- Introduce the net.bpf.tscfg sysctl tree and associated code so as to make one

  aspect of time stamp configuration per interface rather than per BPF
  descriptor. Prior to this, the order in which BPF devices were opened and the
  per descriptor time stamp configuration settings could cause non-deterministic
  and unintended behaviour with respect to time stamping. With the new scheme, a
  BPF attached interface's tscfg sysctl entry can be set to "default", "none",
  "fast", "normal" or "external". Setting "default" means use the system default
  option (set with the net.bpf.tscfg.default sysctl), "none" means do not
  generate time stamps for tapped packets, "fast" means generate time stamps for
  tapped packets using a hz granularity system clock read, "normal" means
  generate time stamps for tapped packets using a full timecounter granularity
  system clock read and "external" (currently unimplemented) means use the time
  stamp provided with the packet from an underlying source.

- Utilise the recently introduced sysclock_getsnapshot() and
  sysclock_snap2bintime() KPIs to ensure the system clock is only read once per
  packet, regardless of the number of BPF descriptors and time stamp formats
  requested. Use the per BPF attached interface time stamp configuration to
  control if sysclock_getsnapshot() is called and whether the system clock read
  is fast or normal. The per BPF descriptor time stamp configuration is then
  used to control how the system clock snapshot is converted to a bintime by
  sysclock_snap2bintime().

- Remove all FAST related BPF descriptor flag variants. Performing a "fast"
  read of the system clock is now controlled per BPF attached interface using
  the net.bpf.tscfg sysctl tree.

- Update the bpf.4 man page.

Committed on behalf of Julien Ridoux and Darryl Veitch from the University of
Melbourne, Australia, as part of the FreeBSD Foundation funded "Feed-Forward
Clock Synchronization Algorithms" project.

For more information, see http://www.synclab.org/radclock/

In collaboration with:	Julien Ridoux (jridoux at unimelb edu au)

2 files changed, 193 insertions, 82 deletions

diff --git a/sys/net/bpf.c b/sys/net/bpf.c
index 3dd2f93..db71e39 100644
--- a/sys/net/bpf.c
+++ b/sys/net/bpf.c
@@ -1,12 +1,17 @@
 /*-
  * Copyright (c) 1990, 1991, 1993
- *	The Regents of the University of California.  All rights reserved.
+ *	The Regents of the University of California.
+ * Copyright (c) 2011 The University of Melbourne.
+ * All rights reserved.
  *
  * This code is derived from the Stanford/CMU enet packet filter,
  * (net/enet.c) distributed as part of 4.3BSD, and code contributed
  * to Berkeley by Steven McCanne and Van Jacobson both of Lawrence
  * Berkeley Laboratory.
  *
+ * Portions of this software were developed by Julien Ridoux at the University
+ * of Melbourne under sponsorship from the FreeBSD Foundation.
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -55,6 +60,7 @@ __FBSDID("$FreeBSD$");
 #include <sys/signalvar.h>
 #include <sys/filio.h>
 #include <sys/sockio.h>
+#include <sys/timeffc.h>
 #include <sys/ttycom.h>
 #include <sys/uio.h>
 
@@ -112,7 +118,7 @@ struct bpf_hdr32 {
 	uint16_t	bh_hdrlen;	/* length of bpf header (this struct
 					   plus alignment padding) */
 };
-#endif
+#endif /* !BURN_BRIDGES */
 
 struct bpf_program32 {
 	u_int bf_len;
@@ -130,7 +136,28 @@ struct bpf_dltlist32 {
 #define	BIOCGDLTLIST32	_IOWR('B', 121, struct bpf_dltlist32)
 #define	BIOCSETWF32	_IOW('B', 123, struct bpf_program32)
 #define	BIOCSETFNR32	_IOW('B', 130, struct bpf_program32)
-#endif
+#endif /* COMPAT_FREEBSD32 */
+
+static const char *bpfiftstypes[] = {
+	"default",
+#define	BPF_TSTAMP_DEFAULT	0
+	"none",
+#define	BPF_TSTAMP_NONE		1
+	"fast",
+#define	BPF_TSTAMP_FAST		2
+	"normal",
+#define	BPF_TSTAMP_NORMAL	3
+	"external"
+#define	BPF_TSTAMP_EXTERNAL	4
+};
+#define	NUM_BPFIFTSTYPES	(sizeof(bpfiftstypes) / sizeof(*bpfiftstypes))
+
+#define	SET_CLOCKCFG_FLAGS(tstype, active, clock, flags) do {		\
+	(flags) = 0;							\
+	(clock) = SYSCLOCK_FBCK;					\
+	if ((tstype) & BPF_T_MONOTONIC)					\
+		(flags) |= FBCLOCK_UPTIME;				\
+} while (0)
 
