summaryrefslogtreecommitdiffstats
path: root/src/filemanager/mountlist.c
diff options
context:
space:
mode:
Diffstat (limited to 'src/filemanager/mountlist.c')
-rw-r--r--src/filemanager/mountlist.c1572
1 files changed, 1572 insertions, 0 deletions
diff --git a/src/filemanager/mountlist.c b/src/filemanager/mountlist.c
new file mode 100644
index 0000000..daa7b80
--- /dev/null
+++ b/src/filemanager/mountlist.c
@@ -0,0 +1,1572 @@
+/*
+ Return a list of mounted file systems
+
+ Copyright (C) 1991-2022
+ Free Software Foundation, Inc.
+
+ This file is part of the Midnight Commander.
+
+ The Midnight Commander is free software: you can redistribute it
+ and/or modify it under the terms of the GNU General Public License as
+ published by the Free Software Foundation, either version 3 of the License,
+ or (at your option) any later version.
+
+ The Midnight Commander 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 General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+/** \file mountlist.c
+ * \brief Source: list of mounted filesystems
+ */
+
+#include <config.h>
+
+#include <limits.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdint.h> /* SIZE_MAX */
+#include <sys/types.h>
+
+#include <errno.h>
+
+/* This header needs to be included before sys/mount.h on *BSD */
+#ifdef HAVE_SYS_PARAM_H
+#include <sys/param.h>
+#endif
+
+#if defined STAT_STATVFS || defined STAT_STATVFS64 /* POSIX 1003.1-2001 (and later) with XSI */
+#include <sys/statvfs.h>
+#else
+/* Don't include backward-compatibility files unless they're needed.
+ Eventually we'd like to remove all this cruft. */
+#include <fcntl.h>
+#include <unistd.h>
+#include <sys/stat.h>
+
+#ifdef MOUNTED_GETFSSTAT /* OSF_1, also (obsolete) Apple Darwin 1.3 */
+#ifdef HAVE_SYS_UCRED_H
+#include <grp.h> /* needed on OSF V4.0 for definition of NGROUPS,
+ NGROUPS is used as an array dimension in ucred.h */
+#include <sys/ucred.h> /* needed by powerpc-apple-darwin1.3.7 */
+#endif
+#ifdef HAVE_SYS_MOUNT_H
+#include <sys/mount.h>
+#endif
+#ifdef HAVE_SYS_FS_TYPES_H
+#include <sys/fs_types.h> /* needed by powerpc-apple-darwin1.3.7 */
+#endif
+#ifdef HAVE_STRUCT_FSSTAT_F_FSTYPENAME
+#define FS_TYPE(Ent) ((Ent).f_fstypename)
+#else
+#define FS_TYPE(Ent) mnt_names[(Ent).f_type]
+#endif
+#endif /* MOUNTED_GETFSSTAT */
+#endif /* STAT_STATVFS || STAT_STATVFS64 */
+
+#ifdef HAVE_SYS_VFS_H
+#include <sys/vfs.h>
+#endif
+#ifdef HAVE_SYS_FS_S5PARAM_H /* Fujitsu UXP/V */
+#include <sys/fs/s5param.h>
+#endif
+#ifdef HAVE_SYS_STATFS_H
+#include <sys/statfs.h>
+#endif
+
+#ifdef MOUNTED_GETMNTENT1 /* glibc, HP-UX, IRIX, Cygwin, Android,
+ also (obsolete) 4.3BSD, SunOS */
+#include <mntent.h>
+#include <sys/types.h>
+#if defined __ANDROID__ /* Android */
+ /* Bionic versions from between 2014-01-09 and 2015-01-08 define MOUNTED to
+ an incorrect value; older Bionic versions don't define it at all. */
+#undef MOUNTED
+#define MOUNTED "/proc/mounts"
+#elif !defined MOUNTED
+#ifdef _PATH_MOUNTED /* GNU libc */
+#define MOUNTED _PATH_MOUNTED
+#endif
+#ifdef MNT_MNTTAB /* HP-UX. */
+#define MOUNTED MNT_MNTTAB
+#endif
+#endif
+#endif
+
+#ifdef MOUNTED_GETMNTINFO /* Mac OS X, FreeBSD, OpenBSD, also (obsolete) 4.4BSD */
+#include <sys/mount.h>
+#endif
+
+#ifdef MOUNTED_GETMNTINFO2 /* NetBSD, Minix */
+#include <sys/statvfs.h>
+#endif
+
+#ifdef MOUNTED_FS_STAT_DEV /* Haiku, also (obsolete) BeOS */
+#include <fs_info.h>
+#include <dirent.h>
+#endif
+
+#ifdef MOUNTED_FREAD_FSTYP /* (obsolete) SVR3 */
+#include <mnttab.h>
+#include <sys/fstyp.h>
+#include <sys/statfs.h>
+#endif
+
+#ifdef MOUNTED_GETEXTMNTENT /* Solaris >= 8 */
+#include <sys/mnttab.h>
+#endif
+
+#ifdef MOUNTED_GETMNTENT2 /* Solaris < 8, also (obsolete) SVR4 */
+#include <sys/mnttab.h>
+#endif
+
+#ifdef MOUNTED_VMOUNT /* AIX */
+#include <fshelp.h>
+#include <sys/vfs.h>
+#endif
+
+#ifdef MOUNTED_INTERIX_STATVFS /* Interix */
+#include <sys/statvfs.h>
+#include <dirent.h>
+#endif
+
+#ifdef HAVE_SYS_MNTENT_H
+/* This is to get MNTOPT_IGNORE on e.g. SVR4. */
+#include <sys/mntent.h>
+#endif
+
+#ifdef MOUNTED_GETMNTENT1
+#if !HAVE_SETMNTENT /* Android <= 4.4 */
+#define setmntent(fp,mode) fopen (fp, mode)
+#endif
+#if !HAVE_ENDMNTENT /* Android <= 4.4 */
+#define endmntent(fp) fclose (fp)
+#endif
+#endif
+
+#ifndef HAVE_HASMNTOPT
+#define hasmntopt(mnt, opt) ((char *) 0)
+#endif
+
+#undef MNT_IGNORE
+#ifdef MNTOPT_IGNORE
+#if defined __sun && defined __SVR4
+/* Solaris defines hasmntopt(struct mnttab *, char *)
+ while it is otherwise hasmntopt(struct mnttab *, const char *). */
+#define MNT_IGNORE(M) hasmntopt (M, (char *) MNTOPT_IGNORE)
+#else
+#define MNT_IGNORE(M) hasmntopt (M, MNTOPT_IGNORE)
+#endif
+#else
+#define MNT_IGNORE(M) 0
+#endif
+
+#ifdef HAVE_INFOMOUNT_QNX
+#include <sys/disk.h>
+#include <sys/fsys.h>
+#endif
+
+#ifdef HAVE_SYS_STATVFS_H /* SVR4. */
+#include <sys/statvfs.h>
+#endif
+
+#include "lib/global.h"
+#include "lib/strutil.h" /* str_verscmp() */
+#include "lib/unixcompat.h" /* makedev */
+#include "mountlist.h"
+
+/*** global variables ****************************************************************************/
+
+/*** file scope macro definitions ****************************************************************/
+
+#if defined (__QNX__) && !defined(__QNXNTO__) && !defined (HAVE_INFOMOUNT_LIST)
+#define HAVE_INFOMOUNT_QNX
+#endif
+
+#if defined(HAVE_INFOMOUNT_LIST) || defined(HAVE_INFOMOUNT_QNX)
+#define HAVE_INFOMOUNT
+#endif
+
+/* The results of opendir() in this file are not used with dirfd and fchdir,
+ therefore save some unnecessary work in fchdir.c. */
+#undef opendir
+#undef closedir
+
+#define ME_DUMMY_0(Fs_name, Fs_type) \
+ (strcmp (Fs_type, "autofs") == 0 \
+ || strcmp (Fs_type, "proc") == 0 \
+ || strcmp (Fs_type, "subfs") == 0 \
+ /* for Linux 2.6/3.x */ \
+ || strcmp (Fs_type, "debugfs") == 0 \
+ || strcmp (Fs_type, "devpts") == 0 \
+ || strcmp (Fs_type, "fusectl") == 0 \
+ || strcmp (Fs_type, "fuse.portal") == 0 \
+ || strcmp (Fs_type, "mqueue") == 0 \
+ || strcmp (Fs_type, "rpc_pipefs") == 0 \
+ || strcmp (Fs_type, "sysfs") == 0 \
+ /* FreeBSD, Linux 2.4 */ \
+ || strcmp (Fs_type, "devfs") == 0 \
+ /* for NetBSD 3.0 */ \
+ || strcmp (Fs_type, "kernfs") == 0 \
+ /* for Irix 6.5 */ \
+ || strcmp (Fs_type, "ignore") == 0)
+
+/* Historically, we have marked as "dummy" any file system of type "none",
+ but now that programs like du need to know about bind-mounted directories,
+ we grant an exception to any with "bind" in its list of mount options.
