[PATCH net-next 0/7] rxrpc: More fixes and adjustments

[PATCH net-next 0/7] rxrpc: More fixes and adjustments

[David Howells]

This set of patches contains some more fixes and adjustments:

 (1) Actually display the retransmission indication previously added to the
     tx_data trace.

 (2) Switch to Congestion Avoidance mode properly at cwnd==ssthresh rather
     than relying on detection during an overshoot and correction.

 (3) Reduce ssthresh to the peer's declared receive window.

 (4) The offset field in rxrpc_skb_priv can be dispensed with and the error
     field is no longer used.  Get rid of them.

 (5) Keep the call timeouts as ktimes rather than jiffies to make it easier
     to deal with RTT-based timeout values in future.  Rounding to jiffies
     is still necessary when the system timer is set.

 (6) Fix the call timer handling to avoid retriggering of expired timeout
     actions.

The patches can be found here also:

	http://git.kernel.org/cgit/linux/kernel/git/dhowells/linux-fs.git/log/?h=rxrpc-rewrite

Tagged thusly:

	git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git
	rxrpc-rewrite-20160930

David
---
David Howells (7):
      rxrpc: Actually display the tx_data trace retransmission note
      rxrpc: Switch to Congestion Avoidance mode at cwnd==ssthresh
      rxrpc: Reduce ssthresh to peer's receive window
      rxrpc: The offset field in struct rxrpc_skb_priv is unnecessary
      rxrpc: Remove error from struct rxrpc_skb_priv as it is unused
      rxrpc: Keep the call timeouts as ktimes rather than jiffies
      rxrpc: Fix the call timer handling


 include/trace/events/rxrpc.h |   28 +++++++-------
 net/rxrpc/ar-internal.h      |   10 ++---
 net/rxrpc/call_event.c       |   87 +++++++++++++++++++++++++-----------------
 net/rxrpc/call_object.c      |   19 +++------
 net/rxrpc/conn_event.c       |    3 +
 net/rxrpc/input.c            |   34 +++++++++-------
 net/rxrpc/local_event.c      |    3 +
 net/rxrpc/misc.c             |   15 +++++--
 net/rxrpc/recvmsg.c          |    6 +--
 net/rxrpc/rxkad.c            |    9 +++-
 net/rxrpc/sendmsg.c          |    8 ++--
 net/rxrpc/sysctl.c           |    8 ++--
 12 files changed, 125 insertions(+), 105 deletions(-)

[PATCH net-next 1/7] rxrpc: Actually display the tx_data trace retransmission note

[David Howells]

Actually display in the tx_data trace the retransmission note added in a
previous patch.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 include/trace/events/rxrpc.h |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/include/trace/events/rxrpc.h b/include/trace/events/rxrpc.h
index 8ba8d76e856a..67f03946ea4a 100644
--- a/include/trace/events/rxrpc.h
+++ b/include/trace/events/rxrpc.h
@@ -280,11 +280,12 @@ TRACE_EVENT(rxrpc_tx_data,
 		    __entry->lose = lose;
 			   ),
 
-	    TP_printk("c=%p DATA %08x q=%08x fl=%02x%s",
+	    TP_printk("c=%p DATA %08x q=%08x fl=%02x%s%s",
 		      __entry->call,
 		      __entry->serial,
 		      __entry->seq,
 		      __entry->flags,
+		      __entry->retrans ? " *RETRANS*" : "",
 		      __entry->lose ? " *LOSE*" : "")
 	    );
 

[PATCH net-next 2/7] rxrpc: Switch to Congestion Avoidance mode at cwnd==ssthresh

[David Howells]

Switch to Congestion Avoidance mode at cwnd == ssthresh rather than relying
on cwnd getting incremented beyond ssthresh and the window size, the mode
being shifted and then cwnd being corrected.

We need to make sure we switch into CA mode so that we stop marking every
packet for ACK.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 net/rxrpc/input.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c
index 1461d30583c9..21746f0f7ae0 100644
--- a/net/rxrpc/input.c
+++ b/net/rxrpc/input.c
@@ -57,7 +57,7 @@ static void rxrpc_congestion_management(struct rxrpc_call *call,
 		call->cong_ssthresh = max_t(unsigned int,
 					    summary->flight_size / 2, 2);
 		cwnd = 1;
-		if (cwnd > call->cong_ssthresh &&
+		if (cwnd >= call->cong_ssthresh &&
 		    call->cong_mode == RXRPC_CALL_SLOW_START) {
 			call->cong_mode = RXRPC_CALL_CONGEST_AVOIDANCE;
 			call->cong_tstamp = skb->tstamp;
@@ -82,7 +82,7 @@ static void rxrpc_congestion_management(struct rxrpc_call *call,
 			goto packet_loss_detected;
 		if (summary->cumulative_acks > 0)
 			cwnd += 1;
-		if (cwnd > call->cong_ssthresh) {
+		if (cwnd >= call->cong_ssthresh) {
 			call->cong_mode = RXRPC_CALL_CONGEST_AVOIDANCE;
 			call->cong_tstamp = skb->tstamp;
 		}
@@ -161,7 +161,7 @@ resume_normality:
 	call->cong_dup_acks = 0;
 	call->cong_extra = 0;
 	call->cong_tstamp = skb->tstamp;
-	if (cwnd <= call->cong_ssthresh)
+	if (cwnd < call->cong_ssthresh)
 		call->cong_mode = RXRPC_CALL_SLOW_START;
 	else
 		call->cong_mode = RXRPC_CALL_CONGEST_AVOIDANCE;

