From: =?ISO-8859-1?q?Ilpo_J=E4rvinen?= <[EMAIL PROTECTED]>
Previously code had IsReno/IsFack defined as macros that were
local to tcp_input.c though sack_ok field has users elsewhere
too for the same purpose. This changes them to static inlines as
preferred according the current coding style and unifies the
access to sack_ok across multiple files. Magic bitops of sack_ok
for FACK and DSACK are also abstracted to functions with
appropriate names.
Note:
- One sack_ok = 1 remains but that's self explanary, i.e., it
enables sack
- Couple of !IsReno cases are changed to tcp_is_sack
- There were no users for IsDSack => I dropped it
Signed-off-by: Ilpo Järvinen <[EMAIL PROTECTED]>
---
include/net/tcp.h | 28 ++++++++++++
net/ipv4/tcp.c | 2 +-
net/ipv4/tcp_input.c | 103 +++++++++++++++++++++++++---------------------
net/ipv4/tcp_minisocks.c | 2 +-
net/ipv4/tcp_output.c | 6 +-
net/ipv4/tcp_timer.c | 2 +-
6 files changed, 90 insertions(+), 53 deletions(-)
diff --git a/include/net/tcp.h b/include/net/tcp.h
index d12154a..dd9749e 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -715,6 +715,34 @@ static inline void tcp_ca_event(struct sock *sk, const
enum tcp_ca_event event)
icsk->icsk_ca_ops->cwnd_event(sk, event);
}
+/* These functions determine how the current flow behaves in respect of SACK
+ * handling. SACK is negotiated with the peer, and therefore it can vary
+ * between different flows.
+ *
+ * tcp_is_sack - SACK enabled
+ * tcp_is_reno - No SACK
+ * tcp_is_fack - FACK enabled, implies SACK enabled
+ */
+static inline int tcp_is_sack(const struct tcp_sock *tp)
+{
+ return tp->rx_opt.sack_ok;
+}
+
+static inline int tcp_is_reno(const struct tcp_sock *tp)
+{
+ return !tcp_is_sack(tp);
+}
+
+static inline int tcp_is_fack(const struct tcp_sock *tp)
+{
+ return tp->rx_opt.sack_ok & 2;
+}
+
+static inline void tcp_enable_fack(struct tcp_sock *tp)
+{
+ tp->rx_opt.sack_ok |= 2;
+}
+
/* This determines how many packets are "in the network" to the best
* of our knowledge. In many cases it is conservative, but where
* detailed information is available from the receiver (via SACK
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index bd4c295..e3de919 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2000,7 +2000,7 @@ void tcp_get_info(struct sock *sk, struct tcp_info *info)
if (tp->rx_opt.tstamp_ok)
info->tcpi_options |= TCPI_OPT_TIMESTAMPS;
- if (tp->rx_opt.sack_ok)
+ if (tcp_is_sack(tp))
info->tcpi_options |= TCPI_OPT_SACK;
if (tp->rx_opt.wscale_ok) {
info->tcpi_options |= TCPI_OPT_WSCALE;
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index f3942c3..3efd4f5 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -108,11 +108,6 @@ int sysctl_tcp_abc __read_mostly;
