[thirdparty/util-linux.git] / libmount / src / utils.c

/* SPDX-License-Identifier: LGPL-2.1-or-later */
/*
 * This file is part of libmount from util-linux project.
 *
 * Copyright (C) 2008-2018 Karel Zak <kzak@redhat.com>
 *
 * libmount 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.
 */

/**
 * SECTION: utils
 * @title: Utils
 * @short_description: misc utils.
 */
#include <ctype.h>
#include <fcntl.h>
#include <pwd.h>
#include <grp.h>
#include <poll.h>
#include <blkid.h>

#include "strutils.h"
#include "pathnames.h"
#include "mountP.h"
#include "mangle.h"
#include "canonicalize.h"
#include "env.h"
#include "match.h"
#include "fileutils.h"
#include "statfs_magic.h"
#include "sysfs.h"

int append_string(char **a, const char *b)
{
        size_t al, bl;
        char *tmp;

        assert(a);

        if (!b || !*b)
                return 0;
        if (!*a) {
                *a = strdup(b);
                return !*a ? -ENOMEM : 0;
        }

        al = strlen(*a);
        bl = strlen(b);

        tmp = realloc(*a, al + bl + 1);
        if (!tmp)
                return -ENOMEM;
        *a = tmp;
        memcpy((*a) + al, b, bl + 1);
        return 0;
}

/*
 * Return 1 if the file is not accessible or empty
 */
int is_file_empty(const char *name)
{
        struct stat st;
        assert(name);

        return (stat(name, &st) != 0 || st.st_size == 0);
}

int mnt_valid_tagname(const char *tagname)
{
        if (tagname && *tagname && (
            strcmp("UUID", tagname) == 0 ||
            strcmp("LABEL", tagname) == 0 ||
            strcmp("PARTUUID", tagname) == 0 ||
            strcmp("PARTLABEL", tagname) == 0))
                return 1;

        return 0;
}

/**
 * mnt_tag_is_valid:
 * @tag: NAME=value string
 *
 * Returns: 1 if the @tag is parsable and tag NAME= is supported by libmount, or 0.
 */
int mnt_tag_is_valid(const char *tag)
{
        char *t = NULL;
        int rc = tag && blkid_parse_tag_string(tag, &t, NULL) == 0
                     && mnt_valid_tagname(t);

        free(t);
        return rc;
}

int mnt_parse_offset(const char *str, size_t len, uintmax_t *res)
{
        char *p;
        int rc = 0;

        if (!str || !*str)
                return -EINVAL;

        p = strndup(str, len);
        if (!p)
                return -errno;

        if (strtosize(p, res))
                rc = -EINVAL;
        free(p);
        return rc;
}

/* used as a callback by bsearch in mnt_fstype_is_pseudofs() */
static int fstype_cmp(const void *v1, const void *v2)
{
        const char *s1 = *(char * const *)v1;
        const char *s2 = *(char * const *)v2;

        return strcmp(s1, s2);
}

int mnt_stat_mountpoint(const char *target, struct stat *st)
{
#ifdef AT_NO_AUTOMOUNT
        return fstatat(AT_FDCWD, target, st, AT_NO_AUTOMOUNT);
#else
        return stat(target, st);
#endif
}

int mnt_lstat_mountpoint(const char *target, struct stat *st)
{
#ifdef AT_NO_AUTOMOUNT
        return fstatat(AT_FDCWD, target, st, AT_NO_AUTOMOUNT | AT_SYMLINK_NOFOLLOW);
#else
        return lstat(target, st);
#endif
}


/*
 * Note that the @target has to be an absolute path (so at least "/").  The
 * @filename returns an allocated buffer with the last path component, for example:
 *
 * mnt_chdir_to_parent("/mnt/test", &buf) ==> chdir("/mnt"), buf="test"
 */
int mnt_chdir_to_parent(const char *target, char **filename)
{
        char *buf, *parent, *last = NULL;
        char cwd[PATH_MAX];
        int rc = -EINVAL;

        if (!target || *target != '/')
                return -EINVAL;

        DBG(UTILS, ul_debug("moving to %s parent", target));

        buf = strdup(target);
        if (!buf)
                return -ENOMEM;

        if (*(buf + 1) != '\0') {
                last = stripoff_last_component(buf);
                if (!last)
                        goto err;
        }

        parent = buf && *buf ? buf : "/";

        if (chdir(parent) == -1) {
                DBG(UTILS, ul_debug("failed to chdir to %s: %m", parent));
                rc = -errno;
                goto err;
        }
        if (!getcwd(cwd, sizeof(cwd))) {
                DBG(UTILS, ul_debug("failed to obtain current directory: %m"));
                rc = -errno;
                goto err;
        }
        if (strcmp(cwd, parent) != 0) {
                DBG(UTILS, ul_debug(
                    "unexpected chdir (expected=%s, cwd=%s)", parent, cwd));
                goto err;
        }

        DBG(CXT, ul_debug(
                "current directory moved to %s [last_component='%s']",
                parent, last));

        if (filename) {
                *filename = buf;

                if (!last || !*last)
                        memcpy(*filename, ".", 2);
                else
                        memmove(*filename, last, strlen(last) + 1);
        } else
                free(buf);
        return 0;
err:
        free(buf);
        return rc;
}

/*
 * Check if @path is on a read-only filesystem independently of file permissions.
 */
int mnt_is_readonly(const char *path)
{
        if (access(path, W_OK) == 0)
                return 0;
        if (errno == EROFS)
                return 1;
        if (errno != EACCES)
                return 0;

