On Wed, 1 May 2024 18:35:29 GMT, Larry Cable <d...@openjdk.org> wrote:
>> Do you mean that it should compare the input PID against the outermost >> (leftmost) PID in the `NSpid` list from `/proc/<pid>/status` and not >> innermost (rightmost) as is done right now? What would be the benefit of >> that? Or did you mean something else? >> >> I'm working on a fix for https://bugs.openjdk.org/browse/JDK-8327114 right >> now, and it occurred to me that there is a tiny risk of `pid != ns_pid` not >> evaluating to `true` even though the processes are in different PID >> namespaces (because two different PID namespaces can have the same PIDs). I >> think it could be mitigated by always trying `/proc/<pid>/root/tmp` first, >> and if it cannot be read, fall back to `/tmp`. > > c.f: /proc/<pid>/ns/pid > > every (Linux) namespace has a unique id, if 2 (or more) processes occupy the > same pid namespace (or any other for that matter) then their ../ns/pid > namespace ids will be the same. **`Files.readSymbolicLink(Path.of("/proc/self/ns/pid"))`** ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/17628#discussion_r1586769059
