tcp cubic代码分析

/*
 * TCP CUBIC: Binary Increase Congestion control for TCP v2.3
 * Home page:
 *      http://netsrv.csc.ncsu.edu/twiki/bin/view/Main/BIC
 * This is from the implementation of CUBIC TCP in
 * Sangtae Ha, Injong Rhee and Lisong Xu,
 *  "CUBIC: A New TCP-Friendly High-Speed TCP Variant"
 *  in ACM SIGOPS Operating System Review, July 2008.
 * Available from:
 *  http://netsrv.csc.ncsu.edu/export/cubic_a_new_tcp_2008.pdf
 *
 * CUBIC integrates a new slow start algorithm, called HyStart.
 * The details of HyStart are presented in
 *  Sangtae Ha and Injong Rhee,
 *  "Taming the Elephants: New TCP Slow Start", NCSU TechReport 2008.
 * Available from:
 *  http://netsrv.csc.ncsu.edu/export/hystart_techreport_2008.pdf
 *
 * All testing results are available from:
 * http://netsrv.csc.ncsu.edu/wiki/index.php/TCP_Testing
 *
 * Unless CUBIC is enabled and congestion window is large
 * this behaves the same as the original Reno.
 */

#include <linux/mm.h>
#include <linux/module.h>
#include <linux/math64.h>
#include <net/tcp.h>

#define BICTCP_BETA_SCALE    1024    /* Scale factor beta calculation
                     * max_cwnd = snd_cwnd * beta
                     */
#define    BICTCP_HZ        10    /* BIC HZ 2^10 = 1024 */

/* Two methods of hybrid slow start */
//Both run independently at the same time and slow start exits when any of them detects an exit point.
//1. ACK train length
//2. Delay increase

#define HYSTART_ACK_TRAIN    0x1
#define HYSTART_DELAY        0x2
/* 注意：这里的delay_min没有放大8倍！
 * 此宏用来计算Delay increase threshold
 * delay_min <= 32ms，则threshold = 2ms
 * 32ms < delay_min < 256ms，则threshold = delay_min / 16 ms
 * delay_min >= 256ms，则threshold = 16ms
 */
/* Number of delay samples for detecting the increase of delay */
#define HYSTART_MIN_SAMPLES    8
#define HYSTART_DELAY_MIN    (2U<<3)
#define HYSTART_DELAY_MAX    (16U<<3)
#define HYSTART_DELAY_THRESH(x)    clamp(x, HYSTART_DELAY_MIN, HYSTART_DELAY_MAX)

static int fast_convergence __read_mostly = 1;
static int beta __read_mostly = 717;    /* = 717/1024 (BICTCP_BETA_SCALE) */
//beta在BIC中为819，而CUBIC中为717，
//会导致在bictcp_recalc_ssthresh中，并且启用了fast convergence，
//cubic: last_max_cwnd = 0.85*snd_cwnd ，而慢启动阈值=0.7*snd_cwnd 。
//bic:   last_max_cwnd = 0.95*snd_cwnd ，而慢启动阈值=0.8*snd_cwnd 。
//这样会导致更早的到达平衡值，对snd_cwnd有很大的影响。



static int initial_ssthresh __read_mostly;
static int bic_scale __read_mostly = 41;
static int tcp_friendliness __read_mostly = 1;



//hybrid slow start的开关
static int hystart __read_mostly = 1;
//HyStart状态描述
//1：packet-train  2: delay   3:both packet-train and delay
//默认2种方法都使用，故设为3
static int hystart_detect __read_mostly = HYSTART_ACK_TRAIN | HYSTART_DELAY;
//设置snd_ssthresh的最小拥塞窗口值,除非cwnd超过了这个值，才能使用HyStart 
static int hystart_low_window __read_mostly = 16;

static u32 cube_rtt_scale __read_mostly;
static u32 beta_scale __read_mostly;
static u64 cube_factor __read_mostly;