#ifdef HAVE_UTIMENSAT
        /*
         * access(2) returns EACCES on read-only FS:
         *
         * - for set-uid application if one component of the path is not
         *   accessible for the current rUID. (Note that euidaccess(2) does not
         *   check for EROFS at all).
         *
         * - for a read-write filesystem with a read-only VFS node (aka -o remount,ro,bind)
         */
        {
                struct timespec times[2];

                DBG(UTILS, ul_debug(" doing utimensat() based write test"));

                times[0].tv_nsec = UTIME_NOW;   /* atime */
                times[1].tv_nsec = UTIME_OMIT;  /* mtime */

                if (utimensat(AT_FDCWD, path, times, 0) == -1)
                        return errno == EROFS;
        }
#endif
        return 0;
}

/**
 * mnt_mangle:
 * @str: string
 *
 * Encode @str to be compatible with fstab/mtab
 *
 * Returns: newly allocated string or NULL in case of error.
 */
char *mnt_mangle(const char *str)
{
        return mangle(str);
}

/**
 * mnt_unmangle:
 * @str: string
 *
 * Decode @str from fstab/mtab
 *
 * Returns: newly allocated string or NULL in case of error.
 */
char *mnt_unmangle(const char *str)
{
        return unmangle(str, NULL);
}

/**
 * mnt_fstype_is_pseudofs:
 * @type: filesystem name
 *
 * Returns: 1 for filesystems like proc, sysfs, ... or 0.
 */
int mnt_fstype_is_pseudofs(const char *type)
{
        /* This array must remain sorted when adding new fstypes */
        static const char *pseudofs[] = {
                "anon_inodefs",
                "apparmorfs",
                "autofs",
                "bdev",
                "binder",
                "binfmt_misc",
                "bpf",
                "cgroup",
                "cgroup2",
                "configfs",
                "cpuset",
                "debugfs",
                "devfs",
                "devpts",
                "devtmpfs",
                "dlmfs",
                "dmabuf",
                "drm",
                "efivarfs",
                "fuse", /* Fallback name of fuse used by many poorly written drivers. */
                "fuse.archivemount", /* Not a true pseudofs (has source), but source is not reported. */
                "fuse.avfsd", /* Not a true pseudofs (has source), but source is not reported. */
                "fuse.dumpfs", /* In fact, it is a netfs, but source is not reported. */
                "fuse.encfs", /* Not a true pseudofs (has source), but source is not reported. */
                "fuse.gvfs-fuse-daemon", /* Old name, not used by gvfs any more. */
                "fuse.gvfsd-fuse",
                "fuse.lxcfs",
                "fuse.rofiles-fuse",
                "fuse.vmware-vmblock",
                "fuse.xwmfs",
                "fusectl",
                "hugetlbfs",
                "ipathfs",
                "mqueue",
                "nfsd",
                "none",
                "nsfs",
                "overlay",
                "pipefs",
                "proc",
                "pstore",
                "ramfs",
                "resctrl",
                "rootfs",
                "rpc_pipefs",
                "securityfs",
                "selinuxfs",
                "smackfs",
                "sockfs",
                "spufs",
                "sysfs",
                "tmpfs",
                "tracefs"
        };

        assert(type);

        return !(bsearch(&type, pseudofs, ARRAY_SIZE(pseudofs),
                                sizeof(char*), fstype_cmp) == NULL);
}

/**
 * mnt_fstype_is_netfs:
 * @type: filesystem name
 *
 * Returns: 1 for filesystems like cifs, nfs, ... or 0.
 */
int mnt_fstype_is_netfs(const char *type)
{
        if (strcmp(type, "cifs")   == 0 ||
            strcmp(type, "smb3")   == 0 ||
            strcmp(type, "smbfs")  == 0 ||
            strncmp(type,"nfs", 3) == 0 ||
            strcmp(type, "afs")    == 0 ||
            strcmp(type, "ncpfs")  == 0 ||
            strcmp(type, "fuse.curlftpfs") == 0 ||
            strcmp(type, "fuse.sshfs") == 0 ||
            strncmp(type,"9p", 2)  == 0)
                return 1;
        return 0;
}

const char *mnt_statfs_get_fstype(struct statfs *vfs)
{
        assert(vfs);

