Merge patch series "pidfs: support bind-mounts"

Christian Brauner <brauner@kernel.org> says:

Allow bind-mounting pidfds. Similar to nsfs let's allow bind-mounts for
pidfds. This allows pidfds to be safely recovered and checked for
process recycling.

Instead of checking d_ops for both nsfs and pidfs we could in a
follow-up patch add a flag argument to struct dentry_operations that
functions similar to file_operations->fop_flags.

* patches from https://lore.kernel.org/r/20241219-work-pidfs-mount-v1-0-dbc56198b839@kernel.org:
  selftests: add pidfd bind-mount tests
  pidfs: allow bind-mounts

Link: https://lore.kernel.org/r/20241219-work-pidfs-mount-v1-0-dbc56198b839@kernel.org
Signed-off-by: Christian Brauner <brauner@kernel.org>

fs/namespace.c

@@ -32,6 +32,7 @@
 #include <linux/fs_context.h>
 #include <linux/shmem_fs.h>
 #include <linux/mnt_idmapping.h>
+#include <linux/pidfs.h>
 #include <linux/nospec.h>
 
 #include "pnode.h"
@@ -2732,8 +2733,13 @@ static struct mount *__do_loopback(struct path *old_path, int recurse)
 	if (IS_MNT_UNBINDABLE(old))
 		return mnt;
 
-	if (!check_mnt(old) && old_path->dentry->d_op != &ns_dentry_operations)
-		return mnt;
+	if (!check_mnt(old)) {
+		const struct dentry_operations *d_op = old_path->dentry->d_op;
+
+		if (d_op != &ns_dentry_operations &&
+		    d_op != &pidfs_dentry_operations)
+			return mnt;
+	}
 
 	if (!recurse && has_locked_children(old, old_path->dentry))
 		return mnt;

fs/pidfs.c

@@ -510,7 +510,7 @@ static char *pidfs_dname(struct dentry *dentry, char *buffer, int buflen)
 	return dynamic_dname(buffer, buflen, "anon_inode:[pidfd]");
 }
 
-static const struct dentry_operations pidfs_dentry_operations = {
+const struct dentry_operations pidfs_dentry_operations = {
 	.d_delete	= always_delete_dentry,
 	.d_dname	= pidfs_dname,
 	.d_prune	= stashed_dentry_prune,

include/linux/pidfs.h

@@ -6,5 +6,6 @@ struct file *pidfs_alloc_file(struct pid *pid, unsigned int flags);
 void __init pidfs_init(void);
 void pidfs_add_pid(struct pid *pid);
 void pidfs_remove_pid(struct pid *pid);
+extern const struct dentry_operations pidfs_dentry_operations;
 
 #endif /* _LINUX_PID_FS_H */

tools/testing/selftests/pidfd/.gitignore

@@ -7,3 +7,4 @@ pidfd_fdinfo_test
 pidfd_getfd_test
 pidfd_setns_test
 pidfd_file_handle_test
+pidfd_bind_mount

tools/testing/selftests/pidfd/Makefile

@@ -3,7 +3,7 @@ CFLAGS += -g $(KHDR_INCLUDES) -pthread -Wall
 
 TEST_GEN_PROGS := pidfd_test pidfd_fdinfo_test pidfd_open_test \
 	pidfd_poll_test pidfd_wait pidfd_getfd_test pidfd_setns_test \
-	pidfd_file_handle_test
+	pidfd_file_handle_test pidfd_bind_mount
 
