[deliverable/linux.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 <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);

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);
}

static void ip_vs_app_inc_destroy(struct ip_vs_app *inc)
{
	kfree(inc->timeout_table);
	kfree(inc);
}

static void ip_vs_app_inc_rcu_free(struct rcu_head *head)
{
	struct ip_vs_app *inc = container_of(head, struct ip_vs_app, rcu_head);

	ip_vs_app_inc_destroy(inc);
}

/*
 *	Allocate/initialize app incarnation and register it in proto apps.
 */
static int
ip_vs_app_inc_new(struct netns_ipvs *ipvs, 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(ipvs, inc);
	if (ret)
		goto out;

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

	return 0;

  out:
	ip_vs_app_inc_destroy(inc);
	return ret;
}


/*
 *	Release app incarnation
 */
static void
ip_vs_app_inc_release(struct netns_ipvs *ipvs, 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(ipvs, inc);

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

	list_del(&inc->a_list);

	call_rcu(&inc->rcu_head, ip_vs_app_inc_rcu_free);
}


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

	result = ip_vs_app_get(inc->app);
	if (result)
		atomic_inc(&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)
{
	atomic_dec(&inc->usecnt);
	ip_vs_app_put(inc->app);
}


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

	mutex_lock(&__ip_vs_app_mutex);

	result = ip_vs_app_inc_new(ipvs, app, proto, port);

	mutex_unlock(&__ip_vs_app_mutex);

	return result;
}


/* Register application for netns */
struct ip_vs_app *register_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app)
{
	struct ip_vs_app *a;
	int err = 0;

	mutex_lock(&__ip_vs_app_mutex);

	list_for_each_entry(a, &ipvs->app_list, a_list) {
		if (!strcmp(app->name, a->name)) {
			err = -EEXIST;
			goto out_unlock;
		}
	}
	a = kmemdup(app, sizeof(*app), GFP_KERNEL);
	if (!a) {
		err = -ENOMEM;
		goto out_unlock;
	}
	INIT_LIST_HEAD(&a->incs_list);
	list_add(&a->a_list, &ipvs->app_list);
	/* increase the module use count */
	ip_vs_use_count_inc();

out_unlock:
	mutex_unlock(&__ip_vs_app_mutex);

	return err ? ERR_PTR(err) : a;
}


/*
 *	ip_vs_app unregistration routine
 *	We are sure there are no app incarnations attached to services
 *	Caller should use synchronize_rcu() or rcu_barrier()
 */
void unregister_ip_vs_app(struct netns_ipvs *ipvs, struct ip_vs_app *app)
{
	struct ip_vs_app *a, *anxt, *inc, *nxt;

	mutex_lock(&__ip_vs_app_mutex);

	list_for_each_entry_safe(a, anxt, &ipvs->app_list, a_list) {
		if (app && strcmp(app->name, a->name))
			continue;
		list_for_each_entry_safe(inc, nxt, &a->incs_list, a_list) {
			ip_vs_app_inc_release(ipvs, inc);
		}

		list_del(&a->a_list);
		kfree(a);

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

	mutex_unlock(&__ip_vs_app_mutex);
}


/*
 *	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 int flag, __u32 seq, int diff)
{
	/* spinlock is to keep updating cp->flags atomic */
	spin_lock_bh(&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_bh(&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(struct netns_ipvs *ipvs, loff_t pos)
{
	struct ip_vs_app *app, *inc;

	list_for_each_entry(app, &ipvs->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)
{
	struct net *net = seq_file_net(seq);
	struct netns_ipvs *ipvs = net_ipvs(net);

	mutex_lock(&__ip_vs_app_mutex);

	return *pos ? ip_vs_app_idx(ipvs, *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;
	struct net *net = seq_file_net(seq);
	struct netns_ipvs *ipvs = net_ipvs(net);

	++*pos;
	if (v == SEQ_START_TOKEN)
		return ip_vs_app_idx(ipvs, 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 != &ipvs->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_net(inode, file, &ip_vs_app_seq_ops,
			    sizeof(struct seq_net_private));
}

