Re: [PATCH net-next] r8169: perform reset synchronously in __rtl8169_resume

[Heiner Kallweit <hkallweit1@gmail.com>]

Am 21.05.2018 um 23:24 schrieb Francois Romieu:
> Heiner Kallweit <hkallweit1@gmail.com> :
> [...]
>> diff --git a/drivers/net/ethernet/realtek/r8169.c b/drivers/net/ethernet/realtek/r8169.c
>> index 75dfac024..1eb4f625a 100644
>> --- a/drivers/net/ethernet/realtek/r8169.c
>> +++ b/drivers/net/ethernet/realtek/r8169.c
>> @@ -7327,9 +7327,9 @@ static void __rtl8169_resume(struct net_device *dev)
>>  	rtl_lock_work(tp);
>>  	napi_enable(&tp->napi);
>>  	set_bit(RTL_FLAG_TASK_ENABLED, tp->wk.flags);
>> +	if (netif_running(dev))
>> +		rtl_reset_work(tp);
>>  	rtl_unlock_work(tp);
>> -
>> -	rtl_schedule_task(tp, RTL_FLAG_TASK_RESET_PENDING);
>>  }
>>  
>>  static int rtl8169_resume(struct device *device)
> 
> netif_running() returns true before rtl_open(): issuing rtl_reset_work()
> synchronously against uninitialized rings right at the start of rtl_open()
> won't work as expected.
> 
rtl8169_runtime_resume() has a check for tp->TxDescArray at the beginning.
Therefore it will return immediately before even calling
__rtl8169_resume(), because tp->TxDescArray is set in rtl_open() only.
So I don't think the described issue exists.

However I have to admit that I wasn't aware that netif_running() is true
already when entering rtl_open(), so thanks for the hint.

> pm_runtime_get_sync() could be issued later in rtl_open() (but I would
> not mind something different with runtime_{resume/suspend} in open/close).
>