+ I.e., those are *not* dummy entries. */
+#ifdef MOUNTED_GETMNTENT1
+#define ME_DUMMY(Fs_name, Fs_type, Bind) \
+ (ME_DUMMY_0 (Fs_name, Fs_type) \
+ || (strcmp (Fs_type, "none") == 0 && !Bind))
+#else
+#define ME_DUMMY(Fs_name, Fs_type) \
+ (ME_DUMMY_0 (Fs_name, Fs_type) || strcmp (Fs_type, "none") == 0)
+#endif
+
+#ifdef __CYGWIN__
+#include <windows.h>
+#define ME_REMOTE me_remote
+/* All cygwin mount points include ':' or start with '//'; so it
+ requires a native Windows call to determine remote disks. */
+static int
+me_remote (char const *fs_name, char const *fs_type)
+{
+ (void) fs_type;
+
+ if (fs_name[0] && fs_name[1] == ':')
+ {
+ char drive[4];
+ sprintf (drive, "%c:\\", fs_name[0]);
+ switch (GetDriveType (drive))
+ {
+ case DRIVE_REMOVABLE:
+ case DRIVE_FIXED:
+ case DRIVE_CDROM:
+ case DRIVE_RAMDISK:
+ return 0;
+ }
+ }
+ return 1;
+}
+#endif
+#ifndef ME_REMOTE
+/* A file system is 'remote' if its Fs_name contains a ':'
+ or if (it is of type (smbfs or smb3 or cifs) and its Fs_name starts with '//')
+ or if it is of any other of the listed types
+ or Fs_name is equal to "-hosts" (used by autofs to mount remote fs).
+ "VM" file systems like prl_fs or vboxsf are not considered remote here. */
+#define ME_REMOTE(Fs_name, Fs_type) \
+ (strchr (Fs_name, ':') != NULL \
+ || ((Fs_name)[0] == '/' \
+ && (Fs_name)[1] == '/' \
+ && (strcmp (Fs_type, "smbfs") == 0 \
+ || strcmp (Fs_type, "smb3") == 0 \
+ || strcmp (Fs_type, "cifs") == 0)) \
+ || strcmp (Fs_type, "acfs") == 0 \
+ || strcmp (Fs_type, "afs") == 0 \
+ || strcmp (Fs_type, "coda") == 0 \
+ || strcmp (Fs_type, "auristorfs") == 0 \
+ || strcmp (Fs_type, "fhgfs") == 0 \
+ || strcmp (Fs_type, "gpfs") == 0 \
+ || strcmp (Fs_type, "ibrix") == 0 \
+ || strcmp (Fs_type, "ocfs2") == 0 \
+ || strcmp (Fs_type, "vxfs") == 0 \
+ || strcmp ("-hosts", Fs_name) == 0)
+#endif
+
+/* Many space usage primitives use all 1 bits to denote a value that is
+ not applicable or unknown. Propagate this information by returning
+ a uintmax_t value that is all 1 bits if X is all 1 bits, even if X
+ is unsigned and narrower than uintmax_t. */
+#define PROPAGATE_ALL_ONES(x) \
+ ((sizeof (x) < sizeof (uintmax_t) \
+ && (~ (x) == (sizeof (x) < sizeof (int) \
+ ? - (1 << (sizeof (x) * CHAR_BIT)) \
+ : 0))) \
+ ? UINTMAX_MAX : (uintmax_t) (x))
+
+/* Extract the top bit of X as an uintmax_t value. */
+#define EXTRACT_TOP_BIT(x) ((x) & ((uintmax_t) 1 << (sizeof (x) * CHAR_BIT - 1)))
+
+/* If a value is negative, many space usage primitives store it into an
+ integer variable by assignment, even if the variable's type is unsigned.
+ So, if a space usage variable X's top bit is set, convert X to the
+ uintmax_t value V such that (- (uintmax_t) V) is the negative of
+ the original value. If X's top bit is clear, just yield X.