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_net,
};
#endif

int __net_init ip_vs_app_net_init(struct netns_ipvs *ipvs)
{
	INIT_LIST_HEAD(&ipvs->app_list);
	proc_create("ip_vs_app", 0, ipvs->net->proc_net, &ip_vs_app_fops);
	return 0;
}

void __net_exit ip_vs_app_net_cleanup(struct netns_ipvs *ipvs)
{
	unregister_ip_vs_app(ipvs, NULL /* all */);
	remove_proc_entry("ip_vs_app", ipvs->net->proc_net);
}
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 <linux/stat.h>
	35	#include <linux/proc_fs.h>
	36	#include <linux/seq_file.h>
57b47a53	37	#include <linux/mutex.h>
1da177e4 LT	38
	39	#include <net/ip_vs.h>
	40
	41	EXPORT_SYMBOL(register_ip_vs_app);
	42	EXPORT_SYMBOL(unregister_ip_vs_app);
	43	EXPORT_SYMBOL(register_ip_vs_app_inc);
	44
736561a0 SH	45	static DEFINE_MUTEX(__ip_vs_app_mutex);
736561a0 SH	46
1da177e4 LT	47	/*
	48	* Get an ip_vs_app object
	49	*/
	50	static inline int ip_vs_app_get(struct ip_vs_app *app)
	51	{
85b60680	52	return try_module_get(app->module);
1da177e4 LT	53	}
	54
	55
	56	static inline void ip_vs_app_put(struct ip_vs_app *app)
	57	{
85b60680	58	module_put(app->module);
1da177e4 LT	59	}
1da177e4 LT	60
363c97d7 JA	61	static void ip_vs_app_inc_destroy(struct ip_vs_app *inc)
	62	{
	63	kfree(inc->timeout_table);
	64	kfree(inc);
	65	}
	66
	67	static void ip_vs_app_inc_rcu_free(struct rcu_head *head)
	68	{
	69	struct ip_vs_app *inc = container_of(head, struct ip_vs_app, rcu_head);
	70
	71	ip_vs_app_inc_destroy(inc);
	72	}
1da177e4 LT	73
	74	/*
	75	* Allocate/initialize app incarnation and register it in proto apps.
	76	*/
	77	static int
a080ce38	78	ip_vs_app_inc_new(struct netns_ipvs ipvs, struct ip_vs_app app, __u16 proto,
ab8a5e84	79	__u16 port)
1da177e4 LT	80	{
	81	struct ip_vs_protocol *pp;
	82	struct ip_vs_app *inc;
	83	int ret;
	84
	85	if (!(pp = ip_vs_proto_get(proto)))
	86	return -EPROTONOSUPPORT;
	87
	88	if (!pp->unregister_app)
	89	return -EOPNOTSUPP;
	90
8b2ed4bb	91	inc = kmemdup(app, sizeof(*inc), GFP_KERNEL);
1da177e4 LT	92	if (!inc)
1da177e4 LT	93	return -ENOMEM;
1da177e4 LT	94	INIT_LIST_HEAD(&inc->p_list);
	95	INIT_LIST_HEAD(&inc->incs_list);
	96	inc->app = app;
	97	inc->port = htons(port);
	98	atomic_set(&inc->usecnt, 0);
	99
	100	if (app->timeouts) {
	101	inc->timeout_table =
	102	ip_vs_create_timeout_table(app->timeouts,
	103	app->timeouts_size);
	104	if (!inc->timeout_table) {
	105	ret = -ENOMEM;
	106	goto out;
	107	}
	108	}
	109
19648918	110	ret = pp->register_app(ipvs, inc);
1da177e4 LT	111	if (ret)
	112	goto out;
	113
	114	list_add(&inc->a_list, &app->incs_list);
26c15cfd JA	115	IP_VS_DBG(9, "%s App %s:%u registered\n",
26c15cfd JA	116	pp->name, inc->name, ntohs(inc->port));
1da177e4 LT	117
	118	return 0;
	119
	120	out:
363c97d7	121	ip_vs_app_inc_destroy(inc);
1da177e4 LT	122	return ret;
	123	}
	124
	125
