2008-01-25 04:13:08 +07:00
|
|
|
#include <linux/bitops.h>
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/bio.h>
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/pagemap.h>
|
|
|
|
#include <linux/page-flags.h>
|
|
|
|
#include <linux/spinlock.h>
|
|
|
|
#include <linux/blkdev.h>
|
|
|
|
#include <linux/swap.h>
|
|
|
|
#include <linux/writeback.h>
|
|
|
|
#include <linux/pagevec.h>
|
2011-05-21 02:50:29 +07:00
|
|
|
#include <linux/prefetch.h>
|
2011-05-26 23:01:56 +07:00
|
|
|
#include <linux/cleancache.h>
|
2008-01-25 04:13:08 +07:00
|
|
|
#include "extent_io.h"
|
|
|
|
#include "extent_map.h"
|
2008-08-20 19:51:49 +07:00
|
|
|
#include "ctree.h"
|
|
|
|
#include "btrfs_inode.h"
|
2011-07-22 20:41:52 +07:00
|
|
|
#include "volumes.h"
|
2011-11-09 19:44:05 +07:00
|
|
|
#include "check-integrity.h"
|
2012-03-13 20:38:00 +07:00
|
|
|
#include "locking.h"
|
2012-06-05 01:03:51 +07:00
|
|
|
#include "rcu-string.h"
|
2013-09-22 11:54:23 +07:00
|
|
|
#include "backref.h"
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
static struct kmem_cache *extent_state_cache;
|
|
|
|
static struct kmem_cache *extent_buffer_cache;
|
2013-05-18 05:30:14 +07:00
|
|
|
static struct bio_set *btrfs_bioset;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2014-07-07 02:09:59 +07:00
|
|
|
static inline bool extent_state_in_tree(const struct extent_state *state)
|
|
|
|
{
|
|
|
|
return !RB_EMPTY_NODE(&state->rb_node);
|
|
|
|
}
|
|
|
|
|
2013-04-22 23:12:31 +07:00
|
|
|
#ifdef CONFIG_BTRFS_DEBUG
|
2008-01-25 04:13:08 +07:00
|
|
|
static LIST_HEAD(buffers);
|
|
|
|
static LIST_HEAD(states);
|
2008-09-08 22:18:08 +07:00
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
static DEFINE_SPINLOCK(leak_lock);
|
2013-04-22 23:12:31 +07:00
|
|
|
|
|
|
|
static inline
|
|
|
|
void btrfs_leak_debug_add(struct list_head *new, struct list_head *head)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&leak_lock, flags);
|
|
|
|
list_add(new, head);
|
|
|
|
spin_unlock_irqrestore(&leak_lock, flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline
|
|
|
|
void btrfs_leak_debug_del(struct list_head *entry)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&leak_lock, flags);
|
|
|
|
list_del(entry);
|
|
|
|
spin_unlock_irqrestore(&leak_lock, flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline
|
|
|
|
void btrfs_leak_debug_check(void)
|
|
|
|
{
|
|
|
|
struct extent_state *state;
|
|
|
|
struct extent_buffer *eb;
|
|
|
|
|
|
|
|
while (!list_empty(&states)) {
|
|
|
|
state = list_entry(states.next, struct extent_state, leak_list);
|
2014-07-07 02:09:59 +07:00
|
|
|
pr_err("BTRFS: state leak: start %llu end %llu state %lu in tree %d refs %d\n",
|
|
|
|
state->start, state->end, state->state,
|
|
|
|
extent_state_in_tree(state),
|
2013-08-20 18:20:07 +07:00
|
|
|
atomic_read(&state->refs));
|
2013-04-22 23:12:31 +07:00
|
|
|
list_del(&state->leak_list);
|
|
|
|
kmem_cache_free(extent_state_cache, state);
|
|
|
|
}
|
|
|
|
|
|
|
|
while (!list_empty(&buffers)) {
|
|
|
|
eb = list_entry(buffers.next, struct extent_buffer, leak_list);
|
2013-12-20 23:37:06 +07:00
|
|
|
printk(KERN_ERR "BTRFS: buffer leak start %llu len %lu "
|
2013-08-20 18:20:07 +07:00
|
|
|
"refs %d\n",
|
|
|
|
eb->start, eb->len, atomic_read(&eb->refs));
|
2013-04-22 23:12:31 +07:00
|
|
|
list_del(&eb->leak_list);
|
|
|
|
kmem_cache_free(extent_buffer_cache, eb);
|
|
|
|
}
|
|
|
|
}
|
2013-04-30 22:22:23 +07:00
|
|
|
|
2013-12-13 22:02:44 +07:00
|
|
|
#define btrfs_debug_check_extent_io_range(tree, start, end) \
|
|
|
|
__btrfs_debug_check_extent_io_range(__func__, (tree), (start), (end))
|
2013-04-30 22:22:23 +07:00
|
|
|
static inline void __btrfs_debug_check_extent_io_range(const char *caller,
|
2013-12-13 22:02:44 +07:00
|
|
|
struct extent_io_tree *tree, u64 start, u64 end)
|
2013-04-30 22:22:23 +07:00
|
|
|
{
|
2013-12-13 22:02:44 +07:00
|
|
|
struct inode *inode;
|
|
|
|
u64 isize;
|
2013-04-30 22:22:23 +07:00
|
|
|
|
2013-12-13 22:02:44 +07:00
|
|
|
if (!tree->mapping)
|
|
|
|
return;
|
2013-04-30 22:22:23 +07:00
|
|
|
|
2013-12-13 22:02:44 +07:00
|
|
|
inode = tree->mapping->host;
|
|
|
|
isize = i_size_read(inode);
|
2013-04-30 22:22:23 +07:00
|
|
|
if (end >= PAGE_SIZE && (end % 2) == 0 && end != isize - 1) {
|
|
|
|
printk_ratelimited(KERN_DEBUG
|
2013-12-20 23:37:06 +07:00
|
|
|
"BTRFS: %s: ino %llu isize %llu odd range [%llu,%llu]\n",
|
2013-08-20 18:20:07 +07:00
|
|
|
caller, btrfs_ino(inode), isize, start, end);
|
2013-04-30 22:22:23 +07:00
|
|
|
}
|
|
|
|
}
|
2013-04-22 23:12:31 +07:00
|
|
|
#else
|
|
|
|
#define btrfs_leak_debug_add(new, head) do {} while (0)
|
|
|
|
#define btrfs_leak_debug_del(entry) do {} while (0)
|
|
|
|
#define btrfs_leak_debug_check() do {} while (0)
|
2013-04-30 22:22:23 +07:00
|
|
|
#define btrfs_debug_check_extent_io_range(c, s, e) do {} while (0)
|
2008-09-08 22:18:08 +07:00
|
|
|
#endif
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
#define BUFFER_LRU_MAX 64
|
|
|
|
|
|
|
|
struct tree_entry {
|
|
|
|
u64 start;
|
|
|
|
u64 end;
|
|
|
|
struct rb_node rb_node;
|
|
|
|
};
|
|
|
|
|
|
|
|
struct extent_page_data {
|
|
|
|
struct bio *bio;
|
|
|
|
struct extent_io_tree *tree;
|
|
|
|
get_extent_t *get_extent;
|
2012-09-26 01:25:58 +07:00
|
|
|
unsigned long bio_flags;
|
2008-11-07 10:02:51 +07:00
|
|
|
|
|
|
|
/* tells writepage not to lock the state bits for this range
|
|
|
|
* it still does the unlocking
|
|
|
|
*/
|
2009-04-21 02:50:09 +07:00
|
|
|
unsigned int extent_locked:1;
|
|
|
|
|
|
|
|
/* tells the submit_bio code to use a WRITE_SYNC */
|
|
|
|
unsigned int sync_io:1;
|
2008-01-25 04:13:08 +07:00
|
|
|
};
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
static noinline void flush_write_bio(void *data);
|
2011-10-04 10:22:32 +07:00
|
|
|
static inline struct btrfs_fs_info *
|
|
|
|
tree_fs_info(struct extent_io_tree *tree)
|
|
|
|
{
|
2013-12-13 22:02:44 +07:00
|
|
|
if (!tree->mapping)
|
|
|
|
return NULL;
|
2011-10-04 10:22:32 +07:00
|
|
|
return btrfs_sb(tree->mapping->host->i_sb);
|
|
|
|
}
|
2012-03-13 20:38:00 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
int __init extent_io_init(void)
|
|
|
|
{
|
2012-09-07 16:00:48 +07:00
|
|
|
extent_state_cache = kmem_cache_create("btrfs_extent_state",
|
2009-04-13 20:33:09 +07:00
|
|
|
sizeof(struct extent_state), 0,
|
|
|
|
SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (!extent_state_cache)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2012-09-07 16:00:48 +07:00
|
|
|
extent_buffer_cache = kmem_cache_create("btrfs_extent_buffer",
|
2009-04-13 20:33:09 +07:00
|
|
|
sizeof(struct extent_buffer), 0,
|
|
|
|
SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (!extent_buffer_cache)
|
|
|
|
goto free_state_cache;
|
2013-05-18 05:30:14 +07:00
|
|
|
|
|
|
|
btrfs_bioset = bioset_create(BIO_POOL_SIZE,
|
|
|
|
offsetof(struct btrfs_io_bio, bio));
|
|
|
|
if (!btrfs_bioset)
|
|
|
|
goto free_buffer_cache;
|
2013-09-20 10:37:07 +07:00
|
|
|
|
|
|
|
if (bioset_integrity_create(btrfs_bioset, BIO_POOL_SIZE))
|
|
|
|
goto free_bioset;
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
return 0;
|
|
|
|
|
2013-09-20 10:37:07 +07:00
|
|
|
free_bioset:
|
|
|
|
bioset_free(btrfs_bioset);
|
|
|
|
btrfs_bioset = NULL;
|
|
|
|
|
2013-05-18 05:30:14 +07:00
|
|
|
free_buffer_cache:
|
|
|
|
kmem_cache_destroy(extent_buffer_cache);
|
|
|
|
extent_buffer_cache = NULL;
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
free_state_cache:
|
|
|
|
kmem_cache_destroy(extent_state_cache);
|
2013-05-18 05:30:14 +07:00
|
|
|
extent_state_cache = NULL;
|
2008-01-25 04:13:08 +07:00
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
void extent_io_exit(void)
|
|
|
|
{
|
2013-04-22 23:12:31 +07:00
|
|
|
btrfs_leak_debug_check();
|
2012-09-26 08:33:07 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Make sure all delayed rcu free are flushed before we
|
|
|
|
* destroy caches.
|
|
|
|
*/
|
|
|
|
rcu_barrier();
|
2008-01-25 04:13:08 +07:00
|
|
|
if (extent_state_cache)
|
|
|
|
kmem_cache_destroy(extent_state_cache);
|
|
|
|
if (extent_buffer_cache)
|
|
|
|
kmem_cache_destroy(extent_buffer_cache);
|
2013-05-18 05:30:14 +07:00
|
|
|
if (btrfs_bioset)
|
|
|
|
bioset_free(btrfs_bioset);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
void extent_io_tree_init(struct extent_io_tree *tree,
|
2011-04-21 04:35:57 +07:00
|
|
|
struct address_space *mapping)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2010-02-24 02:43:04 +07:00
|
|
|
tree->state = RB_ROOT;
|
2008-01-25 04:13:08 +07:00
|
|
|
tree->ops = NULL;
|
|
|
|
tree->dirty_bytes = 0;
|
2008-01-29 21:59:12 +07:00
|
|
|
spin_lock_init(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
tree->mapping = mapping;
|
|
|
|
}
|
|
|
|
|
2008-12-02 21:54:17 +07:00
|
|
|
static struct extent_state *alloc_extent_state(gfp_t mask)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct extent_state *state;
|
|
|
|
|
|
|
|
state = kmem_cache_alloc(extent_state_cache, mask);
|
2008-04-01 22:21:40 +07:00
|
|
|
if (!state)
|
2008-01-25 04:13:08 +07:00
|
|
|
return state;
|
|
|
|
state->state = 0;
|
|
|
|
state->private = 0;
|
2014-07-07 02:09:59 +07:00
|
|
|
RB_CLEAR_NODE(&state->rb_node);
|
2013-04-22 23:12:31 +07:00
|
|
|
btrfs_leak_debug_add(&state->leak_list, &states);
|
2008-01-25 04:13:08 +07:00
|
|
|
atomic_set(&state->refs, 1);
|
|
|
|
init_waitqueue_head(&state->wq);
|
2012-03-01 20:56:26 +07:00
|
|
|
trace_alloc_extent_state(state, mask, _RET_IP_);
|
2008-01-25 04:13:08 +07:00
|
|
|
return state;
|
|
|
|
}
|
|
|
|
|
2010-05-26 07:56:50 +07:00
|
|
|
void free_extent_state(struct extent_state *state)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
if (!state)
|
|
|
|
return;
|
|
|
|
if (atomic_dec_and_test(&state->refs)) {
|
2014-07-07 02:09:59 +07:00
|
|
|
WARN_ON(extent_state_in_tree(state));
|
2013-04-22 23:12:31 +07:00
|
|
|
btrfs_leak_debug_del(&state->leak_list);
|
2012-03-01 20:56:26 +07:00
|
|
|
trace_free_extent_state(state, _RET_IP_);
|
2008-01-25 04:13:08 +07:00
|
|
|
kmem_cache_free(extent_state_cache, state);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-02-12 22:05:53 +07:00
|
|
|
static struct rb_node *tree_insert(struct rb_root *root,
|
|
|
|
struct rb_node *search_start,
|
|
|
|
u64 offset,
|
2013-11-26 22:41:47 +07:00
|
|
|
struct rb_node *node,
|
|
|
|
struct rb_node ***p_in,
|
|
|
|
struct rb_node **parent_in)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2014-02-12 22:05:53 +07:00
|
|
|
struct rb_node **p;
|
2009-01-06 09:25:51 +07:00
|
|
|
struct rb_node *parent = NULL;
|
2008-01-25 04:13:08 +07:00
|
|
|
struct tree_entry *entry;
|
|
|
|
|
2013-11-26 22:41:47 +07:00
|
|
|
if (p_in && parent_in) {
|
|
|
|
p = *p_in;
|
|
|
|
parent = *parent_in;
|
|
|
|
goto do_insert;
|
|
|
|
}
|
|
|
|
|
2014-02-12 22:05:53 +07:00
|
|
|
p = search_start ? &search_start : &root->rb_node;
|
2009-01-06 09:25:51 +07:00
|
|
|
while (*p) {
|
2008-01-25 04:13:08 +07:00
|
|
|
parent = *p;
|
|
|
|
entry = rb_entry(parent, struct tree_entry, rb_node);
|
|
|
|
|
|
|
|
if (offset < entry->start)
|
|
|
|
p = &(*p)->rb_left;
|
|
|
|
else if (offset > entry->end)
|
|
|
|
p = &(*p)->rb_right;
|
|
|
|
else
|
|
|
|
return parent;
|
|
|
|
}
|
|
|
|
|
2013-11-26 22:41:47 +07:00
|
|
|
do_insert:
|
2008-01-25 04:13:08 +07:00
|
|
|
rb_link_node(node, parent, p);
|
|
|
|
rb_insert_color(node, root);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2008-02-02 02:51:59 +07:00
|
|
|
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
|
2013-11-26 22:41:47 +07:00
|
|
|
struct rb_node **prev_ret,
|
|
|
|
struct rb_node **next_ret,
|
|
|
|
struct rb_node ***p_ret,
|
|
|
|
struct rb_node **parent_ret)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2008-02-02 02:51:59 +07:00
|
|
|
struct rb_root *root = &tree->state;
|
2013-11-26 22:41:47 +07:00
|
|
|
struct rb_node **n = &root->rb_node;
|
2008-01-25 04:13:08 +07:00
|
|
|
struct rb_node *prev = NULL;
|
|
|
|
struct rb_node *orig_prev = NULL;
|
|
|
|
struct tree_entry *entry;
|
|
|
|
struct tree_entry *prev_entry = NULL;
|
|
|
|
|
2013-11-26 22:41:47 +07:00
|
|
|
while (*n) {
|
|
|
|
prev = *n;
|
|
|
|
entry = rb_entry(prev, struct tree_entry, rb_node);
|
2008-01-25 04:13:08 +07:00
|
|
|
prev_entry = entry;
|
|
|
|
|
|
|
|
if (offset < entry->start)
|
2013-11-26 22:41:47 +07:00
|
|
|
n = &(*n)->rb_left;
|
2008-01-25 04:13:08 +07:00
|
|
|
else if (offset > entry->end)
|
2013-11-26 22:41:47 +07:00
|
|
|
n = &(*n)->rb_right;
|
2009-01-06 09:25:51 +07:00
|
|
|
else
|
2013-11-26 22:41:47 +07:00
|
|
|
return *n;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2013-11-26 22:41:47 +07:00
|
|
|
if (p_ret)
|
|
|
|
*p_ret = n;
|
|
|
|
if (parent_ret)
|
|
|
|
*parent_ret = prev;
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
if (prev_ret) {
|
|
|
|
orig_prev = prev;
|
2009-01-06 09:25:51 +07:00
|
|
|
while (prev && offset > prev_entry->end) {
|
2008-01-25 04:13:08 +07:00
|
|
|
prev = rb_next(prev);
|
|
|
|
prev_entry = rb_entry(prev, struct tree_entry, rb_node);
|
|
|
|
}
|
|
|
|
*prev_ret = prev;
|
|
|
|
prev = orig_prev;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (next_ret) {
|
|
|
|
prev_entry = rb_entry(prev, struct tree_entry, rb_node);
|
2009-01-06 09:25:51 +07:00
|
|
|
while (prev && offset < prev_entry->start) {
|
2008-01-25 04:13:08 +07:00
|
|
|
prev = rb_prev(prev);
|
|
|
|
prev_entry = rb_entry(prev, struct tree_entry, rb_node);
|
|
|
|
}
|
|
|
|
*next_ret = prev;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2013-11-26 22:41:47 +07:00
|
|
|
static inline struct rb_node *
|
|
|
|
tree_search_for_insert(struct extent_io_tree *tree,
|
|
|
|
u64 offset,
|
|
|
|
struct rb_node ***p_ret,
|
|
|
|
struct rb_node **parent_ret)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2008-01-29 21:59:12 +07:00
|
|
|
struct rb_node *prev = NULL;
|
2008-01-25 04:13:08 +07:00
|
|
|
struct rb_node *ret;
|
2008-01-29 21:59:12 +07:00
|
|
|
|
2013-11-26 22:41:47 +07:00
|
|
|
ret = __etree_search(tree, offset, &prev, NULL, p_ret, parent_ret);
|
2009-01-06 09:25:51 +07:00
|
|
|
if (!ret)
|
2008-01-25 04:13:08 +07:00
|
|
|
return prev;
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-11-26 22:41:47 +07:00
|
|
|
static inline struct rb_node *tree_search(struct extent_io_tree *tree,
|
|
|
|
u64 offset)
|
|
|
|
{
|
|
|
|
return tree_search_for_insert(tree, offset, NULL, NULL);
|
|
|
|
}
|
|
|
|
|
2009-09-12 03:12:44 +07:00
|
|
|
static void merge_cb(struct extent_io_tree *tree, struct extent_state *new,
|
|
|
|
struct extent_state *other)
|
|
|
|
{
|
|
|
|
if (tree->ops && tree->ops->merge_extent_hook)
|
|
|
|
tree->ops->merge_extent_hook(tree->mapping->host, new,
|
|
|
|
other);
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* utility function to look for merge candidates inside a given range.
|
|
|
|
* Any extents with matching state are merged together into a single
|
|
|
|
* extent in the tree. Extents with EXTENT_IO in their state field
|
|
|
|
* are not merged because the end_io handlers need to be able to do
|
|
|
|
* operations on them without sleeping (or doing allocations/splits).
|
|
|
|
*
|
|
|
|
* This should be called with the tree lock held.
