pivot_root: switch to a new mechanism (v2)

This idea came from Andy Lutomirski.  Instead of using a
temporary directory for the pivot_root put-old, use "." both
for new-root and old-root.  Then fchdir into the old root
temporarily in order to unmount the old-root, and finally
chdir back into our '/'.

Drop lxc.pivotdir from the lxc.container.conf manpage.

Warn when we see a lxc.pivotdir entry (but keep it in the
lxc.conf for now).

Signed-off-by: Serge Hallyn <serge.hallyn@ubuntu.com>
Acked-by: Stéphane Graber <stgraber@ubuntu.com>
This commit is contained in:
Serge Hallyn 2014-09-20 03:15:44 +00:00 committed by Stéphane Graber
parent 01db019751
commit 2d489f9e87
3 changed files with 44 additions and 190 deletions

View File

@ -943,20 +943,6 @@ proc proc proc nodev,noexec,nosuid 0 0
</listitem>
</varlistentry>
<varlistentry>
<term>
<option>lxc.pivotdir</option>
</term>
<listitem>
<para>
where to pivot the original root file system under
<option>lxc.rootfs</option>, specified relatively to
that. The default is <filename>mnt</filename>.
It is created if necessary, and also removed after
unmounting everything from it during container setup.
</para>
</listitem>
</varlistentry>
</variablelist>
</refsect2>

