Received-SPF: pass (google.com: domain of linux-kernel+bounces-25650-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Message-ID: <720e045a-1d5e-432e-9da8-5508eaa96720@tuxon.dev>
Date: Mon, 15 Jan 2024 08:08:22 +0200
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH net-next v3 17/19] net: ravb: Return cached statistics if
 the interface is down
To: Sergey Shtylyov <s.shtylyov@omp.ru>, davem@davemloft.net,
 edumazet@google.com, kuba@kernel.org, pabeni@redhat.com,
 richardcochran@gmail.com, p.zabel@pengutronix.de,
 yoshihiro.shimoda.uh@renesas.com, wsa+renesas@sang-engineering.com
Cc: netdev@vger.kernel.org, linux-renesas-soc@vger.kernel.org,
 linux-kernel@vger.kernel.org, geert+renesas@glider.be,
 Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
References: <20240105082339.1468817-1-claudiu.beznea.uj@bp.renesas.com>
 <20240105082339.1468817-18-claudiu.beznea.uj@bp.renesas.com>
 <af5ab82e-5904-c33b-983e-b37844dab3f5@omp.ru>
 <aed6534b-ad5e-4f5c-9861-9a784968adcc@tuxon.dev>
 <28754087-8685-015e-7e8a-d8c3ed26930c@omp.ru>
From: claudiu beznea <claudiu.beznea@tuxon.dev>
Content-Language: en-US
In-Reply-To: <28754087-8685-015e-7e8a-d8c3ed26930c@omp.ru>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit


On 14.01.2024 14:22, Sergey Shtylyov wrote:
> On 1/10/24 4:17 PM, claudiu beznea wrote:
> 
> [...]
> 
>>>> From: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
>>>>
>>>> Return the cached statistics in case the interface is down. There should be
>>>> no drawback to this, as cached statistics are updated in ravb_close().
>>>>
>>>> The commit prepares the code for the addition of runtime PM support.
>>>>
>>>> Suggested-by: Sergey Shtylyov <s.shtylyov@omp.ru>
>>>> Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
>>> [...]
>>>
>>>> diff --git a/drivers/net/ethernet/renesas/ravb_main.c b/drivers/net/ethernet/renesas/ravb_main.c
>>>> index 76035afd4054..168b6208db37 100644
>>>> --- a/drivers/net/ethernet/renesas/ravb_main.c
>>>> +++ b/drivers/net/ethernet/renesas/ravb_main.c
>>>> @@ -2117,6 +2117,9 @@ static struct net_device_stats *ravb_get_stats(struct net_device *ndev)
>>>>  	const struct ravb_hw_info *info = priv->info;
>>>>  	struct net_device_stats *nstats, *stats0, *stats1;
>>>>  
>>>> +	if (!(ndev->flags & IFF_UP))
>>>
>>>    Well, I guess it's OK to read the counters in the reset mode... BUT
>>> won't this race with pm_runtime_put_autosuspend() when its call gets added
>>> to ravb_close()?
>>
>> I re-checked it and, yes, this is true. A sync runtime suspend would be
>> better here. But, as of my current investigation, even with this
> 
>    No, the sync form of the RPM call won't fix the race...
> 
>> ravb_get_stats() can still race with ravb_open()/ravb_close() as they are
>> called though different locking scheme (ravb_open()/ravb_close() is called
>> with rtnl locked while ravb_get_stats() can be called only with
>> dev_base_lock rwlock locked for reading).
>>
>> A mutex in the driver should to help with this.
> 
>    Why don't you want to mimic what the sh_eth driver does?

I thought it can be replaced by the already existing IFF_UP flag from
ndev->flags.

Investigating it further while trying to address this concurrency issue
made me realize that it fits to address the issue you mentioned here.

Thank you,
Claudiu Beznea

> 
>> Thank you,
>> Claudiu Beznea
> 
> [...]
> 
> MBR, Sergey