libknet: fix host info propagation on kronosnetd restart

this avoids lots of incorrect chit/chat and make sure
to send data when required

Signed-off-by: Fabio M. Di Nitto <fdinitto@redhat.com>

diff --git a/libknet/handle.c b/libknet/handle.c
index 0591e469c68858c2fd4bb1393a89c6e6fbd1de35..65ee16c70b397d314fc46d521819df47b8eb224f 100644 (file)
--- a/libknet/handle.c
+++ b/libknet/handle.c
@@ -827,7 +827,6 @@ static void _handle_recv_from_links(knet_handle_t knet_h, int sockfd)
                                src_link = src_host->link +
                                        (knet_hinfo_data->khd_dype.link_up_down.khdt_link_id % KNET_MAX_LINK);
                                /*
-                                * not fully satisfied with this solution yet, but it works
                                 * basically if the node is coming back to life from a crash
                                 * we should receive a host info where local previous status == remote current status
                                 * and so we can detect that node is showing up again
@@ -838,6 +837,17 @@ static void _handle_recv_from_links(knet_handle_t knet_h, int sockfd)
                                        src_link->host_info_up_sent = 0;
                                }
                                src_link->remoteconnected = knet_hinfo_data->khd_dype.link_up_down.khdt_link_status;
+                               if (!src_link->remoteconnected) {
+                                       /*
+                                        * if a host is disconnecting clean, we note that in donnotremoteupdate
+                                        * so that we don't send host info back immediately but we wait
+                                        * for the node to send an update when it's alive again
+                                        */
+                                       src_link->host_info_up_sent = 0;
+                                       src_link->donnotremoteupdate = 1;
+                               } else {
+                                       src_link->donnotremoteupdate = 0;
+                               }
                                log_debug(knet_h, KNET_SUB_LINK, "host message up/down. from host: %s link: %s remote connected: %u",
                                          src_host->name,
                                          src_link->dst_ipaddr,
@@ -910,7 +920,8 @@ static void _handle_dst_link_updates(knet_handle_t knet_h)
                if (dst_host->link[link_idx].connected != 1) /* link is not enabled */
                        continue;
 
-               if (!dst_host->link[link_idx].host_info_up_sent) {
+               if ((!dst_host->link[link_idx].host_info_up_sent) &&
+                   (!dst_host->link[link_idx].donnotremoteupdate)) {
                        send_link_status[send_link_idx] = link_idx;
                        send_link_idx++;
                        /*
@@ -974,6 +985,7 @@ out_unlock:
                        knet_hinfo_data.khd_dype.link_up_down.khdt_link_id = send_link_status[i];
                        _send_host_info(knet_h, &knet_hinfo_data, sizeof(struct knet_hinfo_data));
                        dst_host->link[send_link_status[i]].host_info_up_sent = 1;
+                       dst_host->link[send_link_status[i]].donnotremoteupdate = 0;
                }
        }
 

diff --git a/libknet/libknet.h b/libknet/libknet.h
index 943646706e5cff0a40aae333e2557a744024997a..df03f91e06317ecb2a0ddd026d9128a37f01dd0c 100644 (file)
--- a/libknet/libknet.h
+++ b/libknet/libknet.h
@@ -47,6 +47,7 @@ struct knet_link {
        unsigned int configured:1; /* link is configured and ready to be used */
        unsigned int connected:1;       /* link is enabled for data (local view) */
        unsigned int remoteconnected:1; /* link is enabled for data (peer view) */
+       unsigned int donnotremoteupdate:1;    /* define source of the update */
        unsigned int dynamic; /* see KNET_LINK_DYN_ define above */
        unsigned int dynconnected:1; /* link has been activated by remote dynip */
        uint8_t  priority; /* higher priority == preferred for A/P */