Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Received-SPF: Pass (protection.outlook.com: domain of nvidia.com designates
 216.228.112.34 as permitted sender) receiver=protection.outlook.com;
 client-ip=216.228.112.34; helo=mail.nvidia.com;
Subject: Re: [PATCH] net: mellanox: mlx5: fix error return code of
 mlx5e_stats_flower()
To:     Saeed Mahameed <saeed@kernel.org>,
        Jia-Ju Bai <baijiaju1990@gmail.com>, <leon@kernel.org>,
        <davem@davemloft.net>, <kuba@kernel.org>
CC:     <netdev@vger.kernel.org>, <linux-rdma@vger.kernel.org>,
        <linux-kernel@vger.kernel.org>
References: <20210306134718.17566-1-baijiaju1990@gmail.com>
 <99807217-c2a3-928f-4c8c-2195f3500594@nvidia.com>
 <1fdf71d7-7a54-0b69-3339-e792dd03b2c8@nvidia.com>
 <3a1a2089-a7fa-2a7c-7040-c0aa62b08960@gmail.com>
 <78a83112-2978-42e9-a90e-c8bee0389fd8@nvidia.com>
 <5e0ba8e23231eb15e6f1dbfa9759b3aa94267193.camel@kernel.org>
From:   Roi Dayan <roid@nvidia.com>
Message-ID: <59e7eba0-c26b-6833-8d92-1a4e04ac9e9b@nvidia.com>
Date:   Sun, 14 Mar 2021 11:50:14 +0200
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:78.0) Gecko/20100101
 Thunderbird/78.8.0
MIME-Version: 1.0
In-Reply-To: <5e0ba8e23231eb15e6f1dbfa9759b3aa94267193.camel@kernel.org>
Content-Type: text/plain; charset="utf-8"; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
X-MS-Exchange-CrossTenant-OriginalArrivalTime: 14 Mar 2021 09:50:19.8426
 (UTC)
X-MS-Exchange-CrossTenant-Network-Message-Id: aa28d26f-7932-45b5-4298-08d8e6ce945b
X-MS-Exchange-CrossTenant-Id: 43083d15-7273-40c1-b7db-39efd9ccc17a
X-MS-Exchange-CrossTenant-OriginalAttributedTenantConnectingIp: TenantId=43083d15-7273-40c1-b7db-39efd9ccc17a;Ip=[216.228.112.34];Helo=[mail.nvidia.com]
X-MS-Exchange-CrossTenant-AuthSource: DM6NAM11FT055.eop-nam11.prod.protection.outlook.com
X-MS-Exchange-CrossTenant-AuthAs: Anonymous
X-MS-Exchange-CrossTenant-FromEntityHeader: HybridOnPrem
X-MS-Exchange-Transport-CrossTenantHeadersStamped: DM6PR12MB3273
Precedence: bulk
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org


On 2021-03-12 12:47 AM, Saeed Mahameed wrote:
> On Tue, 2021-03-09 at 11:44 +0200, Roi Dayan wrote:
>>
>>
>> On 2021-03-09 10:32 AM, Jia-Ju Bai wrote:
>>>
>>>
>>> On 2021/3/9 16:24, Roi Dayan wrote:
>>>>
>>>>
>>>> On 2021-03-09 10:20 AM, Roi Dayan wrote:
>>>>>
>>>>>
>>>>> On 2021-03-06 3:47 PM, Jia-Ju Bai wrote:
>>>>>> When mlx5e_tc_get_counter() returns NULL to counter or
>>>>>> mlx5_devcom_get_peer_data() returns NULL to peer_esw, no
>>>>>> error return
>>>>>> code of mlx5e_stats_flower() is assigned.
>>>>>> To fix this bug, err is assigned with -EINVAL in these cases.
>>>>>>
>>>>>> Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
> 
> Hey Jia-Ju, What are the conditions for this robot to raise a flag?
> sometimes it is totally normal to abort a function and return 0.. i am
> just curious to know ?
> 
> 
>>>>>> Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
>>>>>> ---
>>>>>>    drivers/net/ethernet/mellanox/mlx5/core/en_tc.c | 12
>>>>>> +++++++++---
>>>>>>    1 file changed, 9 insertions(+), 3 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
>>>>>> b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
>>>>>> index 0da69b98f38f..1f2c9da7bd35 100644
>>>>>> --- a/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
>>>>>> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_tc.c
>>>>>> @@ -4380,8 +4380,10 @@ int mlx5e_stats_flower(struct
>>>>>> net_device
>>>>>> *dev, struct mlx5e_priv *priv,
>>>>>>        if (mlx5e_is_offloaded_flow(flow) ||
>>>>>> flow_flag_test(flow, CT)) {
>>>>>>            counter = mlx5e_tc_get_counter(flow);
>>>>>> -        if (!counter)
>>>>>> +        if (!counter) {
>>>>>> +            err = -EINVAL;
>>>>>>                goto errout;
>>>>>> +        }
>>>>>>            mlx5_fc_query_cached(counter, &bytes, &packets,
>>>>>> &lastuse);
>>>>>>        }
>>>>>> @@ -4390,8 +4392,10 @@ int mlx5e_stats_flower(struct
>>>>>> net_device
>>>>>> *dev, struct mlx5e_priv *priv,
>>>>>>         * un-offloaded while the other rule is offloaded.
>>>>>>         */
>>>>>>        peer_esw = mlx5_devcom_get_peer_data(devcom,
>>>>>> MLX5_DEVCOM_ESW_OFFLOADS);
>>>>>> -    if (!peer_esw)
>>>>>> +    if (!peer_esw) {
>>>>>> +        err = -EINVAL;
>>>>>
> 
> This is not an error flow, i am curious what are the thresholds of this
> robot ?
> 
>>>>> note here it's not an error. it could be there is no peer esw
>>>>> so just continue with the stats update.
>>>>>
>>>>>>            goto out;
>>>>>> +    }
>>>>>>        if (flow_flag_test(flow, DUP) &&
>>>>>>            flow_flag_test(flow->peer_flow, OFFLOADED)) {
>>>>>> @@ -4400,8 +4404,10 @@ int mlx5e_stats_flower(struct
>>>>>> net_device
>>>>>> *dev, struct mlx5e_priv *priv,
>>>>>>            u64 lastuse2;
>>>>>>            counter = mlx5e_tc_get_counter(flow->peer_flow);
>>>>>> -        if (!counter)
>>>>>> +        if (!counter) {
>>>>>> +            err = -EINVAL;
>>>>
>>>> this change is problematic. the current goto is to do stats
>>>> update with
>>>> the first counter stats we got but if you now want to return an
>>>> error
>>>> then you probably should not do any update at all.
>>>
>>> Thanks for your reply :)
>>> I am not sure whether an error code should be returned here?
>>> If so, flow_stats_update(...) should not be called here?
>>>
>>>
>>> Best wishes,
>>> Jia-Ju Bai
>>>
>>
>> basically flow and peer_flow should be valid and protected from
>> changes,
>> and counter should be valid.
>> it looks like the check here is more of a sanity check if something
>> goes
>> wrong but shouldn't. you can just let it be, update the stats from
>> the
>> first queried counter.
>>
> 
> Roi, let's consider returning an error code here, we shouldn't be
> silently returning if we are not expecting these errors,
> 
> why would mlx5e_stats_flower() be called if stats are not offloaded ?
> 
> Thanks,
> Saeed.
> 
> 

yes we can return an error if peer counter missing.
I just pointed out we should probably not call flow_stats_update() if
we do return an error.
the other option, as today, is updating the stats with first counter
stats and because of that we didn't return an error.