open: introduce openat2(2) syscall

Started by Aleksa Sarai on Nov 5, 2019 · 30 messages from 4 authors · Last reply Nov 14, 2019

Series

9 patches, 35 files

+1,600 −82

Subsystems (21)

Show 18 more

Maintainers (36)

Show 32 more
STALE2563d SUPERSEDED

Status: Superseded by v18 (posted 2019-12-06). [go to v18]

Revision v15 (5 revisions archived).

Revisions (5)
  1. v14 re-posted go to this post
  2. v15 current [diff against previous]
  3. v16 [diff against previous]
  4. v17 [diff against previous] re-posted go to this post
  5. v18 [diff against previous]

open the first message on its own page

  1. [PATCH v15 1/9] namei: LOOKUP_NO_SYMLINKS: block symlink resolution
    Revisions and related patches
    Revisions (5)
    1. v14 re-posted go to this post
    2. v15 current [diff against previous]
    3. v16 [diff against previous]
    4. v17 [diff against previous] re-posted go to this post
    5. v18 [diff against previous] re-posted go to this post
  2. [PATCH v15 2/9] namei: LOOKUP_NO_MAGICLINKS: block magic-link resolution
    Revisions and related patches
    Revisions (5)
    1. v14 re-posted go to this post
    2. v15 current [diff against previous]
    3. v16 [diff against previous]
    4. v17 [diff against previous] re-posted go to this post
    5. v18 [diff against previous] re-posted go to this post
  3. [PATCH v15 3/9] namei: LOOKUP_NO_XDEV: block mountpoint crossing
    Revisions and related patches
    Revisions (5)
    1. v14 re-posted go to this post
    2. v15 current [diff against previous]
    3. v16 [diff against previous]
    4. v17 [diff against previous] re-posted go to this post
    5. v18 [diff against previous]
  4. [PATCH v15 4/9] namei: LOOKUP_BENEATH: O_BENEATH-like scoped resolution
    Revisions and related patches
    Revisions (5)
    1. v14 re-posted go to this post
    2. v15 current [diff against previous]
    3. v16 [diff against previous]
    4. v17 [diff against previous] re-posted go to this post
    5. v18 [diff against previous]
  5. [PATCH v15 5/9] namei: LOOKUP_IN_ROOT: chroot-like scoped resolution
    Revisions and related patches
    Revisions (5)
    1. v14 re-posted go to this post
    2. v15 current [diff against previous]
    3. v16 [diff against previous]
    4. v17 [diff against previous] re-posted go to this post
    5. v18 [diff against previous]
  6. [PATCH v15 6/9] namei: LOOKUP_{IN_ROOT,BENEATH}: permit limited ".." resolution
    Revisions and related patches
    Revisions (5)
    1. v14 re-posted go to this post
    2. v15 current [diff against previous]
    3. v16 [diff against previous]
    4. v17 [diff against previous] re-posted go to this post
    5. v18 [diff against previous]
  7. [PATCH v15 7/9] open: introduce openat2(2) syscall
    Revisions and related patches
    Revisions (4)
    1. v15 current
    2. v16 [diff against previous]
    3. v17 [diff against previous] re-posted go to this post
    4. v18 [diff against previous]
  8. [PATCH v15 8/9] selftests: add openat2(2) selftests
    Revisions and related patches
    Revisions (4)
    1. v15 current
    2. v16 [diff against previous]
    3. v17 [diff against previous] re-posted go to this post
    4. v18 [diff against previous]
  9. [PATCH v15 9/9] Documentation: path-lookup: mention LOOKUP_MAGICLINK_JUMPED
    Revisions and related patches
    Revisions (4)
    1. v15 current
    2. v16 [diff against previous]
    3. v17 [diff against previous] re-posted go to this post
    4. v18 [diff against previous]
Message details and related content
This patchset is being developed here:
  <https://github.com/cyphar/linux/tree/openat2/master>

Patch changelog:
 v15:
  * Fix code style for LOOKUP_IN_ROOT handling in path_init(). [Linus Torvalds]
  * Split out patches for each individual LOOKUP flag.
  * Reword commit messages to give more background information about the
    series, as well as mention the semantics of each flag in more detail.
 v14: <https://lore.kernel.org/lkml/20191010054140.8483-1-cyphar@cyphar.com/>
      <https://lore.kernel.org/lkml/20191026185700.10708-1-cyphar@cyphar.com>
 v13: <https://lore.kernel.org/lkml/20190930183316.10190-1-cyphar@cyphar.com/>
 v12: <https://lore.kernel.org/lkml/20190904201933.10736-1-cyphar@cyphar.com/>
 v11: <https://lore.kernel.org/lkml/20190820033406.29796-1-cyphar@cyphar.com/>
      <https://lore.kernel.org/lkml/20190728010207.9781-1-cyphar@cyphar.com/>
 v10: <https://lore.kernel.org/lkml/20190719164225.27083-1-cyphar@cyphar.com/>
 v09: <https://lore.kernel.org/lkml/20190706145737.5299-1-cyphar@cyphar.com/>
 v08: <https://lore.kernel.org/lkml/20190520133305.11925-1-cyphar@cyphar.com/>
 v07: <https://lore.kernel.org/lkml/20190507164317.13562-1-cyphar@cyphar.com/>
 v06: <https://lore.kernel.org/lkml/20190506165439.9155-1-cyphar@cyphar.com/>
 v05: <https://lore.kernel.org/lkml/20190320143717.2523-1-cyphar@cyphar.com/>
 v04: <https://lore.kernel.org/lkml/20181112142654.341-1-cyphar@cyphar.com/>
 v03: <https://lore.kernel.org/lkml/20181009070230.12884-1-cyphar@cyphar.com/>
 v02: <https://lore.kernel.org/lkml/20181009065300.11053-1-cyphar@cyphar.com/>
 v01: <https://lore.kernel.org/lkml/20180929103453.12025-1-cyphar@cyphar.com/>

For a very long time, extending openat(2) with new features has been
incredibly frustrating. This stems from the fact that openat(2) is
possibly the most famous counter-example to the mantra "don't silently
accept garbage from userspace" -- it doesn't check whether unknown flags
are present[1].

This means that (generally) the addition of new flags to openat(2) has
been fraught with backwards-compatibility issues (O_TMPFILE has to be
defined as __O_TMPFILE|O_DIRECTORY|[O_RDWR or O_WRONLY] to ensure old
kernels gave errors, since it's insecure to silently ignore the
flag[2]). All new security-related flags therefore have a tough road to
being added to openat(2).

Furthermore, the need for some sort of control over VFS's path resolution (to

372 more lines

[PATCH v15 1/9] namei: LOOKUP_NO_SYMLINKS: block symlink resolution

Message details and related content
/* Background. */
Userspace cannot easily resolve a path without resolving symlinks, and
would have to manually resolve each path component with O_PATH and
O_NOFOLLOW. This is clearly inefficient, and can be fairly easy to screw
up (resulting in possible security bugs). Linus has mentioned that Git
has a particular need for this kind of flag[1]. It also resolves a
fairly long-standing perceived deficiency in O_NOFOLLOw -- that it only
blocks the opening of trailing symlinks.

This is part of a refresh of Al's AT_NO_JUMPS patchset[2] (which was a
variation on David Drysdale's O_BENEATH patchset[3], which in turn was
based on the Capsicum project[4]).

/* Userspace API. */
LOOKUP_NO_SYMLINKS will be exposed to userspace through openat2(2).

/* Semantics. */
Unlike most other LOOKUP flags (most notably LOOKUP_FOLLOW),
LOOKUP_NO_SYMLINKS applies to all components of the path.

With LOOKUP_NO_SYMLINKS, any symlink path component encountered during
path resolution will yield -ELOOP. If the trailing component is a
symlink (and no other components were symlinks), then O_PATH|O_NOFOLLOW
will not error out and will instead provide a handle to the trailing
symlink -- without resolving it.

/* Testing. */
LOOKUP_NO_SYMLINKS is tested as part of the openat2(2) selftests.

[1]: https://lore.kernel.org/lkml/CA+55aFyOKM7DW7+0sdDFKdZFXgptb5r1id9=Wvhd8AgSP7qjwQ@mail.gmail.com/
[2]: https://lore.kernel.org/lkml/20170429220414.GT29622@ZenIV.linux.org.uk/
[3]: https://lore.kernel.org/lkml/1415094884-18349-1-git-send-email-drysdale@google.com/
[4]: https://lore.kernel.org/lkml/1404124096-21445-1-git-send-email-drysdale@google.com/

Cc: Christian Brauner <redacted>
Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Aleksa Sarai <redacted>
---
 fs/namei.c            | 3 +++

3 more lines

diff --git a/fs/namei.c b/fs/namei.c
index 671c3c1a3425..4e85d6fa4048 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -1045,6 +1045,9 @@ const char *get_link(struct nameidata *nd)
 	int error;
 	const char *res;
 
+	if (unlikely(nd->flags & LOOKUP_NO_SYMLINKS))
+		return ERR_PTR(-ELOOP);
+
 	if (!(nd->flags & LOOKUP_RCU)) {
 		touch_atime(&last->link);
 		cond_resched();
diff --git a/include/linux/namei.h b/include/linux/namei.h
index 397a08ade6a2..ee2e35af387f 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -39,6 +39,9 @@ enum {LAST_NORM, LAST_ROOT, LAST_DOT, LAST_DOTDOT, LAST_BIND};
 #define LOOKUP_ROOT		0x2000
 #define LOOKUP_ROOT_GRABBED	0x0008
 
+/* Scoping flags for lookup. */
+#define LOOKUP_NO_SYMLINKS	0x020000 /* No symlink crossing. */
+
 extern int path_pts(struct path *path);
 
 extern int user_path_at_empty(int, const char __user *, unsigned, struct path *, int *empty);
-- 
2.23.0

[PATCH v15 2/9] namei: LOOKUP_NO_MAGICLINKS: block magic-link resolution

Message details and related content
/* Background. */
There has always been a special class of symlink-like objects in procfs
(and a few other pseudo-filesystems) which allow for non-lexical
resolution of paths using nd_jump_link(). These "magic-links" do not
follow traditional mount namespace boundaries, and have been used
consistently in container escape attacks because they can be used to
trick unsuspecting privileged processes into resolving unexpected paths.

It is also non-trivial for userspace to unambiguously avoid resolving
magic-links, because they do not have a reliable indication that they
are a magic-link (in order to verify them you'd have to manually open
the path given by readlink(2) and then verify that the two file
descriptors reference the same underlying file, which is plagued with
possible race conditions or supplementary attack scenarios).

It would therefore be very helpful for userspace to be able to avoid
these symlinks easily, thus hopefully removing a tool from attackers'
toolboxes.

This is part of a refresh of Al's AT_NO_JUMPS patchset[1] (which was a
variation on David Drysdale's O_BENEATH patchset[2], which in turn was
based on the Capsicum project[3]).

/* Userspace API. */
LOOKUP_NO_MAGICLINKS will be exposed to userspace through openat2(2).

/* Semantics. */
Unlike most other LOOKUP flags (most notably LOOKUP_FOLLOW),
LOOKUP_NO_MAGICLINKS applies to all components of the path.

With LOOKUP_NO_MAGICLINKS, any magic-link path component encountered
during path resolution will yield -ELOOP. The handling of ~LOOKUP_FOLLOW
for a trailing magic-link is identical to LOOKUP_NO_SYMLINKS.

LOOKUP_NO_SYMLINKS implies LOOKUP_NO_MAGICLINKS.

/* Testing. */
LOOKUP_NO_MAGICLINKS is tested as part of the openat2(2) selftests.

[1]: https://lore.kernel.org/lkml/20170429220414.GT29622@ZenIV.linux.org.uk/

14 more lines

diff --git a/fs/namei.c b/fs/namei.c
index 4e85d6fa4048..1f0d871199e5 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -866,7 +866,7 @@ void nd_jump_link(struct path *path)
 
 	nd->path = *path;
 	nd->inode = nd->path.dentry->d_inode;
-	nd->flags |= LOOKUP_JUMPED;
+	nd->flags |= LOOKUP_JUMPED | LOOKUP_MAGICLINK_JUMPED;
 }
 
 static inline void put_link(struct nameidata *nd)
@@ -1063,6 +1063,7 @@ const char *get_link(struct nameidata *nd)
 		return ERR_PTR(error);
 
 	nd->last_type = LAST_BIND;
+	nd->flags &= ~LOOKUP_MAGICLINK_JUMPED;
 	res = READ_ONCE(inode->i_link);
 	if (!res) {
 		const char * (*get)(struct dentry *, struct inode *,
@@ -1078,6 +1079,10 @@ const char *get_link(struct nameidata *nd)
 		} else {
 			res = get(dentry, inode, &last->done);
 		}
+		if (nd->flags & LOOKUP_MAGICLINK_JUMPED) {
+			if (unlikely(nd->flags & LOOKUP_NO_MAGICLINKS))
+				return ERR_PTR(-ELOOP);
+		}
 		if (IS_ERR_OR_NULL(res))
 			return res;
 	}
diff --git a/include/linux/namei.h b/include/linux/namei.h
index ee2e35af387f..a8b3f93338da 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -38,9 +38,11 @@ enum {LAST_NORM, LAST_ROOT, LAST_DOT, LAST_DOTDOT, LAST_BIND};
 #define LOOKUP_JUMPED		0x1000
 #define LOOKUP_ROOT		0x2000
 #define LOOKUP_ROOT_GRABBED	0x0008
+#define LOOKUP_MAGICLINK_JUMPED	0x10000
 
 /* Scoping flags for lookup. */
 #define LOOKUP_NO_SYMLINKS	0x020000 /* No symlink crossing. */
+#define LOOKUP_NO_MAGICLINKS	0x040000 /* No /proc/$pid/fd/ "symlink" crossing. */
 
 extern int path_pts(struct path *path);
 
-- 
2.23.0

[PATCH v15 3/9] namei: LOOKUP_NO_XDEV: block mountpoint crossing

Message details and related content
/* Background. */
The need to contain path operations within a mountpoint has been a
long-standing usecase that userspace has historically implemented
manually with liberal usage of stat(). find, rsync, tar and
many other programs implement these semantics -- but it'd be much
simpler to have a fool-proof way of refusing to open a path if it
crosses a mountpoint.

This is part of a refresh of Al's AT_NO_JUMPS patchset[1] (which was a
variation on David Drysdale's O_BENEATH patchset[2], which in turn was
based on the Capsicum project[3]).

/* Userspace API. */
LOOKUP_NO_XDEV will be exposed to userspace through openat2(2).

/* Semantics. */
Unlike most other LOOKUP flags (most notably LOOKUP_FOLLOW),
LOOKUP_NO_XDEV applies to all components of the path.

With LOOKUP_NO_XDEV, any path component which crosses a mount-point
during path resolution (including "..") will yield an -EXDEV. Absolute
paths, absolute symlinks, and magic-links will only yield an -EXDEV if
the jump involved changing mount-points.

/* Testing. */
LOOKUP_NO_XDEV is tested as part of the openat2(2) selftests.

[1]: https://lore.kernel.org/lkml/20170429220414.GT29622@ZenIV.linux.org.uk/
[2]: https://lore.kernel.org/lkml/1415094884-18349-1-git-send-email-drysdale@google.com/
[3]: https://lore.kernel.org/lkml/1404124096-21445-1-git-send-email-drysdale@google.com/

Cc: Christian Brauner <redacted>
Suggested-by: David Drysdale <redacted>
Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Suggested-by: Andy Lutomirski <luto@kernel.org>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Aleksa Sarai <redacted>
---
 fs/namei.c            | 34 ++++++++++++++++++++++++++++++----
 include/linux/namei.h |  1 +

2 more lines

diff --git a/fs/namei.c b/fs/namei.c
index 1f0d871199e5..b73ee1601bd4 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -504,6 +504,9 @@ struct nameidata {
 	struct filename	*name;
 	struct nameidata *saved;
 	struct inode	*link_inode;
+	struct {
+		bool same_mnt;
+	} last_magiclink;
 	unsigned	root_seq;
 	int		dfd;
 } __randomize_layout;
@@ -837,6 +840,11 @@ static inline void path_to_nameidata(const struct path *path,
 
 static int nd_jump_root(struct nameidata *nd)
 {
+	if (unlikely(nd->flags & LOOKUP_NO_XDEV)) {
+		/* Absolute path arguments to path_init() are allowed. */
+		if (nd->path.mnt != NULL && nd->path.mnt != nd->root.mnt)
+			return -EXDEV;
+	}
 	if (nd->flags & LOOKUP_RCU) {
 		struct dentry *d;
 		nd->path = nd->root;
@@ -862,6 +870,8 @@ static int nd_jump_root(struct nameidata *nd)
 void nd_jump_link(struct path *path)
 {
 	struct nameidata *nd = current->nameidata;
+
+	nd->last_magiclink.same_mnt = (nd->path.mnt == path->mnt);
 	path_put(&nd->path);
 
 	nd->path = *path;
@@ -1082,6 +1092,10 @@ const char *get_link(struct nameidata *nd)
 		if (nd->flags & LOOKUP_MAGICLINK_JUMPED) {
 			if (unlikely(nd->flags & LOOKUP_NO_MAGICLINKS))
 				return ERR_PTR(-ELOOP);
+			if (unlikely(nd->flags & LOOKUP_NO_XDEV)) {
+				if (!nd->last_magiclink.same_mnt)
+					return ERR_PTR(-EXDEV);
+			}
 		}
 		if (IS_ERR_OR_NULL(res))
 			return res;
@@ -1271,12 +1285,16 @@ static int follow_managed(struct path *path, struct nameidata *nd)
 		break;
 	}
 
-	if (need_mntput && path->mnt == mnt)
-		mntput(path->mnt);
+	if (need_mntput) {
+		if (path->mnt == mnt)
+			mntput(path->mnt);
+		if (unlikely(nd->flags & LOOKUP_NO_XDEV))
+			ret = -EXDEV;
+		else
+			nd->flags |= LOOKUP_JUMPED;
+	}
 	if (ret == -EISDIR || !ret)
 		ret = 1;
-	if (need_mntput)
-		nd->flags |= LOOKUP_JUMPED;
 	if (unlikely(ret < 0))
 		path_put_conditional(path, nd);
 	return ret;
@@ -1333,6 +1351,8 @@ static bool __follow_mount_rcu(struct nameidata *nd, struct path *path,
 		mounted = __lookup_mnt(path->mnt, path->dentry);
 		if (!mounted)
 			break;
+		if (unlikely(nd->flags & LOOKUP_NO_XDEV))
+			return false;
 		path->mnt = &mounted->mnt;
 		path->dentry = mounted->mnt.mnt_root;
 		nd->flags |= LOOKUP_JUMPED;
@@ -1379,6 +1399,8 @@ static int follow_dotdot_rcu(struct nameidata *nd)
 				return -ECHILD;
 			if (&mparent->mnt == nd->path.mnt)
 				break;
+			if (unlikely(nd->flags & LOOKUP_NO_XDEV))
+				return -EXDEV;
 			/* we know that mountpoint was pinned */
 			nd->path.dentry = mountpoint;
 			nd->path.mnt = &mparent->mnt;
@@ -1393,6 +1415,8 @@ static int follow_dotdot_rcu(struct nameidata *nd)
 			return -ECHILD;
 		if (!mounted)
 			break;
+		if (unlikely(nd->flags & LOOKUP_NO_XDEV))
+			return -EXDEV;
 		nd->path.mnt = &mounted->mnt;
 		nd->path.dentry = mounted->mnt.mnt_root;
 		inode = nd->path.dentry->d_inode;
@@ -1491,6 +1515,8 @@ static int follow_dotdot(struct nameidata *nd)
 		}
 		if (!follow_up(&nd->path))
 			break;
+		if (unlikely(nd->flags & LOOKUP_NO_XDEV))
+			return -EXDEV;
 	}
 	follow_mount(&nd->path);
 	nd->inode = nd->path.dentry->d_inode;
diff --git a/include/linux/namei.h b/include/linux/namei.h
index a8b3f93338da..6105c8a59fc8 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -43,6 +43,7 @@ enum {LAST_NORM, LAST_ROOT, LAST_DOT, LAST_DOTDOT, LAST_BIND};
 /* Scoping flags for lookup. */
 #define LOOKUP_NO_SYMLINKS	0x020000 /* No symlink crossing. */
 #define LOOKUP_NO_MAGICLINKS	0x040000 /* No /proc/$pid/fd/ "symlink" crossing. */
+#define LOOKUP_NO_XDEV		0x080000 /* No mountpoint crossing. */
 
 extern int path_pts(struct path *path);
 
-- 
2.23.0

[PATCH v15 4/9] namei: LOOKUP_BENEATH: O_BENEATH-like scoped resolution

Message details and related content
/* Background. */
There are many circumstances when userspace wants to resolve a path and
ensure that it doesn't go outside of a particular root directory during
resolution. Obvious examples include archive extraction tools, as well as
other security-conscious userspace programs. FreeBSD spun out O_BENEATH
from their Capsicum project[1,2], so it also seems reasonable to
implement similar functionality for Linux.

