// SPDX-License-Identifier: GPL-2.0-only /* * xattr.c * * Copyright (C) 2004, 2008 Oracle. All rights reserved. * * CREDITS: * Lots of code in this file is copy from linux/fs/ext3/xattr.c. * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
*/
struct ocfs2_xattr_info { int xi_name_index; constchar *xi_name; int xi_name_len; constvoid *xi_value;
size_t xi_value_len;
};
struct ocfs2_xattr_search { struct buffer_head *inode_bh; /* * xattr_bh point to the block buffer head which has extended attribute * when extended attribute in inode, xattr_bh is equal to inode_bh.
*/ struct buffer_head *xattr_bh; struct ocfs2_xattr_header *header; struct ocfs2_xattr_bucket *bucket; void *base; void *end; struct ocfs2_xattr_entry *here; int not_found;
};
/* Operations on struct ocfs2_xa_entry */ struct ocfs2_xa_loc; struct ocfs2_xa_loc_operations { /* * Journal functions
*/ int (*xlo_journal_access)(handle_t *handle, struct ocfs2_xa_loc *loc, int type); void (*xlo_journal_dirty)(handle_t *handle, struct ocfs2_xa_loc *loc);
/* * Return a pointer to the appropriate buffer in loc->xl_storage * at the given offset from loc->xl_header.
*/ void *(*xlo_offset_pointer)(struct ocfs2_xa_loc *loc, int offset);
/* Can we reuse the existing entry for the new value? */ int (*xlo_can_reuse)(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi);
/* How much space is needed for the new value? */ int (*xlo_check_space)(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi);
/* * Return the offset of the first name+value pair. This is * the start of our downward-filling free space.
*/ int (*xlo_get_free_start)(struct ocfs2_xa_loc *loc);
/* * Remove the name+value at this location. Do whatever is * appropriate with the remaining name+value pairs.
*/ void (*xlo_wipe_namevalue)(struct ocfs2_xa_loc *loc);
/* Fill xl_entry with a new entry */ void (*xlo_add_entry)(struct ocfs2_xa_loc *loc, u32 name_hash);
/* Add name+value storage to an entry */ void (*xlo_add_namevalue)(struct ocfs2_xa_loc *loc, int size);
/* * Initialize the value buf's access and bh fields for this entry. * ocfs2_xa_fill_value_buf() will handle the xv pointer.
*/ void (*xlo_fill_value_buf)(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_value_buf *vb);
};
/* * Describes an xattr entry location. This is a memory structure * tracking the on-disk structure.
*/ struct ocfs2_xa_loc { /* This xattr belongs to this inode */ struct inode *xl_inode;
/* The ocfs2_xattr_header inside the on-disk storage. Not NULL. */ struct ocfs2_xattr_header *xl_header;
/* Bytes from xl_header to the end of the storage */ int xl_size;
/* * The ocfs2_xattr_entry this location describes. If this is * NULL, this location describes the on-disk structure where it * would have been.
*/ struct ocfs2_xattr_entry *xl_entry;
/* * Internal housekeeping
*/
/* Buffer(s) containing this entry */ void *xl_storage;
/* Operations on the storage backing this location */ conststruct ocfs2_xa_loc_operations *xl_ops;
};
/* * Convenience functions to calculate how much space is needed for a * given name+value pair
*/ staticint namevalue_size(int name_len, uint64_t value_len)
{ if (value_len > OCFS2_XATTR_INLINE_SIZE) return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; else return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len);
}
/* * A bucket that has never been written to disk doesn't need to be * read. We just need the buffer_heads. Don't call this for * buckets that are already on disk. ocfs2_read_xattr_bucket() initializes * them fully.
*/ staticint ocfs2_init_xattr_bucket(struct ocfs2_xattr_bucket *bucket,
u64 xb_blkno, intnew)
{ int i, rc = 0;
for (i = 0; i < bucket->bu_blocks; i++) {
bucket->bu_bhs[i] = sb_getblk(bucket->bu_inode->i_sb,
xb_blkno + i); if (!bucket->bu_bhs[i]) {
rc = -ENOMEM;
mlog_errno(rc); break;
}
if (!ocfs2_buffer_uptodate(INODE_CACHE(bucket->bu_inode),
bucket->bu_bhs[i])) { if (new)
ocfs2_set_new_buffer_uptodate(INODE_CACHE(bucket->bu_inode),
bucket->bu_bhs[i]); else {
set_buffer_uptodate(bucket->bu_bhs[i]);
ocfs2_set_buffer_uptodate(INODE_CACHE(bucket->bu_inode),
bucket->bu_bhs[i]);
}
}
}
if (rc)
ocfs2_xattr_bucket_relse(bucket); return rc;
}
/* Read the xattr bucket at xb_blkno */ staticint ocfs2_read_xattr_bucket(struct ocfs2_xattr_bucket *bucket,
u64 xb_blkno)
{ int rc;
/* * If the ecc fails, we return the error but otherwise * leave the filesystem running. We know any error is * local to this block.
*/
rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &xb->xb_check); if (rc) return rc;
/* * Errors after here are fatal
*/
if (!OCFS2_IS_VALID_XATTR_BLOCK(xb)) { return ocfs2_error(sb, "Extended attribute block #%llu has bad signature %.*s\n",
(unsignedlonglong)bh->b_blocknr, 7,
xb->xb_signature);
}
if (le64_to_cpu(xb->xb_blkno) != bh->b_blocknr) { return ocfs2_error(sb, "Extended attribute block #%llu has an invalid xb_blkno of %llu\n",
(unsignedlonglong)bh->b_blocknr,
(unsignedlonglong)le64_to_cpu(xb->xb_blkno));
}
if (le32_to_cpu(xb->xb_fs_generation) != OCFS2_SB(sb)->fs_generation) { return ocfs2_error(sb, "Extended attribute block #%llu has an invalid xb_fs_generation of #%u\n",
(unsignedlonglong)bh->b_blocknr,
le32_to_cpu(xb->xb_fs_generation));
}
static u32 ocfs2_xattr_name_hash(struct inode *inode, constchar *name, int name_len)
{ /* Get hash value of uuid from super block */
u32 hash = OCFS2_SB(inode->i_sb)->uuid_hash; int i;
/* hash extended attribute name */ for (i = 0; i < name_len; i++) {
hash = (hash << OCFS2_HASH_SHIFT) ^
(hash >> (8*sizeof(hash) - OCFS2_HASH_SHIFT)) ^
*name++;
}
int ocfs2_calc_security_init(struct inode *dir, struct ocfs2_security_xattr_info *si, int *want_clusters, int *xattr_credits, struct ocfs2_alloc_context **xattr_ac)
{ int ret = 0; struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); int s_size = ocfs2_xattr_entry_real_size(strlen(si->name),
si->value_len);
/* * The max space of security xattr taken inline is * 256(name) + 80(value) + 16(entry) = 352 bytes, * So reserve one metadata block for it is ok.
*/ if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE ||
s_size > OCFS2_XATTR_FREE_IN_IBODY) {
ret = ocfs2_reserve_new_metadata_blocks(osb, 1, xattr_ac); if (ret) {
mlog_errno(ret); return ret;
}
*xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS;
}
/* reserve clusters for xattr value which will be set in B tree*/ if (si->value_len > OCFS2_XATTR_INLINE_SIZE) { int new_clusters = ocfs2_clusters_for_bytes(dir->i_sb,
si->value_len);
/* * The max space of security xattr taken inline is * 256(name) + 80(value) + 16(entry) = 352 bytes, * The max space of acl xattr taken inline is * 80(value) + 16(entry) * 2(if directory) = 192 bytes, * when blocksize = 512, may reserve one more cluster for * xattr bucket, otherwise reserve one metadata block * for them is ok. * If this is a new directory with inline data, * we choose to reserve the entire inline area for * directory contents and force an external xattr block.
*/ if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE ||
(S_ISDIR(mode) && ocfs2_supports_inline_data(osb)) ||
(s_size + a_size) > OCFS2_XATTR_FREE_IN_IBODY) {
*want_meta = *want_meta + 1;
*xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS;
}
/* * reserve credits and clusters for xattrs which has large value * and have to be set outside
*/ if (si->enable && si->value_len > OCFS2_XATTR_INLINE_SIZE) {
new_clusters = ocfs2_clusters_for_bytes(dir->i_sb,
si->value_len);
*xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb,
new_clusters);
*want_clusters += new_clusters;
} if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL &&
acl_len > OCFS2_XATTR_INLINE_SIZE) { /* for directory, it has DEFAULT and ACCESS two types of acls */
new_clusters = (S_ISDIR(mode) ? 2 : 1) *
ocfs2_clusters_for_bytes(dir->i_sb, acl_len);
*xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb,
new_clusters);
*want_clusters += new_clusters;
}
if (why != RESTART_NONE && clusters_to_add) { /* * We can only fail in case the alloc file doesn't give * up enough clusters.
*/
BUG_ON(why == RESTART_META);
credits = ocfs2_calc_extend_credits(inode->i_sb,
&vb->vb_xv->xr_list);
status = ocfs2_extend_trans(handle, credits); if (status < 0) {
status = -ENOMEM;
mlog_errno(status); break;
}
}
}
staticint ocfs2_xattr_list_entries(struct inode *inode, struct ocfs2_xattr_header *header, char *buffer, size_t buffer_size)
{
size_t result = 0; int i, type, ret; constchar *name;
for (i = 0 ; i < le16_to_cpu(header->xh_count); i++) { struct ocfs2_xattr_entry *entry = &header->xh_entries[i];
type = ocfs2_xattr_get_type(entry);
name = (constchar *)header +
le16_to_cpu(entry->xe_name_offset);
ret = ocfs2_xattr_list_entry(inode->i_sb,
buffer, buffer_size,
&result, type, name,
entry->xe_name_len); if (ret) return ret;
}
return result;
}
int ocfs2_has_inline_xattr_value_outside(struct inode *inode, struct ocfs2_dinode *di)
{ struct ocfs2_xattr_header *xh; int i;
for (i = 0; i < num_clusters * bpc; i++, blkno++) {
ret = ocfs2_read_block(INODE_CACHE(inode), blkno,
&bh, NULL); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_journal_access(handle,
INODE_CACHE(inode),
bh,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret < 0) {
mlog_errno(ret); goto out;
}
/* Give a pointer into the storage for the given offset */ staticvoid *ocfs2_xa_offset_pointer(struct ocfs2_xa_loc *loc, int offset)
{
BUG_ON(offset >= loc->xl_size); return loc->xl_ops->xlo_offset_pointer(loc, offset);
}
/* * Wipe the name+value pair and allow the storage to reclaim it. This * must be followed by either removal of the entry or a call to * ocfs2_xa_add_namevalue().
*/ staticvoid ocfs2_xa_wipe_namevalue(struct ocfs2_xa_loc *loc)
{
loc->xl_ops->xlo_wipe_namevalue(loc);
}
/* * Find lowest offset to a name+value pair. This is the start of our * downward-growing free space.
*/ staticint ocfs2_xa_get_free_start(struct ocfs2_xa_loc *loc)
{ return loc->xl_ops->xlo_get_free_start(loc);
}
/* Can we reuse loc->xl_entry for xi? */ staticint ocfs2_xa_can_reuse_entry(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi)
{ return loc->xl_ops->xlo_can_reuse(loc, xi);
}
/* How much free space is needed to set the new value */ staticint ocfs2_xa_check_space(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi)
{ return loc->xl_ops->xlo_check_space(loc, xi);
}
staticvoid ocfs2_xa_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash)
{
loc->xl_ops->xlo_add_entry(loc, name_hash);
loc->xl_entry->xe_name_hash = cpu_to_le32(name_hash); /* * We can't leave the new entry's xe_name_offset at zero or * add_namevalue() will go nuts. We set it to the size of our * storage so that it can never be less than any other entry.
*/
loc->xl_entry->xe_name_offset = cpu_to_le16(loc->xl_size);
}
staticvoid ocfs2_xa_add_namevalue(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi)
{ int size = namevalue_size_xi(xi); int nameval_offset; char *nameval_buf;
staticvoid ocfs2_xa_fill_value_buf(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_value_buf *vb)
{ int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len);
/* Value bufs are for value trees */
BUG_ON(ocfs2_xattr_is_local(loc->xl_entry));
BUG_ON(namevalue_size_xe(loc->xl_entry) !=
(name_size + OCFS2_XATTR_ROOT_SIZE));
staticint ocfs2_xa_block_can_reuse(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi)
{ /* * Block storage is strict. If the sizes aren't exact, we will * remove the old one and reinsert the new.
