static int exofs_read_kern(struct osd_dev *od, u8 *cred, struct osd_obj_id *obj,
u64 offset, void *p, unsigned length)
{
struct osd_request *or = osd_start_request(od, GFP_KERNEL);
/* struct osd_sense_info osi = {.key = 0};*/
int ret;
if (unlikely(!or)) {
EXOFS_DBGMSG("%s: osd_start_request failed.\n", __func__);
return -ENOMEM;
}
ret = osd_req_read_kern(or, obj, offset, p, length);
if (unlikely(ret)) {
EXOFS_DBGMSG("%s: osd_req_read_kern failed.\n", __func__);
goto out;
}
ret = osd_finalize_request(or, 0, cred, NULL);
if (unlikely(ret)) {
EXOFS_DBGMSG("Failed to osd_finalize_request() => %d\n", ret);
goto out;
}
ret = osd_execute_request(or);
if (unlikely(ret))
EXOFS_DBGMSG("osd_execute_request() => %d\n", ret);
/* osd_req_decode_sense(or, ret); */
out:
osd_end_request(or);
EXOFS_DBGMSG2("read_kern(0x%llx) offset=0x%llx "
"length=0x%llx dev=%p ret=>%d\n",
_LLU(obj->id), _LLU(offset), _LLU(length), od, ret);
return ret;
}
/*
* The main option parsing method. Also makes sure that all of the mandatory
* mount options were set.
*/
static int parse_options(char *options, struct exofs_mountopt *opts)
{
char *p;
substring_t args[MAX_OPT_ARGS];
int option;
bool s_pid = false;
EXOFS_DBGMSG("parse_options %s\n", options);
/* defaults */
memset(opts, 0, sizeof(*opts));
opts->timeout = BLK_DEFAULT_SG_TIMEOUT;
while ((p = strsep(&options, ",")) != NULL) {
int token;
char str[32];
if (!*p)
continue;
token = match_token(p, tokens, args);
switch (token) {
case Opt_name:
opts->dev_name = match_strdup(&args[0]);
if (unlikely(!opts->dev_name)) {
EXOFS_ERR("Error allocating dev_name");
return -ENOMEM;
}
opts->is_osdname = true;
break;
case Opt_pid:
if (0 == match_strlcpy(str, &args[0], sizeof(str)))
return -EINVAL;
opts->pid = simple_strtoull(str, NULL, 0);
if (opts->pid < EXOFS_MIN_PID) {
EXOFS_ERR("Partition ID must be >= %u",
EXOFS_MIN_PID);
return -EINVAL;
}
s_pid = 1;
break;
case Opt_to:
if (match_int(&args[0], &option))
return -EINVAL;
if (option <= 0) {
EXOFS_ERR("Timout must be > 0");
return -EINVAL;
}
opts->timeout = option * HZ;
break;
}
}
if (!s_pid) {
EXOFS_ERR("Need to specify the following options:\n");
EXOFS_ERR(" -o pid=pid_no_to_use\n");
return -EINVAL;
}
return 0;
}
/*
* Write the superblock to the OSD
*/
int exofs_sync_fs(struct super_block *sb, int wait)
{
struct exofs_sb_info *sbi;
struct exofs_fscb *fscb;
struct exofs_io_state *ios;
int ret = -ENOMEM;
lock_super(sb);
sbi = sb->s_fs_info;
fscb = &sbi->s_fscb;
ret = exofs_get_io_state(sbi, &ios);
if (ret)
goto out;
/* Note: We only write the changing part of the fscb. .i.e upto the
* the fscb->s_dev_table_oid member. There is no read-modify-write
* here.
