* [PATCH net-next 0/2] geneve: make geneve_fill_info() RTNL-less
@ 2026-07-01 12:04 Eric Dumazet
2026-07-01 12:04 ` [PATCH net-next 1/2] geneve: convert config to RCU-protected pointer Eric Dumazet
2026-07-01 12:04 ` [PATCH net-next 2/2] geneve: make geneve_fill_info() RTNL independent Eric Dumazet
0 siblings, 2 replies; 3+ messages in thread
From: Eric Dumazet @ 2026-07-01 12:04 UTC (permalink / raw)
To: David S . Miller, Jakub Kicinski, Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Andrew Lunn, netdev,
eric.dumazet, Eric Dumazet
This series makes geneve_fill_info() independent of the RTNL lock by
converting the device configuration to an RCU-protected pointer.
Historically, geneve_changelink() updated the device configuration by
copying the new configuration over the old one using memcpy() under RTNL.
To prevent the transmit/receive data paths from reading torn values during
the copy, geneve_quiesce() was used to pause the data path and wait for
a synchronize_net(), causing packet loss and latency.
By converting the configuration to an RCU-protected pointer, we can
perform atomic updates via RCU swap. This allows data path readers to
safely access the configuration locklessly under RCU read lock, and
removes the need to stop the data path during changelink.
With the RCU infrastructure in place, geneve_fill_info() is then updated
to read the configuration under RCU read lock, removing its dependency
on RTNL.
Eric Dumazet (2):
geneve: convert config to RCU-protected pointer
geneve: make geneve_fill_info() RTNL independent
drivers/net/geneve.c | 390 +++++++++++++++++++++++++------------------
1 file changed, 229 insertions(+), 161 deletions(-)
--
2.55.0.rc0.799.gd6f94ed593-goog
^ permalink raw reply [flat|nested] 3+ messages in thread
* [PATCH net-next 1/2] geneve: convert config to RCU-protected pointer
2026-07-01 12:04 [PATCH net-next 0/2] geneve: make geneve_fill_info() RTNL-less Eric Dumazet
@ 2026-07-01 12:04 ` Eric Dumazet
2026-07-01 12:04 ` [PATCH net-next 2/2] geneve: make geneve_fill_info() RTNL independent Eric Dumazet
1 sibling, 0 replies; 3+ messages in thread
From: Eric Dumazet @ 2026-07-01 12:04 UTC (permalink / raw)
To: David S . Miller, Jakub Kicinski, Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Andrew Lunn, netdev,
eric.dumazet, Eric Dumazet
geneve_changelink() currently updates configuration by copying it over
the old one using memcpy() under RTNL, forcing data path pause via
geneve_quiesce() and synchronize_net() to avoid reading torn values.
Convert geneve->cfg to an RCU-protected pointer, allowing lockless
and safe reads under RCU read lock without synchronization overhead.
Key changes:
- Introduced geneve_config_alloc/free() helpers for lifecycle.
- geneve_configure() allocates config and publishes it via RCU.
- geneve_changelink() performs RCU swap; old config is freed via call_rcu_hurry().
- Allocates new dst_cache during changelink to prevent pcpu sharing.
- Removed geneve_quiesce/unquiesce() and synchronize_net() from changelink.
- Added rcu_barrier() to module exit to wait for pending callbacks.
- Updated data path to use rcu_dereference().
- Updated geneve_fill_info() to use rtnl_dereference() for now.
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
drivers/net/geneve.c | 366 +++++++++++++++++++++++++------------------
1 file changed, 210 insertions(+), 156 deletions(-)
diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c
index 396e1a113cd49e9a43bfe6c7cd78f15ca8963906..c777c1a1fa930d74768a659e16380d719f12a6b2 100644
--- a/drivers/net/geneve.c
+++ b/drivers/net/geneve.c
@@ -82,6 +82,7 @@ struct geneve_config {
u16 port_min;
u16 port_max;
+ struct rcu_head rcu;
/* Must be last --ends in a flexible-array member. */
struct ip_tunnel_info info;
};
@@ -100,7 +101,7 @@ struct geneve_dev {
#endif
struct list_head next; /* geneve's per namespace list */
struct gro_cells gro_cells;
- struct geneve_config cfg;
+ struct geneve_config __rcu *cfg;
};
struct geneve_sock {
@@ -182,8 +183,10 @@ static struct geneve_dev *geneve_lookup(struct geneve_sock *gs,
hash = geneve_net_vni_hash(vni);
vni_list_head = &gs->vni_list[hash];
hlist_for_each_entry_rcu(node, vni_list_head, hlist) {
- if (eq_tun_id_and_vni((u8 *)&node->geneve->cfg.info.key.tun_id, vni) &&
- addr == node->geneve->cfg.info.key.u.ipv4.dst)
