Extend kqueue interface with EVFILT_EXCEPT filter.

This filter, already implemented in macOS and Dragonfly BSD, returns
exceptional conditions like the reception of out-of-band data.

The functionnality is similar to poll(2)'s POLLPRI & POLLRDBAND and
it can be used by the kqfilter-based poll & select implementation.

ok millert@ on a previous version, ok visa@

5 files changed, 60 insertions, 8 deletions

diff --git a/lib/libc/sys/kqueue.2 b/lib/libc/sys/kqueue.2
index fd5cc41c248..66c92b166c7 100644
--- a/lib/libc/sys/kqueue.2
+++ b/lib/libc/sys/kqueue.2
@@ -1,4 +1,4 @@
-.\"	$OpenBSD: kqueue.2,v 1.40 2020/05/31 03:49:44 visa Exp $
+.\"	$OpenBSD: kqueue.2,v 1.41 2020/06/22 13:14:32 mpi Exp $
 .\"
 .\" Copyright (c) 2000 Jonathan Lemon
 .\" All rights reserved.
@@ -26,7 +26,7 @@
 .\"
 .\" $FreeBSD: src/lib/libc/sys/kqueue.2,v 1.18 2001/02/14 08:48:35 guido Exp $
 .\"
-.Dd $Mdocdate: May 31 2020 $
+.Dd $Mdocdate: June 22 2020 $
 .Dt KQUEUE 2
 .Os
 .Sh NAME
@@ -310,6 +310,13 @@ enabled and there is any data to read;
 .Fa data
 contains the number of bytes available.
 .El
+.It Dv EVFILT_EXCEPT
+Takes a descriptor as the identifier, and returns whenever one of the
+specified exceptional conditions has occured on the descriptor.
+Conditions are specified in
+.Fa fflags .
+Currently, a filter can monitor the reception of out-of-band data with
+.Dv NOTE_OOB .
 .It Dv EVFILT_WRITE
 Takes a descriptor as the identifier, and returns whenever
 it is possible to write to the descriptor.
diff --git a/sys/kern/kern_event.c b/sys/kern/kern_event.c
index 4d0e2c88c5f..ddaf08795ec 100644
--- a/sys/kern/kern_event.c
+++ b/sys/kern/kern_event.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: kern_event.c,v 1.139 2020/06/15 15:42:11 mpi Exp $	*/
+/*	$OpenBSD: kern_event.c,v 1.140 2020/06/22 13:14:32 mpi Exp $	*/
 
 /*-
  * Copyright (c) 1999,2000,2001 Jonathan Lemon <jlemon@FreeBSD.org>
@@ -158,6 +158,7 @@ const struct filterops *const sysfilt_ops[] = {
 	&sig_filtops,			/* EVFILT_SIGNAL */
 	&timer_filtops,			/* EVFILT_TIMER */
 	&file_filtops,			/* EVFILT_DEVICE */
+	&file_filtops,			/* EVFILT_EXCEPT */
 };
 
 void
diff --git a/sys/kern/tty_pty.c b/sys/kern/tty_pty.c
index 97f41437f8d..18c6a058005 100644
--- a/sys/kern/tty_pty.c
+++ b/sys/kern/tty_pty.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: tty_pty.c,v 1.100 2020/06/15 15:29:40 mpi Exp $	*/
+/*	$OpenBSD: tty_pty.c,v 1.101 2020/06/22 13:14:32 mpi Exp $	*/
 /*	$NetBSD: tty_pty.c,v 1.33.4.1 1996/06/02 09:08:11 mrg Exp $	*/
 
 /*
@@ -668,6 +668,16 @@ filt_ptcread(struct knote *kn, long hint)
 	tp = pti->pt_tty;
 	kn->kn_data = 0;
 
+	if (kn->kn_sfflags & NOTE_OOB) {
+		/* If in packet or user control mode, check for data. */
+		if (((pti->pt_flags & PF_PKT) && pti->pt_send) ||
+		    ((pti->pt_flags & PF_UCNTL) && pti->pt_ucntl)) {
+			kn->kn_fflags |= NOTE_OOB;
+			kn->kn_data = 1;
+			return (1);
+		}
+		return (0);
+	}
 	if (ISSET(tp->t_state, TS_ISOPEN)) {
 		if (!ISSET(tp->t_state, TS_TTSTOP))
 			kn->kn_data = tp->t_outq.c_cc;
@@ -733,6 +743,13 @@ const struct filterops ptcwrite_filtops = {
 	.f_event	= filt_ptcwrite,
 };
 
