From a5f7a5bb3b7f28ba7e4fa246775b29a0e5537255 Mon Sep 17 00:00:00 2001 From: Jinke Han Date: Tue, 8 Sep 2026 15:37:42 +0800 Subject: x86/kprobes: Fix crash when probing CS CALL instructions When using eBPF to probe CS CALL instructions within a function, a crash can be triggered. The eBPF tool probes offset 257 of the __hrtimer_run_queues() function: <__hrtimer_run_queues+249>: nopl 0x0(%rax,%rax,1) <__hrtimer_run_queues+254>: mov %r14,%rdi <__hrtimer_run_queues+257>: cs call <__x86_indirect_thunk_r12> <__hrtimer_run_queues+263>: mov %eax,%r12d <__hrtimer_run_queues+266>: xchg %ax,%ax <__hrtimer_run_queues+268>: mov %r13,%rdi Which triggers this crash: BUG: unable to handle page fault for address: 00000000000f41c9 #PF: supervisor write access in kernel mode #PF: error_code(0x0002) - not-present page PGD 0 P4D 0 Oops: 0002 [#1] SMP NOPTI CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P RIP: 0010:__hrtimer_run_queues+0x106/0x230 Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is at the 6th byte of the above CS CALL instruction. Since the CS CALL instruction occupies 6 bytes, the exception occurred in the middle of that call instruction. The root cause is that when using eBPF tools to probe in the middle of a function, a kprobe with INT3 is used as the underlying implementation. During single-step emulation of the original CALL instruction, int3_emulate_call() assumes that the probed CALL instruction is 5 bytes long. However, the actual CS-prefixed CALL instruction occupies 6 bytes, so it constructs an incorrect exception return address. When the CPU returns from the kprobe handler, the next instruction to be executed is at the address of the last byte of that CS CALL instruction. Coincidentally, starting from that address, the CPU fetches and decodes a completely different instruction, which ultimately triggers a kernel crash. Fix the issue by using the actual instruction length obtained from the instruction decoder when constructing the exception return address, rather than relying on the hardcoded CALL_INSN_SIZE macro. [ mingo: Refined the changelog ] Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step") Suggested-by: Masami Hiramatsu (Google) Signed-off-by: Jinke Han Signed-off-by: Ingo Molnar Reviewed-by: Masami Hiramatsu (Google) Acked-by: Yafang Shao Acked-by: Borislav Petkov Cc: Peter Zijlstra Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000 --- drivers/net/wireguard/device.c | 475 +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 475 insertions(+) create mode 100644 drivers/net/wireguard/device.c (limited to 'drivers/net/wireguard/device.c') diff --git a/drivers/net/wireguard/device.c b/drivers/net/wireguard/device.c new file mode 100644 index 000000000..67b07ee2d --- /dev/null +++ b/drivers/net/wireguard/device.c @@ -0,0 +1,475 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2015-2019 Jason A. Donenfeld . All Rights Reserved. + */ + +#include "queueing.h" +#include "socket.h" +#include "timers.h" +#include "device.h" +#include "ratelimiter.h" +#include "peer.h" +#include "messages.