+ const struct geneve_config *cfg = rcu_dereference(node->geneve->cfg);
+
+ if (eq_tun_id_and_vni((u8 *)&cfg->info.key.tun_id, vni) &&
+ addr == cfg->info.key.u.ipv4.dst)
return node->geneve;
}
return NULL;
@@ -201,8 +204,10 @@ static struct geneve_dev *geneve6_lookup(struct geneve_sock *gs,
hash = geneve_net_vni_hash(vni);
vni_list_head = &gs->vni_list[hash];
hlist_for_each_entry_rcu(node, vni_list_head, hlist) {
- if (eq_tun_id_and_vni((u8 *)&node->geneve->cfg.info.key.tun_id, vni) &&
- ipv6_addr_equal(&addr6, &node->geneve->cfg.info.key.u.ipv6.dst))
+ const struct geneve_config *cfg = rcu_dereference(node->geneve->cfg);
+
+ if (eq_tun_id_and_vni((u8 *)&cfg->info.key.tun_id, vni) &&
+ ipv6_addr_equal(&addr6, &cfg->info.key.u.ipv6.dst))
return node->geneve;
}
return NULL;
@@ -386,11 +391,6 @@ static int geneve_init(struct net_device *dev)
if (err)
return err;
- err = dst_cache_init(&geneve->cfg.info.dst_cache, GFP_KERNEL);
- if (err) {
- gro_cells_destroy(&geneve->gro_cells);
- return err;
- }
netdev_lockdep_set_classes(dev);
return 0;
}
@@ -399,7 +399,6 @@ static void geneve_uninit(struct net_device *dev)
{
struct geneve_dev *geneve = netdev_priv(dev);
- dst_cache_destroy(&geneve->cfg.info.dst_cache);
gro_cells_destroy(&geneve->gro_cells);
}
@@ -675,10 +674,14 @@ static int geneve_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
inner_proto = geneveh->proto_type;
- if (unlikely((!geneve->cfg.inner_proto_inherit &&
- inner_proto != htons(ETH_P_TEB)))) {
- dev_dstats_rx_dropped(geneve->dev);
- goto drop;
+ {
+ const struct geneve_config *cfg = rcu_dereference(geneve->cfg);
+
+ if (unlikely(!cfg || (!cfg->inner_proto_inherit &&
+ inner_proto != htons(ETH_P_TEB)))) {
+ dev_dstats_rx_dropped(geneve->dev);
+ goto drop;
+ }
}
opts_len = geneveh->opt_len * 4;
@@ -762,9 +765,10 @@ static int geneve_udp_encap_err_lookup(struct sock *sk, struct sk_buff *skb)
}
static struct sock *geneve_create_sock(struct net *net,
- struct geneve_dev *geneve, bool ipv6)
+ struct geneve_dev *geneve,
+ const struct geneve_config *cfg, bool ipv6)
{
- struct ip_tunnel_info *info = &geneve->cfg.info;
+ const struct ip_tunnel_info *info = &cfg->info;
struct udp_port_cfg udp_conf;
struct socket *sock;
int err;
@@ -775,7 +779,7 @@ static struct sock *geneve_create_sock(struct net *net,
if (ipv6) {
udp_conf.family = AF_INET6;
udp_conf.ipv6_v6only = 1;
- udp_conf.use_udp6_rx_checksums = geneve->cfg.use_udp6_rx_checksums;
+ udp_conf.use_udp6_rx_checksums = cfg->use_udp6_rx_checksums;
udp_conf.local_ip6 = info->key.u.ipv6.src;
} else
#endif
@@ -991,7 +995,8 @@ static int geneve_gro_complete(struct sock *sk, struct sk_buff *skb,
/* Create new listen socket if needed */
static struct geneve_sock *geneve_socket_create(struct net *net,
- struct geneve_dev *geneve, bool ipv6)
+ struct geneve_dev *geneve,
+ const struct geneve_config *cfg, bool ipv6)
{
struct geneve_net *gn = net_generic(net, geneve_net_id);
struct udp_tunnel_sock_cfg tunnel_cfg;
@@ -1003,7 +1008,7 @@ static struct geneve_sock *geneve_socket_create(struct net *net,
if (!gs)
return ERR_PTR(-ENOMEM);
- sk = geneve_create_sock(net, geneve, ipv6);
+ sk = geneve_create_sock(net, geneve, cfg, ipv6);
if (IS_ERR(sk)) {
kfree(gs);
return ERR_CAST(sk);
@@ -1060,12 +1065,13 @@ static void geneve_sock_release(struct geneve_dev *geneve)
}
static struct geneve_sock *geneve_find_sock(struct net *net,
- struct geneve_dev *geneve, bool ipv6)
+ struct geneve_dev *geneve,
+ const struct geneve_config *cfg, bool ipv6)
{
struct geneve_net *gn = net_generic(net, geneve_net_id);
- struct ip_tunnel_info *info = &geneve->cfg.info;
+ const struct ip_tunnel_info *info = &cfg->info;
sa_family_t family = ipv6 ? AF_INET6 : AF_INET;
- bool gro_hint = geneve->cfg.gro_hint;
+ bool gro_hint = cfg->gro_hint;
__be16 dst_port = info->key.tp_dst;
struct geneve_sock *gs;
@@ -1095,7 +1101,8 @@ static struct geneve_sock *geneve_find_sock(struct net *net,
return NULL;
}
-static int geneve_sock_add(struct geneve_dev *geneve, bool ipv6)
+static int geneve_sock_add(struct geneve_dev *geneve,
+ const struct geneve_config *cfg, bool ipv6)
{
struct net *net = geneve->net;
struct geneve_dev_node *node;
@@ -1103,19 +1110,19 @@ static int geneve_sock_add(struct geneve_dev *geneve, bool ipv6)
__u8 vni[3];
__u32 hash;
- gs = geneve_find_sock(net, geneve, ipv6);
+ gs = geneve_find_sock(net, geneve, cfg, ipv6);