[PATCH net-next 3/7] rxrpc: Reduce ssthresh to peer's receive window

[David Howells]

When we receive an ACK from the peer that tells us what the peer's receive
window (rwind) is, we should reduce ssthresh to rwind if rwind is smaller
than ssthresh.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 net/rxrpc/input.c |    2 ++
 1 file changed, 2 insertions(+)

diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c
index 21746f0f7ae0..7993473e56bb 100644
--- a/net/rxrpc/input.c
+++ b/net/rxrpc/input.c
@@ -658,6 +658,8 @@ static void rxrpc_input_ackinfo(struct rxrpc_call *call, struct sk_buff *skb,
 	if (rwind > RXRPC_RXTX_BUFF_SIZE - 1)
 		rwind = RXRPC_RXTX_BUFF_SIZE - 1;
 	call->tx_winsize = rwind;
+	if (call->cong_ssthresh > rwind)
+		call->cong_ssthresh = rwind;
 
 	mtu = min(ntohl(ackinfo->rxMTU), ntohl(ackinfo->maxMTU));
 

[PATCH net-next 4/7] rxrpc: The offset field in struct rxrpc_skb_priv is unnecessary

[David Howells]

The offset field in struct rxrpc_skb_priv is unnecessary as the value can
always be calculated.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 net/rxrpc/ar-internal.h |    1 -
 net/rxrpc/conn_event.c  |    3 ++-
 net/rxrpc/input.c       |   23 ++++++++++++-----------
 net/rxrpc/local_event.c |    3 ++-
 net/rxrpc/recvmsg.c     |    6 ++----
 net/rxrpc/rxkad.c       |    9 ++++++---
 6 files changed, 24 insertions(+), 21 deletions(-)

diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h
index 539db54697f9..fd64a2bd1072 100644
--- a/net/rxrpc/ar-internal.h
+++ b/net/rxrpc/ar-internal.h
@@ -144,7 +144,6 @@ struct rxrpc_skb_priv {
 		u8		nr_jumbo;	/* Number of jumbo subpackets */
 	};
 	union {
-		unsigned int	offset;		/* offset into buffer of next read */
 		int		remain;		/* amount of space remaining for next write */
 		u32		error;		/* network error code */
 	};
diff --git a/net/rxrpc/conn_event.c b/net/rxrpc/conn_event.c
index 37609ce89f52..3f9d8d7ec632 100644
--- a/net/rxrpc/conn_event.c
+++ b/net/rxrpc/conn_event.c
@@ -276,7 +276,8 @@ static int rxrpc_process_event(struct rxrpc_connection *conn,
 		return 0;
 
 	case RXRPC_PACKET_TYPE_ABORT:
-		if (skb_copy_bits(skb, sp->offset, &wtmp, sizeof(wtmp)) < 0)
+		if (skb_copy_bits(skb, sizeof(struct rxrpc_wire_header),
+				  &wtmp, sizeof(wtmp)) < 0)
 			return -EPROTO;
 		abort_code = ntohl(wtmp);
 		_proto("Rx ABORT %%%u { ac=%d }", sp->hdr.serial, abort_code);
diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c
index 7993473e56bb..5ba35b4a907b 100644
--- a/net/rxrpc/input.c
+++ b/net/rxrpc/input.c
@@ -358,7 +358,7 @@ static bool rxrpc_receiving_reply(struct rxrpc_call *call)
 static bool rxrpc_validate_jumbo(struct sk_buff *skb)
 {
 	struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
-	unsigned int offset = sp->offset;
+	unsigned int offset = sizeof(struct rxrpc_wire_header);
 	unsigned int len = skb->len;
 	int nr_jumbo = 1;
 	u8 flags = sp->hdr.flags;
@@ -419,7 +419,7 @@ static void rxrpc_input_data(struct rxrpc_call *call, struct sk_buff *skb,
 			     u16 skew)
 {
 	struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
-	unsigned int offset = sp->offset;
+	unsigned int offset = sizeof(struct rxrpc_wire_header);
 	unsigned int ix;
 	rxrpc_serial_t serial = sp->hdr.serial, ack_serial = 0;
 	rxrpc_seq_t seq = sp->hdr.seq, hard_ack;
@@ -746,15 +746,16 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb,
 	} buf;
 	rxrpc_serial_t acked_serial;
 	rxrpc_seq_t first_soft_ack, hard_ack;
-	int nr_acks, offset;
+	int nr_acks, offset, ioffset;
 
 	_enter("");
 
-	if (skb_copy_bits(skb, sp->offset, &buf.ack, sizeof(buf.ack)) < 0) {
+	offset = sizeof(struct rxrpc_wire_header);
+	if (skb_copy_bits(skb, offset, &buf.ack, sizeof(buf.ack)) < 0) {
 		_debug("extraction failure");
 		return rxrpc_proto_abort("XAK", call, 0);
 	}
-	sp->offset += sizeof(buf.ack);
+	offset += sizeof(buf.ack);
 
 	acked_serial = ntohl(buf.ack.serial);
 	first_soft_ack = ntohl(buf.ack.firstPacket);
@@ -792,9 +793,9 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb,
 				  rxrpc_propose_ack_respond_to_ack);
 	}
 
