2020-09-10 02:08:20

by Abhishek Pandit-Subedi

[permalink] [raw]
Subject: [PATCH] Bluetooth: Re-order clearing suspend tasks

Unregister_pm_notifier is a blocking call so suspend tasks should be
cleared beforehand. Otherwise, the notifier will wait for completion
before returning (and we encounter a 2s timeout on resume).

Fixes: 0e9952804ec9c8 (Bluetooth: Clear suspend tasks on unregister)
Signed-off-by: Abhishek Pandit-Subedi <[email protected]>
---
Should have caught that unregister_pm_notifier was blocking last time
but when testing the earlier patch, I got unlucky and saw that the error
message was never hit (the suspend timeout).

When re-testing this patch on the same device, I was able to reproduce
the problem on an older build with the 0e9952804ec9c8 but not on a newer
build with the same patch. Changing the order correctly fixes it
everywhere. Confirmed this by adding debug logs in btusb_disconnect and
hci_suspend_notifier to confirm what order things were getting called.

Sorry about the churn. Next I'm going try to do something about the palm
shaped indentation on my forehead...

net/bluetooth/hci_core.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
index efc0fe2b47dac2..be9cdf5dabe5dc 100644
--- a/net/bluetooth/hci_core.c
+++ b/net/bluetooth/hci_core.c
@@ -3794,8 +3794,8 @@ void hci_unregister_dev(struct hci_dev *hdev)

cancel_work_sync(&hdev->power_on);

- unregister_pm_notifier(&hdev->suspend_notifier);
hci_suspend_clear_tasks(hdev);
+ unregister_pm_notifier(&hdev->suspend_notifier);
cancel_work_sync(&hdev->suspend_prepare);

hci_dev_do_close(hdev);
--
2.28.0.618.gf4bc123cb7-goog