if (gs) {
gs->refcnt++;
goto out;
}
- gs = geneve_socket_create(net, geneve, ipv6);
+ gs = geneve_socket_create(net, geneve, cfg, ipv6);
if (IS_ERR(gs))
return PTR_ERR(gs);
out:
- gs->collect_md = geneve->cfg.collect_md;
- gs->gro_hint = geneve->cfg.gro_hint;
+ gs->collect_md = cfg->collect_md;
+ gs->gro_hint = cfg->gro_hint;
#if IS_ENABLED(CONFIG_IPV6)
if (ipv6) {
rcu_assign_pointer(geneve->sock6, gs);
@@ -1128,7 +1135,7 @@ static int geneve_sock_add(struct geneve_dev *geneve, bool ipv6)
}
node->geneve = geneve;
- tunnel_id_to_vni(geneve->cfg.info.key.tun_id, vni);
+ tunnel_id_to_vni(cfg->info.key.tun_id, vni);
hash = geneve_net_vni_hash(vni);
hlist_add_head_rcu(&node->hlist, &gs->vni_list[hash]);
return 0;
@@ -1137,21 +1144,22 @@ static int geneve_sock_add(struct geneve_dev *geneve, bool ipv6)
static int geneve_open(struct net_device *dev)
{
struct geneve_dev *geneve = netdev_priv(dev);
- bool dualstack = geneve->cfg.dualstack;
+ const struct geneve_config *cfg = rtnl_dereference(geneve->cfg);
+ bool dualstack = cfg->dualstack;
bool ipv4, ipv6;
int ret = 0;
- ipv6 = geneve->cfg.info.mode & IP_TUNNEL_INFO_IPV6 || dualstack;
+ ipv6 = cfg->info.mode & IP_TUNNEL_INFO_IPV6 || dualstack;
ipv4 = !ipv6 || dualstack;
#if IS_ENABLED(CONFIG_IPV6)
if (ipv6) {
- ret = geneve_sock_add(geneve, true);
+ ret = geneve_sock_add(geneve, cfg, true);
if (ret < 0 && ret != -EAFNOSUPPORT)
ipv4 = false;
}
#endif
if (ipv4)
- ret = geneve_sock_add(geneve, false);
+ ret = geneve_sock_add(geneve, cfg, false);
if (ret < 0)
geneve_sock_release(geneve);
@@ -1189,6 +1197,7 @@ static void geneve_build_header(struct genevehdr *geneveh,
}
static int geneve_build_gro_hint_opt(const struct geneve_dev *geneve,
+ const struct geneve_config *cfg,
struct sk_buff *skb)
{
struct geneve_skb_cb *cb = GENEVE_SKB_CB(skb);
@@ -1201,7 +1210,7 @@ static int geneve_build_gro_hint_opt(const struct geneve_dev *geneve,
cb->gro_hint_len = 0;
/* Try to add the GRO hint only in case of double encap. */
- if (!geneve->cfg.gro_hint || !skb->encapsulation)
+ if (!cfg->gro_hint || !skb->encapsulation)
return 0;
/*
@@ -1262,10 +1271,11 @@ static void geneve_put_gro_hint_opt(struct genevehdr *gnvh, int opt_size,
static int geneve_build_skb(struct dst_entry *dst, struct sk_buff *skb,
const struct ip_tunnel_info *info,
- const struct geneve_dev *geneve, int ip_hdr_len)
+ const struct geneve_dev *geneve,
+ const struct geneve_config *cfg, int ip_hdr_len)
{
bool udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags);
- bool inner_proto_inherit = geneve->cfg.inner_proto_inherit;
+ bool inner_proto_inherit = cfg->inner_proto_inherit;
bool xnet = !net_eq(geneve->net, dev_net(geneve->dev));
struct geneve_skb_cb *cb = GENEVE_SKB_CB(skb);
struct genevehdr *gnvh;
@@ -1306,14 +1316,14 @@ static int geneve_build_skb(struct dst_entry *dst, struct sk_buff *skb,
}
static u8 geneve_get_dsfield(struct sk_buff *skb, struct net_device *dev,
+ const struct geneve_config *cfg,
const struct ip_tunnel_info *info,
bool *use_cache)
{
- struct geneve_dev *geneve = netdev_priv(dev);
u8 dsfield;
dsfield = info->key.tos;
- if (dsfield == 1 && !geneve->cfg.collect_md) {
+ if (cfg && dsfield == 1 && !cfg->collect_md) {
dsfield = ip_tunnel_get_dsfield(ip_hdr(skb), skb);
*use_cache = false;
}
@@ -1323,6 +1333,7 @@ static u8 geneve_get_dsfield(struct sk_buff *skb, struct net_device *dev,
static int geneve_xmit_skb(struct sk_buff *skb, struct net_device *dev,
struct geneve_dev *geneve,
+ const struct geneve_config *cfg,
const struct ip_tunnel_info *info)
{
struct geneve_sock *gs4 = rcu_dereference(geneve->sock4);
@@ -1335,35 +1346,35 @@ static int geneve_xmit_skb(struct sk_buff *skb, struct net_device *dev,
__be16 sport;
int err;
- if (skb_vlan_inet_prepare(skb, geneve->cfg.inner_proto_inherit))
+ if (skb_vlan_inet_prepare(skb, cfg->inner_proto_inherit))
return -EINVAL;
if (!gs4)
return -EIO;
use_cache = ip_tunnel_dst_cache_usable(skb, info);
- tos = geneve_get_dsfield(skb, dev, info, &use_cache);
+ tos = geneve_get_dsfield(skb, dev, cfg, info, &use_cache);
sport = udp_flow_src_port(geneve->net, skb,
- geneve->cfg.port_min,
- geneve->cfg.port_max, true);
+ cfg->port_min,
+ cfg->port_max, true);
rt = udp_tunnel_dst_lookup(skb, dev, geneve->net, 0, &saddr,
&info->key,
- sport, geneve->cfg.info.key.tp_dst, tos,
+ sport, cfg->info.key.tp_dst, tos,
use_cache ?
