blob: 85a1f8621b51f09f0893abae9ffab90382efdcfc [file] [log] [blame]
/*
* Copyright (C) 2008 Red Hat. All rights reserved.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public
* License v2 as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* You should have received a copy of the GNU General Public
* License along with this program; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 021110-1307, USA.
*/
#include <linux/pagemap.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/math64.h>
#include <linux/ratelimit.h>
#include "ctree.h"
#include "free-space-cache.h"
#include "transaction.h"
#include "disk-io.h"
#include "extent_io.h"
#include "inode-map.h"
#include "volumes.h"
#define BITS_PER_BITMAP (PAGE_CACHE_SIZE * 8)
#define MAX_CACHE_BYTES_PER_GIG (32 * 1024)
struct btrfs_trim_range {
u64 start;
u64 bytes;
struct list_head list;
};
static int link_free_space(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info);
static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info);
static struct inode *__lookup_free_space_inode(struct btrfs_root *root,
struct btrfs_path *path,
u64 offset)
{
struct btrfs_key key;
struct btrfs_key location;
struct btrfs_disk_key disk_key;
struct btrfs_free_space_header *header;
struct extent_buffer *leaf;
struct inode *inode = NULL;
int ret;
key.objectid = BTRFS_FREE_SPACE_OBJECTID;
key.offset = offset;
key.type = 0;
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
return ERR_PTR(ret);
if (ret > 0) {
btrfs_release_path(path);
return ERR_PTR(-ENOENT);
}
leaf = path->nodes[0];
header = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_free_space_header);
btrfs_free_space_key(leaf, header, &disk_key);
btrfs_disk_key_to_cpu(&location, &disk_key);
btrfs_release_path(path);
inode = btrfs_iget(root->fs_info->sb, &location, root, NULL);
if (!inode)
return ERR_PTR(-ENOENT);
if (IS_ERR(inode))
return inode;
if (is_bad_inode(inode)) {
iput(inode);
return ERR_PTR(-ENOENT);
}
mapping_set_gfp_mask(inode->i_mapping,
mapping_gfp_constraint(inode->i_mapping,
~(__GFP_FS | __GFP_HIGHMEM)));
return inode;
}
struct inode *lookup_free_space_inode(struct btrfs_root *root,
struct btrfs_block_group_cache
*block_group, struct btrfs_path *path)
{
struct inode *inode = NULL;
u32 flags = BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
spin_lock(&block_group->lock);
if (block_group->inode)
inode = igrab(block_group->inode);
spin_unlock(&block_group->lock);
if (inode)
return inode;
inode = __lookup_free_space_inode(root, path,
block_group->key.objectid);
if (IS_ERR(inode))
return inode;
spin_lock(&block_group->lock);
if (!((BTRFS_I(inode)->flags & flags) == flags)) {
btrfs_info(root->fs_info,
"Old style space inode found, converting.");
BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM |
BTRFS_INODE_NODATACOW;
block_group->disk_cache_state = BTRFS_DC_CLEAR;
}
if (!block_group->iref) {
block_group->inode = igrab(inode);
block_group->iref = 1;
}
spin_unlock(&block_group->lock);
return inode;
}
static int __create_free_space_inode(struct btrfs_root *root,
struct btrfs_trans_handle *trans,
struct btrfs_path *path,
u64 ino, u64 offset)
{
struct btrfs_key key;
struct btrfs_disk_key disk_key;
struct btrfs_free_space_header *header;
struct btrfs_inode_item *inode_item;
struct extent_buffer *leaf;
u64 flags = BTRFS_INODE_NOCOMPRESS | BTRFS_INODE_PREALLOC;
int ret;
ret = btrfs_insert_empty_inode(trans, root, path, ino);
if (ret)
return ret;
/* We inline crc's for the free disk space cache */
if (ino != BTRFS_FREE_INO_OBJECTID)
flags |= BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
leaf = path->nodes[0];
inode_item = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_inode_item);
btrfs_item_key(leaf, &disk_key, path->slots[0]);
memset_extent_buffer(leaf, 0, (unsigned long)inode_item,
sizeof(*inode_item));
btrfs_set_inode_generation(leaf, inode_item, trans->transid);
btrfs_set_inode_size(leaf, inode_item, 0);
btrfs_set_inode_nbytes(leaf, inode_item, 0);
btrfs_set_inode_uid(leaf, inode_item, 0);
btrfs_set_inode_gid(leaf, inode_item, 0);
btrfs_set_inode_mode(leaf, inode_item, S_IFREG | 0600);
btrfs_set_inode_flags(leaf, inode_item, flags);
btrfs_set_inode_nlink(leaf, inode_item, 1);
btrfs_set_inode_transid(leaf, inode_item, trans->transid);
btrfs_set_inode_block_group(leaf, inode_item, offset);
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(path);
key.objectid = BTRFS_FREE_SPACE_OBJECTID;
key.offset = offset;
key.type = 0;
ret = btrfs_insert_empty_item(trans, root, path, &key,
sizeof(struct btrfs_free_space_header));
if (ret < 0) {
btrfs_release_path(path);
return ret;
}
leaf = path->nodes[0];
header = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_free_space_header);
memset_extent_buffer(leaf, 0, (unsigned long)header, sizeof(*header));
btrfs_set_free_space_key(leaf, header, &disk_key);
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(path);
return 0;
}
int create_free_space_inode(struct btrfs_root *root,
struct btrfs_trans_handle *trans,
struct btrfs_block_group_cache *block_group,
struct btrfs_path *path)
{
int ret;
u64 ino;
ret = btrfs_find_free_objectid(root, &ino);
if (ret < 0)
return ret;
return __create_free_space_inode(root, trans, path, ino,
block_group->key.objectid);
}
int btrfs_check_trunc_cache_free_space(struct btrfs_root *root,
struct btrfs_block_rsv *rsv)
{
u64 needed_bytes;
int ret;
/* 1 for slack space, 1 for updating the inode */
needed_bytes = btrfs_calc_trunc_metadata_size(root, 1) +
btrfs_calc_trans_metadata_size(root, 1);
spin_lock(&rsv->lock);
if (rsv->reserved < needed_bytes)
ret = -ENOSPC;
else
ret = 0;
spin_unlock(&rsv->lock);
return ret;
}
int btrfs_truncate_free_space_cache(struct btrfs_root *root,
struct btrfs_trans_handle *trans,
struct btrfs_block_group_cache *block_group,
struct inode *inode)
{
int ret = 0;
struct btrfs_path *path = btrfs_alloc_path();
bool locked = false;
if (!path) {
ret = -ENOMEM;
goto fail;
}
if (block_group) {
locked = true;
mutex_lock(&trans->transaction->cache_write_mutex);
if (!list_empty(&block_group->io_list)) {
list_del_init(&block_group->io_list);
btrfs_wait_cache_io(root, trans, block_group,
&block_group->io_ctl, path,
block_group->key.objectid);
btrfs_put_block_group(block_group);
}
/*
* now that we've truncated the cache away, its no longer
* setup or written
*/
spin_lock(&block_group->lock);
block_group->disk_cache_state = BTRFS_DC_CLEAR;
spin_unlock(&block_group->lock);
}
btrfs_free_path(path);
btrfs_i_size_write(inode, 0);
truncate_pagecache(inode, 0);
/*
* We don't need an orphan item because truncating the free space cache
* will never be split across transactions.
