blob: 387f70859f96631af2402f6636167825cff1dd55 [file] [log] [blame]
/*
* Copyright (C) 2012 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <fcntl.h>
#include <ctype.h>
#include <sys/mount.h>
#include <sys/stat.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <libgen.h>
#include <time.h>
#include <sys/swap.h>
#include <dirent.h>
#include <ext4.h>
#include <ext4_sb.h>
#include <ext4_crypt_init_extensions.h>
#include <linux/loop.h>
#include <private/android_filesystem_config.h>
#include <cutils/android_reboot.h>
#include <cutils/partition_utils.h>
#include <cutils/properties.h>
#include <logwrap/logwrap.h>
#include "mincrypt/rsa.h"
#include "mincrypt/sha.h"
#include "mincrypt/sha256.h"
#include "ext4_utils.h"
#include "wipe.h"
#include "fs_mgr_priv.h"
#include "fs_mgr_priv_verity.h"
#define KEY_LOC_PROP "ro.crypto.keyfile.userdata"
#define KEY_IN_FOOTER "footer"
#define E2FSCK_BIN "/system/bin/e2fsck"
#define F2FS_FSCK_BIN "/system/bin/fsck.f2fs"
#define MKSWAP_BIN "/system/bin/mkswap"
#define FSCK_LOG_FILE "/dev/fscklogs/log"
#define ZRAM_CONF_DEV "/sys/block/zram0/disksize"
#define ARRAY_SIZE(a) (sizeof(a) / sizeof(*(a)))
/*
* gettime() - returns the time in seconds of the system's monotonic clock or
* zero on error.
*/
static time_t gettime(void)
{
struct timespec ts;
int ret;
ret = clock_gettime(CLOCK_MONOTONIC, &ts);
if (ret < 0) {
ERROR("clock_gettime(CLOCK_MONOTONIC) failed: %s\n", strerror(errno));
return 0;
}
return ts.tv_sec;
}
static int wait_for_file(const char *filename, int timeout)
{
struct stat info;
time_t timeout_time = gettime() + timeout;
int ret = -1;
while (gettime() < timeout_time && ((ret = stat(filename, &info)) < 0))
usleep(10000);
return ret;
}
static void check_fs(char *blk_device, char *fs_type, char *target)
{
int status;
int ret;
long tmpmnt_flags = MS_NOATIME | MS_NOEXEC | MS_NOSUID;
char tmpmnt_opts[64] = "errors=remount-ro";
char *e2fsck_argv[] = {
E2FSCK_BIN,
"-f",
"-y",
blk_device
};
/* Check for the types of filesystems we know how to check */
if (!strcmp(fs_type, "ext2") || !strcmp(fs_type, "ext3") || !strcmp(fs_type, "ext4")) {
/*
* First try to mount and unmount the filesystem. We do this because
* the kernel is more efficient than e2fsck in running the journal and
* processing orphaned inodes, and on at least one device with a
* performance issue in the emmc firmware, it can take e2fsck 2.5 minutes
* to do what the kernel does in about a second.
*
* After mounting and unmounting the filesystem, run e2fsck, and if an
* error is recorded in the filesystem superblock, e2fsck will do a full
* check. Otherwise, it does nothing. If the kernel cannot mount the
* filesytsem due to an error, e2fsck is still run to do a full check
* fix the filesystem.
*/
errno = 0;
if (!strcmp(fs_type, "ext4")) {
// This option is only valid with ext4
strlcat(tmpmnt_opts, ",nomblk_io_submit", sizeof(tmpmnt_opts));
}
ret = mount(blk_device, target, fs_type, tmpmnt_flags, tmpmnt_opts);
INFO("%s(): mount(%s,%s,%s)=%d: %s\n",
__func__, blk_device, target, fs_type, ret, strerror(errno));
if (!ret) {
int i;
for (i = 0; i < 5; i++) {
// Try to umount 5 times before continuing on.
// Should we try rebooting if all attempts fail?
int result = umount(target);
if (result == 0) {
INFO("%s(): unmount(%s) succeeded\n", __func__, target);
break;
}
ERROR("%s(): umount(%s)=%d: %s\n", __func__, target, result, strerror(errno));
sleep(1);
}
}
/*
* Some system images do not have e2fsck for licensing reasons
* (e.g. recent SDK system images). Detect these and skip the check.