(struct dst_cache *)&info->dst_cache : NULL);
if (IS_ERR(rt))
return PTR_ERR(rt);
- if (geneve->cfg.info.key.u.ipv4.src &&
- saddr != geneve->cfg.info.key.u.ipv4.src) {
+ if (cfg->info.key.u.ipv4.src &&
+ saddr != cfg->info.key.u.ipv4.src) {
dst_release(&rt->dst);
return -EADDRNOTAVAIL;
}
err = skb_tunnel_check_pmtu(skb, &rt->dst,
GENEVE_IPV4_HLEN + info->options_len +
- geneve_build_gro_hint_opt(geneve, skb),
+ geneve_build_gro_hint_opt(geneve, cfg, skb),
netif_is_any_bridge_port(dev));
if (err < 0) {
dst_release(&rt->dst);
@@ -1397,21 +1408,21 @@ static int geneve_xmit_skb(struct sk_buff *skb, struct net_device *dev,
}
tos = ip_tunnel_ecn_encap(tos, ip_hdr(skb), skb);
- if (geneve->cfg.collect_md) {
+ if (cfg->collect_md) {
ttl = key->ttl;
df = test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, key->tun_flags) ?
htons(IP_DF) : 0;
} else {
- if (geneve->cfg.ttl_inherit)
+ if (cfg->ttl_inherit)
ttl = ip_tunnel_get_ttl(ip_hdr(skb), skb);
else
ttl = key->ttl;
ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
- if (geneve->cfg.df == GENEVE_DF_SET) {
+ if (cfg->df == GENEVE_DF_SET) {
df = htons(IP_DF);
- } else if (geneve->cfg.df == GENEVE_DF_INHERIT) {
+ } else if (cfg->df == GENEVE_DF_INHERIT) {
struct ethhdr *eth = skb_eth_hdr(skb);
if (ntohs(eth->h_proto) == ETH_P_IPV6) {
@@ -1425,13 +1436,13 @@ static int geneve_xmit_skb(struct sk_buff *skb, struct net_device *dev,
}
}
- err = geneve_build_skb(&rt->dst, skb, info, geneve,
+ err = geneve_build_skb(&rt->dst, skb, info, geneve, cfg,
sizeof(struct iphdr));
if (unlikely(err))
return err;
udp_tunnel_xmit_skb(rt, gs4->sk, skb, saddr, info->key.u.ipv4.dst,
- tos, ttl, df, sport, geneve->cfg.info.key.tp_dst,
+ tos, ttl, df, sport, cfg->info.key.tp_dst,
!net_eq(geneve->net, dev_net(geneve->dev)),
!test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags),
0);
@@ -1441,6 +1452,7 @@ static int geneve_xmit_skb(struct sk_buff *skb, struct net_device *dev,
#if IS_ENABLED(CONFIG_IPV6)
static int geneve6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
struct geneve_dev *geneve,
+ const struct geneve_config *cfg,
const struct ip_tunnel_info *info)
{
struct geneve_sock *gs6 = rcu_dereference(geneve->sock6);
@@ -1452,35 +1464,35 @@ static int geneve6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
__be16 sport;
int err;
- if (skb_vlan_inet_prepare(skb, geneve->cfg.inner_proto_inherit))
+ if (skb_vlan_inet_prepare(skb, cfg->inner_proto_inherit))
return -EINVAL;
if (!gs6)
return -EIO;
use_cache = ip_tunnel_dst_cache_usable(skb, info);
- prio = geneve_get_dsfield(skb, dev, info, &use_cache);
+ prio = geneve_get_dsfield(skb, dev, cfg, info, &use_cache);
sport = udp_flow_src_port(geneve->net, skb,
- geneve->cfg.port_min,
- geneve->cfg.port_max, true);
+ cfg->port_min,
+ cfg->port_max, true);
dst = udp_tunnel6_dst_lookup(skb, dev, geneve->net, gs6->sk, 0,
&saddr, key, sport,
- geneve->cfg.info.key.tp_dst, prio,
+ cfg->info.key.tp_dst, prio,
use_cache ?
