Like Linux's /etc/fstab file, Android's boot mount location is determined in a specific fstab file. The main function of vold will call process_config to parse it. The fstab file in this example is the fstab.rk30board file in the root directory of the board.
fstab.rk30board
# Android fstab file.
#<src> <mnt_point> <type> <mnt_flags and options> <fs_mgr_flags>
# The filesystem that contains the filesystem checker binary (typically /system) cannot
# specify MF_CHECK, and must come before any filesystems that do specify MF_CHECK
/dev/block/platform/fe330000.sdhci/by-name/system /system ext4 ro,noatime,nodiratime,noauto_da_alloc wait,resize
# use this line below instead to enable verity
#/dev/block/platform/fe330000.sdhci/by-name/system /system ext4 ro,noatime,nodiratime,noauto_da_alloc wait,check,verify
/dev/block/platform/fe330000.sdhci/by-name/cache /cache ext4 noatime,nodiratime,nosuid,nodev,noauto_da_alloc,discard wait,check
/dev/block/platform/fe330000.sdhci/by-name/metadata /metadata ext4 noatime,nodiratime,nosuid,nodev,noauto_da_alloc,discard wait,check
/dev/block/platform/fe330000.sdhci/by-name/userdata /data f2fs noatime,nodiratime,nosuid,nodev,discard,inline_xattr wait,check,notrim,encryptable=/metadata/key_file
#data for f2fs nobarrier
#/dev/block/platform/fe330000.sdhci/by-name/userdata /data f2fs noatime,nodiratime,nosuid,nodev,discard,inline_xattr,nobarrier wait,check,notrim,encryptable=/metadata/key_file
#data for ext4
#/dev/block/platform/fe330000.sdhci/by-name/userdata /data ext4 noatime,nodiratime,nosuid,nodev,noauto_da_alloc,discard,errors=panic wait,check,encryptable=/metadata/key_file
# sdcard
/devices/platform/fe320000.dwmmc/mmc_host* auto auto defaults voldmanaged=sdcard1:auto,encryptable=userdata
# for usb2.0
/devices/platform/*.usb* auto vfat defaults voldmanaged=usb:auto
# for usb3.0
/devices/platform/usb@*/*.dwc3* auto vfat defaults voldmanaged=usb:auto
/dev/block/zram0 none swap defaults zramsize=533413200The DefaultFstabPath function returns the path to the above fstab file. fs_mgr_read_fstab is the main function of parsing. The parsing result is stored in the structure struct fstab. First enter fs_mgr_read_fstab to see.
/system/vold/main.cpp
static int process_config(VolumeManager *vm) {
std::string path(android::vold::DefaultFstabPath());
fstab = fs_mgr_read_fstab(path.c_str());
if (!fstab) {
PLOG(ERROR) << "Failed to open default fstab " << path;
return -1;
}
/* Loop through entries looking for ones that vold manages */
bool has_adoptable = false;
for (int i = 0; i < fstab->num_entries; i++) {
if (fs_mgr_is_voldmanaged(&fstab->recs[i])) {
if (fs_mgr_is_nonremovable(&fstab->recs[i])) {
LOG(WARNING) << "nonremovable no longer supported; ignoring volume";
continue;
}
std::string sysPattern(fstab->recs[i].blk_device);
std::string nickname(fstab->recs[i].label);
int flags = 0;
if (fs_mgr_is_encryptable(&fstab->recs[i])) {
flags |= android::vold::Disk::Flags::kAdoptable;
has_adoptable = true;
}
if (fs_mgr_is_noemulatedsd(&fstab->recs[i])
|| property_get_bool("vold.debug.default_primary", false)) {
flags |= android::vold::Disk::Flags::kDefaultPrimary;
}
vm->addDiskSource(std::shared_ptr<VolumeManager::DiskSource>(
new VolumeManager::DiskSource(sysPattern, nickname, flags)));
}
}
property_set("vold.has_adoptable", has_adoptable ? "1" : "0");
return 0;
}
The value of entries indicates the number of actual mount information lines in the fstab file. After getting the value of entries, it starts to initialize the struct fstab structure. The num_entries member of struct fstab is initialized to the value of entries, fstab_filename is initialized to a string represented by the path of the fstab file, and recs is initialized to a pointer to the struct fstab_rec array with num_entries of elements.