/* Note parameters that are used for precomputing scale factors are read-only */
module_param(fast_convergence, int, 0644);
MODULE_PARM_DESC(fast_convergence, "turn on/off fast convergence");
module_param(beta, int, 0644);
MODULE_PARM_DESC(beta, "beta for multiplicative increase");
module_param(initial_ssthresh, int, 0644);
MODULE_PARM_DESC(initial_ssthresh, "initial value of slow start threshold");
module_param(bic_scale, int, 0444);
MODULE_PARM_DESC(bic_scale, "scale (scaled by 1024) value for bic function (bic_scale/1024)");
module_param(tcp_friendliness, int, 0644);
MODULE_PARM_DESC(tcp_friendliness, "turn on/off tcp friendliness");
module_param(hystart, int, 0644);
MODULE_PARM_DESC(hystart, "turn on/off hybrid slow start algorithm");
module_param(hystart_detect, int, 0644);
MODULE_PARM_DESC(hystart_detect, "hyrbrid slow start detection mechanisms"
         " 1: packet-train 2: delay 3: both packet-train and delay");
module_param(hystart_low_window, int, 0644);
MODULE_PARM_DESC(hystart_low_window, "lower bound cwnd for hybrid slow start");

/* BIC TCP Parameters */
struct bictcp {
    u32    cnt;        /*用来控制snd_cwnd的增长 increase cwnd by 1 after ACKs */
//两个重要的count值:
//第一个是tcp_sock->snd_cwnd_cnt，表示在当前的拥塞窗口中已经
    //发送(经过对方ack包确认)的数据段的个数，
//而第二个是bictcp->cnt，它是cubic拥塞算法的核心，
//主要用来控制在拥塞避免状态的时候，什么时候才能增大拥塞窗口，
//具体实现是通过比较cnt和snd_cwnd_cnt，来决定是否增大拥塞窗口，
    
    u32 last_max_cwnd;    /*上一次的最大拥塞窗口值 last maximum snd_cwnd */
    u32    loss_cwnd;    /* 拥塞状态切换时的拥塞窗口值congestion window at last loss */
    u32    last_cwnd;    /* 上一次的拥塞窗口值 the last snd_cwnd */
    u32    last_time;    /* time when updated last_cwnd */
    u32    bic_origin_point;/*即新的Wmax饱和点，取Wlast_max_cwnd和snd_cwnd较大者 origin point of bic function */
    u32    bic_K;        /*即新Wmax所对应的时间点t，W(bic_K) = Wmax    time to origin point from the beginning of the current epoch */
    u32    delay_min;    /*应该是最小RTT    min delay */
    u32    epoch_start;    /*拥塞状态切换开始的时刻  beginning of an epoch */
    u32    ack_cnt;    /*在一个epoch中的ack包的数量   number of acks */
    u32    tcp_cwnd;    /*按照Reno算法计算得的cwnd    estimated tcp cwnd */
#define ACK_RATIO_SHIFT    4
    u16    delayed_ack;    /* estimate the ratio of Packets/ACKs << 4 */
    u8    sample_cnt;    /*第几个sample    number of samples to decide curr_rtt */
    u8    found;        /* the exit point is found? */
    u32    round_start;    /*针对每个RTT     beginning of each round */
    u32    end_seq;    /*用来标识每个RTT    end_seq of the round */
    u32    last_jiffies;    /*超过2ms则不认为是连续的   last time when the ACK spacing is close */
    u32    curr_rtt;    /*由sampe中最小的决定    the minimum rtt of current round */
};

static inline void bictcp_reset(struct bictcp *ca)
{//论文说Time out时调用
    ca->cnt = 0;
    ca->last_max_cwnd = 0;
    ca->loss_cwnd = 0;
    ca->last_cwnd = 0;
    ca->last_time = 0;
    ca->bic_origin_point = 0;
    ca->bic_K = 0;
    ca->delay_min = 0;
    ca->epoch_start = 0;
    ca->delayed_ack = 2 << ACK_RATIO_SHIFT;
    ca->ack_cnt = 0;
    ca->tcp_cwnd = 0;
    ca->found = 0;
}

static inline void bictcp_hystart_reset(struct sock *sk)
{
    struct tcp_sock *tp = tcp_sk(sk);
    struct bictcp *ca = inet_csk_ca(sk);

    ca->round_start = ca->last_jiffies = jiffies;//记录时间戳
    ca->end_seq = tp->snd_nxt;//记录待发送的下一个序列号
    ca->curr_rtt = 0;
    ca->sample_cnt = 0;

