ip_vs_ctl.c

linux-2.6.15.6

/*
 * IPVS         An implementation of the IP virtual server support for the
 *              LINUX operating system.  IPVS is now implemented as a module
 *              over the NetFilter framework. IPVS can be used to build a
 *              high-performance and highly available server based on a
 *              cluster of servers.
 *
 * Version:     $Id: ip_vs_ctl.c,v 1.36 2003/06/08 09:31:19 wensong Exp $
 *
 * Authors:     Wensong Zhang <[e-mail]>
 *              Peter Kese <[e-mail]>
 *              Julian Anastasov <[e-mail]>
 *
 *              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.
 *
 * Changes:
 *
 */

#include <linux/module.h>
#include <linux/init.h>
#include <linux/types.h>
#include <linux/fs.h>
#include <linux/sysctl.h>
#include <linux/proc_fs.h>
#include <linux/workqueue.h>
#include <linux/swap.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>

#include <linux/netfilter.h>
#include <linux/netfilter_ipv4.h>

#include <net/ip.h>
#include <net/sock.h>

#include <asm/uaccess.h>

#include <net/ip_vs.h>

/* semaphore for IPVS sockopts. And, [gs]etsockopt may sleep. */
static DECLARE_MUTEX(__ip_vs_mutex);

/* lock for service table */
static DEFINE_RWLOCK(__ip_vs_svc_lock);

/* lock for table with the real services */
static DEFINE_RWLOCK(__ip_vs_rs_lock);

/* lock for state and timeout tables */
static DEFINE_RWLOCK(__ip_vs_securetcp_lock);

/* lock for drop entry handling */
static DEFINE_SPINLOCK(__ip_vs_dropentry_lock);

/* lock for drop packet handling */
static DEFINE_SPINLOCK(__ip_vs_droppacket_lock);

/* 1/rate drop and drop-entry variables */
int ip_vs_drop_rate = 0;
int ip_vs_drop_counter = 0;
static atomic_t ip_vs_dropentry = ATOMIC_INIT(0);

/* number of virtual services */
static int ip_vs_num_services = 0;

/* sysctl variables */
static int sysctl_ip_vs_drop_entry = 0;
static int sysctl_ip_vs_drop_packet = 0;
static int sysctl_ip_vs_secure_tcp = 0;
static int sysctl_ip_vs_amemthresh = 1024;
static int sysctl_ip_vs_am_droprate = 10;
int sysctl_ip_vs_cache_bypass = 0;
int sysctl_ip_vs_expire_nodest_conn = 0;
int sysctl_ip_vs_expire_quiescent_template = 0;
int sysctl_ip_vs_sync_threshold[2] = { 3, 50 };
int sysctl_ip_vs_nat_icmp_send = 0;


#ifdef CONFIG_IP_VS_DEBUG
static int sysctl_ip_vs_debug_level = 0;

int ip_vs_get_debug_level(void)
{
	return sysctl_ip_vs_debug_level;
}
#endif

/*
 *	update_defense_level is called from keventd and from sysctl,
 *	so it needs to protect itself from softirqs
 */
static void update_defense_level(void)
{
	struct sysinfo i;
	static int old_secure_tcp = 0;
	int availmem;
	int nomem;
	int to_change = -1;

	/* we only count free and buffered memory (in pages) */
	si_meminfo(&i);
	availmem = i.freeram + i.bufferram;
	/* however in linux 2.5 the i.bufferram is total page cache size,
	   we need adjust it */
	/* si_swapinfo(&i); */
	/* availmem = availmem - (i.totalswap - i.freeswap); */

	nomem = (availmem < sysctl_ip_vs_amemthresh);

	local_bh_disable();

	/* drop_entry */
	spin_lock(&__ip_vs_dropentry_lock);
	switch (sysctl_ip_vs_drop_entry) {
	case 0:
		atomic_set(&ip_vs_dropentry, 0);
		break;
	case 1:
		if (nomem) {
			atomic_set(&ip_vs_dropentry, 1);
			sysctl_ip_vs_drop_entry = 2;
		} else {
			atomic_set(&ip_vs_dropentry, 0);
		}
		break;
	case 2:
		if (nomem) {
			atomic_set(&ip_vs_dropentry, 1);
		} else {
			atomic_set(&ip_vs_dropentry, 0);
			sysctl_ip_vs_drop_entry = 1;
		};
		break;
	case 3:
		atomic_set(&ip_vs_dropentry, 1);
		break;
	}
	spin_unlock(&__ip_vs_dropentry_lock);