(struct dst_cache *)&info->dst_cache : NULL);
if (IS_ERR(dst))
return PTR_ERR(dst);
- if (!ipv6_addr_any(&geneve->cfg.info.key.u.ipv6.src) &&
- !ipv6_addr_equal(&saddr, &geneve->cfg.info.key.u.ipv6.src)) {
+ if (!ipv6_addr_any(&cfg->info.key.u.ipv6.src) &&
+ !ipv6_addr_equal(&saddr, &cfg->info.key.u.ipv6.src)) {
dst_release(dst);
return -EADDRNOTAVAIL;
}
err = skb_tunnel_check_pmtu(skb, dst,
GENEVE_IPV6_HLEN + info->options_len +
- geneve_build_gro_hint_opt(geneve, skb),
+ geneve_build_gro_hint_opt(geneve, cfg, skb),
netif_is_any_bridge_port(dev));
if (err < 0) {
dst_release(dst);
@@ -1513,22 +1525,22 @@ static int geneve6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
}
prio = ip_tunnel_ecn_encap(prio, ip_hdr(skb), skb);
- if (geneve->cfg.collect_md) {
+ if (cfg->collect_md) {
ttl = key->ttl;
} else {
- if (geneve->cfg.ttl_inherit)
+ if (cfg->ttl_inherit)
ttl = ip_tunnel_get_ttl(ip_hdr(skb), skb);
else
ttl = key->ttl;
ttl = ttl ? : ip6_dst_hoplimit(dst);
}
- err = geneve_build_skb(dst, skb, info, geneve, sizeof(struct ipv6hdr));
+ err = geneve_build_skb(dst, skb, info, geneve, cfg, sizeof(struct ipv6hdr));
if (unlikely(err))
return err;
udp_tunnel6_xmit_skb(dst, gs6->sk, skb, dev,
&saddr, &key->u.ipv6.dst, prio, ttl,
- info->key.label, sport, geneve->cfg.info.key.tp_dst,
+ info->key.label, sport, cfg->info.key.tp_dst,
!test_bit(IP_TUNNEL_CSUM_BIT,
info->key.tun_flags),
0);
@@ -1539,28 +1551,36 @@ static int geneve6_xmit_skb(struct sk_buff *skb, struct net_device *dev,
static netdev_tx_t geneve_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct geneve_dev *geneve = netdev_priv(dev);
- struct ip_tunnel_info *info = NULL;
+ const struct ip_tunnel_info *info = NULL;
+ const struct geneve_config *cfg;
int err;
- if (geneve->cfg.collect_md) {
+ rcu_read_lock();
+ cfg = rcu_dereference(geneve->cfg);
+ if (unlikely(!cfg)) {
+ err = -ENODEV;
+ goto tx_err;
+ }
+
+ if (cfg->collect_md) {
info = skb_tunnel_info(skb);
if (unlikely(!info || !(info->mode & IP_TUNNEL_INFO_TX))) {
netdev_dbg(dev, "no tunnel metadata\n");
dev_kfree_skb(skb);
dev_dstats_tx_dropped(dev);
+ rcu_read_unlock();
return NETDEV_TX_OK;
}
} else {
- info = &geneve->cfg.info;
+ info = &cfg->info;
}
- rcu_read_lock();
#if IS_ENABLED(CONFIG_IPV6)
if (info->mode & IP_TUNNEL_INFO_IPV6)
- err = geneve6_xmit_skb(skb, dev, geneve, info);
+ err = geneve6_xmit_skb(skb, dev, geneve, cfg, info);
else
#endif
- err = geneve_xmit_skb(skb, dev, geneve, info);
+ err = geneve_xmit_skb(skb, dev, geneve, cfg, info);
rcu_read_unlock();
if (likely(!err))
@@ -1576,6 +1596,12 @@ static netdev_tx_t geneve_xmit(struct sk_buff *skb, struct net_device *dev)
DEV_STATS_INC(dev, tx_errors);
return NETDEV_TX_OK;
+
+tx_err:
+ rcu_read_unlock();
+ dev_kfree_skb(skb);
+ DEV_STATS_INC(dev, tx_errors);
+ return NETDEV_TX_OK;
}
static int geneve_change_mtu(struct net_device *dev, int new_mtu)
@@ -1593,8 +1619,13 @@ static int geneve_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
{
struct ip_tunnel_info *info = skb_tunnel_info(skb);
struct geneve_dev *geneve = netdev_priv(dev);
+ const struct geneve_config *cfg;
__be16 sport;
+ cfg = rcu_dereference(geneve->cfg);
+ if (unlikely(!cfg))
+ return -ENODEV;
+
if (ip_tunnel_info_af(info) == AF_INET) {
struct rtable *rt;
struct geneve_sock *gs4 = rcu_dereference(geneve->sock4);
@@ -1606,14 +1637,14 @@ static int geneve_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
return -EIO;
use_cache = ip_tunnel_dst_cache_usable(skb, info);
- tos = geneve_get_dsfield(skb, dev, info, &use_cache);
+ tos = geneve_get_dsfield(skb, dev, cfg, info, &use_cache);
sport = udp_flow_src_port(geneve->net, skb,
- geneve->cfg.port_min,
- geneve->cfg.port_max, true);
+ cfg->port_min,
+ cfg->port_max, true);
rt = udp_tunnel_dst_lookup(skb, dev, geneve->net, 0, &saddr,
&info->key,
- sport, geneve->cfg.info.key.tp_dst,
+ sport, cfg->info.key.tp_dst,
tos,
use_cache ? &info->dst_cache : NULL);
if (IS_ERR(rt))
@@ -1633,14 +1664,14 @@ static int geneve_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
return -EIO;
use_cache = ip_tunnel_dst_cache_usable(skb, info);
- prio = geneve_get_dsfield(skb, dev, info, &use_cache);
+ prio = geneve_get_dsfield(skb, dev, cfg, info, &use_cache);
sport = udp_flow_src_port(geneve->net, skb,
- geneve->cfg.port_min,
- geneve->cfg.port_max, true);
+ cfg->port_min,
+ cfg->port_max, true);
dst = udp_tunnel6_dst_lookup(skb, dev, geneve->net, gs6->sk, 0,
&saddr, &info->key, sport,
- geneve->cfg.info.key.tp_dst, prio,
+ cfg->info.key.tp_dst, prio,
use_cache ? &info->dst_cache : NULL);
if (IS_ERR(dst))
return PTR_ERR(dst);
@@ -1653,7 +1684,7 @@ static int geneve_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
}