+const struct filterops ptcexcept_filtops = {
+	.f_flags	= FILTEROP_ISFD,
+	.f_attach	= NULL,
+	.f_detach	= filt_ptcrdetach,
+	.f_event	= filt_ptcread,
+};
+
 int
 ptckqfilter(dev_t dev, struct knote *kn)
 {
@@ -749,6 +766,10 @@ ptckqfilter(dev_t dev, struct knote *kn)
 		klist = &pti->pt_selw.si_note;
 		kn->kn_fop = &ptcwrite_filtops;
 		break;
+	case EVFILT_EXCEPT:
+		klist = &pti->pt_selr.si_note;
+		kn->kn_fop = &ptcexcept_filtops;
+		break;
 	default:
 		return (EINVAL);
 	}
diff --git a/sys/kern/uipc_socket.c b/sys/kern/uipc_socket.c
index 3d6e12e88bc..8f3ac59436c 100644
--- a/sys/kern/uipc_socket.c
+++ b/sys/kern/uipc_socket.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: uipc_socket.c,v 1.246 2020/06/18 14:05:21 mvs Exp $	*/
+/*	$OpenBSD: uipc_socket.c,v 1.247 2020/06/22 13:14:32 mpi Exp $	*/
 /*	$NetBSD: uipc_socket.c,v 1.21 1996/02/04 02:17:52 christos Exp $	*/
 
 /*
@@ -93,6 +93,12 @@ const struct filterops sowrite_filtops = {
 	.f_event	= filt_sowrite,
 };
 
+const struct filterops soexcept_filtops = {
+	.f_flags	= FILTEROP_ISFD,
+	.f_attach	= NULL,
+	.f_detach	= filt_sordetach,
+	.f_event	= filt_soread,
+};
 
 #ifndef SOMINCONN
 #define SOMINCONN 80
@@ -2026,6 +2032,10 @@ soo_kqfilter(struct file *fp, struct knote *kn)
 		kn->kn_fop = &sowrite_filtops;
 		sb = &so->so_snd;
 		break;
+	case EVFILT_EXCEPT:
+		kn->kn_fop = &soexcept_filtops;
+		sb = &so->so_rcv;
+		break;
 	default:
 		return (EINVAL);
 	}
@@ -2052,7 +2062,7 @@ int
 filt_soread(struct knote *kn, long hint)
 {
 	struct socket *so = kn->kn_fp->f_data;
-	int s, rv;
+	int s, rv = 0;
 
 	if ((hint & NOTE_SUBMIT) == 0)
 		s = solock(so);
@@ -2062,7 +2072,13 @@ filt_soread(struct knote *kn, long hint)
 		rv = 0;
 	} else
 #endif /* SOCKET_SPLICE */
-	if (so->so_state & SS_CANTRCVMORE) {
+	if (kn->kn_sfflags & NOTE_OOB) {
+		if (so->so_oobmark || (so->so_state & SS_RCVATMARK)) {
+			kn->kn_fflags |= NOTE_OOB;
+			kn->kn_data -= so->so_oobmark;
+			rv = 1;
+		}
+	} else if (so->so_state & SS_CANTRCVMORE) {
 		kn->kn_flags |= EV_EOF;
 		if (kn->kn_flags & __EV_POLL) {
 			if (so->so_state & SS_ISDISCONNECTED)
diff --git a/sys/sys/event.h b/sys/sys/event.h
index 93a8a904d90..4a46d97af72 100644
--- a/sys/sys/event.h
+++ b/sys/sys/event.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: event.h,v 1.43 2020/06/15 15:42:11 mpi Exp $	*/
+/*	$OpenBSD: event.h,v 1.44 2020/06/22 13:14:32 mpi Exp $	*/
 
 /*-
  * Copyright (c) 1999,2000,2001 Jonathan Lemon <jlemon@FreeBSD.org>
@@ -39,6 +39,7 @@
 #define EVFILT_SIGNAL		(-6)	/* attached to struct process */
 #define EVFILT_TIMER		(-7)	/* timers */
 #define EVFILT_DEVICE		(-8)	/* devices */
+#define EVFILT_EXCEPT		(-9)	/* exceptional conditions */
 
 #define EVFILT_SYSCOUNT		8
 
@@ -87,6 +88,12 @@ struct kevent {
 #define NOTE_EOF	0x0002			/* return on EOF */
 
 /*
+ * data/hint flags for EVFILT_EXCEPT, shared with userspace and with
+ * EVFILT_{READ|WRITE}
+ */
+#define NOTE_OOB	0x0004			/* OOB data on a socket */
+
+/*
  * data/hint flags for EVFILT_VNODE, shared with userspace
  */
 #define	NOTE_DELETE	0x0001			/* vnode was removed */