2018-09-12 09:16:07 +08:00
|
|
|
// SPDX-License-Identifier: GPL-2.0
|
2015-07-08 17:59:36 +08:00
|
|
|
/*
|
|
|
|
* f2fs extent cache support
|
|
|
|
*
|
|
|
|
* Copyright (c) 2015 Motorola Mobility
|
|
|
|
* Copyright (c) 2015 Samsung Electronics
|
|
|
|
* Authors: Jaegeuk Kim <jaegeuk@kernel.org>
|
|
|
|
* Chao Yu <chao2.yu@samsung.com>
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/f2fs_fs.h>
|
|
|
|
|
|
|
|
#include "f2fs.h"
|
|
|
|
#include "node.h"
|
|
|
|
#include <trace/events/f2fs.h>
|
|
|
|
|
2017-04-11 09:25:22 +08:00
|
|
|
static struct rb_entry *__lookup_rb_tree_fast(struct rb_entry *cached_re,
|
|
|
|
unsigned int ofs)
|
|
|
|
{
|
|
|
|
if (cached_re) {
|
|
|
|
if (cached_re->ofs <= ofs &&
|
|
|
|
cached_re->ofs + cached_re->len > ofs) {
|
|
|
|
return cached_re;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2018-10-04 11:18:30 +08:00
|
|
|
static struct rb_entry *__lookup_rb_tree_slow(struct rb_root_cached *root,
|
2017-04-11 09:25:22 +08:00
|
|
|
unsigned int ofs)
|
|
|
|
{
|
2018-10-04 11:18:30 +08:00
|
|
|
struct rb_node *node = root->rb_root.rb_node;
|
2017-04-11 09:25:22 +08:00
|
|
|
struct rb_entry *re;
|
|
|
|
|
|
|
|
while (node) {
|
|
|
|
re = rb_entry(node, struct rb_entry, rb_node);
|
|
|
|
|
|
|
|
if (ofs < re->ofs)
|
|
|
|
node = node->rb_left;
|
|
|
|
else if (ofs >= re->ofs + re->len)
|
|
|
|
node = node->rb_right;
|
|
|
|
else
|
|
|
|
return re;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2018-10-04 11:18:30 +08:00
|
|
|
struct rb_entry *f2fs_lookup_rb_tree(struct rb_root_cached *root,
|
2017-04-11 09:25:22 +08:00
|
|
|
struct rb_entry *cached_re, unsigned int ofs)
|
|
|
|
{
|
|
|
|
struct rb_entry *re;
|
|
|
|
|
|
|
|
re = __lookup_rb_tree_fast(cached_re, ofs);
|
|
|
|
if (!re)
|
|
|
|
return __lookup_rb_tree_slow(root, ofs);
|
|
|
|
|
|
|
|
return re;
|
|
|
|
}
|
|
|
|
|
2020-08-04 21:14:48 +08:00
|
|
|
struct rb_node **f2fs_lookup_rb_tree_ext(struct f2fs_sb_info *sbi,
|
|
|
|
struct rb_root_cached *root,
|
|
|
|
struct rb_node **parent,
|
|
|
|
unsigned long long key, bool *leftmost)
|
|
|
|
{
|
|
|
|
struct rb_node **p = &root->rb_root.rb_node;
|
|
|
|
struct rb_entry *re;
|
|
|
|
|
|
|
|
while (*p) {
|
|
|
|
*parent = *p;
|
|
|
|
re = rb_entry(*parent, struct rb_entry, rb_node);
|
|
|
|
|
|
|
|
if (key < re->key) {
|
|
|
|
p = &(*p)->rb_left;
|
|
|
|
} else {
|
|
|
|
p = &(*p)->rb_right;
|
|
|
|
*leftmost = false;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return p;
|
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-30 00:20:41 +08:00
|
|
|
struct rb_node **f2fs_lookup_rb_tree_for_insert(struct f2fs_sb_info *sbi,
|
2018-10-04 11:18:30 +08:00
|
|
|
struct rb_root_cached *root,
|
|
|
|
struct rb_node **parent,
|
|
|
|
unsigned int ofs, bool *leftmost)
|
2017-04-11 09:25:22 +08:00
|
|
|
{
|
2018-10-04 11:18:30 +08:00
|
|
|
struct rb_node **p = &root->rb_root.rb_node;
|
2017-04-11 09:25:22 +08:00
|
|
|
struct rb_entry *re;
|
|
|
|
|
|
|
|
while (*p) {
|
|
|
|
*parent = *p;
|
|
|
|
re = rb_entry(*parent, struct rb_entry, rb_node);
|
|
|
|
|
2018-10-04 11:18:30 +08:00
|
|
|
if (ofs < re->ofs) {
|
2017-04-11 09:25:22 +08:00
|
|
|
p = &(*p)->rb_left;
|
2018-10-04 11:18:30 +08:00
|
|
|
} else if (ofs >= re->ofs + re->len) {
|
2017-04-11 09:25:22 +08:00
|
|
|
p = &(*p)->rb_right;
|
2018-10-04 11:18:30 +08:00
|
|
|
*leftmost = false;
|
|
|
|
} else {
|
2017-04-11 09:25:22 +08:00
|
|
|
f2fs_bug_on(sbi, 1);
|
2018-10-04 11:18:30 +08:00
|
|
|
}
|
2017-04-11 09:25:22 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
return p;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* lookup rb entry in position of @ofs in rb-tree,
|
|
|
|
* if hit, return the entry, otherwise, return NULL
|
|
|
|
* @prev_ex: extent before ofs
|
|
|
|
* @next_ex: extent after ofs
|
|
|
|
* @insert_p: insert point for new extent at ofs
|
|
|
|
* in order to simpfy the insertion after.
|
|
|
|
* tree must stay unchanged between lookup and insertion.
