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, 399 insertions, 0 deletions

diff --git a/fs/udf/unicode.c b/fs/udf/unicode.c
new file mode 100644
index 000000000..87580ff82
--- /dev/null
+++ b/fs/udf/unicode.c
@@ -0,0 +1,399 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * unicode.c
+ *
+ * PURPOSE
+ *	Routines for converting between UTF-8 and OSTA Compressed Unicode.
+ *      Also handles filename mangling
+ *
+ * DESCRIPTION
+ *	OSTA Compressed Unicode is explained in the OSTA UDF specification.
+ *		http://www.osta.org/
+ *	UTF-8 is explained in the IETF RFC XXXX.
+ *		ftp://ftp.internic.net/rfc/rfcxxxx.txt
+ *
+ */
+
+#include "udfdecl.h"
+
+#include <linux/hex.h>
+#include <linux/kernel.h>
+#include <linux/string.h>	/* for memset */
+#include <linux/nls.h>
+#include <linux/crc-itu-t.h>
+#include <linux/slab.h>
+
+#include "udf_sb.h"
+
+#define PLANE_SIZE 0x10000
+#define UNICODE_MAX 0x10ffff
+#define SURROGATE_MASK 0xfffff800
+#define SURROGATE_PAIR 0x0000d800
+#define SURROGATE_LOW  0x00000400
+#define SURROGATE_CHAR_BITS 10
+#define SURROGATE_CHAR_MASK ((1 << SURROGATE_CHAR_BITS) - 1)
+
+#define ILLEGAL_CHAR_MARK	'_'
+#define EXT_MARK		'.'
+#define CRC_MARK		'#'
+#define EXT_SIZE		5
+/* Number of chars we need to store generated CRC to make filename unique */
+#define CRC_LEN			5
+
+static unicode_t get_utf16_char(const uint8_t *str_i, int str_i_max_len,
+				int str_i_idx, int u_ch, unicode_t *ret)
+{
+	unicode_t c;
+	int start_idx = str_i_idx;
+
+	/* Expand OSTA compressed Unicode to Unicode */
+	c = str_i[str_i_idx++];
+	if (u_ch > 1)
+		c = (c << 8) | str_i[str_i_idx++];
+	if ((c & SURROGATE_MASK) == SURROGATE_PAIR) {
+		unicode_t next;
+
+		/* Trailing surrogate char */
+		if (str_i_idx >= str_i_max_len) {
+			c = UNICODE_MAX + 1;
+			goto out;
+		}
+
+		/* Low surrogate must follow the high one... */
+		if (c & SURROGATE_LOW) {
+			c = UNICODE_MAX + 1;
+			goto out;
+		}
+
+		WARN_ON_ONCE(u_ch != 2);
+		next = str_i[str_i_idx++] << 8;
+		next |= str_i[str_i_idx++];
+		if ((next & SURROGATE_MASK) != SURROGATE_PAIR ||
+		    !(next & SURROGATE_LOW)) {
+			c = UNICODE_MAX + 1;
+			goto out;
+		}
+
+		c = PLANE_SIZE +
+		    ((c & SURROGATE_CHAR_MASK) << SURROGATE_CHAR_BITS) +
+		    (next & SURROGATE_CHAR_MASK);
+	}
+out:
+	*ret = c;
+	return str_i_idx - start_idx;
+}
+
+
+static int udf_name_conv_char(uint8_t *str_o, int str_o_max_len,
+			      int *str_o_idx,
+			      const uint8_t *str_i, int str_i_max_len,
+			      int *str_i_idx,
+			      int u_ch, int *needsCRC,
+			      int (*conv_f)(wchar_t, unsigned char *, int),
+			      int translate)
+{
+	unicode_t c;
+	int illChar = 0;
+	int len, gotch = 0;
+
+	while (!gotch && *str_i_idx < str_i_max_len) {
+		if (*str_o_idx >= str_o_max_len) {
+			*needsCRC = 1;
+			return gotch;
+		}
+
+		len = get_utf16_char(str_i, str_i_max_len, *str_i_idx, u_ch,
+				     &c);
+		/* These chars cannot be converted. Replace them. */
+		if (c == 0 || c > UNICODE_MAX || (conv_f && c > MAX_WCHAR_T) ||
+		    (translate && c == '/')) {
+			illChar = 1;
+			if (!translate)
+				gotch = 1;
+		} else if (illChar)
+			break;
+		else