|
|
|
|
*/
|
2011-07-21 23:56:09 +07:00
|
|
|
static void merge_state(struct extent_io_tree *tree,
|
|
|
|
struct extent_state *state)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct extent_state *other;
|
|
|
|
struct rb_node *other_node;
|
|
|
|
|
2008-09-26 21:05:38 +07:00
|
|
|
if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
|
2011-07-21 23:56:09 +07:00
|
|
|
return;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
other_node = rb_prev(&state->rb_node);
|
|
|
|
if (other_node) {
|
|
|
|
other = rb_entry(other_node, struct extent_state, rb_node);
|
|
|
|
if (other->end == state->start - 1 &&
|
|
|
|
other->state == state->state) {
|
2009-09-12 03:12:44 +07:00
|
|
|
merge_cb(tree, state, other);
|
2008-01-25 04:13:08 +07:00
|
|
|
state->start = other->start;
|
|
|
|
rb_erase(&other->rb_node, &tree->state);
|
2014-07-07 02:09:59 +07:00
|
|
|
RB_CLEAR_NODE(&other->rb_node);
|
2008-01-25 04:13:08 +07:00
|
|
|
free_extent_state(other);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
other_node = rb_next(&state->rb_node);
|
|
|
|
if (other_node) {
|
|
|
|
other = rb_entry(other_node, struct extent_state, rb_node);
|
|
|
|
if (other->start == state->end + 1 &&
|
|
|
|
other->state == state->state) {
|
2009-09-12 03:12:44 +07:00
|
|
|
merge_cb(tree, state, other);
|
2011-06-21 01:53:48 +07:00
|
|
|
state->end = other->end;
|
|
|
|
rb_erase(&other->rb_node, &tree->state);
|
2014-07-07 02:09:59 +07:00
|
|
|
RB_CLEAR_NODE(&other->rb_node);
|
2011-06-21 01:53:48 +07:00
|
|
|
free_extent_state(other);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-07-21 23:56:09 +07:00
|
|
|
static void set_state_cb(struct extent_io_tree *tree,
|
2013-04-29 20:38:46 +07:00
|
|
|
struct extent_state *state, unsigned long *bits)
|
2008-01-30 03:55:23 +07:00
|
|
|
{
|
2011-07-21 23:56:09 +07:00
|
|
|
if (tree->ops && tree->ops->set_bit_hook)
|
|
|
|
tree->ops->set_bit_hook(tree->mapping->host, state, bits);
|
2008-01-30 03:55:23 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
static void clear_state_cb(struct extent_io_tree *tree,
|
2013-04-29 20:38:46 +07:00
|
|
|
struct extent_state *state, unsigned long *bits)
|
2008-01-30 03:55:23 +07:00
|
|
|
{
|
2009-09-12 03:12:44 +07:00
|
|
|
if (tree->ops && tree->ops->clear_bit_hook)
|
|
|
|
tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
|
2008-01-30 03:55:23 +07:00
|
|
|
}
|
|
|
|
|
2011-07-14 10:19:08 +07:00
|
|
|
static void set_state_bits(struct extent_io_tree *tree,
|
2013-04-29 20:38:46 +07:00
|
|
|
struct extent_state *state, unsigned long *bits);
|
2011-07-14 10:19:08 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* insert an extent_state struct into the tree. 'bits' are set on the
|
|
|
|
* struct before it is inserted.
|
|
|
|
*
|
|
|
|
* This may return -EEXIST if the extent is already there, in which case the
|
|
|
|
* state struct is freed.
|
|
|
|
*
|
|
|
|
* The tree lock is not taken internally. This is a utility function and
|
|
|
|
* probably isn't what you want to call (see set/clear_extent_bit).
|
|
|
|
*/
|
|
|
|
static int insert_state(struct extent_io_tree *tree,
|
|
|
|
struct extent_state *state, u64 start, u64 end,
|
2013-11-26 22:41:47 +07:00
|
|
|
struct rb_node ***p,
|
|
|
|
struct rb_node **parent,
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long *bits)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct rb_node *node;
|
|
|
|
|
2012-11-03 17:58:34 +07:00
|
|
|
if (end < start)
|
2013-12-20 23:37:06 +07:00
|
|
|
WARN(1, KERN_ERR "BTRFS: end < start %llu %llu\n",
|
2013-08-20 18:20:07 +07:00
|
|
|
end, start);
|
2008-01-25 04:13:08 +07:00
|
|
|
state->start = start;
|
|
|
|
state->end = end;
|
2009-09-12 03:12:44 +07:00
|
|
|
|
2011-07-14 10:19:08 +07:00
|
|
|
set_state_bits(tree, state, bits);
|
|
|
|
|
2014-02-12 22:05:53 +07:00
|
|
|
node = tree_insert(&tree->state, NULL, end, &state->rb_node, p, parent);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (node) {
|
|
|
|
struct extent_state *found;
|
|
|
|
found = rb_entry(node, struct extent_state, rb_node);
|
2013-12-20 23:37:06 +07:00
|
|
|
printk(KERN_ERR "BTRFS: found node %llu %llu on insert of "
|
2013-08-20 18:20:07 +07:00
|
|
|
"%llu %llu\n",
|
|
|
|
found->start, found->end, start, end);
|
2008-01-25 04:13:08 +07:00
|
|
|
return -EEXIST;
|
|
|
|
}
|
|
|
|
merge_state(tree, state);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2011-07-21 23:56:09 +07:00
|
|
|
static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
|
2009-09-12 03:12:44 +07:00
|
|
|
u64 split)
|
|
|
|
{
|
|
|
|
if (tree->ops && tree->ops->split_extent_hook)
|
2011-07-21 23:56:09 +07:00
|
|
|
tree->ops->split_extent_hook(tree->mapping->host, orig, split);
|
2009-09-12 03:12:44 +07:00
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* split a given extent state struct in two, inserting the preallocated
|
|
|
|
* struct 'prealloc' as the newly created second half. 'split' indicates an
|
|
|
|
* offset inside 'orig' where it should be split.
|
|
|
|
*
|
|
|
|
* Before calling,
|
|
|
|
* the tree has 'orig' at [orig->start, orig->end]. After calling, there
|
|
|
|
* are two extent state structs in the tree:
|
|
|
|
* prealloc: [orig->start, split - 1]
|
|
|
|
* orig: [ split, orig->end ]
|
|
|
|
*
|
|
|
|
* The tree locks are not taken by this function. They need to be held
|
|
|
|
* by the caller.
|
|
|
|
*/
|
|
|
|
static int split_state(struct extent_io_tree *tree, struct extent_state *orig,
|
|
|
|
struct extent_state *prealloc, u64 split)
|
|
|
|
{
|
|
|
|
struct rb_node *node;
|
2009-09-12 03:12:44 +07:00
|
|
|
|
|
|
|
split_cb(tree, orig, split);
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
prealloc->start = orig->start;
|
|
|
|
prealloc->end = split - 1;
|
|
|
|
prealloc->state = orig->state;
|
|
|
|
orig->start = split;
|
|
|
|
|
2014-02-12 22:05:53 +07:00
|
|
|
node = tree_insert(&tree->state, &orig->rb_node, prealloc->end,
|
|
|
|
&prealloc->rb_node, NULL, NULL);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (node) {
|
|
|
|
free_extent_state(prealloc);
|
|
|
|
return -EEXIST;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-03-12 15:39:48 +07:00
|
|
|
static struct extent_state *next_state(struct extent_state *state)
|
|
|
|
{
|
|
|
|
struct rb_node *next = rb_next(&state->rb_node);
|
|
|
|
if (next)
|
|
|
|
return rb_entry(next, struct extent_state, rb_node);
|
|
|
|
else
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* utility function to clear some bits in an extent state struct.
|
2012-04-06 13:35:18 +07:00
|
|
|
* it will optionally wake up any one waiting on this state (wake == 1).
|
2008-01-25 04:13:08 +07:00
|
|
|
*
|
|
|
|
* If no bits are set on the state struct after clearing things, the
|
|
|
|
* struct is freed and removed from the tree
|
|
|
|
*/
|
2012-03-12 15:39:48 +07:00
|
|
|
static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
|
|
|
|
struct extent_state *state,
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long *bits, int wake)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2012-03-12 15:39:48 +07:00
|
|
|
struct extent_state *next;
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long bits_to_clear = *bits & ~EXTENT_CTLBITS;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2010-05-16 21:48:47 +07:00
|
|
|
if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
|
2008-01-25 04:13:08 +07:00
|
|
|
u64 range = state->end - state->start + 1;
|
|
|
|
WARN_ON(range > tree->dirty_bytes);
|
|
|
|
tree->dirty_bytes -= range;
|
|
|
|
}
|
2008-01-30 03:55:23 +07:00
|
|
|
clear_state_cb(tree, state, bits);
|
2009-10-09 00:34:05 +07:00
|
|
|
state->state &= ~bits_to_clear;
|
2008-01-25 04:13:08 +07:00
|
|
|
if (wake)
|
|
|
|
wake_up(&state->wq);
|
2010-05-16 21:48:47 +07:00
|
|
|
if (state->state == 0) {
|
2012-03-12 15:39:48 +07:00
|
|
|
next = next_state(state);
|
2014-07-07 02:09:59 +07:00
|
|
|
if (extent_state_in_tree(state)) {
|
2008-01-25 04:13:08 +07:00
|
|
|
rb_erase(&state->rb_node, &tree->state);
|
2014-07-07 02:09:59 +07:00
|
|
|
RB_CLEAR_NODE(&state->rb_node);
|
2008-01-25 04:13:08 +07:00
|
|
|
free_extent_state(state);
|
|
|
|
} else {
|
|
|
|
WARN_ON(1);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
merge_state(tree, state);
|
2012-03-12 15:39:48 +07:00
|
|
|
next = next_state(state);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
2012-03-12 15:39:48 +07:00
|
|
|
return next;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2011-04-20 13:44:57 +07:00
|
|
|
static struct extent_state *
|
|
|
|
alloc_extent_state_atomic(struct extent_state *prealloc)
|
|
|
|
{
|
|
|
|
if (!prealloc)
|
|
|
|
prealloc = alloc_extent_state(GFP_ATOMIC);
|
|
|
|
|
|
|
|
return prealloc;
|
|
|
|
}
|
|
|
|
|
2013-04-26 03:41:01 +07:00
|
|
|
static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
|
2011-10-04 10:22:32 +07:00
|
|
|
{
|
|
|
|
btrfs_panic(tree_fs_info(tree), err, "Locking error: "
|
|
|
|
"Extent tree was modified by another "
|
|
|
|
"thread while locked.");
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* clear some bits on a range in the tree. This may require splitting
|
|
|
|
* or inserting elements in the tree, so the gfp mask is used to
|
|
|
|
* indicate which allocations or sleeping are allowed.
|
|
|
|
*
|
|
|
|
* pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove
|
|
|
|
* the given range from the tree regardless of state (ie for truncate).
|
|
|
|
*
|
|
|
|
* the range [start, end] is inclusive.
|
|
|
|
*
|
2012-03-01 20:56:29 +07:00
|
|
|
* This takes the tree lock, and returns 0 on success and < 0 on error.
|
2008-01-25 04:13:08 +07:00
|
|
|
*/
|
|
|
|
int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long bits, int wake, int delete,
|
2009-09-03 02:04:12 +07:00
|
|
|
struct extent_state **cached_state,
|
|
|
|
gfp_t mask)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct extent_state *state;
|
2009-09-03 02:04:12 +07:00
|
|
|
struct extent_state *cached;
|
2008-01-25 04:13:08 +07:00
|
|
|
struct extent_state *prealloc = NULL;
|
|
|
|
struct rb_node *node;
|
2009-05-27 20:16:03 +07:00
|
|
|
u64 last_end;
|
2008-01-25 04:13:08 +07:00
|
|
|
int err;
|
2010-02-04 02:33:23 +07:00
|
|
|
int clear = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2013-12-13 22:02:44 +07:00
|
|
|
btrfs_debug_check_extent_io_range(tree, start, end);
|
2013-04-30 22:22:23 +07:00
|
|
|
|
2013-06-22 03:37:03 +07:00
|
|
|
if (bits & EXTENT_DELALLOC)
|
|
|
|
bits |= EXTENT_NORESERVE;
|
|
|
|
|
2010-05-16 21:48:47 +07:00
|
|
|
if (delete)
|
|
|
|
bits |= ~EXTENT_CTLBITS;
|
|
|
|
bits |= EXTENT_FIRST_DELALLOC;
|
|
|
|
|
2010-02-04 02:33:23 +07:00
|
|
|
if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
|
|
|
|
clear = 1;
|
2008-01-25 04:13:08 +07:00
|
|
|
again:
|
|
|
|
if (!prealloc && (mask & __GFP_WAIT)) {
|
|
|
|
prealloc = alloc_extent_state(mask);
|
|
|
|
if (!prealloc)
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_lock(&tree->lock);
|
2009-09-03 02:04:12 +07:00
|
|
|
if (cached_state) {
|
|
|
|
cached = *cached_state;
|
2010-02-04 02:33:23 +07:00
|
|
|
|
|
|
|
if (clear) {
|
|
|
|
*cached_state = NULL;
|
|
|
|
cached_state = NULL;
|
|
|
|
}
|
|
|
|
|
2014-07-07 02:09:59 +07:00
|
|
|
if (cached && extent_state_in_tree(cached) &&
|
|
|
|
cached->start <= start && cached->end > start) {
|
2010-02-04 02:33:23 +07:00
|
|
|
if (clear)
|
|
|
|
atomic_dec(&cached->refs);
|
2009-09-03 02:04:12 +07:00
|
|
|
state = cached;
|
2009-09-24 06:51:09 +07:00
|
|
|
goto hit_next;
|
2009-09-03 02:04:12 +07:00
|
|
|
}
|
2010-02-04 02:33:23 +07:00
|
|
|
if (clear)
|
|
|
|
free_extent_state(cached);
|
2009-09-03 02:04:12 +07:00
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* this search will find the extents that end after
|
|
|
|
* our range starts
|
|
|
|
*/
|
2008-02-02 02:51:59 +07:00
|
|
|
node = tree_search(tree, start);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (!node)
|
|
|
|
goto out;
|
|
|
|
state = rb_entry(node, struct extent_state, rb_node);
|
2009-09-03 02:04:12 +07:00
|
|
|
hit_next:
|
2008-01-25 04:13:08 +07:00
|
|
|
if (state->start > end)
|
|
|
|
goto out;
|
|
|
|
WARN_ON(state->end < start);
|
2009-05-27 20:16:03 +07:00
|
|
|
last_end = state->end;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2012-02-16 17:34:37 +07:00
|
|
|
/* the state doesn't have the wanted bits, go ahead */
|
2012-03-12 15:39:48 +07:00
|
|
|
if (!(state->state & bits)) {
|
|
|
|
state = next_state(state);
|
2012-02-16 17:34:37 +07:00
|
|
|
goto next;
|
2012-03-12 15:39:48 +07:00
|
|
|
}
|
2012-02-16 17:34:37 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* | ---- desired range ---- |
|
|
|
|
* | state | or
|
|
|
|
* | ------------- state -------------- |
|
|
|
|
*
|
|
|
|
* We need to split the extent we found, and may flip
|
|
|
|
* bits on second half.
|
|
|
|
*
|
|
|
|
* If the extent we found extends past our range, we
|
|
|
|
* just split and search again. It'll get split again
|
|
|
|
* the next time though.
|
|
|
|
*
|
|
|
|
* If the extent we found is inside our range, we clear
|
|
|
|
* the desired bit on it.
|
|
|
|
*/
|
|
|
|
|
|
|
|
if (state->start < start) {
|
2011-04-20 13:44:57 +07:00
|
|
|
prealloc = alloc_extent_state_atomic(prealloc);
|
|
|
|
BUG_ON(!prealloc);
|
2008-01-25 04:13:08 +07:00
|
|
|
err = split_state(tree, state, prealloc, start);
|
2011-10-04 10:22:32 +07:00
|
|
|
if (err)
|
|
|
|
extent_io_tree_panic(tree, err);
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
prealloc = NULL;
|
|
|
|
if (err)
|
|
|
|
goto out;
|
|
|
|
if (state->end <= end) {
|
2012-05-10 17:10:39 +07:00
|
|
|
state = clear_state_bit(tree, state, &bits, wake);
|
|
|
|
goto next;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
goto search_again;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* | ---- desired range ---- |
|
|
|
|
* | state |
|
|
|
|
* We need to split the extent, and clear the bit
|
|
|
|
* on the first half
|
|
|
|
*/
|
|
|
|
if (state->start <= end && state->end > end) {
|
2011-04-20 13:44:57 +07:00
|
|
|
prealloc = alloc_extent_state_atomic(prealloc);
|
|
|
|
BUG_ON(!prealloc);
|
2008-01-25 04:13:08 +07:00
|
|
|
err = split_state(tree, state, prealloc, end + 1);
|
2011-10-04 10:22:32 +07:00
|
|
|
if (err)
|
|
|
|
extent_io_tree_panic(tree, err);
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
if (wake)
|
|
|
|
wake_up(&state->wq);
|
2009-09-24 06:51:09 +07:00
|
|
|
|
2012-03-01 20:56:29 +07:00
|
|
|
clear_state_bit(tree, prealloc, &bits, wake);
|
2009-09-12 03:12:44 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
prealloc = NULL;
|
|
|
|
goto out;
|
|
|
|
}
|
2009-09-24 06:51:09 +07:00
|
|
|
|
2012-03-12 15:39:48 +07:00
|
|
|
state = clear_state_bit(tree, state, &bits, wake);
|
2012-02-16 17:34:37 +07:00
|
|
|
next:
|
2009-05-27 20:16:03 +07:00
|
|
|
if (last_end == (u64)-1)
|
|
|
|
goto out;
|
|
|
|
start = last_end + 1;
|
2012-03-12 15:39:48 +07:00
|
|
|
if (start <= end && state && !need_resched())
|
2012-02-16 17:34:36 +07:00
|
|
|
goto hit_next;
|
2008-01-25 04:13:08 +07:00
|
|
|
goto search_again;
|
|
|
|
|
|
|
|
out:
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (prealloc)
|
|
|
|
free_extent_state(prealloc);
|
|
|
|
|
2012-03-01 20:56:29 +07:00
|
|
|
return 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
search_again:
|
|
|
|
if (start > end)
|
|
|
|
goto out;
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (mask & __GFP_WAIT)
|
|
|
|
cond_resched();
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
|
2012-03-01 20:56:26 +07:00
|
|
|
static void wait_on_state(struct extent_io_tree *tree,
|
|
|
|
struct extent_state *state)
|
2008-12-02 18:36:10 +07:00
|
|
|
__releases(tree->lock)
|
|
|
|
__acquires(tree->lock)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
DEFINE_WAIT(wait);
|
|
|
|
prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
schedule();
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_lock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
finish_wait(&state->wq, &wait);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* waits for one or more bits to clear on a range in the state tree.
|
|
|
|
* The range [start, end] is inclusive.
|
|
|
|
* The tree lock is taken by this function
|
|
|
|
*/
|
2013-04-29 20:38:46 +07:00
|
|
|
static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
|
|
|
|
unsigned long bits)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct extent_state *state;
|
|
|
|
struct rb_node *node;
|
|
|
|
|
2013-12-13 22:02:44 +07:00
|
|
|
btrfs_debug_check_extent_io_range(tree, start, end);
|
2013-04-30 22:22:23 +07:00
|
|
|
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_lock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
again:
|
|
|
|
while (1) {
|
|
|
|
/*
|
|
|
|
* this search will find all the extents that end after
|
|
|
|
* our range starts
|
|
|
|
*/
|
2008-02-02 02:51:59 +07:00
|
|
|
node = tree_search(tree, start);
|
2014-03-31 20:53:25 +07:00
|
|
|
process_node:
|
2008-01-25 04:13:08 +07:00
|
|
|
if (!node)
|
|
|
|
break;
|
|
|
|
|
|
|
|
state = rb_entry(node, struct extent_state, rb_node);
|
|
|
|
|
|
|
|
if (state->start > end)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (state->state & bits) {
|
|
|
|
start = state->start;
|
|
|
|
atomic_inc(&state->refs);
|
|
|
|
wait_on_state(tree, state);
|
|
|
|
free_extent_state(state);
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
start = state->end + 1;
|
|
|
|
|
|
|
|
if (start > end)
|
|
|
|
break;
|
|
|
|
|
2014-03-31 20:53:25 +07:00
|
|
|
if (!cond_resched_lock(&tree->lock)) {
|
|
|
|
node = rb_next(node);
|
|
|
|
goto process_node;
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
out:
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2011-07-21 23:56:09 +07:00
|
|
|
static void set_state_bits(struct extent_io_tree *tree,
|
2008-01-25 04:13:08 +07:00
|
|
|
struct extent_state *state,
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long *bits)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long bits_to_set = *bits & ~EXTENT_CTLBITS;
|
2009-09-12 03:12:44 +07:00
|
|
|
|
2011-07-21 23:56:09 +07:00
|
|
|
set_state_cb(tree, state, bits);
|
2010-05-16 21:48:47 +07:00
|
|
|
if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
|
2008-01-25 04:13:08 +07:00
|
|
|
u64 range = state->end - state->start + 1;
|
|
|
|
tree->dirty_bytes += range;
|
|
|
|
}
|
2010-05-16 21:48:47 +07:00
|
|
|
state->state |= bits_to_set;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2009-09-03 02:04:12 +07:00
|
|
|
static void cache_state(struct extent_state *state,
|
|
|
|
struct extent_state **cached_ptr)
|
|
|
|
{
|
|
|
|
if (cached_ptr && !(*cached_ptr)) {
|
|
|
|
if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY)) {
|
|
|
|
*cached_ptr = state;
|
|
|
|
atomic_inc(&state->refs);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
2009-09-03 00:24:36 +07:00
|
|
|
* set some bits on a range in the tree. This may require allocations or
|
|
|
|
* sleeping, so the gfp mask is used to indicate what is allowed.
|
2008-01-25 04:13:08 +07:00
|
|
|
*
|
2009-09-03 00:24:36 +07:00
|
|
|
* If any of the exclusive bits are set, this will fail with -EEXIST if some
|
|
|
|
* part of the range already has the desired bits set. The start of the
|
|
|
|
* existing range is returned in failed_start in this case.
