Re: [PATCH v1 net-next 06/14] neighbour: Free pneigh_entry after RCU grace period.

[Simon Horman <horms@kernel.org>]

On Sat, Jul 12, 2025 at 06:07:51PM +0000, Kuniyuki Iwashima wrote:
> From: Simon Horman <horms@kernel.org>
> Date: Sat, 12 Jul 2025 16:01:59 +0100
> > On Fri, Jul 11, 2025 at 07:06:11PM +0000, Kuniyuki Iwashima wrote:
> > > We will convert RTM_GETNEIGH to RCU.
> > > 
> > > neigh_get() looks up pneigh_entry by pneigh_lookup() and passes
> > > it to pneigh_fill_info().
> > > 
> > > Then, we must ensure that the entry is alive till pneigh_fill_info()
> > > completes, but read_lock_bh(&tbl->lock) in pneigh_lookup() does not
> > > guarantee that.
> > > 
> > > Also, we will convert all readers of tbl->phash_buckets[] to RCU.
> > > 
> > > Let's use call_rcu() to free pneigh_entry and update phash_buckets[]
> > > and ->next by rcu_assign_pointer().
> > > 
> > > pneigh_ifdown_and_unlock() uses list_head to avoid overwriting
> > > ->next and moving RCU iterators to another list.
> > > 
> > > pndisc_destructor() (only IPv6 ndisc uses this) uses a mutex, so it
> > > is not delayed to call_rcu(), where we cannot sleep.  This is fine
> > > because the mcast code works with RCU and ipv6_dev_mc_dec() frees
> > > mcast objects after RCU grace period.
> > > 
> > > While at it, we change the return type of pneigh_ifdown_and_unlock()
> > > to void.
> > > 
> > > Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
> > > ---
> > >  include/net/neighbour.h |  4 ++++
> > >  net/core/neighbour.c    | 51 +++++++++++++++++++++++++----------------
> > >  2 files changed, 35 insertions(+), 20 deletions(-)
> > > 
> > > diff --git a/include/net/neighbour.h b/include/net/neighbour.h
> > > index 7f3d57da5689a..a877e56210b22 100644
> > > --- a/include/net/neighbour.h
> > > +++ b/include/net/neighbour.h
> > > @@ -180,6 +180,10 @@ struct pneigh_entry {
> > >  	possible_net_t		net;
> > >  	struct net_device	*dev;
> > >  	netdevice_tracker	dev_tracker;
> > > +	union {
> > > +		struct list_head	free_node;
> > > +		struct rcu_head		rcu;
> > > +	};
> > >  	u32			flags;
> > >  	u8			protocol;
> > >  	bool			permanent;
> > > diff --git a/net/core/neighbour.c b/net/core/neighbour.c
> > > index 814a45fb1962e..6725a40b2db3a 100644
> > > --- a/net/core/neighbour.c
> > > +++ b/net/core/neighbour.c
> > > @@ -54,9 +54,9 @@ static void neigh_timer_handler(struct timer_list *t);
> > >  static void __neigh_notify(struct neighbour *n, int type, int flags,
> > >  			   u32 pid);
> > >  static void neigh_update_notify(struct neighbour *neigh, u32 nlmsg_pid);
> > > -static int pneigh_ifdown_and_unlock(struct neigh_table *tbl,
> > > -				    struct net_device *dev,
> > > -				    bool skip_perm);
> > > +static void pneigh_ifdown_and_unlock(struct neigh_table *tbl,
> > > +				     struct net_device *dev,
> > > +				     bool skip_perm);
> > >  
> > >  #ifdef CONFIG_PROC_FS
> > >  static const struct seq_operations neigh_stat_seq_ops;
> > > @@ -803,12 +803,20 @@ struct pneigh_entry *pneigh_create(struct neigh_table *tbl,
> > >  
> > >  	write_lock_bh(&tbl->lock);
> > >  	n->next = tbl->phash_buckets[hash_val];
> > > -	tbl->phash_buckets[hash_val] = n;
> > > +	rcu_assign_pointer(tbl->phash_buckets[hash_val], n);
> > 
> > Hi Iwashima-san,
> > 
> > A heads-up that unfortunately Sparse is unhappy about the __rcu annotations
> > here, and elsewhere in this patch (set).
> > 
> > For this patch I see:
> > 
> >   .../neighbour.c:860:33: error: incompatible types in comparison expression (different address spaces):
> >   .../neighbour.c:860:33:    struct pneigh_entry [noderef] __rcu *
> >   .../neighbour.c:860:33:    struct pneigh_entry *
> >   .../neighbour.c:806:9: error: incompatible types in comparison expression (different address spaces):
> >   .../neighbour.c:806:9:    struct pneigh_entry [noderef] __rcu *
> >   .../neighbour.c:806:9:    struct pneigh_entry *
> >   .../neighbour.c:832:25: error: incompatible types in comparison expression (different address spaces):
> >   .../neighbour.c:832:25:    struct pneigh_entry [noderef] __rcu *
> >   .../neighbour.c:832:25:    struct pneigh_entry *
> 
> Thanks for heads-up, Simon!
> 
> This diff below was needed on top of the series, but as I gradually added
> rcu_derefernece_check(), probably I need to churn this patch 6 more.
> 
> Anyway, I'll fix every annotation warning in v2.

Thanks Iwashima-san,

This approach looks good to me.
And, likewise, v2 looks good wrt annotation warnings.