Date:	Mon, 28 Apr 2008 18:57:17 +0300 (EEST)
From:	"Ilpo Järvinen" <ilpo.jarvinen@...sinki.fi>
To:	Andy Furniss <lists@...yfurniss.entadsl.com>
cc:	Netdev <netdev@...r.kernel.org>
Subject: Re: WARNING: at net/ipv4/tcp_input.c

On Mon, 28 Apr 2008, Andy Furniss wrote:

> Since upgrading to 2.6.25 from 2.6.21.1 I've noticed a few of these.

...Thanks for the report, you could actually reproduce it in non-infinite 
time?? ...Finally somebody who might actually have a change to catch 
this shows up... :-)

> timestamps are off everything else tcp is default.
> 
> This is my gateway/torrent/TV server there is nearly always torrent traffic.

I've tried to catch these with torrent but never got anything, it seems to 
be quite sensitive to "network weather", which varies too much for me to 
catch it. Maybe I'll try w/o timestamps if that helps me to reproduce it 
(more likely).

> I've always had issues with the tulip nic (4 port zynx that doesn't like my
> desktop being off - it's connected with a crossover cable. The ports to a
> switch and modem are always on and OK) I haven't seen these warnings on
> 2.6.21.1.
>
> ------------[ cut here ]------------
> WARNING: at net/ipv4/tcp_input.c:2539 tcp_ack+0x1dc2/0x1df0()

We weren't checking this invariant in 2.6.21.1, actually, I made the 
invariant more rigid and added tracking for it to see what is breaking 
things. We still reset the state like the older did but I'd also like to 
find out the cause.

The debug patch below will add considerable amount of processing per ACK 
to verify where this invariant gets broken for the first time, in case 
that's fine with your server, please consider adding that and waiting 
until it spits something out... :-) That WARNING itself is usually too 
late but that's inexpensive enough to be always on to work as early 
warning system (and TCP anyway corrects the inaccurate state if it 
triggersbecause it's trivial to do there, in general these warning only 
pose a theoretical threat to network as a form of packet bursts or cause 
minor, nearly impossible to notice delays if one needs to recover by RTO 
instead of more timely means).


-- 
 i.

---
[PATCH] TCP: debug S+L (for 2.6.25/.26-rcs), v1.4.2

Debugs sacked & lost skb inconstencies in TCP write queue &
verifies fackets_out related variables.
---
 include/net/tcp.h     |   10 +++-
 net/ipv4/tcp_input.c  |   34 ++++++++++++-
 net/ipv4/tcp_ipv4.c   |  134 +++++++++++++++++++++++++++++++++++++++++++++++++
 net/ipv4/tcp_output.c |   20 ++++++--
 4 files changed, 191 insertions(+), 7 deletions(-)

diff --git a/include/net/tcp.h b/include/net/tcp.h
index 4fd3eb2..b7b73fb 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -272,6 +272,9 @@ DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
 #define TCP_ADD_STATS_BH(field, val)	SNMP_ADD_STATS_BH(tcp_statistics, field, val)
 #define TCP_ADD_STATS_USER(field, val)	SNMP_ADD_STATS_USER(tcp_statistics, field, val)
 
+extern void tcp_print_queue(struct sock *sk);
+extern int			tcp_verify_wq(struct sock *sk);
+
 extern void			tcp_v4_err(struct sk_buff *skb, u32);
 
 extern void			tcp_shutdown (struct sock *sk, int how);
@@ -770,7 +773,12 @@ static inline __u32 tcp_current_ssthresh(const struct sock *sk)
 }
 
 /* Use define here intentionally to get WARN_ON location shown at the caller */
-#define tcp_verify_left_out(tp)	WARN_ON(tcp_left_out(tp) > tp->packets_out)
+#define tcp_verify_left_out(tp)		\
+	do {\
+		int res; \
+		res = tcp_verify_wq((struct sock *)tp); \
+		WARN_ON(res || tcp_left_out(tp) > tp->packets_out); \
+	} while(0)
 
 extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
 extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index bbb7d88..7dd7939 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -1428,8 +1428,10 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff *ack_skb,
 	int first_sack_index;
 
 	if (!tp->sacked_out) {
-		if (WARN_ON(tp->fackets_out))
+		if (WARN_ON(tp->fackets_out)) {
+			tcp_verify_left_out(tp);
 			tp->fackets_out = 0;
+		}
 		tcp_highest_sack_reset(sk);
 	}
 
