YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 1 | /* Cluster IP hashmark target |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2 | * (C) 2003-2004 by Harald Welte <laforge@netfilter.org> |
| 3 | * based on ideas of Fabio Olive Leite <olive@unixforge.org> |
| 4 | * |
| 5 | * Development of this code funded by SuSE Linux AG, http://www.suse.com/ |
| 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 version 2 as |
| 9 | * published by the Free Software Foundation. |
| 10 | * |
| 11 | */ |
| 12 | #include <linux/module.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 13 | #include <linux/proc_fs.h> |
| 14 | #include <linux/jhash.h> |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 15 | #include <linux/bitops.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 16 | #include <linux/skbuff.h> |
| 17 | #include <linux/ip.h> |
| 18 | #include <linux/tcp.h> |
| 19 | #include <linux/udp.h> |
| 20 | #include <linux/icmp.h> |
| 21 | #include <linux/if_arp.h> |
| 22 | #include <linux/proc_fs.h> |
| 23 | #include <linux/seq_file.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | #include <linux/netfilter_arp.h> |
Jan Engelhardt | 6709dbb | 2007-02-07 15:11:19 -0800 | [diff] [blame] | 25 | #include <linux/netfilter/x_tables.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 26 | #include <linux/netfilter_ipv4/ip_tables.h> |
| 27 | #include <linux/netfilter_ipv4/ipt_CLUSTERIP.h> |
Patrick McHardy | 587aa64 | 2007-03-14 16:37:25 -0700 | [diff] [blame] | 28 | #include <net/netfilter/nf_conntrack.h> |
| 29 | #include <net/checksum.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 30 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 31 | #define CLUSTERIP_VERSION "0.8" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 32 | |
| 33 | #define DEBUG_CLUSTERIP |
| 34 | |
| 35 | #ifdef DEBUG_CLUSTERIP |
| 36 | #define DEBUGP printk |
| 37 | #else |
| 38 | #define DEBUGP |
| 39 | #endif |
| 40 | |
| 41 | MODULE_LICENSE("GPL"); |
| 42 | MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>"); |
| 43 | MODULE_DESCRIPTION("iptables target for CLUSTERIP"); |
| 44 | |
| 45 | struct clusterip_config { |
| 46 | struct list_head list; /* list of all configs */ |
| 47 | atomic_t refcount; /* reference count */ |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 48 | atomic_t entries; /* number of entries/rules |
| 49 | * referencing us */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 50 | |
Al Viro | 6a19d61 | 2006-09-28 14:22:24 -0700 | [diff] [blame] | 51 | __be32 clusterip; /* the IP address */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 52 | u_int8_t clustermac[ETH_ALEN]; /* the MAC address */ |
| 53 | struct net_device *dev; /* device */ |
| 54 | u_int16_t num_total_nodes; /* total number of nodes */ |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 55 | unsigned long local_nodes; /* node number array */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 56 | |
| 57 | #ifdef CONFIG_PROC_FS |
| 58 | struct proc_dir_entry *pde; /* proc dir entry */ |
| 59 | #endif |
| 60 | enum clusterip_hashmode hash_mode; /* which hashing mode */ |
| 61 | u_int32_t hash_initval; /* hash initialization */ |
| 62 | }; |
| 63 | |
| 64 | static LIST_HEAD(clusterip_configs); |
| 65 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 66 | /* clusterip_lock protects the clusterip_configs list */ |
Patrick McHardy | e45b1be | 2005-06-21 14:01:30 -0700 | [diff] [blame] | 67 | static DEFINE_RWLOCK(clusterip_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 68 | |
| 69 | #ifdef CONFIG_PROC_FS |
Arjan van de Ven | 9a32144 | 2007-02-12 00:55:35 -0800 | [diff] [blame] | 70 | static const struct file_operations clusterip_proc_fops; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 71 | static struct proc_dir_entry *clusterip_procdir; |
| 72 | #endif |
| 73 | |
| 74 | static inline void |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 75 | clusterip_config_get(struct clusterip_config *c) |
| 76 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 77 | atomic_inc(&c->refcount); |
| 78 | } |
| 79 | |
| 80 | static inline void |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 81 | clusterip_config_put(struct clusterip_config *c) |
| 82 | { |
| 83 | if (atomic_dec_and_test(&c->refcount)) |
| 84 | kfree(c); |
| 85 | } |
| 86 | |
| 87 | /* increase the count of entries(rules) using/referencing this config */ |
| 88 | static inline void |
| 89 | clusterip_config_entry_get(struct clusterip_config *c) |
| 90 | { |
| 91 | atomic_inc(&c->entries); |
| 92 | } |
| 93 | |
| 94 | /* decrease the count of entries using/referencing this config. If last |
| 95 | * entry(rule) is removed, remove the config from lists, but don't free it |
| 96 | * yet, since proc-files could still be holding references */ |
| 97 | static inline void |
| 98 | clusterip_config_entry_put(struct clusterip_config *c) |
| 99 | { |
| 100 | if (atomic_dec_and_test(&c->entries)) { |
Patrick McHardy | e45b1be | 2005-06-21 14:01:30 -0700 | [diff] [blame] | 101 | write_lock_bh(&clusterip_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 102 | list_del(&c->list); |
