errcode_t ocfs2_bitmap_alloc_region(ocfs2_bitmap *bitmap,
uint64_t start_bit,
int bitmap_start,
int total_bits,
struct ocfs2_bitmap_region **ret_br)
{
errcode_t ret;
struct ocfs2_bitmap_region *br;
if (total_bits < 0)
return OCFS2_ET_INVALID_BIT;
ret = ocfs2_malloc0(sizeof(struct ocfs2_bitmap_region), &br);
if (ret)
return ret;
br->br_start_bit = start_bit;
br->br_bitmap_start = bitmap_start;
br->br_valid_bits = total_bits;
br->br_total_bits = total_bits + bitmap_start;
br->br_bytes = ocfs2_align_total(br->br_total_bits);
ret = ocfs2_malloc0(br->br_bytes, &br->br_bitmap);
if (ret)
ocfs2_free(&br);
else
*ret_br = br;
return ret;
}
errcode_t io_init_cache(io_channel *channel, size_t nr_blocks)
{
int i;
struct io_cache *ic;
char *dbuf;
struct io_cache_block *icb_list;
errcode_t ret;
ret = ocfs2_malloc0(sizeof(struct io_cache), &ic);
if (ret)
goto out;
ic->ic_nr_blocks = nr_blocks;
ic->ic_lookup = RB_ROOT;
INIT_LIST_HEAD(&ic->ic_lru);
ret = ocfs2_malloc_blocks(channel, nr_blocks, &ic->ic_data_buffer);
if (ret)
goto out;
ic->ic_data_buffer_len = (unsigned long)nr_blocks * channel->io_blksize;
ret = ocfs2_malloc0(sizeof(struct io_cache_block) * nr_blocks,
&ic->ic_metadata_buffer);
if (ret)
goto out;
ic->ic_metadata_buffer_len =
(unsigned long)nr_blocks * sizeof(struct io_cache_block);
icb_list = ic->ic_metadata_buffer;
dbuf = ic->ic_data_buffer;
for (i = 0; i < nr_blocks; i++) {
icb_list[i].icb_blkno = UINT64_MAX;
icb_list[i].icb_buf = dbuf;
dbuf += channel->io_blksize;
list_add_tail(&icb_list[i].icb_list, &ic->ic_lru);
}
ic->ic_use_count = 1;
channel->io_cache = ic;
out:
if (ret)
io_free_cache(ic);
return ret;
}
errcode_t ocfs2_chain_iterate(ocfs2_filesys *fs,
uint64_t blkno,
int (*func)(ocfs2_filesys *fs,
uint64_t gd_blkno,
int chain_num,
void *priv_data),
void *priv_data)
{
int iret = 0;
char *buf;
struct ocfs2_dinode *inode;
errcode_t ret;
struct chain_context ctxt;
ret = ocfs2_malloc_block(fs->fs_io, &buf);
if (ret)
return ret;
ret = ocfs2_read_inode(fs, blkno, buf);
if (ret)
goto out_buf;
inode = (struct ocfs2_dinode *)buf;
ret = OCFS2_ET_INODE_NOT_VALID;
if (!(inode->i_flags & OCFS2_VALID_FL))
goto out_buf;
ret = OCFS2_ET_INODE_CANNOT_BE_ITERATED;
if (!(inode->i_flags & OCFS2_CHAIN_FL))
goto out_buf;
ret = ocfs2_malloc0(fs->fs_blocksize, &ctxt.gd_buf);
if (ret)
goto out_gd_buf;
ctxt.fs = fs;
ctxt.func = func;
ctxt.priv_data = priv_data;
ret = 0;
iret |= chain_iterate_cl(&inode->id2.i_chain, &ctxt);
if (iret & OCFS2_EXTENT_ERROR)
ret = ctxt.errcode;
if (iret & OCFS2_EXTENT_CHANGED) {
/* Do something */
}
out_gd_buf:
if (ctxt.gd_buf)
ocfs2_free(&ctxt.gd_buf);
out_buf:
ocfs2_free(&buf);
return ret;
}
/*
* This function is private to the library. Bitmap subtypes will
* use this to allocate their structure, but their b_ops will
* determine how they work.