    //bictcp_hystart_reset中并没有对ca->found置0。
    //也就是说，只有在初始化时、LOSS状态时、开启hystart的慢启动时。
    //HyStart才会派上用场，其它时间并不使用.
}

static void bictcp_init(struct sock *sk)
{
    bictcp_reset(inet_csk_ca(sk));

    if (hystart)//如果指定hystart
        bictcp_hystart_reset(sk);

    if (!hystart && initial_ssthresh)
        tcp_sk(sk)->snd_ssthresh = initial_ssthresh;
}

/* calculate the cubic root of x using a table lookup followed by one
 * Newton-Raphson iteration.
 * Avg err ~= 0.195%
 */
static u32 cubic_root(u64 a) //用来计算立方根
{
    u32 x, b, shift;
    /*
     * cbrt(x) MSB values for x MSB values in [0..63].
     * Precomputed then refined by hand - Willy Tarreau
     *
     * For x in [0..63],
     *   v = cbrt(x << 18) - 1
     *   cbrt(x) = (v[x] + 10) >> 6
     */
    static const u8 v[] = {
        /* 0x00 */    0,   54,   54,   54,  118,  118,  118,  118,
        /* 0x08 */  123,  129,  134,  138,  143,  147,  151,  156,
        /* 0x10 */  157,  161,  164,  168,  170,  173,  176,  179,
        /* 0x18 */  181,  185,  187,  190,  192,  194,  197,  199,
        /* 0x20 */  200,  202,  204,  206,  209,  211,  213,  215,
        /* 0x28 */  217,  219,  221,  222,  224,  225,  227,  229,
        /* 0x30 */  231,  232,  234,  236,  237,  239,  240,  242,
        /* 0x38 */  244,  245,  246,  248,  250,  251,  252,  254,
    };

    b = fls64(a);
    if (b < 7) {
        /* a in [0..63] */
        return ((u32)v[(u32)a] + 35) >> 6;
    }

    b = ((b * 84) >> 8) - 1;
    shift = (a >> (b * 3));

    x = ((u32)(((u32)v[shift] + 10) << b)) >> 6;

    /*
     * Newton-Raphson iteration
     *                         2
     * x    = ( 2 * x  +  a / x  ) / 3
     *  k+1          k         k
     */
    x = (2 * x + (u32)div64_u64(a, (u64)x * (u64)(x - 1)));
    x = ((x * 341) >> 10);
    return x;
}

/*
 * Compute congestion window to use.
 */  //从快速恢复退出并进入拥塞避免状态之后，更新cnt
static inline void bictcp_update(struct bictcp *ca, u32 cwnd)
{
    u64 offs;//时间差|t - K|
    //delta是cwnd差，bic_target是预测值，t为预测时间
    u32 delta, t, bic_target, max_cnt;

    ca->ack_cnt++;    /*ack包计数器加1   count the number of ACKs */

    if (ca->last_cwnd == cwnd && //当前窗口与历史窗口相同
        (s32)(tcp_time_stamp - ca->last_time) <= HZ / 32)//时间差小于1000/32ms
        return; //直接结束

    ca->last_cwnd = cwnd;//记录进入拥塞避免时的窗口值
    ca->last_time = tcp_time_stamp;//记录进入拥塞避免时的时刻

    if (ca->epoch_start == 0) {//丢包后，开启一个新的时段
        ca->epoch_start = tcp_time_stamp;    /*新时段的开始 record the beginning of an epoch */
        ca->ack_cnt = 1;            /*ack包计数器初始化  start counting */
        ca->tcp_cwnd = cwnd;            /*同步更新 syn with cubic */

        //取max(last_max_cwnd , cwnd)作为当前Wmax饱和点
        if (ca->last_max_cwnd <= cwnd) {
            ca->bic_K = 0;
            ca->bic_origin_point = cwnd;
        } else {
            /* Compute new K based on
             * (wmax-cwnd) * (srtt>>3 / HZ) / c * 2^(3*bictcp_HZ)
             */
            ca->bic_K = cubic_root(cube_factor
                           * (ca->last_max_cwnd - cwnd));
            ca->bic_origin_point = ca->last_max_cwnd;
        }
    }

