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, 462 insertions, 0 deletions

diff --git a/fs/hfs/mdb.c b/fs/hfs/mdb.c
new file mode 100644
index 000000000..277de712f
--- /dev/null
+++ b/fs/hfs/mdb.c
@@ -0,0 +1,462 @@
+/*
+ *  linux/fs/hfs/mdb.c
+ *
+ * Copyright (C) 1995-1997  Paul H. Hargrove
+ * (C) 2003 Ardis Technologies <roman@ardistech.com>
+ * This file may be distributed under the terms of the GNU General Public License.
+ *
+ * This file contains functions for reading/writing the MDB.
+ */
+
+#include <linux/cdrom.h>
+#include <linux/blkdev.h>
+#include <linux/nls.h>
+#include <linux/slab.h>
+
+#include "hfs_fs.h"
+#include "btree.h"
+
+/*================ File-local data types ================*/
+
+/*
+ * The HFS Master Directory Block (MDB).
+ *
+ * Also known as the Volume Information Block (VIB), this structure is
+ * the HFS equivalent of a superblock.
+ *
+ * Reference: _Inside Macintosh: Files_ pages 2-59 through 2-62
+ *
+ * modified for HFS Extended
+ */
+
+static int hfs_get_last_session(struct super_block *sb,
+				sector_t *start, sector_t *size)
+{
+	struct cdrom_device_info *cdi = disk_to_cdi(sb->s_bdev->bd_disk);
+
+	/* default values */
+	*start = 0;
+	*size = bdev_nr_sectors(sb->s_bdev);
+
+	if (HFS_SB(sb)->session >= 0) {
+		struct cdrom_tocentry te;
+	
+		if (!cdi)
+			return -EINVAL;
+
+		te.cdte_track = HFS_SB(sb)->session;
+		te.cdte_format = CDROM_LBA;
+		if (cdrom_read_tocentry(cdi, &te) ||
+		    (te.cdte_ctrl & CDROM_DATA_TRACK) != 4) {
+			pr_err("invalid session number or type of track\n");
+			return -EINVAL;
+		}
+
+		*start = (sector_t)te.cdte_addr.lba << 2;
+	} else if (cdi) {
+		struct cdrom_multisession ms_info;
+
+		ms_info.addr_format = CDROM_LBA;
+		if (cdrom_multisession(cdi, &ms_info) == 0 && ms_info.xa_flag)
+			*start = (sector_t)ms_info.addr.lba << 2;
+	}
+
+	return 0;
+}
+
+bool is_hfs_cnid_counts_valid(struct super_block *sb)
+{
+	struct hfs_sb_info *sbi = HFS_SB(sb);
+	bool corrupted = false;
+
+	if (unlikely(atomic64_read(&sbi->next_id) > U32_MAX)) {
+		pr_warn("next CNID exceeds limit\n");
+		corrupted = true;
+	}
+	if (unlikely(atomic64_read(&sbi->file_count) > U32_MAX)) {
+		pr_warn("file count exceeds limit\n");
+		corrupted = true;
+	}
+	if (unlikely(atomic64_read(&sbi->folder_count) > U32_MAX)) {
+		pr_warn("folder count exceeds limit\n");
+		corrupted = true;
+	}
+
+	return !corrupted;
+}
+
+/*
+ * hfs_sect_offset() - get byte offset within the buffer_head.
+ */
+static unsigned int hfs_sect_offset(struct super_block *sb, sector_t sec)
+{
+	loff_t start = (loff_t)sec << HFS_SECTOR_SIZE_BITS;
+
+	return start & (sb->s_blocksize - 1);
+}
+
+/*
+ * hfs_mdb_publish() - copy the in-memory primary MDB to the on-disk buffer.
+ */
+static void hfs_mdb_publish(struct hfs_sb_info *sbi)
+{
+	lock_buffer(sbi->mdb_bh);
+	memcpy(sbi->mdb_bh->b_data + sbi->mdb_offset, sbi->mdb, HFS_SECTOR_SIZE);
+	mark_buffer_dirty(sbi->mdb_bh);
+	unlock_buffer(sbi->mdb_bh);
+}
+
+/*
+ * hfs_alt_mdb_publish() - copy the in-memory alternate MDB to its buffer.