-	offset = sp->offset + nr_acks + 3;
-	if (skb->len >= offset + sizeof(buf.info)) {
-		if (skb_copy_bits(skb, offset, &buf.info, sizeof(buf.info)) < 0)
+	ioffset = offset + nr_acks + 3;
+	if (skb->len >= ioffset + sizeof(buf.info)) {
+		if (skb_copy_bits(skb, ioffset, &buf.info, sizeof(buf.info)) < 0)
 			return rxrpc_proto_abort("XAI", call, 0);
 		rxrpc_input_ackinfo(call, skb, &buf.info);
 	}
@@ -832,7 +833,7 @@ static void rxrpc_input_ack(struct rxrpc_call *call, struct sk_buff *skb,
 		rxrpc_rotate_tx_window(call, hard_ack, &summary);
 
 	if (nr_acks > 0) {
-		if (skb_copy_bits(skb, sp->offset, buf.acks, nr_acks) < 0)
+		if (skb_copy_bits(skb, offset, buf.acks, nr_acks) < 0)
 			return rxrpc_proto_abort("XSA", call, 0);
 		rxrpc_input_soft_acks(call, buf.acks, first_soft_ack, nr_acks,
 				      &summary);
@@ -880,7 +881,8 @@ static void rxrpc_input_abort(struct rxrpc_call *call, struct sk_buff *skb)
 	_enter("");
 
 	if (skb->len >= 4 &&
-	    skb_copy_bits(skb, sp->offset, &wtmp, sizeof(wtmp)) >= 0)
+	    skb_copy_bits(skb, sizeof(struct rxrpc_wire_header),
+			  &wtmp, sizeof(wtmp)) >= 0)
 		abort_code = ntohl(wtmp);
 
 	_proto("Rx ABORT %%%u { %x }", sp->hdr.serial, abort_code);
@@ -996,7 +998,6 @@ int rxrpc_extract_header(struct rxrpc_skb_priv *sp, struct sk_buff *skb)
 	sp->hdr.securityIndex	= whdr.securityIndex;
 	sp->hdr._rsvd		= ntohs(whdr._rsvd);
 	sp->hdr.serviceId	= ntohs(whdr.serviceId);
-	sp->offset = sizeof(whdr);
 	return 0;
 }
 
diff --git a/net/rxrpc/local_event.c b/net/rxrpc/local_event.c
index 190f68bd9e27..540d3955c1bc 100644
--- a/net/rxrpc/local_event.c
+++ b/net/rxrpc/local_event.c
@@ -95,7 +95,8 @@ void rxrpc_process_local_events(struct rxrpc_local *local)
 
 		switch (sp->hdr.type) {
 		case RXRPC_PACKET_TYPE_VERSION:
-			if (skb_copy_bits(skb, sp->offset, &v, 1) < 0)
+			if (skb_copy_bits(skb, sizeof(struct rxrpc_wire_header),
+					  &v, 1) < 0)
 				return;
 			_proto("Rx VERSION { %02x }", v);
 			if (v == 0)
diff --git a/net/rxrpc/recvmsg.c b/net/rxrpc/recvmsg.c
index 038ae62ddb4d..f05ea0a88076 100644
--- a/net/rxrpc/recvmsg.c
+++ b/net/rxrpc/recvmsg.c
@@ -261,15 +261,13 @@ static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
 			     u8 *_annotation,
 			     unsigned int *_offset, unsigned int *_len)
 {
-	struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
-	unsigned int offset = *_offset;
+	unsigned int offset = sizeof(struct rxrpc_wire_header);
 	unsigned int len = *_len;
 	int ret;
 	u8 annotation = *_annotation;
 
 	/* Locate the subpacket */
-	offset = sp->offset;
-	len = skb->len - sp->offset;
+	len = skb->len - offset;
 	if ((annotation & RXRPC_RX_ANNO_JUMBO) > 0) {
 		offset += (((annotation & RXRPC_RX_ANNO_JUMBO) - 1) *
 			   RXRPC_JUMBO_SUBPKTLEN);
diff --git a/net/rxrpc/rxkad.c b/net/rxrpc/rxkad.c
index 88d080a1a3de..627abed5f999 100644
--- a/net/rxrpc/rxkad.c
+++ b/net/rxrpc/rxkad.c
@@ -771,7 +771,8 @@ static int rxkad_respond_to_challenge(struct rxrpc_connection *conn,
 	}
 
 	abort_code = RXKADPACKETSHORT;
-	if (skb_copy_bits(skb, sp->offset, &challenge, sizeof(challenge)) < 0)
+	if (skb_copy_bits(skb, sizeof(struct rxrpc_wire_header),
+			  &challenge, sizeof(challenge)) < 0)
 		goto protocol_error;
 
 	version = ntohl(challenge.version);
@@ -1028,7 +1029,8 @@ static int rxkad_verify_response(struct rxrpc_connection *conn,
 	_enter("{%d,%x}", conn->debug_id, key_serial(conn->server_key));
 
 	abort_code = RXKADPACKETSHORT;
-	if (skb_copy_bits(skb, sp->offset, &response, sizeof(response)) < 0)
+	if (skb_copy_bits(skb, sizeof(struct rxrpc_wire_header),
+			  &response, sizeof(response)) < 0)
 		goto protocol_error;
 	if (!pskb_pull(skb, sizeof(response)))
 		BUG();
@@ -1057,7 +1059,8 @@ static int rxkad_verify_response(struct rxrpc_connection *conn,
 		return -ENOMEM;
 
 	abort_code = RXKADPACKETSHORT;
-	if (skb_copy_bits(skb, sp->offset, ticket, ticket_len) < 0)
+	if (skb_copy_bits(skb, sizeof(struct rxrpc_wire_header),
+			  ticket, ticket_len) < 0)
 		goto protocol_error_free;
 
 	ret = rxkad_decrypt_ticket(conn, ticket, ticket_len, &session_key,

[PATCH net-next 5/7] rxrpc: Remove error from struct rxrpc_skb_priv as it is unused

