int
search_by_key(struct reiserfs_sb_info *p_s_sbi,
const struct cpu_key * p_s_key, /* Key to search. */
struct path * p_s_search_path, /* This structure was allocated and
initialized by the calling function.
It is filled up by this function. */
int n_stop_level) /* How far down the tree to search. To
stop at leaf level - set to
DISK_LEAF_NODE_LEVEL */
{
int error;
int n_node_level, n_retval;
int n_block_number, expected_level, fs_gen;
struct path_element *p_s_last_element;
struct buf *p_s_bp, *tmp_bp;
/*
* As we add each node to a path we increase its count. This means that
* we must be careful to release all nodes in a path before we either
* discard the path struct or re-use the path struct, as we do here.
*/
decrement_counters_in_path(p_s_search_path);
/*
* With each iteration of this loop we search through the items in the
* current node, and calculate the next current node(next path element)
* for the next iteration of this loop...
*/
n_block_number = SB_ROOT_BLOCK(p_s_sbi);
expected_level = -1;
reiserfs_log(LOG_DEBUG, "root block: #%d\n", n_block_number);
while (1) {
/* Prep path to have another element added to it. */
reiserfs_log(LOG_DEBUG, "path element #%d\n",
p_s_search_path->path_length);
p_s_last_element = PATH_OFFSET_PELEMENT(p_s_search_path,
++p_s_search_path->path_length);
fs_gen = get_generation(p_s_sbi);
/*
* Read the next tree node, and set the last element in the
* path to have a pointer to it.
*/
reiserfs_log(LOG_DEBUG, "reading block #%d\n",
n_block_number);
if ((error = bread(p_s_sbi->s_devvp,
n_block_number * btodb(p_s_sbi->s_blocksize),
p_s_sbi->s_blocksize, NOCRED, &tmp_bp)) != 0) {
reiserfs_log(LOG_DEBUG, "error reading block\n");
p_s_search_path->path_length--;
pathrelse(p_s_search_path);
return (IO_ERROR);
}
reiserfs_log(LOG_DEBUG, "blkno = %ju, lblkno = %ju\n",
(intmax_t)tmp_bp->b_blkno, (intmax_t)tmp_bp->b_lblkno);
/*
* As i didn't found a way to handle the lock correctly,
* i copy the data into a fake buffer
*/
reiserfs_log(LOG_DEBUG, "allocating p_s_bp\n");
p_s_bp = malloc(sizeof *p_s_bp, M_REISERFSPATH, M_WAITOK);
if (!p_s_bp) {
reiserfs_log(LOG_DEBUG, "error allocating memory\n");
p_s_search_path->path_length--;
pathrelse(p_s_search_path);
brelse(tmp_bp);
return (IO_ERROR);
}
reiserfs_log(LOG_DEBUG, "copying struct buf\n");
bcopy(tmp_bp, p_s_bp, sizeof(struct buf));
reiserfs_log(LOG_DEBUG, "allocating p_s_bp->b_data\n");
p_s_bp->b_data = malloc(p_s_sbi->s_blocksize,
M_REISERFSPATH, M_WAITOK);
if (!p_s_bp->b_data) {
reiserfs_log(LOG_DEBUG, "error allocating memory\n");
p_s_search_path->path_length--;
pathrelse(p_s_search_path);
free(p_s_bp, M_REISERFSPATH);
brelse(tmp_bp);
return (IO_ERROR);
}
reiserfs_log(LOG_DEBUG, "copying buffer data\n");
bcopy(tmp_bp->b_data, p_s_bp->b_data, p_s_sbi->s_blocksize);
brelse(tmp_bp);
tmp_bp = NULL;
reiserfs_log(LOG_DEBUG, "...done\n");
p_s_last_element->pe_buffer = p_s_bp;
if (expected_level == -1)
expected_level = SB_TREE_HEIGHT(p_s_sbi);
expected_level--;
reiserfs_log(LOG_DEBUG, "expected level: %d (%d)\n",
expected_level, SB_TREE_HEIGHT(p_s_sbi));
/* XXX */
/*
* It is possible that schedule occurred. We must check
* whether the key to search is still in the tree rooted
* from the current buffer. If not then repeat search
* from the root.
