fs_mgr/fs_mgr.c

/*
 * 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.
 */

/* TO DO:
 *   1. Re-direct fsck output to the kernel log?
 *
 */

#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 <private/android_filesystem_config.h>
#include <cutils/partition_utils.h>
#include <cutils/properties.h>

#include "fs_mgr_priv.h"

#define KEY_LOC_PROP   "ro.crypto.keyfile.userdata"
#define KEY_IN_FOOTER  "footer"

#define E2FSCK_BIN      "/system/bin/e2fsck"

struct flag_list {
    const char *name;
    unsigned flag;
};

static struct flag_list mount_flags[] = {
    { "noatime",    MS_NOATIME },
    { "noexec",     MS_NOEXEC },
    { "nosuid",     MS_NOSUID },
    { "nodev",      MS_NODEV },
    { "nodiratime", MS_NODIRATIME },
    { "ro",         MS_RDONLY },
    { "rw",         0 },
    { "remount",    MS_REMOUNT },
    { "bind",       MS_BIND },
    { "rec",        MS_REC },
    { "unbindable", MS_UNBINDABLE },
    { "private",    MS_PRIVATE },
    { "slave",      MS_SLAVE },
    { "shared",     MS_SHARED },
    { "defaults",   0 },
    { 0,            0 },
};

static struct flag_list fs_mgr_flags[] = {
    { "wait",        MF_WAIT },
    { "check",       MF_CHECK },
    { "encryptable=",MF_CRYPT },
    { "nonremovable",MF_NONREMOVABLE },
    { "voldmanaged=",MF_VOLDMANAGED},
    { "length=",     MF_LENGTH },
    { "recoveryonly",MF_RECOVERYONLY },
    { "defaults",    0 },
    { 0,             0 },
};

/*
 * 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 int parse_flags(char *flags, struct flag_list *fl,
                       char **key_loc, long long *part_length, char **label, int *partnum,
                       char *fs_options, int fs_options_len)
{
    int f = 0;
    int i;
    char *p;
    char *savep;

    /* initialize key_loc to null, if we find an MF_CRYPT flag,
     * then we'll set key_loc to the proper value */
    if (key_loc) {
        *key_loc = NULL;
    }
    /* initialize part_length to 0, if we find an MF_LENGTH flag,
     * then we'll set part_length to the proper value */
    if (part_length) {
        *part_length = 0;
    }
    if (partnum) {
        *partnum = -1;
    }
    if (label) {
        *label = NULL;
    }

    /* initialize fs_options to the null string */
    if (fs_options && (fs_options_len > 0)) {
        fs_options[0] = '\0';
    }

    p = strtok_r(flags, ",", &savep);
    while (p) {
        /* Look for the flag "p" in the flag list "fl"
         * If not found, the loop exits with fl[i].name being null.
         */
        for (i = 0; fl[i].name; i++) {
            if (!strncmp(p, fl[i].name, strlen(fl[i].name))) {
                f |= fl[i].flag;
                if ((fl[i].flag == MF_CRYPT) && key_loc) {
                    /* The encryptable flag is followed by an = and the
                     * location of the keys.  Get it and return it.
                     */
                    *key_loc = strdup(strchr(p, '=') + 1);
                } else if ((fl[i].flag == MF_LENGTH) && part_length) {
                    /* The length flag is followed by an = and the
                     * size of the partition.  Get it and return it.
                     */
                    *part_length = strtoll(strchr(p, '=') + 1, NULL, 0);
                } else if ((fl[i].flag == MF_VOLDMANAGED) && label && partnum) {
                    /* The voldmanaged flag is followed by an = and the
                     * label, a colon and the partition number or the
                     * word "auto", e.g.
                     *   voldmanaged=sdcard:3
                     * Get and return them.
                     */
                    char *label_start;
                    char *label_end;
                    char *part_start;

                    label_start = strchr(p, '=') + 1;
                    label_end = strchr(p, ':');
                    if (label_end) {
                        *label = strndup(label_start,
                                         (int) (label_end - label_start));
                        part_start = strchr(p, ':') + 1;
                        if (!strcmp(part_start, "auto")) {
                            *partnum = -1;
                        } else {
                            *partnum = strtol(part_start, NULL, 0);
                        }
                    } else {
                        ERROR("Warning: voldmanaged= flag malformed\n");
                    }
                }
                break;
            }
        }