|
|
|
|
*/
|
2018-10-04 11:18:30 +08:00
|
|
|
struct rb_entry *f2fs_lookup_rb_tree_ret(struct rb_root_cached *root,
|
2017-04-11 09:25:22 +08:00
|
|
|
struct rb_entry *cached_re,
|
|
|
|
unsigned int ofs,
|
|
|
|
struct rb_entry **prev_entry,
|
|
|
|
struct rb_entry **next_entry,
|
|
|
|
struct rb_node ***insert_p,
|
2017-04-14 23:24:55 +08:00
|
|
|
struct rb_node **insert_parent,
|
2018-10-04 11:18:30 +08:00
|
|
|
bool force, bool *leftmost)
|
2017-04-11 09:25:22 +08:00
|
|
|
{
|
2018-10-04 11:18:30 +08:00
|
|
|
struct rb_node **pnode = &root->rb_root.rb_node;
|
2017-04-11 09:25:22 +08:00
|
|
|
struct rb_node *parent = NULL, *tmp_node;
|
|
|
|
struct rb_entry *re = cached_re;
|
|
|
|
|
|
|
|
*insert_p = NULL;
|
|
|
|
*insert_parent = NULL;
|
|
|
|
*prev_entry = NULL;
|
|
|
|
*next_entry = NULL;
|
|
|
|
|
2018-10-04 11:18:30 +08:00
|
|
|
if (RB_EMPTY_ROOT(&root->rb_root))
|
2017-04-11 09:25:22 +08:00
|
|
|
return NULL;
|
|
|
|
|
|
|
|
if (re) {
|
|
|
|
if (re->ofs <= ofs && re->ofs + re->len > ofs)
|
|
|
|
goto lookup_neighbors;
|
|
|
|
}
|
|
|
|
|
2018-10-04 11:18:30 +08:00
|
|
|
if (leftmost)
|
|
|
|
*leftmost = true;
|
|
|
|
|
2017-04-11 09:25:22 +08:00
|
|
|
while (*pnode) {
|
|
|
|
parent = *pnode;
|
|
|
|
re = rb_entry(*pnode, struct rb_entry, rb_node);
|
|
|
|
|
2018-10-04 11:18:30 +08:00
|
|
|
if (ofs < re->ofs) {
|
2017-04-11 09:25:22 +08:00
|
|
|
pnode = &(*pnode)->rb_left;
|
2018-10-04 11:18:30 +08:00
|
|
|
} else if (ofs >= re->ofs + re->len) {
|
2017-04-11 09:25:22 +08:00
|
|
|
pnode = &(*pnode)->rb_right;
|
2018-10-04 11:18:30 +08:00
|
|
|
if (leftmost)
|
|
|
|
*leftmost = false;
|
|
|
|
} else {
|
2017-04-11 09:25:22 +08:00
|
|
|
goto lookup_neighbors;
|
2018-10-04 11:18:30 +08:00
|
|
|
}
|
2017-04-11 09:25:22 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
*insert_p = pnode;
|
|
|
|
*insert_parent = parent;
|
|
|
|
|
|
|
|
re = rb_entry(parent, struct rb_entry, rb_node);
|
|
|
|
tmp_node = parent;
|
|
|
|
if (parent && ofs > re->ofs)
|
|
|
|
tmp_node = rb_next(parent);
|
|
|
|
*next_entry = rb_entry_safe(tmp_node, struct rb_entry, rb_node);
|
|
|
|
|
|
|
|
tmp_node = parent;
|
|
|
|
if (parent && ofs < re->ofs)
|
|
|
|
tmp_node = rb_prev(parent);
|
|
|
|
*prev_entry = rb_entry_safe(tmp_node, struct rb_entry, rb_node);
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
lookup_neighbors:
|
2017-04-14 23:24:55 +08:00
|
|
|
if (ofs == re->ofs || force) {
|
2017-04-11 09:25:22 +08:00
|
|
|
/* lookup prev node for merging backward later */
|
|
|
|
tmp_node = rb_prev(&re->rb_node);
|
|
|
|
*prev_entry = rb_entry_safe(tmp_node, struct rb_entry, rb_node);
|
|
|
|
}
|
2017-04-14 23:24:55 +08:00
|
|
|
if (ofs == re->ofs + re->len - 1 || force) {
|
2017-04-11 09:25:22 +08:00
|
|
|
/* lookup next node for merging frontward later */
|
|
|
|
tmp_node = rb_next(&re->rb_node);
|
|
|
|
*next_entry = rb_entry_safe(tmp_node, struct rb_entry, rb_node);
|
|
|
|
}
|
|
|
|
return re;
|
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-30 00:20:41 +08:00
|
|
|
bool f2fs_check_rb_tree_consistence(struct f2fs_sb_info *sbi,
|
2020-08-04 21:14:48 +08:00
|
|
|
struct rb_root_cached *root, bool check_key)
|
2017-04-17 18:21:43 +08:00
|
|
|
{
|
|
|
|
#ifdef CONFIG_F2FS_CHECK_FS
|
2018-10-04 11:18:30 +08:00
|
|
|
struct rb_node *cur = rb_first_cached(root), *next;
|
2017-04-17 18:21:43 +08:00
|
|
|
struct rb_entry *cur_re, *next_re;
|
|
|
|
|
|
|
|
if (!cur)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
while (cur) {
|
|
|
|
next = rb_next(cur);
|
|
|
|
if (!next)
|
|
|
|
return true;
|
|
|
|
|
|
|
|
cur_re = rb_entry(cur, struct rb_entry, rb_node);
|
|
|
|
next_re = rb_entry(next, struct rb_entry, rb_node);
|
|
|
|
|
2020-08-04 21:14:48 +08:00
|
|
|
if (check_key) {
|
|
|
|
if (cur_re->key > next_re->key) {
|
|
|
|
f2fs_info(sbi, "inconsistent rbtree, "
|
|
|
|
"cur(%llu) next(%llu)",
|
|
|
|
cur_re->key, next_re->key);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
goto next;
|
|
|
|
}
|
|
|
|
|
2017-04-17 18:21:43 +08:00
|
|
|
if (cur_re->ofs + cur_re->len > next_re->ofs) {
|
2019-06-18 17:48:42 +08:00
|
|
|
f2fs_info(sbi, "inconsistent rbtree, cur(%u, %u) next(%u, %u)",
|
|
|
|
cur_re->ofs, cur_re->len,
|
|
|
|
next_re->ofs, next_re->len);
|
2017-04-17 18:21:43 +08:00
|
|
|
return false;
|
|
|
|
}
|
2020-08-04 21:14:48 +08:00
|
|
|
next:
|
2017-04-17 18:21:43 +08:00
|
|
|
cur = next;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2015-07-08 17:59:36 +08:00
|
|
|
static struct kmem_cache *extent_tree_slab;
|
|
|
|
static struct kmem_cache *extent_node_slab;
|
|
|
|
|
|
|
|
static struct extent_node *__attach_extent_node(struct f2fs_sb_info *sbi,
|
|
|
|
struct extent_tree *et, struct extent_info *ei,
|
2018-10-04 11:18:30 +08:00
|
|
|
struct rb_node *parent, struct rb_node **p,
|
|
|
|
bool leftmost)
|
2015-07-08 17:59:36 +08:00
|
|
|
{
|
|
|
|
struct extent_node *en;
|
|
|
|
|
|
|
|
en = kmem_cache_alloc(extent_node_slab, GFP_ATOMIC);
|
|
|
|
if (!en)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
en->ei = *ei;
|
|
|
|
INIT_LIST_HEAD(&en->list);
|
2016-01-26 20:56:26 +08:00
|
|
|
en->et = et;
|
2015-07-08 17:59:36 +08:00
|
|
|
|
|
|
|
rb_link_node(&en->rb_node, parent, p);
|
2018-10-04 11:18:30 +08:00
|
|
|
rb_insert_color_cached(&en->rb_node, &et->root, leftmost);
|
2016-01-08 20:22:52 +08:00
|
|
|
atomic_inc(&et->node_cnt);
|
2015-07-08 17:59:36 +08:00
|
|
|
atomic_inc(&sbi->total_ext_node);
|
|
|
|
return en;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __detach_extent_node(struct f2fs_sb_info *sbi,
|
|
|
|
struct extent_tree *et, struct extent_node *en)
|
|
|
|
{
|
2018-10-04 11:18:30 +08:00
|
|
|
rb_erase_cached(&en->rb_node, &et->root);
|
2016-01-08 20:22:52 +08:00
|
|
|
atomic_dec(&et->node_cnt);
|
2015-07-08 17:59:36 +08:00
|
|
|
atomic_dec(&sbi->total_ext_node);
|
|
|
|
|
|
|
|
if (et->cached_en == en)
|
|
|
|
et->cached_en = NULL;
|
2016-01-26 20:56:25 +08:00
|
|
|
kmem_cache_free(extent_node_slab, en);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Flow to release an extent_node:
|
|
|
|
* 1. list_del_init
|
|
|
|
* 2. __detach_extent_node
|
|
|
|
* 3. kmem_cache_free.