*/
if (access(E2FSCK_BIN, X_OK)) {
INFO("Not running %s on %s (executable not in system image)\n",
E2FSCK_BIN, blk_device);
} else {
INFO("Running %s on %s\n", E2FSCK_BIN, blk_device);
ret = android_fork_execvp_ext(ARRAY_SIZE(e2fsck_argv), e2fsck_argv,
&status, true, LOG_KLOG | LOG_FILE,
true, FSCK_LOG_FILE, NULL, 0);
if (ret < 0) {
/* No need to check for error in fork, we can't really handle it now */
ERROR("Failed trying to run %s\n", E2FSCK_BIN);
}
}
} else if (!strcmp(fs_type, "f2fs")) {
char *f2fs_fsck_argv[] = {
F2FS_FSCK_BIN,
"-a",
blk_device
};
INFO("Running %s -a %s\n", F2FS_FSCK_BIN, blk_device);
ret = android_fork_execvp_ext(ARRAY_SIZE(f2fs_fsck_argv), f2fs_fsck_argv,
&status, true, LOG_KLOG | LOG_FILE,
true, FSCK_LOG_FILE, NULL, 0);
if (ret < 0) {
/* No need to check for error in fork, we can't really handle it now */
ERROR("Failed trying to run %s\n", F2FS_FSCK_BIN);
}
}
return;
}
static void remove_trailing_slashes(char *n)
{
int len;
len = strlen(n) - 1;
while ((*(n + len) == '/') && len) {
*(n + len) = '\0';
len--;
}
}
/*
* Mark the given block device as read-only, using the BLKROSET ioctl.
* Return 0 on success, and -1 on error.
*/
int fs_mgr_set_blk_ro(const char *blockdev)
{
int fd;
int rc = -1;
int ON = 1;
fd = TEMP_FAILURE_RETRY(open(blockdev, O_RDONLY | O_CLOEXEC));
if (fd < 0) {
// should never happen
return rc;
}
rc = ioctl(fd, BLKROSET, &ON);
close(fd);
return rc;
}
/*
* __mount(): wrapper around the mount() system call which also
* sets the underlying block device to read-only if the mount is read-only.
* See "man 2 mount" for return values.
*/
static int __mount(const char *source, const char *target, const struct fstab_rec *rec)
{
unsigned long mountflags = rec->flags;
int ret;
int save_errno;
/* We need this because sometimes we have legacy symlinks
* that are lingering around and need cleaning up.
*/
struct stat info;
if (!lstat(target, &info))
if ((info.st_mode & S_IFMT) == S_IFLNK)
unlink(target);
mkdir(target, 0755);
ret = mount(source, target, rec->fs_type, mountflags, rec->fs_options);
save_errno = errno;
INFO("%s(source=%s,target=%s,type=%s)=%d\n", __func__, source, target, rec->fs_type, ret);
if ((ret == 0) && (mountflags & MS_RDONLY) != 0) {
fs_mgr_set_blk_ro(source);
}
errno = save_errno;
return ret;
}
static int fs_match(char *in1, char *in2)
{
char *n1;
char *n2;
int ret;
n1 = strdup(in1);
n2 = strdup(in2);
remove_trailing_slashes(n1);
remove_trailing_slashes(n2);
ret = !strcmp(n1, n2);
free(n1);
free(n2);
return ret;
}
static int device_is_debuggable() {
int ret = -1;
char value[PROP_VALUE_MAX];
ret = __system_property_get("ro.debuggable", value);
if (ret < 0)
return ret;
return strcmp(value, "1") ? 0 : 1;
}
static int device_is_secure() {
int ret = -1;
char value[PROP_VALUE_MAX];
ret = __system_property_get("ro.secure", value);
/* If error, we want to fail secure */
if (ret < 0)
return 1;
return strcmp(value, "0") ? 1 : 0;
}
static int device_is_force_encrypted() {
int ret = -1;
char value[PROP_VALUE_MAX];
ret = __system_property_get("ro.vold.forceencryption", value);
if (ret < 0)
return 0;
return strcmp(value, "1") ? 0 : 1;
}
/*
* Tries to mount any of the consecutive fstab entries that match
* the mountpoint of the one given by fstab->recs[start_idx].
*
* end_idx: On return, will be the last rec that was looked at.
* attempted_idx: On return, will indicate which fstab rec
* succeeded. In case of failure, it will be the start_idx.
* Returns
* -1 on failure with errno set to match the 1st mount failure.
* 0 on success.
*/
static int mount_with_alternatives(struct fstab *fstab, int start_idx, int *end_idx, int *attempted_idx)
{
int i;
int mount_errno = 0;
int mounted = 0;
if (!end_idx || !attempted_idx || start_idx >= fstab->num_entries) {
errno = EINVAL;
if (end_idx) *end_idx = start_idx;
if (attempted_idx) *end_idx = start_idx;
return -1;
}
/* Hunt down an fstab entry for the same mount point that might succeed */
for (i = start_idx;
/* We required that fstab entries for the same mountpoint be consecutive */
i < fstab->num_entries && !strcmp(fstab->recs[start_idx].mount_point, fstab->recs[i].mount_point);
i++) {
/*
* Don't try to mount/encrypt the same mount point again.
* Deal with alternate entries for the same point which are required to be all following
* each other.
*/
if (mounted) {
ERROR("%s(): skipping fstab dup mountpoint=%s rec[%d].fs_type=%s already mounted as %s.\n", __func__,
fstab->recs[i].mount_point, i, fstab->recs[i].fs_type, fstab->recs[*attempted_idx].fs_type);
continue;
}
if (fstab->recs[i].fs_mgr_flags & MF_CHECK) {
check_fs(fstab->recs[i].blk_device, fstab->recs[i].fs_type,
fstab->recs[i].mount_point);
}
if (!__mount(fstab->recs[i].blk_device, fstab->recs[i].mount_point, &fstab->recs[i])) {
*attempted_idx = i;
mounted = 1;
if (i != start_idx) {
ERROR("%s(): Mounted %s on %s with fs_type=%s instead of %s\n", __func__,
fstab->recs[i].blk_device, fstab->recs[i].mount_point, fstab->recs[i].fs_type,
fstab->recs[start_idx].fs_type);
}
} else {
/* back up errno for crypto decisions */
mount_errno = errno;
}
}
/* Adjust i for the case where it was still withing the recs[] */
if (i < fstab->num_entries) --i;
*end_idx = i;
if (!mounted) {
*attempted_idx = start_idx;
errno = mount_errno;
return -1;
}
return 0;
}
static int translate_ext_labels(struct fstab_rec *rec)
{
DIR *blockdir = NULL;
struct dirent *ent;
char *label;
size_t label_len;
int ret = -1;
if (strncmp(rec->blk_device, "LABEL=", 6))
return 0;
label = rec->blk_device + 6;
label_len = strlen(label);
if (label_len > 16) {
ERROR("FS label is longer than allowed by filesystem\n");
goto out;
}
blockdir = opendir("/dev/block");
if (!blockdir) {
ERROR("couldn't open /dev/block\n");
goto out;
}
while ((ent = readdir(blockdir))) {
int fd;
char super_buf[1024];
struct ext4_super_block *sb;
if (ent->d_type != DT_BLK)
continue;
fd = openat(dirfd(blockdir), ent->d_name, O_RDONLY);
if (fd < 0) {
ERROR("Cannot open block device /dev/block/%s\n", ent->d_name);
goto out;
}
if (TEMP_FAILURE_RETRY(lseek(fd, 1024, SEEK_SET)) < 0 ||
TEMP_FAILURE_RETRY(read(fd, super_buf, 1024)) != 1024) {
/* Probably a loopback device or something else without a readable
* superblock.