    /* cubic function - calc*/
    /* calculate c * time^3 / rtt,
     *  while considering overflow in calculation of time^3
     * (so time^3 is done by using 64 bit)
     * and without the support of division of 64bit numbers
     * (so all divisions are done by using 32 bit)
     *  also NOTE the unit of those veriables
     *      time  = (t - K) / 2^bictcp_HZ
     *      c = bic_scale >> 10 == 0.04
     * rtt  = (srtt >> 3) / HZ
     * !!! The following code does not have overflow problems,
     * if the cwnd < 1 million packets !!!
     */

    /* change the unit from HZ to bictcp_HZ */
    t = ((tcp_time_stamp + (ca->delay_min>>3) - ca->epoch_start)
         << BICTCP_HZ) / HZ;

     //求| t - bic_K | 
    if (t < ca->bic_K)        // 还未达到Wmax 
        offs = ca->bic_K - t;
    else
        offs = t - ca->bic_K;//已经超过Wmax 

    /* c/rtt * (t-K)^3 */     //计算立方，delta =| W(t) - W(bic_K) | 
    delta = (cube_rtt_scale * offs * offs * offs) >> (10+3*BICTCP_HZ);


     
     //t为预测时间，bic_K为新Wmax所对应的时间，
     //bic_target为cwnd预测值，bic_origin_point为当前Wmax饱和点
    if (t < ca->bic_K)                                    /* below origin*/
        bic_target = ca->bic_origin_point - delta;
    else                                                    /* above origin*/
        bic_target = ca->bic_origin_point + delta;

    /* cubic function - calc bictcp_cnt*/
    if (bic_target > cwnd) {// 相差越多，增长越快，这就是函数形状由来
        ca->cnt = cwnd / (bic_target - cwnd);//
    } else {//目前cwnd已经超出预期了，应该降速
        ca->cnt = 100 * cwnd;              /* very small increment*/
    }



    /* TCP Friendly —如果bic比RENO慢，则提升cwnd增长速度，即减小cnt 
     * 以上次丢包以后的时间t算起，每次RTT增长 3B / ( 2 - B)，那么可以得到 
      * 采用RENO算法的cwnd。 
      * cwnd (RENO) = cwnd + 3B / (2 - B) * ack_cnt / cwnd 
     * B为乘性减少因子，在此算法中为0.3 
     */
    if (tcp_friendliness) {
        u32 scale = beta_scale;
        delta = (cwnd * scale) >> 3; //delta代表多少ACK可使tcp_cwnd++
        while (ca->ack_cnt > delta) {        /* update tcp cwnd */
            ca->ack_cnt -= delta;
            ca->tcp_cwnd++;
        }

        if (ca->tcp_cwnd > cwnd){    /* if bic is slower than tcp */
            delta = ca->tcp_cwnd - cwnd;
            max_cnt = cwnd / delta;
            if (ca->cnt > max_cnt)
                ca->cnt = max_cnt;
        }
    }

    ca->cnt = (ca->cnt << ACK_RATIO_SHIFT) / ca->delayed_ack;
    if (ca->cnt == 0)            /* cannot be zero */
        ca->cnt = 1; //此时代表cwnd远小于bic_target，增长速度最大
}

static void bictcp_cong_avoid(struct sock *sk, u32 ack, u32 in_flight)
{
    struct tcp_sock *tp = tcp_sk(sk);
    struct bictcp *ca = inet_csk_ca(sk);

    //判断发送拥塞窗口是否到达限制，如果到达限制则直接返回。
    if (!tcp_is_cwnd_limited(sk, in_flight))
        return;

    if (tp->snd_cwnd <= tp->snd_ssthresh) {
        //当snd_cwnd<=ssthresh的时候，进入慢启动状态
        if (hystart && after(ack, ca->end_seq))//是否需要reset对应的bictcp的值
            bictcp_hystart_reset(sk);
        tcp_slow_start(tp);//进入slow start状态
    } else {
        //当snd_cwnd>ssthresh的时候，进入拥塞避免状态
        bictcp_update(ca, tp->snd_cwnd);//首先会更新bictcp->cnt
        tcp_cong_avoid_ai(tp, ca->cnt);//然后进入拥塞避免，更新tcp_sock->snd_cwnd_cnt
    }

