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, 387 insertions, 0 deletions

diff --git a/drivers/mtd/ubi/debug.h b/drivers/mtd/ubi/debug.h
new file mode 100644
index 000000000..b2fd97548
--- /dev/null
+++ b/drivers/mtd/ubi/debug.h
@@ -0,0 +1,387 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (c) International Business Machines Corp., 2006
+ *
+ * Author: Artem Bityutskiy (Битюцкий Артём)
+ */
+
+#ifndef __UBI_DEBUG_H__
+#define __UBI_DEBUG_H__
+
+void ubi_dump_flash(struct ubi_device *ubi, int pnum, int offset, int len);
+void ubi_dump_ec_hdr(const struct ubi_ec_hdr *ec_hdr);
+void ubi_dump_vid_hdr(const struct ubi_vid_hdr *vid_hdr);
+
+#include <linux/random.h>
+
+#define ubi_assert(expr)  do {                                               \
+	if (unlikely(!(expr))) {                                             \
+		pr_crit("UBI assert failed in %s at %u (pid %d)\n",          \
+		       __func__, __LINE__, current->pid);                    \
+		dump_stack();                                                \
+	}                                                                    \
+} while (0)
+
+#define ubi_dbg_print_hex_dump(l, ps, pt, r, g, b, len, a)                   \
+		print_hex_dump(l, ps, pt, r, g, b, len, a)
+
+#define ubi_dbg_msg(type, fmt, ...) \
+	pr_debug("UBI DBG " type " (pid %d): " fmt "\n", current->pid,       \
+		 ##__VA_ARGS__)
+
+/* General debugging messages */
+#define dbg_gen(fmt, ...) ubi_dbg_msg("gen", fmt, ##__VA_ARGS__)
+/* Messages from the eraseblock association sub-system */
+#define dbg_eba(fmt, ...) ubi_dbg_msg("eba", fmt, ##__VA_ARGS__)
+/* Messages from the wear-leveling sub-system */
+#define dbg_wl(fmt, ...)  ubi_dbg_msg("wl", fmt, ##__VA_ARGS__)
+/* Messages from the input/output sub-system */
+#define dbg_io(fmt, ...)  ubi_dbg_msg("io", fmt, ##__VA_ARGS__)
+/* Initialization and build messages */
+#define dbg_bld(fmt, ...) ubi_dbg_msg("bld", fmt, ##__VA_ARGS__)
+
+void ubi_dump_vol_info(const struct ubi_volume *vol);
+void ubi_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx);
+void ubi_dump_av(const struct ubi_ainf_volume *av);
+void ubi_dump_aeb(const struct ubi_ainf_peb *aeb, int type);
+void ubi_dump_mkvol_req(const struct ubi_mkvol_req *req);
+int ubi_self_check_all_ff(struct ubi_device *ubi, int pnum, int offset,
+			  int len);
+int ubi_debugfs_init(void);
+void ubi_debugfs_exit(void);
+int ubi_debugfs_init_dev(struct ubi_device *ubi);
+void ubi_debugfs_exit_dev(struct ubi_device *ubi);
+
+/**
+ * The following function is a legacy implementation of UBI fault-injection
+ * hook. When using more powerful fault injection capabilities, the legacy
+ * fault injection interface should be retained.
+ */
+int ubi_dbg_power_cut(struct ubi_device *ubi, int caller);
+
+static inline int ubi_dbg_bitflip(const struct ubi_device *ubi)
+{
+	if (ubi->dbg.emulate_bitflips)
+		return !get_random_u32_below(200);
+	return 0;
+}
+
+static inline int ubi_dbg_write_failure(const struct ubi_device *ubi)
+{
+	if (ubi->dbg.emulate_io_failures)
+		return !get_random_u32_below(500);
+	return 0;
+}
+
+static inline int ubi_dbg_erase_failure(const struct ubi_device *ubi)
+{
+	if (ubi->dbg.emulate_io_failures)
+		return !get_random_u32_below(400);
+	return 0;
+}
+
+/**
+ * MASK_XXX: Mask for emulate_failures in ubi_debug_info.The mask is used to
+ * precisely control the type and process of fault injection.