|
2008-01-25 04:13:08 +07:00
|
|
|
*
|
2009-09-03 00:24:36 +07:00
|
|
|
* [start, end] is inclusive This takes the tree lock.
|
2008-01-25 04:13:08 +07:00
|
|
|
*/
|
2009-09-03 00:24:36 +07:00
|
|
|
|
2012-03-01 20:57:19 +07:00
|
|
|
static int __must_check
|
|
|
|
__set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long bits, unsigned long exclusive_bits,
|
|
|
|
u64 *failed_start, struct extent_state **cached_state,
|
|
|
|
gfp_t mask)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct extent_state *state;
|
|
|
|
struct extent_state *prealloc = NULL;
|
|
|
|
struct rb_node *node;
|
2013-11-26 22:41:47 +07:00
|
|
|
struct rb_node **p;
|
|
|
|
struct rb_node *parent;
|
2008-01-25 04:13:08 +07:00
|
|
|
int err = 0;
|
|
|
|
u64 last_start;
|
|
|
|
u64 last_end;
|
2009-09-24 06:51:09 +07:00
|
|
|
|
2013-12-13 22:02:44 +07:00
|
|
|
btrfs_debug_check_extent_io_range(tree, start, end);
|
2013-04-30 22:22:23 +07:00
|
|
|
|
2010-05-16 21:48:47 +07:00
|
|
|
bits |= EXTENT_FIRST_DELALLOC;
|
2008-01-25 04:13:08 +07:00
|
|
|
again:
|
|
|
|
if (!prealloc && (mask & __GFP_WAIT)) {
|
|
|
|
prealloc = alloc_extent_state(mask);
|
2011-04-20 13:44:57 +07:00
|
|
|
BUG_ON(!prealloc);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_lock(&tree->lock);
|
2009-09-03 02:22:30 +07:00
|
|
|
if (cached_state && *cached_state) {
|
|
|
|
state = *cached_state;
|
2011-06-21 01:53:48 +07:00
|
|
|
if (state->start <= start && state->end > start &&
|
2014-07-07 02:09:59 +07:00
|
|
|
extent_state_in_tree(state)) {
|
2009-09-03 02:22:30 +07:00
|
|
|
node = &state->rb_node;
|
|
|
|
goto hit_next;
|
|
|
|
}
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* this search will find all the extents that end after
|
|
|
|
* our range starts.
|
|
|
|
*/
|
2013-11-26 22:41:47 +07:00
|
|
|
node = tree_search_for_insert(tree, start, &p, &parent);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (!node) {
|
2011-04-20 13:44:57 +07:00
|
|
|
prealloc = alloc_extent_state_atomic(prealloc);
|
|
|
|
BUG_ON(!prealloc);
|
2013-11-26 22:41:47 +07:00
|
|
|
err = insert_state(tree, prealloc, start, end,
|
|
|
|
&p, &parent, &bits);
|
2011-10-04 10:22:32 +07:00
|
|
|
if (err)
|
|
|
|
extent_io_tree_panic(tree, err);
|
|
|
|
|
2013-11-26 22:01:34 +07:00
|
|
|
cache_state(prealloc, cached_state);
|
2008-01-25 04:13:08 +07:00
|
|
|
prealloc = NULL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
state = rb_entry(node, struct extent_state, rb_node);
|
2009-08-05 23:57:59 +07:00
|
|
|
hit_next:
|
2008-01-25 04:13:08 +07:00
|
|
|
last_start = state->start;
|
|
|
|
last_end = state->end;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* | ---- desired range ---- |
|
|
|
|
* | state |
|
|
|
|
*
|
|
|
|
* Just lock what we found and keep going
|
|
|
|
*/
|
|
|
|
if (state->start == start && state->end <= end) {
|
2009-09-03 00:24:36 +07:00
|
|
|
if (state->state & exclusive_bits) {
|
2008-01-25 04:13:08 +07:00
|
|
|
*failed_start = state->start;
|
|
|
|
err = -EEXIST;
|
|
|
|
goto out;
|
|
|
|
}
|
2009-09-24 06:51:09 +07:00
|
|
|
|
2011-07-21 23:56:09 +07:00
|
|
|
set_state_bits(tree, state, &bits);
|
2009-09-03 02:04:12 +07:00
|
|
|
cache_state(state, cached_state);
|
2008-01-25 04:13:08 +07:00
|
|
|
merge_state(tree, state);
|
2009-05-27 20:16:03 +07:00
|
|
|
if (last_end == (u64)-1)
|
|
|
|
goto out;
|
|
|
|
start = last_end + 1;
|
2012-05-10 17:10:39 +07:00
|
|
|
state = next_state(state);
|
|
|
|
if (start < end && state && state->start == start &&
|
|
|
|
!need_resched())
|
|
|
|
goto hit_next;
|
2008-01-25 04:13:08 +07:00
|
|
|
goto search_again;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* | ---- desired range ---- |
|
|
|
|
* | state |
|
|
|
|
* or
|
|
|
|
* | ------------- state -------------- |
|
|
|
|
*
|
|
|
|
* We need to split the extent we found, and may flip bits on
|
|
|
|
* second half.
|
|
|
|
*
|
|
|
|
* If the extent we found extends past our
|
|
|
|
* range, we just split and search again. It'll get split
|
|
|
|
* again the next time though.
|
|
|
|
*
|
|
|
|
* If the extent we found is inside our range, we set the
|
|
|
|
* desired bit on it.
|
|
|
|
*/
|
|
|
|
if (state->start < start) {
|
2009-09-03 00:24:36 +07:00
|
|
|
if (state->state & exclusive_bits) {
|
2008-01-25 04:13:08 +07:00
|
|
|
*failed_start = start;
|
|
|
|
err = -EEXIST;
|
|
|
|
goto out;
|
|
|
|
}
|
2011-04-20 13:44:57 +07:00
|
|
|
|
|
|
|
prealloc = alloc_extent_state_atomic(prealloc);
|
|
|
|
BUG_ON(!prealloc);
|
2008-01-25 04:13:08 +07:00
|
|
|
err = split_state(tree, state, prealloc, start);
|
2011-10-04 10:22:32 +07:00
|
|
|
if (err)
|
|
|
|
extent_io_tree_panic(tree, err);
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
prealloc = NULL;
|
|
|
|
if (err)
|
|
|
|
goto out;
|
|
|
|
if (state->end <= end) {
|
2011-07-21 23:56:09 +07:00
|
|
|
set_state_bits(tree, state, &bits);
|
2009-09-03 02:04:12 +07:00
|
|
|
cache_state(state, cached_state);
|
2008-01-25 04:13:08 +07:00
|
|
|
merge_state(tree, state);
|
2009-05-27 20:16:03 +07:00
|
|
|
if (last_end == (u64)-1)
|
|
|
|
goto out;
|
|
|
|
start = last_end + 1;
|
2012-05-10 17:10:39 +07:00
|
|
|
state = next_state(state);
|
|
|
|
if (start < end && state && state->start == start &&
|
|
|
|
!need_resched())
|
|
|
|
goto hit_next;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
goto search_again;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* | ---- desired range ---- |
|
|
|
|
* | state | or | state |
|
|
|
|
*
|
|
|
|
* There's a hole, we need to insert something in it and
|
|
|
|
* ignore the extent we found.
|
|
|
|
*/
|
|
|
|
if (state->start > start) {
|
|
|
|
u64 this_end;
|
|
|
|
if (end < last_start)
|
|
|
|
this_end = end;
|
|
|
|
else
|
2009-01-06 09:25:51 +07:00
|
|
|
this_end = last_start - 1;
|
2011-04-20 13:44:57 +07:00
|
|
|
|
|
|
|
prealloc = alloc_extent_state_atomic(prealloc);
|
|
|
|
BUG_ON(!prealloc);
|
2011-04-20 13:45:49 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Avoid to free 'prealloc' if it can be merged with
|
|
|
|
* the later extent.
|
|
|
|
*/
|
2008-01-25 04:13:08 +07:00
|
|
|
err = insert_state(tree, prealloc, start, this_end,
|
2013-11-26 22:41:47 +07:00
|
|
|
NULL, NULL, &bits);
|
2011-10-04 10:22:32 +07:00
|
|
|
if (err)
|
|
|
|
extent_io_tree_panic(tree, err);
|
|
|
|
|
2009-09-12 03:12:44 +07:00
|
|
|
cache_state(prealloc, cached_state);
|
|
|
|
prealloc = NULL;
|
2008-01-25 04:13:08 +07:00
|
|
|
start = this_end + 1;
|
|
|
|
goto search_again;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* | ---- desired range ---- |
|
|
|
|
* | state |
|
|
|
|
* We need to split the extent, and set the bit
|
|
|
|
* on the first half
|
|
|
|
*/
|
|
|
|
if (state->start <= end && state->end > end) {
|
2009-09-03 00:24:36 +07:00
|
|
|
if (state->state & exclusive_bits) {
|
2008-01-25 04:13:08 +07:00
|
|
|
*failed_start = start;
|
|
|
|
err = -EEXIST;
|
|
|
|
goto out;
|
|
|
|
}
|
2011-04-20 13:44:57 +07:00
|
|
|
|
|
|
|
prealloc = alloc_extent_state_atomic(prealloc);
|
|
|
|
BUG_ON(!prealloc);
|
2008-01-25 04:13:08 +07:00
|
|
|
err = split_state(tree, state, prealloc, end + 1);
|
2011-10-04 10:22:32 +07:00
|
|
|
if (err)
|
|
|
|
extent_io_tree_panic(tree, err);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2011-07-21 23:56:09 +07:00
|
|
|
set_state_bits(tree, prealloc, &bits);
|
2009-09-03 02:04:12 +07:00
|
|
|
cache_state(prealloc, cached_state);
|
2008-01-25 04:13:08 +07:00
|
|
|
merge_state(tree, prealloc);
|
|
|
|
prealloc = NULL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
goto search_again;
|
|
|
|
|
|
|
|
out:
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (prealloc)
|
|
|
|
free_extent_state(prealloc);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
|
|
|
search_again:
|
|
|
|
if (start > end)
|
|
|
|
goto out;
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (mask & __GFP_WAIT)
|
|
|
|
cond_resched();
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
|
2013-04-29 20:38:46 +07:00
|
|
|
int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
|
|
|
|
unsigned long bits, u64 * failed_start,
|
|
|
|
struct extent_state **cached_state, gfp_t mask)
|
2012-03-01 20:57:19 +07:00
|
|
|
{
|
|
|
|
return __set_extent_bit(tree, start, end, bits, 0, failed_start,
|
|
|
|
cached_state, mask);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2011-09-27 00:56:12 +07:00
|
|
|
/**
|
2012-07-11 14:26:19 +07:00
|
|
|
* convert_extent_bit - convert all bits in a given range from one bit to
|
|
|
|
* another
|
2011-09-27 00:56:12 +07:00
|
|
|
* @tree: the io tree to search
|
|
|
|
* @start: the start offset in bytes
|
|
|
|
* @end: the end offset in bytes (inclusive)
|
|
|
|
* @bits: the bits to set in this range
|
|
|
|
* @clear_bits: the bits to clear in this range
|
2012-09-28 04:07:30 +07:00
|
|
|
* @cached_state: state that we're going to cache
|
2011-09-27 00:56:12 +07:00
|
|
|
* @mask: the allocation mask
|
|
|
|
*
|
|
|
|
* This will go through and set bits for the given range. If any states exist
|
|
|
|
* already in this range they are set with the given bit and cleared of the
|
|
|
|
* clear_bits. This is only meant to be used by things that are mergeable, ie
|
|
|
|
* converting from say DELALLOC to DIRTY. This is not meant to be used with
|
|
|
|
* boundary bits like LOCK.
|
|
|
|
*/
|
|
|
|
int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long bits, unsigned long clear_bits,
|
2012-09-28 04:07:30 +07:00
|
|
|
struct extent_state **cached_state, gfp_t mask)
|
2011-09-27 00:56:12 +07:00
|
|
|
{
|
|
|
|
struct extent_state *state;
|
|
|
|
struct extent_state *prealloc = NULL;
|
|
|
|
struct rb_node *node;
|
2013-11-26 22:41:47 +07:00
|
|
|
struct rb_node **p;
|
|
|
|
struct rb_node *parent;
|
2011-09-27 00:56:12 +07:00
|
|
|
int err = 0;
|
|
|
|
u64 last_start;
|
|
|
|
u64 last_end;
|
|
|
|
|
2013-12-13 22:02:44 +07:00
|
|
|
btrfs_debug_check_extent_io_range(tree, start, end);
|
2013-04-30 22:22:23 +07:00
|
|
|
|
2011-09-27 00:56:12 +07:00
|
|
|
again:
|
|
|
|
if (!prealloc && (mask & __GFP_WAIT)) {
|
|
|
|
prealloc = alloc_extent_state(mask);
|
|
|
|
if (!prealloc)
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_lock(&tree->lock);
|
2012-09-28 04:07:30 +07:00
|
|
|
if (cached_state && *cached_state) {
|
|
|
|
state = *cached_state;
|
|
|
|
if (state->start <= start && state->end > start &&
|
2014-07-07 02:09:59 +07:00
|
|
|
extent_state_in_tree(state)) {
|
2012-09-28 04:07:30 +07:00
|
|
|
node = &state->rb_node;
|
|
|
|
goto hit_next;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-09-27 00:56:12 +07:00
|
|
|
/*
|
|
|
|
* this search will find all the extents that end after
|
|
|
|
* our range starts.
|
|
|
|
*/
|
2013-11-26 22:41:47 +07:00
|
|
|
node = tree_search_for_insert(tree, start, &p, &parent);
|
2011-09-27 00:56:12 +07:00
|
|
|
if (!node) {
|
|
|
|
prealloc = alloc_extent_state_atomic(prealloc);
|
2011-12-08 08:08:40 +07:00
|
|
|
if (!prealloc) {
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
2013-11-26 22:41:47 +07:00
|
|
|
err = insert_state(tree, prealloc, start, end,
|
|
|
|
&p, &parent, &bits);
|
2011-10-04 10:22:32 +07:00
|
|
|
if (err)
|
|
|
|
extent_io_tree_panic(tree, err);
|
2013-11-26 22:01:34 +07:00
|
|
|
cache_state(prealloc, cached_state);
|
|
|
|
prealloc = NULL;
|
2011-09-27 00:56:12 +07:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
state = rb_entry(node, struct extent_state, rb_node);
|
|
|
|
hit_next:
|
|
|
|
last_start = state->start;
|
|
|
|
last_end = state->end;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* | ---- desired range ---- |
|
|
|
|
* | state |
|
|
|
|
*
|
|
|
|
* Just lock what we found and keep going
|
|
|
|
*/
|
|
|
|
if (state->start == start && state->end <= end) {
|
|
|
|
set_state_bits(tree, state, &bits);
|
2012-09-28 04:07:30 +07:00
|
|
|
cache_state(state, cached_state);
|
2012-05-10 17:10:39 +07:00
|
|
|
state = clear_state_bit(tree, state, &clear_bits, 0);
|
2011-09-27 00:56:12 +07:00
|
|
|
if (last_end == (u64)-1)
|
|
|
|
goto out;
|
|
|
|
start = last_end + 1;
|
2012-05-10 17:10:39 +07:00
|
|
|
if (start < end && state && state->start == start &&
|
|
|
|
!need_resched())
|
|
|
|
goto hit_next;
|
2011-09-27 00:56:12 +07:00
|
|
|
goto search_again;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* | ---- desired range ---- |
|
|
|
|
* | state |
|
|
|
|
* or
|
|
|
|
* | ------------- state -------------- |
|
|
|
|
*
|
|
|
|
* We need to split the extent we found, and may flip bits on
|
|
|
|
* second half.
|
|
|
|
*
|
|
|
|
* If the extent we found extends past our
|
|
|
|
* range, we just split and search again. It'll get split
|
|
|
|
* again the next time though.
|
|
|
|
*
|
|
|
|
* If the extent we found is inside our range, we set the
|
|
|
|
* desired bit on it.
|
|
|
|
*/
|
|
|
|
if (state->start < start) {
|
|
|
|
prealloc = alloc_extent_state_atomic(prealloc);
|
2011-12-08 08:08:40 +07:00
|
|
|
if (!prealloc) {
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
2011-09-27 00:56:12 +07:00
|
|
|
err = split_state(tree, state, prealloc, start);
|
2011-10-04 10:22:32 +07:00
|
|
|
if (err)
|
|
|
|
extent_io_tree_panic(tree, err);
|
2011-09-27 00:56:12 +07:00
|
|
|
prealloc = NULL;
|
|
|
|
if (err)
|
|
|
|
goto out;
|
|
|
|
if (state->end <= end) {
|
|
|
|
set_state_bits(tree, state, &bits);
|
2012-09-28 04:07:30 +07:00
|
|
|
cache_state(state, cached_state);
|
2012-05-10 17:10:39 +07:00
|
|
|
state = clear_state_bit(tree, state, &clear_bits, 0);
|
2011-09-27 00:56:12 +07:00
|
|
|
if (last_end == (u64)-1)
|
|
|
|
goto out;
|
|
|
|
start = last_end + 1;
|
2012-05-10 17:10:39 +07:00
|
|
|
if (start < end && state && state->start == start &&
|
|
|
|
!need_resched())
|
|
|
|
goto hit_next;
|
2011-09-27 00:56:12 +07:00
|
|
|
}
|
|
|
|
goto search_again;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* | ---- desired range ---- |
|
|
|
|
* | state | or | state |
|
|
|
|
*
|
|
|
|
* There's a hole, we need to insert something in it and
|
|
|
|
* ignore the extent we found.
|
|
|
|
*/
|
|
|
|
if (state->start > start) {
|
|
|
|
u64 this_end;
|
|
|
|
if (end < last_start)
|
|
|
|
this_end = end;
|
|
|
|
else
|
|
|
|
this_end = last_start - 1;
|
|
|
|
|
|
|
|
prealloc = alloc_extent_state_atomic(prealloc);
|
2011-12-08 08:08:40 +07:00
|
|
|
if (!prealloc) {
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
2011-09-27 00:56:12 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Avoid to free 'prealloc' if it can be merged with
|
|
|
|
* the later extent.
|
|
|
|
*/
|
|
|
|
err = insert_state(tree, prealloc, start, this_end,
|
2013-11-26 22:41:47 +07:00
|
|
|
NULL, NULL, &bits);
|
2011-10-04 10:22:32 +07:00
|
|
|
if (err)
|
|
|
|
extent_io_tree_panic(tree, err);
|
2012-09-28 04:07:30 +07:00
|
|
|
cache_state(prealloc, cached_state);
|
2011-09-27 00:56:12 +07:00
|
|
|
prealloc = NULL;
|
|
|
|
start = this_end + 1;
|
|
|
|
goto search_again;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* | ---- desired range ---- |
|
|
|
|
* | state |
|
|
|
|
* We need to split the extent, and set the bit
|
|
|
|
* on the first half
|
|
|
|
*/
|
|
|
|
if (state->start <= end && state->end > end) {
|
|
|
|
prealloc = alloc_extent_state_atomic(prealloc);
|
2011-12-08 08:08:40 +07:00
|
|
|
if (!prealloc) {
|
|
|
|
err = -ENOMEM;
|
|
|
|
goto out;
|
|
|
|
}
|
2011-09-27 00:56:12 +07:00
|
|
|
|
|
|
|
err = split_state(tree, state, prealloc, end + 1);
|
2011-10-04 10:22:32 +07:00
|
|
|
if (err)
|
|
|
|
extent_io_tree_panic(tree, err);
|
2011-09-27 00:56:12 +07:00
|
|
|
|
|
|
|
set_state_bits(tree, prealloc, &bits);
|
2012-09-28 04:07:30 +07:00
|
|
|
cache_state(prealloc, cached_state);
|
2011-09-27 00:56:12 +07:00
|
|
|
clear_state_bit(tree, prealloc, &clear_bits, 0);
|
|
|
|
prealloc = NULL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
goto search_again;
|
|
|
|
|
|
|
|
out:
|
|
|
|
spin_unlock(&tree->lock);
|
|
|
|
if (prealloc)
|
|
|
|
free_extent_state(prealloc);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
|
|
|
search_again:
|
|
|
|
if (start > end)
|
|
|
|
goto out;
|
|
|
|
spin_unlock(&tree->lock);
|
|
|
|
if (mask & __GFP_WAIT)
|
|
|
|
cond_resched();
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/* wrappers around set/clear extent bit */
|
|
|
|
int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
|
|
|
|
gfp_t mask)
|
|
|
|
{
|
2012-03-01 20:57:19 +07:00
|
|
|
return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
|
2009-09-03 02:04:12 +07:00
|
|
|
NULL, mask);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long bits, gfp_t mask)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2012-03-01 20:57:19 +07:00
|
|
|
return set_extent_bit(tree, start, end, bits, NULL,
|
2009-09-03 02:04:12 +07:00
|
|
|
NULL, mask);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long bits, gfp_t mask)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2009-09-03 02:04:12 +07:00
|
|
|
return clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
|
2010-02-04 02:33:23 +07:00
|
|
|
struct extent_state **cached_state, gfp_t mask)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
return set_extent_bit(tree, start, end,
|
2011-09-29 14:55:28 +07:00
|
|
|
EXTENT_DELALLOC | EXTENT_UPTODATE,
|
2012-03-01 20:57:19 +07:00
|
|
|
NULL, cached_state, mask);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2012-09-06 08:10:51 +07:00
|
|
|
int set_extent_defrag(struct extent_io_tree *tree, u64 start, u64 end,
|
|
|
|
struct extent_state **cached_state, gfp_t mask)
|
|
|
|
{
|
|
|
|
return set_extent_bit(tree, start, end,
|
|
|
|
EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
|
|
|
|
NULL, cached_state, mask);
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
|
|
|
|
gfp_t mask)
|
|
|
|
{
|
|
|
|
return clear_extent_bit(tree, start, end,
|
2009-10-09 00:34:05 +07:00
|
|
|
EXTENT_DIRTY | EXTENT_DELALLOC |
|
2010-05-16 21:48:47 +07:00
|
|
|
EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
|
|
|
|
gfp_t mask)
|
|
|
|
{
|
2012-03-01 20:57:19 +07:00
|
|
|
return set_extent_bit(tree, start, end, EXTENT_NEW, NULL,
|
2009-09-03 02:04:12 +07:00
|
|
|
NULL, mask);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
|
2011-04-06 17:02:20 +07:00
|
|
|
struct extent_state **cached_state, gfp_t mask)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2013-03-28 15:30:28 +07:00
|
|
|
return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
|
2012-03-01 20:57:19 +07:00
|
|
|
cached_state, mask);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2012-05-03 01:00:54 +07:00
|
|
|
int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
|
|
|
|
struct extent_state **cached_state, gfp_t mask)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2009-09-03 02:04:12 +07:00
|
|
|
return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
|
2010-02-04 02:33:23 +07:00
|
|
|
cached_state, mask);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2008-09-30 02:18:18 +07:00
|
|
|
/*
|
|
|
|
* either insert or lock state struct between start and end use mask to tell
|
|
|
|
* us if waiting is desired.
