Skip to contentSkip to conversation

neighbour: Convert RTM_GETNEIGHTBL and RTM_SETNEIGHTBL to RCU.

PATCH Applied Kuniyuki Iwashima 3 participants 14 messages Oct 22, 2025 – Oct 25, 2025

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

cover neighbour: Convert RTM_GETNEIGHTBL and RTM_SETNEIGHTBL to RCU.

Patch 1 & 2 are prep for RCU conversion for RTM_GETNEIGHTBL.

Patch 3 & 4 converts RTM_GETNEIGHTBL and RTM_SETNEIGHTBL to RCU.

Patch 5 converts the neighbour table rwlock to the plain spinlock.

Kuniyuki Iwashima (5):
neighbour: Use RCU list helpers for neigh_parms.list writers.
neighbour: Annotate access to neigh_parms fields.
neighbour: Convert RTM_GETNEIGHTBL to RCU.
neighbour: Convert RTM_SETNEIGHTBL to RCU.
neighbour: Convert rwlock of struct neigh_table to spinlock.

 include/net/neighbour.h |  17 ++++--
 net/atm/clip.c          |   4 +-
 net/core/neighbour.c    | 131 ++++++++++++++++++++--------------------
 net/ipv4/arp.c          |   4 +-
 net/ipv6/ndisc.c        |   8 +--
 5 files changed, 87 insertions(+), 77 deletions(-)
Signature · 1 line
2.51.0.915.g61a8936c21-goog
#1 lore raw

PATCH 1/5 neighbour: Use RCU list helpers for neigh_parms.list writers.

We will convert RTM_GETNEIGHTBL to RCU soon, where we traverse tbl->parms_list under RCU in neightbl_dump_info().

Let's use RCU list helper for neigh_parms in neigh_parms_alloc() and neigh_parms_release().

neigh_table_init() uses the plain list_add() for the default neigh_parm that is embedded in the table and not yet published.

Note that neigh_parms_release() already uses call_rcu() to free neigh_parms.

---
 net/core/neighbour.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
net/core/neighbour.c +2 -2
index dfa997c996896..56ce01db1bcb4 100644
@@ -1764,7 +1764,7 @@ struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
 		}
 
 		write_lock_bh(&tbl->lock);
-		list_add(&p->list, &tbl->parms.list);
+		list_add_rcu(&p->list, &tbl->parms.list);
 		write_unlock_bh(&tbl->lock);
 
 		neigh_parms_data_state_cleanall(p);
@@ -1786,7 +1786,7 @@ void neigh_parms_release(struct neigh_table *tbl, struct neigh_parms *parms)
 	if (!parms || parms == &tbl->parms)
 		return;
 	write_lock_bh(&tbl->lock);
-	list_del(&parms->list);
+	list_del_rcu(&parms->list);
 	parms->dead = 1;
 	write_unlock_bh(&tbl->lock);
 	netdev_put(parms->dev, &parms->dev_tracker);
Signature · 1 line
2.51.0.915.g61a8936c21-goog
Reviewed-by on v1 1/5
#2 lore raw

Re: [PATCH v1 net-next 1/5] neighbour: Use RCU list helpers for neigh_parms.list writers.

Quoted 7 lines · Kuniyuki Iwashima

We will convert RTM_GETNEIGHTBL to RCU soon, where we traverse tbl->parms_list under RCU in neightbl_dump_info().

Let's use RCU list helper for neigh_parms in neigh_parms_alloc() and neigh_parms_release().

Quoted 7 lines

neigh_table_init() uses the plain list_add() for the default neigh_parm that is embedded in the table and not yet published.

Note that neigh_parms_release() already uses call_rcu() to free neigh_parms.

#3 lore raw

PATCH 2/5 neighbour: Annotate access to neigh_parms fields.

In reply to Kuniyuki Iwashima

NEIGH_VAR() is read locklessly in the fast path, and IPv6 ndisc uses NEIGH_VAR_SET() locklessly.

The next patch will convert neightbl_dump_info() to RCU.

Let's annotate accesses to neigh_param with READ_ONCE() and WRITE_ONCE().

Note that ndisc_ifinfo_sysctl_change() uses &NEIGH_VAR() and we cannot use '&' with READ_ONCE(), so NEIGH_VAR_PTR() is introduced.