+ */
+/* Emulate a power cut when writing EC/VID header */
+#define MASK_POWER_CUT_EC			(1 << 0)
+#define MASK_POWER_CUT_VID			(1 << 1)
+/* Emulate a power cut when writing data*/
+#define MASK_POWER_CUT_DATA			(1 << 2)
+/* Emulate bit-flips */
+#define MASK_BITFLIPS				(1 << 3)
+/* Emulate ecc error */
+#define MASK_ECCERR				(1 << 4)
+/* Emulates -EIO during data read */
+#define MASK_READ_FAILURE			(1 << 5)
+#define MASK_READ_FAILURE_EC			(1 << 6)
+#define MASK_READ_FAILURE_VID			(1 << 7)
+/* Emulates -EIO during data write */
+#define MASK_WRITE_FAILURE			(1 << 8)
+/* Emulates -EIO during erase a PEB*/
+#define MASK_ERASE_FAILURE			(1 << 9)
+/* Return UBI_IO_FF when reading EC/VID header */
+#define MASK_IO_FF_EC				(1 << 10)
+#define MASK_IO_FF_VID				(1 << 11)
+/* Return UBI_IO_FF_BITFLIPS when reading EC/VID header */
+#define MASK_IO_FF_BITFLIPS_EC			(1 << 12)
+#define MASK_IO_FF_BITFLIPS_VID			(1 << 13)
+/* Return UBI_IO_BAD_HDR when reading EC/VID header */
+#define MASK_BAD_HDR_EC				(1 << 14)
+#define MASK_BAD_HDR_VID			(1 << 15)
+/* Return UBI_IO_BAD_HDR_EBADMSG when reading EC/VID header */
+#define MASK_BAD_HDR_EBADMSG_EC			(1 << 16)
+#define MASK_BAD_HDR_EBADMSG_VID		(1 << 17)
+
+#ifdef CONFIG_MTD_UBI_FAULT_INJECTION
+
+extern bool should_fail_eccerr(void);
+extern bool should_fail_bitflips(void);
+extern bool should_fail_read_failure(void);
+extern bool should_fail_write_failure(void);
+extern bool should_fail_erase_failure(void);
+extern bool should_fail_power_cut(void);
+extern bool should_fail_io_ff(void);
+extern bool should_fail_io_ff_bitflips(void);
+extern bool should_fail_bad_hdr(void);
+extern bool should_fail_bad_hdr_ebadmsg(void);
+
+static inline bool ubi_dbg_fail_bitflip(const struct ubi_device *ubi)
+{
+	if (ubi->dbg.emulate_failures & MASK_BITFLIPS)
+		return should_fail_bitflips();
+	return false;
+}
+
+static inline bool ubi_dbg_fail_write(const struct ubi_device *ubi)
+{
+	if (ubi->dbg.emulate_failures & MASK_WRITE_FAILURE)
+		return should_fail_write_failure();
+	return false;
+}
+
+static inline bool ubi_dbg_fail_erase(const struct ubi_device *ubi)
+{
+	if (ubi->dbg.emulate_failures & MASK_ERASE_FAILURE)
+		return should_fail_erase_failure();
+	return false;
+}
+
+static inline bool ubi_dbg_fail_power_cut(const struct ubi_device *ubi,
+					  unsigned int caller)
+{
+	if (ubi->dbg.emulate_failures & caller)
+		return should_fail_power_cut();
+	return false;
+}
+
+static inline bool ubi_dbg_fail_read(const struct ubi_device *ubi,
+				     unsigned int caller)
+{
+	if (ubi->dbg.emulate_failures & caller)
+		return should_fail_read_failure();
+	return false;
+}
+
+static inline bool ubi_dbg_fail_eccerr(const struct ubi_device *ubi)
+{
+	if (ubi->dbg.emulate_failures & MASK_ECCERR)
+		return should_fail_eccerr();
+	return false;
+}
+
+static inline bool ubi_dbg_fail_ff(const struct ubi_device *ubi,