*/
close(fd);
continue;
}
sb = (struct ext4_super_block *)super_buf;
if (sb->s_magic != EXT4_SUPER_MAGIC) {
INFO("/dev/block/%s not ext{234}\n", ent->d_name);
continue;
}
if (!strncmp(label, sb->s_volume_name, label_len)) {
char *new_blk_device;
if (asprintf(&new_blk_device, "/dev/block/%s", ent->d_name) < 0) {
ERROR("Could not allocate block device string\n");
goto out;
}
INFO("resolved label %s to %s\n", rec->blk_device, new_blk_device);
free(rec->blk_device);
rec->blk_device = new_blk_device;
ret = 0;
break;
}
}
out:
closedir(blockdir);
return ret;
}
static bool needs_block_encryption(const struct fstab_rec* rec)
{
if (device_is_force_encrypted() && fs_mgr_is_encryptable(rec)) return true;
if (rec->fs_mgr_flags & MF_FORCECRYPT) return true;
if (rec->fs_mgr_flags & MF_CRYPT) {
/* Check for existence of convert_fde breadcrumb file */
char convert_fde_name[PATH_MAX];
snprintf(convert_fde_name, sizeof(convert_fde_name),
"%s/misc/vold/convert_fde", rec->mount_point);
if (access(convert_fde_name, F_OK) == 0) return true;
}
if (rec->fs_mgr_flags & MF_FORCEFDEORFBE) {
/* Check for absence of convert_fbe breadcrumb file */
char convert_fbe_name[PATH_MAX];
snprintf(convert_fbe_name, sizeof(convert_fbe_name),
"%s/convert_fbe", rec->mount_point);
if (access(convert_fbe_name, F_OK) != 0) return true;
}
return false;
}
// Check to see if a mountable volume has encryption requirements
static int handle_encryptable(const struct fstab_rec* rec)
{
/* If this is block encryptable, need to trigger encryption */
if (needs_block_encryption(rec)) {
if (umount(rec->mount_point) == 0) {
return FS_MGR_MNTALL_DEV_NEEDS_ENCRYPTION;
} else {
WARNING("Could not umount %s (%s) - allow continue unencrypted\n",
rec->mount_point, strerror(errno));
return FS_MGR_MNTALL_DEV_NOT_ENCRYPTED;
}
} else if (rec->fs_mgr_flags & (MF_FILEENCRYPTION | MF_FORCEFDEORFBE)) {
// Deal with file level encryption
INFO("%s is file encrypted\n", rec->mount_point);
return FS_MGR_MNTALL_DEV_FILE_ENCRYPTED;
} else if (fs_mgr_is_encryptable(rec)) {
return FS_MGR_MNTALL_DEV_NOT_ENCRYPTED;
} else {
return FS_MGR_MNTALL_DEV_NOT_ENCRYPTABLE;
}
}
/* When multiple fstab records share the same mount_point, it will
* try to mount each one in turn, and ignore any duplicates after a
* first successful mount.
* Returns -1 on error, and FS_MGR_MNTALL_* otherwise.
*/
int fs_mgr_mount_all(struct fstab *fstab, int mount_mode)
{
int i = 0;
int encryptable = FS_MGR_MNTALL_DEV_NOT_ENCRYPTABLE;
int error_count = 0;
int mret = -1;
int mount_errno = 0;
int attempted_idx = -1;
if (!fstab) {
return -1;
}
for (i = 0; i < fstab->num_entries; i++) {
/* Don't mount entries that are managed by vold or not for the mount mode*/
if ((fstab->recs[i].fs_mgr_flags & (MF_VOLDMANAGED | MF_RECOVERYONLY)) ||
((mount_mode == MOUNT_MODE_LATE) && !fs_mgr_is_latemount(&fstab->recs[i])) ||
((mount_mode == MOUNT_MODE_EARLY) && fs_mgr_is_latemount(&fstab->recs[i]))) {
continue;
}
/* Skip swap and raw partition entries such as boot, recovery, etc */
if (!strcmp(fstab->recs[i].fs_type, "swap") ||
!strcmp(fstab->recs[i].fs_type, "emmc") ||
!strcmp(fstab->recs[i].fs_type, "mtd")) {
continue;
}
/* Skip mounting the root partition, as it will already have been mounted */
if (!strcmp(fstab->recs[i].mount_point, "/")) {