@@ -2157,7 +2159,14 @@ static void tcp_mark_head_lost(struct sock *sk, int packets)
 	int err;
 	unsigned int mss;
 
+	tcp_verify_left_out(tp);
+
 	BUG_TRAP(packets <= tp->packets_out);
+	if (packets > tp->packets_out) {
+		pr_err("TCP debug %u/%u-%u/%u %d\n",
+			tp->fackets_out, tp->sacked_out, tp->reordering,
+			tp->packets_out, tp->rx_opt.sack_ok);
+	}
 	if (tp->lost_skb_hint) {
 		skb = tp->lost_skb_hint;
 		cnt = tp->lost_cnt_hint;
@@ -2534,10 +2543,15 @@ static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked, int flag)
 				    (tcp_fackets_out(tp) > tp->reordering));
 	int fast_rexmit = 0;
 
+	tcp_verify_left_out(tp);
+
 	if (WARN_ON(!tp->packets_out && tp->sacked_out))
 		tp->sacked_out = 0;
-	if (WARN_ON(!tp->sacked_out && tp->fackets_out))
+	if (WARN_ON(!tp->sacked_out && tp->fackets_out)) {
+		printk(KERN_ERR "TCP DEBUG %u %u %d %d\n", tp->fackets_out,
+		       tp->packets_out, pkts_acked, tp->rx_opt.sack_ok);
 		tp->fackets_out = 0;
+	}
 
 	/* Now state machine starts.
 	 * A. ECE, hence prohibit cwnd undoing, the reduction is required. */
@@ -2680,6 +2694,10 @@ static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked, int flag)
 	if (do_lost || (tcp_is_fack(tp) && tcp_head_timedout(sk)))
 		tcp_update_scoreboard(sk, fast_rexmit);
 	tcp_cwnd_down(sk, flag);
+
+	WARN_ON(tcp_write_queue_head(sk) == NULL);
+	WARN_ON(!tp->packets_out);
+
 	tcp_xmit_retransmit_queue(sk);
 }
 
@@ -2832,6 +2850,9 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets)
 			tcp_mtup_probe_success(sk, skb);
 		}
 
+		WARN_ON((sacked & (TCPCB_SACKED_ACKED|TCPCB_LOST)) ==
+				  (TCPCB_SACKED_ACKED|TCPCB_LOST));
+
 		if (sacked & TCPCB_RETRANS) {
 			if (sacked & TCPCB_SACKED_RETRANS)
 				tp->retrans_out -= acked_pcount;
@@ -2922,6 +2943,8 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets)
 		}
 	}
 
+	tcp_verify_left_out(tp);
+
 #if FASTRETRANS_DEBUG > 0
 	BUG_TRAP((int)tp->sacked_out >= 0);
 	BUG_TRAP((int)tp->lost_out >= 0);
@@ -3210,6 +3233,8 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag)
 	prior_fackets = tp->fackets_out;
 	prior_in_flight = tcp_packets_in_flight(tp);
 
+	tcp_verify_left_out(tp);
+
 	if (!(flag & FLAG_SLOWPATH) && after(ack, prior_snd_una)) {
 		/* Window is constant, pure forward advance.
 		 * No more checks are required.
@@ -3269,9 +3294,14 @@ static int tcp_ack(struct sock *sk, struct sk_buff *skb, int flag)
 			tcp_cong_avoid(sk, ack, prior_in_flight);
 	}
 
+	if (WARN_ON((icsk->icsk_ca_state == TCP_CA_Open) && tcp_left_out(tp)))
+		tcp_print_queue(sk);
+
 	if ((flag & FLAG_FORWARD_PROGRESS) || !(flag & FLAG_NOT_DUP))
 		dst_confirm(sk->sk_dst_cache);
 
+	tcp_verify_left_out(tp);
+
 	return 1;
 
 no_queue:
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 00156bf..52447b0 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -108,6 +108,140 @@ struct inet_hashinfo __cacheline_aligned tcp_hashinfo = {
 	.lhash_wait  = __WAIT_QUEUE_HEAD_INITIALIZER(tcp_hashinfo.lhash_wait),
 };
 