Note also that NEIGH_VAR_INIT() does not need WRITE_ONCE() as it is before parms is published. Also, the only user hippi_neigh_setup_dev() is no longer called since commit e3804cbebb67 ("net: remove COMPAT_NET_DEV_OPS"), which looks wrong, but probably no one uses HIPPI and RoadRunner.

---
 include/net/neighbour.h | 15 ++++++++++++---
 net/core/neighbour.c    | 17 ++++++-----------
 net/ipv6/ndisc.c        |  8 ++++----
 3 files changed, 22 insertions(+), 18 deletions(-)
include/net/neighbour.h +12 -3
index 4a30bd458c5a9..998ff9eccebb7 100644
@@ -92,15 +92,17 @@ struct neigh_parms {
 static inline void neigh_var_set(struct neigh_parms *p, int index, int val)
 {
 	set_bit(index, p->data_state);
-	p->data[index] = val;
+	WRITE_ONCE(p->data[index], val);
 }
 
-#define NEIGH_VAR(p, attr) ((p)->data[NEIGH_VAR_ ## attr])
+#define __NEIGH_VAR(p, attr) ((p)->data[NEIGH_VAR_ ## attr])
+#define NEIGH_VAR(p, attr) READ_ONCE(__NEIGH_VAR(p, attr))
+#define NEIGH_VAR_PTR(p, attr) (&(__NEIGH_VAR(p, attr)))
 
 /* In ndo_neigh_setup, NEIGH_VAR_INIT should be used.
  * In other cases, NEIGH_VAR_SET should be used.
  */
-#define NEIGH_VAR_INIT(p, attr, val) (NEIGH_VAR(p, attr) = val)
+#define NEIGH_VAR_INIT(p, attr, val) (__NEIGH_VAR(p, attr) = val)
 #define NEIGH_VAR_SET(p, attr, val) neigh_var_set(p, NEIGH_VAR_ ## attr, val)
 
 static inline void neigh_parms_data_state_setall(struct neigh_parms *p)
@@ -378,6 +380,13 @@ struct net *neigh_parms_net(const struct neigh_parms *parms)
 
 unsigned long neigh_rand_reach_time(unsigned long base);
 
+static inline void neigh_set_reach_time(struct neigh_parms *p)
+{
+	unsigned long base = NEIGH_VAR(p, BASE_REACHABLE_TIME);
+
+	WRITE_ONCE(p->reachable_time, neigh_rand_reach_time(base));
+}
+
 void pneigh_enqueue(struct neigh_table *tbl, struct neigh_parms *p,
 		    struct sk_buff *skb);
 struct pneigh_entry *pneigh_lookup(struct neigh_table *tbl, struct net *net,
net/core/neighbour.c +6 -11
index 56ce01db1bcb4..b08a9df31a15b 100644
@@ -995,8 +995,7 @@ static void neigh_periodic_work(struct work_struct *work)
 
 		WRITE_ONCE(tbl->last_rand, jiffies);
 		list_for_each_entry(p, &tbl->parms_list, list)
-			p->reachable_time =
-				neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
+			neigh_set_reach_time(p);
 	}
 
 	if (atomic_read(&tbl->entries) < READ_ONCE(tbl->gc_thresh1))
@@ -1749,8 +1748,7 @@ struct neigh_parms *neigh_parms_alloc(struct net_device *dev,
 	if (p) {
 		p->tbl		  = tbl;
 		refcount_set(&p->refcnt, 1);
-		p->reachable_time =
-				neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
+		neigh_set_reach_time(p);
 		p->qlen = 0;
 		netdev_hold(dev, &p->dev_tracker, GFP_KERNEL);
 		p->dev = dev;
@@ -1810,8 +1808,7 @@ void neigh_table_init(int index, struct neigh_table *tbl)
 	list_add(&tbl->parms.list, &tbl->parms_list);
 	write_pnet(&tbl->parms.net, &init_net);
 	refcount_set(&tbl->parms.refcnt, 1);
-	tbl->parms.reachable_time =
-			  neigh_rand_reach_time(NEIGH_VAR(&tbl->parms, BASE_REACHABLE_TIME));
+	neigh_set_reach_time(&tbl->parms);
 	tbl->parms.qlen = 0;
 