This is part of a refresh of Al's AT_NO_JUMPS patchset[3] (which was a
variation on David Drysdale's O_BENEATH patchset[4], which in turn was
based on the Capsicum project[5]).

/* Userspace API. */
LOOKUP_BENEATH will be exposed to userspace through openat2(2).

/* Semantics. */
Unlike most other LOOKUP flags (most notably LOOKUP_FOLLOW),
LOOKUP_BENEATH applies to all components of the path.

With LOOKUP_BENEATH, any path component which attempts to "escape" the
starting point of the filesystem lookup (the dirfd passed to openat)
will yield -EXDEV. Thus, all absolute paths and symlinks are disallowed.

Due to a security concern brought up by Jann[6], any ".." path
components are also blocked. This restriction will be lifted in a future
patch, but requires more work to ensure that permitting ".." is done
safely.

Magic-link jumps are also blocked, because they can beam the path lookup
across the starting point. It would be possible to detect and block
only the "bad" crossings with path_is_under() checks, but it's unclear
whether it makes sense to permit magic-links at all. However, userspace
is recommended to pass LOOKUP_NO_MAGICLINKS if they want to ensure that
magic-link crossing is entirely disabled.

/* Testing. */
LOOKUP_BENEATH is tested as part of the openat2(2) selftests.

[1]: https://reviews.freebsd.org/D2808
[2]: https://reviews.freebsd.org/D17547

16 more lines

diff --git a/fs/namei.c b/fs/namei.c
index b73ee1601bd4..54fdbdfbeb94 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -644,6 +644,14 @@ static bool legitimize_links(struct nameidata *nd)
 
 static bool legitimize_root(struct nameidata *nd)
 {
+	/*
+	 * For scoped-lookups (where nd->root has been zeroed), we need to
+	 * restart the whole lookup from scratch -- because set_root() is wrong
+	 * for these lookups (nd->dfd is the root, not the filesystem root).
+	 */
+	if (!nd->root.mnt && (nd->flags & LOOKUP_IS_SCOPED))
+		return false;
+	/* Nothing to do if nd->root is zero or is managed by the VFS user. */
 	if (!nd->root.mnt || (nd->flags & LOOKUP_ROOT))
 		return true;
 	nd->flags |= LOOKUP_ROOT_GRABBED;
@@ -779,7 +787,11 @@ static int complete_walk(struct nameidata *nd)
 	int status;
 
 	if (nd->flags & LOOKUP_RCU) {
-		if (!(nd->flags & LOOKUP_ROOT))
+		/*
+		 * We don't want to zero nd->root for scoped-lookups or
+		 * externally-managed nd->root.
+		 */
+		if (!(nd->flags & (LOOKUP_ROOT | LOOKUP_IS_SCOPED)))
 			nd->root.mnt = NULL;
 		if (unlikely(unlazy_walk(nd)))
 			return -ECHILD;
@@ -801,10 +813,18 @@ static int complete_walk(struct nameidata *nd)
 	return status;
 }
 
-static void set_root(struct nameidata *nd)
+static int set_root(struct nameidata *nd)
 {
 	struct fs_struct *fs = current->fs;
 
+	/*
+	 * Jumping to the real root in a scoped-lookup is a BUG in namei, but we
+	 * still have to ensure it doesn't happen because it will cause a breakout
+	 * from the dirfd.
+	 */
+	if (WARN_ON(nd->flags & LOOKUP_IS_SCOPED))
+		return -ENOTRECOVERABLE;
+
 	if (nd->flags & LOOKUP_RCU) {
 		unsigned seq;
 
@@ -817,6 +837,7 @@ static void set_root(struct nameidata *nd)
 		get_fs_root(fs, &nd->root);
 		nd->flags |= LOOKUP_ROOT_GRABBED;
 	}
+	return 0;
 }
 
 static void path_put_conditional(struct path *path, struct nameidata *nd)
@@ -840,11 +861,18 @@ static inline void path_to_nameidata(const struct path *path,
 
 static int nd_jump_root(struct nameidata *nd)
 {
+	if (unlikely(nd->flags & LOOKUP_BENEATH))
+		return -EXDEV;
 	if (unlikely(nd->flags & LOOKUP_NO_XDEV)) {
 		/* Absolute path arguments to path_init() are allowed. */
 		if (nd->path.mnt != NULL && nd->path.mnt != nd->root.mnt)
 			return -EXDEV;
 	}
+	if (!nd->root.mnt) {
+		int error = set_root(nd);
+		if (error)
+			return error;
+	}
 	if (nd->flags & LOOKUP_RCU) {
 		struct dentry *d;
 		nd->path = nd->root;
@@ -1096,15 +1124,17 @@ const char *get_link(struct nameidata *nd)
 				if (!nd->last_magiclink.same_mnt)
 					return ERR_PTR(-EXDEV);
 			}
+			/* Not currently safe for scoped-lookups. */
+			if (unlikely(nd->flags & LOOKUP_IS_SCOPED))
+				return ERR_PTR(-EXDEV);
 		}
 		if (IS_ERR_OR_NULL(res))
 			return res;
 	}
 	if (*res == '/') {
-		if (!nd->root.mnt)
-			set_root(nd);
-		if (unlikely(nd_jump_root(nd)))
-			return ERR_PTR(-ECHILD);
+		error = nd_jump_root(nd);
+		if (unlikely(error))
+			return ERR_PTR(error);
 		while (unlikely(*++res == '/'))
 			;
 	}
@@ -1373,8 +1403,11 @@ static int follow_dotdot_rcu(struct nameidata *nd)
 	struct inode *inode = nd->inode;
 
 	while (1) {
-		if (path_equal(&nd->path, &nd->root))
+		if (path_equal(&nd->path, &nd->root)) {
+			if (unlikely(nd->flags & LOOKUP_BENEATH))
+				return -EXDEV;
 			break;
+		}
 		if (nd->path.dentry != nd->path.mnt->mnt_root) {
 			struct dentry *old = nd->path.dentry;
 			struct dentry *parent = old->d_parent;
@@ -1505,8 +1538,11 @@ static int path_parent_directory(struct path *path)
 static int follow_dotdot(struct nameidata *nd)
 {
 	while(1) {
-		if (path_equal(&nd->path, &nd->root))
+		if (path_equal(&nd->path, &nd->root)) {
+			if (unlikely(nd->flags & LOOKUP_BENEATH))
+				return -EXDEV;
 			break;
+		}
 		if (nd->path.dentry != nd->path.mnt->mnt_root) {
 			int ret = path_parent_directory(&nd->path);
 			if (ret)
@@ -1731,8 +1767,20 @@ static inline int may_lookup(struct nameidata *nd)
 static inline int handle_dots(struct nameidata *nd, int type)
 {
 	if (type == LAST_DOTDOT) {
-		if (!nd->root.mnt)
-			set_root(nd);
+		int error = 0;
+
+		/*
+		 * Scoped-lookup flags resolving ".." is not currently safe --
+		 * races can cause our parent to have moved outside of the root
+		 * and us to skip over it.
+		 */
+		if (unlikely(nd->flags & LOOKUP_IS_SCOPED))
+			return -EXDEV;
+		if (!nd->root.mnt) {
+			error = set_root(nd);
+			if (error)
+				return error;
+		}
 		if (nd->flags & LOOKUP_RCU) {
 			return follow_dotdot_rcu(nd);
 		} else
@@ -2195,6 +2243,7 @@ static int link_path_walk(const char *name, struct nameidata *nd)
 /* must be paired with terminate_walk() */
 static const char *path_init(struct nameidata *nd, unsigned flags)
 {
+	int error;
 	const char *s = nd->name->name;
 
 	if (!*s)
@@ -2227,11 +2276,12 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 	nd->path.dentry = NULL;
 
 	nd->m_seq = read_seqbegin(&mount_lock);
+
+	/* Figure out the starting path and root (if needed). */
 	if (*s == '/') {
-		set_root(nd);
-		if (likely(!nd_jump_root(nd)))
-			return s;
-		return ERR_PTR(-ECHILD);
+		error = nd_jump_root(nd);
+		if (unlikely(error))
+			return ERR_PTR(error);
 	} else if (nd->dfd == AT_FDCWD) {
 		if (flags & LOOKUP_RCU) {
 			struct fs_struct *fs = current->fs;
@@ -2247,7 +2297,6 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 			get_fs_pwd(current->fs, &nd->path);
 			nd->inode = nd->path.dentry->d_inode;
 		}
-		return s;
 	} else {
 		/* Caller must check execute permissions on the starting path component */
 		struct fd f = fdget_raw(nd->dfd);
@@ -2272,8 +2321,18 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 			nd->inode = nd->path.dentry->d_inode;
 		}
 		fdput(f);
-		return s;
 	}
+	/* For scoped-lookups we need to set the root to the dirfd as well. */
+	if (flags & LOOKUP_IS_SCOPED) {
+		nd->root = nd->path;
+		if (flags & LOOKUP_RCU) {
+			nd->root_seq = nd->seq;
+		} else {
+			path_get(&nd->root);
+			nd->flags |= LOOKUP_ROOT_GRABBED;
+		}
+	}
+	return s;
 }
 
 static const char *trailing_symlink(struct nameidata *nd)
diff --git a/include/linux/namei.h b/include/linux/namei.h
index 6105c8a59fc8..12f4f36835c2 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -2,6 +2,7 @@
 #ifndef _LINUX_NAMEI_H
 #define _LINUX_NAMEI_H
 
+#include <linux/fs.h>
 #include <linux/kernel.h>
 #include <linux/path.h>
 #include <linux/fcntl.h>
@@ -44,6 +45,9 @@ enum {LAST_NORM, LAST_ROOT, LAST_DOT, LAST_DOTDOT, LAST_BIND};
 #define LOOKUP_NO_SYMLINKS	0x020000 /* No symlink crossing. */
 #define LOOKUP_NO_MAGICLINKS	0x040000 /* No /proc/$pid/fd/ "symlink" crossing. */
 #define LOOKUP_NO_XDEV		0x080000 /* No mountpoint crossing. */
+#define LOOKUP_BENEATH		0x100000 /* No escaping from starting point. */
+/* LOOKUP_* flags which do scope-related checks based on the dirfd. */
+#define LOOKUP_IS_SCOPED LOOKUP_BENEATH
 
 extern int path_pts(struct path *path);
 
-- 
2.23.0

[PATCH v15 5/9] namei: LOOKUP_IN_ROOT: chroot-like scoped resolution

Message details and related content
/* Background. */
Container runtimes or other administrative management processes will
often interact with root filesystems while in the host mount namespace,
because the cost of doing a chroot(2) on every operation is too
prohibitive (especially in Go, which cannot safely use vfork). However,
a malicious program can trick the management process into doing
operations on files outside of the root filesystem through careful
crafting of symlinks.

Most programs that need this feature have attempted to make this process
safe, by doing all of the path resolution in userspace (with symlinks
being scoped to the root of the malicious root filesystem).
Unfortunately, this method is prone to foot-guns and usually such
implementations have subtle security bugs.

Thus, what userspace needs is a way to resolve a path as though it were
in a chroot(2) -- with all absolute symlinks being resolved relative to
the dirfd root (and ".." components being stuck under the dirfd root).
It is much simpler and more straight-forward to provide this
functionality in-kernel (because it can be done far more cheaply and
correctly).

More classical applications that also have this problem (which have
their own potentially buggy userspace path sanitisation code) include
web servers, archive extraction tools, network file servers, and so on.

/* Userspace API. */
LOOKUP_IN_ROOT will be exposed to userspace through openat2(2).

/* Semantics. */
Unlike most other LOOKUP flags (most notably LOOKUP_FOLLOW),
LOOKUP_IN_ROOT applies to all components of the path.

With LOOKUP_IN_ROOT, any path component which attempts to cross the
starting point of the pathname lookup (the dirfd passed to openat) will
remain at the starting point. Thus, all absolute paths and symlinks will
be scoped within the starting point.

There is a slight change in behaviour regarding pathnames -- if the
pathname is absolute then the dirfd is still used as the root of

27 more lines

diff --git a/fs/namei.c b/fs/namei.c
index 54fdbdfbeb94..a3d199a60708 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -2277,12 +2277,20 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 
 	nd->m_seq = read_seqbegin(&mount_lock);
 
-	/* Figure out the starting path and root (if needed). */
-	if (*s == '/') {
+	/* Absolute pathname -- fetch the root. */
+	if (flags & LOOKUP_IN_ROOT) {
+		/* With LOOKUP_IN_ROOT, act as a relative path. */
+		while (*s == '/')
+			s++;
+	} else if (*s == '/') {
 		error = nd_jump_root(nd);
 		if (unlikely(error))
 			return ERR_PTR(error);
-	} else if (nd->dfd == AT_FDCWD) {
+		return s;
+	}
+
+	/* Relative pathname -- get the starting-point it is relative to. */
+	if (nd->dfd == AT_FDCWD) {
 		if (flags & LOOKUP_RCU) {
 			struct fs_struct *fs = current->fs;
 			unsigned seq;
@@ -2322,6 +2330,7 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 		}
 		fdput(f);
 	}
+
 	/* For scoped-lookups we need to set the root to the dirfd as well. */
 	if (flags & LOOKUP_IS_SCOPED) {
 		nd->root = nd->path;
diff --git a/include/linux/namei.h b/include/linux/namei.h
index 12f4f36835c2..96b374e08230 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -46,8 +46,9 @@ enum {LAST_NORM, LAST_ROOT, LAST_DOT, LAST_DOTDOT, LAST_BIND};
 #define LOOKUP_NO_MAGICLINKS	0x040000 /* No /proc/$pid/fd/ "symlink" crossing. */
 #define LOOKUP_NO_XDEV		0x080000 /* No mountpoint crossing. */
 #define LOOKUP_BENEATH		0x100000 /* No escaping from starting point. */
+#define LOOKUP_IN_ROOT		0x200000 /* Treat dirfd as %current->fs->root. */
 /* LOOKUP_* flags which do scope-related checks based on the dirfd. */
-#define LOOKUP_IS_SCOPED LOOKUP_BENEATH
+#define LOOKUP_IS_SCOPED (LOOKUP_BENEATH | LOOKUP_IN_ROOT)
 
 extern int path_pts(struct path *path);
 
-- 
2.23.0

[PATCH v15 6/9] namei: LOOKUP_{IN_ROOT,BENEATH}: permit limited ".." resolution

Message details and related content
Allow LOOKUP_BENEATH and LOOKUP_IN_ROOT to safely permit ".." resolution
(in the case of LOOKUP_BENEATH the resolution will still fail if ".."
resolution would resolve a path outside of the root -- while
LOOKUP_IN_ROOT will chroot(2)-style scope it). Magic-link jumps are
still disallowed entirely[*].

As Jann explains[1,2], the need for this patch (and the original no-".."
restriction) is explained by observing there is a fairly easy-to-exploit
race condition with chroot(2) (and thus by extension LOOKUP_IN_ROOT and
LOOKUP_BENEATH if ".." is allowed) where a rename(2) of a path can be
used to "skip over" nd->root and thus escape to the filesystem above
nd->root.

  thread1 [attacker]:
    for (;;)
      renameat2(AT_FDCWD, "/a/b/c", AT_FDCWD, "/a/d", RENAME_EXCHANGE);
  thread2 [victim]:
    for (;;)
      openat2(dirb, "b/c/../../etc/shadow",
              { .flags = O_PATH, .resolve = RESOLVE_IN_ROOT } );

With fairly significant regularity, thread2 will resolve to
"/etc/shadow" rather than "/a/b/etc/shadow". There is also a similar
(though somewhat more privileged) attack using MS_MOVE.

With this patch, such cases will be detected *during* ".." resolution
and will return -EAGAIN for userspace to decide to either retry or abort
the lookup. It should be noted that ".." is the weak point of chroot(2)
-- walking *into* a subdirectory tautologically cannot result in you
walking *outside* nd->root (except through a bind-mount or magic-link).
There is also no other way for a directory's parent to change (which is
the primary worry with ".." resolution here) other than a rename or
MS_MOVE.

This is a first-pass implementation, where -EAGAIN will be returned if
any rename or mount occurs anywhere on the host (in any namespace). This
will result in spurious errors, but there isn't a satisfactory
alternative (other than denying ".." altogether).

One other possible alternative (which previous versions of this patch

29 more lines

diff --git a/fs/namei.c b/fs/namei.c
index a3d199a60708..174d69cf9084 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -491,7 +491,7 @@ struct nameidata {
 	struct path	root;
 	struct inode	*inode; /* path.dentry.d_inode */
 	unsigned int	flags;
-	unsigned	seq, m_seq;
+	unsigned	seq, m_seq, r_seq;
 	int		last_type;
 	unsigned	depth;
 	int		total_link_count;
@@ -1769,22 +1769,35 @@ static inline int handle_dots(struct nameidata *nd, int type)
 	if (type == LAST_DOTDOT) {
 		int error = 0;
 
-		/*
-		 * Scoped-lookup flags resolving ".." is not currently safe --
-		 * races can cause our parent to have moved outside of the root
-		 * and us to skip over it.
-		 */
-		if (unlikely(nd->flags & LOOKUP_IS_SCOPED))
-			return -EXDEV;
 		if (!nd->root.mnt) {
 			error = set_root(nd);
 			if (error)
 				return error;
 		}
-		if (nd->flags & LOOKUP_RCU) {
-			return follow_dotdot_rcu(nd);
-		} else
-			return follow_dotdot(nd);
+		if (nd->flags & LOOKUP_RCU)
+			error = follow_dotdot_rcu(nd);
+		else
+			error = follow_dotdot(nd);
+		if (error)
+			return error;
+
+		if (unlikely(nd->flags & LOOKUP_IS_SCOPED)) {
+			bool m_retry = read_seqretry(&mount_lock, nd->m_seq);
+			bool r_retry = read_seqretry(&rename_lock, nd->r_seq);
+
+			/*
+			 * If there was a racing rename or mount along our
+			 * path, then we can't be sure that ".." hasn't jumped
+			 * above nd->root (and so userspace should retry or use
+			 * some fallback).
+			 *
+			 * In future we could do a path_is_under() check here
+			 * instead, but there are O(n*m) performance
+			 * considerations with such a setup.
+			 */
+			if (unlikely(m_retry || r_retry))
+				return -EAGAIN;
+		}
 	}
 	return 0;
 }
@@ -2254,6 +2267,10 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 	nd->last_type = LAST_ROOT; /* if there are only slashes... */
 	nd->flags = flags | LOOKUP_JUMPED | LOOKUP_PARENT;
 	nd->depth = 0;
+
+	nd->m_seq = read_seqbegin(&mount_lock);
+	nd->r_seq = read_seqbegin(&rename_lock);
+
 	if (flags & LOOKUP_ROOT) {
 		struct dentry *root = nd->root.dentry;
 		struct inode *inode = root->d_inode;
@@ -2275,7 +2292,6 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 	nd->path.mnt = NULL;
 	nd->path.dentry = NULL;
 
-	nd->m_seq = read_seqbegin(&mount_lock);
 
 	/* Absolute pathname -- fetch the root. */
 	if (flags & LOOKUP_IN_ROOT) {
-- 
2.23.0

[PATCH v15 7/9] open: introduce openat2(2) syscall

Message details and related content
/* Background. */
For a very long time, extending openat(2) with new features has been
incredibly frustrating. This stems from the fact that openat(2) is
possibly the most famous counter-example to the mantra "don't silently
accept garbage from userspace" -- it doesn't check whether unknown flags
are present[1].