*/ return namevalue_size_xe(loc->xl_entry) ==
namevalue_size_xi(xi);
}
staticint ocfs2_xa_block_get_free_start(struct ocfs2_xa_loc *loc)
{ struct ocfs2_xattr_header *xh = loc->xl_header; int i, count = le16_to_cpu(xh->xh_count); int offset, free_start = loc->xl_size;
for (i = 0; i < count; i++) {
offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); if (offset < free_start)
free_start = offset;
}
return free_start;
}
staticint ocfs2_xa_block_check_space(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi)
{ int count = le16_to_cpu(loc->xl_header->xh_count); int free_start = ocfs2_xa_get_free_start(loc); int needed_space = ocfs2_xi_entry_usage(xi);
/* * Block storage will reclaim the original entry before inserting * the new value, so we only need the difference. If the new * entry is smaller than the old one, we don't need anything.
*/ if (loc->xl_entry) { /* Don't need space if we're reusing! */ if (ocfs2_xa_can_reuse_entry(loc, xi))
needed_space = 0; else
needed_space -= ocfs2_xe_entry_usage(loc->xl_entry);
} if (needed_space < 0)
needed_space = 0; return ocfs2_xa_check_space_helper(needed_space, free_start, count);
}
/* * Block storage for xattrs keeps the name+value pairs compacted. When * we remove one, we have to shift any that preceded it towards the end.
*/ staticvoid ocfs2_xa_block_wipe_namevalue(struct ocfs2_xa_loc *loc)
{ int i, offset; int namevalue_offset, first_namevalue_offset, namevalue_size; struct ocfs2_xattr_entry *entry = loc->xl_entry; struct ocfs2_xattr_header *xh = loc->xl_header; int count = le16_to_cpu(xh->xh_count);
/* Now tell xh->xh_entries about it */ for (i = 0; i < count; i++) {
offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); if (offset <= namevalue_offset)
le16_add_cpu(&xh->xh_entries[i].xe_name_offset,
namevalue_size);
}
/* * Note that we don't update xh_free_start or xh_name_value_len * because they're not used in block-stored xattrs.
*/
}
staticvoid *ocfs2_xa_bucket_offset_pointer(struct ocfs2_xa_loc *loc, int offset)
{ struct ocfs2_xattr_bucket *bucket = loc->xl_storage; int block, block_offset;
/* The header is at the front of the bucket */
block = offset >> loc->xl_inode->i_sb->s_blocksize_bits;
block_offset = offset % loc->xl_inode->i_sb->s_blocksize;
staticint ocfs2_bucket_align_free_start(struct super_block *sb, int free_start, int size)
{ /* * We need to make sure that the name+value pair fits within * one block.
*/ if (((free_start - size) >> sb->s_blocksize_bits) !=
((free_start - 1) >> sb->s_blocksize_bits))
free_start -= free_start % sb->s_blocksize;
return free_start;
}
staticint ocfs2_xa_bucket_check_space(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi)
{ int rc; int count = le16_to_cpu(loc->xl_header->xh_count); int free_start = ocfs2_xa_get_free_start(loc); int needed_space = ocfs2_xi_entry_usage(xi); int size = namevalue_size_xi(xi); struct super_block *sb = loc->xl_inode->i_sb;
/* * Bucket storage does not reclaim name+value pairs it cannot * reuse. They live as holes until the bucket fills, and then * the bucket is defragmented. However, the bucket can reclaim * the ocfs2_xattr_entry.
*/ if (loc->xl_entry) { /* Don't need space if we're reusing! */ if (ocfs2_xa_can_reuse_entry(loc, xi))
needed_space = 0; else
needed_space -= sizeof(struct ocfs2_xattr_entry);
}
BUG_ON(needed_space < 0);
if (free_start < size) { if (needed_space) return -ENOSPC;
} else { /* * First we check if it would fit in the first place. * Below, we align the free start to a block. This may * slide us below the minimum gap. By checking unaligned * first, we avoid that error.
*/
rc = ocfs2_xa_check_space_helper(needed_space, free_start,
count); if (rc) return rc;
free_start = ocfs2_bucket_align_free_start(sb, free_start,
size);
} return ocfs2_xa_check_space_helper(needed_space, free_start, count);
}
staticvoid ocfs2_xa_bucket_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash)
{ struct ocfs2_xattr_header *xh = loc->xl_header; int count = le16_to_cpu(xh->xh_count); int low = 0, high = count - 1, tmp; struct ocfs2_xattr_entry *tmp_xe;
/* * We keep buckets sorted by name_hash, so we need to find * our insert place.
*/ while (low <= high && count) {
tmp = (low + high) / 2;
tmp_xe = &xh->xh_entries[tmp];
/* Values are not allowed to straddle block boundaries */
BUG_ON(block_offset !=
((nameval_offset + size - 1) >> sb->s_blocksize_bits)); /* We expect the bucket to be filled in */
BUG_ON(!bucket->bu_bhs[block_offset]);
/* * The caller of ocfs2_xa_value_truncate() has already called * ocfs2_xa_journal_access on the loc. However, The truncate code * calls ocfs2_extend_trans(). This may commit the previous * transaction and open a new one. If this is a bucket, truncate * could leave only vb->vb_bh set up for journaling. Meanwhile, * the caller is expecting to dirty the entire bucket. So we must * reset the journal work. We do this even if truncate has failed, * as it could have failed after committing the extend.
*/
access_rc = ocfs2_xa_journal_access(ctxt->handle, loc,
OCFS2_JOURNAL_ACCESS_WRITE);
/* Errors in truncate take precedence */ return trunc_rc ? trunc_rc : access_rc;
}
/* * Only zero out the entry if there are more remaining. This is * important for an empty bucket, as it keeps track of the * bucket's hash value. It doesn't hurt empty block storage.
*/ if (count) {
index = ((char *)entry - (char *)&xh->xh_entries) / sizeof(struct ocfs2_xattr_entry);
memmove(&xh->xh_entries[index], &xh->xh_entries[index + 1],
(count - index) * sizeof(struct ocfs2_xattr_entry));
memset(&xh->xh_entries[count], 0, sizeof(struct ocfs2_xattr_entry));
}
}
/* * If we have a problem adjusting the size of an external value during * ocfs2_xa_prepare_entry() or ocfs2_xa_remove(), we may have an xattr * in an intermediate state. For example, the value may be partially * truncated. * * If the value tree hasn't changed, the extend/truncate went nowhere. * We have nothing to do. The caller can treat it as a straight error. * * If the value tree got partially truncated, we now have a corrupted * extended attribute. We're going to wipe its entry and leak the * clusters. Better to leak some storage than leave a corrupt entry. * * If the value tree grew, it obviously didn't grow enough for the * new entry. We're not going to try and reclaim those clusters either. * If there was already an external value there (orig_clusters != 0), * the new clusters are attached safely and we can just leave the old * value in place. If there was no external value there, we remove * the entry. * * This way, the xattr block we store in the journal will be consistent. * If the size change broke because of the journal, no changes will hit * disk anyway.
*/ staticvoid ocfs2_xa_cleanup_value_truncate(struct ocfs2_xa_loc *loc, constchar *what, unsignedint orig_clusters)
{ unsignedint new_clusters = ocfs2_xa_value_clusters(loc); char *nameval_buf = ocfs2_xa_offset_pointer(loc,
le16_to_cpu(loc->xl_entry->xe_name_offset));
if (new_clusters < orig_clusters) {
mlog(ML_ERROR, "Partial truncate while %s xattr %.*s. Leaking " "%u clusters and removing the entry\n",
what, loc->xl_entry->xe_name_len, nameval_buf,
orig_clusters - new_clusters);
ocfs2_xa_remove_entry(loc);
} elseif (!orig_clusters) {
mlog(ML_ERROR, "Unable to allocate an external value for xattr " "%.*s safely. Leaking %u clusters and removing the " "entry\n",
loc->xl_entry->xe_name_len, nameval_buf,
new_clusters - orig_clusters);
ocfs2_xa_remove_entry(loc);
} elseif (new_clusters > orig_clusters)
mlog(ML_ERROR, "Unable to grow xattr %.*s safely. %u new clusters " "have been added, but the value will not be " "modified\n",
loc->xl_entry->xe_name_len, nameval_buf,
new_clusters - orig_clusters);
}
if (!ocfs2_xattr_is_local(loc->xl_entry)) {
orig_clusters = ocfs2_xa_value_clusters(loc);
rc = ocfs2_xa_value_truncate(loc, 0, ctxt); if (rc) {
mlog_errno(rc); /* * Since this is remove, we can return 0 if * ocfs2_xa_cleanup_value_truncate() is going to * wipe the entry anyway. So we check the * cluster count as well.
*/ if (orig_clusters != ocfs2_xa_value_clusters(loc))
rc = 0;
ocfs2_xa_cleanup_value_truncate(loc, "removing",
orig_clusters); goto out;
}
}
/* * Take an existing entry and make it ready for the new value. This * won't allocate space, but it may free space. It should be ready for * ocfs2_xa_prepare_entry() to finish the work.
*/ staticint ocfs2_xa_reuse_entry(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi, struct ocfs2_xattr_set_ctxt *ctxt)
{ int rc = 0; int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len); unsignedint orig_clusters; char *nameval_buf; int xe_local = ocfs2_xattr_is_local(loc->xl_entry); int xi_local = xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE;
/* * Prepares loc->xl_entry to receive the new xattr. This includes * properly setting up the name+value pair region. If loc->xl_entry * already exists, it will take care of modifying it appropriately. * * Note that this modifies the data. You did journal_access already, * right?
*/ staticint ocfs2_xa_prepare_entry(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi,
u32 name_hash, struct ocfs2_xattr_set_ctxt *ctxt)
{ int rc = 0; unsignedint orig_clusters;
__le64 orig_value_size = 0;
rc = ocfs2_xa_check_space(loc, xi); if (rc) goto out;
if (loc->xl_entry) { if (ocfs2_xa_can_reuse_entry(loc, xi)) {
orig_value_size = loc->xl_entry->xe_value_size;
rc = ocfs2_xa_reuse_entry(loc, xi, ctxt); if (rc) goto out; goto alloc_value;
}
/* * If we get here, we have a blank entry. Fill it. We grow our * name+value pair back from the end.
*/
ocfs2_xa_add_namevalue(loc, xi); if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE)
ocfs2_xa_install_value_root(loc);
alloc_value: if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) {
orig_clusters = ocfs2_xa_value_clusters(loc);
rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len, ctxt); if (rc < 0) {
ctxt->set_abort = 1;
ocfs2_xa_cleanup_value_truncate(loc, "growing",
orig_clusters); /* * If we were growing an existing value, * ocfs2_xa_cleanup_value_truncate() won't remove * the entry. We need to restore the original value * size.
*/ if (loc->xl_entry) {
BUG_ON(!orig_value_size);
loc->xl_entry->xe_value_size = orig_value_size;
}
mlog_errno(rc);
}
}
out: return rc;
}
/* * Store the value portion of the name+value pair. This will skip * values that are stored externally. Their tree roots were set up * by ocfs2_xa_prepare_entry().
*/ staticint ocfs2_xa_store_value(struct ocfs2_xa_loc *loc, struct ocfs2_xattr_info *xi, struct ocfs2_xattr_set_ctxt *ctxt)
{ int rc = 0; int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len); char *nameval_buf; struct ocfs2_xattr_value_buf vb;
ret = ocfs2_xa_journal_access(ctxt->handle, loc,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out;
}
/* * From here on out, everything is going to modify the buffer a * little. Errors are going to leave the xattr header in a * sane state. Thus, even with errors we dirty the sucker.
*/
/* Don't worry, we are never called with !xi_value and !xl_entry */ if (!xi->xi_value) {
ret = ocfs2_xa_remove(loc, ctxt); goto out_dirty;
}
ret = ocfs2_xa_prepare_entry(loc, xi, name_hash, ctxt); if (ret) { if (ret != -ENOSPC)
mlog_errno(ret); goto out_dirty;
}
ret = ocfs2_xa_store_value(loc, xi, ctxt); if (ret)
mlog_errno(ret);
/* * In xattr remove, if it is stored outside and refcounted, we may have * the chance to split the refcount tree. So need the allocators.