+void tcp_print_queue(struct sock *sk)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+	struct sk_buff *skb;
+	char s[50+1];
+	char h[50+1];
+	int idx = 0;
+	int i;
+
+	i = 0;
+	tcp_for_write_queue(skb, sk) {
+		if (skb == tcp_send_head(sk))
+			printk(KERN_ERR "head %u %p\n", i, skb);
+		else
+			printk(KERN_ERR "skb %u %p\n", i, skb);
+		i++;
+	}
+
+	tcp_for_write_queue(skb, sk) {
+		if (skb == tcp_send_head(sk))
+			break;
+
+		for (i = 0; i < tcp_skb_pcount(skb); i++) {
+			if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED) {
+				s[idx] = 'S';
+				if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST)
+					s[idx] = 'B';
+
+			} else if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST) {
+				s[idx] = 'L';
+			} else {
+				s[idx] = ' ';
+			}
+			if (s[idx] != ' ' && skb->len < tp->mss_cache)
+				s[idx] += 'a' - 'A';
+
+			if (i == 0) {
+				if (skb == tcp_highest_sack(sk))
+					h[idx] = 'h';
+				else
+					h[idx] = '+';
+			} else {
+				h[idx] = '-';
+			}
+
+			if (++idx >= 50) {
+				s[idx] = 0;
+				h[idx] = 0;
+				printk(KERN_ERR "TCP wq(s) %s\n", s);
+				printk(KERN_ERR "TCP wq(h) %s\n", h);
+				idx = 0;
+			}
+		}
+	}
+	if (idx) {
+		s[idx] = '<';
+		s[idx+1] = 0;
+		h[idx] = '<';
+		h[idx+1] = 0;
+		printk(KERN_ERR "TCP wq(s) %s\n", s);
+		printk(KERN_ERR "TCP wq(h) %s\n", h);
+	}
+	printk(KERN_ERR "l%u s%u f%u p%u seq: su%u hs%u sn%u\n",
+		tp->lost_out, tp->sacked_out, tp->fackets_out, tp->packets_out,
+		tp->snd_una, tcp_highest_sack_seq(tp), tp->snd_nxt);
+}
+
+int tcp_verify_wq(struct sock *sk)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+	u32 lost = 0;
+	u32 sacked = 0;
+	u32 packets = 0;
+	u32 fackets = 0;
+	int hs_valid = 0;
+	int inconsistent = 0;
+	struct sk_buff *skb;
+
+	tcp_for_write_queue(skb, sk) {
+		if (skb == tcp_send_head(sk))
+			break;
+
+		if ((fackets == packets) && (skb == tp->highest_sack))
+			hs_valid = 1;
+
+		packets += tcp_skb_pcount(skb);
+
+		if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED) {
+			sacked += tcp_skb_pcount(skb);
+			if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST)
+				printk(KERN_ERR "Sacked bitmap S+L: %u %u-%u/%u\n",
+					TCP_SKB_CB(skb)->sacked,
+					TCP_SKB_CB(skb)->end_seq - tp->snd_una,
+					TCP_SKB_CB(skb)->seq - tp->snd_una,
+					tp->snd_una);
+			fackets = packets;
+			hs_valid = 0;
+		}
+		if (TCP_SKB_CB(skb)->sacked & TCPCB_LOST)
+			lost += tcp_skb_pcount(skb);
+	}
+
+	if ((fackets == packets) && (tp->highest_sack == tcp_send_head(sk)))
+		hs_valid = 1;
+
+	if ((lost != tp->lost_out) ||
+	    (tcp_is_sack(tp) && (sacked != tp->sacked_out)) ||
+	    ((sacked || (tcp_is_sack(tp) && tp->sacked_out)) && !hs_valid) ||
+	    (packets != tp->packets_out) ||
+	    (fackets != tp->fackets_out) ||
+	    (tcp_is_reno(tp) && tp->packets_out &&
+	     (tcp_left_out(tp) == tp->packets_out) &&
+	     !(TCP_SKB_CB(tcp_write_queue_head(sk))->sacked & TCPCB_LOST)) ||
+	    tcp_left_out(tp) > tp->packets_out) {
+		printk(KERN_ERR "P: %u L: %u vs %u S: %u vs %u F: %u vs %u w: %u-%u (%u)\n",
+			tp->packets_out,
+			lost, tp->lost_out,
+			sacked, tp->sacked_out,
+			fackets, tp->fackets_out,
+			tp->snd_una, tp->snd_nxt,
+			tp->rx_opt.sack_ok);
+		tcp_print_queue(sk);
+		inconsistent = 1;
+	}
+
+	WARN_ON(lost != tp->lost_out);
+	WARN_ON(tcp_is_sack(tp) && (sacked != tp->sacked_out));
+	WARN_ON(packets != tp->packets_out);
+	WARN_ON(fackets != tp->fackets_out);
+	WARN_ON(tcp_left_out(tp) > tp->packets_out);
+
+	return inconsistent;
+}
+
 static inline __u32 tcp_v4_init_sequence(struct sk_buff *skb)
 {
 	return secure_tcp_sequence_number(ip_hdr(skb)->daddr,
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index d29ef79..84113be 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -779,10 +779,9 @@ int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len,
 			tp->lost_out -= diff;
 
