diff options
Diffstat (limited to 'drivers/net/tun.c')
| -rw-r--r-- | drivers/net/tun.c | 458 |
1 files changed, 297 insertions, 161 deletions
diff --git a/drivers/net/tun.c b/drivers/net/tun.c index eba1271b9735..a82b32b40131 100644 --- a/drivers/net/tun.c +++ b/drivers/net/tun.c | |||
| @@ -18,15 +18,11 @@ | |||
| 18 | /* | 18 | /* |
| 19 | * Changes: | 19 | * Changes: |
| 20 | * | 20 | * |
| 21 | * Brian Braunstein <linuxkernel@bristyle.com> 2007/03/23 | ||
| 22 | * Fixed hw address handling. Now net_device.dev_addr is kept consistent | ||
| 23 | * with tun.dev_addr when the address is set by this module. | ||
| 24 | * | ||
| 25 | * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14 | 21 | * Mike Kershaw <dragorn@kismetwireless.net> 2005/08/14 |
| 26 | * Add TUNSETLINK ioctl to set the link encapsulation | 22 | * Add TUNSETLINK ioctl to set the link encapsulation |
| 27 | * | 23 | * |
| 28 | * Mark Smith <markzzzsmith@yahoo.com.au> | 24 | * Mark Smith <markzzzsmith@yahoo.com.au> |
| 29 | * Use random_ether_addr() for tap MAC address. | 25 | * Use random_ether_addr() for tap MAC address. |
| 30 | * | 26 | * |
| 31 | * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20 | 27 | * Harald Roelle <harald.roelle@ifi.lmu.de> 2004/04/20 |
| 32 | * Fixes in packet dropping, queue length setting and queue wakeup. | 28 | * Fixes in packet dropping, queue length setting and queue wakeup. |
| @@ -64,6 +60,7 @@ | |||
| 64 | #include <linux/if_tun.h> | 60 | #include <linux/if_tun.h> |
| 65 | #include <linux/crc32.h> | 61 | #include <linux/crc32.h> |
| 66 | #include <linux/nsproxy.h> | 62 | #include <linux/nsproxy.h> |
| 63 | #include <linux/virtio_net.h> | ||
| 67 | #include <net/net_namespace.h> | 64 | #include <net/net_namespace.h> |
| 68 | #include <net/netns/generic.h> | 65 | #include <net/netns/generic.h> |
| 69 | 66 | ||
| @@ -83,9 +80,16 @@ static int debug; | |||
| 83 | #define DBG1( a... ) | 80 | #define DBG1( a... ) |
| 84 | #endif | 81 | #endif |
| 85 | 82 | ||
| 83 | #define FLT_EXACT_COUNT 8 | ||
| 84 | struct tap_filter { | ||
| 85 | unsigned int count; /* Number of addrs. Zero means disabled */ | ||
| 86 | u32 mask[2]; /* Mask of the hashed addrs */ | ||
| 87 | unsigned char addr[FLT_EXACT_COUNT][ETH_ALEN]; | ||
| 88 | }; | ||
| 89 | |||
| 86 | struct tun_struct { | 90 | struct tun_struct { |
| 87 | struct list_head list; | 91 | struct list_head list; |
| 88 | unsigned long flags; | 92 | unsigned int flags; |
| 89 | int attached; | 93 | int attached; |
| 90 | uid_t owner; | 94 | uid_t owner; |
| 91 | gid_t group; | 95 | gid_t group; |
| @@ -94,19 +98,119 @@ struct tun_struct { | |||
| 94 | struct sk_buff_head readq; | 98 | struct sk_buff_head readq; |
| 95 | 99 | ||
| 96 | struct net_device *dev; | 100 | struct net_device *dev; |
| 101 | struct fasync_struct *fasync; | ||
| 97 | 102 | ||
| 98 | struct fasync_struct *fasync; | 103 | struct tap_filter txflt; |
| 99 | |||
| 100 | unsigned long if_flags; | ||
| 101 | u8 dev_addr[ETH_ALEN]; | ||
| 102 | u32 chr_filter[2]; | ||
| 103 | u32 net_filter[2]; | ||
| 104 | 104 | ||
| 105 | #ifdef TUN_DEBUG | 105 | #ifdef TUN_DEBUG |
| 106 | int debug; | 106 | int debug; |
| 107 | #endif | 107 | #endif |
| 108 | }; | 108 | }; |
| 109 | 109 | ||
| 110 | /* TAP filterting */ | ||
| 111 | static void addr_hash_set(u32 *mask, const u8 *addr) | ||
| 112 | { | ||
| 113 | int n = ether_crc(ETH_ALEN, addr) >> 26; | ||
| 114 | mask[n >> 5] |= (1 << (n & 31)); | ||
| 115 | } | ||
| 116 | |||
| 117 | static unsigned int addr_hash_test(const u32 *mask, const u8 *addr) | ||
| 118 | { | ||
| 119 | int n = ether_crc(ETH_ALEN, addr) >> 26; | ||