This means that (generally) the addition of new flags to openat(2) has
been fraught with backwards-compatibility issues (O_TMPFILE has to be
defined as __O_TMPFILE|O_DIRECTORY|[O_RDWR or O_WRONLY] to ensure old
kernels gave errors, since it's insecure to silently ignore the
flag[2]). All new security-related flags therefore have a tough road to
being added to openat(2).

Userspace also has a hard time figuring out whether a particular flag is
supported on a particular kernel. While it is now possible with
contemporary kernels (thanks to [3]), older kernels will expose unknown
flag bits through fcntl(F_GETFL). Giving a clear -EINVAL during
openat(2) time matches modern syscall designs and is far more
fool-proof.

In addition, the newly-added path resolution restriction LOOKUP flags
(which we would like to expose to user-space) don't feel related to the
pre-existing O_* flag set -- they affect all components of path lookup.
We'd therefore like to add a new flag argument.

Adding a new syscall allows us to finally fix the flag-ignoring problem,
and we can make it extensible enough so that we will hopefully never
need an openat3(2).

/* Syscall Prototype. */
  /*
   * open_how is an extensible structure (similar in interface to
   * clone3(2) or sched_setattr(2)). The size parameter must be set to
   * sizeof(struct open_how), to allow for future extensions. All future
   * extensions will be appended to open_how, with their zero value
   * acting as a no-op default.
   */
  struct open_how { /* ... */ };

101 more lines

diff --git a/CREDITS b/CREDITS
index 031605d46b4d..a048e001d726 100644
--- a/CREDITS
+++ b/CREDITS
@@ -3301,7 +3301,9 @@ S: France
 N: Aleksa Sarai
 E: cyphar@cyphar.com
 W: https://www.cyphar.com/
-D: `pids` cgroup subsystem
+D: /sys/fs/cgroup/pids
+D: openat2(2)
+S: Sydney, Australia
 
 N: Dipankar Sarma
 E: dipankar@in.ibm.com
diff --git a/arch/alpha/kernel/syscalls/syscall.tbl b/arch/alpha/kernel/syscalls/syscall.tbl
index 728fe028c02c..9f374f7d9514 100644
--- a/arch/alpha/kernel/syscalls/syscall.tbl
+++ b/arch/alpha/kernel/syscalls/syscall.tbl
@@ -475,3 +475,4 @@
 543	common	fspick				sys_fspick
 544	common	pidfd_open			sys_pidfd_open
 # 545 reserved for clone3
+547	common	openat2				sys_openat2
diff --git a/arch/arm/tools/syscall.tbl b/arch/arm/tools/syscall.tbl
index 6da7dc4d79cc..4ba54bc7e19a 100644
--- a/arch/arm/tools/syscall.tbl
+++ b/arch/arm/tools/syscall.tbl
@@ -449,3 +449,4 @@
 433	common	fspick				sys_fspick
 434	common	pidfd_open			sys_pidfd_open
 435	common	clone3				sys_clone3
+437	common	openat2				sys_openat2
diff --git a/arch/arm64/include/asm/unistd.h b/arch/arm64/include/asm/unistd.h
index 2629a68b8724..8aa00ccb0b96 100644
--- a/arch/arm64/include/asm/unistd.h
+++ b/arch/arm64/include/asm/unistd.h
@@ -38,7 +38,7 @@
 #define __ARM_NR_compat_set_tls		(__ARM_NR_COMPAT_BASE + 5)
 #define __ARM_NR_COMPAT_END		(__ARM_NR_COMPAT_BASE + 0x800)
 
-#define __NR_compat_syscalls		436
+#define __NR_compat_syscalls		438
 #endif
 
 #define __ARCH_WANT_SYS_CLONE
diff --git a/arch/arm64/include/asm/unistd32.h b/arch/arm64/include/asm/unistd32.h
index 94ab29cf4f00..57f6f592d460 100644
--- a/arch/arm64/include/asm/unistd32.h
+++ b/arch/arm64/include/asm/unistd32.h
@@ -879,6 +879,8 @@ __SYSCALL(__NR_fspick, sys_fspick)
 __SYSCALL(__NR_pidfd_open, sys_pidfd_open)
 #define __NR_clone3 435
 __SYSCALL(__NR_clone3, sys_clone3)
+#define __NR_openat2 437
+__SYSCALL(__NR_openat2, sys_openat2)
 
 /*
  * Please add new compat syscalls above this comment and update
diff --git a/arch/ia64/kernel/syscalls/syscall.tbl b/arch/ia64/kernel/syscalls/syscall.tbl
index 36d5faf4c86c..8d36f2e2dc89 100644
--- a/arch/ia64/kernel/syscalls/syscall.tbl
+++ b/arch/ia64/kernel/syscalls/syscall.tbl
@@ -356,3 +356,4 @@
 433	common	fspick				sys_fspick
 434	common	pidfd_open			sys_pidfd_open
 # 435 reserved for clone3
+437	common	openat2				sys_openat2
diff --git a/arch/m68k/kernel/syscalls/syscall.tbl b/arch/m68k/kernel/syscalls/syscall.tbl
index a88a285a0e5f..2559925f1924 100644
--- a/arch/m68k/kernel/syscalls/syscall.tbl
+++ b/arch/m68k/kernel/syscalls/syscall.tbl
@@ -435,3 +435,4 @@
 433	common	fspick				sys_fspick
 434	common	pidfd_open			sys_pidfd_open
 # 435 reserved for clone3
+437	common	openat2				sys_openat2
diff --git a/arch/microblaze/kernel/syscalls/syscall.tbl b/arch/microblaze/kernel/syscalls/syscall.tbl
index 09b0cd7dab0a..c04385e60833 100644
--- a/arch/microblaze/kernel/syscalls/syscall.tbl
+++ b/arch/microblaze/kernel/syscalls/syscall.tbl
@@ -441,3 +441,4 @@
 433	common	fspick				sys_fspick
 434	common	pidfd_open			sys_pidfd_open
 435	common	clone3				sys_clone3
+437	common	openat2				sys_openat2
diff --git a/arch/mips/kernel/syscalls/syscall_n32.tbl b/arch/mips/kernel/syscalls/syscall_n32.tbl
index e7c5ab38e403..68c9ec06851f 100644
--- a/arch/mips/kernel/syscalls/syscall_n32.tbl
+++ b/arch/mips/kernel/syscalls/syscall_n32.tbl
@@ -374,3 +374,4 @@
 433	n32	fspick				sys_fspick
 434	n32	pidfd_open			sys_pidfd_open
 435	n32	clone3				__sys_clone3
+437	n32	openat2				sys_openat2
diff --git a/arch/mips/kernel/syscalls/syscall_n64.tbl b/arch/mips/kernel/syscalls/syscall_n64.tbl
index 13cd66581f3b..42a72d010050 100644
--- a/arch/mips/kernel/syscalls/syscall_n64.tbl
+++ b/arch/mips/kernel/syscalls/syscall_n64.tbl
@@ -350,3 +350,4 @@
 433	n64	fspick				sys_fspick
 434	n64	pidfd_open			sys_pidfd_open
 435	n64	clone3				__sys_clone3
+437	n64	openat2				sys_openat2
diff --git a/arch/mips/kernel/syscalls/syscall_o32.tbl b/arch/mips/kernel/syscalls/syscall_o32.tbl
index 353539ea4140..f114c4aed0ed 100644
--- a/arch/mips/kernel/syscalls/syscall_o32.tbl
+++ b/arch/mips/kernel/syscalls/syscall_o32.tbl
@@ -423,3 +423,4 @@
 433	o32	fspick				sys_fspick
 434	o32	pidfd_open			sys_pidfd_open
 435	o32	clone3				__sys_clone3
+437	o32	openat2				sys_openat2
diff --git a/arch/parisc/kernel/syscalls/syscall.tbl b/arch/parisc/kernel/syscalls/syscall.tbl
index 285ff516150c..b550ae9a7fea 100644
--- a/arch/parisc/kernel/syscalls/syscall.tbl
+++ b/arch/parisc/kernel/syscalls/syscall.tbl
@@ -433,3 +433,4 @@
 433	common	fspick				sys_fspick
 434	common	pidfd_open			sys_pidfd_open
 435	common	clone3				sys_clone3_wrapper
+437	common	openat2				sys_openat2
diff --git a/arch/powerpc/kernel/syscalls/syscall.tbl b/arch/powerpc/kernel/syscalls/syscall.tbl
index 43f736ed47f2..a8b5ecb5b602 100644
--- a/arch/powerpc/kernel/syscalls/syscall.tbl
+++ b/arch/powerpc/kernel/syscalls/syscall.tbl
@@ -517,3 +517,4 @@
 433	common	fspick				sys_fspick
 434	common	pidfd_open			sys_pidfd_open
 435	nospu	clone3				ppc_clone3
+437	common	openat2				sys_openat2
diff --git a/arch/s390/kernel/syscalls/syscall.tbl b/arch/s390/kernel/syscalls/syscall.tbl
index 3054e9c035a3..16b571c06161 100644
--- a/arch/s390/kernel/syscalls/syscall.tbl
+++ b/arch/s390/kernel/syscalls/syscall.tbl
@@ -438,3 +438,4 @@
 433  common	fspick			sys_fspick			sys_fspick
 434  common	pidfd_open		sys_pidfd_open			sys_pidfd_open
 435  common	clone3			sys_clone3			sys_clone3
+437  common	openat2			sys_openat2			sys_openat2
diff --git a/arch/sh/kernel/syscalls/syscall.tbl b/arch/sh/kernel/syscalls/syscall.tbl
index b5ed26c4c005..a7185cc18626 100644
--- a/arch/sh/kernel/syscalls/syscall.tbl
+++ b/arch/sh/kernel/syscalls/syscall.tbl
@@ -438,3 +438,4 @@
 433	common	fspick				sys_fspick
 434	common	pidfd_open			sys_pidfd_open
 # 435 reserved for clone3
+437	common	openat2				sys_openat2
diff --git a/arch/sparc/kernel/syscalls/syscall.tbl b/arch/sparc/kernel/syscalls/syscall.tbl
index 8c8cc7537fb2..b11c19552022 100644
--- a/arch/sparc/kernel/syscalls/syscall.tbl
+++ b/arch/sparc/kernel/syscalls/syscall.tbl
@@ -481,3 +481,4 @@
 433	common	fspick				sys_fspick
 434	common	pidfd_open			sys_pidfd_open
 # 435 reserved for clone3
+437	common	openat2			sys_openat2
diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl
index 3fe02546aed3..e5c022e9a5c4 100644
--- a/arch/x86/entry/syscalls/syscall_32.tbl
+++ b/arch/x86/entry/syscalls/syscall_32.tbl
@@ -440,3 +440,4 @@
 433	i386	fspick			sys_fspick			__ia32_sys_fspick
 434	i386	pidfd_open		sys_pidfd_open			__ia32_sys_pidfd_open
 435	i386	clone3			sys_clone3			__ia32_sys_clone3
+437	i386	openat2			sys_openat2			__ia32_sys_openat2
diff --git a/arch/x86/entry/syscalls/syscall_64.tbl b/arch/x86/entry/syscalls/syscall_64.tbl
index c29976eca4a8..9035647ef236 100644
--- a/arch/x86/entry/syscalls/syscall_64.tbl
+++ b/arch/x86/entry/syscalls/syscall_64.tbl
@@ -357,6 +357,7 @@
 433	common	fspick			__x64_sys_fspick
 434	common	pidfd_open		__x64_sys_pidfd_open
 435	common	clone3			__x64_sys_clone3/ptregs
+437	common	openat2			__x64_sys_openat2
 
 #
 # x32-specific system call numbers start at 512 to avoid cache impact
diff --git a/arch/xtensa/kernel/syscalls/syscall.tbl b/arch/xtensa/kernel/syscalls/syscall.tbl
index 25f4de729a6d..f0a68013c038 100644
--- a/arch/xtensa/kernel/syscalls/syscall.tbl
+++ b/arch/xtensa/kernel/syscalls/syscall.tbl
@@ -406,3 +406,4 @@
 433	common	fspick				sys_fspick
 434	common	pidfd_open			sys_pidfd_open
 435	common	clone3				sys_clone3
+437	common	openat2				sys_openat2
diff --git a/fs/open.c b/fs/open.c
index b62f5c0923a8..50a46501bcc9 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -955,48 +955,86 @@ struct file *open_with_fake_path(const struct path *path, int flags,
 }
 EXPORT_SYMBOL(open_with_fake_path);
 
-static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
+#define WILL_CREATE(flags)	(flags & (O_CREAT | __O_TMPFILE))
+#define O_PATH_FLAGS		(O_DIRECTORY | O_NOFOLLOW | O_PATH | O_CLOEXEC)
+
+static inline struct open_how build_open_how(int flags, umode_t mode)
+{
+	struct open_how how = {
+		.flags = flags & VALID_OPEN_FLAGS,
+		.mode = mode & S_IALLUGO,
+	};
+
+	/* O_PATH beats everything else. */
+	if (how.flags & O_PATH)
+		how.flags &= O_PATH_FLAGS;
+	/* Modes should only be set for create-like flags. */
+	if (!WILL_CREATE(how.flags))
+		how.mode = 0;
+	return how;
+}
+
+static inline int build_open_flags(const struct open_how *how,
+				   struct open_flags *op)
 {
+	int flags = how->flags;
 	int lookup_flags = 0;
 	int acc_mode = ACC_MODE(flags);
 
+	/* Must never be set by userspace */
+	flags &= ~(FMODE_NONOTIFY | O_CLOEXEC);
+
 	/*
-	 * Clear out all open flags we don't know about so that we don't report
-	 * them in fcntl(F_GETFD) or similar interfaces.
+	 * Older syscalls implicitly clear all of the invalid flags or argument
+	 * values before calling build_open_flags(), but openat2(2) checks all
+	 * of its arguments.
 	 */
-	flags &= VALID_OPEN_FLAGS;
+	if (flags & ~VALID_OPEN_FLAGS)
+		return -EINVAL;
+	if (how->resolve & ~VALID_RESOLVE_FLAGS)
+		return -EINVAL;
+	if (memchr_inv(how->__padding, 0, sizeof(how->__padding)))
+		return -EINVAL;
 
-	if (flags & (O_CREAT | __O_TMPFILE))
-		op->mode = (mode & S_IALLUGO) | S_IFREG;
-	else
+	/* Deal with the mode. */
+	if (WILL_CREATE(flags)) {
+		if (how->mode & ~S_IALLUGO)
+			return -EINVAL;
+		op->mode = how->mode | S_IFREG;
+	} else {
+		if (how->mode != 0)
+			return -EINVAL;
 		op->mode = 0;
-
-	/* Must never be set by userspace */
-	flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
+	}
 
 	/*
-	 * O_SYNC is implemented as __O_SYNC|O_DSYNC.  As many places only
-	 * check for O_DSYNC if the need any syncing at all we enforce it's
-	 * always set instead of having to deal with possibly weird behaviour
-	 * for malicious applications setting only __O_SYNC.
+	 * In order to ensure programs get explicit errors when trying to use
+	 * O_TMPFILE on old kernels, O_TMPFILE is implemented such that it
+	 * looks like (O_DIRECTORY|O_RDWR & ~O_CREAT) to old kernels. But we
+	 * have to require userspace to explicitly set it.
 	 */
-	if (flags & __O_SYNC)
-		flags |= O_DSYNC;
-
 	if (flags & __O_TMPFILE) {
 		if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
 			return -EINVAL;
 		if (!(acc_mode & MAY_WRITE))
 			return -EINVAL;
-	} else if (flags & O_PATH) {
-		/*
-		 * If we have O_PATH in the open flag. Then we
-		 * cannot have anything other than the below set of flags
-		 */
-		flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
+	}
+	if (flags & O_PATH) {
+		/* O_PATH only permits certain other flags to be set. */
+		if (flags & ~O_PATH_FLAGS)
+			return -EINVAL;
 		acc_mode = 0;
 	}
 
+	/*
+	 * O_SYNC is implemented as __O_SYNC|O_DSYNC.  As many places only
+	 * check for O_DSYNC if the need any syncing at all we enforce it's
+	 * always set instead of having to deal with possibly weird behaviour
+	 * for malicious applications setting only __O_SYNC.
+	 */
+	if (flags & __O_SYNC)
+		flags |= O_DSYNC;
+
 	op->open_flag = flags;
 
 	/* O_TRUNC implies we need access checks for write permissions */
@@ -1022,6 +1060,18 @@ static inline int build_open_flags(int flags, umode_t mode, struct open_flags *o
 		lookup_flags |= LOOKUP_DIRECTORY;
 	if (!(flags & O_NOFOLLOW))
 		lookup_flags |= LOOKUP_FOLLOW;
+
+	if (how->resolve & RESOLVE_NO_XDEV)
+		lookup_flags |= LOOKUP_NO_XDEV;
+	if (how->resolve & RESOLVE_NO_MAGICLINKS)
+		lookup_flags |= LOOKUP_NO_MAGICLINKS;
+	if (how->resolve & RESOLVE_NO_SYMLINKS)
+		lookup_flags |= LOOKUP_NO_SYMLINKS;
+	if (how->resolve & RESOLVE_BENEATH)
+		lookup_flags |= LOOKUP_BENEATH;
+	if (how->resolve & RESOLVE_IN_ROOT)
+		lookup_flags |= LOOKUP_IN_ROOT;
+
 	op->lookup_flags = lookup_flags;
 	return 0;
 }
@@ -1040,8 +1090,11 @@ static inline int build_open_flags(int flags, umode_t mode, struct open_flags *o
 struct file *file_open_name(struct filename *name, int flags, umode_t mode)
 {
 	struct open_flags op;
-	int err = build_open_flags(flags, mode, &op);
-	return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op);
+	struct open_how how = build_open_how(flags, mode);
+	int err = build_open_flags(&how, &op);
+	if (err)
+		return ERR_PTR(err);
+	return do_filp_open(AT_FDCWD, name, &op);
 }
 
 /**
@@ -1072,17 +1125,19 @@ struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
 			    const char *filename, int flags, umode_t mode)
 {
 	struct open_flags op;
-	int err = build_open_flags(flags, mode, &op);
+	struct open_how how = build_open_how(flags, mode);
+	int err = build_open_flags(&how, &op);
 	if (err)
 		return ERR_PTR(err);
 	return do_file_open_root(dentry, mnt, filename, &op);
 }
 EXPORT_SYMBOL(file_open_root);
 
-long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
+static long do_sys_openat2(int dfd, const char __user *filename,
+			   struct open_how *how)
 {
 	struct open_flags op;
-	int fd = build_open_flags(flags, mode, &op);
+	int fd = build_open_flags(how, &op);
 	struct filename *tmp;
 
 	if (fd)
@@ -1092,7 +1147,7 @@ long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
 	if (IS_ERR(tmp))
 		return PTR_ERR(tmp);
 