}


//每次发生拥塞状态切换时，就会重新计算慢启动阈值
//做了两件事：重赋值last_max_cwnd、返回新的慢启动阈值
static u32 bictcp_recalc_ssthresh(struct sock *sk)
{//论文说这个函数在Packet loss时调用
    const struct tcp_sock *tp = tcp_sk(sk);
    struct bictcp *ca = inet_csk_ca(sk);

    ca->epoch_start = 0;    /* 发生拥塞状态切换，标志一个epoch结束   end of epoch */

    /* Wmax and fast convergence */
    //当一个新的TCP流加入到网络，
    //网络中已有TCP流需要放弃自己带宽，
    //给新的TCP流提供一定的上升空间。
    //为提高已有TCP流所释放的带宽而引入快速收敛机制。
    if (tp->snd_cwnd < ca->last_max_cwnd && fast_convergence)  
        //snd_cwnd<last_max_cwnd
        //表示已有TCP流所经历的饱和点因为可用带宽改变而正在降低。
        //然后，通过进一步降低Wmax让已有流释放更多带宽。
        //这种行为有效地延长已有流增大其窗口的时间，
        //因为降低后的Wmax强制已有流更早进入平稳状态。
        //这允许新流有更多的时间来赶上其窗口尺寸。
        ca->last_max_cwnd = (tp->snd_cwnd * (BICTCP_BETA_SCALE + beta))
            / (2 * BICTCP_BETA_SCALE); //last_max_cwnd = 0.9 * snd_cwnd 
    else
        ca->last_max_cwnd = tp->snd_cwnd;

    ca->loss_cwnd = tp->snd_cwnd;

    //修改snd_ssthresh，即max(0.7*snd_cwnd，2)
    return max((tp->snd_cwnd * beta) / BICTCP_BETA_SCALE, 2U);
    
}

static u32 bictcp_undo_cwnd(struct sock *sk)
{
    struct bictcp *ca = inet_csk_ca(sk);

    return max(tcp_sk(sk)->snd_cwnd, ca->last_max_cwnd);
}

static void bictcp_state(struct sock *sk, u8 new_state)
{
    if (new_state == TCP_CA_Loss) {//如果处于LOSS状态，丢包处理
        bictcp_reset(inet_csk_ca(sk));
        bictcp_hystart_reset(sk);
    }
}

static void hystart_update(struct sock *sk, u32 delay)
{//会修改snd_ssthresh
    struct tcp_sock *tp = tcp_sk(sk);
    struct bictcp *ca = inet_csk_ca(sk);

    if (!(ca->found & hystart_detect)) {
        u32 curr_jiffies = jiffies;

        /* first detection parameter - ack-train detection */
        if (curr_jiffies - ca->last_jiffies <= msecs_to_jiffies(2)) {
            ca->last_jiffies = curr_jiffies;
            if (curr_jiffies - ca->round_start >= ca->delay_min>>4)
                ca->found |= HYSTART_ACK_TRAIN;
        }

        /* obtain the minimum delay of more than sampling packets */
        if (ca->sample_cnt < HYSTART_MIN_SAMPLES) {
            if (ca->curr_rtt == 0 || ca->curr_rtt > delay)
                ca->curr_rtt = delay;

            ca->sample_cnt++;
        } else {
            if (ca->curr_rtt > ca->delay_min +
                HYSTART_DELAY_THRESH(ca->delay_min>>4))
                ca->found |= HYSTART_DELAY;
        }
        /*
         * Either one of two conditions are met,
         * we exit from slow start immediately.
         */
        if (ca->found & hystart_detect)//found是一个是否退出slow start的标记
            tp->snd_ssthresh = tp->snd_cwnd;//修改snd_ssthresh
    }
}

/* Track delayed acknowledgment ratio using sliding window
 * ratio = (15*ratio + sample) / 16
 */  //基本每次收到ack都会调用这个函数，更新snd_ssthresh和delayed_ack
static void bictcp_acked(struct sock *sk, u32 cnt, s32 rtt_us)
{//论文说这个函数在On each ACK时调用
    const struct inet_connection_sock *icsk = inet_csk(sk);
    const struct tcp_sock *tp = tcp_sk(sk);
    struct bictcp *ca = inet_csk_ca(sk);
    u32 delay;

    if (icsk->icsk_ca_state == TCP_CA_Open) {
        cnt -= ca->delayed_ack >> ACK_RATIO_SHIFT;
        ca->delayed_ack += cnt;
    }