*/ staticint ocfs2_lock_xattr_remove_allocators(struct inode *inode, struct ocfs2_xattr_value_root *xv, struct ocfs2_caching_info *ref_ci, struct buffer_head *ref_root_bh, struct ocfs2_alloc_context **meta_ac, int *ref_credits)
{ int ret, meta_add = 0;
u32 p_cluster, num_clusters; unsignedint ext_flags;
*ref_credits = 0;
ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster,
&num_clusters,
&xv->xr_list,
&ext_flags); if (ret) {
mlog_errno(ret); goto out;
}
if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) goto out;
ret = ocfs2_refcounted_xattr_delete_need(inode, ref_ci,
ref_root_bh, xv,
&meta_add, ref_credits); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(inode->i_sb),
meta_add, meta_ac); if (ret)
mlog_errno(ret);
/* Find the named attribute. */ if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) {
ret = ocfs2_xattr_find_entry(inode, name_index, name, xs); if (ret && ret != -ENODATA) return ret;
xs->not_found = ret;
}
ret = ocfs2_xa_set(&loc, xi, ctxt); if (!ret)
xs->here = loc.xl_entry; elseif ((ret != -ENOSPC) || ctxt->set_abort) goto end; else {
ret = ocfs2_xattr_create_index_block(inode, xs, ctxt); if (ret) goto end;
}
}
if (le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED)
ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt);
end: return ret;
}
/* Check whether the new xattr can be inserted into the inode. */ staticint ocfs2_xattr_can_be_in_inode(struct inode *inode, struct ocfs2_xattr_info *xi, struct ocfs2_xattr_search *xs)
{ struct ocfs2_xattr_entry *last; int free, i;
size_t min_offs = xs->end - xs->base;
if (!xs->header) return 0;
last = xs->header->xh_entries;
for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) {
size_t offs = le16_to_cpu(last->xe_name_offset); if (offs < min_offs)
min_offs = offs;
last += 1;
}
/* * Calculate the clusters we need to write. * No matter whether we replace an old one or add a new one, * we need this for writing.
*/ if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE)
credits += new_clusters *
ocfs2_clusters_to_blocks(inode->i_sb, 1);
if (xis->not_found && xbs->not_found) {
credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
if (!xis->not_found) {
xe = xis->here;
name_offset = le16_to_cpu(xe->xe_name_offset);
name_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
base = xis->base;
credits += OCFS2_INODE_UPDATE_CREDITS;
} else { int i, block_off = 0;
xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data;
xe = xbs->here;
name_offset = le16_to_cpu(xe->xe_name_offset);
name_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
i = xbs->here - xbs->header->xh_entries;
old_in_xb = 1;
if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) {
ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb,
bucket_xh(xbs->bucket),
i, &block_off,
&name_offset);
base = bucket_block(xbs->bucket, block_off);
credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
} else {
base = xbs->base;
credits += OCFS2_XATTR_BLOCK_UPDATE_CREDITS;
}
}
/* * delete a xattr doesn't need metadata and cluster allocation. * so just calculate the credits and return. * * The credits for removing the value tree will be extended * by ocfs2_remove_extent itself.
*/ if (!xi->xi_value) { if (!ocfs2_xattr_is_local(xe))
credits += ocfs2_remove_extent_credits(inode->i_sb);
goto out;
}
/* do cluster allocation guess first. */
value_size = le64_to_cpu(xe->xe_value_size);
if (old_in_xb) { /* * In xattr set, we always try to set the xe in inode first, * so if it can be inserted into inode successfully, the old * one will be removed from the xattr block, and this xattr * will be inserted into inode as a new xattr in inode.
*/ if (ocfs2_xattr_can_be_in_inode(inode, xi, xis)) {
clusters_add += new_clusters;
credits += ocfs2_remove_extent_credits(inode->i_sb) +
OCFS2_INODE_UPDATE_CREDITS; if (!ocfs2_xattr_is_local(xe))
credits += ocfs2_calc_extend_credits(
inode->i_sb,
&def_xv.xv.xr_list); goto out;
}
}
if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { /* the new values will be stored outside. */
u32 old_clusters = 0;
if (!ocfs2_xattr_is_local(xe)) {
old_clusters = ocfs2_clusters_for_bytes(inode->i_sb,
value_size);
xv = (struct ocfs2_xattr_value_root *)
(base + name_offset + name_len);
value_size = OCFS2_XATTR_ROOT_SIZE;
} else
xv = &def_xv.xv;
if (old_clusters >= new_clusters) {
credits += ocfs2_remove_extent_credits(inode->i_sb); goto out;
} else {
meta_add += ocfs2_extend_meta_needed(&xv->xr_list);
clusters_add += new_clusters - old_clusters;
credits += ocfs2_calc_extend_credits(inode->i_sb,
&xv->xr_list); if (value_size >= OCFS2_XATTR_ROOT_SIZE) goto out;
}
} else { /* * Now the new value will be stored inside. So if the new * value is smaller than the size of value root or the old * value, we don't need any allocation, otherwise we have * to guess metadata allocation.
*/ if ((ocfs2_xattr_is_local(xe) &&
(value_size >= xi->xi_value_len)) ||
(!ocfs2_xattr_is_local(xe) &&
OCFS2_XATTR_ROOT_SIZE >= xi->xi_value_len)) goto out;
}
meta_guess: /* calculate metadata allocation. */ if (di->i_xattr_loc) { if (!xbs->xattr_bh) {
ret = ocfs2_read_xattr_block(inode,
le64_to_cpu(di->i_xattr_loc),
&bh); if (ret) {
mlog_errno(ret); goto out;
}
/* * If there is already an xattr tree, good, we can calculate * like other b-trees. Otherwise we may have the chance of * create a tree, the credit calculation is borrowed from * ocfs2_calc_extend_credits with root_el = NULL. And the * new tree will be cluster based, so no meta is needed.
*/ if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { struct ocfs2_extent_list *el =
&xb->xb_attrs.xb_root.xt_list;
meta_add += ocfs2_extend_meta_needed(el);
credits += ocfs2_calc_extend_credits(inode->i_sb,
el);
} else
credits += OCFS2_SUBALLOC_ALLOC + 1;
/* * This cluster will be used either for new bucket or for * new xattr block. * If the cluster size is the same as the bucket size, one * more is needed since we may need to extend the bucket * also.
*/
clusters_add += 1;
credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); if (OCFS2_XATTR_BUCKET_SIZE ==
OCFS2_SB(inode->i_sb)->s_clustersize) {
credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
clusters_add += 1;
}
} else {
credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { struct ocfs2_extent_list *el = &def_xv.xv.xr_list;
meta_add += ocfs2_extend_meta_needed(el);
credits += ocfs2_calc_extend_credits(inode->i_sb,
el);
} else {
meta_add += 1;
}
}
out: if (clusters_need)
*clusters_need = clusters_add; if (meta_need)
*meta_need = meta_add; if (credits_need)
*credits_need = credits;
brelse(bh); return ret;
}
if (meta_add) {
ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add,
&ctxt->meta_ac); if (ret) {
mlog_errno(ret); goto out;
}
}
if (clusters_add) {
ret = ocfs2_reserve_clusters(osb, clusters_add, &ctxt->data_ac); if (ret)
mlog_errno(ret);
}
out: if (ret) { if (ctxt->meta_ac) {
ocfs2_free_alloc_context(ctxt->meta_ac);
ctxt->meta_ac = NULL;
}
/* * We cannot have an error and a non null ctxt->data_ac.
*/
}
if (!xi->xi_value) { /* Remove existing extended attribute */ if (!xis->not_found)
ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); elseif (!xbs->not_found)
ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt);
} else { /* We always try to set extended attribute into inode first*/
ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); if (!ret && !xbs->not_found) { /* * If succeed and that extended attribute existing in * external block, then we will remove it.
*/
xi->xi_value = NULL;
xi->xi_value_len = 0;
ret = ocfs2_extend_trans(ctxt->handle, credits); if (ret) {
mlog_errno(ret); goto out;
}
} /* * If no space in inode, we will set extended attribute * into external block.
*/
ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); if (ret) goto out; if (!xis->not_found) { /* * If succeed and that extended attribute * existing in inode, we will remove it.
*/
xi->xi_value = NULL;
xi->xi_value_len = 0;
xbs->not_found = -ENODATA;
ret = ocfs2_calc_xattr_set_need(inode,
di,
xi,
xis,
xbs,
NULL,
NULL,
&credits); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_extend_trans(ctxt->handle, credits); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_xattr_ibody_set(inode, xi,
xis, ctxt);
}
}
}
if (!ret) { /* Update inode ctime. */
ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode),
xis->inode_bh,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out;
}
/* * This function only called duing creating inode * for init security/acl xattrs of the new inode. * All transanction credits have been reserved in mknod.
*/ int ocfs2_xattr_set_handle(handle_t *handle, struct inode *inode, struct buffer_head *di_bh, int name_index, constchar *name, constvoid *value,
size_t value_len, int flags, struct ocfs2_alloc_context *meta_ac, struct ocfs2_alloc_context *data_ac)
{ struct ocfs2_dinode *di; int ret;
if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) return -EOPNOTSUPP;
/* * In extreme situation, may need xattr bucket when * block size is too small. And we have already reserved * the credits for bucket in mknod.
*/ if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) {
xbs.bucket = ocfs2_xattr_bucket_new(inode); if (!xbs.bucket) {
mlog_errno(-ENOMEM); return -ENOMEM;
}
}
xis.inode_bh = xbs.inode_bh = di_bh;
di = (struct ocfs2_dinode *)di_bh->b_data;
down_write(&OCFS2_I(inode)->ip_xattr_sem);
ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); if (ret) goto cleanup; if (xis.not_found) {
ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs); if (ret) goto cleanup;
}
ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt);
if (!ocfs2_supports_xattr(osb)) return -EOPNOTSUPP;
/* * Only xbs will be used on indexed trees. xis doesn't need a * bucket.
*/
xbs.bucket = ocfs2_xattr_bucket_new(inode); if (!xbs.bucket) {
mlog_errno(-ENOMEM); return -ENOMEM;
}
had_lock = ocfs2_inode_lock_tracker(inode, &di_bh, 1, &oh); if (had_lock < 0) {
ret = had_lock;
mlog_errno(ret); goto cleanup_nolock;
}
xis.inode_bh = xbs.inode_bh = di_bh;
di = (struct ocfs2_dinode *)di_bh->b_data;
down_write(&OCFS2_I(inode)->ip_xattr_sem); /* * Scan inode and external block to find the same name * extended attribute and collect search information.
*/
ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); if (ret) goto cleanup; if (xis.not_found) {
ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs); if (ret) goto cleanup;
}
if (xis.not_found && xbs.not_found) {
ret = -ENODATA; if (flags & XATTR_REPLACE) goto cleanup;
ret = 0; if (!value) goto cleanup;
} else {
ret = -EEXIST; if (flags & XATTR_CREATE) goto cleanup;
}
/* Check whether the value is refcounted and do some preparation. */ if (ocfs2_is_refcount_inode(inode) &&
(!xis.not_found || !xbs.not_found)) {
ret = ocfs2_prepare_refcount_xattr(inode, di, &xi,
&xis, &xbs, &ref_tree,
&ref_meta, &ref_credits); if (ret) {
mlog_errno(ret); goto cleanup;
}
}
inode_lock(tl_inode);
if (ocfs2_truncate_log_needs_flush(osb)) {
ret = __ocfs2_flush_truncate_log(osb); if (ret < 0) {
inode_unlock(tl_inode);
mlog_errno(ret); goto cleanup;
}
}
inode_unlock(tl_inode);
ret = ocfs2_init_xattr_set_ctxt(inode, di, &xi, &xis,
&xbs, &ctxt, ref_meta, &credits); if (ret) {
mlog_errno(ret); goto cleanup;
}
/* we need to update inode's ctime field, so add credit for it. */
credits += OCFS2_INODE_UPDATE_CREDITS;
ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); if (IS_ERR(ctxt.handle)) {
ret = PTR_ERR(ctxt.handle);
mlog_errno(ret); goto out_free_ac;
}
out_free_ac: if (ctxt.data_ac)
ocfs2_free_alloc_context(ctxt.data_ac); if (ctxt.meta_ac)
ocfs2_free_alloc_context(ctxt.meta_ac); if (ocfs2_dealloc_has_cluster(&ctxt.dealloc))
ocfs2_schedule_truncate_log_flush(osb, 1);
ocfs2_run_deallocs(osb, &ctxt.dealloc);
cleanup: if (ref_tree)
ocfs2_unlock_refcount_tree(osb, ref_tree, 1);
up_write(&OCFS2_I(inode)->ip_xattr_sem); if (!value && !ret) {
ret = ocfs2_try_remove_refcount_tree(inode, di_bh); if (ret)
mlog_errno(ret);
}
ocfs2_inode_unlock_tracker(inode, 1, &oh, had_lock);
cleanup_nolock:
brelse(di_bh);
brelse(xbs.xattr_bh);
ocfs2_xattr_bucket_free(xbs.bucket);
return ret;
}
/* * Find the xattr extent rec which may contains name_hash. * e_cpos will be the first name hash of the xattr rec. * el must be the ocfs2_xattr_header.xb_attrs.xb_root.xt_list.