 	tbl->stats = alloc_percpu(struct neigh_statistics);
@@ -2194,7 +2191,7 @@ static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
 			NEIGH_VAR(parms, MCAST_PROBES)) ||
 	    nla_put_u32(skb, NDTPA_MCAST_REPROBES,
 			NEIGH_VAR(parms, MCAST_REPROBES)) ||
-	    nla_put_msecs(skb, NDTPA_REACHABLE_TIME, parms->reachable_time,
+	    nla_put_msecs(skb, NDTPA_REACHABLE_TIME, READ_ONCE(parms->reachable_time),
 			  NDTPA_PAD) ||
 	    nla_put_msecs(skb, NDTPA_BASE_REACHABLE_TIME,
 			  NEIGH_VAR(parms, BASE_REACHABLE_TIME), NDTPA_PAD) ||
@@ -2475,8 +2472,7 @@ static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh,
 				 * only be effective after the next time neigh_periodic_work
 				 * decides to recompute it (can be multiple minutes)
 				 */
-				p->reachable_time =
-					neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
+				neigh_set_reach_time(p);
 				break;
 			case NDTPA_GC_STALETIME:
 				NEIGH_VAR_SET(p, GC_STALETIME,
@@ -3721,8 +3717,7 @@ static int neigh_proc_base_reachable_time(const struct ctl_table *ctl, int write
 		 * only be effective after the next time neigh_periodic_work
 		 * decides to recompute it
 		 */
-		p->reachable_time =
-			neigh_rand_reach_time(NEIGH_VAR(p, BASE_REACHABLE_TIME));
+		neigh_set_reach_time(p);
 	}
 	return ret;
 }
net/ipv6/ndisc.c +4 -4
index f427e41e9c49b..59d17b6f06bfd 100644
@@ -1449,7 +1449,7 @@ static enum skb_drop_reason ndisc_router_discovery(struct sk_buff *skb)
 					      BASE_REACHABLE_TIME, rtime);
 				NEIGH_VAR_SET(in6_dev->nd_parms,
 					      GC_STALETIME, 3 * rtime);
-				in6_dev->nd_parms->reachable_time = neigh_rand_reach_time(rtime);
+				neigh_set_reach_time(in6_dev->nd_parms);
 				in6_dev->tstamp = jiffies;
 				send_ifinfo_notify = true;
 			}
@@ -1948,9 +1948,9 @@ int ndisc_ifinfo_sysctl_change(const struct ctl_table *ctl, int write, void *buf
 		ret = -1;
 
 	if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) {
-		if (ctl->data == &NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME))
-			idev->nd_parms->reachable_time =
-					neigh_rand_reach_time(NEIGH_VAR(idev->nd_parms, BASE_REACHABLE_TIME));
+		if (ctl->data == NEIGH_VAR_PTR(idev->nd_parms, BASE_REACHABLE_TIME))
+			neigh_set_reach_time(idev->nd_parms);
+
 		WRITE_ONCE(idev->tstamp, jiffies);
 		inet6_ifinfo_notify(RTM_NEWLINK, idev);
 		in6_dev_put(idev);
Signature · 1 line
2.51.0.915.g61a8936c21-goog
Reviewed-by on v1 2/5
#4 lore raw

Re: [PATCH v1 net-next 2/5] neighbour: Annotate access to neigh_parms fields.

Quoted 11 lines · Kuniyuki Iwashima

NEIGH_VAR() is read locklessly in the fast path, and IPv6 ndisc uses NEIGH_VAR_SET() locklessly.

The next patch will convert neightbl_dump_info() to RCU.

Let's annotate accesses to neigh_param with READ_ONCE() and WRITE_ONCE().

Note that ndisc_ifinfo_sysctl_change() uses &NEIGH_VAR() and we cannot use '&' with READ_ONCE(), so NEIGH_VAR_PTR() is introduced.

Quoted 6 lines

Note also that NEIGH_VAR_INIT() does not need WRITE_ONCE() as it is before parms is published. Also, the only user hippi_neigh_setup_dev() is no longer called since commit e3804cbebb67 ("net: remove COMPAT_NET_DEV_OPS"), which looks wrong, but probably no one uses HIPPI and RoadRunner.