|
|
|
|
*/
|
2009-09-03 00:24:36 +07:00
|
|
|
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long bits, struct extent_state **cached_state)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
int err;
|
|
|
|
u64 failed_start;
|
|
|
|
while (1) {
|
2012-03-01 20:57:19 +07:00
|
|
|
err = __set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
|
|
|
|
EXTENT_LOCKED, &failed_start,
|
|
|
|
cached_state, GFP_NOFS);
|
2012-03-01 20:57:19 +07:00
|
|
|
if (err == -EEXIST) {
|
2008-01-25 04:13:08 +07:00
|
|
|
wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
|
|
|
|
start = failed_start;
|
2012-03-01 20:57:19 +07:00
|
|
|
} else
|
2008-01-25 04:13:08 +07:00
|
|
|
break;
|
|
|
|
WARN_ON(start > end);
|
|
|
|
}
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2012-03-01 20:57:19 +07:00
|
|
|
int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
|
2009-09-03 00:24:36 +07:00
|
|
|
{
|
2012-03-01 20:57:19 +07:00
|
|
|
return lock_extent_bits(tree, start, end, 0, NULL);
|
2009-09-03 00:24:36 +07:00
|
|
|
}
|
|
|
|
|
2012-03-01 20:57:19 +07:00
|
|
|
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-30 01:49:05 +07:00
|
|
|
{
|
|
|
|
int err;
|
|
|
|
u64 failed_start;
|
|
|
|
|
2012-03-01 20:57:19 +07:00
|
|
|
err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
|
|
|
|
&failed_start, NULL, GFP_NOFS);
|
2008-10-31 01:19:50 +07:00
|
|
|
if (err == -EEXIST) {
|
|
|
|
if (failed_start > start)
|
|
|
|
clear_extent_bit(tree, start, failed_start - 1,
|
2012-03-01 20:57:19 +07:00
|
|
|
EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-30 01:49:05 +07:00
|
|
|
return 0;
|
2008-10-31 01:19:50 +07:00
|
|
|
}
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-30 01:49:05 +07:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2009-09-03 02:04:12 +07:00
|
|
|
int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end,
|
|
|
|
struct extent_state **cached, gfp_t mask)
|
|
|
|
{
|
|
|
|
return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
|
|
|
|
mask);
|
|
|
|
}
|
|
|
|
|
2012-03-01 20:57:19 +07:00
|
|
|
int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2009-09-03 02:04:12 +07:00
|
|
|
return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
|
2012-03-01 20:57:19 +07:00
|
|
|
GFP_NOFS);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2013-03-27 00:07:00 +07:00
|
|
|
int extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end)
|
|
|
|
{
|
|
|
|
unsigned long index = start >> PAGE_CACHE_SHIFT;
|
|
|
|
unsigned long end_index = end >> PAGE_CACHE_SHIFT;
|
|
|
|
struct page *page;
|
|
|
|
|
|
|
|
while (index <= end_index) {
|
|
|
|
page = find_get_page(inode->i_mapping, index);
|
|
|
|
BUG_ON(!page); /* Pages should be in the extent_io_tree */
|
|
|
|
clear_page_dirty_for_io(page);
|
|
|
|
page_cache_release(page);
|
|
|
|
index++;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end)
|
|
|
|
{
|
|
|
|
unsigned long index = start >> PAGE_CACHE_SHIFT;
|
|
|
|
unsigned long end_index = end >> PAGE_CACHE_SHIFT;
|
|
|
|
struct page *page;
|
|
|
|
|
|
|
|
while (index <= end_index) {
|
|
|
|
page = find_get_page(inode->i_mapping, index);
|
|
|
|
BUG_ON(!page); /* Pages should be in the extent_io_tree */
|
|
|
|
account_page_redirty(page);
|
|
|
|
__set_page_dirty_nobuffers(page);
|
|
|
|
page_cache_release(page);
|
|
|
|
index++;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* helper function to set both pages and extents in the tree writeback
|
|
|
|
*/
|
2008-12-02 21:54:17 +07:00
|
|
|
static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
unsigned long index = start >> PAGE_CACHE_SHIFT;
|
|
|
|
unsigned long end_index = end >> PAGE_CACHE_SHIFT;
|
|
|
|
struct page *page;
|
|
|
|
|
|
|
|
while (index <= end_index) {
|
|
|
|
page = find_get_page(tree->mapping, index);
|
2012-03-12 22:03:00 +07:00
|
|
|
BUG_ON(!page); /* Pages should be in the extent_io_tree */
|
2008-01-25 04:13:08 +07:00
|
|
|
set_page_writeback(page);
|
|
|
|
page_cache_release(page);
|
|
|
|
index++;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-09-30 02:18:18 +07:00
|
|
|
/* find the first state struct with 'bits' set after 'start', and
|
|
|
|
* return it. tree->lock must be held. NULL will returned if
|
|
|
|
* nothing was found after 'start'
|
|
|
|
*/
|
2013-04-26 03:41:01 +07:00
|
|
|
static struct extent_state *
|
|
|
|
find_first_extent_bit_state(struct extent_io_tree *tree,
|
2013-04-29 20:38:46 +07:00
|
|
|
u64 start, unsigned long bits)
|
2008-02-19 00:12:38 +07:00
|
|
|
{
|
|
|
|
struct rb_node *node;
|
|
|
|
struct extent_state *state;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* this search will find all the extents that end after
|
|
|
|
* our range starts.
|
|
|
|
*/
|
|
|
|
node = tree_search(tree, start);
|
2009-01-06 09:25:51 +07:00
|
|
|
if (!node)
|
2008-02-19 00:12:38 +07:00
|
|
|
goto out;
|
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (1) {
|
2008-02-19 00:12:38 +07:00
|
|
|
state = rb_entry(node, struct extent_state, rb_node);
|
2009-01-06 09:25:51 +07:00
|
|
|
if (state->end >= start && (state->state & bits))
|
2008-02-19 00:12:38 +07:00
|
|
|
return state;
|
2009-01-06 09:25:51 +07:00
|
|
|
|
2008-02-19 00:12:38 +07:00
|
|
|
node = rb_next(node);
|
|
|
|
if (!node)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2011-07-14 10:19:45 +07:00
|
|
|
/*
|
|
|
|
* find the first offset in the io tree with 'bits' set. zero is
|
|
|
|
* returned if we find something, and *start_ret and *end_ret are
|
|
|
|
* set to reflect the state struct that was found.
|
|
|
|
*
|
2012-04-06 13:35:47 +07:00
|
|
|
* If nothing was found, 1 is returned. If found something, return 0.
|
2011-07-14 10:19:45 +07:00
|
|
|
*/
|
|
|
|
int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
|
2013-04-29 20:38:46 +07:00
|
|
|
u64 *start_ret, u64 *end_ret, unsigned long bits,
|
2012-09-28 04:07:30 +07:00
|
|
|
struct extent_state **cached_state)
|
2011-07-14 10:19:45 +07:00
|
|
|
{
|
|
|
|
struct extent_state *state;
|
2012-09-28 04:07:30 +07:00
|
|
|
struct rb_node *n;
|
2011-07-14 10:19:45 +07:00
|
|
|
int ret = 1;
|
|
|
|
|
|
|
|
spin_lock(&tree->lock);
|
2012-09-28 04:07:30 +07:00
|
|
|
if (cached_state && *cached_state) {
|
|
|
|
state = *cached_state;
|
2014-07-07 02:09:59 +07:00
|
|
|
if (state->end == start - 1 && extent_state_in_tree(state)) {
|
2012-09-28 04:07:30 +07:00
|
|
|
n = rb_next(&state->rb_node);
|
|
|
|
while (n) {
|
|
|
|
state = rb_entry(n, struct extent_state,
|
|
|
|
rb_node);
|
|
|
|
if (state->state & bits)
|
|
|
|
goto got_it;
|
|
|
|
n = rb_next(n);
|
|
|
|
}
|
|
|
|
free_extent_state(*cached_state);
|
|
|
|
*cached_state = NULL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
free_extent_state(*cached_state);
|
|
|
|
*cached_state = NULL;
|
|
|
|
}
|
|
|
|
|
2011-07-14 10:19:45 +07:00
|
|
|
state = find_first_extent_bit_state(tree, start, bits);
|
2012-09-28 04:07:30 +07:00
|
|
|
got_it:
|
2011-07-14 10:19:45 +07:00
|
|
|
if (state) {
|
2012-09-28 04:07:30 +07:00
|
|
|
cache_state(state, cached_state);
|
2011-07-14 10:19:45 +07:00
|
|
|
*start_ret = state->start;
|
|
|
|
*end_ret = state->end;
|
|
|
|
ret = 0;
|
|
|
|
}
|
2012-09-28 04:07:30 +07:00
|
|
|
out:
|
2011-07-14 10:19:45 +07:00
|
|
|
spin_unlock(&tree->lock);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-09-30 02:18:18 +07:00
|
|
|
/*
|
|
|
|
* find a contiguous range of bytes in the file marked as delalloc, not
|
|
|
|
* more than 'max_bytes'. start and end are used to return the range,
|
|
|
|
*
|
|
|
|
* 1 is returned if we find something, 0 if nothing was in the tree
|
|
|
|
*/
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
|
2010-02-03 04:19:11 +07:00
|
|
|
u64 *start, u64 *end, u64 max_bytes,
|
|
|
|
struct extent_state **cached_state)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct rb_node *node;
|
|
|
|
struct extent_state *state;
|
|
|
|
u64 cur_start = *start;
|
|
|
|
u64 found = 0;
|
|
|
|
u64 total_bytes = 0;
|
|
|
|
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_lock(&tree->lock);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* this search will find all the extents that end after
|
|
|
|
* our range starts.
|
|
|
|
*/
|
2008-02-02 02:51:59 +07:00
|
|
|
node = tree_search(tree, cur_start);
|
2008-04-01 22:21:40 +07:00
|
|
|
if (!node) {
|
2008-04-17 22:29:12 +07:00
|
|
|
if (!found)
|
|
|
|
*end = (u64)-1;
|
2008-01-25 04:13:08 +07:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (1) {
|
2008-01-25 04:13:08 +07:00
|
|
|
state = rb_entry(node, struct extent_state, rb_node);
|
2008-09-26 21:05:38 +07:00
|
|
|
if (found && (state->start != cur_start ||
|
|
|
|
(state->state & EXTENT_BOUNDARY))) {
|
2008-01-25 04:13:08 +07:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
if (!(state->state & EXTENT_DELALLOC)) {
|
|
|
|
if (!found)
|
|
|
|
*end = state->end;
|
|
|
|
goto out;
|
|
|
|
}
|
2010-02-03 04:19:11 +07:00
|
|
|
if (!found) {
|
2008-01-25 04:13:08 +07:00
|
|
|
*start = state->start;
|
2010-02-03 04:19:11 +07:00
|
|
|
*cached_state = state;
|
|
|
|
atomic_inc(&state->refs);
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
found++;
|
|
|
|
*end = state->end;
|
|
|
|
cur_start = state->end + 1;
|
|
|
|
node = rb_next(node);
|
|
|
|
total_bytes += state->end - state->start + 1;
|
2013-10-08 09:11:09 +07:00
|
|
|
if (total_bytes >= max_bytes)
|
2013-08-31 01:38:49 +07:00
|
|
|
break;
|
|
|
|
if (!node)
|
2008-01-25 04:13:08 +07:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
out:
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
return found;
|
|
|
|
}
|
|
|
|
|
2012-03-01 20:56:26 +07:00
|
|
|
static noinline void __unlock_for_delalloc(struct inode *inode,
|
|
|
|
struct page *locked_page,
|
|
|
|
u64 start, u64 end)
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct page *pages[16];
|
|
|
|
unsigned long index = start >> PAGE_CACHE_SHIFT;
|
|
|
|
unsigned long end_index = end >> PAGE_CACHE_SHIFT;
|
|
|
|
unsigned long nr_pages = end_index - index + 1;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (index == locked_page->index && end_index == index)
|
2012-03-01 20:56:26 +07:00
|
|
|
return;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (nr_pages > 0) {
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
ret = find_get_pages_contig(inode->i_mapping, index,
|
2008-11-11 21:34:41 +07:00
|
|
|
min_t(unsigned long, nr_pages,
|
|
|
|
ARRAY_SIZE(pages)), pages);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
for (i = 0; i < ret; i++) {
|
|
|
|
if (pages[i] != locked_page)
|
|
|
|
unlock_page(pages[i]);
|
|
|
|
page_cache_release(pages[i]);
|
|
|
|
}
|
|
|
|
nr_pages -= ret;
|
|
|
|
index += ret;
|
|
|
|
cond_resched();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static noinline int lock_delalloc_pages(struct inode *inode,
|
|
|
|
struct page *locked_page,
|
|
|
|
u64 delalloc_start,
|
|
|
|
u64 delalloc_end)
|
|
|
|
{
|
|
|
|
unsigned long index = delalloc_start >> PAGE_CACHE_SHIFT;
|
|
|
|
unsigned long start_index = index;
|
|
|
|
unsigned long end_index = delalloc_end >> PAGE_CACHE_SHIFT;
|
|
|
|
unsigned long pages_locked = 0;
|
|
|
|
struct page *pages[16];
|
|
|
|
unsigned long nrpages;
|
|
|
|
int ret;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/* the caller is responsible for locking the start index */
|
|
|
|
if (index == locked_page->index && index == end_index)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* skip the page at the start index */
|
|
|
|
nrpages = end_index - index + 1;
|
2009-01-06 09:25:51 +07:00
|
|
|
while (nrpages > 0) {
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
ret = find_get_pages_contig(inode->i_mapping, index,
|
2008-11-11 21:34:41 +07:00
|
|
|
min_t(unsigned long,
|
|
|
|
nrpages, ARRAY_SIZE(pages)), pages);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
if (ret == 0) {
|
|
|
|
ret = -EAGAIN;
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
/* now we have an array of pages, lock them all */
|
|
|
|
for (i = 0; i < ret; i++) {
|
|
|
|
/*
|
|
|
|
* the caller is taking responsibility for
|
|
|
|
* locked_page
|
|
|
|
*/
|
2008-11-07 10:02:51 +07:00
|
|
|
if (pages[i] != locked_page) {
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
lock_page(pages[i]);
|
2008-11-10 19:31:30 +07:00
|
|
|
if (!PageDirty(pages[i]) ||
|
|
|
|
pages[i]->mapping != inode->i_mapping) {
|
2008-11-07 10:02:51 +07:00
|
|
|
ret = -EAGAIN;
|
|
|
|
unlock_page(pages[i]);
|
|
|
|
page_cache_release(pages[i]);
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
}
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
page_cache_release(pages[i]);
|
2008-11-07 10:02:51 +07:00
|
|
|
pages_locked++;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
}
|
|
|
|
nrpages -= ret;
|
|
|
|
index += ret;
|
|
|
|
cond_resched();
|
|
|
|
}
|
|
|
|
ret = 0;
|
|
|
|
done:
|
|
|
|
if (ret && pages_locked) {
|
|
|
|
__unlock_for_delalloc(inode, locked_page,
|
|
|
|
delalloc_start,
|
|
|
|
((u64)(start_index + pages_locked - 1)) <<
|
|
|
|
PAGE_CACHE_SHIFT);
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* find a contiguous range of bytes in the file marked as delalloc, not
|
|
|
|
* more than 'max_bytes'. start and end are used to return the range,
|
|
|
|
*
|
|
|
|
* 1 is returned if we find something, 0 if nothing was in the tree
|
|
|
|
*/
|
2013-10-09 23:00:56 +07:00
|
|
|
STATIC u64 find_lock_delalloc_range(struct inode *inode,
|
|
|
|
struct extent_io_tree *tree,
|
|
|
|
struct page *locked_page, u64 *start,
|
|
|
|
u64 *end, u64 max_bytes)
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
{
|
|
|
|
u64 delalloc_start;
|
|
|
|
u64 delalloc_end;
|
|
|
|
u64 found;
|
2009-09-03 02:22:30 +07:00
|
|
|
struct extent_state *cached_state = NULL;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
int ret;
|
|
|
|
int loops = 0;
|
|
|
|
|
|
|
|
again:
|
|
|
|
/* step one, find a bunch of delalloc bytes starting at start */
|
|
|
|
delalloc_start = *start;
|
|
|
|
delalloc_end = 0;
|
|
|
|
found = find_delalloc_range(tree, &delalloc_start, &delalloc_end,
|
2010-02-03 04:19:11 +07:00
|
|
|
max_bytes, &cached_state);
|
2008-10-31 23:46:39 +07:00
|
|
|
if (!found || delalloc_end <= *start) {
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
*start = delalloc_start;
|
|
|
|
*end = delalloc_end;
|
2010-02-03 04:19:11 +07:00
|
|
|
free_extent_state(cached_state);
|
Btrfs: fix crash of compressed writes
The crash[1] is found by xfstests/generic/208 with "-o compress",
it's not reproduced everytime, but it does panic.
The bug is quite interesting, it's actually introduced by a recent commit
(573aecafca1cf7a974231b759197a1aebcf39c2a,
Btrfs: actually limit the size of delalloc range).
Btrfs implements delay allocation, so during writeback, we
(1) get a page A and lock it
(2) search the state tree for delalloc bytes and lock all pages within the range
(3) process the delalloc range, including find disk space and create
ordered extent and so on.
(4) submit the page A.
It runs well in normal cases, but if we're in a racy case, eg.
buffered compressed writes and aio-dio writes,
sometimes we may fail to lock all pages in the 'delalloc' range,
in which case, we need to fall back to search the state tree again with
a smaller range limit(max_bytes = PAGE_CACHE_SIZE - offset).
The mentioned commit has a side effect, that is, in the fallback case,
we can find delalloc bytes before the index of the page we already have locked,
so we're in the case of (delalloc_end <= *start) and return with (found > 0).
This ends with not locking delalloc pages but making ->writepage still
process them, and the crash happens.
This fixes it by just thinking that we find nothing and returning to caller
as the caller knows how to deal with it properly.
[1]:
------------[ cut here ]------------
kernel BUG at mm/page-writeback.c:2170!
[...]
CPU: 2 PID: 11755 Comm: btrfs-delalloc- Tainted: G O 3.11.0+ #8
[...]
RIP: 0010:[<ffffffff810f5093>] [<ffffffff810f5093>] clear_page_dirty_for_io+0x1e/0x83
[...]
