Netdev Archive on lore.kernel.org
help / color / mirror / Atom feed
* [PATCH] Bluetooth: Clear suspend tasks on unregister
@ 2020-08-26 22:47 Abhishek Pandit-Subedi
2020-08-31 15:49 ` Marcel Holtmann
0 siblings, 1 reply; 5+ messages in thread
From: Abhishek Pandit-Subedi @ 2020-08-26 22:47 UTC (permalink / raw)
To: marcel, linux-bluetooth
Cc: chromeos-bluetooth-upstreaming, Abhishek Pandit-Subedi,
David S. Miller, Johan Hedberg, netdev, linux-kernel,
Jakub Kicinski, Miao-chen Chou
While unregistering, make sure to clear the suspend tasks before
cancelling the work. If the unregister is called during resume from
suspend, this will unnecessarily add 2s to the resume time otherwise.
Fixes: 4e8c36c3b0d73d (Bluetooth: Fix suspend notifier race)
Signed-off-by: Abhishek Pandit-Subedi <abhishekpandit@chromium.org>
---
This was discovered with RT8822CE using the btusb driver. This chipset
will reset on resume during system suspend and was unnecessarily adding
2s to every resume. Since we're unregistering anyway, there's no harm in
just clearing the pending events.
net/bluetooth/hci_core.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
index 68bfe57b66250f..ed4cb3479433c0 100644
--- a/net/bluetooth/hci_core.c
+++ b/net/bluetooth/hci_core.c
@@ -3442,6 +3442,16 @@ void hci_copy_identity_address(struct hci_dev *hdev, bdaddr_t *bdaddr,
}
}
+static void hci_suspend_clear_tasks(struct hci_dev *hdev)
+{
+ int i;
+
+ for (i = 0; i < __SUSPEND_NUM_TASKS; ++i)
+ clear_bit(i, hdev->suspend_tasks);
+
+ wake_up(&hdev->suspend_wait_q);
+}
+
static int hci_suspend_wait_event(struct hci_dev *hdev)
{
#define WAKE_COND \
@@ -3785,6 +3795,7 @@ 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);
cancel_work_sync(&hdev->suspend_prepare);
hci_dev_do_close(hdev);
--
2.28.0.297.g1956fa8f8d-goog
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] Bluetooth: Clear suspend tasks on unregister
2020-08-26 22:47 [PATCH] Bluetooth: Clear suspend tasks on unregister Abhishek Pandit-Subedi
@ 2020-08-31 15:49 ` Marcel Holtmann
2020-08-31 16:51 ` Abhishek Pandit-Subedi
0 siblings, 1 reply; 5+ messages in thread
From: Marcel Holtmann @ 2020-08-31 15:49 UTC (permalink / raw)
To: Abhishek Pandit-Subedi
Cc: linux-bluetooth, CrosBT Upstreaming, David S. Miller,
Johan Hedberg, netdev, kernel list, Jakub Kicinski,
Miao-chen Chou
Hi Abhishek,
> While unregistering, make sure to clear the suspend tasks before
> cancelling the work. If the unregister is called during resume from
> suspend, this will unnecessarily add 2s to the resume time otherwise.
>
> Fixes: 4e8c36c3b0d73d (Bluetooth: Fix suspend notifier race)
> Signed-off-by: Abhishek Pandit-Subedi <abhishekpandit@chromium.org>
> ---
> This was discovered with RT8822CE using the btusb driver. This chipset
> will reset on resume during system suspend and was unnecessarily adding
> 2s to every resume. Since we're unregistering anyway, there's no harm in
> just clearing the pending events.
>
> net/bluetooth/hci_core.c | 11 +++++++++++
> 1 file changed, 11 insertions(+)
>
> diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
> index 68bfe57b66250f..ed4cb3479433c0 100644
> --- a/net/bluetooth/hci_core.c
> +++ b/net/bluetooth/hci_core.c
> @@ -3442,6 +3442,16 @@ void hci_copy_identity_address(struct hci_dev *hdev, bdaddr_t *bdaddr,
> }
> }
>
> +static void hci_suspend_clear_tasks(struct hci_dev *hdev)
> +{
> + int i;
> +
> + for (i = 0; i < __SUSPEND_NUM_TASKS; ++i)
> + clear_bit(i, hdev->suspend_tasks);
I prefer i++ instead of ++i.
Regards
Marcel
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] Bluetooth: Clear suspend tasks on unregister
2020-08-31 15:49 ` Marcel Holtmann
@ 2020-08-31 16:51 ` Abhishek Pandit-Subedi
0 siblings, 0 replies; 5+ messages in thread
From: Abhishek Pandit-Subedi @ 2020-08-31 16:51 UTC (permalink / raw)
To: Marcel Holtmann
Cc: linux-bluetooth, CrosBT Upstreaming, David S. Miller,
Johan Hedberg, netdev, kernel list, Jakub Kicinski,
Miao-chen Chou
v2 sent with fix.
On Mon, Aug 31, 2020 at 8:49 AM Marcel Holtmann <marcel@holtmann.org> wrote:
>
> Hi Abhishek,
>
> > While unregistering, make sure to clear the suspend tasks before
> > cancelling the work. If the unregister is called during resume from
> > suspend, this will unnecessarily add 2s to the resume time otherwise.
> >
> > Fixes: 4e8c36c3b0d73d (Bluetooth: Fix suspend notifier race)
> > Signed-off-by: Abhishek Pandit-Subedi <abhishekpandit@chromium.org>
> > ---
> > This was discovered with RT8822CE using the btusb driver. This chipset
> > will reset on resume during system suspend and was unnecessarily adding
> > 2s to every resume. Since we're unregistering anyway, there's no harm in
> > just clearing the pending events.
