| /* vi: set sw=4 ts=4: */ |
| /* |
| * Mini umount implementation for busybox |
| * |
| * Copyright (C) 1999-2004 by Erik Andersen <andersen@codepoet.org> |
| * Copyright (C) 2005 by Rob Landley <rob@landley.net> |
| * |
| * Licensed under GPLv2, see file LICENSE in this source tree. |
| */ |
| //config:config UMOUNT |
| //config: bool "umount (5.1 kb)" |
| //config: default y |
| //config: select PLATFORM_LINUX |
| //config: help |
| //config: When you want to remove a mounted filesystem from its current mount |
| //config: point, for example when you are shutting down the system, the |
| //config: 'umount' utility is the tool to use. If you enabled the 'mount' |
| //config: utility, you almost certainly also want to enable 'umount'. |
| //config: |
| //config:config FEATURE_UMOUNT_ALL |
| //config: bool "Support -a (unmount all)" |
| //config: default y |
| //config: depends on UMOUNT |
| //config: help |
| //config: Support -a option to unmount all currently mounted filesystems. |
| |
| //applet:IF_UMOUNT(APPLET_NOEXEC(umount, umount, BB_DIR_BIN, BB_SUID_DROP, umount)) |
| /* |
| * On one hand, in some weird situations you'd want umount |
| * to not do anything surprising, to behave as a usual fork+execed executable. |
| * |
| * OTOH, there can be situations where execing would not succeed, or even hang |
| * (say, if executable is on a filesystem which is in trouble and accesses to it |
| * block in kernel). |
| * In this case, you might be actually happy if your standalone bbox shell |
| * does not fork+exec, but only forks and calls umount_main() which it already has! |
| * Let's go with NOEXEC. |
| * |
| * bb_common_bufsiz1 usage here is safe wrt NOEXEC: not expecting it to be zeroed. |
| */ |
| |
| //kbuild:lib-$(CONFIG_UMOUNT) += umount.o |
| |
| //usage:#define umount_trivial_usage |
| //usage: "[OPTIONS] FILESYSTEM|DIRECTORY" |
| //usage:#define umount_full_usage "\n\n" |
| //usage: "Unmount file systems\n" |
| //usage: IF_FEATURE_UMOUNT_ALL( |
| //usage: "\n -a Unmount all file systems" IF_FEATURE_MTAB_SUPPORT(" in /etc/mtab") |
| //usage: ) |
| //usage: IF_FEATURE_MTAB_SUPPORT( |
| //usage: "\n -n Don't erase /etc/mtab entries" |
| //usage: ) |
| //usage: "\n -r Try to remount devices as read-only if mount is busy" |
| //usage: "\n -l Lazy umount (detach filesystem)" |
| //usage: "\n -f Force umount (i.e., unreachable NFS server)" |
| //usage: IF_FEATURE_MOUNT_LOOP( |
| //usage: "\n -d Free loop device if it has been used" |
| //usage: ) |
| //usage: "\n -t FSTYPE[,...] Unmount only these filesystem type(s)" |
| //usage: |
| //usage:#define umount_example_usage |
| //usage: "$ umount /dev/hdc1\n" |
| |
| #include <mntent.h> |
| #include <sys/mount.h> |
| #ifndef MNT_DETACH |
| # define MNT_DETACH 0x00000002 |
| #endif |
| #include "libbb.h" |
| #include "common_bufsiz.h" |
| |
| #if defined(__dietlibc__) |
| // TODO: This does not belong here. |
| /* 16.12.2006, Sampo Kellomaki (sampo@iki.fi) |
| * dietlibc-0.30 does not have implementation of getmntent_r() */ |
| static struct mntent *getmntent_r(FILE* stream, struct mntent* result, |
| char* buffer UNUSED_PARAM, int bufsize UNUSED_PARAM) |
| { |
| struct mntent* ment = getmntent(stream); |
| return memcpy(result, ment, sizeof(*ment)); |
| } |
| #endif |
| |
| /* ignored: -c -v -i */ |
| #define OPTION_STRING "fldnrat:" "cvi" |
| #define OPT_FORCE (1 << 0) // Same as MNT_FORCE |
| #define OPT_LAZY (1 << 1) // Same as MNT_DETACH |
| #define OPT_FREELOOP (1 << 2) |
| #define OPT_NO_MTAB (1 << 3) |
| #define OPT_REMOUNT (1 << 4) |
| #define OPT_ALL (ENABLE_FEATURE_UMOUNT_ALL ? (1 << 5) : 0) |
| |
| int umount_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE; |
| int umount_main(int argc UNUSED_PARAM, char **argv) |
| { |
| int doForce; |
| struct mntent me; |
| FILE *fp; |
| char *fstype = NULL; |
| int status = EXIT_SUCCESS; |
| unsigned opt; |
| struct mtab_list { |
| char *dir; |
| char *device; |
| struct mtab_list *next; |
| } *mtl, *m; |
| |
| opt = getopt32(argv, OPTION_STRING, &fstype); |
| //argc -= optind; |
| argv += optind; |
| |
| // MNT_FORCE and MNT_DETACH (from linux/fs.h) must match |
| // OPT_FORCE and OPT_LAZY. |
| BUILD_BUG_ON(OPT_FORCE != MNT_FORCE || OPT_LAZY != MNT_DETACH); |
| doForce = opt & (OPT_FORCE|OPT_LAZY); |
| |
| /* Get a list of mount points from mtab. We read them all in now mostly |
| * for umount -a (so we don't have to worry about the list changing while |
| * we iterate over it, or about getting stuck in a loop on the same failing |
| * entry. Notice that this also naturally reverses the list so that -a |
| * umounts the most recent entries first. */ |
| m = mtl = NULL; |
| |
| // If we're umounting all, then m points to the start of the list and |
| // the argument list should be empty (which will match all). |
| fp = setmntent(bb_path_mtab_file, "r"); |
| if (!fp) { |
| if (opt & OPT_ALL) |
| bb_error_msg_and_die("can't open '%s'", bb_path_mtab_file); |
| } else { |
| setup_common_bufsiz(); |
| while (getmntent_r(fp, &me, bb_common_bufsiz1, COMMON_BUFSIZE)) { |
| /* Match fstype (fstype==NULL matches always) */ |
| if (!fstype_matches(me.mnt_type, fstype)) |
| continue; |
| m = xzalloc(sizeof(*m)); |
| m->next = mtl; |
| m->device = xstrdup(me.mnt_fsname); |
| m->dir = xstrdup(me.mnt_dir); |
| mtl = m; |
| } |
| endmntent(fp); |
| } |
| |
| // If we're not umounting all, we need at least one argument. |
| // Note: "-t FSTYPE" does not imply -a. |
| if (!(opt & OPT_ALL)) { |
| if (!argv[0]) |
| bb_show_usage(); |
| m = NULL; |
| } |
| |
| // Loop through everything we're supposed to umount, and do so. |
| for (;;) { |
| int curstat; |
| char *zapit = *argv; |
| char *path; |
| |
| // Do we already know what to umount this time through the loop? |
| if (m) |
| path = xstrdup(m->dir); |
| // For umount -a, end of mtab means time to exit. |
| else if (opt & OPT_ALL) |
| break; |
| // Use command line argument (and look it up in mtab list) |
| else { |
| if (!zapit) |
| break; |
| argv++; |
| path = xmalloc_realpath(zapit); |
| if (path) { |
| for (m = mtl; m; m = m->next) |
| if (strcmp(path, m->dir) == 0 || strcmp(path, m->device) == 0) |
| break; |
| } |
| } |
| // If we couldn't find this sucker in /etc/mtab, punt by passing our |
| // command line argument straight to the umount syscall. Otherwise, |
| // umount the directory even if we were given the block device. |
| if (m) zapit = m->dir; |
| |
| // umount from util-linux 2.22.2 does not do this: |
| // umount -f uses umount2(MNT_FORCE) immediately, |
| // not trying umount() first. |
| // (Strangely, umount -fl ignores -f: it is equivalent to umount -l. |
| // We do pass both flags in this case) |
| #if 0 |
| // Let's ask the thing nicely to unmount. |
| curstat = umount(zapit); |
| |
| // Unmount with force and/or lazy flags, if necessary. |
| if (curstat && doForce) |
| #endif |
| curstat = umount2(zapit, doForce); |
| |
| // If still can't umount, maybe remount read-only? |
| if (curstat) { |
| if ((opt & OPT_REMOUNT) && errno == EBUSY && m) { |
| // Note! Even if we succeed here, later we should not |
| // free loop device or erase mtab entry! |
| const char *msg = "%s busy - remounted read-only"; |
| curstat = mount(m->device, zapit, NULL, MS_REMOUNT|MS_RDONLY, NULL); |
| if (curstat) { |
| msg = "can't remount %s read-only"; |
| status = EXIT_FAILURE; |
| } |
| bb_error_msg(msg, m->device); |
| } else { |
| status = EXIT_FAILURE; |
| bb_perror_msg("can't unmount %s", zapit); |
| } |
| } else { |
| // De-allocate the loop device. This ioctl should be ignored on |
| // any non-loop block devices. |
| if (ENABLE_FEATURE_MOUNT_LOOP && (opt & OPT_FREELOOP) && m) |
| del_loop(m->device); |
| if (ENABLE_FEATURE_MTAB_SUPPORT && !(opt & OPT_NO_MTAB) && m) |
| erase_mtab(m->dir); |
| } |
| |
| // Find next matching mtab entry for -a or umount /dev |
| // Note this means that "umount /dev/blah" will unmount all instances |
| // of /dev/blah, not just the most recent. |
| if (m) { |
| while ((m = m->next) != NULL) |
| // NB: if m is non-NULL, path is non-NULL as well |
| if ((opt & OPT_ALL) || strcmp(path, m->device) == 0) |
| break; |
| } |
| free(path); |
| } |
| |
| // Free mtab list if necessary |
| if (ENABLE_FEATURE_CLEAN_UP) { |
| while (mtl) { |
| m = mtl->next; |
| free(mtl->device); |
| free(mtl->dir); |
| free(mtl); |
| mtl = m; |
| } |
| } |
| |
| return status; |
| } |