[ 4934.248731] Stack:
[ 4934.248731] ffff8801477e5dc8 ffffea00049b9f00 ffff8801869f9ce8 ffffffffa02b841a
[ 4934.248731] 0000000000000000 0000000000000000 0000000000000fff 0000000000000620
[ 4934.248731] ffff88018db59c78 ffffea0005da8d40 ffffffffa02ff860 00000001810016c0
[ 4934.248731] Call Trace:
[ 4934.248731] [<ffffffffa02b841a>] extent_range_clear_dirty_for_io+0xcf/0xf5 [btrfs]
[ 4934.248731] [<ffffffffa02a8889>] compress_file_range+0x1dc/0x4cb [btrfs]
[ 4934.248731] [<ffffffff8104f7af>] ? detach_if_pending+0x22/0x4b
[ 4934.248731] [<ffffffffa02a8bad>] async_cow_start+0x35/0x53 [btrfs]
[ 4934.248731] [<ffffffffa02c694b>] worker_loop+0x14b/0x48c [btrfs]
[ 4934.248731] [<ffffffffa02c6800>] ? btrfs_queue_worker+0x25c/0x25c [btrfs]
[ 4934.248731] [<ffffffff810608f5>] kthread+0x8d/0x95
[ 4934.248731] [<ffffffff81060868>] ? kthread_freezable_should_stop+0x43/0x43
[ 4934.248731] [<ffffffff814fe09c>] ret_from_fork+0x7c/0xb0
[ 4934.248731] [<ffffffff81060868>] ? kthread_freezable_should_stop+0x43/0x43
[ 4934.248731] Code: ff 85 c0 0f 94 c0 0f b6 c0 59 5b 5d c3 0f 1f 44 00 00 55 48 89 e5 41 54 53 48 89 fb e8 2c de 00 00 49 89 c4 48 8b 03 a8 01 75 02 <0f> 0b 4d 85 e4 74 52 49 8b 84 24 80 00 00 00 f6 40 20 01 75 44
[ 4934.248731] RIP [<ffffffff810f5093>] clear_page_dirty_for_io+0x1e/0x83
[ 4934.248731] RSP <ffff8801869f9c48>
[ 4934.280307] ---[ end trace 36f06d3f8750236a ]---
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
2013-10-01 22:49:49 +07:00
|
|
|
return 0;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
}
|
|
|
|
|
2008-10-31 23:46:39 +07:00
|
|
|
/*
|
|
|
|
* start comes from the offset of locked_page. We have to lock
|
|
|
|
* pages in order, so we can't process delalloc bytes before
|
|
|
|
* locked_page
|
|
|
|
*/
|
2009-01-06 09:25:51 +07:00
|
|
|
if (delalloc_start < *start)
|
2008-10-31 23:46:39 +07:00
|
|
|
delalloc_start = *start;
|
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
/*
|
|
|
|
* make sure to limit the number of pages we try to lock down
|
|
|
|
*/
|
2013-10-08 09:11:09 +07:00
|
|
|
if (delalloc_end + 1 - delalloc_start > max_bytes)
|
|
|
|
delalloc_end = delalloc_start + max_bytes - 1;
|
2009-01-06 09:25:51 +07:00
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
/* step two, lock all the pages after the page that has start */
|
|
|
|
ret = lock_delalloc_pages(inode, locked_page,
|
|
|
|
delalloc_start, delalloc_end);
|
|
|
|
if (ret == -EAGAIN) {
|
|
|
|
/* some of the pages are gone, lets avoid looping by
|
|
|
|
* shortening the size of the delalloc range we're searching
|
|
|
|
*/
|
2009-09-03 02:22:30 +07:00
|
|
|
free_extent_state(cached_state);
|
2014-05-21 19:49:54 +07:00
|
|
|
cached_state = NULL;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
if (!loops) {
|
2013-10-08 09:11:09 +07:00
|
|
|
max_bytes = PAGE_CACHE_SIZE;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
loops = 1;
|
|
|
|
goto again;
|
|
|
|
} else {
|
|
|
|
found = 0;
|
|
|
|
goto out_failed;
|
|
|
|
}
|
|
|
|
}
|
2012-03-12 22:03:00 +07:00
|
|
|
BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
|
|
|
|
/* step three, lock the state bits for the whole range */
|
2012-03-01 20:57:19 +07:00
|
|
|
lock_extent_bits(tree, delalloc_start, delalloc_end, 0, &cached_state);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
|
|
|
|
/* then test to make sure it is all still delalloc */
|
|
|
|
ret = test_range_bit(tree, delalloc_start, delalloc_end,
|
2009-09-03 02:22:30 +07:00
|
|
|
EXTENT_DELALLOC, 1, cached_state);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
if (!ret) {
|
2009-09-03 02:22:30 +07:00
|
|
|
unlock_extent_cached(tree, delalloc_start, delalloc_end,
|
|
|
|
&cached_state, GFP_NOFS);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
__unlock_for_delalloc(inode, locked_page,
|
|
|
|
delalloc_start, delalloc_end);
|
|
|
|
cond_resched();
|
|
|
|
goto again;
|
|
|
|
}
|
2009-09-03 02:22:30 +07:00
|
|
|
free_extent_state(cached_state);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
*start = delalloc_start;
|
|
|
|
*end = delalloc_end;
|
|
|
|
out_failed:
|
|
|
|
return found;
|
|
|
|
}
|
|
|
|
|
2013-07-29 22:20:47 +07:00
|
|
|
int extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
|
|
|
|
struct page *locked_page,
|
|
|
|
unsigned long clear_bits,
|
|
|
|
unsigned long page_ops)
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
{
|
2013-07-29 22:20:47 +07:00
|
|
|
struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
int ret;
|
|
|
|
struct page *pages[16];
|
|
|
|
unsigned long index = start >> PAGE_CACHE_SHIFT;
|
|
|
|
unsigned long end_index = end >> PAGE_CACHE_SHIFT;
|
|
|
|
unsigned long nr_pages = end_index - index + 1;
|
|
|
|
int i;
|
2008-11-07 10:02:51 +07:00
|
|
|
|
2009-09-03 02:04:12 +07:00
|
|
|
clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
|
2013-07-29 22:20:47 +07:00
|
|
|
if (page_ops == 0)
|
2008-11-07 10:02:51 +07:00
|
|
|
return 0;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (nr_pages > 0) {
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
ret = find_get_pages_contig(inode->i_mapping, index,
|
2008-11-11 21:34:41 +07:00
|
|
|
min_t(unsigned long,
|
|
|
|
nr_pages, ARRAY_SIZE(pages)), pages);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
for (i = 0; i < ret; i++) {
|
2009-09-03 03:53:46 +07:00
|
|
|
|
2013-07-29 22:20:47 +07:00
|
|
|
if (page_ops & PAGE_SET_PRIVATE2)
|
2009-09-03 03:53:46 +07:00
|
|
|
SetPagePrivate2(pages[i]);
|
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
if (pages[i] == locked_page) {
|
|
|
|
page_cache_release(pages[i]);
|
|
|
|
continue;
|
|
|
|
}
|
2013-07-29 22:20:47 +07:00
|
|
|
if (page_ops & PAGE_CLEAR_DIRTY)
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
clear_page_dirty_for_io(pages[i]);
|
2013-07-29 22:20:47 +07:00
|
|
|
if (page_ops & PAGE_SET_WRITEBACK)
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
set_page_writeback(pages[i]);
|
2013-07-29 22:20:47 +07:00
|
|
|
if (page_ops & PAGE_END_WRITEBACK)
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
end_page_writeback(pages[i]);
|
2013-07-29 22:20:47 +07:00
|
|
|
if (page_ops & PAGE_UNLOCK)
|
2008-11-07 10:02:51 +07:00
|
|
|
unlock_page(pages[i]);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
page_cache_release(pages[i]);
|
|
|
|
}
|
|
|
|
nr_pages -= ret;
|
|
|
|
index += ret;
|
|
|
|
cond_resched();
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-09-30 02:18:18 +07:00
|
|
|
/*
|
|
|
|
* count the number of bytes in the tree that have a given bit(s)
|
|
|
|
* set. This can be fairly slow, except for EXTENT_DIRTY which is
|
|
|
|
* cached. The total number found is returned.
|
|
|
|
*/
|
2008-01-25 04:13:08 +07:00
|
|
|
u64 count_range_bits(struct extent_io_tree *tree,
|
|
|
|
u64 *start, u64 search_end, u64 max_bytes,
|
2011-02-24 04:23:20 +07:00
|
|
|
unsigned long bits, int contig)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct rb_node *node;
|
|
|
|
struct extent_state *state;
|
|
|
|
u64 cur_start = *start;
|
|
|
|
u64 total_bytes = 0;
|
2011-02-24 04:23:20 +07:00
|
|
|
u64 last = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
int found = 0;
|
|
|
|
|
2013-10-31 12:00:08 +07:00
|
|
|
if (WARN_ON(search_end <= cur_start))
|
2008-01-25 04:13:08 +07:00
|
|
|
return 0;
|
|
|
|
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_lock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (cur_start == 0 && bits == EXTENT_DIRTY) {
|
|
|
|
total_bytes = tree->dirty_bytes;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* this search will find all the extents that end after
|
|
|
|
* our range starts.
|
|
|
|
*/
|
2008-02-02 02:51:59 +07:00
|
|
|
node = tree_search(tree, cur_start);
|
2009-01-06 09:25:51 +07:00
|
|
|
if (!node)
|
2008-01-25 04:13:08 +07:00
|
|
|
goto out;
|
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (1) {
|
2008-01-25 04:13:08 +07:00
|
|
|
state = rb_entry(node, struct extent_state, rb_node);
|
|
|
|
if (state->start > search_end)
|
|
|
|
break;
|
2011-02-24 04:23:20 +07:00
|
|
|
if (contig && found && state->start > last + 1)
|
|
|
|
break;
|
|
|
|
if (state->end >= cur_start && (state->state & bits) == bits) {
|
2008-01-25 04:13:08 +07:00
|
|
|
total_bytes += min(search_end, state->end) + 1 -
|
|
|
|
max(cur_start, state->start);
|
|
|
|
if (total_bytes >= max_bytes)
|
|
|
|
break;
|
|
|
|
if (!found) {
|
2011-05-04 22:11:17 +07:00
|
|
|
*start = max(cur_start, state->start);
|
2008-01-25 04:13:08 +07:00
|
|
|
found = 1;
|
|
|
|
}
|
2011-02-24 04:23:20 +07:00
|
|
|
last = state->end;
|
|
|
|
} else if (contig && found) {
|
|
|
|
break;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
node = rb_next(node);
|
|
|
|
if (!node)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
out:
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
return total_bytes;
|
|
|
|
}
|
2008-12-02 21:54:17 +07:00
|
|
|
|
2008-09-30 02:18:18 +07:00
|
|
|
/*
|
|
|
|
* set the private field for a given byte offset in the tree. If there isn't
|
|
|
|
* an extent_state there already, this does nothing.
|
|
|
|
*/
|
2013-08-15 03:27:46 +07:00
|
|
|
static int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct rb_node *node;
|
|
|
|
struct extent_state *state;
|
|
|
|
int ret = 0;
|
|
|
|
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_lock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* this search will find all the extents that end after
|
|
|
|
* our range starts.
|
|
|
|
*/
|
2008-02-02 02:51:59 +07:00
|
|
|
node = tree_search(tree, start);
|
2008-04-01 22:21:40 +07:00
|
|
|
if (!node) {
|
2008-01-25 04:13:08 +07:00
|
|
|
ret = -ENOENT;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
state = rb_entry(node, struct extent_state, rb_node);
|
|
|
|
if (state->start != start) {
|
|
|
|
ret = -ENOENT;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
state->private = private;
|
|
|
|
out:
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private)
|
|
|
|
{
|
|
|
|
struct rb_node *node;
|
|
|
|
struct extent_state *state;
|
|
|
|
int ret = 0;
|
|
|
|
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_lock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* this search will find all the extents that end after
|
|
|
|
* our range starts.
|
|
|
|
*/
|
2008-02-02 02:51:59 +07:00
|
|
|
node = tree_search(tree, start);
|
2008-04-01 22:21:40 +07:00
|
|
|
if (!node) {
|
2008-01-25 04:13:08 +07:00
|
|
|
ret = -ENOENT;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
state = rb_entry(node, struct extent_state, rb_node);
|
|
|
|
if (state->start != start) {
|
|
|
|
ret = -ENOENT;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
*private = state->private;
|
|
|
|
out:
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* searches a range in the state tree for a given mask.
|
2008-01-29 21:59:12 +07:00
|
|
|
* If 'filled' == 1, this returns 1 only if every extent in the tree
|
2008-01-25 04:13:08 +07:00
|
|
|
* has the bits set. Otherwise, 1 is returned if any bit in the
|
|
|
|
* range is found set.
|
|
|
|
*/
|
|
|
|
int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
|
2013-04-29 20:38:46 +07:00
|
|
|
unsigned long bits, int filled, struct extent_state *cached)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct extent_state *state = NULL;
|
|
|
|
struct rb_node *node;
|
|
|
|
int bitset = 0;
|
|
|
|
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_lock(&tree->lock);
|
2014-07-07 02:09:59 +07:00
|
|
|
if (cached && extent_state_in_tree(cached) && cached->start <= start &&
|
2011-06-21 01:53:48 +07:00
|
|
|
cached->end > start)
|
2009-09-03 02:22:30 +07:00
|
|
|
node = &cached->rb_node;
|
|
|
|
else
|
|
|
|
node = tree_search(tree, start);
|
2008-01-25 04:13:08 +07:00
|
|
|
while (node && start <= end) {
|
|
|
|
state = rb_entry(node, struct extent_state, rb_node);
|
|
|
|
|
|
|
|
if (filled && state->start > start) {
|
|
|
|
bitset = 0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (state->start > end)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (state->state & bits) {
|
|
|
|
bitset = 1;
|
|
|
|
if (!filled)
|
|
|
|
break;
|
|
|
|
} else if (filled) {
|
|
|
|
bitset = 0;
|
|
|
|
break;
|
|
|
|
}
|
2009-09-24 07:23:16 +07:00
|
|
|
|
|
|
|
if (state->end == (u64)-1)
|
|
|
|
break;
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
start = state->end + 1;
|
|
|
|
if (start > end)
|
|
|
|
break;
|
|
|
|
node = rb_next(node);
|
|
|
|
if (!node) {
|
|
|
|
if (filled)
|
|
|
|
bitset = 0;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2008-12-18 02:51:42 +07:00
|
|
|
spin_unlock(&tree->lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
return bitset;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* helper function to set a given page up to date if all the
|
|
|
|
* extents in the tree for that page are up to date
|
|
|
|
*/
|
2012-03-01 20:56:26 +07:00
|
|
|
static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2012-12-21 16:17:45 +07:00
|
|
|
u64 start = page_offset(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
u64 end = start + PAGE_CACHE_SIZE - 1;
|
2009-09-03 02:22:30 +07:00
|
|
|
if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
|
2008-01-25 04:13:08 +07:00
|
|
|
SetPageUptodate(page);
|
|
|
|
}
|
|
|
|
|
2014-09-12 17:44:03 +07:00
|
|
|
int free_io_failure(struct inode *inode, struct io_failure_record *rec)
|
2011-07-22 20:41:52 +07:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
int err = 0;
|
|
|
|
struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
|
|
|
|
|
|
|
|
set_state_private(failure_tree, rec->start, 0);
|
|
|
|
ret = clear_extent_bits(failure_tree, rec->start,
|
|
|
|
rec->start + rec->len - 1,
|
|
|
|
EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
|
|
|
|
if (ret)
|
|
|
|
err = ret;
|
|
|
|
|
2013-01-30 06:40:14 +07:00
|
|
|
ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start,
|
|
|
|
rec->start + rec->len - 1,
|
|
|
|
EXTENT_DAMAGED, GFP_NOFS);
|
|
|
|
if (ret && !err)
|
|
|
|
err = ret;
|
2011-07-22 20:41:52 +07:00
|
|
|
|
|
|
|
kfree(rec);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* this bypasses the standard btrfs submit functions deliberately, as
|
|
|
|
* the standard behavior is to write all copies in a raid setup. here we only
|
|
|
|
* want to write the one bad copy. so we do the mapping for ourselves and issue
|
|
|
|
* submit_bio directly.
|
2012-11-05 21:46:42 +07:00
|
|
|
* to avoid any synchronization issues, wait for the data after writing, which
|
2011-07-22 20:41:52 +07:00
|
|
|
* actually prevents the read that triggered the error from finishing.
|
|
|
|
* currently, there can be no more than two copies of every data bit. thus,
|
|
|
|
* exactly one rewrite is required.
|
|
|
|
*/
|
2014-09-12 17:44:01 +07:00
|
|
|
int repair_io_failure(struct inode *inode, u64 start, u64 length, u64 logical,
|
|
|
|
struct page *page, unsigned int pg_offset, int mirror_num)
|
2011-07-22 20:41:52 +07:00
|
|
|
{
|
2014-09-12 17:44:01 +07:00
|
|
|
struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
|
2011-07-22 20:41:52 +07:00
|
|
|
struct bio *bio;
|
|
|
|
struct btrfs_device *dev;
|
|
|
|
u64 map_length = 0;
|
|
|
|
u64 sector;
|
|
|
|
struct btrfs_bio *bbio = NULL;
|
2013-01-30 06:40:14 +07:00
|
|
|
struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
|
2011-07-22 20:41:52 +07:00
|
|
|
int ret;
|
|
|
|
|
2013-11-04 00:06:39 +07:00
|
|
|
ASSERT(!(fs_info->sb->s_flags & MS_RDONLY));
|
2011-07-22 20:41:52 +07:00
|
|
|
BUG_ON(!mirror_num);
|
|
|
|
|
2013-01-30 06:40:14 +07:00
|
|
|
/* we can't repair anything in raid56 yet */
|
|
|
|
if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
|
|
|
|
return 0;
|
|
|
|
|
2013-05-18 05:30:14 +07:00
|
|
|
bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
|
2011-07-22 20:41:52 +07:00
|
|
|
if (!bio)
|
|
|
|
return -EIO;
|
2013-10-12 05:44:27 +07:00
|
|
|
bio->bi_iter.bi_size = 0;
|
2011-07-22 20:41:52 +07:00
|
|
|
map_length = length;
|
|
|
|
|
2012-11-05 21:46:42 +07:00
|
|
|
ret = btrfs_map_block(fs_info, WRITE, logical,
|
2011-07-22 20:41:52 +07:00
|
|
|
&map_length, &bbio, mirror_num);
|
|
|
|
if (ret) {
|
|
|
|
bio_put(bio);
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
BUG_ON(mirror_num != bbio->mirror_num);
|
|
|
|
sector = bbio->stripes[mirror_num-1].physical >> 9;
|
2013-10-12 05:44:27 +07:00
|
|
|
bio->bi_iter.bi_sector = sector;
|
2011-07-22 20:41:52 +07:00
|
|
|
dev = bbio->stripes[mirror_num-1].dev;
|
|
|
|
kfree(bbio);
|
|
|
|
if (!dev || !dev->bdev || !dev->writeable) {
|
|
|
|
bio_put(bio);
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
bio->bi_bdev = dev->bdev;
|
2014-09-12 17:44:00 +07:00
|
|
|
bio_add_page(bio, page, length, pg_offset);
|
2011-07-22 20:41:52 +07:00
|
|
|
|
2013-11-24 13:33:32 +07:00
|
|
|
if (btrfsic_submit_bio_wait(WRITE_SYNC, bio)) {
|
2011-07-22 20:41:52 +07:00
|
|
|
/* try to remap that extent elsewhere? */
|
|
|
|
bio_put(bio);
|
2012-05-25 21:06:08 +07:00
|
|
|
btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
|
2011-07-22 20:41:52 +07:00
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
|
2013-12-20 23:37:06 +07:00
|
|
|
printk_ratelimited_in_rcu(KERN_INFO
|
2014-09-12 17:44:01 +07:00
|
|
|
"BTRFS: read error corrected: ino %llu off %llu (dev %s sector %llu)\n",
|
|
|
|
btrfs_ino(inode), start,
|
|
|
|
rcu_str_deref(dev->name), sector);
|
2011-07-22 20:41:52 +07:00
|
|
|
bio_put(bio);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-03-27 08:57:36 +07:00
|
|
|
int repair_eb_io_failure(struct btrfs_root *root, struct extent_buffer *eb,
|
|
|
|
int mirror_num)
|
|
|
|
{
|
|
|
|
u64 start = eb->start;
|
|
|
|
unsigned long i, num_pages = num_extent_pages(eb->start, eb->len);
|
2012-04-13 02:55:15 +07:00
|
|
|
int ret = 0;
|
2012-03-27 08:57:36 +07:00
|
|
|
|
2013-11-04 00:06:39 +07:00
|
|
|
if (root->fs_info->sb->s_flags & MS_RDONLY)
|
|
|
|
return -EROFS;
|
|
|
|
|
2012-03-27 08:57:36 +07:00
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
struct page *p = extent_buffer_page(eb, i);
|
2014-09-12 17:44:01 +07:00
|
|
|
|
|
|
|
ret = repair_io_failure(root->fs_info->btree_inode, start,
|
|
|
|
PAGE_CACHE_SIZE, start, p,
|
|
|
|
start - page_offset(p), mirror_num);
|
2012-03-27 08:57:36 +07:00
|
|
|
if (ret)
|
|
|
|
break;
|
|
|
|
start += PAGE_CACHE_SIZE;
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2011-07-22 20:41:52 +07:00
|
|
|
/*
|
|
|
|
* each time an IO finishes, we do a fast check in the IO failure tree
|
|
|
|
* to see if we need to process or clean up an io_failure_record
|
|
|
|
*/
|
2014-09-12 17:44:03 +07:00
|
|
|
int clean_io_failure(struct inode *inode, u64 start, struct page *page,
|
|
|
|
unsigned int pg_offset)
|
2011-07-22 20:41:52 +07:00
|
|
|
{
|
|
|
|
u64 private;
|
|
|
|
u64 private_failure;
|
|
|
|
struct io_failure_record *failrec;
|
2013-11-04 00:06:39 +07:00
|
|
|
struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
|
2011-07-22 20:41:52 +07:00
|
|
|
struct extent_state *state;
|
|
|
|
int num_copies;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
private = 0;
|
|
|
|
ret = count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
|
|
|
|
(u64)-1, 1, EXTENT_DIRTY, 0);
|
|
|
|
if (!ret)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ret = get_state_private(&BTRFS_I(inode)->io_failure_tree, start,
|
|
|
|
&private_failure);
|
|
|
|
if (ret)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
failrec = (struct io_failure_record *)(unsigned long) private_failure;
|
|
|
|
BUG_ON(!failrec->this_mirror);
|
|
|
|
|
|
|
|
if (failrec->in_validation) {
|
|
|
|
/* there was no real error, just free the record */
|
|
|
|
pr_debug("clean_io_failure: freeing dummy error at %llu\n",
|
|
|
|
failrec->start);
|
|
|
|
goto out;
|
|
|
|
}
|
2013-11-04 00:06:39 +07:00
|
|
|
if (fs_info->sb->s_flags & MS_RDONLY)
|
|
|
|
goto out;
|
2011-07-22 20:41:52 +07:00
|
|
|
|
|
|
|
spin_lock(&BTRFS_I(inode)->io_tree.lock);
|
|
|
|
state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
|
|
|
|
failrec->start,
|
|
|
|
EXTENT_LOCKED);
|
|
|
|
spin_unlock(&BTRFS_I(inode)->io_tree.lock);
|
|
|
|
|
2013-07-25 18:22:35 +07:00
|
|
|
if (state && state->start <= failrec->start &&
|
|
|
|
state->end >= failrec->start + failrec->len - 1) {
|
2012-11-05 21:46:42 +07:00
|
|
|
num_copies = btrfs_num_copies(fs_info, failrec->logical,
|
|
|
|
failrec->len);
|
2011-07-22 20:41:52 +07:00
|
|
|
if (num_copies > 1) {
|
2014-09-12 17:44:01 +07:00
|
|
|
repair_io_failure(inode, start, failrec->len,
|
2014-09-12 17:43:58 +07:00
|
|
|
failrec->logical, page,
|
2014-09-12 17:44:01 +07:00
|
|
|
pg_offset, failrec->failed_mirror);
|
2011-07-22 20:41:52 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
2014-09-12 17:43:58 +07:00
|
|
|
free_io_failure(inode, failrec);
|
2011-07-22 20:41:52 +07:00
|
|
|
|
2014-09-12 17:43:58 +07:00
|
|
|
return 0;
|
2011-07-22 20:41:52 +07:00
|
|
|
}
|
|
|
|
|
Btrfs: cleanup the read failure record after write or when the inode is freeing
After the data is written successfully, we should cleanup the read failure record
in that range because
- If we set data COW for the file, the range that the failure record pointed to is
mapped to a new place, so it is invalid.
