[thirdparty/systemd.git] / src / basic / mount-util.c

/***
  This file is part of systemd.

  Copyright 2010 Lennart Poettering

  systemd is free software; you can redistribute it and/or modify it
  under the terms of the GNU Lesser General Public License as published by
  the Free Software Foundation; either version 2.1 of the License, or
  (at your option) any later version.

  systemd is distributed in the hope that it will be useful, but
  WITHOUT ANY WARRANTY; without even the implied warranty of
  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  Lesser General Public License for more details.

  You should have received a copy of the GNU Lesser General Public License
  along with systemd; If not, see <http://www.gnu.org/licenses/>.
***/

#include <errno.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mount.h>
#include <sys/stat.h>
#include <sys/statvfs.h>
#include <unistd.h>

#include "alloc-util.h"
#include "escape.h"
#include "fd-util.h"
#include "fileio.h"
#include "fs-util.h"
#include "hashmap.h"
#include "mount-util.h"
#include "parse-util.h"
#include "path-util.h"
#include "set.h"
#include "stdio-util.h"
#include "string-util.h"
#include "strv.h"

static int fd_fdinfo_mnt_id(int fd, const char *filename, int flags, int *mnt_id) {
        char path[strlen("/proc/self/fdinfo/") + DECIMAL_STR_MAX(int)];
        _cleanup_free_ char *fdinfo = NULL;
        _cleanup_close_ int subfd = -1;
        char *p;
        int r;

        if ((flags & AT_EMPTY_PATH) && isempty(filename))
                xsprintf(path, "/proc/self/fdinfo/%i", fd);
        else {
                subfd = openat(fd, filename, O_CLOEXEC|O_PATH);
                if (subfd < 0)
                        return -errno;

                xsprintf(path, "/proc/self/fdinfo/%i", subfd);
        }

        r = read_full_file(path, &fdinfo, NULL);
        if (r == -ENOENT) /* The fdinfo directory is a relatively new addition */
                return -EOPNOTSUPP;
        if (r < 0)
                return -errno;

        p = startswith(fdinfo, "mnt_id:");
        if (!p) {
                p = strstr(fdinfo, "\nmnt_id:");
                if (!p) /* The mnt_id field is a relatively new addition */
                        return -EOPNOTSUPP;

                p += 8;
        }

        p += strspn(p, WHITESPACE);
        p[strcspn(p, WHITESPACE)] = 0;

        return safe_atoi(p, mnt_id);
}

int fd_is_mount_point(int fd, const char *filename, int flags) {
        union file_handle_union h = FILE_HANDLE_INIT, h_parent = FILE_HANDLE_INIT;
        int mount_id = -1, mount_id_parent = -1;
        bool nosupp = false, check_st_dev = true;
        struct stat a, b;
        int r;

        assert(fd >= 0);
        assert(filename);

        /* First we will try the name_to_handle_at() syscall, which
         * tells us the mount id and an opaque file "handle". It is
         * not supported everywhere though (kernel compile-time
         * option, not all file systems are hooked up). If it works
         * the mount id is usually good enough to tell us whether
         * something is a mount point.
         *
         * If that didn't work we will try to read the mount id from
         * /proc/self/fdinfo/<fd>. This is almost as good as
         * name_to_handle_at(), however, does not return the
         * opaque file handle. The opaque file handle is pretty useful
         * to detect the root directory, which we should always
         * consider a mount point. Hence we use this only as
         * fallback. Exporting the mnt_id in fdinfo is a pretty recent
         * kernel addition.
         *
         * As last fallback we do traditional fstat() based st_dev
         * comparisons. This is how things were traditionally done,
         * but unionfs breaks this since it exposes file
         * systems with a variety of st_dev reported. Also, btrfs
         * subvolumes have different st_dev, even though they aren't
         * real mounts of their own. */

        r = name_to_handle_at(fd, filename, &h.handle, &mount_id, flags);
        if (r < 0) {
                if (errno == ENOSYS)
                        /* This kernel does not support name_to_handle_at()
                         * fall back to simpler logic. */
                        goto fallback_fdinfo;
                else if (errno == EOPNOTSUPP)
                        /* This kernel or file system does not support
                         * name_to_handle_at(), hence let's see if the
                         * upper fs supports it (in which case it is a
                         * mount point), otherwise fallback to the
                         * traditional stat() logic */
                        nosupp = true;
                else
                        return -errno;
        }

        r = name_to_handle_at(fd, "", &h_parent.handle, &mount_id_parent, AT_EMPTY_PATH);
        if (r < 0) {
                if (errno == EOPNOTSUPP) {
                        if (nosupp)
                                /* Neither parent nor child do name_to_handle_at()?
                                   We have no choice but to fall back. */
                                goto fallback_fdinfo;
                        else
                                /* The parent can't do name_to_handle_at() but the
                                 * directory we are interested in can?
                                 * If so, it must be a mount point. */
                                return 1;
                } else
                        return -errno;
        }

        /* The parent can do name_to_handle_at() but the
         * directory we are interested in can't? If so, it
         * must be a mount point. */
        if (nosupp)
                return 1;

        /* If the file handle for the directory we are
         * interested in and its parent are identical, we
         * assume this is the root directory, which is a mount
         * point. */

        if (h.handle.handle_bytes == h_parent.handle.handle_bytes &&
            h.handle.handle_type == h_parent.handle.handle_type &&
            memcmp(h.handle.f_handle, h_parent.handle.f_handle, h.handle.handle_bytes) == 0)
                return 1;

        return mount_id != mount_id_parent;

fallback_fdinfo:
        r = fd_fdinfo_mnt_id(fd, filename, flags, &mount_id);
        if (IN_SET(r, -EOPNOTSUPP, -EACCES))
                goto fallback_fstat;
        if (r < 0)
                return r;

        r = fd_fdinfo_mnt_id(fd, "", AT_EMPTY_PATH, &mount_id_parent);
        if (r < 0)
                return r;

        if (mount_id != mount_id_parent)
                return 1;