    /* Some calls are for duplicates without timetamps */
    if (rtt_us < 0)
        return;

    /* Discard delay samples right after fast recovery */
    if ((s32)(tcp_time_stamp - ca->epoch_start) < HZ)
        return;

    delay = usecs_to_jiffies(rtt_us) << 3;
    if (delay == 0)
        delay = 1;

    /* first time call or link delay decreases */
    if (ca->delay_min == 0 || ca->delay_min > delay)
        ca->delay_min = delay;

    /* hystart triggers when cwnd is larger than some threshold */
    //tp->snd_ssthresh初始值是一个很大的值0x7fffffff
    
    //当拥塞窗口增大到16的时候，
    //调用hystart_update来修改更新snd_ssthresh
    //hystart_update主要用于是否退出slow start
    if (hystart && tp->snd_cwnd <= tp->snd_ssthresh &&
        tp->snd_cwnd >= hystart_low_window)
        hystart_update(sk, delay);
}

static struct tcp_congestion_ops cubictcp = {
    
    .init        = bictcp_init,

    
    //调用ssthresh函数的地方有：tcp_fastretrans_alert(), tcp_enter_cwr(),tcp_enter_frto(), tcp_enter_loss()　
    //看起来每次发生拥塞状态切换的时候，都会调整ssthresh。　　　
　　//修改snd_ssthresh值的地方有bictcp_init,hystart_update以及上面列出的调用ssthresh函数处。
    .ssthresh    = bictcp_recalc_ssthresh,

    //发送方发出一个data包之后，接收方回复一个ack包，发送方收到这个ack包之后，
　　//调用tcp_ack()->tcp_cong_avoid()->bictcp_cong_avoid()来更改拥塞窗口snd_cwnd大小.
    .cong_avoid    = bictcp_cong_avoid,
    
    .set_state    = bictcp_state,

    //调用undo_cwnd函数的地方有:tcp_undo_cwr()用来撤销之前误判导致的"缩小拥塞窗口"
    .undo_cwnd    = bictcp_undo_cwnd,
    
    //调用ptts_acked函数的路径为：tcp_ack() -->tcp_clean_rtx_queue()
    .pkts_acked     = bictcp_acked,
    
    .owner        = THIS_MODULE,
    .name        = "cubic",
};

static int __init cubictcp_register(void)
{
     //bictcp参数的个数不能过多
    BUILD_BUG_ON(sizeof(struct bictcp) > ICSK_CA_PRIV_SIZE);

    /* Precompute a bunch of the scaling factors that are used per-packet
     * based on SRTT of 100ms
     */
     //beta_scale == 8*(1024 + 717) / 3 / (1024 -717 )，大约为15 
    beta_scale = 8*(BICTCP_BETA_SCALE+beta)/ 3 / (BICTCP_BETA_SCALE - beta);

    //cube_rtt_scale == 41*10 = 410 
    cube_rtt_scale = (bic_scale * 10);    /* 1024*c/rtt */

    /* calculate the "K" for (wmax-cwnd) = c/rtt * K^3
     *  so K = cubic_root( (wmax-cwnd)*rtt/c )
     * the unit of K is bictcp_HZ=2^10, not HZ
     *
     *  c = bic_scale >> 10
     *  rtt = 100ms
     *
     * the following code has been designed and tested for
     * cwnd < 1 million packets
     * RTT < 100 seconds
     * HZ < 1,000,00  (corresponding to 10 nano-second)
     */

    /* 1/c * 2^2*bictcp_HZ * srtt */
    cube_factor = 1ull << (10+3*BICTCP_HZ); /* cube_factor == 2^40 */

    /* divide by bic_scale and by constant Srtt (100ms) */
    do_div(cube_factor, bic_scale * 10);//cube_factor == 2^40 / 410

    return tcp_register_congestion_control(&cubictcp);
}

static void __exit cubictcp_unregister(void)
{
    tcp_unregister_congestion_control(&cubictcp);
}

module_init(cubictcp_register);
module_exit(cubictcp_unregister);

MODULE_AUTHOR("Sangtae Ha, Stephen Hemminger");
MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("CUBIC TCP");
MODULE_VERSION("2.3");