- If we set no data COW for the file, and if there is no error during writting,
the corrupted data is corrected, so the failure record can be removed. And if
some errors happen on the mirrors, we also needn't worry about it because the
failure record will be recreated if we read the same place again.
Sometimes, we may fail to correct the data, so the failure records will be left
in the tree, we need free them when we free the inode or the memory leak happens.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Chris Mason <clm@fb.com>
2014-09-12 17:44:04 +07:00
|
|
|
/*
|
|
|
|
* Can be called when
|
|
|
|
* - hold extent lock
|
|
|
|
* - under ordered extent
|
|
|
|
* - the inode is freeing
|
|
|
|
*/
|
|
|
|
void btrfs_free_io_failure_record(struct inode *inode, u64 start, u64 end)
|
|
|
|
{
|
|
|
|
struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
|
|
|
|
struct io_failure_record *failrec;
|
|
|
|
struct extent_state *state, *next;
|
|
|
|
|
|
|
|
if (RB_EMPTY_ROOT(&failure_tree->state))
|
|
|
|
return;
|
|
|
|
|
|
|
|
spin_lock(&failure_tree->lock);
|
|
|
|
state = find_first_extent_bit_state(failure_tree, start, EXTENT_DIRTY);
|
|
|
|
while (state) {
|
|
|
|
if (state->start > end)
|
|
|
|
break;
|
|
|
|
|
|
|
|
ASSERT(state->end <= end);
|
|
|
|
|
|
|
|
next = next_state(state);
|
|
|
|
|
|
|
|
failrec = (struct io_failure_record *)state->private;
|
|
|
|
free_extent_state(state);
|
|
|
|
kfree(failrec);
|
|
|
|
|
|
|
|
state = next;
|
|
|
|
}
|
|
|
|
spin_unlock(&failure_tree->lock);
|
|
|
|
}
|
|
|
|
|
2014-09-12 17:43:59 +07:00
|
|
|
int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
|
|
|
|
struct io_failure_record **failrec_ret)
|
2011-07-22 20:41:52 +07:00
|
|
|
{
|
2014-09-12 17:43:59 +07:00
|
|
|
struct io_failure_record *failrec;
|
2011-07-22 20:41:52 +07:00
|
|
|
u64 private;
|
|
|
|
struct extent_map *em;
|
|
|
|
struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
|
|
|
|
struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
|
|
|
|
struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
|
|
|
|
int ret;
|
|
|
|
u64 logical;
|
|
|
|
|
|
|
|
ret = get_state_private(failure_tree, start, &private);
|
|
|
|
if (ret) {
|
|
|
|
failrec = kzalloc(sizeof(*failrec), GFP_NOFS);
|
|
|
|
if (!failrec)
|
|
|
|
return -ENOMEM;
|
2014-09-12 17:43:59 +07:00
|
|
|
|
2011-07-22 20:41:52 +07:00
|
|
|
failrec->start = start;
|
|
|
|
failrec->len = end - start + 1;
|
|
|
|
failrec->this_mirror = 0;
|
|
|
|
failrec->bio_flags = 0;
|
|
|
|
failrec->in_validation = 0;
|
|
|
|
|
|
|
|
read_lock(&em_tree->lock);
|
|
|
|
em = lookup_extent_mapping(em_tree, start, failrec->len);
|
|
|
|
if (!em) {
|
|
|
|
read_unlock(&em_tree->lock);
|
|
|
|
kfree(failrec);
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
|
2013-11-25 10:22:07 +07:00
|
|
|
if (em->start > start || em->start + em->len <= start) {
|
2011-07-22 20:41:52 +07:00
|
|
|
free_extent_map(em);
|
|
|
|
em = NULL;
|
|
|
|
}
|
|
|
|
read_unlock(&em_tree->lock);
|
2012-10-01 16:07:15 +07:00
|
|
|
if (!em) {
|
2011-07-22 20:41:52 +07:00
|
|
|
kfree(failrec);
|
|
|
|
return -EIO;
|
|
|
|
}
|
2014-09-12 17:43:59 +07:00
|
|
|
|
2011-07-22 20:41:52 +07:00
|
|
|
logical = start - em->start;
|
|
|
|
logical = em->block_start + logical;
|
|
|
|
if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
|
|
|
|
logical = em->block_start;
|
|
|
|
failrec->bio_flags = EXTENT_BIO_COMPRESSED;
|
|
|
|
extent_set_compress_type(&failrec->bio_flags,
|
|
|
|
em->compress_type);
|
|
|
|
}
|
2014-09-12 17:43:59 +07:00
|
|
|
|
|
|
|
pr_debug("Get IO Failure Record: (new) logical=%llu, start=%llu, len=%llu\n",
|
|
|
|
logical, start, failrec->len);
|
|
|
|
|
2011-07-22 20:41:52 +07:00
|
|
|
failrec->logical = logical;
|
|
|
|
free_extent_map(em);
|
|
|
|
|
|
|
|
/* set the bits in the private failure tree */
|
|
|
|
ret = set_extent_bits(failure_tree, start, end,
|
|
|
|
EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
|
|
|
|
if (ret >= 0)
|
|
|
|
ret = set_state_private(failure_tree, start,
|
|
|
|
(u64)(unsigned long)failrec);
|
|
|
|
/* set the bits in the inode's tree */
|
|
|
|
if (ret >= 0)
|
|
|
|
ret = set_extent_bits(tree, start, end, EXTENT_DAMAGED,
|
|
|
|
GFP_NOFS);
|
|
|
|
if (ret < 0) {
|
|
|
|
kfree(failrec);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
failrec = (struct io_failure_record *)(unsigned long)private;
|
2014-09-12 17:43:59 +07:00
|
|
|
pr_debug("Get IO Failure Record: (found) logical=%llu, start=%llu, len=%llu, validation=%d\n",
|
2011-07-22 20:41:52 +07:00
|
|
|
failrec->logical, failrec->start, failrec->len,
|
|
|
|
failrec->in_validation);
|
|
|
|
/*
|
|
|
|
* when data can be on disk more than twice, add to failrec here
|
|
|
|
* (e.g. with a list for failed_mirror) to make
|
|
|
|
* clean_io_failure() clean all those errors at once.
|
|
|
|
*/
|
|
|
|
}
|
2014-09-12 17:43:59 +07:00
|
|
|
|
|
|
|
*failrec_ret = failrec;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_check_repairable(struct inode *inode, struct bio *failed_bio,
|
|
|
|
struct io_failure_record *failrec, int failed_mirror)
|
|
|
|
{
|
|
|
|
int num_copies;
|
|
|
|
|
2012-11-05 20:59:07 +07:00
|
|
|
num_copies = btrfs_num_copies(BTRFS_I(inode)->root->fs_info,
|
|
|
|
failrec->logical, failrec->len);
|
2011-07-22 20:41:52 +07:00
|
|
|
if (num_copies == 1) {
|
|
|
|
/*
|
|
|
|
* we only have a single copy of the data, so don't bother with
|
|
|
|
* all the retry and error correction code that follows. no
|
|
|
|
* matter what the error is, it is very likely to persist.
|
|
|
|
*/
|
2014-09-12 17:43:59 +07:00
|
|
|
pr_debug("Check Repairable: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d\n",
|
2013-07-25 18:22:32 +07:00
|
|
|
num_copies, failrec->this_mirror, failed_mirror);
|
2014-09-12 17:43:59 +07:00
|
|
|
return 0;
|
2011-07-22 20:41:52 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* there are two premises:
|
|
|
|
* a) deliver good data to the caller
|
|
|
|
* b) correct the bad sectors on disk
|
|
|
|
*/
|
|
|
|
if (failed_bio->bi_vcnt > 1) {
|
|
|
|
/*
|
|
|
|
* to fulfill b), we need to know the exact failing sectors, as
|
|
|
|
* we don't want to rewrite any more than the failed ones. thus,
|
|
|
|
* we need separate read requests for the failed bio
|
|
|
|
*
|
|
|
|
* if the following BUG_ON triggers, our validation request got
|
|
|
|
* merged. we need separate requests for our algorithm to work.
|
|
|
|
*/
|
|
|
|
BUG_ON(failrec->in_validation);
|
|
|
|
failrec->in_validation = 1;
|
|
|
|
failrec->this_mirror = failed_mirror;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* we're ready to fulfill a) and b) alongside. get a good copy
|
|
|
|
* of the failed sector and if we succeed, we have setup
|
|
|
|
* everything for repair_io_failure to do the rest for us.
|
|
|
|
*/
|
|
|
|
if (failrec->in_validation) {
|
|
|
|
BUG_ON(failrec->this_mirror != failed_mirror);
|
|
|
|
failrec->in_validation = 0;
|
|
|
|
failrec->this_mirror = 0;
|
|
|
|
}
|
|
|
|
failrec->failed_mirror = failed_mirror;
|
|
|
|
failrec->this_mirror++;
|
|
|
|
if (failrec->this_mirror == failed_mirror)
|
|
|
|
failrec->this_mirror++;
|
|
|
|
}
|
|
|
|
|
2013-07-25 18:22:34 +07:00
|
|
|
if (failrec->this_mirror > num_copies) {
|
2014-09-12 17:43:59 +07:00
|
|
|
pr_debug("Check Repairable: (fail) num_copies=%d, next_mirror %d, failed_mirror %d\n",
|
2011-07-22 20:41:52 +07:00
|
|
|
num_copies, failrec->this_mirror, failed_mirror);
|
2014-09-12 17:43:59 +07:00
|
|
|
return 0;
|
2011-07-22 20:41:52 +07:00
|
|
|
}
|
|
|
|
|
2014-09-12 17:43:59 +07:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
|
|
|
|
struct io_failure_record *failrec,
|
|
|
|
struct page *page, int pg_offset, int icsum,
|
2014-09-12 17:44:03 +07:00
|
|
|
bio_end_io_t *endio_func, void *data)
|
2014-09-12 17:43:59 +07:00
|
|
|
{
|
|
|
|
struct bio *bio;
|
|
|
|
struct btrfs_io_bio *btrfs_failed_bio;
|
|
|
|
struct btrfs_io_bio *btrfs_bio;
|
|
|
|
|
2013-05-18 05:30:14 +07:00
|
|
|
bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
|
2014-09-12 17:43:59 +07:00
|
|
|
if (!bio)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
bio->bi_end_io = endio_func;
|
2013-10-12 05:44:27 +07:00
|
|
|
bio->bi_iter.bi_sector = failrec->logical >> 9;
|
2011-07-22 20:41:52 +07:00
|
|
|
bio->bi_bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
|
2013-10-12 05:44:27 +07:00
|
|
|
bio->bi_iter.bi_size = 0;
|
2014-09-12 17:44:03 +07:00
|
|
|
bio->bi_private = data;
|
2011-07-22 20:41:52 +07:00
|
|
|
|
2013-07-25 18:22:34 +07:00
|
|
|
btrfs_failed_bio = btrfs_io_bio(failed_bio);
|
|
|
|
if (btrfs_failed_bio->csum) {
|
|
|
|
struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
|
|
|
|
u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
|
|
|
|
|
|
|
|
btrfs_bio = btrfs_io_bio(bio);
|
|
|
|
btrfs_bio->csum = btrfs_bio->csum_inline;
|
2014-09-12 17:43:59 +07:00
|
|
|
icsum *= csum_size;
|
|
|
|
memcpy(btrfs_bio->csum, btrfs_failed_bio->csum + icsum,
|
2013-07-25 18:22:34 +07:00
|
|
|
csum_size);
|
|
|
|
}
|
|
|
|
|
2014-09-12 17:43:59 +07:00
|
|
|
bio_add_page(bio, page, failrec->len, pg_offset);
|
|
|
|
|
|
|
|
return bio;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* this is a generic handler for readpage errors (default
|
|
|
|
* readpage_io_failed_hook). if other copies exist, read those and write back
|
|
|
|
* good data to the failed position. does not investigate in remapping the
|
|
|
|
* failed extent elsewhere, hoping the device will be smart enough to do this as
|
|
|
|
* needed
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int bio_readpage_error(struct bio *failed_bio, u64 phy_offset,
|
|
|
|
struct page *page, u64 start, u64 end,
|
|
|
|
int failed_mirror)
|
|
|
|
{
|
|
|
|
struct io_failure_record *failrec;
|
|
|
|
struct inode *inode = page->mapping->host;
|
|
|
|
struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
|
|
|
|
struct bio *bio;
|
|
|
|
int read_mode;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
BUG_ON(failed_bio->bi_rw & REQ_WRITE);
|
|
|
|
|
|
|
|
ret = btrfs_get_io_failure_record(inode, start, end, &failrec);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = btrfs_check_repairable(inode, failed_bio, failrec, failed_mirror);
|
|
|
|
if (!ret) {
|
|
|
|
free_io_failure(inode, failrec);
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (failed_bio->bi_vcnt > 1)
|
|
|
|
read_mode = READ_SYNC | REQ_FAILFAST_DEV;
|
|
|
|
else
|
|
|
|
read_mode = READ_SYNC;
|
|
|
|
|
|
|
|
phy_offset >>= inode->i_sb->s_blocksize_bits;
|
|
|
|
bio = btrfs_create_repair_bio(inode, failed_bio, failrec, page,
|
|
|
|
start - page_offset(page),
|
2014-09-12 17:44:03 +07:00
|
|
|
(int)phy_offset, failed_bio->bi_end_io,
|
|
|
|
NULL);
|
2014-09-12 17:43:59 +07:00
|
|
|
if (!bio) {
|
|
|
|
free_io_failure(inode, failrec);
|
|
|
|
return -EIO;
|
|
|
|
}
|
2011-07-22 20:41:52 +07:00
|
|
|
|
2014-09-12 17:43:59 +07:00
|
|
|
pr_debug("Repair Read Error: submitting new read[%#x] to this_mirror=%d, in_validation=%d\n",
|
|
|
|
read_mode, failrec->this_mirror, failrec->in_validation);
|
2011-07-22 20:41:52 +07:00
|
|
|
|
2012-02-16 08:11:40 +07:00
|
|
|
ret = tree->ops->submit_bio_hook(inode, read_mode, bio,
|
|
|
|
failrec->this_mirror,
|
|
|
|
failrec->bio_flags, 0);
|
2014-09-12 17:43:57 +07:00
|
|
|
if (ret) {
|
2014-09-12 17:43:58 +07:00
|
|
|
free_io_failure(inode, failrec);
|
2014-09-12 17:43:57 +07:00
|
|
|
bio_put(bio);
|
|
|
|
}
|
|
|
|
|
2012-02-16 08:11:40 +07:00
|
|
|
return ret;
|
2011-07-22 20:41:52 +07:00
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/* lots and lots of room for performance fixes in the end_bio funcs */
|
|
|
|
|
2012-02-15 22:23:57 +07:00
|
|
|
int end_extent_writepage(struct page *page, int err, u64 start, u64 end)
|
|
|
|
{
|
|
|
|
int uptodate = (err == 0);
|
|
|
|
struct extent_io_tree *tree;
|
2014-06-12 12:39:58 +07:00
|
|
|
int ret = 0;
|
2012-02-15 22:23:57 +07:00
|
|
|
|
|
|
|
tree = &BTRFS_I(page->mapping->host)->io_tree;
|
|
|
|
|
|
|
|
if (tree->ops && tree->ops->writepage_end_io_hook) {
|
|
|
|
ret = tree->ops->writepage_end_io_hook(page, start,
|
|
|
|
end, NULL, uptodate);
|
|
|
|
if (ret)
|
|
|
|
uptodate = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!uptodate) {
|
|
|
|
ClearPageUptodate(page);
|
|
|
|
SetPageError(page);
|
2014-05-12 11:47:36 +07:00
|
|
|
ret = ret < 0 ? ret : -EIO;
|
|
|
|
mapping_set_error(page->mapping, ret);
|
2012-02-15 22:23:57 +07:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* after a writepage IO is done, we need to:
|
|
|
|
* clear the uptodate bits on error
|
|
|
|
* clear the writeback bits in the extent tree for this IO
|
|
|
|
* end_page_writeback if the page has no more pending IO
|
|
|
|
*
|
|
|
|
* Scheduling is not allowed, so the extent state tree is expected
|
|
|
|
* to have one and only one object corresponding to this IO.
|
|
|
|
*/
|
|
|
|
static void end_bio_extent_writepage(struct bio *bio, int err)
|
|
|
|
{
|
2013-11-08 03:20:26 +07:00
|
|
|
struct bio_vec *bvec;
|
2008-01-25 04:13:08 +07:00
|
|
|
u64 start;
|
|
|
|
u64 end;
|
2013-11-08 03:20:26 +07:00
|
|
|
int i;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2013-11-08 03:20:26 +07:00
|
|
|
bio_for_each_segment_all(bvec, bio, i) {
|
2008-01-25 04:13:08 +07:00
|
|
|
struct page *page = bvec->bv_page;
|
2008-08-20 19:51:49 +07:00
|
|
|
|
2013-05-15 22:38:55 +07:00
|
|
|
/* We always issue full-page reads, but if some block
|
|
|
|
* in a page fails to read, blk_update_request() will
|
|
|
|
* advance bv_offset and adjust bv_len to compensate.
