On Sun, Nov 12, 2023 at 09:30:03PM +0100, Daniel Borkmann wrote:
> Move {l,t,d}stats allocation to the core and let netdevs pick the stats
> type they need. That way the driver doesn't have to bother with error
> handling (allocation failure checking, making sure free happens in the
> right spot, etc) - all happening in the core.
>
> Co-developed-by: Jakub Kicinski <kuba@xxxxxxxxxx>
> Signed-off-by: Jakub Kicinski <kuba@xxxxxxxxxx>
> Signed-off-by: Daniel Borkmann <daniel@xxxxxxxxxxxxx>
> Cc: David Ahern <dsahern@xxxxxxxxxx>
Hi Daniel,
sorry I was a bit to hasty in hitting the send button for my previous
message. I have a some more minor feedback on this series.
> diff --git a/net/core/dev.c b/net/core/dev.c
> index 0d548431f3fa..75db81496db5 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -10049,6 +10049,44 @@ void netif_tx_stop_all_queues(struct net_device *dev)
> }
> EXPORT_SYMBOL(netif_tx_stop_all_queues);
>
> +static int netdev_do_alloc_pcpu_stats(struct net_device *dev)
> +{
> + void __percpu *v;
> +
> + switch (dev->pcpu_stat_type) {
> + case NETDEV_PCPU_STAT_NONE:
> + return 0;
> + case NETDEV_PCPU_STAT_LSTATS:
> + v = dev->lstats = netdev_alloc_pcpu_stats(struct pcpu_lstats);
> + break;
> + case NETDEV_PCPU_STAT_TSTATS:
> + v = dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
> + break;
> + case NETDEV_PCPU_STAT_DSTATS:
> + v = dev->dstats = netdev_alloc_pcpu_stats(struct pcpu_dstats);
> + break;
> + }
> +
> + return v ? 0 : -ENOMEM;
Perhaps this cannot happen, but if none of the cases in the switch
statement are met, then v will be uninitialised here.
As flagged by Smatch.
> +}
> +
...
> @@ -10469,6 +10513,7 @@ void netdev_run_todo(void)
> WARN_ON(rcu_access_pointer(dev->ip_ptr));
> WARN_ON(rcu_access_pointer(dev->ip6_ptr));
>
> + netdev_do_free_pcpu_stats(dev);
> if (dev->priv_destructor)
> dev->priv_destructor(dev);
> if (dev->needs_free_netdev)
nit: the hunk above seems unnecessary; one blank line is enough.
