There is no need the add another sync point at the beginning of restart. The bug was elsewhere - in reparent_thread() the code tested the child's flags for PF_RESTARTING instead of the parent's. Oren. Serge E. Hallyn wrote: > Make sure that all prepared tasks are in sys_restart before the > coordinator proceeds. Otherwise, it was possible for the last > task actually in sys_restart to sync before the straggler was in > sys_restart, sending that task a signal and causing that task to > exit. Then since there are not enough tasks completing restart, > the restart operation ends up hanging, waiting for the killed task. > > Signed-off-by: Serge E. Hallyn <serue@xxxxxxxxxx> > --- _______________________________________________ Containers mailing list Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/containers