 /*
  * bpf_iflist is a list of BPF interface structures, each corresponding to a
@@ -162,6 +189,7 @@ static void	filt_bpfdetach(struct knote *);
 static int	filt_bpfread(struct knote *, long);
 static void	bpf_drvinit(void *);
 static int	bpf_stats_sysctl(SYSCTL_HANDLER_ARGS);
+static int	bpf_tscfg_sysctl_handler(SYSCTL_HANDLER_ARGS);
 
 SYSCTL_NODE(_net, OID_AUTO, bpf, CTLFLAG_RW, 0, "bpf sysctl");
 int bpf_maxinsns = BPF_MAXINSNS;
@@ -172,6 +200,12 @@ SYSCTL_INT(_net_bpf, OID_AUTO, zerocopy_enable, CTLFLAG_RW,
     &bpf_zerocopy_enable, 0, "Enable new zero-copy BPF buffer sessions");
 static SYSCTL_NODE(_net_bpf, OID_AUTO, stats, CTLFLAG_MPSAFE | CTLFLAG_RW,
     bpf_stats_sysctl, "bpf statistics portal");
+static SYSCTL_NODE(_net_bpf, OID_AUTO, tscfg, CTLFLAG_RW, NULL,
+    "Per-interface timestamp configuration");
+static int bpf_default_tstype = BPF_TSTAMP_NORMAL;
+SYSCTL_PROC(_net_bpf_tscfg, OID_AUTO, default,
+    CTLTYPE_STRING | CTLFLAG_RW, NULL, 0, bpf_tscfg_sysctl_handler, "A",
+    "Per-interface system wide default timestamp configuration");
 
 static	d_open_t	bpfopen;
 static	d_read_t	bpfread;
@@ -1759,48 +1793,6 @@ filt_bpfread(struct knote *kn, long hint)
 	return (ready);
 }
 
-#define	BPF_TSTAMP_NONE		0
-#define	BPF_TSTAMP_FAST		1
-#define	BPF_TSTAMP_NORMAL	2
-#define	BPF_TSTAMP_EXTERN	3
-
-static int
-bpf_ts_quality(int tstype)
-{
-
-	if (tstype == BPF_T_NONE)
-		return (BPF_TSTAMP_NONE);
-	if ((tstype & BPF_T_FAST) != 0)
-		return (BPF_TSTAMP_FAST);
-
-	return (BPF_TSTAMP_NORMAL);
-}
-
-static int
-bpf_gettime(struct bintime *bt, int tstype, struct mbuf *m)
-{
-	struct m_tag *tag;
-	int quality;
-
-	quality = bpf_ts_quality(tstype);
-	if (quality == BPF_TSTAMP_NONE)
-		return (quality);
-
-	if (m != NULL) {
-		tag = m_tag_locate(m, MTAG_BPF, MTAG_BPF_TIMESTAMP, NULL);
-		if (tag != NULL) {
-			*bt = *(struct bintime *)(tag + 1);
-			return (BPF_TSTAMP_EXTERN);
-		}
-	}
-	if (quality == BPF_TSTAMP_NORMAL)
-		binuptime(bt);
-	else
-		getbinuptime(bt);
-
-	return (quality);
-}
-
 /*
  * Incoming linkage from device drivers.  Process the packet pkt, of length
  * pktlen, which is stored in a contiguous buffer.  The packet is parsed
@@ -1811,14 +1803,23 @@ void
 bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
 {
 	struct bintime bt;
+	struct sysclock_snap cs;
 	struct bpf_d *d;
+	int tstype, whichclock;
+	u_int clockflags, slen;
 #ifdef BPF_JITTER
 	bpf_jit_filter *bf;
 #endif
-	u_int slen;
-	int gottime;
 