| 120 | return mask[n >> 5] & (1 << (n & 31)); | ||
| 121 | } | ||
| 122 | |||
| 123 | static int update_filter(struct tap_filter *filter, void __user *arg) | ||
| 124 | { | ||
| 125 | struct { u8 u[ETH_ALEN]; } *addr; | ||
| 126 | struct tun_filter uf; | ||
| 127 | int err, alen, n, nexact; | ||
| 128 | |||
| 129 | if (copy_from_user(&uf, arg, sizeof(uf))) | ||
| 130 | return -EFAULT; | ||
| 131 | |||
| 132 | if (!uf.count) { | ||
| 133 | /* Disabled */ | ||
| 134 | filter->count = 0; | ||
| 135 | return 0; | ||
| 136 | } | ||
| 137 | |||
| 138 | alen = ETH_ALEN * uf.count; | ||
| 139 | addr = kmalloc(alen, GFP_KERNEL); | ||
| 140 | if (!addr) | ||
| 141 | return -ENOMEM; | ||
| 142 | |||
| 143 | if (copy_from_user(addr, arg + sizeof(uf), alen)) { | ||
| 144 | err = -EFAULT; | ||
| 145 | goto done; | ||
| 146 | } | ||
| 147 | |||
| 148 | /* The filter is updated without holding any locks. Which is | ||
| 149 | * perfectly safe. We disable it first and in the worst | ||
| 150 | * case we'll accept a few undesired packets. */ | ||
| 151 | filter->count = 0; | ||
| 152 | wmb(); | ||
| 153 | |||
| 154 | /* Use first set of addresses as an exact filter */ | ||
| 155 | for (n = 0; n < uf.count && n < FLT_EXACT_COUNT; n++) | ||
| 156 | memcpy(filter->addr[n], addr[n].u, ETH_ALEN); | ||
| 157 | |||
| 158 | nexact = n; | ||
| 159 | |||
| 160 | /* The rest is hashed */ | ||
| 161 | memset(filter->mask, 0, sizeof(filter->mask)); | ||
| 162 | for (; n < uf.count; n++) | ||
| 163 | addr_hash_set(filter->mask, addr[n].u); | ||
| 164 | |||
| 165 | /* For ALLMULTI just set the mask to all ones. | ||
| 166 | * This overrides the mask populated above. */ | ||
| 167 | if ((uf.flags & TUN_FLT_ALLMULTI)) | ||
| 168 | memset(filter->mask, ~0, sizeof(filter->mask)); | ||
| 169 | |||
| 170 | /* Now enable the filter */ | ||
| 171 | wmb(); | ||
| 172 | filter->count = nexact; | ||
| 173 | |||
| 174 | /* Return the number of exact filters */ | ||
| 175 | err = nexact; | ||
| 176 | |||
| 177 | done: | ||
| 178 | kfree(addr); | ||
| 179 | return err; | ||
| 180 | } | ||
| 181 | |||
| 182 | /* Returns: 0 - drop, !=0 - accept */ | ||
| 183 | static int run_filter(struct tap_filter *filter, const struct sk_buff *skb) | ||
| 184 | { | ||
| 185 | /* Cannot use eth_hdr(skb) here because skb_mac_hdr() is incorrect | ||
| 186 | * at this point. */ | ||
| 187 | struct ethhdr *eh = (struct ethhdr *) skb->data; | ||
| 188 | int i; | ||
| 189 | |||
| 190 | /* Exact match */ | ||
| 191 | for (i = 0; i < filter->count; i++) | ||
| 192 | if (!compare_ether_addr(eh->h_dest, filter->addr[i])) | ||
| 193 | return 1; | ||
| 194 | |||
| 195 | /* Inexact match (multicast only) */ | ||
| 196 | if (is_multicast_ether_addr(eh->h_dest)) | ||
| 197 | return addr_hash_test(filter->mask, eh->h_dest); | ||
| 198 | |||
| 199 | return 0; | ||
| 200 | } | ||
| 201 | |||
| 202 | /* | ||
| 203 | * Checks whether the packet is accepted or not. | ||
| 204 | * Returns: 0 - drop, !=0 - accept | ||
| 205 | */ | ||
| 206 | static int check_filter(struct tap_filter *filter, const struct sk_buff *skb) | ||
| 207 | { | ||
| 208 | if (!filter->count) | ||
| 209 | return 1; | ||
| 210 | |||
| 211 | return run_filter(filter, skb); | ||
| 212 | } | ||
| 213 | |||
| 110 | /* Network device part of the driver */ | 214 | /* Network device part of the driver */ |
| 111 | 215 | ||
| 112 | static unsigned int tun_net_id; | 216 | static unsigned int tun_net_id; |
| @@ -141,7 +245,12 @@ static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev) | |||
| 141 | if (!tun->attached) | 245 | if (!tun->attached) |
| 142 | goto drop; | 246 | goto drop; |