Patrick McHardy | e45b1be | 2005-06-21 14:01:30 -0700 | [diff] [blame] | 103 | write_unlock_bh(&clusterip_lock); |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 104 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 105 | dev_mc_delete(c->dev, c->clustermac, ETH_ALEN, 0); |
| 106 | dev_put(c->dev); |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 107 | |
| 108 | /* In case anyone still accesses the file, the open/close |
| 109 | * functions are also incrementing the refcount on their own, |
| 110 | * so it's safe to remove the entry even if it's in use. */ |
| 111 | #ifdef CONFIG_PROC_FS |
| 112 | remove_proc_entry(c->pde->name, c->pde->parent); |
| 113 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 114 | } |
| 115 | } |
| 116 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 117 | static struct clusterip_config * |
Al Viro | 6a19d61 | 2006-09-28 14:22:24 -0700 | [diff] [blame] | 118 | __clusterip_config_find(__be32 clusterip) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 119 | { |
| 120 | struct list_head *pos; |
| 121 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 122 | list_for_each(pos, &clusterip_configs) { |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 123 | struct clusterip_config *c = list_entry(pos, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 124 | struct clusterip_config, list); |
Jan Engelhardt | 7c4e36b | 2007-07-07 22:19:08 -0700 | [diff] [blame] | 125 | if (c->clusterip == clusterip) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 126 | return c; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 127 | } |
| 128 | |
| 129 | return NULL; |
| 130 | } |
| 131 | |
| 132 | static inline struct clusterip_config * |
Al Viro | 6a19d61 | 2006-09-28 14:22:24 -0700 | [diff] [blame] | 133 | clusterip_config_find_get(__be32 clusterip, int entry) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 134 | { |
| 135 | struct clusterip_config *c; |
| 136 | |
Patrick McHardy | e45b1be | 2005-06-21 14:01:30 -0700 | [diff] [blame] | 137 | read_lock_bh(&clusterip_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 138 | c = __clusterip_config_find(clusterip); |
| 139 | if (!c) { |
Patrick McHardy | e45b1be | 2005-06-21 14:01:30 -0700 | [diff] [blame] | 140 | read_unlock_bh(&clusterip_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 141 | return NULL; |
| 142 | } |
| 143 | atomic_inc(&c->refcount); |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 144 | if (entry) |
| 145 | atomic_inc(&c->entries); |
Patrick McHardy | e45b1be | 2005-06-21 14:01:30 -0700 | [diff] [blame] | 146 | read_unlock_bh(&clusterip_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 147 | |
| 148 | return c; |
| 149 | } |
| 150 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 151 | static void |
| 152 | clusterip_config_init_nodelist(struct clusterip_config *c, |
| 153 | const struct ipt_clusterip_tgt_info *i) |
| 154 | { |
| 155 | int n; |
| 156 | |
Jan Engelhardt | 7c4e36b | 2007-07-07 22:19:08 -0700 | [diff] [blame] | 157 | for (n = 0; n < i->num_local_nodes; n++) |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 158 | set_bit(i->local_nodes[n] - 1, &c->local_nodes); |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 159 | } |
| 160 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 161 | static struct clusterip_config * |
Al Viro | 6a19d61 | 2006-09-28 14:22:24 -0700 | [diff] [blame] | 162 | clusterip_config_init(struct ipt_clusterip_tgt_info *i, __be32 ip, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 163 | struct net_device *dev) |
| 164 | { |
| 165 | struct clusterip_config *c; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 166 | |
Panagiotis Issaris | 0da974f | 2006-07-21 14:51:30 -0700 | [diff] [blame] | 167 | c = kzalloc(sizeof(*c), GFP_ATOMIC); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 168 | if (!c) |
| 169 | return NULL; |
| 170 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 171 | c->dev = dev; |
| 172 | c->clusterip = ip; |
| 173 | memcpy(&c->clustermac, &i->clustermac, ETH_ALEN); |
| 174 | c->num_total_nodes = i->num_total_nodes; |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 175 | clusterip_config_init_nodelist(c, i); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 176 | c->hash_mode = i->hash_mode; |
| 177 | c->hash_initval = i->hash_initval; |
| 178 | atomic_set(&c->refcount, 1); |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 179 | atomic_set(&c->entries, 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 180 | |
| 181 | #ifdef CONFIG_PROC_FS |
Patrick McHardy | 7659258 | 2006-11-29 02:35:42 +0100 | [diff] [blame] | 182 | { |
| 183 | char buffer[16]; |
| 184 | |
| 185 | /* create proc dir entry */ |
| 186 | sprintf(buffer, "%u.%u.%u.%u", NIPQUAD(ip)); |
| 187 | c->pde = create_proc_entry(buffer, S_IWUSR|S_IRUSR, |
| 188 | clusterip_procdir); |
| 189 | if (!c->pde) { |
| 190 | kfree(c); |
| 191 | return NULL; |
| 192 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 193 | } |