#5 lore raw

PATCH 3/5 neighbour: Convert RTM_GETNEIGHTBL to RCU.

In reply to Kuniyuki Iwashima

neightbl_dump_info() calls these functions for each neigh_tables[] entry:

1. neightbl_fill_info() for tbl->parms
2. neightbl_fill_param_info() for tbl->parms_list (except tbl->parms)

Both functions rely on the table lock (read_lock_bh(&tbl->lock)) and RTNL is not needed.

Let's fetch the table under RCU and convert RTM_GETNEIGHTBL to RCU.

Note that the first entry of tbl->parms_list is tbl->parms.list and embedded in neigh_table, so list_next_entry() is safe.

---
 net/core/neighbour.c | 23 +++++++++--------------
 1 file changed, 9 insertions(+), 14 deletions(-)
net/core/neighbour.c +9 -14
index b08a9df31a15b..0c5170438c51e 100644
@@ -2176,7 +2176,7 @@ static int neightbl_fill_parms(struct sk_buff *skb, struct neigh_parms *parms)
 		return -ENOBUFS;
 
 	if ((parms->dev &&
-	     nla_put_u32(skb, NDTPA_IFINDEX, parms->dev->ifindex)) ||
+	     nla_put_u32(skb, NDTPA_IFINDEX, READ_ONCE(parms->dev->ifindex))) ||
 	    nla_put_u32(skb, NDTPA_REFCNT, refcount_read(&parms->refcnt)) ||
 	    nla_put_u32(skb, NDTPA_QUEUE_LENBYTES,
 			NEIGH_VAR(parms, QUEUE_LEN_BYTES)) ||
@@ -2228,8 +2228,6 @@ static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
 		return -EMSGSIZE;
 
 	ndtmsg = nlmsg_data(nlh);
-
-	read_lock_bh(&tbl->lock);
 	ndtmsg->ndtm_family = tbl->family;
 	ndtmsg->ndtm_pad1   = 0;
 	ndtmsg->ndtm_pad2   = 0;
@@ -2255,11 +2253,9 @@ static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
 			.ndtc_proxy_qlen	= READ_ONCE(tbl->proxy_queue.qlen),
 		};
 
-		rcu_read_lock();
 		nht = rcu_dereference(tbl->nht);
 		ndc.ndtc_hash_rnd = nht->hash_rnd[0];
 		ndc.ndtc_hash_mask = ((1 << nht->hash_shift) - 1);
-		rcu_read_unlock();
 
 		if (nla_put(skb, NDTA_CONFIG, sizeof(ndc), &ndc))
 			goto nla_put_failure;
@@ -2297,12 +2293,10 @@ static int neightbl_fill_info(struct sk_buff *skb, struct neigh_table *tbl,
 	if (neightbl_fill_parms(skb, &tbl->parms) < 0)
 		goto nla_put_failure;
 
-	read_unlock_bh(&tbl->lock);
 	nlmsg_end(skb, nlh);
 	return 0;
 
 nla_put_failure:
-	read_unlock_bh(&tbl->lock);
 	nlmsg_cancel(skb, nlh);
 	return -EMSGSIZE;
 }
@@ -2321,8 +2315,6 @@ static int neightbl_fill_param_info(struct sk_buff *skb,
 		return -EMSGSIZE;
 
 	ndtmsg = nlmsg_data(nlh);
-
-	read_lock_bh(&tbl->lock);
 	ndtmsg->ndtm_family = tbl->family;
 	ndtmsg->ndtm_pad1   = 0;
 	ndtmsg->ndtm_pad2   = 0;
@@ -2331,11 +2323,9 @@ static int neightbl_fill_param_info(struct sk_buff *skb,
 	    neightbl_fill_parms(skb, parms) < 0)
 		goto errout;
 
-	read_unlock_bh(&tbl->lock);
 	nlmsg_end(skb, nlh);
 	return 0;
 errout:
-	read_unlock_bh(&tbl->lock);
 	nlmsg_cancel(skb, nlh);
 	return -EMSGSIZE;
 }
@@ -2575,10 +2565,12 @@ static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
 
 	family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
 
+	rcu_read_lock();
+
 	for (tidx = 0; tidx < NEIGH_NR_TABLES; tidx++) {
 		struct neigh_parms *p;
 