| 143 | 247 | ||
| 144 | /* Packet dropping */ | 248 | /* Drop if the filter does not like it. |
| 249 | * This is a noop if the filter is disabled. | ||
| 250 | * Filter can be enabled only for the TAP devices. */ | ||
| 251 | if (!check_filter(&tun->txflt, skb)) | ||
| 252 | goto drop; | ||
| 253 | |||
| 145 | if (skb_queue_len(&tun->readq) >= dev->tx_queue_len) { | 254 | if (skb_queue_len(&tun->readq) >= dev->tx_queue_len) { |
| 146 | if (!(tun->flags & TUN_ONE_QUEUE)) { | 255 | if (!(tun->flags & TUN_ONE_QUEUE)) { |
| 147 | /* Normal queueing mode. */ | 256 | /* Normal queueing mode. */ |
| @@ -158,7 +267,7 @@ static int tun_net_xmit(struct sk_buff *skb, struct net_device *dev) | |||
| 158 | } | 267 | } |
| 159 | } | 268 | } |
| 160 | 269 | ||
| 161 | /* Queue packet */ | 270 | /* Enqueue packet */ |
| 162 | skb_queue_tail(&tun->readq, skb); | 271 | skb_queue_tail(&tun->readq, skb); |
| 163 | dev->trans_start = jiffies; | 272 | dev->trans_start = jiffies; |
| 164 | 273 | ||
| @@ -174,41 +283,14 @@ drop: | |||
| 174 | return 0; | 283 | return 0; |
| 175 | } | 284 | } |
| 176 | 285 | ||
| 177 | /** Add the specified Ethernet address to this multicast filter. */ | 286 | static void tun_net_mclist(struct net_device *dev) |
| 178 | static void | ||
| 179 | add_multi(u32* filter, const u8* addr) | ||
| 180 | { | 287 | { |
| 181 | int bit_nr = ether_crc(ETH_ALEN, addr) >> 26; | 288 | /* |
| 182 | filter[bit_nr >> 5] |= 1 << (bit_nr & 31); | 289 | * This callback is supposed to deal with mc filter in |
| 183 | } | 290 | * _rx_ path and has nothing to do with the _tx_ path. |
| 184 | 291 | * In rx path we always accept everything userspace gives us. | |
| 185 | /** Remove the specified Ethernet addres from this multicast filter. */ | 292 | */ |
| 186 | static void | 293 | return; |
| 187 | del_multi(u32* filter, const u8* addr) | ||
| 188 | { | ||
| 189 | int bit_nr = ether_crc(ETH_ALEN, addr) >> 26; | ||
| 190 | filter[bit_nr >> 5] &= ~(1 << (bit_nr & 31)); | ||
| 191 | } | ||
| 192 | |||
| 193 | /** Update the list of multicast groups to which the network device belongs. | ||
| 194 | * This list is used to filter packets being sent from the character device to | ||
| 195 | * the network device. */ | ||
| 196 | static void | ||
| 197 | tun_net_mclist(struct net_device *dev) | ||
| 198 | { | ||
| 199 | struct tun_struct *tun = netdev_priv(dev); | ||
| 200 | const struct dev_mc_list *mclist; | ||
| 201 | int i; | ||
| 202 | DECLARE_MAC_BUF(mac); | ||
| 203 | DBG(KERN_DEBUG "%s: tun_net_mclist: mc_count %d\n", | ||
| 204 | dev->name, dev->mc_count); | ||
| 205 | memset(tun->chr_filter, 0, sizeof tun->chr_filter); | ||
| 206 | for (i = 0, mclist = dev->mc_list; i < dev->mc_count && mclist != NULL; | ||
| 207 | i++, mclist = mclist->next) { | ||
| 208 | add_multi(tun->net_filter, mclist->dmi_addr); | ||
| 209 | DBG(KERN_DEBUG "%s: tun_net_mclist: %s\n", | ||
| 210 | dev->name, print_mac(mac, mclist->dmi_addr)); | ||
| 211 | } | ||
| 212 | } | 294 | } |
| 213 | 295 | ||
| 214 | #define MIN_MTU 68 | 296 | #define MIN_MTU 68 |
| @@ -244,13 +326,11 @@ static void tun_net_init(struct net_device *dev) | |||
| 244 | 326 | ||
| 245 | case TUN_TAP_DEV: | 327 | case TUN_TAP_DEV: |
| 246 | /* Ethernet TAP Device */ | 328 | /* Ethernet TAP Device */ |
| 247 | dev->set_multicast_list = tun_net_mclist; | ||
| 248 | |||
| 249 | ether_setup(dev); | 329 | ether_setup(dev); |
| 250 | dev->change_mtu = tun_net_change_mtu; | 330 | dev->change_mtu = tun_net_change_mtu; |
| 331 | dev->set_multicast_list = tun_net_mclist; | ||
| 251 | 332 | ||