+				   unsigned int caller)
+{
+	if (ubi->dbg.emulate_failures & caller)
+		return should_fail_io_ff();
+	return false;
+}
+
+static inline bool ubi_dbg_fail_ff_bitflips(const struct ubi_device *ubi,
+					    unsigned int caller)
+{
+	if (ubi->dbg.emulate_failures & caller)
+		return should_fail_io_ff_bitflips();
+	return false;
+}
+
+static inline bool ubi_dbg_fail_bad_hdr(const struct ubi_device *ubi,
+					 unsigned int caller)
+{
+	if (ubi->dbg.emulate_failures & caller)
+		return should_fail_bad_hdr();
+	return false;
+}
+
+static inline bool ubi_dbg_fail_bad_hdr_ebadmsg(const struct ubi_device *ubi,
+						 unsigned int caller)
+{
+	if (ubi->dbg.emulate_failures & caller)
+		return should_fail_bad_hdr_ebadmsg();
+	return false;
+}
+#else /* CONFIG_MTD_UBI_FAULT_INJECTION */
+
+#define ubi_dbg_fail_bitflip(u)             false
+#define ubi_dbg_fail_write(u)               false
+#define ubi_dbg_fail_erase(u)               false
+#define ubi_dbg_fail_power_cut(u, c)        false
+#define ubi_dbg_fail_read(u, c)             false
+#define ubi_dbg_fail_eccerr(u)              false
+#define ubi_dbg_fail_ff(u, c)               false
+#define ubi_dbg_fail_ff_bitflips(u, v)      false
+#define ubi_dbg_fail_bad_hdr(u, c)          false
+#define ubi_dbg_fail_bad_hdr_ebadmsg(u, c)  false
+
+#endif
+
+/**
+ * ubi_dbg_is_power_cut - if it is time to emulate power cut.
+ * @ubi: UBI device description object
+ *
+ * Returns true if power cut should be emulated, otherwise returns false.
+ */
+static inline bool ubi_dbg_is_power_cut(struct ubi_device *ubi,
+					unsigned int caller)
+{
+	if (ubi_dbg_power_cut(ubi, caller))
+		return true;
+	return ubi_dbg_fail_power_cut(ubi, caller);
+}
+
+/**
+ * ubi_dbg_is_bitflip - if it is time to emulate a bit-flip.
+ * @ubi: UBI device description object
+ *
+ * Returns true if a bit-flip should be emulated, otherwise returns false.
+ */
+static inline bool ubi_dbg_is_bitflip(const struct ubi_device *ubi)
+{
+	if (ubi_dbg_bitflip(ubi))
+		return true;
+	return ubi_dbg_fail_bitflip(ubi);
+}
+
+/**
+ * ubi_dbg_is_write_failure - if it is time to emulate a write failure.
+ * @ubi: UBI device description object
+ *
+ * Returns true if a write failure should be emulated, otherwise returns
+ * false.
+ */
+static inline bool ubi_dbg_is_write_failure(const struct ubi_device *ubi)
+{
+	if (ubi_dbg_write_failure(ubi))
+		return true;
+	return ubi_dbg_fail_write(ubi);
+}
+
+/**
+ * ubi_dbg_is_erase_failure - if its time to emulate an erase failure.
+ * @ubi: UBI device description object
+ *
+ * Returns true if an erase failure should be emulated, otherwise returns
+ * false.
+ */
+static inline bool ubi_dbg_is_erase_failure(const struct ubi_device *ubi)
+{
+	if (ubi_dbg_erase_failure(ubi))
+		return true;
+	return ubi_dbg_fail_erase(ubi);
+}
+
+/**
+ * ubi_dbg_is_eccerr - if it is time to emulate ECC error.