Next, traverse each mount information line and initialize the corresponding struct fstab_rec elements one by one. As shown in the fstab file, the contents of < src>, < mnt_point>, < type>, < mnt_flags and options>, < fs_mgr_flags> tags are used to initialize the member variables of struct fstab_rec respectively. The initialization of blk_device, mount_point, and fs_type is relatively simple, just a direct string copy. The flags and fs_mgr_flags members can only be determined after processing the corresponding label content by the parse_flags function.
/system/core/fs_mgr/include/fs_mgr.h
struct fstab {
int num_entries;
struct fstab_rec *recs;
char *fstab_filename;
};/system/core/fs_mgr/include/fs_mgr.h
struct fstab_rec {
char *blk_device;
char *mount_point;
char *fs_type;
unsigned long flags;
char *fs_options;
int fs_mgr_flags;
char *key_loc;
char *verity_loc;
long long length;
char *label;
int partnum;
int swap_prio;
unsigned int zram_size;
}; Take a look at the parse_flags function. Considering that the initialization of different struct fstab_rec members requires different parameter implementations for parse_flags. The following is a case-by-case discussion:
first, let's talk about the initialization of flags and fs_options members. For each comma-separated name of the <mnt_flags and options> tag item, there will be a corresponding flag value in the struct flag_list structure array of mount_flags. If not, it means that the name is unique to the file system. For names that are not unique to the file system, the parse_flags function will OR the flags values corresponding to these names and return them. For filesystem-specific names, the processing method is to copy them into the input parameter fs_options, separated by commas. After the parse_flags function returns to the fs_mgr_read_fstab function, the return value of the parse_flags function is used to initialize the flags member of struct fstab_rec, and the file system-specific options saved in tmp_fs_options are used to initialize the fs_options member.
Let's talk about the initialization of the remaining other struct fstab_rec members. This time, the <fs_mgr_flags> tag item is parsed, and the basis for searching according to the name is the struct flag_list structure array of fs_mgr_flags. Same as the previous paragraph, on the one hand, all the comma-separated names in the <fs_mgr_flags> tag item are used to set flags, on the other hand, some names are used to initialize the input flag_vals, see code for details. After the parse_flags function returns to the fs_mgr_read_fstab function, the returned flags value is used to initialize the fs_mgr_flags member of struct fstab_rec. The remaining members are initialized by the corresponding value of the parameter flag_vals.
/system/core/fs_mgr/fs_mgr_fstab.c
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 },
{ "forceencrypt=",MF_FORCECRYPT },
{ "fileencryption",MF_FILEENCRYPTION },
{ "nonremovable",MF_NONREMOVABLE },
{ "voldmanaged=",MF_VOLDMANAGED},
{ "length=", MF_LENGTH },
{ "recoveryonly",MF_RECOVERYONLY },
{ "swapprio=", MF_SWAPPRIO },
{ "zramsize=", MF_ZRAMSIZE },
{ "verify", MF_VERIFY },
{ "noemulatedsd", MF_NOEMULATEDSD },
{ "notrim", MF_NOTRIM },
{ "formattable", MF_FORMATTABLE },
{ "defaults", 0 },
{ 0, 0 },
};/system/core/fs_mgr/fs_mgr_fstab.c
static int parse_flags(char *flags, struct flag_list *fl,
struct fs_mgr_flag_values *flag_vals,
char *fs_options, int fs_options_len)
{
int f = 0;
int i;
char *p;
char *savep;
/* initialize flag values. If we find a relevant flag, we'll
* update the value */
if (flag_vals) {
memset(flag_vals, 0, sizeof(*flag_vals));
flag_vals->partnum = -1;
flag_vals->swap_prio = -1; /* negative means it wasn't specified. */
}
/* 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) && flag_vals) {
/* The encryptable flag is followed by an = and the
* location of the keys. Get it and return it.
*/
flag_vals->key_loc = strdup(strchr(p, '=') + 1);
} else if ((fl[i].flag == MF_VERIFY) && flag_vals) {
/* If the verify flag is followed by an = and the
* location for the verity state, get it and return it.
