Hi, I need a code review for the fix of CR 6678463.
The root cause of this problem was that the connection was not marked as dead, after it was disconnected, so it can be closed later on by the connection manager's garbage collector. webrev is here: http://cr.opensolaris.org/~dain/6678463/ <http://cr.opensolaris.org/%7Edain/6678463/> The output of 'webrev' is somehow 1 line off. Below is the output of 'hg diff' which is more accurate: > > @@ -3197,9 +3197,12 @@ > * be sending any more RPC replies, we > abort > * the connection, including flushing > * any RPC requests that are in-transit. > + * In either case, mark the entry as > dead so that > + * it can be closed by the connection > manager's > + * garbage collector. > */ > + cm_entry->x_dead = TRUE; > if (cm_entry->x_closing) { > - cm_entry->x_dead = TRUE; > mutex_exit(&connmgr_lock); > have_connmgr_lock = 0; > if (clnt_stop_idle != NULL) Thanks, -Dai
