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/ratelimiter.c | 223 ++++++++++++++++++++++++++++++++++++ 1 file changed, 223 insertions(+) create mode 100644 drivers/net/wireguard/ratelimiter.c (limited to 'drivers/net/wireguard/ratelimiter.c') diff --git a/drivers/net/wireguard/ratelimiter.c b/drivers/net/wireguard/ratelimiter.c new file mode 100644 index 000000000..2b3867f69 --- /dev/null +++ b/drivers/net/wireguard/ratelimiter.c @@ -0,0 +1,223 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (C) 2015-2019 Jason A. Donenfeld . All Rights Reserved. + */ + +#include "ratelimiter.h" +#include +#include +#include +#include + +static struct kmem_cache *entry_cache; +static hsiphash_key_t key; +static spinlock_t table_lock = __SPIN_LOCK_UNLOCKED("ratelimiter_table_lock"); +static DEFINE_MUTEX(init_lock); +static u64 init_refcnt; /* Protected by init_lock, hence not atomic. */ +static atomic_t total_entries = ATOMIC_INIT(0); +static unsigned int max_entries, table_size; +static void wg_ratelimiter_gc_entries(struct work_struct *); +static DECLARE_DEFERRABLE_WORK(gc_work, wg_ratelimiter_gc_entries); +static struct hlist_head *table_v4; +#if IS_ENABLED(CONFIG_IPV6) +static struct hlist_head *table_v6; +#endif + +struct ratelimiter_entry { + u64 last_time_ns, tokens, ip; + void *net; + spinlock_t lock; + struct hlist_node hash; + struct rcu_head rcu; +}; + +enum { + PACKETS_PER_SECOND = 20, + PACKETS_BURSTABLE = 5, + PACKET_COST = NSEC_PER_SEC / PACKETS_PER_SECOND, + TOKEN_MAX = PACKET_COST * PACKETS_BURSTABLE +}; + +static void entry_free(struct rcu_head *rcu) +{ + kmem_cache_free(entry_cache, + container_of(rcu, struct ratelimiter_entry, rcu)); + atomic_dec(&total_entries); +} + +static void entry_uninit(struct ratelimiter_entry *entry) +{ + hlist_del_rcu(&entry->hash); + call_rcu(&entry->rcu, entry_free); +} + +/* Calling this function with a NULL work uninits all entries. */ +static void wg_ratelimiter_gc_entries(struct work_struct *work) +{ + const u64 now = ktime_get_coarse_boottime_ns(); + struct ratelimiter_entry *entry; + struct hlist_node *temp; + unsigned int i; + + for (i = 0; i < table_size; ++i) { + spin_lock(&table_lock); + hlist_for_each_entry_safe(entry, temp, &table_v4[i], hash) { + if (unlikely(!work) || + now - entry->last_time_ns > NSEC_PER_SEC) + entry_uninit(entry); + } +#if IS_ENABLED(CONFIG_IPV6) + hlist_for_each_entry_safe(entry, temp, &table_v6[i], hash) { + if (unlikely(!work) || + now - entry->last_time_ns > NSEC_PER_SEC) + entry_uninit(entry); + } +#endif + spin_unlock(&table_lock); + if (likely(work)) + cond_resched(); + } + if (likely(work)) + queue_delayed_work(system_power_efficient_wq, &gc_work, HZ); +} + +bool wg_ratelimiter_allow(struct sk_buff *skb, struct net *net) +{ + /* We only take the bottom half of the net pointer, so that we can hash + * 3 words in the end. This way, siphash's len param fits into the final + * u32, and we don't incur an extra round. + */ + const u32 net_word = (unsigned long)net; + struct ratelimiter_entry *entry; + struct hlist_head *bucket; + u64 ip; + + if (skb->protocol == htons(ETH_P_IP)) { + ip = (u64 __force)ip_hdr(skb)->saddr; + bucket = &table_v4[hsiphash_2u32(net_word, ip, &key) & + (table_size - 1)]; + } +#if IS_ENABLED(CONFIG_IPV6) + else if (skb->protocol == htons(ETH_P_IPV6)) { + /* Only use 64 bits, so as to ratelimit the whole /64. */ + memcpy(&ip, &ipv6_hdr(skb)->saddr, sizeof(ip)); + bucket = &table_v6[hsiphash_3u32(net_word, ip >> 32, ip, &key) & + (table_size - 1)]; + } +#endif + else + return false; + rcu_read_lock(); + hlist_for_each_entry_rcu(entry, bucket, hash) { + if (entry->net == net && entry->ip == ip) { + u64 now, tokens; + bool ret; + /* Quasi-inspired by nft_limit.c, but this is actually a + * slightly different algorithm. Namely, we incorporate + * the burst as part of the maximum tokens, rather than + * as part of the rate. + */ + spin_lock(&entry->lock); + now = ktime_get_coarse_boottime_ns(); + tokens = min_t(u64, TOKEN_MAX, + entry->tokens + now - + entry->last_time_ns); + entry->last_time_ns = now; + ret = tokens >= PACKET_COST; + entry->tokens = ret ? tokens - PACKET_COST : tokens; + spin_unlock(&entry->lock); + rcu_read_unlock(); + return ret; + } + } + rcu_read_unlock(); + + if (atomic_inc_return(&total_entries) > max_entries) + goto err_oom; + + entry = kmem_cache_alloc(entry_cache, GFP_KERNEL); + if (unlikely(!entry)) + goto err_oom; + + entry->net = net; + entry->ip = ip; + INIT_HLIST_NODE(&entry->hash); + spin_lock_init(&entry->lock); + entry->last_time_ns = ktime_get_coarse_boottime_ns(); + entry->tokens = TOKEN_MAX - PACKET_COST; + spin_lock(&table_lock); + hlist_add_head_rcu(&entry->hash, bucket); + spin_unlock(&table_lock); + return true; + +err_oom: + atomic_dec(&total_entries); + return false; +} + +int wg_ratelimiter_init(void) +{ + mutex_lock(&init_lock); + if (++init_refcnt != 1) + goto out; + + entry_cache = KMEM_CACHE(ratelimiter_entry, 0); + if (!entry_cache) + goto err; + + /* xt_hashlimit.c uses a slightly different algorithm for ratelimiting, + * but what it shares in common is that it uses a massive hashtable. So, + * we borrow their wisdom about good table sizes on different systems + * dependent on RAM. This calculation here comes from there. + */ + table_size = (totalram_pages() > (1U << 30) / PAGE_SIZE) ? 8192 : + max_t(unsigned long, 16, roundup_pow_of_two( + (totalram_pages() << PAGE_SHIFT) / + (1U << 14) / sizeof(struct hlist_head))); + max_entries = table_size * 8; + + table_v4 = kvzalloc_objs(*table_v4, table_size); + if (unlikely(!table_v4)) + goto err_kmemcache; + +#if IS_ENABLED(CONFIG_IPV6) + table_v6 = kvzalloc_objs(*table_v6, table_size); + if (unlikely(!table_v6)) { + kvfree(table_v4); + goto err_kmemcache; + } +#endif + + queue_delayed_work(system_power_efficient_wq, &gc_work, HZ); + get_random_bytes(&key, sizeof(key)); +out: + mutex_unlock(&init_lock); + return 0; + +err_kmemcache: + kmem_cache_destroy(entry_cache); +err: + --init_refcnt; + mutex_unlock(&init_lock); + return -ENOMEM; +} + +void wg_ratelimiter_uninit(void) +{ + mutex_lock(&init_lock); + if (!init_refcnt || --init_refcnt) + goto out; + + cancel_delayed_work_sync(&gc_work); + wg_ratelimiter_gc_entries(NULL); + rcu_barrier(); + kvfree(table_v4); +#if IS_ENABLED(CONFIG_IPV6) + kvfree(table_v6); +#endif + kmem_cache_destroy(entry_cache); +out: + mutex_unlock(&init_lock); +} + +#include "selftest/ratelimiter.c" -- cgit v1.3.1