level-ip

画个图好吧 !
将 linux kernel 中间对应的部分写出来好吧

env setup

ref : https://github.com/saminiir/level-ip/blob/master/Documentation/getting-started.md

sudo mknod /dev/net/tap c 10 200
sudo chmod 0666 /dev/net/tap


# 似乎这才是正确使用 tap 的 : https://stackoverflow.com/questions/15626088/tap-interfaces-and-dev-net-tun-device-using-ip-tuntap-command
# sudo ip tuntap add mode tap tap0
# sudo ip link set dev tap0 up

sudo sysctl -w net.ipv4.ip_forward=1
sudo iptables -I INPUT --source 10.0.0.0/24 -j ACCEPT
sudo iptables -t nat -I POSTROUTING --out-interface wlp2s0  -j MASQUERADE
sudo iptables -I FORWARD --in-interface wlp2s0 --out-interface tap0 -j ACCEPT
sudo iptables -I FORWARD --in-interface tap0 --out-interface wlp2s0 -j ACCEPT

# 其他的配置
sudo iptables -t nat -I POSTROUTING --out-interface  enxd43a650739d8 -j MASQUERADE
sudo iptables -I FORWARD --in-interface enxd43a650739d8 --out-interface tap0 -j ACCEPT
sudo iptables -I FORWARD --in-interface tap0 --out-interface enxd43a650739d8 -j ACCEPT

sudo iptables -I FORWARD --in-interface enxd43a650739d8 --out-interface tap -j ACCEPT
sudo iptables -I FORWARD --in-interface tap --out-interface enxd43a650739d8 -j ACCEPT

replace wlp2s0 with enxd43a650739d, because this is part of output of ifconfig:

enxd43a650739d8: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 1500
        inet 10.170.133.9  netmask 255.255.0.0  broadcast 10.170.255.255
        ether d4:3a:65:07:39:d8  txqueuelen 1000  (Ethernet)
        RX packets 5951876  bytes 7731714225 (7.7 GB)
        RX errors 0  dropped 0  overruns 0  frame 0
        TX packets 3694072  bytes 318623167 (318.6 MB)
        TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0

remove -Werror in ./Makefile

由于不支持 ipv6，也许有用的内容: sudo sysctl -w net.ipv6.conf.all.disable_ipv6=1 sudo sysctl -w net.ipv6.conf.default.disable_ipv6=1 sudo sysctl -w net.ipv6.conf.lo.disable_ipv6=1

注意点 : tools/liblevelip.c 不能包含 /usr/include 的头文件，否则由于符号使用不同类型的重复定义而出错。

Unknown IP header proto 17
Unknown IP header proto 17
Unknown IP header proto 17
Unknown IP header proto 17

使用方法:

在一个终端 : $ ./lvl-ip
在另一个终端 :

配置问题：到底是 tap 还是 tap0

notes

到底如何处理 ofo_queue 的？

struct sk_buff_head tcp_sock::ofo_queue
TCP 的变量，一个窗口就是两个数值就对了啊，实际上，

tcp_data_queue 中间， int expected = skb->seq == tcb->rcv_nxt;

static void tcp_init_segment(struct tcphdr *th, struct iphdr *ih, struct sk_buff *skb)
{
    th->sport = ntohs(th->sport);
    th->dport = ntohs(th->dport);
    th->seq = ntohl(th->seq);
    th->ack_seq = ntohl(th->ack_seq);
    th->win = ntohs(th->win);
    th->csum = ntohs(th->csum);
    th->urp = ntohs(th->urp);

    skb->seq = th->seq; // TODO seq 数值就是代码的长度 ? 感觉哪里有点不对，应该用工具查看一下
    skb->dlen = ip_len(ih) - tcp_hlen(th);
    skb->len = skb->dlen + th->syn + th->fin;
    skb->end_seq = skb->seq + skb->dlen;
    skb->payload = th->data;
}

static struct net_family *families[128] = {
    [AF_INET] = &inet,
};

struct net_family inet = {
    .create = inet_create,
};

static struct sock_type inet_ops[] = {
    {
        .sock_ops = &inet_stream_ops,
        .net_ops = &tcp_ops,
        .type = SOCK_STREAM,
        .protocol = IPPROTO_TCP,
    }
};

static struct sock_ops inet_stream_ops = {
    .connect = &inet_stream_connect,
    .write = &inet_write,
    .read = &inet_read,
    .close = &inet_close,
    .free = &inet_free,
    .abort = &inet_abort,
    .getpeername = &inet_getpeername,
    .getsockname = &inet_getsockname,
};

Why we need unix socket to handle it ?
_read : this is almost the
- inet_read
  - tcp_read
    - tcp_receive
      - tcp_data_dequeue

linux kernel

其他的用户态网络栈

mtcp : 这个是个工业级的

本站所有文章转发 CSDN 将按侵权追究法律责任，其它情况随意。