*/
errcode_t ocfs2_bitmap_new(ocfs2_filesys *fs,
uint64_t total_bits,
const char *description,
struct ocfs2_bitmap_operations *ops,
void *private_data,
ocfs2_bitmap **ret_bitmap)
{
errcode_t ret;
ocfs2_bitmap *bitmap;
if (!ops->set_bit || !ops->clear_bit || !ops->test_bit)
return OCFS2_ET_INVALID_ARGUMENT;
ret = ocfs2_malloc0(sizeof(struct _ocfs2_bitmap), &bitmap);
if (ret)
return ret;
bitmap->b_fs = fs;
bitmap->b_total_bits = total_bits;
bitmap->b_ops = ops;
bitmap->b_regions = RB_ROOT;
bitmap->b_private = private_data;
if (description) {
ret = ocfs2_malloc0(sizeof(char) *
(strlen(description) + 1),
&bitmap->b_description);
if (ret)
goto out_free;
strcpy(bitmap->b_description, description);
}
*ret_bitmap = bitmap;
return 0;
out_free:
ocfs2_free(&bitmap);
return ret;
}
static int features_parse_option(struct tunefs_operation *op, char *arg)
{
int rc = 1;
errcode_t err;
struct feature_op_state *state = NULL;
struct check_supported_context ctxt = {
.sc_string = arg,
};
if (!arg) {
errorf("No features specified\n");
goto out;
}
err = ocfs2_malloc0(sizeof(struct feature_op_state), &state);
if (err) {
tcom_err(err, "while processing feature options");
goto out;
}
state->fo_op = op;
err = ocfs2_parse_feature(arg, &state->fo_feature_set,
&state->fo_reverse_set);
if (err) {
tcom_err(err, "while parsing feature options \"%s\"", arg);
goto out;
}
ctxt.sc_state = state;
ctxt.sc_action = FEATURE_ENABLE;
ocfs2_feature_foreach(&state->fo_feature_set, check_supported_func,
&ctxt);
if (ctxt.sc_error)
goto out;
ctxt.sc_action = FEATURE_DISABLE;
ocfs2_feature_reverse_foreach(&state->fo_reverse_set,
check_supported_func,
&ctxt);
if (ctxt.sc_error)
goto out;
rc = 0;
out:
if (!rc)
op->to_private = state;
else if (state)
ocfs2_free(&state);
return rc;
}
/*
* \a=0x07, \b=0x08, \t=0x09, \n=0x0a, \v=0x0b, \f=0x0c, \r=0x0d
*/
static char * process_escapes(char *queryfmt)
{
char *fmt;
int i, j;
int len;
len = strlen(queryfmt);
if (ocfs2_malloc0(len + 1, &fmt))
return NULL;
for(i = 0, j = 0; i < len; ) {
if (queryfmt[i] != '\\')
fmt[j++] = queryfmt[i++];
else {
switch(queryfmt[i + 1]) {
case 'a':
fmt[j++] = 0x07;
break;
case 'b':
fmt[j++] = 0x08;
break;
case 't':
fmt[j++] = 0x09;
break;
case 'n':
fmt[j++] = 0x0A;
break;
case 'v':
fmt[j++] = 0x0B;
break;
case 'f':
fmt[j++] = 0x0C;
break;
case 'r':
fmt[j++] = 0x0D;
break;
default:
fmt[j++] = queryfmt[i];
fmt[j++] = queryfmt[i + 1];
break;
}
i += 2;
}
}
return fmt;
}
errcode_t ocfs2_lookup_system_inode(ocfs2_filesys *fs, int type,
int slot_num, uint64_t *blkno)
{
errcode_t ret;
char *buf;
ret = ocfs2_malloc0(sizeof(char) * (OCFS2_MAX_FILENAME_LEN + 1), &buf);
if (ret)
return ret;
ocfs2_sprintf_system_inode_name(buf, OCFS2_MAX_FILENAME_LEN,
type, slot_num);
ret = ocfs2_lookup(fs, fs->fs_sysdir_blkno, buf,
strlen(buf), NULL, blkno);
ocfs2_free(&buf);
return ret;
}
/*
* ocfs2_partition_list()
*
*/
static errcode_t ocfs2_partition_list (struct list_head *dev_list)
{
errcode_t ret = 0;
FILE *proc;
char line[256];
char name[256];
char major[256];
char minor[256];
ocfs2_devices *dev;
proc = fopen ("/proc/partitions", "r");
if (proc == NULL) {
ret = OCFS2_ET_IO;
goto bail;
}
while (fgets (line, sizeof(line), proc) != NULL) {
*major = *minor = *name = '\0';
if (sscanf(line, "%*[ ]%[0-9]%*[ ]%[0-9] %*d %99[^ \t\n]",
major, minor, name) != 3)
continue;
ret = ocfs2_malloc0(sizeof(ocfs2_devices), &dev);
if (ret)
goto bail;
snprintf(dev->dev_name, sizeof(dev->dev_name), "/dev/%s", name);
dev->maj_num = strtoul(major, NULL, 0);
dev->min_num = strtoul(minor, NULL, 0);
list_add_tail(&(dev->list), dev_list);
}
bail:
if (proc)
fclose(proc);
return ret;
}
errcode_t io_open(const char *name, int flags, io_channel **channel)
{
errcode_t ret;
io_channel *chan = NULL;
#ifdef __linux__
struct stat stat_buf;
struct utsname ut;
#endif
if (!name || !*name)
return OCFS2_ET_BAD_DEVICE_NAME;
ret = ocfs2_malloc0(sizeof(struct _io_channel), &chan);
if (ret)
return ret;
ret = ocfs2_malloc(strlen(name)+1, &chan->io_name);
if (ret)
goto out_chan;
strcpy(chan->io_name, name);
chan->io_blksize = OCFS2_MIN_BLOCKSIZE;
chan->io_flags = (flags & OCFS2_FLAG_RW) ? O_RDWR : O_RDONLY;
chan->io_nocache = false;
if (!(flags & OCFS2_FLAG_BUFFERED))
chan->io_flags |= O_DIRECT;
chan->io_error = 0;
chan->io_fd = open64(name, chan->io_flags);
if (chan->io_fd < 0) {
/* chan will be freed, don't bother with chan->io_error */
if (errno == ENOENT)
ret = OCFS2_ET_NAMED_DEVICE_NOT_FOUND;
else
ret = OCFS2_ET_IO;
goto out_name;
}
if (!(flags & OCFS2_FLAG_BUFFERED)) {
ret = io_validate_o_direct(chan);
if (ret)
goto out_close; /* FIXME: bindraw here */
}
/* Workaround from e2fsprogs */
#ifdef __linux__
#undef RLIM_INFINITY
#if (defined(__alpha__) || ((defined(__sparc__) || defined(__mips__)) && (SIZEOF_LONG == 4)))
#define RLIM_INFINITY ((unsigned long)(~0UL>>1))
#else
#define RLIM_INFINITY (~0UL)
#endif
/*
* Work around a bug in 2.4.10-2.4.18 kernels where writes to
* block devices are wrongly getting hit by the filesize
* limit. This workaround isn't perfect, since it won't work
* if glibc wasn't built against 2.2 header files. (Sigh.)