+ * @ubi: UBI device description object
+ *
+ * Returns true if a ECC error should be emulated, otherwise returns false.
+ */
+static inline bool ubi_dbg_is_eccerr(const struct ubi_device *ubi)
+{
+	return ubi_dbg_fail_eccerr(ubi);
+}
+
+/**
+ * ubi_dbg_is_read_failure - if it is time to emulate a read failure.
+ * @ubi: UBI device description object
+ *
+ * Returns true if a read failure should be emulated, otherwise returns
+ * false.
+ */
+static inline bool ubi_dbg_is_read_failure(const struct ubi_device *ubi,
+					   unsigned int caller)
+{
+	return ubi_dbg_fail_read(ubi, caller);
+}
+
+/**
+ * ubi_dbg_is_ff - if it is time to emulate that read region is only 0xFF.
+ * @ubi: UBI device description object
+ *
+ * Returns true if read region should be emulated 0xFF, otherwise
+ * returns false.
+ */
+static inline bool ubi_dbg_is_ff(const struct ubi_device *ubi,
+				 unsigned int caller)
+{
+	return ubi_dbg_fail_ff(ubi, caller);
+}
+
+/**
+ * ubi_dbg_is_ff_bitflips - if it is time to emulate that read region is only 0xFF
+ * with error reported by the MTD driver
+ *
+ * @ubi: UBI device description object
+ *
+ * Returns true if read region should be emulated 0xFF and error
+ * reported by the MTD driver, otherwise returns false.
+ */
+static inline bool ubi_dbg_is_ff_bitflips(const struct ubi_device *ubi,
+					  unsigned int caller)
+{
+	return ubi_dbg_fail_ff_bitflips(ubi, caller);
+}
+
+/**
+ * ubi_dbg_is_bad_hdr - if it is time to emulate a bad header
+ * @ubi: UBI device description object
+ *
+ * Returns true if a bad header error should be emulated, otherwise
+ * returns false.
+ */
+static inline bool ubi_dbg_is_bad_hdr(const struct ubi_device *ubi,
+				      unsigned int caller)
+{
+	return ubi_dbg_fail_bad_hdr(ubi, caller);
+}
+
+/**
+ * ubi_dbg_is_bad_hdr_ebadmsg - if it is time to emulate a bad header with
+ * ECC error.
+ *
+ * @ubi: UBI device description object
+ *
+ * Returns true if a bad header with ECC error should be emulated, otherwise
+ * returns false.
+ */
+static inline bool ubi_dbg_is_bad_hdr_ebadmsg(const struct ubi_device *ubi,
+					      unsigned int caller)
+{
+	return ubi_dbg_fail_bad_hdr_ebadmsg(ubi, caller);
+}
+
+/**
+ * ubi_dbg_is_bgt_disabled - if the background thread is disabled.
+ * @ubi: UBI device description object
+ *
+ * Returns non-zero if the UBI background thread is disabled for testing
+ * purposes.
+ */
+static inline int ubi_dbg_is_bgt_disabled(const struct ubi_device *ubi)
+{
+	return ubi->dbg.disable_bgt;
+}
+
+static inline int ubi_dbg_chk_io(const struct ubi_device *ubi)
+{
+	return ubi->dbg.chk_io;
+}
+
+static inline int ubi_dbg_chk_gen(const struct ubi_device *ubi)
+{
+	return ubi->dbg.chk_gen;
+}
+
+static inline int ubi_dbg_chk_fastmap(const struct ubi_device *ubi)
+{
+	return ubi->dbg.chk_fastmap;
+}
+
+static inline void ubi_enable_dbg_chk_fastmap(struct ubi_device *ubi)
+{
+	ubi->dbg.chk_fastmap = 1;
+}
+
+#endif /* !__UBI_DEBUG_H__ */