[net-next-2.6.git] / net / netfilter / ipvs / ip_vs_app.c

/*
 * ip_vs_app.c: Application module support for IPVS
 *
 * Authors:     Wensong Zhang <wensong@linuxvirtualserver.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, or (at your option) any later version.
 *
 * Most code here is taken from ip_masq_app.c in kernel 2.2. The difference
 * is that ip_vs_app module handles the reverse direction (incoming requests
 * and outgoing responses).
 *
 *		IP_MASQ_APP application masquerading module
 *
 * Author:	Juan Jose Ciarlante, <jjciarla@raiz.uncu.edu.ar>
 *
 */

#define KMSG_COMPONENT "IPVS"
#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt

#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/skbuff.h>
#include <linux/in.h>
#include <linux/ip.h>
#include <linux/netfilter.h>
#include <linux/slab.h>
#include <net/net_namespace.h>
#include <net/protocol.h>
#include <net/tcp.h>
#include <asm/system.h>
#include <linux/stat.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/mutex.h>

#include <net/ip_vs.h>

EXPORT_SYMBOL(register_ip_vs_app);
EXPORT_SYMBOL(unregister_ip_vs_app);
EXPORT_SYMBOL(register_ip_vs_app_inc);

/* ipvs application list head */
static LIST_HEAD(ip_vs_app_list);
static DEFINE_MUTEX(__ip_vs_app_mutex);


/*
 *	Get an ip_vs_app object
 */
static inline int ip_vs_app_get(struct ip_vs_app *app)
{
	return try_module_get(app->module);
}


static inline void ip_vs_app_put(struct ip_vs_app *app)
{
	module_put(app->module);
}


/*
 *	Allocate/initialize app incarnation and register it in proto apps.
 */
static int
ip_vs_app_inc_new(struct ip_vs_app *app, __u16 proto, __u16 port)
{
	struct ip_vs_protocol *pp;
	struct ip_vs_app *inc;
	int ret;

	if (!(pp = ip_vs_proto_get(proto)))
		return -EPROTONOSUPPORT;

	if (!pp->unregister_app)
		return -EOPNOTSUPP;

	inc = kmemdup(app, sizeof(*inc), GFP_KERNEL);
	if (!inc)
		return -ENOMEM;
	INIT_LIST_HEAD(&inc->p_list);
	INIT_LIST_HEAD(&inc->incs_list);
	inc->app = app;
	inc->port = htons(port);
	atomic_set(&inc->usecnt, 0);

	if (app->timeouts) {
		inc->timeout_table =
			ip_vs_create_timeout_table(app->timeouts,
						   app->timeouts_size);
		if (!inc->timeout_table) {
			ret = -ENOMEM;
			goto out;
		}
	}

	ret = pp->register_app(inc);
	if (ret)
		goto out;

	list_add(&inc->a_list, &app->incs_list);
	IP_VS_DBG(9, "%s application %s:%u registered\n",
		  pp->name, inc->name, inc->port);

	return 0;

  out:
	kfree(inc->timeout_table);
	kfree(inc);
	return ret;
}


/*
 *	Release app incarnation
 */
static void
ip_vs_app_inc_release(struct ip_vs_app *inc)
{
	struct ip_vs_protocol *pp;

	if (!(pp = ip_vs_proto_get(inc->protocol)))
		return;

	if (pp->unregister_app)
		pp->unregister_app(inc);

	IP_VS_DBG(9, "%s App %s:%u unregistered\n",
		  pp->name, inc->name, inc->port);

	list_del(&inc->a_list);

	kfree(inc->timeout_table);
	kfree(inc);
}


/*
 *	Get reference to app inc (only called from softirq)
 *
 */
int ip_vs_app_inc_get(struct ip_vs_app *inc)
{
	int result;

	atomic_inc(&inc->usecnt);
	if (unlikely((result = ip_vs_app_get(inc->app)) != 1))
		atomic_dec(&inc->usecnt);
	return result;
}


/*
 *	Put the app inc (only called from timer or net softirq)
 */
void ip_vs_app_inc_put(struct ip_vs_app *inc)
{
	ip_vs_app_put(inc->app);
	atomic_dec(&inc->usecnt);
}


/*
 *	Register an application incarnation in protocol applications
 */
int
register_ip_vs_app_inc(struct ip_vs_app *app, __u16 proto, __u16 port)
{
	int result;

	mutex_lock(&__ip_vs_app_mutex);

	result = ip_vs_app_inc_new(app, proto, port);

	mutex_unlock(&__ip_vs_app_mutex);

	return result;
}


/*
 *	ip_vs_app registration routine
 */
int register_ip_vs_app(struct ip_vs_app *app)
{
	/* increase the module use count */
	ip_vs_use_count_inc();

	mutex_lock(&__ip_vs_app_mutex);

	list_add(&app->a_list, &ip_vs_app_list);

	mutex_unlock(&__ip_vs_app_mutex);

	return 0;
}


/*
 *	ip_vs_app unregistration routine
 *	We are sure there are no app incarnations attached to services
 */
void unregister_ip_vs_app(struct ip_vs_app *app)
{
	struct ip_vs_app *inc, *nxt;

	mutex_lock(&__ip_vs_app_mutex);

	list_for_each_entry_safe(inc, nxt, &app->incs_list, a_list) {
		ip_vs_app_inc_release(inc);
	}

	list_del(&app->a_list);

	mutex_unlock(&__ip_vs_app_mutex);

