Merge patch series "fs: lockless mntns lookup"

Christian Brauner <brauner@kernel.org> says:

Currently we take the read lock when looking for a mount namespace to
list mounts in. We can make this lockless. The simple search case can
just use a sequence counter to detect concurrent changes to the rbtree.

For walking the list of mount namespaces sequentially via nsfs we keep a
separate rcu list as rb_prev() and rb_next() aren't usable safely with
rcu.

Since creating mount namespaces is a relatively rare event compared with
querying mounts in a foreign mount namespace this is worth it. Once
libmount and systemd pick up this mechanism to list mounts in foreign
mount namespaces this will be used very frequently.

* patches from https://lore.kernel.org/r/20241213-work-mount-rbtree-lockless-v3-0-6e3cdaf9b280@kernel.org:
  samples: add test-list-all-mounts
  selftests: remove unneeded include
  selftests: add tests for mntns iteration
  seltests: move nsfs into filesystems subfolder
  fs: simplify rwlock to spinlock
  fs: lockless mntns lookup for nsfs
  rculist: add list_bidir_{del,prev}_rcu()
  fs: lockless mntns rbtree lookup
  fs: add mount namespace to rbtree late
  mount: remove inlude/nospec.h include

Link: https://lore.kernel.org/r/20241213-work-mount-rbtree-lockless-v3-0-6e3cdaf9b280@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>

7 files changed, 152 insertions, 3 deletions

diff --git a/tools/testing/selftests/nsfs/.gitignore b/tools/testing/selftests/filesystems/nsfs/.gitignore
index ed79ebdf286e..92a8249006d1 100644
--- a/tools/testing/selftests/nsfs/.gitignore
+++ b/tools/testing/selftests/filesystems/nsfs/.gitignore
@@ -1,3 +1,4 @@
 # SPDX-License-Identifier: GPL-2.0-only
 owner
 pidns
+iterate_mntns
diff --git a/tools/testing/selftests/nsfs/Makefile b/tools/testing/selftests/filesystems/nsfs/Makefile
index dd9bd50b7b93..231aaa7dfd95 100644
--- a/tools/testing/selftests/nsfs/Makefile
+++ b/tools/testing/selftests/filesystems/nsfs/Makefile
@@ -1,6 +1,6 @@
 # SPDX-License-Identifier: GPL-2.0-only
-TEST_GEN_PROGS := owner pidns
+TEST_GEN_PROGS := owner pidns iterate_mntns
 