| 252 | /* random address already created for us by tun_set_iff, use it */ | 333 | random_ether_addr(dev->dev_addr); |
| 253 | memcpy(dev->dev_addr, tun->dev_addr, min(sizeof(tun->dev_addr), sizeof(dev->dev_addr)) ); | ||
| 254 | 334 | ||
| 255 | dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */ | 335 | dev->tx_queue_len = TUN_READQ_SIZE; /* We prefer our own queue length */ |
| 256 | break; | 336 | break; |
| @@ -284,6 +364,7 @@ static __inline__ ssize_t tun_get_user(struct tun_struct *tun, struct iovec *iv, | |||
| 284 | struct tun_pi pi = { 0, __constant_htons(ETH_P_IP) }; | 364 | struct tun_pi pi = { 0, __constant_htons(ETH_P_IP) }; |
| 285 | struct sk_buff *skb; | 365 | struct sk_buff *skb; |
| 286 | size_t len = count, align = 0; | 366 | size_t len = count, align = 0; |
| 367 | struct virtio_net_hdr gso = { 0 }; | ||
| 287 | 368 | ||
| 288 | if (!(tun->flags & TUN_NO_PI)) { | 369 | if (!(tun->flags & TUN_NO_PI)) { |
| 289 | if ((len -= sizeof(pi)) > count) | 370 | if ((len -= sizeof(pi)) > count) |
| @@ -293,6 +374,17 @@ static __inline__ ssize_t tun_get_user(struct tun_struct *tun, struct iovec *iv, | |||
| 293 | return -EFAULT; | 374 | return -EFAULT; |
| 294 | } | 375 | } |
| 295 | 376 | ||
| 377 | if (tun->flags & TUN_VNET_HDR) { | ||
| 378 | if ((len -= sizeof(gso)) > count) | ||
| 379 | return -EINVAL; | ||
| 380 | |||
| 381 | if (memcpy_fromiovec((void *)&gso, iv, sizeof(gso))) | ||
| 382 | return -EFAULT; | ||
| 383 | |||
| 384 | if (gso.hdr_len > len) | ||
| 385 | return -EINVAL; | ||
| 386 | } | ||
| 387 | |||
| 296 | if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) { | 388 | if ((tun->flags & TUN_TYPE_MASK) == TUN_TAP_DEV) { |
| 297 | align = NET_IP_ALIGN; | 389 | align = NET_IP_ALIGN; |
| 298 | if (unlikely(len < ETH_HLEN)) | 390 | if (unlikely(len < ETH_HLEN)) |
| @@ -312,6 +404,16 @@ static __inline__ ssize_t tun_get_user(struct tun_struct *tun, struct iovec *iv, | |||
| 312 | return -EFAULT; | 404 | return -EFAULT; |
| 313 | } | 405 | } |
| 314 | 406 | ||
| 407 | if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) { | ||
| 408 | if (!skb_partial_csum_set(skb, gso.csum_start, | ||
| 409 | gso.csum_offset)) { | ||
| 410 | tun->dev->stats.rx_frame_errors++; | ||
| 411 | kfree_skb(skb); | ||
| 412 | return -EINVAL; | ||
| 413 | } | ||
| 414 | } else if (tun->flags & TUN_NOCHECKSUM) | ||
| 415 | skb->ip_summed = CHECKSUM_UNNECESSARY; | ||
| 416 | |||
| 315 | switch (tun->flags & TUN_TYPE_MASK) { | 417 | switch (tun->flags & TUN_TYPE_MASK) { |
| 316 | case TUN_TUN_DEV: | 418 | case TUN_TUN_DEV: |
| 317 | if (tun->flags & TUN_NO_PI) { | 419 | if (tun->flags & TUN_NO_PI) { |
| @@ -338,8 +440,35 @@ static __inline__ ssize_t tun_get_user(struct tun_struct *tun, struct iovec *iv, | |||
| 338 | break; | 440 | break; |
| 339 | }; | 441 | }; |
| 340 | 442 | ||
| 341 | if (tun->flags & TUN_NOCHECKSUM) | 443 | if (gso.gso_type != VIRTIO_NET_HDR_GSO_NONE) { |
| 342 | skb->ip_summed = CHECKSUM_UNNECESSARY; | 444 | pr_debug("GSO!\n"); |
| 445 | switch (gso.gso_type & ~VIRTIO_NET_HDR_GSO_ECN) { | ||
| 446 | case VIRTIO_NET_HDR_GSO_TCPV4: | ||
| 447 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4; | ||
| 448 | break; | ||
| 449 | case VIRTIO_NET_HDR_GSO_TCPV6: | ||
| 450 | skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6; | ||
| 451 | break; | ||
| 452 | default: | ||
| 453 | tun->dev->stats.rx_frame_errors++; | ||
| 454 | kfree_skb(skb); | ||
| 455 | return -EINVAL; | ||
| 456 | } | ||
| 457 | |||
| 458 | if (gso.gso_type & VIRTIO_NET_HDR_GSO_ECN) | ||
| 459 | skb_shinfo(skb)->gso_type |= SKB_GSO_TCP_ECN; | ||
| 460 | |||
| 461 | skb_shinfo(skb)->gso_size = gso.gso_size; | ||