*
*/
if ((flags & OCFS2_FLAG_RW) &&
(uname(&ut) == 0) &&
((ut.release[0] == '2') && (ut.release[1] == '.') &&
(ut.release[2] == '4') && (ut.release[3] == '.') &&
(ut.release[4] == '1') && (ut.release[5] >= '0') &&
(ut.release[5] < '8')) &&
(fstat(chan->io_fd, &stat_buf) == 0) &&
(S_ISBLK(stat_buf.st_mode))) {
struct rlimit rlim;
rlim.rlim_cur = rlim.rlim_max = (unsigned long) RLIM_INFINITY;
setrlimit(RLIMIT_FSIZE, &rlim);
getrlimit(RLIMIT_FSIZE, &rlim);
if (((unsigned long) rlim.rlim_cur) <
((unsigned long) rlim.rlim_max)) {
rlim.rlim_cur = rlim.rlim_max;
setrlimit(RLIMIT_FSIZE, &rlim);
}
}
#endif
*channel = chan;
return 0;
out_close:
/* Ignore the return, leave the original error */
close(chan->io_fd);
out_name:
ocfs2_free(&chan->io_name);
out_chan:
ocfs2_free(&chan);
*channel = NULL;
return ret;
}
errcode_t o2fsck_check_journals(o2fsck_state *ost)
{
errcode_t ret;
int i;
ocfs2_filesys *fs = ost->ost_fs;
int have_one_good_journal = 0, problem_is_consistent = 1;
errcode_t known_problem = 0;
struct journal_check_context jc = {
.jc_max_slots = OCFS2_RAW_SB(fs->fs_super)->s_max_slots,
};
struct journal_check_info *ji;
ret = ocfs2_malloc0(sizeof(struct journal_check_info) *
jc.jc_max_slots,
&jc.jc_info);
if (ret) {
com_err(whoami, ret, "while checking journals");
goto out;
}
ret = check_journal_walk(ost, check_journals_func, &jc);
if (ret) {
com_err(whoami, ret, "while checking journals");
goto out;
}
/*
* We now know the state of all our journals. If we have at least
* one good journal, we have a sane state to fix the others from.
* We require all our journals to have identical configuration.
* Any inconsistencies (invalid size, bad feature flags) are
* probably corruption or a failed tunefs.
*
* If we don't have a good journal, but all the journals have the
* exact same problem, we may be able to handle it as well. We
* currently know how to handle these problems:
*
* JOURNAL_TOO_SMALL
*
* We simply allocate a default journal size.
*
* UNSUPP_FEATURE & RO_UNSUPP_FEATURE
*
* If one journal has an unsupported feature bit set, it's probably
* corruption. If all the journals have the exact same feature
* bit set, it's certainly a feature we don't understand, and we
* want the user to upgrade their fsck.
*/
for (i = 0; i < jc.jc_max_slots; i++) {
ji = &jc.jc_info[i];
if (!ji->i_error) {
have_one_good_journal = 1;
continue;
}
if ((ji->i_error != OCFS2_ET_JOURNAL_TOO_SMALL) &&
(ji->i_error != OCFS2_ET_UNSUPP_FEATURE) &&
(ji->i_error != OCFS2_ET_RO_UNSUPP_FEATURE)) {
problem_is_consistent = 0;
continue;
}
if (known_problem) {
if (known_problem != ji->i_error)
problem_is_consistent = 0;
continue;
}
known_problem = ji->i_error;
}
if (!have_one_good_journal) {
if (!problem_is_consistent || !known_problem) {
ret = jc.jc_info[0].i_error;
com_err(whoami, ret, "while checking journals");
goto out;
}
if ((known_problem == OCFS2_ET_UNSUPP_FEATURE) ||
(known_problem == OCFS2_ET_RO_UNSUPP_FEATURE)) {
com_err(whoami, known_problem,
"on all journals. Please upgrade to the "
"latest version of fsck.ocfs2");
ret = known_problem;
goto out;
}
if (known_problem != OCFS2_ET_JOURNAL_TOO_SMALL) {
ret = known_problem;
com_err(whoami, ret, "for all journals");
goto out;
}
/* Force a valid cluster count for the journals */
jc.jc_max_clusters =
ocfs2_clusters_in_bytes(fs,
OCFS2_MIN_JOURNAL_SIZE);
}
ret = check_journal_walk(ost, fix_journals_func, &jc);
out:
if (jc.jc_info)
ocfs2_free(&jc.jc_info);