*/ staticint ocfs2_xattr_get_rec(struct inode *inode,
u32 name_hash,
u64 *p_blkno,
u32 *e_cpos,
u32 *num_clusters, struct ocfs2_extent_list *el)
{ int ret = 0, i; struct buffer_head *eb_bh = NULL; struct ocfs2_extent_block *eb; struct ocfs2_extent_rec *rec = NULL;
u64 e_blkno = 0;
if (el->l_tree_depth) {
ret = ocfs2_find_leaf(INODE_CACHE(inode), el, name_hash,
&eb_bh); if (ret) {
mlog_errno(ret); goto out;
}
eb = (struct ocfs2_extent_block *) eb_bh->b_data;
el = &eb->h_list;
if (el->l_tree_depth) {
ret = ocfs2_error(inode->i_sb, "Inode %lu has non zero tree depth in xattr tree block %llu\n",
inode->i_ino,
(unsignedlonglong)eb_bh->b_blocknr); goto out;
}
}
for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) {
rec = &el->l_recs[i];
if (!e_blkno) {
ret = ocfs2_error(inode->i_sb, "Inode %lu has bad extent record (%u, %u, 0) in xattr\n",
inode->i_ino,
le32_to_cpu(rec->e_cpos),
ocfs2_rec_clusters(el, rec)); goto out;
}
staticint ocfs2_find_xe_in_bucket(struct inode *inode, struct ocfs2_xattr_bucket *bucket, int name_index, constchar *name,
u32 name_hash,
u16 *xe_index, int *found)
{ int i, ret = 0, cmp = 1, block_off, new_offset; struct ocfs2_xattr_header *xh = bucket_xh(bucket);
size_t name_len = strlen(name); struct ocfs2_xattr_entry *xe = NULL; char *xe_name;
/* * We don't use binary search in the bucket because there * may be multiple entries with the same name hash.
*/ for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
xe = &xh->xh_entries[i];
if (name_hash > le32_to_cpu(xe->xe_name_hash)) continue; elseif (name_hash < le32_to_cpu(xe->xe_name_hash)) break;
cmp = name_index - ocfs2_xattr_get_type(xe); if (!cmp)
cmp = name_len - xe->xe_name_len; if (cmp) continue;
ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb,
xh,
i,
&block_off,
&new_offset); if (ret) {
mlog_errno(ret); break;
}
/* * Find the specified xattr entry in a series of buckets. * This series start from p_blkno and last for num_clusters. * The ocfs2_xattr_header.xh_num_buckets of the first bucket contains * the num of the valid buckets. * * Return the buffer_head this xattr should reside in. And if the xattr's * hash is in the gap of 2 buckets, return the lower bucket.
*/ staticint ocfs2_xattr_bucket_find(struct inode *inode, int name_index, constchar *name,
u32 name_hash,
u64 p_blkno,
u32 first_hash,
u32 num_clusters, struct ocfs2_xattr_search *xs)
{ int ret, found = 0; struct ocfs2_xattr_header *xh = NULL; struct ocfs2_xattr_entry *xe = NULL;
u16 index = 0;
u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); int low_bucket = 0, bucket, high_bucket; struct ocfs2_xattr_bucket *search;
u64 blkno, lower_blkno = 0;
search = ocfs2_xattr_bucket_new(inode); if (!search) {
ret = -ENOMEM;
mlog_errno(ret); goto out;
}
ret = ocfs2_read_xattr_bucket(search, p_blkno); if (ret) {
mlog_errno(ret); goto out;
}
/* * Check whether the hash of the last entry in our * bucket is larger than the search one. for an empty * bucket, the last one is also the first one.
*/ if (xh->xh_count)
xe = &xh->xh_entries[le16_to_cpu(xh->xh_count) - 1];
/* record lower_blkno which may be the insert place. */
lower_blkno = blkno;
/* the searched xattr should reside in this bucket if exists. */
ret = ocfs2_find_xe_in_bucket(inode, search,
name_index, name, name_hash,
&index, &found); if (ret) {
mlog_errno(ret); goto out;
} break;
}
/* * Record the bucket we have found. * When the xattr's hash value is in the gap of 2 buckets, we will * always set it to the previous bucket.
*/ if (!lower_blkno)
lower_blkno = p_blkno;
/* This should be in cache - we just read it during the search */
ret = ocfs2_read_xattr_bucket(xs->bucket, lower_blkno); if (ret) {
mlog_errno(ret); goto out;
}
for (i = 0; i < num_buckets; i++, blkno += bucket->bu_blocks) {
ret = ocfs2_read_xattr_bucket(bucket, blkno); if (ret) {
mlog_errno(ret); break;
}
/* * The real bucket num in this series of blocks is stored * in the 1st bucket.
*/ if (i == 0)
num_buckets = le16_to_cpu(bucket_xh(bucket)->xh_num_buckets);
trace_ocfs2_iterate_xattr_bucket((unsignedlonglong)blkno,
le32_to_cpu(bucket_xh(bucket)->xh_entries[0].xe_name_hash)); if (func) {
ret = func(inode, bucket, para); if (ret && ret != -ERANGE)
mlog_errno(ret); /* Fall through to bucket_relse() */
}
ocfs2_xattr_bucket_relse(bucket); if (ret) break;
}
staticint ocfs2_list_xattr_bucket(struct inode *inode, struct ocfs2_xattr_bucket *bucket, void *para)
{ int ret = 0, type; struct ocfs2_xattr_tree_list *xl = (struct ocfs2_xattr_tree_list *)para; int i, block_off, new_offset; constchar *name;
for (i = 0 ; i < le16_to_cpu(bucket_xh(bucket)->xh_count); i++) { struct ocfs2_xattr_entry *entry = &bucket_xh(bucket)->xh_entries[i];
type = ocfs2_xattr_get_type(entry);
ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb,
bucket_xh(bucket),
i,
&block_off,
&new_offset); if (ret) break;
name = (constchar *)bucket_block(bucket, block_off) +
new_offset;
ret = ocfs2_xattr_list_entry(inode->i_sb,
xl->buffer,
xl->buffer_size,
&xl->result,
type, name,
entry->xe_name_len); if (ret) break;
}
if (l_hash > r_hash) return 1; if (l_hash < r_hash) return -1; return 0;
}
/* * When the ocfs2_xattr_block is filled up, new bucket will be created * and all the xattr entries will be moved to the new bucket. * The header goes at the start of the bucket, and the names+values are * filled from the end. This is why *target starts as the last buffer. * Note: we need to sort the entries since they are not saved in order * in the ocfs2_xattr_block.
*/ staticvoid ocfs2_cp_xattr_block_to_bucket(struct inode *inode, struct buffer_head *xb_bh, struct ocfs2_xattr_bucket *bucket)
{ int i, blocksize = inode->i_sb->s_blocksize; int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
u16 offset, size, off_change; struct ocfs2_xattr_entry *xe; struct ocfs2_xattr_block *xb =
(struct ocfs2_xattr_block *)xb_bh->b_data; struct ocfs2_xattr_header *xb_xh = &xb->xb_attrs.xb_header; struct ocfs2_xattr_header *xh = bucket_xh(bucket);
u16 count = le16_to_cpu(xb_xh->xh_count); char *src = xb_bh->b_data; char *target = bucket_block(bucket, blks - 1);
for (i = 0; i < blks; i++)
memset(bucket_block(bucket, i), 0, blocksize);
/* * Since the xe_name_offset is based on ocfs2_xattr_header, * there is a offset change corresponding to the change of * ocfs2_xattr_header's position.
*/
off_change = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header);
xe = &xb_xh->xh_entries[count - 1];
offset = le16_to_cpu(xe->xe_name_offset) + off_change;
size = blocksize - offset;
/* copy all the names and values. */
memcpy(target + offset, src + offset, size);
/* Change the xe offset for all the xe because of the move. */
off_change = OCFS2_XATTR_BUCKET_SIZE - blocksize +
offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); for (i = 0; i < count; i++)
le16_add_cpu(&xh->xh_entries[i].xe_name_offset, off_change);
/* * After we move xattr from block to index btree, we have to * update ocfs2_xattr_search to the new xe and base. * * When the entry is in xattr block, xattr_bh indicates the storage place. * While if the entry is in index b-tree, "bucket" indicates the * real place of the xattr.
*/ staticvoid ocfs2_xattr_update_xattr_search(struct inode *inode, struct ocfs2_xattr_search *xs, struct buffer_head *old_bh)
{ char *buf = old_bh->b_data; struct ocfs2_xattr_block *old_xb = (struct ocfs2_xattr_block *)buf; struct ocfs2_xattr_header *old_xh = &old_xb->xb_attrs.xb_header; int i;
/* * XXX: * We can use this lock for now, and maybe move to a dedicated mutex * if performance becomes a problem later.
*/
down_write(&oi->ip_alloc_sem);
ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), xb_bh,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out;
}
ret = __ocfs2_claim_clusters(handle, ctxt->data_ac,
1, 1, &bit_off, &len); if (ret) {
mlog_errno(ret); goto out;
}
/* * The bucket may spread in many blocks, and * we will only touch the 1st block and the last block * in the whole bucket(one for entry and one for data).
*/
blkno = ocfs2_clusters_to_blocks(inode->i_sb, bit_off);
if (l_name_offset < r_name_offset) return 1; if (l_name_offset > r_name_offset) return -1; return 0;
}
/* * defrag a xattr bucket if we find that the bucket has some * holes between name/value pairs. * We will move all the name/value pairs to the end of the bucket * so that we can spare some space for insertion.
*/ staticint ocfs2_defrag_xattr_bucket(struct inode *inode,
handle_t *handle, struct ocfs2_xattr_bucket *bucket)
{ int ret, i;
size_t end, offset, len; struct ocfs2_xattr_header *xh; char *entries, *buf, *bucket_buf = NULL;
u64 blkno = bucket_blkno(bucket);
u16 xh_free_start;
size_t blocksize = inode->i_sb->s_blocksize; struct ocfs2_xattr_entry *xe;
/* * In order to make the operation more efficient and generic, * we copy all the blocks into a contiguous memory and do the * defragment there, so if anything is error, we will not touch * the real block.
*/
bucket_buf = kmalloc(OCFS2_XATTR_BUCKET_SIZE, GFP_NOFS); if (!bucket_buf) {
ret = -EIO; goto out;
}
buf = bucket_buf; for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize)
memcpy(buf, bucket_block(bucket, i), blocksize);
ret = ocfs2_xattr_bucket_journal_access(handle, bucket,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret < 0) {
mlog_errno(ret); goto out;
}
/* * sort all the entries by their offset. * the largest will be the first, so that we can * move them to the end one by one.
*/
sort(entries, le16_to_cpu(xh->xh_count), sizeof(struct ocfs2_xattr_entry),
cmp_xe_offset, NULL);
/* Move all name/values to the end of the bucket. */
xe = xh->xh_entries;
end = OCFS2_XATTR_BUCKET_SIZE; for (i = 0; i < le16_to_cpu(xh->xh_count); i++, xe++) {
offset = le16_to_cpu(xe->xe_name_offset);
len = namevalue_size_xe(xe);
/* * We must make sure that the name/value pair * exist in the same block. So adjust end to * the previous block end if needed.
*/ if (((end - len) / blocksize !=
(end - 1) / blocksize))
end = end - end % blocksize;
memset(bucket_buf + xh_free_start, 0, end - xh_free_start);
xh->xh_free_start = cpu_to_le16(end);
/* sort the entries by their name_hash. */
sort(entries, le16_to_cpu(xh->xh_count), sizeof(struct ocfs2_xattr_entry),
cmp_xe, NULL);
buf = bucket_buf; for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize)
memcpy(bucket_block(bucket, i), buf, blocksize);
ocfs2_xattr_bucket_journal_dirty(handle, bucket);
out:
kfree(bucket_buf); return ret;
}
/* * prev_blkno points to the start of an existing extent. new_blkno * points to a newly allocated extent. Because we know each of our * clusters contains more than bucket, we can easily split one cluster * at a bucket boundary. So we take the last cluster of the existing * extent and split it down the middle. We move the last half of the * buckets in the last cluster of the existing extent over to the new * extent. * * first_bh is the buffer at prev_blkno so we can update the existing * extent's bucket count. header_bh is the bucket were we were hoping * to insert our xattr. If the bucket move places the target in the new * extent, we'll update first_bh and header_bh after modifying the old * extent. * * first_hash will be set as the 1st xe's name_hash in the new extent.
*/ staticint ocfs2_mv_xattr_bucket_cross_cluster(struct inode *inode,
handle_t *handle, struct ocfs2_xattr_bucket *first, struct ocfs2_xattr_bucket *target,
u64 new_blkno,
u32 num_clusters,
u32 *first_hash)
{ int ret; struct super_block *sb = inode->i_sb; int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(sb); int num_buckets = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(sb)); int to_move = num_buckets / 2;
u64 src_blkno;
u64 last_cluster_blkno = bucket_blkno(first) +
((num_clusters - 1) * ocfs2_clusters_to_blocks(sb, 1));
ret = ocfs2_mv_xattr_buckets(inode, handle, bucket_blkno(first),
last_cluster_blkno, new_blkno,
to_move, first_hash); if (ret) {
mlog_errno(ret); goto out;
}
/* This is the first bucket that got moved */
src_blkno = last_cluster_blkno + (to_move * blks_per_bucket);
/* * If the target bucket was part of the moved buckets, we need to * update first and target.
*/ if (bucket_blkno(target) >= src_blkno) { /* Find the block for the new target bucket */
src_blkno = new_blkno +
(bucket_blkno(target) - src_blkno);
/* * These shouldn't fail - the buffers are in the * journal from ocfs2_cp_xattr_bucket().
*/
ret = ocfs2_read_xattr_bucket(first, new_blkno); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_read_xattr_bucket(target, src_blkno); if (ret)
mlog_errno(ret);
}
out: return ret;
}
/* * Find the suitable pos when we divide a bucket into 2. * We have to make sure the xattrs with the same hash value exist * in the same bucket. * * If this ocfs2_xattr_header covers more than one hash value, find a * place where the hash value changes. Try to find the most even split. * The most common case is that all entries have different hash values, * and the first check we make will find a place to split.
*/ staticint ocfs2_xattr_find_divide_pos(struct ocfs2_xattr_header *xh)
{ struct ocfs2_xattr_entry *entries = xh->xh_entries; int count = le16_to_cpu(xh->xh_count); int delta, middle = count / 2;
/* * We start at the middle. Each step gets farther away in both * directions. We therefore hit the change in hash value * nearest to the middle. Note that this loop does not execute for * count < 2.
*/ for (delta = 0; delta < middle; delta++) { /* Let's check delta earlier than middle */ if (cmp_xe(&entries[middle - delta - 1],
&entries[middle - delta])) return middle - delta;
/* For even counts, don't walk off the end */ if ((middle + delta + 1) == count) continue;
/* Now try delta past middle */ if (cmp_xe(&entries[middle + delta],
&entries[middle + delta + 1])) return middle + delta + 1;
}
/* Every entry had the same hash */ return count;
}
/* * Move some xattrs in old bucket(blk) to new bucket(new_blk). * first_hash will record the 1st hash of the new bucket. * * Normally half of the xattrs will be moved. But we have to make * sure that the xattrs with the same hash value are stored in the * same bucket. If all the xattrs in this bucket have the same hash * value, the new bucket will be initialized as an empty one and the * first_hash will be initialized as (hash_value+1).
*/ staticint ocfs2_divide_xattr_bucket(struct inode *inode,
handle_t *handle,
u64 blk,
u64 new_blk,
u32 *first_hash, int new_bucket_head)
{ int ret, i; int count, start, len, name_value_len = 0, name_offset = 0; struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL; struct ocfs2_xattr_header *xh; struct ocfs2_xattr_entry *xe; int blocksize = inode->i_sb->s_blocksize;
s_bucket = ocfs2_xattr_bucket_new(inode);
t_bucket = ocfs2_xattr_bucket_new(inode); if (!s_bucket || !t_bucket) {
ret = -ENOMEM;
mlog_errno(ret); goto out;
}
ret = ocfs2_read_xattr_bucket(s_bucket, blk); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_xattr_bucket_journal_access(handle, s_bucket,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out;
}
/* * Even if !new_bucket_head, we're overwriting t_bucket. Thus, * there's no need to read it.
*/
ret = ocfs2_init_xattr_bucket(t_bucket, new_blk, new_bucket_head); if (ret) {
mlog_errno(ret); goto out;
}
/* * Hey, if we're overwriting t_bucket, what difference does * ACCESS_CREATE vs ACCESS_WRITE make? See the comment in the * same part of ocfs2_cp_xattr_bucket().
*/
ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket,
new_bucket_head ?
OCFS2_JOURNAL_ACCESS_CREATE :
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out;
}
if (start == count) {
xe = &xh->xh_entries[start-1];
/* * initialized a new empty bucket here. * The hash value is set as one larger than * that of the last entry in the previous bucket.
*/ for (i = 0; i < t_bucket->bu_blocks; i++)
memset(bucket_block(t_bucket, i), 0, blocksize);
/* copy the whole bucket to the new first. */
ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket);
/* update the new bucket. */
xh = bucket_xh(t_bucket);
/* * Calculate the total name/value len and xh_free_start for * the old bucket first.
*/
name_offset = OCFS2_XATTR_BUCKET_SIZE;
name_value_len = 0; for (i = 0; i < start; i++) {
xe = &xh->xh_entries[i];
name_value_len += namevalue_size_xe(xe); if (le16_to_cpu(xe->xe_name_offset) < name_offset)
name_offset = le16_to_cpu(xe->xe_name_offset);
}
/* * Now begin the modification to the new bucket. * * In the new bucket, We just move the xattr entry to the beginning * and don't touch the name/value. So there will be some holes in the * bucket, and they will be removed when ocfs2_defrag_xattr_bucket is * called.
*/
xe = &xh->xh_entries[start];
len = sizeof(struct ocfs2_xattr_entry) * (count - start);
trace_ocfs2_divide_xattr_bucket_move(len,
(int)((char *)xe - (char *)xh),
(int)((char *)xh->xh_entries - (char *)xh));
memmove((char *)xh->xh_entries, (char *)xe, len);
xe = &xh->xh_entries[count - start];
len = sizeof(struct ocfs2_xattr_entry) * start;
memset((char *)xe, 0, len);
/* Calculate xh_free_start for the new bucket. */
xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE); for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
xe = &xh->xh_entries[i]; if (le16_to_cpu(xe->xe_name_offset) <
le16_to_cpu(xh->xh_free_start))
xh->xh_free_start = xe->xe_name_offset;
}
set_num_buckets: /* set xh->xh_num_buckets for the new xh. */ if (new_bucket_head)
xh->xh_num_buckets = cpu_to_le16(1); else
xh->xh_num_buckets = 0;
/* store the first_hash of the new bucket. */ if (first_hash)
*first_hash = le32_to_cpu(xh->xh_entries[0].xe_name_hash);
/* * Now only update the 1st block of the old bucket. If we * just added a new empty bucket, there is no need to modify * it.
*/ if (start == count) goto out;
/* * Copy xattr from one bucket to another bucket. * * The caller must make sure that the journal transaction * has enough space for journaling.
*/ staticint ocfs2_cp_xattr_bucket(struct inode *inode,
handle_t *handle,
u64 s_blkno,
u64 t_blkno, int t_is_new)
{ int ret; struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL;
s_bucket = ocfs2_xattr_bucket_new(inode);
t_bucket = ocfs2_xattr_bucket_new(inode); if (!s_bucket || !t_bucket) {
ret = -ENOMEM;
mlog_errno(ret); goto out;
}
ret = ocfs2_read_xattr_bucket(s_bucket, s_blkno); if (ret) goto out;
/* * Even if !t_is_new, we're overwriting t_bucket. Thus, * there's no need to read it.
*/
ret = ocfs2_init_xattr_bucket(t_bucket, t_blkno, t_is_new); if (ret) goto out;
/* * Hey, if we're overwriting t_bucket, what difference does * ACCESS_CREATE vs ACCESS_WRITE make? Well, if we allocated a new * cluster to fill, we came here from * ocfs2_mv_xattr_buckets(), and it is really new - * ACCESS_CREATE is required. But we also might have moved data * out of t_bucket before extending back into it. * ocfs2_add_new_xattr_bucket() can do this - its call to * ocfs2_add_new_xattr_cluster() may have created a new extent * and copied out the end of the old extent. Then it re-extends * the old extent back to create space for new xattrs. That's * how we get here, and the bucket isn't really new.
*/
ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket,
t_is_new ?
OCFS2_JOURNAL_ACCESS_CREATE :
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) goto out;
/* * src_blk points to the start of an existing extent. last_blk points to * last cluster in that extent. to_blk points to a newly allocated * extent. We copy the buckets from the cluster at last_blk to the new * extent. If start_bucket is non-zero, we skip that many buckets before * we start copying. The new extent's xh_num_buckets gets set to the * number of buckets we copied. The old extent's xh_num_buckets shrinks * by the same amount.
*/ staticint ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle,
u64 src_blk, u64 last_blk, u64 to_blk, unsignedint start_bucket,
u32 *first_hash)
{ int i, ret, credits; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); int num_buckets = ocfs2_xattr_buckets_per_cluster(osb); struct ocfs2_xattr_bucket *old_first, *new_first;
/* The first bucket of the original extent */
old_first = ocfs2_xattr_bucket_new(inode); /* The first bucket of the new extent */
new_first = ocfs2_xattr_bucket_new(inode); if (!old_first || !new_first) {
ret = -ENOMEM;
mlog_errno(ret); goto out;
}
ret = ocfs2_read_xattr_bucket(old_first, src_blk); if (ret) {
mlog_errno(ret); goto out;
}
/* * We need to update the first bucket of the old extent and all * the buckets going to the new extent.
*/
credits = ((num_buckets + 1) * blks_per_bucket);
ret = ocfs2_extend_trans(handle, credits); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_xattr_bucket_journal_access(handle, old_first,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out;
}
for (i = 0; i < num_buckets; i++) {
ret = ocfs2_cp_xattr_bucket(inode, handle,
last_blk + (i * blks_per_bucket),
to_blk + (i * blks_per_bucket),
1); if (ret) {
mlog_errno(ret); goto out;
}
}
/* * Get the new bucket ready before we dirty anything * (This actually shouldn't fail, because we already dirtied * it once in ocfs2_cp_xattr_bucket()).
*/
ret = ocfs2_read_xattr_bucket(new_first, to_blk); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_xattr_bucket_journal_access(handle, new_first,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out;
}
/* Now update the headers */
le16_add_cpu(&bucket_xh(old_first)->xh_num_buckets, -num_buckets);
ocfs2_xattr_bucket_journal_dirty(handle, old_first);
/* * Move some xattrs in this cluster to the new cluster. * This function should only be called when bucket size == cluster size. * Otherwise ocfs2_mv_xattr_bucket_cross_cluster should be used instead.
*/ staticint ocfs2_divide_xattr_cluster(struct inode *inode,
handle_t *handle,
u64 prev_blk,
u64 new_blk,
u32 *first_hash)
{
u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); int ret, credits = 2 * blk_per_bucket;
ret = ocfs2_extend_trans(handle, credits); if (ret) {
mlog_errno(ret); return ret;
}
/* Move half of the xattr in start_blk to the next bucket. */ return ocfs2_divide_xattr_bucket(inode, handle, prev_blk,
new_blk, first_hash, 1);
}
/* * Move some xattrs from the old cluster to the new one since they are not * contiguous in ocfs2 xattr tree. * * new_blk starts a new separate cluster, and we will move some xattrs from * prev_blk to it. v_start will be set as the first name hash value in this * new cluster so that it can be used as e_cpos during tree insertion and * don't collide with our original b-tree operations. first_bh and header_bh * will also be updated since they will be used in ocfs2_extend_xattr_bucket * to extend the insert bucket. * * The problem is how much xattr should we move to the new one and when should * we update first_bh and header_bh? * 1. If cluster size > bucket size, that means the previous cluster has more * than 1 bucket, so just move half nums of bucket into the new cluster and * update the first_bh and header_bh if the insert bucket has been moved * to the new cluster. * 2. If cluster_size == bucket_size: * a) If the previous extent rec has more than one cluster and the insert * place isn't in the last cluster, copy the entire last cluster to the * new one. This time, we don't need to update the first_bh and header_bh * since they will not be moved into the new cluster. * b) Otherwise, move the bottom half of the xattrs in the last cluster into * the new one. And we set the extend flag to zero if the insert place is * moved into the new allocated cluster since no extend is needed.
*/ staticint ocfs2_adjust_xattr_cross_cluster(struct inode *inode,
handle_t *handle, struct ocfs2_xattr_bucket *first, struct ocfs2_xattr_bucket *target,
u64 new_blk,
u32 prev_clusters,
u32 *v_start, int *extend)
{ int ret;
if (ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)) > 1) {
ret = ocfs2_mv_xattr_bucket_cross_cluster(inode,
handle,
first, target,
new_blk,
prev_clusters,
v_start); if (ret)
mlog_errno(ret);
} else { /* The start of the last cluster in the first extent */
u64 last_blk = bucket_blkno(first) +
((prev_clusters - 1) *
ocfs2_clusters_to_blocks(inode->i_sb, 1));
if (prev_clusters > 1 && bucket_blkno(target) != last_blk) {
ret = ocfs2_mv_xattr_buckets(inode, handle,
bucket_blkno(first),
last_blk, new_blk, 0,
v_start); if (ret)
mlog_errno(ret);
} else {
ret = ocfs2_divide_xattr_cluster(inode, handle,
last_blk, new_blk,
v_start); if (ret)
mlog_errno(ret);
/* * Add a new cluster for xattr storage. * * If the new cluster is contiguous with the previous one, it will be * appended to the same extent record, and num_clusters will be updated. * If not, we will insert a new extent for it and move some xattrs in * the last cluster into the new allocated one. * We also need to limit the maximum size of a btree leaf, otherwise we'll * lose the benefits of hashing because we'll have to search large leaves. * So now the maximum size is OCFS2_MAX_XATTR_TREE_LEAF_SIZE(or clustersize, * if it's bigger). * * first_bh is the first block of the previous extent rec and header_bh * indicates the bucket we will insert the new xattrs. They will be updated * when the header_bh is moved into the new cluster.
