[net-next,v3] net/core: Export dev_core_stats_*_inc()

  Although there is a kfree_skb_reason() helper function that can be used to
find the reason why this skb is dropped, but most callers didn't increase
one of rx_dropped, tx_dropped, rx_nohandler and rx_otherhost_dropped.

For the users, people are more concerned about why the dropped in ip
is increasing. So we can export dev_core_stats_rx_dropped_inc sets,
which users would trace them know why rx_dropped is increasing.

Export dev_core_stats_{rx_dropped, tx_dropped, rx_nohandler,
rx_otherhost_dropped}_inc for trace. Also, move dev_core_stats()
and netdev_core_stats_alloc() to dev.c, as they are not called
externally.

Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
---
v3: __cold should be added to the netdev_core_stats_alloc().
v2: use __cold instead of inline in dev_core_stats().
v1: https://lore.kernel.org/netdev/20230911082016.3694700-1-yajun.deng@linux.dev/
---
 include/linux/netdevice.h | 32 +++++---------------------------
 net/core/dev.c            | 32 ++++++++++++++++++++++++++++++--
 2 files changed, 35 insertions(+), 29 deletions(-)
  

From: Yajun Deng <yajun.deng@linux.dev>
Date: Thu, 14 Sep 2023 10:37:18 +0800

> Although there is a kfree_skb_reason() helper function that can be used to
> find the reason why this skb is dropped, but most callers didn't increase
> one of rx_dropped, tx_dropped, rx_nohandler and rx_otherhost_dropped.
> 
> For the users, people are more concerned about why the dropped in ip
> is increasing. So we can export dev_core_stats_rx_dropped_inc sets,
> which users would trace them know why rx_dropped is increasing.
> 
> Export dev_core_stats_{rx_dropped, tx_dropped, rx_nohandler,
> rx_otherhost_dropped}_inc for trace. Also, move dev_core_stats()
> and netdev_core_stats_alloc() to dev.c, as they are not called
> externally.

I'd like to hear some arguments against having them static inlines + one
external that I proposed earlier.

> 
> Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
> ---
> v3: __cold should be added to the netdev_core_stats_alloc().
> v2: use __cold instead of inline in dev_core_stats().
> v1: https://lore.kernel.org/netdev/20230911082016.3694700-1-yajun.deng@linux.dev/

...as it's not at least mentioned here in the changelog.

[...]

> diff --git a/net/core/dev.c b/net/core/dev.c
> index ccff2b6ef958..98592e4c1df0 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -10475,7 +10475,7 @@ void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
>  }
>  EXPORT_SYMBOL(netdev_stats_to_stats64);
>  
> -struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev)
> +static __cold struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev)

This is way over 79/80 chars, break the line before "netdev_".

>  {
>  	struct net_device_core_stats __percpu *p;
>  
> @@ -10488,7 +10488,35 @@ struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device
>  	/* This READ_ONCE() pairs with the cmpxchg() above */
>  	return READ_ONCE(dev->core_stats);
>  }
> -EXPORT_SYMBOL(netdev_core_stats_alloc);
> +
> +static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)

Same for the line length.
Also notice that now some of the functions you touch have "dev_" prefix,
others have "netdev_", I'd probably take a look into unifying this.

(note for the maintainers that it would be probably better to leave
 explicit "inline" here, but no bloat-o-meter was provided by the
 author, so I can't say anything for sure)

> +{
> +	/* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
> +	struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
> +
> +	if (likely(p))
> +		return p;
> +
> +	return netdev_core_stats_alloc(dev);
> +}

[...]

Thanks,
Olek
  

On 2023/9/14 23:22, Alexander Lobakin wrote:
> From: Yajun Deng <yajun.deng@linux.dev>
> Date: Thu, 14 Sep 2023 10:37:18 +0800
>
>> Although there is a kfree_skb_reason() helper function that can be used to
>> find the reason why this skb is dropped, but most callers didn't increase
>> one of rx_dropped, tx_dropped, rx_nohandler and rx_otherhost_dropped.
>>
>> For the users, people are more concerned about why the dropped in ip
>> is increasing. So we can export dev_core_stats_rx_dropped_inc sets,
>> which users would trace them know why rx_dropped is increasing.
>>
>> Export dev_core_stats_{rx_dropped, tx_dropped, rx_nohandler,
>> rx_otherhost_dropped}_inc for trace. Also, move dev_core_stats()
>> and netdev_core_stats_alloc() to dev.c, as they are not called
>> externally.
> I'd like to hear some arguments against having them static inlines + one
> external that I proposed earlier.

I'd like to hear suggestions from the maintainers. If the maintainers 
have no objections, I'd like to

use your approach.

>> Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
>> ---
>> v3: __cold should be added to the netdev_core_stats_alloc().
>> v2: use __cold instead of inline in dev_core_stats().
>> v1: https://lore.kernel.org/netdev/20230911082016.3694700-1-yajun.deng@linux.dev/
> ...as it's not at least mentioned here in the changelog.
>
> [...]
>
>> diff --git a/net/core/dev.c b/net/core/dev.c
>> index ccff2b6ef958..98592e4c1df0 100644
>> --- a/net/core/dev.c
>> +++ b/net/core/dev.c
>> @@ -10475,7 +10475,7 @@ void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
>>   }
>>   EXPORT_SYMBOL(netdev_stats_to_stats64);
>>   
>> -struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev)
>> +static __cold struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev)
> This is way over 79/80 chars, break the line before "netdev_".
Okay.
>
>>   {
>>   	struct net_device_core_stats __percpu *p;
>>   
>> @@ -10488,7 +10488,35 @@ struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device
>>   	/* This READ_ONCE() pairs with the cmpxchg() above */
>>   	return READ_ONCE(dev->core_stats);
>>   }
>> -EXPORT_SYMBOL(netdev_core_stats_alloc);
>> +
>> +static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
> Same for the line length.
> Also notice that now some of the functions you touch have "dev_" prefix,
> others have "netdev_", I'd probably take a look into unifying this.
Okay, I'll unify with the "netdev_" prefix.
> (note for the maintainers that it would be probably better to leave
>   explicit "inline" here, but no bloat-o-meter was provided by the
>   author, so I can't say anything for sure)
>
>> +{
>> +	/* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() */
>> +	struct net_device_core_stats __percpu *p = READ_ONCE(dev->core_stats);
>> +
>> +	if (likely(p))
>> +		return p;
>> +
>> +	return netdev_core_stats_alloc(dev);
>> +}
> [...]
>
> Thanks,
> Olek