	/* decrease the module use count */
	ip_vs_use_count_dec();
}


/*
 *	Bind ip_vs_conn to its ip_vs_app (called by cp constructor)
 */
int ip_vs_bind_app(struct ip_vs_conn *cp, struct ip_vs_protocol *pp)
{
	return pp->app_conn_bind(cp);
}


/*
 *	Unbind cp from application incarnation (called by cp destructor)
 */
void ip_vs_unbind_app(struct ip_vs_conn *cp)
{
	struct ip_vs_app *inc = cp->app;

	if (!inc)
		return;

	if (inc->unbind_conn)
		inc->unbind_conn(inc, cp);
	if (inc->done_conn)
		inc->done_conn(inc, cp);
	ip_vs_app_inc_put(inc);
	cp->app = NULL;
}


/*
 *	Fixes th->seq based on ip_vs_seq info.
 */
static inline void vs_fix_seq(const struct ip_vs_seq *vseq, struct tcphdr *th)
{
	__u32 seq = ntohl(th->seq);

	/*
	 *	Adjust seq with delta-offset for all packets after
	 *	the most recent resized pkt seq and with previous_delta offset
	 *	for all packets	before most recent resized pkt seq.
	 */
	if (vseq->delta || vseq->previous_delta) {
		if(after(seq, vseq->init_seq)) {
			th->seq = htonl(seq + vseq->delta);
			IP_VS_DBG(9, "%s(): added delta (%d) to seq\n",
				  __func__, vseq->delta);
		} else {
			th->seq = htonl(seq + vseq->previous_delta);
			IP_VS_DBG(9, "%s(): added previous_delta (%d) to seq\n",
				  __func__, vseq->previous_delta);
		}
	}
}


/*
 *	Fixes th->ack_seq based on ip_vs_seq info.
 */
static inline void
vs_fix_ack_seq(const struct ip_vs_seq *vseq, struct tcphdr *th)
{
	__u32 ack_seq = ntohl(th->ack_seq);

	/*
	 * Adjust ack_seq with delta-offset for
	 * the packets AFTER most recent resized pkt has caused a shift
	 * for packets before most recent resized pkt, use previous_delta
	 */
	if (vseq->delta || vseq->previous_delta) {
		/* since ack_seq is the number of octet that is expected
		   to receive next, so compare it with init_seq+delta */
		if(after(ack_seq, vseq->init_seq+vseq->delta)) {
			th->ack_seq = htonl(ack_seq - vseq->delta);
			IP_VS_DBG(9, "%s(): subtracted delta "
				  "(%d) from ack_seq\n", __func__, vseq->delta);

		} else {
			th->ack_seq = htonl(ack_seq - vseq->previous_delta);
			IP_VS_DBG(9, "%s(): subtracted "
				  "previous_delta (%d) from ack_seq\n",
				  __func__, vseq->previous_delta);
		}
	}
}


/*
 *	Updates ip_vs_seq if pkt has been resized
 *	Assumes already checked proto==IPPROTO_TCP and diff!=0.
 */
static inline void vs_seq_update(struct ip_vs_conn *cp, struct ip_vs_seq *vseq,
				 unsigned flag, __u32 seq, int diff)
{
	/* spinlock is to keep updating cp->flags atomic */
	spin_lock(&cp->lock);
	if (!(cp->flags & flag) || after(seq, vseq->init_seq)) {
		vseq->previous_delta = vseq->delta;
		vseq->delta += diff;
		vseq->init_seq = seq;
		cp->flags |= flag;
	}
	spin_unlock(&cp->lock);
}

static inline int app_tcp_pkt_out(struct ip_vs_conn *cp, struct sk_buff *skb,
				  struct ip_vs_app *app)
{
	int diff;
	const unsigned int tcp_offset = ip_hdrlen(skb);
	struct tcphdr *th;
	__u32 seq;

	if (!skb_make_writable(skb, tcp_offset + sizeof(*th)))
		return 0;

	th = (struct tcphdr *)(skb_network_header(skb) + tcp_offset);

	/*
	 *	Remember seq number in case this pkt gets resized
	 */
	seq = ntohl(th->seq);

	/*
	 *	Fix seq stuff if flagged as so.
	 */
	if (cp->flags & IP_VS_CONN_F_OUT_SEQ)
		vs_fix_seq(&cp->out_seq, th);
	if (cp->flags & IP_VS_CONN_F_IN_SEQ)
		vs_fix_ack_seq(&cp->in_seq, th);

	/*
	 *	Call private output hook function
	 */
	if (app->pkt_out == NULL)
		return 1;

	if (!app->pkt_out(app, cp, skb, &diff))
		return 0;