*/
if (fs_changed(fs_gen, p_s_sbi) &&
(!B_IS_IN_TREE(p_s_bp) ||
B_LEVEL(p_s_bp) != expected_level ||
!key_in_buffer(p_s_search_path, p_s_key, p_s_sbi))) {
reiserfs_log(LOG_DEBUG,
"the key isn't in the tree anymore\n");
decrement_counters_in_path(p_s_search_path);
/*
* Get the root block number so that we can repeat
* the search starting from the root.
*/
n_block_number = SB_ROOT_BLOCK(p_s_sbi);
expected_level = -1;
/* Repeat search from the root */
continue;
}
/*
* Make sure, that the node contents look like a node of
* certain level
*/
if (!is_tree_node(p_s_bp, expected_level)) {
reiserfs_log(LOG_WARNING,
"invalid format found in block %ju. Fsck?",
(intmax_t)p_s_bp->b_blkno);
pathrelse (p_s_search_path);
return (IO_ERROR);
}
/* Ok, we have acquired next formatted node in the tree */
n_node_level = B_LEVEL(p_s_bp);
reiserfs_log(LOG_DEBUG, "block info:\n");
reiserfs_log(LOG_DEBUG, " node level: %d\n",
n_node_level);
reiserfs_log(LOG_DEBUG, " nb of items: %d\n",
B_NR_ITEMS(p_s_bp));
reiserfs_log(LOG_DEBUG, " free space: %d bytes\n",
B_FREE_SPACE(p_s_bp));
reiserfs_log(LOG_DEBUG, "bin_search with :\n"
" p_s_key = (objectid=%d, dirid=%d)\n"
" B_NR_ITEMS(p_s_bp) = %d\n"
" p_s_last_element->pe_position = %d (path_length = %d)\n",
p_s_key->on_disk_key.k_objectid,
p_s_key->on_disk_key.k_dir_id,
B_NR_ITEMS(p_s_bp),
p_s_last_element->pe_position,
p_s_search_path->path_length);
n_retval = bin_search(p_s_key, B_N_PITEM_HEAD(p_s_bp, 0),
B_NR_ITEMS(p_s_bp),
(n_node_level == DISK_LEAF_NODE_LEVEL) ? IH_SIZE : KEY_SIZE,
&(p_s_last_element->pe_position));
reiserfs_log(LOG_DEBUG, "bin_search result: %d\n",
n_retval);
if (n_node_level == n_stop_level) {
reiserfs_log(LOG_DEBUG, "stop level reached (%s)\n",
n_retval == ITEM_FOUND ? "found" : "not found");
return (n_retval);
}
/* We are not in the stop level */
if (n_retval == ITEM_FOUND)
/*
* Item has been found, so we choose the pointer
* which is to the right of the found one
*/
p_s_last_element->pe_position++;
/*
* If item was not found we choose the position which is
* to the left of the found item. This requires no code,
* bin_search did it already.
*/
/*
* So we have chosen a position in the current node which
* is an internal node. Now we calculate child block number
* by position in the node.
