Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.34 as permitted sender) client-ip=23.128.96.34;
Message-ID: <619e3735d99c8642b7b84a151515c2fee99ff694.camel@redhat.com>
Subject: Re: [PATCH net-next v5] net/core: Introduce netdev_core_stats_inc()
 for trace
From:   Paolo Abeni <pabeni@redhat.com>
To:     Yajun Deng <yajun.deng@linux.dev>, davem@davemloft.net,
        edumazet@google.com, kuba@kernel.org
Cc:     netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
        Alexander Lobakin <aleksander.lobakin@intel.com>
Date:   Thu, 28 Sep 2023 10:23:50 +0200
In-Reply-To: <20230919135517.286766-1-yajun.deng@linux.dev>
References: <20230919135517.286766-1-yajun.deng@linux.dev>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.46.4 (3.46.4-1.fc37) 
MIME-Version: 1.0
Precedence: bulk

On Tue, 2023-09-19 at 21:55 +0800, Yajun Deng wrote:
> Although there is a kfree_skb_reason() helper function that can be used t=
o
> 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.
>=20
> For the users, people are more concerned about why the dropped in ip
> is increasing.
>=20
> Introduce netdev_core_stats_inc() for trace. Also, move dev_core_stats()
> and netdev_core_stats_alloc() to dev.c, as they are not called externally=
.
>=20
> Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
> Suggested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
> ---
> v5: Access the per cpu pointer before reach the relevant offset.
> v4: Introduce netdev_core_stats_inc() instead of export dev_core_stats_*_=
inc()
> 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@li=
nux.dev/
> ---
>  include/linux/netdevice.h | 21 ++++-----------------
>  net/core/dev.c            | 26 ++++++++++++++++++++++++--
>  2 files changed, 28 insertions(+), 19 deletions(-)
>=20
> diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
> index db3d8429d50d..4c258d44c7d2 100644
> --- a/include/linux/netdevice.h
> +++ b/include/linux/netdevice.h
> @@ -4001,32 +4001,19 @@ static __always_inline bool __is_skb_forwardable(=
const struct net_device *dev,
>  	return false;
>  }
> =20
> -struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct ne=
t_device *dev);
> -
> -static inline struct net_device_core_stats __percpu *dev_core_stats(stru=
ct net_device *dev)
> -{
> -	/* This READ_ONCE() pairs with the write in netdev_core_stats_alloc() *=
/
> -	struct net_device_core_stats __percpu *p =3D READ_ONCE(dev->core_stats)=
;
> -
> -	if (likely(p))
> -		return p;
> -
> -	return netdev_core_stats_alloc(dev);
> -}
> +void netdev_core_stats_inc(struct net_device *dev, u32 offset);
> =20
>  #define DEV_CORE_STATS_INC(FIELD)						\
>  static inline void dev_core_stats_##FIELD##_inc(struct net_device *dev)	=
	\
>  {										\
> -	struct net_device_core_stats __percpu *p;				\
> -										\
> -	p =3D dev_core_stats(dev);						\
> -	if (p)									\
> -		this_cpu_inc(p->FIELD);						\
> +	netdev_core_stats_inc(dev,						\
> +			offsetof(struct net_device_core_stats, FIELD));		\
>  }
>  DEV_CORE_STATS_INC(rx_dropped)
>  DEV_CORE_STATS_INC(tx_dropped)
>  DEV_CORE_STATS_INC(rx_nohandler)
>  DEV_CORE_STATS_INC(rx_otherhost_dropped)
> +#undef DEV_CORE_STATS_INC
> =20
>  static __always_inline int ____dev_forward_skb(struct net_device *dev,
>  					       struct sk_buff *skb,
> diff --git a/net/core/dev.c b/net/core/dev.c
> index 606a366cc209..4bc0161bc0d6 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -10497,7 +10497,8 @@ void netdev_stats_to_stats64(struct rtnl_link_sta=
ts64 *stats64,
>  }
>  EXPORT_SYMBOL(netdev_stats_to_stats64);
> =20
> -struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct ne=
t_device *dev)
> +static __cold struct net_device_core_stats __percpu *netdev_core_stats_a=
lloc(
> +		struct net_device *dev)
>  {
>  	struct net_device_core_stats __percpu *p;
> =20
> @@ -10510,7 +10511,28 @@ struct net_device_core_stats __percpu *netdev_co=
re_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 *netdev_core_stats(
> +		struct net_device *dev)

I'm sorry for the delayed feedback - conference and traveling in the
way.

It looks like the 'inline' keyword above is a left-over of a previous
revision? The compiler should generate the same code even without it,
right? If so, it should be better drop it.

Cheers,

Paolo