#define FLAG_CA_ALERT (FLAG_DATA_SACKED|FLAG_ECE)
#define FLAG_FORWARD_PROGRESS (FLAG_ACKED|FLAG_DATA_SACKED)
-#define IsReno(tp) ((tp)->rx_opt.sack_ok == 0)
-#define IsFack(tp) ((tp)->rx_opt.sack_ok & 2)
-#define IsDSack(tp) ((tp)->rx_opt.sack_ok & 4)
-#define Is3517Sack(tp) (!IsReno(tp) && !IsFack(tp))
-
#define IsSackFrto() (sysctl_tcp_frto == 0x2)
#define TCP_REMNANT (TCP_FLAG_FIN|TCP_FLAG_URG|TCP_FLAG_SYN|TCP_FLAG_PSH)
@@ -840,6 +835,21 @@ void tcp_enter_cwr(struct sock *sk, const int set_ssthresh)
}
}
+/*
+ * Packet counting of FACK is based on in-order assumptions, therefore TCP
+ * disables it when reordering is detected
+ */
+static void tcp_disable_fack(struct tcp_sock *tp)
+{
+ tp->rx_opt.sack_ok &= ~2;
+}
+
+/* Take a notice that peer is sending DSACKs */
+static void tcp_dsack_seen(struct tcp_sock *tp)
+{
+ tp->rx_opt.sack_ok |= 4;
+}
+
/* Initialize metrics on socket. */
static void tcp_init_metrics(struct sock *sk)
@@ -861,7 +871,7 @@ static void tcp_init_metrics(struct sock *sk)
}
if (dst_metric(dst, RTAX_REORDERING) &&
tp->reordering != dst_metric(dst, RTAX_REORDERING)) {
- tp->rx_opt.sack_ok &= ~2;
+ tcp_disable_fack(tp);
tp->reordering = dst_metric(dst, RTAX_REORDERING);
}
@@ -923,9 +933,9 @@ static void tcp_update_reordering(struct sock *sk, const
int metric,
/* This exciting event is worth to be remembered. 8) */
if (ts)
NET_INC_STATS_BH(LINUX_MIB_TCPTSREORDER);
- else if (IsReno(tp))
+ else if (tcp_is_reno(tp))
NET_INC_STATS_BH(LINUX_MIB_TCPRENOREORDER);
- else if (IsFack(tp))
+ else if (tcp_is_fack(tp))
NET_INC_STATS_BH(LINUX_MIB_TCPFACKREORDER);
else
NET_INC_STATS_BH(LINUX_MIB_TCPSACKREORDER);
@@ -937,8 +947,7 @@ static void tcp_update_reordering(struct sock *sk, const
int metric,
tp->sacked_out,
tp->undo_marker ? tp->undo_retrans : 0);
#endif
- /* Disable FACK yet. */
- tp->rx_opt.sack_ok &= ~2;
+ tcp_disable_fack(tp);
}
}
@@ -1009,7 +1018,7 @@ static int tcp_check_dsack(struct tcp_sock *tp, struct
sk_buff *ack_skb,
if (before(start_seq_0, TCP_SKB_CB(ack_skb)->ack_seq)) {
dup_sack = 1;
- tp->rx_opt.sack_ok |= 4;
+ tcp_dsack_seen(tp);
NET_INC_STATS_BH(LINUX_MIB_TCPDSACKRECV);
} else if (num_sacks > 1) {
u32 end_seq_1 = ntohl(get_unaligned(&sp[1].end_seq));
@@ -1018,7 +1027,7 @@ static int tcp_check_dsack(struct tcp_sock *tp, struct
sk_buff *ack_skb,
if (!after(end_seq_0, end_seq_1) &&
!before(start_seq_0, start_seq_1)) {
dup_sack = 1;
- tp->rx_opt.sack_ok |= 4;
+ tcp_dsack_seen(tp);
NET_INC_STATS_BH(LINUX_MIB_TCPDSACKOFORECV);
}
}
@@ -1319,7 +1328,8 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff
*ack_skb,
* we have to account for reordering! Ugly,
* but should help.