return ret;
}
static errcode_t ocfs2_clear_journal_flag(ocfs2_filesys *fs,
struct ocfs2_dinode *di,
int slot)
{
errcode_t ret = 0;
if (!(di->i_flags & OCFS2_VALID_FL) ||
!(di->i_flags & OCFS2_SYSTEM_FL) ||
!(di->i_flags & OCFS2_JOURNAL_FL))
return OCFS2_ET_INVALID_ARGUMENT;
if (!(di->id1.journal1.ij_flags & OCFS2_JOURNAL_DIRTY_FL))
goto bail;
di->id1.journal1.ij_flags &= ~OCFS2_JOURNAL_DIRTY_FL;
ret = ocfs2_write_inode(fs, di->i_blkno, (char *)di);
if (!ret)
printf("Slot %d's journal dirty flag removed\n", slot);
bail:
return ret;
}
errcode_t ocfs2_extent_iterate_inode(ocfs2_filesys *fs,
struct ocfs2_dinode *inode,
int flags,
char *block_buf,
int (*func)(ocfs2_filesys *fs,
struct ocfs2_extent_rec *rec,
int tree_depth,
uint32_t ccount,
uint64_t ref_blkno,
int ref_recno,
void *priv_data),
void *priv_data)
{
int i;
int iret = 0;
struct ocfs2_extent_list *el;
errcode_t ret;
struct extent_context ctxt;
ret = OCFS2_ET_INODE_NOT_VALID;
if (!(inode->i_flags & OCFS2_VALID_FL))
goto out;
ret = OCFS2_ET_INODE_CANNOT_BE_ITERATED;
if (inode->i_flags & (OCFS2_SUPER_BLOCK_FL |
OCFS2_LOCAL_ALLOC_FL |
OCFS2_CHAIN_FL))
goto out;
el = &inode->id2.i_list;
if (el->l_tree_depth) {
ret = ocfs2_malloc0(sizeof(char *) * el->l_tree_depth,
&ctxt.eb_bufs);
if (ret)
goto out;
if (block_buf) {
ctxt.eb_bufs[0] = block_buf;
} else {
ret = ocfs2_malloc0(fs->fs_blocksize *
el->l_tree_depth,
&ctxt.eb_bufs[0]);
if (ret)
goto out_eb_bufs;
}
for (i = 1; i < el->l_tree_depth; i++) {
ctxt.eb_bufs[i] = ctxt.eb_bufs[0] +
i * fs->fs_blocksize;
}
}
else
ctxt.eb_bufs = NULL;
ctxt.fs = fs;
ctxt.func = func;
ctxt.priv_data = priv_data;
ctxt.flags = flags;
ctxt.ccount = 0;
ctxt.last_eb_blkno = 0;
ctxt.last_eb_cpos = 0;
ret = 0;
iret |= extent_iterate_el(el, 0, &ctxt);
if (iret & OCFS2_EXTENT_ERROR)
ret = ctxt.errcode;
if (iret & OCFS2_EXTENT_ABORT)
goto out_abort;
/* we can only trust ctxt.last_eb_blkno if we walked the whole tree */
if (inode->i_last_eb_blk != ctxt.last_eb_blkno) {
inode->i_last_eb_blk = ctxt.last_eb_blkno;
iret |= OCFS2_EXTENT_CHANGED;
}
out_abort:
if (!ret && (iret & OCFS2_EXTENT_CHANGED))
ret = ocfs2_write_inode(fs, inode->i_blkno, (char *)inode);
out_eb_bufs:
if (ctxt.eb_bufs) {
if (!block_buf && ctxt.eb_bufs[0])
ocfs2_free(&ctxt.eb_bufs[0]);
ocfs2_free(&ctxt.eb_bufs);
}
out:
return ret;
}
/*
* This function will iterate the chain_rec and do the following modifications:
* 1. record all the groups in the chains.
* 2. for every group, do:
* 1) modify Sub Alloc Slot in extent block/inodes accordingly.
* 2) change the GROUP_PARENT according to its future owner.
* 3) link the group to the new slot files.
*/
static errcode_t move_chain_rec(ocfs2_filesys *fs, struct relink_ctxt *ctxt)
{
errcode_t ret = 0;
int i, start, end = 1;
uint64_t blkno, gd_blkno = ctxt->cr->c_blkno;
struct ocfs2_group_desc *gd = NULL;
struct moved_group *group = NULL, *group_head = NULL;
if (gd_blkno == 0)
goto bail;
/* Record the group in the relink_ctxt.
*
* We record the group in a reverse order, so the first group
* will be at the end of the group list. This is useful for
* fsck.ocfs2 when any error happens during the move of groups
* and we can safely move the group also.
*/
while (gd_blkno) {
ret = ocfs2_malloc0(sizeof(struct moved_group), &group);
if (ret)
goto bail;
memset(group, 0, sizeof(struct moved_group));
/* We insert the group first in case of any further error
* will not cause memory leak.