	126	/*
	127	* Release app incarnation
	128	*/
	129	static void
09858708	130	ip_vs_app_inc_release(struct netns_ipvs ipvs, struct ip_vs_app inc)
1da177e4 LT	131	{
	132	struct ip_vs_protocol *pp;
	133
	134	if (!(pp = ip_vs_proto_get(inc->protocol)))
	135	return;
	136
	137	if (pp->unregister_app)
19648918	138	pp->unregister_app(ipvs, inc);
1da177e4 LT	139
1da177e4 LT	140	IP_VS_DBG(9, "%s App %s:%u unregistered\n",
26c15cfd	141	pp->name, inc->name, ntohs(inc->port));
1da177e4 LT	142
	143	list_del(&inc->a_list);
	144
363c97d7	145	call_rcu(&inc->rcu_head, ip_vs_app_inc_rcu_free);
1da177e4 LT	146	}
	147
	148
	149	/*
	150	* Get reference to app inc (only called from softirq)
	151	*
	152	*/
	153	int ip_vs_app_inc_get(struct ip_vs_app *inc)
	154	{
	155	int result;
	156
363c97d7 JA	157	result = ip_vs_app_get(inc->app);
	158	if (result)
	159	atomic_inc(&inc->usecnt);
1da177e4 LT	160	return result;
	161	}
	162
	163
	164	/*
	165	* Put the app inc (only called from timer or net softirq)
	166	*/
	167	void ip_vs_app_inc_put(struct ip_vs_app *inc)
	168	{
1da177e4	169	atomic_dec(&inc->usecnt);
363c97d7	170	ip_vs_app_put(inc->app);
1da177e4 LT	171	}
	172
	173
	174	/*
	175	* Register an application incarnation in protocol applications
	176	*/
	177	int
3250dc9c	178	register_ip_vs_app_inc(struct netns_ipvs ipvs, struct ip_vs_app app, __u16 proto,
ab8a5e84	179	__u16 port)
1da177e4 LT	180	{
	181	int result;
	182
736561a0	183	mutex_lock(&__ip_vs_app_mutex);
1da177e4	184
a080ce38	185	result = ip_vs_app_inc_new(ipvs, app, proto, port);
1da177e4	186
736561a0	187	mutex_unlock(&__ip_vs_app_mutex);
1da177e4 LT	188
	189	return result;
	190	}
	191
	192
be97fdb5	193	/* Register application for netns */
9f8128a5	194	struct ip_vs_app register_ip_vs_app(struct netns_ipvs ipvs, struct ip_vs_app *app)
1da177e4	195	{
be97fdb5 JA	196	struct ip_vs_app *a;
	197	int err = 0;
	198
736561a0	199	mutex_lock(&__ip_vs_app_mutex);
1da177e4	200
be97fdb5 JA	201	list_for_each_entry(a, &ipvs->app_list, a_list) {
	202	if (!strcmp(app->name, a->name)) {
	203	err = -EEXIST;
	204	goto out_unlock;
	205	}
	206	}
	207	a = kmemdup(app, sizeof(*app), GFP_KERNEL);
	208	if (!a) {
	209	err = -ENOMEM;
	210	goto out_unlock;
	211	}
	212	INIT_LIST_HEAD(&a->incs_list);
	213	list_add(&a->a_list, &ipvs->app_list);
	214	/* increase the module use count */
	215	ip_vs_use_count_inc();
1da177e4	216
be97fdb5	217	out_unlock:
736561a0	218	mutex_unlock(&__ip_vs_app_mutex);
1da177e4	219
be97fdb5	220	return err ? ERR_PTR(err) : a;
1da177e4 LT	221	}
	222
	223
	224	/*
	225	* ip_vs_app unregistration routine
	226	* We are sure there are no app incarnations attached to services
363c97d7	227	* Caller should use synchronize_rcu() or rcu_barrier()
1da177e4	228	*/
9f8128a5	229	void unregister_ip_vs_app(struct netns_ipvs ipvs, struct ip_vs_app app)