 include ../lib.mk
 

tools/testing/selftests/pidfd/pidfd_bind_mount.c

@@ -0,0 +1,188 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+// Copyright (c) 2024 Christian Brauner <brauner@kernel.org>
+
+#define _GNU_SOURCE
+#include <fcntl.h>
+#include <limits.h>
+#include <sched.h>
+#include <stdio.h>
+#include <string.h>
+#include <linux/fs.h>
+#include <sys/ioctl.h>
+#include <sys/stat.h>
+#include <sys/mount.h>
+#include <unistd.h>
+
+#include "pidfd.h"
+#include "../kselftest_harness.h"
+
+#ifndef __NR_open_tree
+	#if defined __alpha__
+		#define __NR_open_tree 538
+	#elif defined _MIPS_SIM
+		#if _MIPS_SIM == _MIPS_SIM_ABI32	/* o32 */
+			#define __NR_open_tree 4428
+		#endif
+		#if _MIPS_SIM == _MIPS_SIM_NABI32	/* n32 */
+			#define __NR_open_tree 6428
+		#endif
+		#if _MIPS_SIM == _MIPS_SIM_ABI64	/* n64 */
+			#define __NR_open_tree 5428
+		#endif
+	#elif defined __ia64__
+		#define __NR_open_tree (428 + 1024)
+	#else
+		#define __NR_open_tree 428
+	#endif
+#endif
+
+#ifndef __NR_move_mount
+	#if defined __alpha__
+		#define __NR_move_mount 539
+	#elif defined _MIPS_SIM
+		#if _MIPS_SIM == _MIPS_SIM_ABI32	/* o32 */
+			#define __NR_move_mount 4429
+		#endif
+		#if _MIPS_SIM == _MIPS_SIM_NABI32	/* n32 */
+			#define __NR_move_mount 6429
+		#endif
+		#if _MIPS_SIM == _MIPS_SIM_ABI64	/* n64 */
+			#define __NR_move_mount 5429
+		#endif
+	#elif defined __ia64__
+		#define __NR_move_mount (428 + 1024)
+	#else
+		#define __NR_move_mount 429
+	#endif
+#endif
+
+#ifndef MOVE_MOUNT_F_EMPTY_PATH
+#define MOVE_MOUNT_F_EMPTY_PATH 0x00000004 /* Empty from path permitted */
+#endif
+
+#ifndef MOVE_MOUNT_F_EMPTY_PATH
+#define MOVE_MOUNT_T_EMPTY_PATH 0x00000040 /* Empty to path permitted */
+#endif
+
+static inline int sys_move_mount(int from_dfd, const char *from_pathname,
+                                 int to_dfd, const char *to_pathname,
+                                 unsigned int flags)
+{
+        return syscall(__NR_move_mount, from_dfd, from_pathname, to_dfd,
+                       to_pathname, flags);
+}
+
+#ifndef OPEN_TREE_CLONE
+#define OPEN_TREE_CLONE 1
+#endif
+
+#ifndef OPEN_TREE_CLOEXEC
+#define OPEN_TREE_CLOEXEC O_CLOEXEC
+#endif
+
+#ifndef AT_RECURSIVE
+#define AT_RECURSIVE 0x8000 /* Apply to the entire subtree */
+#endif
+
+static inline int sys_open_tree(int dfd, const char *filename, unsigned int flags)
+{
+	return syscall(__NR_open_tree, dfd, filename, flags);
+}
+
+FIXTURE(pidfd_bind_mount) {
+	char template[PATH_MAX];
+	int fd_tmp;
+	int pidfd;
+	struct stat st1;
+	struct stat st2;
+	__u32 gen1;
+	__u32 gen2;
+	bool must_unmount;
+};
+
+FIXTURE_SETUP(pidfd_bind_mount)