*/
group->next = group_head;
group_head = group;
ret = ocfs2_malloc_block(fs->fs_io, &group->gd_buf);
if (ret)
goto bail;
ret = ocfs2_read_group_desc(fs, gd_blkno, group->gd_buf);
if (ret)
goto bail;
group->blkno = gd_blkno;
gd = (struct ocfs2_group_desc *)group->gd_buf;
gd_blkno = gd->bg_next_group;
}
group = group_head;
while (group) {
gd = (struct ocfs2_group_desc *)group->gd_buf;
end = 1;
/* Modify the "Sub Alloc Slot" in the extent block/inodes. */
while (end < gd->bg_bits) {
start = ocfs2_find_next_bit_set(gd->bg_bitmap,
gd->bg_bits, end);
if (start >= gd->bg_bits)
break;
end = ocfs2_find_next_bit_clear(gd->bg_bitmap,
gd->bg_bits, start);
for (i = start; i < end; i++) {
blkno = group->blkno + i;
ret = change_sub_alloc_slot(fs, blkno, ctxt);
if (ret)
goto bail;
}
}
/* move the group to the new slots. */
ret = move_group(fs, ctxt, group);
if (ret)
goto bail;
group = group->next;
}
bail:
group = group_head;
while (group) {
group_head = group->next;
if (group->gd_buf)
ocfs2_free(&group->gd_buf);
ocfs2_free(&group);
group = group_head;
}
return ret;
}
/*
* allocate ocfs2_image_bitmap_arr and ocfs2 image bitmap blocks. o2image bitmap
* block is of size OCFS2_IMAGE_BITMAP_BLOCKSIZE and ocfs2_image_bitmap_arr
* tracks the bitmap blocks
*/
errcode_t ocfs2_image_alloc_bitmap(ocfs2_filesys *ofs)
{
uint64_t blks, allocsize, leftsize;
struct ocfs2_image_state *ost = ofs->ost;
int indx, i, n;
errcode_t ret;
char *buf;
ost->ost_bmpblks =
((ost->ost_fsblkcnt - 1) / (OCFS2_IMAGE_BITS_IN_BLOCK)) + 1;
ost->ost_bmpblksz = OCFS2_IMAGE_BITMAP_BLOCKSIZE;
blks = ost->ost_bmpblks;
/* allocate memory for an array to track bitmap blocks */
ret = ocfs2_malloc0((blks * sizeof(ocfs2_image_bitmap_arr)),
&ost->ost_bmparr);
if (ret)
return ret;
allocsize = blks * OCFS2_IMAGE_BITMAP_BLOCKSIZE;
leftsize = allocsize;
indx = 0;
/* allocate bitmap blocks and assign blocks to above array */
while (leftsize) {
ret = ocfs2_malloc_blocks(ofs->fs_io,
allocsize/io_get_blksize(ofs->fs_io),
&buf);
if (ret && (ret != -ENOMEM))
goto out;
if (ret == -ENOMEM) {
if (allocsize == OCFS2_IMAGE_BITMAP_BLOCKSIZE)
goto out;
allocsize >>= 1;
if (allocsize % OCFS2_IMAGE_BITMAP_BLOCKSIZE) {
allocsize /= OCFS2_IMAGE_BITMAP_BLOCKSIZE;
allocsize *= OCFS2_IMAGE_BITMAP_BLOCKSIZE;
}
continue;
}
n = allocsize / OCFS2_IMAGE_BITMAP_BLOCKSIZE;
for (i = 0; i < n; i++) {
ost->ost_bmparr[indx].arr_set_bit_cnt = 0;
ost->ost_bmparr[indx].arr_map =
((char *)buf + (i *
OCFS2_IMAGE_BITMAP_BLOCKSIZE));
/* remember buf address to free it later */
if (!i)
ost->ost_bmparr[indx].arr_self = buf;
indx++;
}
leftsize -= allocsize;
if (leftsize <= allocsize)
allocsize = leftsize;
}
out:
/* If allocation failed free and return error */
if (leftsize) {
for (i = 0; i < indx; i++)
if (ost->ost_bmparr[i].arr_self)
ocfs2_free(&ost->ost_bmparr[i].arr_self);
ocfs2_free(&ost->ost_bmparr);
}
return ret;
}
/*
* This routine loads bitmap blocks from an o2image image file into memory.
* This process happens during file open. bitmap blocks reside towards
* the end of the imagefile.
*/
errcode_t ocfs2_image_load_bitmap(ocfs2_filesys *ofs)
{
struct ocfs2_image_state *ost;
struct ocfs2_image_hdr *hdr;
uint64_t blk_off, bits_set;
int i, j, fd;
ssize_t count;
errcode_t ret;
char *blk;
ret = ocfs2_malloc0(sizeof(struct ocfs2_image_state), &ofs->ost);
if (ret)
return ret;
ost = ofs->ost;
ret = ocfs2_malloc_block(ofs->fs_io, &blk);
if (ret)
return ret;
/* read ocfs2 image header */
ret = io_read_block(ofs->fs_io, 0, 1, blk);
if (ret)
goto out;
hdr = (struct ocfs2_image_hdr *)blk;
ocfs2_image_swap_header(hdr);
ret = OCFS2_ET_BAD_MAGIC;
if (hdr->hdr_magic != OCFS2_IMAGE_MAGIC)
goto out;
if (memcmp(hdr->hdr_magic_desc, OCFS2_IMAGE_DESC,
sizeof(OCFS2_IMAGE_DESC)))
goto out;
ret = OCFS2_ET_OCFS_REV;
if (hdr->hdr_version > OCFS2_IMAGE_VERSION)
goto out;
ost->ost_fsblkcnt = hdr->hdr_fsblkcnt;
ost->ost_fsblksz = hdr->hdr_fsblksz;
ost->ost_imgblkcnt = hdr->hdr_imgblkcnt;