        /* Hmm, so, the mount ids are the same. This leaves one
         * special case though for the root file system. For that,
         * let's see if the parent directory has the same inode as we
         * are interested in. Hence, let's also do fstat() checks now,
         * too, but avoid the st_dev comparisons, since they aren't
         * that useful on unionfs mounts. */
        check_st_dev = false;

fallback_fstat:
        /* yay for fstatat() taking a different set of flags than the other
         * _at() above */
        if (flags & AT_SYMLINK_FOLLOW)
                flags &= ~AT_SYMLINK_FOLLOW;
        else
                flags |= AT_SYMLINK_NOFOLLOW;
        if (fstatat(fd, filename, &a, flags) < 0)
                return -errno;

        if (fstatat(fd, "", &b, AT_EMPTY_PATH) < 0)
                return -errno;

        /* A directory with same device and inode as its parent? Must
         * be the root directory */
        if (a.st_dev == b.st_dev &&
            a.st_ino == b.st_ino)
                return 1;

        return check_st_dev && (a.st_dev != b.st_dev);
}

/* flags can be AT_SYMLINK_FOLLOW or 0 */
int path_is_mount_point(const char *t, const char *root, int flags) {
        _cleanup_free_ char *canonical = NULL, *parent = NULL;
        _cleanup_close_ int fd = -1;
        int r;

        assert(t);

        if (path_equal(t, "/"))
                return 1;

        /* we need to resolve symlinks manually, we can't just rely on
         * fd_is_mount_point() to do that for us; if we have a structure like
         * /bin -> /usr/bin/ and /usr is a mount point, then the parent that we
         * look at needs to be /usr, not /. */
        if (flags & AT_SYMLINK_FOLLOW) {
                r = chase_symlinks(t, root, 0, &canonical);
                if (r < 0)
                        return r;

                t = canonical;
        }

        parent = dirname_malloc(t);
        if (!parent)
                return -ENOMEM;

        fd = openat(AT_FDCWD, parent, O_DIRECTORY|O_CLOEXEC|O_PATH);
        if (fd < 0)
                return -errno;

        return fd_is_mount_point(fd, basename(t), flags);
}

int umount_recursive(const char *prefix, int flags) {
        bool again;
        int n = 0, r;

        /* Try to umount everything recursively below a
         * directory. Also, take care of stacked mounts, and keep
         * unmounting them until they are gone. */

        do {
                _cleanup_fclose_ FILE *proc_self_mountinfo = NULL;

                again = false;
                r = 0;

                proc_self_mountinfo = fopen("/proc/self/mountinfo", "re");
                if (!proc_self_mountinfo)
                        return -errno;

                for (;;) {
                        _cleanup_free_ char *path = NULL, *p = NULL;
                        int k;

                        k = fscanf(proc_self_mountinfo,
                                   "%*s "       /* (1) mount id */
                                   "%*s "       /* (2) parent id */
                                   "%*s "       /* (3) major:minor */
                                   "%*s "       /* (4) root */
                                   "%ms "       /* (5) mount point */
                                   "%*s"        /* (6) mount options */
                                   "%*[^-]"     /* (7) optional fields */
                                   "- "         /* (8) separator */
                                   "%*s "       /* (9) file system type */
                                   "%*s"        /* (10) mount source */
                                   "%*s"        /* (11) mount options 2 */
                                   "%*[^\n]",   /* some rubbish at the end */
                                   &path);
                        if (k != 1) {
                                if (k == EOF)
                                        break;

                                continue;
                        }

                        r = cunescape(path, UNESCAPE_RELAX, &p);
                        if (r < 0)
                                return r;

                        if (!path_startswith(p, prefix))
                                continue;

                        if (umount2(p, flags) < 0) {
                                r = log_debug_errno(errno, "Failed to umount %s: %m", p);
                                continue;
                        }

                        log_debug("Successfully unmounted %s", p);

                        again = true;
                        n++;

                        break;
                }

        } while (again);

        return r ? r : n;
}

static int get_mount_flags(const char *path, unsigned long *flags) {
        struct statvfs buf;

        if (statvfs(path, &buf) < 0)
                return -errno;
        *flags = buf.f_flag;
        return 0;
}

int bind_remount_recursive(const char *prefix, bool ro, char **blacklist) {
        _cleanup_set_free_free_ Set *done = NULL;
        _cleanup_free_ char *cleaned = NULL;
        int r;

        /* Recursively remount a directory (and all its submounts) read-only or read-write. If the directory is already
         * mounted, we reuse the mount and simply mark it MS_BIND|MS_RDONLY (or remove the MS_RDONLY for read-write
         * operation). If it isn't we first make it one. Afterwards we apply MS_BIND|MS_RDONLY (or remove MS_RDONLY) to
         * all submounts we can access, too. When mounts are stacked on the same mount point we only care for each
         * individual "top-level" mount on each point, as we cannot influence/access the underlying mounts anyway. We
         * do not have any effect on future submounts that might get propagated, they migt be writable. This includes
         * future submounts that have been triggered via autofs.
         *
         * If the "blacklist" parameter is specified it may contain a list of subtrees to exclude from the
         * remount operation. Note that we'll ignore the blacklist for the top-level path. */

        cleaned = strdup(prefix);
        if (!cleaned)
                return -ENOMEM;

        path_kill_slashes(cleaned);

        done = set_new(&string_hash_ops);
        if (!done)
                return -ENOMEM;

        for (;;) {
                _cleanup_fclose_ FILE *proc_self_mountinfo = NULL;
                _cleanup_set_free_free_ Set *todo = NULL;
                bool top_autofs = false;
                char *x;
                unsigned long orig_flags;

                todo = set_new(&string_hash_ops);
                if (!todo)
                        return -ENOMEM;

                proc_self_mountinfo = fopen("/proc/self/mountinfo", "re");
                if (!proc_self_mountinfo)
                        return -errno;

                for (;;) {
                        _cleanup_free_ char *path = NULL, *p = NULL, *type = NULL;
                        int k;

