[PATCH] wifi: mt76: mt792x: fix scheduler interference in drv own process

[Sean Wang]

Hi, Mingyen

On Tue, May 14, 2024 at 6:52 PM Mingyen Hsieh
<mingyen.hsieh at mediatek.com> wrote:
>
> From: Michael Lo <michael.lo at mediatek.com>
>
> Add some time to wait for LP engine to complete its operation
> before polling pmctrl register.
>
> Signed-off-by: Michael Lo <michael.lo at mediatek.com>
> Signed-off-by: Ming Yen Hsieh <mingyen.hsieh at mediatek.com>
> ---
>  drivers/net/wireless/mediatek/mt76/mt7921/pci.c  | 3 +++
>  drivers/net/wireless/mediatek/mt76/mt7925/pci.c  | 3 +++
>  drivers/net/wireless/mediatek/mt76/mt792x.h      | 1 +
>  drivers/net/wireless/mediatek/mt76/mt792x_core.c | 6 ++++++
>  4 files changed, 13 insertions(+)
>
> diff --git a/drivers/net/wireless/mediatek/mt76/mt7921/pci.c b/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
> index f768e9389ac6..0f7d5a92f3d0 100644
> --- a/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
> +++ b/drivers/net/wireless/mediatek/mt76/mt7921/pci.c
> @@ -339,6 +339,9 @@ static int mt7921_pci_probe(struct pci_dev *pdev,
>         bus_ops->rmw = mt7921_rmw;
>         dev->mt76.bus = bus_ops;
>
> +       if (mt7921_disable_aspm)
> +               dev->disable_aspm = true;
> +
>         ret = mt792xe_mcu_fw_pmctrl(dev);
>         if (ret)
>                 goto err_free_dev;
> diff --git a/drivers/net/wireless/mediatek/mt76/mt7925/pci.c b/drivers/net/wireless/mediatek/mt76/mt7925/pci.c
> index 07b74d492ce1..e7039e312538 100644
> --- a/drivers/net/wireless/mediatek/mt76/mt7925/pci.c
> +++ b/drivers/net/wireless/mediatek/mt76/mt7925/pci.c
> @@ -373,6 +373,9 @@ static int mt7925_pci_probe(struct pci_dev *pdev,
>         bus_ops->rmw = mt7925_rmw;
>         dev->mt76.bus = bus_ops;
>
> +       if (mt7925_disable_aspm)
> +               dev->disable_aspm = true;
> +
>         ret = __mt792x_mcu_fw_pmctrl(dev);
>         if (ret)
>                 goto err_free_dev;
> diff --git a/drivers/net/wireless/mediatek/mt76/mt792x.h b/drivers/net/wireless/mediatek/mt76/mt792x.h
> index 20578497a405..8f913336a668 100644
> --- a/drivers/net/wireless/mediatek/mt76/mt792x.h
> +++ b/drivers/net/wireless/mediatek/mt76/mt792x.h
> @@ -190,6 +190,7 @@ struct mt792x_dev {
>         bool fw_assert:1;
>         bool has_eht:1;
>         bool regd_in_progress:1;
> +       bool disable_aspm:1;
>         wait_queue_head_t wait;
>
>         struct work_struct init_work;
> diff --git a/drivers/net/wireless/mediatek/mt76/mt792x_core.c b/drivers/net/wireless/mediatek/mt76/mt792x_core.c
> index a405af8d9052..a11e5438a36e 100644
> --- a/drivers/net/wireless/mediatek/mt76/mt792x_core.c
> +++ b/drivers/net/wireless/mediatek/mt76/mt792x_core.c
> @@ -766,6 +766,12 @@ int __mt792xe_mcu_drv_pmctrl(struct mt792x_dev *dev)
>
>         for (i = 0; i < MT792x_DRV_OWN_RETRY_COUNT; i++) {
>                 mt76_wr(dev, MT_CONN_ON_LPCTL, PCIE_LPCR_HOST_CLR_OWN);
> +
> +               if (!dev->disable_aspm) {
> +                       /* wait for LP engine to complete its operation */
> +                       usleep_range(2000, 3000);
> +               }
> +

I think we can reduce the extra delay on platforms that don't support
ASPM. Basically, if the platform doesn't support ASPM, we can get rid
of the wait.

>                 if (mt76_poll_msec_tick(dev, MT_CONN_ON_LPCTL,
>                                         PCIE_LPCR_HOST_OWN_SYNC, 0, 50, 1))
>                         break;
> --
> 2.25.1
>
>