-	fd = get_unused_fd_flags(flags);
+	fd = get_unused_fd_flags(how->flags);
 	if (fd >= 0) {
 		struct file *f = do_filp_open(dfd, tmp, &op);
 		if (IS_ERR(f)) {
@@ -1107,12 +1162,16 @@ long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
 	return fd;
 }
 
-SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
+long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
 {
-	if (force_o_largefile())
-		flags |= O_LARGEFILE;
+	struct open_how how = build_open_how(flags, mode);
+	return do_sys_openat2(dfd, filename, &how);
+}
 
-	return do_sys_open(AT_FDCWD, filename, flags, mode);
+
+SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
+{
+	return ksys_open(filename, flags, mode);
 }
 
 SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
@@ -1120,10 +1179,32 @@ SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
 {
 	if (force_o_largefile())
 		flags |= O_LARGEFILE;
-
 	return do_sys_open(dfd, filename, flags, mode);
 }
 
+SYSCALL_DEFINE4(openat2, int, dfd, const char __user *, filename,
+		struct open_how __user *, how, size_t, usize)
+{
+	int err;
+	struct open_how tmp;
+
+	BUILD_BUG_ON(sizeof(struct open_how) < OPEN_HOW_SIZE_VER0);
+	BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_LATEST);
+
+	if (unlikely(usize < OPEN_HOW_SIZE_VER0))
+		return -EINVAL;
+
+	err = copy_struct_from_user(&tmp, sizeof(tmp), how, usize);
+	if (err)
+		return err;
+
+	/* O_LARGEFILE is only allowed for non-O_PATH. */
+	if (!(tmp.flags & O_PATH) && force_o_largefile())
+		tmp.flags |= O_LARGEFILE;
+
+	return do_sys_openat2(dfd, filename, &tmp);
+}
+
 #ifdef CONFIG_COMPAT
 /*
  * Exactly like sys_open(), except that it doesn't set the
diff --git a/include/linux/fcntl.h b/include/linux/fcntl.h
index d019df946cb2..f2eb05bd3af3 100644
--- a/include/linux/fcntl.h
+++ b/include/linux/fcntl.h
@@ -2,15 +2,25 @@
 #ifndef _LINUX_FCNTL_H
 #define _LINUX_FCNTL_H
 
+#include <linux/stat.h>
 #include <uapi/linux/fcntl.h>
 
-/* list of all valid flags for the open/openat flags argument: */
+/* List of all valid flags for the open/openat flags argument: */
 #define VALID_OPEN_FLAGS \
 	(O_RDONLY | O_WRONLY | O_RDWR | O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC | \
 	 O_APPEND | O_NDELAY | O_NONBLOCK | O_NDELAY | __O_SYNC | O_DSYNC | \
 	 FASYNC	| O_DIRECT | O_LARGEFILE | O_DIRECTORY | O_NOFOLLOW | \
 	 O_NOATIME | O_CLOEXEC | O_PATH | __O_TMPFILE)
 
+/* List of all valid flags for the how->upgrade_mask argument: */
+#define VALID_UPGRADE_FLAGS \
+	(UPGRADE_NOWRITE | UPGRADE_NOREAD)
+
+/* List of all valid flags for the how->resolve argument: */
+#define VALID_RESOLVE_FLAGS \
+	(RESOLVE_NO_XDEV | RESOLVE_NO_MAGICLINKS | RESOLVE_NO_SYMLINKS | \
+	 RESOLVE_BENEATH | RESOLVE_IN_ROOT)
+
 #ifndef force_o_largefile
 #define force_o_largefile() (!IS_ENABLED(CONFIG_ARCH_32BIT_OFF_T))
 #endif
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index f7c561c4dcdd..808f103b7a62 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -69,6 +69,7 @@ struct rseq;
 union bpf_attr;
 struct io_uring_params;
 struct clone_args;
+struct open_how;
 
 #include <linux/types.h>
 #include <linux/aio_abi.h>
@@ -439,6 +440,8 @@ asmlinkage long sys_fchownat(int dfd, const char __user *filename, uid_t user,
 asmlinkage long sys_fchown(unsigned int fd, uid_t user, gid_t group);
 asmlinkage long sys_openat(int dfd, const char __user *filename, int flags,
 			   umode_t mode);
+asmlinkage long sys_openat2(int dfd, const char __user *filename,
+			    struct open_how *how, size_t size);
 asmlinkage long sys_close(unsigned int fd);
 asmlinkage long sys_vhangup(void);
 
diff --git a/include/uapi/asm-generic/unistd.h b/include/uapi/asm-generic/unistd.h
index 1fc8faa6e973..d4122c091472 100644
--- a/include/uapi/asm-generic/unistd.h
+++ b/include/uapi/asm-generic/unistd.h
@@ -851,8 +851,11 @@ __SYSCALL(__NR_pidfd_open, sys_pidfd_open)
 __SYSCALL(__NR_clone3, sys_clone3)
 #endif
 
+#define __NR_openat2 437
+__SYSCALL(__NR_openat2, sys_openat2)
+
 #undef __NR_syscalls
-#define __NR_syscalls 436
+#define __NR_syscalls 438
 
 /*
  * 32 bit systems traditionally used different
diff --git a/include/uapi/linux/fcntl.h b/include/uapi/linux/fcntl.h
index 1d338357df8a..5de8b0006a95 100644
--- a/include/uapi/linux/fcntl.h
+++ b/include/uapi/linux/fcntl.h
@@ -93,5 +93,46 @@
 
 #define AT_RECURSIVE		0x8000	/* Apply to the entire subtree */
 
+/*
+ * Arguments for how openat2(2) should open the target path. If @resolve is
+ * zero, then openat2(2) operates very similarly to openat(2).
+ *
+ * However, unlike openat(2), unknown bits in @flags result in -EINVAL rather
+ * than being silently ignored. @mode must be zero unless one of {O_CREAT,
+ * O_TMPFILE} are set, and @upgrade_mask must be zero unless O_PATH is set.
+ *
+ * @flags: O_* flags.
+ * @mode: O_CREAT/O_TMPFILE file mode.
+ * @upgrade_mask: UPGRADE_* flags (to restrict O_PATH re-opening).
+ * @resolve: RESOLVE_* flags.
+ */
+struct open_how {
+	__aligned_u64 flags;
+	__u16 mode;
+	__u16 __padding[3]; /* must be zeroed */
+	__aligned_u64 resolve;
+};
+
+#define OPEN_HOW_SIZE_VER0	24 /* sizeof first published struct */
+#define OPEN_HOW_SIZE_LATEST	OPEN_HOW_SIZE_VER0
+
+/* how->resolve flags for openat2(2). */
+#define RESOLVE_NO_XDEV		0x01 /* Block mount-point crossings
+					(includes bind-mounts). */
+#define RESOLVE_NO_MAGICLINKS	0x02 /* Block traversal through procfs-style
+					"magic-links". */
+#define RESOLVE_NO_SYMLINKS	0x04 /* Block traversal through all symlinks
+					(implies OEXT_NO_MAGICLINKS) */
+#define RESOLVE_BENEATH		0x08 /* Block "lexical" trickery like
+					"..", symlinks, and absolute
+					paths which escape the dirfd. */
+#define RESOLVE_IN_ROOT		0x10 /* Make all jumps to "/" and ".."
+					be scoped inside the dirfd
+					(similar to chroot(2)). */
+
+/* how->upgrade flags for openat2(2). */
+/* First bit is reserved for a future UPGRADE_NOEXEC flag. */
+#define UPGRADE_NOREAD		0x02 /* Block re-opening with MAY_READ. */
+#define UPGRADE_NOWRITE		0x04 /* Block re-opening with MAY_WRITE. */
 
 #endif /* _UAPI_LINUX_FCNTL_H */
-- 
2.23.0

[PATCH v15 8/9] selftests: add openat2(2) selftests

Message details and related content
Test all of the various openat2(2) flags. A small stress-test of a
symlink-rename attack is included to show that the protections against
".."-based attacks are sufficient.

The main things these self-tests are enforcing are:

  * The struct+usize ABI for openat2(2) and copy_struct_from_user() to
    ensure that upgrades will be handled gracefully (in addition,
    ensuring that misaligned structures are also handled correctly).

  * The -EINVAL checks for openat2(2) are all correctly handled to avoid
    userspace passing unknown or conflicting flag sets (most
    importantly, ensuring that invalid flag combinations are checked).

  * All of the RESOLVE_* semantics (including errno values) are
    correctly handled with various combinations of paths and flags.

  * RESOLVE_IN_ROOT correctly protects against the symlink rename(2)
    attack that has been responsible for several CVEs (and likely will
    be responsible for several more).

Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Aleksa Sarai <redacted>
---
 tools/testing/selftests/Makefile              |   1 +
 tools/testing/selftests/openat2/.gitignore    |   1 +
 tools/testing/selftests/openat2/Makefile      |   8 +
 tools/testing/selftests/openat2/helpers.c     | 109 ++++
 tools/testing/selftests/openat2/helpers.h     | 107 ++++
 .../testing/selftests/openat2/openat2_test.c  | 316 +++++++++++
 .../selftests/openat2/rename_attack_test.c    | 160 ++++++
 .../testing/selftests/openat2/resolve_test.c  | 523 ++++++++++++++++++
 8 files changed, 1225 insertions(+)
 create mode 100644 tools/testing/selftests/openat2/.gitignore
 create mode 100644 tools/testing/selftests/openat2/Makefile
 create mode 100644 tools/testing/selftests/openat2/helpers.c
 create mode 100644 tools/testing/selftests/openat2/helpers.h
 create mode 100644 tools/testing/selftests/openat2/openat2_test.c
 create mode 100644 tools/testing/selftests/openat2/rename_attack_test.c
 create mode 100644 tools/testing/selftests/openat2/resolve_test.c