*/
n_block_number = B_N_CHILD_NUM(p_s_bp,
p_s_last_element->pe_position);
}
reiserfs_log(LOG_DEBUG, "done\n");
return (0);
}
/* This function fills up the path from the root to the leaf as it
descends the tree looking for the key. It uses reiserfs_bread to
try to find buffers in the cache given their block number. If it
does not find them in the cache it reads them from disk. For each
node search_by_key finds using reiserfs_bread it then uses
bin_search to look through that node. bin_search will find the
position of the block_number of the next node if it is looking
through an internal node. If it is looking through a leaf node
bin_search will find the position of the item which has key either
equal to given key, or which is the maximal key less than the given
key. search_by_key returns a path that must be checked for the
correctness of the top of the path but need not be checked for the
correctness of the bottom of the path */
int search_by_key(
struct super_block * p_s_sb, /* Super block. */
struct key * p_s_key, /* Key to search. */
struct path * p_s_search_path,/* This structure was allocated and initialized by
the calling function. It is filled up by this
function. */
int * p_n_repeat, /* Whether schedule occured. */
int n_stop_level /* How far down the tree to search.*/
) {
dev_t n_dev = p_s_sb->s_dev;
int n_repeat,
n_block_number = SB_ROOT_BLOCK (p_s_sb),
expected_level = SB_TREE_HEIGHT (p_s_sb),
n_block_size = p_s_sb->s_blocksize;
struct buffer_head * p_s_bh;
struct path_element * p_s_last_element;
int n_retval;
int right_neighbor_of_leaf_node;
#ifdef CONFIG_REISERFS_CHECK
int n_repeat_counter = 0;
#endif
/* As we add each node to a path we increase its count. This
means that we must be careful to release all nodes in a path
before we either discard the path struct or re-use the path
struct, as we do here. */
pathrelse (p_s_search_path);
*p_n_repeat = CARRY_ON;
/* With each iteration of this loop we search through the items in
the current node, and calculate the next current node(next path
element) for the next iteration of this loop.. */
while ( 1 ) {
#ifdef CONFIG_REISERFS_CHECK
if ( !(++n_repeat_counter % 50000) )
printk ("PAP-5100: search_by_key(pid %u): there were %d searches from the tree_root lokking for key %p\n",
current->pid, n_repeat_counter, p_s_key);
#endif
/* prep path to have another element added to it. */
p_s_last_element = PATH_OFFSET_PELEMENT(p_s_search_path, ++p_s_search_path->path_length);
expected_level --;
n_repeat = CARRY_ON;
/* Read the next tree node, and set the last element in the
path to have a pointer to it. */
if ( ! (p_s_bh = p_s_last_element->pe_buffer =
reiserfs_bread(n_dev, n_block_number, n_block_size, &n_repeat)) ) {
p_s_search_path->path_length --;
pathrelse(p_s_search_path);
*p_n_repeat |= n_repeat;
return IO_ERROR;
}
*p_n_repeat |= n_repeat;
/* It is possible that schedule occured. We must check whether
the key to search is still in the tree rooted from the
current buffer. If not then repeat search from the root. */
if ( n_repeat != CARRY_ON &&
(!B_IS_IN_TREE (p_s_bh) || (! key_in_buffer(p_s_search_path, p_s_key, p_s_sb))) ) {
pathrelse (p_s_search_path);
/* Get the root block number so that we can repeat the
search starting from the root. */
n_block_number = SB_ROOT_BLOCK (p_s_sb);
expected_level = SB_TREE_HEIGHT (p_s_sb);
right_neighbor_of_leaf_node = 0;
/* repeat search from the root */
continue;
}
#ifdef CONFIG_REISERFS_CHECK
if ( ! key_in_buffer(p_s_search_path, p_s_key, p_s_sb) )
reiserfs_panic(p_s_sb, "PAP-5130: search_by_key: key is not in the buffer");
if ( cur_tb ) {
/* print_tb (init_mode, init_item_pos, init_pos_in_item, &init_tb, "5140");*/
reiserfs_panic(p_s_sb, "PAP-5140: search_by_key: schedule occurred in do_balance!");
}
#endif
// make sure, that the node contents look like a nod of
// certain level
if (!is_tree_node (p_s_bh, expected_level)) {
print_block (stderr, 0, p_s_bh, 3, -1, -1);
reiserfs_panic ("vs-5150: search_by_key: expeced level %d", expected_level);
pathrelse (p_s_search_path);
return IO_ERROR;
}
/* ok, we have acquired next formatted node in the tree */
n_retval = bin_search (p_s_key, B_N_PITEM_HEAD(p_s_bh, 0), B_NR_ITEMS(p_s_bh),
is_leaf_node (p_s_bh) ? IH_SIZE : KEY_SIZE, &(p_s_last_element->pe_position));
if (node_level (p_s_bh) == n_stop_level)
return n_retval;
/* we are not in the stop level */
if (n_retval == ITEM_FOUND)
/* item has been found, so we choose the pointer which is to the right of the found one */
p_s_last_element->pe_position++;
/* if item was not found we choose the position which is to the left of the found item. This
requires no code, bin_search did it already.*/
/* So we have chosen a position in the current node which is an
internal node. Now we calculate child block number by position in the node. */
n_block_number = B_N_CHILD_NUM(p_s_bh, p_s_last_element->pe_position);
}
}