Repository navigation
Expand file tree
/
Copy pathtcpaccept.py
More file actions
executable file
·273 lines (242 loc) · 8.42 KB
/
Copy pathtcpaccept.py
File metadata and controls
executable file
·273 lines (242 loc) · 8.42 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
#!/usr/bin/python
# @lint-avoid-python-3-compatibility-imports
#
# tcpaccept Trace TCP accept()s.
# For Linux, uses BCC, eBPF. Embedded C.
#
# USAGE: tcpaccept [-h] [-t] [-p PID]
#
# This uses dynamic tracing of the kernel inet_csk_accept() socket function
# (from tcp_prot.accept), and will need to be modified to match kernel changes.
#
# Copyright (c) 2015 Brendan Gregg.
# Licensed under the Apache License, Version 2.0 (the "License")
#
# 13-Oct-2015 Brendan Gregg Created this.
# 14-Feb-2016 " " Switch to bpf_perf_output.
from __future__ import print_function
from bcc import BPF
from socket import inet_ntop, AF_INET, AF_INET6
from struct import pack
import argparse
import ctypes as ct
# arguments
examples = """examples:
./tcpaccept # trace all TCP accept()s
./tcpaccept -t # include timestamps
./tcpaccept -p 181 # only trace PID 181
"""
parser = argparse.ArgumentParser(
description="Trace TCP accepts",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog=examples)
parser.add_argument("-t", "--timestamp", action="store_true",
help="include timestamp on output")
parser.add_argument("-p", "--pid",
help="trace this PID only")
parser.add_argument("--ebpf", action="store_true",
help=argparse.SUPPRESS)
args = parser.parse_args()
debug = 0
# define BPF program
bpf_text = """
#include <uapi/linux/ptrace.h>
#include <net/sock.h>
#include <bcc/proto.h>
// separate data structs for ipv4 and ipv6
struct ipv4_data_t {
// XXX: switch some to u32's when supported
u64 ts_us;
u64 pid;
u32 saddr;
u32 daddr;
u64 ip;
u64 lport;
char task[TASK_COMM_LEN];
};
BPF_PERF_OUTPUT(ipv4_events);
struct ipv6_data_t {
u64 ts_us;
u64 pid;
unsigned __int128 saddr;
unsigned __int128 daddr;
u64 ip;
u64 lport;
char task[TASK_COMM_LEN];
};
BPF_PERF_OUTPUT(ipv6_events);
"""
#
# The following is the code for older kernels(Linux pre-4.16).
# It uses kprobes to instrument inet_csk_accept(). On Linux 4.16 and
# later, the sock:inet_sock_set_state tracepoint should be used instead, as
# is done by the code that follows this.
#
bpf_text_kprobe = """
int kretprobe__inet_csk_accept(struct pt_regs *ctx)
{
struct sock *newsk = (struct sock *)PT_REGS_RC(ctx);
u32 pid = bpf_get_current_pid_tgid();
if (newsk == NULL)
return 0;
// check this is TCP
u8 protocol = 0;
// workaround for reading the sk_protocol bitfield:
// Following comments add by Joe Yin:
// Unfortunately,it can not work since Linux 4.10,
// because the sk_wmem_queued is not following the bitfield of sk_protocol.
// And the following member is sk_gso_max_segs.
// So, we can use this:
// bpf_probe_read(&protocol, 1, (void *)((u64)&newsk->sk_gso_max_segs) - 3);
// In order to diff the pre-4.10 and 4.10+ ,introduce the variables gso_max_segs_offset,sk_lingertime,
// sk_lingertime is closed to the gso_max_segs_offset,and
// the offset between the two members is 4
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)&newsk->sk_gso_max_segs - 3);
else
// pre-4.10 with little endian
protocol = *(u8 *)((u64)&newsk->sk_wmem_queued - 3);
#elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
// 4.10+ with big endian
protocol = *(u8 *)((u64)&newsk->sk_gso_max_segs - 1);
else
// pre-4.10 with big endian
protocol = *(u8 *)((u64)&newsk->sk_wmem_queued - 1);
#else
# error "Fix your compiler's __BYTE_ORDER__?!"
#endif
if (protocol != IPPROTO_TCP)
return 0;
// pull in details
u16 family = 0, lport = 0;
family = newsk->__sk_common.skc_family;
lport = newsk->__sk_common.skc_num;
if (family == AF_INET) {
struct ipv4_data_t data4 = {.pid = pid, .ip = 4};
data4.ts_us = bpf_ktime_get_ns() / 1000;
data4.saddr = newsk->__sk_common.skc_rcv_saddr;
data4.daddr = newsk->__sk_common.skc_daddr;
data4.lport = lport;
bpf_get_current_comm(&data4.task, sizeof(data4.task));
ipv4_events.perf_submit(ctx, &data4, sizeof(data4));
} else if (family == AF_INET6) {
struct ipv6_data_t data6 = {.pid = pid, .ip = 6};
data6.ts_us = bpf_ktime_get_ns() / 1000;
bpf_probe_read(&data6.saddr, sizeof(data6.saddr),
&newsk->__sk_common.skc_v6_rcv_saddr.in6_u.u6_addr32);
bpf_probe_read(&data6.daddr, sizeof(data6.daddr),
&newsk->__sk_common.skc_v6_daddr.in6_u.u6_addr32);
data6.lport = lport;
bpf_get_current_comm(&data6.task, sizeof(data6.task));
ipv6_events.perf_submit(ctx, &data6, sizeof(data6));
}
// else drop
return 0;
}
"""
bpf_text_tracepoint = """
TRACEPOINT_PROBE(sock, inet_sock_set_state)
{
if (args->protocol != IPPROTO_TCP)
return 0;
u32 pid = bpf_get_current_pid_tgid();
// pull in details
u16 family = 0, lport = 0;
family = args->family;
lport = args->sport;
if (family == AF_INET) {
struct ipv4_data_t data4 = {.pid = pid, .ip = 4};
data4.ts_us = bpf_ktime_get_ns() / 1000;
__builtin_memcpy(&data4.saddr, args->saddr, sizeof(data4.saddr));
__builtin_memcpy(&data4.daddr, args->daddr, sizeof(data4.daddr));
data4.lport = lport;
bpf_get_current_comm(&data4.task, sizeof(data4.task));
ipv4_events.perf_submit(args, &data4, sizeof(data4));
} else if (family == AF_INET6) {
struct ipv6_data_t data6 = {.pid = pid, .ip = 6};
data6.ts_us = bpf_ktime_get_ns() / 1000;
__builtin_memcpy(&data6.saddr, args->saddr, sizeof(data6.saddr));
__builtin_memcpy(&data6.daddr, args->daddr, sizeof(data6.daddr));
data6.lport = lport;
bpf_get_current_comm(&data6.task, sizeof(data6.task));
ipv6_events.perf_submit(args, &data6, sizeof(data6));
}
// else drop
return 0;
}
"""
if (BPF.tracepoint_exists("sock", "inet_sock_set_state")):
bpf_text += bpf_text_tracepoint
else:
bpf_text += bpf_text_kprobe
# code substitutions
if args.pid:
bpf_text = bpf_text.replace('FILTER',
'if (pid != %s) { return 0; }' % args.pid)
else:
bpf_text = bpf_text.replace('FILTER', '')
if debug or args.ebpf:
print(bpf_text)
if args.ebpf:
exit()
# event data
TASK_COMM_LEN = 16 # linux/sched.h
class Data_ipv4(ct.Structure):
_fields_ = [
("ts_us", ct.c_ulonglong),
("pid", ct.c_ulonglong),
("saddr", ct.c_uint),
("daddr", ct.c_uint),
("ip", ct.c_ulonglong),
("lport", ct.c_ulonglong),
("task", ct.c_char * TASK_COMM_LEN)
]
class Data_ipv6(ct.Structure):
_fields_ = [
("ts_us", ct.c_ulonglong),
("pid", ct.c_ulonglong),
("saddr", (ct.c_ulonglong * 2)),
("daddr", (ct.c_ulonglong * 2)),
("ip", ct.c_ulonglong),
("lport", ct.c_ulonglong),
("task", ct.c_char * TASK_COMM_LEN)
]
# process event
def print_ipv4_event(cpu, data, size):
event = ct.cast(data, ct.POINTER(Data_ipv4)).contents
global start_ts
if args.timestamp:
if start_ts == 0:
start_ts = event.ts_us
print("%-9.3f" % ((float(event.ts_us) - start_ts) / 1000000), end="")
print("%-6d %-12.12s %-2d %-16s %-16s %-4d" % (event.pid,
event.task.decode(), event.ip,
inet_ntop(AF_INET, pack("I", event.daddr)),
inet_ntop(AF_INET, pack("I", event.saddr)), event.lport))
def print_ipv6_event(cpu, data, size):
event = ct.cast(data, ct.POINTER(Data_ipv6)).contents
global start_ts
if args.timestamp:
if start_ts == 0:
start_ts = event.ts_us
print("%-9.3f" % ((float(event.ts_us) - start_ts) / 1000000), end="")
print("%-6d %-12.12s %-2d %-16s %-16s %-4d" % (event.pid,
event.task.decode(), event.ip, inet_ntop(AF_INET6, event.daddr),
inet_ntop(AF_INET6, event.saddr), event.lport))
# initialize BPF
b = BPF(text=bpf_text)
# header
if args.timestamp:
print("%-9s" % ("TIME(s)"), end="")
print("%-6s %-12s %-2s %-16s %-16s %-4s" % ("PID", "COMM", "IP", "RADDR",
"LADDR", "LPORT"))
start_ts = 0
# read events
b["ipv4_events"].open_perf_buffer(print_ipv4_event)
b["ipv6_events"].open_perf_buffer(print_ipv6_event)
while 1:
b.perf_buffer_poll()