2021-12-07 02:35:13

by Sean Wang

[permalink] [raw]
Subject: [PATCH] mt76: mt7921: fix possible resume failure

From: Sean Wang <[email protected]>

Fix the possible resume failure due to mt76_connac_mcu_set_hif_suspend
timeout.

That is because clearing the flag pm->suspended too early opened up a race
window, where mt7921_poll_tx/rx scheduled a ps_work to put the device in
doze mode, that is unexpected for the device is being resumed from the
suspend state and would make the remaining MCU comamnds in resume handler
failed to execute.

Fixes: ffa1bf97425b ("mt76: mt7921: introduce PM support")
Co-developed-by: YN Chen <[email protected]>
Signed-off-by: YN Chen <[email protected]>
Signed-off-by: Sean Wang <[email protected]>
---
drivers/net/wireless/mediatek/mt76/mt7921/pci.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/net/wireless/mediatek/mt76/mt7921/pci.c b/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
index 5635de3c80b1..21a48f02ab60 100644
--- a/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
+++ b/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
@@ -315,7 +315,6 @@ static int mt7921_pci_resume(struct pci_dev *pdev)
struct mt76_connac_pm *pm = &dev->pm;
int i, err;

- pm->suspended = false;
err = pci_set_power_state(pdev, PCI_D0);
if (err)
return err;
@@ -353,7 +352,11 @@ static int mt7921_pci_resume(struct pci_dev *pdev)
if (!pm->ds_enable)
mt76_connac_mcu_set_deep_sleep(&dev->mt76, false);

- return mt76_connac_mcu_set_hif_suspend(mdev, false);
+ err = mt76_connac_mcu_set_hif_suspend(mdev, false);
+
+ pm->suspended = false;
+
+ return err;
}
#endif /* CONFIG_PM */

--
2.25.1



2021-12-07 10:08:59

by Lorenzo Bianconi

[permalink] [raw]
Subject: Re: [PATCH] mt76: mt7921: fix possible resume failure

> From: Sean Wang <[email protected]>
>
> Fix the possible resume failure due to mt76_connac_mcu_set_hif_suspend
> timeout.
>
> That is because clearing the flag pm->suspended too early opened up a race
> window, where mt7921_poll_tx/rx scheduled a ps_work to put the device in
> doze mode, that is unexpected for the device is being resumed from the
> suspend state and would make the remaining MCU comamnds in resume handler
> failed to execute.

do we have a similar issue in mt7921s_resume()?

>
> Fixes: ffa1bf97425b ("mt76: mt7921: introduce PM support")
> Co-developed-by: YN Chen <[email protected]>
> Signed-off-by: YN Chen <[email protected]>
> Signed-off-by: Sean Wang <[email protected]>
> ---
> drivers/net/wireless/mediatek/mt76/mt7921/pci.c | 7 +++++--
> 1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/wireless/mediatek/mt76/mt7921/pci.c b/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
> index 5635de3c80b1..21a48f02ab60 100644
> --- a/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
> +++ b/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
> @@ -315,7 +315,6 @@ static int mt7921_pci_resume(struct pci_dev *pdev)
> struct mt76_connac_pm *pm = &dev->pm;
> int i, err;
>
> - pm->suspended = false;
> err = pci_set_power_state(pdev, PCI_D0);
> if (err)
> return err;
> @@ -353,7 +352,11 @@ static int mt7921_pci_resume(struct pci_dev *pdev)
> if (!pm->ds_enable)
> mt76_connac_mcu_set_deep_sleep(&dev->mt76, false);
>
> - return mt76_connac_mcu_set_hif_suspend(mdev, false);
> + err = mt76_connac_mcu_set_hif_suspend(mdev, false);
> +
> + pm->suspended = false;
> +
> + return err;
> }
> #endif /* CONFIG_PM */
>
> --
> 2.25.1
>


Attachments:
(No filename) (1.68 kB)
signature.asc (228.00 B)
Download all attachments

2021-12-07 17:22:15

by Sean Wang

[permalink] [raw]
Subject: Re: [PATCH] mt76: mt7921: fix possible resume failure

From: Sean Wang <[email protected]>

>> From: Sean Wang <[email protected]>
>>
>> Fix the possible resume failure due to mt76_connac_mcu_set_hif_suspend
>> timeout.
>>
>> That is because clearing the flag pm->suspended too early opened up a
>> race window, where mt7921_poll_tx/rx scheduled a ps_work to put the
>> device in doze mode, that is unexpected for the device is being
>> resumed from the suspend state and would make the remaining MCU
>> comamnds in resume handler failed to execute.
>
>do we have a similar issue in mt7921s_resume()?