+ */
+static void hfs_alt_mdb_publish(struct hfs_sb_info *sbi)
+{
+	lock_buffer(sbi->alt_mdb_bh);
+	memcpy(sbi->alt_mdb_bh->b_data + sbi->alt_mdb_offset, sbi->alt_mdb,
+	       HFS_SECTOR_SIZE);
+	mark_buffer_dirty(sbi->alt_mdb_bh);
+	unlock_buffer(sbi->alt_mdb_bh);
+}
+
+/*
+ * hfs_mdb_get()
+ *
+ * Build the in-core MDB for a filesystem, including
+ * the B-trees and the volume bitmap.
+ */
+int hfs_mdb_get(struct super_block *sb)
+{
+	struct buffer_head *bh;
+	struct hfs_mdb *mdb, *alt_mdb;
+	unsigned int block;
+	char *ptr;
+	int off2, len, size, sect;
+	sector_t part_start, part_size;
+	loff_t off;
+	__be16 attrib;
+
+	/* set the device driver to 512-byte blocks */
+	size = sb_min_blocksize(sb, HFS_SECTOR_SIZE);
+	if (!size)
+		return -EINVAL;
+
+	if (hfs_get_last_session(sb, &part_start, &part_size))
+		return -EINVAL;
+	while (1) {
+		/* See if this is an HFS filesystem */
+		bh = sb_bread512(sb, part_start + HFS_MDB_BLK, mdb);
+		if (!bh)
+			return -EIO;
+
+		if (mdb->drSigWord == cpu_to_be16(HFS_SUPER_MAGIC))
+			break;
+		brelse(bh);
+
+		/* check for a partition block
+		 * (should do this only for cdrom/loop though)
+		 */
+		if (hfs_part_find(sb, &part_start, &part_size))
+			return -EIO;
+	}
+
+	HFS_SB(sb)->alloc_blksz = size = be32_to_cpu(mdb->drAlBlkSiz);
+	if (!size || (size & (HFS_SECTOR_SIZE - 1))) {
+		pr_err("bad allocation block size %d\n", size);
+		brelse(bh);
+		return -EIO;
+	}
+
+	size = min(HFS_SB(sb)->alloc_blksz, (u32)PAGE_SIZE);
+	/* size must be a multiple of 512 */
+	while (size & (size - 1))
+		size -= HFS_SECTOR_SIZE;
+	sect = be16_to_cpu(mdb->drAlBlSt) + part_start;
+	/* align block size to first sector */
+	while (sect & ((size - 1) >> HFS_SECTOR_SIZE_BITS))
+		size >>= 1;
+	/* align block size to weird alloc size */
+	while (HFS_SB(sb)->alloc_blksz & (size - 1))
+		size >>= 1;
+	brelse(bh);
+	if (!sb_set_blocksize(sb, size)) {
+		pr_err("unable to set blocksize to %u\n", size);
+		return -EIO;
+	}
+
+	bh = sb_bread512(sb, part_start + HFS_MDB_BLK, mdb);
+	if (!bh)
+		return -EIO;
+	if (mdb->drSigWord != cpu_to_be16(HFS_SUPER_MAGIC)) {
+		brelse(bh);
+		return -EIO;
+	}
+
+	mdb = kmemdup(mdb, HFS_SECTOR_SIZE, GFP_KERNEL);
+	if (!mdb) {
+		brelse(bh);
+		return -ENOMEM;
+	}
+
+	HFS_SB(sb)->mdb_bh = bh;
+	HFS_SB(sb)->mdb_offset = hfs_sect_offset(sb, part_start + HFS_MDB_BLK);
+	HFS_SB(sb)->mdb = mdb;
+
+	/* These parameters are read from the MDB, and never written */
+	HFS_SB(sb)->part_start = part_start;
+	HFS_SB(sb)->fs_ablocks = be16_to_cpu(mdb->drNmAlBlks);
+	HFS_SB(sb)->fs_div = HFS_SB(sb)->alloc_blksz >> sb->s_blocksize_bits;
+	HFS_SB(sb)->clumpablks = be32_to_cpu(mdb->drClpSiz) /
+				 HFS_SB(sb)->alloc_blksz;