        switch (vfs->f_type) {
        case STATFS_ADFS_MAGIC:         return "adfs";
        case STATFS_AFFS_MAGIC:         return "affs";
        case STATFS_AFS_MAGIC:          return "afs";
        case STATFS_AUTOFS_MAGIC:       return "autofs";
        case STATFS_BDEVFS_MAGIC:       return "bdev";
        case STATFS_BEFS_MAGIC:         return "befs";
        case STATFS_BFS_MAGIC:          return "befs";
        case STATFS_BINFMTFS_MAGIC:     return "binfmt_misc";
        case STATFS_BTRFS_MAGIC:        return "btrfs";
        case STATFS_CEPH_MAGIC:         return "ceph";
        case STATFS_CGROUP_MAGIC:       return "cgroup";
        case STATFS_CIFS_MAGIC:         return "cifs";
        case STATFS_CODA_MAGIC:         return "coda";
        case STATFS_CONFIGFS_MAGIC:     return "configfs";
        case STATFS_CRAMFS_MAGIC:       return "cramfs";
        case STATFS_DEBUGFS_MAGIC:      return "debugfs";
        case STATFS_DEVPTS_MAGIC:       return "devpts";
        case STATFS_ECRYPTFS_MAGIC:     return "ecryptfs";
        case STATFS_EFIVARFS_MAGIC:     return "efivarfs";
        case STATFS_EFS_MAGIC:          return "efs";
        case STATFS_EXOFS_MAGIC:        return "exofs";
        case STATFS_EXT4_MAGIC:         return "ext4";     /* all extN use the same magic */
        case STATFS_F2FS_MAGIC:         return "f2fs";
        case STATFS_FUSE_MAGIC:         return "fuse";
        case STATFS_FUTEXFS_MAGIC:      return "futexfs";
        case STATFS_GFS2_MAGIC:         return "gfs2";
        case STATFS_HFSPLUS_MAGIC:      return "hfsplus";
        case STATFS_HOSTFS_MAGIC:       return "hostfs";
        case STATFS_HPFS_MAGIC:         return "hpfs";
        case STATFS_HPPFS_MAGIC:        return "hppfs";
        case STATFS_HUGETLBFS_MAGIC:    return "hugetlbfs";
        case STATFS_ISOFS_MAGIC:        return "iso9660";
        case STATFS_JFFS2_MAGIC:        return "jffs2";
        case STATFS_JFS_MAGIC:          return "jfs";
        case STATFS_LOGFS_MAGIC:        return "logfs";
        case STATFS_MINIX2_MAGIC:
        case STATFS_MINIX2_MAGIC2:
        case STATFS_MINIX3_MAGIC:
        case STATFS_MINIX_MAGIC:
        case STATFS_MINIX_MAGIC2:       return "minix";
        case STATFS_MQUEUE_MAGIC:       return "mqueue";
        case STATFS_MSDOS_MAGIC:        return "vfat";
        case STATFS_NCP_MAGIC:          return "ncp";
        case STATFS_NFS_MAGIC:          return "nfs";
        case STATFS_NILFS_MAGIC:        return "nilfs2";
        case STATFS_NTFS_MAGIC:         return "ntfs";
        case STATFS_OCFS2_MAGIC:        return "ocfs2";
        case STATFS_OMFS_MAGIC:         return "omfs";
        case STATFS_OPENPROMFS_MAGIC:   return "openpromfs";
        case STATFS_PIPEFS_MAGIC:       return "pipefs";
        case STATFS_PROC_MAGIC:         return "proc";
        case STATFS_PSTOREFS_MAGIC:     return "pstore";
        case STATFS_QNX4_MAGIC:         return "qnx4";
        case STATFS_QNX6_MAGIC:         return "qnx6";
        case STATFS_RAMFS_MAGIC:        return "ramfs";
        case STATFS_REISERFS_MAGIC:     return "reiser4";
        case STATFS_ROMFS_MAGIC:        return "romfs";
        case STATFS_SECURITYFS_MAGIC:   return "securityfs";
        case STATFS_SELINUXFS_MAGIC:    return "selinuxfs";
        case STATFS_SMACKFS_MAGIC:      return "smackfs";
        case STATFS_SMB_MAGIC:          return "smb";
        case STATFS_SOCKFS_MAGIC:       return "sockfs";
        case STATFS_SQUASHFS_MAGIC:     return "squashfs";
        case STATFS_SYSFS_MAGIC:        return "sysfs";
        case STATFS_TMPFS_MAGIC:        return "tmpfs";
        case STATFS_UBIFS_MAGIC:        return "ubifs";
        case STATFS_UDF_MAGIC:          return "udf";
        case STATFS_UFS2_MAGIC:
        case STATFS_UFS_MAGIC:          return "ufs";
        case STATFS_V9FS_MAGIC:         return "9p";
        case STATFS_VXFS_MAGIC:         return "vxfs";
        case STATFS_XENFS_MAGIC:        return "xenfs";
        case STATFS_XFS_MAGIC:          return "xfs";
        default:
                break;
        }

        return NULL;
}

int is_procfs_fd(int fd)
{
        struct statfs sfs;

        return fstatfs(fd, &sfs) == 0 && sfs.f_type == STATFS_PROC_MAGIC;
}

/**
 * mnt_match_fstype:
 * @type: filesystem type
 * @pattern: filesystem name or comma delimited list of names
 *
 * The @pattern list of filesystems can be prefixed with a global
 * "no" prefix to invert matching of the whole list. The "no" could
 * also be used for individual items in the @pattern list. So,
 * "nofoo,bar" has the same meaning as "nofoo,nobar".
 *
 * "bar"  : "nofoo,bar"         -> False   (global "no" prefix)
 *
 * "bar"  : "foo,bar"           -> True
 *
 * "bar" : "foo,nobar"          -> False
 *
 * Returns: 1 if type is matching, else 0. This function also returns
 *          0 if @pattern is NULL and @type is non-NULL.
 */
int mnt_match_fstype(const char *type, const char *pattern)
{
        return match_fstype(type, pattern);
}

void mnt_free_filesystems(char **filesystems)
{
        char **p;

        if (!filesystems)
                return;
        for (p = filesystems; *p; p++)
                free(*p);
        free(filesystems);
}

static int add_filesystem(char ***filesystems, char *name)
{
        int n = 0;

        assert(filesystems);
        assert(name);

        if (*filesystems) {
                char **p;
                for (n = 0, p = *filesystems; *p; p++, n++) {
                        if (strcmp(*p, name) == 0)
                                return 0;
                }
        }

