Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.37 as permitted sender) client-ip=23.128.96.37;
Message-ID: <dfe6cf9b-7cf9-f938-8ab1-a2d4c7741c64@linux.dev>
Date:   Thu, 14 Sep 2023 10:44:14 +0800
MIME-Version: 1.0
Subject: Re: [PATCH] net/core: Export dev_core_stats_rx_dropped_inc sets
Content-Language: en-US
To:     Alexander Lobakin <aleksander.lobakin@intel.com>
Cc:     davem@davemloft.net, edumazet@google.com, kuba@kernel.org,
        pabeni@redhat.com, horms@kernel.org, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <20230911082016.3694700-1-yajun.deng@linux.dev>
 <9a1de9b3-b3cf-d26b-388e-c98294580bca@intel.com>
 <599cdff8-4865-3ade-0439-36e337891ca0@linux.dev>
 <e6de7c0b-97cc-090e-a331-dc566019f54e@intel.com>
From:   Yajun Deng <yajun.deng@linux.dev>
In-Reply-To: <e6de7c0b-97cc-090e-a331-dc566019f54e@intel.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 2023/9/13 17:58, Alexander Lobakin wrote:
> From: Yajun Deng <yajun.deng@linux.dev>
> Date: Wed, 13 Sep 2023 10:08:08 +0800
>
>> On 2023/9/13 00:22, Alexander Lobakin wrote:
>>> From: Yajun Deng <yajun.deng@linux.dev>
>>> Date: Mon, 11 Sep 2023 16:20:16 +0800
> [...]
>
>>> EXPORT_SYMBOL_GPL(dev_core_stats_inc); // Why not GPL BTW?
>> This may be a better option.
>>
>> Just because EXPORT_SYMBOL(netdev_core_stats_alloc) before,  but I think
>>
>> EXPORT_SYMBOL_GPL is better.
> Ah I see. BTW, if you will still define increment functions as
> externals, there will be no reason to export netdev_core_stats_alloc()
> or even make it non-static at all.
>
>>   
>>> And then build inlines:
>>>
>>> #define DEV_CORE_STATS_INC(FIELD)				\
>>> static inline void						\
>>> dev_core_stats_##FIELD##_inc(struct net_device *dev)		\
>>> {								\
>>> 	dev_core_stats_inc(dev,					\
>>> 		offsetof(struct net_device_core_stats, FIELD));	\
>>> }
>>>
>>> DEV_CORE_STATS_INC(rx_dropped);
>>> ...
>>>
>>> OR even just make them macros
>>>
>>> #define __DEV_CORE_STATS_INC(dev, field)			\
>>> 	dev_core_stats_inc(dev,					\
>>> 		offsetof(struct net_device_core_stats, field))
>>>
>>> #define dev_core_stats_rx_dropped_inc(dev)			\
>>> 	__DEV_CORE_STATS_INC(dev, rx_dropped)
>>> ...
>> I would like the former.  Keep it the same as before.
> By "the former" you mean to build static inlines or externals? Seems
> like the first one, but I got confused by your "the same as before" :D
>
>>
>>> Just don't copy that awful Thunderbird's line wrap and don't assume this
>>> code builds and works and that is something finished/polished.
>>>
>>> You'll be able to trace functions and you'll be able to understand which
>>> counter has been incremented by checking the second argument, i.e. the
>>> field offset (IIRC tracing shows you arguments).
>>> And that way you wouldn't geometrically increase the number of symbol
>>> exports and deal with its consequences.
>> I agree that.
> Ok, after this one I guess you meant "I'd like to use your approach with
> static inlines".

Finally, I give up this approach.

The new function dev_core_stats_inc() didn't called by external modules 
directly.

So EXPORT_SYMBOL_GPL(dev_core_stats_inc) can be removed by anyone.


>>>>   
>>>>   /**
>>>>    *	dev_get_stats	- get network device statistics
>>> Thanks,
>>> Olek
> Thanks,
> Olek