| 462 | if (skb_shinfo(skb)->gso_size == 0) { | ||
| 463 | tun->dev->stats.rx_frame_errors++; | ||
| 464 | kfree_skb(skb); | ||
| 465 | return -EINVAL; | ||
| 466 | } | ||
| 467 | |||
| 468 | /* Header must be checked, and gso_segs computed. */ | ||
| 469 | skb_shinfo(skb)->gso_type |= SKB_GSO_DODGY; | ||
| 470 | skb_shinfo(skb)->gso_segs = 0; | ||
| 471 | } | ||
| 343 | 472 | ||
| 344 | netif_rx_ni(skb); | 473 | netif_rx_ni(skb); |
| 345 | tun->dev->last_rx = jiffies; | 474 | tun->dev->last_rx = jiffies; |
| @@ -385,6 +514,39 @@ static __inline__ ssize_t tun_put_user(struct tun_struct *tun, | |||
| 385 | total += sizeof(pi); | 514 | total += sizeof(pi); |
| 386 | } | 515 | } |
| 387 | 516 | ||
| 517 | if (tun->flags & TUN_VNET_HDR) { | ||
| 518 | struct virtio_net_hdr gso = { 0 }; /* no info leak */ | ||
| 519 | if ((len -= sizeof(gso)) < 0) | ||
| 520 | return -EINVAL; | ||
| 521 | |||
| 522 | if (skb_is_gso(skb)) { | ||
| 523 | struct skb_shared_info *sinfo = skb_shinfo(skb); | ||
| 524 | |||
| 525 | /* This is a hint as to how much should be linear. */ | ||
| 526 | gso.hdr_len = skb_headlen(skb); | ||
| 527 | gso.gso_size = sinfo->gso_size; | ||
| 528 | if (sinfo->gso_type & SKB_GSO_TCPV4) | ||
| 529 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV4; | ||
| 530 | else if (sinfo->gso_type & SKB_GSO_TCPV6) | ||
| 531 | gso.gso_type = VIRTIO_NET_HDR_GSO_TCPV6; | ||
| 532 | else | ||
| 533 | BUG(); | ||
| 534 | if (sinfo->gso_type & SKB_GSO_TCP_ECN) | ||
| 535 | gso.gso_type |= VIRTIO_NET_HDR_GSO_ECN; | ||
| 536 | } else | ||
| 537 | gso.gso_type = VIRTIO_NET_HDR_GSO_NONE; | ||
| 538 | |||
| 539 | if (skb->ip_summed == CHECKSUM_PARTIAL) { | ||
| 540 | gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; | ||
| 541 | gso.csum_start = skb->csum_start - skb_headroom(skb); | ||
| 542 | gso.csum_offset = skb->csum_offset; | ||
| 543 | } /* else everything is zero */ | ||
| 544 | |||
| 545 | if (unlikely(memcpy_toiovec(iv, (void *)&gso, sizeof(gso)))) | ||
| 546 | return -EFAULT; | ||
| 547 | total += sizeof(gso); | ||
| 548 | } | ||
| 549 | |||
| 388 | len = min_t(int, skb->len, len); | 550 | len = min_t(int, skb->len, len); |
| 389 | 551 | ||
| 390 | skb_copy_datagram_iovec(skb, 0, iv, len); | 552 | skb_copy_datagram_iovec(skb, 0, iv, len); |
| @@ -404,7 +566,6 @@ static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv, | |||
| 404 | DECLARE_WAITQUEUE(wait, current); | 566 | DECLARE_WAITQUEUE(wait, current); |
| 405 | struct sk_buff *skb; | 567 | struct sk_buff *skb; |
| 406 | ssize_t len, ret = 0; | 568 | ssize_t len, ret = 0; |
| 407 | DECLARE_MAC_BUF(mac); | ||
| 408 | 569 | ||
| 409 | if (!tun) | 570 | if (!tun) |
| 410 | return -EBADFD; | 571 | return -EBADFD; |
| @@ -417,10 +578,6 @@ static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv, | |||
| 417 | 578 | ||
| 418 | add_wait_queue(&tun->read_wait, &wait); | 579 | add_wait_queue(&tun->read_wait, &wait); |
| 419 | while (len) { | 580 | while (len) { |
| 420 | const u8 ones[ ETH_ALEN] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; | ||
| 421 | u8 addr[ ETH_ALEN]; | ||
| 422 | int bit_nr; | ||
| 423 | |||
| 424 | current->state = TASK_INTERRUPTIBLE; | 581 | current->state = TASK_INTERRUPTIBLE; |
| 425 | 582 | ||
| 426 | /* Read frames from the queue */ | 583 | /* Read frames from the queue */ |
| @@ -440,36 +597,9 @@ static ssize_t tun_chr_aio_read(struct kiocb *iocb, const struct iovec *iv, | |||
| 440 | } | 597 | } |
| 441 | netif_wake_queue(tun->dev); | 598 | netif_wake_queue(tun->dev); |
| 442 | 599 | ||
| 443 | /** Decide whether to accept this packet. This code is designed to | 600 | ret = tun_put_user(tun, skb, (struct iovec *) iv, len); |