[David Howells]

Remove error from struct rxrpc_skb_priv as it is no longer used.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 net/rxrpc/ar-internal.h |    1 -
 1 file changed, 1 deletion(-)

diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h
index fd64a2bd1072..141c1458e719 100644
--- a/net/rxrpc/ar-internal.h
+++ b/net/rxrpc/ar-internal.h
@@ -145,7 +145,6 @@ struct rxrpc_skb_priv {
 	};
 	union {
 		int		remain;		/* amount of space remaining for next write */
-		u32		error;		/* network error code */
 	};
 
 	struct rxrpc_host_header hdr;		/* RxRPC packet header from this packet */

[PATCH net-next 6/7] rxrpc: Keep the call timeouts as ktimes rather than jiffies

[David Howells]

Keep that call timeouts as ktimes rather than jiffies so that they can be
expressed as functions of RTT.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 include/trace/events/rxrpc.h |   25 +++++++-------
 net/rxrpc/ar-internal.h      |    8 ++---
 net/rxrpc/call_event.c       |   73 ++++++++++++++++++++++--------------------
 net/rxrpc/call_object.c      |   16 +++------
 net/rxrpc/input.c            |    3 +-
 net/rxrpc/misc.c             |   15 ++++++---
 net/rxrpc/sendmsg.c          |    8 ++---
 net/rxrpc/sysctl.c           |    8 ++---
 8 files changed, 82 insertions(+), 74 deletions(-)

diff --git a/include/trace/events/rxrpc.h b/include/trace/events/rxrpc.h
index 67f03946ea4a..0383e5e9a0f3 100644
--- a/include/trace/events/rxrpc.h
+++ b/include/trace/events/rxrpc.h
@@ -453,17 +453,18 @@ TRACE_EVENT(rxrpc_rtt_rx,
 
 TRACE_EVENT(rxrpc_timer,
 	    TP_PROTO(struct rxrpc_call *call, enum rxrpc_timer_trace why,
-		     unsigned long now),
+		     ktime_t now, unsigned long now_j),
 
-	    TP_ARGS(call, why, now),
+	    TP_ARGS(call, why, now, now_j),
 
 	    TP_STRUCT__entry(
 		    __field(struct rxrpc_call *,		call		)
 		    __field(enum rxrpc_timer_trace,		why		)
-		    __field(unsigned long,			now		)
-		    __field(unsigned long,			expire_at	)
-		    __field(unsigned long,			ack_at		)
-		    __field(unsigned long,			resend_at	)
+		    __field_struct(ktime_t,			now		)
+		    __field_struct(ktime_t,			expire_at	)
+		    __field_struct(ktime_t,			ack_at		)
+		    __field_struct(ktime_t,			resend_at	)
+		    __field(unsigned long,			now_j		)
 		    __field(unsigned long,			timer		)
 			     ),
 
@@ -474,17 +475,17 @@ TRACE_EVENT(rxrpc_timer,
 		    __entry->expire_at	= call->expire_at;
 		    __entry->ack_at	= call->ack_at;
 		    __entry->resend_at	= call->resend_at;
+		    __entry->now_j	= now_j;
 		    __entry->timer	= call->timer.expires;
 			   ),
 
-	    TP_printk("c=%p %s now=%lx x=%ld a=%ld r=%ld t=%ld",
+	    TP_printk("c=%p %s x=%lld a=%lld r=%lld t=%ld",
 		      __entry->call,
 		      rxrpc_timer_traces[__entry->why],
-		      __entry->now,
-		      __entry->expire_at - __entry->now,
-		      __entry->ack_at - __entry->now,
-		      __entry->resend_at - __entry->now,
-		      __entry->timer - __entry->now)
+		      ktime_to_ns(ktime_sub(__entry->expire_at, __entry->now)),
+		      ktime_to_ns(ktime_sub(__entry->ack_at, __entry->now)),
+		      ktime_to_ns(ktime_sub(__entry->resend_at, __entry->now)),
+		      __entry->timer - __entry->now_j)
 	    );
 
 TRACE_EVENT(rxrpc_rx_lose,
diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h
index 141c1458e719..d38dffd78085 100644
--- a/net/rxrpc/ar-internal.h
+++ b/net/rxrpc/ar-internal.h
@@ -464,9 +464,9 @@ struct rxrpc_call {
 	struct rxrpc_connection	*conn;		/* connection carrying call */
 	struct rxrpc_peer	*peer;		/* Peer record for remote address */
 	struct rxrpc_sock __rcu	*socket;	/* socket responsible */
-	unsigned long		ack_at;		/* When deferred ACK needs to happen */
-	unsigned long		resend_at;	/* When next resend needs to happen */
-	unsigned long		expire_at;	/* When the call times out */
+	ktime_t			ack_at;		/* When deferred ACK needs to happen */
+	ktime_t			resend_at;	/* When next resend needs to happen */
+	ktime_t			expire_at;	/* When the call times out */
 	struct timer_list	timer;		/* Combined event timer */
 	struct work_struct	processor;	/* Event processor */
 	rxrpc_notify_rx_t	notify_rx;	/* kernel service Rx notification function */
@@ -805,7 +805,7 @@ int rxrpc_reject_call(struct rxrpc_sock *);
 /*
  * call_event.c
  */
-void rxrpc_set_timer(struct rxrpc_call *, enum rxrpc_timer_trace);
+void rxrpc_set_timer(struct rxrpc_call *, enum rxrpc_timer_trace, ktime_t);
 void rxrpc_propose_ACK(struct rxrpc_call *, u8, u16, u32, bool, bool,
 		       enum rxrpc_propose_ack_trace);
 void rxrpc_process_call(struct work_struct *);
diff --git a/net/rxrpc/call_event.c b/net/rxrpc/call_event.c
index 1f6c7633b964..9ff3bb3ffb41 100644
--- a/net/rxrpc/call_event.c
+++ b/net/rxrpc/call_event.c
@@ -24,28 +24,40 @@
 /*
  * Set the timer
  */
-void rxrpc_set_timer(struct rxrpc_call *call, enum rxrpc_timer_trace why)
+void rxrpc_set_timer(struct rxrpc_call *call, enum rxrpc_timer_trace why,
+		     ktime_t now)
 {
-	unsigned long t, now = jiffies;
+	unsigned long t_j, now_j = jiffies;
+	ktime_t t;
 
