Linux Container Development
 help / color / mirror / Atom feed
From: "Hallgrimur H. Gunnarsson" <halg@google.com>
To: Christian Brauner <brauner@kernel.org>,
	Serge Hallyn <sergeh@kernel.org>,
	 Andy Lutomirski <luto@kernel.org>
Cc: "Paul Moore" <paul@paul-moore.com>,
	"James Morris" <jmorris@namei.org>,
	"David Howells" <dhowells@redhat.com>,
	"Jarkko Sakkinen" <jarkko@kernel.org>,
	"Kees Cook" <kees@kernel.org>, "Jonathan Corbet" <corbet@lwn.net>,
	"Shuah Khan" <skhan@linuxfoundation.org>,
	"Randy Dunlap" <rdunlap@infradead.org>,
	"Mickaël Salaün" <mic@digikod.net>,
	"Jonathan Calmels" <jcalmels@3xx0.net>,
	"Jann Horn" <jannh@google.com>,
	"Günther Noack" <gnoack@google.com>,
	"KP Singh" <kpsingh@kernel.org>,
	"Casey Schaufler" <casey@schaufler-ca.com>,
	linux-fsdevel@vger.kernel.org,
	linux-security-module@vger.kernel.org, linux-api@vger.kernel.org,
	containers@lists.linux.dev, keyrings@vger.kernel.org,
	linux-doc@vger.kernel.org, linux-kselftest@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	"Hallgrimur H. Gunnarsson" <halg@google.com>
Subject: [PATCH v2 0/4] userns: per-namespace capability bounding set
Date: Thu, 19 Nov 2026 10:42:04 +0000	[thread overview]
Message-ID: <20261119104208.782900-1-halg@google.com> (raw)

When an unprivileged process creates a new user namespace, it is granted
a full set of capabilities (CAP_FULL_SET) scoped to that namespace.
While isolated to the namespace, these capabilities allow unprivileged
local users to reach deep kernel subsystems (most notably networking
configuration and raw sockets via CAP_NET_ADMIN and CAP_NET_RAW) that
historically assumed a trusted caller and have accounted for a large
share of local privilege escalation vulnerabilities.

Many sandboxes and workloads use unprivileged user namespaces strictly
for UID/GID, mount, PID, or network isolation and never need high-risk
capabilities like CAP_NET_ADMIN. Likewise, system administrators,
service managers, and container runtimes often want to enforce a
capability ceiling across descendant user namespaces while still
permitting exceptions for authorized workloads.

Prior work by Jonathan Calmels [1, 2] proposed a per-thread userns
capability set (cap_userns), while v1 of this series [3] proposed a
capability bounding set (cap_bset) in struct user_namespace configured
via a global sysctl and clone3().

During v1 review, both Christian Brauner and Andy Lutomirski had
reservations about the sysctl and clone3() approach, and Andy pointed
out the need to cover the single-deep container case. Christian then
proposed combining per-cred cap_userns with the per-namespace cap_bset
ceiling [4].

This series implements that combined approach:

1. Per-namespace cap_bset and per-cred cap_userns (Patch 1)
   Introduce user_namespace->cap_bset and cred->cap_userns, controlled
   via prctl(PR_CAP_USERNS) and SECBIT_CAP_USERNS_STRICT (named _STRICT
   rather than SECURE_NO_USERNS_CAP_RAISE as it also disables the
   default cap_permitted exemption). Clamp namespace creation,
   set_cred_user_ns(), and parent-owner checks in cap_capable_helper()
   against these ceilings.

   One refinement over the v1 review sketch: set_cred_user_ns() also
   intersects cred->cap_userns with user_ns->cap_bset on namespace
   entry. Since userns transitions only move downward in the hierarchy,
   excluded bits are permanently inert anyway. Clearing them keeps
   CapUns and PR_CAP_USERNS_IS_SET accurate after setns() and gives all
   six task capability masks a uniform user_ns->cap_bset subset
   invariant.

2. Ptrace and dumpability bounding (Patch 2)
   Require the tracee's cred_userns_bset() to be a subset of the
   tracer's in the same-namespace ptrace check (complementing the
   cross-userns owner check in cap_capable_helper()), and update
   cred_cap_issubset() so widening the effective userns ceiling resets
   mm->dumpable and pdeath_signal.

3. Introspection via /proc/<pid>/status (Patch 3)
   Expose cred->cap_userns as CapUns in /proc/<pid>/status. Leave
   AUDIT_CAPSET unchanged, as capset() never modifies cap_userns
   (matching cap_bset, where prctl() changes are covered by syscall
   auditing).

4. Selftests (Patch 4)
   Add kselftests covering PR_CAP_USERNS, CapUns, nested monotonicity,
   creator capability retention, SECBIT_CAP_USERNS_STRICT, setns()
   clamping, parent-owner bounding, ptrace restrictions, and dumpability
   resets.

