[net-next-2.6.git] / net / ipv4 / tcp_probe.c

/*
 * tcpprobe - Observe the TCP flow with kprobes.
 *
 * The idea for this came from Werner Almesberger's umlsim
 * Copyright (C) 2004, Stephen Hemminger <shemminger@osdl.org>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

#include <linux/kernel.h>
#include <linux/kprobes.h>
#include <linux/socket.h>
#include <linux/tcp.h>
#include <linux/proc_fs.h>
#include <linux/module.h>
#include <linux/ktime.h>
#include <linux/time.h>
#include <net/net_namespace.h>

#include <net/tcp.h>

MODULE_AUTHOR("Stephen Hemminger <shemminger@linux-foundation.org>");
MODULE_DESCRIPTION("TCP cwnd snooper");
MODULE_LICENSE("GPL");
MODULE_VERSION("1.1");

static int port __read_mostly = 0;
MODULE_PARM_DESC(port, "Port to match (0=all)");
module_param(port, int, 0);

static int bufsize __read_mostly = 4096;
MODULE_PARM_DESC(bufsize, "Log buffer size in packets (4096)");
module_param(bufsize, int, 0);

static int full __read_mostly;
MODULE_PARM_DESC(full, "Full log (1=every ack packet received,  0=only cwnd changes)");
module_param(full, int, 0);

static const char procname[] = "tcpprobe";

struct tcp_log {
	ktime_t tstamp;
	__be32	saddr, daddr;
	__be16	sport, dport;
	u16	length;
	u32	snd_nxt;
	u32	snd_una;
	u32	snd_wnd;
	u32	snd_cwnd;
	u32	ssthresh;
	u32	srtt;
};

static struct {
	spinlock_t	lock;
	wait_queue_head_t wait;
	ktime_t		start;
	u32		lastcwnd;

	unsigned long	head, tail;
	struct tcp_log	*log;
} tcp_probe;


static inline int tcp_probe_used(void)
{
	return (tcp_probe.head - tcp_probe.tail) % bufsize;
}

static inline int tcp_probe_avail(void)
{
	return bufsize - tcp_probe_used();
}

/*
 * Hook inserted to be called before each receive packet.
 * Note: arguments must match tcp_rcv_established()!
 */
static int jtcp_rcv_established(struct sock *sk, struct sk_buff *skb,
			       struct tcphdr *th, unsigned len)
{
	const struct tcp_sock *tp = tcp_sk(sk);
	const struct inet_sock *inet = inet_sk(sk);

	/* Only update if port matches */
	if ((port == 0 || ntohs(inet->dport) == port || ntohs(inet->sport) == port)
	    && (full || tp->snd_cwnd != tcp_probe.lastcwnd)) {

		spin_lock(&tcp_probe.lock);
		/* If log fills, just silently drop */
		if (tcp_probe_avail() > 1) {
			struct tcp_log *p = tcp_probe.log + tcp_probe.head;

			p->tstamp = ktime_get();
			p->saddr = inet->saddr;
			p->sport = inet->sport;
			p->daddr = inet->daddr;
			p->dport = inet->dport;
			p->length = skb->len;
			p->snd_nxt = tp->snd_nxt;
			p->snd_una = tp->snd_una;
			p->snd_cwnd = tp->snd_cwnd;
			p->snd_wnd = tp->snd_wnd;
			p->ssthresh = tcp_current_ssthresh(sk);
			p->srtt = tp->srtt >> 3;

			tcp_probe.head = (tcp_probe.head + 1) % bufsize;
		}
		tcp_probe.lastcwnd = tp->snd_cwnd;
		spin_unlock(&tcp_probe.lock);

		wake_up(&tcp_probe.wait);
	}

	jprobe_return();
	return 0;
}

static struct jprobe tcp_jprobe = {
	.kp = {
		.symbol_name	= "tcp_rcv_established",
	},
	.entry	= jtcp_rcv_established,
};

static int tcpprobe_open(struct inode * inode, struct file * file)
{
	/* Reset (empty) log */
	spin_lock_bh(&tcp_probe.lock);
	tcp_probe.head = tcp_probe.tail = 0;
	tcp_probe.start = ktime_get();
	spin_unlock_bh(&tcp_probe.lock);

	return 0;
}

static int tcpprobe_sprint(char *tbuf, int n)
{
	const struct tcp_log *p
		= tcp_probe.log + tcp_probe.tail % bufsize;
	struct timespec tv
		= ktime_to_timespec(ktime_sub(p->tstamp, tcp_probe.start));

