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, 203 insertions, 0 deletions

diff --git a/fs/bfs/file.c b/fs/bfs/file.c
new file mode 100644
index 000000000..d33d6bde9
--- /dev/null
+++ b/fs/bfs/file.c
@@ -0,0 +1,203 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *	fs/bfs/file.c
+ *	BFS file operations.
+ *	Copyright (C) 1999-2018 Tigran Aivazian <aivazian.tigran@gmail.com>
+ *
+ *	Make the file block allocation algorithm understand the size
+ *	of the underlying block device.
+ *	Copyright (C) 2007 Dmitri Vorobiev <dmitri.vorobiev@gmail.com>
+ *
+ */
+
+#include <linux/fs.h>
+#include <linux/mpage.h>
+#include <linux/buffer_head.h>
+#include "bfs.h"
+
+#undef DEBUG
+
+#ifdef DEBUG
+#define dprintf(x...)	printf(x)
+#else
+#define dprintf(x...)
+#endif
+
+const struct file_operations bfs_file_operations = {
+	.llseek 	= generic_file_llseek,
+	.read_iter	= generic_file_read_iter,
+	.write_iter	= generic_file_write_iter,
+	.mmap_prepare	= generic_file_mmap_prepare,
+	.splice_read	= filemap_splice_read,
+};
+
+static int bfs_move_block(unsigned long from, unsigned long to,
+					struct super_block *sb)
+{
+	struct buffer_head *bh, *new;
+
+	bh = sb_bread(sb, from);
+	if (!bh)
+		return -EIO;
+	new = sb_getblk(sb, to);
+	memcpy(new->b_data, bh->b_data, bh->b_size);
+	mark_buffer_dirty(new);
+	bforget(bh);
+	brelse(new);
+	return 0;
+}
+
+static int bfs_move_blocks(struct super_block *sb, unsigned long start,
+				unsigned long end, unsigned long where)
+{
+	unsigned long i;
+
+	dprintf("%08lx-%08lx->%08lx\n", start, end, where);
+	for (i = start; i <= end; i++)
+		if(bfs_move_block(i, where + i, sb)) {
+			dprintf("failed to move block %08lx -> %08lx\n", i,
+								where + i);
+			return -EIO;
+		}
+	return 0;
+}
+
+static int bfs_get_block(struct inode *inode, sector_t block,
+			struct buffer_head *bh_result, int create)
+{
+	unsigned long phys;
+	int err;
+	struct super_block *sb = inode->i_sb;
+	struct bfs_sb_info *info = BFS_SB(sb);
+	struct bfs_inode_info *bi = BFS_I(inode);
+
+	phys = bi->i_sblock + block;
+	if (!create) {
+		if (phys <= bi->i_eblock) {
+			dprintf("c=%d, b=%08lx, phys=%09lx (granted)\n",
+                                create, (unsigned long)block, phys);
+			map_bh(bh_result, sb, phys);
+		}
+		return 0;
+	}
+
+	/*
+	 * If the file is not empty and the requested block is within the
+	 * range of blocks allocated for this file, we can grant it.
+	 */
+	if (bi->i_sblock && (phys <= bi->i_eblock)) {
+		dprintf("c=%d, b=%08lx, phys=%08lx (interim block granted)\n", 
+				create, (unsigned long)block, phys);
+		map_bh(bh_result, sb, phys);
+		return 0;
+	}
+
+	/* The file will be extended, so let's see if there is enough space. */
+	if (phys >= info->si_blocks)
+		return -ENOSPC;
+
+	/* The rest has to be protected against itself. */
+	mutex_lock(&info->bfs_lock);
+
+	/*
+	 * If the last data block for this file is the last allocated
+	 * block, we can extend the file trivially, without moving it
+	 * anywhere.
+	 */
+	if (bi->i_eblock == info->si_lf_eblk) {
+		dprintf("c=%d, b=%08lx, phys=%08lx (simple extension)\n", 
+				create, (unsigned long)block, phys);
+		map_bh(bh_result, sb, phys);
+		info->si_freeb -= phys - bi->i_eblock;
+		info->si_lf_eblk = bi->i_eblock = phys;
+		mark_inode_dirty(inode);
+		err = 0;
+		goto out;
+	}
+
+	/* Ok, we have to move this entire file to the next free block. */
+	phys = info->si_lf_eblk + 1;
+	if (phys + block >= info->si_blocks) {
+		err = -ENOSPC;
+		goto out;
+	}
+
+	if (bi->i_sblock) {
+		err = bfs_move_blocks(inode->i_sb, bi->i_sblock, 
+						bi->i_eblock, phys);
+		if (err) {
+			dprintf("failed to move ino=%08lx -> fs corruption\n",
+								inode->i_ino);
+			goto out;
+		}
+	} else
+		err = 0;
+
+	dprintf("c=%d, b=%08lx, phys=%08lx (moved)\n",
+                create, (unsigned long)block, phys);
+	bi->i_sblock = phys;
+	phys += block;
+	info->si_lf_eblk = bi->i_eblock = phys;
+
+	/*
+	 * This assumes nothing can write the inode back while we are here
+	 * and thus update inode->i_blocks! (XXX)
+	 */
+	info->si_freeb -= bi->i_eblock - bi->i_sblock + 1 - inode->i_blocks;
+	mark_inode_dirty(inode);
+	map_bh(bh_result, sb, phys);
+out:
+	mutex_unlock(&info->bfs_lock);
+	return err;
+}
+
+static int bfs_writepages(struct address_space *mapping,
+		struct writeback_control *wbc)
+{
+	return mpage_writepages(mapping, wbc, bfs_get_block);
+}
+
+static int bfs_read_folio(struct file *file, struct folio *folio)
+{
+	return block_read_full_folio(folio, bfs_get_block);
+}
+
+static void bfs_write_failed(struct address_space *mapping, loff_t to)
+{
+	struct inode *inode = mapping->host;
+
+	if (to > inode->i_size)
+		truncate_pagecache(inode, inode->i_size);
+}
+
+static int bfs_write_begin(const struct kiocb *iocb,
+			   struct address_space *mapping,
+			   loff_t pos, unsigned len,
+			   struct folio **foliop, void **fsdata)
+{
+	int ret;
+
+	ret = block_write_begin(mapping, pos, len, foliop, bfs_get_block);
+	if (unlikely(ret))
+		bfs_write_failed(mapping, pos + len);
+
+	return ret;
+}
+
+static sector_t bfs_bmap(struct address_space *mapping, sector_t block)
+{
+	return generic_block_bmap(mapping, block, bfs_get_block);
+}
+
+const struct address_space_operations bfs_aops = {
+	.dirty_folio	= block_dirty_folio,
+	.invalidate_folio = block_invalidate_folio,
+	.read_folio	= bfs_read_folio,
+	.writepages	= bfs_writepages,
+	.write_begin	= bfs_write_begin,
+	.write_end	= generic_write_end,
+	.migrate_folio	= buffer_migrate_folio,
+	.bmap		= bfs_bmap,
+};
+
+const struct inode_operations bfs_file_inops;