Changes in v2:
- Replace the global sysctl, CLONE_NEWUSER_BSET / struct clone_args
  extension, and NS_USER_GET_INFO ioctl with per-cred cap_userns,
  prctl(PR_CAP_USERNS), SECBIT_CAP_USERNS_STRICT, and CapUns in
  /proc/<pid>/status.
- Drop CLONE_NEWNET_LOOPBACK and UNSHARE_NEWNET_LOOPBACK to handle in a
  separate series.
- Clamp set_cred_user_ns() (including cred->cap_userns) against
  user_ns->cap_bset and cred_userns_bset(cred).
- In cap_capable_helper(), bound implicit parent-owner capabilities
  by the owned namespace's cap_bset, and grant them only when that
  cap_bset is a subset of cred_userns_bset(cred).
- Bound same-namespace ptrace checks in cap_ptrace_access_check() and
  cap_ptrace_traceme() and credential subset checks in
  cred_cap_issubset() by userns capability ceilings.
- Rebase onto v7.4-rc1, preserving SECURE_ALL_UNPRIVILEGED handling in
  set_cred_user_ns().
- Link to v1: https://lore.kernel.org/r/20261028221705.3915340-1-halg@google.com

[1] https://lore.kernel.org/all/20240609104355.442002-1-jcalmels@3xx0.net/
[2] https://lwn.net/Articles/978846/
[3] https://lore.kernel.org/all/20261028221705.3915340-1-halg@google.com/
[4] https://lore.kernel.org/all/20261111-behagen-faktisch-entgegen-e2e714373895@brauner/

Hallgrimur H. Gunnarsson (4):
  userns: introduce capability bounding set for user namespaces
  capability: bound ptrace and cred subset checks by userns ceilings
  proc: expose userns capability mask as CapUns in /proc/<pid>/status
  selftests/capabilities: add userns capability bounding set tests

 Documentation/filesystems/proc.rst            |    2 +
 Documentation/security/credentials.rst        |   12 +-
 fs/proc/array.c                               |    4 +-
 include/linux/cred.h                          |   22 +
 include/linux/user_namespace.h                |   12 +
 include/uapi/linux/prctl.h                    |    7 +
 include/uapi/linux/securebits.h               |   16 +-
 init/init_task.c                              |    1 +
 kernel/cred.c                                 |   13 +-
 kernel/user.c                                 |    1 +
 kernel/user_namespace.c                       |   33 +-
 security/commoncap.c                          |   80 +-
 security/keys/process_keys.c                  |    1 +
 .../testing/selftests/capabilities/.gitignore |    1 +
 tools/testing/selftests/capabilities/Makefile |    2 +-
 .../capabilities/userns_bounding_set_test.c   | 1064 +++++++++++++++++
 16 files changed, 1246 insertions(+), 25 deletions(-)
 create mode 100644 tools/testing/selftests/capabilities/userns_bounding_set_test.c


base-commit: 6628e7d4c6b25c8d18f1345e3a20d21829aae7e4
-- 
2.56.1.708.g77570a6a1d-goog


             reply	other threads:[~2026-11-19 10:42 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-19 10:42 Hallgrimur H. Gunnarsson [this message]
2026-11-19 10:42 ` [PATCH v2 1/4] userns: introduce capability bounding set for user namespaces Hallgrimur H. Gunnarsson
2026-11-19 10:42 ` [PATCH v2 2/4] capability: bound ptrace and cred subset checks by userns ceilings Hallgrimur H. Gunnarsson
2026-11-19 10:42 ` [PATCH v2 3/4] proc: expose userns capability mask as CapUns in /proc/<pid>/status Hallgrimur H. Gunnarsson
2026-11-19 10:42 ` [PATCH v2 4/4] selftests/capabilities: add userns capability bounding set tests Hallgrimur H. Gunnarsson

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261119104208.782900-1-halg@google.com \
    --to=halg@google.com \
    --cc=brauner@kernel.org \
    --cc=casey@schaufler-ca.com \
    --cc=containers@lists.linux.dev \
    --cc=corbet@lwn.net \
    --cc=dhowells@redhat.com \
    --cc=gnoack@google.com \
    --cc=jannh@google.com \
    --cc=jarkko@kernel.org \
    --cc=jcalmels@3xx0.net \
    --cc=jmorris@namei.org \
    --cc=kees@kernel.org \
    --cc=keyrings@vger.kernel.org \
    --cc=kpsingh@kernel.org \
    --cc=linux-api@vger.kernel.org \
    --cc=linux-doc@vger.kernel.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-kselftest@vger.kernel.org \
    --cc=linux-security-module@vger.kernel.org \
    --cc=luto@kernel.org \
    --cc=mic@digikod.net \
    --cc=paul@paul-moore.com \
    --cc=rdunlap@infradead.org \
    --cc=sergeh@kernel.org \
    --cc=skhan@linuxfoundation.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox