old_tcpack.py

#!/usr/bin/env python

from __future__ import print_function
from bcc import BPF
import argparse
from time import strftime
from socket import inet_ntop, AF_INET, AF_INET6
from struct import pack
import ctypes as ct
from time import sleep
from bcc import tcp
import argparse
from subprocess import call
from os import kill, getpid, path
from signal import SIGKILL
import sys
from tcptools import check_filename, valid_function_name

# arguments
examples = """examples:
    ./tcpack             # trace all ACKs
    ./tcpack -p  5205    # only trace port 5205
    ./tcpack -dp 5205    # only trace remote port 5205
    ./tcpack -sp 5205    # only trace local port 5205
    ./tcpack -s 5        # only trace one ACK in every 2^5 ACKs
    ./tcpack -o  [fname] # print the information into /usr/local/bcc/fname
"""

parser = argparse.ArgumentParser(
    description="Trace the TCP metrics with ACKs",
    formatter_class=argparse.RawDescriptionHelpFormatter,
    epilog=examples)
parser.add_argument("-p", "--port", 
    help="TCP port")
parser.add_argument("-sp", "--sport", 
    help="TCP source port")
parser.add_argument("-dp", "--dport",
    help="TCP destination port")
parser.add_argument("-s", "--sample",
    help="Trace sampling")
parser.add_argument("-o", "--output", nargs='?', const="tcpack",
    help="Output file")

args = parser.parse_args()

# define BPF program
bpf_text = """
#include <uapi/linux/ptrace.h>
#define KBUILD_MODNAME "foo"
#include <linux/tcp.h>
#include <uapi/linux/tcp.h>
#include <uapi/linux/ip.h>
#include <net/sock.h>
#include <net/inet_connection_sock.h>
#include <bcc/proto.h>


struct ipv4_flow_info {
    u64 init_time;
    u16 mss;
    u32 fastRe;
    u32 timeout;
    u32 last_cwnd;
    u32 max_bytes_inflight;
};
BPF_HASH(flows_info, struct sock *, struct ipv4_flow_info);


// only consider data structs for ipv4
struct ipv4_data_t {
    u32 pid;
    u32 saddr;
    u32 daddr;
    u16 sport;
    u16 dport;
    u32 seq;
    u32 ack;
    u8 state;
    u8 tcpflags;
    u32 snd_cwnd;
    u32 rcv_wnd;
    u32	total_retrans;
    u32 fastRe;
    u32 timeout;
    u64 bytes_acked;
    u64 bytes_received;
    u32 srtt;
    u64 srtt_sum;
    u32 srtt_counter;
    // flight size
    u32 packets_out;
    u64 duration;
    u32 bytes_inflight;
};
BPF_PERF_OUTPUT(ipv4_events);

static struct tcphdr *skb_to_tcphdr(const struct sk_buff *skb)
{
    // unstable API. verify logic in tcp_hdr() -> skb_transport_header().
    return (struct tcphdr *)(skb->head + skb->transport_header);
}

static inline struct iphdr *skb_to_iphdr(const struct sk_buff *skb)
{
    // unstable API. verify logic in ip_hdr() -> skb_network_header().
    return (struct iphdr *)(skb->head + skb->network_header);
}


int trace_tcp_ack(struct pt_regs *ctx, struct sock *sk, struct sk_buff *skb)
{
    if (sk == NULL)
        return 0;

    // pull in details from the packet headers and the sock struct
    u16 family = sk->__sk_common.skc_family;
    if (family != AF_INET)
        return 0;

    u32 pid = bpf_get_current_pid_tgid();
    char state = sk->__sk_common.skc_state;
    u32 ack = 0, seq = 0, snd_cwnd = 0;
    u16 sport = 0, dport = 0;
    struct tcphdr *tcp = skb_to_tcphdr(skb);
    struct iphdr *ip = skb_to_iphdr(skb);
    struct tcp_sock *tp = (struct tcp_sock *)sk;
    u8 tcpflags = ((u_int8_t *)tcp)[13];
    sport = tcp->source;
    dport = tcp->dest;
    sport = ntohs(sport);
    dport = ntohs(dport);
    seq = tcp->seq;
    seq = ntohl(seq);
    ack = tcp->ack_seq;
    ack = ntohl(ack);

    SAMPLING
    FILTER_PORT
    FILTER_DPORT
    FILTER_SPORT

    struct ipv4_flow_info *finfo, zero = {};
    finfo = flows_info.lookup_or_init(&sk, &zero);
    struct ipv4_data_t data4 = {};
    u32 srtt = 0;
    data4.pid = pid;
    data4.saddr = ip->saddr;
    data4.daddr = ip->daddr;
    data4.dport = dport;
    data4.sport = sport;
    data4.seq = seq;
    data4.ack = ack;
    data4.state = state;
    data4.tcpflags = tcpflags;
    data4.snd_cwnd = tp->snd_cwnd;
    data4.rcv_wnd = tp->rcv_wnd;
    data4.bytes_acked = tp->bytes_acked;
    data4.bytes_received = tp->bytes_received;
    data4.total_retrans = tp->total_retrans;
    data4.fastRe = finfo->fastRe;
    data4.timeout = finfo->timeout;
    data4.srtt = tp->srtt_us;
    data4.srtt_counter += 1;
    data4.srtt_sum += tp->srtt_us;
    data4.packets_out = tp->packets_out;
    data4.duration = bpf_ktime_get_ns() - finfo->init_time;
    data4.bytes_inflight = tp->snd_nxt - tp->snd_una; 

