bindsnoop.py

#!/usr/bin/python
#
# bindsnoop       Trace IPv4 and IPv6 binds()s.
#               For Linux, uses BCC, eBPF. Embedded C.
#
# based on tcpconnect utility from Brendan Gregg's suite.
#
# USAGE: bindsnoop [-h] [-t] [-E] [-p PID] [-P PORT[,PORT ...]] [-w]
#             [--count] [--cgroupmap mappath] [--mntnsmap mappath]
#
# bindsnoop reports socket options set before the bind call
# that would impact this system call behavior:
# SOL_IP     IP_FREEBIND              F....
# SOL_IP     IP_TRANSPARENT           .T...
# SOL_IP     IP_BIND_ADDRESS_NO_PORT  ..N..
# SOL_SOCKET SO_REUSEADDR             ...R.
# SOL_SOCKET SO_REUSEPORT             ....r
#
# SO_BINDTODEVICE interface is reported as "BOUND_IF" index
#
# This uses dynamic tracing of kernel functions, and will need to be updated
# to match kernel changes.
#
# Copyright (c) 2020-present Facebook.
# Licensed under the Apache License, Version 2.0 (the "License")
#
# 14-Feb-2020   Pavel Dubovitsky   Created this.

from __future__ import print_function, absolute_import, unicode_literals
from bcc import BPF, DEBUG_SOURCE
from bcc.containers import filter_by_containers
from bcc.utils import printb
import argparse
import re
from os import strerror
from socket import (
    inet_ntop, AF_INET, AF_INET6, __all__ as socket_all, __dict__ as socket_dct
)
from struct import pack
from time import sleep

# arguments
examples = """examples:
    ./bindsnoop           # trace all TCP bind()s
    ./bindsnoop -t        # include timestamps
    ./tcplife -w        # wider columns (fit IPv6)
    ./bindsnoop -p 181    # only trace PID 181
    ./bindsnoop -P 80     # only trace port 80
    ./bindsnoop -P 80,81  # only trace port 80 and 81
    ./bindsnoop -U        # include UID
    ./bindsnoop -u 1000   # only trace UID 1000
    ./bindsnoop -E        # report bind errors
    ./bindsnoop --count   # count bind per src ip
    ./bindsnoop --cgroupmap mappath  # only trace cgroups in this BPF map
    ./bindsnoop --mntnsmap  mappath  # only trace mount namespaces in the map

it is reporting socket options set before the bins call
impacting system call behavior:
 SOL_IP     IP_FREEBIND              F....
 SOL_IP     IP_TRANSPARENT           .T...
 SOL_IP     IP_BIND_ADDRESS_NO_PORT  ..N..
 SOL_SOCKET SO_REUSEADDR             ...R.
 SOL_SOCKET SO_REUSEPORT             ....r

 SO_BINDTODEVICE interface is reported as "IF" index
"""
parser = argparse.ArgumentParser(
    description="Trace TCP binds",
    formatter_class=argparse.RawDescriptionHelpFormatter,
    epilog=examples)
parser.add_argument("-t", "--timestamp", action="store_true",
    help="include timestamp on output")
parser.add_argument("-w", "--wide", action="store_true",
    help="wide column output (fits IPv6 addresses)")
parser.add_argument("-p", "--pid",
    help="trace this PID only")
parser.add_argument("-P", "--port",
    help="comma-separated list of ports to trace.")
parser.add_argument("-E", "--errors", action="store_true",
    help="include errors in the output.")
parser.add_argument("-U", "--print-uid", action="store_true",
    help="include UID on output")
parser.add_argument("-u", "--uid",
    help="trace this UID only")
parser.add_argument("--count", action="store_true",
    help="count binds per src ip and port")
parser.add_argument("--cgroupmap",
    help="trace cgroups in this BPF map only")
parser.add_argument("--mntnsmap",
    help="trace mount namespaces in this BPF map only")
parser.add_argument("--ebpf", action="store_true",
    help=argparse.SUPPRESS)
parser.add_argument("--debug-source", action="store_true",
    help=argparse.SUPPRESS)
args = parser.parse_args()

# define BPF program
bpf_text = """
#include <uapi/linux/ptrace.h>
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wtautological-compare"
#include <net/sock.h>
#pragma clang diagnostic pop
#include <net/inet_sock.h>
#include <net/net_namespace.h>
#include <bcc/proto.h>

BPF_HASH(currsock, u32, struct socket *);

