Merge branch 'master' of git://eden-feed.erg.abdn.ac.uk/net-next-2.6

6 files changed, 90 insertions, 96 deletions

diff --git a/net/dccp/ccids/ccid3.c b/net/dccp/ccids/ccid3.c
index a1929f33d70..f6756e0c9e6 100644
--- a/net/dccp/ccids/ccid3.c
+++ b/net/dccp/ccids/ccid3.c
@@ -794,7 +794,7 @@ static void ccid3_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb)
 {
 	struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
 	enum ccid3_fback_type do_feedback = CCID3_FBACK_NONE;
-	const u32 ndp = dccp_sk(sk)->dccps_options_received.dccpor_ndp;
+	const u64 ndp = dccp_sk(sk)->dccps_options_received.dccpor_ndp;
 	const bool is_data_packet = dccp_data_packet(skb);
 
 	if (unlikely(hcrx->ccid3hcrx_state == TFRC_RSTATE_NO_DATA)) {
@@ -825,18 +825,16 @@ static void ccid3_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb)
 	}
 
 	/*
-	 * Handle pending losses and otherwise check for new loss
+	 * Perform loss detection and handle pending losses
 	 */
-	if (tfrc_rx_hist_loss_pending(&hcrx->ccid3hcrx_hist) &&
-	    tfrc_rx_handle_loss(&hcrx->ccid3hcrx_hist,
-				&hcrx->ccid3hcrx_li_hist,
-				skb, ndp, ccid3_first_li, sk) ) {
+	if (tfrc_rx_handle_loss(&hcrx->ccid3hcrx_hist, &hcrx->ccid3hcrx_li_hist,
+				skb, ndp, ccid3_first_li, sk)) {
 		do_feedback = CCID3_FBACK_PARAM_CHANGE;
 		goto done_receiving;
 	}
 
-	if (tfrc_rx_hist_new_loss_indicated(&hcrx->ccid3hcrx_hist, skb, ndp))
-		goto update_records;
+	if (tfrc_rx_hist_loss_pending(&hcrx->ccid3hcrx_hist))
+		return; /* done receiving */
 
 	/*
 	 * Handle data packets: RTT sampling and monitoring p
diff --git a/net/dccp/ccids/lib/loss_interval.c b/net/dccp/ccids/lib/loss_interval.c
index 849e181e698..bcd6ac415bb 100644
--- a/net/dccp/ccids/lib/loss_interval.c
+++ b/net/dccp/ccids/lib/loss_interval.c
@@ -90,14 +90,14 @@ u8 tfrc_lh_update_i_mean(struct tfrc_loss_hist *lh, struct sk_buff *skb)
 {
 	struct tfrc_loss_interval *cur = tfrc_lh_peek(lh);
 	u32 old_i_mean = lh->i_mean;
-	s64 length;
+	s64 len;
 
 	if (cur == NULL)			/* not initialised */
 		return 0;
 
-	length = dccp_delta_seqno(cur->li_seqno, DCCP_SKB_CB(skb)->dccpd_seq);
+	len = dccp_delta_seqno(cur->li_seqno, DCCP_SKB_CB(skb)->dccpd_seq) + 1;
 
-	if (length - cur->li_length <= 0)	/* duplicate or reordered */
+	if (len - (s64)cur->li_length <= 0)	/* duplicate or reordered */
 		return 0;
 
 	if (SUB16(dccp_hdr(skb)->dccph_ccval, cur->li_ccval) > 4)
@@ -114,7 +114,7 @@ u8 tfrc_lh_update_i_mean(struct tfrc_loss_hist *lh, struct sk_buff *skb)
 	if (tfrc_lh_length(lh) == 1)		/* due to RFC 3448, 6.3.1 */
 		return 0;
 
-	cur->li_length = length;
+	cur->li_length = len;
 	tfrc_lh_calc_i_mean(lh);
 
 	return (lh->i_mean < old_i_mean);
@@ -159,7 +159,7 @@ int tfrc_lh_interval_add(struct tfrc_loss_hist *lh, struct tfrc_rx_hist *rh,
 	else {
 		cur->li_length = dccp_delta_seqno(cur->li_seqno, new->li_seqno);
 		new->li_length = dccp_delta_seqno(new->li_seqno,
-				  tfrc_rx_hist_last_rcv(rh)->tfrchrx_seqno);
+				  tfrc_rx_hist_last_rcv(rh)->tfrchrx_seqno) + 1;
 		if (lh->counter > (2*LIH_SIZE))
 			lh->counter -= LIH_SIZE;
 
