x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

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) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 577 insertions, 0 deletions

diff --git a/net/xfrm/espintcp.c b/net/xfrm/espintcp.c
new file mode 100644
index 000000000..674aedc5a
--- /dev/null
+++ b/net/xfrm/espintcp.c
@@ -0,0 +1,577 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <net/tcp.h>
+#include <net/strparser.h>
+#include <net/xfrm.h>
+#include <net/esp.h>
+#include <net/espintcp.h>
+#include <linux/skmsg.h>
+#include <net/inet_common.h>
+#include <trace/events/sock.h>
+#include <net/hotdata.h>
+
+static void handle_nonesp(struct espintcp_ctx *ctx, struct sk_buff *skb,
+			  struct sock *sk)
+{
+	if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf ||
+	    !sk_rmem_schedule(sk, skb, skb->truesize)) {
+		XFRM_INC_STATS(sock_net(sk), LINUX_MIB_XFRMINERROR);
+		kfree_skb(skb);
+		return;
+	}
+
+	skb_set_owner_r(skb, sk);
+
+	memset(skb->cb, 0, sizeof(skb->cb));
+	skb_queue_tail(&ctx->ike_queue, skb);
+	ctx->saved_data_ready(sk);
+}
+
+static void handle_esp(struct sk_buff *skb, struct sock *sk)
+{
+	struct tcp_skb_cb *tcp_cb = (struct tcp_skb_cb *)skb->cb;
+
+	skb_reset_transport_header(skb);
+
+	/* restore IP CB, we need at least IP6CB->nhoff */
+	memmove(skb->cb, &tcp_cb->header, sizeof(tcp_cb->header));
+
+	rcu_read_lock();
+	skb->dev = dev_get_by_index_rcu(sock_net(sk), skb->skb_iif);
+	if (!skb->dev) {
+		XFRM_INC_STATS(sock_net(sk), LINUX_MIB_XFRMINERROR);
+		kfree_skb(skb);
+		goto out;
+	}
+	local_bh_disable();
+#if IS_ENABLED(CONFIG_IPV6)
+	if (sk->sk_family == AF_INET6)
+		xfrm6_rcv_encap(skb, IPPROTO_ESP, 0, TCP_ENCAP_ESPINTCP);
+	else
+#endif
+		xfrm4_rcv_encap(skb, IPPROTO_ESP, 0, TCP_ENCAP_ESPINTCP);
+	local_bh_enable();
+out:
+	rcu_read_unlock();
+}
+
+static void espintcp_rcv(struct strparser *strp, struct sk_buff *skb)
+{
+	struct espintcp_ctx *ctx = container_of(strp, struct espintcp_ctx,
+						strp);
+	struct strp_msg *rxm = strp_msg(skb);
+	int len = rxm->full_len - 2;
+	u32 nonesp_marker;
+	int err;
+
+	/* keepalive packet? */
+	if (unlikely(len == 1)) {
+		u8 data;
+
+		err = skb_copy_bits(skb, rxm->offset + 2, &data, 1);
+		if (err < 0) {
+			XFRM_INC_STATS(sock_net(strp->sk), LINUX_MIB_XFRMINHDRERROR);
+			kfree_skb(skb);
+			return;
+		}
+
+		if (data == 0xff) {
+			kfree_skb(skb);
+			return;
+		}
+	}
+
+	/* drop other short messages */
+	if (unlikely(len <= sizeof(nonesp_marker))) {
+		XFRM_INC_STATS(sock_net(strp->sk), LINUX_MIB_XFRMINHDRERROR);
+		kfree_skb(skb);
+		return;
+	}
+
+	err = skb_copy_bits(skb, rxm->offset + 2, &nonesp_marker,
+			    sizeof(nonesp_marker));
+	if (err < 0) {
+		XFRM_INC_STATS(sock_net(strp->sk), LINUX_MIB_XFRMINHDRERROR);
