This PR closes the sockets used by a mountpoint when leaving the related thread: since a "destroy" request joins the thread, this ensures ports are free once a response is sent. Before, sockets would only be closed in the mountpoint free callback, so when reference counters gave the green light, which could take longer and cause the issue described in #3111.
lminiero
commented
1 year ago
This PR closes the sockets used by a mountpoint when leaving the related thread: since a "destroy" request joins the thread, this ensures ports are free once a response is sent. Before, sockets would only be closed in the mountpoint free callback, so when reference counters gave the green light, which could take longer and cause the issue described in #3111.