->old_sig_block_mask, NULL);
+	}
+}
+
+
 //	#pragma mark - private kernel API
 
 
@@ -2318,6 +2328,8 @@ thread_at_kernel_exit(void)
 
 	disable_interrupts();
 
+	update_thread_sigmask_on_exit(thread);
+
 	// track kernel time
 	bigtime_t now = system_time();
 	SpinLocker threadTimeLocker(thread->time_lock);
@@ -2339,6 +2351,8 @@ thread_at_kernel_exit_no_signals(void)
 	TRACE(("thread_at_kernel_exit_no_signals: exit thread %" B_PRId32 "\n",
 		thread->id));
 
+	update_thread_sigmask_on_exit(thread);
+
 	// track kernel time
 	bigtime_t now = system_time();
 	SpinLocker threadTimeLocker(thread->time_lock);

src/system/kernel/wait_for_objects.cpp

@@ -475,15 +475,21 @@ common_select(int numFDs, fd_set *readSet, fd_set *writeSet, fd_set *errorSet,
 
 	// set new signal mask
 	sigset_t oldSigMask;
-	if (sigMask != NULL)
+	if (sigMask != NULL) {
 		sigprocmask(SIG_SETMASK, sigMask, &oldSigMask);
+		if (!kernel) {
+			Thread *thread = thread_get_current_thread();
+			thread->old_sig_block_mask = oldSigMask;
+			thread->flags |= THREAD_FLAGS_OLD_SIGMASK;
+		}
+	}
 
 	// wait for something to happen
 	status = acquire_sem_etc(sync->sem, 1,
 		B_CAN_INTERRUPT | (timeout >= 0 ? B_ABSOLUTE_TIMEOUT : 0), timeout);
 
 	// restore the old signal mask
-	if (sigMask != NULL)
+	if (sigMask != NULL && kernel)
 		sigprocmask(SIG_SETMASK, &oldSigMask, NULL);
 
 	PRINT(("common_select(): acquire_sem_etc() returned: %lx\n", status));
@@ -927,7 +933,12 @@ _user_select(int numFDs, fd_set *userReadSet, fd_set *userWriteSet,
 	sigset_t sigMask;
 	int result;
 
-	syscall_restart_handle_timeout_pre(timeout);
+	if (timeout >= 0) {
+		timeout += system_time();
+		// deal with overflow
+		if (timeout < 0)
+			timeout = B_INFINITE_TIMEOUT;
+	}
 
 	if (numFDs < 0 || !check_max_fds(numFDs))
 		return B_BAD_VALUE;
@@ -994,8 +1005,7 @@ _user_select(int numFDs, fd_set *userReadSet, fd_set *userWriteSet,
 			|| (errorSet != NULL
 				&& user_memcpy(userErrorSet, errorSet, bytes) < B_OK))) {
 		result = B_BAD_ADDRESS;
-	} else
+	}
-		syscall_restart_handle_timeout_post(result, timeout);
 
 err:
 	free(readSet);
@@ -1013,16 +1023,19 @@ _user_poll(struct pollfd *userfds, int numFDs, bigtime_t timeout)
 	size_t bytes;
 	int result;
 
-	syscall_restart_handle_timeout_pre(timeout);
+	if (timeout >= 0) {
+		timeout += system_time();
+		// deal with overflow
+		if (timeout < 0)
+			timeout = B_INFINITE_TIMEOUT;
+	}
 
 	if (numFDs < 0)
 		return B_BAD_VALUE;
 
 	if (numFDs == 0) {
 		// special case: no FDs
-		result = common_poll(NULL, 0, timeout, false);
+		return common_poll(NULL, 0, timeout, false);
-		return result < 0
-			? syscall_restart_handle_timeout_post(result, timeout) : result;
 	}
 
 	if (!check_max_fds(numFDs))
@@ -1047,9 +1060,7 @@ _user_poll(struct pollfd *userfds, int numFDs, bigtime_t timeout)
 	if (numFDs > 0 && user_memcpy(userfds, fds, bytes) != 0) {
 		if (result >= 0)
 			result = B_BAD_ADDRESS;
-	} else
+	}
-		syscall_restart_handle_timeout_post(result, timeout);
-
 err:
 	free(fds);