                        k = fscanf(proc_self_mountinfo,
                                   "%*s "       /* (1) mount id */
                                   "%*s "       /* (2) parent id */
                                   "%*s "       /* (3) major:minor */
                                   "%*s "       /* (4) root */
                                   "%ms "       /* (5) mount point */
                                   "%*s"        /* (6) mount options (superblock) */
                                   "%*[^-]"     /* (7) optional fields */
                                   "- "         /* (8) separator */
                                   "%ms "       /* (9) file system type */
                                   "%*s"        /* (10) mount source */
                                   "%*s"        /* (11) mount options (bind mount) */
                                   "%*[^\n]",   /* some rubbish at the end */
                                   &path,
                                   &type);
                        if (k != 2) {
                                if (k == EOF)
                                        break;

                                continue;
                        }

                        r = cunescape(path, UNESCAPE_RELAX, &p);
                        if (r < 0)
                                return r;

                        if (!path_startswith(p, cleaned))
                                continue;

                        /* Ignore this mount if it is blacklisted, but only if it isn't the top-level mount we shall
                         * operate on. */
                        if (!path_equal(cleaned, p)) {
                                bool blacklisted = false;
                                char **i;

                                STRV_FOREACH(i, blacklist) {

                                        if (path_equal(*i, cleaned))
                                                continue;

                                        if (!path_startswith(*i, cleaned))
                                                continue;

                                        if (path_startswith(p, *i)) {
                                                blacklisted = true;
                                                log_debug("Not remounting %s, because blacklisted by %s, called for %s", p, *i, cleaned);
                                                break;
                                        }
                                }
                                if (blacklisted)
                                        continue;
                        }

                        /* Let's ignore autofs mounts.  If they aren't
                         * triggered yet, we want to avoid triggering
                         * them, as we don't make any guarantees for
                         * future submounts anyway.  If they are
                         * already triggered, then we will find
                         * another entry for this. */
                        if (streq(type, "autofs")) {
                                top_autofs = top_autofs || path_equal(cleaned, p);
                                continue;
                        }

                        if (!set_contains(done, p)) {
                                r = set_consume(todo, p);
                                p = NULL;
                                if (r == -EEXIST)
                                        continue;
                                if (r < 0)
                                        return r;
                        }
                }

                /* If we have no submounts to process anymore and if
                 * the root is either already done, or an autofs, we
                 * are done */
                if (set_isempty(todo) &&
                    (top_autofs || set_contains(done, cleaned)))
                        return 0;

                if (!set_contains(done, cleaned) &&
                    !set_contains(todo, cleaned)) {
                        /* The prefix directory itself is not yet a mount, make it one. */
                        if (mount(cleaned, cleaned, NULL, MS_BIND|MS_REC, NULL) < 0)
                                return -errno;

                        orig_flags = 0;
                        (void) get_mount_flags(cleaned, &orig_flags);
                        orig_flags &= ~MS_RDONLY;

                        if (mount(NULL, prefix, NULL, orig_flags|MS_BIND|MS_REMOUNT|(ro ? MS_RDONLY : 0), NULL) < 0)
                                return -errno;

                        log_debug("Made top-level directory %s a mount point.", prefix);

                        x = strdup(cleaned);
                        if (!x)
                                return -ENOMEM;

                        r = set_consume(done, x);
                        if (r < 0)
                                return r;
                }

                while ((x = set_steal_first(todo))) {

                        r = set_consume(done, x);
                        if (r == -EEXIST || r == 0)
                                continue;
                        if (r < 0)
                                return r;

                        /* Deal with mount points that are obstructed by a later mount */
                        r = path_is_mount_point(x, NULL, 0);
                        if (r == -ENOENT || r == 0)
                                continue;
                        if (r < 0)
                                return r;

                        /* Try to reuse the original flag set */
                        orig_flags = 0;
                        (void) get_mount_flags(x, &orig_flags);
                        orig_flags &= ~MS_RDONLY;

                        if (mount(NULL, x, NULL, orig_flags|MS_BIND|MS_REMOUNT|(ro ? MS_RDONLY : 0), NULL) < 0)
                                return -errno;

                        log_debug("Remounted %s read-only.", x);
                }
        }
}

int mount_move_root(const char *path) {
        assert(path);

        if (chdir(path) < 0)
                return -errno;

        if (mount(path, "/", NULL, MS_MOVE, NULL) < 0)
                return -errno;

        if (chroot(".") < 0)
                return -errno;

        if (chdir("/") < 0)
                return -errno;

        return 0;
}

bool fstype_is_network(const char *fstype) {
        static const char table[] =
                "afs\0"
                "cifs\0"
                "smbfs\0"
                "sshfs\0"
                "ncpfs\0"
                "ncp\0"
                "nfs\0"
                "nfs4\0"
                "gfs\0"
                "gfs2\0"
                "glusterfs\0"
                "pvfs2\0" /* OrangeFS */
                "ocfs2\0"
                "lustre\0"
                ;

        const char *x;

        x = startswith(fstype, "fuse.");
        if (x)
                fstype = x;

        return nulstr_contains(table, fstype);
}

int repeat_unmount(const char *path, int flags) {
        bool done = false;

        assert(path);

        /* If there are multiple mounts on a mount point, this
         * removes them all */

        for (;;) {
                if (umount2(path, flags) < 0) {

                        if (errno == EINVAL)
                                return done;

                        return -errno;
                }

                done = true;
        }
}

const char* mode_to_inaccessible_node(mode_t mode) {
        /* This function maps a node type to the correspondent inaccessible node type.
         * Character and block inaccessible devices may not be created (because major=0 and minor=0),
         * in such case we map character and block devices to the inaccessible node type socket. */
        switch(mode & S_IFMT) {
                case S_IFREG:
                        return "/run/systemd/inaccessible/reg";
                case S_IFDIR:
                        return "/run/systemd/inaccessible/dir";
                case S_IFCHR:
                        if (access("/run/systemd/inaccessible/chr", F_OK) == 0)
                                return "/run/systemd/inaccessible/chr";
                        return "/run/systemd/inaccessible/sock";
                case S_IFBLK:
                        if (access("/run/systemd/inaccessible/blk", F_OK) == 0)
                                return "/run/systemd/inaccessible/blk";
                        return "/run/systemd/inaccessible/sock";
                case S_IFIFO:
                        return "/run/systemd/inaccessible/fifo";
                case S_IFSOCK:
                        return "/run/systemd/inaccessible/sock";
        }
        return NULL;
}