|
|
|
|
*/
|
|
|
|
static void __release_extent_node(struct f2fs_sb_info *sbi,
|
|
|
|
struct extent_tree *et, struct extent_node *en)
|
|
|
|
{
|
|
|
|
spin_lock(&sbi->extent_lock);
|
2016-01-26 20:56:26 +08:00
|
|
|
f2fs_bug_on(sbi, list_empty(&en->list));
|
|
|
|
list_del_init(&en->list);
|
2016-01-26 20:56:25 +08:00
|
|
|
spin_unlock(&sbi->extent_lock);
|
|
|
|
|
|
|
|
__detach_extent_node(sbi, et, en);
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
static struct extent_tree *__grab_extent_tree(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
|
|
|
struct extent_tree *et;
|
|
|
|
nid_t ino = inode->i_ino;
|
|
|
|
|
2017-02-23 19:39:59 +08:00
|
|
|
mutex_lock(&sbi->extent_tree_lock);
|
2015-07-08 17:59:36 +08:00
|
|
|
et = radix_tree_lookup(&sbi->extent_tree_root, ino);
|
|
|
|
if (!et) {
|
|
|
|
et = f2fs_kmem_cache_alloc(extent_tree_slab, GFP_NOFS);
|
|
|
|
f2fs_radix_tree_insert(&sbi->extent_tree_root, ino, et);
|
|
|
|
memset(et, 0, sizeof(struct extent_tree));
|
|
|
|
et->ino = ino;
|
2018-10-04 11:18:30 +08:00
|
|
|
et->root = RB_ROOT_CACHED;
|
2015-07-08 17:59:36 +08:00
|
|
|
et->cached_en = NULL;
|
|
|
|
rwlock_init(&et->lock);
|
2016-01-01 07:02:16 +08:00
|
|
|
INIT_LIST_HEAD(&et->list);
|
2016-01-08 20:22:52 +08:00
|
|
|
atomic_set(&et->node_cnt, 0);
|
2015-12-22 11:20:15 +08:00
|
|
|
atomic_inc(&sbi->total_ext_tree);
|
2015-12-22 11:25:50 +08:00
|
|
|
} else {
|
|
|
|
atomic_dec(&sbi->total_zombie_tree);
|
2016-01-01 07:02:16 +08:00
|
|
|
list_del_init(&et->list);
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
2017-02-23 19:39:59 +08:00
|
|
|
mutex_unlock(&sbi->extent_tree_lock);
|
2015-07-08 17:59:36 +08:00
|
|
|
|
|
|
|
/* never died until evict_inode */
|
|
|
|
F2FS_I(inode)->extent_tree = et;
|
|
|
|
|
|
|
|
return et;
|
|
|
|
}
|
|
|
|
|
2015-08-19 19:14:15 +08:00
|
|
|
static struct extent_node *__init_extent_tree(struct f2fs_sb_info *sbi,
|
|
|
|
struct extent_tree *et, struct extent_info *ei)
|
2015-07-08 17:59:36 +08:00
|
|
|
{
|
2018-10-04 11:18:30 +08:00
|
|
|
struct rb_node **p = &et->root.rb_root.rb_node;
|
2015-07-08 17:59:36 +08:00
|
|
|
struct extent_node *en;
|
|
|
|
|
2018-10-04 11:18:30 +08:00
|
|
|
en = __attach_extent_node(sbi, et, ei, NULL, p, true);
|
2015-07-08 17:59:36 +08:00
|
|
|
if (!en)
|
|
|
|
return NULL;
|
2015-08-19 19:14:15 +08:00
|
|
|
|
|
|
|
et->largest = en->ei;
|
2015-07-08 17:59:36 +08:00
|
|
|
et->cached_en = en;
|
|
|
|
return en;
|
|
|
|
}
|
|
|
|
|
|
|
|
static unsigned int __free_extent_tree(struct f2fs_sb_info *sbi,
|
2016-01-26 20:56:26 +08:00
|
|
|
struct extent_tree *et)
|
2015-07-08 17:59:36 +08:00
|
|
|
{
|
|
|
|
struct rb_node *node, *next;
|
|
|
|
struct extent_node *en;
|
2016-01-08 20:22:52 +08:00
|
|
|
unsigned int count = atomic_read(&et->node_cnt);
|
2015-07-08 17:59:36 +08:00
|
|
|
|
2018-10-04 11:18:30 +08:00
|
|
|
node = rb_first_cached(&et->root);
|
2015-07-08 17:59:36 +08:00
|
|
|
while (node) {
|
|
|
|
next = rb_next(node);
|
|
|
|
en = rb_entry(node, struct extent_node, rb_node);
|
2016-01-26 20:56:26 +08:00
|
|
|
__release_extent_node(sbi, et, en);
|
2015-07-08 17:59:36 +08:00
|
|
|
node = next;
|
|
|
|
}
|
|
|
|
|
2016-01-08 20:22:52 +08:00
|
|
|
return count - atomic_read(&et->node_cnt);
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
|
|
|
|
2018-09-10 16:18:25 +08:00
|
|
|
static void __drop_largest_extent(struct extent_tree *et,
|
2015-09-17 18:24:17 +08:00
|
|
|
pgoff_t fofs, unsigned int len)
|
2015-07-08 17:59:36 +08:00
|
|
|
{
|
2018-09-10 16:18:25 +08:00
|
|
|
if (fofs < et->largest.fofs + et->largest.len &&
|
|
|
|
fofs + len > et->largest.fofs) {
|
|
|
|
et->largest.len = 0;
|
|
|
|
et->largest_updated = true;
|
2016-05-21 00:52:20 +08:00
|
|
|
}
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
|
|
|
|
2015-12-29 03:39:06 +08:00
|
|
|
/* return true, if inode page is changed */
|
2020-06-28 10:58:17 +08:00
|
|
|
static void __f2fs_init_extent_tree(struct inode *inode, struct page *ipage)
|
2015-07-08 17:59:36 +08:00
|
|
|
{
|
|
|
|
struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
2020-06-28 10:58:17 +08:00
|
|
|
struct f2fs_extent *i_ext = ipage ? &F2FS_INODE(ipage)->i_ext : NULL;
|
2015-07-08 17:59:36 +08:00
|
|
|
struct extent_tree *et;
|
|
|
|
struct extent_node *en;
|
|
|
|
struct extent_info ei;
|
|
|
|
|
2015-12-29 03:39:06 +08:00
|
|
|
if (!f2fs_may_extent_tree(inode)) {
|
|
|
|
/* drop largest extent */
|
|
|
|
if (i_ext && i_ext->len) {
|
2020-06-28 10:58:17 +08:00
|
|
|
f2fs_wait_on_page_writeback(ipage, NODE, true, true);
|
2015-12-29 03:39:06 +08:00
|
|
|
i_ext->len = 0;
|
2020-06-28 10:58:17 +08:00
|
|
|
set_page_dirty(ipage);
|
|
|
|
return;
|
2015-12-29 03:39:06 +08:00
|
|
|
}
|
2020-06-28 10:58:17 +08:00
|
|
|
return;
|
2015-12-29 03:39:06 +08:00
|
|
|
}
|
2015-07-08 17:59:36 +08:00
|
|
|
|
|
|
|
et = __grab_extent_tree(inode);
|
|
|
|
|
2015-12-29 03:39:06 +08:00
|
|
|
if (!i_ext || !i_ext->len)
|
2020-06-28 10:58:17 +08:00
|
|
|
return;
|
2015-07-08 17:59:36 +08:00
|
|
|
|
2016-05-04 23:19:47 +08:00
|
|
|
get_extent_info(&ei, i_ext);
|
2015-07-08 17:59:36 +08:00
|
|
|
|
|
|
|
write_lock(&et->lock);
|
2016-01-08 20:22:52 +08:00
|
|
|
if (atomic_read(&et->node_cnt))
|
2015-07-08 17:59:36 +08:00
|
|
|
goto out;
|
|
|
|
|
2015-08-19 19:14:15 +08:00