	/*
	 *	Update ip_vs seq stuff if len has changed.
	 */
	if (diff != 0)
		vs_seq_update(cp, &cp->out_seq,
			      IP_VS_CONN_F_OUT_SEQ, seq, diff);

	return 1;
}

/*
 *	Output pkt hook. Will call bound ip_vs_app specific function
 *	called by ipvs packet handler, assumes previously checked cp!=NULL
 *	returns false if it can't handle packet (oom)
 */
int ip_vs_app_pkt_out(struct ip_vs_conn *cp, struct sk_buff *skb)
{
	struct ip_vs_app *app;

	/*
	 *	check if application module is bound to
	 *	this ip_vs_conn.
	 */
	if ((app = cp->app) == NULL)
		return 1;

	/* TCP is complicated */
	if (cp->protocol == IPPROTO_TCP)
		return app_tcp_pkt_out(cp, skb, app);

	/*
	 *	Call private output hook function
	 */
	if (app->pkt_out == NULL)
		return 1;

	return app->pkt_out(app, cp, skb, NULL);
}


static inline int app_tcp_pkt_in(struct ip_vs_conn *cp, struct sk_buff *skb,
				 struct ip_vs_app *app)
{
	int diff;
	const unsigned int tcp_offset = ip_hdrlen(skb);
	struct tcphdr *th;
	__u32 seq;

	if (!skb_make_writable(skb, tcp_offset + sizeof(*th)))
		return 0;

	th = (struct tcphdr *)(skb_network_header(skb) + tcp_offset);

	/*
	 *	Remember seq number in case this pkt gets resized
	 */
	seq = ntohl(th->seq);

	/*
	 *	Fix seq stuff if flagged as so.
	 */
	if (cp->flags & IP_VS_CONN_F_IN_SEQ)
		vs_fix_seq(&cp->in_seq, th);
	if (cp->flags & IP_VS_CONN_F_OUT_SEQ)
		vs_fix_ack_seq(&cp->out_seq, th);

	/*
	 *	Call private input hook function
	 */
	if (app->pkt_in == NULL)
		return 1;

	if (!app->pkt_in(app, cp, skb, &diff))
		return 0;

	/*
	 *	Update ip_vs seq stuff if len has changed.
	 */
	if (diff != 0)
		vs_seq_update(cp, &cp->in_seq,
			      IP_VS_CONN_F_IN_SEQ, seq, diff);

	return 1;
}

/*
 *	Input pkt hook. Will call bound ip_vs_app specific function
 *	called by ipvs packet handler, assumes previously checked cp!=NULL.
 *	returns false if can't handle packet (oom).
 */
int ip_vs_app_pkt_in(struct ip_vs_conn *cp, struct sk_buff *skb)
{
	struct ip_vs_app *app;

	/*
	 *	check if application module is bound to
	 *	this ip_vs_conn.
	 */
	if ((app = cp->app) == NULL)
		return 1;

	/* TCP is complicated */
	if (cp->protocol == IPPROTO_TCP)
		return app_tcp_pkt_in(cp, skb, app);

	/*
	 *	Call private input hook function
	 */
	if (app->pkt_in == NULL)
		return 1;

	return app->pkt_in(app, cp, skb, NULL);
}


#ifdef CONFIG_PROC_FS
/*
 *	/proc/net/ip_vs_app entry function
 */

static struct ip_vs_app *ip_vs_app_idx(loff_t pos)
{
	struct ip_vs_app *app, *inc;

	list_for_each_entry(app, &ip_vs_app_list, a_list) {
		list_for_each_entry(inc, &app->incs_list, a_list) {
			if (pos-- == 0)
				return inc;
		}
	}
	return NULL;

}

static void *ip_vs_app_seq_start(struct seq_file *seq, loff_t *pos)
{
	mutex_lock(&__ip_vs_app_mutex);

	return *pos ? ip_vs_app_idx(*pos - 1) : SEQ_START_TOKEN;
}

static void *ip_vs_app_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
	struct ip_vs_app *inc, *app;
	struct list_head *e;

	++*pos;
	if (v == SEQ_START_TOKEN)
		return ip_vs_app_idx(0);

	inc = v;
	app = inc->app;

	if ((e = inc->a_list.next) != &app->incs_list)
		return list_entry(e, struct ip_vs_app, a_list);

	/* go on to next application */
	for (e = app->a_list.next; e != &ip_vs_app_list; e = e->next) {
		app = list_entry(e, struct ip_vs_app, a_list);
		list_for_each_entry(inc, &app->incs_list, a_list) {
			return inc;
		}
	}
	return NULL;
}

static void ip_vs_app_seq_stop(struct seq_file *seq, void *v)
{
	mutex_unlock(&__ip_vs_app_mutex);
}

static int ip_vs_app_seq_show(struct seq_file *seq, void *v)
{
	if (v == SEQ_START_TOKEN)
		seq_puts(seq, "prot port    usecnt name\n");
	else {
		const struct ip_vs_app *inc = v;

		seq_printf(seq, "%-3s  %-7u %-6d %-17s\n",
			   ip_vs_proto_name(inc->protocol),
			   ntohs(inc->port),
			   atomic_read(&inc->usecnt),
			   inc->name);
	}
	return 0;
}

static const struct seq_operations ip_vs_app_seq_ops = {
	.start = ip_vs_app_seq_start,
	.next  = ip_vs_app_seq_next,
	.stop  = ip_vs_app_seq_stop,
	.show  = ip_vs_app_seq_show,
};

static int ip_vs_app_open(struct inode *inode, struct file *file)
{
	return seq_open(file, &ip_vs_app_seq_ops);
}

static const struct file_operations ip_vs_app_fops = {
	.owner	 = THIS_MODULE,
	.open	 = ip_vs_app_open,
	.read	 = seq_read,
	.llseek  = seq_lseek,
	.release = seq_release,
};
#endif