        #define MYCHUNK 16

        if (n == 0 || !((n + 1) % MYCHUNK)) {
                size_t items = ((n + 1 + MYCHUNK) / MYCHUNK) * MYCHUNK;
                char **x = realloc(*filesystems, items * sizeof(char *));

                if (!x)
                        goto err;
                *filesystems = x;
        }
        name = strdup(name);
        (*filesystems)[n] = name;
        (*filesystems)[n + 1] = NULL;
        if (!name)
                goto err;
        return 0;
err:
        mnt_free_filesystems(*filesystems);
        return -ENOMEM;
}

static int get_filesystems(const char *filename, char ***filesystems, const char *pattern)
{
        int rc = 0;
        FILE *f;
        char line[129];

        f = fopen(filename, "r" UL_CLOEXECSTR);
        if (!f)
                return 1;

        DBG(UTILS, ul_debug("reading filesystems list from: %s", filename));

        while (fgets(line, sizeof(line), f)) {
                char name[sizeof(line)];

                if (*line == '#' || strncmp(line, "nodev", 5) == 0)
                        continue;
                if (sscanf(line, " %128[^\n ]\n", name) != 1)
                        continue;
                if (strcmp(name, "*") == 0) {
                        rc = 1;
                        break;          /* end of the /etc/filesystems */
                }
                if (pattern && !mnt_match_fstype(name, pattern))
                        continue;
                rc = add_filesystem(filesystems, name);
                if (rc)
                        break;
        }

        fclose(f);
        return rc;
}

/*
 * Always check the @filesystems pointer!
 *
 * man mount:
 *
 * ...mount will try to read the file /etc/filesystems, or, if that does not
 * exist, /proc/filesystems. All of the filesystem  types  listed  there  will
 * be tried,  except  for  those  that  are  labeled  "nodev"  (e.g.,  devpts,
 * proc  and  nfs).  If /etc/filesystems ends in a line with a single * only,
 * mount will read /proc/filesystems afterwards.
 */
int mnt_get_filesystems(char ***filesystems, const char *pattern)
{
        int rc;

        if (!filesystems)
                return -EINVAL;

        *filesystems = NULL;

        rc = get_filesystems(_PATH_FILESYSTEMS, filesystems, pattern);
        if (rc != 1)
                return rc;

        rc = get_filesystems(_PATH_PROC_FILESYSTEMS, filesystems, pattern);
        if (rc == 1 && *filesystems)
                rc = 0;                 /* /proc/filesystems not found */

        return rc;
}

/*
 * Returns an allocated string with username or NULL.
 */
char *mnt_get_username(const uid_t uid)
{
        struct passwd pwd;
        struct passwd *res;
        char *buf, *username = NULL;

        buf = malloc(UL_GETPW_BUFSIZ);
        if (!buf)
                return NULL;

        if (!getpwuid_r(uid, &pwd, buf, UL_GETPW_BUFSIZ, &res) && res)
                username = strdup(pwd.pw_name);

        free(buf);
        return username;
}

int mnt_get_uid(const char *username, uid_t *uid)
{
        int rc = -1;
        struct passwd pwd;
        struct passwd *pw;
        char *buf;

        if (!username || !uid)
                return -EINVAL;

        buf = malloc(UL_GETPW_BUFSIZ);
        if (!buf)
                return -ENOMEM;

        if (!getpwnam_r(username, &pwd, buf, UL_GETPW_BUFSIZ, &pw) && pw) {
                *uid= pw->pw_uid;
                rc = 0;
        } else {
                DBG(UTILS, ul_debug(
                        "cannot convert '%s' username to UID", username));
                rc = errno ? -errno : -EINVAL;
        }

        free(buf);
        return rc;
}

int mnt_get_gid(const char *groupname, gid_t *gid)
{
        int rc = -1;
        struct group grp;
        struct group *gr;
        char *buf;

        if (!groupname || !gid)
                return -EINVAL;

        buf = malloc(UL_GETPW_BUFSIZ);
        if (!buf)
                return -ENOMEM;

        if (!getgrnam_r(groupname, &grp, buf, UL_GETPW_BUFSIZ, &gr) && gr) {
                *gid= gr->gr_gid;
                rc = 0;
        } else {
                DBG(UTILS, ul_debug(
                        "cannot convert '%s' groupname to GID", groupname));
                rc = errno ? -errno : -EINVAL;
        }

        free(buf);
        return rc;
}

int mnt_in_group(gid_t gid)
{
        int rc = 0, n, i;
        gid_t *grps = NULL;

        if (getgid() == gid)
                return 1;

        n = getgroups(0, NULL);
        if (n <= 0)
                goto done;

        grps = malloc(n * sizeof(*grps));
        if (!grps)
                goto done;

        if (getgroups(n, grps) == n) {
                for (i = 0; i < n; i++) {
                        if (grps[i] == gid) {
                                rc = 1;
                                break;
                        }
                }
        }
done:
        free(grps);
        return rc;
}

static int try_write(const char *filename, const char *directory)
{
        int rc = 0;

        if (!filename)
                return -EINVAL;

        DBG(UTILS, ul_debug("try write %s dir: %s", filename, directory));

#ifdef HAVE_EACCESS
        /* Try eaccess() first, because open() is overkill, may be monitored by
         * audit and we don't want to fill logs by our checks...
         */
        if (eaccess(filename, R_OK|W_OK) == 0) {
                DBG(UTILS, ul_debug(" access OK"));
                return 0;
        }

        if (errno != ENOENT) {
                DBG(UTILS, ul_debug(" access FAILED"));
                return -errno;
        }

        if (directory) {
                /* file does not exist; try if directory is writable */
                if (eaccess(directory, R_OK|W_OK) != 0)
                        rc = -errno;

                DBG(UTILS, ul_debug(" access %s [%s]", rc ? "FAILED" : "OK", directory));
                return rc;
        }
#endif

