[PATCH] Battery: sysfs_remove_battery(): possible circular locking

[Sergey Senozhatsky <sergey.senozhatsky@gmail.com>]

Commit 9c921c22a7f33397a6774d7fa076db9b6a0fd669
Author: Lan Tianyu <tianyu.lan@intel.com>

    ACPI / Battery: Resolve the race condition in the sysfs_remove_battery()

introduced battery locking to sysfs_remove_battery(). That lead to a possible 
deadlock warning:

[14818.477170] =======================================================
[14818.477200] [ INFO: possible circular locking dependency detected ]
[14818.477221] 3.1.0-dbg-07865-g1280ea8-dirty #668
[14818.477236] -------------------------------------------------------
[14818.477257] s2ram/1599 is trying to acquire lock:
[14818.477276]  (s_active#8){++++.+}, at: [<ffffffff81169147>] sysfs_addrm_finish+0x31/0x5a
[14818.477323] 
[14818.477325] but task is already holding lock:
[14818.477350]  (&battery->lock){+.+.+.}, at: [<ffffffffa0047278>] sysfs_remove_battery+0x10/0x4b [battery]
[14818.477395] 
[14818.477397] which lock already depends on the new lock.
[...]
[14818.479121] stack backtrace:
[14818.479148] Pid: 1599, comm: s2ram Not tainted 3.1.0-dbg-07865-g1280ea8-dirty #668
[14818.479175] Call Trace:
[14818.479198]  [<ffffffff814828c3>] print_circular_bug+0x293/0x2a4
[14818.479228]  [<ffffffff81070cb5>] __lock_acquire+0xfe4/0x164b
[14818.479260]  [<ffffffff81169147>] ? sysfs_addrm_finish+0x31/0x5a
[14818.479288]  [<ffffffff810718d2>] lock_acquire+0x138/0x1ac
[14818.479316]  [<ffffffff81169147>] ? sysfs_addrm_finish+0x31/0x5a
[14818.479345]  [<ffffffff81168a79>] sysfs_deactivate+0x9b/0xec
[14818.479373]  [<ffffffff81169147>] ? sysfs_addrm_finish+0x31/0x5a
[14818.479405]  [<ffffffff81169147>] sysfs_addrm_finish+0x31/0x5a
[14818.479433]  [<ffffffff81167bc5>] sysfs_hash_and_remove+0x54/0x77
[14818.479461]  [<ffffffff811681b9>] sysfs_remove_file+0x12/0x14
[14818.479488]  [<ffffffff81385bf8>] device_remove_file+0x12/0x14
[14818.479516]  [<ffffffff81386504>] device_del+0x119/0x17c
[14818.479542]  [<ffffffff81386575>] device_unregister+0xe/0x1a
[14818.479570]  [<ffffffff813c6ef9>] power_supply_unregister+0x23/0x27
[14818.479601]  [<ffffffffa004729c>] sysfs_remove_battery+0x34/0x4b [battery]
[14818.479632]  [<ffffffffa004778f>] battery_notify+0x2c/0x3a [battery]
[14818.479662]  [<ffffffff8148fe82>] notifier_call_chain+0x74/0xa1
[14818.479692]  [<ffffffff810624b4>] __blocking_notifier_call_chain+0x6c/0x89
[14818.479722]  [<ffffffff810624e0>] blocking_notifier_call_chain+0xf/0x11
[14818.479751]  [<ffffffff8107e40e>] pm_notifier_call_chain+0x15/0x27
[14818.479770]  [<ffffffff8107ee1a>] enter_state+0xa7/0xd5
[14818.479782]  [<ffffffff8107e341>] state_store+0xaa/0xc0
[14818.479795]  [<ffffffff8107e297>] ? pm_async_store+0x45/0x45
[14818.479807]  [<ffffffff81248837>] kobj_attr_store+0x17/0x19
[14818.479820]  [<ffffffff81167e27>] sysfs_write_file+0x103/0x13f
[14818.479834]  [<ffffffff81109037>] vfs_write+0xad/0x13d
[14818.479847]  [<ffffffff811092b2>] sys_write+0x45/0x6c
[14818.479860]  [<ffffffff81492f92>] system_call_fastpath+0x16/0x1b


I've changed `the marker' from `battery->bat.dev' to `battery->bat.name', so 
the basic idea should remain the same, now we just can release battery->lock 
more quicker, before device_remove_file() call.

Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>

---

 drivers/acpi/battery.c |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/acpi/battery.c b/drivers/acpi/battery.c
index 87c0a8d..398cbfb 100644
--- a/drivers/acpi/battery.c
+++ b/drivers/acpi/battery.c
@@ -574,15 +574,17 @@ static int sysfs_add_battery(struct acpi_battery *battery)
 static void sysfs_remove_battery(struct acpi_battery *battery)
 {
 	mutex_lock(&battery->lock);
-	if (!battery->bat.dev) {
+	if (!battery->bat.name) {
 		mutex_unlock(&battery->lock);
 		return;
 	}
 
+	battery->bat.name = NULL;
+	mutex_unlock(&battery->lock);
+
 	device_remove_file(battery->bat.dev, &alarm_attr);
 	power_supply_unregister(&battery->bat);
 	battery->bat.dev = NULL;
-	mutex_unlock(&battery->lock);
 }
 
 /*