	return snprintf(tbuf, n,
			"%lu.%09lu %pI4:%u %pI4:%u %d %#x %#x %u %u %u %u\n",
			(unsigned long) tv.tv_sec,
			(unsigned long) tv.tv_nsec,
			&p->saddr, ntohs(p->sport),
			&p->daddr, ntohs(p->dport),
			p->length, p->snd_nxt, p->snd_una,
			p->snd_cwnd, p->ssthresh, p->snd_wnd, p->srtt);
}

static ssize_t tcpprobe_read(struct file *file, char __user *buf,
			     size_t len, loff_t *ppos)
{
	int error = 0;
	size_t cnt = 0;

	if (!buf)
		return -EINVAL;

	while (cnt < len) {
		char tbuf[128];
		int width;

		/* Wait for data in buffer */
		error = wait_event_interruptible(tcp_probe.wait,
						 tcp_probe_used() > 0);
		if (error)
			break;

		spin_lock_bh(&tcp_probe.lock);
		if (tcp_probe.head == tcp_probe.tail) {
			/* multiple readers race? */
			spin_unlock_bh(&tcp_probe.lock);
			continue;
		}

		width = tcpprobe_sprint(tbuf, sizeof(tbuf));

		if (cnt + width < len)
			tcp_probe.tail = (tcp_probe.tail + 1) % bufsize;

		spin_unlock_bh(&tcp_probe.lock);

		/* if record greater than space available
		   return partial buffer (so far) */
		if (cnt + width >= len)
			break;

		if (copy_to_user(buf + cnt, tbuf, width))
			return -EFAULT;
		cnt += width;
	}

	return cnt == 0 ? error : cnt;
}

static const struct file_operations tcpprobe_fops = {
	.owner	 = THIS_MODULE,
	.open	 = tcpprobe_open,
	.read    = tcpprobe_read,
};

static __init int tcpprobe_init(void)
{
	int ret = -ENOMEM;

	init_waitqueue_head(&tcp_probe.wait);
	spin_lock_init(&tcp_probe.lock);

	if (bufsize < 0)
		return -EINVAL;

	tcp_probe.log = kcalloc(bufsize, sizeof(struct tcp_log), GFP_KERNEL);
	if (!tcp_probe.log)
		goto err0;

	if (!proc_net_fops_create(&init_net, procname, S_IRUSR, &tcpprobe_fops))
		goto err0;

	ret = register_jprobe(&tcp_jprobe);
	if (ret)
		goto err1;

	pr_info("TCP probe registered (port=%d)\n", port);
	return 0;
 err1:
	proc_net_remove(&init_net, procname);
 err0:
	kfree(tcp_probe.log);
	return ret;
}
module_init(tcpprobe_init);