if ((fstab->recs[i].fs_mgr_flags & MS_RDONLY) != 0) {
fs_mgr_set_blk_ro(fstab->recs[i].blk_device);
}
continue;
}
/* Translate LABEL= file system labels into block devices */
if (!strcmp(fstab->recs[i].fs_type, "ext2") ||
!strcmp(fstab->recs[i].fs_type, "ext3") ||
!strcmp(fstab->recs[i].fs_type, "ext4")) {
int tret = translate_ext_labels(&fstab->recs[i]);
if (tret < 0) {
ERROR("Could not translate label to block device\n");
continue;
}
}
if (fstab->recs[i].fs_mgr_flags & MF_WAIT) {
wait_for_file(fstab->recs[i].blk_device, WAIT_TIMEOUT);
}
if ((fstab->recs[i].fs_mgr_flags & MF_VERIFY) && device_is_secure()) {
int rc = fs_mgr_setup_verity(&fstab->recs[i]);
if (device_is_debuggable() && rc == FS_MGR_SETUP_VERITY_DISABLED) {
INFO("Verity disabled");
} else if (rc != FS_MGR_SETUP_VERITY_SUCCESS) {
ERROR("Could not set up verified partition, skipping!\n");
continue;
}
}
int last_idx_inspected;
int top_idx = i;
mret = mount_with_alternatives(fstab, i, &last_idx_inspected, &attempted_idx);
i = last_idx_inspected;
mount_errno = errno;
/* Deal with encryptability. */
if (!mret) {
int status = handle_encryptable(&fstab->recs[attempted_idx]);
if (status == FS_MGR_MNTALL_FAIL) {
/* Fatal error - no point continuing */
return status;
}
if (status != FS_MGR_MNTALL_DEV_NOT_ENCRYPTABLE) {
if (encryptable != FS_MGR_MNTALL_DEV_NOT_ENCRYPTABLE) {
// Log and continue
ERROR("Only one encryptable/encrypted partition supported\n");
}
encryptable = status;
}
/* Success! Go get the next one */
continue;
}
/* mount(2) returned an error, handle the encryptable/formattable case */
bool wiped = partition_wiped(fstab->recs[top_idx].blk_device);
if (mret && mount_errno != EBUSY && mount_errno != EACCES &&
fs_mgr_is_formattable(&fstab->recs[top_idx]) && wiped) {
/* top_idx and attempted_idx point at the same partition, but sometimes
* at two different lines in the fstab. Use the top one for formatting
* as that is the preferred one.
*/
ERROR("%s(): %s is wiped and %s %s is formattable. Format it.\n", __func__,
fstab->recs[top_idx].blk_device, fstab->recs[top_idx].mount_point,
fstab->recs[top_idx].fs_type);
if (fs_mgr_is_encryptable(&fstab->recs[top_idx]) &&
strcmp(fstab->recs[top_idx].key_loc, KEY_IN_FOOTER)) {
int fd = open(fstab->recs[top_idx].key_loc, O_WRONLY, 0644);
if (fd >= 0) {
INFO("%s(): also wipe %s\n", __func__, fstab->recs[top_idx].key_loc);
wipe_block_device(fd, get_file_size(fd));
close(fd);
} else {
ERROR("%s(): %s wouldn't open (%s)\n", __func__,
fstab->recs[top_idx].key_loc, strerror(errno));
}
}
if (fs_mgr_do_format(&fstab->recs[top_idx]) == 0) {
/* Let's replay the mount actions. */
i = top_idx - 1;
continue;
} else {
ERROR("%s(): Format failed. Suggest recovery...\n", __func__);
encryptable = FS_MGR_MNTALL_DEV_NEEDS_RECOVERY;
continue;
}
}
if (mret && mount_errno != EBUSY && mount_errno != EACCES &&
fs_mgr_is_encryptable(&fstab->recs[attempted_idx])) {
if (wiped) {
ERROR("%s(): %s is wiped and %s %s is encryptable. Suggest recovery...\n", __func__,
fstab->recs[attempted_idx].blk_device, fstab->recs[attempted_idx].mount_point,
fstab->recs[attempted_idx].fs_type);
encryptable = FS_MGR_MNTALL_DEV_NEEDS_RECOVERY;
continue;
} else {
/* Need to mount a tmpfs at this mountpoint for now, and set
* properties that vold will query later for decrypting
*/
ERROR("%s(): possibly an encryptable blkdev %s for mount %s type %s )\n", __func__,
fstab->recs[attempted_idx].blk_device, fstab->recs[attempted_idx].mount_point,
fstab->recs[attempted_idx].fs_type);
if (fs_mgr_do_tmpfs_mount(fstab->recs[attempted_idx].mount_point) < 0) {
++error_count;
continue;
}
}
encryptable = FS_MGR_MNTALL_DEV_MIGHT_BE_ENCRYPTED;
} else {
if (fs_mgr_is_nofail(&fstab->recs[attempted_idx])) {
ERROR("Ignoring failure to mount an un-encryptable or wiped partition on"
"%s at %s options: %s error: %s\n",
fstab->recs[attempted_idx].blk_device, fstab->recs[attempted_idx].mount_point,
fstab->recs[attempted_idx].fs_options, strerror(mount_errno));
} else {
ERROR("Failed to mount an un-encryptable or wiped partition on"
"%s at %s options: %s error: %s\n",
fstab->recs[attempted_idx].blk_device, fstab->recs[attempted_idx].mount_point,
fstab->recs[attempted_idx].fs_options, strerror(mount_errno));
++error_count;
}
continue;
}
}
if (error_count) {
return -1;
} else {
return encryptable;
}
}
/* If tmp_mount_point is non-null, mount the filesystem there. This is for the
* tmp mount we do to check the user password
* If multiple fstab entries are to be mounted on "n_name", it will try to mount each one
* in turn, and stop on 1st success, or no more match.
*/
int fs_mgr_do_mount(struct fstab *fstab, char *n_name, char *n_blk_device,
char *tmp_mount_point)
{
int i = 0;
int ret = FS_MGR_DOMNT_FAILED;
int mount_errors = 0;
int first_mount_errno = 0;
char *m;
if (!fstab) {
return ret;
}
for (i = 0; i < fstab->num_entries; i++) {
if (!fs_match(fstab->recs[i].mount_point, n_name)) {
continue;
}
/* We found our match */
/* If this swap or a raw partition, report an error */
if (!strcmp(fstab->recs[i].fs_type, "swap") ||
!strcmp(fstab->recs[i].fs_type, "emmc") ||
!strcmp(fstab->recs[i].fs_type, "mtd")) {
ERROR("Cannot mount filesystem of type %s on %s\n",
fstab->recs[i].fs_type, n_blk_device);
goto out;
}
/* First check the filesystem if requested */
if (fstab->recs[i].fs_mgr_flags & MF_WAIT) {
wait_for_file(n_blk_device, WAIT_TIMEOUT);
}
if (fstab->recs[i].fs_mgr_flags & MF_CHECK) {
check_fs(n_blk_device, fstab->recs[i].fs_type,
fstab->recs[i].mount_point);
}
if ((fstab->recs[i].fs_mgr_flags & MF_VERIFY) && device_is_secure()) {
int rc = fs_mgr_setup_verity(&fstab->recs[i]);
if (device_is_debuggable() && rc == FS_MGR_SETUP_VERITY_DISABLED) {
INFO("Verity disabled");
} else if (rc != FS_MGR_SETUP_VERITY_SUCCESS) {
ERROR("Could not set up verified partition, skipping!\n");
continue;
}
}
/* Now mount it where requested */
if (tmp_mount_point) {
m = tmp_mount_point;
} else {
m = fstab->recs[i].mount_point;
}
if (__mount(n_blk_device, m, &fstab->recs[i])) {
if (!first_mount_errno) first_mount_errno = errno;
mount_errors++;
continue;
} else {
ret = 0;
goto out;
}
}
if (mount_errors) {
ERROR("Cannot mount filesystem on %s at %s. error: %s\n",
n_blk_device, m, strerror(first_mount_errno));
if (first_mount_errno == EBUSY) {
ret = FS_MGR_DOMNT_BUSY;
} else {
ret = FS_MGR_DOMNT_FAILED;
}
} else {
/* We didn't find a match, say so and return an error */
ERROR("Cannot find mount point %s in fstab\n", fstab->recs[i].mount_point);
}
out:
return ret;
}
/*
* mount a tmpfs filesystem at the given point.
* return 0 on success, non-zero on failure.