    ipv4_events.perf_submit(ctx, &data4, sizeof(data4));

    return 0;
}


int trace_tcp_set_state(struct pt_regs *ctx, struct sock *sk, int state)
{
    if (state == TCP_ESTABLISHED) {
        u64 ts = bpf_ktime_get_ns();
        struct tcp_sock *tp = (struct tcp_sock *)sk;
        struct ipv4_flow_info *finfo, zero = {};
        finfo = flows_info.lookup_or_init(&sk, &zero);
        finfo->init_time = ts;
        finfo->mss == tp->advmss;
    } else if (state == TCP_CLOSE) {
        flows_info.delete(&sk);
    }

    return 0;
} 


int trace_tcp_enter_recovery(struct pt_regs *ctx, struct sock *sk)
{
    struct ipv4_flow_info *finfo, zero = {};
    finfo = flows_info.lookup_or_init(&sk, &zero);
    finfo->fastRe += 1;

    return 0;
}


int trace_tcp_enter_loss(struct pt_regs *ctx, struct sock *sk)
{
    struct ipv4_flow_info *finfo, zero = {};
    finfo = flows_info.lookup_or_init(&sk, &zero);
    finfo->timeout += 1;

    return 0;
}
"""

# code substitutions
if args.port:
    bpf_text = bpf_text.replace('FILTER_PORT',
        'if (sport != %s && dport != %s) { return 0; }' % (args.port, args.port))
else:
    bpf_text = bpf_text.replace('FILTER_PORT', '')
if args.sport:
    bpf_text = bpf_text.replace('FILTER_SPORT',
        'if (sport != %s) { return 0; }' % args.sport)
else:
    bpf_text = bpf_text.replace('FILTER_SPORT', '')
    
if args.dport:
    bpf_text = bpf_text.replace('FILTER_DPORT',
        'if (dport != %s) { return 0; }' % args.dport)
else:
    bpf_text = bpf_text.replace('FILTER_DPORT', '')
if args.sample:
    bpf_text = bpf_text.replace('SAMPLING',
        'if (((seq+ack) << (32-%s) >> (32-%s)) != ((0x01 << %s) - 1)) { return 0;}' % (args.sample, args.sample, args.sample))
else:
    bpf_text = bpf_text.replace('SAMPLING', '')
if args.output:
    if check_filename(args.output):
        output_dir = "/usr/local/bcc/"
        if not path.isdir(output_dir):
            call(["mkdir", "-p", output_dir])
        output_file = output_dir + args.output
        sys.stdout = open(output_file, "w+", buffering=1)
    else:
        print("The output filename is invalid. Exit...")
        exit()

# event data
class Data_ipv4(ct.Structure):
    _fields_ = [
        ("pid", ct.c_uint),
        ("saddr", ct.c_uint),
        ("daddr", ct.c_uint),
        ("sport", ct.c_ushort),
        ("dport", ct.c_ushort),
        ("seq", ct.c_uint),
        ("ack", ct.c_uint),
        ("state", ct.c_ubyte),
        ("tcpflags", ct.c_ubyte),
        ("snd_cwnd", ct.c_uint),
        ("rcv_wnd", ct.c_uint),
        ("total_retrans", ct.c_uint),
        ("fastRe", ct.c_uint),
        ("timeout", ct.c_uint),
        ("bytes_acked", ct.c_ulonglong),
        ("bytes_received", ct.c_ulonglong),
        ("srtt", ct.c_uint),
        ("srtt_sum", ct.c_ulonglong),
        ("srtt_counter", ct.c_uint),
        ("packets_out", ct.c_uint),
        ("duration", ct.c_ulonglong),
        ("bytes_inflight", ct.c_uint),
    ]

# process event
def print_ipv4_event(cpu, data, size):
    event = ct.cast(data, ct.POINTER(Data_ipv4)).contents
    print("3 %-20s -> %-20s %-10s %-10s %-8s %-8s %-12s (%s)" % (
        "%s:%d" % (inet_ntop(AF_INET, pack('I', event.saddr)), event.sport),
        "%s:%d" % (inet_ntop(AF_INET, pack('I', event.daddr)), event.dport),
        "%d" % (event.seq),
        "%d" % (event.ack),
        "%d" % (event.srtt >> 3),
        "%d" % (event.snd_cwnd),
        tcp.tcpstate[event.state], tcp.flags2str(event.tcpflags)))

# initialize BPF
b = BPF(text=bpf_text)
trace_prefix = "trace_"
functions_list = ["tcp_ack", "tcp_set_state", "tcp_enter_recovery", "tcp_enter_loss"]
for i in range(len(functions_list)):
    function = functions_list[i]
    trace_function = trace_prefix + function
    function = valid_function_name(function)
    if function == None:
        exit()
    if b.get_kprobe_functions(function):
        b.attach_kprobe(event=function, fn_name=trace_function)
    else:
        print("ERROR: %s() kernel function not found or traceable." % (function))
        exit()

# header
if not args.output:
    print("%-20s -> %-20s %-10s %-10s %-8s %-8s %-12s (%s)" % ("SADDR:SPORT", "DADDR:DPORT", "SEQ", "ACK", "RTT(us)", "CWnd", "STATE", "FLAGS"))

# read events
b["ipv4_events"].open_perf_buffer(print_ipv4_event)
while 1:
    try:
        b.perf_buffer_poll()
    except KeyboardInterrupt:
        kill(getpid(), SIGKILL)