Re: [PATCH net-next v18 20/25] ovpn: implement peer add/get/dump/delete via netlink

2025-01-22, 00:26:50 +0100, Antonio Quartulli wrote:

On 20/01/2025 15:52, Antonio Quartulli wrote:

quoted

On 17/01/2025 12:48, Sabrina Dubroca wrote:
[...]

quoted

-------- 8< --------

diff --git a/drivers/net/ovpn/netlink.c b/drivers/net/ovpn/netlink.c
index 72357bb5f30b..19aa4ee6d468 100644
--- a/drivers/net/ovpn/netlink.c
+++ b/drivers/net/ovpn/netlink.c

@@ -733,6 +733,9 @@ int ovpn_nl_peer_del_doit(struct sk_buff *skb,

struct genl_info *info)
      netdev_dbg(ovpn->dev, "del peer %u\n", peer->id);
      ret = ovpn_peer_del(peer, OVPN_DEL_PEER_REASON_USERSPACE);
+    if (ret >= 0 && peer->sock)
+        wait_for_completion(&peer->sock_detach);
+
      ovpn_peer_put(peer);
      return ret;

diff --git a/drivers/net/ovpn/peer.c b/drivers/net/ovpn/peer.c
index b032390047fe..6120521d0c32 100644
--- a/drivers/net/ovpn/peer.c
+++ b/drivers/net/ovpn/peer.c

@@ -92,6 +92,7 @@ struct ovpn_peer *ovpn_peer_new(struct ovpn_priv

*ovpn, u32 id)
      ovpn_peer_stats_init(&peer->vpn_stats);
      ovpn_peer_stats_init(&peer->link_stats);
      INIT_WORK(&peer->keepalive_work, ovpn_peer_keepalive_send);
+    init_completion(&peer->sock_detach);
      ret = dst_cache_init(&peer->dst_cache, GFP_KERNEL);
      if (ret < 0) {

diff --git a/drivers/net/ovpn/peer.h b/drivers/net/ovpn/peer.h
index 7a062cc5a5a4..8c54bf5709ef 100644
--- a/drivers/net/ovpn/peer.h
+++ b/drivers/net/ovpn/peer.h

@@ -112,6 +112,7 @@ struct ovpn_peer {

      struct rcu_head rcu;
      struct work_struct remove_work;
      struct work_struct keepalive_work;
+    struct completion sock_detach;
  };
  /**

diff --git a/drivers/net/ovpn/socket.c b/drivers/net/ovpn/socket.c
index a5c3bc834a35..7cefac42c3be 100644
--- a/drivers/net/ovpn/socket.c
+++ b/drivers/net/ovpn/socket.c

@@ -31,6 +31,8 @@ static void ovpn_socket_release_kref(struct kref *kref)

      sockfd_put(sock->sock);
      kfree_rcu(sock, rcu);
+
+    complete(&sock->peer->sock_detach);

I am moving this line to ovpn_socket_put(), right after kref_put() so that
every peer going through the socket release will get their complete()
invoked.

If the peer happens to be the last one owning the socket, kref_put() will
first do the detach and only then complete() gets called.

This requires ovpn_socket_release/put() to take the peer as argument, but
that's ok.

This way we should achieve what we needed.

This seems like it would, thanks.

-- 
Sabrina