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>
32 MODULE_AUTHOR("Stephen Hemminger <shemminger@linux-foundation.org>");
33 MODULE_DESCRIPTION("TCP cwnd snooper");
34 MODULE_LICENSE("GPL");
35 MODULE_VERSION("1.1");
37 static int port __read_mostly = 0;
38 MODULE_PARM_DESC(port, "Port to match (0=all)");
39 module_param(port, int, 0);
41 static int bufsize __read_mostly = 4096;
42 MODULE_PARM_DESC(bufsize, "Log buffer size in packets (4096)");
43 module_param(bufsize, int, 0);
45 static int full __read_mostly;
46 MODULE_PARM_DESC(full, "Full log (1=every ack packet received, 0=only cwnd changes)");
47 module_param(full, int, 0);
49 static const char procname[] = "tcpprobe";
66 wait_queue_head_t wait;
70 unsigned long head, tail;
75 static inline int tcp_probe_used(void)
77 return (tcp_probe.head - tcp_probe.tail) % bufsize;
80 static inline int tcp_probe_avail(void)
82 return bufsize - tcp_probe_used();
86 * Hook inserted to be called before each receive packet.
87 * Note: arguments must match tcp_rcv_established()!
89 static int jtcp_rcv_established(struct sock *sk, struct sk_buff *skb,
90 struct tcphdr *th, unsigned len)
92 const struct tcp_sock *tp = tcp_sk(sk);
93 const struct inet_sock *inet = inet_sk(sk);
95 /* Only update if port matches */
96 if ((port == 0 || ntohs(inet->dport) == port || ntohs(inet->sport) == port)
97 && (full || tp->snd_cwnd != tcp_probe.lastcwnd)) {
99 spin_lock(&tcp_probe.lock);
100 /* If log fills, just silently drop */
101 if (tcp_probe_avail() > 1) {
102 struct tcp_log *p = tcp_probe.log + tcp_probe.head;
104 p->tstamp = ktime_get();
105 p->saddr = inet->saddr;
106 p->sport = inet->sport;
107 p->daddr = inet->daddr;
108 p->dport = inet->dport;
109 p->length = skb->len;
110 p->snd_nxt = tp->snd_nxt;
111 p->snd_una = tp->snd_una;
112 p->snd_cwnd = tp->snd_cwnd;
113 p->snd_wnd = tp->snd_wnd;
114 p->ssthresh = tcp_current_ssthresh(sk);
115 p->srtt = tp->srtt >> 3;
117 tcp_probe.head = (tcp_probe.head + 1) % bufsize;
119 tcp_probe.lastcwnd = tp->snd_cwnd;
120 spin_unlock(&tcp_probe.lock);
122 wake_up(&tcp_probe.wait);
129 static struct jprobe tcp_jprobe = {
131 .symbol_name = "tcp_rcv_established",
133 .entry = jtcp_rcv_established,
136 static int tcpprobe_open(struct inode * inode, struct file * file)
138 /* Reset (empty) log */
139 spin_lock_bh(&tcp_probe.lock);
140 tcp_probe.head = tcp_probe.tail = 0;
141 tcp_probe.start = ktime_get();
142 spin_unlock_bh(&tcp_probe.lock);
147 static int tcpprobe_sprint(char *tbuf, int n)
149 const struct tcp_log *p
150 = tcp_probe.log + tcp_probe.tail % bufsize;
152 = ktime_to_timespec(ktime_sub(p->tstamp, tcp_probe.start));
154 return snprintf(tbuf, n,
155 "%lu.%09lu %d.%d.%d.%d:%u %d.%d.%d.%d:%u"
156 " %d %#x %#x %u %u %u %u\n",
157 (unsigned long) tv.tv_sec,
158 (unsigned long) tv.tv_nsec,
159 NIPQUAD(p->saddr), ntohs(p->sport),
160 NIPQUAD(p->daddr), ntohs(p->dport),
161 p->length, p->snd_nxt, p->snd_una,
162 p->snd_cwnd, p->ssthresh, p->snd_wnd, p->srtt);
165 static ssize_t tcpprobe_read(struct file *file, char __user *buf,
166 size_t len, loff_t *ppos)
168 int error = 0, cnt = 0;
177 /* Wait for data in buffer */
178 error = wait_event_interruptible(tcp_probe.wait,
179 tcp_probe_used() > 0);
183 spin_lock_bh(&tcp_probe.lock);
184 if (tcp_probe.head == tcp_probe.tail) {
185 /* multiple readers race? */
186 spin_unlock_bh(&tcp_probe.lock);
190 width = tcpprobe_sprint(tbuf, sizeof(tbuf));
193 tcp_probe.tail = (tcp_probe.tail + 1) % bufsize;
195 spin_unlock_bh(&tcp_probe.lock);
197 /* if record greater than space available
198 return partial buffer (so far) */
202 error = copy_to_user(buf + cnt, tbuf, width);
208 return cnt == 0 ? error : cnt;
211 static const struct file_operations tcpprobe_fops = {
212 .owner = THIS_MODULE,
213 .open = tcpprobe_open,
214 .read = tcpprobe_read,
217 static __init int tcpprobe_init(void)
221 init_waitqueue_head(&tcp_probe.wait);
222 spin_lock_init(&tcp_probe.lock);
227 tcp_probe.log = kcalloc(sizeof(struct tcp_log), bufsize, GFP_KERNEL);
231 if (!proc_net_fops_create(procname, S_IRUSR, &tcpprobe_fops))
234 ret = register_jprobe(&tcp_jprobe);
238 pr_info("TCP probe registered (port=%d)\n", port);
241 proc_net_remove(procname);
243 kfree(tcp_probe.log);
246 module_init(tcpprobe_init);
248 static __exit void tcpprobe_exit(void)
250 proc_net_remove(procname);
251 unregister_jprobe(&tcp_jprobe);
252 kfree(tcp_probe.log);
254 module_exit(tcpprobe_exit);