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, 185 insertions, 0 deletions

diff --git a/lib/base64.c b/lib/base64.c
new file mode 100644
index 000000000..325c7332b
--- /dev/null
+++ b/lib/base64.c
@@ -0,0 +1,185 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * base64.c - Base64 with support for multiple variants
+ *
+ * Copyright (c) 2020 Hannes Reinecke, SUSE
+ *
+ * Based on the base64url routines from fs/crypto/fname.c
+ * (which are using the URL-safe Base64 encoding),
+ * modified to support multiple Base64 variants.
+ */
+
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/export.h>
+#include <linux/string.h>
+#include <linux/base64.h>
+
+static const char base64_tables[][65] = {
+	[BASE64_STD] = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/",
+	[BASE64_URLSAFE] = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_",
+	[BASE64_IMAP] = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+,",
+};
+
+/*
+ * Initialize the base64 reverse mapping for a single character
+ * This macro maps a character to its corresponding base64 value,
+ * returning -1 if the character is invalid.
+ * char 'A'-'Z' maps to 0-25, 'a'-'z' maps to 26-51, '0'-'9' maps to 52-61,
+ * ch_62 maps to 62, ch_63 maps to 63, and other characters return -1
+ */
+#define INIT_1(v, ch_62, ch_63) \
+	[v] = (v) >= 'A' && (v) <= 'Z' ? (v) - 'A' \
+		: (v) >= 'a' && (v) <= 'z' ? (v) - 'a' + 26 \
+		: (v) >= '0' && (v) <= '9' ? (v) - '0' + 52 \
+		: (v) == (ch_62) ? 62 : (v) == (ch_63) ? 63 : -1
+
+/*
+ * Recursive macros to generate multiple Base64 reverse mapping table entries.
+ * Each macro generates a sequence of entries in the lookup table:
+ * INIT_2 generates 2 entries, INIT_4 generates 4, INIT_8 generates 8, and so on up to INIT_32.
+ */
+#define INIT_2(v, ...) INIT_1(v, __VA_ARGS__), INIT_1((v) + 1, __VA_ARGS__)
+#define INIT_4(v, ...) INIT_2(v, __VA_ARGS__), INIT_2((v) + 2, __VA_ARGS__)
+#define INIT_8(v, ...) INIT_4(v, __VA_ARGS__), INIT_4((v) + 4, __VA_ARGS__)
+#define INIT_16(v, ...) INIT_8(v, __VA_ARGS__), INIT_8((v) + 8, __VA_ARGS__)
+#define INIT_32(v, ...) INIT_16(v, __VA_ARGS__), INIT_16((v) + 16, __VA_ARGS__)
+
+#define BASE64_REV_INIT(ch_62, ch_63) { \
+	[0 ... 0x1f] = -1, \
+	INIT_32(0x20, ch_62, ch_63), \
+	INIT_32(0x40, ch_62, ch_63), \
+	INIT_32(0x60, ch_62, ch_63), \
+	[0x80 ... 0xff] = -1 }
+
+static const s8 base64_rev_maps[][256] = {
+	[BASE64_STD] = BASE64_REV_INIT('+', '/'),
+	[BASE64_URLSAFE] = BASE64_REV_INIT('-', '_'),
+	[BASE64_IMAP] = BASE64_REV_INIT('+', ',')
+};
+
+#undef BASE64_REV_INIT
+#undef INIT_32
+#undef INIT_16
+#undef INIT_8
+#undef INIT_4
+#undef INIT_2
+#undef INIT_1
+/**
+ * base64_encode() - Base64-encode some binary data
+ * @src: the binary data to encode
+ * @srclen: the length of @src in bytes
+ * @dst: (output) the Base64-encoded string.  Not NUL-terminated.
+ * @padding: whether to append '=' padding characters
+ * @variant: which base64 variant to use
+ *
+ * Encodes data using the selected Base64 variant.
+ *
+ * Return: the length of the resulting Base64-encoded string in bytes.
+ */
+int base64_encode(const u8 *src, int srclen, char *dst, bool padding, enum base64_variant variant)
+{
+	u32 ac = 0;
+	char *cp = dst;
+	const char *base64_table = base64_tables[variant];
+
+	while (srclen >= 3) {
+		ac = src[0] << 16 | src[1] << 8 | src[2];
+		*cp++ = base64_table[ac >> 18];
+		*cp++ = base64_table[(ac >> 12) & 0x3f];
+		*cp++ = base64_table[(ac >> 6) & 0x3f];
+		*cp++ = base64_table[ac & 0x3f];
+
+		src += 3;
+		srclen -= 3;
+	}
+
+	switch (srclen) {
+	case 2:
+		ac = src[0] << 16 | src[1] << 8;
+		*cp++ = base64_table[ac >> 18];
+		*cp++ = base64_table[(ac >> 12) & 0x3f];
+		*cp++ = base64_table[(ac >> 6) & 0x3f];
+		if (padding)
+			*cp++ = '=';
+		break;
+	case 1:
+		ac = src[0] << 16;
+		*cp++ = base64_table[ac >> 18];
+		*cp++ = base64_table[(ac >> 12) & 0x3f];
+		if (padding) {
+			*cp++ = '=';
+			*cp++ = '=';
+		}
+		break;
+	}
+	return cp - dst;
+}
+EXPORT_SYMBOL_GPL(base64_encode);
+
+/**
+ * base64_decode() - Base64-decode a string
+ * @src: the string to decode.  Doesn't need to be NUL-terminated.
+ * @srclen: the length of @src in bytes
+ * @dst: (output) the decoded binary data
+ * @padding: whether the input is expected to include '=' padding characters
+ * @variant: which base64 variant to use
+ *
+ * Decodes a string using the selected Base64 variant.
+ *
+ * Return: the length of the resulting decoded binary data in bytes,
+ *	   or -1 if the string isn't a valid Base64 string.
+ */
+int base64_decode(const char *src, int srclen, u8 *dst, bool padding, enum base64_variant variant)
+{
+	u8 *bp = dst;
+	s8 input[4];
+	s32 val;
+	const u8 *s = (const u8 *)src;
+	const s8 *base64_rev_tables = base64_rev_maps[variant];
+
+	while (srclen >= 4) {
+		input[0] = base64_rev_tables[s[0]];
+		input[1] = base64_rev_tables[s[1]];
+		input[2] = base64_rev_tables[s[2]];
+		input[3] = base64_rev_tables[s[3]];
+
+		val = input[0] << 18 | input[1] << 12 | input[2] << 6 | input[3];
+
+		if (unlikely(val < 0)) {
+			if (!padding || srclen != 4 || s[3] != '=')
+				return -1;
+			padding = 0;
+			srclen = s[2] == '=' ? 2 : 3;
+			break;
+		}
+
+		*bp++ = val >> 16;
+		*bp++ = val >> 8;
+		*bp++ = val;
+
+		s += 4;
+		srclen -= 4;
+	}
+
+	if (likely(!srclen))
+		return bp - dst;
+	if (padding || srclen == 1)
+		return -1;
+
+	val = (base64_rev_tables[s[0]] << 12) | (base64_rev_tables[s[1]] << 6);
+
+	if (srclen == 2) {
+		if (val & 0x800003ff)
+			return -1;
+		*bp++ = val >> 10;
+	} else {
+		val |= base64_rev_tables[s[2]];
+		if (val & 0x80000003)
+			return -1;
+		*bp++ = val >> 10;
+		*bp++ = val >> 2;
+	}
+	return bp - dst;
+}
+EXPORT_SYMBOL_GPL(base64_decode);