-		tbl = rcu_dereference_rtnl(neigh_tables[tidx]);
+		tbl = rcu_dereference(neigh_tables[tidx]);
 		if (!tbl)
 			continue;
 
@@ -2592,7 +2584,7 @@ static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
 
 		nidx = 0;
 		p = list_next_entry(&tbl->parms, list);
-		list_for_each_entry_from(p, &tbl->parms_list, list) {
+		list_for_each_entry_from_rcu(p, &tbl->parms_list, list) {
 			if (!net_eq(neigh_parms_net(p), net))
 				continue;
 
@@ -2612,6 +2604,8 @@ static int neightbl_dump_info(struct sk_buff *skb, struct netlink_callback *cb)
 		neigh_skip = 0;
 	}
 out:
+	rcu_read_unlock();
+
 	cb->args[0] = tidx;
 	cb->args[1] = nidx;
 
@@ -3913,7 +3907,8 @@ static const struct rtnl_msg_handler neigh_rtnl_msg_handlers[] __initconst = {
 	{.msgtype = RTM_DELNEIGH, .doit = neigh_delete},
 	{.msgtype = RTM_GETNEIGH, .doit = neigh_get, .dumpit = neigh_dump_info,
 	 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
-	{.msgtype = RTM_GETNEIGHTBL, .dumpit = neightbl_dump_info},
+	{.msgtype = RTM_GETNEIGHTBL, .dumpit = neightbl_dump_info,
+	 .flags = RTNL_FLAG_DUMP_UNLOCKED},
 	{.msgtype = RTM_SETNEIGHTBL, .doit = neightbl_set},
 };
 
Signature · 1 line
2.51.0.915.g61a8936c21-goog
Reviewed-by on v1 3/5
#6 lore raw

Re: [PATCH v1 net-next 3/5] neighbour: Convert RTM_GETNEIGHTBL to RCU.

Quoted 10 lines · Kuniyuki Iwashima

neightbl_dump_info() calls these functions for each neigh_tables[] entry:

1. neightbl_fill_info() for tbl->parms
2. neightbl_fill_param_info() for tbl->parms_list (except tbl->parms)

Both functions rely on the table lock (read_lock_bh(&tbl->lock)) and RTNL is not needed.

Quoted 7 lines

Let's fetch the table under RCU and convert RTM_GETNEIGHTBL to RCU.

Note that the first entry of tbl->parms_list is tbl->parms.list and embedded in neigh_table, so list_next_entry() is safe.

---

#7 lore raw

PATCH 4/5 neighbour: Convert RTM_SETNEIGHTBL to RCU.

In reply to Kuniyuki Iwashima

neightbl_set() fetches neigh_tables[] and updates attributes under write_lock_bh(&tbl->lock), so RTNL is not needed.

neigh_table_clear() synchronises RCU only, and rcu_dereference_rtnl() protects nothing here.

If we released RCU after fetching neigh_tables[], there would be no synchronisation to block neigh_table_clear() further, so RCU is held until the end of the function.

Another option would be to protect neigh_tables[] user with SRCU and add synchronize_srcu() in neigh_table_clear().

But, holding RCU should be fine as we hold write_lock_bh() for the rest of neightbl_set() anyway.

Let's perform RTM_SETNEIGHTBL under RCU and drop RTNL.

---
 net/core/neighbour.c | 19 ++++++++++++++-----
 1 file changed, 14 insertions(+), 5 deletions(-)
net/core/neighbour.c +14 -5
index 0c5170438c51e..807a0d2457728 100644
@@ -2362,9 +2362,9 @@ static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh,
 			struct netlink_ext_ack *extack)
 {
 	struct net *net = sock_net(skb->sk);
+	struct nlattr *tb[NDTA_MAX + 1];
 	struct neigh_table *tbl;
 	struct ndtmsg *ndtmsg;
-	struct nlattr *tb[NDTA_MAX+1];
 	bool found = false;
 	int err, tidx;
 
@@ -2380,20 +2380,27 @@ static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh,
 
 	ndtmsg = nlmsg_data(nlh);
 