+ Use PROPAGATE_TOP_BIT if the original value might be negative;
+ otherwise, use PROPAGATE_ALL_ONES. */
+#define PROPAGATE_TOP_BIT(x) ((x) | ~ (EXTRACT_TOP_BIT (x) - 1))
+
+#ifdef STAT_STATVFS
+#if ! (defined __linux__ && (defined __GLIBC__ || defined __UCLIBC__))
+/* The FRSIZE fallback is not required in this case. */
+#undef STAT_STATFS2_FRSIZE
+#else
+#include <sys/utsname.h>
+#include <sys/statfs.h>
+#define STAT_STATFS2_BSIZE 1
+#endif
+#endif
+
+/*** file scope type declarations ****************************************************************/
+
+/* A mount table entry. */
+struct mount_entry
+{
+ char *me_devname; /* Device node name, including "/dev/". */
+ char *me_mountdir; /* Mount point directory name. */
+ char *me_mntroot; /* Directory on filesystem of device used
+ as root for the (bind) mount. */
+ char *me_type; /* "nfs", "4.2", etc. */
+ dev_t me_dev; /* Device number of me_mountdir. */
+ unsigned int me_dummy:1; /* Nonzero for dummy file systems. */
+ unsigned int me_remote:1; /* Nonzero for remote fileystems. */
+ unsigned int me_type_malloced:1; /* Nonzero if me_type was malloced. */
+};
+
+struct fs_usage
+{
+ uintmax_t fsu_blocksize; /* Size of a block. */
+ uintmax_t fsu_blocks; /* Total blocks. */
+ uintmax_t fsu_bfree; /* Free blocks available to superuser. */
+ uintmax_t fsu_bavail; /* Free blocks available to non-superuser. */
+ int fsu_bavail_top_bit_set; /* 1 if fsu_bavail represents a value < 0. */
+ uintmax_t fsu_files; /* Total file nodes. */
+ uintmax_t fsu_ffree; /* Free file nodes. */
+};
+
+/*** file scope variables ************************************************************************/
+
+#ifdef HAVE_INFOMOUNT_LIST
+static GSList *mc_mount_list = NULL;
+#endif /* HAVE_INFOMOUNT_LIST */
+
+/*** file scope functions ************************************************************************/
+/* --------------------------------------------------------------------------------------------- */
+
+#ifdef STAT_STATVFS
+/* Return true if statvfs works. This is false for statvfs on systems
+ with GNU libc on Linux kernels before 2.6.36, which stats all
+ preceding entries in /proc/mounts; that makes df hang if even one
+ of the corresponding file systems is hard-mounted but not available. */
+static int
+statvfs_works (void)
+{
+#if ! (defined __linux__ && (defined __GLIBC__ || defined __UCLIBC__))
+ return 1;
+#else
+ static int statvfs_works_cache = -1;
+ struct utsname name;
+
+ if (statvfs_works_cache < 0)
+ statvfs_works_cache = (uname (&name) == 0 && 0 <= str_verscmp (name.release, "2.6.36"));
+ return statvfs_works_cache;
+#endif
+}
+#endif
+
+/* --------------------------------------------------------------------------------------------- */
+
+#ifdef HAVE_INFOMOUNT_LIST
+static void
+free_mount_entry (struct mount_entry *me)
+{
+ if (me == NULL)
+ return;
+ g_free (me->me_devname);
+ g_free (me->me_mountdir);
+ g_free (me->me_mntroot);
+ if (me->me_type_malloced)
+ g_free (me->me_type);
+ g_free (me);
+}
+
+/* --------------------------------------------------------------------------------------------- */
+
+#ifdef MOUNTED_GETMNTINFO /* Mac OS X, FreeBSD, OpenBSD, also (obsolete) 4.4BSD */
+
+#ifndef HAVE_STRUCT_STATFS_F_FSTYPENAME
+static char *
+fstype_to_string (short int t)
+{
+ switch (t)
+ {
+#ifdef MOUNT_PC
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_PC:
+ return "pc";
+#endif
+#ifdef MOUNT_MFS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_MFS:
+ return "mfs";
+#endif
+#ifdef MOUNT_LO
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_LO:
+ return "lo";
+#endif
+#ifdef MOUNT_TFS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_TFS:
+ return "tfs";
+#endif
+#ifdef MOUNT_TMP
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_TMP:
+ return "tmp";
+#endif
+#ifdef MOUNT_UFS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_UFS:
+ return "ufs";
+#endif
+#ifdef MOUNT_NFS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_NFS:
+ return "nfs";
+#endif
+#ifdef MOUNT_MSDOS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_MSDOS:
+ return "msdos";
+#endif
+#ifdef MOUNT_LFS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_LFS:
+ return "lfs";
+#endif
+#ifdef MOUNT_LOFS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_LOFS:
+ return "lofs";
+#endif
+#ifdef MOUNT_FDESC
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_FDESC:
+ return "fdesc";
+#endif
+#ifdef MOUNT_PORTAL
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_PORTAL:
+ return "portal";
+#endif
+#ifdef MOUNT_NULL
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_NULL:
+ return "null";
+#endif
+#ifdef MOUNT_UMAP
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_UMAP:
+ return "umap";
+#endif
+#ifdef MOUNT_KERNFS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_KERNFS:
+ return "kernfs";
+#endif
+#ifdef MOUNT_PROCFS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_PROCFS:
+ return "procfs";
+#endif
+#ifdef MOUNT_AFS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_AFS:
+ return "afs";
+#endif
+#ifdef MOUNT_CD9660
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_CD9660:
+ return "cd9660";
+#endif
+#ifdef MOUNT_UNION
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_UNION:
+ return "union";
+#endif
+#ifdef MOUNT_DEVFS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_DEVFS:
+ return "devfs";
+#endif
+#ifdef MOUNT_EXT2FS
+ /* cppcheck-suppress syntaxError */
+ case MOUNT_EXT2FS:
+ return "ext2fs";
+#endif
+ default:
+ return "?";
+ }
+}
+#endif /* ! HAVE_STRUCT_STATFS_F_FSTYPENAME */
+
+/* --------------------------------------------------------------------------------------------- */
+
+static char *
+fsp_to_string (const struct statfs *fsp)
+{
+#ifdef HAVE_STRUCT_STATFS_F_FSTYPENAME
+ return (char *) (fsp->f_fstypename);
+#else
+ return fstype_to_string (fsp->f_type);
+#endif
+}
+#endif /* MOUNTED_GETMNTINFO */
+
+/* --------------------------------------------------------------------------------------------- */
+
+#ifdef MOUNTED_VMOUNT /* AIX */
+static char *
+fstype_to_string (int t)
+{
+ struct vfs_ent *e;
+
+ e = getvfsbytype (t);
+ if (!e || !e->vfsent_name)
+ return "none";
+ else
+ return e->vfsent_name;
+}
+#endif /* MOUNTED_VMOUNT */
+
+/* --------------------------------------------------------------------------------------------- */
+
+#if defined MOUNTED_GETMNTENT1 || defined MOUNTED_GETMNTENT2
+
+/* Return the device number from MOUNT_OPTIONS, if possible.
+ Otherwise return (dev_t) -1. */
+
+/* --------------------------------------------------------------------------------------------- */
+
+static dev_t
+dev_from_mount_options (char const *mount_options)
+{
+ /* GNU/Linux allows file system implementations to define their own
+ meaning for "dev=" mount options, so don't trust the meaning
+ here. */
+#ifndef __linux__
+ static char const dev_pattern[] = ",dev=";
+ char const *devopt = strstr (mount_options, dev_pattern);
+
+ if (devopt)
+ {
+ char const *optval = devopt + sizeof (dev_pattern) - 1;
+ char *optvalend;
+ unsigned long int dev;
+ errno = 0;
+ dev = strtoul (optval, &optvalend, 16);
+ if (optval != optvalend
+ && (*optvalend == '\0' || *optvalend == ',')
+ && !(dev == ULONG_MAX && errno == ERANGE) && dev == (dev_t) dev)
+ return dev;
+ }
+#endif
+
+ (void) mount_options;
+ return -1;
+}
+
+#endif
+
+/* --------------------------------------------------------------------------------------------- */
+
+#if defined MOUNTED_GETMNTENT1 && (defined __linux__ || defined __ANDROID__) /* GNU/Linux, Android */
+
+/* Unescape the paths in mount tables.