        DBG(UTILS, ul_debug(" doing open-write test"));

        int fd = open(filename, O_RDWR|O_CREAT|O_CLOEXEC,
                    S_IWUSR|S_IRUSR|S_IRGRP|S_IROTH);
        if (fd < 0)
                rc = -errno;
        else
                close(fd);

        return rc;
}

/**
 * mnt_has_regular_mtab:
 * @mtab: returns path to mtab
 * @writable: returns 1 if the file is writable
 *
 * If the file does not exist and @writable argument is not NULL, then it will
 * try to create the file.
 *
 * Returns: 1 if /etc/mtab is a regular file, and 0 in case of error (check
 *          errno for more details).
 */
int mnt_has_regular_mtab(const char **mtab, int *writable)
{
        struct stat st;
        int rc;
        const char *filename = mtab && *mtab ? *mtab : mnt_get_mtab_path();

        if (writable)
                *writable = 0;
        if (mtab && !*mtab)
                *mtab = filename;

        DBG(UTILS, ul_debug("mtab: %s", filename));

        rc = lstat(filename, &st);

        if (rc == 0) {
                /* file exists */
                if (S_ISREG(st.st_mode)) {
                        if (writable)
                                *writable = !try_write(filename, NULL);
                        DBG(UTILS, ul_debug("%s: writable", filename));
                        return 1;
                }
                goto done;
        }

        /* try to create the file */
        if (writable) {
                *writable = !try_write(filename, NULL);
                if (*writable) {
                        DBG(UTILS, ul_debug("%s: writable", filename));
                        return 1;
                }
        }

done:
        DBG(UTILS, ul_debug("%s: irregular/non-writable", filename));
        return 0;
}

/*
 * Don't export this to libmount API -- utab is private library stuff.
 *
 * If the file does not exist and @writable argument is not NULL, then it will
 * try to create the directory (e.g. /run/mount) and the file.
 *
 * Returns: 1 if utab is a regular file, and 0 in case of
 *          error (check errno for more details).
 */
int mnt_has_regular_utab(const char **utab, int *writable)
{
        struct stat st;
        int rc;
        const char *filename = utab && *utab ? *utab : mnt_get_utab_path();

        if (writable)
                *writable = 0;
        if (utab && !*utab)
                *utab = filename;

        DBG(UTILS, ul_debug("utab: %s", filename));

        rc = lstat(filename, &st);

        if (rc == 0) {
                /* file exists */
                if (S_ISREG(st.st_mode)) {
                        if (writable)
                                *writable = !try_write(filename, NULL);
                        return 1;
                }
                goto done;      /* it's not a regular file */
        }

        if (writable) {
                char *dirname = strdup(filename);

                if (!dirname)
                        goto done;

                stripoff_last_component(dirname);       /* remove filename */

                rc = mkdir(dirname, S_IWUSR|
                                    S_IRUSR|S_IRGRP|S_IROTH|
                                    S_IXUSR|S_IXGRP|S_IXOTH);
                if (rc && errno != EEXIST) {
                        free(dirname);
                        goto done;                      /* probably EACCES */
                }

                *writable = !try_write(filename, dirname);
                free(dirname);
                if (*writable)
                        return 1;
        }
done:
        DBG(UTILS, ul_debug("%s: irregular/non-writable file", filename));
        return 0;
}

/**
 * mnt_get_swaps_path:
 *
 * Returns: path to /proc/swaps or $LIBMOUNT_SWAPS.
 */
const char *mnt_get_swaps_path(void)
{
        const char *p = safe_getenv("LIBMOUNT_SWAPS");
        return p ? : _PATH_PROC_SWAPS;
}

/**
 * mnt_get_fstab_path:
 *
 * Returns: path to /etc/fstab or $LIBMOUNT_FSTAB.
 */
const char *mnt_get_fstab_path(void)
{
        const char *p = safe_getenv("LIBMOUNT_FSTAB");
        return p ? : _PATH_MNTTAB;
}

/**
 * mnt_get_mtab_path:
 *
 * This function returns the *default* location of the mtab file. The result does
 * not have to be writable. See also mnt_has_regular_mtab().
 *
 * Returns: path to /etc/mtab or $LIBMOUNT_MTAB.
 */
const char *mnt_get_mtab_path(void)
{
        const char *p = safe_getenv("LIBMOUNT_MTAB");
        return p ? : _PATH_MOUNTED;
}

/*
 * Don't export this to libmount API -- utab is private library stuff.
 *
 * Returns: path to /run/mount/utab (or /dev/.mount/utab) or $LIBMOUNT_UTAB.
 */
const char *mnt_get_utab_path(void)
{
        struct stat st;
        const char *p = safe_getenv("LIBMOUNT_UTAB");

        if (p)
                return p;

        if (stat(MNT_RUNTIME_TOPDIR, &st) == 0)
                return MNT_PATH_UTAB;

        return MNT_PATH_UTAB_OLD;
}


/* returns file descriptor or -errno, @name returns a unique filename
 */
int mnt_open_uniq_filename(const char *filename, char **name)
{
        int rc, fd;
        char *n;
        mode_t oldmode;

        if (!filename)
                return -EINVAL;
        if (name)
                *name = NULL;

        rc = asprintf(&n, "%s.XXXXXX", filename);
        if (rc <= 0)
                return -errno;