| 444 | * behave identically to an Ethernet interface. Accept the packet if | 601 | kfree_skb(skb); |
| 445 | * - we are promiscuous. | 602 | break; |
| 446 | * - the packet is addressed to us. | ||
| 447 | * - the packet is broadcast. | ||
| 448 | * - the packet is multicast and | ||
| 449 | * - we are multicast promiscous. | ||
| 450 | * - we belong to the multicast group. | ||
| 451 | */ | ||
| 452 | skb_copy_from_linear_data(skb, addr, min_t(size_t, sizeof addr, | ||
| 453 | skb->len)); | ||
| 454 | bit_nr = ether_crc(sizeof addr, addr) >> 26; | ||
| 455 | if ((tun->if_flags & IFF_PROMISC) || | ||
| 456 | memcmp(addr, tun->dev_addr, sizeof addr) == 0 || | ||
| 457 | memcmp(addr, ones, sizeof addr) == 0 || | ||
| 458 | (((addr[0] == 1 && addr[1] == 0 && addr[2] == 0x5e) || | ||
| 459 | (addr[0] == 0x33 && addr[1] == 0x33)) && | ||
| 460 | ((tun->if_flags & IFF_ALLMULTI) || | ||
| 461 | (tun->chr_filter[bit_nr >> 5] & (1 << (bit_nr & 31)))))) { | ||
| 462 | DBG(KERN_DEBUG "%s: tun_chr_readv: accepted: %s\n", | ||
| 463 | tun->dev->name, print_mac(mac, addr)); | ||
| 464 | ret = tun_put_user(tun, skb, (struct iovec *) iv, len); | ||
| 465 | kfree_skb(skb); | ||
| 466 | break; | ||
| 467 | } else { | ||
| 468 | DBG(KERN_DEBUG "%s: tun_chr_readv: rejected: %s\n", | ||
| 469 | tun->dev->name, print_mac(mac, addr)); | ||
| 470 | kfree_skb(skb); | ||
| 471 | continue; | ||
| 472 | } | ||
| 473 | } | 603 | } |
| 474 | 604 | ||
| 475 | current->state = TASK_RUNNING; | 605 | current->state = TASK_RUNNING; |
| @@ -565,12 +695,7 @@ static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) | |||
| 565 | tun = netdev_priv(dev); | 695 | tun = netdev_priv(dev); |
| 566 | tun->dev = dev; | 696 | tun->dev = dev; |
| 567 | tun->flags = flags; | 697 | tun->flags = flags; |
| 568 | /* Be promiscuous by default to maintain previous behaviour. */ | 698 | tun->txflt.count = 0; |
| 569 | tun->if_flags = IFF_PROMISC; | ||
| 570 | /* Generate random Ethernet address. */ | ||
| 571 | *(__be16 *)tun->dev_addr = htons(0x00FF); | ||
| 572 | get_random_bytes(tun->dev_addr + sizeof(u16), 4); | ||
| 573 | memset(tun->chr_filter, 0, sizeof tun->chr_filter); | ||
| 574 | 699 | ||
| 575 | tun_net_init(dev); | 700 | tun_net_init(dev); |
| 576 | 701 | ||
| @@ -599,6 +724,11 @@ static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) | |||
| 599 | else | 724 | else |
| 600 | tun->flags &= ~TUN_ONE_QUEUE; | 725 | tun->flags &= ~TUN_ONE_QUEUE; |
| 601 | 726 | ||
| 727 | if (ifr->ifr_flags & IFF_VNET_HDR) | ||
| 728 | tun->flags |= TUN_VNET_HDR; | ||
| 729 | else | ||
| 730 | tun->flags &= ~TUN_VNET_HDR; | ||
| 731 | |||
| 602 | file->private_data = tun; | 732 | file->private_data = tun; |
| 603 | tun->attached = 1; | 733 | tun->attached = 1; |
| 604 | get_net(dev_net(tun->dev)); | 734 | get_net(dev_net(tun->dev)); |
| @@ -618,12 +748,53 @@ static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) | |||
| 618 | return err; | 748 | return err; |
| 619 | } | 749 | } |
| 620 | 750 | ||
| 751 | /* This is like a cut-down ethtool ops, except done via tun fd so no | ||
| 752 | * privs required. */ | ||
| 753 | static int set_offload(struct net_device *dev, unsigned long arg) | ||
| 754 | { | ||
| 755 | unsigned int old_features, features; | ||
| 756 | |||
| 757 | old_features = dev->features; | ||
| 758 | /* Unset features, set them as we chew on the arg. */ | ||
| 759 | features = (old_features & ~(NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST | ||
| 760 | |NETIF_F_TSO_ECN|NETIF_F_TSO|NETIF_F_TSO6)); | ||
| 761 | |||