 		/* Adjust Reno SACK estimate. */
-		if (tcp_is_reno(tp) && diff > 0) {
+		if (tcp_is_reno(tp) && diff > 0)
 			tcp_dec_pcount_approx_int(&tp->sacked_out, diff);
-			tcp_verify_left_out(tp);
-		}
+
 		tcp_adjust_fackets_out(sk, skb, diff);
 	}
 
@@ -790,6 +789,8 @@ int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32 len,
 	skb_header_release(buff);
 	tcp_insert_write_queue_after(skb, buff, sk);
 
+	tcp_verify_left_out(tp);
+
 	return 0;
 }
 
@@ -1471,6 +1472,7 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle)
 	} else if (result > 0) {
 		sent_pkts = 1;
 	}
+	tcp_verify_left_out(tp);
 
 	while ((skb = tcp_send_head(sk))) {
 		unsigned int limit;
@@ -1772,6 +1774,7 @@ static void tcp_retrans_try_collapse(struct sock *sk, struct sk_buff *skb,
 	tcp_clear_retrans_hints_partial(tp);
 
 	sk_wmem_free_skb(sk, next_skb);
+	tcp_verify_left_out(tp);
 }
 
 /* Do a simple retransmit without using the backoff mechanisms in
@@ -1895,10 +1898,13 @@ int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb)
 	    (TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN) &&
 	    tp->snd_una == (TCP_SKB_CB(skb)->end_seq - 1)) {
 		if (!pskb_trim(skb, 0)) {
+			WARN_ON(TCP_SKB_CB(skb)->sacked & TCPCB_LOST);
 			/* Reuse, even though it does some unnecessary work */
 			tcp_init_nondata_skb(skb, TCP_SKB_CB(skb)->end_seq - 1,
 					     TCP_SKB_CB(skb)->flags);
 			skb->ip_summed = CHECKSUM_NONE;
+
+			tcp_verify_left_out(tp);
 		}
 	}
 
@@ -1981,8 +1987,10 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 			 * packet to be MSS sized and all the
 			 * packet counting works out.
 			 */
-			if (tcp_packets_in_flight(tp) >= tp->snd_cwnd)
+			if (tcp_packets_in_flight(tp) >= tp->snd_cwnd) {
+				tcp_verify_left_out(tp);
 				return;
+			}
 
 			if (sacked & TCPCB_LOST) {
 				if (!(sacked & (TCPCB_SACKED_ACKED|TCPCB_SACKED_RETRANS))) {
@@ -2008,6 +2016,8 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 		}
 	}
 
+	tcp_verify_left_out(tp);
+
 	/* OK, demanded retransmission is finished. */
 
 	/* Forward retransmissions are possible only during Recovery. */
@@ -2065,6 +2075,8 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
 
 		NET_INC_STATS_BH(LINUX_MIB_TCPFORWARDRETRANS);
 	}
+
+	tcp_verify_left_out(tp);
 }
 
 /* Send a fin.  The caller locks the socket for us.  This cannot be
-- 
1.5.2.2

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists