(Yup, it is less than 40 KiB.) Signed-off-by: Tommi Saviranta <wnd@xxxxxx> --- diff --git a/net/dccp/ccids/ccid4.c b/net/dccp/ccids/ccid4.c new file mode 100644 index 0000000..7610c6e --- /dev/null +++ b/net/dccp/ccids/ccid4.c @@ -0,0 +1,984 @@ +/* + * net/dccp/ccids/ccid4.c + * + * Copyright (c) 2007 The University of Aberdeen, Scotland, UK + * Copyright (c) 2005-7 The University of Waikato, Hamilton, New Zealand. + * Copyright (c) 2005-7 Ian McDonald <ian.mcdonald@xxxxxxxxxxx> + * + * An implementation of the DCCP protocol + * + * This code has been developed by the University of Waikato WAND + * research group. For further information please see http://www.wand.net.nz/ + * + * This code also uses code from Lulea University, rereleased as GPL by its + * authors: + * Copyright (c) 2003 Nils-Erik Mattsson, Joacim Haggmark, Magnus Erixzon + * + * Changes to meet Linux coding standards, to make it meet latest ccid4 draft + * and to make it work as a loadable module in the DCCP stack written by + * Arnaldo Carvalho de Melo <acme@xxxxxxxxxxxxxxxx>. + * + * Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@xxxxxxxxxxxxxxxx> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ +#include "../dccp.h" +#include "ccid4.h" + +#ifdef CONFIG_IP_DCCP_CCID4_DEBUG +static int ccid4_debug; +#define ccid4_pr_debug(format, a...) DCCP_PR_DEBUG(ccid4_debug, format, ##a) +#else +#define ccid4_pr_debug(format, a...) +#endif + +DECLARE_TFRC_TX_CACHE(ccid4_tx_hist); + +/* + * Transmitter Half-Connection Routines + */ +#ifdef CONFIG_IP_DCCP_CCID4_DEBUG +static const char *ccid4_tx_state_name(enum ccid4_hc_tx_states state) +{ + static char *ccid4_state_names[] = { + [TFRC_SSTATE_NO_SENT] = "NO_SENT", + [TFRC_SSTATE_NO_FBACK] = "NO_FBACK", + [TFRC_SSTATE_FBACK] = "FBACK", + [TFRC_SSTATE_TERM] = "TERM", + }; + + return ccid4_state_names[state]; +} +#endif + +static void ccid4_hc_tx_set_state(struct sock *sk, + enum ccid4_hc_tx_states state) +{ + struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + enum ccid4_hc_tx_states oldstate = hctx->ccid4hctx_state; + + ccid4_pr_debug("%s(%p) %-8.8s -> %s\n", + dccp_role(sk), sk, ccid4_tx_state_name(oldstate), + ccid4_tx_state_name(state)); + WARN_ON(state == oldstate); + hctx->ccid4hctx_state = state; +} + +/* + * Compute the initial sending rate X_init in the manner of RFC 3390: + * + * X_init = min(4 * s, max(2 * s, 4380 bytes)) / RTT + * + * Note that RFC 3390 uses MSS, RFC 4342 refers to RFC 3390, and rfc3448bis + * (rev-02) clarifies the use of RFC 3390 with regard to the above formula. + * For consistency with other parts of the code, X_init is scaled by 2^6. + */ +static inline u64 rfc3390_initial_rate(struct sock *sk) +{ + const struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + const __u32 w_init = min_t(__u32, 4 * hctx->ccid4hctx_s, + max_t(__u32, 2 * hctx->ccid4hctx_s, 4380)); + + return scaled_div(w_init << 6, hctx->ccid4hctx_rtt); +} + +/* + * Recalculate t_ipi and delta (should be called whenever X changes) + */ +static inline void ccid4_update_send_interval(struct ccid4_hc_tx_sock *hctx) +{ + /* Calculate new t_ipi = s / X_inst (X_inst is in 64 * bytes/second) */ + hctx->ccid4hctx_t_ipi = scaled_div32(((u64)hctx->ccid4hctx_s) << 6, + hctx->ccid4hctx_x); + + /* Calculate new delta by delta = min(t_ipi / 2, t_gran / 2) */ + hctx->ccid4hctx_delta = min_t(u32, hctx->ccid4hctx_t_ipi / 2, + TFRC_OPSYS_HALF_TIME_GRAN); + + ccid4_pr_debug("t_ipi=%u, delta=%u, s=%u, X=%u\n", + hctx->ccid4hctx_t_ipi, hctx->ccid4hctx_delta, + hctx->ccid4hctx_s, (unsigned)(hctx->ccid4hctx_x >> 6)); + +} + +static u32 ccid4_hc_tx_idle_rtt(struct ccid4_hc_tx_sock *hctx, ktime_t now) +{ + u32 delta = ktime_us_delta(now, hctx->ccid4hctx_t_last_win_count); + + return delta/hctx->ccid4hctx_rtt; +} + +/** + * ccid4_hc_tx_update_x - Update allowed sending rate X + * @stamp: most recent time if available - can be left NULL. + * This function tracks draft rfc3448bis, check there for latest details. + * + * Note: X and X_recv are both stored in units of 64 * bytes/second, to support + * fine-grained resolution of sending rates. This requires scaling by 2^6 + * throughout the code. Only X_calc is unscaled (in bytes/second). + * + */ +static void ccid4_hc_tx_update_x(struct sock *sk, ktime_t *stamp) + +{ + struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + __u64 min_rate = 2 * hctx->ccid4hctx_x_recv; + const __u64 old_x = hctx->ccid4hctx_x; + ktime_t now = stamp? *stamp : ktime_get_real(); + + /* + * Handle IDLE periods: do not reduce below RFC3390 initial sending rate + * when idling [RFC 4342, 5.1]. Definition of idling is from rfc3448bis: + * a sender is idle if it has not sent anything over a 2-RTT-period. + * For consistency with X and X_recv, min_rate is also scaled by 2^6. + */ + if (ccid4_hc_tx_idle_rtt(hctx, now) >= 2) { + min_rate = rfc3390_initial_rate(sk); + min_rate = max(min_rate, 2 * hctx->ccid4hctx_x_recv); + } + + if (hctx->ccid4hctx_p > 0) { + + hctx->ccid4hctx_x = min(((__u64)hctx->ccid4hctx_x_calc) << 6, + min_rate); + hctx->ccid4hctx_x = max(hctx->ccid4hctx_x, + (((__u64)hctx->ccid4hctx_s) << 6) / + TFRC_T_MBI); + + } else if (ktime_us_delta(now, hctx->ccid4hctx_t_ld) + - (s64)hctx->ccid4hctx_rtt >= 0) { + + hctx->ccid4hctx_x = + max(min(2 * hctx->ccid4hctx_x, min_rate), + scaled_div(((__u64)hctx->ccid4hctx_s) << 6, + hctx->ccid4hctx_rtt)); + hctx->ccid4hctx_t_ld = now; + } + + if (hctx->ccid4hctx_x != old_x) { + ccid4_pr_debug("X_prev=%u, X_now=%u, X_calc=%u, " + "X_recv=%u\n", (unsigned)(old_x >> 6), + (unsigned)(hctx->ccid4hctx_x >> 6), + hctx->ccid4hctx_x_calc, + (unsigned)(hctx->ccid4hctx_x_recv >> 6)); + + ccid4_update_send_interval(hctx); + } +} + +/* + * Track the mean packet size `s' (cf. RFC 4342, 5.3 and RFC 3448, 4.1) + * @len: DCCP packet payload size in bytes + */ +static inline void ccid4_hc_tx_update_s(struct ccid4_hc_tx_sock *hctx, int len) +{ + const u16 old_s = hctx->ccid4hctx_s; + + hctx->ccid4hctx_s = tfrc_ewma(hctx->ccid4hctx_s, len, 9); + + if (hctx->ccid4hctx_s != old_s) + ccid4_update_send_interval(hctx); +} + +/* + * Update Window Counter using the algorithm from [RFC 4342, 8.1]. + * The algorithm is not applicable if RTT < 4 microseconds. + */ +static inline void ccid4_hc_tx_update_win_count(struct ccid4_hc_tx_sock *hctx, + ktime_t now) +{ + u32 quarter_rtts; + + if (unlikely(hctx->ccid4hctx_rtt < 4)) /* avoid divide-by-zero */ + return; + + quarter_rtts = ktime_us_delta(now, hctx->ccid4hctx_t_last_win_count); + quarter_rtts /= hctx->ccid4hctx_rtt / 4; + + if (quarter_rtts > 0) { + hctx->ccid4hctx_t_last_win_count = now; + hctx->ccid4hctx_last_win_count += min_t(u32, quarter_rtts, 5); + hctx->ccid4hctx_last_win_count &= 0xF; /* mod 16 */ + } +} + +static void ccid4_hc_tx_no_feedback_timer(unsigned long data) +{ + struct sock *sk = (struct sock *)data; + struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + unsigned long t_nfb = USEC_PER_SEC / 5; + + bh_lock_sock(sk); + if (sock_owned_by_user(sk)) { + /* Try again later. */ + /* XXX: set some sensible MIB */ + goto restart_timer; + } + + ccid4_pr_debug("%s(%p, state=%s) - entry \n", dccp_role(sk), sk, + ccid4_tx_state_name(hctx->ccid4hctx_state)); + + if (hctx->ccid4hctx_state == TFRC_SSTATE_FBACK) + ccid4_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK); + else if (hctx->ccid4hctx_state != TFRC_SSTATE_NO_FBACK) + goto out; + + /* + * Determine new allowed sending rate X as per draft rfc3448bis, 4.4 + */ + if (hctx->ccid4hctx_t_rto == 0 || /* no feedback received yet */ + hctx->ccid4hctx_p == 0 ) { + + /* halve send rate directly */ + hctx->ccid4hctx_x = max(hctx->ccid4hctx_x / 2, + (((__u64)hctx->ccid4hctx_s) << 6) / + TFRC_T_MBI); + ccid4_update_send_interval(hctx); + } else { + /* + * Modify the cached value of X_recv + * + * If (X_calc > 2 * X_recv) + * X_recv = max(X_recv / 2, s / (2 * t_mbi)); + * Else + * X_recv = X_calc / 4; + * + * Note that X_recv is scaled by 2^6 while X_calc is not + */ + BUG_ON(hctx->ccid4hctx_p && !hctx->ccid4hctx_x_calc); + + if (hctx->ccid4hctx_x_calc > (hctx->ccid4hctx_x_recv >> 5)) + hctx->ccid4hctx_x_recv = + max(hctx->ccid4hctx_x_recv / 2, + (((__u64)hctx->ccid4hctx_s) << 6) / + (2 * TFRC_T_MBI)); + else { + hctx->ccid4hctx_x_recv = hctx->ccid4hctx_x_calc; + hctx->ccid4hctx_x_recv <<= 4; + } + ccid4_hc_tx_update_x(sk, NULL); + } + ccid4_pr_debug("Reduced X to %llu/64 bytes/sec\n", + (unsigned long long)hctx->ccid4hctx_x); + + /* + * Set new timeout for the nofeedback timer. + * See comments in packet_recv() regarding the value of t_RTO. + */ + if (unlikely(hctx->ccid4hctx_t_rto == 0)) /* no feedback yet */ + t_nfb = TFRC_INITIAL_TIMEOUT; + else + t_nfb = max(hctx->ccid4hctx_t_rto, 2 * hctx->ccid4hctx_t_ipi); + +restart_timer: + sk_reset_timer(sk, &hctx->ccid4hctx_no_feedback_timer, + jiffies + usecs_to_jiffies(t_nfb)); +out: + bh_unlock_sock(sk); + sock_put(sk); +} + +/* + * returns + * > 0: delay (in msecs) that should pass before actually sending + * = 0: can send immediately + * < 0: error condition; do not send packet + */ +static int ccid4_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb) +{ + struct dccp_sock *dp = dccp_sk(sk); + struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + ktime_t now = ktime_get_real(); + s64 delay; + + /* + * This function is called only for Data and DataAck packets. Sending + * zero-sized Data(Ack)s is theoretically possible, but for congestion + * control this case is pathological - ignore it. + */ + if (unlikely(skb->len == 0)) + return -EBADMSG; + + switch (hctx->ccid4hctx_state) { + case TFRC_SSTATE_NO_SENT: + sk_reset_timer(sk, &hctx->ccid4hctx_no_feedback_timer, + (jiffies + + usecs_to_jiffies(TFRC_INITIAL_TIMEOUT))); + hctx->ccid4hctx_last_win_count = 0; + hctx->ccid4hctx_t_last_win_count = now; + + /* Set t_0 for initial packet */ + hctx->ccid4hctx_t_nom = now; + + hctx->ccid4hctx_s = skb->len; + + /* + * Use initial RTT sample when available: recommended by erratum + * to RFC 4342. This implements the initialisation procedure of + * draft rfc3448bis, section 4.2. Remember, X is scaled by 2^6. + */ + if (dp->dccps_syn_rtt) { + ccid4_pr_debug("SYN RTT = %uus\n", dp->dccps_syn_rtt); + hctx->ccid4hctx_rtt = dp->dccps_syn_rtt; + hctx->ccid4hctx_x = rfc3390_initial_rate(sk); + hctx->ccid4hctx_t_ld = now; + } else { + /* Sender does not have RTT sample: X_pps = 1 pkt/sec */ + hctx->ccid4hctx_x = hctx->ccid4hctx_s; + hctx->ccid4hctx_x <<= 6; + } + ccid4_update_send_interval(hctx); + + ccid4_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK); + break; + case TFRC_SSTATE_NO_FBACK: + case TFRC_SSTATE_FBACK: + delay = ktime_us_delta(hctx->ccid4hctx_t_nom, now); + ccid4_pr_debug("delay=%ld\n", (long)delay); + /* + * Scheduling of packet transmissions [RFC 3448, 4.6] + * + * if (t_now > t_nom - delta) + * // send the packet now + * else + * // send the packet in (t_nom - t_now) milliseconds. + */ + if (delay - (s64)hctx->ccid4hctx_delta >= 1000) + return (u32)delay / 1000L; + + ccid4_hc_tx_update_win_count(hctx, now); + break; + case TFRC_SSTATE_TERM: + DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk); + return -EINVAL; + } + + /* prepare to send now (add options etc.) */ + dp->dccps_hc_tx_insert_options = 1; + DCCP_SKB_CB(skb)->dccpd_ccval = hctx->ccid4hctx_last_win_count; + + /* set the nominal send time for the next following packet */ + hctx->ccid4hctx_t_nom = ktime_add_us(hctx->ccid4hctx_t_nom, + hctx->ccid4hctx_t_ipi); + return 0; +} + +static void ccid4_hc_tx_packet_sent(struct sock *sk, int more, + unsigned int len) +{ + struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + + ccid4_hc_tx_update_s(hctx, len); + + if (tfrc_tx_hist_add(&hctx->ccid4hctx_hist, dccp_sk(sk)->dccps_gss)) + DCCP_CRIT("packet history - out of memory!"); +} + +static void ccid4_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb) +{ + struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + struct ccid4_options_received *opt_recv; + ktime_t t_send, now; + unsigned long t_nfb; + u32 pinv, r_sample; + + /* we are only interested in ACKs */ + if (!(DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_ACK || + DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_DATAACK)) + return; + + /* ... and only in the established state */ + switch (hctx->ccid4hctx_state) { + case TFRC_SSTATE_NO_FBACK: /* fall through */ + case TFRC_SSTATE_FBACK: break; + default: return; + } + + /* estimate RTT from history if ACK number is valid */ + if (! tfrc_tx_hist_when(&t_send, &hctx->ccid4hctx_hist, + DCCP_SKB_CB(skb)->dccpd_ack_seq)) { + DCCP_WARN("%s(%p): %s with bogus ACK-%llu\n", dccp_role(sk), sk, + dccp_packet_name(DCCP_SKB_CB(skb)->dccpd_type), + (unsigned long long)DCCP_SKB_CB(skb)->dccpd_ack_seq); + return; + } + + opt_recv = &hctx->ccid4hctx_options_received; + + /* Update receive rate in units of 64 * bytes/second */ + hctx->ccid4hctx_x_recv = opt_recv->ccid4or_receive_rate; + hctx->ccid4hctx_x_recv <<= 6; + + /* Update loss event rate (scaled by 1e6), cf. RFC 4342, 8.5 */ + pinv = opt_recv->ccid4or_loss_event_rate; + hctx->ccid4hctx_p = (pinv == ~0U || pinv == 0)? 0 : scaled_div(1, pinv); + + /* + * Calculate new RTT sample and update moving average + */ + now = ktime_get_real(); + r_sample = dccp_sample_rtt(sk, ktime_us_delta(now, t_send)); + hctx->ccid4hctx_rtt = tfrc_ewma(hctx->ccid4hctx_rtt, r_sample, 9); + + /* + * Update allowed sending rate X as per draft rfc3448bis, 4.2/4.3 + */ + if (hctx->ccid4hctx_state == TFRC_SSTATE_NO_FBACK) { + + ccid4_hc_tx_set_state(sk, TFRC_SSTATE_FBACK); + + if (hctx->ccid4hctx_t_rto == 0) { + /* + * Initial feedback packet: Larger Initial Windows (4.2) + */ + hctx->ccid4hctx_x = rfc3390_initial_rate(sk); + hctx->ccid4hctx_t_ld = now; + + ccid4_update_send_interval(hctx); + + goto done_computing_x; + + } else if (hctx->ccid4hctx_p == 0) { + /* + * First feedback after nofeedback timer expiry (4.3) + */ + goto done_computing_x; + } + } + /* perform step (4) of draft rfc3448bis, section 4.3 */ + if (hctx->ccid4hctx_p > 0) + hctx->ccid4hctx_x_calc = tfrc_calc_x(hctx->ccid4hctx_s, + hctx->ccid4hctx_rtt, + hctx->ccid4hctx_p); + ccid4_hc_tx_update_x(sk, &now); + +done_computing_x: + ccid4_pr_debug("%s(%p), RTT=%uus (sample=%uus), s=%u, p=%u, X_calc=%u, " + "X_recv=%u, X=%u\n", dccp_role(sk), sk, + hctx->ccid4hctx_rtt, r_sample, hctx->ccid4hctx_s, + hctx->ccid4hctx_p, hctx->ccid4hctx_x_calc, + (unsigned)(hctx->ccid4hctx_x_recv >> 6), + (unsigned)(hctx->ccid4hctx_x >> 6)); + + /* unschedule no feedback timer */ + sk_stop_timer(sk, &hctx->ccid4hctx_no_feedback_timer); + + /* + * As we have calculated new ipi, delta, t_nom it is possible + * that we now can send a packet, so wake up dccp_wait_for_ccid + */ + sk->sk_write_space(sk); + + /* + * Update timeout interval for the nofeedback timer. + * We use a configuration option to increase the lower bound. + * This can help avoid triggering the nofeedback timer too + * often ('spinning') on LANs with small RTTs. + */ + hctx->ccid4hctx_t_rto = max_t(u32, 4 * hctx->ccid4hctx_rtt, + CONFIG_IP_DCCP_CCID4_RTO * + (USEC_PER_SEC/1000)); + /* + * Schedule no feedback timer to expire in + * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi) + */ + t_nfb = max(hctx->ccid4hctx_t_rto, 2 * hctx->ccid4hctx_t_ipi); + + sk_reset_timer(sk, &hctx->ccid4hctx_no_feedback_timer, + jiffies + usecs_to_jiffies(t_nfb)); + + ccid4_pr_debug("Scheduled no feedback timer to expire in %lu jiffies" + "(%luus)\n", usecs_to_jiffies(t_nfb), t_nfb); +} + +static int ccid4_hc_tx_parse_options(struct sock *sk, unsigned char option, + unsigned char len, u16 idx, + unsigned char *value) +{ + int rc = 0; + const struct dccp_sock *dp = dccp_sk(sk); + struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + struct ccid4_options_received *opt_recv; + u32 opt_val; + + opt_recv = &hctx->ccid4hctx_options_received; + + if (opt_recv->ccid4or_seqno != dp->dccps_gsr) { + opt_recv->ccid4or_seqno = dp->dccps_gsr; + opt_recv->ccid4or_loss_event_rate = ~0; + opt_recv->ccid4or_loss_intervals_idx = 0; + opt_recv->ccid4or_loss_intervals_len = 0; + opt_recv->ccid4or_receive_rate = 0; + } + + switch (option) { + case TFRC_OPT_LOSS_EVENT_RATE: + if (unlikely(len != 4)) { + DCCP_WARN("%s(%p), invalid len %d " + "for TFRC_OPT_LOSS_EVENT_RATE\n", + dccp_role(sk), sk, len); + rc = -EINVAL; + } else { + opt_val = get_unaligned((u32 *)value); + opt_recv->ccid4or_loss_event_rate = ntohl(opt_val); + ccid4_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n", + dccp_role(sk), sk, + opt_recv->ccid4or_loss_event_rate); + } + break; + case TFRC_OPT_LOSS_INTERVALS: + opt_recv->ccid4or_loss_intervals_idx = idx; + opt_recv->ccid4or_loss_intervals_len = len; + ccid4_pr_debug("%s(%p), LOSS_INTERVALS=(%u, %u)\n", + dccp_role(sk), sk, + opt_recv->ccid4or_loss_intervals_idx, + opt_recv->ccid4or_loss_intervals_len); + break; + case TFRC_OPT_RECEIVE_RATE: + if (unlikely(len != 4)) { + DCCP_WARN("%s(%p), invalid len %d " + "for TFRC_OPT_RECEIVE_RATE\n", + dccp_role(sk), sk, len); + rc = -EINVAL; + } else { + opt_val = get_unaligned((u32 *)value); + opt_recv->ccid4or_receive_rate = ntohl(opt_val); + ccid4_pr_debug("%s(%p), RECEIVE_RATE=%u\n", + dccp_role(sk), sk, + opt_recv->ccid4or_receive_rate); + } + break; + } + + return rc; +} + +static int ccid4_hc_tx_init(struct ccid *ccid, struct sock *sk) +{ + struct ccid4_hc_tx_sock *hctx = ccid_priv(ccid); + + hctx->ccid4hctx_state = TFRC_SSTATE_NO_SENT; + tfrc_tx_hist_init(&hctx->ccid4hctx_hist, ccid4_tx_hist); + + hctx->ccid4hctx_no_feedback_timer.function = + ccid4_hc_tx_no_feedback_timer; + hctx->ccid4hctx_no_feedback_timer.data = (unsigned long)sk; + init_timer(&hctx->ccid4hctx_no_feedback_timer); + + return 0; +} + +static void ccid4_hc_tx_exit(struct sock *sk) +{ + struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + + ccid4_hc_tx_set_state(sk, TFRC_SSTATE_TERM); + sk_stop_timer(sk, &hctx->ccid4hctx_no_feedback_timer); + + /* Empty packet history */ + tfrc_tx_hist_cleanup(&hctx->ccid4hctx_hist); +} + +static void ccid4_hc_tx_get_info(struct sock *sk, struct tcp_info *info) +{ + const struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + + /* Listen socks doesn't have a private CCID block */ + if (sk->sk_state == DCCP_LISTEN) + return; + + info->tcpi_rto = hctx->ccid4hctx_t_rto; + info->tcpi_rtt = hctx->ccid4hctx_rtt; +} + +static int ccid4_hc_tx_getsockopt(struct sock *sk, const int optname, int len, + u32 __user *optval, int __user *optlen) +{ + const struct ccid4_hc_tx_sock *hctx = ccid4_hc_tx_sk(sk); + const void *val; + + /* Listen socks doesn't have a private CCID block */ + if (sk->sk_state == DCCP_LISTEN) + return -EINVAL; + + switch (optname) { + case DCCP_SOCKOPT_CCID_TX_INFO: + if (len < sizeof(hctx->ccid4hctx_tfrc)) + return -EINVAL; + len = sizeof(hctx->ccid4hctx_tfrc); + val = &hctx->ccid4hctx_tfrc; + break; + default: + return -ENOPROTOOPT; + } + + if (put_user(len, optlen) || copy_to_user(optval, val, len)) + return -EFAULT; + + return 0; +} + +/* + * Receiver Half-Connection Routines + */ +#ifdef CONFIG_IP_DCCP_CCID4_DEBUG +static const char *ccid4_rx_state_name(enum ccid4_hc_rx_states state) +{ + static char *ccid4_rx_state_names[] = { + [TFRC_RSTATE_NO_DATA] = "NO_DATA", + [TFRC_RSTATE_DATA] = "DATA", + [TFRC_RSTATE_TERM] = "TERM", + }; + + return ccid4_rx_state_names[state]; +} +#endif + +static void ccid4_hc_rx_set_state(struct sock *sk, + enum ccid4_hc_rx_states state) +{ + struct ccid4_hc_rx_sock *hcrx = ccid4_hc_rx_sk(sk); + enum ccid4_hc_rx_states oldstate = hcrx->ccid4hcrx_state; + + ccid4_pr_debug("%s(%p) %-8.8s -> %s\n", + dccp_role(sk), sk, ccid4_rx_state_name(oldstate), + ccid4_rx_state_name(state)); + WARN_ON(state == oldstate); + hcrx->ccid4hcrx_state = state; +} + +static inline void ccid4_hc_rx_update_s(struct ccid4_hc_rx_sock *hcrx, int len) +{ + if (likely(len > 0)) /* don't update on empty packets (e.g. ACKs) */ + hcrx->ccid4hcrx_s = tfrc_ewma(hcrx->ccid4hcrx_s, len, 9); +} + +static void ccid4_hc_rx_send_feedback(struct sock *sk, struct sk_buff *skb, + enum ccid4_fback_type fbtype) +{ + struct ccid4_hc_rx_sock *hcrx = ccid4_hc_rx_sk(sk); + struct dccp_sock *dp = dccp_sk(sk); + ktime_t now = ktime_get_real(); + s64 delta = 0; + + if (unlikely(hcrx->ccid4hcrx_state == TFRC_RSTATE_TERM)) + return; + + switch (fbtype) { + case FBACK_INITIAL: + hcrx->ccid4hcrx_x_recv = 0; + hcrx->ccid4hcrx_pinv = ~0U; /* see RFC 4342, 8.5 */ + break; + case FBACK_PARAM_CHANGE: + /* + * When parameters change (new loss or p > p_prev), we do not + * have a reliable estimate for R_m of [RFC 3448, 6.2] and so + * need to reuse the previous value of X_recv. However, when + * X_recv was 0 (due to early loss), this would kill X down to + * s/t_mbi (i.e. one packet in 64 seconds). + * To avoid such drastic reduction, we approximate X_recv as + * the number of bytes since last feedback. + * This is a safe fallback, since X is bounded above by X_calc. + */ + if (hcrx->ccid4hcrx_x_recv > 0) + break; + /* fall through */ + case FBACK_PERIODIC: + delta = ktime_us_delta(now, hcrx->ccid4hcrx_last_feedback); + if (delta <= 0) + DCCP_BUG("delta (%ld) <= 0", (long)delta); + else + hcrx->ccid4hcrx_x_recv = + scaled_div32(hcrx->ccid4hcrx_bytes_recv, delta); + break; + default: + return; + } + ccid4_pr_debug("Interval %ldusec, X_recv=%u, 1/p=%u\n", (long)delta, + hcrx->ccid4hcrx_x_recv, hcrx->ccid4hcrx_pinv); + + hcrx->ccid4hcrx_last_feedback = now; + hcrx->ccid4hcrx_last_counter = dccp_hdr(skb)->dccph_ccval; + hcrx->ccid4hcrx_bytes_recv = 0; + + dp->dccps_hc_rx_insert_options = 1; + dccp_send_ack(sk); +} + +static int ccid4_hc_rx_insert_options(struct sock *sk, struct sk_buff *skb) +{ + const struct ccid4_hc_rx_sock *hcrx = ccid4_hc_rx_sk(sk); + __be32 x_recv, pinv; + + if (!(sk->sk_state == DCCP_OPEN || sk->sk_state == DCCP_PARTOPEN)) + return 0; + + if (dccp_packet_without_ack(skb)) + return 0; + + x_recv = htonl(hcrx->ccid4hcrx_x_recv); + pinv = htonl(hcrx->ccid4hcrx_pinv); + + if (dccp_insert_option(sk, skb, TFRC_OPT_LOSS_EVENT_RATE, + &pinv, sizeof(pinv)) || + dccp_insert_option(sk, skb, TFRC_OPT_RECEIVE_RATE, + &x_recv, sizeof(x_recv))) + return -1; + + return 0; +} + +/** ccid4_first_li - Implements [RFC 3448, 6.3.1] + * + * Determine the length of the first loss interval via inverse lookup. + * Assume that X_recv can be computed by the throughput equation + * s + * X_recv = -------- + * R * fval + * Find some p such that f(p) = fval; return 1/p (scaled). + */ +static u32 ccid4_first_li(struct sock *sk) +{ + struct ccid4_hc_rx_sock *hcrx = ccid4_hc_rx_sk(sk); + u32 x_recv, p, delta; + u64 fval; + + if (hcrx->ccid4hcrx_rtt == 0) { + DCCP_WARN("No RTT estimate available, using fallback RTT\n"); + hcrx->ccid4hcrx_rtt = DCCP_FALLBACK_RTT; + } + + delta = ktime_to_us(net_timedelta(hcrx->ccid4hcrx_last_feedback)); + x_recv = scaled_div32(hcrx->ccid4hcrx_bytes_recv, delta); + if (x_recv == 0) { /* would also trigger divide-by-zero */ + DCCP_WARN("X_recv==0\n"); + if ((x_recv = hcrx->ccid4hcrx_x_recv) == 0) { + DCCP_BUG("stored value of X_recv is zero"); + return ~0U; + } + } + + fval = scaled_div(hcrx->ccid4hcrx_s, hcrx->ccid4hcrx_rtt); + fval = scaled_div32(fval, x_recv); + p = tfrc_calc_x_reverse_lookup(fval); + + ccid4_pr_debug("%s(%p), receive rate=%u bytes/s, implied " + "loss rate=%u\n", dccp_role(sk), sk, x_recv, p); + + return p == 0? ~0U : scaled_div(1, p); +} + +static void ccid4_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb) +{ + struct ccid4_hc_rx_sock *hcrx = ccid4_hc_rx_sk(sk); + enum ccid4_fback_type do_feedback = FBACK_NONE; + u32 sample, ndp = dccp_sk(sk)->dccps_options_received.dccpor_ndp, + payload_size = skb->len - dccp_hdr(skb)->dccph_doff * 4; + u8 is_data_packet = dccp_data_packet(skb); + + spin_lock(&hcrx->ccid4hcrx_hist.lock); + + if (unlikely(hcrx->ccid4hcrx_state == TFRC_RSTATE_NO_DATA)) { + if (is_data_packet) { + do_feedback = FBACK_INITIAL; + ccid4_hc_rx_set_state(sk, TFRC_RSTATE_DATA); + ccid4_hc_rx_update_s(hcrx, payload_size); + } + goto update_records; + } + + if (tfrc_rx_duplicate(&hcrx->ccid4hcrx_hist, skb)) + goto done_receiving; + + if (is_data_packet) { + ccid4_hc_rx_update_s(hcrx, payload_size); + hcrx->ccid4hcrx_bytes_recv += payload_size; + } + + /* + * Handle pending losses and otherwise check for new loss + */ + if (tfrc_rx_loss_pending(&hcrx->ccid4hcrx_hist) && + tfrc_rx_handle_loss(&hcrx->ccid4hcrx_hist, + &hcrx->ccid4hcrx_li_hist, + skb, ndp, ccid4_first_li, sk) ) { + do_feedback = FBACK_PARAM_CHANGE; + goto done_receiving; + } + + if (tfrc_rx_new_loss_indicated(&hcrx->ccid4hcrx_hist, skb, ndp)) + goto update_records; + + /* + * Handle data packets: RTT sampling and monitoring p + */ + if (unlikely(!is_data_packet)) + goto update_records; + + if (! tfrc_lh_is_initialised(&hcrx->ccid4hcrx_li_hist)) { + /* + * Empty loss history: no loss so far, hence p stays 0. + * Sample RTT values, since an RTT estimate is required for the + * computation of p when the first loss occurs; RFC 3448, 6.3.1. + */ + sample = tfrc_rx_sample_rtt(&hcrx->ccid4hcrx_hist, skb); + if (sample != 0) + hcrx->ccid4hcrx_rtt = + tfrc_ewma(hcrx->ccid4hcrx_rtt, sample, 9); + + } else if (tfrc_lh_update_i_mean(&hcrx->ccid4hcrx_li_hist, skb)) + /* + * Step (3) of [RFC 3448, 6.1]: Recompute I_mean and, if I_mean + * has decreased (resp. p has increased), send feedback now. + */ + do_feedback = FBACK_PARAM_CHANGE; + + /* check if the periodic once-per-RTT feedback is due; RFC 4342, 10.3 */ + if (SUB16(dccp_hdr(skb)->dccph_ccval, hcrx->ccid4hcrx_last_counter) > 3) + do_feedback = FBACK_PERIODIC; + +update_records: + tfrc_rx_hist_update(&hcrx->ccid4hcrx_hist, skb, ndp); + +done_receiving: + spin_unlock(&hcrx->ccid4hcrx_hist.lock); + + if (do_feedback) + ccid4_hc_rx_send_feedback(sk, skb, do_feedback); +} + +static int ccid4_hc_rx_init(struct ccid *ccid, struct sock *sk) +{ + struct ccid4_hc_rx_sock *hcrx = ccid_priv(ccid); + + ccid4_pr_debug("entry\n"); + + tfrc_lh_init(&hcrx->ccid4hcrx_li_hist); + hcrx->ccid4hcrx_state = TFRC_RSTATE_NO_DATA; + + return tfrc_rx_hist_init(&hcrx->ccid4hcrx_hist); +} + +static void ccid4_hc_rx_exit(struct sock *sk) +{ + struct ccid4_hc_rx_sock *hcrx = ccid4_hc_rx_sk(sk); + + ccid4_hc_rx_set_state(sk, TFRC_RSTATE_TERM); + + tfrc_rx_hist_cleanup(&hcrx->ccid4hcrx_hist); + tfrc_lh_cleanup(&hcrx->ccid4hcrx_li_hist); +} + +static void ccid4_hc_rx_get_info(struct sock *sk, struct tcp_info *info) +{ + const struct ccid4_hc_rx_sock *hcrx = ccid4_hc_rx_sk(sk); + + /* Listen socks doesn't have a private CCID block */ + if (sk->sk_state == DCCP_LISTEN) + return; + + info->tcpi_ca_state = hcrx->ccid4hcrx_state; + info->tcpi_options |= TCPI_OPT_TIMESTAMPS; + info->tcpi_rcv_rtt = hcrx->ccid4hcrx_rtt; +} + +static int ccid4_hc_rx_getsockopt(struct sock *sk, const int optname, int len, + u32 __user *optval, int __user *optlen) +{ + const struct ccid4_hc_rx_sock *hcrx = ccid4_hc_rx_sk(sk); + const void *val; + struct tfrc_rx_info rx_info; + + /* Listen socks don't have a private CCID block */ + if (sk->sk_state == DCCP_LISTEN) + return -EINVAL; + + switch (optname) { + case DCCP_SOCKOPT_CCID_RX_INFO: + if (len < sizeof(rx_info)) + return -EINVAL; + len = sizeof(rx_info); + val = &rx_info; + rx_info.tfrcrx_x_recv = hcrx->ccid4hcrx_x_recv; + rx_info.tfrcrx_rtt = hcrx->ccid4hcrx_rtt; + rx_info.tfrcrx_p = hcrx->ccid4hcrx_pinv == 0? ~0U + : scaled_div(1, hcrx->ccid4hcrx_pinv); + break; + default: + return -ENOPROTOOPT; + } + + if (put_user(len, optlen) || copy_to_user(optval, val, len)) + return -EFAULT; + + return 0; +} + +static struct ccid_operations ccid4 = { + .ccid_id = DCCPC_CCID4, + .ccid_name = "ccid4", + .ccid_owner = THIS_MODULE, + .ccid_hc_tx_obj_size = sizeof(struct ccid4_hc_tx_sock), + .ccid_hc_tx_init = ccid4_hc_tx_init, + .ccid_hc_tx_exit = ccid4_hc_tx_exit, + .ccid_hc_tx_send_packet = ccid4_hc_tx_send_packet, + .ccid_hc_tx_packet_sent = ccid4_hc_tx_packet_sent, + .ccid_hc_tx_packet_recv = ccid4_hc_tx_packet_recv, + .ccid_hc_tx_parse_options = ccid4_hc_tx_parse_options, + .ccid_hc_rx_obj_size = sizeof(struct ccid4_hc_rx_sock), + .ccid_hc_rx_init = ccid4_hc_rx_init, + .ccid_hc_rx_exit = ccid4_hc_rx_exit, + .ccid_hc_rx_insert_options = ccid4_hc_rx_insert_options, + .ccid_hc_rx_packet_recv = ccid4_hc_rx_packet_recv, + .ccid_hc_rx_get_info = ccid4_hc_rx_get_info, + .ccid_hc_tx_get_info = ccid4_hc_tx_get_info, + .ccid_hc_rx_getsockopt = ccid4_hc_rx_getsockopt, + .ccid_hc_tx_getsockopt = ccid4_hc_tx_getsockopt, +}; + +#ifdef CONFIG_IP_DCCP_CCID4_DEBUG +module_param(ccid4_debug, bool, 0444); +MODULE_PARM_DESC(ccid4_debug, "Enable debug messages"); +#endif + +static __init int ccid4_module_init(void) +{ + int rc = -ENOBUFS; + + if (tfrc_tx_cache_init(&ccid4_tx_hist, "ccid4")) + goto out; + + rc = ccid_register(&ccid4); + if (rc != 0) + tfrc_tx_cache_cleanup(ccid4_tx_hist); +out: + return rc; +} +module_init(ccid4_module_init); + +static __exit void ccid4_module_exit(void) +{ + ccid_unregister(&ccid4); + + if (ccid4_tx_hist != NULL) + tfrc_tx_cache_cleanup(ccid4_tx_hist); +} +module_exit(ccid4_module_exit); + +MODULE_AUTHOR("Ian McDonald <ian.mcdonald@xxxxxxxxxxx>, " + "Arnaldo Carvalho de Melo <acme@xxxxxxxxxxxxxxxxxx>"); +MODULE_DESCRIPTION("DCCP TFRC CCID4 CCID"); +MODULE_LICENSE("GPL"); +MODULE_ALIAS("net-dccp-ccid-4"); diff --git a/net/dccp/ccids/ccid4.h b/net/dccp/ccids/ccid4.h new file mode 100644 index 0000000..d233f6d --- /dev/null +++ b/net/dccp/ccids/ccid4.h @@ -0,0 +1,177 @@ +#ifndef _DCCP_CCID4_H_ +#define _DCCP_CCID4_H_ +/* + * net/dccp/ccids/ccid4.h + * + * Copyright (c) 2007 The University of Aberdeen, Scotland, UK + * Copyright (c) 2005-7 The University of Waikato, Hamilton, New Zealand. + * + * An implementation of the DCCP protocol + * + * This code has been developed by the University of Waikato WAND + * research group. For further information please see http://www.wand.net.nz/ + * or e-mail Ian McDonald - ian.mcdonald@xxxxxxxxxxx + * + * This code also uses code from Lulea University, rereleased as GPL by its + * authors: + * Copyright (c) 2003 Nils-Erik Mattsson, Joacim Haggmark, Magnus Erixzon + * + * Changes to meet Linux coding standards, to make it meet latest ccid4 draft + * and to make it work as a loadable module in the DCCP stack written by + * Arnaldo Carvalho de Melo <acme@xxxxxxxxxxxxxxxx>. + * + * Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@xxxxxxxxxxxxxxxx> + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ +#include <linux/ktime.h> +#include <linux/list.h> +#include <linux/types.h> +#include <linux/tfrc.h> +#include <asm/unaligned.h> +#include "lib/tfrc.h" +#include "../ccid.h" + +/* Two seconds as per RFC 3448 4.2 */ +#define TFRC_INITIAL_TIMEOUT (2 * USEC_PER_SEC) + +/* In usecs - half the scheduling granularity as per RFC3448 4.6 */ +#define TFRC_OPSYS_HALF_TIME_GRAN (USEC_PER_SEC / (2 * HZ)) + +/* Parameter t_mbi from [RFC 3448, 4.3]: backoff interval in seconds */ +#define TFRC_T_MBI 64 + +enum ccid4_options { + TFRC_OPT_LOSS_EVENT_RATE = 192, + TFRC_OPT_LOSS_INTERVALS = 193, + TFRC_OPT_RECEIVE_RATE = 194, +}; + +struct ccid4_options_received { + u64 ccid4or_seqno:48, + ccid4or_loss_intervals_idx:16; + u16 ccid4or_loss_intervals_len; + u32 ccid4or_loss_event_rate; + u32 ccid4or_receive_rate; +}; + +/* TFRC sender states */ +enum ccid4_hc_tx_states { + TFRC_SSTATE_NO_SENT = 1, + TFRC_SSTATE_NO_FBACK, + TFRC_SSTATE_FBACK, + TFRC_SSTATE_TERM, +}; + +/** struct ccid4_hc_tx_sock - CCID4 sender half-connection socket + * + * @ccid4hctx_x - Current sending rate in 64 * bytes per second + * @ccid4hctx_x_recv - Receive rate in 64 * bytes per second + * @ccid4hctx_x_calc - Calculated rate in bytes per second + * @ccid4hctx_rtt - Estimate of current round trip time in usecs + * @ccid4hctx_p - Current loss event rate (0-1) scaled by 1000000 + * @ccid4hctx_s - Packet size in bytes + * @ccid4hctx_t_rto - Nofeedback Timer setting in usecs + * @ccid4hctx_t_ipi - Interpacket (send) interval (RFC 3448, 4.6) in usecs + * @ccid4hctx_state - Sender state, one of %ccid4_hc_tx_states + * @ccid4hctx_last_win_count - Last window counter sent + * @ccid4hctx_t_last_win_count - Timestamp of earliest packet + * with last_win_count value sent + * @ccid4hctx_no_feedback_timer - Handle to no feedback timer + * @ccid4hctx_t_ld - Time last doubled during slow start + * @ccid4hctx_t_nom - Nominal send time of next packet + * @ccid4hctx_delta - Send timer delta (RFC 3448, 4.6) in usecs + * @ccid4hctx_hist - Packet history + * @ccid4hctx_options_received - Parsed set of retrieved options + */ +struct ccid4_hc_tx_sock { + struct tfrc_tx_info ccid4hctx_tfrc; +#define ccid4hctx_x ccid4hctx_tfrc.tfrctx_x +#define ccid4hctx_x_recv ccid4hctx_tfrc.tfrctx_x_recv +#define ccid4hctx_x_calc ccid4hctx_tfrc.tfrctx_x_calc +#define ccid4hctx_rtt ccid4hctx_tfrc.tfrctx_rtt +#define ccid4hctx_p ccid4hctx_tfrc.tfrctx_p +#define ccid4hctx_t_rto ccid4hctx_tfrc.tfrctx_rto +#define ccid4hctx_t_ipi ccid4hctx_tfrc.tfrctx_ipi + u16 ccid4hctx_s; + enum ccid4_hc_tx_states ccid4hctx_state:8; + u8 ccid4hctx_last_win_count; + ktime_t ccid4hctx_t_last_win_count; + struct timer_list ccid4hctx_no_feedback_timer; + ktime_t ccid4hctx_t_ld; + ktime_t ccid4hctx_t_nom; + u32 ccid4hctx_delta; + struct tfrc_tx_hist_head ccid4hctx_hist; + struct ccid4_options_received ccid4hctx_options_received; +}; + +static inline struct ccid4_hc_tx_sock *ccid4_hc_tx_sk(const struct sock *sk) +{ + void *ccid4_tx_priv = ccid_priv(dccp_sk(sk)->dccps_hc_tx_ccid); + + BUG_ON(ccid4_tx_priv == NULL); + return ccid4_tx_priv; +} + +/* CCID4 receiver states */ +enum ccid4_hc_rx_states { + TFRC_RSTATE_NO_DATA = 1, + TFRC_RSTATE_DATA, + TFRC_RSTATE_TERM = 127, +}; + +/* CCID4 feedback types */ +enum ccid4_fback_type { + FBACK_NONE = 0, + FBACK_INITIAL, + FBACK_PERIODIC, + FBACK_PARAM_CHANGE +}; + +/** struct ccid4_hc_rx_sock - CCID4 receiver half-connection socket + * + * @ccid4hcrx_last_counter - Tracks window counter (RFC 4342, 8.1) + * @ccid4hcrx_state - Receiver state, one of %ccid4_hc_rx_states + * @ccid4hcrx_bytes_recv - Total sum of DCCP payload bytes + * @ccid4hcrx_x_recv - Receiver estimate of send rate (RFC 3448, sec. 4.3) + * @ccid4hcrx_rtt - Receiver estimate of RTT + * @ccid4hcrx_last_feedback - Time at which last feedback was sent + * @ccid4hcrx_hist - Packet history, exported by TFRC module + * @ccid4hcrx_li_hist - Loss Interval database, exported by TFRC module + * @ccid4hcrx_s - Received packet size in bytes + * @ccid4hcrx_pinv - Inverse of Loss Event Rate (RFC 4342, sec. 8.5) + */ +struct ccid4_hc_rx_sock { + u8 ccid4hcrx_last_counter:4; + enum ccid4_hc_rx_states ccid4hcrx_state:8; + u32 ccid4hcrx_bytes_recv; + u32 ccid4hcrx_x_recv; + u32 ccid4hcrx_rtt; + ktime_t ccid4hcrx_last_feedback; + struct tfrc_rx_hist ccid4hcrx_hist; + struct tfrc_loss_hist ccid4hcrx_li_hist; + u16 ccid4hcrx_s; +#define ccid4hcrx_pinv ccid4hcrx_li_hist.i_mean +}; + +static inline struct ccid4_hc_rx_sock *ccid4_hc_rx_sk(const struct sock *sk) +{ + void *ccid4_rx_priv = ccid_priv(dccp_sk(sk)->dccps_hc_rx_ccid); + + BUG_ON(ccid4_rx_priv == NULL); + return ccid4_rx_priv; +} + +#endif /* _DCCP_CCID4_H_ */ - To unsubscribe from this list: send the line "unsubscribe dccp" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