ost->ost_bmpblksz = hdr->hdr_bmpblksz;
ret = ocfs2_image_alloc_bitmap(ofs);
if (ret)
return ret;
/* load bitmap blocks ocfs2 image state */
bits_set = 0;
fd = io_get_fd(ofs->fs_io);
blk_off = (ost->ost_imgblkcnt + 1) * ost->ost_fsblksz;
for (i = 0; i < ost->ost_bmpblks; i++) {
ost->ost_bmparr[i].arr_set_bit_cnt = bits_set;
/*
* we don't use io_read_block as ocfs2 image bitmap block size
* could be different from filesystem block size
*/
count = pread64(fd, ost->ost_bmparr[i].arr_map,
ost->ost_bmpblksz, blk_off);
if (count < 0) {
ret = OCFS2_ET_IO;
goto out;
}
/* add bits set in this bitmap */
for (j = 0; j < (ost->ost_bmpblksz * 8); j++)
if (ocfs2_test_bit(j, ost->ost_bmparr[i].arr_map))
bits_set++;
blk_off += ost->ost_bmpblksz;
}
out:
if (blk)
ocfs2_free(&blk);
return ret;
}
static int walk_tree_func(struct ocfs2_dir_entry *dentry,
int offset,
int blocksize,
char *buf,
void *priv_data)
{
errcode_t ret;
int len, oldval;
int reti = 0;
char *old_path, *path;
struct walk_path *wp = priv_data;
if (!strncmp(dentry->name, ".", dentry->name_len) ||
!strncmp(dentry->name, "..", dentry->name_len))
return 0;
len = strlen(wp->path);
if (len + dentry->name_len > 4095) {
fprintf(stderr, "name is too long in %s\n", wp->path);
return OCFS2_DIRENT_ABORT;
}
ret = ocfs2_malloc0(4096, &path);
if (ret) {
com_err(wp->argv0, ret,
"while allocating path memory in %s\n",
wp->path);
return OCFS2_DIRENT_ABORT;
}
memcpy(path, wp->path, len);
memcpy(path + len, dentry->name, dentry->name_len);
if (dentry->file_type == OCFS2_FT_DIR)
path[len + dentry->name_len] = '/';
if (wp->check_dups) {
ret = ocfs2_bitmap_test(wp->dup_map, dentry->inode,
&oldval);
if (oldval) {
fprintf(stdout, "Dup! %20"PRIu64" %s\n",
(uint64_t)dentry->inode, path);
}
goto out;
}
oldval = 0;
ret = ocfs2_bitmap_set(wp->inode_map, dentry->inode, &oldval);
if (ret) {
com_err(wp->argv0, ret,
"while setting bitmap bit %"PRIu64"\n",
(uint64_t)dentry->inode);
reti = OCFS2_DIRENT_ABORT;
goto out;
}
if (oldval) {
wp->has_dups = 1;
ret = ocfs2_bitmap_set(wp->dup_map, dentry->inode,
NULL);
if (ret) {
com_err(wp->argv0, ret,
"while setting dup bit %"PRIu64"\n",
(uint64_t)dentry->inode);
reti = OCFS2_DIRENT_ABORT;
goto out;
}
}
if (!wp->quiet)
fprintf(stdout, "%20"PRIu64" %s\n",
(uint64_t)dentry->inode, path);
if (dentry->file_type == OCFS2_FT_DIR) {
old_path = wp->path;
wp->path = path;
ret = ocfs2_dir_iterate(wp->fs, dentry->inode, 0, NULL,
walk_tree_func, wp);
if (ret) {
com_err(wp->argv0, ret,
"while walking %s", wp->path);
reti = OCFS2_DIRENT_ABORT;
}
wp->path = old_path;
}
out:
ocfs2_free(&path);
return reti;
}
/* Try and replay the slots journals if they're dirty. This only returns
* a non-zero error if the caller should not continue. */
errcode_t o2fsck_replay_journals(ocfs2_filesys *fs, int *replayed)
{
errcode_t err = 0, ret = 0;
struct journal_info *jis = NULL, *ji;
journal_superblock_t *jsb;
char *buf = NULL;
int journal_trouble = 0;
uint16_t i, max_slots;
ocfs2_bitmap *used_blocks = NULL;
max_slots = OCFS2_RAW_SB(fs->fs_super)->s_max_slots;
ret = ocfs2_block_bitmap_new(fs, "journal blocks",
&used_blocks);
if (ret) {
com_err(whoami, ret, "while allocating journal block bitmap");
goto out;
}
ret = ocfs2_malloc_blocks(fs->fs_io, 1, &buf);
if (ret) {
com_err(whoami, ret, "while allocating room to read journal "
"blocks");
goto out;
}
ret = ocfs2_malloc0(sizeof(struct journal_info) * max_slots, &jis);
if (ret) {
com_err(whoami, ret, "while allocating an array of block "
"numbers for journal replay");
goto out;
}
printf("Checking each slot's journal.\n");
for (i = 0, ji = jis; i < max_slots; i++, ji++) {
ji->ji_used_blocks = used_blocks;
ji->ji_revoke = RB_ROOT;
ji->ji_slot = i;
/* sets ji->ji_replay */
err = prep_journal_info(fs, i, ji);
if (err) {
printf("Slot %d seems to have a corrupt journal.\n",
i);
journal_trouble = 1;
continue;
}
if (!ji->ji_replay) {
verbosef("slot %d is clean\n", i);
continue;
}
err = walk_journal(fs, i, ji, buf, 0);
if (err) {