| 194 | c->pde->proc_fops = &clusterip_proc_fops; |
| 195 | c->pde->data = c; |
| 196 | #endif |
| 197 | |
Patrick McHardy | e45b1be | 2005-06-21 14:01:30 -0700 | [diff] [blame] | 198 | write_lock_bh(&clusterip_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 199 | list_add(&c->list, &clusterip_configs); |
Patrick McHardy | e45b1be | 2005-06-21 14:01:30 -0700 | [diff] [blame] | 200 | write_unlock_bh(&clusterip_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 201 | |
| 202 | return c; |
| 203 | } |
| 204 | |
Patrick McHardy | 7659258 | 2006-11-29 02:35:42 +0100 | [diff] [blame] | 205 | #ifdef CONFIG_PROC_FS |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 206 | static int |
| 207 | clusterip_add_node(struct clusterip_config *c, u_int16_t nodenum) |
| 208 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 209 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 210 | if (nodenum == 0 || |
| 211 | nodenum > c->num_total_nodes) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 212 | return 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 213 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 214 | /* check if we already have this number in our bitfield */ |
| 215 | if (test_and_set_bit(nodenum - 1, &c->local_nodes)) |
| 216 | return 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 217 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 218 | return 0; |
| 219 | } |
| 220 | |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 221 | static bool |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 222 | clusterip_del_node(struct clusterip_config *c, u_int16_t nodenum) |
| 223 | { |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 224 | if (nodenum == 0 || |
| 225 | nodenum > c->num_total_nodes) |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 226 | return true; |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 227 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 228 | if (test_and_clear_bit(nodenum - 1, &c->local_nodes)) |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 229 | return false; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 230 | |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 231 | return true; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 232 | } |
Patrick McHardy | 7659258 | 2006-11-29 02:35:42 +0100 | [diff] [blame] | 233 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 234 | |
| 235 | static inline u_int32_t |
Jan Engelhardt | a47362a | 2007-07-07 22:16:55 -0700 | [diff] [blame] | 236 | clusterip_hashfn(const struct sk_buff *skb, |
| 237 | const struct clusterip_config *config) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 238 | { |
Jan Engelhardt | a47362a | 2007-07-07 22:16:55 -0700 | [diff] [blame] | 239 | const struct iphdr *iph = ip_hdr(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 240 | unsigned long hashval; |
| 241 | u_int16_t sport, dport; |
Jan Engelhardt | a47362a | 2007-07-07 22:16:55 -0700 | [diff] [blame] | 242 | const u_int16_t *ports; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 243 | |
| 244 | switch (iph->protocol) { |
| 245 | case IPPROTO_TCP: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 246 | case IPPROTO_UDP: |
Patrick McHardy | a8d0f95 | 2007-02-07 15:07:43 -0800 | [diff] [blame] | 247 | case IPPROTO_UDPLITE: |
Patrick McHardy | 957dc80 | 2006-05-29 18:19:56 -0700 | [diff] [blame] | 248 | case IPPROTO_SCTP: |
| 249 | case IPPROTO_DCCP: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 250 | case IPPROTO_ICMP: |
Jan Engelhardt | a47362a | 2007-07-07 22:16:55 -0700 | [diff] [blame] | 251 | ports = (const void *)iph+iph->ihl*4; |
Patrick McHardy | 957dc80 | 2006-05-29 18:19:56 -0700 | [diff] [blame] | 252 | sport = ports[0]; |
| 253 | dport = ports[1]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 254 | break; |
| 255 | default: |
Jan Engelhardt | 7c4e36b | 2007-07-07 22:19:08 -0700 | [diff] [blame] | 256 | if (net_ratelimit()) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 257 | printk(KERN_NOTICE "CLUSTERIP: unknown protocol `%u'\n", |
| 258 | iph->protocol); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 259 | sport = dport = 0; |
| 260 | } |
| 261 | |
| 262 | switch (config->hash_mode) { |
| 263 | case CLUSTERIP_HASHMODE_SIP: |
| 264 | hashval = jhash_1word(ntohl(iph->saddr), |
| 265 | config->hash_initval); |
| 266 | break; |
| 267 | case CLUSTERIP_HASHMODE_SIP_SPT: |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 268 | hashval = jhash_2words(ntohl(iph->saddr), sport, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 269 | config->hash_initval); |
| 270 | break; |
| 271 | case CLUSTERIP_HASHMODE_SIP_SPT_DPT: |
| 272 | hashval = jhash_3words(ntohl(iph->saddr), sport, dport, |
| 273 | config->hash_initval); |
| 274 | break; |
| 275 | default: |
| 276 | /* to make gcc happy */ |
| 277 | hashval = 0; |
| 278 | /* This cannot happen, unless the check function wasn't called |
| 279 | * at rule load time */ |
| 280 | printk("CLUSTERIP: unknown mode `%u'\n", config->hash_mode); |
| 281 | BUG(); |