static __exit void tcpprobe_exit(void)
{
	proc_net_remove(&init_net, procname);
	unregister_jprobe(&tcp_jprobe);
	kfree(tcp_probe.log);
}
module_exit(tcpprobe_exit);
Commit	Line	Data
a42e9d6c SH	1	/*
	2	* tcpprobe - Observe the TCP flow with kprobes.
	3	*
	4	* The idea for this came from Werner Almesberger's umlsim
	5	* Copyright (C) 2004, Stephen Hemminger <shemminger@osdl.org>
	6	*
	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
662ad4f8	9	* the Free Software Foundation; either version 2 of the License.
a42e9d6c SH	10	*
	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.
	15	*
	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.
	19	*/
	20
	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>
85795d64 SH	27	#include <linux/ktime.h>
85795d64 SH	28	#include <linux/time.h>
457c4cbc	29	#include <net/net_namespace.h>
a42e9d6c SH	30
	31	#include <net/tcp.h>
	32
65ebe634	33	MODULE_AUTHOR("Stephen Hemminger <shemminger@linux-foundation.org>");
a42e9d6c SH	34	MODULE_DESCRIPTION("TCP cwnd snooper");
a42e9d6c SH	35	MODULE_LICENSE("GPL");
662ad4f8	36	MODULE_VERSION("1.1");
a42e9d6c	37
85795d64	38	static int port __read_mostly = 0;
a42e9d6c SH	39	MODULE_PARM_DESC(port, "Port to match (0=all)");
	40	module_param(port, int, 0);
	41
662ad4f8 SH	42	static int bufsize __read_mostly = 4096;
662ad4f8 SH	43	MODULE_PARM_DESC(bufsize, "Log buffer size in packets (4096)");
a42e9d6c SH	44	module_param(bufsize, int, 0);
a42e9d6c SH	45
85795d64 SH	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);
	49
a42e9d6c SH	50	static const char procname[] = "tcpprobe";
a42e9d6c SH	51
662ad4f8 SH	52	struct tcp_log {
	53	ktime_t tstamp;
	54	__be32 saddr, daddr;
	55	__be16 sport, dport;
	56	u16 length;
	57	u32 snd_nxt;
	58	u32 snd_una;
	59	u32 snd_wnd;
	60	u32 snd_cwnd;
	61	u32 ssthresh;
	62	u32 srtt;
	63	};
	64
	65	static struct {
85795d64	66	spinlock_t lock;
a42e9d6c	67	wait_queue_head_t wait;
85795d64 SH	68	ktime_t start;
85795d64 SH	69	u32 lastcwnd;
a42e9d6c	70
662ad4f8 SH	71	unsigned long head, tail;
	72	struct tcp_log *log;
	73	} tcp_probe;
	74
14a49e1f	75
662ad4f8	76	static inline int tcp_probe_used(void)
a42e9d6c	77	{
662ad4f8 SH	78	return (tcp_probe.head - tcp_probe.tail) % bufsize;
	79	}
	80
	81	static inline int tcp_probe_avail(void)
	82	{
	83	return bufsize - tcp_probe_used();
14a49e1f	84	}
a42e9d6c	85
85795d64 SH	86	/*
	87	* Hook inserted to be called before each receive packet.
	88	* Note: arguments must match tcp_rcv_established()!
	89	*/
	90	static int jtcp_rcv_established(struct sock sk, struct sk_buff skb,
	91	struct tcphdr *th, unsigned len)
a42e9d6c SH	92	{
	93	const struct tcp_sock *tp = tcp_sk(sk);
	94	const struct inet_sock *inet = inet_sk(sk);
	95
85795d64 SH	96	/* Only update if port matches */
85795d64 SH	97	if ((port == 0 \|\| ntohs(inet->dport) == port \|\| ntohs(inet->sport) == port)
662ad4f8 SH	98	&& (full \|\| tp->snd_cwnd != tcp_probe.lastcwnd)) {
	99
	100	spin_lock(&tcp_probe.lock);
	101	/* If log fills, just silently drop */
	102	if (tcp_probe_avail() > 1) {
	103	struct tcp_log *p = tcp_probe.log + tcp_probe.head;
	104
	105	p->tstamp = ktime_get();
	106	p->saddr = inet->saddr;
	107	p->sport = inet->sport;
	108	p->daddr = inet->daddr;
	109	p->dport = inet->dport;
	110	p->length = skb->len;
	111	p->snd_nxt = tp->snd_nxt;
	112	p->snd_una = tp->snd_una;
	113	p->snd_cwnd = tp->snd_cwnd;
	114	p->snd_wnd = tp->snd_wnd;
b3b0b681	115	p->ssthresh = tcp_current_ssthresh(sk);
662ad4f8 SH	116	p->srtt = tp->srtt >> 3;
	117
	118	tcp_probe.head = (tcp_probe.head + 1) % bufsize;
	119	}
	120	tcp_probe.lastcwnd = tp->snd_cwnd;
	121	spin_unlock(&tcp_probe.lock);
	122
	123	wake_up(&tcp_probe.wait);
a42e9d6c SH	124	}
	125
	126	jprobe_return();
	127	return 0;
	128	}
	129
662ad4f8	130	static struct jprobe tcp_jprobe = {