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +static LIST_HEAD(device_list); + +static int wg_open(struct net_device *dev) +{ + struct in_device *dev_v4 = __in_dev_get_rtnl(dev); + struct inet6_dev *dev_v6 = __in6_dev_get(dev); + struct wg_device *wg = netdev_priv(dev); + struct wg_peer *peer; + int ret; + + if (dev_v4) { + /* At some point we might put this check near the ip_rt_send_ + * redirect call of ip_forward in net/ipv4/ip_forward.c, similar + * to the current secpath check. + */ + IN_DEV_CONF_SET(dev_v4, SEND_REDIRECTS, false); + IPV4_DEVCONF_ALL(dev_net(dev), SEND_REDIRECTS) = false; + } + if (dev_v6) + dev_v6->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_NONE; + + mutex_lock(&wg->device_update_lock); + ret = wg_socket_init(wg, wg->incoming_port); + if (ret < 0) + goto out; + list_for_each_entry(peer, &wg->peer_list, peer_list) { + wg_packet_send_staged_packets(peer); + if (peer->persistent_keepalive_interval) + wg_packet_send_keepalive(peer); + } +out: + mutex_unlock(&wg->device_update_lock); + return ret; +} + +static int wg_pm_notification(struct notifier_block *nb, unsigned long action, void *data) +{ + struct wg_device *wg; + struct wg_peer *peer; + + /* If the machine is constantly suspending and resuming, as part of + * its normal operation rather than as a somewhat rare event, then we + * don't actually want to clear keys. + */ + if (IS_ENABLED(CONFIG_PM_AUTOSLEEP) || + IS_ENABLED(CONFIG_PM_USERSPACE_AUTOSLEEP)) + return 0; + + if (action != PM_HIBERNATION_PREPARE && action != PM_SUSPEND_PREPARE) + return 0; + + rtnl_lock(); + list_for_each_entry(wg, &device_list, device_list) { + mutex_lock(&wg->device_update_lock); + list_for_each_entry(peer, &wg->peer_list, peer_list) { + timer_delete(&peer->timer_zero_key_material); + wg_noise_handshake_clear(&peer->handshake); + wg_noise_keypairs_clear(&peer->keypairs); + } + mutex_unlock(&wg->device_update_lock); + } + rtnl_unlock(); + rcu_barrier(); + return 0; +} + +static struct notifier_block pm_notifier = { .notifier_call = wg_pm_notification }; + +static int wg_vm_notification(struct notifier_block *nb, unsigned long action, void *data) +{ + struct wg_device *wg; + struct wg_peer *peer; + + rtnl_lock(); + list_for_each_entry(wg, &device_list, device_list) { + mutex_lock(&wg->device_update_lock); + list_for_each_entry(peer, &wg->peer_list, peer_list) + wg_noise_expire_current_peer_keypairs(peer); + mutex_unlock(&wg->device_update_lock); + } + rtnl_unlock(); + return 0; +} + +static struct notifier_block vm_notifier = { .notifier_call = wg_vm_notification }; + +static int wg_stop(struct net_device *dev) +{ + struct wg_device *wg = netdev_priv(dev); + struct wg_peer *peer; + struct sk_buff *skb; + + mutex_lock(&wg->device_update_lock); + list_for_each_entry(peer, &wg->peer_list, peer_list) { + wg_packet_purge_staged_packets(peer); + wg_timers_stop(peer); + wg_noise_handshake_clear(&peer->handshake); + wg_noise_keypairs_clear(&peer->keypairs); + wg_noise_reset_last_sent_handshake(&peer->last_sent_handshake); + } + mutex_unlock(&wg->device_update_lock); + while ((skb = ptr_ring_consume(&wg->handshake_queue.ring)) != NULL) + kfree_skb(skb); + atomic_set(&wg->handshake_queue_len, 0); + wg_socket_reinit(wg, NULL, NULL); + return 0; +} + +static netdev_tx_t wg_xmit(struct sk_buff *skb, struct net_device *dev) +{ + struct wg_device *wg = netdev_priv(dev); + struct