-	gottime = BPF_TSTAMP_NONE;
+	tstype = bp->tstype;
+	if (tstype == BPF_TSTAMP_DEFAULT)
+		tstype = bpf_default_tstype;
+
+	if (tstype == BPF_TSTAMP_NORMAL || tstype == BPF_TSTAMP_FAST)
+		sysclock_getsnapshot(&cs, tstype == BPF_TSTAMP_FAST ? 1 : 0);
+	else
+		bzero(&bt, sizeof(bt));
+
 	BPFIF_LOCK(bp);
 	LIST_FOREACH(d, &bp->bif_dlist, bd_next) {
 		BPFD_LOCK(d);
@@ -1838,8 +1839,16 @@ bpf_tap(struct bpf_if *bp, u_char *pkt, u_int pktlen)
 		slen = bpf_filter(d->bd_rfilter, pkt, pktlen, pktlen);
 		if (slen != 0) {
 			d->bd_fcount++;
-			if (gottime < bpf_ts_quality(d->bd_tstamp))
-				gottime = bpf_gettime(&bt, d->bd_tstamp, NULL);
+			if (tstype == BPF_TSTAMP_NORMAL ||
+			    tstype == BPF_TSTAMP_FAST) {
+				whichclock = -1;
+				SET_CLOCKCFG_FLAGS(d->bd_tstamp,
+				    cs.sysclock_active, whichclock, clockflags);
+				KASSERT(whichclock >= 0, ("Bogus BPF tstamp "
+				    "configuration: 0x%04x", d->bd_tstamp));
+				sysclock_snap2bintime(&cs, &bt, whichclock,
+				    clockflags);
+			}
 #ifdef MAC
 			if (mac_bpfdesc_check_receive(d, bp->bif_ifp) == 0)
 #endif
@@ -1862,12 +1871,13 @@ void
 bpf_mtap(struct bpf_if *bp, struct mbuf *m)
 {
 	struct bintime bt;
+	struct sysclock_snap cs;
 	struct bpf_d *d;
+	u_int clockflags, pktlen, slen;
+	int tstype, whichclock;
 #ifdef BPF_JITTER
 	bpf_jit_filter *bf;
 #endif
-	u_int pktlen, slen;
-	int gottime;
 
 	/* Skip outgoing duplicate packets. */
 	if ((m->m_flags & M_PROMISC) != 0 && m->m_pkthdr.rcvif == NULL) {
@@ -1875,9 +1885,22 @@ bpf_mtap(struct bpf_if *bp, struct mbuf *m)
 		return;
 	}
 
+	tstype = bp->tstype;
+	if (tstype == BPF_TSTAMP_DEFAULT)
+		tstype = bpf_default_tstype;
+
+	if (tstype == BPF_TSTAMP_NORMAL || tstype == BPF_TSTAMP_FAST)
+		sysclock_getsnapshot(&cs, tstype == BPF_TSTAMP_FAST ?
+		    1 : 0);
+#ifdef notyet
+	else if (tstype == BPF_TSTAMP_EXTERNAL)
+		/* XXX: Convert external tstamp to bintime. */
+#endif
+	else
+		bzero(&bt, sizeof(bt));
+
 	pktlen = m_length(m, NULL);
 