*/
- if (IsFack(tp) && state.lost_retrans && icsk->icsk_ca_state ==
TCP_CA_Recovery) {
+ if (tcp_is_fack(tp) && state.lost_retrans &&
+ icsk->icsk_ca_state == TCP_CA_Recovery) {
struct sk_buff *skb;
tcp_for_write_queue(skb, sk) {
@@ -1331,7 +1341,7 @@ tcp_sacktag_write_queue(struct sock *sk, struct sk_buff
*ack_skb,
continue;
if ((TCP_SKB_CB(skb)->sacked&TCPCB_SACKED_RETRANS) &&
after(state.lost_retrans, TCP_SKB_CB(skb)->ack_seq)
&&
- (IsFack(tp) ||
+ (tcp_is_fack(tp) ||
!before(state.lost_retrans,
TCP_SKB_CB(skb)->ack_seq + tp->reordering *
tp->mss_cache))) {
@@ -1529,7 +1539,7 @@ static void tcp_enter_frto_loss(struct sock *sk, int
allowed_segments, int flag)
tp->lost_out = 0;
tp->retrans_out = 0;
- if (IsReno(tp))
+ if (tcp_is_reno(tp))
tcp_reset_reno_sack(tp);
tcp_for_write_queue(skb, sk) {
@@ -1669,8 +1679,7 @@ static int tcp_check_sack_reneging(struct sock *sk)
static inline int tcp_fackets_out(struct tcp_sock *tp)
{
- return (IsReno(tp) || Is3517Sack(tp)) ? tp->sacked_out + 1 :
- tp->fackets_out;
+ return tcp_is_fack(tp) ? tp->fackets_out : tp->sacked_out + 1;
}
static inline int tcp_skb_timedout(struct sock *sk, struct sk_buff *skb)
@@ -1799,7 +1808,7 @@ static int tcp_time_to_recover(struct sock *sk)
/* Trick#3 : when we use RFC2988 timer restart, fast
* retransmit can be triggered by timeout of queue head.
*/
- if (IsFack(tp) && tcp_head_timedout(sk))
+ if (tcp_is_fack(tp) && tcp_head_timedout(sk))
return 1;
/* Trick#4: It is still not OK... But will it be useful to delay
@@ -1924,7 +1933,7 @@ static struct sk_buff *tcp_update_scoreboard_fack(struct
sock *sk,
}
timedout_continue = NULL;
- if (IsFack(tp) && tcp_skb_timedout(sk, skb) &&
+ if (tcp_is_fack(tp) && tcp_skb_timedout(sk, skb) &&
(tp->fackets_out < tp->packets_out)) {
timedout_continue = tcp_write_queue_next(sk, skb);
if (!after(entry_seq, tp->high_seq)) {
@@ -1935,7 +1944,7 @@ static struct sk_buff *tcp_update_scoreboard_fack(struct
sock *sk,
}
} else {
- unsigned int reord_count = IsFack(tp) ? 0 : 1;
+ unsigned int reord_count = tcp_is_fack(tp) ? 0 : 1;
skb = tcp_write_queue_find(sk, entry_seq);
/* If this ever becomes expensive, it can be delayed */
@@ -1948,12 +1957,12 @@ static struct sk_buff
*tcp_update_scoreboard_fack(struct sock *sk,
(TCP_SKB_CB(skb)->sacked & TCPCB_LOST))
goto backwards_walk_done;
- if (IsFack(tp) && tcp_skb_timedout(sk, skb))
+ if (tcp_is_fack(tp) && tcp_skb_timedout(sk, skb))
break;
else
timedout_continue = NULL;
- if (IsFack(tp) ||
+ if (tcp_is_fack(tp) ||
(TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
reord_count += tcp_skb_pcount(skb);
if (reord_count > tp->reordering) {
@@ -1989,7 +1998,7 @@ static struct sk_buff *tcp_update_scoreboard_fack(struct
sock *sk,
}
/* Phase III: Nothing is still marked?, mark head then */
- if ((IsFack(tp) || fast_rexmit) && !tp->lost_out)
+ if ((tcp_is_fack(tp) || fast_rexmit) && !tp->lost_out)
tcp_mark_head_lost_single(sk);
backwards_walk_done:
@@ -2003,9 +2012,9 @@ static void tcp_update_scoreboard(struct sock *sk, u32
sack_entry_seq,
struct tcp_sock *tp = tcp_sk(sk);
struct sk_buff *skb = NULL;
- if (!IsReno(tp)) {
+ if (tcp_is_sack(tp)) {
if (!tp->sacked_out) {
- if (IsFack(tp) || fast_rexmit) {
+ if (tcp_is_fack(tp) || fast_rexmit) {
tcp_mark_head_lost_single(sk);
skb = tcp_write_queue_head(sk);
skb = tcp_write_queue_next(sk, skb);
@@ -2015,7 +2024,7 @@ static void tcp_update_scoreboard(struct sock *sk, u32
sack_entry_seq,
fast_rexmit);
/* Continue with timedout work */
- if (IsFack(tp) && (skb != NULL))
+ if (tcp_is_fack(tp) && (skb != NULL))
tcp_timedout_mark_forward(sk, skb);
} else
@@ -2139,7 +2148,7 @@ static int tcp_try_undo_recovery(struct sock *sk)
NET_INC_STATS_BH(LINUX_MIB_TCPFULLUNDO);
tp->undo_marker = 0;
}
- if (tp->snd_una == tp->high_seq && IsReno(tp)) {
+ if (tp->snd_una == tp->high_seq && tcp_is_reno(tp)) {
/* Hold old state until something *above* high_seq
* is ACKed. For Reno it is MUST to prevent false
* fast retransmits (RFC2582). SACK TCP is safe. */
@@ -2169,7 +2178,7 @@ static int tcp_try_undo_partial(struct sock *sk, int
acked)
{
struct tcp_sock *tp = tcp_sk(sk);
/* Partial ACK arrived. Force Hoe's retransmit. */
- int failed = IsReno(tp) || (tcp_fackets_out(tp) > tp->reordering);
+ int failed = tcp_is_reno(tp) || (tcp_fackets_out(tp) > tp->reordering);
if (tcp_may_undo(tp)) {
/* Plain luck! Hole if filled with delayed
@@ -2214,7 +2223,7 @@ static int tcp_try_undo_loss(struct sock *sk)
NET_INC_STATS_BH(LINUX_MIB_TCPLOSSUNDO);
inet_csk(sk)->icsk_retransmits = 0;
tp->undo_marker = 0;
- if (!IsReno(tp))
+ if (tcp_is_sack(tp))
tcp_set_ca_state(sk, TCP_CA_Open);
return 1;
}
@@ -2321,7 +2330,7 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una,
return;
/* C. Process data loss notification, provided it is valid. */
- if (IsFack(tp) && (flag&FLAG_DATA_LOST) &&
+ if (tcp_is_fack(tp) && (flag&FLAG_DATA_LOST) &&
before(tp->snd_una, tp->high_seq) &&
icsk->icsk_ca_state != TCP_CA_Open &&
tp->fackets_out > tp->reordering) {
@@ -2359,14 +2368,14 @@ tcp_fastretrans_alert(struct sock *sk, u32
prior_snd_una,
if (!tp->undo_marker ||
/* For SACK case do not Open to allow to undo
* catching for all duplicate ACKs. */
- IsReno(tp) || tp->snd_una != tp->high_seq) {
+ tcp_is_reno(tp) || tp->snd_una != tp->high_seq) {
tp->undo_marker = 0;
tcp_set_ca_state(sk, TCP_CA_Open);
}
break;
case TCP_CA_Recovery:
- if (IsReno(tp))
+ if (tcp_is_reno(tp))
tcp_reset_reno_sack(tp);
if (tcp_try_undo_recovery(sk))
return;
@@ -2379,11 +2388,11 @@ tcp_fastretrans_alert(struct sock *sk, u32
prior_snd_una,
switch (icsk->icsk_ca_state) {
case TCP_CA_Recovery:
if (prior_snd_una == tp->snd_una) {
- if (IsReno(tp) && is_dupack)
+ if (tcp_is_reno(tp) && is_dupack)
tcp_add_reno_sack(sk);
} else {
int acked = prior_packets - tp->packets_out;
- if (IsReno(tp))
+ if (tcp_is_reno(tp))
tcp_remove_reno_sacks(sk, acked);
is_dupack = tcp_try_undo_partial(sk, acked);
}
@@ -2400,7 +2409,7 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una,
return;
/* Loss is undone; fall through to processing in Open state. */