1da177e4	230	{
be97fdb5	231	struct ip_vs_app a, anxt, inc, nxt;
1da177e4	232
736561a0	233	mutex_lock(&__ip_vs_app_mutex);
1da177e4	234
be97fdb5 JA	235	list_for_each_entry_safe(a, anxt, &ipvs->app_list, a_list) {
	236	if (app && strcmp(app->name, a->name))
	237	continue;
	238	list_for_each_entry_safe(inc, nxt, &a->incs_list, a_list) {
09858708	239	ip_vs_app_inc_release(ipvs, inc);
be97fdb5	240	}
1da177e4	241
be97fdb5 JA	242	list_del(&a->a_list);
be97fdb5 JA	243	kfree(a);
1da177e4	244
be97fdb5 JA	245	/* decrease the module use count */
	246	ip_vs_use_count_dec();
	247	}
1da177e4	248
be97fdb5	249	mutex_unlock(&__ip_vs_app_mutex);
1da177e4 LT	250	}
	251
	252
1da177e4 LT	253	/*
	254	* Bind ip_vs_conn to its ip_vs_app (called by cp constructor)
	255	*/
ab8a5e84 HS	256	int ip_vs_bind_app(struct ip_vs_conn *cp,
ab8a5e84 HS	257	struct ip_vs_protocol *pp)
1da177e4 LT	258	{
	259	return pp->app_conn_bind(cp);
	260	}
	261
	262
	263	/*
	264	* Unbind cp from application incarnation (called by cp destructor)
	265	*/
	266	void ip_vs_unbind_app(struct ip_vs_conn *cp)
	267	{
	268	struct ip_vs_app *inc = cp->app;
	269
	270	if (!inc)
	271	return;
	272
	273	if (inc->unbind_conn)
	274	inc->unbind_conn(inc, cp);
	275	if (inc->done_conn)
	276	inc->done_conn(inc, cp);
	277	ip_vs_app_inc_put(inc);
	278	cp->app = NULL;
	279	}
	280
	281
	282	/*
	283	* Fixes th->seq based on ip_vs_seq info.
	284	*/
	285	static inline void vs_fix_seq(const struct ip_vs_seq vseq, struct tcphdr th)
	286	{
	287	__u32 seq = ntohl(th->seq);
	288
	289	/*
	290	* Adjust seq with delta-offset for all packets after
	291	* the most recent resized pkt seq and with previous_delta offset
	292	* for all packets before most recent resized pkt seq.
	293	*/
	294	if (vseq->delta \|\| vseq->previous_delta) {
	295	if(after(seq, vseq->init_seq)) {
	296	th->seq = htonl(seq + vseq->delta);
1e3e238e HE	297	IP_VS_DBG(9, "%s(): added delta (%d) to seq\n",
1e3e238e HE	298	__func__, vseq->delta);
1da177e4 LT	299	} else {
1da177e4 LT	300	th->seq = htonl(seq + vseq->previous_delta);
1e3e238e HE	301	IP_VS_DBG(9, "%s(): added previous_delta (%d) to seq\n",
1e3e238e HE	302	__func__, vseq->previous_delta);
1da177e4 LT	303	}
	304	}
	305	}
	306
	307
	308	/*
	309	* Fixes th->ack_seq based on ip_vs_seq info.
	310	*/
	311	static inline void
	312	vs_fix_ack_seq(const struct ip_vs_seq vseq, struct tcphdr th)
	313	{
	314	__u32 ack_seq = ntohl(th->ack_seq);
	315
	316	/*
	317	* Adjust ack_seq with delta-offset for
	318	* the packets AFTER most recent resized pkt has caused a shift
	319	* for packets before most recent resized pkt, use previous_delta
	320	*/
	321	if (vseq->delta \|\| vseq->previous_delta) {
	322	/* since ack_seq is the number of octet that is expected
	323	to receive next, so compare it with init_seq+delta */
