ATA: Fix port removal ordering
authorRafael J. Wysocki <rafael.j.wysocki@intel.com>
Mon, 25 Nov 2013 12:19:01 +0000 (13:19 +0100)
committerTejun Heo <tj@kernel.org>
Wed, 27 Nov 2013 18:55:16 +0000 (13:55 -0500)
commitc5700766975c69d27150256444db63fbfd103791
tree36b8e6a522905eb81285e18bb2d1a8b7439046a5
parent6d5278a68a75891db1df5ae1ecf83d288fc58c65
ATA: Fix port removal ordering

After commit bcdde7e221a8 (sysfs: make __sysfs_remove_dir() recursive)
Mika Westerberg sees traces analogous to the one below in Thunderbolt
hot-remove testing:

 WARNING: CPU: 0 PID: 4 at fs/sysfs/group.c:214 sysfs_remove_group+0xc6/0xd0()
 sysfs group ffffffff81c6f1e0 not found for kobject 'host7'
 Modules linked in:
 CPU: 0 PID: 4 Comm: kworker/0:0 Not tainted 3.12.0+ #13
 Hardware name:                  /D33217CK, BIOS GKPPT10H.86A.0042.2013.0422.1439 04/22/2013
 Workqueue: kacpi_hotplug acpi_hotplug_work_fn
  0000000000000009 ffff8801002459b0 ffffffff817daab1 ffff8801002459f8
  ffff8801002459e8 ffffffff810436b8 0000000000000000 ffffffff81c6f1e0
  ffff88006d440358 ffff88006d440188 ffff88006e8b4c28 ffff880100245a48
 Call Trace:
  [<ffffffff817daab1>] dump_stack+0x45/0x56
  [<ffffffff810436b8>] warn_slowpath_common+0x78/0xa0
  [<ffffffff81043727>] warn_slowpath_fmt+0x47/0x50
  [<ffffffff811ad319>] ? sysfs_get_dirent_ns+0x49/0x70
  [<ffffffff811ae526>] sysfs_remove_group+0xc6/0xd0
  [<ffffffff81432f7e>] dpm_sysfs_remove+0x3e/0x50
  [<ffffffff8142a0d0>] device_del+0x40/0x1b0
  [<ffffffff8142a24d>] device_unregister+0xd/0x20
  [<ffffffff8144131a>] scsi_remove_host+0xba/0x110
  [<ffffffff8145f526>] ata_host_detach+0xc6/0x100
  [<ffffffff8145f578>] ata_pci_remove_one+0x18/0x20
  [<ffffffff812e8f48>] pci_device_remove+0x28/0x60
  [<ffffffff8142d854>] __device_release_driver+0x64/0xd0
  [<ffffffff8142d8de>] device_release_driver+0x1e/0x30
  [<ffffffff8142d257>] bus_remove_device+0xf7/0x140
  [<ffffffff8142a1b1>] device_del+0x121/0x1b0
  [<ffffffff812e43d4>] pci_stop_bus_device+0x94/0xa0
  [<ffffffff812e437b>] pci_stop_bus_device+0x3b/0xa0
  [<ffffffff812e437b>] pci_stop_bus_device+0x3b/0xa0
  [<ffffffff812e44dd>] pci_stop_and_remove_bus_device+0xd/0x20
  [<ffffffff812fc743>] trim_stale_devices+0x73/0xe0
  [<ffffffff812fc78b>] trim_stale_devices+0xbb/0xe0
  [<ffffffff812fc78b>] trim_stale_devices+0xbb/0xe0
  [<ffffffff812fcb6e>] acpiphp_check_bridge+0x7e/0xd0
  [<ffffffff812fd90d>] hotplug_event+0xcd/0x160
  [<ffffffff812fd9c5>] hotplug_event_work+0x25/0x60
  [<ffffffff81316749>] acpi_hotplug_work_fn+0x17/0x22
  [<ffffffff8105cf3a>] process_one_work+0x17a/0x430
  [<ffffffff8105db29>] worker_thread+0x119/0x390
  [<ffffffff8105da10>] ? manage_workers.isra.25+0x2a0/0x2a0
  [<ffffffff81063a5d>] kthread+0xcd/0xf0
  [<ffffffff81063990>] ? kthread_create_on_node+0x180/0x180
  [<ffffffff817eb33c>] ret_from_fork+0x7c/0xb0
  [<ffffffff81063990>] ? kthread_create_on_node+0x180/0x180

The source of this problem is that SCSI hosts are removed from
ATA ports after calling ata_tport_delete() which removes the
port's sysfs directory, among other things.  Now, after commit
bcdde7e221a8, the sysfs directory is removed along with all of
its subdirectories that include the SCSI host's sysfs directory
and its subdirectories at this point.  Consequently, when
device_del() is finally called for any child device of the SCSI
host and tries to remove its "power" group (which is already
gone then), it triggers the above warning.

To make the warnings go away, change the removal ordering in
ata_port_detach() so that the SCSI host is removed from the
port before ata_tport_delete() is called.

References: https://bugzilla.kernel.org/show_bug.cgi?id=65281
Reported-and-tested-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
drivers/ata/libata-core.c