+			gotch = 1;
+		*str_i_idx += len;
+	}
+	if (illChar) {
+		*needsCRC = 1;
+		c = ILLEGAL_CHAR_MARK;
+		gotch = 1;
+	}
+	if (gotch) {
+		if (conv_f) {
+			len = conv_f(c, &str_o[*str_o_idx],
+				     str_o_max_len - *str_o_idx);
+		} else {
+			len = utf32_to_utf8(c, &str_o[*str_o_idx],
+					    str_o_max_len - *str_o_idx);
+			if (len < 0)
+				len = -ENAMETOOLONG;
+		}
+		/* Valid character? */
+		if (len >= 0)
+			*str_o_idx += len;
+		else if (len == -ENAMETOOLONG) {
+			*needsCRC = 1;
+			gotch = 0;
+		} else {
+			str_o[(*str_o_idx)++] = ILLEGAL_CHAR_MARK;
+			*needsCRC = 1;
+		}
+	}
+	return gotch;
+}
+
+static int udf_name_from_CS0(struct super_block *sb,
+			     uint8_t *str_o, int str_max_len,
+			     const uint8_t *ocu, int ocu_len,
+			     int translate)
+{
+	uint32_t c;
+	uint8_t cmp_id;
+	int idx, len;
+	int u_ch;
+	int needsCRC = 0;
+	int ext_i_len, ext_max_len;
+	int str_o_len = 0;	/* Length of resulting output */
+	int ext_o_len = 0;	/* Extension output length */
+	int ext_crc_len = 0;	/* Extension output length if used with CRC */
+	int i_ext = -1;		/* Extension position in input buffer */
+	int o_crc = 0;		/* Rightmost possible output pos for CRC+ext */
+	unsigned short valueCRC;
+	uint8_t ext[EXT_SIZE * NLS_MAX_CHARSET_SIZE + 1];
+	uint8_t crc[CRC_LEN];
+	int (*conv_f)(wchar_t, unsigned char *, int);
+
+	if (str_max_len <= 0)
+		return 0;
+
+	if (ocu_len == 0) {
+		memset(str_o, 0, str_max_len);
+		return 0;
+	}
+
+	if (UDF_SB(sb)->s_nls_map)
+		conv_f = UDF_SB(sb)->s_nls_map->uni2char;
+	else
+		conv_f = NULL;
+
+	cmp_id = ocu[0];
+	if (cmp_id != 8 && cmp_id != 16) {
+		memset(str_o, 0, str_max_len);
+		pr_err("unknown compression code (%u)\n", cmp_id);
+		return -EINVAL;
+	}
+	u_ch = cmp_id >> 3;
+
+	ocu++;
+	ocu_len--;
+
+	if (ocu_len % u_ch) {
+		pr_err("incorrect filename length (%d)\n", ocu_len + 1);
+		return -EINVAL;
+	}
+
+	if (translate) {
+		/* Look for extension */
+		for (idx = ocu_len - u_ch, ext_i_len = 0;
+		     (idx >= 0) && (ext_i_len < EXT_SIZE);
+		     idx -= u_ch, ext_i_len++) {
+			c = ocu[idx];
+			if (u_ch > 1)
+				c = (c << 8) | ocu[idx + 1];
+
+			if (c == EXT_MARK) {
+				if (ext_i_len)
+					i_ext = idx;
+				break;
+			}
+		}
+		if (i_ext >= 0) {
+			/* Convert extension */
+			ext_max_len = min_t(int, sizeof(ext), str_max_len);
+			ext[ext_o_len++] = EXT_MARK;
+			idx = i_ext + u_ch;
+			while (udf_name_conv_char(ext, ext_max_len, &ext_o_len,
+						  ocu, ocu_len, &idx,
+						  u_ch, &needsCRC,
+						  conv_f, translate)) {
+				if ((ext_o_len + CRC_LEN) < str_max_len)
+					ext_crc_len = ext_o_len;
+			}
+		}
+	}
+
+	idx = 0;
+	while (1) {
+		if (translate && (idx == i_ext)) {
+			if (str_o_len > (str_max_len - ext_o_len))
+				needsCRC = 1;
+			break;
+		}
+
+		if (!udf_name_conv_char(str_o, str_max_len, &str_o_len,
+					ocu, ocu_len, &idx,
+					u_ch, &needsCRC, conv_f, translate))
+			break;
+
+		if (translate &&
+		    (str_o_len <= (str_max_len - ext_o_len - CRC_LEN)))
+			o_crc = str_o_len;
+	}
+
+	if (translate) {
+		if (str_o_len > 0 && str_o_len <= 2 && str_o[0] == '.' &&
+		    (str_o_len == 1 || str_o[1] == '.'))