* We don't need to check for -EAGAIN because we're a free space
* cache inode
*/
ret = btrfs_truncate_inode_items(trans, root, inode,
0, BTRFS_EXTENT_DATA_KEY);
if (ret)
goto fail;
ret = btrfs_update_inode(trans, root, inode);
fail:
if (locked)
mutex_unlock(&trans->transaction->cache_write_mutex);
if (ret)
btrfs_abort_transaction(trans, root, ret);
return ret;
}
static int readahead_cache(struct inode *inode)
{
struct file_ra_state *ra;
unsigned long last_index;
ra = kzalloc(sizeof(*ra), GFP_NOFS);
if (!ra)
return -ENOMEM;
file_ra_state_init(ra, inode->i_mapping);
last_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;
page_cache_sync_readahead(inode->i_mapping, ra, NULL, 0, last_index);
kfree(ra);
return 0;
}
static int io_ctl_init(struct btrfs_io_ctl *io_ctl, struct inode *inode,
struct btrfs_root *root, int write)
{
int num_pages;
int check_crcs = 0;
num_pages = DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE);
if (btrfs_ino(inode) != BTRFS_FREE_INO_OBJECTID)
check_crcs = 1;
/* Make sure we can fit our crcs into the first page */
if (write && check_crcs &&
(num_pages * sizeof(u32)) >= PAGE_CACHE_SIZE)
return -ENOSPC;
memset(io_ctl, 0, sizeof(struct btrfs_io_ctl));
io_ctl->pages = kcalloc(num_pages, sizeof(struct page *), GFP_NOFS);
if (!io_ctl->pages)
return -ENOMEM;
io_ctl->num_pages = num_pages;
io_ctl->root = root;
io_ctl->check_crcs = check_crcs;
io_ctl->inode = inode;
return 0;
}
static void io_ctl_free(struct btrfs_io_ctl *io_ctl)
{
kfree(io_ctl->pages);
io_ctl->pages = NULL;
}
static void io_ctl_unmap_page(struct btrfs_io_ctl *io_ctl)
{
if (io_ctl->cur) {
io_ctl->cur = NULL;
io_ctl->orig = NULL;
}
}
static void io_ctl_map_page(struct btrfs_io_ctl *io_ctl, int clear)
{
ASSERT(io_ctl->index < io_ctl->num_pages);
io_ctl->page = io_ctl->pages[io_ctl->index++];
io_ctl->cur = page_address(io_ctl->page);
io_ctl->orig = io_ctl->cur;
io_ctl->size = PAGE_CACHE_SIZE;
if (clear)
memset(io_ctl->cur, 0, PAGE_CACHE_SIZE);
}
static void io_ctl_drop_pages(struct btrfs_io_ctl *io_ctl)
{
int i;
io_ctl_unmap_page(io_ctl);
for (i = 0; i < io_ctl->num_pages; i++) {
if (io_ctl->pages[i]) {
ClearPageChecked(io_ctl->pages[i]);
unlock_page(io_ctl->pages[i]);
page_cache_release(io_ctl->pages[i]);
}
}
}
static int io_ctl_prepare_pages(struct btrfs_io_ctl *io_ctl, struct inode *inode,
int uptodate)
{
struct page *page;
gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
int i;
for (i = 0; i < io_ctl->num_pages; i++) {
page = find_or_create_page(inode->i_mapping, i, mask);
if (!page) {
io_ctl_drop_pages(io_ctl);
return -ENOMEM;
}
io_ctl->pages[i] = page;
if (uptodate && !PageUptodate(page)) {
btrfs_readpage(NULL, page);
lock_page(page);
if (!PageUptodate(page)) {
btrfs_err(BTRFS_I(inode)->root->fs_info,
"error reading free space cache");
io_ctl_drop_pages(io_ctl);
return -EIO;
}
}
}
for (i = 0; i < io_ctl->num_pages; i++) {
clear_page_dirty_for_io(io_ctl->pages[i]);
set_page_extent_mapped(io_ctl->pages[i]);
}
return 0;
}
static void io_ctl_set_generation(struct btrfs_io_ctl *io_ctl, u64 generation)
{
__le64 *val;
io_ctl_map_page(io_ctl, 1);
/*
* Skip the csum areas. If we don't check crcs then we just have a
* 64bit chunk at the front of the first page.
*/
if (io_ctl->check_crcs) {
io_ctl->cur += (sizeof(u32) * io_ctl->num_pages);
io_ctl->size -= sizeof(u64) + (sizeof(u32) * io_ctl->num_pages);
} else {
io_ctl->cur += sizeof(u64);
io_ctl->size -= sizeof(u64) * 2;
}
val = io_ctl->cur;
*val = cpu_to_le64(generation);
io_ctl->cur += sizeof(u64);
}
static int io_ctl_check_generation(struct btrfs_io_ctl *io_ctl, u64 generation)
{
__le64 *gen;
/*
* Skip the crc area. If we don't check crcs then we just have a 64bit
* chunk at the front of the first page.
*/
if (io_ctl->check_crcs) {
io_ctl->cur += sizeof(u32) * io_ctl->num_pages;
io_ctl->size -= sizeof(u64) +
(sizeof(u32) * io_ctl->num_pages);
} else {
io_ctl->cur += sizeof(u64);
io_ctl->size -= sizeof(u64) * 2;
}
gen = io_ctl->cur;
if (le64_to_cpu(*gen) != generation) {
btrfs_err_rl(io_ctl->root->fs_info,
"space cache generation (%llu) does not match inode (%llu)",
*gen, generation);
io_ctl_unmap_page(io_ctl);
return -EIO;
}
io_ctl->cur += sizeof(u64);
return 0;
}
static void io_ctl_set_crc(struct btrfs_io_ctl *io_ctl, int index)
{
u32 *tmp;
u32 crc = ~(u32)0;
unsigned offset = 0;
if (!io_ctl->check_crcs) {
io_ctl_unmap_page(io_ctl);
return;
}
if (index == 0)
offset = sizeof(u32) * io_ctl->num_pages;
crc = btrfs_csum_data(io_ctl->orig + offset, crc,
PAGE_CACHE_SIZE - offset);
btrfs_csum_final(crc, (char *)&crc);
io_ctl_unmap_page(io_ctl);
tmp = page_address(io_ctl->pages[0]);
tmp += index;
*tmp = crc;
}
static int io_ctl_check_crc(struct btrfs_io_ctl *io_ctl, int index)
{
u32 *tmp, val;
u32 crc = ~(u32)0;
unsigned offset = 0;
if (!io_ctl->check_crcs) {
io_ctl_map_page(io_ctl, 0);
return 0;
}
if (index == 0)
offset = sizeof(u32) * io_ctl->num_pages;
tmp = page_address(io_ctl->pages[0]);
tmp += index;
val = *tmp;
io_ctl_map_page(io_ctl, 0);
crc = btrfs_csum_data(io_ctl->orig + offset, crc,
PAGE_CACHE_SIZE - offset);
btrfs_csum_final(crc, (char *)&crc);
if (val != crc) {
btrfs_err_rl(io_ctl->root->fs_info,
"csum mismatch on free space cache");
io_ctl_unmap_page(io_ctl);
return -EIO;
}
return 0;
}
static int io_ctl_add_entry(struct btrfs_io_ctl *io_ctl, u64 offset, u64 bytes,
void *bitmap)
{
struct btrfs_free_space_entry *entry;
if (!io_ctl->cur)
return -ENOSPC;
entry = io_ctl->cur;
entry->offset = cpu_to_le64(offset);
entry->bytes = cpu_to_le64(bytes);
entry->type = (bitmap) ? BTRFS_FREE_SPACE_BITMAP :
BTRFS_FREE_SPACE_EXTENT;
io_ctl->cur += sizeof(struct btrfs_free_space_entry);
io_ctl->size -= sizeof(struct btrfs_free_space_entry);
if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
return 0;
io_ctl_set_crc(io_ctl, io_ctl->index - 1);
/* No more pages to map */
if (io_ctl->index >= io_ctl->num_pages)
return 0;
/* map the next page */
io_ctl_map_page(io_ctl, 1);
return 0;
}
static int io_ctl_add_bitmap(struct btrfs_io_ctl *io_ctl, void *bitmap)
{
if (!io_ctl->cur)
return -ENOSPC;
/*
* If we aren't at the start of the current page, unmap this one and
* map the next one if there is any left.
*/
if (io_ctl->cur != io_ctl->orig) {
io_ctl_set_crc(io_ctl, io_ctl->index - 1);
if (io_ctl->index >= io_ctl->num_pages)
return -ENOSPC;
io_ctl_map_page(io_ctl, 0);
}
memcpy(io_ctl->cur, bitmap, PAGE_CACHE_SIZE);
io_ctl_set_crc(io_ctl, io_ctl->index - 1);
if (io_ctl->index < io_ctl->num_pages)
io_ctl_map_page(io_ctl, 0);
return 0;
}
static void io_ctl_zero_remaining_pages(struct btrfs_io_ctl *io_ctl)
{
/*
* If we're not on the boundary we know we've modified the page and we
* need to crc the page.
*/
if (io_ctl->cur != io_ctl->orig)
io_ctl_set_crc(io_ctl, io_ctl->index - 1);
else
io_ctl_unmap_page(io_ctl);
while (io_ctl->index < io_ctl->num_pages) {
io_ctl_map_page(io_ctl, 1);
io_ctl_set_crc(io_ctl, io_ctl->index - 1);
}
}
static int io_ctl_read_entry(struct btrfs_io_ctl *io_ctl,
struct btrfs_free_space *entry, u8 *type)
{
struct btrfs_free_space_entry *e;
int ret;
if (!io_ctl->cur) {
ret = io_ctl_check_crc(io_ctl, io_ctl->index);
if (ret)
return ret;
}
e = io_ctl->cur;
entry->offset = le64_to_cpu(e->offset);
entry->bytes = le64_to_cpu(e->bytes);
*type = e->type;
io_ctl->cur += sizeof(struct btrfs_free_space_entry);
io_ctl->size -= sizeof(struct btrfs_free_space_entry);
if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
return 0;
io_ctl_unmap_page(io_ctl);
return 0;
}
static int io_ctl_read_bitmap(struct btrfs_io_ctl *io_ctl,
struct btrfs_free_space *entry)
{
int ret;
ret = io_ctl_check_crc(io_ctl, io_ctl->index);
if (ret)
return ret;
memcpy(entry->bitmap, io_ctl->cur, PAGE_CACHE_SIZE);
io_ctl_unmap_page(io_ctl);
return 0;
}
/*
* Since we attach pinned extents after the fact we can have contiguous sections
* of free space that are split up in entries. This poses a problem with the
* tree logging stuff since it could have allocated across what appears to be 2
* entries since we would have merged the entries when adding the pinned extents
* back to the free space cache. So run through the space cache that we just
* loaded and merge contiguous entries. This will make the log replay stuff not
* blow up and it will make for nicer allocator behavior.