diff --git a/net/dccp/ccids/lib/packet_history.c b/net/dccp/ccids/lib/packet_history.c
index 20af1a69342..6cc108afdc3 100644
--- a/net/dccp/ccids/lib/packet_history.c
+++ b/net/dccp/ccids/lib/packet_history.c
@@ -153,7 +153,7 @@ void tfrc_rx_packet_history_exit(void)
 
 static inline void tfrc_rx_hist_entry_from_skb(struct tfrc_rx_hist_entry *entry,
 					       const struct sk_buff *skb,
-					       const u32 ndp)
+					       const u64 ndp)
 {
 	const struct dccp_hdr *dh = dccp_hdr(skb);
 
@@ -166,7 +166,7 @@ static inline void tfrc_rx_hist_entry_from_skb(struct tfrc_rx_hist_entry *entry,
 
 void tfrc_rx_hist_add_packet(struct tfrc_rx_hist *h,
 			     const struct sk_buff *skb,
-			     const u32 ndp)
+			     const u64 ndp)
 {
 	struct tfrc_rx_hist_entry *entry = tfrc_rx_hist_last_rcv(h);
 
@@ -206,31 +206,39 @@ static void tfrc_rx_hist_swap(struct tfrc_rx_hist *h, const u8 a, const u8 b)
  *
  * In the descriptions, `Si' refers to the sequence number of entry number i,
  * whose NDP count is `Ni' (lower case is used for variables).
- * Note: All __after_loss functions expect that a test against duplicates has
- *       been performed already: the seqno of the skb must not be less than the
- *       seqno of loss_prev; and it must not equal that of any valid hist_entry.
+ * Note: All __xxx_loss functions expect that a test against duplicates has been
+ *       performed already: the seqno of the skb must not be less than the seqno
+ *       of loss_prev; and it must not equal that of any valid history entry.
  */
+static void __do_track_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u64 n1)
+{
+	u64 s0 = tfrc_rx_hist_loss_prev(h)->tfrchrx_seqno,
+	    s1 = DCCP_SKB_CB(skb)->dccpd_seq;
+
+	if (!dccp_loss_free(s0, s1, n1)) {	/* gap between S0 and S1 */
+		h->loss_count = 1;
+		tfrc_rx_hist_entry_from_skb(tfrc_rx_hist_entry(h, 1), skb, n1);
+	}
+}
+
 static void __one_after_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u32 n2)
 {
 	u64 s0 = tfrc_rx_hist_loss_prev(h)->tfrchrx_seqno,
 	    s1 = tfrc_rx_hist_entry(h, 1)->tfrchrx_seqno,
 	    s2 = DCCP_SKB_CB(skb)->dccpd_seq;
-	int n1 = tfrc_rx_hist_entry(h, 1)->tfrchrx_ndp,
-	   d12 = dccp_delta_seqno(s1, s2), d2;
 
-	if (d12 > 0) {			/* S1  <  S2 */
+	if (likely(dccp_delta_seqno(s1, s2) > 0)) {	/* S1  <  S2 */
 		h->loss_count = 2;
 		tfrc_rx_hist_entry_from_skb(tfrc_rx_hist_entry(h, 2), skb, n2);
 		return;
 	}
 
 	/* S0  <  S2  <  S1 */
-	d2 = dccp_delta_seqno(s0, s2);
 
-	if (d2 == 1 || n2 >= d2) {	/* S2 is direct successor of S0 */
-		int d21 = -d12;
+	if (dccp_loss_free(s0, s2, n2)) {
+		u64 n1 = tfrc_rx_hist_entry(h, 1)->tfrchrx_ndp;
 
-		if (d21 == 1 || n1 >= d21) {
+		if (dccp_loss_free(s2, s1, n1)) {
 			/* hole is filled: S0, S2, and S1 are consecutive */
 			h->loss_count = 0;
 			h->loss_start = tfrc_rx_hist_index(h, 1);
@@ -238,9 +246,9 @@ static void __one_after_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u32 n2
 			/* gap between S2 and S1: just update loss_prev */
 			tfrc_rx_hist_entry_from_skb(tfrc_rx_hist_loss_prev(h), skb, n2);
 