1 more lines

diff --git a/tools/testing/selftests/Makefile b/tools/testing/selftests/Makefile
index 4cdbae6f4e61..28996856ed5e 100644
--- a/tools/testing/selftests/Makefile
+++ b/tools/testing/selftests/Makefile
@@ -37,6 +37,7 @@ TARGETS += powerpc
 TARGETS += proc
 TARGETS += pstore
 TARGETS += ptrace
+TARGETS += openat2
 TARGETS += rseq
 TARGETS += rtc
 TARGETS += seccomp
diff --git a/tools/testing/selftests/openat2/.gitignore b/tools/testing/selftests/openat2/.gitignore
new file mode 100644
index 000000000000..bd68f6c3fd07
--- /dev/null
+++ b/tools/testing/selftests/openat2/.gitignore
@@ -0,0 +1 @@
+/*_test
diff --git a/tools/testing/selftests/openat2/Makefile b/tools/testing/selftests/openat2/Makefile
new file mode 100644
index 000000000000..4b93b1417b86
--- /dev/null
+++ b/tools/testing/selftests/openat2/Makefile
@@ -0,0 +1,8 @@
+# SPDX-License-Identifier: GPL-2.0-or-later
+
+CFLAGS += -Wall -O2 -g -fsanitize=address -fsanitize=undefined
+TEST_GEN_PROGS := openat2_test resolve_test rename_attack_test
+
+include ../lib.mk
+
+$(TEST_GEN_PROGS): helpers.c
diff --git a/tools/testing/selftests/openat2/helpers.c b/tools/testing/selftests/openat2/helpers.c
new file mode 100644
index 000000000000..e9a6557ab16f
--- /dev/null
+++ b/tools/testing/selftests/openat2/helpers.c
@@ -0,0 +1,109 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Author: Aleksa Sarai <cyphar@cyphar.com>
+ * Copyright (C) 2018-2019 SUSE LLC.
+ */
+
+#define _GNU_SOURCE
+#include <errno.h>
+#include <fcntl.h>
+#include <stdbool.h>
+#include <string.h>
+#include <syscall.h>
+#include <limits.h>
+
+#include "helpers.h"
+
+bool needs_openat2(const struct open_how *how)
+{
+	return how->resolve != 0;
+}
+
+int raw_openat2(int dfd, const char *path, void *how, size_t size)
+{
+	int ret = syscall(__NR_openat2, dfd, path, how, size);
+	return ret >= 0 ? ret : -errno;
+}
+
+int sys_openat2(int dfd, const char *path, struct open_how *how)
+{
+	return raw_openat2(dfd, path, how, sizeof(*how));
+}
+
+int sys_openat(int dfd, const char *path, struct open_how *how)
+{
+	int ret = openat(dfd, path, how->flags, how->mode);
+	return ret >= 0 ? ret : -errno;
+}
+
+int sys_renameat2(int olddirfd, const char *oldpath,
+		  int newdirfd, const char *newpath, unsigned int flags)
+{
+	int ret = syscall(__NR_renameat2, olddirfd, oldpath,
+					  newdirfd, newpath, flags);
+	return ret >= 0 ? ret : -errno;
+}
+
+int touchat(int dfd, const char *path)
+{
+	int fd = openat(dfd, path, O_CREAT);
+	if (fd >= 0)
+		close(fd);
+	return fd;
+}
+
+char *fdreadlink(int fd)
+{
+	char *target, *tmp;
+
+	E_asprintf(&tmp, "/proc/self/fd/%d", fd);
+
+	target = malloc(PATH_MAX);
+	if (!target)
+		ksft_exit_fail_msg("fdreadlink: malloc failed\n");
+	memset(target, 0, PATH_MAX);
+
+	E_readlink(tmp, target, PATH_MAX);
+	free(tmp);
+	return target;
+}
+
+bool fdequal(int fd, int dfd, const char *path)
+{
+	char *fdpath, *dfdpath, *other;
+	bool cmp;
+
+	fdpath = fdreadlink(fd);
+	dfdpath = fdreadlink(dfd);
+
+	if (!path)
+		E_asprintf(&other, "%s", dfdpath);
+	else if (*path == '/')
+		E_asprintf(&other, "%s", path);
+	else
+		E_asprintf(&other, "%s/%s", dfdpath, path);
+
+	cmp = !strcmp(fdpath, other);
+
+	free(fdpath);
+	free(dfdpath);
+	free(other);
+	return cmp;
+}
+
+bool openat2_supported = false;
+
+void __attribute__((constructor)) init(void)
+{
+	struct open_how how = {};
+	int fd;
+
+	BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_VER0);
+
+	/* Check openat2(2) support. */
+	fd = sys_openat2(AT_FDCWD, ".", &how);
+	openat2_supported = (fd >= 0);
+
+	if (fd >= 0)
+		close(fd);
+}
diff --git a/tools/testing/selftests/openat2/helpers.h b/tools/testing/selftests/openat2/helpers.h
new file mode 100644
index 000000000000..43ca5ceab6e3
--- /dev/null
+++ b/tools/testing/selftests/openat2/helpers.h
@@ -0,0 +1,107 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Author: Aleksa Sarai <cyphar@cyphar.com>
+ * Copyright (C) 2018-2019 SUSE LLC.
+ */
+
+#ifndef __RESOLVEAT_H__
+#define __RESOLVEAT_H__
+
+#define _GNU_SOURCE
+#include <stdint.h>
+#include <errno.h>
+#include <linux/types.h>
+#include "../kselftest.h"
+
+#define ARRAY_LEN(X) (sizeof (X) / sizeof (*(X)))
+#define BUILD_BUG_ON(e) ((void)(sizeof(struct { int:(-!!(e)); })))
+
+#ifndef SYS_openat2
+#ifndef __NR_openat2
+#define __NR_openat2 437
+#endif /* __NR_openat2 */
+#define SYS_openat2 __NR_openat2
+#endif /* SYS_openat2 */
+
+/*
+ * Arguments for how openat2(2) should open the target path. If @resolve is
+ * zero, then openat2(2) operates very similarly to openat(2).
+ *
+ * However, unlike openat(2), unknown bits in @flags result in -EINVAL rather
+ * than being silently ignored. @mode must be zero unless one of {O_CREAT,
+ * O_TMPFILE} are set.
+ *
+ * @flags: O_* flags.
+ * @mode: O_CREAT/O_TMPFILE file mode.
+ * @resolve: RESOLVE_* flags.
+ */
+struct open_how {
+	__aligned_u64 flags;
+	__u16 mode;
+	__u16 __padding[3]; /* must be zeroed */
+	__aligned_u64 resolve;
+};
+
+#define OPEN_HOW_SIZE_VER0	24 /* sizeof first published struct */
+#define OPEN_HOW_SIZE_LATEST	OPEN_HOW_SIZE_VER0
+
+bool needs_openat2(const struct open_how *how);
+
+#ifndef RESOLVE_IN_ROOT
+/* how->resolve flags for openat2(2). */
+#define RESOLVE_NO_XDEV		0x01 /* Block mount-point crossings
+					(includes bind-mounts). */
+#define RESOLVE_NO_MAGICLINKS	0x02 /* Block traversal through procfs-style
+					"magic-links". */
+#define RESOLVE_NO_SYMLINKS	0x04 /* Block traversal through all symlinks
+					(implies OEXT_NO_MAGICLINKS) */
+#define RESOLVE_BENEATH		0x08 /* Block "lexical" trickery like
+					"..", symlinks, and absolute
+					paths which escape the dirfd. */
+#define RESOLVE_IN_ROOT		0x10 /* Make all jumps to "/" and ".."
+					be scoped inside the dirfd
+					(similar to chroot(2)). */
+#endif /* RESOLVE_IN_ROOT */
+
+#define E_func(func, ...)						\
+	do {								\
+		if (func(__VA_ARGS__) < 0)				\
+			ksft_exit_fail_msg("%s:%d %s failed\n", \
+					   __FILE__, __LINE__, #func);\
+	} while (0)
+
+#define E_asprintf(...)		E_func(asprintf,	__VA_ARGS__)
+#define E_chmod(...)		E_func(chmod,		__VA_ARGS__)
+#define E_dup2(...)		E_func(dup2,		__VA_ARGS__)
+#define E_fchdir(...)		E_func(fchdir,		__VA_ARGS__)
+#define E_fstatat(...)		E_func(fstatat,		__VA_ARGS__)
+#define E_kill(...)		E_func(kill,		__VA_ARGS__)
+#define E_mkdirat(...)		E_func(mkdirat,		__VA_ARGS__)
+#define E_mount(...)		E_func(mount,		__VA_ARGS__)
+#define E_prctl(...)		E_func(prctl,		__VA_ARGS__)
+#define E_readlink(...)		E_func(readlink,	__VA_ARGS__)
+#define E_setresuid(...)	E_func(setresuid,	__VA_ARGS__)
+#define E_symlinkat(...)	E_func(symlinkat,	__VA_ARGS__)
+#define E_touchat(...)		E_func(touchat,		__VA_ARGS__)
+#define E_unshare(...)		E_func(unshare,		__VA_ARGS__)
+
+#define E_assert(expr, msg, ...)					\
+	do {								\
+		if (!(expr))						\
+			ksft_exit_fail_msg("ASSERT(%s:%d) failed (%s): " msg "\n", \
+					   __FILE__, __LINE__, #expr, ##__VA_ARGS__); \
+	} while (0)
+
+int raw_openat2(int dfd, const char *path, void *how, size_t size);
+int sys_openat2(int dfd, const char *path, struct open_how *how);
+int sys_openat(int dfd, const char *path, struct open_how *how);
+int sys_renameat2(int olddirfd, const char *oldpath,
+		  int newdirfd, const char *newpath, unsigned int flags);
+
+int touchat(int dfd, const char *path);
+char *fdreadlink(int fd);
+bool fdequal(int fd, int dfd, const char *path);
+
+extern bool openat2_supported;
+
+#endif /* __RESOLVEAT_H__ */
diff --git a/tools/testing/selftests/openat2/openat2_test.c b/tools/testing/selftests/openat2/openat2_test.c
new file mode 100644
index 000000000000..8a641acb0d6c
--- /dev/null
+++ b/tools/testing/selftests/openat2/openat2_test.c
@@ -0,0 +1,316 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Author: Aleksa Sarai <cyphar@cyphar.com>
+ * Copyright (C) 2018-2019 SUSE LLC.
+ */
+
+#define _GNU_SOURCE
+#include <fcntl.h>
+#include <sched.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <sys/mount.h>
+#include <stdlib.h>
+#include <stdbool.h>
+#include <string.h>
+
+#include "../kselftest.h"
+#include "helpers.h"
+
+/*
+ * O_LARGEFILE is set to 0 by glibc.
+ * XXX: This is wrong on {mips, parisc, powerpc, sparc}.
+ */
+#undef	O_LARGEFILE
+#define	O_LARGEFILE 0x8000
+
+struct open_how_ext {
+	struct open_how inner;
+	uint32_t extra1;
+	char pad1[128];
+	uint32_t extra2;
+	char pad2[128];
+	uint32_t extra3;
+};
+
+struct struct_test {
+	const char *name;
+	struct open_how_ext arg;
+	size_t size;
+	int err;
+};
+
+#define NUM_OPENAT2_STRUCT_TESTS 9
+#define NUM_OPENAT2_STRUCT_VARIATIONS 13
+
+void test_openat2_struct(void)
+{
+	int misalignments[] = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 11, 17, 87 };
+
+	struct struct_test tests[] = {
+		/* Normal struct. */
+		{ .name = "normal struct",
+		  .arg.inner.flags = O_RDONLY,
+		  .size = sizeof(struct open_how) },
+		/* Bigger struct, with zeroed out end. */
+		{ .name = "bigger struct (zeroed out)",
+		  .arg.inner.flags = O_RDONLY,
+		  .size = sizeof(struct open_how_ext) },
+
+		/* Normal struct with broken padding. */
+		{ .name = "normal struct (non-zero padding[0])",
+		  .arg.inner.flags = O_RDONLY,
+		  .arg.inner.__padding = {0xa0, 0x00},
+		  .size = sizeof(struct open_how_ext), .err = -EINVAL },
+		{ .name = "normal struct (non-zero padding[1])",
+		  .arg.inner.flags = O_RDONLY,
+		  .arg.inner.__padding = {0x00, 0x1a},
+		  .size = sizeof(struct open_how_ext), .err = -EINVAL },
+
+		/* TODO: Once expanded, check zero-padding. */
+
+		/* Smaller than version-0 struct. */
+		{ .name = "zero-sized 'struct'",
+		  .arg.inner.flags = O_RDONLY, .size = 0, .err = -EINVAL },
+		{ .name = "smaller-than-v0 struct",
+		  .arg.inner.flags = O_RDONLY,
+		  .size = OPEN_HOW_SIZE_VER0 - 1, .err = -EINVAL },
+
+		/* Bigger struct, with non-zero trailing bytes. */
+		{ .name = "bigger struct (non-zero data in first 'future field')",
+		  .arg.inner.flags = O_RDONLY, .arg.extra1 = 0xdeadbeef,
+		  .size = sizeof(struct open_how_ext), .err = -E2BIG },
+		{ .name = "bigger struct (non-zero data in middle of 'future fields')",
+		  .arg.inner.flags = O_RDONLY, .arg.extra2 = 0xfeedcafe,
+		  .size = sizeof(struct open_how_ext), .err = -E2BIG },
+		{ .name = "bigger struct (non-zero data at end of 'future fields')",
+		  .arg.inner.flags = O_RDONLY, .arg.extra3 = 0xabad1dea,
+		  .size = sizeof(struct open_how_ext), .err = -E2BIG },
+	};
+
+	BUILD_BUG_ON(ARRAY_LEN(misalignments) != NUM_OPENAT2_STRUCT_VARIATIONS);
+	BUILD_BUG_ON(ARRAY_LEN(tests) != NUM_OPENAT2_STRUCT_TESTS);
+
+	for (int i = 0; i < ARRAY_LEN(tests); i++) {
+		struct struct_test *test = &tests[i];
+		struct open_how_ext how_ext = test->arg;
+
+		for (int j = 0; j < ARRAY_LEN(misalignments); j++) {
+			int fd, misalign = misalignments[j];
+			char *fdpath = NULL;
+			bool failed;
+			void (*resultfn)(const char *msg, ...) = ksft_test_result_pass;
+
+			void *copy = NULL, *how_copy = &how_ext;
+
+			if (!openat2_supported) {
+				ksft_print_msg("openat2(2) unsupported\n");
+				resultfn = ksft_test_result_skip;
+				goto skip;
+			}
+
+			if (misalign) {
+				/*
+				 * Explicitly misalign the structure copying it with the given
+				 * (mis)alignment offset. The other data is set to be non-zero to
+				 * make sure that non-zero bytes outside the struct aren't checked
+				 *
+				 * This is effectively to check that is_zeroed_user() works.
+				 */
+				copy = malloc(misalign + sizeof(how_ext));
+				how_copy = copy + misalign;
+				memset(copy, 0xff, misalign);
+				memcpy(how_copy, &how_ext, sizeof(how_ext));
+			}
+
+			fd = raw_openat2(AT_FDCWD, ".", how_copy, test->size);
+			if (test->err >= 0)
+				failed = (fd < 0);
+			else
+				failed = (fd != test->err);
+			if (fd >= 0) {
+				fdpath = fdreadlink(fd);
+				close(fd);
+			}
+
+			if (failed) {
+				resultfn = ksft_test_result_fail;
+
+				ksft_print_msg("openat2 unexpectedly returned ");
+				if (fdpath)
+					ksft_print_msg("%d['%s']\n", fd, fdpath);
+				else
+					ksft_print_msg("%d (%s)\n", fd, strerror(-fd));
+			}
+
+skip:
+			if (test->err >= 0)
+				resultfn("openat2 with %s argument [misalign=%d] succeeds\n",
+					 test->name, misalign);
+			else
+				resultfn("openat2 with %s argument [misalign=%d] fails with %d (%s)\n",
+					 test->name, misalign, test->err,
+					 strerror(-test->err));
+
+			free(copy);
+			free(fdpath);
+			fflush(stdout);
+		}
+	}
+}
+
+struct flag_test {
+	const char *name;
+	struct open_how how;
+	int err;
+};
+
+#define NUM_OPENAT2_FLAG_TESTS 21
+
+void test_openat2_flags(void)
+{
+	struct flag_test tests[] = {
+		/* O_TMPFILE is incompatible with O_PATH and O_CREAT. */
+		{ .name = "incompatible flags (O_TMPFILE | O_PATH)",
+		  .how.flags = O_TMPFILE | O_PATH | O_RDWR, .err = -EINVAL },
+		{ .name = "incompatible flags (O_TMPFILE | O_CREAT)",
+		  .how.flags = O_TMPFILE | O_CREAT | O_RDWR, .err = -EINVAL },
+
+		/* O_PATH only permits certain other flags to be set ... */
+		{ .name = "compatible flags (O_PATH | O_CLOEXEC)",
+		  .how.flags = O_PATH | O_CLOEXEC },
+		{ .name = "compatible flags (O_PATH | O_DIRECTORY)",
+		  .how.flags = O_PATH | O_DIRECTORY },
+		{ .name = "compatible flags (O_PATH | O_NOFOLLOW)",
+		  .how.flags = O_PATH | O_NOFOLLOW },
+		/* ... and others are absolutely not permitted. */
+		{ .name = "incompatible flags (O_PATH | O_RDWR)",
+		  .how.flags = O_PATH | O_RDWR, .err = -EINVAL },
+		{ .name = "incompatible flags (O_PATH | O_CREAT)",
+		  .how.flags = O_PATH | O_CREAT, .err = -EINVAL },
+		{ .name = "incompatible flags (O_PATH | O_EXCL)",
+		  .how.flags = O_PATH | O_EXCL, .err = -EINVAL },
+		{ .name = "incompatible flags (O_PATH | O_NOCTTY)",
+		  .how.flags = O_PATH | O_NOCTTY, .err = -EINVAL },
+		{ .name = "incompatible flags (O_PATH | O_DIRECT)",
+		  .how.flags = O_PATH | O_DIRECT, .err = -EINVAL },
+		{ .name = "incompatible flags (O_PATH | O_LARGEFILE)",
+		  .how.flags = O_PATH | O_LARGEFILE, .err = -EINVAL },
+
+		/* ->mode must only be set with O_{CREAT,TMPFILE}. */
+		{ .name = "non-zero how.mode and O_RDONLY",
+		  .how.flags = O_RDONLY, .how.mode = 0600, .err = -EINVAL },
+		{ .name = "non-zero how.mode and O_PATH",
+		  .how.flags = O_PATH,   .how.mode = 0600, .err = -EINVAL },
+		{ .name = "valid how.mode and O_CREAT",
+		  .how.flags = O_CREAT,  .how.mode = 0600 },
+		{ .name = "valid how.mode and O_TMPFILE",
+		  .how.flags = O_TMPFILE | O_RDWR, .how.mode = 0600 },
+		/* ->mode must only contain 0777 bits. */
+		{ .name = "invalid how.mode and O_CREAT",
+		  .how.flags = O_CREAT,
+		  .how.mode = 0xFFFF, .err = -EINVAL },
+		{ .name = "invalid how.mode and O_TMPFILE",
+		  .how.flags = O_TMPFILE | O_RDWR,
+		  .how.mode = 0x1337, .err = -EINVAL },
+
+		/* ->resolve must only contain RESOLVE_* flags. */
+		{ .name = "invalid how.resolve and O_RDONLY",
+		  .how.flags = O_RDONLY,
+		  .how.resolve = 0x1337, .err = -EINVAL },
+		{ .name = "invalid how.resolve and O_CREAT",
+		  .how.flags = O_CREAT,
+		  .how.resolve = 0x1337, .err = -EINVAL },
+		{ .name = "invalid how.resolve and O_TMPFILE",
+		  .how.flags = O_TMPFILE | O_RDWR,
+		  .how.resolve = 0x1337, .err = -EINVAL },
+		{ .name = "invalid how.resolve and O_PATH",
+		  .how.flags = O_PATH,
+		  .how.resolve = 0x1337, .err = -EINVAL },
+	};
+
+	BUILD_BUG_ON(ARRAY_LEN(tests) != NUM_OPENAT2_FLAG_TESTS);
+
+	for (int i = 0; i < ARRAY_LEN(tests); i++) {
+		int fd, fdflags = -1;
+		char *path, *fdpath = NULL;
+		bool failed = false;
+		struct flag_test *test = &tests[i];
+		void (*resultfn)(const char *msg, ...) = ksft_test_result_pass;
+
+		if (!openat2_supported) {
+			ksft_print_msg("openat2(2) unsupported\n");
+			resultfn = ksft_test_result_skip;
+			goto skip;
+		}
+
+		path = (test->how.flags & O_CREAT) ? "/tmp/ksft.openat2_tmpfile" : ".";
+		unlink(path);
+
+		fd = sys_openat2(AT_FDCWD, path, &test->how);
+		if (test->err >= 0)
+			failed = (fd < 0);
+		else
+			failed = (fd != test->err);
+		if (fd >= 0) {
+			int otherflags;
+
+			fdpath = fdreadlink(fd);
+			fdflags = fcntl(fd, F_GETFL);
+			otherflags = fcntl(fd, F_GETFD);
+			close(fd);
+
+			E_assert(fdflags >= 0, "fcntl F_GETFL of new fd");
+			E_assert(otherflags >= 0, "fcntl F_GETFD of new fd");
+
+			/* O_CLOEXEC isn't shown in F_GETFL. */
+			if (otherflags & FD_CLOEXEC)
+				fdflags |= O_CLOEXEC;
+			/* O_CREAT is hidden from F_GETFL. */
+			if (test->how.flags & O_CREAT)
+				fdflags |= O_CREAT;
+			if (!(test->how.flags & O_LARGEFILE))
+				fdflags &= ~O_LARGEFILE;
+			failed |= (fdflags != test->how.flags);
+		}
+
+		if (failed) {
+			resultfn = ksft_test_result_fail;
+
+			ksft_print_msg("openat2 unexpectedly returned ");
+			if (fdpath)
+				ksft_print_msg("%d['%s'] with %X (!= %X)\n",
+					       fd, fdpath, fdflags,
+					       test->how.flags);
+			else
+				ksft_print_msg("%d (%s)\n", fd, strerror(-fd));
+		}
+
+skip:
+		if (test->err >= 0)
+			resultfn("openat2 with %s succeeds\n", test->name);
+		else
+			resultfn("openat2 with %s fails with %d (%s)\n",
+				 test->name, test->err, strerror(-test->err));
+
+		free(fdpath);
+		fflush(stdout);
+	}
+}
+
+#define NUM_TESTS (NUM_OPENAT2_STRUCT_VARIATIONS * NUM_OPENAT2_STRUCT_TESTS + \
+		   NUM_OPENAT2_FLAG_TESTS)
+
+int main(int argc, char **argv)
+{
+	ksft_print_header();
+	ksft_set_plan(NUM_TESTS);
+
+	test_openat2_struct();
+	test_openat2_flags();
+
+	if (ksft_get_fail_cnt() + ksft_get_error_cnt() > 0)
+		ksft_exit_fail();
+	else
+		ksft_exit_pass();
+}
diff --git a/tools/testing/selftests/openat2/rename_attack_test.c b/tools/testing/selftests/openat2/rename_attack_test.c
new file mode 100644
index 000000000000..0a770728b436
--- /dev/null
+++ b/tools/testing/selftests/openat2/rename_attack_test.c
@@ -0,0 +1,160 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Author: Aleksa Sarai <cyphar@cyphar.com>
+ * Copyright (C) 2018-2019 SUSE LLC.
+ */
+
+#define _GNU_SOURCE
+#include <errno.h>
+#include <fcntl.h>
+#include <sched.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <sys/mount.h>
+#include <sys/mman.h>
+#include <sys/prctl.h>
+#include <signal.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdbool.h>
+#include <string.h>
+#include <syscall.h>
+#include <limits.h>
+#include <unistd.h>
+
+#include "../kselftest.h"
+#include "helpers.h"
+
+/* Construct a test directory with the following structure:
+ *
+ * root/
+ * |-- a/
+ * |   `-- c/
+ * `-- b/
+ */
+int setup_testdir(void)
+{
+	int dfd;
+	char dirname[] = "/tmp/ksft-openat2-rename-attack.XXXXXX";
+
+	/* Make the top-level directory. */
+	if (!mkdtemp(dirname))
+		ksft_exit_fail_msg("setup_testdir: failed to create tmpdir\n");
+	dfd = open(dirname, O_PATH | O_DIRECTORY);
+	if (dfd < 0)
+		ksft_exit_fail_msg("setup_testdir: failed to open tmpdir\n");
+
+	E_mkdirat(dfd, "a", 0755);
+	E_mkdirat(dfd, "b", 0755);
+	E_mkdirat(dfd, "a/c", 0755);
+
+	return dfd;
+}
+
+/* Swap @dirfd/@a and @dirfd/@b constantly. Parent must kill this process. */
+pid_t spawn_attack(int dirfd, char *a, char *b)
+{
+	pid_t child = fork();
+	if (child != 0)
+		return child;
+
+	/* If the parent (the test process) dies, kill ourselves too. */
+	E_prctl(PR_SET_PDEATHSIG, SIGKILL);
+
+	/* Swap @a and @b. */
+	for (;;)
+		renameat2(dirfd, a, dirfd, b, RENAME_EXCHANGE);
+	exit(1);
+}
+
+#define NUM_RENAME_TESTS 2
+#define ROUNDS 400000
+
+const char *flagname(int resolve)
+{
+	switch (resolve) {
+	case RESOLVE_IN_ROOT:
+		return "RESOLVE_IN_ROOT";
+	case RESOLVE_BENEATH:
+		return "RESOLVE_BENEATH";
+	}
+	return "(unknown)";
+}
+
+void test_rename_attack(int resolve)
+{
+	int dfd, afd;
+	pid_t child;
+	void (*resultfn)(const char *msg, ...) = ksft_test_result_pass;
+	int escapes = 0, other_errs = 0, exdevs = 0, eagains = 0, successes = 0;
+
+	struct open_how how = {
+		.flags = O_PATH,
+		.resolve = resolve,
+	};
+
+	if (!openat2_supported) {
+		how.resolve = 0;
+		ksft_print_msg("openat2(2) unsupported -- using openat(2) instead\n");
+	}
+
+	dfd = setup_testdir();
+	afd = openat(dfd, "a", O_PATH);
+	if (afd < 0)
+		ksft_exit_fail_msg("test_rename_attack: failed to open 'a'\n");
+
+	child = spawn_attack(dfd, "a/c", "b");
+
+	for (int i = 0; i < ROUNDS; i++) {
+		int fd;
+		char *victim_path = "c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../../c/../..";
+
+		if (openat2_supported)
+			fd = sys_openat2(afd, victim_path, &how);
+		else
+			fd = sys_openat(afd, victim_path, &how);
+
+		if (fd < 0) {
+			if (fd == -EAGAIN)
+				eagains++;
+			else if (fd == -EXDEV)
+				exdevs++;
+			else if (fd == -ENOENT)
+				escapes++; /* escaped outside and got ENOENT... */
+			else
+				other_errs++; /* unexpected error */
+		} else {
+			if (fdequal(fd, afd, NULL))
+				successes++;
+			else
+				escapes++; /* we got an unexpected fd */
+		}
+		close(fd);
+	}
+
+	if (escapes > 0)
+		resultfn = ksft_test_result_fail;
+	ksft_print_msg("non-escapes: EAGAIN=%d EXDEV=%d E<other>=%d success=%d\n",
+		       eagains, exdevs, other_errs, successes);
