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, 120 insertions, 0 deletions

diff --git a/fs/ntfs/bdev-io.c b/fs/ntfs/bdev-io.c
new file mode 100644
index 000000000..4f27eed3b
--- /dev/null
+++ b/fs/ntfs/bdev-io.c
@@ -0,0 +1,120 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * NTFS block device I/O.
+ *
+ * Copyright (c) 2026 LG Electronics Co., Ltd.
+ */
+
+#include <linux/blkdev.h>
+
+#include "ntfs.h"
+
+/*
+ * ntfs_bdev_read - Read data directly from block device using bio
+ * @bdev:	block device to read from
+ * @data:	destination buffer
+ * @start:	starting byte offset on the block device
+ * @size:	number of bytes to read
+ *
+ * Reads @size bytes starting from byte offset @start directly from the block
+ * device using one or more BIOs. This function bypasses the page cache
+ * completely and performs synchronous I/O with REQ_META | REQ_SYNC flags set.
+ *
+ * The @start offset must be sector-aligned (512 bytes). If it is not aligned,
+ * the function will return -EINVAL.
+ *
+ * If the destination buffer @data is not a vmalloc address, it falls back
+ * to the more efficient bdev_rw_virt() helper.
+ *
+ * Return: 0 on success, negative error code on failure.
+ */
+int ntfs_bdev_read(struct block_device *bdev, char *data, loff_t start, size_t size)
+{
+	unsigned int done = 0, added;
+	int error;
+	struct bio *bio;
+	blk_opf_t op;
+	sector_t sector = ntfs_bytes_to_bio_sector(start);
+
+	if (start & (SECTOR_SIZE - 1))
+		return -EINVAL;
+
+	op = REQ_OP_READ | REQ_META | REQ_SYNC;
+	if (!is_vmalloc_addr(data))
+		return bdev_rw_virt(bdev, sector, data, size, op);
+
+	bio = bio_alloc(bdev,
+			bio_max_segs(DIV_ROUND_UP(size, PAGE_SIZE)),
+			op, GFP_KERNEL);
+	bio->bi_iter.bi_sector = sector;
+
+	do {
+		added = bio_add_vmalloc_chunk(bio, data + done, size - done);
+		if (!added) {
+			struct bio	*prev = bio;
+
+			bio = bio_alloc(prev->bi_bdev,
+					bio_max_segs(DIV_ROUND_UP(size - done, PAGE_SIZE)),
+					prev->bi_opf, GFP_KERNEL);
+			bio->bi_iter.bi_sector = bio_end_sector(prev);
+			bio_chain(prev, bio);
+			submit_bio(prev);
+		}
+		done += added;
+	} while (done < size);
+
+	error = submit_bio_wait(bio);
+	bio_put(bio);
+
+	if ((op & REQ_OP_MASK) == REQ_OP_READ)
+		invalidate_kernel_vmap_range(data, size);
+	return error;
+}
+
+/*
+ * ntfs_bdev_write - Update block device contents via page cache
+ * @sb:		super block of the mounted NTFS filesystem
+ * @buf:	source buffer containing data to write
+ * @start:	starting byte offset on the block device
+ * @size:	number of bytes to write
+ *
+ * Writes @size bytes from @buf to the block device (sb->s_bdev) starting
+ * at byte offset @start. The write is performed entirely through the page
+ * cache of the block device's address space.
+ */
+int ntfs_bdev_write(struct super_block *sb, void *buf, loff_t start, size_t size)
+{
+	pgoff_t idx, idx_end;
+	loff_t offset, end = start + size;
+	u32 from, to, buf_off = 0;
+	struct folio *folio;
+
+	idx = start >> PAGE_SHIFT;
+	idx_end = end >> PAGE_SHIFT;
+	from = start & ~PAGE_MASK;
+
+	if (idx == idx_end)
+		idx_end++;
+
+	for (; idx < idx_end; idx++, from = 0) {
+		u32 len;
+
+		folio = read_mapping_folio(sb->s_bdev->bd_mapping, idx, NULL);
+		if (IS_ERR(folio)) {
+			ntfs_error(sb, "Unable to read %ld page", idx);
+			return PTR_ERR(folio);
+		}
+
+		offset = (loff_t)idx << PAGE_SHIFT;
+		to = min_t(u32, end - offset, PAGE_SIZE);
+		len = to - from;
+
+		memcpy_to_folio(folio, from, buf + buf_off, len);
+		buf_off += len;
+		folio_mark_uptodate(folio);
+		folio_mark_dirty(folio);
+		folio_put(folio);
+	}
+
+	return 0;
+}