info->key.tp_src = sport;
- info->key.tp_dst = geneve->cfg.info.key.tp_dst;
+ info->key.tp_dst = cfg->info.key.tp_dst;
return 0;
}
@@ -1709,7 +1740,49 @@ static void geneve_offload_rx_ports(struct net_device *dev, bool push)
}
}
+static struct geneve_config *geneve_config_alloc(const struct geneve_config *src)
+{
+ struct geneve_config *cfg;
+ int err;
+
+ cfg = kmemdup(src, sizeof(*src), GFP_KERNEL);
+ if (!cfg)
+ return ERR_PTR(-ENOMEM);
+
+ cfg->info.dst_cache.cache = NULL;
+ err = dst_cache_init(&cfg->info.dst_cache, GFP_KERNEL);
+ if (err) {
+ kfree(cfg);
+ return ERR_PTR(err);
+ }
+
+ return cfg;
+}
+
+static void geneve_config_free(struct geneve_config *cfg)
+{
+ if (cfg) {
+ dst_cache_destroy(&cfg->info.dst_cache);
+ kfree(cfg);
+ }
+}
+
+static void geneve_config_free_rcu(struct rcu_head *head)
+{
+ struct geneve_config *cfg = container_of(head, struct geneve_config, rcu);
+
+ geneve_config_free(cfg);
+}
+
/* Initialize the device structure. */
+static void geneve_free_dev(struct net_device *dev)
+{
+ struct geneve_dev *geneve = netdev_priv(dev);
+ struct geneve_config *cfg = rcu_dereference_protected(geneve->cfg, 1);
+
+ geneve_config_free(cfg);
+}
+
static void geneve_setup(struct net_device *dev)
{
ether_setup(dev);
@@ -1717,6 +1790,8 @@ static void geneve_setup(struct net_device *dev)
dev->netdev_ops = &geneve_netdev_ops;
dev->ethtool_ops = &geneve_ethtool_ops;
dev->needs_free_netdev = true;
+ dev->priv_destructor = geneve_free_dev;
+
SET_NETDEV_DEVTYPE(dev, &geneve_type);
@@ -1878,15 +1953,17 @@ static struct geneve_dev *geneve_find_dev(struct geneve_net *gn,
*tun_on_same_port = false;
*tun_collect_md = false;
list_for_each_entry(geneve, &gn->geneve_list, next) {
- if (info->key.tp_dst == geneve->cfg.info.key.tp_dst &&
- (cfg->dualstack || geneve->cfg.dualstack ||
- geneve_saddr_conflict(info, &geneve->cfg.info))) {
- *tun_collect_md |= geneve->cfg.collect_md;
+ const struct geneve_config *gcfg = rtnl_dereference(geneve->cfg);
+
+ if (info->key.tp_dst == gcfg->info.key.tp_dst &&
+ (cfg->dualstack || gcfg->dualstack ||
+ geneve_saddr_conflict(info, &gcfg->info))) {
+ *tun_collect_md |= gcfg->collect_md;
*tun_on_same_port = true;
}
- if (info->key.tun_id == geneve->cfg.info.key.tun_id &&
- info->key.tp_dst == geneve->cfg.info.key.tp_dst &&
- !memcmp(&info->key.u, &geneve->cfg.info.key.u, sizeof(info->key.u)))
+ if (info->key.tun_id == gcfg->info.key.tun_id &&
+ info->key.tp_dst == gcfg->info.key.tp_dst &&
+ !memcmp(&info->key.u, &gcfg->info.key.u, sizeof(info->key.u)))
t = geneve;
}
return t;
@@ -1922,6 +1999,7 @@ static int geneve_configure(struct net *net, struct net_device *dev,
struct geneve_dev *t, *geneve = netdev_priv(dev);
const struct ip_tunnel_info *info = &cfg->info;
bool tun_collect_md, tun_on_same_port;
+ struct geneve_config *new_cfg;
int err, encap_len;
if (cfg->collect_md && !is_tnl_info_zero(info)) {
@@ -1962,10 +2040,13 @@ static int geneve_configure(struct net *net, struct net_device *dev,
}
}
- dst_cache_reset(&geneve->cfg.info.dst_cache);
- memcpy(&geneve->cfg, cfg, sizeof(*cfg));
+ new_cfg = geneve_config_alloc(cfg);
+ if (IS_ERR(new_cfg))
+ return PTR_ERR(new_cfg);
- if (geneve->cfg.inner_proto_inherit) {
+ rcu_assign_pointer(geneve->cfg, new_cfg);
+
+ if (cfg->inner_proto_inherit) {
dev->header_ops = NULL;
dev->type = ARPHRD_NONE;
dev->hard_header_len = 0;
@@ -1975,10 +2056,15 @@ static int geneve_configure(struct net *net, struct net_device *dev,
err = register_netdevice(dev);
if (err)
- return err;
+ goto err_free_cfg;
list_add(&geneve->next, &gn->geneve_list);
return 0;
+
+err_free_cfg:
+ geneve_config_free(new_cfg);
+ RCU_INIT_POINTER(geneve->cfg, NULL);
+ return err;
}
static void init_tnl_info(struct ip_tunnel_info *info, __u16 dst_port)
@@ -2323,81 +2409,47 @@ static int geneve_newlink(struct net_device *dev,
return 0;
}
-/* Quiesces the geneve device data path for both TX and RX.
- *
- * On transmit geneve checks for non-NULL geneve_sock before it proceeds.
- * So, if we set that socket to NULL under RCU and wait for synchronize_net()
- * to complete for the existing set of in-flight packets to be transmitted,
- * then we would have quiesced the transmit data path. All the future packets
- * will get dropped until we unquiesce the data path.
- *
- * On receive geneve dereference the geneve_sock stashed in the socket. So,
- * if we set that to NULL under RCU and wait for synchronize_net() to
- * complete, then we would have quiesced the receive data path.
+/* Update the device configuration under RTNL.
+ * We use RCU swap to update the configuration atomically, so the data path
+ * (both TX and RX) can continue running without interruption or packet loss.