*/ staticint ocfs2_add_new_xattr_cluster(struct inode *inode, struct buffer_head *root_bh, struct ocfs2_xattr_bucket *first, struct ocfs2_xattr_bucket *target,
u32 *num_clusters,
u32 prev_cpos, int *extend, struct ocfs2_xattr_set_ctxt *ctxt)
{ int ret;
u16 bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
u32 prev_clusters = *num_clusters;
u32 clusters_to_add = 1, bit_off, num_bits, v_start = 0;
u64 block;
handle_t *handle = ctxt->handle; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct ocfs2_extent_tree et;
if (bucket_blkno(first) + (prev_clusters * bpc) == block &&
(prev_clusters + num_bits) << osb->s_clustersize_bits <=
OCFS2_MAX_XATTR_TREE_LEAF_SIZE) { /* * If this cluster is contiguous with the old one and * adding this new cluster, we don't surpass the limit of * OCFS2_MAX_XATTR_TREE_LEAF_SIZE, cool. We will let it be * initialized and used like other buckets in the previous * cluster. * So add it as a contiguous one. The caller will handle * its init process.
*/
v_start = prev_cpos + prev_clusters;
*num_clusters = prev_clusters + num_bits;
} else {
ret = ocfs2_adjust_xattr_cross_cluster(inode,
handle,
first,
target,
block,
prev_clusters,
&v_start,
extend); if (ret) {
mlog_errno(ret); goto leave;
}
}
/* * We are given an extent. 'first' is the bucket at the very front of * the extent. The extent has space for an additional bucket past * bucket_xh(first)->xh_num_buckets. 'target_blkno' is the block number * of the target bucket. We wish to shift every bucket past the target * down one, filling in that additional space. When we get back to the * target, we split the target between itself and the now-empty bucket * at target+1 (aka, target_blkno + blks_per_bucket).
*/ staticint ocfs2_extend_xattr_bucket(struct inode *inode,
handle_t *handle, struct ocfs2_xattr_bucket *first,
u64 target_blk,
u32 num_clusters)
{ int ret, credits; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
u64 end_blk;
u16 new_bucket = le16_to_cpu(bucket_xh(first)->xh_num_buckets);
/* The extent must have room for an additional bucket */
BUG_ON(new_bucket >=
(num_clusters * ocfs2_xattr_buckets_per_cluster(osb)));
/* end_blk points to the last existing bucket */
end_blk = bucket_blkno(first) + ((new_bucket - 1) * blk_per_bucket);
/* * end_blk is the start of the last existing bucket. * Thus, (end_blk - target_blk) covers the target bucket and * every bucket after it up to, but not including, the last * existing bucket. Then we add the last existing bucket, the * new bucket, and the first bucket (3 * blk_per_bucket).
*/
credits = (end_blk - target_blk) + (3 * blk_per_bucket);
ret = ocfs2_extend_trans(handle, credits); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_xattr_bucket_journal_access(handle, first,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out;
}
while (end_blk != target_blk) {
ret = ocfs2_cp_xattr_bucket(inode, handle, end_blk,
end_blk + blk_per_bucket, 0); if (ret) goto out;
end_blk -= blk_per_bucket;
}
/* Move half of the xattr in target_blkno to the next bucket. */
ret = ocfs2_divide_xattr_bucket(inode, handle, target_blk,
target_blk + blk_per_bucket, NULL, 0);
/* * Add new xattr bucket in an extent record and adjust the buckets * accordingly. xb_bh is the ocfs2_xattr_block, and target is the * bucket we want to insert into. * * In the easy case, we will move all the buckets after target down by * one. Half of target's xattrs will be moved to the next bucket. * * If current cluster is full, we'll allocate a new one. This may not * be contiguous. The underlying calls will make sure that there is * space for the insert, shifting buckets around if necessary. * 'target' may be moved by those calls.
*/ staticint ocfs2_add_new_xattr_bucket(struct inode *inode, struct buffer_head *xb_bh, struct ocfs2_xattr_bucket *target, struct ocfs2_xattr_set_ctxt *ctxt)
{ struct ocfs2_xattr_block *xb =
(struct ocfs2_xattr_block *)xb_bh->b_data; struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root; struct ocfs2_extent_list *el = &xb_root->xt_list;
u32 name_hash =
le32_to_cpu(bucket_xh(target)->xh_entries[0].xe_name_hash); struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); int ret, num_buckets, extend = 1;
u64 p_blkno;
u32 e_cpos, num_clusters; /* The bucket at the front of the extent */ struct ocfs2_xattr_bucket *first;
/* The first bucket of the original extent */
first = ocfs2_xattr_bucket_new(inode); if (!first) {
ret = -ENOMEM;
mlog_errno(ret); goto out;
}
ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &e_cpos,
&num_clusters, el); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_read_xattr_bucket(first, p_blkno); if (ret) {
mlog_errno(ret); goto out;
}
num_buckets = ocfs2_xattr_buckets_per_cluster(osb) * num_clusters; if (num_buckets == le16_to_cpu(bucket_xh(first)->xh_num_buckets)) { /* * This can move first+target if the target bucket moves * to the new extent.
*/
ret = ocfs2_add_new_xattr_cluster(inode,
xb_bh,
first,
target,
&num_clusters,
e_cpos,
&extend,
ctxt); if (ret) {
mlog_errno(ret); goto out;
}
}
if (extend) {
ret = ocfs2_extend_xattr_bucket(inode,
ctxt->handle,
first,
bucket_blkno(target),
num_clusters); if (ret)
mlog_errno(ret);
}
out:
ocfs2_xattr_bucket_free(first);
return ret;
}
/* * Truncate the specified xe_off entry in xattr bucket. * bucket is indicated by header_bh and len is the new length. * Both the ocfs2_xattr_value_root and the entry will be updated here. * * Copy the new updated xe and xe_value_root to new_xe and new_xv if needed.
*/ staticint ocfs2_xattr_bucket_value_truncate(struct inode *inode, struct ocfs2_xattr_bucket *bucket, int xe_off, int len, struct ocfs2_xattr_set_ctxt *ctxt)
{ int ret, offset;
u64 value_blk; struct ocfs2_xattr_entry *xe; struct ocfs2_xattr_header *xh = bucket_xh(bucket);
size_t blocksize = inode->i_sb->s_blocksize; struct ocfs2_xattr_value_buf vb = {
.vb_access = ocfs2_journal_access,
};
/* * From here on out we have to dirty the bucket. The generic * value calls only modify one of the bucket's bhs, but we need * to send the bucket at once. So if they error, they *could* have * modified something. We have to assume they did, and dirty * the whole bucket. This leaves us in a consistent state.
*/
trace_ocfs2_xattr_bucket_value_truncate(
(unsignedlonglong)bucket_blkno(bucket), xe_off, len);
ret = ocfs2_xattr_value_truncate(inode, &vb, len, ctxt); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_xattr_bucket_journal_access(ctxt->handle, bucket,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out;
}
/* * check whether the xattr bucket is filled up with the same hash value. * If we want to insert the xattr with the same hash, return -ENOSPC. * If we want to insert a xattr with different hash value, go ahead * and ocfs2_divide_xattr_bucket will handle this.
*/ staticint ocfs2_check_xattr_bucket_collision(struct inode *inode, struct ocfs2_xattr_bucket *bucket, constchar *name)
{ struct ocfs2_xattr_header *xh = bucket_xh(bucket);
u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name));
if (name_hash != le32_to_cpu(xh->xh_entries[0].xe_name_hash)) return 0;
if (xh->xh_entries[le16_to_cpu(xh->xh_count) - 1].xe_name_hash ==
xh->xh_entries[0].xe_name_hash) {
mlog(ML_ERROR, "Too much hash collision in xattr bucket %llu, " "hash = %u\n",
(unsignedlonglong)bucket_blkno(bucket),
le32_to_cpu(xh->xh_entries[0].xe_name_hash)); return -ENOSPC;
}
return 0;
}
/* * Try to set the entry in the current bucket. If we fail, the caller * will handle getting us another bucket.
*/ staticint ocfs2_xattr_set_entry_bucket(struct inode *inode, struct ocfs2_xattr_info *xi, struct ocfs2_xattr_search *xs, struct ocfs2_xattr_set_ctxt *ctxt)
{ int ret; struct ocfs2_xa_loc loc;
trace_ocfs2_xattr_set_entry_bucket(xi->xi_name);
ocfs2_init_xattr_bucket_xa_loc(&loc, xs->bucket,
xs->not_found ? NULL : xs->here);
ret = ocfs2_xa_set(&loc, xi, ctxt); if (!ret) {
xs->here = loc.xl_entry; goto out;
} if (ret != -ENOSPC) {
mlog_errno(ret); goto out;
}
/* Ok, we need space. Let's try defragmenting the bucket. */
ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle,
xs->bucket); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_xa_set(&loc, xi, ctxt); if (!ret) {
xs->here = loc.xl_entry; goto out;
} if (ret != -ENOSPC)
mlog_errno(ret);
ret = ocfs2_xattr_set_entry_bucket(inode, xi, xs, ctxt); if (!ret) goto out; if (ret != -ENOSPC) {
mlog_errno(ret); goto out;
}
/* Ack, need more space. Let's try to get another bucket! */
/* * We do not allow for overlapping ranges between buckets. And * the maximum number of collisions we will allow for then is * one bucket's worth, so check it here whether we need to * add a new bucket for the insert.
*/
ret = ocfs2_check_xattr_bucket_collision(inode,
xs->bucket,
xi->xi_name); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_add_new_xattr_bucket(inode,
xs->xattr_bh,
xs->bucket,
ctxt); if (ret) {
mlog_errno(ret); goto out;
}
/* * ocfs2_add_new_xattr_bucket() will have updated * xs->bucket if it moved, but it will not have updated * any of the other search fields. Thus, we drop it and * re-search. Everything should be cached, so it'll be * quick.
*/
ocfs2_xattr_bucket_relse(xs->bucket);
ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh,
xi->xi_name_index,
xi->xi_name, xs); if (ret && ret != -ENODATA) goto out;
xs->not_found = ret;
/* Ok, we have a new bucket, let's try again */
ret = ocfs2_xattr_set_entry_bucket(inode, xi, xs, ctxt); if (ret && (ret != -ENOSPC))
mlog_errno(ret);
for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
xe = &xh->xh_entries[i]; if (ocfs2_xattr_is_local(xe)) continue;
ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket,
i, &xv, NULL); if (ret) {
mlog_errno(ret); break;
}
ret = ocfs2_lock_xattr_remove_allocators(inode, xv,
args->ref_ci,
args->ref_root_bh,
&ctxt.meta_ac,
&ref_credits);
ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); if (IS_ERR(ctxt.handle)) {
ret = PTR_ERR(ctxt.handle);
mlog_errno(ret); break;
}
ret = ocfs2_xattr_bucket_value_truncate(inode, bucket,
i, 0, &ctxt);
ocfs2_commit_trans(osb, ctxt.handle); if (ctxt.meta_ac) {
ocfs2_free_alloc_context(ctxt.meta_ac);
ctxt.meta_ac = NULL;
} if (ret) {
mlog_errno(ret); break;
}
}
if (ctxt.meta_ac)
ocfs2_free_alloc_context(ctxt.meta_ac);
ocfs2_schedule_truncate_log_flush(osb, 1);
ocfs2_run_deallocs(osb, &ctxt.dealloc); return ret;
}
/* * Whenever we modify a xattr value root in the bucket(e.g, CoW * or change the extent record flag), we need to recalculate * the metaecc for the whole bucket. So it is done here. * * Note: * We have to give the extra credits for the caller.
*/ staticint ocfs2_xattr_bucket_post_refcount(struct inode *inode,
handle_t *handle, void *para)
{ int ret; struct ocfs2_xattr_bucket *bucket =
(struct ocfs2_xattr_bucket *)para;
ret = ocfs2_xattr_bucket_journal_access(handle, bucket,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); return ret;
}
ocfs2_xattr_bucket_journal_dirty(handle, bucket);
return 0;
}
/* * Special action we need if the xattr value is refcounted. * * 1. If the xattr is refcounted, lock the tree. * 2. CoW the xattr if we are setting the new value and the value * will be stored outside. * 3. In other case, decrease_refcount will work for us, so just * lock the refcount tree, calculate the meta and credits is OK. * * We have to do CoW before ocfs2_init_xattr_set_ctxt since * currently CoW is a completed transaction, while this function * will also lock the allocators and let us deadlock. So we will * CoW the whole xattr value.
