So everybody jumped on GRO, while my concern was more the napi->dev
thing.

I pointed to gro because the commit was about gro, but gro was fine
IMHO.

Are we sure netpoll is ok ?

If a napi is shared by several net_device, should'nt we change
netpoll_poll_lock() ?

static inline void *netpoll_poll_lock(struct napi_struct *napi)
{
        struct net_device *dev = napi->dev;

        if (dev && dev->npinfo) {
                spin_lock(&napi->poll_lock);
                napi->poll_owner = smp_processor_id();
                return napi;
        }
        return NULL;
}

---->

static inline void *netpoll_poll_lock(struct napi_struct *napi)
{
        if (atomic_read(&napi->poll_count)) {
                spin_lock(&napi->poll_lock);
                napi->poll_owner = smp_processor_id();
                return napi;
        }
        return NULL;
}