+	resultfn("rename attack with %s (%d runs, got %d escapes)\n",
+		 flagname(resolve), ROUNDS, escapes);
+
+	/* Should be killed anyway, but might as well make sure. */
+	E_kill(child, SIGKILL);
+}
+
+#define NUM_TESTS NUM_RENAME_TESTS
+
+int main(int argc, char **argv)
+{
+	ksft_print_header();
+	ksft_set_plan(NUM_TESTS);
+
+	test_rename_attack(RESOLVE_BENEATH);
+	test_rename_attack(RESOLVE_IN_ROOT);
+
+	if (ksft_get_fail_cnt() + ksft_get_error_cnt() > 0)
+		ksft_exit_fail();
+	else
+		ksft_exit_pass();
+}
diff --git a/tools/testing/selftests/openat2/resolve_test.c b/tools/testing/selftests/openat2/resolve_test.c
new file mode 100644
index 000000000000..7a94b1da8e7b
--- /dev/null
+++ b/tools/testing/selftests/openat2/resolve_test.c
@@ -0,0 +1,523 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Author: Aleksa Sarai <cyphar@cyphar.com>
+ * Copyright (C) 2018-2019 SUSE LLC.
+ */
+
+#define _GNU_SOURCE
+#include <fcntl.h>
+#include <sched.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <sys/mount.h>
+#include <stdlib.h>
+#include <stdbool.h>
+#include <string.h>
+
+#include "../kselftest.h"
+#include "helpers.h"
+
+/*
+ * Construct a test directory with the following structure:
+ *
+ * root/
+ * |-- procexe -> /proc/self/exe
+ * |-- procroot -> /proc/self/root
+ * |-- root/
+ * |-- mnt/ [mountpoint]
+ * |   |-- self -> ../mnt/
+ * |   `-- absself -> /mnt/
+ * |-- etc/
+ * |   `-- passwd
+ * |-- creatlink -> /newfile3
+ * |-- reletc -> etc/
+ * |-- relsym -> etc/passwd
+ * |-- absetc -> /etc/
+ * |-- abssym -> /etc/passwd
+ * |-- abscheeky -> /cheeky
+ * `-- cheeky/
+ *     |-- absself -> /
+ *     |-- self -> ../../root/
+ *     |-- garbageself -> /../../root/
+ *     |-- passwd -> ../cheeky/../cheeky/../etc/../etc/passwd
+ *     |-- abspasswd -> /../cheeky/../cheeky/../etc/../etc/passwd
+ *     |-- dotdotlink -> ../../../../../../../../../../../../../../etc/passwd
+ *     `-- garbagelink -> /../../../../../../../../../../../../../../etc/passwd
+ */
+int setup_testdir(void)
+{
+	int dfd, tmpfd;
+	char dirname[] = "/tmp/ksft-openat2-testdir.XXXXXX";
+
+	/* Unshare and make /tmp a new directory. */
+	E_unshare(CLONE_NEWNS);
+	E_mount("", "/tmp", "", MS_PRIVATE, "");
+
+	/* Make the top-level directory. */
+	if (!mkdtemp(dirname))
+		ksft_exit_fail_msg("setup_testdir: failed to create tmpdir\n");
+	dfd = open(dirname, O_PATH | O_DIRECTORY);
+	if (dfd < 0)
+		ksft_exit_fail_msg("setup_testdir: failed to open tmpdir\n");
+
+	/* A sub-directory which is actually used for tests. */
+	E_mkdirat(dfd, "root", 0755);
+	tmpfd = openat(dfd, "root", O_PATH | O_DIRECTORY);
+	if (tmpfd < 0)
+		ksft_exit_fail_msg("setup_testdir: failed to open tmpdir\n");
+	close(dfd);
+	dfd = tmpfd;
+
+	E_symlinkat("/proc/self/exe", dfd, "procexe");
+	E_symlinkat("/proc/self/root", dfd, "procroot");
+	E_mkdirat(dfd, "root", 0755);
+
+	/* There is no mountat(2), so use chdir. */
+	E_mkdirat(dfd, "mnt", 0755);
+	E_fchdir(dfd);
+	E_mount("tmpfs", "./mnt", "tmpfs", MS_NOSUID | MS_NODEV, "");
+	E_symlinkat("../mnt/", dfd, "mnt/self");
+	E_symlinkat("/mnt/", dfd, "mnt/absself");
+
+	E_mkdirat(dfd, "etc", 0755);
+	E_touchat(dfd, "etc/passwd");
+
+	E_symlinkat("/newfile3", dfd, "creatlink");
+	E_symlinkat("etc/", dfd, "reletc");
+	E_symlinkat("etc/passwd", dfd, "relsym");
+	E_symlinkat("/etc/", dfd, "absetc");
+	E_symlinkat("/etc/passwd", dfd, "abssym");
+	E_symlinkat("/cheeky", dfd, "abscheeky");
+
+	E_mkdirat(dfd, "cheeky", 0755);
+
+	E_symlinkat("/", dfd, "cheeky/absself");
+	E_symlinkat("../../root/", dfd, "cheeky/self");
+	E_symlinkat("/../../root/", dfd, "cheeky/garbageself");
+
+	E_symlinkat("../cheeky/../etc/../etc/passwd", dfd, "cheeky/passwd");
+	E_symlinkat("/../cheeky/../etc/../etc/passwd", dfd, "cheeky/abspasswd");
+
+	E_symlinkat("../../../../../../../../../../../../../../etc/passwd",
+		    dfd, "cheeky/dotdotlink");
+	E_symlinkat("/../../../../../../../../../../../../../../etc/passwd",
+		    dfd, "cheeky/garbagelink");
+
+	return dfd;
+}
+
+struct basic_test {
+	const char *name;
+	const char *dir;
+	const char *path;
+	struct open_how how;
+	bool pass;
+	union {
+		int err;
+		const char *path;
+	} out;
+};
+
+#define NUM_OPENAT2_OPATH_TESTS 88
+
+void test_openat2_opath_tests(void)
+{
+	int rootfd, hardcoded_fd;
+	char *procselfexe, *hardcoded_fdpath;
+
+	E_asprintf(&procselfexe, "/proc/%d/exe", getpid());
+	rootfd = setup_testdir();
+
+	hardcoded_fd = open("/dev/null", O_RDONLY);
+	E_assert(hardcoded_fd >= 0, "open fd to hardcode");
+	E_asprintf(&hardcoded_fdpath, "self/fd/%d", hardcoded_fd);
+
+	struct basic_test tests[] = {
+		/** RESOLVE_BENEATH **/
+		/* Attempts to cross dirfd should be blocked. */
+		{ .name = "[beneath] jump to /",
+		  .path = "/",			.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] absolute link to $root",
+		  .path = "cheeky/absself",	.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] chained absolute links to $root",
+		  .path = "abscheeky/absself",	.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] jump outside $root",
+		  .path = "..",			.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] temporary jump outside $root",
+		  .path = "../root/",		.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] symlink temporary jump outside $root",
+		  .path = "cheeky/self",	.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] chained symlink temporary jump outside $root",
+		  .path = "abscheeky/self",	.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] garbage links to $root",
+		  .path = "cheeky/garbageself",	.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] chained garbage links to $root",
+		  .path = "abscheeky/garbageself", .how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		/* Only relative paths that stay inside dirfd should work. */
+		{ .name = "[beneath] ordinary path to 'root'",
+		  .path = "root",		.how.resolve = RESOLVE_BENEATH,
+		  .out.path = "root",		.pass = true },
+		{ .name = "[beneath] ordinary path to 'etc'",
+		  .path = "etc",		.how.resolve = RESOLVE_BENEATH,
+		  .out.path = "etc",		.pass = true },
+		{ .name = "[beneath] ordinary path to 'etc/passwd'",
+		  .path = "etc/passwd",		.how.resolve = RESOLVE_BENEATH,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[beneath] relative symlink inside $root",
+		  .path = "relsym",		.how.resolve = RESOLVE_BENEATH,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[beneath] chained-'..' relative symlink inside $root",
+		  .path = "cheeky/passwd",	.how.resolve = RESOLVE_BENEATH,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[beneath] absolute symlink component outside $root",
+		  .path = "abscheeky/passwd",	.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] absolute symlink target outside $root",
+		  .path = "abssym",		.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] absolute path outside $root",
+		  .path = "/etc/passwd",	.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] cheeky absolute path outside $root",
+		  .path = "cheeky/abspasswd",	.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] chained cheeky absolute path outside $root",
+		  .path = "abscheeky/abspasswd", .how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		/* Tricky paths should fail. */
+		{ .name = "[beneath] tricky '..'-chained symlink outside $root",
+		  .path = "cheeky/dotdotlink",	.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] tricky absolute + '..'-chained symlink outside $root",
+		  .path = "abscheeky/dotdotlink", .how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] tricky garbage link outside $root",
+		  .path = "cheeky/garbagelink",	.how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[beneath] tricky absolute + garbage link outside $root",
+		  .path = "abscheeky/garbagelink", .how.resolve = RESOLVE_BENEATH,
+		  .out.err = -EXDEV,		.pass = false },
+
+		/** RESOLVE_IN_ROOT **/
+		/* All attempts to cross the dirfd will be scoped-to-root. */
+		{ .name = "[in_root] jump to /",
+		  .path = "/",			.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = NULL,		.pass = true },
+		{ .name = "[in_root] absolute symlink to /root",
+		  .path = "cheeky/absself",	.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = NULL,		.pass = true },
+		{ .name = "[in_root] chained absolute symlinks to /root",
+		  .path = "abscheeky/absself",	.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = NULL,		.pass = true },
+		{ .name = "[in_root] '..' at root",
+		  .path = "..",			.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = NULL,		.pass = true },
+		{ .name = "[in_root] '../root' at root",
+		  .path = "../root/",		.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "root",		.pass = true },
+		{ .name = "[in_root] relative symlink containing '..' above root",
+		  .path = "cheeky/self",	.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "root",		.pass = true },
+		{ .name = "[in_root] garbage link to /root",
+		  .path = "cheeky/garbageself",	.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "root",		.pass = true },
+		{ .name = "[in_root] chainged garbage links to /root",
+		  .path = "abscheeky/garbageself", .how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "root",		.pass = true },
+		{ .name = "[in_root] relative path to 'root'",
+		  .path = "root",		.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "root",		.pass = true },
+		{ .name = "[in_root] relative path to 'etc'",
+		  .path = "etc",		.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc",		.pass = true },
+		{ .name = "[in_root] relative path to 'etc/passwd'",
+		  .path = "etc/passwd",		.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] relative symlink to 'etc/passwd'",
+		  .path = "relsym",		.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] chained-'..' relative symlink to 'etc/passwd'",
+		  .path = "cheeky/passwd",	.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] chained-'..' absolute + relative symlink to 'etc/passwd'",
+		  .path = "abscheeky/passwd",	.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] absolute symlink to 'etc/passwd'",
+		  .path = "abssym",		.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] absolute path 'etc/passwd'",
+		  .path = "/etc/passwd",	.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] cheeky absolute path 'etc/passwd'",
+		  .path = "cheeky/abspasswd",	.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] chained cheeky absolute path 'etc/passwd'",
+		  .path = "abscheeky/abspasswd", .how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] tricky '..'-chained symlink outside $root",
+		  .path = "cheeky/dotdotlink",	.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] tricky absolute + '..'-chained symlink outside $root",
+		  .path = "abscheeky/dotdotlink", .how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] tricky absolute path + absolute + '..'-chained symlink outside $root",
+		  .path = "/../../../../abscheeky/dotdotlink", .how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] tricky garbage link outside $root",
+		  .path = "cheeky/garbagelink",	.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] tricky absolute + garbage link outside $root",
+		  .path = "abscheeky/garbagelink", .how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		{ .name = "[in_root] tricky absolute path + absolute + garbage link outside $root",
+		  .path = "/../../../../abscheeky/garbagelink", .how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "etc/passwd",	.pass = true },
+		/* O_CREAT should handle trailing symlinks correctly. */
+		{ .name = "[in_root] O_CREAT of relative path inside $root",
+		  .path = "newfile1",		.how.flags = O_CREAT,
+						.how.mode = 0700,
+						.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "newfile1",	.pass = true },
+		{ .name = "[in_root] O_CREAT of absolute path",
+		  .path = "/newfile2",		.how.flags = O_CREAT,
+						.how.mode = 0700,
+						.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "newfile2",	.pass = true },
+		{ .name = "[in_root] O_CREAT of tricky symlink outside root",
+		  .path = "/creatlink",		.how.flags = O_CREAT,
+						.how.mode = 0700,
+						.how.resolve = RESOLVE_IN_ROOT,
+		  .out.path = "newfile3",	.pass = true },
+
+		/** RESOLVE_NO_XDEV **/
+		/* Crossing *down* into a mountpoint is disallowed. */
+		{ .name = "[no_xdev] cross into $mnt",
+		  .path = "mnt",		.how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[no_xdev] cross into $mnt/",
+		  .path = "mnt/",		.how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[no_xdev] cross into $mnt/.",
+		  .path = "mnt/.",		.how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,		.pass = false },
+		/* Crossing *up* out of a mountpoint is disallowed. */
+		{ .name = "[no_xdev] goto mountpoint root",
+		  .dir = "mnt", .path = ".",	.how.resolve = RESOLVE_NO_XDEV,
+		  .out.path = "mnt",		.pass = true },
+		{ .name = "[no_xdev] cross up through '..'",
+		  .dir = "mnt", .path = "..",	.how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[no_xdev] temporary cross up through '..'",
+		  .dir = "mnt", .path = "../mnt", .how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[no_xdev] temporary relative symlink cross up",
+		  .dir = "mnt", .path = "self",	.how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[no_xdev] temporary absolute symlink cross up",
+		  .dir = "mnt", .path = "absself", .how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,		.pass = false },
+		/* Jumping to "/" is ok, but later components cannot cross. */
+		{ .name = "[no_xdev] jump to / directly",
+		  .dir = "mnt", .path = "/",	.how.resolve = RESOLVE_NO_XDEV,
+		  .out.path = "/",		.pass = true },
+		{ .name = "[no_xdev] jump to / (from /) directly",
+		  .dir = "/", .path = "/",	.how.resolve = RESOLVE_NO_XDEV,
+		  .out.path = "/",		.pass = true },
+		{ .name = "[no_xdev] jump to / then proc",
+		  .path = "/proc/1",		.how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,		.pass = false },
+		{ .name = "[no_xdev] jump to / then tmp",
+		  .path = "/tmp",		.how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,		.pass = false },
+		/* Magic-links are blocked since they can switch vfsmounts. */
+		{ .name = "[no_xdev] cross through magic-link to self/root",
+		  .dir = "/proc", .path = "self/root", 	.how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,			.pass = false },
+		{ .name = "[no_xdev] cross through magic-link to self/cwd",
+		  .dir = "/proc", .path = "self/cwd",	.how.resolve = RESOLVE_NO_XDEV,
+		  .out.err = -EXDEV,			.pass = false },
+		/* Except magic-link jumps inside the same vfsmount. */
+		{ .name = "[no_xdev] jump through magic-link to same procfs",
+		  .dir = "/proc", .path = hardcoded_fdpath, .how.resolve = RESOLVE_NO_XDEV,
+		  .out.path = "/proc",			    .pass = true, },
+
+		/** RESOLVE_NO_MAGICLINKS **/
+		/* Regular symlinks should work. */
+		{ .name = "[no_magiclinks] ordinary relative symlink",
+		  .path = "relsym",		.how.resolve = RESOLVE_NO_MAGICLINKS,
+		  .out.path = "etc/passwd",	.pass = true },
+		/* Magic-links should not work. */
+		{ .name = "[no_magiclinks] symlink to magic-link",
+		  .path = "procexe",		.how.resolve = RESOLVE_NO_MAGICLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_magiclinks] normal path to magic-link",
+		  .path = "/proc/self/exe",	.how.resolve = RESOLVE_NO_MAGICLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_magiclinks] normal path to magic-link with O_NOFOLLOW",
+		  .path = "/proc/self/exe",	.how.flags = O_NOFOLLOW,
+						.how.resolve = RESOLVE_NO_MAGICLINKS,
+		  .out.path = procselfexe,	.pass = true },
+		{ .name = "[no_magiclinks] symlink to magic-link path component",
+		  .path = "procroot/etc",	.how.resolve = RESOLVE_NO_MAGICLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_magiclinks] magic-link path component",
+		  .path = "/proc/self/root/etc", .how.resolve = RESOLVE_NO_MAGICLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_magiclinks] magic-link path component with O_NOFOLLOW",
+		  .path = "/proc/self/root/etc", .how.flags = O_NOFOLLOW,
+						 .how.resolve = RESOLVE_NO_MAGICLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+
+		/** RESOLVE_NO_SYMLINKS **/
+		/* Normal paths should work. */
+		{ .name = "[no_symlinks] ordinary path to '.'",
+		  .path = ".",			.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.path = NULL,		.pass = true },
+		{ .name = "[no_symlinks] ordinary path to 'root'",
+		  .path = "root",		.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.path = "root",		.pass = true },
+		{ .name = "[no_symlinks] ordinary path to 'etc'",
+		  .path = "etc",		.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.path = "etc",		.pass = true },
+		{ .name = "[no_symlinks] ordinary path to 'etc/passwd'",
+		  .path = "etc/passwd",		.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.path = "etc/passwd",	.pass = true },
+		/* Regular symlinks are blocked. */
+		{ .name = "[no_symlinks] relative symlink target",
+		  .path = "relsym",		.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_symlinks] relative symlink component",
+		  .path = "reletc/passwd",	.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_symlinks] absolute symlink target",
+		  .path = "abssym",		.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_symlinks] absolute symlink component",
+		  .path = "absetc/passwd",	.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_symlinks] cheeky garbage link",
+		  .path = "cheeky/garbagelink",	.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_symlinks] cheeky absolute + garbage link",
+		  .path = "abscheeky/garbagelink", .how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_symlinks] cheeky absolute + absolute symlink",
+		  .path = "abscheeky/absself",	.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		/* Trailing symlinks with NO_FOLLOW. */
+		{ .name = "[no_symlinks] relative symlink with O_NOFOLLOW",
+		  .path = "relsym",		.how.flags = O_NOFOLLOW,
+						.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.path = "relsym",		.pass = true },
+		{ .name = "[no_symlinks] absolute symlink with O_NOFOLLOW",
+		  .path = "abssym",		.how.flags = O_NOFOLLOW,
+						.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.path = "abssym",		.pass = true },
+		{ .name = "[no_symlinks] trailing symlink with O_NOFOLLOW",
+		  .path = "cheeky/garbagelink",	.how.flags = O_NOFOLLOW,
+						.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.path = "cheeky/garbagelink", .pass = true },
+		{ .name = "[no_symlinks] multiple symlink components with O_NOFOLLOW",
+		  .path = "abscheeky/absself",	.how.flags = O_NOFOLLOW,
+						.how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+		{ .name = "[no_symlinks] multiple symlink (and garbage link) components with O_NOFOLLOW",
+		  .path = "abscheeky/garbagelink", .how.flags = O_NOFOLLOW,
+						   .how.resolve = RESOLVE_NO_SYMLINKS,
+		  .out.err = -ELOOP,		.pass = false },
+	};
+
+	BUILD_BUG_ON(ARRAY_LEN(tests) != NUM_OPENAT2_OPATH_TESTS);
+
+	for (int i = 0; i < ARRAY_LEN(tests); i++) {
+		int dfd, fd;
+		char *fdpath = NULL;
+		bool failed;
+		void (*resultfn)(const char *msg, ...) = ksft_test_result_pass;
+		struct basic_test *test = &tests[i];
+
+		if (!openat2_supported) {
+			ksft_print_msg("openat2(2) unsupported\n");
+			resultfn = ksft_test_result_skip;
+			goto skip;
+		}
+
+		/* Auto-set O_PATH. */
+		if (!(test->how.flags & O_CREAT))
+			test->how.flags |= O_PATH;
+
+		if (test->dir)
+			dfd = openat(rootfd, test->dir, O_PATH | O_DIRECTORY);
+		else
+			dfd = dup(rootfd);
+		E_assert(dfd, "failed to openat root '%s': %m", test->dir);
+
+		E_dup2(dfd, hardcoded_fd);
+
+		fd = sys_openat2(dfd, test->path, &test->how);
+		if (test->pass)
+			failed = (fd < 0 || !fdequal(fd, rootfd, test->out.path));
+		else
+			failed = (fd != test->out.err);
+		if (fd >= 0) {
+			fdpath = fdreadlink(fd);
+			close(fd);
+		}
+		close(dfd);
+
+		if (failed) {
+			resultfn = ksft_test_result_fail;
+
+			ksft_print_msg("openat2 unexpectedly returned ");
+			if (fdpath)
+				ksft_print_msg("%d['%s']\n", fd, fdpath);
+			else
+				ksft_print_msg("%d (%s)\n", fd, strerror(-fd));
+		}
+
+skip:
+		if (test->pass)
+			resultfn("%s gives path '%s'\n", test->name,
+				 test->out.path ?: ".");
+		else
+			resultfn("%s fails with %d (%s)\n", test->name,
+				 test->out.err, strerror(-test->out.err));
+
+		fflush(stdout);
+		free(fdpath);
+	}
+
+	free(procselfexe);
+	close(rootfd);
+
+	free(hardcoded_fdpath);
+	close(hardcoded_fd);
+}
+
+#define NUM_TESTS NUM_OPENAT2_OPATH_TESTS
+
+int main(int argc, char **argv)
+{
+	ksft_print_header();
+	ksft_set_plan(NUM_TESTS);
+
+	/* NOTE: We should be checking for CAP_SYS_ADMIN here... */
+	if (geteuid() != 0)
+		ksft_exit_skip("all tests require euid == 0\n");
+
+	test_openat2_opath_tests();
+
+	if (ksft_get_fail_cnt() + ksft_get_error_cnt() > 0)
+		ksft_exit_fail();
+	else
+		ksft_exit_pass();
+}
-- 
2.23.0