-	gottime = BPF_TSTAMP_NONE;
 	BPFIF_LOCK(bp);
 	LIST_FOREACH(d, &bp->bif_dlist, bd_next) {
 		if (BPF_CHECK_DIRECTION(d, m->m_pkthdr.rcvif, bp->bif_ifp))
@@ -1894,8 +1917,16 @@ bpf_mtap(struct bpf_if *bp, struct mbuf *m)
 		slen = bpf_filter(d->bd_rfilter, (u_char *)m, pktlen, 0);
 		if (slen != 0) {
 			d->bd_fcount++;
-			if (gottime < bpf_ts_quality(d->bd_tstamp))
-				gottime = bpf_gettime(&bt, d->bd_tstamp, m);
+			if (tstype == BPF_TSTAMP_NORMAL ||
+			    tstype == BPF_TSTAMP_FAST) {
+				whichclock = -1;
+				SET_CLOCKCFG_FLAGS(d->bd_tstamp,
+				    cs.sysclock_active, whichclock, clockflags);
+				KASSERT(whichclock >= 0, ("Bogus BPF tstamp "
+				    "configuration: 0x%04x", d->bd_tstamp));
+				sysclock_snap2bintime(&cs, &bt, whichclock,
+				    clockflags);
+			}
 #ifdef MAC
 			if (mac_bpfdesc_check_receive(d, bp->bif_ifp) == 0)
 #endif
@@ -1915,10 +1946,11 @@ void
 bpf_mtap2(struct bpf_if *bp, void *data, u_int dlen, struct mbuf *m)
 {
 	struct bintime bt;
+	struct sysclock_snap cs;
 	struct mbuf mb;
 	struct bpf_d *d;
-	u_int pktlen, slen;
-	int gottime;
+	u_int clockflags, pktlen, slen;
+	int tstype, whichclock;
 
 	/* Skip outgoing duplicate packets. */
 	if ((m->m_flags & M_PROMISC) != 0 && m->m_pkthdr.rcvif == NULL) {
@@ -1926,6 +1958,20 @@ bpf_mtap2(struct bpf_if *bp, void *data, u_int dlen, struct mbuf *m)
 		return;
 	}
 
+	tstype = bp->tstype;
+	if (tstype == BPF_TSTAMP_DEFAULT)
+		tstype = bpf_default_tstype;
+
+	if (tstype == BPF_TSTAMP_NORMAL || tstype == BPF_TSTAMP_FAST)
+		sysclock_getsnapshot(&cs, tstype == BPF_TSTAMP_FAST ?
+		    1 : 0);
+#ifdef notyet
+	else if (tstype == BPF_TSTAMP_EXTERNAL)
+		/* XXX: Convert extern tstamp to bintime. */
+#endif
+	else
+		bzero(&bt, sizeof(bt));
+
 	pktlen = m_length(m, NULL);
 	/*
 	 * Craft on-stack mbuf suitable for passing to bpf_filter.
@@ -1937,7 +1983,6 @@ bpf_mtap2(struct bpf_if *bp, void *data, u_int dlen, struct mbuf *m)
 	mb.m_len = dlen;
 	pktlen += dlen;
 
-	gottime = BPF_TSTAMP_NONE;
 	BPFIF_LOCK(bp);
 	LIST_FOREACH(d, &bp->bif_dlist, bd_next) {
 		if (BPF_CHECK_DIRECTION(d, m->m_pkthdr.rcvif, bp->bif_ifp))
@@ -1947,8 +1992,16 @@ bpf_mtap2(struct bpf_if *bp, void *data, u_int dlen, struct mbuf *m)
 		slen = bpf_filter(d->bd_rfilter, (u_char *)&mb, pktlen, 0);
 		if (slen != 0) {
 			d->bd_fcount++;
-			if (gottime < bpf_ts_quality(d->bd_tstamp))
-				gottime = bpf_gettime(&bt, d->bd_tstamp, m);
+			if (tstype == BPF_TSTAMP_NORMAL ||
+			    tstype == BPF_TSTAMP_FAST) {
+				whichclock = -1;
+				SET_CLOCKCFG_FLAGS(d->bd_tstamp,
+				    cs.sysclock_active, whichclock, clockflags);
+				KASSERT(whichclock >= 0, ("Bogus BPF tstamp "
+				    "configuration: 0x%04x", d->bd_tstamp));
+				sysclock_snap2bintime(&cs, &bt, whichclock,
+				    clockflags);
+			}
 #ifdef MAC
 			if (mac_bpfdesc_check_receive(d, bp->bif_ifp) == 0)
 #endif
@@ -1962,11 +2015,6 @@ bpf_mtap2(struct bpf_if *bp, void *data, u_int dlen, struct mbuf *m)
 