/*
 *	Replace a segment of data with a new segment
 */
int ip_vs_skb_replace(struct sk_buff *skb, gfp_t pri,
		      char *o_buf, int o_len, char *n_buf, int n_len)
{
	int diff;
	int o_offset;
	int o_left;

	EnterFunction(9);

	diff = n_len - o_len;
	o_offset = o_buf - (char *)skb->data;
	/* The length of left data after o_buf+o_len in the skb data */
	o_left = skb->len - (o_offset + o_len);

	if (diff <= 0) {
		memmove(o_buf + n_len, o_buf + o_len, o_left);
		memcpy(o_buf, n_buf, n_len);
		skb_trim(skb, skb->len + diff);
	} else if (diff <= skb_tailroom(skb)) {
		skb_put(skb, diff);
		memmove(o_buf + n_len, o_buf + o_len, o_left);
		memcpy(o_buf, n_buf, n_len);
	} else {
		if (pskb_expand_head(skb, skb_headroom(skb), diff, pri))
			return -ENOMEM;
		skb_put(skb, diff);
		memmove(skb->data + o_offset + n_len,
			skb->data + o_offset + o_len, o_left);
		skb_copy_to_linear_data_offset(skb, o_offset, n_buf, n_len);
	}

	/* must update the iph total length here */
	ip_hdr(skb)->tot_len = htons(skb->len);

	LeaveFunction(9);
	return 0;
}


int __init ip_vs_app_init(void)
{
	/* we will replace it with proc_net_ipvs_create() soon */
	proc_net_fops_create(&init_net, "ip_vs_app", 0, &ip_vs_app_fops);
	return 0;
}