+	if (!HFS_SB(sb)->clumpablks)
+		HFS_SB(sb)->clumpablks = 1;
+	HFS_SB(sb)->fs_start = (be16_to_cpu(mdb->drAlBlSt) + part_start) >>
+			       (sb->s_blocksize_bits - HFS_SECTOR_SIZE_BITS);
+
+	/* These parameters are read from and written to the MDB */
+	HFS_SB(sb)->free_ablocks = be16_to_cpu(mdb->drFreeBks);
+	atomic64_set(&HFS_SB(sb)->next_id, be32_to_cpu(mdb->drNxtCNID));
+	HFS_SB(sb)->root_files = be16_to_cpu(mdb->drNmFls);
+	HFS_SB(sb)->root_dirs = be16_to_cpu(mdb->drNmRtDirs);
+	atomic64_set(&HFS_SB(sb)->file_count, be32_to_cpu(mdb->drFilCnt));
+	atomic64_set(&HFS_SB(sb)->folder_count, be32_to_cpu(mdb->drDirCnt));
+
+	if (!is_hfs_cnid_counts_valid(sb)) {
+		pr_warn("filesystem possibly corrupted, running fsck.hfs is recommended. Mounting read-only.\n");
+		sb->s_flags |= SB_RDONLY;
+	}
+
+	/* TRY to get the alternate (backup) MDB. */
+	sect = part_start + part_size - 2;
+	bh = sb_bread512(sb, sect, alt_mdb);
+	if (bh) {
+		if (alt_mdb->drSigWord == cpu_to_be16(HFS_SUPER_MAGIC)) {
+			alt_mdb = kmemdup(alt_mdb, HFS_SECTOR_SIZE, GFP_KERNEL);
+			if (alt_mdb) {
+				HFS_SB(sb)->alt_mdb_bh = bh;
+				HFS_SB(sb)->alt_mdb_offset =
+					hfs_sect_offset(sb, sect);
+				HFS_SB(sb)->alt_mdb = alt_mdb;
+			} else {
+				brelse(bh);
+			}
+		} else
+			brelse(bh);
+	}
+
+	if (!HFS_SB(sb)->alt_mdb) {
+		pr_warn("unable to locate alternate MDB\n");
+		pr_warn("continuing without an alternate MDB\n");
+	}
+
+	HFS_SB(sb)->bitmap = kzalloc(8192, GFP_KERNEL);
+	if (!HFS_SB(sb)->bitmap)
+		return -EIO;
+
+	/* read in the bitmap */
+	block = be16_to_cpu(mdb->drVBMSt) + part_start;
+	off = (loff_t)block << HFS_SECTOR_SIZE_BITS;
+	size = (HFS_SB(sb)->fs_ablocks + 8) / 8;
+	ptr = (u8 *)HFS_SB(sb)->bitmap;
+	while (size) {
+		bh = sb_bread(sb, off >> sb->s_blocksize_bits);
+		if (!bh) {
+			pr_err("unable to read volume bitmap\n");
+			return -EIO;
+		}
+		off2 = off & (sb->s_blocksize - 1);
+		len = min((int)sb->s_blocksize - off2, size);
+		memcpy(ptr, bh->b_data + off2, len);
+		brelse(bh);
+		ptr += len;
+		off += len;
+		size -= len;
+	}
+
+	HFS_SB(sb)->ext_tree = hfs_btree_open(sb, HFS_EXT_CNID, hfs_ext_keycmp);
+	if (!HFS_SB(sb)->ext_tree) {
+		pr_err("unable to open extent tree\n");
+		return -EIO;
+	}
+	HFS_SB(sb)->cat_tree = hfs_btree_open(sb, HFS_CAT_CNID, hfs_cat_keycmp);
+	if (!HFS_SB(sb)->cat_tree) {
+		pr_err("unable to open catalog tree\n");
+		return -EIO;
+	}
+
+	attrib = mdb->drAtrb;
+	if (!(attrib & cpu_to_be16(HFS_SB_ATTRIB_UNMNT))) {
+		pr_warn("filesystem was not cleanly unmounted, running fsck.hfs is recommended.	Mounting read-only.\n");
+		sb->s_flags |= SB_RDONLY;
+	}
+	if ((attrib & cpu_to_be16(HFS_SB_ATTRIB_SLOCK))) {
