Re: [PATCH] wifi: rtl8xxxu: don't call dev_kfree_skb() under spin_lock_irqsave()

[Yang Yingliang <yangyingliang@huawei.com>]

On 2022/12/8 9:41, Ping-Ke Shih wrote:
>> -----Original Message-----
>> From: Yang Yingliang <yangyingliang@huawei.com>
>> Sent: Thursday, December 8, 2022 9:26 AM
>> To: Ping-Ke Shih <pkshih@realtek.com>; Jes.Sorensen@gmail.com; kvalo@kernel.org
>> Cc: linux-wireless@vger.kernel.org; yangyingliang@huawei.com
>> Subject: Re: [PATCH] wifi: rtl8xxxu: don't call dev_kfree_skb() under spin_lock_irqsave()
>>
>>
>> On 2022/12/8 8:38, Ping-Ke Shih wrote:
>>>> -----Original Message-----
>>>> From: Yang Yingliang <yangyingliang@huawei.com>
>>>> Sent: Wednesday, December 7, 2022 10:38 PM
>>>> To: Jes.Sorensen@gmail.com; kvalo@kernel.org
>>>> Cc: linux-wireless@vger.kernel.org
>>>> Subject: [PATCH] wifi: rtl8xxxu: don't call dev_kfree_skb() under spin_lock_irqsave()
>>>>
>>>> It is not allowed to call consume_skb() from hardware interrupt context
>>>                               ^^^^^^^^^^^^^ kfree_skb()?
>>> because this patch is to replace dev_kfree_skb().
>>>
>>>> or with interrupts being disabled. So replace dev_kfree_skb() with
>>>> dev_consume_skb_irq() under spin_lock_irqsave(). Compile tested only.
>>>>
>>>> Fixes: 26f1fad29ad9 ("New driver: rtl8xxxu (mac80211)")
>>>> Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
>>>> ---
>>>>    drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c | 2 +-
>>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c
>>>> b/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c
>>>> index ac641a56efb0..d0600af5bef4 100644
>>>> --- a/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c
>>>> +++ b/drivers/net/wireless/realtek/rtl8xxxu/rtl8xxxu_core.c
>>>> @@ -5274,7 +5274,7 @@ static void rtl8xxxu_queue_rx_urb(struct rtl8xxxu_priv *priv,
>>>>    		pending = priv->rx_urb_pending_count;
>>>>    	} else {
>>>>    		skb = (struct sk_buff *)rx_urb->urb.context;
>>>> -		dev_kfree_skb(skb);
>>>> +		dev_consume_skb_irq(skb);
>>> Why not dev_kfree_skb_irq() instead? any reason?
>> #define dev_kfree_skb(a)        consume_skb(a)
>> dev_kfree_skb() is consume_skb(), so use dev_consume_skb_irq() instead.
>>
>> static inline void dev_kfree_skb_irq(struct sk_buff *skb)
>> {
>>           __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
>> }
>>
>> static inline void dev_consume_skb_irq(struct sk_buff *skb)
>> {
>>           __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
>> }
>> They have different free reasons.
>>
> It falls into this case because of 'priv->shutdown', so DROPPED reason makes
> sense, no? Or I misunderstand the reason?
You are right, it's better to use dev_kfree_skb_irq(), because this is 
called when it's
stopped and need to drop the SKB, I will send a v2 to change it.

Thanks,
Yang
>
> --
> Ping-Ke
>