| 282 | break; |
| 283 | } |
| 284 | |
| 285 | /* node numbers are 1..n, not 0..n */ |
Jan Engelhardt | 7c4e36b | 2007-07-07 22:19:08 -0700 | [diff] [blame] | 286 | return (hashval % config->num_total_nodes) + 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 287 | } |
| 288 | |
| 289 | static inline int |
Jan Engelhardt | a47362a | 2007-07-07 22:16:55 -0700 | [diff] [blame] | 290 | clusterip_responsible(const struct clusterip_config *config, u_int32_t hash) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 291 | { |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 292 | return test_bit(hash - 1, &config->local_nodes); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 293 | } |
| 294 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 295 | /*********************************************************************** |
| 296 | * IPTABLES TARGET |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 297 | ***********************************************************************/ |
| 298 | |
| 299 | static unsigned int |
| 300 | target(struct sk_buff **pskb, |
| 301 | const struct net_device *in, |
| 302 | const struct net_device *out, |
| 303 | unsigned int hooknum, |
Patrick McHardy | c498673 | 2006-03-20 18:02:56 -0800 | [diff] [blame] | 304 | const struct xt_target *target, |
Patrick McHardy | fe1cb10 | 2006-08-22 00:35:47 -0700 | [diff] [blame] | 305 | const void *targinfo) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 306 | { |
| 307 | const struct ipt_clusterip_tgt_info *cipinfo = targinfo; |
Patrick McHardy | 587aa64 | 2007-03-14 16:37:25 -0700 | [diff] [blame] | 308 | struct nf_conn *ct; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 309 | enum ip_conntrack_info ctinfo; |
Patrick McHardy | 587aa64 | 2007-03-14 16:37:25 -0700 | [diff] [blame] | 310 | u_int32_t hash; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 311 | |
| 312 | /* don't need to clusterip_config_get() here, since refcount |
| 313 | * is only decremented by destroy() - and ip_tables guarantees |
| 314 | * that the ->target() function isn't called after ->destroy() */ |
| 315 | |
Patrick McHardy | 587aa64 | 2007-03-14 16:37:25 -0700 | [diff] [blame] | 316 | ct = nf_ct_get(*pskb, &ctinfo); |
| 317 | if (ct == NULL) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 318 | printk(KERN_ERR "CLUSTERIP: no conntrack!\n"); |
| 319 | /* FIXME: need to drop invalid ones, since replies |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 320 | * to outgoing connections of other nodes will be |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 321 | * marked as INVALID */ |
| 322 | return NF_DROP; |
| 323 | } |
| 324 | |
| 325 | /* special case: ICMP error handling. conntrack distinguishes between |
| 326 | * error messages (RELATED) and information requests (see below) */ |
Arnaldo Carvalho de Melo | eddc9ec | 2007-04-20 22:47:35 -0700 | [diff] [blame] | 327 | if (ip_hdr(*pskb)->protocol == IPPROTO_ICMP |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 328 | && (ctinfo == IP_CT_RELATED |
Harald Welte | 5d927eb | 2005-06-22 12:37:50 -0700 | [diff] [blame] | 329 | || ctinfo == IP_CT_RELATED+IP_CT_IS_REPLY)) |
Jan Engelhardt | 6709dbb | 2007-02-07 15:11:19 -0800 | [diff] [blame] | 330 | return XT_CONTINUE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 331 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 332 | /* ip_conntrack_icmp guarantees us that we only have ICMP_ECHO, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 333 | * TIMESTAMP, INFO_REQUEST or ADDRESS type icmp packets from here |
| 334 | * on, which all have an ID field [relevant for hashing]. */ |
| 335 | |
| 336 | hash = clusterip_hashfn(*pskb, cipinfo->config); |
| 337 | |
| 338 | switch (ctinfo) { |
| 339 | case IP_CT_NEW: |
Patrick McHardy | 587aa64 | 2007-03-14 16:37:25 -0700 | [diff] [blame] | 340 | ct->mark = hash; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 341 | break; |
| 342 | case IP_CT_RELATED: |
| 343 | case IP_CT_RELATED+IP_CT_IS_REPLY: |
| 344 | /* FIXME: we don't handle expectations at the |
| 345 | * moment. they can arrive on a different node than |
| 346 | * the master connection (e.g. FTP passive mode) */ |
| 347 | case IP_CT_ESTABLISHED: |
| 348 | case IP_CT_ESTABLISHED+IP_CT_IS_REPLY: |
| 349 | break; |
| 350 | default: |
| 351 | break; |
| 352 | } |
| 353 | |
| 354 | #ifdef DEBUG_CLUSTERP |
| 355 | DUMP_TUPLE(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); |
| 356 | #endif |
Patrick McHardy | 587aa64 | 2007-03-14 16:37:25 -0700 | [diff] [blame] | 357 | DEBUGP("hash=%u ct_hash=%u ", hash, ct->mark); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 358 | if (!clusterip_responsible(cipinfo->config, hash)) { |
| 359 | DEBUGP("not responsible\n"); |
| 360 | return NF_DROP; |
| 361 | } |
| 362 | DEBUGP("responsible\n"); |
| 363 | |
| 364 | /* despite being received via linklayer multicast, this is |
| 365 | * actually a unicast IP packet. TCP doesn't like PACKET_MULTICAST */ |
| 366 | (*pskb)->pkt_type = PACKET_HOST; |
| 367 | |