        /* This is for very old glibc and for compatibility with Posix, which says
         * nothing about mkstemp() mode. All sane glibc use secure mode (0600).
         */
        oldmode = umask(S_IRGRP|S_IWGRP|S_IXGRP|
                        S_IROTH|S_IWOTH|S_IXOTH);
        fd = mkostemp(n, O_RDWR|O_CREAT|O_EXCL|O_CLOEXEC);
        if (fd < 0)
                fd = -errno;
        umask(oldmode);

        if (fd >= 0 && name)
                *name = n;
        else
                free(n);

        return fd;
}

/**
 * mnt_get_mountpoint:
 * @path: pathname
 *
 * This function finds the mountpoint that a given path resides in. @path
 * should be canonicalized. The returned pointer should be freed by the caller.
 *
 * WARNING: the function compares st_dev of the @path elements. This traditional
 * way may be insufficient on filesystems like Linux "overlay". See also
 * mnt_table_find_target().
 *
 * Returns: allocated string with the target of the mounted device or NULL on error
 */
char *mnt_get_mountpoint(const char *path)
{
        char *mnt;
        struct stat st;
        dev_t dir, base;

        if (!path)
                return NULL;

        mnt = strdup(path);
        if (!mnt)
                return NULL;
        if (*mnt == '/' && *(mnt + 1) == '\0')
                goto done;

        if (mnt_stat_mountpoint(mnt, &st))
                goto err;
        base = st.st_dev;

        do {
                char *p = stripoff_last_component(mnt);

                if (!p)
                        break;
                if (mnt_stat_mountpoint(*mnt ? mnt : "/", &st))
                        goto err;
                dir = st.st_dev;
                if (dir != base) {
                        if (p > mnt)
                                *(p - 1) = '/';
                        goto done;
                }
                base = dir;
        } while (mnt && *(mnt + 1) != '\0');

        memcpy(mnt, "/", 2);
done:
        DBG(UTILS, ul_debug("%s mountpoint is %s", path, mnt));
        return mnt;
err:
        free(mnt);
        return NULL;
}

/*
 * Search for @name kernel command parameter.
 *
 * Returns newly allocated string with a parameter argument if the @name is
 * specified as "name=" or returns pointer to @name or returns NULL if not
 * found.  If it is specified more than once, we grab the last copy.
 *
 * For example cmdline: "aaa bbb=BBB ccc"
 *
 *      @name is "aaa"  --returns--> "aaa" (pointer to @name)
 *      @name is "bbb=" --returns--> "BBB" (allocated)
 *      @name is "foo"  --returns--> NULL
 *
 * Note: It is not really feasible to parse the command line exactly the same
 * as the kernel does since we don't know which options are valid.  We can use
 * the -- marker though and not walk past that.
 */
char *mnt_get_kernel_cmdline_option(const char *name)
{
        FILE *f;
        size_t len;
        int val = 0;
        char *p, *res = NULL, *mem = NULL;
        char buf[BUFSIZ];       /* see kernel include/asm-generic/setup.h: COMMAND_LINE_SIZE */
        const char *path = _PATH_PROC_CMDLINE;

        if (!name || !name[0])
                return NULL;

#ifdef TEST_PROGRAM
        path = getenv("LIBMOUNT_KERNEL_CMDLINE");
        if (!path)
                path = _PATH_PROC_CMDLINE;
#endif
        f = fopen(path, "r" UL_CLOEXECSTR);
        if (!f)
                return NULL;

        p = fgets(buf, sizeof(buf), f);
        fclose(f);

        if (!p || !*p || *p == '\n')
                return NULL;

        p = strstr(p, " -- ");
        if (p) {
                /* no more kernel args after this */
                *p = '\0';
        } else {
                len = strlen(buf);
                buf[len - 1] = '\0';    /* remove last '\n' */
        }

        len = strlen(name);
        if (name[len - 1] == '=')
                val = 1;

        for (p = buf; p && *p; p++) {
                if (!(p = strstr(p, name)))
                        break;                  /* not found the option */
                if (p != buf && !isblank(*(p - 1)))
                        continue;               /* no space before the option */
                if (!val && *(p + len) != '\0' && !isblank(*(p + len)))
                        continue;               /* no space after the option */
                if (val) {
                        char *v = p + len;
                        int end;

                        while (*p && !isblank(*p))      /* jump to the end of the argument */
                                p++;
                        end = (*p == '\0');
                        *p = '\0';
                        free(mem);
                        res = mem = strdup(v);
                        if (end)
                                break;
                } else
                        res = (char *) name;    /* option without '=' */
                /* don't break -- keep scanning for more options */
        }

        return res;
}

/**
 * mnt_guess_system_root:
 * @devno: device number or zero
 * @cache: paths cache or NULL
 * @path: returns allocated path
 *
 * Converts @devno to the real device name if devno major number is greater
 * than zero, otherwise use root= kernel cmdline option to get device name.
 *
 * The function uses /sys to convert devno to device name.
 *
 * Returns: 0 = success, 1 = not found, <0 = error
 *
 * Since: 2.34
 */
int mnt_guess_system_root(dev_t devno, struct libmnt_cache *cache, char **path)
{
        char buf[PATH_MAX];
        char *dev = NULL, *spec = NULL;
        unsigned int x, y;
        int allocated = 0;

        assert(path);

        DBG(UTILS, ul_debug("guessing system root [devno %u:%u]", major(devno), minor(devno)));

        /* The pseudo-fs, net-fs or btrfs devno is useless, otherwise it
         * usually matches with the source device, let's try to use it.
         */
        if (major(devno) > 0) {
                dev = sysfs_devno_to_devpath(devno, buf, sizeof(buf));
                if (dev) {
                        DBG(UTILS, ul_debug("  devno converted to %s", dev));
                        goto done;
                }
        }