*/
static void merge_space_tree(struct btrfs_free_space_ctl *ctl)
{
struct btrfs_free_space *e, *prev = NULL;
struct rb_node *n;
again:
spin_lock(&ctl->tree_lock);
for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
e = rb_entry(n, struct btrfs_free_space, offset_index);
if (!prev)
goto next;
if (e->bitmap || prev->bitmap)
goto next;
if (prev->offset + prev->bytes == e->offset) {
unlink_free_space(ctl, prev);
unlink_free_space(ctl, e);
prev->bytes += e->bytes;
kmem_cache_free(btrfs_free_space_cachep, e);
link_free_space(ctl, prev);
prev = NULL;
spin_unlock(&ctl->tree_lock);
goto again;
}
next:
prev = e;
}
spin_unlock(&ctl->tree_lock);
}
static int __load_free_space_cache(struct btrfs_root *root, struct inode *inode,
struct btrfs_free_space_ctl *ctl,
struct btrfs_path *path, u64 offset)
{
struct btrfs_free_space_header *header;
struct extent_buffer *leaf;
struct btrfs_io_ctl io_ctl;
struct btrfs_key key;
struct btrfs_free_space *e, *n;
LIST_HEAD(bitmaps);
u64 num_entries;
u64 num_bitmaps;
u64 generation;
u8 type;
int ret = 0;
/* Nothing in the space cache, goodbye */
if (!i_size_read(inode))
return 0;
key.objectid = BTRFS_FREE_SPACE_OBJECTID;
key.offset = offset;
key.type = 0;
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
return 0;
else if (ret > 0) {
btrfs_release_path(path);
return 0;
}
ret = -1;
leaf = path->nodes[0];
header = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_free_space_header);
num_entries = btrfs_free_space_entries(leaf, header);
num_bitmaps = btrfs_free_space_bitmaps(leaf, header);
generation = btrfs_free_space_generation(leaf, header);
btrfs_release_path(path);
if (!BTRFS_I(inode)->generation) {
btrfs_info(root->fs_info,
"The free space cache file (%llu) is invalid. skip it\n",
offset);
return 0;
}
if (BTRFS_I(inode)->generation != generation) {
btrfs_err(root->fs_info,
"free space inode generation (%llu) "
"did not match free space cache generation (%llu)",
BTRFS_I(inode)->generation, generation);
return 0;
}
if (!num_entries)
return 0;
ret = io_ctl_init(&io_ctl, inode, root, 0);
if (ret)
return ret;
ret = readahead_cache(inode);
if (ret)
goto out;
ret = io_ctl_prepare_pages(&io_ctl, inode, 1);
if (ret)
goto out;
ret = io_ctl_check_crc(&io_ctl, 0);
if (ret)
goto free_cache;
ret = io_ctl_check_generation(&io_ctl, generation);
if (ret)
goto free_cache;
while (num_entries) {
e = kmem_cache_zalloc(btrfs_free_space_cachep,
GFP_NOFS);
if (!e)
goto free_cache;
ret = io_ctl_read_entry(&io_ctl, e, &type);
if (ret) {
kmem_cache_free(btrfs_free_space_cachep, e);
goto free_cache;
}
if (!e->bytes) {
kmem_cache_free(btrfs_free_space_cachep, e);
goto free_cache;
}
if (type == BTRFS_FREE_SPACE_EXTENT) {
spin_lock(&ctl->tree_lock);
ret = link_free_space(ctl, e);
spin_unlock(&ctl->tree_lock);
if (ret) {
btrfs_err(root->fs_info,
"Duplicate entries in free space cache, dumping");
kmem_cache_free(btrfs_free_space_cachep, e);
goto free_cache;
}
} else {
ASSERT(num_bitmaps);
num_bitmaps--;
e->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
if (!e->bitmap) {
kmem_cache_free(
btrfs_free_space_cachep, e);
goto free_cache;
}
spin_lock(&ctl->tree_lock);
ret = link_free_space(ctl, e);
ctl->total_bitmaps++;
ctl->op->recalc_thresholds(ctl);
spin_unlock(&ctl->tree_lock);
if (ret) {
btrfs_err(root->fs_info,
"Duplicate entries in free space cache, dumping");
kmem_cache_free(btrfs_free_space_cachep, e);
goto free_cache;
}
list_add_tail(&e->list, &bitmaps);
}
num_entries--;
}
io_ctl_unmap_page(&io_ctl);
/*
* We add the bitmaps at the end of the entries in order that
* the bitmap entries are added to the cache.
*/
list_for_each_entry_safe(e, n, &bitmaps, list) {
list_del_init(&e->list);
ret = io_ctl_read_bitmap(&io_ctl, e);
if (ret)
goto free_cache;
}
io_ctl_drop_pages(&io_ctl);
merge_space_tree(ctl);
ret = 1;
out:
io_ctl_free(&io_ctl);
return ret;
free_cache:
io_ctl_drop_pages(&io_ctl);
__btrfs_remove_free_space_cache(ctl);
goto out;
}
int load_free_space_cache(struct btrfs_fs_info *fs_info,
struct btrfs_block_group_cache *block_group)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
struct btrfs_root *root = fs_info->tree_root;
struct inode *inode;
struct btrfs_path *path;
int ret = 0;
bool matched;
u64 used = btrfs_block_group_used(&block_group->item);
/*
* If this block group has been marked to be cleared for one reason or
* another then we can't trust the on disk cache, so just return.
*/
spin_lock(&block_group->lock);
if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
spin_unlock(&block_group->lock);
return 0;
}
spin_unlock(&block_group->lock);
path = btrfs_alloc_path();
if (!path)
return 0;
path->search_commit_root = 1;
path->skip_locking = 1;
inode = lookup_free_space_inode(root, block_group, path);
if (IS_ERR(inode)) {
btrfs_free_path(path);
return 0;
}
/* We may have converted the inode and made the cache invalid. */
spin_lock(&block_group->lock);
if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
spin_unlock(&block_group->lock);
btrfs_free_path(path);
goto out;
}
spin_unlock(&block_group->lock);
ret = __load_free_space_cache(fs_info->tree_root, inode, ctl,
path, block_group->key.objectid);
btrfs_free_path(path);
if (ret <= 0)
goto out;
spin_lock(&ctl->tree_lock);
matched = (ctl->free_space == (block_group->key.offset - used -
block_group->bytes_super));
spin_unlock(&ctl->tree_lock);
if (!matched) {
__btrfs_remove_free_space_cache(ctl);
btrfs_warn(fs_info, "block group %llu has wrong amount of free space",
block_group->key.objectid);
ret = -1;
}
out:
if (ret < 0) {
/* This cache is bogus, make sure it gets cleared */
spin_lock(&block_group->lock);
block_group->disk_cache_state = BTRFS_DC_CLEAR;
spin_unlock(&block_group->lock);
ret = 0;
btrfs_warn(fs_info, "failed to load free space cache for block group %llu, rebuild it now",
block_group->key.objectid);
}
iput(inode);
return ret;
}
static noinline_for_stack
int write_cache_extent_entries(struct btrfs_io_ctl *io_ctl,
struct btrfs_free_space_ctl *ctl,
struct btrfs_block_group_cache *block_group,
int *entries, int *bitmaps,
struct list_head *bitmap_list)
{
int ret;
struct btrfs_free_cluster *cluster = NULL;
struct btrfs_free_cluster *cluster_locked = NULL;
struct rb_node *node = rb_first(&ctl->free_space_offset);
struct btrfs_trim_range *trim_entry;
/* Get the cluster for this block_group if it exists */
if (block_group && !list_empty(&block_group->cluster_list)) {
cluster = list_entry(block_group->cluster_list.next,
struct btrfs_free_cluster,
block_group_list);
}
if (!node && cluster) {
cluster_locked = cluster;
spin_lock(&cluster_locked->lock);
node = rb_first(&cluster->root);
cluster = NULL;
}
/* Write out the extent entries */
while (node) {
struct btrfs_free_space *e;
e = rb_entry(node, struct btrfs_free_space, offset_index);
*entries += 1;
ret = io_ctl_add_entry(io_ctl, e->offset, e->bytes,
e->bitmap);
if (ret)
goto fail;
if (e->bitmap) {
list_add_tail(&e->list, bitmap_list);
*bitmaps += 1;
}
node = rb_next(node);
if (!node && cluster) {
node = rb_first(&cluster->root);
cluster_locked = cluster;
spin_lock(&cluster_locked->lock);
cluster = NULL;
}
}
if (cluster_locked) {
spin_unlock(&cluster_locked->lock);
cluster_locked = NULL;
}
/*
* Make sure we don't miss any range that was removed from our rbtree
* because trimming is running. Otherwise after a umount+mount (or crash
* after committing the transaction) we would leak free space and get
* an inconsistent free space cache report from fsck.