*/ staticint ocfs2_prepare_refcount_xattr(struct inode *inode, struct ocfs2_dinode *di, struct ocfs2_xattr_info *xi, struct ocfs2_xattr_search *xis, struct ocfs2_xattr_search *xbs, struct ocfs2_refcount_tree **ref_tree, int *meta_add, int *credits)
{ int ret = 0; struct ocfs2_xattr_block *xb; struct ocfs2_xattr_entry *xe; char *base;
u32 p_cluster, num_clusters; unsignedint ext_flags; int name_offset, name_len; struct ocfs2_xattr_value_buf vb; struct ocfs2_xattr_bucket *bucket = NULL; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct ocfs2_post_refcount refcount; struct ocfs2_post_refcount *p = NULL; struct buffer_head *ref_root_bh = NULL;
if (!xis->not_found) {
xe = xis->here;
name_offset = le16_to_cpu(xe->xe_name_offset);
name_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
base = xis->base;
vb.vb_bh = xis->inode_bh;
vb.vb_access = ocfs2_journal_access_di;
} else { int i, block_off = 0;
xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data;
xe = xbs->here;
name_offset = le16_to_cpu(xe->xe_name_offset);
name_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
i = xbs->here - xbs->header->xh_entries;
if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) {
ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb,
bucket_xh(xbs->bucket),
i, &block_off,
&name_offset); if (ret) {
mlog_errno(ret); goto out;
}
base = bucket_block(xbs->bucket, block_off);
vb.vb_bh = xbs->bucket->bu_bhs[block_off];
vb.vb_access = ocfs2_journal_access;
if (ocfs2_meta_ecc(osb)) { /*create parameters for ocfs2_post_refcount. */
bucket = xbs->bucket;
refcount.credits = bucket->bu_blocks;
refcount.para = bucket;
refcount.func =
ocfs2_xattr_bucket_post_refcount;
p = &refcount;
}
} else {
base = xbs->base;
vb.vb_bh = xbs->xattr_bh;
vb.vb_access = ocfs2_journal_access_xb;
}
}
ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster,
&num_clusters, &vb.vb_xv->xr_list,
&ext_flags); if (ret) {
mlog_errno(ret); goto out;
}
/* * We just need to check the 1st extent record, since we always * CoW the whole xattr. So there shouldn't be a xattr with * some REFCOUNT extent recs after the 1st one.
*/ if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) goto out;
ret = ocfs2_lock_refcount_tree(osb, le64_to_cpu(di->i_refcount_loc),
1, ref_tree, &ref_root_bh); if (ret) {
mlog_errno(ret); goto out;
}
/* * If we are deleting the xattr or the new size will be stored inside, * cool, leave it there, the xattr truncate process will remove them * for us(it still needs the refcount tree lock and the meta, credits). * And the worse case is that every cluster truncate will split the * refcount tree, and make the original extent become 3. So we will need * 2 * cluster more extent recs at most.
*/ if (!xi->xi_value || xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE) {
ret = ocfs2_refcounted_xattr_delete_need(inode,
&(*ref_tree)->rf_ci,
ref_root_bh, vb.vb_xv,
meta_add, credits); if (ret)
mlog_errno(ret); goto out;
}
ret = ocfs2_refcount_cow_xattr(inode, di, &vb,
*ref_tree, ref_root_bh, 0,
le32_to_cpu(vb.vb_xv->xr_clusters), p); if (ret)
mlog_errno(ret);
out:
brelse(ref_root_bh); return ret;
}
/* * Add the REFCOUNTED flags for all the extent rec in ocfs2_xattr_value_root. * The physical clusters will be added to refcount tree.
*/ staticint ocfs2_xattr_value_attach_refcount(struct inode *inode, struct ocfs2_xattr_value_root *xv, struct ocfs2_extent_tree *value_et, struct ocfs2_caching_info *ref_ci, struct buffer_head *ref_root_bh, struct ocfs2_cached_dealloc_ctxt *dealloc, struct ocfs2_post_refcount *refcount)
{ int ret = 0;
u32 clusters = le32_to_cpu(xv->xr_clusters);
u32 cpos, p_cluster, num_clusters; struct ocfs2_extent_list *el = &xv->xr_list; unsignedint ext_flags;
cpos = 0; while (cpos < clusters) {
ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster,
&num_clusters, el, &ext_flags); if (ret) {
mlog_errno(ret); break;
}
cpos += num_clusters; if ((ext_flags & OCFS2_EXT_REFCOUNTED)) continue;
BUG_ON(!p_cluster);
ret = ocfs2_add_refcount_flag(inode, value_et,
ref_ci, ref_root_bh,
cpos - num_clusters,
p_cluster, num_clusters,
dealloc, refcount); if (ret) {
mlog_errno(ret); break;
}
}
return ret;
}
/* * Given a normal ocfs2_xattr_header, refcount all the entries which * have value stored outside. * Used for xattrs stored in inode and ocfs2_xattr_block.
*/ staticint ocfs2_xattr_attach_refcount_normal(struct inode *inode, struct ocfs2_xattr_value_buf *vb, struct ocfs2_xattr_header *header, struct ocfs2_caching_info *ref_ci, struct buffer_head *ref_root_bh, struct ocfs2_cached_dealloc_ctxt *dealloc)
{
struct ocfs2_xattr_entry *xe; struct ocfs2_xattr_value_root *xv; struct ocfs2_extent_tree et; int i, ret = 0;
for (i = 0; i < le16_to_cpu(header->xh_count); i++) {
xe = &header->xh_entries[i];
if (ocfs2_xattr_is_local(xe)) continue;
xv = (struct ocfs2_xattr_value_root *)((void *)header +
le16_to_cpu(xe->xe_name_offset) +
OCFS2_XATTR_SIZE(xe->xe_name_len));
if (bh)
*bh = bucket->bu_bhs[block_off];
out: return ret;
}
/* * For a given xattr bucket, refcount all the entries which * have value stored outside.
*/ staticint ocfs2_xattr_bucket_value_refcount(struct inode *inode, struct ocfs2_xattr_bucket *bucket, void *para)
{ int i, ret = 0; struct ocfs2_extent_tree et; struct ocfs2_xattr_tree_value_refcount_para *ref =
(struct ocfs2_xattr_tree_value_refcount_para *)para; struct ocfs2_xattr_header *xh =
(struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; struct ocfs2_xattr_entry *xe; struct ocfs2_xattr_value_buf vb = {
.vb_access = ocfs2_journal_access,
}; struct ocfs2_post_refcount refcount = {
.credits = bucket->bu_blocks,
.para = bucket,
.func = ocfs2_xattr_bucket_post_refcount,
}; struct ocfs2_post_refcount *p = NULL;
/* We only need post_refcount if we support metaecc. */ if (ocfs2_meta_ecc(OCFS2_SB(inode->i_sb)))
p = &refcount;
trace_ocfs2_xattr_bucket_value_refcount(
(unsignedlonglong)bucket_blkno(bucket),
le16_to_cpu(xh->xh_count)); for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
xe = &xh->xh_entries[i];
if (ocfs2_xattr_is_local(xe)) continue;
ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, i,
&vb.vb_xv, &vb.vb_bh); if (ret) {
mlog_errno(ret); break;
}
if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) {
ret = ocfs2_xattr_inline_attach_refcount(inode, fe_bh,
ref_ci, ref_root_bh,
dealloc); if (ret) {
mlog_errno(ret); goto out;
}
}
if (!di->i_xattr_loc) goto out;
ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc),
&blk_bh); if (ret < 0) {
mlog_errno(ret); goto out;
}
ret = ocfs2_xattr_block_attach_refcount(inode, blk_bh, ref_ci,
ref_root_bh, dealloc); if (ret)
mlog_errno(ret);
brelse(blk_bh);
out:
return ret;
}
typedefint (should_xattr_reflinked)(struct ocfs2_xattr_entry *xe); /* * Store the information we need in xattr reflink. * old_bh and new_bh are inode bh for the old and new inode.
*/ struct ocfs2_xattr_reflink { struct inode *old_inode; struct inode *new_inode; struct buffer_head *old_bh; struct buffer_head *new_bh; struct ocfs2_caching_info *ref_ci; struct buffer_head *ref_root_bh; struct ocfs2_cached_dealloc_ctxt *dealloc;
should_xattr_reflinked *xattr_reflinked;
};
/* * Given a xattr header and xe offset, * return the proper xv and the corresponding bh. * xattr in inode, block and xattr tree have different implementations.
*/ typedefint (get_xattr_value_root)(struct super_block *sb, struct buffer_head *bh, struct ocfs2_xattr_header *xh, int offset, struct ocfs2_xattr_value_root **xv, struct buffer_head **ret_bh, void *para);
/* * Calculate all the xattr value root metadata stored in this xattr header and * credits we need if we create them from the scratch. * We use get_xattr_value_root so that all types of xattr container can use it.
*/ staticint ocfs2_value_metas_in_xattr_header(struct super_block *sb, struct buffer_head *bh, struct ocfs2_xattr_header *xh, int *metas, int *credits, int *num_recs,
get_xattr_value_root *func, void *para)
{ int i, ret = 0; struct ocfs2_xattr_value_root *xv; struct ocfs2_xattr_entry *xe;
for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
xe = &xh->xh_entries[i]; if (ocfs2_xattr_is_local(xe)) continue;
ret = func(sb, bh, xh, i, &xv, NULL, para); if (ret) {
mlog_errno(ret); break;
}
/* * If the value is a tree with depth > 1, We don't go deep * to the extent block, so just calculate a maximum record num.
*/ if (!xv->xr_list.l_tree_depth)
*num_recs += le16_to_cpu(xv->xr_list.l_next_free_rec); else
*num_recs += ocfs2_clusters_for_bytes(sb,
XATTR_SIZE_MAX);
}
return ret;
}
/* Used by xattr inode and block to return the right xv and buffer_head. */ staticint ocfs2_get_xattr_value_root(struct super_block *sb, struct buffer_head *bh, struct ocfs2_xattr_header *xh, int offset, struct ocfs2_xattr_value_root **xv, struct buffer_head **ret_bh, void *para)
{ struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset];
/* * Lock the meta_ac and calculate how much credits we need for reflink xattrs. * It is only used for inline xattr and xattr block.
*/ staticint ocfs2_reflink_lock_xattr_allocators(struct ocfs2_super *osb, struct ocfs2_xattr_header *xh, struct buffer_head *ref_root_bh, int *credits, struct ocfs2_alloc_context **meta_ac)
{ int ret, meta_add = 0, num_recs = 0; struct ocfs2_refcount_block *rb =
(struct ocfs2_refcount_block *)ref_root_bh->b_data;
*credits = 0;
ret = ocfs2_value_metas_in_xattr_header(osb->sb, NULL, xh,
&meta_add, credits, &num_recs,
ocfs2_get_xattr_value_root,
NULL); if (ret) {
mlog_errno(ret); goto out;
}
/* * We need to add/modify num_recs in refcount tree, so just calculate * an approximate number we need for refcount tree change. * Sometimes we need to split the tree, and after split, half recs * will be moved to the new block, and a new block can only provide * half number of recs. So we multiple new blocks by 2.
*/
num_recs = num_recs / ocfs2_refcount_recs_per_rb(osb->sb) * 2;
meta_add += num_recs;
*credits += num_recs + num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL)
*credits += le16_to_cpu(rb->rf_list.l_tree_depth) *
le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; else
*credits += 1;
ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, meta_ac); if (ret)
mlog_errno(ret);
out: return ret;
}
/* * Given a xattr header, reflink all the xattrs in this container. * It can be used for inode, block and bucket. * * NOTE: * Before we call this function, the caller has memcpy the xattr in * old_xh to the new_xh. * * If args.xattr_reflinked is set, call it to decide whether the xe should * be reflinked or not. If not, remove it from the new xattr header.
*/ staticint ocfs2_reflink_xattr_header(handle_t *handle, struct ocfs2_xattr_reflink *args, struct buffer_head *old_bh, struct ocfs2_xattr_header *xh, struct buffer_head *new_bh, struct ocfs2_xattr_header *new_xh, struct ocfs2_xattr_value_buf *vb, struct ocfs2_alloc_context *meta_ac,
get_xattr_value_root *func, void *para)
{ int ret = 0, i, j; struct super_block *sb = args->old_inode->i_sb; struct buffer_head *value_bh; struct ocfs2_xattr_entry *xe, *last; struct ocfs2_xattr_value_root *xv, *new_xv; struct ocfs2_extent_tree data_et;
u32 clusters, cpos, p_cluster, num_clusters; unsignedint ext_flags = 0;
/* * We don't want j to increase in the next round since * it is already moved ahead.
