net: Clean up af_ops->{get_link_af_size,fill_link_af}().
Subsystems (4)
cover net: Clean up af_ops->{get_link_af_size,fill_link_af}().
Since commit 5fa85a09390c ("net: core: rcu-ify rtnl af_ops"), af_ops->{get_link_af_size,fill_link_af}() are called under RCU.
Patch 1 makes the context clear and patch 2 removes unnecessary rcu_read_lock().
Kuniyuki Iwashima (2):
ipv4: Use RCU helper in inet_get_link_af_size() and
inet_fill_link_af().
net: bridge: Remove rcu_read_lock() in br_get_link_af_size_filtered().
net/bridge/br_netlink.c | 2 -- net/ipv4/devinet.c | 6 ++---- 2 files changed, 2 insertions(+), 6 deletions(-)
Signature · 1 line
2.30.2
PATCH 1/2 ipv4: Use RCU helper in inet_get_link_af_size() and inet_fill_link_af().
Since commit 5fa85a09390c ("net: core: rcu-ify rtnl af_ops"), af_ops->{get_link_af_size,fill_link_af}() are called under RCU.
Instead of using rcu_dereference_rtnl(), let's make the context clear by using RCU helpers.
--- net/ipv4/devinet.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-)
net/ipv4/devinet.c +2 -4
@@ -1950,9 +1950,7 @@ static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh,static size_t inet_get_link_af_size (const struct net_device *dev,u32 ext_filter_mask) { -struct in_device *in_dev =rcu_dereference_rtnl (dev->ip_ptr); - -if (!in_dev) +if (!rcu_access_pointer (dev->ip_ptr))return 0 ;return nla_total_size (IPV4_DEVCONF_MAX *4 );/* IFLA_INET_CONF */
@@ -1961,7 +1959,7 @@ static size_t inet_get_link_af_size(const struct net_device *dev,static int inet_fill_link_af (struct sk_buff *skb,const struct net_device *dev,u32 ext_filter_mask) { -struct in_device *in_dev =rcu_dereference_rtnl (dev->ip_ptr); +struct in_device *in_dev =rcu_dereference (dev->ip_ptr);struct nlattr *nla;int i;
Signature · 1 line
2.30.2
PATCH 2/2 net: bridge: Remove rcu_read_lock() in br_get_link_af_size_filtered().
In reply to Kuniyuki IwashimaSince commit 5fa85a09390c ("net: core: rcu-ify rtnl af_ops"), af_ops->get_link_af_size() is called under RCU.
Let's remove unnecessary rcu_read_lock() in br_get_link_af_size_filtered().
--- net/bridge/br_netlink.c | 2 -- 1 file changed, 2 deletions(-)
net/bridge/br_netlink.c +0 -2
@@ -101,7 +101,6 @@ static size_t br_get_link_af_size_filtered(const struct net_device *dev,size_t vinfo_sz =0 ;int num_vlan_infos; -rcu_read_lock ();if (netif_is_bridge_port (dev)) { p =br_port_get_check_rcu (dev);if (p)
@@ -111,7 +110,6 @@ static size_t br_get_link_af_size_filtered(const struct net_device *dev, vg =br_vlan_group_rcu (br); } num_vlan_infos =br_get_num_vlan_infos (vg, filter_mask); -rcu_read_unlock ();if (p && (p->flags &BR_VLAN_TUNNEL )) vinfo_sz +=br_get_vlan_tunnel_info_size (vg);
Signature · 1 line
2.30.2
Jakub Kicinski
maintainerQuoted 5 lines · On Thu, 15 Aug 2024 14:11:35 -0700 Kuniyuki Iwashima wrote:
Since commit
5fa85a09390c("net: core: rcu-ify rtnlaf_ops"), af_ops->{get_link_af_size,fill_link_af}() are called under RCU.Patch 1 makes the context clear and patch 2 removes unnecessary
rcu_read_lock().
Tests violently disagree.
https://netdev.bots.linux.dev/contest.html?branch=net-next-2024-08-16--00-00&pw-n=0&pass=0
Signature · 1 line
pw-bot: cr
From: Jakub Kicinski <[email protected]>
Date: Thu, 15 Aug 2024 18:36:51 -0700
Quoted 10 lines · On Thu, 15 Aug 2024 14:11:35 -0700 Kuniyuki Iwashima wrote:
Quoted 5 lines
Since commit
5fa85a09390c("net: core: rcu-ify rtnlaf_ops"), af_ops->{get_link_af_size,fill_link_af}() are called under RCU.Patch 1 makes the context clear and patch 2 removes unnecessary
rcu_read_lock().Tests violently disagree.
https://netdev.bots.linux.dev/contest.html?branch=net-next-2024-08-16--00-00&pw-n=0&pass=0
Sorry, I missed the birdge one is called under RTNL. I'll drop patch 2.