        if (!fl[i].name) {
            if (fs_options) {
                /* It's not a known flag, so it must be a filesystem specific
                 * option.  Add it to fs_options if it was passed in.
                 */
                strlcat(fs_options, p, fs_options_len);
                strlcat(fs_options, ",", fs_options_len);
            } else {
                /* fs_options was not passed in, so if the flag is unknown
                 * it's an error.
                 */
                ERROR("Warning: unknown flag %s\n", p);
            }
        }
        p = strtok_r(NULL, ",", &savep);
    }

out:
    if (fs_options && fs_options[0]) {
        /* remove the last trailing comma from the list of options */
        fs_options[strlen(fs_options) - 1] = '\0';
    }

    return f;
}

/* Read a line of text till the next newline character.
 * If no newline is found before the buffer is full, continue reading till a new line is seen,
 * then return an empty buffer.  This effectively ignores lines that are too long.
 * On EOF, return null.
 */
static char *fs_getline(char *buf, int size, FILE *file)
{
    int cnt = 0;
    int eof = 0;
    int eol = 0;
    int c;

    if (size < 1) {
        return NULL;
    }

    while (cnt < (size - 1)) {
        c = getc(file);
        if (c == EOF) {
            eof = 1;
            break;
        }

        *(buf + cnt) = c;
        cnt++;

        if (c == '\n') {
            eol = 1;
            break;
        }
    }

    /* Null terminate what we've read */
    *(buf + cnt) = '\0';

    if (eof) {
        if (cnt) {
            return buf;
        } else {
            return NULL;
        }
    } else if (eol) {
        return buf;
    } else {
        /* The line is too long.  Read till a newline or EOF.
         * If EOF, return null, if newline, return an empty buffer.
         */
        while(1) {
            c = getc(file);
            if (c == EOF) {
                return NULL;
            } else if (c == '\n') {
                *buf = '\0';
                return buf;
            }
        }
    }
}

struct fstab *fs_mgr_read_fstab(const char *fstab_path)
{
    FILE *fstab_file;
    int cnt, entries;
    int len;
    char line[256];
    const char *delim = " \t";
    char *save_ptr, *p;
    struct fstab *fstab;
    struct fstab_rec *recs;
    char *key_loc;
    long long part_length;
    char *label;
    int partnum;
#define FS_OPTIONS_LEN 1024
    char tmp_fs_options[FS_OPTIONS_LEN];

    fstab_file = fopen(fstab_path, "r");
    if (!fstab_file) {
        ERROR("Cannot open file %s\n", fstab_path);
        return 0;
    }

    entries = 0;
    while (fs_getline(line, sizeof(line), fstab_file)) {
        /* if the last character is a newline, shorten the string by 1 byte */
        len = strlen(line);
        if (line[len - 1] == '\n') {
            line[len - 1] = '\0';
        }
        /* Skip any leading whitespace */
        p = line;
        while (isspace(*p)) {
            p++;
        }
        /* ignore comments or empty lines */
        if (*p == '#' || *p == '\0')
            continue;
        entries++;
    }

    if (!entries) {
        ERROR("No entries found in fstab\n");
        return 0;
    }

    /* Allocate and init the fstab structure */
    fstab = calloc(1, sizeof(struct fstab));
    fstab->num_entries = entries;
    fstab->fstab_filename = strdup(fstab_path);
    fstab->recs = calloc(fstab->num_entries, sizeof(struct fstab_rec));

    fseek(fstab_file, 0, SEEK_SET);

    cnt = 0;
    while (fs_getline(line, sizeof(line), fstab_file)) {
        /* if the last character is a newline, shorten the string by 1 byte */
        len = strlen(line);
        if (line[len - 1] == '\n') {
            line[len - 1] = '\0';
        }

        /* Skip any leading whitespace */
        p = line;
        while (isspace(*p)) {
            p++;
        }
        /* ignore comments or empty lines */
        if (*p == '#' || *p == '\0')
            continue;