default:
- if (IsReno(tp)) {
+ if (tcp_is_reno(tp)) {
if (tp->snd_una != prior_snd_una)
tcp_reset_reno_sack(tp);
if (is_dupack)
@@ -2428,7 +2437,7 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una,
/* Otherwise enter Recovery state */
- if (IsReno(tp))
+ if (tcp_is_reno(tp))
NET_INC_STATS_BH(LINUX_MIB_TCPRENORECOVERY);
else
NET_INC_STATS_BH(LINUX_MIB_TCPSACKRECOVERY);
@@ -2451,7 +2460,7 @@ tcp_fastretrans_alert(struct sock *sk, u32 prior_snd_una,
fast_rexmit = 1;
}
- if (is_dupack || (IsFack(tp) && tcp_head_timedout(sk)))
+ if (is_dupack || (tcp_is_fack(tp) && tcp_head_timedout(sk)))
tcp_update_scoreboard(sk, mark_lost_entry_seq, fast_rexmit);
tcp_cwnd_down(sk);
tcp_xmit_retransmit_queue(sk);
@@ -2691,7 +2700,7 @@ static int tcp_clean_rtx_queue(struct sock *sk, __s32
*seq_rtt_p)
BUG_TRAP((int)tp->sacked_out >= 0);
BUG_TRAP((int)tp->lost_out >= 0);
BUG_TRAP((int)tp->retrans_out >= 0);
- if (!tp->packets_out && tp->rx_opt.sack_ok) {
+ if (!tp->packets_out && tcp_is_sack(tp)) {
const struct inet_connection_sock *icsk = inet_csk(sk);
if (tp->lost_out) {
printk(KERN_DEBUG "Leak l=%u %d\n",
@@ -2871,7 +2880,7 @@ static int tcp_process_frto(struct sock *sk, u32
prior_snd_una, int flag)
return 1;
}
- if (!IsSackFrto() || IsReno(tp)) {
+ if (!IsSackFrto() || tcp_is_reno(tp)) {
/* RFC4138 shortcoming in step 2; should also have case c):
* ACK isn't duplicate nor advances window, e.g., opposite dir
* data, winupdate
@@ -3356,7 +3365,7 @@ static void tcp_fin(struct sk_buff *skb, struct sock *sk,
struct tcphdr *th)
* Probably, we should reset in this case. For now drop them.
*/
__skb_queue_purge(&tp->out_of_order_queue);
- if (tp->rx_opt.sack_ok)
+ if (tcp_is_sack(tp))
tcp_sack_reset(&tp->rx_opt);
sk_stream_mem_reclaim(sk);
@@ -3386,7 +3395,7 @@ static inline int tcp_sack_extend(struct tcp_sack_block
*sp, u32 seq, u32 end_se
static void tcp_dsack_set(struct tcp_sock *tp, u32 seq, u32 end_seq)
{
- if (tp->rx_opt.sack_ok && sysctl_tcp_dsack) {
+ if (tcp_is_sack(tp) && sysctl_tcp_dsack) {
if (before(seq, tp->rcv_nxt))
NET_INC_STATS_BH(LINUX_MIB_TCPDSACKOLDSENT);
else
@@ -3416,7 +3425,7 @@ static void tcp_send_dupack(struct sock *sk, struct
sk_buff *skb)
NET_INC_STATS_BH(LINUX_MIB_DELAYEDACKLOST);
tcp_enter_quickack_mode(sk);
- if (tp->rx_opt.sack_ok && sysctl_tcp_dsack) {
+ if (tcp_is_sack(tp) && sysctl_tcp_dsack) {
u32 end_seq = TCP_SKB_CB(skb)->end_seq;
if (after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt))
@@ -3732,7 +3741,7 @@ drop:
if (!skb_peek(&tp->out_of_order_queue)) {
/* Initial out of order segment, build 1 SACK. */
- if (tp->rx_opt.sack_ok) {
+ if (tcp_is_sack(tp)) {
tp->rx_opt.num_sacks = 1;
tp->rx_opt.dsack = 0;
tp->rx_opt.eff_sacks = 1;
@@ -3797,7 +3806,7 @@ drop:
}
add_sack:
- if (tp->rx_opt.sack_ok)
+ if (tcp_is_sack(tp))
tcp_sack_new_ofo_skb(sk, seq, end_seq);
}
}
@@ -3986,7 +3995,7 @@ static int tcp_prune_queue(struct sock *sk)
* is in a sad state like this, we care only about integrity
* of the connection not performance.