|
|
|
en = __init_extent_tree(sbi, et, &ei);
|
2015-07-08 17:59:36 +08:00
|
|
|
if (en) {
|
|
|
|
spin_lock(&sbi->extent_lock);
|
|
|
|
list_add_tail(&en->list, &sbi->extent_list);
|
|
|
|
spin_unlock(&sbi->extent_lock);
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
write_unlock(&et->lock);
|
|
|
|
}
|
|
|
|
|
2020-06-28 10:58:17 +08:00
|
|
|
void f2fs_init_extent_tree(struct inode *inode, struct page *ipage)
|
2017-05-19 15:06:12 +08:00
|
|
|
{
|
2020-06-28 10:58:17 +08:00
|
|
|
__f2fs_init_extent_tree(inode, ipage);
|
2017-05-19 15:06:12 +08:00
|
|
|
|
|
|
|
if (!F2FS_I(inode)->extent_tree)
|
|
|
|
set_inode_flag(inode, FI_NO_EXTENT);
|
|
|
|
}
|
|
|
|
|
2015-07-08 17:59:36 +08:00
|
|
|
static bool f2fs_lookup_extent_tree(struct inode *inode, pgoff_t pgofs,
|
|
|
|
struct extent_info *ei)
|
|
|
|
{
|
|
|
|
struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
|
|
|
struct extent_tree *et = F2FS_I(inode)->extent_tree;
|
|
|
|
struct extent_node *en;
|
|
|
|
bool ret = false;
|
|
|
|
|
|
|
|
f2fs_bug_on(sbi, !et);
|
|
|
|
|
|
|
|
trace_f2fs_lookup_extent_tree_start(inode, pgofs);
|
|
|
|
|
|
|
|
read_lock(&et->lock);
|
|
|
|
|
|
|
|
if (et->largest.fofs <= pgofs &&
|
|
|
|
et->largest.fofs + et->largest.len > pgofs) {
|
|
|
|
*ei = et->largest;
|
|
|
|
ret = true;
|
2015-08-19 19:12:20 +08:00
|
|
|
stat_inc_largest_node_hit(sbi);
|
2015-07-08 17:59:36 +08:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-30 00:20:41 +08:00
|
|
|
en = (struct extent_node *)f2fs_lookup_rb_tree(&et->root,
|
2017-04-11 09:25:22 +08:00
|
|
|
(struct rb_entry *)et->cached_en, pgofs);
|
|
|
|
if (!en)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (en == et->cached_en)
|
|
|
|
stat_inc_cached_node_hit(sbi);
|
|
|
|
else
|
|
|
|
stat_inc_rbtree_node_hit(sbi);
|
|
|
|
|
|
|
|
*ei = en->ei;
|
|
|
|
spin_lock(&sbi->extent_lock);
|
|
|
|
if (!list_empty(&en->list)) {
|
|
|
|
list_move_tail(&en->list, &sbi->extent_list);
|
|
|
|
et->cached_en = en;
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
2017-04-11 09:25:22 +08:00
|
|
|
spin_unlock(&sbi->extent_lock);
|
|
|
|
ret = true;
|
2015-07-08 17:59:36 +08:00
|
|
|
out:
|
2015-07-15 17:29:49 +08:00
|
|
|
stat_inc_total_hit(sbi);
|
2015-07-08 17:59:36 +08:00
|
|
|
read_unlock(&et->lock);
|
|
|
|
|
|
|
|
trace_f2fs_lookup_extent_tree_end(inode, pgofs, ei);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2018-09-10 16:18:25 +08:00
|
|
|
static struct extent_node *__try_merge_extent_node(struct f2fs_sb_info *sbi,
|
2015-07-15 18:05:17 +08:00
|
|
|
struct extent_tree *et, struct extent_info *ei,
|
|
|
|
struct extent_node *prev_ex,
|
2015-08-19 19:15:09 +08:00
|
|
|
struct extent_node *next_ex)
|
2015-07-15 18:05:17 +08:00
|
|
|
{
|
|
|
|
struct extent_node *en = NULL;
|
|
|
|
|
|
|
|
if (prev_ex && __is_back_mergeable(ei, &prev_ex->ei)) {
|
|
|
|
prev_ex->ei.len += ei->len;
|
|
|
|
ei = &prev_ex->ei;
|
|
|
|
en = prev_ex;
|
|
|
|
}
|
2015-08-19 19:15:09 +08:00
|
|
|
|
2015-07-15 18:05:17 +08:00
|
|
|
if (next_ex && __is_front_mergeable(ei, &next_ex->ei)) {
|
|
|
|
next_ex->ei.fofs = ei->fofs;
|
|
|
|
next_ex->ei.blk = ei->blk;
|
|
|
|
next_ex->ei.len += ei->len;
|
2016-12-19 20:10:48 +08:00
|
|
|
if (en)
|
|
|
|
__release_extent_node(sbi, et, prev_ex);
|
|
|
|
|
2015-07-15 18:05:17 +08:00
|
|
|
en = next_ex;
|
|
|
|
}
|
2015-08-19 19:15:09 +08:00
|
|
|
|
2016-01-27 01:12:50 +08:00
|
|
|
if (!en)
|
|
|
|
return NULL;
|
|
|
|
|
2018-09-10 16:18:25 +08:00
|
|
|
__try_update_largest_extent(et, en);
|
2016-01-27 01:12:50 +08:00
|
|
|
|
|
|
|
spin_lock(&sbi->extent_lock);
|
2016-01-27 01:20:05 +08:00
|
|
|
if (!list_empty(&en->list)) {
|
2016-01-27 01:12:50 +08:00
|
|
|
list_move_tail(&en->list, &sbi->extent_list);
|
2016-01-27 01:20:05 +08:00
|
|
|
et->cached_en = en;
|
|
|
|
}
|
2016-01-27 01:12:50 +08:00
|
|
|
spin_unlock(&sbi->extent_lock);
|
2015-08-19 19:15:09 +08:00
|
|
|
return en;
|
|
|
|
}
|
|
|
|
|
2018-09-10 16:18:25 +08:00
|
|
|
static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi,
|
2015-08-19 19:15:09 +08:00
|
|
|
struct extent_tree *et, struct extent_info *ei,
|
|
|
|
struct rb_node **insert_p,
|
2018-10-04 11:18:30 +08:00
|
|
|
struct rb_node *insert_parent,
|
|
|
|
bool leftmost)
|
2015-08-19 19:15:09 +08:00
|
|
|
{
|
2018-02-22 02:13:40 +08:00
|
|
|
struct rb_node **p;
|
2015-08-19 19:15:09 +08:00
|
|
|
struct rb_node *parent = NULL;
|
|
|
|
struct extent_node *en = NULL;
|
2015-07-15 18:05:17 +08:00
|
|
|
|
|
|
|
if (insert_p && insert_parent) {
|
|
|
|
parent = insert_parent;
|
|
|
|
p = insert_p;
|
|
|
|
goto do_insert;
|
|
|
|
}
|
|
|
|
|
2018-10-04 11:18:30 +08:00
|
|
|
leftmost = true;
|
|
|
|
|
|
|
|
p = f2fs_lookup_rb_tree_for_insert(sbi, &et->root, &parent,
|
|
|
|
ei->fofs, &leftmost);
|
2015-07-15 18:05:17 +08:00
|
|
|
do_insert:
|
2018-10-04 11:18:30 +08:00
|
|
|
en = __attach_extent_node(sbi, et, ei, parent, p, leftmost);
|
2015-07-15 18:05:17 +08:00
|
|
|
if (!en)
|
|
|
|
return NULL;
|
2015-08-19 19:15:09 +08:00
|
|
|
|
2018-09-10 16:18:25 +08:00
|
|
|
__try_update_largest_extent(et, en);
|
2016-01-27 01:12:50 +08:00
|
|
|
|
|
|
|
/* update in global extent list */
|
|
|
|
spin_lock(&sbi->extent_lock);
|
|
|
|
list_add_tail(&en->list, &sbi->extent_list);
|
2016-01-27 01:20:05 +08:00
|
|
|
et->cached_en = en;
|