#define FLAG(name) (flags & name ? STRINGIFY(name) "|" : "")
static char* mount_flags_to_string(long unsigned flags) {
        char *x;
        _cleanup_free_ char *y = NULL;
        long unsigned overflow;

        overflow = flags & ~(MS_RDONLY |
                             MS_NOSUID |
                             MS_NODEV |
                             MS_NOEXEC |
                             MS_SYNCHRONOUS |
                             MS_REMOUNT |
                             MS_MANDLOCK |
                             MS_DIRSYNC |
                             MS_NOATIME |
                             MS_NODIRATIME |
                             MS_BIND |
                             MS_MOVE |
                             MS_REC |
                             MS_SILENT |
                             MS_POSIXACL |
                             MS_UNBINDABLE |
                             MS_PRIVATE |
                             MS_SLAVE |
                             MS_SHARED |
                             MS_RELATIME |
                             MS_KERNMOUNT |
                             MS_I_VERSION |
                             MS_STRICTATIME |
                             MS_LAZYTIME);

        if (flags == 0 || overflow != 0)
                if (asprintf(&y, "%lx", overflow) < 0)
                        return NULL;

        x = strjoin(FLAG(MS_RDONLY),
                    FLAG(MS_NOSUID),
                    FLAG(MS_NODEV),
                    FLAG(MS_NOEXEC),
                    FLAG(MS_SYNCHRONOUS),
                    FLAG(MS_REMOUNT),
                    FLAG(MS_MANDLOCK),
                    FLAG(MS_DIRSYNC),
                    FLAG(MS_NOATIME),
                    FLAG(MS_NODIRATIME),
                    FLAG(MS_BIND),
                    FLAG(MS_MOVE),
                    FLAG(MS_REC),
                    FLAG(MS_SILENT),
                    FLAG(MS_POSIXACL),
                    FLAG(MS_UNBINDABLE),
                    FLAG(MS_PRIVATE),
                    FLAG(MS_SLAVE),
                    FLAG(MS_SHARED),
                    FLAG(MS_RELATIME),
                    FLAG(MS_KERNMOUNT),
                    FLAG(MS_I_VERSION),
                    FLAG(MS_STRICTATIME),
                    FLAG(MS_LAZYTIME),
                    y);
        if (!x)
                return NULL;
        if (!y)
                x[strlen(x) - 1] = '\0'; /* truncate the last | */
        return x;
}

int mount_verbose(
                int error_log_level,
                const char *what,
                const char *where,
                const char *type,
                unsigned long flags,
                const char *options) {

        _cleanup_free_ char *fl = NULL;

        fl = mount_flags_to_string(flags);

        if ((flags & MS_REMOUNT) && !what && !type)
                log_debug("Remounting %s (%s \"%s\")...",
                          where, strnull(fl), strempty(options));
        else if (!what && !type)
                log_debug("Mounting %s (%s \"%s\")...",
                          where, strnull(fl), strempty(options));
        else if ((flags & MS_BIND) && !type)
                log_debug("Bind-mounting %s on %s (%s \"%s\")...",
                          what, where, strnull(fl), strempty(options));
        else
                log_debug("Mounting %s on %s (%s \"%s\")...",
                          strna(type), where, strnull(fl), strempty(options));
        if (mount(what, where, type, flags, options) < 0)
                return log_full_errno(error_log_level, errno,
                                      "Failed to mount %s on %s (%s \"%s\"): %m",
                                      strna(type), where, strnull(fl), strempty(options));
        return 0;
}