+		kfree_skb(skb);
+		return;
+	}
+
+	/* remove header, leave non-ESP marker/SPI */
+	if (!pskb_pull(skb, rxm->offset + 2)) {
+		XFRM_INC_STATS(sock_net(strp->sk), LINUX_MIB_XFRMINERROR);
+		kfree_skb(skb);
+		return;
+	}
+
+	if (pskb_trim(skb, rxm->full_len - 2) != 0) {
+		XFRM_INC_STATS(sock_net(strp->sk), LINUX_MIB_XFRMINERROR);
+		kfree_skb(skb);
+		return;
+	}
+
+	if (nonesp_marker == 0)
+		handle_nonesp(ctx, skb, strp->sk);
+	else
+		handle_esp(skb, strp->sk);
+}
+
+static int espintcp_parse(struct strparser *strp, struct sk_buff *skb)
+{
+	struct strp_msg *rxm = strp_msg(skb);
+	__be16 blen;
+	u16 len;
+	int err;
+
+	if (skb->len < rxm->offset + 2)
+		return 0;
+
+	err = skb_copy_bits(skb, rxm->offset, &blen, sizeof(blen));
+	if (err < 0)
+		return err;
+
+	len = be16_to_cpu(blen);
+	if (len < 2)
+		return -EINVAL;
+
+	return len;
+}
+
+static int espintcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
+			    int flags)
+{
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+	struct sk_buff *skb;
+	int err = 0;
+	int copied;
+	int off = 0;
+
+	skb = __skb_recv_datagram(sk, &ctx->ike_queue, flags, &off, &err);
+	if (!skb) {
+		if (err == -EAGAIN && sk->sk_shutdown & RCV_SHUTDOWN)
+			return 0;
+		return err;
+	}
+
+	copied = len;
+	if (copied > skb->len)
+		copied = skb->len;
+	else if (copied < skb->len)
+		msg->msg_flags |= MSG_TRUNC;
+
+	err = skb_copy_datagram_msg(skb, 0, msg, copied);
+	if (unlikely(err)) {
+		kfree_skb(skb);
+		return err;
+	}
+
+	if (flags & MSG_TRUNC)
+		copied = skb->len;
+	kfree_skb(skb);
+	return copied;
+}
+
+int espintcp_queue_out(struct sock *sk, struct sk_buff *skb)
+{
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+
+	if (skb_queue_len(&ctx->out_queue) >=
+	    READ_ONCE(net_hotdata.max_backlog)) {
+		kfree_skb(skb);
+		return -ENOBUFS;
+	}
+
+	__skb_queue_tail(&ctx->out_queue, skb);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(espintcp_queue_out);
+
+/* espintcp length field is 2B and length includes the length field's size */
+#define MAX_ESPINTCP_MSG (((1 << 16) - 1) - 2)
+
+static int espintcp_sendskb_locked(struct sock *sk, struct espintcp_msg *emsg,
+				   int flags)
+{
+	do {
+		int ret;
+
+		ret = skb_send_sock_locked(sk, emsg->skb,
+					   emsg->offset, emsg->len);
+		if (ret < 0)
+			return ret;
+
+		emsg->len -= ret;
+		emsg->offset += ret;
+	} while (emsg->len > 0);
+
+	kfree_skb(emsg->skb);
+	memset(emsg, 0, sizeof(*emsg));
+
+	return 0;
+}
+
+static int espintcp_sendskmsg_locked(struct sock *sk,
+				     struct espintcp_msg *emsg, int flags)
+{
+	struct msghdr msghdr = {
+		.msg_flags = flags | MSG_SPLICE_PAGES | MSG_MORE,
+	};
+	struct sk_msg *skmsg = &emsg->skmsg;
+	bool more = flags & MSG_MORE;
+	struct scatterlist *sg;
+	int ret;
+