*/
char *start = strchr(p, '=');
if (start) {
flag_vals->verity_loc = strdup(start + 1);
}
} else if ((fl[i].flag == MF_FORCECRYPT) && flag_vals) {
/* The forceencrypt flag is followed by an = and the
* location of the keys. Get it and return it.
*/
flag_vals->key_loc = strdup(strchr(p, '=') + 1);
} else if ((fl[i].flag == MF_LENGTH) && flag_vals) {
/* The length flag is followed by an = and the
* size of the partition. Get it and return it.
*/
flag_vals->part_length = strtoll(strchr(p, '=') + 1, NULL, 0);
} else if ((fl[i].flag == MF_VOLDMANAGED) && flag_vals) {
/* 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) {
flag_vals->label = strndup(label_start,
(int) (label_end - label_start));
part_start = strchr(p, ':') + 1;
if (!strcmp(part_start, "auto")) {
flag_vals->partnum = -1;
} else {
flag_vals->partnum = strtol(part_start, NULL, 0);
}
} else {
ERROR("Warning: voldmanaged= flag malformed\n");
}
} else if ((fl[i].flag == MF_SWAPPRIO) && flag_vals) {
flag_vals->swap_prio = strtoll(strchr(p, '=') + 1, NULL, 0);
} else if ((fl[i].flag == MF_ZRAMSIZE) && flag_vals) {
int is_percent = !!strrchr(p, '%');
unsigned int val = strtoll(strchr(p, '=') + 1, NULL, 0);
if (is_percent)
flag_vals->zram_size = calculate_zram_size(val);
else
flag_vals->zram_size = val;
}
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);
}
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;
}/system/core/fs_mgr/fs_mgr_fstab.c
struct fstab *fs_mgr_read_fstab(const char *fstab_path)
{
FILE *fstab_file;
int cnt, entries;
ssize_t len;
size_t alloc_len = 0;
char *line = NULL;
const char *delim = " \t";
char *save_ptr, *p;
struct fstab *fstab = NULL;
struct fs_mgr_flag_values flag_vals;
#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 ((len = getline(&line, &alloc_len, fstab_file)) != -1) {
/* if the last character is a newline, shorten the string by 1 byte */
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");
goto err;
}
/* 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 ((len = getline(&line, &alloc_len, fstab_file)) != -1) {
/* if the last character is a newline, shorten the string by 1 byte */
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");
goto err;
}
fstab->recs[cnt].blk_device = strdup(p);
if (!(p = strtok_r(NULL, delim, &save_ptr))) {
ERROR("Error parsing mount_point\n");
goto err;
}
fstab->recs[cnt].mount_point = strdup(p);
if (!(p = strtok_r(NULL, delim, &save_ptr))) {
ERROR("Error parsing fs_type\n");
goto err;
}
fstab->recs[cnt].fs_type = strdup(p);
if (!(p = strtok_r(NULL, delim, &save_ptr))) {
ERROR("Error parsing mount_flags\n");
goto err;
}
tmp_fs_options[0] = '\0';
fstab->recs[cnt].flags = parse_flags(p, mount_flags, 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");
goto err;
}
fstab->recs[cnt].fs_mgr_flags = parse_flags(p, fs_mgr_flags,
&flag_vals, NULL, 0);
fstab->recs[cnt].key_loc = flag_vals.key_loc;
fstab->recs[cnt].verity_loc = flag_vals.verity_loc;
fstab->recs[cnt].length = flag_vals.part_length;
fstab->recs[cnt].label = flag_vals.label;
fstab->recs[cnt].partnum = flag_vals.partnum;
fstab->recs[cnt].swap_prio = flag_vals.swap_prio;
fstab->recs[cnt].zram_size = flag_vals.zram_size;
cnt++;
}
fclose(fstab_file);
free(line);
return fstab;
err:
fclose(fstab_file);
free(line);
if (fstab)
fs_mgr_free_fstab(fstab);
return NULL;
}Going back to the process_config function, now that you have a struct fstab representing the mount situation, and then processing the mount points controlled by vold (with the "voldmanaged=" flag). According to the mounted device path, label value (that is, the string between '=' and ':' after "voldmanaged"), and flag value (obtained by code logic), construct a DiskSource object, and push its pointer to mDiskSources of VolumeManager . vold will receive uevent events from the kernel from these DiskSource objects.