*/
int fs_mgr_do_tmpfs_mount(char *n_name)
{
int ret;
ret = mount("tmpfs", n_name, "tmpfs",
MS_NOATIME | MS_NOSUID | MS_NODEV, CRYPTO_TMPFS_OPTIONS);
if (ret < 0) {
ERROR("Cannot mount tmpfs filesystem at %s\n", n_name);
return -1;
}
/* Success */
return 0;
}
int fs_mgr_unmount_all(struct fstab *fstab)
{
int i = 0;
int ret = 0;
if (!fstab) {
return -1;
}
while (fstab->recs[i].blk_device) {
if (umount(fstab->recs[i].mount_point)) {
ERROR("Cannot unmount filesystem at %s\n", fstab->recs[i].mount_point);
ret = -1;
}
i++;
}
return ret;
}
/* This must be called after mount_all, because the mkswap command needs to be
* available.
*/
int fs_mgr_swapon_all(struct fstab *fstab)
{
int i = 0;
int flags = 0;
int err = 0;
int ret = 0;
int status;
char *mkswap_argv[2] = {
MKSWAP_BIN,
NULL
};
if (!fstab) {
return -1;
}
for (i = 0; i < fstab->num_entries; i++) {
/* Skip non-swap entries */
if (strcmp(fstab->recs[i].fs_type, "swap")) {
continue;
}
if (fstab->recs[i].zram_size > 0) {
/* A zram_size was specified, so we need to configure the
* device. There is no point in having multiple zram devices
* on a system (all the memory comes from the same pool) so
* we can assume the device number is 0.
*/
FILE *zram_fp;
zram_fp = fopen(ZRAM_CONF_DEV, "r+");
if (zram_fp == NULL) {
ERROR("Unable to open zram conf device %s\n", ZRAM_CONF_DEV);
ret = -1;
continue;
}
fprintf(zram_fp, "%d\n", fstab->recs[i].zram_size);
fclose(zram_fp);
}
if (fstab->recs[i].fs_mgr_flags & MF_WAIT) {
wait_for_file(fstab->recs[i].blk_device, WAIT_TIMEOUT);
}
/* Initialize the swap area */
mkswap_argv[1] = fstab->recs[i].blk_device;
err = android_fork_execvp_ext(ARRAY_SIZE(mkswap_argv), mkswap_argv,
&status, true, LOG_KLOG, false, NULL,
NULL, 0);
if (err) {
ERROR("mkswap failed for %s\n", fstab->recs[i].blk_device);
ret = -1;
continue;
}
/* If -1, then no priority was specified in fstab, so don't set
* SWAP_FLAG_PREFER or encode the priority */
if (fstab->recs[i].swap_prio >= 0) {
flags = (fstab->recs[i].swap_prio << SWAP_FLAG_PRIO_SHIFT) &
SWAP_FLAG_PRIO_MASK;
flags |= SWAP_FLAG_PREFER;
} else {
flags = 0;
}
err = swapon(fstab->recs[i].blk_device, flags);
if (err) {
ERROR("swapon failed for %s\n", fstab->recs[i].blk_device);
ret = -1;
}
}
return ret;
}
/*
* key_loc must be at least PROPERTY_VALUE_MAX bytes long
*
* real_blk_device must be at least PROPERTY_VALUE_MAX bytes long
*/
int fs_mgr_get_crypt_info(struct fstab *fstab, char *key_loc, char *real_blk_device, int size)
{
int i = 0;
if (!fstab) {
return -1;
}
/* Initialize return values to null strings */
if (key_loc) {
*key_loc = '\0';
}
if (real_blk_device) {
*real_blk_device = '\0';
}
/* Look for the encryptable partition to find the data */
for (i = 0; i < fstab->num_entries; i++) {
/* Don't deal with vold managed enryptable partitions here */
if (fstab->recs[i].fs_mgr_flags & MF_VOLDMANAGED) {
continue;
}
if (!(fstab->recs[i].fs_mgr_flags
& (MF_CRYPT | MF_FORCECRYPT | MF_FORCEFDEORFBE))) {
continue;
}
/* We found a match */
if (key_loc) {
strlcpy(key_loc, fstab->recs[i].key_loc, size);
}
if (real_blk_device) {
strlcpy(real_blk_device, fstab->recs[i].blk_device, size);
}
break;
}
return 0;
}