[Bug 223745] PROC_REAP_KILL misses descendants if another process has become a reaper in between

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=223745

            Bug ID: 223745
           Summary: PROC_REAP_KILL misses descendants if another process
                    has become a reaper in between
           Product: Base System
           Version: CURRENT
          Hardware: Any
                OS: Any
            Status: New
          Severity: Affects Only Me
          Priority: ---
         Component: kern
          Assignee: freebsd-bugs at FreeBSD.org
          Reporter: muh.muhten at gmail.com

PROC_REAP_KILL sends a signal to processes that (when REAPER_KILL_CHILDREN is
not set) _are to be reaped by_ the reaper of the specified process.

If that reaper has a descendant which has acquired reaper status, then the
descendant reaper's descendants will not be considered descendants of the
reaper of the specified process.

An implication is that, if a reaper kills all its descendants with SIGKILL, it
can still have children reparented to it due to their reaper getting killed.

I don't see many situations where it would be useful to signal a paraphyletic
subtree of a reaper's descendant. On the other hand, it would be useful to be
signal a subtree in such a way that a process cannot "escape" being killed. The
latter is sort of achieved by repeatedly signalling all descendants of a reaper
until no children can be killed, but the call has not failed on any particular
process (i.e. rk_fpid == -1); however, this requires that the reaper be
specified by its own pid (and not that of a descendant, since that descendant
will die) and on the whole does not seem reasonable.

I suggest a flag for PROC_REAP_KILL to recursively kill the descendants of
reaper descendants of reapers.

-- 
You are receiving this mail because:
You are the assignee for the bug.