Jan Engelhardt | 6709dbb | 2007-02-07 15:11:19 -0800 | [diff] [blame] | 368 | return XT_CONTINUE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 369 | } |
| 370 | |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 371 | static bool |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 372 | checkentry(const char *tablename, |
Harald Welte | 2e4e6a1 | 2006-01-12 13:30:04 -0800 | [diff] [blame] | 373 | const void *e_void, |
Patrick McHardy | c498673 | 2006-03-20 18:02:56 -0800 | [diff] [blame] | 374 | const struct xt_target *target, |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 375 | void *targinfo, |
| 376 | unsigned int hook_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 377 | { |
| 378 | struct ipt_clusterip_tgt_info *cipinfo = targinfo; |
Harald Welte | 2e4e6a1 | 2006-01-12 13:30:04 -0800 | [diff] [blame] | 379 | const struct ipt_entry *e = e_void; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 380 | |
| 381 | struct clusterip_config *config; |
| 382 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 383 | if (cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP && |
| 384 | cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP_SPT && |
| 385 | cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP_SPT_DPT) { |
| 386 | printk(KERN_WARNING "CLUSTERIP: unknown mode `%u'\n", |
| 387 | cipinfo->hash_mode); |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 388 | return false; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 389 | |
| 390 | } |
Al Viro | 6a19d61 | 2006-09-28 14:22:24 -0700 | [diff] [blame] | 391 | if (e->ip.dmsk.s_addr != htonl(0xffffffff) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 392 | || e->ip.dst.s_addr == 0) { |
| 393 | printk(KERN_ERR "CLUSTERIP: Please specify destination IP\n"); |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 394 | return false; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 395 | } |
| 396 | |
| 397 | /* FIXME: further sanity checks */ |
| 398 | |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 399 | config = clusterip_config_find_get(e->ip.dst.s_addr, 1); |
| 400 | if (config) { |
| 401 | if (cipinfo->config != NULL) { |
| 402 | /* Case A: This is an entry that gets reloaded, since |
| 403 | * it still has a cipinfo->config pointer. Simply |
| 404 | * increase the entry refcount and return */ |
| 405 | if (cipinfo->config != config) { |
| 406 | printk(KERN_ERR "CLUSTERIP: Reloaded entry " |
| 407 | "has invalid config pointer!\n"); |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 408 | return false; |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 409 | } |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 410 | } else { |
| 411 | /* Case B: This is a new rule referring to an existing |
| 412 | * clusterip config. */ |
| 413 | cipinfo->config = config; |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 414 | } |
| 415 | } else { |
| 416 | /* Case C: This is a completely new clusterip config */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 417 | if (!(cipinfo->flags & CLUSTERIP_FLAG_NEW)) { |
| 418 | printk(KERN_WARNING "CLUSTERIP: no config found for %u.%u.%u.%u, need 'new'\n", NIPQUAD(e->ip.dst.s_addr)); |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 419 | return false; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 420 | } else { |
| 421 | struct net_device *dev; |
| 422 | |
| 423 | if (e->ip.iniface[0] == '\0') { |
| 424 | printk(KERN_WARNING "CLUSTERIP: Please specify an interface name\n"); |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 425 | return false; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 426 | } |
| 427 | |
| 428 | dev = dev_get_by_name(e->ip.iniface); |
| 429 | if (!dev) { |
| 430 | printk(KERN_WARNING "CLUSTERIP: no such interface %s\n", e->ip.iniface); |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 431 | return false; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 432 | } |
| 433 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 434 | config = clusterip_config_init(cipinfo, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 435 | e->ip.dst.s_addr, dev); |
| 436 | if (!config) { |
| 437 | printk(KERN_WARNING "CLUSTERIP: cannot allocate config\n"); |
| 438 | dev_put(dev); |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 439 | return false; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 440 | } |
| 441 | dev_mc_add(config->dev,config->clustermac, ETH_ALEN, 0); |
| 442 | } |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 443 | cipinfo->config = config; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 444 | } |
| 445 | |
Yasuyuki Kozakai | 11078c3 | 2006-12-12 00:29:02 -0800 | [diff] [blame] | 446 | if (nf_ct_l3proto_try_module_get(target->family) < 0) { |
| 447 | printk(KERN_WARNING "can't load conntrack support for " |
| 448 | "proto=%d\n", target->family); |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 449 | return false; |
Yasuyuki Kozakai | 11078c3 | 2006-12-12 00:29:02 -0800 | [diff] [blame] | 450 | } |
| 451 | |