|
|
|
|
* Print a warning for nonzero offsets, and an error
|
|
|
|
* if they don't add up to a full page. */
|
2013-12-20 23:37:06 +07:00
|
|
|
if (bvec->bv_offset || bvec->bv_len != PAGE_CACHE_SIZE) {
|
|
|
|
if (bvec->bv_offset + bvec->bv_len != PAGE_CACHE_SIZE)
|
|
|
|
btrfs_err(BTRFS_I(page->mapping->host)->root->fs_info,
|
|
|
|
"partial page write in btrfs with offset %u and length %u",
|
|
|
|
bvec->bv_offset, bvec->bv_len);
|
|
|
|
else
|
|
|
|
btrfs_info(BTRFS_I(page->mapping->host)->root->fs_info,
|
|
|
|
"incomplete page write in btrfs with offset %u and "
|
|
|
|
"length %u",
|
|
|
|
bvec->bv_offset, bvec->bv_len);
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2013-05-15 22:38:55 +07:00
|
|
|
start = page_offset(page);
|
|
|
|
end = start + bvec->bv_offset + bvec->bv_len - 1;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2012-02-15 22:23:57 +07:00
|
|
|
if (end_extent_writepage(page, err, start, end))
|
|
|
|
continue;
|
2008-01-29 21:59:12 +07:00
|
|
|
|
2013-05-15 22:38:55 +07:00
|
|
|
end_page_writeback(page);
|
2013-11-08 03:20:26 +07:00
|
|
|
}
|
2008-09-24 22:48:04 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
bio_put(bio);
|
|
|
|
}
|
|
|
|
|
2013-07-25 18:22:35 +07:00
|
|
|
static void
|
|
|
|
endio_readpage_release_extent(struct extent_io_tree *tree, u64 start, u64 len,
|
|
|
|
int uptodate)
|
|
|
|
{
|
|
|
|
struct extent_state *cached = NULL;
|
|
|
|
u64 end = start + len - 1;
|
|
|
|
|
|
|
|
if (uptodate && tree->track_uptodate)
|
|
|
|
set_extent_uptodate(tree, start, end, &cached, GFP_ATOMIC);
|
|
|
|
unlock_extent_cached(tree, start, end, &cached, GFP_ATOMIC);
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* after a readpage IO is done, we need to:
|
|
|
|
* clear the uptodate bits on error
|
|
|
|
* set the uptodate bits if things worked
|
|
|
|
* set the page up to date if all extents in the tree are uptodate
|
|
|
|
* clear the lock bit in the extent tree
|
|
|
|
* unlock the page if there are no other extents locked for it
|
|
|
|
*
|
|
|
|
* Scheduling is not allowed, so the extent state tree is expected
|
|
|
|
* to have one and only one object corresponding to this IO.
|
|
|
|
*/
|
|
|
|
static void end_bio_extent_readpage(struct bio *bio, int err)
|
|
|
|
{
|
2013-11-08 03:20:26 +07:00
|
|
|
struct bio_vec *bvec;
|
2008-01-25 04:13:08 +07:00
|
|
|
int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
|
2013-07-25 18:22:34 +07:00
|
|
|
struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
|
2008-08-20 19:51:49 +07:00
|
|
|
struct extent_io_tree *tree;
|
2013-07-25 18:22:34 +07:00
|
|
|
u64 offset = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
u64 start;
|
|
|
|
u64 end;
|
2013-07-25 18:22:34 +07:00
|
|
|
u64 len;
|
2013-07-25 18:22:35 +07:00
|
|
|
u64 extent_start = 0;
|
|
|
|
u64 extent_len = 0;
|
2012-04-16 20:42:26 +07:00
|
|
|
int mirror;
|
2008-01-25 04:13:08 +07:00
|
|
|
int ret;
|
2013-11-08 03:20:26 +07:00
|
|
|
int i;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
Btrfs: move data checksumming into a dedicated tree
Btrfs stores checksums for each data block. Until now, they have
been stored in the subvolume trees, indexed by the inode that is
referencing the data block. This means that when we read the inode,
we've probably read in at least some checksums as well.
But, this has a few problems:
* The checksums are indexed by logical offset in the file. When
compression is on, this means we have to do the expensive checksumming
on the uncompressed data. It would be faster if we could checksum
the compressed data instead.
* If we implement encryption, we'll be checksumming the plain text and
storing that on disk. This is significantly less secure.
* For either compression or encryption, we have to get the plain text
back before we can verify the checksum as correct. This makes the raid
layer balancing and extent moving much more expensive.
* It makes the front end caching code more complex, as we have touch
the subvolume and inodes as we cache extents.
* There is potentitally one copy of the checksum in each subvolume
referencing an extent.
The solution used here is to store the extent checksums in a dedicated
tree. This allows us to index the checksums by phyiscal extent
start and length. It means:
* The checksum is against the data stored on disk, after any compression
or encryption is done.
* The checksum is stored in a central location, and can be verified without
following back references, or reading inodes.
This makes compression significantly faster by reducing the amount of
data that needs to be checksummed. It will also allow much faster
raid management code in general.
The checksums are indexed by a key with a fixed objectid (a magic value
in ctree.h) and offset set to the starting byte of the extent. This
allows us to copy the checksum items into the fsync log tree directly (or
any other tree), without having to invent a second format for them.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-09 04:58:54 +07:00
|
|
|
if (err)
|
|
|
|
uptodate = 0;
|
|
|
|
|
2013-11-08 03:20:26 +07:00
|
|
|
bio_for_each_segment_all(bvec, bio, i) {
|
2008-01-25 04:13:08 +07:00
|
|
|
struct page *page = bvec->bv_page;
|
2013-06-18 04:14:39 +07:00
|
|
|
struct inode *inode = page->mapping->host;
|
2011-04-06 17:02:20 +07:00
|
|
|
|
2012-09-12 03:23:05 +07:00
|
|
|
pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
|
2014-09-12 17:43:56 +07:00
|
|
|
"mirror=%u\n", (u64)bio->bi_iter.bi_sector, err,
|
2013-05-18 05:30:14 +07:00
|
|
|
io_bio->mirror_num);
|
2013-06-18 04:14:39 +07:00
|
|
|
tree = &BTRFS_I(inode)->io_tree;
|
2008-08-20 19:51:49 +07:00
|
|
|
|
2013-05-15 22:38:55 +07:00
|
|
|
/* We always issue full-page reads, but if some block
|
|
|
|
* in a page fails to read, blk_update_request() will
|
|
|
|
* advance bv_offset and adjust bv_len to compensate.
|
|
|
|
* Print a warning for nonzero offsets, and an error
|
|
|
|
* if they don't add up to a full page. */
|
2013-12-20 23:37:06 +07:00
|
|
|
if (bvec->bv_offset || bvec->bv_len != PAGE_CACHE_SIZE) {
|
|
|
|
if (bvec->bv_offset + bvec->bv_len != PAGE_CACHE_SIZE)
|
|
|
|
btrfs_err(BTRFS_I(page->mapping->host)->root->fs_info,
|
|
|
|
"partial page read in btrfs with offset %u and length %u",
|
|
|
|
bvec->bv_offset, bvec->bv_len);
|
|
|
|
else
|
|
|
|
btrfs_info(BTRFS_I(page->mapping->host)->root->fs_info,
|
|
|
|
"incomplete page read in btrfs with offset %u and "
|
|
|
|
"length %u",
|
|
|
|
bvec->bv_offset, bvec->bv_len);
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2013-05-15 22:38:55 +07:00
|
|
|
start = page_offset(page);
|
|
|
|
end = start + bvec->bv_offset + bvec->bv_len - 1;
|
2013-07-25 18:22:34 +07:00
|
|
|
len = bvec->bv_len;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2013-05-18 05:30:14 +07:00
|
|
|
mirror = io_bio->mirror_num;
|
2013-07-25 18:22:33 +07:00
|
|
|
if (likely(uptodate && tree->ops &&
|
|
|
|
tree->ops->readpage_end_io_hook)) {
|
2013-07-25 18:22:34 +07:00
|
|
|
ret = tree->ops->readpage_end_io_hook(io_bio, offset,
|
|
|
|
page, start, end,
|
|
|
|
mirror);
|
2012-08-27 21:30:03 +07:00
|
|
|
if (ret)
|
2008-01-25 04:13:08 +07:00
|
|
|
uptodate = 0;
|
2012-08-27 21:30:03 +07:00
|
|
|
else
|
2014-09-12 17:44:01 +07:00
|
|
|
clean_io_failure(inode, start, page, 0);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
2012-03-27 08:57:36 +07:00
|
|
|
|
2013-07-25 18:22:33 +07:00
|
|
|
if (likely(uptodate))
|
|
|
|
goto readpage_ok;
|
|
|
|
|
|
|
|
if (tree->ops && tree->ops->readpage_io_failed_hook) {
|
2012-04-16 20:42:26 +07:00
|
|
|
ret = tree->ops->readpage_io_failed_hook(page, mirror);
|
2012-03-27 08:57:36 +07:00
|
|
|
if (!ret && !err &&
|
|
|
|
test_bit(BIO_UPTODATE, &bio->bi_flags))
|
|
|
|
uptodate = 1;
|
2013-07-25 18:22:33 +07:00
|
|
|
} else {
|
2011-12-01 21:30:36 +07:00
|
|
|
/*
|
|
|
|
* The generic bio_readpage_error handles errors the
|
|
|
|
* following way: If possible, new read requests are
|
|
|
|
* created and submitted and will end up in
|
|
|
|
* end_bio_extent_readpage as well (if we're lucky, not
|
|
|
|
* in the !uptodate case). In that case it returns 0 and
|
|
|
|
* we just go on with the next page in our bio. If it
|
|
|
|
* can't handle the error it will return -EIO and we
|
|
|
|
* remain responsible for that page.
|
|
|
|
*/
|
2013-07-25 18:22:34 +07:00
|
|
|
ret = bio_readpage_error(bio, offset, page, start, end,
|
|
|
|
mirror);
|
2008-04-10 03:28:12 +07:00
|
|
|
if (ret == 0) {
|
2008-04-17 22:29:12 +07:00
|
|
|
uptodate =
|
|
|
|
test_bit(BIO_UPTODATE, &bio->bi_flags);
|
Btrfs: move data checksumming into a dedicated tree
Btrfs stores checksums for each data block. Until now, they have
been stored in the subvolume trees, indexed by the inode that is
referencing the data block. This means that when we read the inode,
we've probably read in at least some checksums as well.
But, this has a few problems:
* The checksums are indexed by logical offset in the file. When
compression is on, this means we have to do the expensive checksumming
on the uncompressed data. It would be faster if we could checksum
the compressed data instead.
* If we implement encryption, we'll be checksumming the plain text and
storing that on disk. This is significantly less secure.
* For either compression or encryption, we have to get the plain text
back before we can verify the checksum as correct. This makes the raid
layer balancing and extent moving much more expensive.
* It makes the front end caching code more complex, as we have touch
the subvolume and inodes as we cache extents.
* There is potentitally one copy of the checksum in each subvolume
referencing an extent.
The solution used here is to store the extent checksums in a dedicated
tree. This allows us to index the checksums by phyiscal extent
start and length. It means:
* The checksum is against the data stored on disk, after any compression
or encryption is done.
* The checksum is stored in a central location, and can be verified without
following back references, or reading inodes.
This makes compression significantly faster by reducing the amount of
data that needs to be checksummed. It will also allow much faster
raid management code in general.
The checksums are indexed by a key with a fixed objectid (a magic value
in ctree.h) and offset set to the starting byte of the extent. This
allows us to copy the checksum items into the fsync log tree directly (or
any other tree), without having to invent a second format for them.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-09 04:58:54 +07:00
|
|
|
if (err)
|
|
|
|
uptodate = 0;
|
2014-08-19 22:33:13 +07:00
|
|
|
offset += len;
|
2008-04-10 03:28:12 +07:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
}
|
2013-07-25 18:22:33 +07:00
|
|
|
readpage_ok:
|
2013-07-25 18:22:35 +07:00
|
|
|
if (likely(uptodate)) {
|
2013-06-18 04:14:39 +07:00
|
|
|
loff_t i_size = i_size_read(inode);
|
|
|
|
pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
|
2014-08-19 22:32:22 +07:00
|
|
|
unsigned off;
|
2013-06-18 04:14:39 +07:00
|
|
|
|
|
|
|
/* Zero out the end if this page straddles i_size */
|
2014-08-19 22:32:22 +07:00
|
|
|
off = i_size & (PAGE_CACHE_SIZE-1);
|
|
|
|
if (page->index == end_index && off)
|
|
|
|
zero_user_segment(page, off, PAGE_CACHE_SIZE);
|
2013-05-15 22:38:55 +07:00
|
|
|
SetPageUptodate(page);
|
2008-01-29 21:59:12 +07:00
|
|
|
} else {
|
2013-05-15 22:38:55 +07:00
|
|
|
ClearPageUptodate(page);
|
|
|
|
SetPageError(page);
|
2008-01-29 21:59:12 +07:00
|
|
|
}
|
2013-05-15 22:38:55 +07:00
|
|
|
unlock_page(page);
|
2013-07-25 18:22:34 +07:00
|
|
|
offset += len;
|
2013-07-25 18:22:35 +07:00
|
|
|
|
|
|
|
if (unlikely(!uptodate)) {
|
|
|
|
if (extent_len) {
|
|
|
|
endio_readpage_release_extent(tree,
|
|
|
|
extent_start,
|
|
|
|
extent_len, 1);
|
|
|
|
extent_start = 0;
|
|
|
|
extent_len = 0;
|
|
|
|
}
|
|
|
|
endio_readpage_release_extent(tree, start,
|
|
|
|
end - start + 1, 0);
|
|
|
|
} else if (!extent_len) {
|
|
|
|
extent_start = start;
|
|
|
|
extent_len = end + 1 - start;
|
|
|
|
} else if (extent_start + extent_len == start) {
|
|
|
|
extent_len += end + 1 - start;
|
|
|
|
} else {
|
|
|
|
endio_readpage_release_extent(tree, extent_start,
|
|
|
|
extent_len, uptodate);
|
|
|
|
extent_start = start;
|
|
|
|
extent_len = end + 1 - start;
|
|
|
|
}
|
2013-11-08 03:20:26 +07:00
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2013-07-25 18:22:35 +07:00
|
|
|
if (extent_len)
|
|
|
|
endio_readpage_release_extent(tree, extent_start, extent_len,
|
|
|
|
uptodate);
|
2013-07-25 18:22:34 +07:00
|
|
|
if (io_bio->end_io)
|
|
|
|
io_bio->end_io(io_bio, err);
|
2008-01-25 04:13:08 +07:00
|
|
|
bio_put(bio);
|
|
|
|
}
|
|
|
|
|
2013-05-18 05:30:14 +07:00
|
|
|
/*
|
|
|
|
* this allocates from the btrfs_bioset. We're returning a bio right now
|
|
|
|
* but you can call btrfs_io_bio for the appropriate container_of magic
|
|
|
|
*/
|
2010-11-22 10:02:55 +07:00
|
|
|
struct bio *
|
|
|
|
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
|
|
|
|
gfp_t gfp_flags)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2013-07-25 18:22:34 +07:00
|
|
|
struct btrfs_io_bio *btrfs_bio;
|
2008-01-25 04:13:08 +07:00
|
|
|
struct bio *bio;
|
|
|
|
|
2013-05-18 05:30:14 +07:00
|
|
|
bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
if (bio == NULL && (current->flags & PF_MEMALLOC)) {
|
2013-05-18 05:30:14 +07:00
|
|
|
while (!bio && (nr_vecs /= 2)) {
|
|
|
|
bio = bio_alloc_bioset(gfp_flags,
|
|
|
|
nr_vecs, btrfs_bioset);
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
if (bio) {
|
|
|
|
bio->bi_bdev = bdev;
|
2013-10-12 05:44:27 +07:00
|
|
|
bio->bi_iter.bi_sector = first_sector;
|
2013-07-25 18:22:34 +07:00
|
|
|
btrfs_bio = btrfs_io_bio(bio);
|
|
|
|
btrfs_bio->csum = NULL;
|
|
|
|
btrfs_bio->csum_allocated = NULL;
|
|
|
|
btrfs_bio->end_io = NULL;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
return bio;
|
|
|
|
}
|
|
|
|
|
2013-05-18 05:30:14 +07:00
|
|
|
struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask)
|
|
|
|
{
|
2014-09-12 17:43:54 +07:00
|
|
|
struct btrfs_io_bio *btrfs_bio;
|
|
|
|
struct bio *new;
|
2013-05-18 05:30:14 +07:00
|
|
|
|
2014-09-12 17:43:54 +07:00
|
|
|
new = bio_clone_bioset(bio, gfp_mask, btrfs_bioset);
|
|
|
|
if (new) {
|
|
|
|
btrfs_bio = btrfs_io_bio(new);
|
|
|
|
btrfs_bio->csum = NULL;
|
|
|
|
btrfs_bio->csum_allocated = NULL;
|
|
|
|
btrfs_bio->end_io = NULL;
|
|
|
|
}
|
|
|
|
return new;
|
|
|
|
}
|
2013-05-18 05:30:14 +07:00
|
|
|
|
|
|
|
/* this also allocates from the btrfs_bioset */
|
|
|
|
struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
|
|
|
|
{
|
2013-07-25 18:22:34 +07:00
|
|
|
struct btrfs_io_bio *btrfs_bio;
|
|
|
|
struct bio *bio;
|
|
|
|
|
|
|
|
bio = bio_alloc_bioset(gfp_mask, nr_iovecs, btrfs_bioset);
|
|
|
|
if (bio) {
|
|
|
|
btrfs_bio = btrfs_io_bio(bio);
|
|
|
|
btrfs_bio->csum = NULL;
|
|
|
|
btrfs_bio->csum_allocated = NULL;
|
|
|
|
btrfs_bio->end_io = NULL;
|
|
|
|
}
|
|
|
|
return bio;
|
2013-05-18 05:30:14 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2011-10-04 10:23:14 +07:00
|
|
|
static int __must_check submit_one_bio(int rw, struct bio *bio,
|
|
|
|
int mirror_num, unsigned long bio_flags)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
int ret = 0;
|
2008-01-29 21:59:12 +07:00
|
|
|
struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
|
|
|
|
struct page *page = bvec->bv_page;
|
|
|
|
struct extent_io_tree *tree = bio->bi_private;
|
|
|
|
u64 start;
|
|
|
|
|
2012-12-21 16:17:45 +07:00
|
|
|
start = page_offset(page) + bvec->bv_offset;
|
2008-01-29 21:59:12 +07:00
|
|
|
|
2008-08-20 19:51:49 +07:00
|
|
|
bio->bi_private = NULL;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
bio_get(bio);
|
|
|
|
|
2008-02-21 00:07:25 +07:00
|
|
|
if (tree->ops && tree->ops->submit_bio_hook)
|
2011-01-26 13:21:39 +07:00
|
|
|
ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
|
2010-05-25 20:48:28 +07:00
|
|
|
mirror_num, bio_flags, start);
|
2008-03-25 02:01:56 +07:00
|
|
|
else
|
2011-11-09 19:44:05 +07:00
|
|
|
btrfsic_submit_bio(rw, bio);
|
2011-07-22 20:41:52 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
if (bio_flagged(bio, BIO_EOPNOTSUPP))
|
|
|
|
ret = -EOPNOTSUPP;
|
|
|
|
bio_put(bio);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-07-16 05:29:37 +07:00
|
|
|
static int merge_bio(int rw, struct extent_io_tree *tree, struct page *page,
|
2011-10-04 10:23:13 +07:00
|
|
|
unsigned long offset, size_t size, struct bio *bio,
|
|
|
|
unsigned long bio_flags)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
if (tree->ops && tree->ops->merge_bio_hook)
|
2009-07-16 05:29:37 +07:00
|
|
|
ret = tree->ops->merge_bio_hook(rw, page, offset, size, bio,
|
2011-10-04 10:23:13 +07:00
|
|
|
bio_flags);
|
|
|
|
BUG_ON(ret < 0);
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
static int submit_extent_page(int rw, struct extent_io_tree *tree,
|
|
|
|
struct page *page, sector_t sector,
|
|
|
|
size_t size, unsigned long offset,
|
|
|
|
struct block_device *bdev,
|
|
|
|
struct bio **bio_ret,
|
|
|
|
unsigned long max_pages,
|
2008-04-10 03:28:12 +07:00
|
|
|
bio_end_io_t end_io_func,
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
int mirror_num,
|
|
|
|
unsigned long prev_bio_flags,
|
|
|
|
unsigned long bio_flags)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
struct bio *bio;
|
|
|
|
int nr;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
int contig = 0;
|
|
|
|
int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED;
|
|
|
|
int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
