[PATCH net-next v5] net/core: Introduce netdev_core_stats_inc() for trace

Yajun Deng posted 1 patch 9 months, 4 weeks ago
include/linux/netdevice.h | 21 ++++-----------------
net/core/dev.c            | 26 ++++++++++++++++++++++++--
2 files changed, 28 insertions(+), 19 deletions(-)
[PATCH net-next v5] net/core: Introduce netdev_core_stats_inc() for trace
Posted by Yajun Deng 9 months, 4 weeks ago
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.

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.

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@linux.dev/
---
 include/linux/netdevice.h | 21 ++++-----------------
 net/core/dev.c            | 26 ++++++++++++++++++++++++--
 2 files changed, 28 insertions(+), 19 deletions(-)

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;
 }
 
-struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
-
-static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
-{
-	/* 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);
-}
+void netdev_core_stats_inc(struct net_device *dev, u32 offset);
 
 #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 = 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
 
 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_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)
 {
 	struct net_device_core_stats __percpu *p;
 
@@ -10510,7 +10511,28 @@ 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 *netdev_core_stats(
+		struct net_device *dev)
+{
+	/* 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);
+}
+
+void netdev_core_stats_inc(struct net_device *dev, u32 offset)
+{
+	struct net_device_core_stats __percpu *p;
+
+	p = netdev_core_stats(dev);
+	if (p)
+		(*(unsigned long *)((void *)this_cpu_ptr(p) + offset))++;
+}
+EXPORT_SYMBOL_GPL(netdev_core_stats_inc);
 
 /**
  *	dev_get_stats	- get network device statistics
-- 
2.25.1
Re: [PATCH net-next v5] net/core: Introduce netdev_core_stats_inc() for trace
Posted by Paolo Abeni 9 months, 2 weeks ago
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 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.
> 
> 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.
> 
> 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@linux.dev/
> ---
>  include/linux/netdevice.h | 21 ++++-----------------
>  net/core/dev.c            | 26 ++++++++++++++++++++++++--
>  2 files changed, 28 insertions(+), 19 deletions(-)
> 
> 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;
>  }
>  
> -struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
> -
> -static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
> -{
> -	/* 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);
> -}
> +void netdev_core_stats_inc(struct net_device *dev, u32 offset);
>  
>  #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 = 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
>  
>  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_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)
>  {
>  	struct net_device_core_stats __percpu *p;
>  
> @@ -10510,7 +10511,28 @@ 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 *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
Re: [PATCH net-next v5] net/core: Introduce netdev_core_stats_inc() for trace
Posted by Yajun Deng 9 months, 2 weeks ago
On 2023/9/28 16:23, Paolo Abeni wrote:
> 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 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.
>>
>> 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.
>>
>> 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@linux.dev/
>> ---
>>   include/linux/netdevice.h | 21 ++++-----------------
>>   net/core/dev.c            | 26 ++++++++++++++++++++++++--
>>   2 files changed, 28 insertions(+), 19 deletions(-)
>>
>> 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;
>>   }
>>   
>> -struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
>> -
>> -static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
>> -{
>> -	/* 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);
>> -}
>> +void netdev_core_stats_inc(struct net_device *dev, u32 offset);
>>   
>>   #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 = 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
>>   
>>   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_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)
>>   {
>>   	struct net_device_core_stats __percpu *p;
>>   
>> @@ -10510,7 +10511,28 @@ 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 *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.


If so, should I merge netdev_core_stats and netdev_core_stats_inc 
together, as it didn't called by others. Like:

void netdev_core_stats_inc(struct net_device *dev, u32 offset)
{
         /* 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 (unlikely(!p))
                 p = netdev_core_stats_alloc(dev);

         if (p)
                 (*(unsigned long *)((void *)this_cpu_ptr(p) + offset))++;
}
EXPORT_SYMBOL_GPL(netdev_core_stats_inc);


>
> Cheers,
>
> Paolo
>
Re: [PATCH net-next v5] net/core: Introduce netdev_core_stats_inc() for trace
Posted by Paolo Abeni 9 months, 2 weeks ago
On Thu, 2023-09-28 at 17:21 +0800, Yajun Deng wrote:
> On 2023/9/28 16:23, Paolo Abeni wrote:
> > 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 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.
> > > 
> > > 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.
> > > 
> > > 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@linux.dev/
> > > ---
> > >   include/linux/netdevice.h | 21 ++++-----------------
> > >   net/core/dev.c            | 26 ++++++++++++++++++++++++--
> > >   2 files changed, 28 insertions(+), 19 deletions(-)
> > > 
> > > 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;
> > >   }
> > >   
> > > -struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
> > > -
> > > -static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
> > > -{
> > > -	/* 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);
> > > -}
> > > +void netdev_core_stats_inc(struct net_device *dev, u32 offset);
> > >   
> > >   #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 = 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
> > >   
> > >   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_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)
> > >   {
> > >   	struct net_device_core_stats __percpu *p;
> > >   
> > > @@ -10510,7 +10511,28 @@ 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 *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.
> 
> 
> If so, should I merge netdev_core_stats and netdev_core_stats_inc 
> together, as it didn't called by others. Like:
> 
> void netdev_core_stats_inc(struct net_device *dev, u32 offset)
> {
>          /* 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 (unlikely(!p))
>                  p = netdev_core_stats_alloc(dev);
> 
>          if (p)
>                  (*(unsigned long *)((void *)this_cpu_ptr(p) + offset))++;
> }
> EXPORT_SYMBOL_GPL(netdev_core_stats_inc);

Makes sense to me. I initially thought the code would be more readable
with 2 separate functions, but the above LGTM.

Cheers,

Paolo
Re: [PATCH net-next v5] net/core: Introduce netdev_core_stats_inc() for trace
Posted by Yajun Deng 9 months, 3 weeks ago
Hi, MAINTAINERS,

Kindly ping...

Thanks


On 2023/9/19 21:55, Yajun Deng wrote:
> 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.
>
> 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.
>
> 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@linux.dev/
> ---
>   include/linux/netdevice.h | 21 ++++-----------------
>   net/core/dev.c            | 26 ++++++++++++++++++++++++--
>   2 files changed, 28 insertions(+), 19 deletions(-)
>
> 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;
>   }
>   
> -struct net_device_core_stats __percpu *netdev_core_stats_alloc(struct net_device *dev);
> -
> -static inline struct net_device_core_stats __percpu *dev_core_stats(struct net_device *dev)
> -{
> -	/* 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);
> -}
> +void netdev_core_stats_inc(struct net_device *dev, u32 offset);
>   
>   #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 = 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
>   
>   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_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)
>   {
>   	struct net_device_core_stats __percpu *p;
>   
> @@ -10510,7 +10511,28 @@ 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 *netdev_core_stats(
> +		struct net_device *dev)
> +{
> +	/* 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);
> +}
> +
> +void netdev_core_stats_inc(struct net_device *dev, u32 offset)
> +{
> +	struct net_device_core_stats __percpu *p;
> +
> +	p = netdev_core_stats(dev);
> +	if (p)
> +		(*(unsigned long *)((void *)this_cpu_ptr(p) + offset))++;
> +}
> +EXPORT_SYMBOL_GPL(netdev_core_stats_inc);
>   
>   /**
>    *	dev_get_stats	- get network device statistics