ipt_clusterip.c

linux 内核源代码

	if (e->ip.dmsk.s_addr != htonl(0xffffffff)
	    || e->ip.dst.s_addr == 0) {
		printk(KERN_ERR "CLUSTERIP: Please specify destination IP\n");
		return false;
	}

	/* FIXME: further sanity checks */

	config = clusterip_config_find_get(e->ip.dst.s_addr, 1);
	if (!config) {
		if (!(cipinfo->flags & CLUSTERIP_FLAG_NEW)) {
			printk(KERN_WARNING "CLUSTERIP: no config found for %u.%u.%u.%u, need 'new'\n", NIPQUAD(e->ip.dst.s_addr));
			return false;
		} else {
			struct net_device *dev;

			if (e->ip.iniface[0] == '\0') {
				printk(KERN_WARNING "CLUSTERIP: Please specify an interface name\n");
				return false;
			}

			dev = dev_get_by_name(&init_net, e->ip.iniface);
			if (!dev) {
				printk(KERN_WARNING "CLUSTERIP: no such interface %s\n", e->ip.iniface);
				return false;
			}

			config = clusterip_config_init(cipinfo,
							e->ip.dst.s_addr, dev);
			if (!config) {
				printk(KERN_WARNING "CLUSTERIP: cannot allocate config\n");
				dev_put(dev);
				return false;
			}
			dev_mc_add(config->dev,config->clustermac, ETH_ALEN, 0);
		}
	}
	cipinfo->config = config;

	if (nf_ct_l3proto_try_module_get(target->family) < 0) {
		printk(KERN_WARNING "can't load conntrack support for "
				    "proto=%d\n", target->family);
		return false;
	}

	return true;
}

/* drop reference count of cluster config when rule is deleted */
static void destroy(const struct xt_target *target, void *targinfo)
{
	struct ipt_clusterip_tgt_info *cipinfo = targinfo;

	/* if no more entries are referencing the config, remove it
	 * from the list and destroy the proc entry */
	clusterip_config_entry_put(cipinfo->config);

	clusterip_config_put(cipinfo->config);

	nf_ct_l3proto_module_put(target->family);
}

#ifdef CONFIG_COMPAT
struct compat_ipt_clusterip_tgt_info
{
	u_int32_t	flags;
	u_int8_t	clustermac[6];
	u_int16_t	num_total_nodes;
	u_int16_t	num_local_nodes;
	u_int16_t	local_nodes[CLUSTERIP_MAX_NODES];
	u_int32_t	hash_mode;
	u_int32_t	hash_initval;
	compat_uptr_t	config;
};
#endif /* CONFIG_COMPAT */

static struct xt_target clusterip_tgt __read_mostly = {
	.name		= "CLUSTERIP",
	.family		= AF_INET,
	.target		= target,
	.checkentry	= checkentry,
	.destroy	= destroy,
	.targetsize	= sizeof(struct ipt_clusterip_tgt_info),
#ifdef CONFIG_COMPAT
	.compatsize	= sizeof(struct compat_ipt_clusterip_tgt_info),
#endif /* CONFIG_COMPAT */
	.me		= THIS_MODULE
};


/***********************************************************************
 * ARP MANGLING CODE
 ***********************************************************************/

/* hardcoded for 48bit ethernet and 32bit ipv4 addresses */
struct arp_payload {
	u_int8_t src_hw[ETH_ALEN];
	__be32 src_ip;
	u_int8_t dst_hw[ETH_ALEN];
	__be32 dst_ip;
} __attribute__ ((packed));