 #undef	BPF_CHECK_DIRECTION
 
-#undef	BPF_TSTAMP_NONE
-#undef	BPF_TSTAMP_FAST
-#undef	BPF_TSTAMP_NORMAL
-#undef	BPF_TSTAMP_EXTERN
-
 static int
 bpf_hdrlen(struct bpf_d *d)
 {
@@ -1998,15 +2046,9 @@ bpf_hdrlen(struct bpf_d *d)
 static void
 bpf_bintime2ts(struct bintime *bt, struct bpf_ts *ts, int tstype)
 {
-	struct bintime bt2;
 	struct timeval tsm;
 	struct timespec tsn;
 
-	if ((tstype & BPF_T_MONOTONIC) == 0) {
-		bt2 = *bt;
-		bintime_add(&bt2, &boottimebin);
-		bt = &bt2;
-	}
 	switch (BPF_T_FORMAT(tstype)) {
 	case BPF_T_MICROTIME:
 		bintime2timeval(bt, &tsm);
@@ -2200,6 +2242,64 @@ bpf_freed(struct bpf_d *d)
 }
 
 /*
+ * Show or change the per bpf_if or system wide default timestamp configuration.
+ */
+static int
+bpf_tscfg_sysctl_handler(SYSCTL_HANDLER_ARGS)
+{
+	char tstype_name[16];
+	struct bpf_if *bp;
+	int error, tstype;
+
+	bp = (struct bpf_if *)arg1;
+
+	if (req->newptr == NULL) {
+		/*
+		 * Return the name of the BPF interface's timestamp setting, or
+		 * the system wide default if bp is NULL.
+		 */
+		strlcpy(tstype_name,
+		    bpfiftstypes[bp ? bp->tstype : bpf_default_tstype],
+		    sizeof(tstype_name));
+		error = sysctl_handle_string(oidp, tstype_name,
+		    sizeof(tstype_name), req);
+	} else {
+		/*
+		 * Change the timestamp configuration for this BPF interface or
+		 * the system wide default setting.
+		 */
+		error = EINVAL;
+		for (tstype = 0; tstype < NUM_BPFIFTSTYPES; tstype++) {
+			if (strncmp((char *)req->newptr, bpfiftstypes[tstype],
+			    strlen(bpfiftstypes[tstype])) == 0) {
+				/* User specified type found in bpfiftstypes. */
+				if (strcmp(oidp->oid_name, "default") == 0) {
+					/*
+					 * Don't allow BPF_TSTAMP_DEFAULT to be
+					 * assigned to the
+					 * "net.bpf.tscfg.default" OID.
+					 */
+					if (tstype != BPF_TSTAMP_DEFAULT) {
+						bpf_default_tstype = tstype;
+						error = 0;
+					}
+				} else {
+					/*
+					 * Valid tstype for
+					 * "net.bpf.tscfg.<iface>" OID.
+					 */
+					bp->tstype = tstype;
+					error = 0;
+				}
+				break;
+			}
+		}
+	}
+
+	return (error);
+}
+
+/*
  * Attach an interface to bpf.  dlt is the link layer type; hdrlen is the
  * fixed size of the link header (variable length headers not yet supported).
  */
@@ -2225,6 +2325,17 @@ bpfattach2(struct ifnet *ifp, u_int dlt, u_int hdrlen, struct bpf_if **driverp)
 	if (bp == NULL)
 		panic("bpfattach");
 
+	bp->tscfgoid = SYSCTL_ADD_PROC(NULL,
+	    SYSCTL_STATIC_CHILDREN(_net_bpf_tscfg), OID_AUTO, ifp->if_xname,
+	    CTLTYPE_STRING | CTLFLAG_RW, bp, sizeof(bp),
+	    bpf_tscfg_sysctl_handler, "A",
+	    "Interface BPF timestamp configuration");
+	if (bp->tscfgoid == NULL) {
+		free(bp, M_BPF);
+		panic("bpfattach tscfgoid");
+	}
+
+	bp->tstype = BPF_TSTAMP_DEFAULT;
 	LIST_INIT(&bp->bif_dlist);
 	bp->bif_ifp = ifp;
 	bp->bif_dlt = dlt;
@@ -2278,6 +2389,7 @@ bpfdetach(struct ifnet *ifp)
 		BPFD_UNLOCK(d);
 	}
 