+ STR is updated in place. */
+static void
+unescape_tab (char *str)
+{
+ size_t i, j = 0;
+ size_t len;
+
+ len = strlen (str) + 1;
+
+ for (i = 0; i < len; i++)
+ {
+ if (str[i] == '\\' && (i + 4 < len)
+ && str[i + 1] >= '0' && str[i + 1] <= '3'
+ && str[i + 2] >= '0' && str[i + 2] <= '7' && str[i + 3] >= '0' && str[i + 3] <= '7')
+ {
+ str[j++] = (str[i + 1] - '0') * 64 + (str[i + 2] - '0') * 8 + (str[i + 3] - '0');
+ i += 3;
+ }
+ else
+ str[j++] = str[i];
+ }
+}
+#endif
+
+/* --------------------------------------------------------------------------------------------- */
+
+/* Return a list of the currently mounted file systems, or NULL on error.
+ Add each entry to the tail of the list so that they stay in order. */
+
+static GSList *
+read_file_system_list (void)
+{
+ GSList *mount_list = NULL;
+ struct mount_entry *me;
+
+#ifdef MOUNTED_GETMNTENT1 /* glibc, HP-UX, IRIX, Cygwin, Android,
+ also (obsolete) 4.3BSD, SunOS */
+ {
+ FILE *fp;
+
+#if defined __linux__ || defined __ANDROID__
+ /* Try parsing mountinfo first, as that make device IDs available.
+ Note we could use libmount routines to simplify this parsing a little
+ (and that code is in previous versions of this function), however
+ libmount depends on libselinux which pulls in many dependencies. */
+ char const *mountinfo = "/proc/self/mountinfo";
+
+ fp = fopen (mountinfo, "r");
+ if (fp != NULL)
+ {
+ char *line = NULL;
+ size_t buf_size = 0;
+
+ while (getline (&line, &buf_size, fp) != -1)
+ {
+ unsigned int devmaj, devmin;
+ int target_s, target_e, type_s, type_e;
+ int source_s, source_e, mntroot_s, mntroot_e;
+ char test;
+ char *dash;
+ int rc;
+
+ rc = sscanf (line, "%*u " /* id - discarded */
+ "%*u " /* parent - discarded */
+ "%u:%u " /* dev major:minor */
+ "%n%*s%n " /* mountroot */
+ "%n%*s%n" /* target, start and end */
+ "%c", /* more data... */
+ &devmaj, &devmin, &mntroot_s, &mntroot_e, &target_s, &target_e, &test);
+
+ if (rc != 3 && rc != 7) /* 7 if %n included in count. */
+ continue;
+
+ /* skip optional fields, terminated by " - " */
+ dash = strstr (line + target_e, " - ");
+ if (dash == NULL)
+ continue;
+
+ rc = sscanf (dash, " - " /* */
+ "%n%*s%n " /* FS type, start and end */
+ "%n%*s%n " /* source, start and end */
+ "%c", /* more data... */
+ &type_s, &type_e, &source_s, &source_e, &test);
+ if (rc != 1 && rc != 5) /* 5 if %n included in count. */
+ continue;
+
+ /* manipulate the sub-strings in place. */
+ line[mntroot_e] = '\0';
+ line[target_e] = '\0';
+ dash[type_e] = '\0';
+ dash[source_e] = '\0';
+ unescape_tab (dash + source_s);
+ unescape_tab (line + target_s);
+ unescape_tab (line + mntroot_s);
+
+ me = g_malloc (sizeof *me);
+
+ me->me_devname = g_strdup (dash + source_s);
+ me->me_mountdir = g_strdup (line + target_s);
+ me->me_mntroot = g_strdup (line + mntroot_s);
+ me->me_type = g_strdup (dash + type_s);
+ me->me_type_malloced = 1;
+ me->me_dev = makedev (devmaj, devmin);
+ /* we pass "false" for the "Bind" option as that's only
+ significant when the Fs_type is "none" which will not be
+ the case when parsing "/proc/self/mountinfo", and only
+ applies for static /etc/mtab files. */
+ me->me_dummy = ME_DUMMY (me->me_devname, me->me_type, FALSE);
+ me->me_remote = ME_REMOTE (me->me_devname, me->me_type);
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+
+ free (line);
+
+ if (ferror (fp) != 0)
+ {
+ int saved_errno = errno;
+
+ fclose (fp);
+ errno = saved_errno;
+ goto free_then_fail;
+ }
+
+ if (fclose (fp) == EOF)
+ goto free_then_fail;
+ }
+ else /* fallback to /proc/self/mounts (/etc/mtab). */
+#endif /* __linux __ || __ANDROID__ */
+ {
+ struct mntent *mnt;
+ const char *table = MOUNTED;
+
+ fp = setmntent (table, "r");
+ if (fp == NULL)
+ return NULL;
+
+ while ((mnt = getmntent (fp)) != NULL)
+ {
+ gboolean bind;
+
+ bind = hasmntopt (mnt, "bind") != NULL;
+
+ me = g_malloc (sizeof (*me));
+ me->me_devname = g_strdup (mnt->mnt_fsname);
+ me->me_mountdir = g_strdup (mnt->mnt_dir);
+ me->me_mntroot = NULL;
+ me->me_type = g_strdup (mnt->mnt_type);
+ me->me_type_malloced = 1;
+ me->me_dummy = ME_DUMMY (me->me_devname, me->me_type, bind);
+ me->me_remote = ME_REMOTE (me->me_devname, me->me_type);
+ me->me_dev = dev_from_mount_options (mnt->mnt_opts);
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+
+ if (endmntent (fp) == 0)
+ goto free_then_fail;
+ }
+ }
+#endif /* MOUNTED_GETMNTENT1. */
+
+#ifdef MOUNTED_GETMNTINFO /* Mac OS X, FreeBSD, OpenBSD, also (obsolete) 4.4BSD */
+ {
+ struct statfs *fsp;
+ int entries;
+
+ entries = getmntinfo (&fsp, MNT_NOWAIT);
+ if (entries < 0)
+ return NULL;
+ for (; entries-- > 0; fsp++)
+ {
+ char *fs_type = fsp_to_string (fsp);
+
+ me = g_malloc (sizeof (*me));
+ me->me_devname = g_strdup (fsp->f_mntfromname);
+ me->me_mountdir = g_strdup (fsp->f_mntonname);
+ me->me_mntroot = NULL;
+ me->me_type = fs_type;
+ me->me_type_malloced = 0;
+ me->me_dummy = ME_DUMMY (me->me_devname, me->me_type);
+ me->me_remote = ME_REMOTE (me->me_devname, me->me_type);
+ me->me_dev = (dev_t) (-1); /* Magic; means not known yet. */
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+ }
+#endif /* MOUNTED_GETMNTINFO */
+
+#ifdef MOUNTED_GETMNTINFO2 /* NetBSD, Minix */
+ {
+ struct statvfs *fsp;
+ int entries;
+
+ entries = getmntinfo (&fsp, MNT_NOWAIT);
+ if (entries < 0)
+ return NULL;
+ for (; entries-- > 0; fsp++)
+ {
+ me = g_malloc (sizeof (*me));
+ me->me_devname = g_strdup (fsp->f_mntfromname);
+ me->me_mountdir = g_strdup (fsp->f_mntonname);
+ me->me_mntroot = NULL;
+ me->me_type = g_strdup (fsp->f_fstypename);
+ me->me_type_malloced = 1;
+ me->me_dummy = ME_DUMMY (me->me_devname, me->me_type);
+ me->me_remote = ME_REMOTE (me->me_devname, me->me_type);
+ me->me_dev = (dev_t) (-1); /* Magic; means not known yet. */
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+ }
+#endif /* MOUNTED_GETMNTINFO2 */
+
+#if defined MOUNTED_FS_STAT_DEV /* Haiku, also (obsolete) BeOS */
+ {
+ /* The next_dev() and fs_stat_dev() system calls give the list of
+ all file systems, including the information returned by statvfs()
+ (fs type, total blocks, free blocks etc.), but without the mount
+ point. But on BeOS all file systems except / are mounted in the
+ rootfs, directly under /.
+ The directory name of the mount point is often, but not always,
+ identical to the volume name of the device.
+ We therefore get the list of subdirectories of /, and the list
+ of all file systems, and match the two lists. */
+
+ DIR *dirp;
+ struct rootdir_entry
+ {
+ char *name;
+ dev_t dev;
+ ino_t ino;
+ struct rootdir_entry *next;
+ };
+ struct rootdir_entry *rootdir_list;
+ struct rootdir_entry **rootdir_tail;
+ int32 pos;
+ dev_t dev;
+ fs_info fi;
+
+ /* All volumes are mounted in the rootfs, directly under /. */
+ rootdir_list = NULL;
+ rootdir_tail = &rootdir_list;
+ dirp = opendir (PATH_SEP_STR);
+ if (dirp)
+ {
+ struct dirent *d;
+
+ while ((d = readdir (dirp)) != NULL)
+ {
+ char *name;
+ struct stat statbuf;
+
+ if (DIR_IS_DOT (d->d_name))
+ continue;
+
+ if (DIR_IS_DOTDOT (d->d_name))
+ name = g_strdup (PATH_SEP_STR);
+ else
+ name = g_strconcat (PATH_SEP_STR, d->d_name, (char *) NULL);
+
+ if (lstat (name, &statbuf) >= 0 && S_ISDIR (statbuf.st_mode))
+ {
+ struct rootdir_entry *re = g_malloc (sizeof (*re));
+ re->name = name;
+ re->dev = statbuf.st_dev;
+ re->ino = statbuf.st_ino;
+
+ /* Add to the linked list. */
+ *rootdir_tail = re;
+ rootdir_tail = &re->next;
+ }
+ else
+ g_free (name);
+ }
+ closedir (dirp);
+ }
+ *rootdir_tail = NULL;
+
+ for (pos = 0; (dev = next_dev (&pos)) >= 0;)
+ if (fs_stat_dev (dev, &fi) >= 0)
+ {
+ /* Note: fi.dev == dev. */
+ struct rootdir_entry *re;
+
+ for (re = rootdir_list; re; re = re->next)
+ if (re->dev == fi.dev && re->ino == fi.root)
+ break;
+
+ me = g_malloc (sizeof (*me));
+ me->me_devname =
+ g_strdup (fi.device_name[0] != '\0' ? fi.device_name : fi.fsh_name);
+ me->me_mountdir = g_strdup (re != NULL ? re->name : fi.fsh_name);
+ me->me_mntroot = NULL;
+ me->me_type = g_strdup (fi.fsh_name);
+ me->me_type_malloced = 1;
+ me->me_dev = fi.dev;
+ me->me_dummy = 0;
+ me->me_remote = (fi.flags & B_FS_IS_SHARED) != 0;
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+
+ while (rootdir_list != NULL)
+ {
+ struct rootdir_entry *re = rootdir_list;
+
+ rootdir_list = re->next;
+ g_free (re->name);
+ g_free (re);
+ }
+ }
+#endif /* MOUNTED_FS_STAT_DEV */
+
+#ifdef MOUNTED_GETFSSTAT /* OSF/1, also (obsolete) Apple Darwin 1.3 */
+ {
+ int numsys, counter;
+ size_t bufsize;
+ struct statfs *stats;
+
+ numsys = getfsstat (NULL, 0L, MNT_NOWAIT);
+ if (numsys < 0)
+ return NULL;
+ if (SIZE_MAX / sizeof (*stats) <= numsys)
+ {
+ fprintf (stderr, "%s\n", _("Memory exhausted!"));
+ exit (EXIT_FAILURE);
+ }
+
+ bufsize = (1 + numsys) * sizeof (*stats);
+ stats = g_malloc (bufsize);
+ numsys = getfsstat (stats, bufsize, MNT_NOWAIT);
+
+ if (numsys < 0)
+ {
+ g_free (stats);
+ return NULL;
+ }
+
+ for (counter = 0; counter < numsys; counter++)
+ {
+ me = g_malloc (sizeof (*me));
+ me->me_devname = g_strdup (stats[counter].f_mntfromname);
+ me->me_mountdir = g_strdup (stats[counter].f_mntonname);
+ me->me_mntroot = NULL;
+ me->me_type = g_strdup (FS_TYPE (stats[counter]));
+ me->me_type_malloced = 1;
+ me->me_dummy = ME_DUMMY (me->me_devname, me->me_type);
+ me->me_remote = ME_REMOTE (me->me_devname, me->me_type);
+ me->me_dev = (dev_t) (-1); /* Magic; means not known yet. */
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+
+ g_free (stats);
+ }
+#endif /* MOUNTED_GETFSSTAT */
+
+#if defined MOUNTED_FREAD_FSTYP /* (obsolete) SVR3 */
+ {
+ struct mnttab mnt;
+ char *table = "/etc/mnttab";
+ FILE *fp;
+
+ fp = fopen (table, "r");
+ if (fp == NULL)
+ return NULL;
+
+ while (fread (&mnt, sizeof (mnt), 1, fp) > 0)
+ {
+ me = g_malloc (sizeof (*me));
+ me->me_devname = g_strdup (mnt.mt_dev);
+ me->me_mountdir = g_strdup (mnt.mt_filsys);
+ me->me_mntroot = NULL;
+ me->me_dev = (dev_t) (-1); /* Magic; means not known yet. */
+ me->me_type = "";
+ me->me_type_malloced = 0;
+ {
+ struct statfs fsd;
+ char typebuf[FSTYPSZ];
+
+ if (statfs (me->me_mountdir, &fsd, sizeof (fsd), 0) != -1
+ && sysfs (GETFSTYP, fsd.f_fstyp, typebuf) != -1)
+ {
+ me->me_type = g_strdup (typebuf);
+ me->me_type_malloced = 1;
+ }
+ }
+ me->me_dummy = ME_DUMMY (me->me_devname, me->me_type);
+ me->me_remote = ME_REMOTE (me->me_devname, me->me_type);
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+
+ if (ferror (fp))
+ {
+ /* The last fread() call must have failed. */
+ int saved_errno = errno;
+
+ fclose (fp);
+ errno = saved_errno;
+ goto free_then_fail;
+ }
+
+ if (fclose (fp) == EOF)
+ goto free_then_fail;
+ }
+#endif /* MOUNTED_FREAD_FSTYP */
+
+#ifdef MOUNTED_GETEXTMNTENT /* Solaris >= 8 */
+ {
+ struct extmnttab mnt;
+ const char *table = MNTTAB;
+ FILE *fp;
+ int ret;
+
+ /* No locking is needed, because the contents of /etc/mnttab is generated by the kernel. */
+
+ errno = 0;
+ fp = fopen (table, "r");
+ if (fp == NULL)
+ ret = errno;
+ else
+ {
+ while ((ret = getextmntent (fp, &mnt, 1)) == 0)
+ {
+ me = g_malloc (sizeof *me);
+ me->me_devname = g_strdup (mnt.mnt_special);
+ me->me_mountdir = g_strdup (mnt.mnt_mountp);
+ me->me_mntroot = NULL;
+ me->me_type = g_strdup (mnt.mnt_fstype);
+ me->me_type_malloced = 1;
+ me->me_dummy = MNT_IGNORE (&mnt) != 0;
+ me->me_remote = ME_REMOTE (me->me_devname, me->me_type);
+ me->me_dev = makedev (mnt.mnt_major, mnt.mnt_minor);
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+
+ ret = fclose (fp) == EOF ? errno : 0 < ret ? 0 : -1;
+ /* Here ret = -1 means success, ret >= 0 means failure. */
+ }
+
+ if (ret >= 0)
+ {
+ errno = ret;
+ goto free_then_fail;
+ }
+ }
+#endif /* MOUNTED_GETEXTMNTENT */
+
+#ifdef MOUNTED_GETMNTENT2 /* Solaris < 8, also (obsolete) SVR4 */
+ {
+ struct mnttab mnt;
+ const char *table = MNTTAB;
+ FILE *fp;
+ int ret;
+ int lockfd = -1;
+
+#if defined F_RDLCK && defined F_SETLKW
+ /* MNTTAB_LOCK is a macro name of our own invention; it's not present in
+ e.g. Solaris 2.6. If the SVR4 folks ever define a macro
+ for this file name, we should use their macro name instead.