#ifdef DEBUG
static void arp_print(struct arp_payload *payload)
{
#define HBUFFERLEN 30
	char hbuffer[HBUFFERLEN];
	int j,k;
	const char hexbuf[]= "0123456789abcdef";

	for (k=0, j=0; k < HBUFFERLEN-3 && j < ETH_ALEN; j++) {
		hbuffer[k++]=hexbuf[(payload->src_hw[j]>>4)&15];
		hbuffer[k++]=hexbuf[payload->src_hw[j]&15];
		hbuffer[k++]=':';
	}
	hbuffer[--k]='\0';

	printk("src %u.%u.%u.%u@%s, dst %u.%u.%u.%u\n",
		NIPQUAD(payload->src_ip), hbuffer,
		NIPQUAD(payload->dst_ip));
}
#endif

static unsigned int
arp_mangle(unsigned int hook,
	   struct sk_buff *skb,
	   const struct net_device *in,
	   const struct net_device *out,
	   int (*okfn)(struct sk_buff *))
{
	struct arphdr *arp = arp_hdr(skb);
	struct arp_payload *payload;
	struct clusterip_config *c;

	/* we don't care about non-ethernet and non-ipv4 ARP */
	if (arp->ar_hrd != htons(ARPHRD_ETHER)
	    || arp->ar_pro != htons(ETH_P_IP)
	    || arp->ar_pln != 4 || arp->ar_hln != ETH_ALEN)
		return NF_ACCEPT;

	/* we only want to mangle arp requests and replies */
	if (arp->ar_op != htons(ARPOP_REPLY)
	    && arp->ar_op != htons(ARPOP_REQUEST))
		return NF_ACCEPT;

	payload = (void *)(arp+1);

	/* if there is no clusterip configuration for the arp reply's
	 * source ip, we don't want to mangle it */
	c = clusterip_config_find_get(payload->src_ip, 0);
	if (!c)
		return NF_ACCEPT;

	/* normally the linux kernel always replies to arp queries of
	 * addresses on different interfacs.  However, in the CLUSTERIP case
	 * this wouldn't work, since we didn't subscribe the mcast group on
	 * other interfaces */
	if (c->dev != out) {
		pr_debug("CLUSTERIP: not mangling arp reply on different "
			 "interface: cip'%s'-skb'%s'\n",
			 c->dev->name, out->name);
		clusterip_config_put(c);
		return NF_ACCEPT;
	}

	/* mangle reply hardware address */
	memcpy(payload->src_hw, c->clustermac, arp->ar_hln);

#ifdef DEBUG
	pr_debug(KERN_DEBUG "CLUSTERIP mangled arp reply: ");
	arp_print(payload);
#endif

	clusterip_config_put(c);

	return NF_ACCEPT;
}

static struct nf_hook_ops cip_arp_ops = {
	.hook = arp_mangle,
	.pf = NF_ARP,
	.hooknum = NF_ARP_OUT,
	.priority = -1
};

/***********************************************************************
 * PROC DIR HANDLING
 ***********************************************************************/

#ifdef CONFIG_PROC_FS

struct clusterip_seq_position {
	unsigned int pos;	/* position */
	unsigned int weight;	/* number of bits set == size */
	unsigned int bit;	/* current bit */
	unsigned long val;	/* current value */
};

static void *clusterip_seq_start(struct seq_file *s, loff_t *pos)
{
	struct proc_dir_entry *pde = s->private;
	struct clusterip_config *c = pde->data;
	unsigned int weight;
	u_int32_t local_nodes;
	struct clusterip_seq_position *idx;

	/* FIXME: possible race */
	local_nodes = c->local_nodes;
	weight = hweight32(local_nodes);
	if (*pos >= weight)
		return NULL;

	idx = kmalloc(sizeof(struct clusterip_seq_position), GFP_KERNEL);
	if (!idx)
		return ERR_PTR(-ENOMEM);

	idx->pos = *pos;
	idx->weight = weight;
	idx->bit = ffs(local_nodes);
	idx->val = local_nodes;
	clear_bit(idx->bit - 1, &idx->val);

	return idx;
}

static void *clusterip_seq_next(struct seq_file *s, void *v, loff_t *pos)
{
	struct clusterip_seq_position *idx = (struct clusterip_seq_position *)v;

