| Message ID | 20200624113351.1.Ic6b1fca2b1b3fe989db21ceae76bab80bd87d387@changeid (mailing list archive) |
|---|---|
| State | Accepted |
| Headers | show |
| Series | Bluetooth: Don't restart scanning if paused | expand |
Hi Abhishek, > When restarting LE scanning, check if it's currently paused before > enabling passive scanning. > > Signed-off-by: Abhishek Pandit-Subedi <abhishekpandit@chromium.org> > --- > When running suspend stress tests on Chromebooks, we discovered > instances where the Chromebook didn't enter the deepest idle states > (i.e. S0ix). After some debugging, we found that passive scanning was > being enabled AFTER the suspend notifier had run (and disabled all > scanning). > > For this fix, I simply looked at all the places where we call > HCI_OP_LE_SET_SCAN_ENABLE and added a guard clause for suspend. With > this fix, we were able to get through 100+ iterations of the suspend > stress test without any problems entering S0ix. > > > net/bluetooth/hci_request.c | 10 ++++++++++ > 1 file changed, 10 insertions(+) patch has been applied to bluetooth-next tree. Regards Marcel
diff --git a/net/bluetooth/hci_request.c b/net/bluetooth/hci_request.c index 86ae4b953a011e..116207009dde01 100644 --- a/net/bluetooth/hci_request.c +++ b/net/bluetooth/hci_request.c @@ -819,6 +819,11 @@ static void hci_req_start_scan(struct hci_request *req, u8 type, u16 interval, { struct hci_dev *hdev = req->hdev; + if (hdev->scanning_paused) { + bt_dev_dbg(hdev, "Scanning is paused for suspend"); + return; + } + /* Use ext scanning if set ext scan param and ext scan enable is * supported */ @@ -2657,6 +2662,11 @@ static int le_scan_restart(struct hci_request *req, unsigned long opt) if (!hci_dev_test_flag(hdev, HCI_LE_SCAN)) return 0; + if (hdev->scanning_paused) { + bt_dev_dbg(hdev, "Scanning is paused for suspend"); + return 0; + } + hci_req_add_le_scan_disable(req); if (use_ext_scan(hdev)) {
When restarting LE scanning, check if it's currently paused before enabling passive scanning. Signed-off-by: Abhishek Pandit-Subedi <abhishekpandit@chromium.org> --- When running suspend stress tests on Chromebooks, we discovered instances where the Chromebook didn't enter the deepest idle states (i.e. S0ix). After some debugging, we found that passive scanning was being enabled AFTER the suspend notifier had run (and disabled all scanning). For this fix, I simply looked at all the places where we call HCI_OP_LE_SET_SCAN_ENABLE and added a guard clause for suspend. With this fix, we were able to get through 100+ iterations of the suspend stress test without any problems entering S0ix. net/bluetooth/hci_request.c | 10 ++++++++++ 1 file changed, 10 insertions(+)