sk_buff_head packets; + struct wg_peer *peer; + struct sk_buff *next; + sa_family_t family; + u32 mtu; + int ret; + + if (unlikely(!wg_check_packet_protocol(skb))) { + ret = -EPROTONOSUPPORT; + net_dbg_ratelimited("%s: Invalid IP packet\n", dev->name); + goto err; + } + + peer = wg_allowedips_lookup_dst(&wg->peer_allowedips, skb); + if (unlikely(!peer)) { + ret = -ENOKEY; + if (skb->protocol == htons(ETH_P_IP)) + net_dbg_ratelimited("%s: No peer has allowed IPs matching %pI4\n", + dev->name, &ip_hdr(skb)->daddr); + else if (skb->protocol == htons(ETH_P_IPV6)) + net_dbg_ratelimited("%s: No peer has allowed IPs matching %pI6\n", + dev->name, &ipv6_hdr(skb)->daddr); + goto err_icmp; + } + + family = READ_ONCE(peer->endpoint.addr.sa_family); + if (unlikely(family != AF_INET && family != AF_INET6)) { + ret = -EDESTADDRREQ; + net_dbg_ratelimited("%s: No valid endpoint has been configured or discovered for peer %llu\n", + dev->name, peer->internal_id); + goto err_peer; + } + + mtu = skb_valid_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu; + + __skb_queue_head_init(&packets); + if (!skb_is_gso(skb)) { + skb_mark_not_on_list(skb); + } else { + struct sk_buff *segs = skb_gso_segment(skb, 0); + + if (IS_ERR(segs)) { + ret = PTR_ERR(segs); + goto err_peer; + } + dev_kfree_skb(skb); + skb = segs; + } + + skb_list_walk_safe(skb, skb, next) { + skb_mark_not_on_list(skb); + + skb = skb_share_check(skb, GFP_ATOMIC); + if (unlikely(!skb)) + continue; + + /* We only need to keep the original dst around for icmp, + * so at this point we're in a position to drop it. + */ + skb_dst_drop(skb); + + PACKET_CB(skb)->mtu = mtu; + + __skb_queue_tail(&packets, skb); + } + + spin_lock_bh(&peer->staged_packet_queue.lock); + /* If the queue is getting too big, we start removing the oldest packets + * until it's small again. We do this before adding the new packet, so + * we don't remove GSO segments that are in excess. + */ + while (skb_queue_len(&peer->staged_packet_queue) > MAX_STAGED_PACKETS) { + dev_kfree_skb(__skb_dequeue(&peer->staged_packet_queue)); + DEV_STATS_INC(dev, tx_dropped); + } + skb_queue_splice_tail(&packets, &peer->staged_packet_queue); + spin_unlock_bh(&peer->staged_packet_queue.lock); + + wg_packet_send_staged_packets(peer); + + wg_peer_put(peer); + return NETDEV_TX_OK; + +err_peer: + wg_peer_put(peer); +err_icmp: + if (skb->protocol == htons(ETH_P_IP)) + icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0); + else if (skb->protocol == htons(ETH_P_IPV6)) + icmpv6_ndo_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0); +err: + DEV_STATS_INC(dev, tx_errors); + kfree_skb(skb); + return ret; +} + +static const struct net_device_ops netdev_ops = { + .ndo_open = wg_open, + .ndo_stop = wg_stop, + .ndo_start_xmit = wg_xmit, +}; + +static void wg_destruct(struct net_device *dev) +{ + struct wg_device *wg = netdev_priv(dev); + + rtnl_lock(); + list_del(&wg->device_list); + rtnl_unlock(); + mutex_lock(&wg->device_update_lock); + rcu_assign_pointer(wg->creating_net, NULL); + wg->incoming_port = 0; + wg_socket_reinit(wg, NULL, NULL); + /* The final references are cleared in the below calls to destroy_workqueue. */ + wg_peer_remove_all(wg); + destroy_workqueue(wg->handshake_receive_wq); + destroy_workqueue(wg->handshake_send_wq); + destroy_workqueue(wg->packet_crypt_wq); + wg_packet_queue_free(&wg->handshake_queue, true); + wg_packet_queue_free(&wg->decrypt_queue, false); + wg_packet_queue_free(&wg->encrypt_queue, false); + rcu_barrier(); /* Wait for all the peers to be actually freed. */ + wg_ratelimiter_uninit(); + memzero_explicit(&wg->static_identity, sizeof(wg->static_identity)); + kvfree(wg->index_hashtable); + kvfree(wg->peer_hashtable); + mutex_unlock(&wg->device_update_lock); + + pr_debug("%s: Interface destroyed\n", dev->name); + free_netdev(dev); +} + +static const struct device_type device_type = { .name = KBUILD_MODNAME }; + +static void wg_setup(struct net_device *dev) +{ + struct wg_device *wg = netdev_priv(dev); + enum { WG_NETDEV_FEATURES = NETIF_F_HW_CSUM | NETIF_F_RXCSUM | + NETIF_F_SG | NETIF_F_GSO | + NETIF_F_GSO_SOFTWARE | NETIF_F_HIGHDMA }; + const int overhead = MESSAGE_MINIMUM_LENGTH + sizeof(struct udphdr) + + max(sizeof(struct ipv6hdr), sizeof(struct iphdr)); + + dev->netdev_ops = &netdev_ops; + dev->header_ops = &ip_tunnel_header_ops; + dev->hard_header_len = 0; + dev->addr_len = 0; + dev->needed_headroom = DATA_PACKET_HEAD_ROOM; + dev->needed_tailroom = noise_encrypted_len(MESSAGE_PADDING_MULTIPLE); + dev->type = ARPHRD_NONE; + dev->flags = IFF_POINTOPOINT | IFF_NOARP; + dev->priv_flags |= IFF_NO_QUEUE; + dev->lltx = true; + dev->features |= WG_NETDEV_FEATURES; + dev->hw_features |= WG_NETDEV_FEATURES; + dev->hw_enc_features |= WG_NETDEV_FEATURES; + dev->mtu = ETH_DATA_LEN - overhead; + dev->max_mtu = round_down(INT_MAX, MESSAGE_PADDING_MULTIPLE) - overhead; + dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS; + + SET_NETDEV_DEVTYPE(dev, &device_type); + + /* We need to keep the dst around in case of icmp replies. */ + netif_keep_dst(dev); + + netif_set_tso_max_size(dev, GSO_MAX_SIZE); + + wg->dev = dev; +} + +static int wg_newlink(struct net_device *dev, + struct rtnl_newlink_params *params, + struct netlink_ext_ack *extack) +{ + struct net *link_net = rtnl_newlink_link_net(params); + struct wg_device *wg = netdev_priv(dev); + int ret = -ENOMEM; + + rcu_assign_pointer(wg->creating_net, link_net); + init_rwsem(&wg->static_identity.lock); + mutex_init(&wg->socket_update_lock); + mutex_init(&wg->device_update_lock); + wg_allowedips_init(&wg->peer_allowedips); + wg_cookie_checker_init(&wg->cookie_checker, wg); + INIT_LIST_HEAD(&wg->peer_list); + wg->device_update_gen = 1; + + wg->peer_hashtable = wg_pubkey_hashtable_alloc(); + if (!wg->peer_hashtable) + return ret; + + wg->index_hashtable = wg_index_hashtable_alloc(); + if (!wg->index_hashtable) + goto err_free_peer_hashtable; + + wg->handshake_receive_wq = alloc_workqueue("wg-kex-%s", + WQ_CPU_INTENSIVE | WQ_FREEZABLE | WQ_PERCPU, 0, + dev->name); + if (!wg->handshake_receive_wq) + goto err_free_index_hashtable; + + wg->handshake_send_wq = alloc_workqueue("wg-kex-%s", + WQ_UNBOUND | WQ_FREEZABLE, 0, dev->name); + if (!wg->handshake_send_wq) + goto err_destroy_handshake_receive; + + wg->packet_crypt_wq = alloc_workqueue("wg-crypt-%s", + WQ_CPU_INTENSIVE | WQ_MEM_RECLAIM | WQ_PERCPU, 0, + dev->name); + if (!wg->packet_crypt_wq) + goto err_destroy_handshake_send; + + ret = wg_packet_queue_init(&wg->encrypt_queue, wg_packet_encrypt_worker, + MAX_QUEUED_PACKETS); + if (ret < 0) + goto err_destroy_packet_crypt; + + ret = wg_packet_queue_init(&wg->decrypt_queue, wg_packet_decrypt_worker, + MAX_QUEUED_PACKETS); + if (ret < 0) + goto err_free_encrypt_queue; + + ret = wg_packet_queue_init(&wg->handshake_queue, wg_packet_handshake_receive_worker, + MAX_QUEUED_INCOMING_HANDSHAKES); + if (ret < 0) + goto err_free_decrypt_queue; + + ret = wg_ratelimiter_init(); + if (ret < 0) + goto err_free_handshake_queue; + + netif_threaded_enable(dev); + ret = register_netdevice(dev); + if (ret < 0) + goto err_uninit_ratelimiter; + + list_add(&wg->device_list, &device_list); + + /* We wait until the end to assign priv_destructor, so that + * register_netdevice doesn't call it for us if it fails. + */ + dev->priv_destructor = wg_destruct; + + pr_debug("%s: Interface created\n", dev->name); + return ret; + +err_uninit_ratelimiter: + wg_ratelimiter_uninit(); +err_free_handshake_queue: + wg_packet_queue_free(&wg->handshake_queue, false); +err_free_decrypt_queue: + wg_packet_queue_free(&wg->decrypt_queue, false); +err_free_encrypt_queue: + wg_packet_queue_free(&wg->encrypt_queue, false); +err_destroy_packet_crypt: + destroy_workqueue(wg->packet_crypt_wq); +err_destroy_handshake_send: + destroy_workqueue(wg->handshake_send_wq); +err_destroy_handshake_receive: + destroy_workqueue(wg->handshake_receive_wq); +err_free_index_hashtable: + kvfree(wg->index_hashtable); +err_free_peer_hashtable: + kvfree(wg->peer_hashtable); + return ret; +} + +static struct rtnl_link_ops link_ops __read_mostly = { + .kind = KBUILD_MODNAME, + .priv_size = sizeof(struct wg_device), + .setup = wg_setup, + .newlink = wg_newlink, +}; + +static void __net_exit wg_netns_exit_rtnl(struct net *net, struct list_head *dev_kill_list) +{ + struct wg_device *wg; + struct wg_peer *peer; + + list_for_each_entry(wg, &device_list, device_list) { + if (rcu_access_pointer(wg->creating_net) == net) { + pr_debug("%s: Creating namespace exiting\n", wg->dev->name); + netif_carrier_off(wg->dev); + mutex_lock(&wg->device_update_lock); + rcu_assign_pointer(wg->creating_net, NULL); + wg_socket_reinit(wg, NULL, NULL); + list_for_each_entry(peer, &wg->peer_list, peer_list) + wg_socket_clear_peer_endpoint_src(peer); + mutex_unlock(&wg->device_update_lock); + } + } +} + +static struct pernet_operations pernet_ops __read_mostly = { + .exit_rtnl = wg_netns_exit_rtnl +}; + +int __init wg_device_init(void) +{ + int ret; + + ret = register_pm_notifier(&pm_notifier); + if (ret) + return ret; + + ret = register_random_vmfork_notifier(&vm_notifier); + if (ret) + goto error_pm; + + ret = register_pernet_device(&pernet_ops); + if (ret) + goto error_vm; + + ret = rtnl_link_register(&link_ops); + if (ret) + goto error_pernet; + + return 0; + +error_pernet: + unregister_pernet_device(&pernet_ops); +error_vm: + unregister_random_vmfork_notifier(&vm_notifier); +error_pm: + unregister_pm_notifier(&pm_notifier); + return ret; +} + +void wg_device_uninit(void) +{ + rtnl_link_unregister(&link_ops); + unregister_pernet_device(&pernet_ops); + unregister_random_vmfork_notifier(&vm_notifier); + unregister_pm_notifier(&pm_notifier); + rcu_barrier(); +} -- cgit v1.3.1