-	} else {			/* hole between S0 and S2 */
+	} else {	/* gap between S0 and S2 */
 		/*
-		 * Reorder history to insert S2 between S0 and s1
+		 * Reorder history to insert S2 between S0 and S1
 		 */
 		tfrc_rx_hist_swap(h, 0, 3);
 		h->loss_start = tfrc_rx_hist_index(h, 3);
@@ -256,22 +264,18 @@ static int __two_after_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u32 n3)
 	    s1 = tfrc_rx_hist_entry(h, 1)->tfrchrx_seqno,
 	    s2 = tfrc_rx_hist_entry(h, 2)->tfrchrx_seqno,
 	    s3 = DCCP_SKB_CB(skb)->dccpd_seq;
-	int n1 = tfrc_rx_hist_entry(h, 1)->tfrchrx_ndp,
-	   d23 = dccp_delta_seqno(s2, s3), d13, d3, d31;
 
-	if (d23 > 0) {			/* S2  <  S3 */
+	if (likely(dccp_delta_seqno(s2, s3) > 0)) {	/* S2  <  S3 */
 		h->loss_count = 3;
 		tfrc_rx_hist_entry_from_skb(tfrc_rx_hist_entry(h, 3), skb, n3);
 		return 1;
 	}
 
 	/* S3  <  S2 */
-	d13 = dccp_delta_seqno(s1, s3);
 
-	if (d13 > 0) {
+	if (dccp_delta_seqno(s1, s3) > 0) {		/* S1  <  S3  <  S2 */
 		/*
-		 * The sequence number order is S1, S3, S2
-		 * Reorder history to insert entry between S1 and S2
+		 * Reorder history to insert S3 between S1 and S2
 		 */
 		tfrc_rx_hist_swap(h, 2, 3);
 		tfrc_rx_hist_entry_from_skb(tfrc_rx_hist_entry(h, 2), skb, n3);
@@ -280,17 +284,15 @@ static int __two_after_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u32 n3)
 	}
 
 	/* S0  <  S3  <  S1 */
-	d31 = -d13;
-	d3  = dccp_delta_seqno(s0, s3);
 
-	if (d3 == 1 || n3 >= d3) {	/* S3 is a successor of S0 */
+	if (dccp_loss_free(s0, s3, n3)) {
+		u64 n1 = tfrc_rx_hist_entry(h, 1)->tfrchrx_ndp;
 
-		if (d31 == 1 || n1 >= d31) {
+		if (dccp_loss_free(s3, s1, n1)) {
 			/* hole between S0 and S1 filled by S3 */
-			int  d2 = dccp_delta_seqno(s1, s2),
-			     n2 = tfrc_rx_hist_entry(h, 2)->tfrchrx_ndp;
+			u64 n2 = tfrc_rx_hist_entry(h, 2)->tfrchrx_ndp;
 
-			if (d2 == 1 || n2 >= d2) {
+			if (dccp_loss_free(s1, s2, n2)) {
 				/* entire hole filled by S0, S3, S1, S2 */
 				h->loss_start = tfrc_rx_hist_index(h, 2);
 				h->loss_count = 0;
@@ -307,8 +309,8 @@ static int __two_after_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u32 n3)
 	}
 
 	/*
-	 * The remaining case: S3 is not a successor of S0.
-	 * Sequence order is S0, S3, S1, S2; reorder to insert between S0 and S1
+	 * The remaining case:  S0  <  S3  <  S1  <  S2;  gap between S0 and S3
+	 * Reorder history to insert S3 between S0 and S1.
 	 */
 	tfrc_rx_hist_swap(h, 0, 3);
 	h->loss_start = tfrc_rx_hist_index(h, 3);
@@ -318,33 +320,25 @@ static int __two_after_loss(struct tfrc_rx_hist *h, struct sk_buff *skb, u32 n3)
 	return 1;
 }
 
-/* return the signed modulo-2^48 sequence number distance from entry e1 to e2 */
-static s64 tfrc_rx_hist_delta_seqno(struct tfrc_rx_hist *h, u8 e1, u8 e2)
-{
-	DCCP_BUG_ON(e1 > h->loss_count || e2 > h->loss_count);
-
-	return dccp_delta_seqno(tfrc_rx_hist_entry(h, e1)->tfrchrx_seqno,
-				tfrc_rx_hist_entry(h, e2)->tfrchrx_seqno);
-}
-
 /* recycle RX history records to continue loss detection if necessary */
 static void __three_after_loss(struct tfrc_rx_hist *h)
 {
 	/*
-	 * The distance between S0 and S1 is always greater than 1 and the NDP
-	 * count of S1 is smaller than this distance. Otherwise there would
-	 * have been no loss. Hence it is only necessary to see whether there
-	 * are further missing data packets between S1/S2 and S2/S3.
+	 * At this stage we know already that there is a gap between S0 and S1
+	 * (since S0 was the highest sequence number received before detecting
+	 * the loss). To recycle the loss record, it is	thus only necessary to
+	 * check for other possible gaps between S1/S2 and between S2/S3.
 	 */
-	int d2 = tfrc_rx_hist_delta_seqno(h, 1, 2),
-	    d3 = tfrc_rx_hist_delta_seqno(h, 2, 3),
-	    n2 = tfrc_rx_hist_entry(h, 2)->tfrchrx_ndp,
+	u64 s1 = tfrc_rx_hist_entry(h, 1)->tfrchrx_seqno,
+	    s2 = tfrc_rx_hist_entry(h, 2)->tfrchrx_seqno,
+	    s3 = tfrc_rx_hist_entry(h, 3)->tfrchrx_seqno;
+	u64 n2 = tfrc_rx_hist_entry(h, 2)->tfrchrx_ndp,
 	    n3 = tfrc_rx_hist_entry(h, 3)->tfrchrx_ndp;
 