	*pos = ++idx->pos;
	if (*pos >= idx->weight) {
		kfree(v);
		return NULL;
	}
	idx->bit = ffs(idx->val);
	clear_bit(idx->bit - 1, &idx->val);
	return idx;
}

static void clusterip_seq_stop(struct seq_file *s, void *v)
{
	kfree(v);
}

static int clusterip_seq_show(struct seq_file *s, void *v)
{
	struct clusterip_seq_position *idx = (struct clusterip_seq_position *)v;

	if (idx->pos != 0)
		seq_putc(s, ',');

	seq_printf(s, "%u", idx->bit);

	if (idx->pos == idx->weight - 1)
		seq_putc(s, '\n');

	return 0;
}

static const struct seq_operations clusterip_seq_ops = {
	.start	= clusterip_seq_start,
	.next	= clusterip_seq_next,
	.stop	= clusterip_seq_stop,
	.show	= clusterip_seq_show,
};

static int clusterip_proc_open(struct inode *inode, struct file *file)
{
	int ret = seq_open(file, &clusterip_seq_ops);

	if (!ret) {
		struct seq_file *sf = file->private_data;
		struct proc_dir_entry *pde = PDE(inode);
		struct clusterip_config *c = pde->data;

		sf->private = pde;

		clusterip_config_get(c);
	}

	return ret;
}

static int clusterip_proc_release(struct inode *inode, struct file *file)
{
	struct proc_dir_entry *pde = PDE(inode);
	struct clusterip_config *c = pde->data;
	int ret;

	ret = seq_release(inode, file);

	if (!ret)
		clusterip_config_put(c);

	return ret;
}

static ssize_t clusterip_proc_write(struct file *file, const char __user *input,
				size_t size, loff_t *ofs)
{
#define PROC_WRITELEN	10
	char buffer[PROC_WRITELEN+1];
	struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode);
	struct clusterip_config *c = pde->data;
	unsigned long nodenum;

	if (copy_from_user(buffer, input, PROC_WRITELEN))
		return -EFAULT;

	if (*buffer == '+') {
		nodenum = simple_strtoul(buffer+1, NULL, 10);
		if (clusterip_add_node(c, nodenum))
			return -ENOMEM;
	} else if (*buffer == '-') {
		nodenum = simple_strtoul(buffer+1, NULL,10);
		if (clusterip_del_node(c, nodenum))
			return -ENOENT;
	} else
		return -EIO;

	return size;
}

static const struct file_operations clusterip_proc_fops = {
	.owner	 = THIS_MODULE,
	.open	 = clusterip_proc_open,
	.read	 = seq_read,
	.write	 = clusterip_proc_write,
	.llseek	 = seq_lseek,
	.release = clusterip_proc_release,
};

#endif /* CONFIG_PROC_FS */

static int __init ipt_clusterip_init(void)
{
	int ret;

	ret = xt_register_target(&clusterip_tgt);
	if (ret < 0)
		return ret;

	ret = nf_register_hook(&cip_arp_ops);
	if (ret < 0)
		goto cleanup_target;

#ifdef CONFIG_PROC_FS
	clusterip_procdir = proc_mkdir("ipt_CLUSTERIP", init_net.proc_net);
	if (!clusterip_procdir) {
		printk(KERN_ERR "CLUSTERIP: Unable to proc dir entry\n");
		ret = -ENOMEM;
		goto cleanup_hook;
	}
#endif /* CONFIG_PROC_FS */

	printk(KERN_NOTICE "ClusterIP Version %s loaded successfully\n",
		CLUSTERIP_VERSION);
	return 0;

#ifdef CONFIG_PROC_FS
cleanup_hook:
	nf_unregister_hook(&cip_arp_ops);
#endif /* CONFIG_PROC_FS */
cleanup_target:
	xt_unregister_target(&clusterip_tgt);
	return ret;
}

static void __exit ipt_clusterip_fini(void)
{
	printk(KERN_NOTICE "ClusterIP Version %s unloading\n",
		CLUSTERIP_VERSION);
#ifdef CONFIG_PROC_FS
	remove_proc_entry(clusterip_procdir->name, clusterip_procdir->parent);
#endif
	nf_unregister_hook(&cip_arp_ops);
	xt_unregister_target(&clusterip_tgt);
}

module_init(ipt_clusterip_init);
module_exit(ipt_clusterip_fini);