+ (Why not just lock MNTTAB directly? We don't know.) */
+#ifndef MNTTAB_LOCK
+#define MNTTAB_LOCK "/etc/.mnttab.lock"
+#endif
+ lockfd = open (MNTTAB_LOCK, O_RDONLY);
+ if (lockfd >= 0)
+ {
+ struct flock flock;
+
+ flock.l_type = F_RDLCK;
+ flock.l_whence = SEEK_SET;
+ flock.l_start = 0;
+ flock.l_len = 0;
+ while (fcntl (lockfd, F_SETLKW, &flock) == -1)
+ if (errno != EINTR)
+ {
+ int saved_errno = errno;
+ close (lockfd);
+ errno = saved_errno;
+ return NULL;
+ }
+ }
+ else if (errno != ENOENT)
+ return NULL;
+#endif
+
+ errno = 0;
+ fp = fopen (table, "r");
+ if (fp == NULL)
+ ret = errno;
+ else
+ {
+ while ((ret = getmntent (fp, &mnt)) == 0)
+ {
+ me = g_malloc (sizeof (*me));
+ me->me_devname = g_strdup (mnt.mnt_special);
+ me->me_mountdir = g_strdup (mnt.mnt_mountp);
+ me->me_mntroot = NULL;
+ me->me_type = g_strdup (mnt.mnt_fstype);
+ me->me_type_malloced = 1;
+ me->me_dummy = MNT_IGNORE (&mnt) != 0;
+ me->me_remote = ME_REMOTE (me->me_devname, me->me_type);
+ me->me_dev = dev_from_mount_options (mnt.mnt_mntopts);
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+
+ ret = fclose (fp) == EOF ? errno : 0 < ret ? 0 : -1;
+ /* Here ret = -1 means success, ret >= 0 means failure. */
+ }
+
+ if (lockfd >= 0 && close (lockfd) != 0)
+ ret = errno;
+
+ if (ret >= 0)
+ {
+ errno = ret;
+ goto free_then_fail;
+ }
+ }
+#endif /* MOUNTED_GETMNTENT2. */
+
+#ifdef MOUNTED_VMOUNT /* AIX */
+ {
+ int bufsize;
+ void *entries;
+ char *thisent;
+ struct vmount *vmp;
+ int n_entries;
+ int i;
+
+ /* Ask how many bytes to allocate for the mounted file system info. */
+ entries = &bufsize;
+ if (mntctl (MCTL_QUERY, sizeof (bufsize), entries) != 0)
+ return NULL;
+ entries = g_malloc (bufsize);
+
+ /* Get the list of mounted file systems. */
+ n_entries = mntctl (MCTL_QUERY, bufsize, entries);
+ if (n_entries < 0)
+ {
+ int saved_errno = errno;
+
+ g_free (entries);
+ errno = saved_errno;
+ return NULL;
+ }
+
+ for (i = 0, thisent = entries; i < n_entries; i++, thisent += vmp->vmt_length)
+ {
+ char *options, *ignore;
+
+ vmp = (struct vmount *) thisent;
+ me = g_malloc (sizeof (*me));
+ if (vmp->vmt_flags & MNT_REMOTE)
+ {
+ char *host, *dir;
+
+ me->me_remote = 1;
+ /* Prepend the remote dirname. */
+ host = thisent + vmp->vmt_data[VMT_HOSTNAME].vmt_off;
+ dir = thisent + vmp->vmt_data[VMT_OBJECT].vmt_off;
+ me->me_devname = g_strconcat (host, ":", dir, (char *) NULL);
+ }
+ else
+ {
+ me->me_remote = 0;
+ me->me_devname = g_strdup (thisent + vmp->vmt_data[VMT_OBJECT].vmt_off);
+ }
+ me->me_mountdir = g_strdup (thisent + vmp->vmt_data[VMT_STUB].vmt_off);
+ me->me_mntroot = NULL;
+ me->me_type = g_strdup (fstype_to_string (vmp->vmt_gfstype));
+ me->me_type_malloced = 1;
+ options = thisent + vmp->vmt_data[VMT_ARGS].vmt_off;
+ ignore = strstr (options, "ignore");
+ me->me_dummy = (ignore
+ && (ignore == options || ignore[-1] == ',')
+ && (ignore[sizeof ("ignore") - 1] == ','
+ || ignore[sizeof ("ignore") - 1] == '\0'));
+ me->me_dev = (dev_t) (-1); /* vmt_fsid might be the info we want. */
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+ g_free (entries);
+ }
+#endif /* MOUNTED_VMOUNT. */
+
+#ifdef MOUNTED_INTERIX_STATVFS /* Interix */
+ {
+ DIR *dirp = opendir ("/dev/fs");
+ char node[9 + NAME_MAX];
+
+ if (!dirp)
+ goto free_then_fail;
+
+ while (1)
+ {
+ struct statvfs dev;
+ struct dirent entry;
+ struct dirent *result;
+
+ if (readdir_r (dirp, &entry, &result) || result == NULL)
+ break;
+
+ strcpy (node, "/dev/fs/");
+ strcat (node, entry.d_name);
+
+ if (statvfs (node, &dev) == 0)
+ {
+ me = g_malloc (sizeof *me);
+ me->me_devname = g_strdup (dev.f_mntfromname);
+ me->me_mountdir = g_strdup (dev.f_mntonname);
+ me->me_mntroot = NULL;
+ me->me_type = g_strdup (dev.f_fstypename);
+ me->me_type_malloced = 1;
+ me->me_dummy = ME_DUMMY (me->me_devname, me->me_type);
+ me->me_remote = ME_REMOTE (me->me_devname, me->me_type);
+ me->me_dev = (dev_t) (-1); /* Magic; means not known yet. */
+
+ mount_list = g_slist_prepend (mount_list, me);
+ }
+ }
+ closedir (dirp);
+ }
+#endif /* MOUNTED_INTERIX_STATVFS */
+
+ return g_slist_reverse (mount_list);
+
+ free_then_fail:
+ {
+ int saved_errno = errno;
+
+ g_slist_free_full (mount_list, (GDestroyNotify) free_mount_entry);
+
+ errno = saved_errno;
+ return NULL;
+ }
+}
+#endif /* HAVE_INFOMOUNT_LIST */
+
+/* --------------------------------------------------------------------------------------------- */
+
+#ifdef HAVE_INFOMOUNT_QNX
+/**
+ ** QNX has no [gs]etmnt*(), [gs]etfs*(), or /etc/mnttab, but can do
+ ** this via the following code.