2016-01-27 01:12:50 +08:00
|
|
|
spin_unlock(&sbi->extent_lock);
|
2015-07-15 18:05:17 +08:00
|
|
|
return en;
|
|
|
|
}
|
|
|
|
|
2017-02-25 17:29:54 +08:00
|
|
|
static void f2fs_update_extent_tree_range(struct inode *inode,
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
pgoff_t fofs, block_t blkaddr, unsigned int len)
|
2015-07-08 17:59:36 +08:00
|
|
|
{
|
|
|
|
struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
|
|
|
struct extent_tree *et = F2FS_I(inode)->extent_tree;
|
2015-09-17 18:42:06 +08:00
|
|
|
struct extent_node *en = NULL, *en1 = NULL;
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
struct extent_node *prev_en = NULL, *next_en = NULL;
|
2015-07-08 17:59:36 +08:00
|
|
|
struct extent_info ei, dei, prev;
|
2015-07-15 18:05:17 +08:00
|
|
|
struct rb_node **insert_p = NULL, *insert_parent = NULL;
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
unsigned int end = fofs + len;
|
|
|
|
unsigned int pos = (unsigned int)fofs;
|
2018-09-10 16:18:25 +08:00
|
|
|
bool updated = false;
|
f2fs: fix to initialize variable to avoid UBSAN/smatch warning
As Dan Carpenter as below:
The patch df634f444ee9: "f2fs: use rb_*_cached friends" from Oct 4,
2018, leads to the following static checker warning:
fs/f2fs/extent_cache.c:606 f2fs_update_extent_tree_range()
error: uninitialized symbol 'leftmost'.
And also Eric Biggers, and Kyungtae Kim reported, there is an UBSAN
warning described as below:
We report a bug in linux-4.20.2: "UBSAN: Undefined behaviour in
fs/f2fs/extent_cache.c"
kernel config: https://kt0755.github.io/etc/config_v4.20_stable
repro: https://kt0755.github.io/etc/repro.4a3e7.c (f2fs is mounted on
/mnt/f2fs/)
This arose in f2fs_update_extent_tree_range (fs/f2fs/extent_cache.c:605).
It seems that, for some reason, its last argument became "24"
although that was supposed to be bool type.
=========================================
UBSAN: Undefined behaviour in fs/f2fs/extent_cache.c:605:4
load of value 24 is not a valid value for type '_Bool'
CPU: 0 PID: 6774 Comm: syz-executor5 Not tainted 4.20.2 #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
Call Trace:
__dump_stack lib/dump_stack.c:77 [inline]
dump_stack+0xb1/0x118 lib/dump_stack.c:113
ubsan_epilogue+0x12/0x94 lib/ubsan.c:159
__ubsan_handle_load_invalid_value+0x17a/0x1be lib/ubsan.c:457
f2fs_update_extent_tree_range+0x1d4a/0x1d50 fs/f2fs/extent_cache.c:605
f2fs_update_extent_cache+0x2b6/0x350 fs/f2fs/extent_cache.c:804
f2fs_update_data_blkaddr+0x61/0x70 fs/f2fs/data.c:656
f2fs_outplace_write_data+0x1d6/0x4b0 fs/f2fs/segment.c:3140
f2fs_convert_inline_page+0x86d/0x2060 fs/f2fs/inline.c:163
f2fs_convert_inline_inode+0x6b5/0xad0 fs/f2fs/inline.c:208
f2fs_preallocate_blocks+0x78b/0xb00 fs/f2fs/data.c:982
f2fs_file_write_iter+0x31b/0xf40 fs/f2fs/file.c:3062
call_write_iter include/linux/fs.h:1857 [inline]
new_sync_write fs/read_write.c:474 [inline]
__vfs_write+0x538/0x6e0 fs/read_write.c:487
vfs_write+0x1b3/0x520 fs/read_write.c:549
ksys_write+0xde/0x1c0 fs/read_write.c:598
__do_sys_write fs/read_write.c:610 [inline]
__se_sys_write fs/read_write.c:607 [inline]
__x64_sys_write+0x7e/0xc0 fs/read_write.c:607
do_syscall_64+0xbe/0x4f0 arch/x86/entry/common.c:290
entry_SYSCALL_64_after_hwframe+0x49/0xbe
RIP: 0033:0x4497b9
Code: e8 8c 9f 02 00 48 83 c4 18 c3 0f 1f 80 00 00 00 00 48 89 f8 48
89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d
01 f0 ff ff 0f 83 9b 6b fc ff c3 66 2e 0f 1f 84 00 00 00 00
RSP: 002b:00007f1ea15edc68 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
RAX: ffffffffffffffda RBX: 00007f1ea15ee6cc RCX: 00000000004497b9
RDX: 0000000000001000 RSI: 0000000020000140 RDI: 0000000000000013
RBP: 000000000071bea0 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 00000000ffffffff
R13: 000000000000bb50 R14: 00000000006f4bf0 R15: 00007f1ea15ee700
=========================================
As I checked, this uninitialized variable won't cause extent cache
corruption, but in order to avoid such kind of warning of both UBSAN
and smatch, fix to initialize related variable.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Reported-by: Eric Biggers <ebiggers@google.com>
Reported-by: Kyungtae Kim <kt0755@gmail.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2019-01-16 09:51:28 +08:00
|
|
|
bool leftmost = false;
|
2015-07-08 17:59:36 +08:00
|
|
|
|
|
|
|
if (!et)
|
2017-02-25 17:29:54 +08:00
|
|
|
return;
|
2015-07-08 17:59:36 +08:00
|
|
|
|
2015-09-06 17:50:13 +08:00
|
|
|
trace_f2fs_update_extent_tree_range(inode, fofs, blkaddr, len);
|
|
|
|
|
2015-07-08 17:59:36 +08:00
|
|
|
write_lock(&et->lock);
|
|
|
|
|
2016-05-21 01:13:22 +08:00
|
|
|
if (is_inode_flag_set(inode, FI_NO_EXTENT)) {
|
2015-07-08 17:59:36 +08:00
|
|
|
write_unlock(&et->lock);
|
2017-02-25 17:29:54 +08:00
|
|
|
return;
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
prev = et->largest;
|
|
|
|
dei.len = 0;
|
|
|
|
|
2015-09-17 18:42:06 +08:00
|
|
|
/*
|
|
|
|
* drop largest extent before lookup, in case it's already
|
|
|
|
* been shrunk from extent tree
|
|
|
|
*/
|
2018-09-10 16:18:25 +08:00
|
|
|
__drop_largest_extent(et, fofs, len);
|
2015-07-08 17:59:36 +08:00
|
|
|
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
/* 1. lookup first extent node in range [fofs, fofs + len - 1] */