// separate data structs for ipv4 and ipv6
struct ipv4_bind_data_t {
    u64 ts_us;
    u32 pid;
    u32 uid;
    u64 ip;
    u32 saddr;
    u32 bound_dev_if;
    int return_code;
    u16 sport;
    u8 socket_options;
    u8 protocol;
    char task[TASK_COMM_LEN];
};
BPF_PERF_OUTPUT(ipv4_bind_events);

struct ipv6_bind_data_t {
    // int128 would be aligned on 16 bytes boundary, better to go first
    unsigned __int128 saddr;
    u64 ts_us;
    u32 pid;
    u32 uid;
    u64 ip;
    u32 bound_dev_if;
    int return_code;
    u16 sport;
    u8 socket_options;
    u8 protocol;
    char task[TASK_COMM_LEN];
};
BPF_PERF_OUTPUT(ipv6_bind_events);

// separate flow keys per address family
struct ipv4_flow_key_t {
    u32 saddr;
    u16 sport;
};
BPF_HASH(ipv4_count, struct ipv4_flow_key_t);

struct ipv6_flow_key_t {
    unsigned __int128 saddr;
    u16 sport;
};
BPF_HASH(ipv6_count, struct ipv6_flow_key_t);

// bind options for event reporting
union bind_options {
    u8 data;
    struct {
        u8 freebind:1;
        u8 transparent:1;
        u8 bind_address_no_port:1;
        u8 reuseaddress:1;
        u8 reuseport:1;
    } fields;
};

// TODO: add reporting for the original bind arguments
int bindsnoop_entry(struct pt_regs *ctx, struct socket *socket)
{
    u64 pid_tgid = bpf_get_current_pid_tgid();
    u32 pid = pid_tgid >> 32;
    u32 tid = pid_tgid;
    FILTER_PID

    u32 uid = bpf_get_current_uid_gid();

    FILTER_UID

    if (container_should_be_filtered()) {
        return 0;
    }

    // stash the sock ptr for lookup on return
    currsock.update(&tid, &socket);

    return 0;
};


static int bindsnoop_return(struct pt_regs *ctx, short ipver)
{
    int ret = PT_REGS_RC(ctx);
    u64 pid_tgid = bpf_get_current_pid_tgid();
    u32 pid = pid_tgid >> 32;
    u32 tid = pid_tgid;

    struct socket **skpp;
    skpp = currsock.lookup(&tid);
    if (skpp == 0) {
        return 0;   // missed entry
    }

    int ignore_errors = 1;
    FILTER_ERRORS
    if (ret != 0 && ignore_errors) {
        // failed to bind
        currsock.delete(&tid);
        return 0;
    }

    // pull in details
    struct socket *skp_ = *skpp;
    struct sock *skp = skp_->sk;

    struct inet_sock *sockp = (struct inet_sock *)skp;

    u16 sport = 0;
    bpf_probe_read_kernel(&sport, sizeof(sport), &sockp->inet_sport);
    sport = ntohs(sport);

    FILTER_PORT

    union bind_options opts = {0};
    u8 bitfield;
    // fetching freebind, transparent, and bind_address_no_port bitfields
    // via the next struct member, rcv_tos
    bitfield = (u8) *(&sockp->rcv_tos - 2) & 0xFF;
    // IP_FREEBIND (sockp->freebind)
    opts.fields.freebind = bitfield >> 2 & 0x01;
    // IP_TRANSPARENT (sockp->transparent)
    opts.fields.transparent = bitfield >> 5 & 0x01;
    // IP_BIND_ADDRESS_NO_PORT (sockp->bind_address_no_port)
    opts.fields.bind_address_no_port = *(&sockp->rcv_tos - 1) & 0x01;

    // SO_REUSEADDR and SO_REUSEPORT are bitfields that
    // cannot be accessed directly, fetched via the next struct member,
    // __sk_common.skc_bound_dev_if
    bitfield = *((u8*)&skp->__sk_common.skc_bound_dev_if - 1);
    // SO_REUSEADDR (skp->reuse)
    // it is 4 bit, but we are interested in the lowest one
    opts.fields.reuseaddress = bitfield & 0x0F;
    // SO_REUSEPORT (skp->reuseport)
    opts.fields.reuseport = bitfield >> 4 & 0x01;

    // workaround for reading the sk_protocol bitfield (from tcpaccept.py):
    u8 protocol;
    int gso_max_segs_offset = offsetof(struct sock, sk_gso_max_segs);
    int sk_lingertime_offset = offsetof(struct sock, sk_lingertime);
    if (sk_lingertime_offset - gso_max_segs_offset == 4)
        // 4.10+ with little endian
#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
        protocol = *(u8 *)((u64)&skp->sk_gso_max_segs - 3);
    else
        // pre-4.10 with little endian
        protocol = *(u8 *)((u64)&skp->sk_wmem_queued - 3);
#elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
        // 4.10+ with big endian
        protocol = *(u8 *)((u64)&skp->sk_gso_max_segs - 1);
    else
        // pre-4.10 with big endian
        protocol = *(u8 *)((u64)&skp->sk_wmem_queued - 1);
#else
# error "Fix your compiler's __BYTE_ORDER__?!"
#endif

    if (ipver == 4) {
        IPV4_CODE
    } else /* 6 */ {
        IPV6_CODE
    }

    currsock.delete(&tid);

    return 0;
}

int bindsnoop_v4_return(struct pt_regs *ctx)
{
    return bindsnoop_return(ctx, 4);
}

int bindsnoop_v6_return(struct pt_regs *ctx)
{
    return bindsnoop_return(ctx, 6);
}
"""

struct_init = {
    'ipv4': {
        'count': """
               struct ipv4_flow_key_t flow_key = {};
               flow_key.saddr = skp->__sk_common.skc_rcv_saddr;
               flow_key.sport = sport;
               ipv4_count.increment(flow_key);""",
        'trace': """
               struct ipv4_bind_data_t data4 = {.pid = pid, .ip = ipver};
               data4.uid = bpf_get_current_uid_gid();
               data4.ts_us = bpf_ktime_get_ns() / 1000;
               bpf_probe_read_kernel(
                 &data4.saddr, sizeof(data4.saddr), &sockp->inet_saddr);
               data4.return_code = ret;
               data4.sport = sport;
               data4.bound_dev_if = skp->__sk_common.skc_bound_dev_if;
               data4.socket_options = opts.data;
               data4.protocol = protocol;
               bpf_get_current_comm(&data4.task, sizeof(data4.task));
               ipv4_bind_events.perf_submit(ctx, &data4, sizeof(data4));"""
    },
    'ipv6': {
        'count': """
               struct ipv6_flow_key_t flow_key = {};
               bpf_probe_read_kernel(&flow_key.saddr, sizeof(flow_key.saddr),
                   skp->__sk_common.skc_v6_rcv_saddr.in6_u.u6_addr32);
               flow_key.sport = sport;
               ipv6_count.increment(flow_key);""",
        'trace': """
               struct ipv6_bind_data_t data6 = {.pid = pid, .ip = ipver};
               data6.uid = bpf_get_current_uid_gid();
               data6.ts_us = bpf_ktime_get_ns() / 1000;
               bpf_probe_read_kernel(&data6.saddr, sizeof(data6.saddr),
                   skp->__sk_common.skc_v6_rcv_saddr.in6_u.u6_addr32);
               data6.return_code = ret;
               data6.sport = sport;
               data6.bound_dev_if = skp->__sk_common.skc_bound_dev_if;
               data6.socket_options = opts.data;
               data6.protocol = protocol;
               bpf_get_current_comm(&data6.task, sizeof(data6.task));
               ipv6_bind_events.perf_submit(ctx, &data6, sizeof(data6));"""
    },
}

# code substitutions
if args.count:
    bpf_text = bpf_text.replace("IPV4_CODE", struct_init['ipv4']['count'])
    bpf_text = bpf_text.replace("IPV6_CODE", struct_init['ipv6']['count'])
else:
    bpf_text = bpf_text.replace("IPV4_CODE", struct_init['ipv4']['trace'])
    bpf_text = bpf_text.replace("IPV6_CODE", struct_init['ipv6']['trace'])

if args.pid:
    bpf_text = bpf_text.replace('FILTER_PID',
        'if (pid != %s) { return 0; }' % args.pid)
if args.port:
    sports = [int(sport) for sport in args.port.split(',')]
    sports_if = ' && '.join(['sport != %d' % sport for sport in sports])
    bpf_text = bpf_text.replace('FILTER_PORT',
        'if (%s) { currsock.delete(&tid); return 0; }' % sports_if)
if args.uid:
    bpf_text = bpf_text.replace('FILTER_UID',
        'if (uid != %s) { return 0; }' % args.uid)
if args.errors:
    bpf_text = bpf_text.replace('FILTER_ERRORS', 'ignore_errors = 0;')
bpf_text = filter_by_containers(args) + bpf_text
bpf_text = bpf_text.replace('FILTER_PID', '')
bpf_text = bpf_text.replace('FILTER_PORT', '')
bpf_text = bpf_text.replace('FILTER_UID', '')
bpf_text = bpf_text.replace('FILTER_ERRORS', '')

# selecting output format - 80 characters or wide, fitting IPv6 addresses
header_fmt = "%8s %-12.12s %-4s %-15s %-5s %5s %2s"
output_fmt = b"%8d %-12.12s %-4.4s %-15.15s %5d %-5s %2d"
error_header_fmt = "%3s "
error_output_fmt = b"%3s "
error_value_fmt = str
if args.wide:
    header_fmt = "%10s %-12.12s %-4s %-39s %-5s %5s %2s"
    output_fmt = b"%10d %-12.12s %-4s %-39s %5d %-5s %2d"
    error_header_fmt = "%-25s "
    error_output_fmt = b"%-25s "
    error_value_fmt = strerror

if args.ebpf:
    print(bpf_text)
    exit()