int umount_verbose(const char *what) {
        log_debug("Umounting %s...", what);
        if (umount(what) < 0)
                return log_error_errno(errno, "Failed to unmount %s: %m", what);
        return 0;
}
Commit	Line	Data
4349cd7c LP	1	/***
	2	This file is part of systemd.
	3
	4	Copyright 2010 Lennart Poettering
	5
	6	systemd is free software; you can redistribute it and/or modify it
	7	under the terms of the GNU Lesser General Public License as published by
	8	the Free Software Foundation; either version 2.1 of the License, or
	9	(at your option) any later version.
	10
	11	systemd is distributed in the hope that it will be useful, but
	12	WITHOUT ANY WARRANTY; without even the implied warranty of
	13	MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	14	Lesser General Public License for more details.
	15
	16	You should have received a copy of the GNU Lesser General Public License
	17	along with systemd; If not, see <http://www.gnu.org/licenses/>.
	18	***/
	19
11c3a366 TA	20	#include <errno.h>
11c3a366 TA	21	#include <stdlib.h>
4349cd7c LP	22	#include <string.h>
4349cd7c LP	23	#include <sys/mount.h>
11c3a366	24	#include <sys/stat.h>
4349cd7c	25	#include <sys/statvfs.h>
11c3a366	26	#include <unistd.h>
4349cd7c	27
b5efdb8a	28	#include "alloc-util.h"
4349cd7c LP	29	#include "escape.h"
	30	#include "fd-util.h"
	31	#include "fileio.h"
e1873695	32	#include "fs-util.h"
93cc7779	33	#include "hashmap.h"
4349cd7c LP	34	#include "mount-util.h"
	35	#include "parse-util.h"
	36	#include "path-util.h"
	37	#include "set.h"
15a5e950	38	#include "stdio-util.h"
4349cd7c	39	#include "string-util.h"
6b7c9f8b	40	#include "strv.h"
4349cd7c LP	41
	42	static int fd_fdinfo_mnt_id(int fd, const char filename, int flags, int mnt_id) {
	43	char path[strlen("/proc/self/fdinfo/") + DECIMAL_STR_MAX(int)];
	44	_cleanup_free_ char *fdinfo = NULL;
	45	_cleanup_close_ int subfd = -1;
	46	char *p;
	47	int r;
	48
	49	if ((flags & AT_EMPTY_PATH) && isempty(filename))
	50	xsprintf(path, "/proc/self/fdinfo/%i", fd);
	51	else {
c4b69156	52	subfd = openat(fd, filename, O_CLOEXEC\|O_PATH);
4349cd7c LP	53	if (subfd < 0)
	54	return -errno;
	55
	56	xsprintf(path, "/proc/self/fdinfo/%i", subfd);
	57	}
	58
	59	r = read_full_file(path, &fdinfo, NULL);
	60	if (r == -ENOENT) /* The fdinfo directory is a relatively new addition */
	61	return -EOPNOTSUPP;
	62	if (r < 0)
	63	return -errno;
	64
	65	p = startswith(fdinfo, "mnt_id:");
	66	if (!p) {
	67	p = strstr(fdinfo, "\nmnt_id:");
	68	if (!p) /* The mnt_id field is a relatively new addition */
	69	return -EOPNOTSUPP;
	70
	71	p += 8;
	72	}
	73
	74	p += strspn(p, WHITESPACE);
	75	p[strcspn(p, WHITESPACE)] = 0;
	76
	77	return safe_atoi(p, mnt_id);
	78	}
	79
4349cd7c LP	80	int fd_is_mount_point(int fd, const char *filename, int flags) {
	81	union file_handle_union h = FILE_HANDLE_INIT, h_parent = FILE_HANDLE_INIT;
	82	int mount_id = -1, mount_id_parent = -1;
	83	bool nosupp = false, check_st_dev = true;
	84	struct stat a, b;
	85	int r;
	86
	87	assert(fd >= 0);
	88	assert(filename);
	89
	90	/* First we will try the name_to_handle_at() syscall, which
	91	* tells us the mount id and an opaque file "handle". It is
	92	* not supported everywhere though (kernel compile-time
	93	* option, not all file systems are hooked up). If it works
	94	* the mount id is usually good enough to tell us whether
	95	* something is a mount point.
	96	*
	97	* If that didn't work we will try to read the mount id from
	98	* /proc/self/fdinfo/<fd>. This is almost as good as
	99	* name_to_handle_at(), however, does not return the
	100	* opaque file handle. The opaque file handle is pretty useful
	101	* to detect the root directory, which we should always
	102	* consider a mount point. Hence we use this only as
	103	* fallback. Exporting the mnt_id in fdinfo is a pretty recent
	104	* kernel addition.
	105	*
	106	* As last fallback we do traditional fstat() based st_dev
	107	* comparisons. This is how things were traditionally done,
61233823	108	* but unionfs breaks this since it exposes file
4349cd7c LP	109	* systems with a variety of st_dev reported. Also, btrfs
	110	* subvolumes have different st_dev, even though they aren't
	111	* real mounts of their own. */
	112
	113	r = name_to_handle_at(fd, filename, &h.handle, &mount_id, flags);
	114	if (r < 0) {
	115	if (errno == ENOSYS)
	116	/* This kernel does not support name_to_handle_at()
	117	* fall back to simpler logic. */
	118	goto fallback_fdinfo;
	119	else if (errno == EOPNOTSUPP)
	120	/* This kernel or file system does not support
	121	* name_to_handle_at(), hence let's see if the
	122	* upper fs supports it (in which case it is a
	123	* mount point), otherwise fallback to the
	124	* traditional stat() logic */
	125	nosupp = true;
	126	else
	127	return -errno;
	128	}
	129
	130	r = name_to_handle_at(fd, "", &h_parent.handle, &mount_id_parent, AT_EMPTY_PATH);
	131	if (r < 0) {
	132	if (errno == EOPNOTSUPP) {
	133	if (nosupp)
	134	/* Neither parent nor child do name_to_handle_at()?
	135	We have no choice but to fall back. */
	136	goto fallback_fdinfo;
	137	else
	138	/* The parent can't do name_to_handle_at() but the
	139	* directory we are interested in can?
	140	* If so, it must be a mount point. */