> >
> > net/bluetooth/hci_core.c | 11 +++++++++++
> > 1 file changed, 11 insertions(+)
> >
> > diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
> > index 68bfe57b66250f..ed4cb3479433c0 100644
> > --- a/net/bluetooth/hci_core.c
> > +++ b/net/bluetooth/hci_core.c
> > @@ -3442,6 +3442,16 @@ void hci_copy_identity_address(struct hci_dev *hdev, bdaddr_t *bdaddr,
> > }
> > }
> >
> > +static void hci_suspend_clear_tasks(struct hci_dev *hdev)
> > +{
> > + int i;
> > +
> > + for (i = 0; i < __SUSPEND_NUM_TASKS; ++i)
> > + clear_bit(i, hdev->suspend_tasks);
>
> I prefer i++ instead of ++i.
>
> Regards
>
> Marcel
>
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH] Bluetooth: Clear suspend tasks on unregister
@ 2020-08-26 22:26 Abhishek Pandit-Subedi
2020-08-26 22:48 ` Abhishek Pandit-Subedi
0 siblings, 1 reply; 5+ messages in thread
From: Abhishek Pandit-Subedi @ 2020-08-26 22:26 UTC (permalink / raw)
To: marcel, linux-bluetooth
Cc: chromeos-bluetooth-upstreaming, Abhishek Pandit-Subedi,
David S. Miller, Johan Hedberg, netdev, linux-kernel,
Jakub Kicinski
While unregistering, make sure to clear the suspend tasks before
cancelling the work. If the unregister is called during resume from
suspend, this will unnecessarily add 2s to the resume time otherwise.
Signed-off-by: Abhishek Pandit-Subedi <abhishekpandit@chromium.org>
---
This was discovered with RT8822CE using the btusb driver. This chipset
will reset on resume during system suspend and was unnecessarily adding
2s to every resume. Since we're unregistering anyway, there's no harm in
just clearing the pending events.
net/bluetooth/hci_core.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
index 68bfe57b66250f..ed4cb3479433c0 100644
--- a/net/bluetooth/hci_core.c
+++ b/net/bluetooth/hci_core.c
@@ -3442,6 +3442,16 @@ void hci_copy_identity_address(struct hci_dev *hdev, bdaddr_t *bdaddr,
}
}
+static void hci_suspend_clear_tasks(struct hci_dev *hdev)
+{
+ int i;
+
+ for (i = 0; i < __SUSPEND_NUM_TASKS; ++i)
+ clear_bit(i, hdev->suspend_tasks);
+
+ wake_up(&hdev->suspend_wait_q);
+}
+
static int hci_suspend_wait_event(struct hci_dev *hdev)
{
#define WAKE_COND \
@@ -3785,6 +3795,7 @@ 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);
cancel_work_sync(&hdev->suspend_prepare);
hci_dev_do_close(hdev);
--
2.28.0.297.g1956fa8f8d-goog
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [PATCH] Bluetooth: Clear suspend tasks on unregister
2020-08-26 22:26 Abhishek Pandit-Subedi
@ 2020-08-26 22:48 ` Abhishek Pandit-Subedi
0 siblings, 0 replies; 5+ messages in thread
From: Abhishek Pandit-Subedi @ 2020-08-26 22:48 UTC (permalink / raw)
To: Marcel Holtmann, Bluez mailing list
Cc: ChromeOS Bluetooth Upstreaming, David S. Miller, Johan Hedberg,
netdev, LKML, Jakub Kicinski
Please disregard this earlier email without the Fixes tag.
On Wed, Aug 26, 2020 at 3:26 PM Abhishek Pandit-Subedi
<abhishekpandit@chromium.org> wrote:
>
> While unregistering, make sure to clear the suspend tasks before
> cancelling the work. If the unregister is called during resume from
> suspend, this will unnecessarily add 2s to the resume time otherwise.
>
> Signed-off-by: Abhishek Pandit-Subedi <abhishekpandit@chromium.org>
> ---
> This was discovered with RT8822CE using the btusb driver. This chipset
> will reset on resume during system suspend and was unnecessarily adding
> 2s to every resume. Since we're unregistering anyway, there's no harm in
> just clearing the pending events.
>
> net/bluetooth/hci_core.c | 11 +++++++++++
> 1 file changed, 11 insertions(+)
>
> diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
> index 68bfe57b66250f..ed4cb3479433c0 100644
> --- a/net/bluetooth/hci_core.c
> +++ b/net/bluetooth/hci_core.c
> @@ -3442,6 +3442,16 @@ void hci_copy_identity_address(struct hci_dev *hdev, bdaddr_t *bdaddr,
> }
> }
>
> +static void hci_suspend_clear_tasks(struct hci_dev *hdev)
> +{
> + int i;
> +
> + for (i = 0; i < __SUSPEND_NUM_TASKS; ++i)
> + clear_bit(i, hdev->suspend_tasks);
> +
> + wake_up(&hdev->suspend_wait_q);
> +}
> +
> static int hci_suspend_wait_event(struct hci_dev *hdev)
> {
> #define WAKE_COND \
> @@ -3785,6 +3795,7 @@ 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);
> cancel_work_sync(&hdev->suspend_prepare);
>
> hci_dev_do_close(hdev);
> --
> 2.28.0.297.g1956fa8f8d-goog
>
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2020-08-31 16:52 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-08-26 22:47 [PATCH] Bluetooth: Clear suspend tasks on unregister Abhishek Pandit-Subedi
2020-08-31 15:49 ` Marcel Holtmann
2020-08-31 16:51 ` Abhishek Pandit-Subedi
-- strict thread matches above, loose matches on Subject: below --
2020-08-26 22:26 Abhishek Pandit-Subedi
2020-08-26 22:48 ` Abhishek Pandit-Subedi
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).