On 06/20, Eric W. Biederman wrote: > > Oleg Nesterov <oleg@xxxxxxxxxx> writes: > > > And. I do not think these series can fix the discussed problems. ns->dead > > definitely can't, no? > > I'm am fairly confident that we have the signal sending races fixed so > we can reasonably expect having sent SIGKILL to all processes in a pid > namespace Sorry, didn't notice this part... Which races? I am talking about the current problems with pid_ns_release_proc(), we have at least 3 bugs, from the 2/2 changelog: - Nobody does mntput() if copy_process() fails after pid_ns_prepare_proc(). - proc_flush_task() checks upid->nr == 1 to verify we are init, this is wrong if a multi-threaded init does exec. - As Louis pointed out, this namespace can have the detached EXIT_DEAD tasks which can use ns->proc_mnt after this mntput(). Oleg. _______________________________________________ Containers mailing list Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/containers
