diff options
Diffstat (limited to 'net/ipv4')
| -rw-r--r-- | net/ipv4/netfilter/nf_nat_rule.c | 2 | ||||
| -rw-r--r-- | net/ipv4/tcp_output.c | 22 | ||||
| -rw-r--r-- | net/ipv4/tcp_vegas.c | 82 |
3 files changed, 23 insertions, 83 deletions
diff --git a/net/ipv4/netfilter/nf_nat_rule.c b/net/ipv4/netfilter/nf_nat_rule.c index bea54a6..8d489e7 100644 --- a/net/ipv4/netfilter/nf_nat_rule.c +++ b/net/ipv4/netfilter/nf_nat_rule.c @@ -61,7 +61,7 @@ static struct static struct xt_table nat_table = { .name = "nat", .valid_hooks = NAT_VALID_HOOKS, - .lock = __RW_LOCK_UNLOCKED(__nat_table.lock), + .lock = __RW_LOCK_UNLOCKED(nat_table.lock), .me = THIS_MODULE, .af = AF_INET, }; diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c index 85b07eb..fe3b4bd 100644 --- a/net/ipv4/tcp_output.c +++ b/net/ipv4/tcp_output.c @@ -722,8 +722,7 @@ static void tcp_queue_skb(struct sock *sk, struct sk_buff *skb) static void tcp_set_skb_tso_segs(struct sock *sk, struct sk_buff *skb, unsigned int mss_now) { - if (skb->len <= mss_now || !sk_can_gso(sk) || - tcp_urg_mode(tcp_sk(sk))) { + if (skb->len <= mss_now || !sk_can_gso(sk)) { /* Avoid the costly divide in the normal * non-TSO case. */ @@ -1029,10 +1028,6 @@ unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu) /* Compute the current effective MSS, taking SACKs and IP options, * and even PMTU discovery events into account. - * - * LARGESEND note: !tcp_urg_mode is overkill, only frames up to snd_up - * cannot be large. However, taking into account rare use of URG, this - * is not a big flaw. */ unsigned int tcp_current_mss(struct sock *sk, int large_allowed) { @@ -1047,7 +1042,7 @@ unsigned int tcp_current_mss(struct sock *sk, int large_allowed) mss_now = tp->mss_cache; - if (large_allowed && sk_can_gso(sk) && !tcp_urg_mode(tp)) + if (large_allowed && sk_can_gso(sk)) doing_tso = 1; if (dst) { @@ -1164,9 +1159,7 @@ static int tcp_init_tso_segs(struct sock *sk, struct sk_buff *skb, { int tso_segs = tcp_skb_pcount(skb); - if (!tso_segs || - (tso_segs > 1 && (tcp_skb_mss(skb) != mss_now || - tcp_urg_mode(tcp_sk(sk))))) { + if (!tso_segs || (tso_segs > 1 && tcp_skb_mss(skb) != mss_now)) { tcp_set_skb_tso_segs(sk, skb, mss_now); tso_segs = tcp_skb_pcount(skb); } @@ -1519,6 +1512,10 @@ static int tcp_mtu_probe(struct sock *sk) * send_head. This happens as incoming acks open up the remote * window for us. * + * LARGESEND note: !tcp_urg_mode is overkill, only frames between + * snd_up-64k-mss .. snd_up cannot be large. However, taking into + * account rare use of URG, this is not a big flaw. + * * Returns 1, if no segments are in flight and we have queued segments, but * cannot send anything now because of SWS or another problem. */ @@ -1570,7 +1567,7 @@ static int tcp_write_xmit(struct sock *sk, unsigned int mss_now, int nonagle) } limit = mss_now; - if (tso_segs > 1) + if (tso_segs > 1 && !tcp_urg_mode(tp)) limit = tcp_mss_split_point(sk, skb, mss_now, cwnd_quota); @@ -1619,6 +1616,7 @@ void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss, */ void tcp_push_one(struct sock *sk, unsigned int mss_now) { + struct tcp_sock *tp = tcp_sk(sk); struct sk_buff *skb = tcp_send_head(sk); unsigned int tso_segs, cwnd_quota; @@ -1633,7 +1631,7 @@ void tcp_push_one(struct sock *sk, unsigned int mss_now) BUG_ON(!tso_segs); limit = mss_now; - if (tso_segs > 1) + if (tso_segs > 1 && !tcp_urg_mode(tp)) limit = tcp_mss_split_point(sk, skb, mss_now, cwnd_quota); diff --git a/net/ipv4/tcp_vegas.c b/net/ipv4/tcp_vegas.c index 14504da..a453aac 100644 --- a/net/ipv4/tcp_vegas.c +++ b/net/ipv4/tcp_vegas.c @@ -40,18 +40,14 @@ #include "tcp_vegas.h" -/* Default values of the Vegas variables, in fixed-point representation - * with V_PARAM_SHIFT bits to the right of the binary point. - */ -#define V_PARAM_SHIFT 1 -static int alpha = 2<<V_PARAM_SHIFT; -static int beta = 4<<V_PARAM_SHIFT; -static int gamma = 1<<V_PARAM_SHIFT; +static int alpha = 2; +static int beta = 4; +static int gamma = 1; module_param(alpha, int, 0644); -MODULE_PARM_DESC(alpha, "lower bound of packets in network (scale by 2)"); +MODULE_PARM_DESC(alpha, "lower bound of packets in network"); module_param(beta, int, 0644); -MODULE_PARM_DESC(beta, "upper bound of packets in network (scale by 2)"); +MODULE_PARM_DESC(beta, "upper