+	do {
+		struct bio_vec bvec;
+
+		sg = &skmsg->sg.data[skmsg->sg.start];
+		if (sg_is_last(sg) && !more)
+			msghdr.msg_flags &= ~MSG_MORE;
+
+		bvec_set_page(&bvec, sg_page(sg), sg->length, sg->offset);
+		iov_iter_bvec(&msghdr.msg_iter, ITER_SOURCE, &bvec, 1, sg->length);
+		ret = tcp_sendmsg_locked(sk, &msghdr, sg->length);
+		if (ret < 0)
+			return ret;
+
+		sk_msg_free_partial(sk, skmsg, ret);
+	} while (skmsg->sg.size);
+
+	memset(emsg, 0, sizeof(*emsg));
+
+	return 0;
+}
+
+static int espintcp_push_msgs(struct sock *sk, int flags)
+{
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+	struct espintcp_msg *emsg = &ctx->partial;
+	int err;
+
+	if (!emsg->len)
+		return 0;
+
+	if (ctx->tx_running)
+		return -EAGAIN;
+	ctx->tx_running = 1;
+
+	if (emsg->skb)
+		err = espintcp_sendskb_locked(sk, emsg, flags);
+	else
+		err = espintcp_sendskmsg_locked(sk, emsg, flags);
+	if (err == -EAGAIN) {
+		ctx->tx_running = 0;
+		return flags & MSG_DONTWAIT ? -EAGAIN : 0;
+	}
+	if (!err)
+		memset(emsg, 0, sizeof(*emsg));
+
+	ctx->tx_running = 0;
+
+	return err;
+}
+
+int espintcp_push_skb(struct sock *sk, struct sk_buff *skb)
+{
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+	struct espintcp_msg *emsg = &ctx->partial;
+	unsigned int len;
+	int offset;
+
+	if (sk->sk_state != TCP_ESTABLISHED) {
+		kfree_skb(skb);
+		return -ECONNRESET;
+	}
+
+	offset = skb_transport_offset(skb);
+	len = skb->len - offset;
+
+	espintcp_push_msgs(sk, 0);
+
+	if (emsg->len) {
+		kfree_skb(skb);
+		return -ENOBUFS;
+	}
+
+	skb_set_owner_w(skb, sk);
+
+	emsg->offset = offset;
+	emsg->len = len;
+	emsg->skb = skb;
+
+	espintcp_push_msgs(sk, 0);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(espintcp_push_skb);
+
+static int espintcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
+{
+	long timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+	struct espintcp_msg *emsg = &ctx->partial;
+	struct iov_iter pfx_iter;
+	struct kvec pfx_iov = {};
+	size_t msglen = size + 2;
+	char buf[2] = {0};
+	int err, end;
+
+	if (msg->msg_flags & ~MSG_DONTWAIT)
+		return -EOPNOTSUPP;
+
+	if (size > MAX_ESPINTCP_MSG)
+		return -EMSGSIZE;
+
+	if (msg->msg_controllen)
+		return -EOPNOTSUPP;
+
+	lock_sock(sk);
+
+	err = espintcp_push_msgs(sk, msg->msg_flags & MSG_DONTWAIT);
+	if (err < 0) {
+		if (err != -EAGAIN || !(msg->msg_flags & MSG_DONTWAIT))
+			err = -ENOBUFS;
+		goto unlock;
+	}
+	if (emsg->len) {
+		err = -ENOBUFS;
+		goto unlock;
+	}
+
+	sk_msg_init(&emsg->skmsg);
+	while (1) {
+		/* only -ENOMEM is possible since we don't coalesce */
+		err = sk_msg_alloc(sk, &emsg->skmsg, msglen, 0);
+		if (!err)
+			break;
+
+		err = sk_stream_wait_memory(sk, &timeo);
+		if (err)
+			goto fail;
+	}
+
+	*((__be16 *)buf) = cpu_to_be16(msglen);