-	if (d2 == 1 || n2 >= d2) {	/* S2 is successor to S1 */
+	if (dccp_loss_free(s1, s2, n2)) {
 
-		if (d3 == 1 || n3 >= d3) {
-			/* S3 is successor of S2: entire hole is filled */
+		if (dccp_loss_free(s2, s3, n3)) {
+			/* no gap between S2 and S3: entire hole is filled */
 			h->loss_start = tfrc_rx_hist_index(h, 3);
 			h->loss_count = 0;
 		} else {
@@ -353,7 +347,7 @@ static void __three_after_loss(struct tfrc_rx_hist *h)
 			h->loss_count = 1;
 		}
 
-	} else {			/* gap between S1 and S2 */
+	} else {	/* gap between S1 and S2 */
 		h->loss_start = tfrc_rx_hist_index(h, 1);
 		h->loss_count = 2;
 	}
@@ -370,15 +364,20 @@ static void __three_after_loss(struct tfrc_rx_hist *h)
  *  Chooses action according to pending loss, updates LI database when a new
  *  loss was detected, and does required post-processing. Returns 1 when caller
  *  should send feedback, 0 otherwise.
+ *  Since it also takes care of reordering during loss detection and updates the
+ *  records accordingly, the caller should not perform any more RX history
+ *  operations when loss_count is greater than 0 after calling this function.
  */
 int tfrc_rx_handle_loss(struct tfrc_rx_hist *h,
 			struct tfrc_loss_hist *lh,
-			struct sk_buff *skb, u32 ndp,
+			struct sk_buff *skb, const u64 ndp,
 			u32 (*calc_first_li)(struct sock *), struct sock *sk)
 {
 	int is_new_loss = 0;
 
-	if (h->loss_count == 1) {
+	if (h->loss_count == 0) {
+		__do_track_loss(h, skb, ndp);
+	} else if (h->loss_count == 1) {
 		__one_after_loss(h, skb, ndp);
 	} else if (h->loss_count != 2) {
 		DCCP_BUG("invalid loss_count %d", h->loss_count);
diff --git a/net/dccp/ccids/lib/packet_history.h b/net/dccp/ccids/lib/packet_history.h
index c7eeda49cb2..461cc91cce8 100644
--- a/net/dccp/ccids/lib/packet_history.h
+++ b/net/dccp/ccids/lib/packet_history.h
@@ -64,7 +64,7 @@ struct tfrc_rx_hist_entry {
 	u64		 tfrchrx_seqno:48,
 			 tfrchrx_ccval:4,
 			 tfrchrx_type:4;
-	u32		 tfrchrx_ndp; /* In fact it is from 8 to 24 bits */
+	u64		 tfrchrx_ndp:48;
 	ktime_t		 tfrchrx_tstamp;
 };
 
@@ -118,41 +118,21 @@ static inline struct tfrc_rx_hist_entry *
 	return h->ring[h->loss_start];
 }
 
-/* initialise loss detection and disable RTT sampling */
-static inline void tfrc_rx_hist_loss_indicated(struct tfrc_rx_hist *h)
-{
-	h->loss_count = 1;
-}
-
 /* indicate whether previously a packet was detected missing */
-static inline int tfrc_rx_hist_loss_pending(const struct tfrc_rx_hist *h)
-{
-	return h->loss_count;
-}
-
-/* any data packets missing between last reception and skb ? */
-static inline int tfrc_rx_hist_new_loss_indicated(struct tfrc_rx_hist *h,
-						  const struct sk_buff *skb,
-						  u32 ndp)
+static inline bool tfrc_rx_hist_loss_pending(const struct tfrc_rx_hist *h)
 {
-	int delta = dccp_delta_seqno(tfrc_rx_hist_last_rcv(h)->tfrchrx_seqno,
-				     DCCP_SKB_CB(skb)->dccpd_seq);
-
-	if (delta > 1 && ndp < delta)
-		tfrc_rx_hist_loss_indicated(h);
-
-	return tfrc_rx_hist_loss_pending(h);
+	return h->loss_count > 0;
 }
 