	141	return 1;
	142	} else
	143	return -errno;
	144	}
	145
	146	/* The parent can do name_to_handle_at() but the
	147	* directory we are interested in can't? If so, it
	148	* must be a mount point. */
	149	if (nosupp)
	150	return 1;
	151
	152	/* If the file handle for the directory we are
	153	* interested in and its parent are identical, we
	154	* assume this is the root directory, which is a mount
	155	* point. */
	156
	157	if (h.handle.handle_bytes == h_parent.handle.handle_bytes &&
	158	h.handle.handle_type == h_parent.handle.handle_type &&
	159	memcmp(h.handle.f_handle, h_parent.handle.f_handle, h.handle.handle_bytes) == 0)
	160	return 1;
	161
	162	return mount_id != mount_id_parent;
	163
	164	fallback_fdinfo:
	165	r = fd_fdinfo_mnt_id(fd, filename, flags, &mount_id);
548bd573	166	if (IN_SET(r, -EOPNOTSUPP, -EACCES))
4349cd7c LP	167	goto fallback_fstat;
	168	if (r < 0)
	169	return r;
	170
	171	r = fd_fdinfo_mnt_id(fd, "", AT_EMPTY_PATH, &mount_id_parent);
	172	if (r < 0)
	173	return r;
	174
	175	if (mount_id != mount_id_parent)
	176	return 1;
	177
	178	/* Hmm, so, the mount ids are the same. This leaves one
	179	* special case though for the root file system. For that,
	180	* let's see if the parent directory has the same inode as we
	181	* are interested in. Hence, let's also do fstat() checks now,
	182	* too, but avoid the st_dev comparisons, since they aren't
	183	* that useful on unionfs mounts. */
	184	check_st_dev = false;
	185
	186	fallback_fstat:
	187	/* yay for fstatat() taking a different set of flags than the other
	188	* _at() above */
	189	if (flags & AT_SYMLINK_FOLLOW)
	190	flags &= ~AT_SYMLINK_FOLLOW;
	191	else
	192	flags \|= AT_SYMLINK_NOFOLLOW;
	193	if (fstatat(fd, filename, &a, flags) < 0)
	194	return -errno;
	195
	196	if (fstatat(fd, "", &b, AT_EMPTY_PATH) < 0)
	197	return -errno;
	198
	199	/* A directory with same device and inode as its parent? Must
	200	* be the root directory */
	201	if (a.st_dev == b.st_dev &&
	202	a.st_ino == b.st_ino)
	203	return 1;
	204
	205	return check_st_dev && (a.st_dev != b.st_dev);
	206	}
	207
	208	/* flags can be AT_SYMLINK_FOLLOW or 0 */
e1873695	209	int path_is_mount_point(const char t, const char root, int flags) {
4349cd7c	210	_cleanup_free_ char canonical = NULL, parent = NULL;
e1873695 LP	211	_cleanup_close_ int fd = -1;
e1873695 LP	212	int r;
4349cd7c LP	213
	214	assert(t);
	215
	216	if (path_equal(t, "/"))
	217	return 1;
	218
	219	/* we need to resolve symlinks manually, we can't just rely on
	220	* fd_is_mount_point() to do that for us; if we have a structure like
	221	* /bin -> /usr/bin/ and /usr is a mount point, then the parent that we
	222	* look at needs to be /usr, not /. */
	223	if (flags & AT_SYMLINK_FOLLOW) {
c4f4fce7	224	r = chase_symlinks(t, root, 0, &canonical);
e1873695 LP	225	if (r < 0)
e1873695 LP	226	return r;
4349cd7c LP	227
	228	t = canonical;
	229	}
	230
	231	parent = dirname_malloc(t);
	232	if (!parent)
	233	return -ENOMEM;
	234
c4b69156	235	fd = openat(AT_FDCWD, parent, O_DIRECTORY\|O_CLOEXEC\|O_PATH);
4349cd7c LP	236	if (fd < 0)
	237	return -errno;
	238
	239	return fd_is_mount_point(fd, basename(t), flags);
	240	}
	241
	242	int umount_recursive(const char *prefix, int flags) {
	243	bool again;
	244	int n = 0, r;
	245
	246	/* Try to umount everything recursively below a
	247	* directory. Also, take care of stacked mounts, and keep
	248	* unmounting them until they are gone. */
	249
	250	do {
	251	_cleanup_fclose_ FILE *proc_self_mountinfo = NULL;
	252
	253	again = false;
	254	r = 0;
	255
	256	proc_self_mountinfo = fopen("/proc/self/mountinfo", "re");
	257	if (!proc_self_mountinfo)
	258	return -errno;
	259
	260	for (;;) {
	261	_cleanup_free_ char path = NULL, p = NULL;
	262	int k;
	263
	264	k = fscanf(proc_self_mountinfo,
	265	"%s " / (1) mount id */
	266	"%s " / (2) parent id */
	267	"%s " / (3) major:minor */
	268	"%s " / (4) root */
	269	"%ms " /* (5) mount point */
	270	"%s" / (6) mount options */
	271	"%[^-]" / (7) optional fields */
	272	"- " /* (8) separator */
	273	"%s " / (9) file system type */
	274	"%s" / (10) mount source */
	275	"%s" / (11) mount options 2 */
	276	"%[^\n]", / some rubbish at the end */
	277	&path);
	278	if (k != 1) {
	279	if (k == EOF)
	280	break;
	281
	282	continue;
	283	}
	284
	285	r = cunescape(path, UNESCAPE_RELAX, &p);
	286	if (r < 0)
	287	return r;
	288
	289	if (!path_startswith(p, prefix))
	290	continue;
	291
	292	if (umount2(p, flags) < 0) {
6b7c9f8b	293	r = log_debug_errno(errno, "Failed to umount %s: %m", p);
4349cd7c LP	294	continue;
	295	}
	296
6b7c9f8b LP	297	log_debug("Successfully unmounted %s", p);
6b7c9f8b LP	298
4349cd7c LP	299	again = true;
	300	n++;
	301
	302	break;
	303	}
	304
	305	} while (again);
	306
	307	return r ? r : n;
	308	}
	309
	310	static int get_mount_flags(const char path, unsigned long flags) {
	311	struct statvfs buf;
	312
	313	if (statvfs(path, &buf) < 0)
	314	return -errno;
	315	*flags = buf.f_flag;
	316	return 0;
	317	}
	318
6b7c9f8b	319	int bind_remount_recursive(const char prefix, bool ro, char *blacklist) {
4349cd7c LP	320	_cleanup_set_free_free_ Set *done = NULL;
	321	_cleanup_free_ char *cleaned = NULL;
	322	int r;
	323
