Re: [PATCH net-next v7] net/core: Introduce netdev_core_stats_inc()

From: Eric Dumazet
Date: Sat Oct 07 2023 - 01:30:22 EST


On Sat, Oct 7, 2023 at 7:06 AM Yajun Deng <yajun.deng@xxxxxxxxx> 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.
>
...

> +
> +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);
> + unsigned long *field;
> +
> + if (unlikely(!p))
> + p = netdev_core_stats_alloc(dev);
> +
> + if (p) {
> + field = (unsigned long *)((void *)this_cpu_ptr(p) + offset);
> + WRITE_ONCE(*field, READ_ONCE(*field) + 1);

This is broken...

As I explained earlier, dev_core_stats_xxxx(dev) can be called from
many different contexts:

1) process contexts, where preemption and migration are allowed.
2) interrupt contexts.

Adding WRITE_ONCE()/READ_ONCE() is not solving potential races.

I _think_ I already gave you how to deal with this ?

Please try instead:

+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);
+ unsigned long __percpu *field;
+
+ if (unlikely(!p)) {
+ p = netdev_core_stats_alloc(dev);
+ if (!p)
+ return;
+ }
+ field = (__force unsigned long __percpu *)((__force void *)p + offset);
+ this_cpu_inc(*field);
+}