	324	if(after(ack_seq, vseq->init_seq+vseq->delta)) {
	325	th->ack_seq = htonl(ack_seq - vseq->delta);
1e3e238e HE	326	IP_VS_DBG(9, "%s(): subtracted delta "
1e3e238e HE	327	"(%d) from ack_seq\n", __func__, vseq->delta);
1da177e4 LT	328
	329	} else {
	330	th->ack_seq = htonl(ack_seq - vseq->previous_delta);
1e3e238e	331	IP_VS_DBG(9, "%s(): subtracted "
1da177e4	332	"previous_delta (%d) from ack_seq\n",
1e3e238e	333	__func__, vseq->previous_delta);
1da177e4 LT	334	}
	335	}
	336	}
	337
	338
	339	/*
	340	* Updates ip_vs_seq if pkt has been resized
	341	* Assumes already checked proto==IPPROTO_TCP and diff!=0.
	342	*/
	343	static inline void vs_seq_update(struct ip_vs_conn cp, struct ip_vs_seq vseq,
95c96174	344	unsigned int flag, __u32 seq, int diff)
1da177e4 LT	345	{
1da177e4 LT	346	/* spinlock is to keep updating cp->flags atomic */
ac69269a	347	spin_lock_bh(&cp->lock);
1da177e4 LT	348	if (!(cp->flags & flag) \|\| after(seq, vseq->init_seq)) {
	349	vseq->previous_delta = vseq->delta;
	350	vseq->delta += diff;
	351	vseq->init_seq = seq;
	352	cp->flags \|= flag;
	353	}
ac69269a	354	spin_unlock_bh(&cp->lock);
1da177e4 LT	355	}
1da177e4 LT	356
3db05fea	357	static inline int app_tcp_pkt_out(struct ip_vs_conn cp, struct sk_buff skb,
1da177e4 LT	358	struct ip_vs_app *app)
	359	{
	360	int diff;
3db05fea	361	const unsigned int tcp_offset = ip_hdrlen(skb);
1da177e4 LT	362	struct tcphdr *th;
	363	__u32 seq;
	364
3db05fea	365	if (!skb_make_writable(skb, tcp_offset + sizeof(*th)))
1da177e4 LT	366	return 0;
1da177e4 LT	367
3db05fea	368	th = (struct tcphdr *)(skb_network_header(skb) + tcp_offset);
1da177e4 LT	369
	370	/*
	371	* Remember seq number in case this pkt gets resized
	372	*/
	373	seq = ntohl(th->seq);
	374
	375	/*
	376	* Fix seq stuff if flagged as so.
	377	*/
	378	if (cp->flags & IP_VS_CONN_F_OUT_SEQ)
	379	vs_fix_seq(&cp->out_seq, th);
	380	if (cp->flags & IP_VS_CONN_F_IN_SEQ)
	381	vs_fix_ack_seq(&cp->in_seq, th);
	382
	383	/*
	384	* Call private output hook function
	385	*/
	386	if (app->pkt_out == NULL)
	387	return 1;
	388
3db05fea	389	if (!app->pkt_out(app, cp, skb, &diff))
1da177e4 LT	390	return 0;
	391
	392	/*
	393	* Update ip_vs seq stuff if len has changed.
	394	*/
	395	if (diff != 0)
	396	vs_seq_update(cp, &cp->out_seq,
	397	IP_VS_CONN_F_OUT_SEQ, seq, diff);
	398
	399	return 1;
	400	}
	401
	402	/*
	403	* Output pkt hook. Will call bound ip_vs_app specific function
	404	* called by ipvs packet handler, assumes previously checked cp!=NULL
	405	* returns false if it can't handle packet (oom)
	406	*/
3db05fea	407	int ip_vs_app_pkt_out(struct ip_vs_conn cp, struct sk_buff skb)
1da177e4 LT	408	{
	409	struct ip_vs_app *app;
	410
	411	/*
	412	* check if application module is bound to
	413	* this ip_vs_conn.
	414	*/
	415	if ((app = cp->app) == NULL)
	416	return 1;
	417
	418	/* TCP is complicated */