| 762 | if (arg & TUN_F_CSUM) { | ||
| 763 | features |= NETIF_F_HW_CSUM|NETIF_F_SG|NETIF_F_FRAGLIST; | ||
| 764 | arg &= ~TUN_F_CSUM; | ||
| 765 | |||
| 766 | if (arg & (TUN_F_TSO4|TUN_F_TSO6)) { | ||
| 767 | if (arg & TUN_F_TSO_ECN) { | ||
| 768 | features |= NETIF_F_TSO_ECN; | ||
| 769 | arg &= ~TUN_F_TSO_ECN; | ||
| 770 | } | ||
| 771 | if (arg & TUN_F_TSO4) | ||
| 772 | features |= NETIF_F_TSO; | ||
| 773 | if (arg & TUN_F_TSO6) | ||
| 774 | features |= NETIF_F_TSO6; | ||
| 775 | arg &= ~(TUN_F_TSO4|TUN_F_TSO6); | ||
| 776 | } | ||
| 777 | } | ||
| 778 | |||
| 779 | /* This gives the user a way to test for new features in future by | ||
| 780 | * trying to set them. */ | ||
| 781 | if (arg) | ||
| 782 | return -EINVAL; | ||
| 783 | |||
| 784 | dev->features = features; | ||
| 785 | if (old_features != dev->features) | ||
| 786 | netdev_features_change(dev); | ||
| 787 | |||
| 788 | return 0; | ||
| 789 | } | ||
| 790 | |||
| 621 | static int tun_chr_ioctl(struct inode *inode, struct file *file, | 791 | static int tun_chr_ioctl(struct inode *inode, struct file *file, |
| 622 | unsigned int cmd, unsigned long arg) | 792 | unsigned int cmd, unsigned long arg) |
| 623 | { | 793 | { |
| 624 | struct tun_struct *tun = file->private_data; | 794 | struct tun_struct *tun = file->private_data; |
| 625 | void __user* argp = (void __user*)arg; | 795 | void __user* argp = (void __user*)arg; |
| 626 | struct ifreq ifr; | 796 | struct ifreq ifr; |
| 797 | int ret; | ||
| 627 | DECLARE_MAC_BUF(mac); | 798 | DECLARE_MAC_BUF(mac); |
| 628 | 799 | ||
| 629 | if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89) | 800 | if (cmd == TUNSETIFF || _IOC_TYPE(cmd) == 0x89) |
| @@ -647,6 +818,15 @@ static int tun_chr_ioctl(struct inode *inode, struct file *file, | |||
| 647 | return 0; | 818 | return 0; |
| 648 | } | 819 | } |
| 649 | 820 | ||
| 821 | if (cmd == TUNGETFEATURES) { | ||
| 822 | /* Currently this just means: "what IFF flags are valid?". | ||
| 823 | * This is needed because we never checked for invalid flags on | ||
| 824 | * TUNSETIFF. */ | ||
| 825 | return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE | | ||
| 826 | IFF_VNET_HDR, | ||
| 827 | (unsigned int __user*)argp); | ||
| 828 | } | ||
| 829 | |||
| 650 | if (!tun) | 830 | if (!tun) |
| 651 | return -EBADFD; | 831 | return -EBADFD; |
| 652 | 832 | ||
| @@ -690,9 +870,6 @@ static int tun_chr_ioctl(struct inode *inode, struct file *file, | |||
| 690 | break; | 870 | break; |
| 691 | 871 | ||
| 692 | case TUNSETLINK: | 872 | case TUNSETLINK: |
| 693 | { | ||
| 694 | int ret; | ||
| 695 | |||
| 696 | /* Only allow setting the type when the interface is down */ | 873 | /* Only allow setting the type when the interface is down */ |
| 697 | rtnl_lock(); | 874 | rtnl_lock(); |
| 698 | if (tun->dev->flags & IFF_UP) { | 875 | if (tun->dev->flags & IFF_UP) { |
| @@ -706,85 +883,44 @@ static int tun_chr_ioctl(struct inode *inode, struct file *file, | |||
| 706 | } | 883 | } |
| 707 | rtnl_unlock(); | 884 | rtnl_unlock(); |
| 708 | return ret; | 885 | return ret; |
| 709 | } | ||
| 710 | 886 | ||
| 711 | #ifdef TUN_DEBUG | 887 | #ifdef TUN_DEBUG |
| 712 | case TUNSETDEBUG: | 888 | case TUNSETDEBUG: |
| 713 | tun->debug = arg; | 889 | tun->debug = arg; |
| 714 | break; | 890 | break; |
| 715 | #endif | 891 | #endif |
| 892 | case TUNSETOFFLOAD: | ||
| 893 | rtnl_lock(); | ||
| 894 | ret = set_offload(tun->dev, arg); | ||
| 895 | rtnl_unlock(); | ||
| 896 | return ret; | ||
| 716 | 897 | ||
| 717 | case SIOCGIFFLAGS: | 898 | case TUNSETTXFILTER: |
| 718 | ifr.ifr_flags = tun->if_flags; | 899 | /* Can be set only for TAPs */ |