void ip_vs_app_cleanup(void)
{
	proc_net_remove(&init_net, "ip_vs_app");
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* ip_vs_app.c: Application module support for IPVS
	3	*
1da177e4 LT	4	* Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
	5	*
	6	* This program is free software; you can redistribute it and/or
	7	* modify it under the terms of the GNU General Public License
	8	* as published by the Free Software Foundation; either version
	9	* 2 of the License, or (at your option) any later version.
	10	*
	11	* Most code here is taken from ip_masq_app.c in kernel 2.2. The difference
	12	* is that ip_vs_app module handles the reverse direction (incoming requests
	13	* and outgoing responses).
	14	*
	15	* IP_MASQ_APP application masquerading module
	16	*
	17	* Author: Juan Jose Ciarlante, <jjciarla@raiz.uncu.edu.ar>
	18	*
	19	*/
	20
9aada7ac HE	21	#define KMSG_COMPONENT "IPVS"
	22	#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
	23
1da177e4 LT	24	#include <linux/module.h>
	25	#include <linux/kernel.h>
	26	#include <linux/skbuff.h>
	27	#include <linux/in.h>
	28	#include <linux/ip.h>
af1e1cf0	29	#include <linux/netfilter.h>
5a0e3ad6	30	#include <linux/slab.h>
457c4cbc	31	#include <net/net_namespace.h>
1da177e4	32	#include <net/protocol.h>
c752f073	33	#include <net/tcp.h>
1da177e4 LT	34	#include <asm/system.h>
	35	#include <linux/stat.h>
	36	#include <linux/proc_fs.h>
	37	#include <linux/seq_file.h>
57b47a53	38	#include <linux/mutex.h>
1da177e4 LT	39
	40	#include <net/ip_vs.h>
	41
	42	EXPORT_SYMBOL(register_ip_vs_app);
	43	EXPORT_SYMBOL(unregister_ip_vs_app);
	44	EXPORT_SYMBOL(register_ip_vs_app_inc);
	45
	46	/* ipvs application list head */
	47	static LIST_HEAD(ip_vs_app_list);
57b47a53	48	static DEFINE_MUTEX(__ip_vs_app_mutex);
1da177e4 LT	49
	50
	51	/*
	52	* Get an ip_vs_app object
	53	*/
	54	static inline int ip_vs_app_get(struct ip_vs_app *app)
	55	{
85b60680	56	return try_module_get(app->module);
1da177e4 LT	57	}
	58
	59
	60	static inline void ip_vs_app_put(struct ip_vs_app *app)
	61	{
85b60680	62	module_put(app->module);
1da177e4 LT	63	}
	64
	65
	66	/*
	67	* Allocate/initialize app incarnation and register it in proto apps.
	68	*/
	69	static int
	70	ip_vs_app_inc_new(struct ip_vs_app *app, __u16 proto, __u16 port)
	71	{
	72	struct ip_vs_protocol *pp;
	73	struct ip_vs_app *inc;
	74	int ret;
	75
	76	if (!(pp = ip_vs_proto_get(proto)))
	77	return -EPROTONOSUPPORT;
	78
	79	if (!pp->unregister_app)
	80	return -EOPNOTSUPP;
	81
8b2ed4bb	82	inc = kmemdup(app, sizeof(*inc), GFP_KERNEL);
1da177e4 LT	83	if (!inc)
1da177e4 LT	84	return -ENOMEM;
1da177e4 LT	85	INIT_LIST_HEAD(&inc->p_list);
	86	INIT_LIST_HEAD(&inc->incs_list);
	87	inc->app = app;
	88	inc->port = htons(port);
	89	atomic_set(&inc->usecnt, 0);
	90
	91	if (app->timeouts) {
	92	inc->timeout_table =
	93	ip_vs_create_timeout_table(app->timeouts,
	94	app->timeouts_size);
	95	if (!inc->timeout_table) {
	96	ret = -ENOMEM;
	97	goto out;
	98	}
	99	}
	100
	101	ret = pp->register_app(inc);
	102	if (ret)
	103	goto out;
	104
	105	list_add(&inc->a_list, &app->incs_list);
	106	IP_VS_DBG(9, "%s application %s:%u registered\n",
	107	pp->name, inc->name, inc->port);
	108
	109	return 0;
	110
	111	out:
a51482bd	112	kfree(inc->timeout_table);
1da177e4 LT	113	kfree(inc);
	114	return ret;
	115	}
	116
	117
	118	/*
	119	* Release app incarnation
	120	*/
	121	static void
	122	ip_vs_app_inc_release(struct ip_vs_app *inc)
	123	{
	124	struct ip_vs_protocol *pp;
	125
	126	if (!(pp = ip_vs_proto_get(inc->protocol)))
	127	return;
	128
	129	if (pp->unregister_app)
	130	pp->unregister_app(inc);
	131
	132	IP_VS_DBG(9, "%s App %s:%u unregistered\n",
	133	pp->name, inc->name, inc->port);
	134
	135	list_del(&inc->a_list);
	136
a51482bd	137	kfree(inc->timeout_table);
1da177e4 LT	138	kfree(inc);
	139	}
	140
	141
	142	/*
	143	* Get reference to app inc (only called from softirq)
	144	*
	145	*/
	146	int ip_vs_app_inc_get(struct ip_vs_app *inc)
	147	{
	148	int result;
	149
	150	atomic_inc(&inc->usecnt);
	151	if (unlikely((result = ip_vs_app_get(inc->app)) != 1))
	152	atomic_dec(&inc->usecnt);
	153	return result;
	154	}
	155
	156
	157	/*
	158	* Put the app inc (only called from timer or net softirq)
	159	*/