We don't have the same issue found in mt7921s_resume because the driver
guarantees all MCU commands (even in mt7921s_resume) always work in the active
state.

But we can have the same patch for mt7921s just to make code consistent with
mt7921e.

>
>>
>> Fixes: ffa1bf97425b ("mt76: mt7921: introduce PM support")
>> Co-developed-by: YN Chen <[email protected]>
>> Signed-off-by: YN Chen <[email protected]>
>> Signed-off-by: Sean Wang <[email protected]>
>> ---
>> drivers/net/wireless/mediatek/mt76/mt7921/pci.c | 7 +++++--
>> 1 file changed, 5 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
>> b/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
>> index 5635de3c80b1..21a48f02ab60 100644
>> --- a/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
>> +++ b/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
>> @@ -315,7 +315,6 @@ static int mt7921_pci_resume(struct pci_dev *pdev)
>> struct mt76_connac_pm *pm = &dev->pm;
>> int i, err;
>>
>> - pm->suspended = false;
>> err = pci_set_power_state(pdev, PCI_D0);
>> if (err)
>> return err;
>> @@ -353,7 +352,11 @@ static int mt7921_pci_resume(struct pci_dev *pdev)
>> if (!pm->ds_enable)
>> mt76_connac_mcu_set_deep_sleep(&dev->mt76, false);
>>
>> - return mt76_connac_mcu_set_hif_suspend(mdev, false);
>> + err = mt76_connac_mcu_set_hif_suspend(mdev, false);
>> +
>> + pm->suspended = false;
>> +
>> + return err;
>> }
>> #endif /* CONFIG_PM */
>>
>> --
>> 2.25.1
>>
>

2021-12-07 20:08:14

by Lorenzo Bianconi

[permalink] [raw]
Subject: Re: [PATCH] mt76: mt7921: fix possible resume failure

> From: Sean Wang <[email protected]>
>
> >> From: Sean Wang <[email protected]>
> >>
> >> Fix the possible resume failure due to mt76_connac_mcu_set_hif_suspend
> >> timeout.
> >>
> >> That is because clearing the flag pm->suspended too early opened up a
> >> race window, where mt7921_poll_tx/rx scheduled a ps_work to put the
> >> device in doze mode, that is unexpected for the device is being
> >> resumed from the suspend state and would make the remaining MCU
> >> comamnds in resume handler failed to execute.
> >
> >do we have a similar issue in mt7921s_resume()?
>
> We don't have the same issue found in mt7921s_resume because the driver
> guarantees all MCU commands (even in mt7921s_resume) always work in the active
> state.
>
> But we can have the same patch for mt7921s just to make code consistent with
> mt7921e.

ack, agree.

Regards,
Lorenzo

>
> >
> >>
> >> Fixes: ffa1bf97425b ("mt76: mt7921: introduce PM support")
> >> Co-developed-by: YN Chen <[email protected]>
> >> Signed-off-by: YN Chen <[email protected]>
> >> Signed-off-by: Sean Wang <[email protected]>
> >> ---
> >> drivers/net/wireless/mediatek/mt76/mt7921/pci.c | 7 +++++--
> >> 1 file changed, 5 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
> >> b/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
> >> index 5635de3c80b1..21a48f02ab60 100644
> >> --- a/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
> >> +++ b/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
> >> @@ -315,7 +315,6 @@ static int mt7921_pci_resume(struct pci_dev *pdev)
> >> struct mt76_connac_pm *pm = &dev->pm;
> >> int i, err;
> >>
> >> - pm->suspended = false;
> >> err = pci_set_power_state(pdev, PCI_D0);
> >> if (err)
> >> return err;
> >> @@ -353,7 +352,11 @@ static int mt7921_pci_resume(struct pci_dev *pdev)
> >> if (!pm->ds_enable)
> >> mt76_connac_mcu_set_deep_sleep(&dev->mt76, false);
> >>
> >> - return mt76_connac_mcu_set_hif_suspend(mdev, false);
> >> + err = mt76_connac_mcu_set_hif_suspend(mdev, false);
> >> +
> >> + pm->suspended = false;
> >> +
> >> + return err;
> >> }
> >> #endif /* CONFIG_PM */
> >>
> >> --
> >> 2.25.1
> >>
> >
>


Attachments:
(No filename) (2.15 kB)
signature.asc (228.00 B)
Download all attachments