bound of packets in network"); module_param(gamma, int, 0644); MODULE_PARM_DESC(gamma, "limit on increase (scale by 2)"); @@ -172,49 +168,13 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack, u32 in_flight) return; } - /* The key players are v_beg_snd_una and v_beg_snd_nxt. - * - * These are so named because they represent the approximate values - * of snd_una and snd_nxt at the beginning of the current RTT. More - * precisely, they represent the amount of data sent during the RTT. - * At the end of the RTT, when we receive an ACK for v_beg_snd_nxt, - * we will calculate that (v_beg_snd_nxt - v_beg_snd_una) outstanding - * bytes of data have been ACKed during the course of the RTT, giving - * an "actual" rate of: - * - * (v_beg_snd_nxt - v_beg_snd_una) / (rtt duration) - * - * Unfortunately, v_beg_snd_una is not exactly equal to snd_una, - * because delayed ACKs can cover more than one segment, so they - * don't line up nicely with the boundaries of RTTs. - * - * Another unfortunate fact of life is that delayed ACKs delay the - * advance of the left edge of our send window, so that the number - * of bytes we send in an RTT is often less than our cwnd will allow. - * So we keep track of our cwnd separately, in v_beg_snd_cwnd. - */ - if (after(ack, vegas->beg_snd_nxt)) { /* Do the Vegas once-per-RTT cwnd adjustment. */ - u32 old_wnd, old_snd_cwnd; - - - /* Here old_wnd is essentially the window of data that was - * sent during the previous RTT, and has all - * been acknowledged in the course of the RTT that ended - * with the ACK we just received. Likewise, old_snd_cwnd - * is the cwnd during the previous RTT. - */ - old_wnd = (vegas->beg_snd_nxt - vegas->beg_snd_una) / - tp->mss_cache; - old_snd_cwnd = vegas->beg_snd_cwnd; /* Save the extent of the current window so we can use this * at the end of the next RTT. */ - vegas->beg_snd_una = vegas->beg_snd_nxt; vegas->beg_snd_nxt = tp->snd_nxt; - vegas->beg_snd_cwnd = tp->snd_cwnd; /* We do the Vegas calculations only if we got enough RTT * samples that we can be reasonably sure that we got @@ -252,22 +212,14 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack, u32 in_flight) * * This is: * (actual rate in segments) * baseRTT - * We keep it as a fixed point number with - * V_PARAM_SHIFT bits to the right of the binary point. */ - target_cwnd = ((u64)old_wnd * vegas->baseRTT); - target_cwnd <<= V_PARAM_SHIFT; - do_div(target_cwnd, rtt); + target_cwnd = tp->snd_cwnd * vegas->baseRTT / rtt; /* Calculate the difference between the window we had, * and the window we would like to have. This quantity * is the "Diff" from the Arizona Vegas papers. - * - * Again, this is a fixed point number with - * V_PARAM_SHIFT bits to the right of the binary - * point. */ - diff = (old_wnd << V_PARAM_SHIFT) - target_cwnd; + diff = tp->snd_cwnd * (rtt-vegas->baseRTT) / vegas->baseRTT; if (diff > gamma && tp->snd_ssthresh > 2 ) { /* Going too fast. Time to slow down @@ -282,16 +234,13 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack, u32 in_flight) * truncation robs us of full link * utilization. */ - tp->snd_cwnd = min(tp->snd_cwnd, - ((u32)target_cwnd >> - V_PARAM_SHIFT)+1); + tp->snd_cwnd = min(tp->snd_cwnd, (u32)target_cwnd+1); } else if (tp->snd_cwnd <= tp->snd_ssthresh) { /* Slow start. */ tcp_slow_start(tp); } else { /* Congestion avoidance. */ - u32 next_snd_cwnd; /* Figure out where we would like cwnd * to be. @@ -300,32 +249,25 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack, u32 in_flight) /* The old window was too fast, so * we slow down. */ - next_snd_cwnd = old_snd_cwnd - 1; + tp->snd_cwnd--; } else if (diff < alpha) { /* We don't have enough extra packets * in the network, so speed up. */ - next_snd_cwnd = old_snd_cwnd + 1; + tp->snd_cwnd++; } else { /* Sending just as fast as we * should be. */ - next_snd_cwnd = old_snd_cwnd; } - - /* Adjust cwnd upward or downward, toward the - * desired value. - */ - if (next_snd_cwnd > tp->snd_cwnd) - tp->snd_cwnd++; - else if (next_snd_cwnd < tp->snd_cwnd) - tp->snd_cwnd--; } if (tp->snd_cwnd < 2) tp->snd_cwnd = 2; else if (tp->snd_cwnd > tp->snd_cwnd_clamp) tp->snd_cwnd = tp->snd_cwnd_clamp; + + tp->snd_ssthresh = tcp_current_ssthresh(sk); } /* Wipe the slate clean for the next RTT. */ |