*/
-static void geneve_quiesce(struct geneve_dev *geneve, struct geneve_sock **gs4,
- struct geneve_sock **gs6)
-{
- *gs4 = rtnl_dereference(geneve->sock4);
- rcu_assign_pointer(geneve->sock4, NULL);
- if (*gs4)
- rcu_assign_sk_user_data((*gs4)->sk, NULL);
-#if IS_ENABLED(CONFIG_IPV6)
- *gs6 = rtnl_dereference(geneve->sock6);
- rcu_assign_pointer(geneve->sock6, NULL);
- if (*gs6)
- rcu_assign_sk_user_data((*gs6)->sk, NULL);
-#else
- *gs6 = NULL;
-#endif
- synchronize_net();
-}
-/* Resumes the geneve device data path for both TX and RX. */
-static void geneve_unquiesce(struct geneve_dev *geneve, struct geneve_sock *gs4,
- struct geneve_sock __maybe_unused *gs6)
-{
- rcu_assign_pointer(geneve->sock4, gs4);
- if (gs4)
- rcu_assign_sk_user_data(gs4->sk, gs4);
-#if IS_ENABLED(CONFIG_IPV6)
- rcu_assign_pointer(geneve->sock6, gs6);
- if (gs6)
- rcu_assign_sk_user_data(gs6->sk, gs6);
-#endif
-}
static int geneve_changelink(struct net_device *dev, struct nlattr *tb[],
struct nlattr *data[],
struct netlink_ext_ack *extack)
{
struct geneve_dev *geneve = netdev_priv(dev);
- struct geneve_sock *gs4, *gs6;
- struct geneve_config cfg;
+ struct geneve_config *old_cfg = rtnl_dereference(geneve->cfg);
+ struct geneve_config *cfg;
int err;
/* If the geneve device is configured for metadata (or externally
* controlled, for example, OVS), then nothing can be changed.
*/
- if (geneve->cfg.collect_md)
+ if (old_cfg->collect_md)
return -EOPNOTSUPP;
/* Start with the existing info. */
- memcpy(&cfg, &geneve->cfg, sizeof(cfg));
- err = geneve_nl2info(tb, data, extack, &cfg, true);
+ cfg = geneve_config_alloc(old_cfg);
+ if (IS_ERR(cfg))
+ return PTR_ERR(cfg);
+
+ err = geneve_nl2info(tb, data, extack, cfg, true);
if (err)
- return err;
+ goto err_free_cfg;
- if (!geneve_dst_addr_equal(&geneve->cfg.info, &cfg.info)) {
- dst_cache_reset(&cfg.info.dst_cache);
- geneve_link_config(dev, &cfg.info, tb);
- }
+ if (!geneve_dst_addr_equal(&old_cfg->info, &cfg->info))
+ geneve_link_config(dev, &cfg->info, tb);
- geneve_quiesce(geneve, &gs4, &gs6);
- memcpy(&geneve->cfg, &cfg, sizeof(cfg));
- geneve_unquiesce(geneve, gs4, gs6);
+ rcu_assign_pointer(geneve->cfg, cfg);
+ call_rcu_hurry(&old_cfg->rcu, geneve_config_free_rcu);
return 0;
+
+err_free_cfg:
+ geneve_config_free(cfg);
+ return err;
}
static void geneve_dellink(struct net_device *dev, struct list_head *head)
@@ -2432,12 +2484,13 @@ static size_t geneve_get_size(const struct net_device *dev)
static int geneve_fill_info(struct sk_buff *skb, const struct net_device *dev)
{
struct geneve_dev *geneve = netdev_priv(dev);
- struct ip_tunnel_info *info = &geneve->cfg.info;
- bool ttl_inherit = geneve->cfg.ttl_inherit;
- bool metadata = geneve->cfg.collect_md;
+ struct geneve_config *cfg = rtnl_dereference(geneve->cfg);
+ struct ip_tunnel_info *info = &cfg->info;
+ bool ttl_inherit = cfg->ttl_inherit;
+ bool metadata = cfg->collect_md;
struct ifla_geneve_port_range ports = {
- .low = htons(geneve->cfg.port_min),
- .high = htons(geneve->cfg.port_max),
+ .low = htons(cfg->port_min),
+ .high = htons(cfg->port_max),
};
__u8 tmp_vni[3];
__u32 vni;
@@ -2468,17 +2521,17 @@ static int geneve_fill_info(struct sk_buff *skb, const struct net_device *dev)
#endif
}
- if (!geneve->cfg.dualstack) {
+ if (!cfg->dualstack) {
if (ip_tunnel_info_af(info) == AF_INET) {
if ((info->key.u.ipv4.src ||
- geneve->cfg.collect_md) &&
+ metadata) &&
nla_put_in_addr(skb, IFLA_GENEVE_LOCAL,
info->key.u.ipv4.src))
goto nla_put_failure;
#if IS_ENABLED(CONFIG_IPV6)
} else {
if ((!ipv6_addr_any(&info->key.u.ipv6.src) ||
- geneve->cfg.collect_md) &&
+ metadata) &&
nla_put_in6_addr(skb, IFLA_GENEVE_LOCAL6,
&info->key.u.ipv6.src))
goto nla_put_failure;
@@ -2491,7 +2544,7 @@ static int geneve_fill_info(struct sk_buff *skb, const struct net_device *dev)
nla_put_be32(skb, IFLA_GENEVE_LABEL, info->key.label))
goto nla_put_failure;
- if (nla_put_u8(skb, IFLA_GENEVE_DF, geneve->cfg.df))
+ if (nla_put_u8(skb, IFLA_GENEVE_DF, cfg->df))
goto nla_put_failure;
if (nla_put_be16(skb, IFLA_GENEVE_PORT, info->key.tp_dst))
@@ -2502,21 +2555,21 @@ static int geneve_fill_info(struct sk_buff *skb, const struct net_device *dev)