*/
list_for_each_entry(trim_entry, &ctl->trimming_ranges, list) {
ret = io_ctl_add_entry(io_ctl, trim_entry->start,
trim_entry->bytes, NULL);
if (ret)
goto fail;
*entries += 1;
}
return 0;
fail:
if (cluster_locked)
spin_unlock(&cluster_locked->lock);
return -ENOSPC;
}
static noinline_for_stack int
update_cache_item(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct inode *inode,
struct btrfs_path *path, u64 offset,
int entries, int bitmaps)
{
struct btrfs_key key;
struct btrfs_free_space_header *header;
struct extent_buffer *leaf;
int ret;
key.objectid = BTRFS_FREE_SPACE_OBJECTID;
key.offset = offset;
key.type = 0;
ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
if (ret < 0) {
clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL,
GFP_NOFS);
goto fail;
}
leaf = path->nodes[0];
if (ret > 0) {
struct btrfs_key found_key;
ASSERT(path->slots[0]);
path->slots[0]--;
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
if (found_key.objectid != BTRFS_FREE_SPACE_OBJECTID ||
found_key.offset != offset) {
clear_extent_bit(&BTRFS_I(inode)->io_tree, 0,
inode->i_size - 1,
EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0,
NULL, GFP_NOFS);
btrfs_release_path(path);
goto fail;
}
}
BTRFS_I(inode)->generation = trans->transid;
header = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_free_space_header);
btrfs_set_free_space_entries(leaf, header, entries);
btrfs_set_free_space_bitmaps(leaf, header, bitmaps);
btrfs_set_free_space_generation(leaf, header, trans->transid);
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(path);
return 0;
fail:
return -1;
}
static noinline_for_stack int
write_pinned_extent_entries(struct btrfs_root *root,
struct btrfs_block_group_cache *block_group,
struct btrfs_io_ctl *io_ctl,
int *entries)
{
u64 start, extent_start, extent_end, len;
struct extent_io_tree *unpin = NULL;
int ret;
if (!block_group)
return 0;
/*
* We want to add any pinned extents to our free space cache
* so we don't leak the space
*
* We shouldn't have switched the pinned extents yet so this is the
* right one
*/
unpin = root->fs_info->pinned_extents;
start = block_group->key.objectid;
while (start < block_group->key.objectid + block_group->key.offset) {
ret = find_first_extent_bit(unpin, start,
&extent_start, &extent_end,
EXTENT_DIRTY, NULL);
if (ret)
return 0;
/* This pinned extent is out of our range */
if (extent_start >= block_group->key.objectid +
block_group->key.offset)
return 0;
extent_start = max(extent_start, start);
extent_end = min(block_group->key.objectid +
block_group->key.offset, extent_end + 1);
len = extent_end - extent_start;
*entries += 1;
ret = io_ctl_add_entry(io_ctl, extent_start, len, NULL);
if (ret)
return -ENOSPC;
start = extent_end;
}
return 0;
}
static noinline_for_stack int
write_bitmap_entries(struct btrfs_io_ctl *io_ctl, struct list_head *bitmap_list)
{
struct list_head *pos, *n;
int ret;
/* Write out the bitmaps */
list_for_each_safe(pos, n, bitmap_list) {
struct btrfs_free_space *entry =
list_entry(pos, struct btrfs_free_space, list);
ret = io_ctl_add_bitmap(io_ctl, entry->bitmap);
if (ret)
return -ENOSPC;
list_del_init(&entry->list);
}
return 0;
}
static int flush_dirty_cache(struct inode *inode)
{
int ret;
ret = btrfs_wait_ordered_range(inode, 0, (u64)-1);
if (ret)
clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL,
GFP_NOFS);
return ret;
}
static void noinline_for_stack
cleanup_bitmap_list(struct list_head *bitmap_list)
{
struct list_head *pos, *n;
list_for_each_safe(pos, n, bitmap_list) {
struct btrfs_free_space *entry =
list_entry(pos, struct btrfs_free_space, list);
list_del_init(&entry->list);
}
}
static void noinline_for_stack
cleanup_write_cache_enospc(struct inode *inode,
struct btrfs_io_ctl *io_ctl,
struct extent_state **cached_state,
struct list_head *bitmap_list)
{
io_ctl_drop_pages(io_ctl);
unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
i_size_read(inode) - 1, cached_state,
GFP_NOFS);
}
int btrfs_wait_cache_io(struct btrfs_root *root,
struct btrfs_trans_handle *trans,
struct btrfs_block_group_cache *block_group,
struct btrfs_io_ctl *io_ctl,
struct btrfs_path *path, u64 offset)
{
int ret;
struct inode *inode = io_ctl->inode;
if (!inode)
return 0;
if (block_group)
root = root->fs_info->tree_root;
/* Flush the dirty pages in the cache file. */
ret = flush_dirty_cache(inode);
if (ret)
goto out;
/* Update the cache item to tell everyone this cache file is valid. */
ret = update_cache_item(trans, root, inode, path, offset,
io_ctl->entries, io_ctl->bitmaps);
out:
io_ctl_free(io_ctl);
if (ret) {
invalidate_inode_pages2(inode->i_mapping);
BTRFS_I(inode)->generation = 0;
if (block_group) {
#ifdef DEBUG
btrfs_err(root->fs_info,
"failed to write free space cache for block group %llu",
block_group->key.objectid);
#endif
}
}
btrfs_update_inode(trans, root, inode);
if (block_group) {
/* the dirty list is protected by the dirty_bgs_lock */
spin_lock(&trans->transaction->dirty_bgs_lock);
/* the disk_cache_state is protected by the block group lock */
spin_lock(&block_group->lock);
/*
* only mark this as written if we didn't get put back on
* the dirty list while waiting for IO. Otherwise our
* cache state won't be right, and we won't get written again
*/
if (!ret && list_empty(&block_group->dirty_list))
block_group->disk_cache_state = BTRFS_DC_WRITTEN;
else if (ret)
block_group->disk_cache_state = BTRFS_DC_ERROR;
spin_unlock(&block_group->lock);
spin_unlock(&trans->transaction->dirty_bgs_lock);
io_ctl->inode = NULL;
iput(inode);
}
return ret;
}
/**
* __btrfs_write_out_cache - write out cached info to an inode
* @root - the root the inode belongs to
* @ctl - the free space cache we are going to write out
* @block_group - the block_group for this cache if it belongs to a block_group
* @trans - the trans handle
* @path - the path to use
* @offset - the offset for the key we'll insert
*
* This function writes out a free space cache struct to disk for quick recovery
* on mount. This will return 0 if it was successful in writing the cache out,
* or an errno if it was not.
*/
static int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode,
struct btrfs_free_space_ctl *ctl,
struct btrfs_block_group_cache *block_group,
struct btrfs_io_ctl *io_ctl,
struct btrfs_trans_handle *trans,
struct btrfs_path *path, u64 offset)
{
struct extent_state *cached_state = NULL;
LIST_HEAD(bitmap_list);
int entries = 0;
int bitmaps = 0;
int ret;
int must_iput = 0;
if (!i_size_read(inode))
return -EIO;
WARN_ON(io_ctl->pages);
ret = io_ctl_init(io_ctl, inode, root, 1);
if (ret)
return ret;
if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA)) {
down_write(&block_group->data_rwsem);
spin_lock(&block_group->lock);
if (block_group->delalloc_bytes) {
block_group->disk_cache_state = BTRFS_DC_WRITTEN;
spin_unlock(&block_group->lock);
up_write(&block_group->data_rwsem);
BTRFS_I(inode)->generation = 0;
ret = 0;
must_iput = 1;
goto out;
}
spin_unlock(&block_group->lock);
}
/* Lock all pages first so we can lock the extent safely. */
ret = io_ctl_prepare_pages(io_ctl, inode, 0);
if (ret)
goto out;
lock_extent_bits(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
0, &cached_state);
io_ctl_set_generation(io_ctl, trans->transid);
mutex_lock(&ctl->cache_writeout_mutex);
/* Write out the extent entries in the free space cache */
spin_lock(&ctl->tree_lock);
ret = write_cache_extent_entries(io_ctl, ctl,
block_group, &entries, &bitmaps,
&bitmap_list);
if (ret)
goto out_nospc_locked;
/*
* Some spaces that are freed in the current transaction are pinned,
* they will be added into free space cache after the transaction is
* committed, we shouldn't lose them.
*
* If this changes while we are working we'll get added back to
* the dirty list and redo it. No locking needed
*/
ret = write_pinned_extent_entries(root, block_group, io_ctl, &entries);
if (ret)
goto out_nospc_locked;
/*
* At last, we write out all the bitmaps and keep cache_writeout_mutex
* locked while doing it because a concurrent trim can be manipulating
* or freeing the bitmap.