# L4 protocol resolver
class L4Proto:
    def __init__(self):
        self.num2str = {}
        proto_re = re.compile("IPPROTO_(.*)")
        for attr in socket_all:
            proto_match = proto_re.match(attr)
            if proto_match:
                self.num2str[socket_dct[attr]] = proto_match.group(1)

    def proto2str(self, proto):
        return self.num2str.get(proto, "UNKNOWN")

l4 = L4Proto()

# bind options:
# SOL_IP     IP_FREEBIND              F....
# SOL_IP     IP_TRANSPARENT           .T...
# SOL_IP     IP_BIND_ADDRESS_NO_PORT  ..N..
# SOL_SOCKET SO_REUSEADDR             ...R.
# SOL_SOCKET SO_REUSEPORT             ....r
def opts2str(bitfield):
    str_options = ""
    bit = 1
    for opt in "FTNRr":
        str_options += opt if bitfield & bit else "."
        bit *= 2
    return str_options.encode()


# process events
def print_ipv4_bind_event(cpu, data, size):
    event = b["ipv4_bind_events"].event(data)
    global start_ts
    if args.timestamp:
        if start_ts == 0:
            start_ts = event.ts_us
        printb(b"%-9.6f " % ((float(event.ts_us) - start_ts) / 1000000), nl="")
    if args.print_uid:
        printb(b"%6d " % event.uid, nl="")
    if args.errors:
        printb(
            error_output_fmt % error_value_fmt(event.return_code).encode(),
            nl="",
        )
    printb(output_fmt % (event.pid, event.task,
        l4.proto2str(event.protocol).encode(),
        inet_ntop(AF_INET, pack("I", event.saddr)).encode(),
        event.sport, opts2str(event.socket_options), event.bound_dev_if))


def print_ipv6_bind_event(cpu, data, size):
    event = b["ipv6_bind_events"].event(data)
    global start_ts
    if args.timestamp:
        if start_ts == 0:
            start_ts = event.ts_us
        printb(b"%-9.6f " % ((float(event.ts_us) - start_ts) / 1000000), nl="")
    if args.print_uid:
        printb(b"%6d " % event.uid, nl="")
    if args.errors:
        printb(
            error_output_fmt % error_value_fmt(event.return_code).encode(),
            nl="",
        )
    printb(output_fmt % (event.pid, event.task,
        l4.proto2str(event.protocol).encode(),
        inet_ntop(AF_INET6, event.saddr).encode(),
        event.sport, opts2str(event.socket_options), event.bound_dev_if))


def depict_cnt(counts_tab, l3prot='ipv4'):
    for k, v in sorted(
        counts_tab.items(), key=lambda counts: counts[1].value, reverse=True
    ):
        depict_key = ""
        if l3prot == 'ipv4':
            depict_key = "%-32s %20s" % (
                (inet_ntop(AF_INET, pack('I', k.saddr))), k.sport
            )
        else:
            depict_key = "%-32s %20s" % (
                (inet_ntop(AF_INET6, k.saddr)), k.sport
            )
        print("%s     %-10d" % (depict_key, v.value))


# initialize BPF
b = BPF(text=bpf_text)
b.attach_kprobe(event="inet_bind", fn_name="bindsnoop_entry")
b.attach_kprobe(event="inet6_bind", fn_name="bindsnoop_entry")
b.attach_kretprobe(event="inet_bind", fn_name="bindsnoop_v4_return")
b.attach_kretprobe(event="inet6_bind", fn_name="bindsnoop_v6_return")

print("Tracing binds ... Hit Ctrl-C to end")
if args.count:
    try:
        while 1:
            sleep(99999999)
    except KeyboardInterrupt:
        pass

    # header
    print("\n%-32s %20s     %-10s" % (
        "LADDR", "LPORT", "BINDS"))
    depict_cnt(b["ipv4_count"])
    depict_cnt(b["ipv6_count"], l3prot='ipv6')
# read events
else:
    # header
    if args.timestamp:
        print("%-9s " % ("TIME(s)"), end="")
    if args.print_uid:
        print("%6s " % ("UID"), end="")
    if args.errors:
        print(error_header_fmt % ("RC"), end="")
    print(header_fmt % ("PID", "COMM", "PROT", "ADDR", "PORT", "OPTS", "IF"))

    start_ts = 0

    # read events
    b["ipv4_bind_events"].open_perf_buffer(print_ipv4_bind_event)
    b["ipv6_bind_events"].open_perf_buffer(print_ipv6_bind_event)
    while 1:
        try:
            b.perf_buffer_poll()
        except KeyboardInterrupt:
            exit()