	160	void ip_vs_app_inc_put(struct ip_vs_app *inc)
	161	{
	162	ip_vs_app_put(inc->app);
	163	atomic_dec(&inc->usecnt);
	164	}
	165
	166
	167	/*
	168	* Register an application incarnation in protocol applications
	169	*/
	170	int
	171	register_ip_vs_app_inc(struct ip_vs_app *app, __u16 proto, __u16 port)
	172	{
	173	int result;
	174
57b47a53	175	mutex_lock(&__ip_vs_app_mutex);
1da177e4 LT	176
	177	result = ip_vs_app_inc_new(app, proto, port);
	178
57b47a53	179	mutex_unlock(&__ip_vs_app_mutex);
1da177e4 LT	180
	181	return result;
	182	}
	183
	184
	185	/*
	186	* ip_vs_app registration routine
	187	*/
	188	int register_ip_vs_app(struct ip_vs_app *app)
	189	{
	190	/* increase the module use count */
	191	ip_vs_use_count_inc();
	192
57b47a53	193	mutex_lock(&__ip_vs_app_mutex);
1da177e4 LT	194
	195	list_add(&app->a_list, &ip_vs_app_list);
	196
57b47a53	197	mutex_unlock(&__ip_vs_app_mutex);
1da177e4 LT	198
	199	return 0;
	200	}
	201
	202
	203	/*
	204	* ip_vs_app unregistration routine
	205	* We are sure there are no app incarnations attached to services
	206	*/
	207	void unregister_ip_vs_app(struct ip_vs_app *app)
	208	{
	209	struct ip_vs_app inc, nxt;
	210
57b47a53	211	mutex_lock(&__ip_vs_app_mutex);
1da177e4 LT	212
	213	list_for_each_entry_safe(inc, nxt, &app->incs_list, a_list) {
	214	ip_vs_app_inc_release(inc);
	215	}
	216
	217	list_del(&app->a_list);
	218
57b47a53	219	mutex_unlock(&__ip_vs_app_mutex);
1da177e4 LT	220
	221	/* decrease the module use count */
	222	ip_vs_use_count_dec();
	223	}
	224
	225
1da177e4 LT	226	/*
	227	* Bind ip_vs_conn to its ip_vs_app (called by cp constructor)
	228	*/
	229	int ip_vs_bind_app(struct ip_vs_conn cp, struct ip_vs_protocol pp)
	230	{
	231	return pp->app_conn_bind(cp);
	232	}
	233
	234
	235	/*
	236	* Unbind cp from application incarnation (called by cp destructor)
	237	*/
	238	void ip_vs_unbind_app(struct ip_vs_conn *cp)
	239	{
	240	struct ip_vs_app *inc = cp->app;
	241
	242	if (!inc)
	243	return;
	244
	245	if (inc->unbind_conn)
	246	inc->unbind_conn(inc, cp);
	247	if (inc->done_conn)
	248	inc->done_conn(inc, cp);
	249	ip_vs_app_inc_put(inc);
	250	cp->app = NULL;
	251	}
	252
	253
	254	/*
	255	* Fixes th->seq based on ip_vs_seq info.
	256	*/
	257	static inline void vs_fix_seq(const struct ip_vs_seq vseq, struct tcphdr th)
	258	{
	259	__u32 seq = ntohl(th->seq);
	260
	261	/*
	262	* Adjust seq with delta-offset for all packets after
	263	* the most recent resized pkt seq and with previous_delta offset
	264	* for all packets before most recent resized pkt seq.
	265	*/
	266	if (vseq->delta \|\| vseq->previous_delta) {
	267	if(after(seq, vseq->init_seq)) {
	268	th->seq = htonl(seq + vseq->delta);
1e3e238e HE	269	IP_VS_DBG(9, "%s(): added delta (%d) to seq\n",
1e3e238e HE	270	__func__, vseq->delta);
1da177e4 LT	271	} else {
1da177e4 LT	272	th->seq = htonl(seq + vseq->previous_delta);
1e3e238e HE	273	IP_VS_DBG(9, "%s(): added previous_delta (%d) to seq\n",
1e3e238e HE	274	__func__, vseq->previous_delta);
1da177e4 LT	275	}
	276	}
	277	}
	278
	279
	280	/*
	281	* Fixes th->ack_seq based on ip_vs_seq info.
	282	*/
	283	static inline void
	284	vs_fix_ack_seq(const struct ip_vs_seq vseq, struct tcphdr th)
	285	{
	286	__u32 ack_seq = ntohl(th->ack_seq);
	287
	288	/*
	289	* Adjust ack_seq with delta-offset for
	290	* the packets AFTER most recent resized pkt has caused a shift
	291	* for packets before most recent resized pkt, use previous_delta
	292	*/
	293	if (vseq->delta \|\| vseq->previous_delta) {
	294	/* since ack_seq is the number of octet that is expected
	295	to receive next, so compare it with init_seq+delta */
	296	if(after(ack_seq, vseq->init_seq+vseq->delta)) {
	297	th->ack_seq = htonl(ack_seq - vseq->delta);
1e3e238e HE	298	IP_VS_DBG(9, "%s(): subtracted delta "
1e3e238e HE	299	"(%d) from ack_seq\n", __func__, vseq->delta);
1da177e4 LT	300
	301	} else {
	302	th->ack_seq = htonl(ack_seq - vseq->previous_delta);
1e3e238e	303	IP_VS_DBG(9, "%s(): subtracted "
1da177e4	304	"previous_delta (%d) from ack_seq\n",
1e3e238e	305	__func__, vseq->previous_delta);
1da177e4 LT	306	}
	307	}
	308	}
	309
	310
	311	/*
	312	* Updates ip_vs_seq if pkt has been resized
	313	* Assumes already checked proto==IPPROTO_TCP and diff!=0.
	314	*/
