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, 179 insertions, 0 deletions

diff --git a/include/uapi/cxl/features.h b/include/uapi/cxl/features.h
new file mode 100644
index 000000000..490606d76
--- /dev/null
+++ b/include/uapi/cxl/features.h
@@ -0,0 +1,179 @@
+/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
+/*
+ * Copyright (c) 2024,2025, Intel Corporation
+ *
+ * These are definitions for the mailbox command interface of CXL subsystem.
+ */
+#ifndef _UAPI_CXL_FEATURES_H_
+#define _UAPI_CXL_FEATURES_H_
+
+#include <linux/types.h>
+
+typedef unsigned char __uapi_uuid_t[16];
+
+#ifdef __KERNEL__
+#include <linux/uuid.h>
+/*
+ * Note, __uapi_uuid_t is 1-byte aligned on modern compilers and 4-byte
+ * aligned on others. Ensure that __uapi_uuid_t in a struct is placed at
+ * a 4-byte aligned offset, or the structure is packed, to ensure
+ * consistent padding.
+ */
+static_assert(sizeof(__uapi_uuid_t) == sizeof(uuid_t));
+#define __uapi_uuid_t uuid_t
+#endif
+
+/*
+ * struct cxl_mbox_get_sup_feats_in - Get Supported Features input
+ *
+ * @count: bytes of Feature data to return in output
+ * @start_idx: index of first requested Supported Feature Entry, 0 based.
+ * @reserved: reserved field, must be 0s.
+ *
+ * Get Supported Features (0x500h) CXL r3.2 8.2.9.6.1 command.
+ * Input block for Get support Feature
+ */
+struct cxl_mbox_get_sup_feats_in {
+	__le32 count;
+	__le16 start_idx;
+	__u8 reserved[2];
+} __attribute__ ((__packed__));
+
+/* CXL spec r3.2 Table 8-87 command effects */
+#define CXL_CMD_CONFIG_CHANGE_COLD_RESET	BIT(0)
+#define CXL_CMD_CONFIG_CHANGE_IMMEDIATE		BIT(1)
+#define CXL_CMD_DATA_CHANGE_IMMEDIATE		BIT(2)
+#define CXL_CMD_POLICY_CHANGE_IMMEDIATE		BIT(3)
+#define CXL_CMD_LOG_CHANGE_IMMEDIATE		BIT(4)
+#define CXL_CMD_SECURITY_STATE_CHANGE		BIT(5)
+#define CXL_CMD_BACKGROUND			BIT(6)
+#define CXL_CMD_BGCMD_ABORT_SUPPORTED		BIT(7)
+#define CXL_CMD_EFFECTS_VALID			BIT(9)
+#define CXL_CMD_CONFIG_CHANGE_CONV_RESET	BIT(10)
+#define CXL_CMD_CONFIG_CHANGE_CXL_RESET		BIT(11)
+#define CXL_CMD_EFFECTS_RESERVED		GENMASK(15, 12)
+
+/*
+ * struct cxl_feat_entry - Supported Feature Entry
+ * @uuid: UUID of the Feature
+ * @id: id to identify the feature. 0 based
+ * @get_feat_size: max bytes required for Get Feature command for this Feature
+ * @set_feat_size: max bytes required for Set Feature command for this Feature
+ * @flags: attribute flags
+ * @get_feat_ver: Get Feature version
+ * @set_feat_ver: Set Feature version
+ * @effects: Set Feature command effects
+ * @reserved: reserved, must be 0
+ *
+ * CXL spec r3.2 Table 8-109
+ * Get Supported Features Supported Feature Entry
+ */
+struct cxl_feat_entry {
+	__uapi_uuid_t uuid;
+	__le16 id;
+	__le16 get_feat_size;
+	__le16 set_feat_size;
+	__le32 flags;
+	__u8 get_feat_ver;
+	__u8 set_feat_ver;
+	__le16 effects;
+	__u8 reserved[18];
+} __attribute__ ((__packed__));
+
+/* @flags field for 'struct cxl_feat_entry' */
+#define CXL_FEATURE_F_CHANGEABLE		BIT(0)
+#define CXL_FEATURE_F_PERSIST_FW_UPDATE		BIT(4)
+#define CXL_FEATURE_F_DEFAULT_SEL		BIT(5)
+#define CXL_FEATURE_F_SAVED_SEL			BIT(6)
+
+/*
+ * struct cxl_mbox_get_sup_feats_out - Get Supported Features output
+ * @num_entries: number of Supported Feature Entries returned
+ * @supported_feats: number of supported Features
+ * @reserved: reserved, must be 0s.
