dm: wait until embedded kobject is released before destroying a device
authorMikulas Patocka <mpatocka@redhat.com>
Tue, 7 Jan 2014 04:01:22 +0000 (23:01 -0500)
committerBen Hutchings <ben@decadent.org.uk>
Tue, 1 Apr 2014 23:58:41 +0000 (00:58 +0100)
commit1765b496a0a8b2d8d9c840fa748986f0ca93198f
tree8ef4d60a63b7df3f632fdd0b2e39293782ee88c3
parent220ff586da578215e1cef7f2d89ef35b65965ce5
dm: wait until embedded kobject is released before destroying a device

commit be35f486108227e10fe5d96fd42fb2b344c59983 upstream.

There may be other parts of the kernel holding a reference on the dm
kobject.  We must wait until all references are dropped before
deallocating the mapped_device structure.

The dm_kobject_release method signals that all references are dropped
via completion.  But dm_kobject_release doesn't free the kobject (which
is embedded in the mapped_device structure).

This is the sequence of operations:
* when destroying a DM device, call kobject_put from dm_sysfs_exit
* wait until all users stop using the kobject, when it happens the
  release method is called
* the release method signals the completion and should return without
  delay
* the dm device removal code that waits on the completion continues
* the dm device removal code drops the dm_mod reference the device had
* the dm device removal code frees the mapped_device structure that
  contains the kobject

Using kobject this way should avoid the module unload race that was
mentioned at the beginning of this thread:
https://lkml.org/lkml/2014/1/4/83

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
drivers/md/dm-sysfs.c
drivers/md/dm.c
drivers/md/dm.h