*/
ret = write_bitmap_entries(io_ctl, &bitmap_list);
spin_unlock(&ctl->tree_lock);
mutex_unlock(&ctl->cache_writeout_mutex);
if (ret)
goto out_nospc;
/* Zero out the rest of the pages just to make sure */
io_ctl_zero_remaining_pages(io_ctl);
/* Everything is written out, now we dirty the pages in the file. */
ret = btrfs_dirty_pages(root, inode, io_ctl->pages, io_ctl->num_pages,
0, i_size_read(inode), &cached_state);
if (ret)
goto out_nospc;
if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
up_write(&block_group->data_rwsem);
/*
* Release the pages and unlock the extent, we will flush
* them out later
*/
io_ctl_drop_pages(io_ctl);
unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
i_size_read(inode) - 1, &cached_state, GFP_NOFS);
/*
* at this point the pages are under IO and we're happy,
* The caller is responsible for waiting on them and updating the
* the cache and the inode
*/
io_ctl->entries = entries;
io_ctl->bitmaps = bitmaps;
ret = btrfs_fdatawrite_range(inode, 0, (u64)-1);
if (ret)
goto out;
return 0;
out:
io_ctl->inode = NULL;
io_ctl_free(io_ctl);
if (ret) {
invalidate_inode_pages2(inode->i_mapping);
BTRFS_I(inode)->generation = 0;
}
btrfs_update_inode(trans, root, inode);
if (must_iput)
iput(inode);
return ret;
out_nospc_locked:
cleanup_bitmap_list(&bitmap_list);
spin_unlock(&ctl->tree_lock);
mutex_unlock(&ctl->cache_writeout_mutex);
out_nospc:
cleanup_write_cache_enospc(inode, io_ctl, &cached_state, &bitmap_list);
if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
up_write(&block_group->data_rwsem);
goto out;
}
int btrfs_write_out_cache(struct btrfs_root *root,
struct btrfs_trans_handle *trans,
struct btrfs_block_group_cache *block_group,
struct btrfs_path *path)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
struct inode *inode;
int ret = 0;
root = root->fs_info->tree_root;
spin_lock(&block_group->lock);
if (block_group->disk_cache_state < BTRFS_DC_SETUP) {
spin_unlock(&block_group->lock);
return 0;
}
spin_unlock(&block_group->lock);
inode = lookup_free_space_inode(root, block_group, path);
if (IS_ERR(inode))
return 0;
ret = __btrfs_write_out_cache(root, inode, ctl, block_group,
&block_group->io_ctl, trans,
path, block_group->key.objectid);
if (ret) {
#ifdef DEBUG
btrfs_err(root->fs_info,
"failed to write free space cache for block group %llu",
block_group->key.objectid);
#endif
spin_lock(&block_group->lock);
block_group->disk_cache_state = BTRFS_DC_ERROR;
spin_unlock(&block_group->lock);
block_group->io_ctl.inode = NULL;
iput(inode);
}
/*
* if ret == 0 the caller is expected to call btrfs_wait_cache_io
* to wait for IO and put the inode
*/
return ret;
}
static inline unsigned long offset_to_bit(u64 bitmap_start, u32 unit,
u64 offset)
{
ASSERT(offset >= bitmap_start);
offset -= bitmap_start;
return (unsigned long)(div_u64(offset, unit));
}
static inline unsigned long bytes_to_bits(u64 bytes, u32 unit)
{
return (unsigned long)(div_u64(bytes, unit));
}
static inline u64 offset_to_bitmap(struct btrfs_free_space_ctl *ctl,
u64 offset)
{
u64 bitmap_start;
u32 bytes_per_bitmap;
bytes_per_bitmap = BITS_PER_BITMAP * ctl->unit;
bitmap_start = offset - ctl->start;
bitmap_start = div_u64(bitmap_start, bytes_per_bitmap);
bitmap_start *= bytes_per_bitmap;
bitmap_start += ctl->start;
return bitmap_start;
}
static int tree_insert_offset(struct rb_root *root, u64 offset,
struct rb_node *node, int bitmap)
{
struct rb_node **p = &root->rb_node;
struct rb_node *parent = NULL;
struct btrfs_free_space *info;
while (*p) {
parent = *p;
info = rb_entry(parent, struct btrfs_free_space, offset_index);
if (offset < info->offset) {
p = &(*p)->rb_left;
} else if (offset > info->offset) {
p = &(*p)->rb_right;
} else {
/*
* we could have a bitmap entry and an extent entry
* share the same offset. If this is the case, we want
* the extent entry to always be found first if we do a
* linear search through the tree, since we want to have
* the quickest allocation time, and allocating from an
* extent is faster than allocating from a bitmap. So
* if we're inserting a bitmap and we find an entry at
* this offset, we want to go right, or after this entry
* logically. If we are inserting an extent and we've
* found a bitmap, we want to go left, or before
* logically.
*/
if (bitmap) {
if (info->bitmap) {
WARN_ON_ONCE(1);
return -EEXIST;
}
p = &(*p)->rb_right;
} else {
if (!info->bitmap) {
WARN_ON_ONCE(1);
return -EEXIST;
}
p = &(*p)->rb_left;
}
}
}
rb_link_node(node, parent, p);
rb_insert_color(node, root);
return 0;
}
/*
* searches the tree for the given offset.
*
* fuzzy - If this is set, then we are trying to make an allocation, and we just
* want a section that has at least bytes size and comes at or after the given
* offset.
*/
static struct btrfs_free_space *
tree_search_offset(struct btrfs_free_space_ctl *ctl,
u64 offset, int bitmap_only, int fuzzy)
{
struct rb_node *n = ctl->free_space_offset.rb_node;
struct btrfs_free_space *entry, *prev = NULL;
/* find entry that is closest to the 'offset' */
while (1) {
if (!n) {
entry = NULL;
break;
}
entry = rb_entry(n, struct btrfs_free_space, offset_index);
prev = entry;
if (offset < entry->offset)
n = n->rb_left;
else if (offset > entry->offset)
n = n->rb_right;
else
break;
}
if (bitmap_only) {
if (!entry)
return NULL;
if (entry->bitmap)
return entry;
/*
* bitmap entry and extent entry may share same offset,
* in that case, bitmap entry comes after extent entry.
*/
n = rb_next(n);
if (!n)
return NULL;
entry = rb_entry(n, struct btrfs_free_space, offset_index);
if (entry->offset != offset)
return NULL;
WARN_ON(!entry->bitmap);
return entry;
} else if (entry) {
if (entry->bitmap) {
/*
* if previous extent entry covers the offset,
* we should return it instead of the bitmap entry
*/
n = rb_prev(&entry->offset_index);
if (n) {
prev = rb_entry(n, struct btrfs_free_space,
offset_index);
if (!prev->bitmap &&
prev->offset + prev->bytes > offset)
entry = prev;
}
}
return entry;
}
if (!prev)
return NULL;
/* find last entry before the 'offset' */
entry = prev;
if (entry->offset > offset) {
n = rb_prev(&entry->offset_index);
if (n) {
entry = rb_entry(n, struct btrfs_free_space,
offset_index);
ASSERT(entry->offset <= offset);
} else {
if (fuzzy)
return entry;
else
return NULL;
}
}
if (entry->bitmap) {
n = rb_prev(&entry->offset_index);
if (n) {
prev = rb_entry(n, struct btrfs_free_space,
offset_index);
if (!prev->bitmap &&
prev->offset + prev->bytes > offset)
return prev;
}
if (entry->offset + BITS_PER_BITMAP * ctl->unit > offset)
return entry;
} else if (entry->offset + entry->bytes > offset)
return entry;
if (!fuzzy)
return NULL;
while (1) {
if (entry->bitmap) {
if (entry->offset + BITS_PER_BITMAP *
ctl->unit > offset)
break;
} else {
if (entry->offset + entry->bytes > offset)
break;
}
n = rb_next(&entry->offset_index);
if (!n)
return NULL;
entry = rb_entry(n, struct btrfs_free_space, offset_index);
}
return entry;
}
static inline void
__unlink_free_space(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info)
{
rb_erase(&info->offset_index, &ctl->free_space_offset);
ctl->free_extents--;
}
static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info)
{
__unlink_free_space(ctl, info);
ctl->free_space -= info->bytes;
}
static int link_free_space(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info)
{
int ret = 0;
ASSERT(info->bytes || info->bitmap);
ret = tree_insert_offset(&ctl->free_space_offset, info->offset,
&info->offset_index, (info->bitmap != NULL));
if (ret)
return ret;
ctl->free_space += info->bytes;
ctl->free_extents++;
return ret;
}
static void recalculate_thresholds(struct btrfs_free_space_ctl *ctl)
{
struct btrfs_block_group_cache *block_group = ctl->private;
u64 max_bytes;
u64 bitmap_bytes;
u64 extent_bytes;
u64 size = block_group->key.offset;
u32 bytes_per_bg = BITS_PER_BITMAP * ctl->unit;
u32 max_bitmaps = div_u64(size + bytes_per_bg - 1, bytes_per_bg);
max_bitmaps = max_t(u32, max_bitmaps, 1);
ASSERT(ctl->total_bitmaps <= max_bitmaps);
/*
* The goal is to keep the total amount of memory used per 1gb of space
* at or below 32k, so we need to adjust how much memory we allow to be
* used by extent based free space tracking
*/
if (size < 1024 * 1024 * 1024)
max_bytes = MAX_CACHE_BYTES_PER_GIG;
else
max_bytes = MAX_CACHE_BYTES_PER_GIG *
div_u64(size, 1024 * 1024 * 1024);
/*
* we want to account for 1 more bitmap than what we have so we can make
* sure we don't go over our overall goal of MAX_CACHE_BYTES_PER_GIG as
* we add more bitmaps.
*/
bitmap_bytes = (ctl->total_bitmaps + 1) * PAGE_CACHE_SIZE;
if (bitmap_bytes >= max_bytes) {
ctl->extents_thresh = 0;
return;
}
/*
* we want the extent entry threshold to always be at most 1/2 the max
* bytes we can have, or whatever is less than that.
*/
extent_bytes = max_bytes - bitmap_bytes;
extent_bytes = min_t(u64, extent_bytes, max_bytes >> 1);
ctl->extents_thresh =
div_u64(extent_bytes, sizeof(struct btrfs_free_space));
}
static inline void __bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info,
u64 offset, u64 bytes)
{
unsigned long start, count;
start = offset_to_bit(info->offset, ctl->unit, offset);
count = bytes_to_bits(bytes, ctl->unit);
ASSERT(start + count <= BITS_PER_BITMAP);
bitmap_clear(info->bitmap, start, count);
info->bytes -= bytes;
}
static void bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info, u64 offset,
u64 bytes)
{
__bitmap_clear_bits(ctl, info, offset, bytes);
ctl->free_space -= bytes;
}
static void bitmap_set_bits(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info, u64 offset,
u64 bytes)
{
unsigned long start, count;
start = offset_to_bit(info->offset, ctl->unit, offset);
count = bytes_to_bits(bytes, ctl->unit);
ASSERT(start + count <= BITS_PER_BITMAP);
bitmap_set(info->bitmap, start, count);
info->bytes += bytes;
ctl->free_space += bytes;
}
/*
* If we can not find suitable extent, we will use bytes to record
* the size of the max extent.