+{
+	self->fd_tmp = -EBADF;
+	self->must_unmount = false;
+	ASSERT_EQ(unshare(CLONE_NEWNS), 0);
+	ASSERT_LE(snprintf(self->template, PATH_MAX, "%s", P_tmpdir "/pidfd_bind_mount_XXXXXX"), PATH_MAX);
+	self->fd_tmp = mkstemp(self->template);
+	ASSERT_GE(self->fd_tmp, 0);
+	self->pidfd = sys_pidfd_open(getpid(), 0);
+	ASSERT_GE(self->pidfd, 0);
+	ASSERT_GE(fstat(self->pidfd, &self->st1), 0);
+	ASSERT_EQ(ioctl(self->pidfd, FS_IOC_GETVERSION, &self->gen1), 0);
+}
+
+FIXTURE_TEARDOWN(pidfd_bind_mount)
+{
+	ASSERT_EQ(close(self->fd_tmp), 0);
+	if (self->must_unmount)
+		ASSERT_EQ(umount2(self->template, 0), 0);
+	ASSERT_EQ(unlink(self->template), 0);
+}
+
+/*
+ * Test that a detached mount can be created for a pidfd and then
+ * attached to the filesystem hierarchy.
+ */
+TEST_F(pidfd_bind_mount, bind_mount)
+{
+	int fd_tree;
+
+	fd_tree = sys_open_tree(self->pidfd, "", OPEN_TREE_CLONE | OPEN_TREE_CLOEXEC | AT_EMPTY_PATH);
+	ASSERT_GE(fd_tree, 0);
+
+	ASSERT_EQ(move_mount(fd_tree, "", self->fd_tmp, "", MOVE_MOUNT_F_EMPTY_PATH | MOVE_MOUNT_T_EMPTY_PATH), 0);
+	self->must_unmount = true;
+
+	ASSERT_EQ(close(fd_tree), 0);
+}
+
+/* Test that a pidfd can be reopened through procfs. */
+TEST_F(pidfd_bind_mount, reopen)
+{
+	int pidfd;
+	char proc_path[PATH_MAX];
+
+	sprintf(proc_path, "/proc/self/fd/%d", self->pidfd);
+	pidfd = open(proc_path, O_RDONLY | O_NOCTTY | O_CLOEXEC);
+	ASSERT_GE(pidfd, 0);
+
+	ASSERT_GE(fstat(self->pidfd, &self->st2), 0);
+	ASSERT_EQ(ioctl(self->pidfd, FS_IOC_GETVERSION, &self->gen2), 0);
+
+	ASSERT_TRUE(self->st1.st_dev == self->st2.st_dev && self->st1.st_ino == self->st2.st_ino);
+	ASSERT_TRUE(self->gen1 == self->gen2);
+
+	ASSERT_EQ(close(pidfd), 0);
+}
+
+/*
+ * Test that a detached mount can be created for a pidfd and then
+ * attached to the filesystem hierarchy and reopened.
+ */
+TEST_F(pidfd_bind_mount, bind_mount_reopen)
+{
+	int fd_tree, fd_pidfd_mnt;
+
+	fd_tree = sys_open_tree(self->pidfd, "", OPEN_TREE_CLONE | OPEN_TREE_CLOEXEC | AT_EMPTY_PATH);
+	ASSERT_GE(fd_tree, 0);
+
+	ASSERT_EQ(move_mount(fd_tree, "", self->fd_tmp, "", MOVE_MOUNT_F_EMPTY_PATH | MOVE_MOUNT_T_EMPTY_PATH), 0);
+	self->must_unmount = true;
+
+	fd_pidfd_mnt = openat(-EBADF, self->template, O_RDONLY | O_NOCTTY | O_CLOEXEC);
+	ASSERT_GE(fd_pidfd_mnt, 0);
+
+	ASSERT_GE(fstat(fd_tree, &self->st2), 0);
+	ASSERT_EQ(ioctl(fd_pidfd_mnt, FS_IOC_GETVERSION, &self->gen2), 0);
+
+	ASSERT_TRUE(self->st1.st_dev == self->st2.st_dev && self->st1.st_ino == self->st2.st_ino);
+	ASSERT_TRUE(self->gen1 == self->gen2);
+
+	ASSERT_EQ(close(fd_tree), 0);
+	ASSERT_EQ(close(fd_pidfd_mnt), 0);
+}
+
+TEST_HARNESS_MAIN