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, 197 insertions, 0 deletions

diff --git a/drivers/thunderbolt/trace.h b/drivers/thunderbolt/trace.h
new file mode 100644
index 000000000..6d0776514
--- /dev/null
+++ b/drivers/thunderbolt/trace.h
@@ -0,0 +1,197 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Thunderbolt tracing support
+ *
+ * Copyright (C) 2024, Intel Corporation
+ * Author: Mika Westerberg <mika.westerberg@linux.intel.com>
+ *	   Gil Fine <gil.fine@intel.com>
+ */
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM thunderbolt
+
+#if !defined(TB_TRACE_H_) || defined(TRACE_HEADER_MULTI_READ)
+#define TB_TRACE_H_
+
+#include <linux/trace_seq.h>
+#include <linux/tracepoint.h>
+
+#include "tb_msgs.h"
+
+#define tb_cfg_type_name(type)		{ type, #type }
+#define show_type_name(val)					\
+	__print_symbolic(val,					\
+		tb_cfg_type_name(TB_CFG_PKG_READ),		\
+		tb_cfg_type_name(TB_CFG_PKG_WRITE),		\
+		tb_cfg_type_name(TB_CFG_PKG_ERROR),		\
+		tb_cfg_type_name(TB_CFG_PKG_NOTIFY_ACK),	\
+		tb_cfg_type_name(TB_CFG_PKG_EVENT),		\
+		tb_cfg_type_name(TB_CFG_PKG_XDOMAIN_REQ),	\
+		tb_cfg_type_name(TB_CFG_PKG_XDOMAIN_RESP),	\
+		tb_cfg_type_name(TB_CFG_PKG_OVERRIDE),		\
+		tb_cfg_type_name(TB_CFG_PKG_RESET),		\
+		tb_cfg_type_name(TB_CFG_PKG_ICM_EVENT),		\
+		tb_cfg_type_name(TB_CFG_PKG_ICM_CMD),		\
+		tb_cfg_type_name(TB_CFG_PKG_ICM_RESP))
+
+#ifndef TB_TRACE_HELPERS
+#define TB_TRACE_HELPERS
+static inline const char *show_data_read_write(struct trace_seq *p,
+					       const u32 *data)
+{
+	const struct cfg_read_pkg *msg = (const struct cfg_read_pkg *)data;
+	const char *ret = trace_seq_buffer_ptr(p);
+
+	trace_seq_printf(p, "offset=%#x, len=%u, port=%d, config=%#x, seq=%d, ",
+			 msg->addr.offset, msg->addr.length, msg->addr.port,
+			 msg->addr.space, msg->addr.seq);
+
+	return ret;
+}
+
+static inline const char *show_data_error(struct trace_seq *p, const u32 *data)
+{
+	const struct cfg_error_pkg *msg = (const struct cfg_error_pkg *)data;
+	const char *ret = trace_seq_buffer_ptr(p);
+
+	trace_seq_printf(p, "error=%#x, port=%d, plug=%#x, ", msg->error,
+			 msg->port, msg->pg);
+
+	return ret;
+}
+
+static inline const char *show_data_event(struct trace_seq *p, const u32 *data)
+{
+	const struct cfg_event_pkg *msg = (const struct cfg_event_pkg *)data;
+	const char *ret = trace_seq_buffer_ptr(p);
+
+	trace_seq_printf(p, "port=%d, unplug=%#x, ", msg->port, msg->unplug);
+
+	return ret;
+}
+
+static inline const char *show_route(struct trace_seq *p, const u32 *data)
+{
+	const struct tb_cfg_header *header = (const struct tb_cfg_header *)data;
+	const char *ret = trace_seq_buffer_ptr(p);
+
+	trace_seq_printf(p, "route=%llx, ", tb_cfg_get_route(header));
+
+	return ret;
+}
+
+static inline const char *show_data(struct trace_seq *p, u8 type,
+				    const u32 *data, u32 length)
+{
+	const char *ret = trace_seq_buffer_ptr(p);
+	const char *prefix = "";
+	int i;
+
+	switch (type) {
+	case TB_CFG_PKG_READ:
+	case TB_CFG_PKG_WRITE:
+		show_route(p, data);
+		show_data_read_write(p, data);
+		break;
+
+	case TB_CFG_PKG_ERROR:
+		show_route(p, data);
+		show_data_error(p, data);
+		break;
+
+	case TB_CFG_PKG_EVENT:
+		show_route(p, data);
+		show_data_event(p, data);
+		break;
+
+	case TB_CFG_PKG_ICM_EVENT:
+	case TB_CFG_PKG_ICM_CMD:
+	case TB_CFG_PKG_ICM_RESP:
+		/* ICM messages always target the host router */
+		trace_seq_puts(p, "route=0, ");
+		break;
+
+	default:
+		show_route(p, data);
+		break;
+	}
+
+	trace_seq_printf(p, "data=[");
+	for (i = 0; i < length; i++) {
+		trace_seq_printf(p, "%s0x%08x", prefix, data[i]);
+		prefix = ", ";
+	}
+	trace_seq_printf(p, "]");
+	trace_seq_putc(p, 0);
+
+	return ret;
+}
+#endif
+
+DECLARE_EVENT_CLASS(tb_raw,
+	TP_PROTO(int index, u8 type, const void *data, size_t size),
+	TP_ARGS(index, type, data, size),
+	TP_STRUCT__entry(
+		__field(int, index)
+		__field(u8, type)
+		__field(size_t, size)
+		__dynamic_array(u32, data, size / 4)
+	),
+	TP_fast_assign(
+		__entry->index = index;
+		__entry->type = type;
+		__entry->size = size / 4;
+		memcpy(__get_dynamic_array(data), data, size);
+	),
+	TP_printk("type=%s, size=%zd, domain=%d, %s",
+		  show_type_name(__entry->type), __entry->size, __entry->index,
+		  show_data(p, __entry->type, __get_dynamic_array(data),
+			    __entry->size)
+	)
+);
+
+DEFINE_EVENT(tb_raw, tb_tx,
+	TP_PROTO(int index, u8 type, const void *data, size_t size),
+	TP_ARGS(index, type, data, size)
+);
+
+DEFINE_EVENT(tb_raw, tb_event,
+	TP_PROTO(int index, u8 type, const void *data, size_t size),
+	TP_ARGS(index, type, data, size)
+);
+
+TRACE_EVENT(tb_rx,
+	TP_PROTO(int index, u8 type, const void *data, size_t size, bool dropped),
+	TP_ARGS(index, type, data, size, dropped),
+	TP_STRUCT__entry(
+		__field(int, index)
+		__field(u8, type)
+		__field(size_t, size)
+		__dynamic_array(u32, data, size / 4)
+		__field(bool, dropped)
+	),
+	TP_fast_assign(
+		__entry->index = index;
+		__entry->type = type;
+		__entry->size = size / 4;
+		memcpy(__get_dynamic_array(data), data, size);
+		__entry->dropped = dropped;
+	),
+	TP_printk("type=%s, dropped=%u, size=%zd, domain=%d, %s",
+		  show_type_name(__entry->type), __entry->dropped,
+		  __entry->size, __entry->index,
+		  show_data(p, __entry->type, __get_dynamic_array(data),
+			    __entry->size)
+	)
+);
+
+#endif /* TB_TRACE_H_ */
+
+#undef TRACE_INCLUDE_PATH
+#define TRACE_INCLUDE_PATH .
+
+#undef TRACE_INCLUDE_FILE
+#define TRACE_INCLUDE_FILE trace
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>