/* Called from locking and called from ocfs2_clear_inode. Dump the
* cache for a given inode.
*
* This function is a few more lines longer than necessary due to some
* accounting done here, but I think it's worth tracking down those
* bugs sooner -- Mark */
void ocfs2_metadata_cache_purge(struct ocfs2_caching_info *ci)
{
unsigned int tree, to_purge, purged;
struct rb_root root = RB_ROOT;
BUG_ON(!ci || !ci->ci_ops);
ocfs2_metadata_cache_lock(ci);
tree = !(ci->ci_flags & OCFS2_CACHE_FL_INLINE);
to_purge = ci->ci_num_cached;
trace_ocfs2_metadata_cache_purge(
(unsigned long long)ocfs2_metadata_cache_owner(ci),
to_purge, tree);
/* If we're a tree, save off the root so that we can safely
* initialize the cache. We do the work to free tree members
* without the spinlock. */
if (tree)
root = ci->ci_cache.ci_tree;
ocfs2_metadata_cache_reset(ci, 0);
ocfs2_metadata_cache_unlock(ci);
purged = ocfs2_purge_copied_metadata_tree(&root);
/* If possible, track the number wiped so that we can more
* easily detect counting errors. Unfortunately, this is only
* meaningful for trees. */
if (tree && purged != to_purge)
mlog(ML_ERROR, "Owner %llu, count = %u, purged = %u\n",
(unsigned long long)ocfs2_metadata_cache_owner(ci),
to_purge, purged);
}
/* Requires ip_lock */
static void ocfs2_append_cache_array(struct ocfs2_caching_info *ci,
sector_t block)
{
BUG_ON(ci->ci_num_cached >= OCFS2_CACHE_INFO_MAX_ARRAY);
trace_ocfs2_append_cache_array(
(unsigned long long)ocfs2_metadata_cache_owner(ci),
(unsigned long long)block, ci->ci_num_cached);
ci->ci_cache.ci_array[ci->ci_num_cached] = block;
ci->ci_num_cached++;
}
static int ocfs2_buffer_cached(struct ocfs2_caching_info *ci,
struct buffer_head *bh)
{
int index = -1;
struct ocfs2_meta_cache_item *item = NULL;
ocfs2_metadata_cache_lock(ci);
trace_ocfs2_buffer_cached_begin(
(unsigned long long)ocfs2_metadata_cache_owner(ci),
(unsigned long long) bh->b_blocknr,
!!(ci->ci_flags & OCFS2_CACHE_FL_INLINE));
if (ci->ci_flags & OCFS2_CACHE_FL_INLINE)
index = ocfs2_search_cache_array(ci, bh->b_blocknr);
else
item = ocfs2_search_cache_tree(ci, bh->b_blocknr);
ocfs2_metadata_cache_unlock(ci);
trace_ocfs2_buffer_cached_end(index, item);
return (index != -1) || (item != NULL);
}
static int ocfs2_buffer_cached(struct ocfs2_caching_info *ci,
struct buffer_head *bh)
{
int index = -1;
struct ocfs2_meta_cache_item *item = NULL;
ocfs2_metadata_cache_lock(ci);
mlog(0, "Owner %llu, query block %llu (inline = %u)\n",
(unsigned long long)ocfs2_metadata_cache_owner(ci),
(unsigned long long) bh->b_blocknr,
!!(ci->ci_flags & OCFS2_CACHE_FL_INLINE));
if (ci->ci_flags & OCFS2_CACHE_FL_INLINE)
index = ocfs2_search_cache_array(ci, bh->b_blocknr);
else
item = ocfs2_search_cache_tree(ci, bh->b_blocknr);
ocfs2_metadata_cache_unlock(ci);
mlog(0, "index = %d, item = %p\n", index, item);
return (index != -1) || (item != NULL);
}
static int __ocfs2_move_extent(handle_t *handle,
struct ocfs2_move_extents_context *context,
u32 cpos, u32 len, u32 p_cpos, u32 new_p_cpos,
int ext_flags)
{
int ret = 0, index;
struct inode *inode = context->inode;
struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
struct ocfs2_extent_rec *rec, replace_rec;
struct ocfs2_path *path = NULL;
struct ocfs2_extent_list *el;
u64 ino = ocfs2_metadata_cache_owner(context->et.et_ci);
u64 old_blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cpos);
ret = ocfs2_duplicate_clusters_by_page(handle, inode, cpos,
p_cpos, new_p_cpos, len);
if (ret) {
mlog_errno(ret);
goto out;
}
memset(&replace_rec, 0, sizeof(replace_rec));
replace_rec.e_cpos = cpu_to_le32(cpos);
replace_rec.e_leaf_clusters = cpu_to_le16(len);
replace_rec.e_blkno = cpu_to_le64(ocfs2_clusters_to_blocks(inode->i_sb,
new_p_cpos));
path = ocfs2_new_path_from_et(&context->et);
if (!path) {
ret = -ENOMEM;
mlog_errno(ret);
goto out;
}
ret = ocfs2_find_path(INODE_CACHE(inode), path, cpos);
if (ret) {
mlog_errno(ret);
goto out;
}
el = path_leaf_el(path);
index = ocfs2_search_extent_list(el, cpos);
if (index == -1) {
ret = ocfs2_error(inode->i_sb,
"Inode %llu has an extent at cpos %u which can no longer be found\n",
(unsigned long long)ino, cpos);
goto out;
}
rec = &el->l_recs[index];
BUG_ON(ext_flags != rec->e_flags);
/*
* after moving/defraging to new location, the extent is not going
* to be refcounted anymore.