+	pfx_iov.iov_base = buf;
+	pfx_iov.iov_len = sizeof(buf);
+	iov_iter_kvec(&pfx_iter, ITER_SOURCE, &pfx_iov, 1, pfx_iov.iov_len);
+
+	err = sk_msg_memcopy_from_iter(sk, &pfx_iter, &emsg->skmsg,
+				       pfx_iov.iov_len);
+	if (err < 0)
+		goto fail;
+
+	err = sk_msg_memcopy_from_iter(sk, &msg->msg_iter, &emsg->skmsg, size);
+	if (err < 0)
+		goto fail;
+
+	end = emsg->skmsg.sg.end;
+	emsg->len = size;
+	sk_msg_iter_var_prev(end);
+	sg_mark_end(sk_msg_elem(&emsg->skmsg, end));
+
+	tcp_rate_check_app_limited(sk);
+
+	err = espintcp_push_msgs(sk, msg->msg_flags & MSG_DONTWAIT);
+	/* this message could be partially sent, keep it */
+
+	release_sock(sk);
+
+	return size;
+
+fail:
+	sk_msg_free(sk, &emsg->skmsg);
+	memset(emsg, 0, sizeof(*emsg));
+unlock:
+	release_sock(sk);
+	return err;
+}
+
+static struct proto espintcp_prot __ro_after_init;
+static struct proto_ops espintcp_ops __ro_after_init;
+static struct proto espintcp6_prot;
+static struct proto_ops espintcp6_ops;
+static DEFINE_MUTEX(tcpv6_prot_mutex);
+
+static void espintcp_data_ready(struct sock *sk)
+{
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+
+	trace_sk_data_ready(sk);
+
+	strp_data_ready(&ctx->strp);
+}
+
+static void espintcp_tx_work(struct work_struct *work)
+{
+	struct espintcp_ctx *ctx = container_of(work,
+						struct espintcp_ctx, work);
+	struct sock *sk = ctx->strp.sk;
+
+	lock_sock(sk);
+	if (!ctx->tx_running)
+		espintcp_push_msgs(sk, 0);
+	release_sock(sk);
+}
+
+static void espintcp_write_space(struct sock *sk)
+{
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+
+	schedule_work(&ctx->work);
+	ctx->saved_write_space(sk);
+}
+
+static void espintcp_destruct(struct sock *sk)
+{
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+
+	ctx->saved_destruct(sk);
+	kfree(ctx);
+}
+
+bool tcp_is_ulp_esp(struct sock *sk)
+{
+	return sk->sk_prot == &espintcp_prot || sk->sk_prot == &espintcp6_prot;
+}
+EXPORT_SYMBOL_GPL(tcp_is_ulp_esp);
+
+static void build_protos(struct proto *espintcp_prot,
+			 struct proto_ops *espintcp_ops,
+			 const struct proto *orig_prot,
+			 const struct proto_ops *orig_ops);
+static int espintcp_init_sk(struct sock *sk)
+{
+	struct inet_connection_sock *icsk = inet_csk(sk);
+	struct strp_callbacks cb = {
+		.rcv_msg = espintcp_rcv,
+		.parse_msg = espintcp_parse,
+	};
+	struct espintcp_ctx *ctx;
+	int err;
+
+	/* sockmap is not compatible with espintcp */
+	if (sk->sk_user_data)
+		return -EBUSY;
+
+	ctx = kzalloc_obj(*ctx);
+	if (!ctx)
+		return -ENOMEM;
+
+	err = strp_init(&ctx->strp, sk, &cb);
+	if (err)
+		goto free;
+
+	__sk_dst_reset(sk);
+
+	strp_check_rcv(&ctx->strp);
+	skb_queue_head_init(&ctx->ike_queue);
+	skb_queue_head_init(&ctx->out_queue);
+
+	if (sk->sk_family == AF_INET) {