3a872d89	131	.kp = {
85795d64	132	.symbol_name = "tcp_rcv_established",
3a872d89	133	},
9e367d85	134	.entry = jtcp_rcv_established,
a42e9d6c SH	135	};
a42e9d6c SH	136
a42e9d6c SH	137	static int tcpprobe_open(struct inode * inode, struct file * file)
a42e9d6c SH	138	{
662ad4f8 SH	139	/* Reset (empty) log */
	140	spin_lock_bh(&tcp_probe.lock);
	141	tcp_probe.head = tcp_probe.tail = 0;
	142	tcp_probe.start = ktime_get();
	143	spin_unlock_bh(&tcp_probe.lock);
	144
a42e9d6c SH	145	return 0;
	146	}
	147
662ad4f8 SH	148	static int tcpprobe_sprint(char *tbuf, int n)
	149	{
	150	const struct tcp_log *p
	151	= tcp_probe.log + tcp_probe.tail % bufsize;
	152	struct timespec tv
	153	= ktime_to_timespec(ktime_sub(p->tstamp, tcp_probe.start));
	154
	155	return snprintf(tbuf, n,
673d57e7	156	"%lu.%09lu %pI4:%u %pI4:%u %d %#x %#x %u %u %u %u\n",
662ad4f8 SH	157	(unsigned long) tv.tv_sec,
662ad4f8 SH	158	(unsigned long) tv.tv_nsec,
673d57e7 HH	159	&p->saddr, ntohs(p->sport),
673d57e7 HH	160	&p->daddr, ntohs(p->dport),
662ad4f8 SH	161	p->length, p->snd_nxt, p->snd_una,
	162	p->snd_cwnd, p->ssthresh, p->snd_wnd, p->srtt);
	163	}
	164
a42e9d6c SH	165	static ssize_t tcpprobe_read(struct file file, char __user buf,
	166	size_t len, loff_t *ppos)
	167	{
a2025b8b RK	168	int error = 0;
a2025b8b RK	169	size_t cnt = 0;
a42e9d6c	170
a2025b8b	171	if (!buf)
a42e9d6c SH	172	return -EINVAL;
a42e9d6c SH	173
662ad4f8 SH	174	while (cnt < len) {
	175	char tbuf[128];
	176	int width;
	177
	178	/* Wait for data in buffer */
	179	error = wait_event_interruptible(tcp_probe.wait,
	180	tcp_probe_used() > 0);
	181	if (error)
	182	break;
a42e9d6c	183
662ad4f8 SH	184	spin_lock_bh(&tcp_probe.lock);
	185	if (tcp_probe.head == tcp_probe.tail) {
	186	/* multiple readers race? */
	187	spin_unlock_bh(&tcp_probe.lock);
	188	continue;
	189	}
a42e9d6c	190
662ad4f8	191	width = tcpprobe_sprint(tbuf, sizeof(tbuf));
a42e9d6c	192
8d390efd	193	if (cnt + width < len)
662ad4f8	194	tcp_probe.tail = (tcp_probe.tail + 1) % bufsize;
a42e9d6c	195
662ad4f8 SH	196	spin_unlock_bh(&tcp_probe.lock);
	197
	198	/* if record greater than space available
	199	return partial buffer (so far) */
8d390efd	200	if (cnt + width >= len)
662ad4f8 SH	201	break;
662ad4f8 SH	202
8d390efd TQ	203	if (copy_to_user(buf + cnt, tbuf, width))
8d390efd TQ	204	return -EFAULT;
662ad4f8 SH	205	cnt += width;
662ad4f8 SH	206	}
a42e9d6c	207
662ad4f8	208	return cnt == 0 ? error : cnt;
a42e9d6c SH	209	}
a42e9d6c SH	210
9a32144e	211	static const struct file_operations tcpprobe_fops = {
a42e9d6c SH	212	.owner = THIS_MODULE,
	213	.open = tcpprobe_open,
	214	.read = tcpprobe_read,
	215	};
	216
	217	static __init int tcpprobe_init(void)
	218	{
	219	int ret = -ENOMEM;
	220
662ad4f8 SH	221	init_waitqueue_head(&tcp_probe.wait);
	222	spin_lock_init(&tcp_probe.lock);
	223
	224	if (bufsize < 0)
	225	return -EINVAL;
	226
3d8ea1fd	227	tcp_probe.log = kcalloc(bufsize, sizeof(struct tcp_log), GFP_KERNEL);
662ad4f8 SH	228	if (!tcp_probe.log)
662ad4f8 SH	229	goto err0;
a42e9d6c	230
457c4cbc	231	if (!proc_net_fops_create(&init_net, procname, S_IRUSR, &tcpprobe_fops))
a42e9d6c SH	232	goto err0;
a42e9d6c SH	233
662ad4f8	234	ret = register_jprobe(&tcp_jprobe);
a42e9d6c SH	235	if (ret)
	236	goto err1;
	237
662ad4f8	238	pr_info("TCP probe registered (port=%d)\n", port);
a42e9d6c SH	239	return 0;
a42e9d6c SH	240	err1:
457c4cbc	241	proc_net_remove(&init_net, procname);
a42e9d6c	242	err0:
662ad4f8	243	kfree(tcp_probe.log);
a42e9d6c SH	244	return ret;
	245	}
	246	module_init(tcpprobe_init);
	247
	248	static __exit void tcpprobe_exit(void)
	249	{
457c4cbc	250	proc_net_remove(&init_net, procname);
662ad4f8 SH	251	unregister_jprobe(&tcp_jprobe);
662ad4f8 SH	252	kfree(tcp_probe.log);
a42e9d6c SH	253	}
a42e9d6c SH	254	module_exit(tcpprobe_exit);