On Wed, Feb 22, 2023 at 12:18:36PM +0100, Lorenzo Bianconi wrote:
> Thx for working on this. I would say it is a more general issue since we have
> dbdc support in mt7915/mt7996/mt7915 devices. I think we can move the check in
> mac80211.c, what do you think? Something like (please note this patch is not
> tested):
Thanks for looking into this and improving the solution. Your patch
didn't apply cleanly to torvald's master, but it was easy to rebase.
> diff --git a/drivers/net/wireless/mediatek/mt76/mac80211.c
> b/drivers/net/wireless/mediatek/mt76/mac80211.c
> index b117e4467c87..34abf70f44af 100644
> --- a/drivers/net/wireless/mediatek/mt76/mac80211.c
> +++ b/drivers/net/wireless/mediatek/mt76/mac80211.c
> @@ -539,6 +539,7 @@ int mt76_register_phy(struct mt76_phy *phy, bool vht,
> if (ret)
> return ret;
>
> + set_bit(MT76_STATE_REGISTERED, &phy->state);
> phy->dev->phys[phy->band_idx] = phy;
>
> return 0;
> @@ -549,6 +550,9 @@ void mt76_unregister_phy(struct mt76_phy *phy)
> {
> struct mt76_dev *dev = phy->dev;
>
> + if (!test_bit(MT76_STATE_REGISTERED, &phy->state))
> + return;
> +
> if (IS_ENABLED(CONFIG_MT76_LEDS))
> mt76_led_cleanup(phy);
> mt76_tx_status_check(dev, true);
> @@ -719,6 +723,7 @@ int mt76_register_device(struct mt76_dev *dev, bool vht,
> return ret;
>
> WARN_ON(mt76_worker_setup(hw, &dev->tx_worker, NULL, "tx"));
> + set_bit(MT76_STATE_REGISTERED, &phy->state);
> sched_set_fifo_low(dev->tx_worker.task);
>
> return 0;
> @@ -729,6 +734,9 @@ void mt76_unregister_device(struct mt76_dev *dev)
> {
> struct ieee80211_hw *hw = dev->hw;
>
> + if (!test_bit(MT76_STATE_REGISTERED, &dev->phy.state))
> + return;
> +
> if (IS_ENABLED(CONFIG_MT76_LEDS))
> mt76_led_cleanup(&dev->phy);
> mt76_tx_status_check(dev, true);
> diff --git a/drivers/net/wireless/mediatek/mt76/mt76.h
> b/drivers/net/wireless/mediatek/mt76/mt76.h
> index 4ffb6be63571..38c311760d15 100644
> --- a/drivers/net/wireless/mediatek/mt76/mt76.h
> +++ b/drivers/net/wireless/mediatek/mt76/mt76.h
> @@ -402,6 +402,7 @@ struct mt76_tx_cb {
>
> enum {
> MT76_STATE_INITIALIZED,
> + MT76_STATE_REGISTERED,
> MT76_STATE_RUNNING,
> MT76_STATE_MCU_RUNNING,
> MT76_SCANNING,
>
Tested-by: Helmut Grohne <hel...@freexian.com>
I confirm that it really solves the problem on MT7921 PCIe hardware.
Helmut