printf("Slot %d's journal can not be replayed.\n", i);
journal_trouble = 1;
}
}
for (i = 0, ji = jis; i < max_slots; i++, ji++) {
if (!ji->ji_replay)
continue;
printf("Replaying slot %d's journal.\n", i);
err = walk_journal(fs, i, ji, buf, 1);
if (err) {
journal_trouble = 1;
continue;
}
jsb = ji->ji_jsb;
/* reset the journal */
jsb->s_start = 0;
if (ji->ji_set_final_seq)
jsb->s_sequence = ji->ji_final_seq + 1;
/* we don't write back a clean 'mounted' bit here. That would
* have to also include having recovered the orphan dir. we
* updated s_start, though, so we won't replay the journal
* again. */
err = ocfs2_write_journal_superblock(fs,
ji->ji_jsb_block,
(char *)ji->ji_jsb);
if (err) {
com_err(whoami, err, "while writing slot %d's journal "
"super block", i);
journal_trouble = 1;
} else {
printf("Slot %d's journal replayed successfully.\n",
i);
*replayed = 1;
}
}
/* this is awkward, but we want fsck -n to tell us as much as it
* can so we don't want to ask to proceed here. */
if (journal_trouble)
printf("*** There were problems replaying journals. Be "
"careful in telling fsck to make repairs to this "
"filesystem.\n");
ret = 0;
out:
if (jis) {
for (i = 0, ji = jis; i < max_slots; i++, ji++) {
if (ji->ji_jsb)
ocfs2_free(&ji->ji_jsb);
if (ji->ji_cinode)
ocfs2_free_cached_inode(fs,
ji->ji_cinode);
revoke_free_all(&ji->ji_revoke);
}
ocfs2_free(&jis);
}
if (buf)
ocfs2_free(&buf);
if (used_blocks)
ocfs2_bitmap_free(used_blocks);
return ret;
}
errcode_t ocfs2_open(const char *name, int flags,
unsigned int superblock, unsigned int block_size,
ocfs2_filesys **ret_fs)
{
ocfs2_filesys *fs;
errcode_t ret;
int i, len;
char *ptr;
unsigned char *raw_uuid;
ret = ocfs2_malloc0(sizeof(ocfs2_filesys), &fs);
if (ret)
return ret;
fs->fs_flags = flags;
fs->fs_umask = 022;
ret = io_open(name, (flags & (OCFS2_FLAG_RO | OCFS2_FLAG_RW |
OCFS2_FLAG_BUFFERED)),
&fs->fs_io);
if (ret)
goto out;
ret = ocfs2_malloc(strlen(name)+1, &fs->fs_devname);
if (ret)
goto out;
strcpy(fs->fs_devname, name);
/*
* If OCFS2_FLAG_IMAGE_FILE is specified, it needs to be handled
* differently
*/
if (flags & OCFS2_FLAG_IMAGE_FILE) {
ret = ocfs2_image_load_bitmap(fs);
if (ret)
goto out;
if (!superblock)
superblock = fs->ost->ost_superblocks[0];
if (!block_size)
block_size = fs->ost->ost_fsblksz;
}
/* image file is not a device */
if (!(flags & OCFS2_FLAG_IMAGE_FILE)) {
if (io_is_device_readonly(fs->fs_io))
fs->fs_flags |= OCFS2_FLAG_HARD_RO;
}
/*
* If OCFS2_FLAG_NO_REV_CHECK is specified, fsck (or someone
* like it) is asking to ignore the OCFS vol_header at
* block 0.
*/
if (!(flags & OCFS2_FLAG_NO_REV_CHECK)) {
ret = ocfs2_validate_ocfs1_header(fs);
if (ret)
goto out;
}
if (superblock) {
ret = OCFS2_ET_INVALID_ARGUMENT;
if (!block_size)
goto out;
io_set_blksize(fs->fs_io, block_size);
ret = ocfs2_read_super(fs, (uint64_t)superblock, NULL);
} else {
superblock = OCFS2_SUPER_BLOCK_BLKNO;
if (block_size) {
io_set_blksize(fs->fs_io, block_size);
ret = ocfs2_read_super(fs, (uint64_t)superblock, NULL);
} else {
for (block_size = io_get_blksize(fs->fs_io);
block_size <= OCFS2_MAX_BLOCKSIZE;
block_size <<= 1) {
io_set_blksize(fs->fs_io, block_size);
ret = ocfs2_read_super(fs, (uint64_t)superblock,
NULL);
if ((ret == OCFS2_ET_BAD_MAGIC) ||
(ret == OCFS2_ET_IO))
continue;
break;
}
}
}
if (ret)
goto out;
fs->fs_blocksize = block_size;
if (superblock == OCFS2_SUPER_BLOCK_BLKNO) {
ret = ocfs2_malloc_block(fs->fs_io, &fs->fs_orig_super);
if (ret)
goto out;
memcpy((char *)fs->fs_orig_super,
(char *)fs->fs_super, fs->fs_blocksize);
}
#if 0
ret = OCFS2_ET_REV_TOO_HIGH;
if (fs->fs_super->id2.i_super.s_major_rev_level >
OCFS2_LIB_CURRENT_REV)
goto out;
#endif
if (flags & OCFS2_FLAG_STRICT_COMPAT_CHECK) {
ret = OCFS2_ET_UNSUPP_FEATURE;
if (OCFS2_RAW_SB(fs->fs_super)->s_feature_compat &
~OCFS2_LIB_FEATURE_COMPAT_SUPP)
goto out;
/* We need to check s_tunefs_flag also to make sure
* fsck.ocfs2 won't try to clean up an aborted tunefs
* that it doesn't know.