 	read_lock_bh(&call->state_lock);
 
 	if (call->state < RXRPC_CALL_COMPLETE) {
 		t = call->expire_at;
-		if (time_before_eq(t, now))
+		if (!ktime_after(t, now))
 			goto out;
 
-		if (time_after(call->resend_at, now) &&
-		    time_before(call->resend_at, t))
+		if (ktime_after(call->resend_at, now) &&
+		    ktime_before(call->resend_at, t))
 			t = call->resend_at;
 
-		if (time_after(call->ack_at, now) &&
-		    time_before(call->ack_at, t))
+		if (ktime_after(call->ack_at, now) &&
+		    ktime_before(call->ack_at, t))
 			t = call->ack_at;
 
-		if (call->timer.expires != t || !timer_pending(&call->timer)) {
-			mod_timer(&call->timer, t);
-			trace_rxrpc_timer(call, why, now);
+		t_j = nsecs_to_jiffies(ktime_to_ns(ktime_sub(t, now)));
+		t_j += jiffies;
+
+		/* We have to make sure that the calculated jiffies value falls
+		 * at or after the nsec value, or we may loop ceaselessly
+		 * because the timer times out, but we haven't reached the nsec
+		 * timeout yet.
+		 */
+		t_j++;
+
+		if (call->timer.expires != t_j || !timer_pending(&call->timer)) {
+			mod_timer(&call->timer, t_j);
+			trace_rxrpc_timer(call, why, now, now_j);
 		}
 	}
 
@@ -62,7 +74,8 @@ static void __rxrpc_propose_ACK(struct rxrpc_call *call, u8 ack_reason,
 				enum rxrpc_propose_ack_trace why)
 {
 	enum rxrpc_propose_ack_outcome outcome = rxrpc_propose_ack_use;
-	unsigned long now, ack_at, expiry = rxrpc_soft_ack_delay;
+	unsigned int expiry = rxrpc_soft_ack_delay;
+	ktime_t now, ack_at;
 	s8 prior = rxrpc_ack_priority[ack_reason];
 
 	/* Update DELAY, IDLE, REQUESTED and PING_RESPONSE ACK serial
@@ -111,7 +124,6 @@ static void __rxrpc_propose_ACK(struct rxrpc_call *call, u8 ack_reason,
 		break;
 	}
 
-	now = jiffies;
 	if (test_bit(RXRPC_CALL_EV_ACK, &call->events)) {
 		_debug("already scheduled");
 	} else if (immediate || expiry == 0) {
@@ -120,11 +132,11 @@ static void __rxrpc_propose_ACK(struct rxrpc_call *call, u8 ack_reason,
 		    background)
 			rxrpc_queue_call(call);
 	} else {
-		ack_at = now + expiry;
-		_debug("deferred ACK %ld < %ld", expiry, call->ack_at - now);
-		if (time_before(ack_at, call->ack_at)) {
+		now = ktime_get_real();
+		ack_at = ktime_add_ms(now, expiry);
+		if (ktime_before(ack_at, call->ack_at)) {
 			call->ack_at = ack_at;
-			rxrpc_set_timer(call, rxrpc_timer_set_for_ack);
+			rxrpc_set_timer(call, rxrpc_timer_set_for_ack, now);
 		}
 	}
 
@@ -157,12 +169,12 @@ static void rxrpc_congestion_timeout(struct rxrpc_call *call)
 /*
  * Perform retransmission of NAK'd and unack'd packets.
  */
-static void rxrpc_resend(struct rxrpc_call *call)
+static void rxrpc_resend(struct rxrpc_call *call, ktime_t now)
 {
 	struct rxrpc_skb_priv *sp;
 	struct sk_buff *skb;
 	rxrpc_seq_t cursor, seq, top;
-	ktime_t now = ktime_get_real(), max_age, oldest, resend_at, ack_ts;
+	ktime_t max_age, oldest, ack_ts;
 	int ix;
 	u8 annotation, anno_type, retrans = 0, unacked = 0;
 
@@ -212,14 +224,7 @@ static void rxrpc_resend(struct rxrpc_call *call)
 				       ktime_to_ns(ktime_sub(skb->tstamp, max_age)));
 	}
 