	315	static inline void vs_seq_update(struct ip_vs_conn cp, struct ip_vs_seq vseq,
	316	unsigned flag, __u32 seq, int diff)
	317	{
	318	/* spinlock is to keep updating cp->flags atomic */
	319	spin_lock(&cp->lock);
	320	if (!(cp->flags & flag) \|\| after(seq, vseq->init_seq)) {
	321	vseq->previous_delta = vseq->delta;
	322	vseq->delta += diff;
	323	vseq->init_seq = seq;
	324	cp->flags \|= flag;
	325	}
	326	spin_unlock(&cp->lock);
	327	}
	328
3db05fea	329	static inline int app_tcp_pkt_out(struct ip_vs_conn cp, struct sk_buff skb,
1da177e4 LT	330	struct ip_vs_app *app)
	331	{
	332	int diff;
3db05fea	333	const unsigned int tcp_offset = ip_hdrlen(skb);
1da177e4 LT	334	struct tcphdr *th;
	335	__u32 seq;
	336
3db05fea	337	if (!skb_make_writable(skb, tcp_offset + sizeof(*th)))
1da177e4 LT	338	return 0;
1da177e4 LT	339
3db05fea	340	th = (struct tcphdr *)(skb_network_header(skb) + tcp_offset);
1da177e4 LT	341
	342	/*
	343	* Remember seq number in case this pkt gets resized
	344	*/
	345	seq = ntohl(th->seq);
	346
	347	/*
	348	* Fix seq stuff if flagged as so.
	349	*/
	350	if (cp->flags & IP_VS_CONN_F_OUT_SEQ)
	351	vs_fix_seq(&cp->out_seq, th);
	352	if (cp->flags & IP_VS_CONN_F_IN_SEQ)
	353	vs_fix_ack_seq(&cp->in_seq, th);
	354
	355	/*
	356	* Call private output hook function
	357	*/
	358	if (app->pkt_out == NULL)
	359	return 1;
	360
3db05fea	361	if (!app->pkt_out(app, cp, skb, &diff))
1da177e4 LT	362	return 0;
	363
	364	/*
	365	* Update ip_vs seq stuff if len has changed.
	366	*/
	367	if (diff != 0)
	368	vs_seq_update(cp, &cp->out_seq,
	369	IP_VS_CONN_F_OUT_SEQ, seq, diff);
	370
	371	return 1;
	372	}
	373
	374	/*
	375	* Output pkt hook. Will call bound ip_vs_app specific function
	376	* called by ipvs packet handler, assumes previously checked cp!=NULL
	377	* returns false if it can't handle packet (oom)
	378	*/
3db05fea	379	int ip_vs_app_pkt_out(struct ip_vs_conn cp, struct sk_buff skb)
1da177e4 LT	380	{
	381	struct ip_vs_app *app;
	382
	383	/*
	384	* check if application module is bound to
	385	* this ip_vs_conn.
	386	*/
	387	if ((app = cp->app) == NULL)
	388	return 1;
	389
	390	/* TCP is complicated */
	391	if (cp->protocol == IPPROTO_TCP)
3db05fea	392	return app_tcp_pkt_out(cp, skb, app);
1da177e4 LT	393
	394	/*
	395	* Call private output hook function
	396	*/
	397	if (app->pkt_out == NULL)
	398	return 1;
	399
3db05fea	400	return app->pkt_out(app, cp, skb, NULL);
1da177e4 LT	401	}
	402
	403
3db05fea	404	static inline int app_tcp_pkt_in(struct ip_vs_conn cp, struct sk_buff skb,
1da177e4 LT	405	struct ip_vs_app *app)
	406	{
	407	int diff;
3db05fea	408	const unsigned int tcp_offset = ip_hdrlen(skb);
1da177e4 LT	409	struct tcphdr *th;
	410	__u32 seq;
	411
3db05fea	412	if (!skb_make_writable(skb, tcp_offset + sizeof(*th)))
1da177e4 LT	413	return 0;
1da177e4 LT	414
3db05fea	415	th = (struct tcphdr *)(skb_network_header(skb) + tcp_offset);
1da177e4 LT	416
	417	/*
	418	* Remember seq number in case this pkt gets resized
	419	*/
	420	seq = ntohl(th->seq);
	421
	422	/*
	423	* Fix seq stuff if flagged as so.
	424	*/
	425	if (cp->flags & IP_VS_CONN_F_IN_SEQ)
	426	vs_fix_seq(&cp->in_seq, th);
	427	if (cp->flags & IP_VS_CONN_F_OUT_SEQ)
	428	vs_fix_ack_seq(&cp->out_seq, th);
	429
	430	/*
	431	* Call private input hook function
	432	*/
	433	if (app->pkt_in == NULL)
	434	return 1;
	435
3db05fea	436	if (!app->pkt_in(app, cp, skb, &diff))
1da177e4 LT	437	return 0;
	438
	439	/*
	440	* Update ip_vs seq stuff if len has changed.
	441	*/
	442	if (diff != 0)
	443	vs_seq_update(cp, &cp->in_seq,
	444	IP_VS_CONN_F_IN_SEQ, seq, diff);
	445
	446	return 1;
	447	}
	448
	449	/*
	450	* Input pkt hook. Will call bound ip_vs_app specific function
	451	* called by ipvs packet handler, assumes previously checked cp!=NULL.
	452	* returns false if can't handle packet (oom).
	453	*/
3db05fea	454	int ip_vs_app_pkt_in(struct ip_vs_conn cp, struct sk_buff skb)
1da177e4 LT	455	{
	456	struct ip_vs_app *app;
	457
	458	/*
	459	* check if application module is bound to
	460	* this ip_vs_conn.
	461	*/
	462	if ((app = cp->app) == NULL)
	463	return 1;
	464
	465	/* TCP is complicated */
	466	if (cp->protocol == IPPROTO_TCP)
3db05fea	467	return app_tcp_pkt_in(cp, skb, app);
1da177e4 LT	468
	469	/*