+	sysctl_remove_oid(bp->tscfgoid, 1, 0);
 	mtx_destroy(&bp->bif_mtx);
 	free(bp, M_BPF);
 }
diff --git a/sys/net/bpf.h b/sys/net/bpf.h
index d4c369d..055ca5a 100644
--- a/sys/net/bpf.h
+++ b/sys/net/bpf.h
@@ -1,12 +1,17 @@
 /*-
  * Copyright (c) 1990, 1991, 1993
- *	The Regents of the University of California.  All rights reserved.
+ *	The Regents of the University of California.
+ * Copyright (c) 2011 The University of Melbourne.
+ * All rights reserved.
  *
  * This code is derived from the Stanford/CMU enet packet filter,
  * (net/enet.c) distributed as part of 4.3BSD, and code contributed
  * to Berkeley by Steven McCanne and Van Jacobson both of Lawrence
  * Berkeley Laboratory.
  *
+ * Portions of this software were developed by Julien Ridoux at the University
+ * of Melbourne under sponsorship from the FreeBSD Foundation.
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
  * are met:
@@ -166,25 +171,17 @@ enum bpf_direction {
 #define	BPF_T_NONE		0x0003
 #define	BPF_T_FORMAT_MASK	0x0003
 #define	BPF_T_NORMAL		0x0000
-#define	BPF_T_FAST		0x0100
-#define	BPF_T_MONOTONIC		0x0200
-#define	BPF_T_MONOTONIC_FAST	(BPF_T_FAST | BPF_T_MONOTONIC)
-#define	BPF_T_FLAG_MASK		0x0300
+#define	BPF_T_MONOTONIC		0x0100
+#define	BPF_T_FLAG_MASK		0x0100
 #define	BPF_T_FORMAT(t)		((t) & BPF_T_FORMAT_MASK)
 #define	BPF_T_FLAG(t)		((t) & BPF_T_FLAG_MASK)
 #define	BPF_T_VALID(t)						\
     ((t) == BPF_T_NONE || (BPF_T_FORMAT(t) != BPF_T_NONE &&	\
     ((t) & ~(BPF_T_FORMAT_MASK | BPF_T_FLAG_MASK)) == 0))
 
-#define	BPF_T_MICROTIME_FAST		(BPF_T_MICROTIME | BPF_T_FAST)
-#define	BPF_T_NANOTIME_FAST		(BPF_T_NANOTIME | BPF_T_FAST)
-#define	BPF_T_BINTIME_FAST		(BPF_T_BINTIME | BPF_T_FAST)
 #define	BPF_T_MICROTIME_MONOTONIC	(BPF_T_MICROTIME | BPF_T_MONOTONIC)
 #define	BPF_T_NANOTIME_MONOTONIC	(BPF_T_NANOTIME | BPF_T_MONOTONIC)
 #define	BPF_T_BINTIME_MONOTONIC		(BPF_T_BINTIME | BPF_T_MONOTONIC)
-#define	BPF_T_MICROTIME_MONOTONIC_FAST	(BPF_T_MICROTIME | BPF_T_MONOTONIC_FAST)
-#define	BPF_T_NANOTIME_MONOTONIC_FAST	(BPF_T_NANOTIME | BPF_T_MONOTONIC_FAST)
-#define	BPF_T_BINTIME_MONOTONIC_FAST	(BPF_T_BINTIME | BPF_T_MONOTONIC_FAST)
 
 /*
  * Structure prepended to each packet.
@@ -1100,6 +1097,8 @@ struct bpf_if {
 	u_int bif_hdrlen;		/* length of link header */
 	struct ifnet *bif_ifp;		/* corresponding interface */
 	struct mtx	bif_mtx;	/* mutex for interface */
+	struct sysctl_oid *tscfgoid;	/* timestamp sysctl oid for interface */
+	int tstype;			/* timestamp setting for interface */
 };
 
 void	 bpf_bufheld(struct bpf_d *d);