Re: [net-next,v25,6/6] eea: introduce callback for ndo_get_stats64

[Jakub Kicinski <kuba@kernel.org>]

On Tue,  3 Feb 2026 20:08:41 -0800 Jakub Kicinski wrote:
> > +static void eea_stats(struct net_device *netdev, struct rtnl_link_stats64 *tot)
> > +{
> > +	struct eea_net *enet = netdev_priv(netdev);
> > +	u64 packets, bytes;
> > +	u32 start;
> > +	int i;
> > +
> > +	spin_lock_bh(&enet->stats_lock);  
> 
> Can this cause a deadlock? The enet_bind_new_q_and_cfg() and
> eea_free_rxtx_q_mem() functions use spin_lock(&enet->stats_lock), but
> eea_stats() uses spin_lock_bh(&enet->stats_lock).
> 
> If a thread holds stats_lock via spin_lock() (softirqs still enabled) and a
> softirq fires that calls eea_stats() through ndo_get_stats64, the
> spin_lock_bh() will attempt to acquire the already-held lock.
> 
> Should all acquisition sites use spin_lock_bh()/spin_unlock_bh() for
> consistency?

AI suggestion is semi-right here, just remove the _bh()
I don't think we actually read stats from BH.
We read them under RCU, which is atomic context but not BH.