[PATCH] md: don't hand out the array before md_alloc() has added mddev->kobj

[Yogesh Gaur]

md_alloc() publishes the gendisk before it is done setting the mddev up:

	disk->private_data = mddev;
	...
	error = add_disk(disk);
	if (error)
		goto out_put_disk;

	kobject_init(&mddev->kobj, &md_ktype);
	error = kobject_add(&mddev->kobj, &disk_to_dev(disk)->kobj, "%s", "md");

add_disk() makes /dev/mdN openable, and md_open() only refuses the open
when MD_CLOSING is set.  mddev comes from mddev_alloc() and is zeroed,
so anything that gets in between add_disk() and kobject_add() sees an
mddev->kobj that has never been through kobject_init() - no ktype, no
kref, state_initialized clear.

syzbot opens the array in that window and issues ADD_NEW_DISK:

  kobject: '(null)' (ffff8880120640f0): is not initialized, yet kobject_get() is being called.
  WARNING: lib/kobject.c:642 at kobject_add_internal+0xea/0xcd0 lib/kobject.c:225
   kobject_add_varg lib/kobject.c:374 [inline]
   kobject_add+0x163/0x240 lib/kobject.c:426
   bind_rdev_to_array+0x80c/0xdd0 drivers/md/md.c:2621
   md_add_new_disk+0xe3b/0x1850 drivers/md/md.c:7684
   md_ioctl+0x200a/0x2610 drivers/md/md.c:8499

bind_rdev_to_array() is not the only way in.  md_run() calls
sysfs_create_group(&mddev->kobj, &md_redundancy_group), and
internal_create_group() has its own WARN_ON(!kobj->sd), so RUN_ARRAY in
the same window warns too.  Guarding the individual callers would mean
finding all of them; the window itself is what should not be reachable.

Refuse the open until md_alloc() has added the kobject.  mddev->kobj.sd
is NULL until kobject_add() creates the directory, and stays set for the
rest of the mddev's life: commit ca39f7502425 ("md: fix mddev->kobj
lifetime") dropped the explicit kobject_del() and lets the final put do
the removal.  md.c already tests mddev->kobj.sd for "is this array
published yet" in mddev_unlock() and md_run().

This cannot deadlock against add_disk() itself: device_add_disk() only
opens the disk to scan partitions when get_capacity(disk) is non-zero,
and md_alloc() does not set the capacity - md_run() does, long after.

Fixes: ca39f7502425 ("md: fix mddev->kobj lifetime")
Reported-by: syzbot+95eeb4ada2349a2170ea@xxxxxxxxxxxxxxxxxxxxxxxxx
Closes: https://syzkaller.appspot.com/bug?extid=95eeb4ada2349a2170ea
Cc: stable@xxxxxxxxxxxxxxx
Assisted-by: LLM
---
 drivers/md/md.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/md/md.c b/drivers/md/md.c
index 680b34a63cb3..e3326b78aef7 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -8610,6 +8610,15 @@ static int md_open(struct gendisk *disk, blk_mode_t mode)
 	if (test_bit(MD_CLOSING, &mddev->flags))
 		goto out_unlock;
 
+	/*
+	 * md_alloc() publishes the gendisk with add_disk() before it has
+	 * initialised and added mddev->kobj, so an array opened in that window
+	 * would let ioctls reach a kobject that is still all zeroes.  Wait for
+	 * md_alloc() to finish rather than handing out the device early.
+	 */
+	if (!mddev->kobj.sd)
+		goto out_unlock;
+
 	atomic_inc(&mddev->openers);
 	mutex_unlock(&mddev->open_mutex);
 
-- 
2.55.0.windows.5