+		pr_warn("filesystem is marked locked, mounting read-only.\n");
+		sb->s_flags |= SB_RDONLY;
+	}
+	if (!sb_rdonly(sb)) {
+		/* Mark the volume uncleanly unmounted in case we crash */
+		attrib &= cpu_to_be16(~HFS_SB_ATTRIB_UNMNT);
+		attrib |= cpu_to_be16(HFS_SB_ATTRIB_INCNSTNT);
+		mdb->drAtrb = attrib;
+		be32_add_cpu(&mdb->drWrCnt, 1);
+		mdb->drLsMod = hfs_mtime();
+
+		hfs_mdb_publish(HFS_SB(sb));
+		sync_dirty_buffer(HFS_SB(sb)->mdb_bh);
+	}
+
+	return 0;
+}
+
+/*
+ * hfs_mdb_commit()
+ *
+ * Description:
+ *   This updates the MDB on disk.
+ *   It does not check, if the superblock has been modified, or
+ *   if the filesystem has been mounted read-only. It is mainly
+ *   called by hfs_sync_fs() and flush_mdb().
+ * Input Variable(s):
+ *   struct hfs_mdb *mdb: Pointer to the hfs MDB
+ *   int backup;
+ * Output Variable(s):
+ *   NONE
+ * Returns:
+ *   0 on success, -EIO if the MDB or alternate MDB buffer is no longer
+ *   valid (e.g. after a prior write error), in which case the volume is
+ *   remounted read-only.
+ * Preconditions:
+ *   'mdb' points to a "valid" (struct hfs_mdb).
+ * Postconditions:
+ *   The HFS MDB and on disk will be updated, by copying the possibly
+ *   modified fields from the in memory MDB (in native byte order) to
+ *   the disk block buffer.
+ *   If 'backup' is non-zero then the alternate MDB is also written
+ *   and the function doesn't return until it is actually on disk.
+ */
+int hfs_mdb_commit(struct super_block *sb)
+{
+	struct hfs_mdb *mdb = HFS_SB(sb)->mdb;
+	int ret = 0;
+
+	if (sb_rdonly(sb))
+		return 0;
+
+	if (!buffer_uptodate(HFS_SB(sb)->mdb_bh)) {
+		pr_err("primary MDB is corrupt, mounting read-only\n");
+		sb->s_flags |= SB_RDONLY;
+		return -EIO;
+	}
+
+	if (test_and_clear_bit(HFS_FLG_MDB_DIRTY, &HFS_SB(sb)->flags)) {
+		/* These parameters may have been modified, so write them back */
+		mdb->drLsMod = hfs_mtime();
+		mdb->drFreeBks = cpu_to_be16(HFS_SB(sb)->free_ablocks);
+		mdb->drNxtCNID =
+			cpu_to_be32((u32)atomic64_read(&HFS_SB(sb)->next_id));
+		mdb->drNmFls = cpu_to_be16(HFS_SB(sb)->root_files);
+		mdb->drNmRtDirs = cpu_to_be16(HFS_SB(sb)->root_dirs);
+		mdb->drFilCnt =
+			cpu_to_be32((u32)atomic64_read(&HFS_SB(sb)->file_count));
+		mdb->drDirCnt =
+			cpu_to_be32((u32)atomic64_read(&HFS_SB(sb)->folder_count));
+
+		hfs_inode_write_fork(HFS_SB(sb)->ext_tree->inode, mdb->drXTExtRec,
+				     &mdb->drXTFlSize, NULL);
+		hfs_inode_write_fork(HFS_SB(sb)->cat_tree->inode, mdb->drCTExtRec,
+				     &mdb->drCTFlSize, NULL);
+
+		/* write MDB to disk */
+		hfs_mdb_publish(HFS_SB(sb));
+		sync_dirty_buffer(HFS_SB(sb)->mdb_bh);
+	}
+
+	/* write the backup MDB, not returning until it is written.
