| Message ID | a5a118807f06bded3feea4ba35168e9240c31a3b.1717690115.git.gnault@redhat.com (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net,v2] vxlan: Pull inner IP header in vxlan_xmit_one(). | expand |
On Thu, Jun 06, 2024 at 06:13:59PM +0200, Guillaume Nault wrote: > + if (!(flags & VXLAN_F_GPE) || skb->protocol == ETH_P_TEB) { > + if (!skb_vlan_inet_prepare(skb)) > + goto drop; Obviously wrong. Need to rest, sorry. I'll post v3 later.
Hi Guillaume, kernel test robot noticed the following build warnings: [auto build test WARNING on net/main] url: https://github.com/intel-lab-lkp/linux/commits/Guillaume-Nault/vxlan-Pull-inner-IP-header-in-vxlan_xmit_one/20240607-002253 base: net/main patch link: https://lore.kernel.org/r/a5a118807f06bded3feea4ba35168e9240c31a3b.1717690115.git.gnault%40redhat.com patch subject: [PATCH net v2] vxlan: Pull inner IP header in vxlan_xmit_one(). config: i386-randconfig-062-20240610 (https://download.01.org/0day-ci/archive/20240610/202406100922.x6iXjoXS-lkp@intel.com/config) compiler: gcc-8 (Ubuntu 8.4.0-3ubuntu2) 8.4.0 reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20240610/202406100922.x6iXjoXS-lkp@intel.com/reproduce) If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags | Reported-by: kernel test robot <lkp@intel.com> | Closes: https://lore.kernel.org/oe-kbuild-all/202406100922.x6iXjoXS-lkp@intel.com/ sparse warnings: (new ones prefixed by >>) drivers/net/vxlan/vxlan_core.c:393:34: sparse: sparse: incorrect type in argument 2 (different base types) @@ expected unsigned int [usertype] b @@ got restricted __be32 [usertype] vni @@ drivers/net/vxlan/vxlan_core.c:393:34: sparse: expected unsigned int [usertype] b drivers/net/vxlan/vxlan_core.c:393:34: sparse: got restricted __be32 [usertype] vni >> drivers/net/vxlan/vxlan_core.c:2358:42: sparse: sparse: restricted __be16 degrades to integer drivers/net/vxlan/vxlan_core.c: note: in included file (through include/net/net_namespace.h, include/linux/netdevice.h, include/net/inet_sock.h, ...): include/linux/skbuff.h:2743:13: sparse: sparse: self-comparison always evaluates to false vim +2358 drivers/net/vxlan/vxlan_core.c 2333 2334 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, 2335 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc) 2336 { 2337 struct dst_cache *dst_cache; 2338 struct ip_tunnel_info *info; 2339 struct ip_tunnel_key *pkey; 2340 struct ip_tunnel_key key; 2341 struct vxlan_dev *vxlan = netdev_priv(dev); 2342 const struct iphdr *old_iph; 2343 struct vxlan_metadata _md; 2344 struct vxlan_metadata *md = &_md; 2345 unsigned int pkt_len = skb->len; 2346 __be16 src_port = 0, dst_port; 2347 struct dst_entry *ndst = NULL; 2348 int addr_family; 2349 __u8 tos, ttl; 2350 int ifindex; 2351 int err; 2352 u32 flags = vxlan->cfg.flags; 2353 bool use_cache; 2354 bool udp_sum = false; 2355 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); 2356 __be32 vni = 0; 2357 > 2358 if (!(flags & VXLAN_F_GPE) || skb->protocol == ETH_P_TEB) { 2359 if (!skb_vlan_inet_prepare(skb)) 2360 goto drop; 2361 } else { 2362 if (!pskb_inet_may_pull(skb)) 2363 goto drop; 2364 } 2365 2366 old_iph = ip_hdr(skb); 2367 2368 info = skb_tunnel_info(skb); 2369 use_cache = ip_tunnel_dst_cache_usable(skb, info); 2370 2371 if (rdst) { 2372 memset(&key, 0, sizeof(key)); 2373 pkey = &key; 2374 2375 if (vxlan_addr_any(&rdst->remote_ip)) { 2376 if (did_rsc) { 2377 /* short-circuited back to local bridge */ 2378 vxlan_encap_bypass(skb, vxlan, vxlan, 2379 default_vni, true); 2380 return; 2381 } 2382 goto drop; 2383 } 2384 2385 addr_family = vxlan->cfg.saddr.sa.sa_family; 2386 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; 2387 vni = (rdst->remote_vni) ? : default_vni; 2388 ifindex = rdst->remote_ifindex; 2389 2390 if (addr_family == AF_INET) { 2391 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr; 2392 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr; 2393 } else { 2394 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr; 2395 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr; 2396 } 2397 2398 dst_cache = &rdst->dst_cache; 2399 md->gbp = skb->mark; 2400 if (flags & VXLAN_F_TTL_INHERIT) { 2401 ttl = ip_tunnel_get_ttl(old_iph, skb); 2402 } else { 2403 ttl = vxlan->cfg.ttl; 2404 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip)) 2405 ttl = 1; 2406 } 2407 tos = vxlan->cfg.tos; 2408 if (tos == 1) 2409 tos = ip_tunnel_get_dsfield(old_iph, skb); 2410 if (tos && !info) 2411 use_cache = false; 2412 2413 if (addr_family == AF_INET) 2414 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); 2415 else 2416 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); 2417 #if IS_ENABLED(CONFIG_IPV6) 2418 switch (vxlan->cfg.label_policy) { 2419 case VXLAN_LABEL_FIXED: 2420 key.label = vxlan->cfg.label; 2421 break; 2422 case VXLAN_LABEL_INHERIT: 2423 key.label = ip_tunnel_get_flowlabel(old_iph, skb); 2424 break; 2425 default: 2426 DEBUG_NET_WARN_ON_ONCE(1); 2427 goto drop; 2428 } 2429 #endif 2430 } else { 2431 if (!info) { 2432 WARN_ONCE(1, "%s: Missing encapsulation instructions\n", 2433 dev->name); 2434 goto drop; 2435 } 2436 pkey = &info->key; 2437 addr_family = ip_tunnel_info_af(info); 2438 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; 2439 vni = tunnel_id_to_key32(info->key.tun_id); 2440 ifindex = 0; 2441 dst_cache = &info->dst_cache; 2442 if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) { 2443 if (info->options_len < sizeof(*md)) 2444 goto drop; 2445 md = ip_tunnel_info_opts(info); 2446 } 2447 ttl = info->key.ttl; 2448 tos = info->key.tos; 2449 udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags); 2450 } 2451 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, 2452 vxlan->cfg.port_max, true); 2453 2454 rcu_read_lock(); 2455 if (addr_family == AF_INET) { 2456 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock); 2457 struct rtable *rt; 2458 __be16 df = 0; 2459 __be32 saddr; 2460 2461 if (!ifindex) 2462 ifindex = sock4->sock->sk->sk_bound_dev_if; 2463 2464 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex, 2465 &saddr, pkey, src_port, dst_port, 2466 tos, use_cache ? dst_cache : NULL); 2467 if (IS_ERR(rt)) { 2468 err = PTR_ERR(rt); 2469 goto tx_error; 2470 } 2471 2472 if (!info) { 2473 /* Bypass encapsulation if the destination is local */ 2474 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET, 2475 dst_port, ifindex, vni, 2476 &rt->dst, rt->rt_flags); 2477 if (err) 2478 goto out_unlock; 2479 2480 if (vxlan->cfg.df == VXLAN_DF_SET) { 2481 df = htons(IP_DF); 2482 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) { 2483 struct ethhdr *eth = eth_hdr(skb); 2484 2485 if (ntohs(eth->h_proto) == ETH_P_IPV6 || 2486 (ntohs(eth->h_proto) == ETH_P_IP && 2487 old_iph->frag_off & htons(IP_DF))) 2488 df = htons(IP_DF); 2489 } 2490 } else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, 2491 info->key.tun_flags)) { 2492 df = htons(IP_DF); 2493 } 2494 2495 ndst = &rt->dst; 2496 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE), 2497 netif_is_any_bridge_port(dev)); 2498 if (err < 0) { 2499 goto tx_error; 2500 } else if (err) { 2501 if (info) { 2502 struct ip_tunnel_info *unclone; 2503 2504 unclone = skb_tunnel_info_unclone(skb); 2505 if (unlikely(!unclone)) 2506 goto tx_error; 2507 2508 unclone->key.u.ipv4.src = pkey->u.ipv4.dst; 2509 unclone->key.u.ipv4.dst = saddr; 2510 } 2511 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2512 dst_release(ndst); 2513 goto out_unlock; 2514 } 2515 2516 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2517 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); 2518 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr), 2519 vni, md, flags, udp_sum); 2520 if (err < 0) 2521 goto tx_error; 2522 2523 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr, 2524 pkey->u.ipv4.dst, tos, ttl, df, 2525 src_port, dst_port, xnet, !udp_sum); 