	/* drop_packet */
	spin_lock(&__ip_vs_droppacket_lock);
	switch (sysctl_ip_vs_drop_packet) {
	case 0:
		ip_vs_drop_rate = 0;
		break;
	case 1:
		if (nomem) {
			ip_vs_drop_rate = ip_vs_drop_counter
				= sysctl_ip_vs_amemthresh /
				(sysctl_ip_vs_amemthresh-availmem);
			sysctl_ip_vs_drop_packet = 2;
		} else {
			ip_vs_drop_rate = 0;
		}
		break;
	case 2:
		if (nomem) {
			ip_vs_drop_rate = ip_vs_drop_counter
				= sysctl_ip_vs_amemthresh /
				(sysctl_ip_vs_amemthresh-availmem);
		} else {
			ip_vs_drop_rate = 0;
			sysctl_ip_vs_drop_packet = 1;
		}
		break;
	case 3:
		ip_vs_drop_rate = sysctl_ip_vs_am_droprate;
		break;
	}
	spin_unlock(&__ip_vs_droppacket_lock);

	/* secure_tcp */
	write_lock(&__ip_vs_securetcp_lock);
	switch (sysctl_ip_vs_secure_tcp) {
	case 0:
		if (old_secure_tcp >= 2)
			to_change = 0;
		break;
	case 1:
		if (nomem) {
			if (old_secure_tcp < 2)
				to_change = 1;
			sysctl_ip_vs_secure_tcp = 2;
		} else {
			if (old_secure_tcp >= 2)
				to_change = 0;
		}
		break;
	case 2:
		if (nomem) {
			if (old_secure_tcp < 2)
				to_change = 1;
		} else {
			if (old_secure_tcp >= 2)
				to_change = 0;
			sysctl_ip_vs_secure_tcp = 1;
		}
		break;
	case 3:
		if (old_secure_tcp < 2)
			to_change = 1;
		break;
	}
	old_secure_tcp = sysctl_ip_vs_secure_tcp;
	if (to_change >= 0)
		ip_vs_protocol_timeout_change(sysctl_ip_vs_secure_tcp>1);
	write_unlock(&__ip_vs_securetcp_lock);

	local_bh_enable();
}


/*
 *	Timer for checking the defense
 */
#define DEFENSE_TIMER_PERIOD	1*HZ
static void defense_work_handler(void *data);
static DECLARE_WORK(defense_work, defense_work_handler, NULL);

static void defense_work_handler(void *data)
{
	update_defense_level();
	if (atomic_read(&ip_vs_dropentry))
		ip_vs_random_dropentry();

	schedule_delayed_work(&defense_work, DEFENSE_TIMER_PERIOD);
}

int
ip_vs_use_count_inc(void)
{
	return try_module_get(THIS_MODULE);
}

void
ip_vs_use_count_dec(void)
{
	module_put(THIS_MODULE);
}


/*
 *	Hash table: for virtual service lookups
 */
#define IP_VS_SVC_TAB_BITS 8
#define IP_VS_SVC_TAB_SIZE (1 << IP_VS_SVC_TAB_BITS)
#define IP_VS_SVC_TAB_MASK (IP_VS_SVC_TAB_SIZE - 1)

/* the service table hashed by <protocol, addr, port> */
static struct list_head ip_vs_svc_table[IP_VS_SVC_TAB_SIZE];
/* the service table hashed by fwmark */
static struct list_head ip_vs_svc_fwm_table[IP_VS_SVC_TAB_SIZE];

/*
 *	Hash table: for real service lookups
 */
#define IP_VS_RTAB_BITS 4
#define IP_VS_RTAB_SIZE (1 << IP_VS_RTAB_BITS)
#define IP_VS_RTAB_MASK (IP_VS_RTAB_SIZE - 1)

static struct list_head ip_vs_rtable[IP_VS_RTAB_SIZE];

/*
 *	Trash for destinations
 */
static LIST_HEAD(ip_vs_dest_trash);

/*
 *	FTP & NULL virtual service counters
 */
static atomic_t ip_vs_ftpsvc_counter = ATOMIC_INIT(0);
static atomic_t ip_vs_nullsvc_counter = ATOMIC_INIT(0);


/*
 *	Returns hash value for virtual service
 */
static __inline__ unsigned
ip_vs_svc_hashkey(unsigned proto, __u32 addr, __u16 port)
{
	register unsigned porth = ntohs(port);

	return (proto^ntohl(addr)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
		& IP_VS_SVC_TAB_MASK;
}

/*
 *	Returns hash value of fwmark for virtual service lookup
 */
static __inline__ unsigned ip_vs_svc_fwm_hashkey(__u32 fwmark)
{
	return fwmark & IP_VS_SVC_TAB_MASK;
}

/*
 *	Hashes a service in the ip_vs_svc_table by <proto,addr,port>
 *	or in the ip_vs_svc_fwm_table by fwmark.
 *	Should be called with locked tables.
 */
static int ip_vs_svc_hash(struct ip_vs_service *svc)
{
	unsigned hash;

	if (svc->flags & IP_VS_SVC_F_HASHED) {
		IP_VS_ERR("ip_vs_svc_hash(): request for already hashed, "
			  "called from %p\n", __builtin_return_address(0));
		return 0;
	}

	if (svc->fwmark == 0) {
		/*
		 *  Hash it by <protocol,addr,port> in ip_vs_svc_table
		 */
		hash = ip_vs_svc_hashkey(svc->protocol, svc->addr, svc->port);
		list_add(&svc->s_list, &ip_vs_svc_table[hash]);
	} else {
		/*
		 *  Hash it by fwmark in ip_vs_svc_fwm_table
		 */
		hash = ip_vs_svc_fwm_hashkey(svc->fwmark);
		list_add(&svc->f_list, &ip_vs_svc_fwm_table[hash]);
	}

	svc->flags |= IP_VS_SVC_F_HASHED;
	/* increase its refcnt because it is referenced by the svc table */
	atomic_inc(&svc->refcnt);
	return 1;
}


/*
 *	Unhashes a service from ip_vs_svc_table/ip_vs_svc_fwm_table.
 *	Should be called with locked tables.
 */
static int ip_vs_svc_unhash(struct ip_vs_service *svc)
{
	if (!(svc->flags & IP_VS_SVC_F_HASHED)) {
		IP_VS_ERR("ip_vs_svc_unhash(): request for unhash flagged, "
			  "called from %p\n", __builtin_return_address(0));
		return 0;
	}

	if (svc->fwmark == 0) {
		/* Remove it from the ip_vs_svc_table table */
		list_del(&svc->s_list);
	} else {
		/* Remove it from the ip_vs_svc_fwm_table table */
		list_del(&svc->f_list);
	}

	svc->flags &= ~IP_VS_SVC_F_HASHED;
	atomic_dec(&svc->refcnt);
	return 1;
}


/*
 *	Get service by {proto,addr,port} in the service table.
 */
static __inline__ struct ip_vs_service *
__ip_vs_service_get(__u16 protocol, __u32 vaddr, __u16 vport)
{
	unsigned hash;
	struct ip_vs_service *svc;

	/* Check for "full" addressed entries */
	hash = ip_vs_svc_hashkey(protocol, vaddr, vport);

	list_for_each_entry(svc, &ip_vs_svc_table[hash], s_list){
		if ((svc->addr == vaddr)
		    && (svc->port == vport)
		    && (svc->protocol == protocol)) {
			/* HIT */
			atomic_inc(&svc->usecnt);
			return svc;
		}
	}

	return NULL;
}


/*
 *	Get service by {fwmark} in the service table.
 */
static __inline__ struct ip_vs_service *__ip_vs_svc_fwm_get(__u32 fwmark)
{
	unsigned hash;
	struct ip_vs_service *svc;

	/* Check for fwmark addressed entries */
	hash = ip_vs_svc_fwm_hashkey(fwmark);

	list_for_each_entry(svc, &ip_vs_svc_fwm_table[hash], f_list) {
		if (svc->fwmark == fwmark) {
			/* HIT */
			atomic_inc(&svc->usecnt);
			return svc;
		}
	}

	return NULL;
}

struct ip_vs_service *
ip_vs_service_get(__u32 fwmark, __u16 protocol, __u32 vaddr, __u16 vport)
{
	struct ip_vs_service *svc;

	read_lock(&__ip_vs_svc_lock);

	/*
	 *	Check the table hashed by fwmark first
	 */
	if (fwmark && (svc = __ip_vs_svc_fwm_get(fwmark)))
		goto out;

	/*
	 *	Check the table hashed by <protocol,addr,port>
	 *	for "full" addressed entries
	 */
	svc = __ip_vs_service_get(protocol, vaddr, vport);

	if (svc == NULL
	    && protocol == IPPROTO_TCP
	    && atomic_read(&ip_vs_ftpsvc_counter)
	    && (vport == FTPDATA || ntohs(vport) >= PROT_SOCK)) {
		/*
		 * Check if ftp service entry exists, the packet
		 * might belong to FTP data connections.
		 */
		svc = __ip_vs_service_get(protocol, vaddr, FTPPORT);
	}

	if (svc == NULL
	    && atomic_read(&ip_vs_nullsvc_counter)) {
		/*
		 * Check if the catch-all port (port zero) exists
		 */
		svc = __ip_vs_service_get(protocol, vaddr, 0);
	}

  out:
	read_unlock(&__ip_vs_svc_lock);

	IP_VS_DBG(6, "lookup service: fwm %u %s %u.%u.%u.%u:%u %s\n",
		  fwmark, ip_vs_proto_name(protocol),
		  NIPQUAD(vaddr), ntohs(vport),
		  svc?"hit":"not hit");

	return svc;
}


static inline void
__ip_vs_bind_svc(struct ip_vs_dest *dest, struct ip_vs_service *svc)
{
	atomic_inc(&svc->refcnt);
	dest->svc = svc;
}

static inline void
__ip_vs_unbind_svc(struct ip_vs_dest *dest)
{
	struct ip_vs_service *svc = dest->svc;

	dest->svc = NULL;
	if (atomic_dec_and_test(&svc->refcnt))
		kfree(svc);
}


/*
 *	Returns hash value for real service
 */
static __inline__ unsigned ip_vs_rs_hashkey(__u32 addr, __u16 port)
{
	register unsigned porth = ntohs(port);

	return (ntohl(addr)^(porth>>IP_VS_RTAB_BITS)^porth)
		& IP_VS_RTAB_MASK;
}

/*
 *	Hashes ip_vs_dest in ip_vs_rtable by <proto,addr,port>.
 *	should be called with locked tables.
 */
static int ip_vs_rs_hash(struct ip_vs_dest *dest)
{
	unsigned hash;

	if (!list_empty(&dest->d_list)) {
		return 0;
	}

	/*
	 *	Hash by proto,addr,port,
	 *	which are the parameters of the real service.
	 */
	hash = ip_vs_rs_hashkey(dest->addr, dest->port);
	list_add(&dest->d_list, &ip_vs_rtable[hash]);

	return 1;
}

/*
 *	UNhashes ip_vs_dest from ip_vs_rtable.
 *	should be called with locked tables.
 */
static int ip_vs_rs_unhash(struct ip_vs_dest *dest)
{
	/*
	 * Remove it from the ip_vs_rtable table.
	 */
	if (!list_empty(&dest->d_list)) {
		list_del(&dest->d_list);
		INIT_LIST_HEAD(&dest->d_list);
	}

	return 1;
}

/*
 *	Lookup real service by <proto,addr,port> in the real service table.
 */
struct ip_vs_dest *
ip_vs_lookup_real_service(__u16 protocol, __u32 daddr, __u16 dport)
{
	unsigned hash;
	struct ip_vs_dest *dest;

	/*
	 *	Check for "full" addressed entries
	 *	Return the first found entry
	 */
	hash = ip_vs_rs_hashkey(daddr, dport);

	read_lock(&__ip_vs_rs_lock);
	list_for_each_entry(dest, &ip_vs_rtable[hash], d_list) {
		if ((dest->addr == daddr)
		    && (dest->port == dport)
		    && ((dest->protocol == protocol) ||
			dest->vfwmark)) {
			/* HIT */
			read_unlock(&__ip_vs_rs_lock);
			return dest;
		}
	}
	read_unlock(&__ip_vs_rs_lock);

	return NULL;
}

/*
 *	Lookup destination by {addr,port} in the given service
 */
static struct ip_vs_dest *
ip_vs_lookup_dest(struct ip_vs_service *svc, __u32 daddr, __u16 dport)
{
	struct ip_vs_dest *dest;

	/*
	 * Find the destination for the given service
	 */
	list_for_each_entry(dest, &svc->destinations, n_list) {
		if ((dest->addr == daddr) && (dest->port == dport)) {
			/* HIT */
			return dest;
		}
	}

	return NULL;
}


/*
 *  Lookup dest by {svc,addr,port} in the destination trash.
 *  The destination trash is used to hold the destinations that are removed
 *  from the service table but are still referenced by some conn entries.
 *  The reason to add the destination trash is when the dest is temporary
 *  down (either by administrator or by monitor program), the dest can be
 *  picked back from the trash, the remaining connections to the dest can
 *  continue, and the counting information of the dest is also useful for
 *  scheduling.
 */
static struct ip_vs_dest *
ip_vs_trash_get_dest(struct ip_vs_service *svc, __u32 daddr, __u16 dport)
{
	struct ip_vs_dest *dest, *nxt;

	/*
	 * Find the destination in trash
	 */
	list_for_each_entry_safe(dest, nxt, &ip_vs_dest_trash, n_list) {
		IP_VS_DBG(3, "Destination %u/%u.%u.%u.%u:%u still in trash, "
			  "refcnt=%d\n",