+	rcu_read_lock();
+
 	for (tidx = 0; tidx < NEIGH_NR_TABLES; tidx++) {
-		tbl = rcu_dereference_rtnl(neigh_tables[tidx]);
+		tbl = rcu_dereference(neigh_tables[tidx]);
 		if (!tbl)
 			continue;
+
 		if (ndtmsg->ndtm_family && tbl->family != ndtmsg->ndtm_family)
 			continue;
+
 		if (nla_strcmp(tb[NDTA_NAME], tbl->id) == 0) {
 			found = true;
 			break;
 		}
 	}
 
-	if (!found)
-		return -ENOENT;
+	if (!found) {
+		rcu_read_unlock();
+		err = -ENOENT;
+		goto errout;
+	}
 
 	/*
 	 * We acquire tbl->lock to be nice to the periodic timers and
@@ -2519,6 +2526,7 @@ static int neightbl_set(struct sk_buff *skb, struct nlmsghdr *nlh,
 
 errout_tbl_lock:
 	write_unlock_bh(&tbl->lock);
+	rcu_read_unlock();
 errout:
 	return err;
 }
@@ -3909,7 +3917,8 @@ static const struct rtnl_msg_handler neigh_rtnl_msg_handlers[] __initconst = {
 	 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
 	{.msgtype = RTM_GETNEIGHTBL, .dumpit = neightbl_dump_info,
 	 .flags = RTNL_FLAG_DUMP_UNLOCKED},
-	{.msgtype = RTM_SETNEIGHTBL, .doit = neightbl_set},
+	{.msgtype = RTM_SETNEIGHTBL, .doit = neightbl_set,
+	 .flags = RTNL_FLAG_DOIT_UNLOCKED},
 };
 
 static int __init neigh_init(void)
Signature · 1 line
2.51.0.915.g61a8936c21-goog
Reviewed-by on v1 4/5
#8 lore raw

Re: [PATCH v1 net-next 4/5] neighbour: Convert RTM_SETNEIGHTBL to RCU.

Quoted 14 lines · Kuniyuki Iwashima

neightbl_set() fetches neigh_tables[] and updates attributes under write_lock_bh(&tbl->lock), so RTNL is not needed.

neigh_table_clear() synchronises RCU only, and rcu_dereference_rtnl() protects nothing here.

If we released RCU after fetching neigh_tables[], there would be no synchronisation to block neigh_table_clear() further, so RCU is held until the end of the function.

Another option would be to protect neigh_tables[] user with SRCU and add synchronize_srcu() in neigh_table_clear().

Quoted 7 lines

But, holding RCU should be fine as we hold write_lock_bh() for the rest of neightbl_set() anyway.

Let's perform RTM_SETNEIGHTBL under RCU and drop RTNL.

---

#9 lore raw

PATCH 5/5 neighbour: Convert rwlock of struct neigh_table to spinlock.

In reply to Kuniyuki Iwashima

Only neigh_for_each() and neigh_seq_start/stop() are on the reader side of neigh_table.lock.

Let's convert rwlock to the plain spinlock.

---
 include/net/neighbour.h |  2 +-
 net/atm/clip.c          |  4 +--
 net/core/neighbour.c    | 68 +++++++++++++++++++++--------------------
 net/ipv4/arp.c          |  4 +--
 4 files changed, 40 insertions(+), 38 deletions(-)
include/net/neighbour.h +1 -1
index 998ff9eccebb7..2dfee6d4258af 100644
@@ -238,7 +238,7 @@ struct neigh_table {
 	atomic_t		gc_entries;
 	struct list_head	gc_list;
 	struct list_head	managed_list;
-	rwlock_t		lock;
+	spinlock_t		lock;
 	unsigned long		last_rand;
 	struct neigh_statistics	__percpu *stats;
 	struct neigh_hash_table __rcu *nht;
net/atm/clip.c +2 -2
index f7a5565e794ef..8f152e5fa6594 100644
@@ -168,10 +168,10 @@ static int neigh_check_cb(struct neighbour *n)
 