[PATCH v15 9/9] Documentation: path-lookup: mention LOOKUP_MAGICLINK_JUMPED

Message details and related content
Now that we have a special flag to signify magic-link jumps, mention it
within the path-lookup docs. And now that "magic link" is the correct
term for nd_jump_link()-style symlinks, clean up references to this type
of "symlink".

Signed-off-by: Aleksa Sarai <redacted>
---
 Documentation/filesystems/path-lookup.rst | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)
diff --git a/Documentation/filesystems/path-lookup.rst b/Documentation/filesystems/path-lookup.rst
index 434a07b0002b..2c32795389bd 100644
--- a/Documentation/filesystems/path-lookup.rst
+++ b/Documentation/filesystems/path-lookup.rst
@@ -405,6 +405,10 @@ is requested.  Keeping a reference in the ``nameidata`` ensures that
 only one root is in effect for the entire path walk, even if it races
 with a ``chroot()`` system call.
 
+It should be noted that in the case of ``LOOKUP_IN_ROOT`` or
+``LOOKUP_BENEATH``, the effective root becomes the directory file descriptor
+passed to ``openat2()`` (which exposes these ``LOOKUP_`` flags).
+
 The root is needed when either of two conditions holds: (1) either the
 pathname or a symbolic link starts with a "'/'", or (2) a "``..``"
 component is being handled, since "``..``" from the root must always stay
@@ -1149,7 +1153,7 @@ so ``NULL`` is returned to indicate that the symlink can be released and
 the stack frame discarded.
 
 The other case involves things in ``/proc`` that look like symlinks but
-aren't really::
+aren't really (and are therefore commonly referred to as "magic-links")::
 
      $ ls -l /proc/self/fd/1
      lrwx------ 1 neilb neilb 64 Jun 13 10:19 /proc/self/fd/1 -> /dev/pts/4
@@ -1310,12 +1314,14 @@ longer needed.
 ``LOOKUP_JUMPED`` means that the current dentry was chosen not because
 it had the right name but for some other reason.  This happens when
 following "``..``", following a symlink to ``/``, crossing a mount point
-or accessing a "``/proc/$PID/fd/$FD``" symlink.  In this case the
-filesystem has not been asked to revalidate the name (with
-``d_revalidate()``).  In such cases the inode may still need to be
-revalidated, so ``d_op->d_weak_revalidate()`` is called if
+or accessing a "``/proc/$PID/fd/$FD``" symlink (also known as a "magic
+link"). In this case the filesystem has not been asked to revalidate the
+name (with ``d_revalidate()``).  In such cases the inode may still need
+to be revalidated, so ``d_op->d_weak_revalidate()`` is called if
 ``LOOKUP_JUMPED`` is set when the look completes - which may be at the
-final component or, when creating, unlinking, or renaming, at the penultimate component.
+final component or, when creating, unlinking, or renaming, at the
+penultimate component. ``LOOKUP_MAGICLINK_JUMPED`` is set alongside
+``LOOKUP_JUMPED`` if a magic-link was traversed.
 