Jan Engelhardt | e1931b7 | 2007-07-07 22:16:26 -0700 | [diff] [blame] | 452 | return true; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 453 | } |
| 454 | |
| 455 | /* drop reference count of cluster config when rule is deleted */ |
Patrick McHardy | efa7416 | 2006-08-22 00:36:37 -0700 | [diff] [blame] | 456 | static void destroy(const struct xt_target *target, void *targinfo) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 457 | { |
Patrick McHardy | c498673 | 2006-03-20 18:02:56 -0800 | [diff] [blame] | 458 | struct ipt_clusterip_tgt_info *cipinfo = targinfo; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 459 | |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 460 | /* if no more entries are referencing the config, remove it |
| 461 | * from the list and destroy the proc entry */ |
| 462 | clusterip_config_entry_put(cipinfo->config); |
| 463 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 464 | clusterip_config_put(cipinfo->config); |
Yasuyuki Kozakai | 11078c3 | 2006-12-12 00:29:02 -0800 | [diff] [blame] | 465 | |
| 466 | nf_ct_l3proto_module_put(target->family); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 467 | } |
| 468 | |
Patrick McHardy | 9f15c53 | 2007-07-07 22:22:02 -0700 | [diff] [blame^] | 469 | static struct xt_target clusterip_tgt __read_mostly = { |
Patrick McHardy | 1d5cd90 | 2006-03-20 18:01:14 -0800 | [diff] [blame] | 470 | .name = "CLUSTERIP", |
Jan Engelhardt | 6709dbb | 2007-02-07 15:11:19 -0800 | [diff] [blame] | 471 | .family = AF_INET, |
Patrick McHardy | 1d5cd90 | 2006-03-20 18:01:14 -0800 | [diff] [blame] | 472 | .target = target, |
| 473 | .targetsize = sizeof(struct ipt_clusterip_tgt_info), |
| 474 | .checkentry = checkentry, |
| 475 | .destroy = destroy, |
| 476 | .me = THIS_MODULE |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 477 | }; |
| 478 | |
| 479 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 480 | /*********************************************************************** |
| 481 | * ARP MANGLING CODE |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 482 | ***********************************************************************/ |
| 483 | |
| 484 | /* hardcoded for 48bit ethernet and 32bit ipv4 addresses */ |
| 485 | struct arp_payload { |
| 486 | u_int8_t src_hw[ETH_ALEN]; |
Al Viro | 6a19d61 | 2006-09-28 14:22:24 -0700 | [diff] [blame] | 487 | __be32 src_ip; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 488 | u_int8_t dst_hw[ETH_ALEN]; |
Al Viro | 6a19d61 | 2006-09-28 14:22:24 -0700 | [diff] [blame] | 489 | __be32 dst_ip; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 490 | } __attribute__ ((packed)); |
| 491 | |
| 492 | #ifdef CLUSTERIP_DEBUG |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 493 | static void arp_print(struct arp_payload *payload) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 494 | { |
| 495 | #define HBUFFERLEN 30 |
| 496 | char hbuffer[HBUFFERLEN]; |
| 497 | int j,k; |
| 498 | const char hexbuf[]= "0123456789abcdef"; |
| 499 | |
| 500 | for (k=0, j=0; k < HBUFFERLEN-3 && j < ETH_ALEN; j++) { |
| 501 | hbuffer[k++]=hexbuf[(payload->src_hw[j]>>4)&15]; |
| 502 | hbuffer[k++]=hexbuf[payload->src_hw[j]&15]; |
| 503 | hbuffer[k++]=':'; |
| 504 | } |
| 505 | hbuffer[--k]='\0'; |
| 506 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 507 | printk("src %u.%u.%u.%u@%s, dst %u.%u.%u.%u\n", |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 508 | NIPQUAD(payload->src_ip), hbuffer, |
| 509 | NIPQUAD(payload->dst_ip)); |
| 510 | } |
| 511 | #endif |
| 512 | |
| 513 | static unsigned int |
| 514 | arp_mangle(unsigned int hook, |
| 515 | struct sk_buff **pskb, |
| 516 | const struct net_device *in, |
| 517 | const struct net_device *out, |
| 518 | int (*okfn)(struct sk_buff *)) |
| 519 | { |
Arnaldo Carvalho de Melo | d0a92be | 2007-03-12 20:56:31 -0300 | [diff] [blame] | 520 | struct arphdr *arp = arp_hdr(*pskb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 521 | struct arp_payload *payload; |
| 522 | struct clusterip_config *c; |
| 523 | |
| 524 | /* we don't care about non-ethernet and non-ipv4 ARP */ |
| 525 | if (arp->ar_hrd != htons(ARPHRD_ETHER) |
| 526 | || arp->ar_pro != htons(ETH_P_IP) |
| 527 | || arp->ar_pln != 4 || arp->ar_hln != ETH_ALEN) |
| 528 | return NF_ACCEPT; |
| 529 | |
Harald Welte | 4095ebf | 2005-06-28 12:49:30 -0700 | [diff] [blame] | 530 | /* we only want to mangle arp requests and replies */ |
| 531 | if (arp->ar_op != htons(ARPOP_REPLY) |
| 532 | && arp->ar_op != htons(ARPOP_REQUEST)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 533 | return NF_ACCEPT; |
| 534 | |
| 535 | payload = (void *)(arp+1); |
| 536 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 537 | /* if there is no clusterip configuration for the arp reply's |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 538 | * source ip, we don't want to mangle it */ |
KOVACS Krisztian | 4451362 | 2005-09-16 16:59:46 -0700 | [diff] [blame] | 539 | c = clusterip_config_find_get(payload->src_ip, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 540 | if (!c) |
| 541 | return NF_ACCEPT; |