+		sk->sk_prot = &espintcp_prot;
+		sk->sk_socket->ops = &espintcp_ops;
+	} else {
+		mutex_lock(&tcpv6_prot_mutex);
+		if (!espintcp6_prot.recvmsg)
+			build_protos(&espintcp6_prot, &espintcp6_ops, sk->sk_prot, sk->sk_socket->ops);
+		mutex_unlock(&tcpv6_prot_mutex);
+
+		sk->sk_prot = &espintcp6_prot;
+		sk->sk_socket->ops = &espintcp6_ops;
+	}
+	ctx->saved_data_ready = sk->sk_data_ready;
+	ctx->saved_write_space = sk->sk_write_space;
+	ctx->saved_destruct = sk->sk_destruct;
+	sk->sk_data_ready = espintcp_data_ready;
+	sk->sk_write_space = espintcp_write_space;
+	sk->sk_destruct = espintcp_destruct;
+	rcu_assign_pointer(icsk->icsk_ulp_data, ctx);
+	INIT_WORK(&ctx->work, espintcp_tx_work);
+
+	/* avoid using task_frag */
+	sk->sk_allocation = GFP_ATOMIC;
+	sk->sk_use_task_frag = false;
+
+	return 0;
+
+free:
+	kfree(ctx);
+	return err;
+}
+
+static void espintcp_release(struct sock *sk)
+{
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+	struct sk_buff_head queue;
+	struct sk_buff *skb;
+
+	__skb_queue_head_init(&queue);
+	skb_queue_splice_init(&ctx->out_queue, &queue);
+
+	while ((skb = __skb_dequeue(&queue)))
+		espintcp_push_skb(sk, skb);
+
+	tcp_release_cb(sk);
+}
+
+static void espintcp_close(struct sock *sk, long timeout)
+{
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+	struct espintcp_msg *emsg = &ctx->partial;
+
+	strp_stop(&ctx->strp);
+
+	sk->sk_prot = &tcp_prot;
+
+	synchronize_rcu();
+
+	disable_work_sync(&ctx->work);
+	strp_done(&ctx->strp);
+
+	skb_queue_purge(&ctx->out_queue);
+	skb_queue_purge(&ctx->ike_queue);
+
+	if (emsg->len) {
+		if (emsg->skb)
+			kfree_skb(emsg->skb);
+		else
+			sk_msg_free(sk, &emsg->skmsg);
+	}
+
+	tcp_close(sk, timeout);
+}
+
+static __poll_t espintcp_poll(struct file *file, struct socket *sock,
+			      poll_table *wait)
+{
+	struct sock *sk = sock->sk;
+	struct espintcp_ctx *ctx = espintcp_getctx(sk);
+
+	return datagram_poll_queue(file, sock, wait, &ctx->ike_queue);
+}
+
+static void build_protos(struct proto *espintcp_prot,
+			 struct proto_ops *espintcp_ops,
+			 const struct proto *orig_prot,
+			 const struct proto_ops *orig_ops)
+{
+	memcpy(espintcp_prot, orig_prot, sizeof(struct proto));
+	memcpy(espintcp_ops, orig_ops, sizeof(struct proto_ops));
+	espintcp_prot->sendmsg = espintcp_sendmsg;
+	espintcp_prot->recvmsg = espintcp_recvmsg;
+	espintcp_prot->close = espintcp_close;
+	espintcp_prot->release_cb = espintcp_release;
+	espintcp_ops->poll = espintcp_poll;
+}
+
+static struct tcp_ulp_ops espintcp_ulp __read_mostly = {
+	.name = "espintcp",
+	.owner = THIS_MODULE,
+	.init = espintcp_init_sk,
+};
+
+void __init espintcp_init(void)
+{
+	build_protos(&espintcp_prot, &espintcp_ops, &tcp_prot, &inet_stream_ops);
+
+	tcp_register_ulp(&espintcp_ulp);
+}