*/
static int search_bitmap(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *bitmap_info, u64 *offset,
u64 *bytes, bool for_alloc)
{
unsigned long found_bits = 0;
unsigned long max_bits = 0;
unsigned long bits, i;
unsigned long next_zero;
unsigned long extent_bits;
/*
* Skip searching the bitmap if we don't have a contiguous section that
* is large enough for this allocation.
*/
if (for_alloc &&
bitmap_info->max_extent_size &&
bitmap_info->max_extent_size < *bytes) {
*bytes = bitmap_info->max_extent_size;
return -1;
}
i = offset_to_bit(bitmap_info->offset, ctl->unit,
max_t(u64, *offset, bitmap_info->offset));
bits = bytes_to_bits(*bytes, ctl->unit);
for_each_set_bit_from(i, bitmap_info->bitmap, BITS_PER_BITMAP) {
if (for_alloc && bits == 1) {
found_bits = 1;
break;
}
next_zero = find_next_zero_bit(bitmap_info->bitmap,
BITS_PER_BITMAP, i);
extent_bits = next_zero - i;
if (extent_bits >= bits) {
found_bits = extent_bits;
break;
} else if (extent_bits > max_bits) {
max_bits = extent_bits;
}
i = next_zero;
}
if (found_bits) {
*offset = (u64)(i * ctl->unit) + bitmap_info->offset;
*bytes = (u64)(found_bits) * ctl->unit;
return 0;
}
*bytes = (u64)(max_bits) * ctl->unit;
bitmap_info->max_extent_size = *bytes;
return -1;
}
/* Cache the size of the max extent in bytes */
static struct btrfs_free_space *
find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes,
unsigned long align, u64 *max_extent_size)
{
struct btrfs_free_space *entry;
struct rb_node *node;
u64 tmp;
u64 align_off;
int ret;
if (!ctl->free_space_offset.rb_node)
goto out;
entry = tree_search_offset(ctl, offset_to_bitmap(ctl, *offset), 0, 1);
if (!entry)
goto out;
for (node = &entry->offset_index; node; node = rb_next(node)) {
entry = rb_entry(node, struct btrfs_free_space, offset_index);
if (entry->bytes < *bytes) {
if (entry->bytes > *max_extent_size)
*max_extent_size = entry->bytes;
continue;
}
/* make sure the space returned is big enough
* to match our requested alignment
*/
if (*bytes >= align) {
tmp = entry->offset - ctl->start + align - 1;
tmp = div64_u64(tmp, align);
tmp = tmp * align + ctl->start;
align_off = tmp - entry->offset;
} else {
align_off = 0;
tmp = entry->offset;
}
if (entry->bytes < *bytes + align_off) {
if (entry->bytes > *max_extent_size)
*max_extent_size = entry->bytes;
continue;
}
if (entry->bitmap) {
u64 size = *bytes;
ret = search_bitmap(ctl, entry, &tmp, &size, true);
if (!ret) {
*offset = tmp;
*bytes = size;
return entry;
} else if (size > *max_extent_size) {
*max_extent_size = size;
}
continue;
}
*offset = tmp;
*bytes = entry->bytes - align_off;
return entry;
}
out:
return NULL;
}
static void add_new_bitmap(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info, u64 offset)
{
info->offset = offset_to_bitmap(ctl, offset);
info->bytes = 0;
INIT_LIST_HEAD(&info->list);
link_free_space(ctl, info);
ctl->total_bitmaps++;
ctl->op->recalc_thresholds(ctl);
}
static void free_bitmap(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *bitmap_info)
{
unlink_free_space(ctl, bitmap_info);
kfree(bitmap_info->bitmap);
kmem_cache_free(btrfs_free_space_cachep, bitmap_info);
ctl->total_bitmaps--;
ctl->op->recalc_thresholds(ctl);
}
static noinline int remove_from_bitmap(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *bitmap_info,
u64 *offset, u64 *bytes)
{
u64 end;
u64 search_start, search_bytes;
int ret;
again:
end = bitmap_info->offset + (u64)(BITS_PER_BITMAP * ctl->unit) - 1;
/*
* We need to search for bits in this bitmap. We could only cover some
* of the extent in this bitmap thanks to how we add space, so we need
* to search for as much as it as we can and clear that amount, and then
* go searching for the next bit.
*/
search_start = *offset;
search_bytes = ctl->unit;
search_bytes = min(search_bytes, end - search_start + 1);
ret = search_bitmap(ctl, bitmap_info, &search_start, &search_bytes,
false);
if (ret < 0 || search_start != *offset)
return -EINVAL;
/* We may have found more bits than what we need */
search_bytes = min(search_bytes, *bytes);
/* Cannot clear past the end of the bitmap */
search_bytes = min(search_bytes, end - search_start + 1);
bitmap_clear_bits(ctl, bitmap_info, search_start, search_bytes);
*offset += search_bytes;
*bytes -= search_bytes;
if (*bytes) {
struct rb_node *next = rb_next(&bitmap_info->offset_index);
if (!bitmap_info->bytes)
free_bitmap(ctl, bitmap_info);
/*
* no entry after this bitmap, but we still have bytes to
* remove, so something has gone wrong.
*/
if (!next)
return -EINVAL;
bitmap_info = rb_entry(next, struct btrfs_free_space,
offset_index);
/*
* if the next entry isn't a bitmap we need to return to let the
* extent stuff do its work.
*/
if (!bitmap_info->bitmap)
return -EAGAIN;
/*
* Ok the next item is a bitmap, but it may not actually hold
* the information for the rest of this free space stuff, so
* look for it, and if we don't find it return so we can try
* everything over again.
*/
search_start = *offset;
search_bytes = ctl->unit;
ret = search_bitmap(ctl, bitmap_info, &search_start,
&search_bytes, false);
if (ret < 0 || search_start != *offset)
return -EAGAIN;
goto again;
} else if (!bitmap_info->bytes)
free_bitmap(ctl, bitmap_info);
return 0;
}
static u64 add_bytes_to_bitmap(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info, u64 offset,
u64 bytes)
{
u64 bytes_to_set = 0;
u64 end;
end = info->offset + (u64)(BITS_PER_BITMAP * ctl->unit);
bytes_to_set = min(end - offset, bytes);
bitmap_set_bits(ctl, info, offset, bytes_to_set);
/*
* We set some bytes, we have no idea what the max extent size is
* anymore.
*/
info->max_extent_size = 0;
return bytes_to_set;
}
static bool use_bitmap(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info)
{
struct btrfs_block_group_cache *block_group = ctl->private;
bool forced = false;
#ifdef CONFIG_BTRFS_DEBUG
if (btrfs_should_fragment_free_space(block_group->fs_info->extent_root,
block_group))
forced = true;
#endif
/*
* If we are below the extents threshold then we can add this as an
* extent, and don't have to deal with the bitmap
*/
if (!forced && ctl->free_extents < ctl->extents_thresh) {
/*
* If this block group has some small extents we don't want to
* use up all of our free slots in the cache with them, we want
* to reserve them to larger extents, however if we have plent
* of cache left then go ahead an dadd them, no sense in adding
* the overhead of a bitmap if we don't have to.
*/
if (info->bytes <= block_group->sectorsize * 4) {
if (ctl->free_extents * 2 <= ctl->extents_thresh)
return false;
} else {
return false;
}
}
/*
* The original block groups from mkfs can be really small, like 8
* megabytes, so don't bother with a bitmap for those entries. However
* some block groups can be smaller than what a bitmap would cover but
* are still large enough that they could overflow the 32k memory limit,
* so allow those block groups to still be allowed to have a bitmap
* entry.
*/
if (((BITS_PER_BITMAP * ctl->unit) >> 1) > block_group->key.offset)
return false;
return true;
}
static struct btrfs_free_space_op free_space_op = {
.recalc_thresholds = recalculate_thresholds,
.use_bitmap = use_bitmap,
};
static int insert_into_bitmap(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info)
{
struct btrfs_free_space *bitmap_info;
struct btrfs_block_group_cache *block_group = NULL;
int added = 0;
u64 bytes, offset, bytes_added;
int ret;
bytes = info->bytes;
offset = info->offset;
if (!ctl->op->use_bitmap(ctl, info))
return 0;
if (ctl->op == &free_space_op)
block_group = ctl->private;
again:
/*
* Since we link bitmaps right into the cluster we need to see if we
* have a cluster here, and if so and it has our bitmap we need to add
* the free space to that bitmap.
*/
if (block_group && !list_empty(&block_group->cluster_list)) {
struct btrfs_free_cluster *cluster;
struct rb_node *node;
struct btrfs_free_space *entry;
cluster = list_entry(block_group->cluster_list.next,
struct btrfs_free_cluster,
block_group_list);
spin_lock(&cluster->lock);
node = rb_first(&cluster->root);
if (!node) {
spin_unlock(&cluster->lock);
goto no_cluster_bitmap;
}
entry = rb_entry(node, struct btrfs_free_space, offset_index);
if (!entry->bitmap) {
spin_unlock(&cluster->lock);
goto no_cluster_bitmap;
}
if (entry->offset == offset_to_bitmap(ctl, offset)) {
bytes_added = add_bytes_to_bitmap(ctl, entry,
offset, bytes);
bytes -= bytes_added;
offset += bytes_added;
}
spin_unlock(&cluster->lock);
if (!bytes) {
ret = 1;
goto out;
}
}
no_cluster_bitmap:
bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
1, 0);
if (!bitmap_info) {
ASSERT(added == 0);
goto new_bitmap;
}
bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes);
bytes -= bytes_added;
offset += bytes_added;
added = 0;
if (!bytes) {
ret = 1;
goto out;
} else
goto again;
new_bitmap:
if (info && info->bitmap) {
add_new_bitmap(ctl, info, offset);
added = 1;
info = NULL;
goto again;
} else {
spin_unlock(&ctl->tree_lock);
/* no pre-allocated info, allocate a new one */
if (!info) {
info = kmem_cache_zalloc(btrfs_free_space_cachep,
GFP_NOFS);
if (!info) {
spin_lock(&ctl->tree_lock);
ret = -ENOMEM;
goto out;
}
}
/* allocate the bitmap */
info->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
spin_lock(&ctl->tree_lock);
if (!info->bitmap) {
ret = -ENOMEM;
goto out;
}
goto again;
}
out:
if (info) {
if (info->bitmap)
kfree(info->bitmap);
kmem_cache_free(btrfs_free_space_cachep, info);
}
return ret;
}
static bool try_merge_free_space(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info, bool update_stat)
{
struct btrfs_free_space *left_info;
struct btrfs_free_space *right_info;
bool merged = false;
u64 offset = info->offset;
u64 bytes = info->bytes;
/*
* first we want to see if there is free space adjacent to the range we
* are adding, if there is remove that struct and add a new one to
* cover the entire range
*/
right_info = tree_search_offset(ctl, offset + bytes, 0, 0);
if (right_info && rb_prev(&right_info->offset_index))
left_info = rb_entry(rb_prev(&right_info->offset_index),
struct btrfs_free_space, offset_index);
else
left_info = tree_search_offset(ctl, offset - 1, 0, 0);
if (right_info && !right_info->bitmap) {
if (update_stat)
unlink_free_space(ctl, right_info);
else
__unlink_free_space(ctl, right_info);
info->bytes += right_info->bytes;
kmem_cache_free(btrfs_free_space_cachep, right_info);
merged = true;
}
if (left_info && !left_info->bitmap &&
left_info->offset + left_info->bytes == offset) {
if (update_stat)
unlink_free_space(ctl, left_info);
else
__unlink_free_space(ctl, left_info);
info->offset = left_info->offset;
info->bytes += left_info->bytes;
kmem_cache_free(btrfs_free_space_cachep, left_info);
merged = true;
}
return merged;
}
static bool steal_from_bitmap_to_end(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info,
bool update_stat)
{
struct btrfs_free_space *bitmap;
unsigned long i;
unsigned long j;
const u64 end = info->offset + info->bytes;
const u64 bitmap_offset = offset_to_bitmap(ctl, end);
u64 bytes;
bitmap = tree_search_offset(ctl, bitmap_offset, 1, 0);
if (!bitmap)
return false;
i = offset_to_bit(bitmap->offset, ctl->unit, end);
j = find_next_zero_bit(bitmap->bitmap, BITS_PER_BITMAP, i);
if (j == i)
return false;
bytes = (j - i) * ctl->unit;
info->bytes += bytes;
if (update_stat)
bitmap_clear_bits(ctl, bitmap, end, bytes);
else
__bitmap_clear_bits(ctl, bitmap, end, bytes);
if (!bitmap->bytes)
free_bitmap(ctl, bitmap);
return true;
}
static bool steal_from_bitmap_to_front(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info,
bool update_stat)
{
struct btrfs_free_space *bitmap;
u64 bitmap_offset;
unsigned long i;
unsigned long j;
unsigned long prev_j;
u64 bytes;
bitmap_offset = offset_to_bitmap(ctl, info->offset);
/* If we're on a boundary, try the previous logical bitmap. */
if (bitmap_offset == info->offset) {
if (info->offset == 0)
return false;
bitmap_offset = offset_to_bitmap(ctl, info->offset - 1);
}
bitmap = tree_search_offset(ctl, bitmap_offset, 1, 0);
if (!bitmap)
return false;
i = offset_to_bit(bitmap->offset, ctl->unit, info->offset) - 1;
j = 0;
prev_j = (unsigned long)-1;
for_each_clear_bit_from(j, bitmap->bitmap, BITS_PER_BITMAP) {
if (j > i)
break;
prev_j = j;
}
if (prev_j == i)
return false;
if (prev_j == (unsigned long)-1)
bytes = (i + 1) * ctl->unit;
else
bytes = (i - prev_j) * ctl->unit;
info->offset -= bytes;
info->bytes += bytes;
if (update_stat)
bitmap_clear_bits(ctl, bitmap, info->offset, bytes);
else
__bitmap_clear_bits(ctl, bitmap, info->offset, bytes);
if (!bitmap->bytes)
free_bitmap(ctl, bitmap);
return true;
}
/*
* We prefer always to allocate from extent entries, both for clustered and
* non-clustered allocation requests. So when attempting to add a new extent
* entry, try to see if there's adjacent free space in bitmap entries, and if
* there is, migrate that space from the bitmaps to the extent.
* Like this we get better chances of satisfying space allocation requests
* because we attempt to satisfy them based on a single cache entry, and never
* on 2 or more entries - even if the entries represent a contiguous free space
* region (e.g. 1 extent entry + 1 bitmap entry starting where the extent entry
* ends).
*/
static void steal_from_bitmap(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *info,
bool update_stat)
{
/*
* Only work with disconnected entries, as we can change their offset,
* and must be extent entries.
*/
ASSERT(!info->bitmap);
ASSERT(RB_EMPTY_NODE(&info->offset_index));
if (ctl->total_bitmaps > 0) {
bool stole_end;
bool stole_front = false;
stole_end = steal_from_bitmap_to_end(ctl, info, update_stat);
if (ctl->total_bitmaps > 0)
stole_front = steal_from_bitmap_to_front(ctl, info,
update_stat);
if (stole_end || stole_front)
try_merge_free_space(ctl, info, update_stat);
}
}
int __btrfs_add_free_space(struct btrfs_free_space_ctl *ctl,
u64 offset, u64 bytes)
{
struct btrfs_free_space *info;
int ret = 0;
info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
if (!info)
return -ENOMEM;
info->offset = offset;
info->bytes = bytes;
RB_CLEAR_NODE(&info->offset_index);
spin_lock(&ctl->tree_lock);
if (try_merge_free_space(ctl, info, true))
goto link;
/*
* There was no extent directly to the left or right of this new
* extent then we know we're going to have to allocate a new extent, so
* before we do that see if we need to drop this into a bitmap
*/
ret = insert_into_bitmap(ctl, info);
if (ret < 0) {
goto out;
} else if (ret) {
ret = 0;
goto out;
}
link:
/*
* Only steal free space from adjacent bitmaps if we're sure we're not
* going to add the new free space to existing bitmap entries - because
* that would mean unnecessary work that would be reverted. Therefore
* attempt to steal space from bitmaps if we're adding an extent entry.
*/
steal_from_bitmap(ctl, info, true);
ret = link_free_space(ctl, info);
if (ret)
kmem_cache_free(btrfs_free_space_cachep, info);
out:
spin_unlock(&ctl->tree_lock);
if (ret) {
printk(KERN_CRIT "BTRFS: unable to add free space :%d\n", ret);
ASSERT(ret != -EEXIST);
}
return ret;
}
int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
u64 offset, u64 bytes)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
struct btrfs_free_space *info;
int ret;
bool re_search = false;
spin_lock(&ctl->tree_lock);
again:
ret = 0;
if (!bytes)
goto out_lock;
info = tree_search_offset(ctl, offset, 0, 0);
if (!info) {
/*
* oops didn't find an extent that matched the space we wanted
* to remove, look for a bitmap instead
*/
info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
1, 0);
if (!info) {
/*
* If we found a partial bit of our free space in a
* bitmap but then couldn't find the other part this may
* be a problem, so WARN about it.