| 542 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 543 | /* normally the linux kernel always replies to arp queries of |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 544 | * addresses on different interfacs. However, in the CLUSTERIP case |
| 545 | * this wouldn't work, since we didn't subscribe the mcast group on |
| 546 | * other interfaces */ |
| 547 | if (c->dev != out) { |
| 548 | DEBUGP("CLUSTERIP: not mangling arp reply on different " |
| 549 | "interface: cip'%s'-skb'%s'\n", c->dev->name, out->name); |
| 550 | clusterip_config_put(c); |
| 551 | return NF_ACCEPT; |
| 552 | } |
| 553 | |
| 554 | /* mangle reply hardware address */ |
| 555 | memcpy(payload->src_hw, c->clustermac, arp->ar_hln); |
| 556 | |
| 557 | #ifdef CLUSTERIP_DEBUG |
| 558 | DEBUGP(KERN_DEBUG "CLUSTERIP mangled arp reply: "); |
| 559 | arp_print(payload); |
| 560 | #endif |
| 561 | |
| 562 | clusterip_config_put(c); |
| 563 | |
| 564 | return NF_ACCEPT; |
| 565 | } |
| 566 | |
| 567 | static struct nf_hook_ops cip_arp_ops = { |
| 568 | .hook = arp_mangle, |
| 569 | .pf = NF_ARP, |
| 570 | .hooknum = NF_ARP_OUT, |
| 571 | .priority = -1 |
| 572 | }; |
| 573 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 574 | /*********************************************************************** |
| 575 | * PROC DIR HANDLING |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 576 | ***********************************************************************/ |
| 577 | |
| 578 | #ifdef CONFIG_PROC_FS |
| 579 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 580 | struct clusterip_seq_position { |
| 581 | unsigned int pos; /* position */ |
| 582 | unsigned int weight; /* number of bits set == size */ |
| 583 | unsigned int bit; /* current bit */ |
| 584 | unsigned long val; /* current value */ |
| 585 | }; |
| 586 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 587 | static void *clusterip_seq_start(struct seq_file *s, loff_t *pos) |
| 588 | { |
| 589 | struct proc_dir_entry *pde = s->private; |
| 590 | struct clusterip_config *c = pde->data; |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 591 | unsigned int weight; |
| 592 | u_int32_t local_nodes; |
| 593 | struct clusterip_seq_position *idx; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 594 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 595 | /* FIXME: possible race */ |
| 596 | local_nodes = c->local_nodes; |
| 597 | weight = hweight32(local_nodes); |
| 598 | if (*pos >= weight) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 599 | return NULL; |
| 600 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 601 | idx = kmalloc(sizeof(struct clusterip_seq_position), GFP_KERNEL); |
| 602 | if (!idx) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 603 | return ERR_PTR(-ENOMEM); |
| 604 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 605 | idx->pos = *pos; |
| 606 | idx->weight = weight; |
| 607 | idx->bit = ffs(local_nodes); |
| 608 | idx->val = local_nodes; |
| 609 | clear_bit(idx->bit - 1, &idx->val); |
| 610 | |
| 611 | return idx; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 612 | } |
| 613 | |
| 614 | static void *clusterip_seq_next(struct seq_file *s, void *v, loff_t *pos) |
| 615 | { |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 616 | struct clusterip_seq_position *idx = (struct clusterip_seq_position *)v; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 617 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 618 | *pos = ++idx->pos; |
| 619 | if (*pos >= idx->weight) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 620 | kfree(v); |
| 621 | return NULL; |
| 622 | } |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 623 | idx->bit = ffs(idx->val); |
| 624 | clear_bit(idx->bit - 1, &idx->val); |
| 625 | return idx; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 626 | } |
| 627 | |
| 628 | static void clusterip_seq_stop(struct seq_file *s, void *v) |
| 629 | { |
| 630 | kfree(v); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 631 | } |
| 632 | |
| 633 | static int clusterip_seq_show(struct seq_file *s, void *v) |
| 634 | { |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 635 | struct clusterip_seq_position *idx = (struct clusterip_seq_position *)v; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 636 | |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 637 | if (idx->pos != 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 638 | seq_putc(s, ','); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 639 | |
KOVACS Krisztian | 136e92b | 2005-09-16 17:00:04 -0700 | [diff] [blame] | 640 | seq_printf(s, "%u", idx->bit); |
| 641 | |
| 642 | if (idx->pos == idx->weight - 1) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 643 | seq_putc(s, '\n'); |
| 644 | |
| 645 | return 0; |
| 646 | } |
| 647 | |
| 648 | static struct seq_operations clusterip_seq_ops = { |
| 649 | .start = clusterip_seq_start, |
| 650 | .next = clusterip_seq_next, |
| 651 | .stop = clusterip_seq_stop, |
| 652 | .show = clusterip_seq_show, |
| 653 | }; |
| 654 | |