-	resend_at = ktime_add_ms(oldest, rxrpc_resend_timeout);
-	call->resend_at = jiffies +
-		nsecs_to_jiffies(ktime_to_ns(ktime_sub(resend_at, now))) +
-		1; /* We have to make sure that the calculated jiffies value
-		    * falls at or after the nsec value, or we shall loop
-		    * ceaselessly because the timer times out, but we haven't
-		    * reached the nsec timeout yet.
-		    */
+	call->resend_at = ktime_add_ms(oldest, rxrpc_resend_timeout);
 
 	if (unacked)
 		rxrpc_congestion_timeout(call);
@@ -229,7 +234,7 @@ static void rxrpc_resend(struct rxrpc_call *call)
 	 * retransmitting data.
 	 */
 	if (!retrans) {
-		rxrpc_set_timer(call, rxrpc_timer_set_for_resend);
+		rxrpc_set_timer(call, rxrpc_timer_set_for_resend, now);
 		spin_unlock_bh(&call->lock);
 		ack_ts = ktime_sub(now, call->acks_latest_ts);
 		if (ktime_to_ns(ack_ts) < call->peer->rtt)
@@ -301,7 +306,7 @@ void rxrpc_process_call(struct work_struct *work)
 {
 	struct rxrpc_call *call =
 		container_of(work, struct rxrpc_call, processor);
-	unsigned long now;
+	ktime_t now;
 
 	rxrpc_see_call(call);
 
@@ -320,15 +325,15 @@ recheck_state:
 		goto out_put;
 	}
 
-	now = jiffies;
-	if (time_after_eq(now, call->expire_at)) {
+	now = ktime_get_real();
+	if (ktime_before(call->expire_at, now)) {
 		rxrpc_abort_call("EXP", call, 0, RX_CALL_TIMEOUT, ETIME);
 		set_bit(RXRPC_CALL_EV_ABORT, &call->events);
 		goto recheck_state;
 	}
 
 	if (test_and_clear_bit(RXRPC_CALL_EV_ACK, &call->events) ||
-	    time_after_eq(now, call->ack_at)) {
+	    ktime_before(call->ack_at, now)) {
 		call->ack_at = call->expire_at;
 		if (call->ackr_reason) {
 			rxrpc_send_call_packet(call, RXRPC_PACKET_TYPE_ACK);
@@ -337,12 +342,12 @@ recheck_state:
 	}
 
 	if (test_and_clear_bit(RXRPC_CALL_EV_RESEND, &call->events) ||
-	    time_after_eq(now, call->resend_at)) {
-		rxrpc_resend(call);
+	    ktime_before(call->resend_at, now)) {
+		rxrpc_resend(call, now);
 		goto recheck_state;
 	}
 
-	rxrpc_set_timer(call, rxrpc_timer_set_for_resend);
+	rxrpc_set_timer(call, rxrpc_timer_set_for_resend, now);
 
 	/* other events may have been raised since we started checking */
 	if (call->events && call->state < RXRPC_CALL_COMPLETE) {
diff --git a/net/rxrpc/call_object.c b/net/rxrpc/call_object.c
index d4b3293b78fa..456ab752d473 100644
--- a/net/rxrpc/call_object.c
+++ b/net/rxrpc/call_object.c
@@ -19,11 +19,6 @@
 #include <net/af_rxrpc.h>
 #include "ar-internal.h"
 
-/*
- * Maximum lifetime of a call (in jiffies).
- */
-unsigned int rxrpc_max_call_lifetime = 60 * HZ;
-
 const char *const rxrpc_call_states[NR__RXRPC_CALL_STATES] = {
 	[RXRPC_CALL_UNINITIALISED]		= "Uninit  ",
 	[RXRPC_CALL_CLIENT_AWAIT_CONN]		= "ClWtConn",
@@ -77,7 +72,8 @@ static void rxrpc_call_timer_expired(unsigned long _call)
 	_enter("%d", call->debug_id);
 
 	if (call->state < RXRPC_CALL_COMPLETE) {
-		trace_rxrpc_timer(call, rxrpc_timer_expired, jiffies);
+		trace_rxrpc_timer(call, rxrpc_timer_expired,
+				  ktime_get_real(), jiffies);
 		rxrpc_queue_call(call);
 	}
 }
@@ -207,14 +203,14 @@ static struct rxrpc_call *rxrpc_alloc_client_call(struct sockaddr_rxrpc *srx,
  */
 static void rxrpc_start_call_timer(struct rxrpc_call *call)
 {
-	unsigned long expire_at;
+	ktime_t now = ktime_get_real(), expire_at;
 
-	expire_at = jiffies + rxrpc_max_call_lifetime;
+	expire_at = ktime_add_ms(now, rxrpc_max_call_lifetime);
 	call->expire_at = expire_at;
 	call->ack_at = expire_at;
 	call->resend_at = expire_at;
-	call->timer.expires = expire_at + 1;
-	rxrpc_set_timer(call, rxrpc_timer_begin);
+	call->timer.expires = jiffies + LONG_MAX / 2;
+	rxrpc_set_timer(call, rxrpc_timer_begin, now);
 }
 
 /*
diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c
index 5ba35b4a907b..3ad9f75031e3 100644
--- a/net/rxrpc/input.c
+++ b/net/rxrpc/input.c
@@ -328,7 +328,8 @@ static bool rxrpc_receiving_reply(struct rxrpc_call *call)
 		call->resend_at = call->expire_at;
 		call->ack_at = call->expire_at;
 		spin_unlock_bh(&call->lock);
-		rxrpc_set_timer(call, rxrpc_timer_init_for_reply);
+		rxrpc_set_timer(call, rxrpc_timer_init_for_reply,
+				ktime_get_real());
 	}
 
 	if (!test_bit(RXRPC_CALL_TX_LAST, &call->flags))
diff --git a/net/rxrpc/misc.c b/net/rxrpc/misc.c
index 47dddacdbb91..9d1c721bc4e8 100644
--- a/net/rxrpc/misc.c
+++ b/net/rxrpc/misc.c
@@ -21,28 +21,33 @@
 unsigned int rxrpc_max_backlog __read_mostly = 10;
 
