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, 91 insertions, 0 deletions

diff --git a/fs/ntfs/mft.h b/fs/ntfs/mft.h
new file mode 100644
index 000000000..75a51a98d
--- /dev/null
+++ b/fs/ntfs/mft.h
@@ -0,0 +1,91 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Defines for mft record handling in NTFS Linux kernel driver.
+ *
+ * Copyright (c) 2001-2004 Anton Altaparmakov
+ */
+
+#ifndef _LINUX_NTFS_MFT_H
+#define _LINUX_NTFS_MFT_H
+
+#include <linux/highmem.h>
+#include <linux/pagemap.h>
+
+#include "inode.h"
+
+struct mft_record *map_mft_record(struct ntfs_inode *ni);
+void unmap_mft_record(struct ntfs_inode *ni);
+struct mft_record *map_extent_mft_record(struct ntfs_inode *base_ni, u64 mref,
+		struct ntfs_inode **ntfs_ino);
+
+static inline void unmap_extent_mft_record(struct ntfs_inode *ni)
+{
+	unmap_mft_record(ni);
+}
+
+void __mark_mft_record_dirty(struct ntfs_inode *ni);
+
+/*
+ * mark_mft_record_dirty - set the mft record and the page containing it dirty
+ * @ni:		ntfs inode describing the mapped mft record
+ *
+ * Set the mapped (extent) mft record of the (base or extent) ntfs inode @ni,
+ * as well as the page containing the mft record, dirty.  Also, mark the base
+ * vfs inode dirty.  This ensures that any changes to the mft record are
+ * written out to disk.
+ *
+ * NOTE:  Do not do anything if the mft record is already marked dirty.
+ */
+static inline void mark_mft_record_dirty(struct ntfs_inode *ni)
+{
+	if (!NInoTestSetDirty(ni))
+		__mark_mft_record_dirty(ni);
+}
+
+int ntfs_sync_mft_mirror(struct ntfs_volume *vol, const u64 mft_no,
+		struct mft_record *m);
+int write_mft_record_nolock(struct ntfs_inode *ni, struct mft_record *m, int sync);
+
+/*
+ * write_mft_record - write out a mapped (extent) mft record
+ * @ni:		ntfs inode describing the mapped (extent) mft record
+ * @m:		mapped (extent) mft record to write
+ * @sync:	if true, wait for i/o completion
+ *
+ * This is just a wrapper for write_mft_record_nolock() (see mft.c), which
+ * locks the page for the duration of the write.  This ensures that there are
+ * no race conditions between writing the mft record via the dirty inode code
+ * paths and via the page cache write back code paths or between writing
+ * neighbouring mft records residing in the same page.
+ *
+ * Locking the page also serializes us against ->read_folio() if the page is not
+ * uptodate.
+ *
+ * On success, clean the mft record and return 0.  On error, leave the mft
+ * record dirty and return -errno.
+ */
+static inline int write_mft_record(struct ntfs_inode *ni, struct mft_record *m, int sync)
+{
+	struct folio *folio = ni->folio;
+	int err;
+
+	folio_lock(folio);
+	err = write_mft_record_nolock(ni, m, sync);
+	folio_unlock(folio);
+
+	return err;
+}
+
+int ntfs_mft_record_alloc(struct ntfs_volume *vol, const int mode,
+		struct ntfs_inode **ni, struct ntfs_inode *base_ni,
+		struct mft_record **ni_mrec);
+int ntfs_mft_record_free(struct ntfs_volume *vol, struct ntfs_inode *ni);
+int ntfs_mft_records_write(const struct ntfs_volume *vol, const u64 mref,
+		const s64 count, struct mft_record *b);
+int ntfs_mft_record_check(const struct ntfs_volume *vol, struct mft_record *m,
+			  u64 mft_no);
+int ntfs_mft_writepages(struct address_space *mapping,
+		struct writeback_control *wbc);
+void ntfs_mft_mark_dirty(struct folio *folio);
+
+#endif /* _LINUX_NTFS_MFT_H */