*/
- if (tp->rx_opt.sack_ok)
+ if (tcp_is_sack(tp))
tcp_sack_reset(&tp->rx_opt);
sk_stream_mem_reclaim(sk);
}
@@ -4687,8 +4696,8 @@ static int tcp_rcv_synsent_state_process(struct sock *sk,
struct sk_buff *skb,
tp->tcp_header_len = sizeof(struct tcphdr);
}
- if (tp->rx_opt.sack_ok && sysctl_tcp_fack)
- tp->rx_opt.sack_ok |= 2;
+ if (tcp_is_sack(tp) && sysctl_tcp_fack)
+ tcp_enable_fack(tp);
tcp_mtup_init(sk);
tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
diff --git a/net/ipv4/tcp_minisocks.c b/net/ipv4/tcp_minisocks.c
index 0ad4f36..2bd0625 100644
--- a/net/ipv4/tcp_minisocks.c
+++ b/net/ipv4/tcp_minisocks.c
@@ -446,7 +446,7 @@ struct sock *tcp_create_openreq_child(struct sock *sk,
struct request_sock *req,
newtp->rx_opt.tstamp_ok = ireq->tstamp_ok;
if ((newtp->rx_opt.sack_ok = ireq->sack_ok) != 0) {
if (sysctl_tcp_fack)
- newtp->rx_opt.sack_ok |= 2;
+ tcp_enable_fack(newtp);
}
newtp->window_clamp = req->window_clamp;
newtp->rcv_ssthresh = req->rcv_wnd;
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index c8be1c8..25804a3 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -737,7 +737,7 @@ int tcp_fragment(struct sock *sk, struct sk_buff *skb, u32
len, unsigned int mss
if (diff > 0) {
/* Adjust Reno SACK estimate. */
- if (!tp->rx_opt.sack_ok) {
+ if (tcp_is_reno(tp)) {
tcp_dec_pcount_approx_int(&tp->sacked_out,
diff);
tcp_sync_left_out(tp);
}
@@ -1722,7 +1722,7 @@ static void tcp_retrans_try_collapse(struct sock *sk,
struct sk_buff *skb, int m
if (TCP_SKB_CB(next_skb)->sacked&TCPCB_LOST)
tp->lost_out -= tcp_skb_pcount(next_skb);
/* Reno case is special. Sigh... */
- if (!tp->rx_opt.sack_ok && tp->sacked_out)
+ if (tcp_is_reno(tp) && tp->sacked_out)
tcp_dec_pcount_approx(&tp->sacked_out, next_skb);
/* Not quite right: it can be > snd.fack, but
@@ -1970,7 +1970,7 @@ void tcp_xmit_retransmit_queue(struct sock *sk)
return;
/* No forward retransmissions in Reno are possible. */
- if (!tp->rx_opt.sack_ok)
+ if (tcp_is_reno(tp))
return;
/* Yeah, we have to make difficult choice between forward transmission
diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index 2ca97b2..a664a76 100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c
@@ -317,7 +317,7 @@ static void tcp_retransmit_timer(struct sock *sk)
if (icsk->icsk_retransmits == 0) {
if (icsk->icsk_ca_state == TCP_CA_Disorder ||
icsk->icsk_ca_state == TCP_CA_Recovery) {
- if (tp->rx_opt.sack_ok) {
+ if (tcp_is_sack(tp)) {
if (icsk->icsk_ca_state == TCP_CA_Recovery)
NET_INC_STATS_BH(LINUX_MIB_TCPSACKRECOVERYFAIL);
else
--
1.5.0.6
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