+ * @ents: Supported Feature Entries array
+ *
+ * CXL spec r3.2 Table 8-108
+ * Get supported Features Output Payload
+ */
+struct cxl_mbox_get_sup_feats_out {
+	__struct_group(cxl_mbox_get_sup_feats_out_hdr, hdr, /* no attrs */,
+		__le16 num_entries;
+		__le16 supported_feats;
+		__u8 reserved[4];
+	);
+	struct cxl_feat_entry ents[] __counted_by_le(num_entries);
+} __attribute__ ((__packed__));
+
+/*
+ * Get Feature CXL spec r3.2 Spec 8.2.9.6.2
+ */
+
+/*
+ * struct cxl_mbox_get_feat_in - Get Feature input
+ * @uuid: UUID for Feature
+ * @offset: offset of the first byte in Feature data for output payload
+ * @count: count in bytes of Feature data returned
+ * @selection: 0 current value, 1 default value, 2 saved value
+ *
+ * CXL spec r3.2 section 8.2.9.6.2 Table 8-99
+ */
+struct cxl_mbox_get_feat_in {
+	__uapi_uuid_t uuid;
+	__le16 offset;
+	__le16 count;
+	__u8 selection;
+} __attribute__ ((__packed__));
+
+/*
+ * enum cxl_get_feat_selection - selection field of Get Feature input
+ */
+enum cxl_get_feat_selection {
+	CXL_GET_FEAT_SEL_CURRENT_VALUE,
+	CXL_GET_FEAT_SEL_DEFAULT_VALUE,
+	CXL_GET_FEAT_SEL_SAVED_VALUE,
+	CXL_GET_FEAT_SEL_MAX
+};
+
+/*
+ * Set Feature CXL spec r3.2  8.2.9.6.3
+ */
+
+/*
+ * struct cxl_mbox_set_feat_in - Set Features input
+ * @uuid: UUID for Feature
+ * @flags: set feature flags
+ * @offset: byte offset of Feature data to update
+ * @version: Feature version of the data in Feature Data
+ * @rsvd: reserved, must be 0s.
+ * @feat_data: raw byte stream of Features data to update
+ *
+ * CXL spec r3.2 section 8.2.9.6.3 Table 8-101
+ */
+struct cxl_mbox_set_feat_in {
+	__struct_group(cxl_mbox_set_feat_hdr, hdr, /* no attrs */,
+		__uapi_uuid_t uuid;
+		__le32 flags;
+		__le16 offset;
+		__u8 version;
+		__u8 rsvd[9];
+	);
+	__u8 feat_data[];
+}  __packed;
+
+/*
+ * enum cxl_set_feat_flag_data_transfer - Set Feature flags field
+ */
+enum cxl_set_feat_flag_data_transfer {
+	CXL_SET_FEAT_FLAG_FULL_DATA_TRANSFER = 0,
+	CXL_SET_FEAT_FLAG_INITIATE_DATA_TRANSFER,
+	CXL_SET_FEAT_FLAG_CONTINUE_DATA_TRANSFER,
+	CXL_SET_FEAT_FLAG_FINISH_DATA_TRANSFER,
+	CXL_SET_FEAT_FLAG_ABORT_DATA_TRANSFER,
+	CXL_SET_FEAT_FLAG_DATA_TRANSFER_MAX
+};
+
+#define CXL_SET_FEAT_FLAG_DATA_TRANSFER_MASK		GENMASK(2, 0)
+#define CXL_SET_FEAT_FLAG_DATA_SAVED_ACROSS_RESET	BIT(3)
+
+#endif