View File

@ -1025,199 +1025,66 @@ static int setup_tty(const struct lxc_rootfs *rootfs,
return 0;
}
static int setup_rootfs_pivot_root_cb(char *buffer, void *data)
{
struct lxc_list *mountlist, *listentry, *iterator;
char *pivotdir, *mountpoint, *mountentry, *saveptr = NULL;
int found;
void **cbparm;
mountentry = buffer;
cbparm = (void **)data;
mountlist = cbparm[0];
pivotdir = cbparm[1];
/* parse entry, first field is mountname, ignore */
mountpoint = strtok_r(mountentry, " ", &saveptr);
if (!mountpoint)
return -1;
/* second field is mountpoint */
mountpoint = strtok_r(NULL, " ", &saveptr);
if (!mountpoint)
return -1;
/* only consider mountpoints below old root fs */
if (strncmp(mountpoint, pivotdir, strlen(pivotdir)))
return 0;
/* filter duplicate mountpoints */
found = 0;
lxc_list_for_each(iterator, mountlist) {
if (!strcmp(iterator->elem, mountpoint)) {
found = 1;
break;
}
}
if (found)
return 0;
/* add entry to list */
listentry = malloc(sizeof(*listentry));
if (!listentry) {
SYSERROR("malloc for mountpoint listentry failed");
return -1;
}
listentry->elem = strdup(mountpoint);
if (!listentry->elem) {
SYSERROR("strdup failed");
free(listentry);
return -1;
}
lxc_list_add_tail(mountlist, listentry);
return 0;
}
static int umount_oldrootfs(const char *oldrootfs)
{
char path[MAXPATHLEN];
void *cbparm[2];
struct lxc_list mountlist, *iterator, *next;
int ok, still_mounted, last_still_mounted;
int rc;
/* read and parse /proc/mounts in old root fs */
lxc_list_init(&mountlist);
/* oldrootfs is on the top tree directory now */
rc = snprintf(path, sizeof(path), "/%s", oldrootfs);
if (rc >= sizeof(path)) {
ERROR("rootfs name too long");
return -1;
}
cbparm[0] = &mountlist;
cbparm[1] = strdup(path);
if (!cbparm[1]) {
SYSERROR("strdup failed");
return -1;
}
rc = snprintf(path, sizeof(path), "%s/proc/mounts", oldrootfs);
if (rc >= sizeof(path)) {
ERROR("container proc/mounts name too long");
return -1;
}
ok = lxc_file_for_each_line(path,
setup_rootfs_pivot_root_cb, &cbparm);
if (ok < 0) {
SYSERROR("failed to read or parse mount list '%s'", path);
return -1;
}
/* umount filesystems until none left or list no longer shrinks */
still_mounted = 0;
do {
last_still_mounted = still_mounted;
still_mounted = 0;
lxc_list_for_each_safe(iterator, &mountlist, next) {
/* umount normally */
if (!umount(iterator->elem)) {
DEBUG("umounted '%s'", (char *)iterator->elem);
lxc_list_del(iterator);
continue;
}
still_mounted++;
}
} while (still_mounted > 0 && still_mounted != last_still_mounted);
lxc_list_for_each(iterator, &mountlist) {
/* let's try a lazy umount */
if (!umount2(iterator->elem, MNT_DETACH)) {
INFO("lazy unmount of '%s'", (char *)iterator->elem);
continue;
}
/* be more brutal (nfs) */
if (!umount2(iterator->elem, MNT_FORCE)) {
INFO("forced unmount of '%s'", (char *)iterator->elem);
continue;
}
WARN("failed to unmount '%s'", (char *)iterator->elem);
}
return 0;
}
static int setup_rootfs_pivot_root(const char *rootfs, const char *pivotdir)
{
char path[MAXPATHLEN];
int remove_pivotdir = 0;
int rc;
int oldroot = -1, newroot = -1;
oldroot = open("/", O_DIRECTORY | O_RDONLY);
if (oldroot < 0) {
SYSERROR("Error opening old-/ for fchdir");
return -1;
}
newroot = open(rootfs, O_DIRECTORY | O_RDONLY);
if (newroot < 0) {
SYSERROR("Error opening new-/ for fchdir");
goto fail;
}
/* change into new root fs */
if (chdir(rootfs)) {
if (fchdir(newroot)) {
SYSERROR("can't chdir to new rootfs '%s'", rootfs);
return -1;
goto fail;
}
if (!pivotdir)
pivotdir = "lxc_putold";
/* compute the full path to pivotdir under rootfs */
rc = snprintf(path, sizeof(path), "%s/%s", rootfs, pivotdir);
if (rc >= sizeof(path)) {
ERROR("pivot dir name too long");
return -1;
}
if (access(path, F_OK)) {
if (mkdir_p(path, 0755) < 0) {
SYSERROR("failed to create pivotdir '%s'", path);
return -1;
}
remove_pivotdir = 1;
DEBUG("created '%s' directory", path);
}
DEBUG("mountpoint for old rootfs is '%s'", path);
/* pivot_root into our new root fs */
if (pivot_root(".", path)) {
if (pivot_root(".", ".")) {
SYSERROR("pivot_root syscall failed");
return -1;
goto fail;
}
if (chdir("/")) {
SYSERROR("can't chdir to / after pivot_root");
return -1;
/*
* at this point the old-root is mounted on top of our new-root
* To unmounted it we must not be chdir'd into it, so escape back
* to old-root
*/
if (fchdir(oldroot) < 0) {
SYSERROR("Error entering oldroot");
goto fail;
}
if (umount2("/", MNT_DETACH) < 0) {
SYSERROR("Error detaching old root");
goto fail;
}
if (fchdir(newroot) < 0) {
SYSERROR("Error re-entering newroot");
goto fail;
}
close(oldroot);
close(newroot);
DEBUG("pivot_root syscall to '%s' successful", rootfs);
/* we switch from absolute path to relative path */
if (umount_oldrootfs(pivotdir))
return -1;
/* remove temporary mount point, we don't consider the removing
* as fatal */
if (remove_pivotdir && rmdir(pivotdir))
WARN("can't remove mountpoint '%s': %m", pivotdir);
return 0;
fail:
if (oldroot != -1)
close(oldroot);
if (newroot != -1)
close(newroot);
return -1;
}
/*

View File

@ -1662,6 +1662,7 @@ static int config_rootfs_options(const char *key, const char *value,
static int config_pivotdir(const char *key, const char *value,
struct lxc_conf *lxc_conf)
{
WARN("lxc.pivotdir is ignored. It will soon become an error.");
return config_path_item(&lxc_conf->rootfs.pivot, value);
}