 /*
+ * Maximum lifetime of a call (in mx).
+ */
+unsigned int rxrpc_max_call_lifetime = 60 * 1000;
+
+/*
  * How long to wait before scheduling ACK generation after seeing a
- * packet with RXRPC_REQUEST_ACK set (in jiffies).
+ * packet with RXRPC_REQUEST_ACK set (in ms).
  */
 unsigned int rxrpc_requested_ack_delay = 1;
 
 /*
- * How long to wait before scheduling an ACK with subtype DELAY (in jiffies).
+ * How long to wait before scheduling an ACK with subtype DELAY (in ms).
  *
  * We use this when we've received new data packets.  If those packets aren't
  * all consumed within this time we will send a DELAY ACK if an ACK was not
  * requested to let the sender know it doesn't need to resend.
  */
-unsigned int rxrpc_soft_ack_delay = 1 * HZ;
+unsigned int rxrpc_soft_ack_delay = 1 * 1000;
 
 /*
- * How long to wait before scheduling an ACK with subtype IDLE (in jiffies).
+ * How long to wait before scheduling an ACK with subtype IDLE (in ms).
  *
  * We use this when we've consumed some previously soft-ACK'd packets when
  * further packets aren't immediately received to decide when to send an IDLE
  * ACK let the other end know that it can free up its Tx buffer space.
  */
-unsigned int rxrpc_idle_ack_delay = 0.5 * HZ;
+unsigned int rxrpc_idle_ack_delay = 0.5 * 1000;
 
 /*
  * Receive window size in packets.  This indicates the maximum number of
diff --git a/net/rxrpc/sendmsg.c b/net/rxrpc/sendmsg.c
index d8dfdce874d8..3322543d460a 100644
--- a/net/rxrpc/sendmsg.c
+++ b/net/rxrpc/sendmsg.c
@@ -149,13 +149,13 @@ static void rxrpc_queue_packet(struct rxrpc_call *call, struct sk_buff *skb,
 		_debug("need instant resend %d", ret);
 		rxrpc_instant_resend(call, ix);
 	} else {
-		unsigned long resend_at;
+		ktime_t now = ktime_get_real(), resend_at;
 
-		resend_at = jiffies + msecs_to_jiffies(rxrpc_resend_timeout);
+		resend_at = ktime_add_ms(now, rxrpc_resend_timeout);
 
-		if (time_before(resend_at, call->resend_at)) {
+		if (ktime_before(resend_at, call->resend_at)) {
 			call->resend_at = resend_at;
-			rxrpc_set_timer(call, rxrpc_timer_set_for_send);
+			rxrpc_set_timer(call, rxrpc_timer_set_for_send, now);
 		}
 	}
 
diff --git a/net/rxrpc/sysctl.c b/net/rxrpc/sysctl.c
index 13d1df03ebac..34c706d2f79c 100644
--- a/net/rxrpc/sysctl.c
+++ b/net/rxrpc/sysctl.c
@@ -35,7 +35,7 @@ static struct ctl_table rxrpc_sysctl_table[] = {
 		.data		= &rxrpc_requested_ack_delay,
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec_ms_jiffies,
+		.proc_handler	= proc_dointvec,
 		.extra1		= (void *)&zero,
 	},
 	{
@@ -43,7 +43,7 @@ static struct ctl_table rxrpc_sysctl_table[] = {
 		.data		= &rxrpc_soft_ack_delay,
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec_ms_jiffies,
+		.proc_handler	= proc_dointvec,
 		.extra1		= (void *)&one,
 	},
 	{
@@ -51,7 +51,7 @@ static struct ctl_table rxrpc_sysctl_table[] = {
 		.data		= &rxrpc_idle_ack_delay,
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec_ms_jiffies,
+		.proc_handler	= proc_dointvec,
 		.extra1		= (void *)&one,
 	},
 	{
@@ -85,7 +85,7 @@ static struct ctl_table rxrpc_sysctl_table[] = {
 		.data		= &rxrpc_max_call_lifetime,
 		.maxlen		= sizeof(unsigned int),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec_jiffies,
+		.proc_handler	= proc_dointvec,
 		.extra1		= (void *)&one,
 	},
 

[PATCH net-next 7/7] rxrpc: Fix the call timer handling

[David Howells]

The call timer's concept of a call timeout (of which there are three) that
is inactive is that it is the timeout has the same expiration time as the
call expiration timeout (the expiration timer is never inactive).  However,
I'm not resetting the timeouts when they expire, leading to repeated
processing of expired timeouts when other timeout events occur.

Fix this by:

 (1) Move the timer expiry detection into rxrpc_set_timer() inside the
     locked section.  This means that if a timeout is set that will expire
     immediately, we deal with it immediately.

 (2) If a timeout is at or before now then it has expired.  When an expiry
     is detected, an event is raised, the timeout is automatically
     inactivated and the event processor is queued.

 (3) If a timeout is at or after the expiry timeout then it is inactive.
     Inactive timeouts do not contribute to the timer setting.

 (4) The call timer callback can now just call rxrpc_set_timer() to handle
     things.

