Comment 3 for bug 1551894

Hrm, this looks like it might be a legit regression. 4.4.0-8 passes the test, while 4.4.0-9.X is failing. In both instances, /proc/sys/kernel/yama/ptrace_scope is set to 1. It looks like cousin processes are allowed to ptrace each other, which yama's ptrace restrictions should prevent.

Looking at the git commits between tags Ubuntu-4.4.0-8.23 and Ubuntu-4.4.0-9.24, the following commits stand out as being ptrace relevent:

  commit 969624b7c1c8c9784651eb97431e6f2bbb7a024c
  Author: Jann Horn <email address hidden>
  Date: Wed Jan 20 15:00:04 2016 -0800
  ptrace: use fsuid, fsgid, effective creds for fs access checks
  upstream commit caaee6234d05a58c5b4d05e7bf766131b810a657 upstream.

and

  commit a76b8ce7ad1f65a96638f161ff83075de04ec9cc
  Author: Jann Horn <email address hidden>
  Date: Sat Dec 12 21:12:41 2015 +0100
  UBUNTU: SAUCE: (noup) ptrace: being capable wrt a process requires mapped uids/gids
  upstream reference https://lkml.org/lkml/2015/12/12/259

But it's not obvious to me why either commit would break this.