+			needsCRC = 1;
+		if (needsCRC) {
+			str_o_len = o_crc;
+			valueCRC = crc_itu_t(0, ocu, ocu_len);
+			crc[0] = CRC_MARK;
+			crc[1] = hex_asc_upper_hi(valueCRC >> 8);
+			crc[2] = hex_asc_upper_lo(valueCRC >> 8);
+			crc[3] = hex_asc_upper_hi(valueCRC);
+			crc[4] = hex_asc_upper_lo(valueCRC);
+			len = min_t(int, CRC_LEN, str_max_len - str_o_len);
+			memcpy(&str_o[str_o_len], crc, len);
+			str_o_len += len;
+			ext_o_len = ext_crc_len;
+		}
+		if (ext_o_len > 0) {
+			memcpy(&str_o[str_o_len], ext, ext_o_len);
+			str_o_len += ext_o_len;
+		}
+	}
+
+	return str_o_len;
+}
+
+static int udf_name_to_CS0(struct super_block *sb,
+			   uint8_t *ocu, int ocu_max_len,
+			   const uint8_t *str_i, int str_len)
+{
+	int i, len;
+	unsigned int max_val;
+	int u_len, u_ch;
+	unicode_t uni_char;
+	int (*conv_f)(const unsigned char *, int, wchar_t *);
+
+	if (ocu_max_len <= 0)
+		return 0;
+
+	if (UDF_SB(sb)->s_nls_map)
+		conv_f = UDF_SB(sb)->s_nls_map->char2uni;
+	else
+		conv_f = NULL;
+
+	memset(ocu, 0, ocu_max_len);
+	ocu[0] = 8;
+	max_val = 0xff;
+	u_ch = 1;
+
+try_again:
+	u_len = 1;
+	for (i = 0; i < str_len; i += len) {
+		/* Name didn't fit? */
+		if (u_len + u_ch > ocu_max_len)
+			return 0;
+		if (conv_f) {
+			wchar_t wchar;
+
+			len = conv_f(&str_i[i], str_len - i, &wchar);
+			if (len > 0)
+				uni_char = wchar;
+		} else {
+			len = utf8_to_utf32(&str_i[i], str_len - i,
+					    &uni_char);
+		}
+		/* Invalid character, deal with it */
+		if (len <= 0 || uni_char > UNICODE_MAX) {
+			len = 1;
+			uni_char = '?';
+		}
+
+		if (uni_char > max_val) {
+			unicode_t c;
+
+			if (max_val == 0xff) {
+				max_val = 0xffff;
+				ocu[0] = 0x10;
+				u_ch = 2;
+				goto try_again;
+			}
+			/*
+			 * Use UTF-16 encoding for chars outside we
+			 * cannot encode directly.
+			 */
+			if (u_len + 2 * u_ch > ocu_max_len)
+				return 0;
+
+			uni_char -= PLANE_SIZE;
+			c = SURROGATE_PAIR |
+			    ((uni_char >> SURROGATE_CHAR_BITS) &
+			     SURROGATE_CHAR_MASK);
+			ocu[u_len++] = (uint8_t)(c >> 8);
+			ocu[u_len++] = (uint8_t)(c & 0xff);
+			uni_char = SURROGATE_PAIR | SURROGATE_LOW |
+					(uni_char & SURROGATE_CHAR_MASK);
+		}
+
+		if (max_val == 0xffff)
+			ocu[u_len++] = (uint8_t)(uni_char >> 8);
+		ocu[u_len++] = (uint8_t)(uni_char & 0xff);
+	}
+
+	return u_len;
+}
+
+/*
+ * Convert CS0 dstring to output charset. Warning: This function may truncate
+ * input string if it is too long as it is used for informational strings only
+ * and it is better to truncate the string than to refuse mounting a media.
+ */
+int udf_dstrCS0toChar(struct super_block *sb, uint8_t *utf_o, int o_len,
+		      const uint8_t *ocu_i, int i_len)
+{
+	int s_len = 0;
+
+	if (i_len > 0) {
+		s_len = ocu_i[i_len - 1];
+		if (s_len >= i_len) {
+			pr_warn("incorrect dstring lengths (%d/%d),"
+				" truncating\n", s_len, i_len);
+			s_len = i_len - 1;
+			/* 2-byte encoding? Need to round properly... */
+			if (ocu_i[0] == 16)
+				s_len -= (s_len - 1) & 2;
+		}
+	}
+
+	return udf_name_from_CS0(sb, utf_o, o_len, ocu_i, s_len, 0);
+}
+
+int udf_get_filename(struct super_block *sb, const uint8_t *sname, int slen,
+		     uint8_t *dname, int dlen)
+{
+	int ret;
+
+	if (!slen)
+		return -EIO;
+
+	if (dlen <= 0)
+		return 0;
+
+	ret = udf_name_from_CS0(sb, dname, dlen, sname, slen, 1);
+	/* Zero length filename isn't valid... */
+	if (ret == 0)
+		ret = -EINVAL;
+	return ret;
+}
+
+int udf_put_filename(struct super_block *sb, const uint8_t *sname, int slen,
+		     uint8_t *dname, int dlen)
+{
+	return udf_name_to_CS0(sb, dname, dlen, sname, slen);
+}
+