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, 226 insertions, 0 deletions

diff --git a/net/rds/tcp_send.c b/net/rds/tcp_send.c
new file mode 100644
index 000000000..7c52acc74
--- /dev/null
+++ b/net/rds/tcp_send.c
@@ -0,0 +1,226 @@
+/*
+ * Copyright (c) 2006, 2017 Oracle and/or its affiliates. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ */
+#include <linux/kernel.h>
+#include <linux/in.h>
+#include <net/tcp.h>
+
+#include "rds_single_path.h"
+#include "rds.h"
+#include "tcp.h"
+
+void rds_tcp_xmit_path_prepare(struct rds_conn_path *cp)
+{
+	struct rds_tcp_connection *tc = cp->cp_transport_data;
+
+	tcp_sock_set_cork(tc->t_sock->sk, true);
+}
+
+void rds_tcp_xmit_path_complete(struct rds_conn_path *cp)
+{
+	struct rds_tcp_connection *tc = cp->cp_transport_data;
+
+	tcp_sock_set_cork(tc->t_sock->sk, false);
+}
+
+/* the core send_sem serializes this with other xmit and shutdown */
+static int rds_tcp_sendmsg(struct socket *sock, void *data, unsigned int len)
+{
+	struct kvec vec = {
+		.iov_base = data,
+		.iov_len = len,
+	};
+	struct msghdr msg = {
+		.msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL,
+	};
+
+	return kernel_sendmsg(sock, &msg, &vec, 1, vec.iov_len);
+}
+
+/* the core send_sem serializes this with other xmit and shutdown */
+int rds_tcp_xmit(struct rds_connection *conn, struct rds_message *rm,
+		 unsigned int hdr_off, unsigned int sg, unsigned int off)
+{
+	struct rds_conn_path *cp = rm->m_inc.i_conn_path;
+	struct rds_tcp_connection *tc = cp->cp_transport_data;
+	struct msghdr msg = {};
+	struct bio_vec bvec;
+	int done = 0;
+	int ret = 0;
+
+	if (hdr_off == 0) {
+		/*
+		 * m_ack_seq is set to the sequence number of the last byte of
+		 * header and data.  see rds_tcp_is_acked().
+		 */
+		tc->t_last_sent_nxt = rds_tcp_write_seq(tc);
+		rm->m_ack_seq = tc->t_last_sent_nxt +
+				sizeof(struct rds_header) +
+				be32_to_cpu(rm->m_inc.i_hdr.h_len) - 1;
+		smp_mb__before_atomic();
+		set_bit(RDS_MSG_HAS_ACK_SEQ, &rm->m_flags);
+		tc->t_last_expected_una = rm->m_ack_seq + 1;
+
+		if (test_bit(RDS_MSG_RETRANSMITTED, &rm->m_flags))
+			rm->m_inc.i_hdr.h_flags |= RDS_FLAG_RETRANSMITTED;
+
+		rdsdebug("rm %p tcp nxt %u ack_seq %llu\n",
+			 rm, rds_tcp_write_seq(tc),
+			 (unsigned long long)rm->m_ack_seq);
+	}
+
+	if (hdr_off < sizeof(struct rds_header)) {
+		/* see rds_tcp_write_space() */
+		set_bit(SOCK_NOSPACE, &tc->t_sock->sk->sk_socket->flags);
+
+		ret = rds_tcp_sendmsg(tc->t_sock,
+				      (void *)&rm->m_inc.i_hdr + hdr_off,
+				      sizeof(rm->m_inc.i_hdr) - hdr_off);
+		if (ret < 0)
+			goto out;
+		done += ret;
+		if (hdr_off + done != sizeof(struct rds_header))
+			goto out;
+	}
+
+	while (sg < rm->data.op_nents) {
+		msg.msg_flags = MSG_SPLICE_PAGES | MSG_DONTWAIT | MSG_NOSIGNAL;
+		if (sg + 1 < rm->data.op_nents)
+			msg.msg_flags |= MSG_MORE;
+
+		bvec_set_page(&bvec, sg_page(&rm->data.op_sg[sg]),
+			      rm->data.op_sg[sg].length - off,
+			      rm->data.op_sg[sg].offset + off);
+		iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1,
+			      rm->data.op_sg[sg].length - off);
+		ret = sock_sendmsg(tc->t_sock, &msg);
+		rdsdebug("tcp sendpage %p:%u:%u ret %d\n", (void *)sg_page(&rm->data.op_sg[sg]),
+			 rm->data.op_sg[sg].offset + off, rm->data.op_sg[sg].length - off,
+			 ret);
+		if (ret <= 0)
+			break;
+
+		off += ret;
+		done += ret;
+		if (off == rm->data.op_sg[sg].length) {
+			off = 0;
+			sg++;
+		}
+	}
+
+out:
+	if (ret <= 0) {
+		/* write_space will hit after EAGAIN, all else fatal */
+		if (ret == -EAGAIN) {
+			rds_tcp_stats_inc(s_tcp_sndbuf_full);
+			ret = 0;
+		} else {
+			/* No need to disconnect/reconnect if path_drop
+			 * has already been triggered, because, e.g., of
+			 * an incoming RST.