        /* Let's try to use root= kernel command line option
         */
        spec = mnt_get_kernel_cmdline_option("root=");
        if (!spec)
                goto done;

        /* maj:min notation */
        if (sscanf(spec, "%u:%u", &x, &y) == 2) {
                dev = sysfs_devno_to_devpath(makedev(x, y), buf, sizeof(buf));
                if (dev) {
                        DBG(UTILS, ul_debug("  root=%s converted to %s", spec, dev));
                        goto done;
                }

        /* hexhex notation */
        } else if (isxdigit_string(spec)) {
                char *end = NULL;
                uint32_t n;

                errno = 0;
                n = strtoul(spec, &end, 16);

                if (errno || spec == end || (end && *end))
                        DBG(UTILS, ul_debug("  failed to parse root='%s'", spec));
                else {
                        /* kernel new_decode_dev() */
                        x = (n & 0xfff00) >> 8;
                        y = (n & 0xff) | ((n >> 12) & 0xfff00);
                        dev = sysfs_devno_to_devpath(makedev(x, y), buf, sizeof(buf));
                        if (dev) {
                                DBG(UTILS, ul_debug("  root=%s converted to %s", spec, dev));
                                goto done;
                        }
                }

        /* devname or PARTUUID= etc. */
        } else {
                DBG(UTILS, ul_debug("  converting root='%s'", spec));

                dev = mnt_resolve_spec(spec, cache);
                if (dev && !cache)
                        allocated = 1;
        }
done:
        free(spec);
        if (dev) {
                *path = allocated ? dev : strdup(dev);
                if (!*path)
                        return -ENOMEM;
                return 0;
        }

        return 1;
}

#if defined(HAVE_FMEMOPEN) || defined(TEST_PROGRAM)

/*
 * This function tries to minimize possible races when we read
 * /proc/#/{mountinfo,mount} files.
 *
 * The idea is to minimize number of read()s and check by poll() that during
 * the read the mount table has not been modified. If yes, than re-read it
 * (with some limitations to avoid never ending loop).
 *
 * Returns: <0 error, 0 success, 1 too many attempts
 */
static int read_procfs_file(int fd, char **buf, size_t *bufsiz)
{
        size_t bufmax = 0;
        int rc = 0, tries = 0, ninters = 0;
        char *bufptr = NULL;

        assert(buf);
        assert(bufsiz);

        *bufsiz = 0;
        *buf = NULL;

        do {
                ssize_t ret;

                if (!bufptr || bufmax == *bufsiz) {
                        char *tmp;

                        bufmax = bufmax ? bufmax * 2 : (16 * 1024);
                        tmp = realloc(*buf, bufmax);
                        if (!tmp)
                                break;
                        *buf = tmp;
                        bufptr = tmp + *bufsiz;
                }

                errno = 0;
                ret = read(fd, bufptr, bufmax - *bufsiz);

                if (ret < 0) {
                        /* error */
                        if ((errno == EAGAIN || errno == EINTR) && (ninters++ < 5)) {
                                xusleep(200000);
                                continue;
                        }
                        break;

                } if (ret > 0) {
                        /* success -- verify no event during read */
                        struct pollfd fds[] = {
                                { .fd = fd, .events = POLLPRI }
                        };

                        rc = poll(fds, 1, 0);
                        if (rc < 0)
                                break;          /* poll() error */
                        if (rc > 0) {
                                /* event -- read all again */
                                if (lseek(fd, 0, SEEK_SET) != 0)
                                        break;
                                *bufsiz = 0;
                                bufptr = *buf;
                                tries++;

                                if (tries > 10)
                                        /* busy system? -- wait */
                                        xusleep(10000);
                                continue;
                        }

                        /* successful read() without active poll() */
                        (*bufsiz) += (size_t) ret;
                        bufptr += ret;
                        tries = ninters = 0;
                } else {
                        /* end-of-file */
                        goto success;
                }
        } while (tries <= 100);

        rc = errno ? -errno : 1;
        free(*buf);
        return rc;

success:
        return 0;
}

/*
 * Create FILE stream for data from read_procfs_file()
 */
FILE *mnt_get_procfs_memstream(int fd, char **membuf)
{
        size_t sz = 0;
        off_t cur;

        *membuf = NULL;

        /* in case of error, rewind to the original position */
        cur = lseek(fd, 0, SEEK_CUR);

        if (read_procfs_file(fd, membuf, &sz) == 0 && sz > 0) {
                FILE *memf = fmemopen(*membuf, sz, "r");
                if (memf)
                        return memf;    /* success */

                free(*membuf);
                *membuf = NULL;
        }

        /* error */
        if (cur != (off_t) -1)
                lseek(fd, cur, SEEK_SET);
        return NULL;
}
#else
FILE *mnt_get_procfs_memstream(int fd __attribute((__unused__)),
                               char **membuf __attribute((__unused__)))
{
        return NULL;
}
#endif /* HAVE_FMEMOPEN */