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-30 00:20:41 +08:00
|
|
|
en = (struct extent_node *)f2fs_lookup_rb_tree_ret(&et->root,
|
2017-04-11 09:25:22 +08:00
|
|
|
(struct rb_entry *)et->cached_en, fofs,
|
|
|
|
(struct rb_entry **)&prev_en,
|
|
|
|
(struct rb_entry **)&next_en,
|
2018-10-04 11:18:30 +08:00
|
|
|
&insert_p, &insert_parent, false,
|
|
|
|
&leftmost);
|
2015-09-17 18:42:06 +08:00
|
|
|
if (!en)
|
|
|
|
en = next_en;
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
|
|
|
|
/* 2. invlidate all extent nodes in range [fofs, fofs + len - 1] */
|
2015-09-17 18:42:06 +08:00
|
|
|
while (en && en->ei.fofs < end) {
|
|
|
|
unsigned int org_end;
|
|
|
|
int parts = 0; /* # of parts current extent split into */
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
|
2015-09-17 18:42:06 +08:00
|
|
|
next_en = en1 = NULL;
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
|
|
|
|
dei = en->ei;
|
2015-09-17 18:42:06 +08:00
|
|
|
org_end = dei.fofs + dei.len;
|
|
|
|
f2fs_bug_on(sbi, pos >= org_end);
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
|
2015-09-17 18:42:06 +08:00
|
|
|
if (pos > dei.fofs && pos - dei.fofs >= F2FS_MIN_EXTENT_LEN) {
|
|
|
|
en->ei.len = pos - en->ei.fofs;
|
|
|
|
prev_en = en;
|
|
|
|
parts = 1;
|
|
|
|
}
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
|
2015-09-17 18:42:06 +08:00
|
|
|
if (end < org_end && org_end - end >= F2FS_MIN_EXTENT_LEN) {
|
|
|
|
if (parts) {
|
|
|
|
set_extent_info(&ei, end,
|
|
|
|
end - dei.fofs + dei.blk,
|
|
|
|
org_end - end);
|
2018-09-10 16:18:25 +08:00
|
|
|
en1 = __insert_extent_tree(sbi, et, &ei,
|
2018-10-04 11:18:30 +08:00
|
|
|
NULL, NULL, true);
|
2015-09-17 18:42:06 +08:00
|
|
|
next_en = en1;
|
|
|
|
} else {
|
|
|
|
en->ei.fofs = end;
|
|
|
|
en->ei.blk += end - dei.fofs;
|
|
|
|
en->ei.len -= end - dei.fofs;
|
|
|
|
next_en = en;
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
}
|
2015-09-17 18:42:06 +08:00
|
|
|
parts++;
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
}
|
|
|
|
|
2015-09-17 18:42:06 +08:00
|
|
|
if (!next_en) {
|
|
|
|
struct rb_node *node = rb_next(&en->rb_node);
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
|
2016-12-20 21:57:42 +08:00
|
|
|
next_en = rb_entry_safe(node, struct extent_node,
|
|
|
|
rb_node);
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
|
|
|
|
2015-09-22 21:07:47 +08:00
|
|
|
if (parts)
|
2018-09-10 16:18:25 +08:00
|
|
|
__try_update_largest_extent(et, en);
|
2015-09-22 21:07:47 +08:00
|
|
|
else
|
2016-01-26 20:56:25 +08:00
|
|
|
__release_extent_node(sbi, et, en);
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
|
|
|
|
/*
|
2015-09-17 18:42:06 +08:00
|
|
|
* if original extent is split into zero or two parts, extent
|
|
|
|
* tree has been altered by deletion or insertion, therefore
|
|
|
|
* invalidate pointers regard to tree.
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
*/
|
2015-09-17 18:42:06 +08:00
|
|
|
if (parts != 1) {
|
|
|
|
insert_p = NULL;
|
|
|
|
insert_parent = NULL;
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
2015-09-17 18:42:06 +08:00
|
|
|
en = next_en;
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* 3. update extent in extent cache */
|
|
|
|
if (blkaddr) {
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
|
|
|
|
set_extent_info(&ei, fofs, blkaddr, len);
|
2018-09-10 16:18:25 +08:00
|
|
|
if (!__try_merge_extent_node(sbi, et, &ei, prev_en, next_en))
|
|
|
|
__insert_extent_tree(sbi, et, &ei,
|
2018-10-04 11:18:30 +08:00
|
|
|
insert_p, insert_parent, leftmost);
|
2015-07-08 17:59:36 +08:00
|
|
|
|
|
|
|
/* give up extent_cache, if split and small updates happen */
|
|
|
|
if (dei.len >= 1 &&
|
|
|
|
prev.len < F2FS_MIN_EXTENT_LEN &&
|
|
|
|
et->largest.len < F2FS_MIN_EXTENT_LEN) {
|
2018-09-10 16:18:25 +08:00
|
|
|
et->largest.len = 0;
|
|
|
|
et->largest_updated = true;
|
2016-05-21 01:13:22 +08:00
|
|
|
set_inode_flag(inode, FI_NO_EXTENT);
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
}
|
2015-07-08 17:59:36 +08:00
|
|
|
|
2016-05-21 01:13:22 +08:00
|
|
|
if (is_inode_flag_set(inode, FI_NO_EXTENT))
|
2016-01-26 20:56:26 +08:00
|
|
|
__free_extent_tree(sbi, et);
|
2015-07-08 17:59:36 +08:00
|
|
|
|
2018-09-10 16:18:25 +08:00
|
|
|
if (et->largest_updated) {
|
|
|
|
et->largest_updated = false;
|
|
|
|
updated = true;
|
|
|
|
}
|
|
|
|
|
2015-07-08 17:59:36 +08:00
|
|
|
write_unlock(&et->lock);
|
2018-09-10 16:18:25 +08:00
|
|
|
|
|
|
|
if (updated)
|
|
|
|
f2fs_mark_inode_dirty_sync(inode, true);
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
unsigned int f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink)
|
|
|
|
{
|
2016-01-01 07:02:16 +08:00
|
|
|
struct extent_tree *et, *next;
|
2016-01-26 20:56:26 +08:00
|
|
|
struct extent_node *en;
|
2015-07-08 17:59:36 +08:00
|
|
|
unsigned int node_cnt = 0, tree_cnt = 0;
|
|
|
|
int remained;
|
|
|
|
|
|
|
|
if (!test_opt(sbi, EXTENT_CACHE))
|
|
|
|
return 0;
|
|
|
|
|
2015-12-22 11:25:50 +08:00
|
|
|
if (!atomic_read(&sbi->total_zombie_tree))
|
|
|
|
goto free_node;
|
|
|
|
|
2017-02-23 19:39:59 +08:00
|
|
|
if (!mutex_trylock(&sbi->extent_tree_lock))
|
2015-07-08 17:59:36 +08:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
/* 1. remove unreferenced extent tree */
|
2016-01-01 07:02:16 +08:00
|
|
|