2526 #if IS_ENABLED(CONFIG_IPV6) 2527 } else { 2528 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock); 2529 struct in6_addr saddr; 2530 2531 if (!ifindex) 2532 ifindex = sock6->sock->sk->sk_bound_dev_if; 2533 2534 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock, 2535 ifindex, &saddr, pkey, 2536 src_port, dst_port, tos, 2537 use_cache ? dst_cache : NULL); 2538 if (IS_ERR(ndst)) { 2539 err = PTR_ERR(ndst); 2540 ndst = NULL; 2541 goto tx_error; 2542 } 2543 2544 if (!info) { 2545 u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags; 2546 2547 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6, 2548 dst_port, ifindex, vni, 2549 ndst, rt6i_flags); 2550 if (err) 2551 goto out_unlock; 2552 } 2553 2554 err = skb_tunnel_check_pmtu(skb, ndst, 2555 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6), 2556 netif_is_any_bridge_port(dev)); 2557 if (err < 0) { 2558 goto tx_error; 2559 } else if (err) { 2560 if (info) { 2561 struct ip_tunnel_info *unclone; 2562 2563 unclone = skb_tunnel_info_unclone(skb); 2564 if (unlikely(!unclone)) 2565 goto tx_error; 2566 2567 unclone->key.u.ipv6.src = pkey->u.ipv6.dst; 2568 unclone->key.u.ipv6.dst = saddr; 2569 } 2570 2571 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false); 2572 dst_release(ndst); 2573 goto out_unlock; 2574 } 2575 2576 tos = ip_tunnel_ecn_encap(tos, old_iph, skb); 2577 ttl = ttl ? : ip6_dst_hoplimit(ndst); 2578 skb_scrub_packet(skb, xnet); 2579 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), 2580 vni, md, flags, udp_sum); 2581 if (err < 0) 2582 goto tx_error; 2583 2584 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev, 2585 &saddr, &pkey->u.ipv6.dst, tos, ttl, 2586 pkey->label, src_port, dst_port, !udp_sum); 2587 #endif 2588 } 2589 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len); 2590 out_unlock: 2591 rcu_read_unlock(); 2592 return; 2593 2594 drop: 2595 dev_core_stats_tx_dropped_inc(dev); 2596 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0); 2597 dev_kfree_skb(skb); 2598 return; 2599 2600 tx_error: 2601 rcu_read_unlock(); 2602 if (err == -ELOOP) 2603 DEV_STATS_INC(dev, collisions); 2604 else if (err == -ENETUNREACH) 2605 DEV_STATS_INC(dev, tx_carrier_errors); 2606 dst_release(ndst); 2607 DEV_STATS_INC(dev, tx_errors); 2608 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0); 2609 kfree_skb(skb); 2610 } 2611
diff --git a/drivers/net/vxlan/vxlan_core.c b/drivers/net/vxlan/vxlan_core.c index 567cb3faab70..b510855c2b2a 100644 --- a/drivers/net/vxlan/vxlan_core.c +++ b/drivers/net/vxlan/vxlan_core.c @@ -2339,7 +2339,7 @@ void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, struct ip_tunnel_key *pkey; struct ip_tunnel_key key; struct vxlan_dev *vxlan = netdev_priv(dev); - const struct iphdr *old_iph = ip_hdr(skb); + const struct iphdr *old_iph; struct vxlan_metadata _md; struct vxlan_metadata *md = &_md; unsigned int pkt_len = skb->len; @@ -2355,6 +2355,16 @@ void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); __be32 vni = 0; + if (!(flags & VXLAN_F_GPE) || skb->protocol == ETH_P_TEB) { + if (!skb_vlan_inet_prepare(skb)) + goto drop; + } else { + if (!pskb_inet_may_pull(skb)) + goto drop; + } + + old_iph = ip_hdr(skb); + info = skb_tunnel_info(skb); use_cache = ip_tunnel_dst_cache_usable(skb, info);
Ensure the inner IP header is part of the skb's linear data before setting old_iph. Otherwise, on a non-linear skb, old_iph could point outside of the packet data. VXLAN-GPE can carry IP packets directly. Use pskb_inet_may_pull() in that case. Otherwise use skb_vlan_inet_prepare() to handle Ethernet header and potential VLANs. Fixes: d342894c5d2f ("vxlan: virtual extensible lan") Signed-off-by: Guillaume Nault <gnault@redhat.com> --- v2: * Handle the case of VXLAN-GPE carrying Ethernet frames (Paolo) * s/fragmented skb/non-linear skb/ (Eric) drivers/net/vxlan/vxlan_core.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-)