6b7c9f8b LP	324	/* Recursively remount a directory (and all its submounts) read-only or read-write. If the directory is already
	325	* mounted, we reuse the mount and simply mark it MS_BIND\|MS_RDONLY (or remove the MS_RDONLY for read-write
	326	* operation). If it isn't we first make it one. Afterwards we apply MS_BIND\|MS_RDONLY (or remove MS_RDONLY) to
	327	* all submounts we can access, too. When mounts are stacked on the same mount point we only care for each
	328	* individual "top-level" mount on each point, as we cannot influence/access the underlying mounts anyway. We
	329	* do not have any effect on future submounts that might get propagated, they migt be writable. This includes
	330	* future submounts that have been triggered via autofs.
	331	*
	332	* If the "blacklist" parameter is specified it may contain a list of subtrees to exclude from the
	333	* remount operation. Note that we'll ignore the blacklist for the top-level path. */
4349cd7c LP	334
	335	cleaned = strdup(prefix);
	336	if (!cleaned)
	337	return -ENOMEM;
	338
	339	path_kill_slashes(cleaned);
	340
	341	done = set_new(&string_hash_ops);
	342	if (!done)
	343	return -ENOMEM;
	344
	345	for (;;) {
	346	_cleanup_fclose_ FILE *proc_self_mountinfo = NULL;
	347	_cleanup_set_free_free_ Set *todo = NULL;
	348	bool top_autofs = false;
	349	char *x;
	350	unsigned long orig_flags;
	351
	352	todo = set_new(&string_hash_ops);
	353	if (!todo)
	354	return -ENOMEM;
	355
	356	proc_self_mountinfo = fopen("/proc/self/mountinfo", "re");
	357	if (!proc_self_mountinfo)
	358	return -errno;
	359
	360	for (;;) {
	361	_cleanup_free_ char path = NULL, p = NULL, *type = NULL;
	362	int k;
	363
	364	k = fscanf(proc_self_mountinfo,
	365	"%s " / (1) mount id */
	366	"%s " / (2) parent id */
	367	"%s " / (3) major:minor */
	368	"%s " / (4) root */
	369	"%ms " /* (5) mount point */
	370	"%s" / (6) mount options (superblock) */
	371	"%[^-]" / (7) optional fields */
	372	"- " /* (8) separator */
	373	"%ms " /* (9) file system type */
	374	"%s" / (10) mount source */
	375	"%s" / (11) mount options (bind mount) */
	376	"%[^\n]", / some rubbish at the end */
	377	&path,
	378	&type);
	379	if (k != 2) {
	380	if (k == EOF)
	381	break;
	382
	383	continue;
	384	}
	385
	386	r = cunescape(path, UNESCAPE_RELAX, &p);
	387	if (r < 0)
	388	return r;
	389
6b7c9f8b LP	390	if (!path_startswith(p, cleaned))
	391	continue;
	392
	393	/* Ignore this mount if it is blacklisted, but only if it isn't the top-level mount we shall
	394	* operate on. */
	395	if (!path_equal(cleaned, p)) {
	396	bool blacklisted = false;
	397	char **i;
	398
	399	STRV_FOREACH(i, blacklist) {
	400
	401	if (path_equal(*i, cleaned))
	402	continue;
	403
	404	if (!path_startswith(*i, cleaned))
	405	continue;
	406
	407	if (path_startswith(p, *i)) {
	408	blacklisted = true;
	409	log_debug("Not remounting %s, because blacklisted by %s, called for %s", p, *i, cleaned);
	410	break;
	411	}
	412	}
	413	if (blacklisted)
	414	continue;
	415	}
	416
4349cd7c LP	417	/* Let's ignore autofs mounts. If they aren't
	418	* triggered yet, we want to avoid triggering
	419	* them, as we don't make any guarantees for
	420	* future submounts anyway. If they are
	421	* already triggered, then we will find
	422	* another entry for this. */
	423	if (streq(type, "autofs")) {
	424	top_autofs = top_autofs \|\| path_equal(cleaned, p);
	425	continue;
	426	}
	427
6b7c9f8b	428	if (!set_contains(done, p)) {
4349cd7c LP	429	r = set_consume(todo, p);
4349cd7c LP	430	p = NULL;
4349cd7c LP	431	if (r == -EEXIST)
	432	continue;
	433	if (r < 0)
	434	return r;
	435	}
	436	}
	437
	438	/* If we have no submounts to process anymore and if
	439	* the root is either already done, or an autofs, we
	440	* are done */
	441	if (set_isempty(todo) &&
	442	(top_autofs \|\| set_contains(done, cleaned)))
	443	return 0;
	444
	445	if (!set_contains(done, cleaned) &&
	446	!set_contains(todo, cleaned)) {
6b7c9f8b	447	/* The prefix directory itself is not yet a mount, make it one. */
4349cd7c LP	448	if (mount(cleaned, cleaned, NULL, MS_BIND\|MS_REC, NULL) < 0)
	449	return -errno;
	450
	451	orig_flags = 0;
	452	(void) get_mount_flags(cleaned, &orig_flags);
	453	orig_flags &= ~MS_RDONLY;
	454
	455	if (mount(NULL, prefix, NULL, orig_flags\|MS_BIND\|MS_REMOUNT\|(ro ? MS_RDONLY : 0), NULL) < 0)
	456	return -errno;
	457
6b7c9f8b LP	458	log_debug("Made top-level directory %s a mount point.", prefix);
6b7c9f8b LP	459
4349cd7c LP	460	x = strdup(cleaned);
	461	if (!x)
	462	return -ENOMEM;
	463
	464	r = set_consume(done, x);
	465	if (r < 0)
	466	return r;
	467	}
	468
	469	while ((x = set_steal_first(todo))) {
	470
	471	r = set_consume(done, x);
	472	if (r == -EEXIST \|\| r == 0)
	473	continue;
	474	if (r < 0)
	475	return r;
	476
6b7c9f8b	477	/* Deal with mount points that are obstructed by a later mount */
e1873695	478	r = path_is_mount_point(x, NULL, 0);
98df8089 AC	479	if (r == -ENOENT \|\| r == 0)
	480	continue;
	481	if (r < 0)
	482	return r;
	483
	484	/* Try to reuse the original flag set */
4349cd7c LP	485	orig_flags = 0;
	486	(void) get_mount_flags(x, &orig_flags);
	487	orig_flags &= ~MS_RDONLY;
	488
98df8089 AC	489	if (mount(NULL, x, NULL, orig_flags\|MS_BIND\|MS_REMOUNT\|(ro ? MS_RDONLY : 0), NULL) < 0)
98df8089 AC	490	return -errno;