+ ** Note that, as this is based on CWD, it only fills one mount_entry
+ ** structure. See my_statfs() below for the "other side" of this hack.
+ */
+
+static GSList *
+read_file_system_list (void)
+{
+ struct _disk_entry de;
+ struct statfs fs;
+ int i, fd;
+ char *tp, dev[_POSIX_NAME_MAX], dir[_POSIX_PATH_MAX];
+ struct mount_entry *me = NULL;
+ static GSList *list = NULL;
+
+ if (list != NULL)
+ {
+ me = (struct mount_entry *) list->data;
+
+ g_free (me->me_devname);
+ g_free (me->me_mountdir);
+ g_free (me->me_mntroot);
+ g_free (me->me_type);
+ }
+ else
+ {
+ me = (struct mount_entry *) g_malloc (sizeof (struct mount_entry));
+ list = g_slist_prepend (list, me);
+ }
+
+ if (!getcwd (dir, _POSIX_PATH_MAX))
+ return (NULL);
+
+ fd = open (dir, O_RDONLY);
+ if (fd == -1)
+ return (NULL);
+
+ i = disk_get_entry (fd, &de);
+
+ close (fd);
+
+ if (i == -1)
+ return (NULL);
+
+ switch (de.disk_type)
+ {
+ case _UNMOUNTED:
+ tp = "unmounted";
+ break;
+ case _FLOPPY:
+ tp = "Floppy";
+ break;
+ case _HARD:
+ tp = "Hard";
+ break;
+ case _RAMDISK:
+ tp = "Ram";
+ break;
+ case _REMOVABLE:
+ tp = "Removable";
+ break;
+ case _TAPE:
+ tp = "Tape";
+ break;
+ case _CDROM:
+ tp = "CDROM";
+ break;
+ default:
+ tp = "unknown";
+ }
+
+ if (fsys_get_mount_dev (dir, &dev) == -1)
+ return (NULL);
+
+ if (fsys_get_mount_pt (dev, &dir) == -1)
+ return (NULL);
+
+ me->me_devname = g_strdup (dev);
+ me->me_mountdir = g_strdup (dir);
+ me->me_mntroot = NULL;
+ me->me_type = g_strdup (tp);
+ me->me_dev = de.disk_type;
+
+#ifdef DEBUG
+ fprintf (stderr,
+ "disk_get_entry():\n\tdisk_type=%d (%s)\n\tdriver_name='%-*.*s'\n\tdisk_drv=%d\n",
+ de.disk_type, tp, _DRIVER_NAME_LEN, _DRIVER_NAME_LEN, de.driver_name, de.disk_drv);
+ fprintf (stderr, "fsys_get_mount_dev():\n\tdevice='%s'\n", dev);
+ fprintf (stderr, "fsys_get_mount_pt():\n\tmount point='%s'\n", dir);
+#endif /* DEBUG */
+
+ return (list);
+}
+#endif /* HAVE_INFOMOUNT_QNX */
+
+/* --------------------------------------------------------------------------------------------- */
+
+#ifdef HAVE_INFOMOUNT
+/* Fill in the fields of FSP with information about space usage for
+ the file system on which FILE resides.
+ DISK is the device on which FILE is mounted, for space-getting
+ methods that need to know it.
+ Return 0 if successful, -1 if not. When returning -1, ensure that
+ ERRNO is either a system error value, or zero if DISK is NULL
+ on a system that requires a non-NULL value. */
+static int
+get_fs_usage (char const *file, char const *disk, struct fs_usage *fsp)
+{
+#ifdef STAT_STATVFS /* POSIX, except pre-2.6.36 glibc/Linux */
+
+ if (statvfs_works ())
+ {
+ struct statvfs vfsd;
+
+ if (statvfs (file, &vfsd) < 0)
+ return -1;
+
+ /* f_frsize isn't guaranteed to be supported. */
+ fsp->fsu_blocksize = (vfsd.f_frsize
+ ? PROPAGATE_ALL_ONES (vfsd.f_frsize)
+ : PROPAGATE_ALL_ONES (vfsd.f_bsize));
+
+ fsp->fsu_blocks = PROPAGATE_ALL_ONES (vfsd.f_blocks);
+ fsp->fsu_bfree = PROPAGATE_ALL_ONES (vfsd.f_bfree);
+ fsp->fsu_bavail = PROPAGATE_TOP_BIT (vfsd.f_bavail);
+ fsp->fsu_bavail_top_bit_set = EXTRACT_TOP_BIT (vfsd.f_bavail) != 0;
+ fsp->fsu_files = PROPAGATE_ALL_ONES (vfsd.f_files);
+ fsp->fsu_ffree = PROPAGATE_ALL_ONES (vfsd.f_ffree);
+ }
+ else
+#endif
+
+ {
+#if defined STAT_STATVFS64 /* AIX */
+
+ struct statvfs64 fsd;
+
+ if (statvfs64 (file, &fsd) < 0)
+ return -1;
+
+ /* f_frsize isn't guaranteed to be supported. */
+ /* *INDENT-OFF* */
+ fsp->fsu_blocksize = fsd.f_frsize
+ ? PROPAGATE_ALL_ONES (fsd.f_frsize)
+ : PROPAGATE_ALL_ONES (fsd.f_bsize);
+ /* *INDENT-ON* */
+
+#elif defined STAT_STATFS3_OSF1 /* OSF/1 */
+
+ struct statfs fsd;
+
+ if (statfs (file, &fsd, sizeof (struct statfs)) != 0)
+ return -1;
+
+ fsp->fsu_blocksize = PROPAGATE_ALL_ONES (fsd.f_fsize);
+
+#elif defined STAT_STATFS2_FRSIZE /* 2.6 < glibc/Linux < 2.6.36 */
+
+ struct statfs fsd;
+
+ if (statfs (file, &fsd) < 0)
+ return -1;
+
+ fsp->fsu_blocksize = PROPAGATE_ALL_ONES (fsd.f_frsize);
+
+#elif defined STAT_STATFS2_BSIZE /* glibc/Linux < 2.6, 4.3BSD, SunOS 4, \
+ Mac OS X < 10.4, FreeBSD < 5.0, \
+ NetBSD < 3.0, OpenBSD < 4.4 */
+
+ struct statfs fsd;
+
+ if (statfs (file, &fsd) < 0)
+ return -1;
+
+ fsp->fsu_blocksize = PROPAGATE_ALL_ONES (fsd.f_bsize);
+
+#ifdef STATFS_TRUNCATES_BLOCK_COUNTS
+
+ /* In SunOS 4.1.2, 4.1.3, and 4.1.3_U1, the block counts in the
+ struct statfs are truncated to 2GB. These conditions detect that
+ truncation, presumably without botching the 4.1.1 case, in which
+ the values are not truncated. The correct counts are stored in
+ undocumented spare fields. */
+ if (fsd.f_blocks == 0x7fffffff / fsd.f_bsize && fsd.f_spare[0] > 0)