 Final-component flags
 ~~~~~~~~~~~~~~~~~~~~~
-- 
2.23.0
Message details and related content
On 2019-11-05, Aleksa Sarai [off-list ref] wrote:
This patchset is being developed here:
  <https://github.com/cyphar/linux/tree/openat2/master>

Patch changelog:
 v15:
  * Fix code style for LOOKUP_IN_ROOT handling in path_init(). [Linus Torvalds]
  * Split out patches for each individual LOOKUP flag.
  * Reword commit messages to give more background information about the
    series, as well as mention the semantics of each flag in more detail.
 v14: <https://lore.kernel.org/lkml/20191010054140.8483-1-cyphar@cyphar.com/>
      <https://lore.kernel.org/lkml/20191026185700.10708-1-cyphar@cyphar.com>
 v13: <https://lore.kernel.org/lkml/20190930183316.10190-1-cyphar@cyphar.com/>
 v12: <https://lore.kernel.org/lkml/20190904201933.10736-1-cyphar@cyphar.com/>
 v11: <https://lore.kernel.org/lkml/20190820033406.29796-1-cyphar@cyphar.com/>
      <https://lore.kernel.org/lkml/20190728010207.9781-1-cyphar@cyphar.com/>
 v10: <https://lore.kernel.org/lkml/20190719164225.27083-1-cyphar@cyphar.com/>
 v09: <https://lore.kernel.org/lkml/20190706145737.5299-1-cyphar@cyphar.com/>
 v08: <https://lore.kernel.org/lkml/20190520133305.11925-1-cyphar@cyphar.com/>
 v07: <https://lore.kernel.org/lkml/20190507164317.13562-1-cyphar@cyphar.com/>
 v06: <https://lore.kernel.org/lkml/20190506165439.9155-1-cyphar@cyphar.com/>
 v05: <https://lore.kernel.org/lkml/20190320143717.2523-1-cyphar@cyphar.com/>
 v04: <https://lore.kernel.org/lkml/20181112142654.341-1-cyphar@cyphar.com/>
 v03: <https://lore.kernel.org/lkml/20181009070230.12884-1-cyphar@cyphar.com/>
 v02: <https://lore.kernel.org/lkml/20181009065300.11053-1-cyphar@cyphar.com/>
 v01: <https://lore.kernel.org/lkml/20180929103453.12025-1-cyphar@cyphar.com/>

For a very long time, extending openat(2) with new features has been
incredibly frustrating. This stems from the fact that openat(2) is
possibly the most famous counter-example to the mantra "don't silently
accept garbage from userspace" -- it doesn't check whether unknown flags
are present[1].

This means that (generally) the addition of new flags to openat(2) has
been fraught with backwards-compatibility issues (O_TMPFILE has to be
defined as __O_TMPFILE|O_DIRECTORY|[O_RDWR or O_WRONLY] to ensure old
kernels gave errors, since it's insecure to silently ignore the
flag[2]). All new security-related flags therefore have a tough road to
being added to openat(2).

Furthermore, the need for some sort of control over VFS's path resolution (to

369 more quoted lines

Ping -- this patch hasn't been touched for a week. Thanks.

-- 
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>
Message details and related content
On Tue, Nov 12, 2019 at 12:24:04AM +1100, Aleksa Sarai wrote:
On 2019-11-05, Aleksa Sarai [off-list ref] wrote:
quoted
This patchset is being developed here:
  <https://github.com/cyphar/linux/tree/openat2/master>

Patch changelog:
 v15:
  * Fix code style for LOOKUP_IN_ROOT handling in path_init(). [Linus Torvalds]
  * Split out patches for each individual LOOKUP flag.
  * Reword commit messages to give more background information about the
    series, as well as mention the semantics of each flag in more detail.
[...]
Ping -- this patch hasn't been touched for a week. Thanks.
If I've been following correctly, everyone is happy with this series.
(i.e. Linus's comment appear to have been addressed.)

Perhaps the next question is should this go via a pull request by you to
Linus directly during the v5.5 merge window, via akpm, via akpm, via
Christian, or some other path? Besides Linus, it's not been clear who
should "claim" this series. :)

-- 
Kees Cook
Message details and related content
On Tue, Nov 12, 2019 at 03:01:26PM -0800, Kees Cook wrote:
On Tue, Nov 12, 2019 at 12:24:04AM +1100, Aleksa Sarai wrote:
quoted
On 2019-11-05, Aleksa Sarai [off-list ref] wrote:
quoted
This patchset is being developed here:
  <https://github.com/cyphar/linux/tree/openat2/master>

Patch changelog:
 v15:
  * Fix code style for LOOKUP_IN_ROOT handling in path_init(). [Linus Torvalds]
  * Split out patches for each individual LOOKUP flag.
  * Reword commit messages to give more background information about the
    series, as well as mention the semantics of each flag in more detail.
[...]
Ping -- this patch hasn't been touched for a week. Thanks.
If I've been following correctly, everyone is happy with this series.
(i.e. Linus's comment appear to have been addressed.)

Perhaps the next question is should this go via a pull request by you to
Linus directly during the v5.5 merge window, via akpm, via akpm, via
Christian, or some other path? Besides Linus, it's not been clear who
should "claim" this series. :)
I like this series and the same with the copy_struct_from_user() part of
it I've taken I'm happy to stuff this into a dedicated branch, merge it
into my for-next and send it for v5.5.
Though I'd _much_ rather see Al pick this up or have him give his
blessing first.

Christian
Message details and related content
On 2019-11-12, Kees Cook [off-list ref] wrote:
On Tue, Nov 12, 2019 at 12:24:04AM +1100, Aleksa Sarai wrote:
quoted
On 2019-11-05, Aleksa Sarai [off-list ref] wrote:
quoted
This patchset is being developed here:
  <https://github.com/cyphar/linux/tree/openat2/master>

Patch changelog:
 v15:
  * Fix code style for LOOKUP_IN_ROOT handling in path_init(). [Linus Torvalds]
  * Split out patches for each individual LOOKUP flag.
  * Reword commit messages to give more background information about the
    series, as well as mention the semantics of each flag in more detail.
[...]
Ping -- this patch hasn't been touched for a week. Thanks.
If I've been following correctly, everyone is happy with this series.
(i.e. Linus's comment appear to have been addressed.)

Perhaps the next question is should this go via a pull request by you to
Linus directly during the v5.5 merge window, via akpm, via akpm, via
Christian, or some other path? Besides Linus, it's not been clear who
should "claim" this series. :)
Given the namei changes, I wanted to avoid stepping on Al's toes. Though
he did review the series a few versions ago, the discussion didn't focus
on the openat2(2) semantics (which have also changed since then). I'm
not sure whether to interpret the silence to mean he's satisfied with
things as they are, or if he hasn't had more time to look at the series.

As for which tree it should be routed to, I don't mind -- Christian is
the most straight-forward choice (but if Al wants to route it, that's
fine with me too).

-- 
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>

Re: [PATCH v15 2/9] namei: LOOKUP_NO_MAGICLINKS: block magic-link resolution

Message details and related content
On Tue, Nov 05, 2019 at 08:05:46PM +1100, Aleksa Sarai wrote:
quoted hunk
@@ -1078,6 +1079,10 @@ const char *get_link(struct nameidata *nd)
 		} else {
 			res = get(dentry, inode, &last->done);
 		}
+		if (nd->flags & LOOKUP_MAGICLINK_JUMPED) {
+			if (unlikely(nd->flags & LOOKUP_NO_MAGICLINKS))
+				return ERR_PTR(-ELOOP);
+		}
Minor nit - the first check probably wants unlikely() more than the
second one; it's probably noise anyway, but most of the symlinks
traversed are not going to be procfs ones, so you get test + branch
taken most of the time.

OTOH, that just might compile into
	fetch nd->flags
	and with LOOKUP_MAGICLINK_JUMPED | LOOKUP_NO_MAGICLINKS
	compare with the same constant
	unlikely branch when equal

Anyway, that's no more than a minor nit and can be dealt with later (if
at all)

Re: [PATCH v15 3/9] namei: LOOKUP_NO_XDEV: block mountpoint crossing

Message details and related content
On Tue, Nov 05, 2019 at 08:05:47PM +1100, Aleksa Sarai wrote:
quoted hunk
@@ -862,6 +870,8 @@ static int nd_jump_root(struct nameidata *nd)
 void nd_jump_link(struct path *path)
 {
 	struct nameidata *nd = current->nameidata;
+
+	nd->last_magiclink.same_mnt = (nd->path.mnt == path->mnt);
 	path_put(&nd->path);
 
 	nd->path = *path;
@@ -1082,6 +1092,10 @@ const char *get_link(struct nameidata *nd)
 		if (nd->flags & LOOKUP_MAGICLINK_JUMPED) {
 			if (unlikely(nd->flags & LOOKUP_NO_MAGICLINKS))
 				return ERR_PTR(-ELOOP);
+			if (unlikely(nd->flags & LOOKUP_NO_XDEV)) {
+				if (!nd->last_magiclink.same_mnt)
+					return ERR_PTR(-EXDEV);
+			}
 		}
Ugh...  Wouldn't it be better to take that logics (some equivalent thereof)
into nd_jump_link()?  Or just have nd_jump_link() return an error...

I mean, look at the callers of nd_jump_link().
static const char *policy_get_link(struct dentry *dentry,
                                   struct inode *inode,
                                   struct delayed_call *done)
{
        struct aa_ns *ns;
        struct path path;  

        if (!dentry)   
                return ERR_PTR(-ECHILD);
        ns = aa_get_current_ns();
        path.mnt = mntget(aafs_mnt);
        path.dentry = dget(ns_dir(ns));
        nd_jump_link(&path); 
        aa_put_ns(ns);

        return NULL;
}
- very close to the end of ->get_link() instance.

static const char *proc_pid_get_link(struct dentry *dentry,
                                     struct inode *inode,
                                     struct delayed_call *done)
{ 
        struct path path;
        int error = -EACCES;

        if (!dentry)
                return ERR_PTR(-ECHILD);

        /* Are we allowed to snoop on the tasks file descriptors? */
        if (!proc_fd_access_allowed(inode))
                goto out;

        error = PROC_I(inode)->op.proc_get_link(dentry, &path);
        if (error)

38 more lines

diff --git a/fs/namei.c b/fs/namei.c
index 671c3c1a3425..8ec924813c30 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -859,7 +859,7 @@ static int nd_jump_root(struct nameidata *nd)
  * Helper to directly jump to a known parsed path from ->get_link,
  * caller must have taken a reference to path beforehand.
  */
-void nd_jump_link(struct path *path)
+const char *nd_jump_link(struct path *path)
 {
 	struct nameidata *nd = current->nameidata;
 	path_put(&nd->path);
@@ -867,6 +867,7 @@ void nd_jump_link(struct path *path)
 	nd->path = *path;
 	nd->inode = nd->path.dentry->d_inode;
 	nd->flags |= LOOKUP_JUMPED;
+	return NULL;
 }
 
 static inline void put_link(struct nameidata *nd)
diff --git a/fs/proc/base.c b/fs/proc/base.c
index ebea9501afb8..ac4e57a3dfa5 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1626,8 +1626,7 @@ static const char *proc_pid_get_link(struct dentry *dentry,
 	if (error)
 		goto out;
 
-	nd_jump_link(&path);
-	return NULL;
+	return nd_jump_link(&path);
 out:
 	return ERR_PTR(error);
 }
diff --git a/fs/proc/namespaces.c b/fs/proc/namespaces.c
index dd2b35f78b09..dde0c501b2f3 100644
--- a/fs/proc/namespaces.c
+++ b/fs/proc/namespaces.c
@@ -54,7 +54,7 @@ static const char *proc_ns_get_link(struct dentry *dentry,
 	if (ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
 		error = ns_get_path(&ns_path, task, ns_ops);
 		if (!error)
-			nd_jump_link(&ns_path);
+			error = nd_jump_link(&ns_path);
 	}
 	put_task_struct(task);
 	return error;
diff --git a/include/linux/namei.h b/include/linux/namei.h
index 397a08ade6a2..f3e8438e5631 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -68,7 +68,7 @@ extern int follow_up(struct path *);
 extern struct dentry *lock_rename(struct dentry *, struct dentry *);
 extern void unlock_rename(struct dentry *, struct dentry *);
 
-extern void nd_jump_link(struct path *path);
+extern const char *nd_jump_link(struct path *path);
 
 static inline void nd_terminate_link(void *name, size_t len, size_t maxlen)
 {
diff --git a/security/apparmor/apparmorfs.c b/security/apparmor/apparmorfs.c
index 45d13b6462aa..98aef94b4777 100644
--- a/security/apparmor/apparmorfs.c
+++ b/security/apparmor/apparmorfs.c
@@ -2453,18 +2453,16 @@ static const char *policy_get_link(struct dentry *dentry,
 				   struct inode *inode,
 				   struct delayed_call *done)
 {
-	struct aa_ns *ns;
-	struct path path;
-
-	if (!dentry)
-		return ERR_PTR(-ECHILD);
-	ns = aa_get_current_ns();
-	path.mnt = mntget(aafs_mnt);
-	path.dentry = dget(ns_dir(ns));
-	nd_jump_link(&path);
-	aa_put_ns(ns);
-
-	return NULL;
+	const char *err = ERR_PTR(-ECHILD);
+
+	if (dentry) {
+		struct aa_ns *ns = aa_get_current_ns();
+		struct path path = {.mnt = mntget(aafs_mnt),
+				    .dentry = ns_dir(ns)};
+		err = nd_jump_link(&path);
+		aa_put_ns(ns);
+	}
+	return err;
 }
 
 static int policy_readlink(struct dentry *dentry, char __user *buffer,

Re: [PATCH v15 4/9] namei: LOOKUP_BENEATH: O_BENEATH-like scoped resolution

Message details and related content
On Tue, Nov 05, 2019 at 08:05:48PM +1100, Aleksa Sarai wrote:

Minor nit here - I'd split "move the conditional call of set_root()
into nd_jump_root()" into a separate patch before that one.  Makes
for fewer distractions in this one.  I'd probably fold "and be
ready for errors other than -ECHILD" into the same preliminary
patch.
quoted hunk
+			/* Not currently safe for scoped-lookups. */
+			if (unlikely(nd->flags & LOOKUP_IS_SCOPED))
+				return ERR_PTR(-EXDEV);
Also a candidate for doing in nd_jump_link()...
quoted hunk
@@ -1373,8 +1403,11 @@ static int follow_dotdot_rcu(struct nameidata *nd)
 	struct inode *inode = nd->inode;
 
 	while (1) {
-		if (path_equal(&nd->path, &nd->root))
+		if (path_equal(&nd->path, &nd->root)) {
+			if (unlikely(nd->flags & LOOKUP_BENEATH))
+				return -EXDEV;
Umm...  Are you sure it's not -ECHILD?

Re: [PATCH v15 5/9] namei: LOOKUP_IN_ROOT: chroot-like scoped resolution

Message details and related content
On Tue, Nov 05, 2019 at 08:05:49PM +1100, Aleksa Sarai wrote:
quoted hunk
@@ -2277,12 +2277,20 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 
 	nd->m_seq = read_seqbegin(&mount_lock);
 
-	/* Figure out the starting path and root (if needed). */
-	if (*s == '/') {
+	/* Absolute pathname -- fetch the root. */
+	if (flags & LOOKUP_IN_ROOT) {
+		/* With LOOKUP_IN_ROOT, act as a relative path. */
+		while (*s == '/')
+			s++;
Er...  Why bother skipping slashes?  I mean, not only link_path_walk()
will skip them just fine, you are actually risking breakage in this:
                if (*s && unlikely(!d_can_lookup(dentry))) {
                        fdput(f);
                        return ERR_PTR(-ENOTDIR);
                }
which is downstream from there with you patch, AFAICS.

Re: [PATCH v15 6/9] namei: LOOKUP_{IN_ROOT,BENEATH}: permit limited ".." resolution

Message details and related content
On Tue, Nov 05, 2019 at 08:05:50PM +1100, Aleksa Sarai wrote:
One other possible alternative (which previous versions of this patch
used) would be to check with path_is_under() if there was a racing
rename or mount (after re-taking the relevant seqlocks). While this does
work, it results in possible O(n*m) behaviour if there are many renames
or mounts occuring *anywhere on the system*.
BTW, do you realize that open-by-fhandle (or working nfsd, for that matter)
will trigger arseloads of write_seqlock(&rename_lock) simply on d_splice_alias()
bringing disconnected subtrees in contact with parent?

Re: [PATCH v15 7/9] open: introduce openat2(2) syscall

Message details and related content
On Tue, Nov 05, 2019 at 08:05:51PM +1100, Aleksa Sarai wrote:
  
quoted hunk
+/*
+ * Arguments for how openat2(2) should open the target path. If @resolve is
+ * zero, then openat2(2) operates very similarly to openat(2).
+ *
+ * However, unlike openat(2), unknown bits in @flags result in -EINVAL rather
+ * than being silently ignored. @mode must be zero unless one of {O_CREAT,
+ * O_TMPFILE} are set, and @upgrade_mask must be zero unless O_PATH is set.
+ *
+ * @flags: O_* flags.
+ * @mode: O_CREAT/O_TMPFILE file mode.
+ * @upgrade_mask: UPGRADE_* flags (to restrict O_PATH re-opening).
???
quoted hunk
+ * @resolve: RESOLVE_* flags.
+ */
+struct open_how {
+	__aligned_u64 flags;
+	__u16 mode;
+	__u16 __padding[3]; /* must be zeroed */
+	__aligned_u64 resolve;
+};

Re: [PATCH v15 7/9] open: introduce openat2(2) syscall

Message details and related content
On 2019-11-13, Al Viro [off-list ref] wrote:
On Tue, Nov 05, 2019 at 08:05:51PM +1100, Aleksa Sarai wrote:
quoted
+/*
+ * Arguments for how openat2(2) should open the target path. If @resolve is
+ * zero, then openat2(2) operates very similarly to openat(2).
+ *
+ * However, unlike openat(2), unknown bits in @flags result in -EINVAL rather
+ * than being silently ignored. @mode must be zero unless one of {O_CREAT,
+ * O_TMPFILE} are set, and @upgrade_mask must be zero unless O_PATH is set.
+ *
+ * @flags: O_* flags.
+ * @mode: O_CREAT/O_TMPFILE file mode.
+ * @upgrade_mask: UPGRADE_* flags (to restrict O_PATH re-opening).
???
Sorry, that was left over from a previous revision (where the magic-link
re-opening restrictions were part of this series).
quoted
+ * @resolve: RESOLVE_* flags.
+ */
+struct open_how {
+	__aligned_u64 flags;
+	__u16 mode;
+	__u16 __padding[3]; /* must be zeroed */
+	__aligned_u64 resolve;
+};

-- 
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>

Re: [PATCH v15 5/9] namei: LOOKUP_IN_ROOT: chroot-like scoped resolution

Message details and related content
On 2019-11-13, Al Viro [off-list ref] wrote:
On Tue, Nov 05, 2019 at 08:05:49PM +1100, Aleksa Sarai wrote:
quoted
@@ -2277,12 +2277,20 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 
 	nd->m_seq = read_seqbegin(&mount_lock);
 
-	/* Figure out the starting path and root (if needed). */
-	if (*s == '/') {
+	/* Absolute pathname -- fetch the root. */
+	if (flags & LOOKUP_IN_ROOT) {
+		/* With LOOKUP_IN_ROOT, act as a relative path. */
+		while (*s == '/')
+			s++;
Er...  Why bother skipping slashes?  I mean, not only link_path_walk()
will skip them just fine, you are actually risking breakage in this:
                if (*s && unlikely(!d_can_lookup(dentry))) {
                        fdput(f);
                        return ERR_PTR(-ENOTDIR);
                }
which is downstream from there with you patch, AFAICS.
I switched to stripping the slashes at your suggestion a few revisions
ago[1], and had (wrongly) assumed we needed to handle "/" somehow in
path_init(). But you're quite right about link_path_walk() -- and I'd be
more than happy to drop it.

[1]: https://lore.kernel.org/lkml/20190712125552.GL17978@ZenIV.linux.org.uk/

-- 
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>

Re: [PATCH v15 5/9] namei: LOOKUP_IN_ROOT: chroot-like scoped resolution

Message details and related content
On Wed, Nov 13, 2019 at 01:44:14PM +1100, Aleksa Sarai wrote:
On 2019-11-13, Al Viro [off-list ref] wrote:
quoted
On Tue, Nov 05, 2019 at 08:05:49PM +1100, Aleksa Sarai wrote:
quoted
@@ -2277,12 +2277,20 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 
 	nd->m_seq = read_seqbegin(&mount_lock);
 
-	/* Figure out the starting path and root (if needed). */
-	if (*s == '/') {
+	/* Absolute pathname -- fetch the root. */
+	if (flags & LOOKUP_IN_ROOT) {
+		/* With LOOKUP_IN_ROOT, act as a relative path. */
+		while (*s == '/')
+			s++;
Er...  Why bother skipping slashes?  I mean, not only link_path_walk()
will skip them just fine, you are actually risking breakage in this:
                if (*s && unlikely(!d_can_lookup(dentry))) {
                        fdput(f);
                        return ERR_PTR(-ENOTDIR);
                }
which is downstream from there with you patch, AFAICS.
I switched to stripping the slashes at your suggestion a few revisions
ago[1], and had (wrongly) assumed we needed to handle "/" somehow in
path_init(). But you're quite right about link_path_walk() -- and I'd be
more than happy to drop it.
That, IIRC, was about untangling the weirdness around multiple calls of
dirfd_path_init() and basically went "we might want just strip the slashes
in case of that flag very early in the entire thing, so that later the
normal logics for absolute/relative would DTRT".  Since your check is
right next to checking for absolute pathnames (and not in the very
beginning of path_init()), we might as well turn the check for
absolute pathname into *s == '/' && !(flags & LOOKUP_IN_ROOT) and be
done with that.

Re: [PATCH v15 5/9] namei: LOOKUP_IN_ROOT: chroot-like scoped resolution

Message details and related content
On 2019-11-13, Al Viro [off-list ref] wrote:
On Wed, Nov 13, 2019 at 01:44:14PM +1100, Aleksa Sarai wrote:
quoted
On 2019-11-13, Al Viro [off-list ref] wrote:
quoted
On Tue, Nov 05, 2019 at 08:05:49PM +1100, Aleksa Sarai wrote:
quoted
@@ -2277,12 +2277,20 @@ static const char *path_init(struct nameidata *nd, unsigned flags)
 
 	nd->m_seq = read_seqbegin(&mount_lock);
 
-	/* Figure out the starting path and root (if needed). */
-	if (*s == '/') {
+	/* Absolute pathname -- fetch the root. */
+	if (flags & LOOKUP_IN_ROOT) {
+		/* With LOOKUP_IN_ROOT, act as a relative path. */
+		while (*s == '/')
+			s++;
Er...  Why bother skipping slashes?  I mean, not only link_path_walk()
will skip them just fine, you are actually risking breakage in this:
                if (*s && unlikely(!d_can_lookup(dentry))) {
                        fdput(f);
                        return ERR_PTR(-ENOTDIR);
                }
which is downstream from there with you patch, AFAICS.
I switched to stripping the slashes at your suggestion a few revisions
ago[1], and had (wrongly) assumed we needed to handle "/" somehow in
path_init(). But you're quite right about link_path_walk() -- and I'd be
more than happy to drop it.
That, IIRC, was about untangling the weirdness around multiple calls of
dirfd_path_init() and basically went "we might want just strip the slashes
in case of that flag very early in the entire thing, so that later the
normal logics for absolute/relative would DTRT".
Ah okay, I'd misunderstood the point you were making in that thread.
Since your check is right next to checking for absolute pathnames (and
not in the very beginning of path_init()), we might as well turn the
check for absolute pathname into *s == '/' && !(flags &
LOOKUP_IN_ROOT) and be done with that.
Yup, agreed.

-- 
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>

Re: [PATCH v15 4/9] namei: LOOKUP_BENEATH: O_BENEATH-like scoped resolution

Message details and related content
On 2019-11-13, Al Viro [off-list ref] wrote:
Minor nit here - I'd split "move the conditional call of set_root()
into nd_jump_root()" into a separate patch before that one.  Makes
for fewer distractions in this one.  I'd probably fold "and be
ready for errors other than -ECHILD" into the same preliminary
patch.
Will do.
quoted
+			/* Not currently safe for scoped-lookups. */
+			if (unlikely(nd->flags & LOOKUP_IS_SCOPED))
+				return ERR_PTR(-EXDEV);
Also a candidate for doing in nd_jump_link()...
quoted
@@ -1373,8 +1403,11 @@ static int follow_dotdot_rcu(struct nameidata *nd)
 	struct inode *inode = nd->inode;
 
 	while (1) {
-		if (path_equal(&nd->path, &nd->root))
+		if (path_equal(&nd->path, &nd->root)) {
+			if (unlikely(nd->flags & LOOKUP_BENEATH))
+				return -EXDEV;
Umm...  Are you sure it's not -ECHILD?
It wouldn't hurt to be -ECHILD -- though it's not clear to me how likely
a success would be in REF-walk if the parent components didn't already
trigger an unlazy_walk() in RCU-walk.

I guess that also means LOOKUP_NO_XDEV should trigger -ECHILD in
follow_dotdot_rcu()?

-- 
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>

Re: [PATCH v15 6/9] namei: LOOKUP_{IN_ROOT,BENEATH}: permit limited ".." resolution

Message details and related content
On 2019-11-13, Al Viro [off-list ref] wrote:
On Tue, Nov 05, 2019 at 08:05:50PM +1100, Aleksa Sarai wrote:
quoted
One other possible alternative (which previous versions of this patch
used) would be to check with path_is_under() if there was a racing
rename or mount (after re-taking the relevant seqlocks). While this does
work, it results in possible O(n*m) behaviour if there are many renames
or mounts occuring *anywhere on the system*.
BTW, do you realize that open-by-fhandle (or working nfsd, for that matter)
will trigger arseloads of write_seqlock(&rename_lock) simply on d_splice_alias()
bringing disconnected subtrees in contact with parent?
I wasn't aware of that -- that makes path_is_under() even less viable.
I'll reword it to be clearer that path_is_under() isn't a good idea and
why we went with -EAGAIN over an in-kernel retry.

-- 
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>

Re: [PATCH v15 3/9] namei: LOOKUP_NO_XDEV: block mountpoint crossing

Message details and related content
On 2019-11-13, Al Viro [off-list ref] wrote:
On Tue, Nov 05, 2019 at 08:05:47PM +1100, Aleksa Sarai wrote:
quoted
@@ -862,6 +870,8 @@ static int nd_jump_root(struct nameidata *nd)
 void nd_jump_link(struct path *path)
 {
 	struct nameidata *nd = current->nameidata;
+
+	nd->last_magiclink.same_mnt = (nd->path.mnt == path->mnt);
 	path_put(&nd->path);
 
 	nd->path = *path;
@@ -1082,6 +1092,10 @@ const char *get_link(struct nameidata *nd)
 		if (nd->flags & LOOKUP_MAGICLINK_JUMPED) {
 			if (unlikely(nd->flags & LOOKUP_NO_MAGICLINKS))
 				return ERR_PTR(-ELOOP);
+			if (unlikely(nd->flags & LOOKUP_NO_XDEV)) {
+				if (!nd->last_magiclink.same_mnt)
+					return ERR_PTR(-EXDEV);
+			}
 		}
Ugh...  Wouldn't it be better to take that logics (some equivalent thereof)
into nd_jump_link()?  Or just have nd_jump_link() return an error...
This could be done, but the reason for stashing it away in
last_magiclink is because of the future magic-link re-opening patches
which can't be implemented like that without putting the open_flags
inside nameidata (which was decided to be too ugly a while ago).

My point being that I could implement it this way for this series, but
I'd have to implement something like last_magiclink when I end up
re-posting the magic-link stuff in a few weeks.

Looking at all the nd_jump_link() users, the other option is to just
disallow magic-link crossings entirely for LOOKUP_NO_XDEV. The only
thing allowing them permits is to resolve file descriptors that are
pointing to the same procfs mount -- and it's unclear to me how useful
that really is (apparmorfs and nsfs will always give -EXDEV because
aafs_mnt and nsfs_mnt are internal kernel vfsmounts).

-- 
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>

Re: [PATCH v15 4/9] namei: LOOKUP_BENEATH: O_BENEATH-like scoped resolution

Message details and related content
On 2019-11-13, Aleksa Sarai [off-list ref] wrote:
On 2019-11-13, Al Viro [off-list ref] wrote:
quoted
Minor nit here - I'd split "move the conditional call of set_root()
into nd_jump_root()" into a separate patch before that one.  Makes
for fewer distractions in this one.  I'd probably fold "and be
ready for errors other than -ECHILD" into the same preliminary
patch.
Will do.
quoted
quoted
+			/* Not currently safe for scoped-lookups. */
+			if (unlikely(nd->flags & LOOKUP_IS_SCOPED))
+				return ERR_PTR(-EXDEV);
Also a candidate for doing in nd_jump_link()...
quoted
@@ -1373,8 +1403,11 @@ static int follow_dotdot_rcu(struct nameidata *nd)
 	struct inode *inode = nd->inode;
 
 	while (1) {
-		if (path_equal(&nd->path, &nd->root))
+		if (path_equal(&nd->path, &nd->root)) {
+			if (unlikely(nd->flags & LOOKUP_BENEATH))
+				return -EXDEV;
Umm...  Are you sure it's not -ECHILD?
It wouldn't hurt to be -ECHILD -- though it's not clear to me how likely
a success would be in REF-walk if the parent components didn't already
trigger an unlazy_walk() in RCU-walk.

I guess that also means LOOKUP_NO_XDEV should trigger -ECHILD in
follow_dotdot_rcu()?
Scratch the last question -- AFAICS we don't need to do that for
LOOKUP_NO_XDEV because we check against mount_lock so it's very unlikely
that -ECHILD will have any benefit.

-- 
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>

Re: [PATCH v15 3/9] namei: LOOKUP_NO_XDEV: block mountpoint crossing

Message details and related content
On Thu, Nov 14, 2019 at 03:49:45PM +1100, Aleksa Sarai wrote:
On 2019-11-13, Al Viro [off-list ref] wrote:
quoted
On Tue, Nov 05, 2019 at 08:05:47PM +1100, Aleksa Sarai wrote:
quoted
@@ -862,6 +870,8 @@ static int nd_jump_root(struct nameidata *nd)
 void nd_jump_link(struct path *path)
 {
 	struct nameidata *nd = current->nameidata;
+
+	nd->last_magiclink.same_mnt = (nd->path.mnt == path->mnt);
 	path_put(&nd->path);
 
 	nd->path = *path;
@@ -1082,6 +1092,10 @@ const char *get_link(struct nameidata *nd)
 		if (nd->flags & LOOKUP_MAGICLINK_JUMPED) {
 			if (unlikely(nd->flags & LOOKUP_NO_MAGICLINKS))
 				return ERR_PTR(-ELOOP);
+			if (unlikely(nd->flags & LOOKUP_NO_XDEV)) {
+				if (!nd->last_magiclink.same_mnt)
+					return ERR_PTR(-EXDEV);
+			}
 		}
Ugh...  Wouldn't it be better to take that logics (some equivalent thereof)
into nd_jump_link()?  Or just have nd_jump_link() return an error...
This could be done, but the reason for stashing it away in
last_magiclink is because of the future magic-link re-opening patches
which can't be implemented like that without putting the open_flags
inside nameidata (which was decided to be too ugly a while ago).

My point being that I could implement it this way for this series, but
I'd have to implement something like last_magiclink when I end up
re-posting the magic-link stuff in a few weeks.

Looking at all the nd_jump_link() users, the other option is to just
disallow magic-link crossings entirely for LOOKUP_NO_XDEV. The only
thing allowing them permits is to resolve file descriptors that are
pointing to the same procfs mount -- and it's unclear to me how useful
that really is (apparmorfs and nsfs will always give -EXDEV because
aafs_mnt and nsfs_mnt are internal kernel vfsmounts).
I would rather keep the entire if (nd->flags & LOOKUP_MAGICLINK_JUMPED)
out of the get_link().  If you want to generate some error if
nd_jump_link() has been called, just do it right there.  The fewer
pieces of state need to be carried around, the better...

And as for opening them...  Why would you need full open_flags in there?
Details, please...

Re: [PATCH v15 3/9] namei: LOOKUP_NO_XDEV: block mountpoint crossing

Message details and related content
On 2019-11-14, Al Viro [off-list ref] wrote:
On Thu, Nov 14, 2019 at 03:49:45PM +1100, Aleksa Sarai wrote:
quoted
On 2019-11-13, Al Viro [off-list ref] wrote:
quoted
On Tue, Nov 05, 2019 at 08:05:47PM +1100, Aleksa Sarai wrote:
quoted
@@ -862,6 +870,8 @@ static int nd_jump_root(struct nameidata *nd)
 void nd_jump_link(struct path *path)
 {
 	struct nameidata *nd = current->nameidata;
+
+	nd->last_magiclink.same_mnt = (nd->path.mnt == path->mnt);
 	path_put(&nd->path);
 
 	nd->path = *path;
@@ -1082,6 +1092,10 @@ const char *get_link(struct nameidata *nd)
 		if (nd->flags & LOOKUP_MAGICLINK_JUMPED) {
 			if (unlikely(nd->flags & LOOKUP_NO_MAGICLINKS))
 				return ERR_PTR(-ELOOP);
+			if (unlikely(nd->flags & LOOKUP_NO_XDEV)) {
+				if (!nd->last_magiclink.same_mnt)
+					return ERR_PTR(-EXDEV);
+			}
 		}
Ugh...  Wouldn't it be better to take that logics (some equivalent thereof)
into nd_jump_link()?  Or just have nd_jump_link() return an error...
This could be done, but the reason for stashing it away in
last_magiclink is because of the future magic-link re-opening patches
which can't be implemented like that without putting the open_flags
inside nameidata (which was decided to be too ugly a while ago).

My point being that I could implement it this way for this series, but
I'd have to implement something like last_magiclink when I end up
re-posting the magic-link stuff in a few weeks.

Looking at all the nd_jump_link() users, the other option is to just
disallow magic-link crossings entirely for LOOKUP_NO_XDEV. The only
thing allowing them permits is to resolve file descriptors that are
pointing to the same procfs mount -- and it's unclear to me how useful
that really is (apparmorfs and nsfs will always give -EXDEV because

6 more quoted lines

Sure, I can make nd_jump_link() give -ELOOP and drop the current need
for LOOKUP_MAGICLINK_JUMPED -- if necessary we can re-add it for the
magic-link reopening patches.
And as for opening them...  Why would you need full open_flags in there?
Details, please...
I was referring to [1] which has been dropped from this series. I
misspoke -- you don't need the full open_flags, you just need acc_mode
in nameidata -- but from memory you (understandably) weren't in favour
of that either because it further muddled the open semantics with namei.

So the solution I went with was to stash away the i_mode of the
magiclink in nd->last_magiclink.mode (though to avoid a race which Jann
found, you actually need to recalculate it when you call nd_jump_link()
but that's a different topic) and then check it in trailing_magiclink().

However, I've since figured out that we need to restrict things like
bind-mounts and truncate() because they can be used to get around the
restrictions. I dropped that patch from this series so that I could work
on implementing the restrictions for the other relevant VFS syscalls
separately from openat2 (upgrade_mask will be re-added to open_how with
those patches).

My point was that AFAICS we will either have to have nd->acc_mode (or
something similar) or have nd->last_magiclink in order to implement the
magic-link reopening hardening.

[1]: https://lore.kernel.org/lkml/20190930183316.10190-2-cyphar@cyphar.com/

-- 
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>
Keyboard shortcuts
hback out one level
jnext message in thread / next series-diff position
kprevious message in thread / previous series-diff position
ldrill in
Escclose help / fold thread tree
?toggle this help