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, 157 insertions, 0 deletions

diff --git a/fs/hfsplus/part_tbl.c b/fs/hfsplus/part_tbl.c
new file mode 100644
index 000000000..9ec21664e
--- /dev/null
+++ b/fs/hfsplus/part_tbl.c
@@ -0,0 +1,157 @@
+/*
+ * linux/fs/hfsplus/part_tbl.c
+ *
+ * Copyright (C) 1996-1997  Paul H. Hargrove
+ * This file may be distributed under the terms of
+ * the GNU General Public License.
+ *
+ * Original code to handle the new style Mac partition table based on
+ * a patch contributed by Holger Schemel (aeglos@valinor.owl.de).
+ *
+ * In function preconditions the term "valid" applied to a pointer to
+ * a structure means that the pointer is non-NULL and the structure it
+ * points to has all fields initialized to consistent values.
+ *
+ */
+
+#include <linux/slab.h>
+#include "hfsplus_fs.h"
+
+/* offsets to various blocks */
+#define HFS_DD_BLK		0 /* Driver Descriptor block */
+#define HFS_PMAP_BLK		1 /* First block of partition map */
+#define HFS_MDB_BLK		2 /* Block (w/i partition) of MDB */
+
+/* magic numbers for various disk blocks */
+#define HFS_DRVR_DESC_MAGIC	0x4552 /* "ER": driver descriptor map */
+#define HFS_OLD_PMAP_MAGIC	0x5453 /* "TS": old-type partition map */
+#define HFS_NEW_PMAP_MAGIC	0x504D /* "PM": new-type partition map */
+#define HFS_SUPER_MAGIC		0x4244 /* "BD": HFS MDB (super block) */
+#define HFS_MFS_SUPER_MAGIC	0xD2D7 /* MFS MDB (super block) */
+
+/*
+ * The new style Mac partition map
+ *
+ * For each partition on the media there is a physical block (512-byte
+ * block) containing one of these structures.  These blocks are
+ * contiguous starting at block 1.
+ */
+struct new_pmap {
+	__be16	pmSig;		/* signature */
+	__be16	reSigPad;	/* padding */
+	__be32	pmMapBlkCnt;	/* partition blocks count */
+	__be32	pmPyPartStart;	/* physical block start of partition */
+	__be32	pmPartBlkCnt;	/* physical block count of partition */
+	u8	pmPartName[32];	/* (null terminated?) string
+				   giving the name of this
+				   partition */
+	u8	pmPartType[32];	/* (null terminated?) string
+				   giving the type of this
+				   partition */
+	/* a bunch more stuff we don't need */
+} __packed;
+
+/*
+ * The old style Mac partition map
+ *
+ * The partition map consists for a 2-byte signature followed by an
+ * array of these structures.  The map is terminated with an all-zero
+ * one of these.
+ */
+struct old_pmap {
+	__be16		pdSig;	/* Signature bytes */
+	struct old_pmap_entry {
+		__be32	pdStart;
+		__be32	pdSize;
+		__be32	pdFSID;
+	}	pdEntry[42];
+} __packed;
+
+static int hfs_parse_old_pmap(struct super_block *sb, struct old_pmap *pm,
+		sector_t *part_start, sector_t *part_size)
+{
+	struct hfsplus_sb_info *sbi = HFSPLUS_SB(sb);
+	int i;
+
+	for (i = 0; i < 42; i++) {
+		struct old_pmap_entry *p = &pm->pdEntry[i];
+
+		if (p->pdStart && p->pdSize &&
+		    p->pdFSID == cpu_to_be32(0x54465331)/*"TFS1"*/ &&
+		    (sbi->part < 0 || sbi->part == i)) {
+			*part_start += be32_to_cpu(p->pdStart);
+			*part_size = be32_to_cpu(p->pdSize);
+			return 0;
+		}
+	}
+
+	return -ENOENT;
+}
+
+static int hfs_parse_new_pmap(struct super_block *sb, void *buf,
+		struct new_pmap *pm, sector_t *part_start, sector_t *part_size)
+{
+	struct hfsplus_sb_info *sbi = HFSPLUS_SB(sb);
+	int size = be32_to_cpu(pm->pmMapBlkCnt);
+	int buf_size = hfsplus_min_io_size(sb);
+	int res;
+	int i = 0;
+
+	do {
+		if (!memcmp(pm->pmPartType, "Apple_HFS", 9) &&
+		    (sbi->part < 0 || sbi->part == i)) {
+			*part_start += be32_to_cpu(pm->pmPyPartStart);
+			*part_size = be32_to_cpu(pm->pmPartBlkCnt);
+			return 0;
+		}
+
+		if (++i >= size)
+			return -ENOENT;
+
+		pm = (struct new_pmap *)((u8 *)pm + HFSPLUS_SECTOR_SIZE);
+		if ((u8 *)pm - (u8 *)buf >= buf_size) {
+			res = hfsplus_submit_bio(sb,
+						 *part_start + HFS_PMAP_BLK + i,
+						 buf, (void **)&pm, REQ_OP_READ);
+			if (res)
+				return res;
+		}
+	} while (pm->pmSig == cpu_to_be16(HFS_NEW_PMAP_MAGIC));
+
+	return -ENOENT;
+}
+
+/*
+ * Parse the partition map looking for the start and length of a
+ * HFS/HFS+ partition.
+ */
+int hfs_part_find(struct super_block *sb,
+		sector_t *part_start, sector_t *part_size)
+{
+	void *buf, *data;
+	int res;
+
+	buf = kmalloc(hfsplus_min_io_size(sb), GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+
+	res = hfsplus_submit_bio(sb, *part_start + HFS_PMAP_BLK,
+				 buf, &data, REQ_OP_READ);
+	if (res)
+		goto out;
+
+	switch (be16_to_cpu(*((__be16 *)data))) {
+	case HFS_OLD_PMAP_MAGIC:
+		res = hfs_parse_old_pmap(sb, data, part_start, part_size);
+		break;
+	case HFS_NEW_PMAP_MAGIC:
+		res = hfs_parse_new_pmap(sb, buf, data, part_start, part_size);
+		break;
+	default:
+		res = -ENOENT;
+		break;
+	}
+out:
+	kfree(buf);
+	return res;
+}