 (5) The call processor work function now checks the event flags rather
     than checking the timeouts directly.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 net/rxrpc/call_event.c  |   26 ++++++++++++++++++--------
 net/rxrpc/call_object.c |    7 ++-----
 2 files changed, 20 insertions(+), 13 deletions(-)

diff --git a/net/rxrpc/call_event.c b/net/rxrpc/call_event.c
index 9ff3bb3ffb41..4f00476630b9 100644
--- a/net/rxrpc/call_event.c
+++ b/net/rxrpc/call_event.c
@@ -29,6 +29,7 @@ void rxrpc_set_timer(struct rxrpc_call *call, enum rxrpc_timer_trace why,
 {
 	unsigned long t_j, now_j = jiffies;
 	ktime_t t;
+	bool queue = false;
 
 	read_lock_bh(&call->state_lock);
 
@@ -37,13 +38,21 @@ void rxrpc_set_timer(struct rxrpc_call *call, enum rxrpc_timer_trace why,
 		if (!ktime_after(t, now))
 			goto out;
 
-		if (ktime_after(call->resend_at, now) &&
-		    ktime_before(call->resend_at, t))
+		if (!ktime_after(call->resend_at, now)) {
+			call->resend_at = call->expire_at;
+			if (!test_and_set_bit(RXRPC_CALL_EV_RESEND, &call->events))
+				queue = true;
+		} else if (ktime_before(call->resend_at, t)) {
 			t = call->resend_at;
+		}
 
-		if (ktime_after(call->ack_at, now) &&
-		    ktime_before(call->ack_at, t))
+		if (!ktime_after(call->ack_at, now)) {
+			call->ack_at = call->expire_at;
+			if (!test_and_set_bit(RXRPC_CALL_EV_ACK, &call->events))
+				queue = true;
+		} else if (ktime_before(call->ack_at, t)) {
 			t = call->ack_at;
+		}
 
 		t_j = nsecs_to_jiffies(ktime_to_ns(ktime_sub(t, now)));
 		t_j += jiffies;
@@ -59,6 +68,9 @@ void rxrpc_set_timer(struct rxrpc_call *call, enum rxrpc_timer_trace why,
 			mod_timer(&call->timer, t_j);
 			trace_rxrpc_timer(call, why, now, now_j);
 		}
+
+		if (queue)
+			rxrpc_queue_call(call);
 	}
 
 out:
@@ -332,8 +344,7 @@ recheck_state:
 		goto recheck_state;
 	}
 
-	if (test_and_clear_bit(RXRPC_CALL_EV_ACK, &call->events) ||
-	    ktime_before(call->ack_at, now)) {
+	if (test_and_clear_bit(RXRPC_CALL_EV_ACK, &call->events)) {
 		call->ack_at = call->expire_at;
 		if (call->ackr_reason) {
 			rxrpc_send_call_packet(call, RXRPC_PACKET_TYPE_ACK);
@@ -341,8 +352,7 @@ recheck_state:
 		}
 	}
 
-	if (test_and_clear_bit(RXRPC_CALL_EV_RESEND, &call->events) ||
-	    ktime_before(call->resend_at, now)) {
+	if (test_and_clear_bit(RXRPC_CALL_EV_RESEND, &call->events)) {
 		rxrpc_resend(call, now);
 		goto recheck_state;
 	}
diff --git a/net/rxrpc/call_object.c b/net/rxrpc/call_object.c
index 456ab752d473..364b42dc3dce 100644
--- a/net/rxrpc/call_object.c
+++ b/net/rxrpc/call_object.c
@@ -71,11 +71,8 @@ static void rxrpc_call_timer_expired(unsigned long _call)
 
 	_enter("%d", call->debug_id);
 
-	if (call->state < RXRPC_CALL_COMPLETE) {
-		trace_rxrpc_timer(call, rxrpc_timer_expired,
-				  ktime_get_real(), jiffies);
-		rxrpc_queue_call(call);
-	}
+	if (call->state < RXRPC_CALL_COMPLETE)
+		rxrpc_set_timer(call, rxrpc_timer_expired, ktime_get_real());
 }
 
 /*

Re: [PATCH net-next 0/7] rxrpc: More fixes and adjustments

[David Miller]

From: David Howells <dhowells@redhat.com>
Date: Fri, 30 Sep 2016 22:40:11 +0100

> 
> This set of patches contains some more fixes and adjustments:
> 
>  (1) Actually display the retransmission indication previously added to the
>      tx_data trace.
> 
>  (2) Switch to Congestion Avoidance mode properly at cwnd==ssthresh rather
>      than relying on detection during an overshoot and correction.
> 
>  (3) Reduce ssthresh to the peer's declared receive window.
> 
>  (4) The offset field in rxrpc_skb_priv can be dispensed with and the error
>      field is no longer used.  Get rid of them.
> 
>  (5) Keep the call timeouts as ktimes rather than jiffies to make it easier
>      to deal with RTT-based timeout values in future.  Rounding to jiffies
>      is still necessary when the system timer is set.
> 
>  (6) Fix the call timer handling to avoid retriggering of expired timeout
>      actions.
 ...
> 	git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git
> 	rxrpc-rewrite-20160930

Pulled, thanks David.

Re: [PATCH net-next 0/7] rxrpc: More fixes and adjustments

[David Howells]

Hi Dave,

What's the procedure for sending more fixes at this point?  Do I send them
against net-next, since net doesn't contain the things to be fixed yet, or do
I wait for net to catch up?

David

Re: [PATCH net-next 0/7] rxrpc: More fixes and adjustments

[David Miller]

From: David Howells <dhowells@redhat.com>
Date: Tue, 04 Oct 2016 17:01:30 +0100

> What's the procedure for sending more fixes at this point?  Do I send them
> against net-next, since net doesn't contain the things to be fixed yet, or do
> I wait for net to catch up?

I'm sending Linus a pull request later today for the stuff in net-next,
and once he takes that all the trees will be synced up and you can start
sending fixes.

Thanks.