thermal: Fix deadlock in thermal thermal_zone_device_check

commit 163b00cde7cf2206e248789d2780121ad5e6a70b upstream.

1851799e1d29 ("thermal: Fix use-after-free when unregistering thermal zone
device") changed cancel_delayed_work to cancel_delayed_work_sync to avoid
a use-after-free issue. However, cancel_delayed_work_sync could be called
insides the WQ causing deadlock.

[54109.642398] c0   1162 kworker/u17:1   D    0 11030      2 0x00000000
[54109.642437] c0   1162 Workqueue: thermal_passive_wq thermal_zone_device_check
[54109.642447] c0   1162 Call trace:
[54109.642456] c0   1162  __switch_to+0x138/0x158
[54109.642467] c0   1162  __schedule+0xba4/0x1434
[54109.642480] c0   1162  schedule_timeout+0xa0/0xb28
[54109.642492] c0   1162  wait_for_common+0x138/0x2e8
[54109.642511] c0   1162  flush_work+0x348/0x40c
[54109.642522] c0   1162  __cancel_work_timer+0x180/0x218
[54109.642544] c0   1162  handle_thermal_trip+0x2c4/0x5a4
[54109.642553] c0   1162  thermal_zone_device_update+0x1b4/0x25c
[54109.642563] c0   1162  thermal_zone_device_check+0x18/0x24
[54109.642574] c0   1162  process_one_work+0x3cc/0x69c
[54109.642583] c0   1162  worker_thread+0x49c/0x7c0
[54109.642593] c0   1162  kthread+0x17c/0x1b0
[54109.642602] c0   1162  ret_from_fork+0x10/0x18
[54109.643051] c0   1162 kworker/u17:2   D    0 16245      2 0x00000000
[54109.643067] c0   1162 Workqueue: thermal_passive_wq thermal_zone_device_check
[54109.643077] c0   1162 Call trace:
[54109.643085] c0   1162  __switch_to+0x138/0x158
[54109.643095] c0   1162  __schedule+0xba4/0x1434
[54109.643104] c0   1162  schedule_timeout+0xa0/0xb28
[54109.643114] c0   1162  wait_for_common+0x138/0x2e8
[54109.643122] c0   1162  flush_work+0x348/0x40c
[54109.643131] c0   1162  __cancel_work_timer+0x180/0x218
[54109.643141] c0   1162  handle_thermal_trip+0x2c4/0x5a4
[54109.643150] c0   1162  thermal_zone_device_update+0x1b4/0x25c
[54109.643159] c0   1162  thermal_zone_device_check+0x18/0x24
[54109.643167] c0   1162  process_one_work+0x3cc/0x69c
[54109.643177] c0   1162  worker_thread+0x49c/0x7c0
[54109.643186] c0   1162  kthread+0x17c/0x1b0
[54109.643195] c0   1162  ret_from_fork+0x10/0x18
[54109.644500] c0   1162 cat             D    0  7766      1 0x00000001
[54109.644515] c0   1162 Call trace:
[54109.644524] c0   1162  __switch_to+0x138/0x158
[54109.644536] c0   1162  __schedule+0xba4/0x1434
[54109.644546] c0   1162  schedule_preempt_disabled+0x80/0xb0
[54109.644555] c0   1162  __mutex_lock+0x3a8/0x7f0
[54109.644563] c0   1162  __mutex_lock_slowpath+0x14/0x20
[54109.644575] c0   1162  thermal_zone_get_temp+0x84/0x360
[54109.644586] c0   1162  temp_show+0x30/0x78
[54109.644609] c0   1162  dev_attr_show+0x5c/0xf0
[54109.644628] c0   1162  sysfs_kf_seq_show+0xcc/0x1a4
[54109.644636] c0   1162  kernfs_seq_show+0x48/0x88
[54109.644656] c0   1162  seq_read+0x1f4/0x73c
[54109.644664] c0   1162  kernfs_fop_read+0x84/0x318
[54109.644683] c0   1162  __vfs_read+0x50/0x1bc
[54109.644692] c0   1162  vfs_read+0xa4/0x140
[54109.644701] c0   1162  SyS_read+0xbc/0x144
[54109.644708] c0   1162  el0_svc_naked+0x34/0x38
[54109.845800] c0   1162 D 720.000s 1->7766->7766 cat [panic]

Fixes: 1851799e1d29 ("thermal: Fix use-after-free when unregistering thermal zone device")
Cc: stable@vger.kernel.org
Signed-off-by: Wei Wang <wvw@google.com>
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/drivers/thermal/thermal_core.c b/drivers/thermal/thermal_core.c
index 57603be42c50772f49c9d57775a348b56f5088d3..90c033b4ec98f2a1baae72c76015870d6e5e3d0a 100644 (file)
--- a/drivers/thermal/thermal_core.c
+++ b/drivers/thermal/thermal_core.c
@@ -402,7 +402,7 @@ static void thermal_zone_device_set_polling(struct thermal_zone_device *tz,
                mod_delayed_work(system_freezable_wq, &tz->poll_queue,
                                 msecs_to_jiffies(delay));
        else
-               cancel_delayed_work_sync(&tz->poll_queue);
+               cancel_delayed_work(&tz->poll_queue);
 }
 
 static void monitor_thermal_zone(struct thermal_zone_device *tz)
@@ -2073,7 +2073,7 @@ void thermal_zone_device_unregister(struct thermal_zone_device *tz)
 
        mutex_unlock(&thermal_list_lock);
 
-       thermal_zone_device_set_polling(tz, 0);
+       cancel_delayed_work_sync(&tz->poll_queue);
 
        if (tz->type[0])
                device_remove_file(&tz->device, &dev_attr_type);