+	 * we only do this when either the catalog or extents overflow
+	 * files grow. */
+	if (test_and_clear_bit(HFS_FLG_ALT_MDB_DIRTY, &HFS_SB(sb)->flags) &&
+	    HFS_SB(sb)->alt_mdb) {
+		if (!buffer_uptodate(HFS_SB(sb)->alt_mdb_bh)) {
+			pr_err("alternate MDB is corrupt, mounting read-only\n");
+			sb->s_flags |= SB_RDONLY;
+			ret = -EIO;
+			goto out;
+		}
+
+		memcpy(HFS_SB(sb)->alt_mdb, mdb, HFS_SECTOR_SIZE);
+		HFS_SB(sb)->alt_mdb->drAtrb |= cpu_to_be16(HFS_SB_ATTRIB_UNMNT);
+		HFS_SB(sb)->alt_mdb->drAtrb &= cpu_to_be16(~HFS_SB_ATTRIB_INCNSTNT);
+
+		hfs_alt_mdb_publish(HFS_SB(sb));
+		sync_dirty_buffer(HFS_SB(sb)->alt_mdb_bh);
+	}
+
+	if (test_and_clear_bit(HFS_FLG_BITMAP_DIRTY, &HFS_SB(sb)->flags)) {
+		struct buffer_head *bh;
+		sector_t block;
+		char *ptr;
+		int off, size, len;
+
+		block = be16_to_cpu(HFS_SB(sb)->mdb->drVBMSt) + HFS_SB(sb)->part_start;
+		off = (block << HFS_SECTOR_SIZE_BITS) & (sb->s_blocksize - 1);
+		block >>= sb->s_blocksize_bits - HFS_SECTOR_SIZE_BITS;
+		size = (HFS_SB(sb)->fs_ablocks + 7) / 8;
+		ptr = (u8 *)HFS_SB(sb)->bitmap;
+		while (size) {
+			bh = sb_bread(sb, block);
+			if (!bh) {
+				pr_err("unable to read volume bitmap\n");
+				break;
+			}
+			len = min((int)sb->s_blocksize - off, size);
+
+			lock_buffer(bh);
+			memcpy(bh->b_data + off, ptr, len);
+			unlock_buffer(bh);
+
+			mark_buffer_dirty(bh);
+			brelse(bh);
+			block++;
+			off = 0;
+			ptr += len;
+			size -= len;
+		}
+	}
+out:
+	return ret;
+}
+
+void hfs_mdb_close(struct super_block *sb)
+{
+	/* update volume attributes */
+	if (sb_rdonly(sb))
+		return;
+
+	if (!buffer_uptodate(HFS_SB(sb)->mdb_bh))
+		return;
+
+	HFS_SB(sb)->mdb->drAtrb |= cpu_to_be16(HFS_SB_ATTRIB_UNMNT);
+	HFS_SB(sb)->mdb->drAtrb &= cpu_to_be16(~HFS_SB_ATTRIB_INCNSTNT);
+	hfs_mdb_publish(HFS_SB(sb));
+}
+
+/*
+ * hfs_mdb_put()
+ *
+ * Release the resources associated with the in-core MDB.  */
+void hfs_mdb_put(struct super_block *sb)
+{
+	/* free the B-trees */
+	hfs_btree_close(HFS_SB(sb)->ext_tree);
+	hfs_btree_close(HFS_SB(sb)->cat_tree);
+
+	/* free the buffers holding the primary and alternate MDBs */
+	brelse(HFS_SB(sb)->mdb_bh);
+	brelse(HFS_SB(sb)->alt_mdb_bh);
+	kfree(HFS_SB(sb)->mdb);
+	kfree(HFS_SB(sb)->alt_mdb);
+
+	unload_nls(HFS_SB(sb)->nls_io);
+	unload_nls(HFS_SB(sb)->nls_disk);
+
+	kfree(HFS_SB(sb)->bitmap);
+}