+			 */
+			if (rds_conn_path_up(cp)) {
+				pr_warn("RDS/tcp: send to %pI6c on cp [%d]"
+					"returned %d, "
+					"disconnecting and reconnecting\n",
+					&conn->c_faddr, cp->cp_index, ret);
+				rds_conn_path_drop(cp, false);
+			}
+		}
+	}
+	if (done == 0)
+		done = ret;
+	return done;
+}
+
+/*
+ * rm->m_ack_seq is set to the tcp sequence number that corresponds to the
+ * last byte of the message, including the header.  This means that the
+ * entire message has been received if rm->m_ack_seq is "before" the next
+ * unacked byte of the TCP sequence space.  We have to do very careful
+ * wrapping 32bit comparisons here.
+ */
+int rds_tcp_is_acked(struct rds_message *rm, uint64_t ack)
+{
+	if (!test_bit(RDS_MSG_HAS_ACK_SEQ, &rm->m_flags))
+		return 0;
+	return (__s32)((u32)rm->m_ack_seq - (u32)ack) < 0;
+}
+
+void rds_tcp_write_space(struct sock *sk)
+{
+	void (*write_space)(struct sock *sk);
+	struct rds_conn_path *cp;
+	struct rds_tcp_connection *tc;
+
+	read_lock_bh(&sk->sk_callback_lock);
+	cp = sk->sk_user_data;
+	if (!cp) {
+		write_space = sk->sk_write_space;
+		goto out;
+	}
+
+	tc = cp->cp_transport_data;
+	rdsdebug("write_space for tc %p\n", tc);
+	write_space = tc->t_orig_write_space;
+	rds_tcp_stats_inc(s_tcp_write_space_calls);
+
+	rdsdebug("tcp una %u\n", rds_tcp_snd_una(tc));
+	tc->t_last_seen_una = rds_tcp_snd_una(tc);
+	rds_send_path_drop_acked(cp, rds_tcp_snd_una(tc), rds_tcp_is_acked);
+
+	rcu_read_lock();
+	if ((refcount_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf &&
+	    !rds_destroy_pending(cp->cp_conn))
+		queue_delayed_work(cp->cp_wq, &cp->cp_send_w, 0);
+	rcu_read_unlock();
+
+out:
+	read_unlock_bh(&sk->sk_callback_lock);
+
+	/*
+	 * write_space is only called when data leaves tcp's send queue if
+	 * SOCK_NOSPACE is set.  We set SOCK_NOSPACE every time we put
+	 * data in tcp's send queue because we use write_space to parse the
+	 * sequence numbers and notice that rds messages have been fully
+	 * received.
+	 *
+	 * tcp's write_space clears SOCK_NOSPACE if the send queue has more
+	 * than a certain amount of space. So we need to set it again *after*
+	 * we call tcp's write_space or else we might only get called on the
+	 * first of a series of incoming tcp acks.
+	 */
+	write_space(sk);
+
+	if (sk->sk_socket)
+		set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
+}