#ifdef TEST_PROGRAM
static int test_proc_read(struct libmnt_test *ts, int argc, char *argv[])
{
        char *buf = NULL;
        char *filename = argv[1];
        size_t bufsiz = 0;
        int rc = 0, fd = open(filename, O_RDONLY);

        if (fd <= 0) {
                warn("%s: cannot open", filename);
                return -errno;
        }

        rc = read_procfs_file(fd, &buf, &bufsiz);
        close(fd);

        switch (rc) {
        case 0:
                fwrite(buf, 1, bufsiz, stdout);
                free(buf);
                break;
        case 1:
                warnx("too many attempts");
                break;
        default:
                warn("%s: cannot read", filename);
                break;
        }

        return rc;
}

static int test_match_fstype(struct libmnt_test *ts, int argc, char *argv[])
{
        char *type = argv[1];
        char *pattern = argv[2];

        printf("%s\n", mnt_match_fstype(type, pattern) ? "MATCH" : "NOT-MATCH");
        return 0;
}

static int test_match_options(struct libmnt_test *ts, int argc, char *argv[])
{
        char *optstr = argv[1];
        char *pattern = argv[2];

        printf("%s\n", mnt_match_options(optstr, pattern) ? "MATCH" : "NOT-MATCH");
        return 0;
}

static int test_startswith(struct libmnt_test *ts, int argc, char *argv[])
{
        char *optstr = argv[1];
        char *pattern = argv[2];

        printf("%s\n", startswith(optstr, pattern) ? "YES" : "NOT");
        return 0;
}

static int test_endswith(struct libmnt_test *ts, int argc, char *argv[])
{
        char *optstr = argv[1];
        char *pattern = argv[2];

        printf("%s\n", endswith(optstr, pattern) ? "YES" : "NOT");
        return 0;
}

static int test_appendstr(struct libmnt_test *ts, int argc, char *argv[])
{
        char *str = strdup(argv[1]);
        const char *ap = argv[2];

        append_string(&str, ap);
        printf("new string: '%s'\n", str);

        free(str);
        return 0;
}

static int test_mountpoint(struct libmnt_test *ts, int argc, char *argv[])
{
        char *path = canonicalize_path(argv[1]),
             *mnt = path ? mnt_get_mountpoint(path) :  NULL;

        printf("%s: %s\n", argv[1], mnt ? : "unknown");
        free(mnt);
        free(path);
        return 0;
}

static int test_filesystems(struct libmnt_test *ts, int argc, char *argv[])
{
        char **filesystems = NULL;
        int rc;

        rc = mnt_get_filesystems(&filesystems, argc ? argv[1] : NULL);
        if (!rc) {
                char **p;
                for (p = filesystems; *p; p++)
                        printf("%s\n", *p);
                mnt_free_filesystems(filesystems);
        }
        return rc;
}

static int test_chdir(struct libmnt_test *ts, int argc, char *argv[])
{
        int rc;
        char *path = canonicalize_path(argv[1]),
             *last = NULL;

        if (!path)
                return -errno;

        rc = mnt_chdir_to_parent(path, &last);
        if (!rc) {
                printf("path='%s', abs='%s', last='%s'\n",
                                argv[1], path, last);
        }
        free(path);
        free(last);
        return rc;
}

static int test_kernel_cmdline(struct libmnt_test *ts, int argc, char *argv[])
{
        char *name = argv[1];
        char *res;

        res = mnt_get_kernel_cmdline_option(name);
        if (!res)
                printf("'%s' not found\n", name);
        else if (res == name)
                printf("'%s' found\n", name);
        else {
                printf("'%s' found, argument: '%s'\n", name, res);
                free(res);
        }

        return 0;
}


static int test_guess_root(struct libmnt_test *ts, int argc, char *argv[])
{
        int rc;
        char *real;
        dev_t devno = 0;

        if (argc) {
                unsigned int x, y;

                if (sscanf(argv[1], "%u:%u", &x, &y) != 2)
                        return -EINVAL;
                devno = makedev(x, y);
        }

        rc = mnt_guess_system_root(devno, NULL, &real);
        if (rc < 0)
                return rc;
        if (rc == 1)
                fputs("not found\n", stdout);
        else {
                printf("%s\n", real);
                free(real);
        }
        return 0;
}

static int test_mkdir(struct libmnt_test *ts, int argc, char *argv[])
{
        int rc;

        rc = mkdir_p(argv[1], S_IRWXU |
                         S_IRGRP | S_IXGRP |
                         S_IROTH | S_IXOTH);
        if (rc)
                printf("mkdir %s failed\n", argv[1]);
        return rc;
}

static int test_statfs_type(struct libmnt_test *ts, int argc, char *argv[])
{
        struct statfs vfs;
        int rc;

        rc = statfs(argv[1], &vfs);
        if (rc)
                printf("%s: statfs failed: %m\n", argv[1]);
        else
                printf("%-30s: statfs type: %-12s [0x%lx]\n", argv[1],
                                mnt_statfs_get_fstype(&vfs),
                                (long) vfs.f_type);
        return rc;
}


int main(int argc, char *argv[])
{
        struct libmnt_test tss[] = {
        { "--match-fstype",  test_match_fstype,    "<type> <pattern>     FS types matching" },
        { "--match-options", test_match_options,   "<options> <pattern>  options matching" },
        { "--filesystems",   test_filesystems,     "[<pattern>] list /{etc,proc}/filesystems" },
        { "--starts-with",   test_startswith,      "<string> <prefix>" },
        { "--ends-with",     test_endswith,        "<string> <prefix>" },
        { "--append-string", test_appendstr,       "<string> <appendix>" },
        { "--mountpoint",    test_mountpoint,      "<path>" },
        { "--cd-parent",     test_chdir,           "<path>" },
        { "--kernel-cmdline",test_kernel_cmdline,  "<option> | <option>=" },
        { "--guess-root",    test_guess_root,      "[<maj:min>]" },
        { "--mkdir",         test_mkdir,           "<path>" },
        { "--statfs-type",   test_statfs_type,     "<path>" },
        { "--read-procfs",   test_proc_read,       "<path>" },

        { NULL }
        };

        return mnt_run_test(tss, argc, argv);
}

#endif /* TEST_PROGRAM */