list_for_each_entry_safe(et, next, &sbi->zombie_list, list) {
|
2016-01-08 20:24:00 +08:00
|
|
|
if (atomic_read(&et->node_cnt)) {
|
|
|
|
write_lock(&et->lock);
|
2016-01-26 20:56:26 +08:00
|
|
|
node_cnt += __free_extent_tree(sbi, et);
|
2016-01-08 20:24:00 +08:00
|
|
|
write_unlock(&et->lock);
|
|
|
|
}
|
2016-01-26 20:56:26 +08:00
|
|
|
f2fs_bug_on(sbi, atomic_read(&et->node_cnt));
|
2016-01-01 07:02:16 +08:00
|
|
|
list_del_init(&et->list);
|
|
|
|
radix_tree_delete(&sbi->extent_tree_root, et->ino);
|
|
|
|
kmem_cache_free(extent_tree_slab, et);
|
|
|
|
atomic_dec(&sbi->total_ext_tree);
|
|
|
|
atomic_dec(&sbi->total_zombie_tree);
|
|
|
|
tree_cnt++;
|
2015-07-08 17:59:36 +08:00
|
|
|
|
2016-01-01 07:02:16 +08:00
|
|
|
if (node_cnt + tree_cnt >= nr_shrink)
|
|
|
|
goto unlock_out;
|
2016-01-20 07:31:48 +08:00
|
|
|
cond_resched();
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
2017-02-23 19:39:59 +08:00
|
|
|
mutex_unlock(&sbi->extent_tree_lock);
|
2015-07-08 17:59:36 +08:00
|
|
|
|
2015-12-22 11:25:50 +08:00
|
|
|
free_node:
|
2015-07-08 17:59:36 +08:00
|
|
|
/* 2. remove LRU extent entries */
|
2017-02-23 19:39:59 +08:00
|
|
|
if (!mutex_trylock(&sbi->extent_tree_lock))
|
2015-07-08 17:59:36 +08:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
remained = nr_shrink - (node_cnt + tree_cnt);
|
|
|
|
|
|
|
|
spin_lock(&sbi->extent_lock);
|
2016-01-26 20:56:26 +08:00
|
|
|
for (; remained > 0; remained--) {
|
|
|
|
if (list_empty(&sbi->extent_list))
|
2015-07-08 17:59:36 +08:00
|
|
|
break;
|
2016-01-26 20:56:26 +08:00
|
|
|
en = list_first_entry(&sbi->extent_list,
|
|
|
|
struct extent_node, list);
|
|
|
|
et = en->et;
|
|
|
|
if (!write_trylock(&et->lock)) {
|
|
|
|
/* refresh this extent node's position in extent list */
|
|
|
|
list_move_tail(&en->list, &sbi->extent_list);
|
|
|
|
continue;
|
|
|
|
}
|
2015-07-08 17:59:36 +08:00
|
|
|
|
2016-01-26 20:56:26 +08:00
|
|
|
list_del_init(&en->list);
|
|
|
|
spin_unlock(&sbi->extent_lock);
|
2016-01-08 20:24:00 +08:00
|
|
|
|
2016-01-26 20:56:26 +08:00
|
|
|
__detach_extent_node(sbi, et, en);
|
2015-07-08 17:59:36 +08:00
|
|
|
|
2016-01-26 20:56:26 +08:00
|
|
|
write_unlock(&et->lock);
|
|
|
|
node_cnt++;
|
|
|
|
spin_lock(&sbi->extent_lock);
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
2016-01-26 20:56:26 +08:00
|
|
|
spin_unlock(&sbi->extent_lock);
|
|
|
|
|
2015-07-08 17:59:36 +08:00
|
|
|
unlock_out:
|
2017-02-23 19:39:59 +08:00
|
|
|
mutex_unlock(&sbi->extent_tree_lock);
|
2015-07-08 17:59:36 +08:00
|
|
|
out:
|
|
|
|
trace_f2fs_shrink_extent_tree(sbi, node_cnt, tree_cnt);
|
|
|
|
|
|
|
|
return node_cnt + tree_cnt;
|
|
|
|
}
|
|
|
|
|
|
|
|
unsigned int f2fs_destroy_extent_node(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
|
|
|
struct extent_tree *et = F2FS_I(inode)->extent_tree;
|
|
|
|
unsigned int node_cnt = 0;
|
|
|
|
|
2016-01-08 20:24:00 +08:00
|
|
|
if (!et || !atomic_read(&et->node_cnt))
|
2015-07-08 17:59:36 +08:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
write_lock(&et->lock);
|
2016-01-26 20:56:26 +08:00
|
|
|
node_cnt = __free_extent_tree(sbi, et);
|
2015-07-08 17:59:36 +08:00
|
|
|
write_unlock(&et->lock);
|
|
|
|
|
|
|
|
return node_cnt;
|
|
|
|
}
|
|
|
|
|
2016-07-13 02:07:52 +08:00
|
|
|
void f2fs_drop_extent_tree(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
|
|
|
struct extent_tree *et = F2FS_I(inode)->extent_tree;
|
2018-09-10 16:18:25 +08:00
|
|
|
bool updated = false;
|
2016-07-13 02:07:52 +08:00
|
|
|
|
2018-01-27 17:29:49 +08:00
|
|
|
if (!f2fs_may_extent_tree(inode))
|
|
|
|
return;
|
|
|
|
|
2016-07-13 02:07:52 +08:00
|
|
|
set_inode_flag(inode, FI_NO_EXTENT);
|
|
|
|
|
|
|
|
write_lock(&et->lock);
|
|
|
|
__free_extent_tree(sbi, et);
|
2018-09-10 16:18:25 +08:00
|
|
|
if (et->largest.len) {
|
|
|
|
et->largest.len = 0;
|
|
|
|
updated = true;
|
|
|
|
}
|
2016-07-13 02:07:52 +08:00
|
|
|
write_unlock(&et->lock);
|
2018-09-10 16:18:25 +08:00
|
|
|
if (updated)
|
|
|
|
f2fs_mark_inode_dirty_sync(inode, true);
|
2016-07-13 02:07:52 +08:00
|
|
|
}
|
|
|
|
|
2015-07-08 17:59:36 +08:00
|
|
|
void f2fs_destroy_extent_tree(struct inode *inode)
|
|
|
|
{
|
|
|
|
struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
|
|
|
|
struct extent_tree *et = F2FS_I(inode)->extent_tree;
|
|
|
|
unsigned int node_cnt = 0;
|
|
|
|
|
|
|
|
if (!et)
|
|
|
|
return;
|
|
|
|
|
2016-01-08 20:22:52 +08:00
|
|
|
if (inode->i_nlink && !is_bad_inode(inode) &&
|
|
|
|
atomic_read(&et->node_cnt)) {
|
2017-02-23 19:39:59 +08:00
|
|
|
mutex_lock(&sbi->extent_tree_lock);
|
2016-01-01 07:02:16 +08:00
|
|
|
list_add_tail(&et->list, &sbi->zombie_list);
|
2015-12-22 11:25:50 +08:00
|
|
|
atomic_inc(&sbi->total_zombie_tree);
|
2017-02-23 19:39:59 +08:00
|
|
|
mutex_unlock(&sbi->extent_tree_lock);
|
2015-07-08 17:59:36 +08:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* free all extent info belong to this extent tree */
|
|
|
|
node_cnt = f2fs_destroy_extent_node(inode);
|
|
|
|
|
|
|
|
/* delete extent tree entry in radix tree */
|
2017-02-23 19:39:59 +08:00
|
|
|
mutex_lock(&sbi->extent_tree_lock);
|
2016-01-08 20:22:52 +08:00
|
|
|
f2fs_bug_on(sbi, atomic_read(&et->node_cnt));
|
2015-07-08 17:59:36 +08:00
|
|
|
radix_tree_delete(&sbi->extent_tree_root, inode->i_ino);
|
|
|
|
kmem_cache_free(extent_tree_slab, et);
|