*/
j--; continue;
}
if (ocfs2_xattr_is_local(xe)) continue;
ret = func(sb, old_bh, xh, i, &xv, NULL, para); if (ret) {
mlog_errno(ret); break;
}
ret = func(sb, new_bh, new_xh, j, &new_xv, &value_bh, para); if (ret) {
mlog_errno(ret); break;
}
/* * For the xattr which has l_tree_depth = 0, all the extent * recs have already be copied to the new xh with the * propriate OCFS2_EXT_REFCOUNTED flag we just need to * increase the refount count int the refcount tree. * * For the xattr which has l_tree_depth > 0, we need * to initialize it to the empty default value root, * and then insert the extents one by one.
*/ if (xv->xr_list.l_tree_depth) {
memcpy(new_xv, &def_xv, OCFS2_XATTR_ROOT_SIZE);
vb->vb_xv = new_xv;
vb->vb_bh = value_bh;
ocfs2_init_xattr_value_extent_tree(&data_et,
INODE_CACHE(args->new_inode), vb);
}
clusters = le32_to_cpu(xv->xr_clusters);
cpos = 0; while (cpos < clusters) {
ret = ocfs2_xattr_get_clusters(args->old_inode,
cpos,
&p_cluster,
&num_clusters,
&xv->xr_list,
&ext_flags); if (ret) {
mlog_errno(ret); goto out;
}
BUG_ON(!p_cluster);
if (xv->xr_list.l_tree_depth) {
ret = ocfs2_insert_extent(handle,
&data_et, cpos,
ocfs2_clusters_to_blocks(
args->old_inode->i_sb,
p_cluster),
num_clusters, ext_flags,
meta_ac); if (ret) {
mlog_errno(ret); goto out;
}
}
ret = ocfs2_increase_refcount(handle, args->ref_ci,
args->ref_root_bh,
p_cluster, num_clusters,
meta_ac, args->dealloc); if (ret) {
mlog_errno(ret); goto out;
}
ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh,
&credits, &meta_ac); if (ret) {
mlog_errno(ret); return ret;
}
/* One more credits in case we need to add xattr flags in new inode. */
handle = ocfs2_start_trans(osb, credits + 1); if (IS_ERR(handle)) {
ret = PTR_ERR(handle);
mlog_errno(ret); goto out;
}
if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) {
ret = ocfs2_journal_access_di(handle,
INODE_CACHE(args->new_inode),
args->new_bh,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out_commit;
}
}
ret = ocfs2_journal_access_xb(handle, INODE_CACHE(args->new_inode),
new_blk_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); goto out_commit;
}
/* * NOTE: * We have to handle the case that both old bucket and new bucket * will call this function to get the right ret_bh. * So The caller must give us the right bh.
*/ staticint ocfs2_get_reflink_xattr_value_root(struct super_block *sb, struct buffer_head *bh, struct ocfs2_xattr_header *xh, int offset, struct ocfs2_xattr_value_root **xv, struct buffer_head **ret_bh, void *para)
{ struct ocfs2_reflink_xattr_tree_args *args =
(struct ocfs2_reflink_xattr_tree_args *)para; struct ocfs2_xattr_bucket *bucket;
/* Add the credits for this bucket first. */
metas->credits += bucket->bu_blocks; return ocfs2_value_metas_in_xattr_header(inode->i_sb, bucket->bu_bhs[0],
xh, &metas->num_metas,
&metas->credits, &metas->num_recs,
ocfs2_value_tree_metas_in_bucket,
bucket);
}
/* * Given a xattr extent rec starting from blkno and having len clusters, * iterate all the buckets calculate how much metadata we need for reflinking * all the ocfs2_xattr_value_root and lock the allocators accordingly.
*/ staticint ocfs2_lock_reflink_xattr_rec_allocators( struct ocfs2_reflink_xattr_tree_args *args, struct ocfs2_extent_tree *xt_et,
u64 blkno, u32 len, int *credits, struct ocfs2_alloc_context **meta_ac, struct ocfs2_alloc_context **data_ac)
{ int ret, num_free_extents; struct ocfs2_value_tree_metas metas; struct ocfs2_super *osb = OCFS2_SB(args->reflink->old_inode->i_sb); struct ocfs2_refcount_block *rb;
memset(&metas, 0, sizeof(metas));
ret = ocfs2_iterate_xattr_buckets(args->reflink->old_inode, blkno, len,
ocfs2_calc_value_tree_metas, &metas); if (ret) {
mlog_errno(ret); goto out;
}
*credits = metas.credits;
/* * Calculate we need for refcount tree change. * * We need to add/modify num_recs in refcount tree, so just calculate * an approximate number we need for refcount tree change. * Sometimes we need to split the tree, and after split, half recs * will be moved to the new block, and a new block can only provide * half number of recs. So we multiple new blocks by 2. * In the end, we have to add credits for modifying the already * existed refcount block.
*/
rb = (struct ocfs2_refcount_block *)args->reflink->ref_root_bh->b_data;
metas.num_recs =
(metas.num_recs + ocfs2_refcount_recs_per_rb(osb->sb) - 1) /
ocfs2_refcount_recs_per_rb(osb->sb) * 2;
metas.num_metas += metas.num_recs;
*credits += metas.num_recs +
metas.num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL)
*credits += le16_to_cpu(rb->rf_list.l_tree_depth) *
le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; else
*credits += 1;
/* count in the xattr tree change. */
num_free_extents = ocfs2_num_free_extents(xt_et); if (num_free_extents < 0) {
ret = num_free_extents;
mlog_errno(ret); goto out;
}
if (num_free_extents < len)
metas.num_metas += ocfs2_extend_meta_needed(xt_et->et_root_el);
if (metas.num_metas) {
ret = ocfs2_reserve_new_metadata_blocks(osb, metas.num_metas,
meta_ac); if (ret) {
mlog_errno(ret); goto out;
}
}
if (len) {
ret = ocfs2_reserve_clusters(osb, len, data_ac); if (ret)
mlog_errno(ret);
}
out: if (ret) { if (*meta_ac) {
ocfs2_free_alloc_context(*meta_ac);
*meta_ac = NULL;
}
}
return ret;
}
staticint ocfs2_reflink_xattr_bucket(handle_t *handle,
u64 blkno, u64 new_blkno, u32 clusters,
u32 *cpos, int num_buckets, struct ocfs2_alloc_context *meta_ac, struct ocfs2_alloc_context *data_ac, struct ocfs2_reflink_xattr_tree_args *args)
{ int i, j, ret = 0; struct super_block *sb = args->reflink->old_inode->i_sb; int bpb = args->old_bucket->bu_blocks; struct ocfs2_xattr_value_buf vb = {
.vb_access = ocfs2_journal_access,
};
for (i = 0; i < num_buckets; i++, blkno += bpb, new_blkno += bpb) {
ret = ocfs2_read_xattr_bucket(args->old_bucket, blkno); if (ret) {
mlog_errno(ret); break;
}
ret = ocfs2_init_xattr_bucket(args->new_bucket, new_blkno, 1); if (ret) {
mlog_errno(ret); break;
}
ret = ocfs2_xattr_bucket_journal_access(handle,
args->new_bucket,
OCFS2_JOURNAL_ACCESS_CREATE); if (ret) {
mlog_errno(ret); break;
}
/* * Record the start cpos so that we can use it to initialize * our xattr tree we also set the xh_num_bucket for the new * bucket.
*/ if (i == 0) {
*cpos = le32_to_cpu(bucket_xh(args->new_bucket)->
xh_entries[0].xe_name_hash);
bucket_xh(args->new_bucket)->xh_num_buckets =
cpu_to_le16(num_buckets);
}
ret = ocfs2_reflink_xattr_header(handle, args->reflink,
args->old_bucket->bu_bhs[0],
bucket_xh(args->old_bucket),
args->new_bucket->bu_bhs[0],
bucket_xh(args->new_bucket),
&vb, meta_ac,
ocfs2_get_reflink_xattr_value_root,
args); if (ret) {
mlog_errno(ret); break;
}
/* * Re-access and dirty the bucket to calculate metaecc. * Because we may extend the transaction in reflink_xattr_header * which will let the already accessed block gone.
*/
ret = ocfs2_xattr_bucket_journal_access(handle,
args->new_bucket,
OCFS2_JOURNAL_ACCESS_WRITE); if (ret) {
mlog_errno(ret); break;
}
ret = ocfs2_reflink_xattr_bucket(handle, blkno,
new_blkno, num_clusters,
&reflink_cpos, reflink_buckets,
meta_ac, data_ac, args); if (ret) {
mlog_errno(ret); goto out;
}
/* * For the 1st allocated cluster, we make it use the same cpos * so that the xattr tree looks the same as the original one * in the most case.
*/ if (!first_inserted) {
reflink_cpos = cpos;
first_inserted = 1;
}
ret = ocfs2_insert_extent(handle, et, reflink_cpos, new_blkno,
num_clusters, 0, meta_ac); if (ret)
mlog_errno(ret);
ret = ocfs2_lock_reflink_xattr_rec_allocators(args, &et, blkno,
len, &credits,
&meta_ac, &data_ac); if (ret) {
mlog_errno(ret); goto out;
}
handle = ocfs2_start_trans(osb, credits); if (IS_ERR(handle)) {
ret = PTR_ERR(handle);
mlog_errno(ret); goto out;
}
ret = ocfs2_reflink_xattr_buckets(handle, inode, args, &et,
meta_ac, data_ac,
blkno, cpos, len); if (ret)
mlog_errno(ret);
ocfs2_commit_trans(osb, handle);
out: if (meta_ac)
ocfs2_free_alloc_context(meta_ac); if (data_ac)
ocfs2_free_alloc_context(data_ac); return ret;
}
/* * Create reflinked xattr buckets. * We will add bucket one by one, and refcount all the xattrs in the bucket * if they are stored outside.
*/ staticint ocfs2_reflink_xattr_tree(struct ocfs2_xattr_reflink *args, struct buffer_head *blk_bh, struct buffer_head *new_blk_bh)
{ int ret; struct ocfs2_reflink_xattr_tree_args para;
if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)
indexed = 1;
ret = ocfs2_create_empty_xattr_block(args->new_inode, args->new_bh,
&new_blk_bh, indexed); if (ret) {
mlog_errno(ret); goto out;
}
if (!indexed)
ret = ocfs2_reflink_xattr_block(args, blk_bh, new_blk_bh); else
ret = ocfs2_reflink_xattr_tree(args, blk_bh, new_blk_bh); if (ret)
mlog_errno(ret);
out:
brelse(new_blk_bh); return ret;
}
staticint ocfs2_reflink_xattr_no_security(struct ocfs2_xattr_entry *xe)
{ int type = ocfs2_xattr_get_type(xe);
return type != OCFS2_XATTR_INDEX_SECURITY &&
type != OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS &&
type != OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT;
}
if (ocfs2_dealloc_has_cluster(&dealloc)) {
ocfs2_schedule_truncate_log_flush(OCFS2_SB(old_inode->i_sb), 1);
ocfs2_run_deallocs(OCFS2_SB(old_inode->i_sb), &dealloc);
}
out: return ret;
}
/* * Initialize security and acl for a already created inode. * Used for reflink a non-preserve-security file. * * It uses common api like ocfs2_xattr_set, so the caller * must not hold any lock expect i_rwsem.
*/ int ocfs2_init_security_and_acl(struct inode *dir, struct inode *inode, conststruct qstr *qstr)
{ int ret = 0; struct buffer_head *dir_bh = NULL;
ret = ocfs2_init_security_get(inode, dir, qstr, NULL); if (ret) {
mlog_errno(ret); goto leave;
}
ret = ocfs2_inode_lock(dir, &dir_bh, 0); if (ret) {
mlog_errno(ret); goto leave;
}
ret = ocfs2_init_acl(NULL, inode, dir, NULL, dir_bh, NULL, NULL); if (ret)
mlog_errno(ret);
int ocfs2_init_security_get(struct inode *inode, struct inode *dir, conststruct qstr *qstr, struct ocfs2_security_xattr_info *si)
{ int ret;
/* check whether ocfs2 support feature xattr */ if (!ocfs2_supports_xattr(OCFS2_SB(dir->i_sb))) return -EOPNOTSUPP; if (si) {
ret = security_inode_init_security(inode, dir, qstr,
&ocfs2_initxattrs, si); /* * security_inode_init_security() does not return -EOPNOTSUPP, * we have to check the xattr ourselves.
*/ if (!ret && !si->name)
si->enable = 0;
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.