Re: [PATCH v2] Bluetooth: Shutdown controller after workqueues are flushed or cancelled

[Kai-Heng Feng <kai.heng.feng@canonical.com>]

Hi Mattijs,

On Fri, Aug 6, 2021 at 4:51 PM Mattijs Korpershoek
<mkorpershoek@baylibre.com> wrote:
>
> Hi Kai-Heng,
>
> Kai-Heng Feng <kai.heng.feng@canonical.com> writes:
>
> > Hi Mattijs,
> >
> > On Thu, Aug 5, 2021 at 2:55 PM Mattijs Korpershoek
> > <mkorpershoek@baylibre.com> wrote:
> >>
> >> Hi Kai-Heng,
> >>
> >> Thanks for your patch,
> >>
> >> Kai-Heng Feng <kai.heng.feng@canonical.com> writes:
> >>
> >
> > [snipped]
> >
> >> I confirm this diff works for me:
> >>
> >> root@i500-pumpkin:~# hciconfig hci0 up
> >> root@i500-pumpkin:~# hciconfig hci0 down
> >> root@i500-pumpkin:~# hciconfig hci0 up
> >> root@i500-pumpkin:~# hciconfig hci0
> >> hci0:   Type: Primary  Bus: SDIO
> >>         BD Address: 00:0C:E7:55:FF:12  ACL MTU: 1021:8  SCO MTU: 244:4
> >>         UP RUNNING
> >>         RX bytes:11268 acl:0 sco:0 events:829 errors:0
> >>         TX bytes:182569 acl:0 sco:0 commands:829 errors:0
> >>
> >> root@i500-pumpkin:~# hcitool scan
> >> Scanning ...
> >>         <redacted>       Pixel 3 XL
> >>
> >> Tested-by: Mattijs Korpershoek <mkorpershoek@baylibre.com>
> >
> > I found that btmtksdio_flush() only cancels the work instead of doing
> > flush_work(). That probably explains why putting ->shutdown right
> > before ->flush doesn't work.
> > So can you please test the following again:
> > diff --git a/drivers/bluetooth/btmtksdio.c b/drivers/bluetooth/btmtksdio.c
> > index 9872ef18f9fea..b33c05ad2150b 100644
> > --- a/drivers/bluetooth/btmtksdio.c
> > +++ b/drivers/bluetooth/btmtksdio.c
> > @@ -649,9 +649,9 @@ static int btmtksdio_flush(struct hci_dev *hdev)
> >  {
> >         struct btmtksdio_dev *bdev = hci_get_drvdata(hdev);
> >
> > -       skb_queue_purge(&bdev->txq);
> > +       flush_work(&bdev->tx_work);
> >
> > -       cancel_work_sync(&bdev->tx_work);
> > +       skb_queue_purge(&bdev->txq);
> >
> >         return 0;
> >  }
> > diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
> > index 2560ed2f144d4..a61e610a400cb 100644
> >
> > --- a/net/bluetooth/hci_core.c
> > +++ b/net/bluetooth/hci_core.c
> > @@ -1785,6 +1785,14 @@ int hci_dev_do_close(struct hci_dev *hdev)
> >         aosp_do_close(hdev);
> >         msft_do_close(hdev);
> >
> > +       if (!hci_dev_test_flag(hdev, HCI_UNREGISTER) &&
> > +           !hci_dev_test_flag(hdev, HCI_USER_CHANNEL) &&
> > +           test_bit(HCI_UP, &hdev->flags)) {
> > +               /* Execute vendor specific shutdown routine */
> > +               if (hdev->shutdown)
> > +                       hdev->shutdown(hdev);
> > +       }
> > +
> >         if (hdev->flush)
> >                 hdev->flush(hdev);
> >
> > @@ -1798,14 +1806,6 @@ int hci_dev_do_close(struct hci_dev *hdev)
> >                 clear_bit(HCI_INIT, &hdev->flags);
> >         }
> >
> > -       if (!hci_dev_test_flag(hdev, HCI_UNREGISTER) &&
> > -           !hci_dev_test_flag(hdev, HCI_USER_CHANNEL) &&
> > -           test_bit(HCI_UP, &hdev->flags)) {
> > -               /* Execute vendor specific shutdown routine */
> > -               if (hdev->shutdown)
> > -                       hdev->shutdown(hdev);
> > -       }
> > -
> >         /* flush cmd  work */
> >         flush_work(&hdev->cmd_work);
> I've tried this but I have the same (broken) symptoms as before.
>
> Here are some logs of v3:
> dmesg: https://pastebin.com/1x4UHkzy
> ftrace: https://pastebin.com/Lm1d6AWy

Thanks for your testing. I think I finally got it:
btmtksdio_shutdown()
-> mtk_hci_wmt_sync()
 -> __hci_cmd_send()
  then waiting for BTMTKSDIO_TX_WAIT_VND_EVT, which is cleared in
btmtksdio_recv_event():
btmtksdio_recv_event()
-> hci_recv_frame()
 -> queue_work(hdev->workqueue, &hdev->rx_work);

That means it has to be done before the following drain_workqueue() call.
Can you please see if moving the ->shutdown() part right before
drain_workqueue() can fix the issue?

Kai-Heng

>
> Mattijs
>
> >
> > Kai-Heng