*/
ios->length = offsetof(struct exofs_fscb, s_dev_table_oid);
memset(fscb, 0, ios->length);
fscb->s_nextid = cpu_to_le64(sbi->s_nextid);
fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles);
fscb->s_magic = cpu_to_le16(sb->s_magic);
fscb->s_newfs = 0;
fscb->s_version = EXOFS_FSCB_VER;
ios->obj.id = EXOFS_SUPER_ID;
ios->offset = 0;
ios->kern_buff = fscb;
ios->cred = sbi->s_cred;
ret = exofs_sbi_write(ios);
if (unlikely(ret)) {
EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
goto out;
}
sb->s_dirt = 0;
out:
EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret);
exofs_put_io_state(ios);
unlock_super(sb);
return ret;
}
static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
unsigned table_count)
{
struct exofs_sb_info *sbi = *psbi;
struct osd_dev *fscb_od;
struct osd_obj_id obj = {.partition = sbi->s_pid,
.id = EXOFS_DEVTABLE_ID};
struct exofs_device_table *dt;
unsigned table_bytes = table_count * sizeof(dt->dt_dev_table[0]) +
sizeof(*dt);
unsigned numdevs, i;
int ret;
dt = kmalloc(table_bytes, GFP_KERNEL);
if (unlikely(!dt)) {
EXOFS_ERR("ERROR: allocating %x bytes for device table\n",
table_bytes);
return -ENOMEM;
}
fscb_od = sbi->s_ods[0];
sbi->s_ods[0] = NULL;
sbi->s_numdevs = 0;
ret = exofs_read_kern(fscb_od, sbi->s_cred, &obj, 0, dt, table_bytes);
if (unlikely(ret)) {
EXOFS_ERR("ERROR: reading device table\n");
goto out;
}
numdevs = le64_to_cpu(dt->dt_num_devices);
if (unlikely(!numdevs)) {
ret = -EINVAL;
goto out;
}
WARN_ON(table_count != numdevs);
ret = _read_and_match_data_map(sbi, numdevs, dt);
if (unlikely(ret))
goto out;
if (likely(numdevs > 1)) {
unsigned size = numdevs * sizeof(sbi->s_ods[0]);
sbi = krealloc(sbi, sizeof(*sbi) + size, GFP_KERNEL);
if (unlikely(!sbi)) {
ret = -ENOMEM;
goto out;
}
memset(&sbi->s_ods[1], 0, size - sizeof(sbi->s_ods[0]));
*psbi = sbi;
}
for (i = 0; i < numdevs; i++) {
struct exofs_fscb fscb;
struct osd_dev_info odi;
struct osd_dev *od;
if (exofs_devs_2_odi(&dt->dt_dev_table[i], &odi)) {
EXOFS_ERR("ERROR: Read all-zeros device entry\n");
ret = -EINVAL;
goto out;
}
printk(KERN_NOTICE "Add device[%d]: osd_name-%s\n",
i, odi.osdname);
/* On all devices the device table is identical. The user can
* specify any one of the participating devices on the command
* line. We always keep them in device-table order.
*/
if (fscb_od && osduld_device_same(fscb_od, &odi)) {
sbi->s_ods[i] = fscb_od;
++sbi->s_numdevs;
fscb_od = NULL;
continue;
}
od = osduld_info_lookup(&odi);
if (unlikely(IS_ERR(od))) {
ret = PTR_ERR(od);
EXOFS_ERR("ERROR: device requested is not found "
"osd_name-%s =>%d\n", odi.osdname, ret);
goto out;
}
sbi->s_ods[i] = od;
++sbi->s_numdevs;
/* Read the fscb of the other devices to make sure the FS
* partition is there.
*/
ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb,
sizeof(fscb));
if (unlikely(ret)) {
EXOFS_ERR("ERROR: Malformed participating device "
"error reading fscb osd_name-%s\n",
odi.osdname);
goto out;
}
/* TODO: verify other information is correct and FS-uuid
* matches. Benny what did you say about device table
* generation and old devices?
*/
}
out:
kfree(dt);
if (unlikely(!ret && fscb_od)) {
EXOFS_ERR(
"ERROR: Bad device-table container device not present\n");
osduld_put_device(fscb_od);
ret = -EINVAL;
}
return ret;
}
/*
* Read the superblock from the OSD and fill in the fields
*/
static int exofs_fill_super(struct super_block *sb, void *data, int silent)
{
struct inode *root;
struct exofs_mountopt *opts = data;
struct exofs_sb_info *sbi; /*extended info */
struct osd_dev *od; /* Master device */
struct exofs_fscb fscb; /*on-disk superblock info */
struct osd_obj_id obj;
unsigned table_count;
int ret;
sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
if (!sbi)
return -ENOMEM;
/* use mount options to fill superblock */
od = osduld_path_lookup(opts->dev_name);
if (IS_ERR(od)) {
ret = PTR_ERR(od);
goto free_sbi;
}
sbi->s_ods[0] = od;
sbi->s_numdevs = 1;
sbi->s_pid = opts->pid;
sbi->s_timeout = opts->timeout;
/* fill in some other data by hand */
memset(sb->s_id, 0, sizeof(sb->s_id));
strcpy(sb->s_id, "exofs");
sb->s_blocksize = EXOFS_BLKSIZE;
sb->s_blocksize_bits = EXOFS_BLKSHIFT;
sb->s_maxbytes = MAX_LFS_FILESIZE;
atomic_set(&sbi->s_curr_pending, 0);
sb->s_bdev = NULL;
sb->s_dev = 0;
obj.partition = sbi->s_pid;
obj.id = EXOFS_SUPER_ID;
exofs_make_credential(sbi->s_cred, &obj);
ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb));
if (unlikely(ret))
goto free_sbi;
sb->s_magic = le16_to_cpu(fscb.s_magic);
sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles);
/* make sure what we read from the object store is correct */
if (sb->s_magic != EXOFS_SUPER_MAGIC) {
if (!silent)
EXOFS_ERR("ERROR: Bad magic value\n");
ret = -EINVAL;
goto free_sbi;
}
if (le32_to_cpu(fscb.s_version) != EXOFS_FSCB_VER) {
EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n",
EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version));
ret = -EINVAL;
goto free_sbi;
}
/* start generation numbers from a random point */
get_random_bytes(&sbi->s_next_generation, sizeof(u32));
spin_lock_init(&sbi->s_next_gen_lock);
table_count = le64_to_cpu(fscb.s_dev_table_count);
if (table_count) {
ret = exofs_read_lookup_dev_table(&sbi, table_count);
if (unlikely(ret))
goto free_sbi;
}
/* set up operation vectors */
sb->s_fs_info = sbi;
sb->s_op = &exofs_sops;
sb->s_export_op = &exofs_export_ops;
root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF);
if (IS_ERR(root)) {
EXOFS_ERR("ERROR: exofs_iget failed\n");
ret = PTR_ERR(root);
goto free_sbi;
}
sb->s_root = d_alloc_root(root);
if (!sb->s_root) {
iput(root);
EXOFS_ERR("ERROR: get root inode failed\n");
ret = -ENOMEM;
goto free_sbi;
}
if (!S_ISDIR(root->i_mode)) {
dput(sb->s_root);
sb->s_root = NULL;
EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n",
root->i_mode);
ret = -EINVAL;
goto free_sbi;
}
_exofs_print_device("Mounting", opts->dev_name, sbi->s_ods[0],
sbi->s_pid);
return 0;
free_sbi:
EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n",
opts->dev_name, sbi->s_pid, ret);
exofs_free_sbi(sbi);
return ret;
}
/*
* Set up the superblock (calls exofs_fill_super eventually)
*/
static int exofs_get_sb(struct file_system_type *type,
int flags, const char *dev_name,
void *data, struct vfsmount *mnt)
{
struct exofs_mountopt opts;
int ret;
ret = parse_options(data, &opts);
if (ret)
return ret;
opts.dev_name = dev_name;
return get_sb_nodev(type, flags, &opts, exofs_fill_super, mnt);
}
/*
* Return information about the file system state in the buffer. This is used
* by the 'df' command, for example.
*/
static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
{
struct super_block *sb = dentry->d_sb;
struct exofs_sb_info *sbi = sb->s_fs_info;
struct exofs_io_state *ios;
struct osd_attr attrs[] = {
ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION,
OSD_ATTR_PI_USED_CAPACITY, sizeof(__be64)),
};
uint64_t capacity = ULLONG_MAX;
uint64_t used = ULLONG_MAX;
uint8_t cred_a[OSD_CAP_LEN];
int ret;
ret = exofs_get_io_state(sbi, &ios);
if (ret) {
EXOFS_DBGMSG("exofs_get_io_state failed.\n");
return ret;
}
exofs_make_credential(cred_a, &ios->obj);
ios->cred = sbi->s_cred;
ios->in_attr = attrs;
ios->in_attr_len = ARRAY_SIZE(attrs);
ret = exofs_sbi_read(ios);
if (unlikely(ret))
goto out;
ret = extract_attr_from_ios(ios, &attrs[0]);
if (likely(!ret)) {
capacity = get_unaligned_be64(attrs[0].val_ptr);
if (unlikely(!capacity))
capacity = ULLONG_MAX;
} else
EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
ret = extract_attr_from_ios(ios, &attrs[1]);
if (likely(!ret))
used = get_unaligned_be64(attrs[1].val_ptr);
else
EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n");
/* fill in the stats buffer */
buf->f_type = EXOFS_SUPER_MAGIC;
buf->f_bsize = EXOFS_BLKSIZE;
buf->f_blocks = capacity >> 9;
buf->f_bfree = (capacity - used) >> 9;
buf->f_bavail = buf->f_bfree;
buf->f_files = sbi->s_numfiles;
buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles;
buf->f_namelen = EXOFS_NAME_LEN;
out:
exofs_put_io_state(ios);
return ret;
}
static const struct super_operations exofs_sops = {
.alloc_inode = exofs_alloc_inode,
.destroy_inode = exofs_destroy_inode,
.write_inode = exofs_write_inode,
.delete_inode = exofs_delete_inode,
.put_super = exofs_put_super,
.write_super = exofs_write_super,
.sync_fs = exofs_sync_fs,
.statfs = exofs_statfs,
};
/******************************************************************************
* EXPORT OPERATIONS
*****************************************************************************/
struct dentry *exofs_get_parent(struct dentry *child)
{
unsigned long ino = exofs_parent_ino(child);
if (!ino)
return NULL;
return d_obtain_alias(exofs_iget(child->d_inode->i_sb, ino));
}