	419	if (cp->protocol == IPPROTO_TCP)
3db05fea	420	return app_tcp_pkt_out(cp, skb, app);
1da177e4 LT	421
	422	/*
	423	* Call private output hook function
	424	*/
	425	if (app->pkt_out == NULL)
	426	return 1;
	427
3db05fea	428	return app->pkt_out(app, cp, skb, NULL);
1da177e4 LT	429	}
	430
	431
3db05fea	432	static inline int app_tcp_pkt_in(struct ip_vs_conn cp, struct sk_buff skb,
1da177e4 LT	433	struct ip_vs_app *app)
	434	{
	435	int diff;
3db05fea	436	const unsigned int tcp_offset = ip_hdrlen(skb);
1da177e4 LT	437	struct tcphdr *th;
	438	__u32 seq;
	439
3db05fea	440	if (!skb_make_writable(skb, tcp_offset + sizeof(*th)))
1da177e4 LT	441	return 0;
1da177e4 LT	442
3db05fea	443	th = (struct tcphdr *)(skb_network_header(skb) + tcp_offset);
1da177e4 LT	444
	445	/*
	446	* Remember seq number in case this pkt gets resized
	447	*/
	448	seq = ntohl(th->seq);
	449
	450	/*
	451	* Fix seq stuff if flagged as so.
	452	*/
	453	if (cp->flags & IP_VS_CONN_F_IN_SEQ)
	454	vs_fix_seq(&cp->in_seq, th);
	455	if (cp->flags & IP_VS_CONN_F_OUT_SEQ)
	456	vs_fix_ack_seq(&cp->out_seq, th);
	457
	458	/*
	459	* Call private input hook function
	460	*/
	461	if (app->pkt_in == NULL)
	462	return 1;
	463
3db05fea	464	if (!app->pkt_in(app, cp, skb, &diff))
1da177e4 LT	465	return 0;
	466
	467	/*
	468	* Update ip_vs seq stuff if len has changed.
	469	*/
	470	if (diff != 0)
	471	vs_seq_update(cp, &cp->in_seq,
	472	IP_VS_CONN_F_IN_SEQ, seq, diff);
	473
	474	return 1;
	475	}
	476
	477	/*
	478	* Input pkt hook. Will call bound ip_vs_app specific function
	479	* called by ipvs packet handler, assumes previously checked cp!=NULL.
	480	* returns false if can't handle packet (oom).
	481	*/
3db05fea	482	int ip_vs_app_pkt_in(struct ip_vs_conn cp, struct sk_buff skb)
1da177e4 LT	483	{
	484	struct ip_vs_app *app;
	485
	486	/*
	487	* check if application module is bound to
	488	* this ip_vs_conn.
	489	*/
	490	if ((app = cp->app) == NULL)
	491	return 1;
	492
	493	/* TCP is complicated */
	494	if (cp->protocol == IPPROTO_TCP)
3db05fea	495	return app_tcp_pkt_in(cp, skb, app);
1da177e4 LT	496
	497	/*
	498	* Call private input hook function
	499	*/
	500	if (app->pkt_in == NULL)
	501	return 1;
	502
3db05fea	503	return app->pkt_in(app, cp, skb, NULL);
1da177e4 LT	504	}
	505
	506
	507	#ifdef CONFIG_PROC_FS
	508	/*
	509	* /proc/net/ip_vs_app entry function
	510	*/
	511
ab8a5e84	512	static struct ip_vs_app ip_vs_app_idx(struct netns_ipvs ipvs, loff_t pos)
1da177e4 LT	513	{
	514	struct ip_vs_app app, inc;
	515
ab8a5e84	516	list_for_each_entry(app, &ipvs->app_list, a_list) {
1da177e4 LT	517	list_for_each_entry(inc, &app->incs_list, a_list) {
	518	if (pos-- == 0)
	519	return inc;
	520	}
	521	}
	522	return NULL;
	523
	524	}
	525
	526	static void ip_vs_app_seq_start(struct seq_file seq, loff_t *pos)
	527	{
