Coredump regression with e37fdb785a5f ("exec: Use secureexec for setting dumpability")

From: Laura Abbott
Date: Mon Dec 18 2017 - 18:17:42 EST


Fedora got a bug report
of coredumps failing for systemd units. If systemd units segfault, no
core is produced.

commit e37fdb785a5f95ecadf43b773c97f676500ac7b8 (refs/bisect/bad)
Author: Kees Cook <keescook@xxxxxxxxxxxx>
Date: Tue Jul 18 15:25:31 2017 -0700

exec: Use secureexec for setting dumpability
The examination of "current" to decide dumpability is wrong. This was a
check of and euid/uid (or egid/gid) mismatch in the existing process,
not the newly created one. This appears to stretch back into even the
"history.git" tree. Luckily, dumpability is later set in commit_creds().
In earlier kernel versions before creds existed, similar checks also
existed late in the exec flow, covering up the mistake as far back as I
could find.
Note that because the commit_creds() check examines differences of euid,
uid, egid, gid, and capabilities between the old and new creds, it would
look like the setup_new_exec() dumpability test could be entirely removed.
However, the secureexec test may cover a different set of tests (specific
to the LSMs) than what commit_creds() checks for. So, fix this test to
use secureexec (the removed euid tests are redundant to the commoncap
secureexec checks now).
Cc: David Howells <dhowells@xxxxxxxxxx>
Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
Acked-by: Serge Hallyn <serge@xxxxxxxxxx>
Reviewed-by: James Morris <james.l.morris@xxxxxxxxxx>

This was still broken as of -rc3 although interestingly enough this worked
on my rawhide system. Any ideas?