 extern void tfrc_rx_hist_add_packet(struct tfrc_rx_hist *h,
-				    const struct sk_buff *skb, const u32 ndp);
+				    const struct sk_buff *skb, const u64 ndp);
 
 extern int tfrc_rx_hist_duplicate(struct tfrc_rx_hist *h, struct sk_buff *skb);
 
 struct tfrc_loss_hist;
 extern int  tfrc_rx_handle_loss(struct tfrc_rx_hist *h,
 				struct tfrc_loss_hist *lh,
-				struct sk_buff *skb, u32 ndp,
+				struct sk_buff *skb, const u64 ndp,
 				u32 (*first_li)(struct sock *sk),
 				struct sock *sk);
 extern u32 tfrc_rx_hist_sample_rtt(struct tfrc_rx_hist *h,
diff --git a/net/dccp/dccp.h b/net/dccp/dccp.h
index 1b2cea244e1..32617e0576c 100644
--- a/net/dccp/dccp.h
+++ b/net/dccp/dccp.h
@@ -153,6 +153,21 @@ static inline u64 max48(const u64 seq1, const u64 seq2)
 	return after48(seq1, seq2) ? seq1 : seq2;
 }
 
+/**
+ * dccp_loss_free  -  Evaluates condition for data loss from RFC 4340, 7.7.1
+ * @s1:	 start sequence number
+ * @s2:  end sequence number
+ * @ndp: NDP count on packet with sequence number @s2
+ * Returns true if the sequence range s1...s2 has no data loss.
+ */
+static inline bool dccp_loss_free(const u64 s1, const u64 s2, const u64 ndp)
+{
+	s64 delta = dccp_delta_seqno(s1, s2);
+
+	BUG_TRAP(delta >= 0);
+	return (u64)delta <= ndp + 1;
+}
+
 enum {
 	DCCP_MIB_NUM = 0,
 	DCCP_MIB_ACTIVEOPENS,			/* ActiveOpens */
diff --git a/net/dccp/options.c b/net/dccp/options.c
index 43bc24e761d..dc7c158a2f4 100644
--- a/net/dccp/options.c
+++ b/net/dccp/options.c
@@ -124,12 +124,12 @@ int dccp_parse_options(struct sock *sk, struct dccp_request_sock *dreq,
 				mandatory = 1;
 			break;
 		case DCCPO_NDP_COUNT:
-			if (len > 3)
+			if (len > 6)
 				goto out_invalid_option;
 
 			opt_recv->dccpor_ndp = dccp_decode_value_var(value, len);
-			dccp_pr_debug("%s rx opt: NDP count=%d\n", dccp_role(sk),
-				      opt_recv->dccpor_ndp);
+			dccp_pr_debug("%s opt: NDP count=%llu\n", dccp_role(sk),
+				      (unsigned long long)opt_recv->dccpor_ndp);
 			break;
 		case DCCPO_CHANGE_L:
 			/* fall through */
@@ -307,9 +307,11 @@ static void dccp_encode_value_var(const u32 value, unsigned char *to,
 		*to++ = (value & 0xFF);
 }
 
-static inline int dccp_ndp_len(const int ndp)
+static inline u8 dccp_ndp_len(const u64 ndp)
 {
-	return likely(ndp <= 0xFF) ? 1 : ndp <= 0xFFFF ? 2 : 3;
+	if (likely(ndp <= 0xFF))
+		return 1;
+	return likely(ndp <= USHORT_MAX) ? 2 : (ndp <= UINT_MAX ? 4 : 6);
 }
 
 int dccp_insert_option(struct sock *sk, struct sk_buff *skb,
@@ -336,7 +338,7 @@ EXPORT_SYMBOL_GPL(dccp_insert_option);
 static int dccp_insert_option_ndp(struct sock *sk, struct sk_buff *skb)
 {
 	struct dccp_sock *dp = dccp_sk(sk);
-	int ndp = dp->dccps_ndp_count;
+	u64 ndp = dp->dccps_ndp_count;
 
 	if (dccp_non_data_packet(skb))
 		++dp->dccps_ndp_count;