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, 151 insertions, 0 deletions

diff --git a/net/llc/llc_if.c b/net/llc/llc_if.c
new file mode 100644
index 000000000..1514362e6
--- /dev/null
+++ b/net/llc/llc_if.c
@@ -0,0 +1,151 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * llc_if.c - Defines LLC interface to upper layer
+ *
+ * Copyright (c) 1997 by Procom Technology, Inc.
+ * 		 2001-2003 by Arnaldo Carvalho de Melo <acme@conectiva.com.br>
+ */
+#include <linux/gfp.h>
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/netdevice.h>
+#include <linux/errno.h>
+#include <net/llc_if.h>
+#include <net/llc_sap.h>
+#include <net/llc_s_ev.h>
+#include <net/llc_conn.h>
+#include <net/sock.h>
+#include <net/llc_c_ev.h>
+#include <net/llc_c_ac.h>
+#include <net/llc_c_st.h>
+#include <net/tcp_states.h>
+
+/**
+ *	llc_build_and_send_pkt - Connection data sending for upper layers.
+ *	@sk: connection
+ *	@skb: packet to send
+ *
+ *	This function is called when upper layer wants to send data using
+ *	connection oriented communication mode. During sending data, connection
+ *	will be locked and received frames and expired timers will be queued.
+ *	Returns 0 for success, -ECONNABORTED when the connection already
+ *	closed and -EBUSY when sending data is not permitted in this state or
+ *	LLC has send an I pdu with p bit set to 1 and is waiting for it's
+ *	response.
+ *
+ *	This function always consumes a reference to the skb.
+ */
+int llc_build_and_send_pkt(struct sock *sk, struct sk_buff *skb)
+{
+	struct llc_conn_state_ev *ev;
+	int rc = -ECONNABORTED;
+	struct llc_sock *llc = llc_sk(sk);
+
+	if (unlikely(llc->state == LLC_CONN_STATE_ADM))
+		goto out_free;
+	rc = -EBUSY;
+	if (unlikely(llc_data_accept_state(llc->state) || /* data_conn_refuse */
+		     llc->p_flag)) {
+		llc->failed_data_req = 1;
+		goto out_free;
+	}
+	ev = llc_conn_ev(skb);
+	ev->type      = LLC_CONN_EV_TYPE_PRIM;
+	ev->prim      = LLC_DATA_PRIM;
+	ev->prim_type = LLC_PRIM_TYPE_REQ;
+	skb->dev      = llc->dev;
+	return llc_conn_state_process(sk, skb);
+
+out_free:
+	kfree_skb(skb);
+	return rc;
+}
+
+/**
+ *	llc_establish_connection - Called by upper layer to establish a conn
+ *	@sk: connection
+ *	@lmac: local mac address
+ *	@dmac: destination mac address
+ *	@dsap: destination sap
+ *
+ *	Upper layer calls this to establish an LLC connection with a remote
+ *	machine. This function packages a proper event and sends it connection
+ *	component state machine. Success or failure of connection
+ *	establishment will inform to upper layer via calling it's confirm
+ *	function and passing proper information.
+ */
+int llc_establish_connection(struct sock *sk, const u8 *lmac, u8 *dmac, u8 dsap)
+{
+	int rc = -EISCONN;
+	struct llc_addr laddr, daddr;
+	struct sk_buff *skb;
+	struct llc_sock *llc = llc_sk(sk);
+	struct sock *existing;
+
+	laddr.lsap = llc->sap->laddr.lsap;
+	daddr.lsap = dsap;
+	memcpy(daddr.mac, dmac, sizeof(daddr.mac));
+	memcpy(laddr.mac, lmac, sizeof(laddr.mac));
+	existing = llc_lookup_established(llc->sap, &daddr, &laddr, sock_net(sk));
+	if (existing) {
+		if (existing->sk_state == TCP_ESTABLISHED) {
+			sk = existing;
+			goto out_put;
+		} else
+			sock_put(existing);
+	}
+	sock_hold(sk);
+	rc = -ENOMEM;
+	skb = alloc_skb(0, GFP_ATOMIC);
+	if (skb) {
+		struct llc_conn_state_ev *ev = llc_conn_ev(skb);
+
+		ev->type      = LLC_CONN_EV_TYPE_PRIM;
+		ev->prim      = LLC_CONN_PRIM;
+		ev->prim_type = LLC_PRIM_TYPE_REQ;
+		skb_set_owner_w(skb, sk);
+		rc = llc_conn_state_process(sk, skb);
+	}
+out_put:
+	sock_put(sk);
+	return rc;
+}
+
+/**
+ *	llc_send_disc - Called by upper layer to close a connection
+ *	@sk: connection to be closed
+ *
+ *	Upper layer calls this when it wants to close an established LLC
+ *	connection with a remote machine. This function packages a proper event
+ *	and sends it to connection component state machine. Returns 0 for
+ *	success, 1 otherwise.
+ */
+int llc_send_disc(struct sock *sk)
+{
+	u16 rc = 1;
+	struct llc_conn_state_ev *ev;
+	struct sk_buff *skb;
+
+	sock_hold(sk);
+	if (sk->sk_type != SOCK_STREAM || sk->sk_state != TCP_ESTABLISHED ||
+	    llc_sk(sk)->state == LLC_CONN_STATE_ADM ||
+	    llc_sk(sk)->state == LLC_CONN_OUT_OF_SVC)
+		goto out;
+	/*
+	 * Postpone unassigning the connection from its SAP and returning the
+	 * connection until all ACTIONs have been completely executed
+	 */
+	skb = alloc_skb(0, GFP_ATOMIC);
+	if (!skb)
+		goto out;
+	skb_set_owner_w(skb, sk);
+	sk->sk_state  = TCP_CLOSING;
+	ev	      = llc_conn_ev(skb);
+	ev->type      = LLC_CONN_EV_TYPE_PRIM;
+	ev->prim      = LLC_DISC_PRIM;
+	ev->prim_type = LLC_PRIM_TYPE_REQ;
+	rc = llc_conn_state_process(sk, skb);
+out:
+	sock_put(sk);
+	return rc;
+}