2015-12-22 11:20:15 +08:00
|
|
|
atomic_dec(&sbi->total_ext_tree);
|
2017-02-23 19:39:59 +08:00
|
|
|
mutex_unlock(&sbi->extent_tree_lock);
|
2015-07-08 17:59:36 +08:00
|
|
|
|
|
|
|
F2FS_I(inode)->extent_tree = NULL;
|
|
|
|
|
|
|
|
trace_f2fs_destroy_extent_tree(inode, node_cnt);
|
|
|
|
}
|
|
|
|
|
|
|
|
bool f2fs_lookup_extent_cache(struct inode *inode, pgoff_t pgofs,
|
|
|
|
struct extent_info *ei)
|
|
|
|
{
|
|
|
|
if (!f2fs_may_extent_tree(inode))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
return f2fs_lookup_extent_tree(inode, pgofs, ei);
|
|
|
|
}
|
|
|
|
|
|
|
|
void f2fs_update_extent_cache(struct dnode_of_data *dn)
|
|
|
|
{
|
|
|
|
pgoff_t fofs;
|
2016-02-24 17:16:47 +08:00
|
|
|
block_t blkaddr;
|
2015-07-08 17:59:36 +08:00
|
|
|
|
|
|
|
if (!f2fs_may_extent_tree(dn->inode))
|
|
|
|
return;
|
|
|
|
|
2016-02-24 17:16:47 +08:00
|
|
|
if (dn->data_blkaddr == NEW_ADDR)
|
|
|
|
blkaddr = NULL_ADDR;
|
|
|
|
else
|
|
|
|
blkaddr = dn->data_blkaddr;
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-30 00:20:41 +08:00
|
|
|
fofs = f2fs_start_bidx_of_node(ofs_of_node(dn->node_page), dn->inode) +
|
2016-01-26 15:39:35 +08:00
|
|
|
dn->ofs_in_node;
|
2016-05-21 07:32:49 +08:00
|
|
|
f2fs_update_extent_tree_range(dn->inode, fofs, blkaddr, 1);
|
f2fs: update extent tree in batches
This patch introduce a new helper f2fs_update_extent_tree_range which can
do extent mapping update at a specified range.
The main idea is:
1) punch all mapping info in extent node(s) which are at a specified range;
2) try to merge new extent mapping with adjacent node, or failing that,
insert the mapping into extent tree as a new node.
In order to see the benefit, I add a function for stating time stamping
count as below:
uint64_t rdtsc(void)
{
uint32_t lo, hi;
__asm__ __volatile__ ("rdtsc" : "=a" (lo), "=d" (hi));
return (uint64_t)hi << 32 | lo;
}
My test environment is: ubuntu, intel i7-3770, 16G memory, 256g micron ssd.
truncation path: update extent cache from truncate_data_blocks_range
non-truncataion path: update extent cache from other paths
total: all update paths
a) Removing 128MB file which has one extent node mapping whole range of
file:
1. dd if=/dev/zero of=/mnt/f2fs/128M bs=1M count=128
2. sync
3. rm /mnt/f2fs/128M
Before:
total count average
truncation: 7651022 32768 233.49
Patched:
total count average
truncation: 3321 33 100.64
b) fsstress:
fsstress -d /mnt/f2fs -l 5 -n 100 -p 20
Test times: 5 times.
Before:
total count average
truncation: 5812480.6 20911.6 277.95
non-truncation: 7783845.6 13440.8 579.12
total: 13596326.2 34352.4 395.79
Patched:
total count average
truncation: 1281283.0 3041.6 421.25
non-truncation: 7355844.4 13662.8 538.38
total: 8637127.4 16704.4 517.06
1) For the updates in truncation path:
- we can see updating in batches leads total tsc and update count reducing
explicitly;
- besides, for a single batched updating, punching multiple extent nodes
in a loop, result in executing more operations, so our average tsc
increase intensively.
2) For the updates in non-truncation path:
- there is a little improvement, that is because for the scenario that we
just need to update in the head or tail of extent node, new interface
optimize to update info in extent node directly, rather than removing
original extent node for updating and then inserting that updated one
into cache as new node.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2015-08-26 20:34:48 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
void f2fs_update_extent_cache_range(struct dnode_of_data *dn,
|
|
|
|
pgoff_t fofs, block_t blkaddr, unsigned int len)
|
|
|
|
|
|
|
|
{
|
|
|
|
if (!f2fs_may_extent_tree(dn->inode))
|
|
|
|
return;
|
|
|
|
|
2016-05-21 07:32:49 +08:00
|
|
|
f2fs_update_extent_tree_range(dn->inode, fofs, blkaddr, len);
|
2015-07-08 17:59:36 +08:00
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-30 00:20:41 +08:00
|
|
|
void f2fs_init_extent_cache_info(struct f2fs_sb_info *sbi)
|
2015-07-08 17:59:36 +08:00
|
|
|
{
|
|
|
|
INIT_RADIX_TREE(&sbi->extent_tree_root, GFP_NOIO);
|
2017-02-23 19:39:59 +08:00
|
|
|
mutex_init(&sbi->extent_tree_lock);
|
2015-07-08 17:59:36 +08:00
|
|
|
INIT_LIST_HEAD(&sbi->extent_list);
|
|
|
|
spin_lock_init(&sbi->extent_lock);
|
2015-12-22 11:20:15 +08:00
|
|
|
atomic_set(&sbi->total_ext_tree, 0);
|
2016-01-01 07:02:16 +08:00
|
|
|
INIT_LIST_HEAD(&sbi->zombie_list);
|
2015-12-22 11:25:50 +08:00
|
|
|
atomic_set(&sbi->total_zombie_tree, 0);
|
2015-07-08 17:59:36 +08:00
|
|
|
atomic_set(&sbi->total_ext_node, 0);
|
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-30 00:20:41 +08:00
|
|
|
int __init f2fs_create_extent_cache(void)
|
2015-07-08 17:59:36 +08:00
|
|
|
{
|
|
|
|
extent_tree_slab = f2fs_kmem_cache_create("f2fs_extent_tree",
|
|
|
|
sizeof(struct extent_tree));
|
|
|
|
if (!extent_tree_slab)
|
|
|
|
return -ENOMEM;
|
|
|
|
extent_node_slab = f2fs_kmem_cache_create("f2fs_extent_node",
|
|
|
|
sizeof(struct extent_node));
|
|
|
|
if (!extent_node_slab) {
|
|
|
|
kmem_cache_destroy(extent_tree_slab);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-30 00:20:41 +08:00
|
|
|
void f2fs_destroy_extent_cache(void)
|
2015-07-08 17:59:36 +08:00
|
|
|
{
|
|
|
|
kmem_cache_destroy(extent_node_slab);
|
|
|
|
kmem_cache_destroy(extent_tree_slab);
|
|
|
|
}
|