 CFLAGS := -Wall -Werror
 
-include ../lib.mk
+include ../../lib.mk
diff --git a/tools/testing/selftests/nsfs/config b/tools/testing/selftests/filesystems/nsfs/config
index 598d0a225fc9..598d0a225fc9 100644
--- a/tools/testing/selftests/nsfs/config
+++ b/tools/testing/selftests/filesystems/nsfs/config
diff --git a/tools/testing/selftests/filesystems/nsfs/iterate_mntns.c b/tools/testing/selftests/filesystems/nsfs/iterate_mntns.c
new file mode 100644
index 000000000000..457cf76f3c5f
--- /dev/null
+++ b/tools/testing/selftests/filesystems/nsfs/iterate_mntns.c
@@ -0,0 +1,149 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+// Copyright (c) 2024 Christian Brauner <brauner@kernel.org>
+
+#define _GNU_SOURCE
+#include <fcntl.h>
+#include <sched.h>
+#include <stdio.h>
+#include <string.h>
+#include <sys/stat.h>
+#include <sys/mount.h>
+#include <unistd.h>
+
+#include "../../kselftest_harness.h"
+
+#define MNT_NS_COUNT 11
+#define MNT_NS_LAST_INDEX 10
+
+struct mnt_ns_info {
+	__u32 size;
+	__u32 nr_mounts;
+	__u64 mnt_ns_id;
+};
+
+#define MNT_NS_INFO_SIZE_VER0 16 /* size of first published struct */
+
+/* Get information about namespace. */
+#define NS_MNT_GET_INFO _IOR(0xb7, 10, struct mnt_ns_info)
+/* Get next namespace. */
+#define NS_MNT_GET_NEXT _IOR(0xb7, 11, struct mnt_ns_info)
+/* Get previous namespace. */
+#define NS_MNT_GET_PREV _IOR(0xb7, 12, struct mnt_ns_info)
+
+FIXTURE(iterate_mount_namespaces) {
+	int fd_mnt_ns[MNT_NS_COUNT];
+	__u64 mnt_ns_id[MNT_NS_COUNT];
+};
+
+FIXTURE_SETUP(iterate_mount_namespaces)
+{
+	for (int i = 0; i < MNT_NS_COUNT; i++)
+		self->fd_mnt_ns[i] = -EBADF;
+
+	/*
+	 * Creating a new user namespace let's us guarantee that we only see
+	 * mount namespaces that we did actually create.
+	 */
+	ASSERT_EQ(unshare(CLONE_NEWUSER), 0);
+
+	for (int i = 0; i < MNT_NS_COUNT; i++) {
+		struct mnt_ns_info info = {};
+
+		ASSERT_EQ(unshare(CLONE_NEWNS), 0);
+		self->fd_mnt_ns[i] = open("/proc/self/ns/mnt", O_RDONLY | O_CLOEXEC);
+		ASSERT_GE(self->fd_mnt_ns[i], 0);
+		ASSERT_EQ(ioctl(self->fd_mnt_ns[i], NS_MNT_GET_INFO, &info), 0);
+		self->mnt_ns_id[i] = info.mnt_ns_id;
+	}
+}
+
+FIXTURE_TEARDOWN(iterate_mount_namespaces)
+{
+	for (int i = 0; i < MNT_NS_COUNT; i++) {
+		if (self->fd_mnt_ns[i] < 0)
+			continue;
+		ASSERT_EQ(close(self->fd_mnt_ns[i]), 0);
+	}
+}
+
+TEST_F(iterate_mount_namespaces, iterate_all_forward)
+{
+	int fd_mnt_ns_cur, count = 0;
+
+	fd_mnt_ns_cur = fcntl(self->fd_mnt_ns[0], F_DUPFD_CLOEXEC);
+	ASSERT_GE(fd_mnt_ns_cur, 0);
+
+	for (;; count++) {
+		struct mnt_ns_info info = {};
+		int fd_mnt_ns_next;
+
+		fd_mnt_ns_next = ioctl(fd_mnt_ns_cur, NS_MNT_GET_NEXT, &info);
+		if (fd_mnt_ns_next < 0 && errno == ENOENT)
+			break;
+		ASSERT_GE(fd_mnt_ns_next, 0);
+		ASSERT_EQ(close(fd_mnt_ns_cur), 0);
+		fd_mnt_ns_cur = fd_mnt_ns_next;
+	}
+	ASSERT_EQ(count, MNT_NS_LAST_INDEX);
+}
+
+TEST_F(iterate_mount_namespaces, iterate_all_backwards)
+{
+	int fd_mnt_ns_cur, count = 0;
+
+	fd_mnt_ns_cur = fcntl(self->fd_mnt_ns[MNT_NS_LAST_INDEX], F_DUPFD_CLOEXEC);
+	ASSERT_GE(fd_mnt_ns_cur, 0);
+
+	for (;; count++) {
+		struct mnt_ns_info info = {};
+		int fd_mnt_ns_prev;
+
+		fd_mnt_ns_prev = ioctl(fd_mnt_ns_cur, NS_MNT_GET_PREV, &info);
+		if (fd_mnt_ns_prev < 0 && errno == ENOENT)
+			break;
+		ASSERT_GE(fd_mnt_ns_prev, 0);
+		ASSERT_EQ(close(fd_mnt_ns_cur), 0);
+		fd_mnt_ns_cur = fd_mnt_ns_prev;
+	}
+	ASSERT_EQ(count, MNT_NS_LAST_INDEX);
+}
+
+TEST_F(iterate_mount_namespaces, iterate_forward)
+{
+	int fd_mnt_ns_cur;
+
+	ASSERT_EQ(setns(self->fd_mnt_ns[0], CLONE_NEWNS), 0);
+
+	fd_mnt_ns_cur = self->fd_mnt_ns[0];
+	for (int i = 1; i < MNT_NS_COUNT; i++) {
+		struct mnt_ns_info info = {};
+		int fd_mnt_ns_next;
+
+		fd_mnt_ns_next = ioctl(fd_mnt_ns_cur, NS_MNT_GET_NEXT, &info);
+		ASSERT_GE(fd_mnt_ns_next, 0);
+		ASSERT_EQ(close(fd_mnt_ns_cur), 0);
+		fd_mnt_ns_cur = fd_mnt_ns_next;
+		ASSERT_EQ(info.mnt_ns_id, self->mnt_ns_id[i]);
+	}
+}
+
+TEST_F(iterate_mount_namespaces, iterate_backward)
+{
+	int fd_mnt_ns_cur;
+
+	ASSERT_EQ(setns(self->fd_mnt_ns[MNT_NS_LAST_INDEX], CLONE_NEWNS), 0);
+
+	fd_mnt_ns_cur = self->fd_mnt_ns[MNT_NS_LAST_INDEX];
+	for (int i = MNT_NS_LAST_INDEX - 1; i >= 0; i--) {
+		struct mnt_ns_info info = {};
+		int fd_mnt_ns_prev;
+
+		fd_mnt_ns_prev = ioctl(fd_mnt_ns_cur, NS_MNT_GET_PREV, &info);
+		ASSERT_GE(fd_mnt_ns_prev, 0);
+		ASSERT_EQ(close(fd_mnt_ns_cur), 0);
+		fd_mnt_ns_cur = fd_mnt_ns_prev;
+		ASSERT_EQ(info.mnt_ns_id, self->mnt_ns_id[i]);
+	}
+}
+
+TEST_HARNESS_MAIN
diff --git a/tools/testing/selftests/nsfs/owner.c b/tools/testing/selftests/filesystems/nsfs/owner.c
index 96a976c74550..96a976c74550 100644
--- a/tools/testing/selftests/nsfs/owner.c
+++ b/tools/testing/selftests/filesystems/nsfs/owner.c
diff --git a/tools/testing/selftests/nsfs/pidns.c b/tools/testing/selftests/filesystems/nsfs/pidns.c
index e3c772c6a7c7..e3c772c6a7c7 100644
--- a/tools/testing/selftests/nsfs/pidns.c
+++ b/tools/testing/selftests/filesystems/nsfs/pidns.c
diff --git a/tools/testing/selftests/pidfd/pidfd.h b/tools/testing/selftests/pidfd/pidfd.h
index 88d6830ee004..3a96053e52e7 100644
--- a/tools/testing/selftests/pidfd/pidfd.h
+++ b/tools/testing/selftests/pidfd/pidfd.h
@@ -12,7 +12,6 @@
 #include <stdlib.h>
 #include <string.h>
 #include <syscall.h>
-#include <sys/mount.h>
 #include <sys/types.h>
 #include <sys/wait.h>