#if IS_ENABLED(CONFIG_IPV6)
if (nla_put_u8(skb, IFLA_GENEVE_UDP_ZERO_CSUM6_RX,
- !geneve->cfg.use_udp6_rx_checksums))
+ !cfg->use_udp6_rx_checksums))
goto nla_put_failure;
#endif
if (nla_put_u8(skb, IFLA_GENEVE_TTL_INHERIT, ttl_inherit))
goto nla_put_failure;
- if (geneve->cfg.inner_proto_inherit &&
+ if (cfg->inner_proto_inherit &&
nla_put_flag(skb, IFLA_GENEVE_INNER_PROTO_INHERIT))
goto nla_put_failure;
if (nla_put(skb, IFLA_GENEVE_PORT_RANGE, sizeof(ports), &ports))
goto nla_put_failure;
- if (geneve->cfg.gro_hint &&
+ if (cfg->gro_hint &&
nla_put_flag(skb, IFLA_GENEVE_GRO_HINT))
goto nla_put_failure;
@@ -2671,6 +2724,7 @@ static void __exit geneve_cleanup_module(void)
rtnl_link_unregister(&geneve_link_ops);
unregister_netdevice_notifier(&geneve_notifier_block);
unregister_pernet_subsys(&geneve_net_ops);
+ rcu_barrier();
}
module_exit(geneve_cleanup_module);
--
2.55.0.rc0.799.gd6f94ed593-goog
^ permalink raw reply related [flat|nested] 3+ messages in thread
* [PATCH net-next 2/2] geneve: make geneve_fill_info() RTNL independent
2026-07-01 12:04 [PATCH net-next 0/2] geneve: make geneve_fill_info() RTNL-less Eric Dumazet
2026-07-01 12:04 ` [PATCH net-next 1/2] geneve: convert config to RCU-protected pointer Eric Dumazet
@ 2026-07-01 12:04 ` Eric Dumazet
1 sibling, 0 replies; 3+ messages in thread
From: Eric Dumazet @ 2026-07-01 12:04 UTC (permalink / raw)
To: David S . Miller, Jakub Kicinski, Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Andrew Lunn, netdev,
eric.dumazet, Eric Dumazet
Now that geneve->cfg is an RCU-protected pointer, we can update
geneve_fill_info() to read the configuration under RCU read lock
instead of relying on RTNL.
Also add const qualifiers to the dereferenced pointers where appropriate.
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
drivers/net/geneve.c | 36 +++++++++++++++++++++++++-----------
1 file changed, 25 insertions(+), 11 deletions(-)
diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c
index c777c1a1fa930d74768a659e16380d719f12a6b2..7197f90eeb0f3a381fac5be4a96d6570ca4fccb1 100644
--- a/drivers/net/geneve.c
+++ b/drivers/net/geneve.c
@@ -2483,17 +2483,28 @@ static size_t geneve_get_size(const struct net_device *dev)
static int geneve_fill_info(struct sk_buff *skb, const struct net_device *dev)
{
- struct geneve_dev *geneve = netdev_priv(dev);
- struct geneve_config *cfg = rtnl_dereference(geneve->cfg);
- struct ip_tunnel_info *info = &cfg->info;
- bool ttl_inherit = cfg->ttl_inherit;
- bool metadata = cfg->collect_md;
- struct ifla_geneve_port_range ports = {
- .low = htons(cfg->port_min),
- .high = htons(cfg->port_max),
- };
+ const struct geneve_dev *geneve = netdev_priv(dev);
+ struct ifla_geneve_port_range ports;
+ const struct geneve_config *cfg;
+ const struct ip_tunnel_info *info;
+ bool ttl_inherit;
+ bool metadata;
__u8 tmp_vni[3];
__u32 vni;
+ int err = 0;
+
+ rcu_read_lock();
+ cfg = rcu_dereference(geneve->cfg);
+ if (!cfg) {
+ err = -ENODEV;
+ goto out;
+ }
+
+ info = &cfg->info;
+ ttl_inherit = cfg->ttl_inherit;
+ metadata = cfg->collect_md;
+ ports.low = htons(cfg->port_min);
+ ports.high = htons(cfg->port_max);
tunnel_id_to_vni(info->key.tun_id, tmp_vni);
vni = (tmp_vni[0] << 16) | (tmp_vni[1] << 8) | tmp_vni[2];
@@ -2573,10 +2584,13 @@ static int geneve_fill_info(struct sk_buff *skb, const struct net_device *dev)
nla_put_flag(skb, IFLA_GENEVE_GRO_HINT))
goto nla_put_failure;
- return 0;
+out:
+ rcu_read_unlock();
+ return err;
nla_put_failure:
- return -EMSGSIZE;
+ err = -EMSGSIZE;
+ goto out;
}
static struct rtnl_link_ops geneve_link_ops __read_mostly = {
--
2.55.0.rc0.799.gd6f94ed593-goog
^ permalink raw reply related [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-07-01 12:05 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-07-01 12:04 [PATCH net-next 0/2] geneve: make geneve_fill_info() RTNL-less Eric Dumazet
2026-07-01 12:04 ` [PATCH net-next 1/2] geneve: convert config to RCU-protected pointer Eric Dumazet
2026-07-01 12:04 ` [PATCH net-next 2/2] geneve: make geneve_fill_info() RTNL independent Eric Dumazet
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox