2 * tcpprobe - Observe the TCP flow with kprobes.
4 * The idea for this came from Werner Almesberger's umlsim
5 * Copyright (C) 2004, Stephen Hemminger <shemminger@osdl.org>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 #include <linux/kernel.h>
22 #include <linux/kprobes.h>
23 #include <linux/socket.h>
24 #include <linux/tcp.h>
25 #include <linux/proc_fs.h>
26 #include <linux/module.h>
27 #include <linux/ktime.h>
28 #include <linux/time.h>
29 #include <net/net_namespace.h>
33 MODULE_AUTHOR("Stephen Hemminger <shemminger@linux-foundation.org>");
34 MODULE_DESCRIPTION("TCP cwnd snooper");
35 MODULE_LICENSE("GPL");
36 MODULE_VERSION("1.1");
38 static int port __read_mostly
= 0;
39 MODULE_PARM_DESC(port
, "Port to match (0=all)");
40 module_param(port
, int, 0);
42 static unsigned int bufsize __read_mostly
= 4096;
43 MODULE_PARM_DESC(bufsize
, "Log buffer size in packets (4096)");
44 module_param(bufsize
, uint
, 0);
46 static int full __read_mostly
;
47 MODULE_PARM_DESC(full
, "Full log (1=every ack packet received, 0=only cwnd changes)");
48 module_param(full
, int, 0);
50 static const char procname
[] = "tcpprobe";
67 wait_queue_head_t wait
;
71 unsigned long head
, tail
;
76 static inline int tcp_probe_used(void)
78 return (tcp_probe
.head
- tcp_probe
.tail
) & (bufsize
- 1);
81 static inline int tcp_probe_avail(void)
83 return bufsize
- tcp_probe_used() - 1;
87 * Hook inserted to be called before each receive packet.
88 * Note: arguments must match tcp_rcv_established()!
90 static int jtcp_rcv_established(struct sock
*sk
, struct sk_buff
*skb
,
91 struct tcphdr
*th
, unsigned len
)
93 const struct tcp_sock
*tp
= tcp_sk(sk
);
94 const struct inet_sock
*inet
= inet_sk(sk
);
96 /* Only update if port matches */
97 if ((port
== 0 || ntohs(inet
->inet_dport
) == port
||
98 ntohs(inet
->inet_sport
) == port
) &&
99 (full
|| tp
->snd_cwnd
!= tcp_probe
.lastcwnd
)) {
101 spin_lock(&tcp_probe
.lock
);
102 /* If log fills, just silently drop */
103 if (tcp_probe_avail() > 1) {
104 struct tcp_log
*p
= tcp_probe
.log
+ tcp_probe
.head
;
106 p
->tstamp
= ktime_get();
107 p
->saddr
= inet
->inet_saddr
;
108 p
->sport
= inet
->inet_sport
;
109 p
->daddr
= inet
->inet_daddr
;
110 p
->dport
= inet
->inet_dport
;
111 p
->length
= skb
->len
;
112 p
->snd_nxt
= tp
->snd_nxt
;
113 p
->snd_una
= tp
->snd_una
;
114 p
->snd_cwnd
= tp
->snd_cwnd
;
115 p
->snd_wnd
= tp
->snd_wnd
;
116 p
->ssthresh
= tcp_current_ssthresh(sk
);
117 p
->srtt
= tp
->srtt
>> 3;
119 tcp_probe
.head
= (tcp_probe
.head
+ 1) & (bufsize
- 1);
121 tcp_probe
.lastcwnd
= tp
->snd_cwnd
;
122 spin_unlock(&tcp_probe
.lock
);
124 wake_up(&tcp_probe
.wait
);
131 static struct jprobe tcp_jprobe
= {
133 .symbol_name
= "tcp_rcv_established",
135 .entry
= jtcp_rcv_established
,
138 static int tcpprobe_open(struct inode
* inode
, struct file
* file
)
140 /* Reset (empty) log */
141 spin_lock_bh(&tcp_probe
.lock
);
142 tcp_probe
.head
= tcp_probe
.tail
= 0;
143 tcp_probe
.start
= ktime_get();
144 spin_unlock_bh(&tcp_probe
.lock
);
149 static int tcpprobe_sprint(char *tbuf
, int n
)
151 const struct tcp_log
*p
152 = tcp_probe
.log
+ tcp_probe
.tail
;
154 = ktime_to_timespec(ktime_sub(p
->tstamp
, tcp_probe
.start
));
156 return snprintf(tbuf
, n
,
157 "%lu.%09lu %pI4:%u %pI4:%u %d %#x %#x %u %u %u %u\n",
158 (unsigned long) tv
.tv_sec
,
159 (unsigned long) tv
.tv_nsec
,
160 &p
->saddr
, ntohs(p
->sport
),
161 &p
->daddr
, ntohs(p
->dport
),
162 p
->length
, p
->snd_nxt
, p
->snd_una
,
163 p
->snd_cwnd
, p
->ssthresh
, p
->snd_wnd
, p
->srtt
);
166 static ssize_t
tcpprobe_read(struct file
*file
, char __user
*buf
,
167 size_t len
, loff_t
*ppos
)
179 /* Wait for data in buffer */
180 error
= wait_event_interruptible(tcp_probe
.wait
,
181 tcp_probe_used() > 0);
185 spin_lock_bh(&tcp_probe
.lock
);
186 if (tcp_probe
.head
== tcp_probe
.tail
) {
187 /* multiple readers race? */
188 spin_unlock_bh(&tcp_probe
.lock
);
192 width
= tcpprobe_sprint(tbuf
, sizeof(tbuf
));
194 if (cnt
+ width
< len
)
195 tcp_probe
.tail
= (tcp_probe
.tail
+ 1) & (bufsize
- 1);
197 spin_unlock_bh(&tcp_probe
.lock
);
199 /* if record greater than space available
200 return partial buffer (so far) */
201 if (cnt
+ width
>= len
)
204 if (copy_to_user(buf
+ cnt
, tbuf
, width
))
209 return cnt
== 0 ? error
: cnt
;
212 static const struct file_operations tcpprobe_fops
= {
213 .owner
= THIS_MODULE
,
214 .open
= tcpprobe_open
,
215 .read
= tcpprobe_read
,
218 static __init
int tcpprobe_init(void)
222 init_waitqueue_head(&tcp_probe
.wait
);
223 spin_lock_init(&tcp_probe
.lock
);
228 bufsize
= roundup_pow_of_two(bufsize
);
229 tcp_probe
.log
= kcalloc(bufsize
, sizeof(struct tcp_log
), GFP_KERNEL
);
233 if (!proc_net_fops_create(&init_net
, procname
, S_IRUSR
, &tcpprobe_fops
))
236 ret
= register_jprobe(&tcp_jprobe
);
240 pr_info("TCP probe registered (port=%d) bufsize=%u\n", port
, bufsize
);
243 proc_net_remove(&init_net
, procname
);
245 kfree(tcp_probe
.log
);
248 module_init(tcpprobe_init
);
250 static __exit
void tcpprobe_exit(void)
252 proc_net_remove(&init_net
, procname
);
253 unregister_jprobe(&tcp_jprobe
);
254 kfree(tcp_probe
.log
);
256 module_exit(tcpprobe_exit
);