*/
replace_rec.e_flags = ext_flags & ~OCFS2_EXT_REFCOUNTED;
ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode),
context->et.et_root_bh,
OCFS2_JOURNAL_ACCESS_WRITE);
if (ret) {
mlog_errno(ret);
goto out;
}
ret = ocfs2_split_extent(handle, &context->et, path, index,
&replace_rec, context->meta_ac,
&context->dealloc);
if (ret) {
mlog_errno(ret);
goto out;
}
ocfs2_journal_dirty(handle, context->et.et_root_bh);
context->new_phys_cpos = new_p_cpos;
/*
* need I to append truncate log for old clusters?
*/
if (old_blkno) {
if (ext_flags & OCFS2_EXT_REFCOUNTED)
ret = ocfs2_decrease_refcount(inode, handle,
ocfs2_blocks_to_clusters(osb->sb,
old_blkno),
len, context->meta_ac,
&context->dealloc, 1);
else
ret = ocfs2_truncate_log_append(osb, handle,
old_blkno, len);
}
ocfs2_update_inode_fsync_trans(handle, inode, 0);
out:
ocfs2_free_path(path);
return ret;
}
ci->ci_num_cached++;
}
/* By now the caller should have checked that the item does *not*
* exist in the tree.
* Requires ip_lock. */
static void __ocfs2_insert_cache_tree(struct ocfs2_caching_info *ci,
struct ocfs2_meta_cache_item *new)
{
sector_t block = new->c_block;
struct rb_node *parent = NULL;
struct rb_node **p = &ci->ci_cache.ci_tree.rb_node;
struct ocfs2_meta_cache_item *tmp;
trace_ocfs2_insert_cache_tree(
(unsigned long long)ocfs2_metadata_cache_owner(ci),
(unsigned long long)block, ci->ci_num_cached);
while(*p) {
parent = *p;
tmp = rb_entry(parent, struct ocfs2_meta_cache_item, c_node);
if (block < tmp->c_block)
p = &(*p)->rb_left;
else if (block > tmp->c_block)
p = &(*p)->rb_right;
else {
/* This should never happen! */
mlog(ML_ERROR, "Duplicate block %llu cached!\n",
(unsigned long long) block);
int ocfs2_read_blocks(struct ocfs2_caching_info *ci, u64 block, int nr,
struct buffer_head *bhs[], int flags,
int (*validate)(struct super_block *sb,
struct buffer_head *bh))
{
int status = 0;
int i, ignore_cache = 0;
struct buffer_head *bh;
struct super_block *sb = ocfs2_metadata_cache_get_super(ci);
trace_ocfs2_read_blocks_begin(ci, (unsigned long long)block, nr, flags);
BUG_ON(!ci);
BUG_ON((flags & OCFS2_BH_READAHEAD) &&
(flags & OCFS2_BH_IGNORE_CACHE));
if (bhs == NULL) {
status = -EINVAL;
mlog_errno(status);
goto bail;
}
if (nr < 0) {
mlog(ML_ERROR, "asked to read %d blocks!\n", nr);
status = -EINVAL;
mlog_errno(status);
goto bail;
}
if (nr == 0) {
status = 0;
goto bail;
}
ocfs2_metadata_cache_io_lock(ci);
for (i = 0 ; i < nr ; i++) {
if (bhs[i] == NULL) {
bhs[i] = sb_getblk(sb, block++);
if (bhs[i] == NULL) {
ocfs2_metadata_cache_io_unlock(ci);
status = -ENOMEM;
mlog_errno(status);
goto bail;
}
}
bh = bhs[i];
ignore_cache = (flags & OCFS2_BH_IGNORE_CACHE);
/* There are three read-ahead cases here which we need to
* be concerned with. All three assume a buffer has
* previously been submitted with OCFS2_BH_READAHEAD
* and it hasn't yet completed I/O.