ab8a5e84 HS	528	struct net *net = seq_file_net(seq);
	529	struct netns_ipvs *ipvs = net_ipvs(net);
	530
736561a0	531	mutex_lock(&__ip_vs_app_mutex);
1da177e4	532
ab8a5e84	533	return pos ? ip_vs_app_idx(ipvs, pos - 1) : SEQ_START_TOKEN;
1da177e4 LT	534	}
	535
	536	static void ip_vs_app_seq_next(struct seq_file seq, void v, loff_t pos)
	537	{
	538	struct ip_vs_app inc, app;
	539	struct list_head *e;
ab8a5e84 HS	540	struct net *net = seq_file_net(seq);
ab8a5e84 HS	541	struct netns_ipvs *ipvs = net_ipvs(net);
1da177e4 LT	542
	543	++*pos;
	544	if (v == SEQ_START_TOKEN)
ab8a5e84	545	return ip_vs_app_idx(ipvs, 0);
1da177e4 LT	546
	547	inc = v;
	548	app = inc->app;
	549
	550	if ((e = inc->a_list.next) != &app->incs_list)
	551	return list_entry(e, struct ip_vs_app, a_list);
	552
	553	/* go on to next application */
ab8a5e84	554	for (e = app->a_list.next; e != &ipvs->app_list; e = e->next) {
1da177e4 LT	555	app = list_entry(e, struct ip_vs_app, a_list);
	556	list_for_each_entry(inc, &app->incs_list, a_list) {
	557	return inc;
	558	}
	559	}
	560	return NULL;
	561	}
	562
	563	static void ip_vs_app_seq_stop(struct seq_file seq, void v)
	564	{
736561a0	565	mutex_unlock(&__ip_vs_app_mutex);
1da177e4 LT	566	}
	567
	568	static int ip_vs_app_seq_show(struct seq_file seq, void v)
	569	{
	570	if (v == SEQ_START_TOKEN)
	571	seq_puts(seq, "prot port usecnt name\n");
	572	else {
	573	const struct ip_vs_app *inc = v;
	574
	575	seq_printf(seq, "%-3s %-7u %-6d %-17s\n",
	576	ip_vs_proto_name(inc->protocol),
	577	ntohs(inc->port),
	578	atomic_read(&inc->usecnt),
	579	inc->name);
	580	}
	581	return 0;
	582	}
	583
56b3d975	584	static const struct seq_operations ip_vs_app_seq_ops = {
1da177e4 LT	585	.start = ip_vs_app_seq_start,
	586	.next = ip_vs_app_seq_next,
	587	.stop = ip_vs_app_seq_stop,
	588	.show = ip_vs_app_seq_show,
	589	};
	590
	591	static int ip_vs_app_open(struct inode inode, struct file file)
	592	{
ab8a5e84 HS	593	return seq_open_net(inode, file, &ip_vs_app_seq_ops,
ab8a5e84 HS	594	sizeof(struct seq_net_private));
1da177e4 LT	595	}
1da177e4 LT	596
9a32144e	597	static const struct file_operations ip_vs_app_fops = {
1da177e4 LT	598	.owner = THIS_MODULE,
	599	.open = ip_vs_app_open,
	600	.read = seq_read,
	601	.llseek = seq_lseek,
0f08190f	602	.release = seq_release_net,
1da177e4 LT	603	};
	604	#endif
	605
b5dd212c	606	int __net_init ip_vs_app_net_init(struct netns_ipvs *ipvs)
1da177e4	607	{
ab8a5e84	608	INIT_LIST_HEAD(&ipvs->app_list);
f6ca9f46	609	proc_create("ip_vs_app", 0, ipvs->net->proc_net, &ip_vs_app_fops);
1da177e4 LT	610	return 0;
	611	}
	612
b5dd212c	613	void __net_exit ip_vs_app_net_cleanup(struct netns_ipvs *ipvs)
61b1ab45	614	{
9f8128a5	615	unregister_ip_vs_app(ipvs, NULL /* all */);
f6ca9f46	616	remove_proc_entry("ip_vs_app", ipvs->net->proc_net);
61b1ab45	617	}