 static void idle_timer_check(struct timer_list *unused)
 {
-	write_lock(&arp_tbl.lock);
+	spin_lock(&arp_tbl.lock);
 	__neigh_for_each_release(&arp_tbl, neigh_check_cb);
 	mod_timer(&idle_timer, jiffies + CLIP_CHECK_INTERVAL * HZ);
-	write_unlock(&arp_tbl.lock);
+	spin_unlock(&arp_tbl.lock);
 }
 
 static int clip_arp_rcv(struct sk_buff *skb)
net/core/neighbour.c +35 -33

240 lines not on this page · raw message

net/ipv4/arp.c +2 -2
index 833f2cf97178e..f3bfecf8a2341 100644
@@ -1217,10 +1217,10 @@ int arp_invalidate(struct net_device *dev, __be32 ip, bool force)
 			err = neigh_update(neigh, NULL, NUD_FAILED,
 					   NEIGH_UPDATE_F_OVERRIDE|
 					   NEIGH_UPDATE_F_ADMIN, 0);
-		write_lock_bh(&tbl->lock);
+		spin_lock_bh(&tbl->lock);
 		neigh_release(neigh);
 		neigh_remove_one(neigh);
-		write_unlock_bh(&tbl->lock);
+		spin_unlock_bh(&tbl->lock);
 	}
 
 	return err;
Signature · 1 line
2.51.0.915.g61a8936c21-goog
#10 lore raw

Re: [PATCH v1 net-next 5/5] neighbour: Convert rwlock of struct neigh_table to spinlock.

Quoted 8 lines · Kuniyuki Iwashima

Only neigh_for_each() and neigh_seq_start/stop() are on the reader side of neigh_table.lock.

Let's convert rwlock to the plain spinlock.

Do we still need _bh prefix ?

#11 lore raw

Re: [PATCH v1 net-next 5/5] neighbour: Convert rwlock of struct neigh_table to spinlock.

Quoted 6 lines · Eric Dumazet
Quoted 4 lines · Kuniyuki Iwashima

Only neigh_for_each() and neigh_seq_start/stop() are on the reader side of neigh_table.lock.

Quoted 6 lines
Quoted 4 lines

Let's convert rwlock to the plain spinlock.

Do we still need _bh prefix ?

Yes, I think _bh is just for IPv6 ndisc calling neigh_update().

Reviewed-by
#12 lore raw

Re: [PATCH v1 net-next 5/5] neighbour: Convert rwlock of struct neigh_table to spinlock.

Quoted 8 lines · Kuniyuki Iwashima
Quoted 6 lines · Eric Dumazet
Quoted 4 lines · Kuniyuki Iwashima

Only neigh_for_each() and neigh_seq_start/stop() are on the reader side of neigh_table.lock.

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

Let's convert rwlock to the plain spinlock.

Do we still need _bh prefix ?

Yes, I think _bh is just for IPv6 ndisc calling neigh_update().

OK, this is presumably to protect tbl->gc_list.

This list could use a separate spinlock, given neigh_forced_gc() can spin for up to 1ms.

Thanks !

#13 lore raw
In reply to Kuniyuki Iwashima
This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Quoted 8 lines · On Wed, 22 Oct 2025 05:39:44 +0000 you wrote:

Patch 1 & 2 are prep for RCU conversion for RTM_GETNEIGHTBL.

Patch 3 & 4 converts RTM_GETNEIGHTBL and RTM_SETNEIGHTBL to RCU.

Patch 5 converts the neighbour table rwlock to the plain spinlock.

[...]

Here is the summary with links:
- [v1,net-next,1/5] neighbour: Use RCU list helpers for neigh_parms.list writers.

    https://git.kernel.org/netdev/net-next/c/06d6322280d9
  - [v1,net-next,2/5] neighbour: Annotate access to neigh_parms fields.
    https://git.kernel.org/netdev/net-next/c/35d7c7087033
  - [v1,net-next,3/5] neighbour: Convert RTM_GETNEIGHTBL to RCU.
    https://git.kernel.org/netdev/net-next/c/4ae34be50064
  - [v1,net-next,4/5] neighbour: Convert RTM_SETNEIGHTBL to RCU.
    https://git.kernel.org/netdev/net-next/c/55a6046b48a8
  - [v1,net-next,5/5] neighbour: Convert rwlock of struct neigh_table to spinlock.
    https://git.kernel.org/netdev/net-next/c/3064d0fe02af

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html