*/
WARN_ON(re_search);
goto out_lock;
}
}
re_search = false;
if (!info->bitmap) {
unlink_free_space(ctl, info);
if (offset == info->offset) {
u64 to_free = min(bytes, info->bytes);
info->bytes -= to_free;
info->offset += to_free;
if (info->bytes) {
ret = link_free_space(ctl, info);
WARN_ON(ret);
} else {
kmem_cache_free(btrfs_free_space_cachep, info);
}
offset += to_free;
bytes -= to_free;
goto again;
} else {
u64 old_end = info->bytes + info->offset;
info->bytes = offset - info->offset;
ret = link_free_space(ctl, info);
WARN_ON(ret);
if (ret)
goto out_lock;
/* Not enough bytes in this entry to satisfy us */
if (old_end < offset + bytes) {
bytes -= old_end - offset;
offset = old_end;
goto again;
} else if (old_end == offset + bytes) {
/* all done */
goto out_lock;
}
spin_unlock(&ctl->tree_lock);
ret = btrfs_add_free_space(block_group, offset + bytes,
old_end - (offset + bytes));
WARN_ON(ret);
goto out;
}
}
ret = remove_from_bitmap(ctl, info, &offset, &bytes);
if (ret == -EAGAIN) {
re_search = true;
goto again;
}
out_lock:
spin_unlock(&ctl->tree_lock);
out:
return ret;
}
void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
u64 bytes)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
struct btrfs_free_space *info;
struct rb_node *n;
int count = 0;
for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
info = rb_entry(n, struct btrfs_free_space, offset_index);
if (info->bytes >= bytes && !block_group->ro)
count++;
btrfs_crit(block_group->fs_info,
"entry offset %llu, bytes %llu, bitmap %s",
info->offset, info->bytes,
(info->bitmap) ? "yes" : "no");
}
btrfs_info(block_group->fs_info, "block group has cluster?: %s",
list_empty(&block_group->cluster_list) ? "no" : "yes");
btrfs_info(block_group->fs_info,
"%d blocks of free space at or bigger than bytes is", count);
}
void btrfs_init_free_space_ctl(struct btrfs_block_group_cache *block_group)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
spin_lock_init(&ctl->tree_lock);
ctl->unit = block_group->sectorsize;
ctl->start = block_group->key.objectid;
ctl->private = block_group;
ctl->op = &free_space_op;
INIT_LIST_HEAD(&ctl->trimming_ranges);
mutex_init(&ctl->cache_writeout_mutex);
/*
* we only want to have 32k of ram per block group for keeping
* track of free space, and if we pass 1/2 of that we want to
* start converting things over to using bitmaps
*/
ctl->extents_thresh = ((1024 * 32) / 2) /
sizeof(struct btrfs_free_space);
}
/*
* for a given cluster, put all of its extents back into the free
* space cache. If the block group passed doesn't match the block group
* pointed to by the cluster, someone else raced in and freed the
* cluster already. In that case, we just return without changing anything
*/
static int
__btrfs_return_cluster_to_free_space(
struct btrfs_block_group_cache *block_group,
struct btrfs_free_cluster *cluster)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
struct btrfs_free_space *entry;
struct rb_node *node;
spin_lock(&cluster->lock);
if (cluster->block_group != block_group)
goto out;
cluster->block_group = NULL;
cluster->window_start = 0;
list_del_init(&cluster->block_group_list);
node = rb_first(&cluster->root);
while (node) {
bool bitmap;
entry = rb_entry(node, struct btrfs_free_space, offset_index);
node = rb_next(&entry->offset_index);
rb_erase(&entry->offset_index, &cluster->root);
RB_CLEAR_NODE(&entry->offset_index);
bitmap = (entry->bitmap != NULL);
if (!bitmap) {
try_merge_free_space(ctl, entry, false);
steal_from_bitmap(ctl, entry, false);
}
tree_insert_offset(&ctl->free_space_offset,
entry->offset, &entry->offset_index, bitmap);
}
cluster->root = RB_ROOT;
out:
spin_unlock(&cluster->lock);
btrfs_put_block_group(block_group);
return 0;
}
static void __btrfs_remove_free_space_cache_locked(
struct btrfs_free_space_ctl *ctl)
{
struct btrfs_free_space *info;
struct rb_node *node;
while ((node = rb_last(&ctl->free_space_offset)) != NULL) {
info = rb_entry(node, struct btrfs_free_space, offset_index);
if (!info->bitmap) {
unlink_free_space(ctl, info);
kmem_cache_free(btrfs_free_space_cachep, info);
} else {
free_bitmap(ctl, info);
}
cond_resched_lock(&ctl->tree_lock);
}
}
void __btrfs_remove_free_space_cache(struct btrfs_free_space_ctl *ctl)
{
spin_lock(&ctl->tree_lock);
__btrfs_remove_free_space_cache_locked(ctl);
spin_unlock(&ctl->tree_lock);
}
void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
struct btrfs_free_cluster *cluster;
struct list_head *head;
spin_lock(&ctl->tree_lock);
while ((head = block_group->cluster_list.next) !=
&block_group->cluster_list) {
cluster = list_entry(head, struct btrfs_free_cluster,
block_group_list);
WARN_ON(cluster->block_group != block_group);
__btrfs_return_cluster_to_free_space(block_group, cluster);
cond_resched_lock(&ctl->tree_lock);
}
__btrfs_remove_free_space_cache_locked(ctl);
spin_unlock(&ctl->tree_lock);
}
u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
u64 offset, u64 bytes, u64 empty_size,
u64 *max_extent_size)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
struct btrfs_free_space *entry = NULL;
u64 bytes_search = bytes + empty_size;
u64 ret = 0;
u64 align_gap = 0;
u64 align_gap_len = 0;
spin_lock(&ctl->tree_lock);
entry = find_free_space(ctl, &offset, &bytes_search,
block_group->full_stripe_len, max_extent_size);
if (!entry)
goto out;
ret = offset;
if (entry->bitmap) {
bitmap_clear_bits(ctl, entry, offset, bytes);
if (!entry->bytes)
free_bitmap(ctl, entry);
} else {
unlink_free_space(ctl, entry);
align_gap_len = offset - entry->offset;
align_gap = entry->offset;
entry->offset = offset + bytes;
WARN_ON(entry->bytes < bytes + align_gap_len);
entry->bytes -= bytes + align_gap_len;
if (!entry->bytes)
kmem_cache_free(btrfs_free_space_cachep, entry);
else
link_free_space(ctl, entry);
}
out:
spin_unlock(&ctl->tree_lock);
if (align_gap_len)
__btrfs_add_free_space(ctl, align_gap, align_gap_len);
return ret;
}
/*
* given a cluster, put all of its extents back into the free space
* cache. If a block group is passed, this function will only free
* a cluster that belongs to the passed block group.
*
* Otherwise, it'll get a reference on the block group pointed to by the
* cluster and remove the cluster from it.
*/
int btrfs_return_cluster_to_free_space(
struct btrfs_block_group_cache *block_group,
struct btrfs_free_cluster *cluster)
{
struct btrfs_free_space_ctl *ctl;
int ret;
/* first, get a safe pointer to the block group */
spin_lock(&cluster->lock);
if (!block_group) {
block_group = cluster->block_group;
if (!block_group) {
spin_unlock(&cluster->lock);
return 0;
}
} else if (cluster->block_group != block_group) {
/* someone else has already freed it don't redo their work */
spin_unlock(&cluster->lock);
return 0;
}
atomic_inc(&block_group->count);
spin_unlock(&cluster->lock);
ctl = block_group->free_space_ctl;
/* now return any extents the cluster had on it */
spin_lock(&ctl->tree_lock);
ret = __btrfs_return_cluster_to_free_space(block_group, cluster);
spin_unlock(&ctl->tree_lock);
/* finally drop our ref */
btrfs_put_block_group(block_group);
return ret;
}
static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
struct btrfs_free_cluster *cluster,
struct btrfs_free_space *entry,
u64 bytes, u64 min_start,
u64 *max_extent_size)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
int err;
u64 search_start = cluster->window_start;
u64 search_bytes = bytes;
u64 ret = 0;
search_start = min_start;
search_bytes = bytes;
err = search_bitmap(ctl, entry, &search_start, &search_bytes, true);
if (err) {
if (search_bytes > *max_extent_size)
*max_extent_size = search_bytes;
return 0;
}
ret = search_start;
__bitmap_clear_bits(ctl, entry, ret, bytes);
return ret;
}
/*
* given a cluster, try to allocate 'bytes' from it, returns 0
* if it couldn't find anything suitably large, or a logical disk offset
* if things worked out
*/
u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
struct btrfs_free_cluster *cluster, u64 bytes,
u64 min_start, u64 *max_extent_size)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
struct btrfs_free_space *entry = NULL;
struct rb_node *node;
u64 ret = 0;
spin_lock(&cluster->lock);
if (bytes > cluster->max_size)
goto out;
if (cluster->block_group != block_group)
goto out;
node = rb_first(&cluster->root);
if (!node)
goto out;
entry = rb_entry(node, struct btrfs_free_space, offset_index);
while (1) {
if (entry->bytes < bytes && entry->bytes > *max_extent_size)
*max_extent_size = entry->bytes;
if (entry->bytes < bytes ||
(!entry->bitmap && entry->offset < min_start)) {
node = rb_next(&entry->offset_index);
if (!node)
break;
entry = rb_entry(node, struct btrfs_free_space,
offset_index);
continue;
}
if (entry->bitmap) {
ret = btrfs_alloc_from_bitmap(block_group,
cluster, entry, bytes,
cluster->window_start,
max_extent_size);
if (ret == 0) {
node = rb_next(&entry->offset_index);
if (!node)
break;
entry = rb_entry(node, struct btrfs_free_space,
offset_index);
continue;
}
cluster->window_start += bytes;
} else {
ret = entry->offset;
entry->offset += bytes;
entry->bytes -= bytes;
}
if (entry->bytes == 0)
rb_erase(&entry->offset_index, &cluster->root);
break;
}
out:
spin_unlock(&cluster->lock);
if (!ret)
return 0;
spin_lock(&ctl->tree_lock);
ctl->free_space -= bytes;
if (entry->bytes == 0) {
ctl->free_extents--;
if (entry->bitmap) {