4349cd7c	491
6b7c9f8b	492	log_debug("Remounted %s read-only.", x);
4349cd7c LP	493	}
	494	}
	495	}
	496
	497	int mount_move_root(const char *path) {
	498	assert(path);
	499
	500	if (chdir(path) < 0)
	501	return -errno;
	502
	503	if (mount(path, "/", NULL, MS_MOVE, NULL) < 0)
	504	return -errno;
	505
	506	if (chroot(".") < 0)
	507	return -errno;
	508
	509	if (chdir("/") < 0)
	510	return -errno;
	511
	512	return 0;
	513	}
4e036b7a LP	514
	515	bool fstype_is_network(const char *fstype) {
	516	static const char table[] =
	517	"afs\0"
	518	"cifs\0"
	519	"smbfs\0"
	520	"sshfs\0"
	521	"ncpfs\0"
	522	"ncp\0"
	523	"nfs\0"
	524	"nfs4\0"
	525	"gfs\0"
	526	"gfs2\0"
a44cb5cb ZJS	527	"glusterfs\0"
a44cb5cb ZJS	528	"pvfs2\0" /* OrangeFS */
0a86e681	529	"ocfs2\0"
67ae4366	530	"lustre\0"
a44cb5cb	531	;
4e036b7a LP	532
	533	const char *x;
	534
	535	x = startswith(fstype, "fuse.");
	536	if (x)
	537	fstype = x;
	538
	539	return nulstr_contains(table, fstype);
	540	}
3f2c0bec LP	541
	542	int repeat_unmount(const char *path, int flags) {
	543	bool done = false;
	544
	545	assert(path);
	546
	547	/* If there are multiple mounts on a mount point, this
	548	* removes them all */
	549
	550	for (;;) {
	551	if (umount2(path, flags) < 0) {
	552
	553	if (errno == EINVAL)
	554	return done;
	555
	556	return -errno;
	557	}
	558
	559	done = true;
	560	}
	561	}
c4b41707 AP	562
c4b41707 AP	563	const char* mode_to_inaccessible_node(mode_t mode) {
b3d1d516 AP	564	/* This function maps a node type to the correspondent inaccessible node type.
	565	* Character and block inaccessible devices may not be created (because major=0 and minor=0),
	566	* in such case we map character and block devices to the inaccessible node type socket. */
c4b41707 AP	567	switch(mode & S_IFMT) {
	568	case S_IFREG:
	569	return "/run/systemd/inaccessible/reg";
	570	case S_IFDIR:
	571	return "/run/systemd/inaccessible/dir";
	572	case S_IFCHR:
b3d1d516 AP	573	if (access("/run/systemd/inaccessible/chr", F_OK) == 0)
	574	return "/run/systemd/inaccessible/chr";
	575	return "/run/systemd/inaccessible/sock";
c4b41707	576	case S_IFBLK:
b3d1d516 AP	577	if (access("/run/systemd/inaccessible/blk", F_OK) == 0)
	578	return "/run/systemd/inaccessible/blk";
	579	return "/run/systemd/inaccessible/sock";
c4b41707 AP	580	case S_IFIFO:
	581	return "/run/systemd/inaccessible/fifo";
	582	case S_IFSOCK:
	583	return "/run/systemd/inaccessible/sock";
	584	}
	585	return NULL;
	586	}
60e76d48 ZJS	587
	588	#define FLAG(name) (flags & name ? STRINGIFY(name) "\|" : "")
	589	static char* mount_flags_to_string(long unsigned flags) {
	590	char *x;
	591	_cleanup_free_ char *y = NULL;
	592	long unsigned overflow;
	593
	594	overflow = flags & ~(MS_RDONLY \|
	595	MS_NOSUID \|
	596	MS_NODEV \|
	597	MS_NOEXEC \|
	598	MS_SYNCHRONOUS \|
	599	MS_REMOUNT \|
	600	MS_MANDLOCK \|
	601	MS_DIRSYNC \|
	602	MS_NOATIME \|
	603	MS_NODIRATIME \|
	604	MS_BIND \|
	605	MS_MOVE \|
	606	MS_REC \|
	607	MS_SILENT \|
	608	MS_POSIXACL \|
	609	MS_UNBINDABLE \|
	610	MS_PRIVATE \|
	611	MS_SLAVE \|
	612	MS_SHARED \|
	613	MS_RELATIME \|
	614	MS_KERNMOUNT \|
	615	MS_I_VERSION \|
	616	MS_STRICTATIME \|
	617	MS_LAZYTIME);
	618
	619	if (flags == 0 \|\| overflow != 0)
	620	if (asprintf(&y, "%lx", overflow) < 0)
	621	return NULL;
	622
	623	x = strjoin(FLAG(MS_RDONLY),
	624	FLAG(MS_NOSUID),
	625	FLAG(MS_NODEV),
	626	FLAG(MS_NOEXEC),
	627	FLAG(MS_SYNCHRONOUS),
	628	FLAG(MS_REMOUNT),
	629	FLAG(MS_MANDLOCK),
	630	FLAG(MS_DIRSYNC),
	631	FLAG(MS_NOATIME),
	632	FLAG(MS_NODIRATIME),
	633	FLAG(MS_BIND),
	634	FLAG(MS_MOVE),
	635	FLAG(MS_REC),
	636	FLAG(MS_SILENT),
	637	FLAG(MS_POSIXACL),
	638	FLAG(MS_UNBINDABLE),
	639	FLAG(MS_PRIVATE),
	640	FLAG(MS_SLAVE),
	641	FLAG(MS_SHARED),
	642	FLAG(MS_RELATIME),
	643	FLAG(MS_KERNMOUNT),
	644	FLAG(MS_I_VERSION),
	645	FLAG(MS_STRICTATIME),
	646	FLAG(MS_LAZYTIME),
605405c6	647	y);
60e76d48 ZJS	648	if (!x)
	649	return NULL;
	650	if (!y)
	651	x[strlen(x) - 1] = '\0'; /* truncate the last \| */
	652	return x;
	653	}
	654
	655	int mount_verbose(
	656	int error_log_level,
	657	const char *what,
	658	const char *where,
	659	const char *type,
	660	unsigned long flags,
	661	const char *options) {
	662
	663	_cleanup_free_ char *fl = NULL;
	664
	665	fl = mount_flags_to_string(flags);
	666
	667	if ((flags & MS_REMOUNT) && !what && !type)
	668	log_debug("Remounting %s (%s \"%s\")...",
	669	where, strnull(fl), strempty(options));
	670	else if (!what && !type)
	671	log_debug("Mounting %s (%s \"%s\")...",
	672	where, strnull(fl), strempty(options));
	673	else if ((flags & MS_BIND) && !type)
	674	log_debug("Bind-mounting %s on %s (%s \"%s\")...",
	675	what, where, strnull(fl), strempty(options));
	676	else
	677	log_debug("Mounting %s on %s (%s \"%s\")...",
	678	strna(type), where, strnull(fl), strempty(options));
	679	if (mount(what, where, type, flags, options) < 0)
	680	return log_full_errno(error_log_level, errno,
	681	"Failed to mount %s on %s (%s \"%s\"): %m",
	682	strna(type), where, strnull(fl), strempty(options));
	683	return 0;
	684	}
	685
	686	int umount_verbose(const char *what) {
	687	log_debug("Umounting %s...", what);
	688	if (umount(what) < 0)
	689	return log_error_errno(errno, "Failed to unmount %s: %m", what);
	690	return 0;
	691	}