| 719 | if (copy_to_user( argp, &ifr, sizeof ifr)) | 900 | if ((tun->flags & TUN_TYPE_MASK) != TUN_TAP_DEV) |
| 720 | return -EFAULT; | 901 | return -EINVAL; |
| 721 | return 0; | 902 | rtnl_lock(); |
| 722 | 903 | ret = update_filter(&tun->txflt, (void *) __user arg); | |
| 723 | case SIOCSIFFLAGS: | 904 | rtnl_unlock(); |
| 724 | /** Set the character device's interface flags. Currently only | 905 | return ret; |
| 725 | * IFF_PROMISC and IFF_ALLMULTI are used. */ | ||
| 726 | tun->if_flags = ifr.ifr_flags; | ||
| 727 | DBG(KERN_INFO "%s: interface flags 0x%lx\n", | ||
| 728 | tun->dev->name, tun->if_flags); | ||
| 729 | return 0; | ||
| 730 | 906 | ||
| 731 | case SIOCGIFHWADDR: | 907 | case SIOCGIFHWADDR: |
| 732 | /* Note: the actual net device's address may be different */ | 908 | /* Get hw addres */ |
| 733 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev_addr, | 909 | memcpy(ifr.ifr_hwaddr.sa_data, tun->dev->dev_addr, ETH_ALEN); |
| 734 | min(sizeof ifr.ifr_hwaddr.sa_data, sizeof tun->dev_addr)); | 910 | ifr.ifr_hwaddr.sa_family = tun->dev->type; |
| 735 | if (copy_to_user( argp, &ifr, sizeof ifr)) | 911 | if (copy_to_user(argp, &ifr, sizeof ifr)) |
| 736 | return -EFAULT; | 912 | return -EFAULT; |
| 737 | return 0; | 913 | return 0; |
| 738 | 914 | ||
| 739 | case SIOCSIFHWADDR: | 915 | case SIOCSIFHWADDR: |
| 740 | { | 916 | /* Set hw address */ |
| 741 | /* try to set the actual net device's hw address */ | 917 | DBG(KERN_DEBUG "%s: set hw address: %s\n", |
| 742 | int ret; | 918 | tun->dev->name, print_mac(mac, ifr.ifr_hwaddr.sa_data)); |
| 743 | 919 | ||
| 744 | rtnl_lock(); | 920 | rtnl_lock(); |
| 745 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); | 921 | ret = dev_set_mac_address(tun->dev, &ifr.ifr_hwaddr); |
| 746 | rtnl_unlock(); | 922 | rtnl_unlock(); |
| 747 | 923 | return ret; | |
| 748 | if (ret == 0) { | ||
| 749 | /** Set the character device's hardware address. This is used when | ||
| 750 | * filtering packets being sent from the network device to the character | ||
| 751 | * device. */ | ||
| 752 | memcpy(tun->dev_addr, ifr.ifr_hwaddr.sa_data, | ||
| 753 | min(sizeof ifr.ifr_hwaddr.sa_data, sizeof tun->dev_addr)); | ||
| 754 | DBG(KERN_DEBUG "%s: set hardware address: %x:%x:%x:%x:%x:%x\n", | ||
| 755 | tun->dev->name, | ||
| 756 | tun->dev_addr[0], tun->dev_addr[1], tun->dev_addr[2], | ||
| 757 | tun->dev_addr[3], tun->dev_addr[4], tun->dev_addr[5]); | ||
| 758 | } | ||
| 759 | |||
| 760 | return ret; | ||
| 761 | } | ||
| 762 | |||
| 763 | case SIOCADDMULTI: | ||
| 764 | /** Add the specified group to the character device's multicast filter | ||
| 765 | * list. */ | ||
| 766 | rtnl_lock(); | ||
| 767 | netif_tx_lock_bh(tun->dev); | ||
| 768 | add_multi(tun->chr_filter, ifr.ifr_hwaddr.sa_data); | ||
| 769 | netif_tx_unlock_bh(tun->dev); | ||
| 770 | rtnl_unlock(); | ||
| 771 | |||
| 772 | DBG(KERN_DEBUG "%s: add multi: %s\n", | ||
| 773 | tun->dev->name, print_mac(mac, ifr.ifr_hwaddr.sa_data)); | ||
| 774 | return 0; | ||
| 775 | |||
| 776 | case SIOCDELMULTI: | ||
| 777 | /** Remove the specified group from the character device's multicast | ||
| 778 | * filter list. */ | ||
| 779 | rtnl_lock(); | ||
| 780 | netif_tx_lock_bh(tun->dev); | ||
| 781 | del_multi(tun->chr_filter, ifr.ifr_hwaddr.sa_data); | ||
| 782 | netif_tx_unlock_bh(tun->dev); | ||
| 783 | rtnl_unlock(); | ||
| 784 | |||
| 785 | DBG(KERN_DEBUG "%s: del multi: %s\n", | ||
| 786 | tun->dev->name, print_mac(mac, ifr.ifr_hwaddr.sa_data)); | ||
| 787 | return 0; | ||
| 788 | 924 | ||
| 789 | default: | 925 | default: |
| 790 | return -EINVAL; | 926 | return -EINVAL; |