*/
if (OCFS2_HAS_INCOMPAT_FEATURE(OCFS2_RAW_SB(fs->fs_super),
OCFS2_FEATURE_INCOMPAT_TUNEFS_INPROG) &&
(OCFS2_RAW_SB(fs->fs_super)->s_tunefs_flag &
~OCFS2_LIB_ABORTED_TUNEFS_SUPP))
goto out;
}
ret = OCFS2_ET_UNSUPP_FEATURE;
if (OCFS2_RAW_SB(fs->fs_super)->s_feature_incompat &
~OCFS2_LIB_FEATURE_INCOMPAT_SUPP)
goto out;
ret = OCFS2_ET_RO_UNSUPP_FEATURE;
if ((flags & OCFS2_FLAG_RW) &&
(OCFS2_RAW_SB(fs->fs_super)->s_feature_ro_compat &
~OCFS2_LIB_FEATURE_RO_COMPAT_SUPP))
goto out;
ret = OCFS2_ET_UNSUPP_FEATURE;
if (!(flags & OCFS2_FLAG_HEARTBEAT_DEV_OK) &&
(OCFS2_RAW_SB(fs->fs_super)->s_feature_incompat &
OCFS2_FEATURE_INCOMPAT_HEARTBEAT_DEV))
goto out;
ret = OCFS2_ET_CORRUPT_SUPERBLOCK;
if (!OCFS2_RAW_SB(fs->fs_super)->s_blocksize_bits)
goto out;
if (fs->fs_super->i_blkno != superblock)
goto out;
if ((OCFS2_RAW_SB(fs->fs_super)->s_clustersize_bits < 12) ||
(OCFS2_RAW_SB(fs->fs_super)->s_clustersize_bits > 20))
goto out;
if (!OCFS2_RAW_SB(fs->fs_super)->s_root_blkno ||
!OCFS2_RAW_SB(fs->fs_super)->s_system_dir_blkno)
goto out;
if (OCFS2_RAW_SB(fs->fs_super)->s_max_slots > OCFS2_MAX_SLOTS)
goto out;
ret = ocfs2_malloc0(OCFS2_RAW_SB(fs->fs_super)->s_max_slots *
sizeof(ocfs2_cached_inode *),
&fs->fs_inode_allocs);
if (ret)
goto out;
ret = ocfs2_malloc0(OCFS2_RAW_SB(fs->fs_super)->s_max_slots *
sizeof(ocfs2_cached_inode *),
&fs->fs_eb_allocs);
if (ret)
goto out;
ret = OCFS2_ET_UNEXPECTED_BLOCK_SIZE;
if (block_size !=
(1U << OCFS2_RAW_SB(fs->fs_super)->s_blocksize_bits))
goto out;
fs->fs_clustersize =
1 << OCFS2_RAW_SB(fs->fs_super)->s_clustersize_bits;
/* FIXME: Read the system dir */
fs->fs_root_blkno =
OCFS2_RAW_SB(fs->fs_super)->s_root_blkno;
fs->fs_sysdir_blkno =
OCFS2_RAW_SB(fs->fs_super)->s_system_dir_blkno;
fs->fs_clusters = fs->fs_super->i_clusters;
fs->fs_blocks = ocfs2_clusters_to_blocks(fs, fs->fs_clusters);
fs->fs_first_cg_blkno =
OCFS2_RAW_SB(fs->fs_super)->s_first_cluster_group;
raw_uuid = OCFS2_RAW_SB(fs->fs_super)->s_uuid;
for (i = 0, ptr = fs->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
/* print with null */
len = snprintf(ptr, 3, "%02X", raw_uuid[i]);
if (len != 2) {
ret = OCFS2_ET_INTERNAL_FAILURE;
goto out;
}
/* then only advace past the last char */
ptr += 2;
}
*ret_fs = fs;
return 0;
out:
if (fs->fs_inode_allocs)
ocfs2_free(&fs->fs_inode_allocs);
ocfs2_freefs(fs);
return ret;
}
errcode_t ocfs2_extent_iterate_xattr(ocfs2_filesys *fs,
struct ocfs2_extent_list *el,
uint64_t last_eb_blk,
int flags,
int (*func)(ocfs2_filesys *fs,
struct ocfs2_extent_rec *rec,
int tree_depth,
uint32_t ccount,
uint64_t ref_blkno,
int ref_recno,
void *priv_data),
void *priv_data,
int *changed)
{
int i;
int iret = 0;
errcode_t ret;
struct extent_context ctxt;
if (el->l_tree_depth) {
ret = ocfs2_malloc0(sizeof(char *) * el->l_tree_depth,
&ctxt.eb_bufs);
if (ret)
goto out;
ret = ocfs2_malloc0(fs->fs_blocksize *
el->l_tree_depth,
&ctxt.eb_bufs[0]);
if (ret)
goto out_eb_bufs;
for (i = 1; i < el->l_tree_depth; i++) {
ctxt.eb_bufs[i] = ctxt.eb_bufs[0] +
i * fs->fs_blocksize;
}
} else
ctxt.eb_bufs = NULL;
ctxt.fs = fs;
ctxt.func = func;
ctxt.priv_data = priv_data;
ctxt.flags = flags;
ctxt.ccount = 0;
ctxt.last_eb_blkno = 0;
ctxt.last_eb_cpos = 0;
ret = 0;
iret |= extent_iterate_el(el, 0, &ctxt);
if (iret & OCFS2_EXTENT_ERROR)
ret = ctxt.errcode;
if (iret & OCFS2_EXTENT_ABORT)
goto out_abort;
if (last_eb_blk != ctxt.last_eb_blkno) {
last_eb_blk = ctxt.last_eb_blkno;
iret |= OCFS2_EXTENT_CHANGED;
}
out_abort:
if (!ret && (iret & OCFS2_EXTENT_CHANGED))
*changed = 1;
out_eb_bufs:
if (ctxt.eb_bufs) {
if (ctxt.eb_bufs[0])
ocfs2_free(&ctxt.eb_bufs[0]);
ocfs2_free(&ctxt.eb_bufs);
}
out:
return ret;
}