+ {
+ fsd.f_blocks = fsd.f_spare[0];
+ fsd.f_bfree = fsd.f_spare[1];
+ fsd.f_bavail = fsd.f_spare[2];
+ }
+#endif /* STATFS_TRUNCATES_BLOCK_COUNTS */
+
+#elif defined STAT_STATFS2_FSIZE /* 4.4BSD and older NetBSD */
+
+ struct statfs fsd;
+
+ if (statfs (file, &fsd) < 0)
+ return -1;
+
+ fsp->fsu_blocksize = PROPAGATE_ALL_ONES (fsd.f_fsize);
+
+#elif defined STAT_STATFS4 /* SVR3, old Irix */
+
+ struct statfs fsd;
+
+ if (statfs (file, &fsd, sizeof (fsd), 0) < 0)
+ return -1;
+
+ /* Empirically, the block counts on most SVR3 and SVR3-derived
+ systems seem to always be in terms of 512-byte blocks,
+ no matter what value f_bsize has. */
+ fsp->fsu_blocksize = 512;
+#endif
+
+#if (defined STAT_STATVFS64 || defined STAT_STATFS3_OSF1 \
+ || defined STAT_STATFS2_FRSIZE || defined STAT_STATFS2_BSIZE \
+ || defined STAT_STATFS2_FSIZE || defined STAT_STATFS4)
+
+ fsp->fsu_blocks = PROPAGATE_ALL_ONES (fsd.f_blocks);
+ fsp->fsu_bfree = PROPAGATE_ALL_ONES (fsd.f_bfree);
+ fsp->fsu_bavail = PROPAGATE_TOP_BIT (fsd.f_bavail);
+ fsp->fsu_bavail_top_bit_set = EXTRACT_TOP_BIT (fsd.f_bavail) != 0;
+ fsp->fsu_files = PROPAGATE_ALL_ONES (fsd.f_files);
+ fsp->fsu_ffree = PROPAGATE_ALL_ONES (fsd.f_ffree);
+
+#endif
+ }
+
+ (void) disk; /* avoid argument-unused warning */
+
+ return 0;
+}
+#endif /* HAVE_INFOMOUNT */
+
+/* --------------------------------------------------------------------------------------------- */
+/*** public functions ****************************************************************************/
+/* --------------------------------------------------------------------------------------------- */
+
+void
+free_my_statfs (void)
+{
+#ifdef HAVE_INFOMOUNT_LIST
+ g_clear_slist (&mc_mount_list, (GDestroyNotify) free_mount_entry);
+#endif /* HAVE_INFOMOUNT_LIST */
+}
+
+/* --------------------------------------------------------------------------------------------- */
+
+void
+init_my_statfs (void)
+{
+#ifdef HAVE_INFOMOUNT_LIST
+ free_my_statfs ();
+ mc_mount_list = read_file_system_list ();
+#endif /* HAVE_INFOMOUNT_LIST */
+}
+
+/* --------------------------------------------------------------------------------------------- */
+
+void
+my_statfs (struct my_statfs *myfs_stats, const char *path)
+{
+#ifdef HAVE_INFOMOUNT_LIST
+ size_t len = 0;
+ struct mount_entry *entry = NULL;
+ GSList *temp;
+ struct fs_usage fs_use;
+
+ for (temp = mc_mount_list; temp != NULL; temp = g_slist_next (temp))
+ {
+ struct mount_entry *me;
+ size_t i;
+
+ me = (struct mount_entry *) temp->data;
+ i = strlen (me->me_mountdir);
+ if (i > len && (strncmp (path, me->me_mountdir, i) == 0) &&
+ (entry == NULL || IS_PATH_SEP (path[i]) || path[i] == '\0'))
+ {
+ len = i;
+ entry = me;
+ }
+ }
+
+ if (entry != NULL)
+ {
+ memset (&fs_use, 0, sizeof (fs_use));
+ get_fs_usage (entry->me_mountdir, NULL, &fs_use);
+
+ myfs_stats->type = entry->me_dev;
+ myfs_stats->typename = entry->me_type;
+ myfs_stats->mpoint = entry->me_mountdir;
+ myfs_stats->mroot = entry->me_mntroot;
+ myfs_stats->device = entry->me_devname;
+ myfs_stats->avail =
+ ((uintmax_t) (getuid ()? fs_use.fsu_bavail : fs_use.fsu_bfree) *
+ fs_use.fsu_blocksize) >> 10;
+ myfs_stats->total = ((uintmax_t) fs_use.fsu_blocks * fs_use.fsu_blocksize) >> 10;
+ myfs_stats->nfree = (uintmax_t) fs_use.fsu_ffree;
+ myfs_stats->nodes = (uintmax_t) fs_use.fsu_files;
+ }
+ else
+#endif /* HAVE_INFOMOUNT_LIST */
+
+#ifdef HAVE_INFOMOUNT_QNX
+ /*
+ ** This is the "other side" of the hack to read_file_system_list() above.
+ ** It's not the most efficient approach, but consumes less memory. It
+ ** also accommodates QNX's ability to mount filesystems on the fly.
+ */
+ struct mount_entry *entry;
+ struct fs_usage fs_use;
+
+ entry = read_file_system_list ();
+ if (entry != NULL)
+ {
+ get_fs_usage (entry->me_mountdir, NULL, &fs_use);
+
+ myfs_stats->type = entry->me_dev;
+ myfs_stats->typename = entry->me_type;
+ myfs_stats->mpoint = entry->me_mountdir;
+ myfs_stats->mroot = entry->me_mntroot;
+ myfs_stats->device = entry->me_devname;
+
+ myfs_stats->avail = ((uintmax_t) fs_use.fsu_bfree * fs_use.fsu_blocksize) >> 10;
+ myfs_stats->total = ((uintmax_t) fs_use.fsu_blocks * fs_use.fsu_blocksize) >> 10;
+ myfs_stats->nfree = (uintmax_t) fs_use.fsu_ffree;
+ myfs_stats->nodes = (uintmax_t) fs_use.fsu_files;
+ }
+ else
+#endif /* HAVE_INFOMOUNT_QNX */
+ {
+ myfs_stats->type = 0;
+ myfs_stats->mpoint = "unknown";
+ myfs_stats->device = "unknown";
+ myfs_stats->avail = 0;
+ myfs_stats->total = 0;
+ myfs_stats->nfree = 0;
+ myfs_stats->nodes = 0;
+ }
+}
+
+/* --------------------------------------------------------------------------------------------- */