| 655 | static int clusterip_proc_open(struct inode *inode, struct file *file) |
| 656 | { |
| 657 | int ret = seq_open(file, &clusterip_seq_ops); |
| 658 | |
| 659 | if (!ret) { |
| 660 | struct seq_file *sf = file->private_data; |
| 661 | struct proc_dir_entry *pde = PDE(inode); |
| 662 | struct clusterip_config *c = pde->data; |
| 663 | |
| 664 | sf->private = pde; |
| 665 | |
| 666 | clusterip_config_get(c); |
| 667 | } |
| 668 | |
| 669 | return ret; |
| 670 | } |
| 671 | |
| 672 | static int clusterip_proc_release(struct inode *inode, struct file *file) |
| 673 | { |
| 674 | struct proc_dir_entry *pde = PDE(inode); |
| 675 | struct clusterip_config *c = pde->data; |
| 676 | int ret; |
| 677 | |
| 678 | ret = seq_release(inode, file); |
| 679 | |
| 680 | if (!ret) |
| 681 | clusterip_config_put(c); |
| 682 | |
| 683 | return ret; |
| 684 | } |
| 685 | |
| 686 | static ssize_t clusterip_proc_write(struct file *file, const char __user *input, |
| 687 | size_t size, loff_t *ofs) |
| 688 | { |
| 689 | #define PROC_WRITELEN 10 |
| 690 | char buffer[PROC_WRITELEN+1]; |
Josef Sipek | 6df81ab | 2006-12-08 02:37:24 -0800 | [diff] [blame] | 691 | struct proc_dir_entry *pde = PDE(file->f_path.dentry->d_inode); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 692 | struct clusterip_config *c = pde->data; |
| 693 | unsigned long nodenum; |
| 694 | |
| 695 | if (copy_from_user(buffer, input, PROC_WRITELEN)) |
| 696 | return -EFAULT; |
| 697 | |
| 698 | if (*buffer == '+') { |
| 699 | nodenum = simple_strtoul(buffer+1, NULL, 10); |
| 700 | if (clusterip_add_node(c, nodenum)) |
| 701 | return -ENOMEM; |
| 702 | } else if (*buffer == '-') { |
| 703 | nodenum = simple_strtoul(buffer+1, NULL,10); |
| 704 | if (clusterip_del_node(c, nodenum)) |
| 705 | return -ENOENT; |
| 706 | } else |
| 707 | return -EIO; |
| 708 | |
| 709 | return size; |
| 710 | } |
| 711 | |
Arjan van de Ven | 9a32144 | 2007-02-12 00:55:35 -0800 | [diff] [blame] | 712 | static const struct file_operations clusterip_proc_fops = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 713 | .owner = THIS_MODULE, |
| 714 | .open = clusterip_proc_open, |
| 715 | .read = seq_read, |
| 716 | .write = clusterip_proc_write, |
| 717 | .llseek = seq_lseek, |
| 718 | .release = clusterip_proc_release, |
| 719 | }; |
| 720 | |
| 721 | #endif /* CONFIG_PROC_FS */ |
| 722 | |
Patrick McHardy | 32292a7 | 2006-04-06 14:11:30 -0700 | [diff] [blame] | 723 | static int __init ipt_clusterip_init(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 724 | { |
| 725 | int ret; |
| 726 | |
Jan Engelhardt | 6709dbb | 2007-02-07 15:11:19 -0800 | [diff] [blame] | 727 | ret = xt_register_target(&clusterip_tgt); |
Patrick McHardy | 32292a7 | 2006-04-06 14:11:30 -0700 | [diff] [blame] | 728 | if (ret < 0) |
| 729 | return ret; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 730 | |
Patrick McHardy | 32292a7 | 2006-04-06 14:11:30 -0700 | [diff] [blame] | 731 | ret = nf_register_hook(&cip_arp_ops); |
| 732 | if (ret < 0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 733 | goto cleanup_target; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 734 | |
| 735 | #ifdef CONFIG_PROC_FS |
| 736 | clusterip_procdir = proc_mkdir("ipt_CLUSTERIP", proc_net); |
| 737 | if (!clusterip_procdir) { |
| 738 | printk(KERN_ERR "CLUSTERIP: Unable to proc dir entry\n"); |
| 739 | ret = -ENOMEM; |
| 740 | goto cleanup_hook; |
| 741 | } |
| 742 | #endif /* CONFIG_PROC_FS */ |
| 743 | |
| 744 | printk(KERN_NOTICE "ClusterIP Version %s loaded successfully\n", |
| 745 | CLUSTERIP_VERSION); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 746 | return 0; |
| 747 | |
Patrick McHardy | 7659258 | 2006-11-29 02:35:42 +0100 | [diff] [blame] | 748 | #ifdef CONFIG_PROC_FS |
Patrick McHardy | 32292a7 | 2006-04-06 14:11:30 -0700 | [diff] [blame] | 749 | cleanup_hook: |
| 750 | nf_unregister_hook(&cip_arp_ops); |
Patrick McHardy | 7659258 | 2006-11-29 02:35:42 +0100 | [diff] [blame] | 751 | #endif /* CONFIG_PROC_FS */ |
Patrick McHardy | 32292a7 | 2006-04-06 14:11:30 -0700 | [diff] [blame] | 752 | cleanup_target: |
Jan Engelhardt | 6709dbb | 2007-02-07 15:11:19 -0800 | [diff] [blame] | 753 | xt_unregister_target(&clusterip_tgt); |
Patrick McHardy | 32292a7 | 2006-04-06 14:11:30 -0700 | [diff] [blame] | 754 | return ret; |
| 755 | } |
| 756 | |
| 757 | static void __exit ipt_clusterip_fini(void) |
| 758 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 759 | printk(KERN_NOTICE "ClusterIP Version %s unloading\n", |
| 760 | CLUSTERIP_VERSION); |
| 761 | #ifdef CONFIG_PROC_FS |
| 762 | remove_proc_entry(clusterip_procdir->name, clusterip_procdir->parent); |
| 763 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 764 | nf_unregister_hook(&cip_arp_ops); |
Jan Engelhardt | 6709dbb | 2007-02-07 15:11:19 -0800 | [diff] [blame] | 765 | xt_unregister_target(&clusterip_tgt); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 766 | } |
| 767 | |
Andrew Morton | 65b4b4e | 2006-03-28 16:37:06 -0800 | [diff] [blame] | 768 | module_init(ipt_clusterip_init); |
| 769 | module_exit(ipt_clusterip_fini); |