        /* If a non-comment entry is greater than the size we allocated, give an
         * error and quit.  This can happen in the unlikely case the file changes
         * between the two reads.
         */
        if (cnt >= entries) {
            ERROR("Tried to process more entries than counted\n");
            break;
        }

        if (!(p = strtok_r(line, delim, &save_ptr))) {
            ERROR("Error parsing mount source\n");
            return 0;
        }
        fstab->recs[cnt].blk_device = strdup(p);

        if (!(p = strtok_r(NULL, delim, &save_ptr))) {
            ERROR("Error parsing mount_point\n");
            return 0;
        }
        fstab->recs[cnt].mount_point = strdup(p);

        if (!(p = strtok_r(NULL, delim, &save_ptr))) {
            ERROR("Error parsing fs_type\n");
            return 0;
        }
        fstab->recs[cnt].fs_type = strdup(p);

        if (!(p = strtok_r(NULL, delim, &save_ptr))) {
            ERROR("Error parsing mount_flags\n");
            return 0;
        }
        tmp_fs_options[0] = '\0';
        fstab->recs[cnt].flags = parse_flags(p, mount_flags,
                                       NULL, NULL, NULL, NULL,
                                       tmp_fs_options, FS_OPTIONS_LEN);

        /* fs_options are optional */
        if (tmp_fs_options[0]) {
            fstab->recs[cnt].fs_options = strdup(tmp_fs_options);
        } else {
            fstab->recs[cnt].fs_options = NULL;
        }

        if (!(p = strtok_r(NULL, delim, &save_ptr))) {
            ERROR("Error parsing fs_mgr_options\n");
            return 0;
        }
        fstab->recs[cnt].fs_mgr_flags = parse_flags(p, fs_mgr_flags,
                                              &key_loc, &part_length,
                                              &label, &partnum,
                                              NULL, 0);
        fstab->recs[cnt].key_loc = key_loc;
        fstab->recs[cnt].length = part_length;
        fstab->recs[cnt].label = label;
        fstab->recs[cnt].partnum = partnum;
        cnt++;
    }
    fclose(fstab_file);

    return fstab;
}

void fs_mgr_free_fstab(struct fstab *fstab)
{
    int i;

    for (i = 0; i < fstab->num_entries; i++) {
        /* Free the pointers return by strdup(3) */
        free(fstab->recs[i].blk_device);
        free(fstab->recs[i].mount_point);
        free(fstab->recs[i].fs_type);
        free(fstab->recs[i].fs_options);
        free(fstab->recs[i].key_loc);
        free(fstab->recs[i].label);
        i++;
    }

    /* Free the fstab_recs array created by calloc(3) */
    free(fstab->recs);

    /* Free the fstab filename */
    free(fstab->fstab_filename);

    /* Free fstab */
    free(fstab);
}

static void check_fs(char *blk_device, char *fs_type, char *target)
{
    pid_t pid;
    int status;
    int ret;
    long tmpmnt_flags = MS_NOATIME | MS_NOEXEC | MS_NOSUID;
    char *tmpmnt_opts = "nomblk_io_submit,errors=remount-ro";

    /* 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.
         */
        ret = mount(blk_device, target, fs_type, tmpmnt_flags, tmpmnt_opts);
        if (!ret) {
            umount(target);
        }

        INFO("Running %s on %s\n", E2FSCK_BIN, blk_device);
        pid = fork();
        if (pid > 0) {
            /* Parent, wait for the child to return */
            waitpid(pid, &status, 0);
        } else if (pid == 0) {
            /* child, run checker */
            execlp(E2FSCK_BIN, E2FSCK_BIN, "-y", blk_device, (char *)NULL);

            /* Only gets here on error */
            ERROR("Cannot run fs_mgr binary %s\n", E2FSCK_BIN);
        } else {
            /* No need to check for error in fork, we can't really handle it now */
            ERROR("Fork failed trying to run %s\n", E2FSCK_BIN);
        }
    }

    return;
}

static void remove_trailing_slashes(char *n)
{
    int len;

    len = strlen(n) - 1;
    while ((*(n + len) == '/') && len) {
      *(n + len) = '\0';
      len--;
    }
}

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;
}

int fs_mgr_mount_all(struct fstab *fstab)
{
    int i = 0;
    int encrypted = 0;
    int ret = -1;
    int mret;

    if (!fstab) {
        return ret;
    }

    for (i = 0; i < fstab->num_entries; i++) {
        /* Don't mount entries that are managed by vold */
        if (fstab->recs[i].fs_mgr_flags & (MF_VOLDMANAGED | MF_RECOVERYONLY)) {
            continue;
        }

        /* Skip raw partition entries such as boot, recovery, etc */
        if (!strcmp(fstab->recs[i].fs_type, "emmc") ||
            !strcmp(fstab->recs[i].fs_type, "mtd")) {
            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_CHECK) {
            check_fs(fstab->recs[i].blk_device, fstab->recs[i].fs_type,
                     fstab->recs[i].mount_point);
        }

        mret = mount(fstab->recs[i].blk_device, fstab->recs[i].mount_point,
                     fstab->recs[i].fs_type, fstab->recs[i].flags,
                     fstab->recs[i].fs_options);
        if (!mret) {
            /* Success!  Go get the next one */
            continue;
        }

        /* mount(2) returned an error, check if it's encrypted and deal with it */
        if ((fstab->recs[i].fs_mgr_flags & MF_CRYPT) &&
            !partition_wiped(fstab->recs[i].blk_device)) {
            /* Need to mount a tmpfs at this mountpoint for now, and set
             * properties that vold will query later for decrypting
             */
            if (mount("tmpfs", fstab->recs[i].mount_point, "tmpfs",
                  MS_NOATIME | MS_NOSUID | MS_NODEV, CRYPTO_TMPFS_OPTIONS) < 0) {
                ERROR("Cannot mount tmpfs filesystem for encrypted fs at %s\n",
                        fstab->recs[i].mount_point);
                goto out;
            }
            encrypted = 1;
        } else {
            ERROR("Cannot mount filesystem on %s at %s\n",
                    fstab->recs[i].blk_device, fstab->recs[i].mount_point);
            goto out;
        }
    }

    if (encrypted) {
        ret = 1;
    } else {
        ret = 0;
    }

out:
    return ret;
}

/* If tmp_mount_point is non-null, mount the filesystem there.  This is for the
 * tmp mount we do to check the user password
 */
int fs_mgr_do_mount(struct fstab *fstab, char *n_name, char *n_blk_device,
                    char *tmp_mount_point)
{
    int i = 0;
    int ret = -1;
    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 is a raw partition, report an error */
        if (!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);
        }

        /* 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].fs_type,
                  fstab->recs[i].flags, fstab->recs[i].fs_options)) {
            ERROR("Cannot mount filesystem on %s at %s\n",
                    n_blk_device, m);
            goto out;
        } else {
            ret = 0;
            goto out;
        }
    }

    /* 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;
}
/*
 * 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)) {
            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;
}

/* Add an entry to the fstab, and return 0 on success or -1 on error */
int fs_mgr_add_entry(struct fstab *fstab,
                     const char *mount_point, const char *fs_type,
                     const char *blk_device, long long length)
{
    struct fstab_rec *new_fstab_recs;
    int n = fstab->num_entries;

    new_fstab_recs = (struct fstab_rec *)
                     realloc(fstab->recs, sizeof(struct fstab_rec) * (n + 1));

    if (!new_fstab_recs) {
        return -1;
    }

    /* A new entry was added, so initialize it */
     memset(&new_fstab_recs[n], 0, sizeof(struct fstab_rec));
     new_fstab_recs[n].mount_point = strdup(mount_point);
     new_fstab_recs[n].fs_type = strdup(fs_type);
     new_fstab_recs[n].blk_device = strdup(blk_device);
     new_fstab_recs[n].length = 0;

     /* Update the fstab struct */
     fstab->recs = new_fstab_recs;
     fstab->num_entries++;

     return 0;
}

struct fstab_rec *fs_mgr_get_entry_for_mount_point(struct fstab *fstab, const char *path)
{
    int i;

    if (!fstab) {
        return NULL;
    }

    for (i = 0; i < fstab->num_entries; i++) {
        int len = strlen(fstab->recs[i].mount_point);
        if (strncmp(path, fstab->recs[i].mount_point, len) == 0 &&
            (path[len] == '\0' || path[len] == '/')) {
            return &fstab->recs[i];
        }
    }

    return NULL;
}

int fs_mgr_is_voldmanaged(struct fstab_rec *fstab)
{
    return fstab->fs_mgr_flags & MF_VOLDMANAGED;
}

int fs_mgr_is_nonremovable(struct fstab_rec *fstab)
{
    return fstab->fs_mgr_flags & MF_NONREMOVABLE;
}

int fs_mgr_is_encryptable(struct fstab_rec *fstab)
{
    return fstab->fs_mgr_flags & MF_CRYPT;
}