	470	* Call private input hook function
	471	*/
	472	if (app->pkt_in == NULL)
	473	return 1;
	474
3db05fea	475	return app->pkt_in(app, cp, skb, NULL);
1da177e4 LT	476	}
	477
	478
	479	#ifdef CONFIG_PROC_FS
	480	/*
	481	* /proc/net/ip_vs_app entry function
	482	*/
	483
	484	static struct ip_vs_app *ip_vs_app_idx(loff_t pos)
	485	{
	486	struct ip_vs_app app, inc;
	487
	488	list_for_each_entry(app, &ip_vs_app_list, a_list) {
	489	list_for_each_entry(inc, &app->incs_list, a_list) {
	490	if (pos-- == 0)
	491	return inc;
	492	}
	493	}
	494	return NULL;
	495
	496	}
	497
	498	static void ip_vs_app_seq_start(struct seq_file seq, loff_t *pos)
	499	{
57b47a53	500	mutex_lock(&__ip_vs_app_mutex);
1da177e4 LT	501
	502	return pos ? ip_vs_app_idx(pos - 1) : SEQ_START_TOKEN;
	503	}
	504
	505	static void ip_vs_app_seq_next(struct seq_file seq, void v, loff_t pos)
	506	{
	507	struct ip_vs_app inc, app;
	508	struct list_head *e;
	509
	510	++*pos;
	511	if (v == SEQ_START_TOKEN)
	512	return ip_vs_app_idx(0);
	513
	514	inc = v;
	515	app = inc->app;
	516
	517	if ((e = inc->a_list.next) != &app->incs_list)
	518	return list_entry(e, struct ip_vs_app, a_list);
	519
	520	/* go on to next application */
	521	for (e = app->a_list.next; e != &ip_vs_app_list; e = e->next) {
	522	app = list_entry(e, struct ip_vs_app, a_list);
	523	list_for_each_entry(inc, &app->incs_list, a_list) {
	524	return inc;
	525	}
	526	}
	527	return NULL;
	528	}
	529
	530	static void ip_vs_app_seq_stop(struct seq_file seq, void v)
	531	{
57b47a53	532	mutex_unlock(&__ip_vs_app_mutex);
1da177e4 LT	533	}
	534
	535	static int ip_vs_app_seq_show(struct seq_file seq, void v)
	536	{
	537	if (v == SEQ_START_TOKEN)
	538	seq_puts(seq, "prot port usecnt name\n");
	539	else {
	540	const struct ip_vs_app *inc = v;
	541
	542	seq_printf(seq, "%-3s %-7u %-6d %-17s\n",
	543	ip_vs_proto_name(inc->protocol),
	544	ntohs(inc->port),
	545	atomic_read(&inc->usecnt),
	546	inc->name);
	547	}
	548	return 0;
	549	}
	550
56b3d975	551	static const struct seq_operations ip_vs_app_seq_ops = {
1da177e4 LT	552	.start = ip_vs_app_seq_start,
	553	.next = ip_vs_app_seq_next,
	554	.stop = ip_vs_app_seq_stop,
	555	.show = ip_vs_app_seq_show,
	556	};
	557
	558	static int ip_vs_app_open(struct inode inode, struct file file)
	559	{
	560	return seq_open(file, &ip_vs_app_seq_ops);
	561	}
	562
9a32144e	563	static const struct file_operations ip_vs_app_fops = {
1da177e4 LT	564	.owner = THIS_MODULE,
	565	.open = ip_vs_app_open,
	566	.read = seq_read,
	567	.llseek = seq_lseek,
	568	.release = seq_release,
	569	};
	570	#endif
	571
	572
	573	/*
	574	* Replace a segment of data with a new segment
	575	*/
dd0fc66f	576	int ip_vs_skb_replace(struct sk_buff *skb, gfp_t pri,
1da177e4 LT	577	char o_buf, int o_len, char n_buf, int n_len)
1da177e4 LT	578	{
1da177e4 LT	579	int diff;
	580	int o_offset;
	581	int o_left;
	582
	583	EnterFunction(9);
	584
	585	diff = n_len - o_len;
	586	o_offset = o_buf - (char *)skb->data;
	587	/* The length of left data after o_buf+o_len in the skb data */
	588	o_left = skb->len - (o_offset + o_len);
	589
	590	if (diff <= 0) {
	591	memmove(o_buf + n_len, o_buf + o_len, o_left);
	592	memcpy(o_buf, n_buf, n_len);
	593	skb_trim(skb, skb->len + diff);
	594	} else if (diff <= skb_tailroom(skb)) {
	595	skb_put(skb, diff);
	596	memmove(o_buf + n_len, o_buf + o_len, o_left);
	597	memcpy(o_buf, n_buf, n_len);
	598	} else {
	599	if (pskb_expand_head(skb, skb_headroom(skb), diff, pri))
	600	return -ENOMEM;
	601	skb_put(skb, diff);
	602	memmove(skb->data + o_offset + n_len,
	603	skb->data + o_offset + o_len, o_left);
27d7ff46	604	skb_copy_to_linear_data_offset(skb, o_offset, n_buf, n_len);
1da177e4 LT	605	}
	606
	607	/* must update the iph total length here */
eddc9ec5	608	ip_hdr(skb)->tot_len = htons(skb->len);
1da177e4 LT	609
	610	LeaveFunction(9);
	611	return 0;
	612	}
	613
	614
048cf48b	615	int __init ip_vs_app_init(void)
1da177e4 LT	616	{
1da177e4 LT	617	/* we will replace it with proc_net_ipvs_create() soon */
457c4cbc	618	proc_net_fops_create(&init_net, "ip_vs_app", 0, &ip_vs_app_fops);
1da177e4 LT	619	return 0;
	620	}
	621
	622
	623	void ip_vs_app_cleanup(void)
	624	{
457c4cbc	625	proc_net_remove(&init_net, "ip_vs_app");
1da177e4	626	}