Cap retained guest-to-host TCP payload across flows so a stalled sink cannot grow the supervisor toward the per-flow limit times the maximum flow count.
PR #292 rejected '.'/'..' and added O_NOFOLLOW on the legacy
createDirRecursively walk, but mnt_newapi createDirAt still used
multi-component mkdirat(), which follows intermediate symlinks.
After a -s symlink mount pointing outside the staging root, later
mounts with destinations under that symlink (e.g. link/pwned) could
create directories or attach mounts on the host before pivot_root.
Walk destinations one component at a time with O_NOFOLLOW, resolve
parent fd + basename for leaf symlinkat/move_mount/open_tree/openat,
and use AT_SYMLINK_NOFOLLOW in existsAs*At helpers.
Mount destinations were joined as newroot+"/"+dst and walked with
mkdirat/openat without rejecting "."/".." components or following
symlinks safely. An attacker who can influence mount dst (CLI, config,
or prefix_dst_env) could create directories, symlinks, or mounts outside
the intended staging root on the host before pivot_root.
Add isSafeContainmentPath(), validate destinations in both legacy and
new mount APIs, use O_NOFOLLOW while walking parents, and include a
standalone regression test.
It provides lightweight IP-level connectivity for jailed processes via a TUN device and an epoll-based NAT proxy thread, as an alternative to pasta.
Supports TCP, UDP, and ICMP proxying with per-flow tracking, configurable firewall/redirect rules, SOCKS5 encapsulation for TCP and UDP.