*
* 1) The current request is sync to disk. This rarely
* happens these days, and never when performance
* matters - the code can just wait on the buffer
* lock and re-submit.
*
* 2) The current request is cached, but not
* readahead. ocfs2_buffer_uptodate() will return
* false anyway, so we'll wind up waiting on the
* buffer lock to do I/O. We re-check the request
* with after getting the lock to avoid a re-submit.
*
* 3) The current request is readahead (and so must
* also be a caching one). We short circuit if the
* buffer is locked (under I/O) and if it's in the
* uptodate cache. The re-check from #2 catches the
* case that the previous read-ahead completes just
* before our is-it-in-flight check.
*/
if (!ignore_cache && !ocfs2_buffer_uptodate(ci, bh)) {
trace_ocfs2_read_blocks_from_disk(
(unsigned long long)bh->b_blocknr,
(unsigned long long)ocfs2_metadata_cache_owner(ci));
/* We're using ignore_cache here to say
* "go to disk" */
ignore_cache = 1;
}
trace_ocfs2_read_blocks_bh((unsigned long long)bh->b_blocknr,
ignore_cache, buffer_jbd(bh), buffer_dirty(bh));
if (buffer_jbd(bh)) {
continue;
}
if (ignore_cache) {
if (buffer_dirty(bh)) {
/* This should probably be a BUG, or
* at least return an error. */
continue;
}
/* A read-ahead request was made - if the
* buffer is already under read-ahead from a
* previously submitted request than we are
* done here. */
if ((flags & OCFS2_BH_READAHEAD)
&& ocfs2_buffer_read_ahead(ci, bh))
continue;
lock_buffer(bh);
if (buffer_jbd(bh)) {
#ifdef CATCH_BH_JBD_RACES
mlog(ML_ERROR, "block %llu had the JBD bit set "
"while I was in lock_buffer!",
(unsigned long long)bh->b_blocknr);
BUG();
#else
unlock_buffer(bh);
continue;
#endif
}
/* Re-check ocfs2_buffer_uptodate() as a
* previously read-ahead buffer may have
* completed I/O while we were waiting for the
* buffer lock. */
if (!(flags & OCFS2_BH_IGNORE_CACHE)
&& !(flags & OCFS2_BH_READAHEAD)
&& ocfs2_buffer_uptodate(ci, bh)) {
unlock_buffer(bh);
continue;
}
clear_buffer_uptodate(bh);
get_bh(bh); /* for end_buffer_read_sync() */
if (validate)
set_buffer_needs_validate(bh);
bh->b_end_io = end_buffer_read_sync;
submit_bh(REQ_OP_READ, 0, bh);
continue;
}
}
status = 0;
for (i = (nr - 1); i >= 0; i--) {
bh = bhs[i];
if (!(flags & OCFS2_BH_READAHEAD)) {
if (status) {
/* Clear the rest of the buffers on error */
put_bh(bh);
bhs[i] = NULL;
continue;
}
/* We know this can't have changed as we hold the
* owner sem. Avoid doing any work on the bh if the
* journal has it. */
if (!buffer_jbd(bh))
wait_on_buffer(bh);
if (!buffer_uptodate(bh)) {
/* Status won't be cleared from here on out,
* so we can safely record this and loop back
* to cleanup the other buffers. Don't need to
* remove the clustered uptodate information
* for this bh as it's not marked locally
* uptodate. */
status = -EIO;
put_bh(bh);
bhs[i] = NULL;
continue;
}
if (buffer_needs_validate(bh)) {
/* We never set NeedsValidate if the
* buffer was held by the journal, so
* that better not have changed */
BUG_ON(buffer_jbd(bh));
clear_buffer_needs_validate(bh);
status = validate(sb, bh);
if (status) {
put_bh(bh);
bhs[i] = NULL;
continue;
}
}
}
/* Always set the buffer in the cache, even if it was
* a forced read, or read-ahead which hasn't yet
* completed. */
ocfs2_set_buffer_uptodate(ci, bh);
}
ocfs2_metadata_cache_io_unlock(ci);
trace_ocfs2_read_blocks_end((unsigned long long)block, nr,
flags, ignore_cache);
bail:
return status;
}