|
2008-11-11 21:34:41 +07:00
|
|
|
size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
if (bio_ret && *bio_ret) {
|
|
|
|
bio = *bio_ret;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
if (old_compressed)
|
2013-10-12 05:44:27 +07:00
|
|
|
contig = bio->bi_iter.bi_sector == sector;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
else
|
2012-09-26 05:05:12 +07:00
|
|
|
contig = bio_end_sector(bio) == sector;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
|
|
|
|
if (prev_bio_flags != bio_flags || !contig ||
|
2009-07-16 05:29:37 +07:00
|
|
|
merge_bio(rw, tree, page, offset, page_size, bio, bio_flags) ||
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
bio_add_page(bio, page, page_size, offset) < page_size) {
|
|
|
|
ret = submit_one_bio(rw, bio, mirror_num,
|
|
|
|
prev_bio_flags);
|
2012-03-12 22:03:00 +07:00
|
|
|
if (ret < 0)
|
|
|
|
return ret;
|
2008-01-25 04:13:08 +07:00
|
|
|
bio = NULL;
|
|
|
|
} else {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
}
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
if (this_compressed)
|
|
|
|
nr = BIO_MAX_PAGES;
|
|
|
|
else
|
|
|
|
nr = bio_get_nr_vecs(bdev);
|
|
|
|
|
2010-11-22 10:02:55 +07:00
|
|
|
bio = btrfs_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
|
2011-02-01 16:17:35 +07:00
|
|
|
if (!bio)
|
|
|
|
return -ENOMEM;
|
2008-01-29 21:59:12 +07:00
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
bio_add_page(bio, page, page_size, offset);
|
2008-01-25 04:13:08 +07:00
|
|
|
bio->bi_end_io = end_io_func;
|
|
|
|
bio->bi_private = tree;
|
2008-01-29 21:59:12 +07:00
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
if (bio_ret)
|
2008-01-25 04:13:08 +07:00
|
|
|
*bio_ret = bio;
|
2009-01-06 09:25:51 +07:00
|
|
|
else
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-04-26 03:41:01 +07:00
|
|
|
static void attach_extent_buffer_page(struct extent_buffer *eb,
|
|
|
|
struct page *page)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
if (!PagePrivate(page)) {
|
|
|
|
SetPagePrivate(page);
|
|
|
|
page_cache_get(page);
|
2012-03-08 04:20:05 +07:00
|
|
|
set_page_private(page, (unsigned long)eb);
|
|
|
|
} else {
|
|
|
|
WARN_ON(page->private != (unsigned long)eb);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-03-08 04:20:05 +07:00
|
|
|
void set_page_extent_mapped(struct page *page)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2012-03-08 04:20:05 +07:00
|
|
|
if (!PagePrivate(page)) {
|
|
|
|
SetPagePrivate(page);
|
|
|
|
page_cache_get(page);
|
|
|
|
set_page_private(page, EXTENT_PAGE_PRIVATE);
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2013-07-25 18:22:37 +07:00
|
|
|
static struct extent_map *
|
|
|
|
__get_extent_map(struct inode *inode, struct page *page, size_t pg_offset,
|
|
|
|
u64 start, u64 len, get_extent_t *get_extent,
|
|
|
|
struct extent_map **em_cached)
|
|
|
|
{
|
|
|
|
struct extent_map *em;
|
|
|
|
|
|
|
|
if (em_cached && *em_cached) {
|
|
|
|
em = *em_cached;
|
2014-02-25 21:15:12 +07:00
|
|
|
if (extent_map_in_tree(em) && start >= em->start &&
|
2013-07-25 18:22:37 +07:00
|
|
|
start < extent_map_end(em)) {
|
|
|
|
atomic_inc(&em->refs);
|
|
|
|
return em;
|
|
|
|
}
|
|
|
|
|
|
|
|
free_extent_map(em);
|
|
|
|
*em_cached = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
em = get_extent(inode, page, pg_offset, start, len, 0);
|
|
|
|
if (em_cached && !IS_ERR_OR_NULL(em)) {
|
|
|
|
BUG_ON(*em_cached);
|
|
|
|
atomic_inc(&em->refs);
|
|
|
|
*em_cached = em;
|
|
|
|
}
|
|
|
|
return em;
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* basic readpage implementation. Locked extent state structs are inserted
|
|
|
|
* into the tree that are removed when the IO is done (by the end_io
|
|
|
|
* handlers)
|
2012-03-12 22:03:00 +07:00
|
|
|
* XXX JDM: This needs looking at to ensure proper page locking
|
2008-01-25 04:13:08 +07:00
|
|
|
*/
|
2013-07-25 18:22:36 +07:00
|
|
|
static int __do_readpage(struct extent_io_tree *tree,
|
|
|
|
struct page *page,
|
|
|
|
get_extent_t *get_extent,
|
2013-07-25 18:22:37 +07:00
|
|
|
struct extent_map **em_cached,
|
2013-07-25 18:22:36 +07:00
|
|
|
struct bio **bio, int mirror_num,
|
|
|
|
unsigned long *bio_flags, int rw)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct inode *inode = page->mapping->host;
|
2012-12-21 16:17:45 +07:00
|
|
|
u64 start = page_offset(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
u64 page_end = start + PAGE_CACHE_SIZE - 1;
|
|
|
|
u64 end;
|
|
|
|
u64 cur = start;
|
|
|
|
u64 extent_offset;
|
|
|
|
u64 last_byte = i_size_read(inode);
|
|
|
|
u64 block_start;
|
|
|
|
u64 cur_end;
|
|
|
|
sector_t sector;
|
|
|
|
struct extent_map *em;
|
|
|
|
struct block_device *bdev;
|
|
|
|
int ret;
|
|
|
|
int nr = 0;
|
2013-08-07 01:42:50 +07:00
|
|
|
int parent_locked = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
|
2011-04-19 19:29:38 +07:00
|
|
|
size_t pg_offset = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
size_t iosize;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
size_t disk_io_size;
|
2008-01-25 04:13:08 +07:00
|
|
|
size_t blocksize = inode->i_sb->s_blocksize;
|
2013-08-07 01:42:50 +07:00
|
|
|
unsigned long this_bio_flag = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
set_page_extent_mapped(page);
|
|
|
|
|
2013-07-25 18:22:36 +07:00
|
|
|
end = page_end;
|
2011-05-26 23:01:56 +07:00
|
|
|
if (!PageUptodate(page)) {
|
|
|
|
if (cleancache_get_page(page) == 0) {
|
|
|
|
BUG_ON(blocksize != PAGE_SIZE);
|
2013-07-25 18:22:36 +07:00
|
|
|
unlock_extent(tree, start, end);
|
2011-05-26 23:01:56 +07:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
if (page->index == last_byte >> PAGE_CACHE_SHIFT) {
|
|
|
|
char *userpage;
|
|
|
|
size_t zero_offset = last_byte & (PAGE_CACHE_SIZE - 1);
|
|
|
|
|
|
|
|
if (zero_offset) {
|
|
|
|
iosize = PAGE_CACHE_SIZE - zero_offset;
|
2011-11-25 22:14:28 +07:00
|
|
|
userpage = kmap_atomic(page);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
memset(userpage + zero_offset, 0, iosize);
|
|
|
|
flush_dcache_page(page);
|
2011-11-25 22:14:28 +07:00
|
|
|
kunmap_atomic(userpage);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
}
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
while (cur <= end) {
|
2013-02-11 23:33:00 +07:00
|
|
|
unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
if (cur >= last_byte) {
|
|
|
|
char *userpage;
|
2011-04-06 17:02:20 +07:00
|
|
|
struct extent_state *cached = NULL;
|
|
|
|
|
2011-04-19 19:29:38 +07:00
|
|
|
iosize = PAGE_CACHE_SIZE - pg_offset;
|
2011-11-25 22:14:28 +07:00
|
|
|
userpage = kmap_atomic(page);
|
2011-04-19 19:29:38 +07:00
|
|
|
memset(userpage + pg_offset, 0, iosize);
|
2008-01-25 04:13:08 +07:00
|
|
|
flush_dcache_page(page);
|
2011-11-25 22:14:28 +07:00
|
|
|
kunmap_atomic(userpage);
|
2008-01-25 04:13:08 +07:00
|
|
|
set_extent_uptodate(tree, cur, cur + iosize - 1,
|
2011-04-06 17:02:20 +07:00
|
|
|
&cached, GFP_NOFS);
|
2013-08-07 01:42:50 +07:00
|
|
|
if (!parent_locked)
|
|
|
|
unlock_extent_cached(tree, cur,
|
|
|
|
cur + iosize - 1,
|
|
|
|
&cached, GFP_NOFS);
|
2008-01-25 04:13:08 +07:00
|
|
|
break;
|
|
|
|
}
|
2013-07-25 18:22:37 +07:00
|
|
|
em = __get_extent_map(inode, page, pg_offset, cur,
|
|
|
|
end - cur + 1, get_extent, em_cached);
|
2011-04-19 23:00:01 +07:00
|
|
|
if (IS_ERR_OR_NULL(em)) {
|
2008-01-25 04:13:08 +07:00
|
|
|
SetPageError(page);
|
2013-08-07 01:42:50 +07:00
|
|
|
if (!parent_locked)
|
|
|
|
unlock_extent(tree, cur, end);
|
2008-01-25 04:13:08 +07:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
extent_offset = cur - em->start;
|
|
|
|
BUG_ON(extent_map_end(em) <= cur);
|
|
|
|
BUG_ON(end < cur);
|
|
|
|
|
2010-12-17 13:21:50 +07:00
|
|
|
if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
|
2013-08-07 01:42:50 +07:00
|
|
|
this_bio_flag |= EXTENT_BIO_COMPRESSED;
|
2010-12-17 13:21:50 +07:00
|
|
|
extent_set_compress_type(&this_bio_flag,
|
|
|
|
em->compress_type);
|
|
|
|
}
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
iosize = min(extent_map_end(em) - cur, end - cur + 1);
|
|
|
|
cur_end = min(extent_map_end(em) - 1, end);
|
2013-02-26 15:10:22 +07:00
|
|
|
iosize = ALIGN(iosize, blocksize);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
if (this_bio_flag & EXTENT_BIO_COMPRESSED) {
|
|
|
|
disk_io_size = em->block_len;
|
|
|
|
sector = em->block_start >> 9;
|
|
|
|
} else {
|
|
|
|
sector = (em->block_start + extent_offset) >> 9;
|
|
|
|
disk_io_size = iosize;
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
bdev = em->bdev;
|
|
|
|
block_start = em->block_start;
|
2008-10-31 01:25:28 +07:00
|
|
|
if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
|
|
|
|
block_start = EXTENT_MAP_HOLE;
|
2008-01-25 04:13:08 +07:00
|
|
|
free_extent_map(em);
|
|
|
|
em = NULL;
|
|
|
|
|
|
|
|
/* we've found a hole, just zero and go on */
|
|
|
|
if (block_start == EXTENT_MAP_HOLE) {
|
|
|
|
char *userpage;
|
2011-04-06 17:02:20 +07:00
|
|
|
struct extent_state *cached = NULL;
|
|
|
|
|
2011-11-25 22:14:28 +07:00
|
|
|
userpage = kmap_atomic(page);
|
2011-04-19 19:29:38 +07:00
|
|
|
memset(userpage + pg_offset, 0, iosize);
|
2008-01-25 04:13:08 +07:00
|
|
|
flush_dcache_page(page);
|
2011-11-25 22:14:28 +07:00
|
|
|
kunmap_atomic(userpage);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
set_extent_uptodate(tree, cur, cur + iosize - 1,
|
2011-04-06 17:02:20 +07:00
|
|
|
&cached, GFP_NOFS);
|
|
|
|
unlock_extent_cached(tree, cur, cur + iosize - 1,
|
|
|
|
&cached, GFP_NOFS);
|
2008-01-25 04:13:08 +07:00
|
|
|
cur = cur + iosize;
|
2011-04-19 19:29:38 +07:00
|
|
|
pg_offset += iosize;
|
2008-01-25 04:13:08 +07:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
/* the get_extent function already copied into the page */
|
2009-09-03 02:22:30 +07:00
|
|
|
if (test_range_bit(tree, cur, cur_end,
|
|
|
|
EXTENT_UPTODATE, 1, NULL)) {
|
2008-09-06 03:09:51 +07:00
|
|
|
check_page_uptodate(tree, page);
|
2013-08-07 01:42:50 +07:00
|
|
|
if (!parent_locked)
|
|
|
|
unlock_extent(tree, cur, cur + iosize - 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
cur = cur + iosize;
|
2011-04-19 19:29:38 +07:00
|
|
|
pg_offset += iosize;
|
2008-01-25 04:13:08 +07:00
|
|
|
continue;
|
|
|
|
}
|
2008-01-29 21:59:12 +07:00
|
|
|
/* we have an inline extent but it didn't get marked up
|
|
|
|
* to date. Error out
|
|
|
|
*/
|
|
|
|
if (block_start == EXTENT_MAP_INLINE) {
|
|
|
|
SetPageError(page);
|
2013-08-07 01:42:50 +07:00
|
|
|
if (!parent_locked)
|
|
|
|
unlock_extent(tree, cur, cur + iosize - 1);
|
2008-01-29 21:59:12 +07:00
|
|
|
cur = cur + iosize;
|
2011-04-19 19:29:38 +07:00
|
|
|
pg_offset += iosize;
|
2008-01-29 21:59:12 +07:00
|
|
|
continue;
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2013-02-11 23:33:00 +07:00
|
|
|
pnr -= page->index;
|
2013-04-20 06:49:09 +07:00
|
|
|
ret = submit_extent_page(rw, tree, page,
|
2011-04-19 19:29:38 +07:00
|
|
|
sector, disk_io_size, pg_offset,
|
2008-07-24 20:41:53 +07:00
|
|
|
bdev, bio, pnr,
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
end_bio_extent_readpage, mirror_num,
|
|
|
|
*bio_flags,
|
|
|
|
this_bio_flag);
|
2013-02-11 23:33:00 +07:00
|
|
|
if (!ret) {
|
|
|
|
nr++;
|
|
|
|
*bio_flags = this_bio_flag;
|
|
|
|
} else {
|
2008-01-25 04:13:08 +07:00
|
|
|
SetPageError(page);
|
2013-08-07 01:42:50 +07:00
|
|
|
if (!parent_locked)
|
|
|
|
unlock_extent(tree, cur, cur + iosize - 1);
|
2012-10-06 03:40:32 +07:00
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
cur = cur + iosize;
|
2011-04-19 19:29:38 +07:00
|
|
|
pg_offset += iosize;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
2011-05-26 23:01:56 +07:00
|
|
|
out:
|
2008-01-25 04:13:08 +07:00
|
|
|
if (!nr) {
|
|
|
|
if (!PageError(page))
|
|
|
|
SetPageUptodate(page);
|
|
|
|
unlock_page(page);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-07-25 18:22:36 +07:00
|
|
|
static inline void __do_contiguous_readpages(struct extent_io_tree *tree,
|
|
|
|
struct page *pages[], int nr_pages,
|
|
|
|
u64 start, u64 end,
|
|
|
|
get_extent_t *get_extent,
|
2013-07-25 18:22:37 +07:00
|
|
|
struct extent_map **em_cached,
|
2013-07-25 18:22:36 +07:00
|
|
|
struct bio **bio, int mirror_num,
|
|
|
|
unsigned long *bio_flags, int rw)
|
|
|
|
{
|
|
|
|
struct inode *inode;
|
|
|
|
struct btrfs_ordered_extent *ordered;
|
|
|
|
int index;
|
|
|
|
|
|
|
|
inode = pages[0]->mapping->host;
|
|
|
|
while (1) {
|
|
|
|
lock_extent(tree, start, end);
|
|
|
|
ordered = btrfs_lookup_ordered_range(inode, start,
|
|
|
|
end - start + 1);
|
|
|
|
if (!ordered)
|
|
|
|
break;
|
|
|
|
unlock_extent(tree, start, end);
|
|
|
|
btrfs_start_ordered_extent(inode, ordered, 1);
|
|
|
|
btrfs_put_ordered_extent(ordered);
|
|
|
|
}
|
|
|
|
|
|
|
|
for (index = 0; index < nr_pages; index++) {
|
2013-07-25 18:22:37 +07:00
|
|
|
__do_readpage(tree, pages[index], get_extent, em_cached, bio,
|
|
|
|
mirror_num, bio_flags, rw);
|
2013-07-25 18:22:36 +07:00
|
|
|
page_cache_release(pages[index]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __extent_readpages(struct extent_io_tree *tree,
|
|
|
|
struct page *pages[],
|
|
|
|
int nr_pages, get_extent_t *get_extent,
|
2013-07-25 18:22:37 +07:00
|
|
|
struct extent_map **em_cached,
|
2013-07-25 18:22:36 +07:00
|
|
|
struct bio **bio, int mirror_num,
|
|
|
|
unsigned long *bio_flags, int rw)
|
|
|
|
{
|
2013-08-14 23:12:25 +07:00
|
|
|
u64 start = 0;
|
2013-07-25 18:22:36 +07:00
|
|
|
u64 end = 0;
|
|
|
|
u64 page_start;
|
|
|
|
int index;
|
2013-08-14 23:12:25 +07:00
|
|
|
int first_index = 0;
|
2013-07-25 18:22:36 +07:00
|
|
|
|
|
|
|
for (index = 0; index < nr_pages; index++) {
|
|
|
|
page_start = page_offset(pages[index]);
|
|
|
|
if (!end) {
|
|
|
|
start = page_start;
|
|
|
|
end = start + PAGE_CACHE_SIZE - 1;
|
|
|
|
first_index = index;
|
|
|
|
} else if (end + 1 == page_start) {
|
|
|
|
end += PAGE_CACHE_SIZE;
|
|
|
|
} else {
|
|
|
|
__do_contiguous_readpages(tree, &pages[first_index],
|
|
|
|
index - first_index, start,
|
2013-07-25 18:22:37 +07:00
|
|
|
end, get_extent, em_cached,
|
|
|
|
bio, mirror_num, bio_flags,
|
|
|
|
rw);
|
2013-07-25 18:22:36 +07:00
|
|
|
start = page_start;
|
|
|
|
end = start + PAGE_CACHE_SIZE - 1;
|
|
|
|
first_index = index;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (end)
|
|
|
|
__do_contiguous_readpages(tree, &pages[first_index],
|
|
|
|
index - first_index, start,
|
2013-07-25 18:22:37 +07:00
|
|
|
end, get_extent, em_cached, bio,
|
2013-07-25 18:22:36 +07:00
|
|
|
mirror_num, bio_flags, rw);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __extent_read_full_page(struct extent_io_tree *tree,
|
|
|
|
struct page *page,
|
|
|
|
get_extent_t *get_extent,
|
|
|
|
struct bio **bio, int mirror_num,
|
|
|
|
unsigned long *bio_flags, int rw)
|
|
|
|
{
|
|
|
|
struct inode *inode = page->mapping->host;
|
|
|
|
struct btrfs_ordered_extent *ordered;
|
|
|
|
u64 start = page_offset(page);
|
|
|
|
u64 end = start + PAGE_CACHE_SIZE - 1;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
lock_extent(tree, start, end);
|
|
|
|
ordered = btrfs_lookup_ordered_extent(inode, start);
|
|
|
|
if (!ordered)
|
|
|
|
break;
|
|
|
|
unlock_extent(tree, start, end);
|
|
|
|
btrfs_start_ordered_extent(inode, ordered, 1);
|
|
|
|
btrfs_put_ordered_extent(ordered);
|
|
|
|
}
|
|
|
|
|
2013-07-25 18:22:37 +07:00
|
|
|
ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
|
|
|
|
bio_flags, rw);
|
2013-07-25 18:22:36 +07:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
|
2011-06-14 01:02:58 +07:00
|
|
|
get_extent_t *get_extent, int mirror_num)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct bio *bio = NULL;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
unsigned long bio_flags = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
int ret;
|
|
|
|
|
2011-06-14 01:02:58 +07:00
|
|
|
ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
|
2013-04-20 06:49:09 +07:00
|
|
|
&bio_flags, READ);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (bio)
|
2011-06-14 01:02:58 +07:00
|
|
|
ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
|
2008-01-25 04:13:08 +07:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-08-07 01:42:50 +07:00
|
|
|
int extent_read_full_page_nolock(struct extent_io_tree *tree, struct page *page,
|
|
|
|
get_extent_t *get_extent, int mirror_num)
|
|
|
|
{
|
|
|
|
struct bio *bio = NULL;
|
|
|
|
unsigned long bio_flags = EXTENT_BIO_PARENT_LOCKED;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = __do_readpage(tree, page, get_extent, NULL, &bio, mirror_num,
|
|
|
|
&bio_flags, READ);
|
|
|
|
if (bio)
|
|
|
|
ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-04-21 02:50:09 +07:00
|
|
|
static noinline void update_nr_written(struct page *page,
|
|
|
|
struct writeback_control *wbc,
|
|
|
|
unsigned long nr_written)
|
|
|
|
{
|
|
|
|
wbc->nr_to_write -= nr_written;
|
|
|
|
if (wbc->range_cyclic || (wbc->nr_to_write > 0 &&
|
|
|
|
wbc->range_start == 0 && wbc->range_end == LLONG_MAX))
|
|
|
|
page->mapping->writeback_index = page->index + nr_written;
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
2014-05-22 03:35:51 +07:00
|
|
|
* helper for __extent_writepage, doing all of the delayed allocation setup.
|
|
|
|
*
|
|
|
|
* This returns 1 if our fill_delalloc function did all the work required
|
|
|
|
* to write the page (copy into inline extent). In this case the IO has
|
|
|
|
* been started and the page is already unlocked.
|
|
|
|
*
|
|
|
|
* This returns 0 if all went well (page still locked)
|
|
|
|
* This returns < 0 if there were errors (page still locked)
|
2008-01-25 04:13:08 +07:00
|
|
|
*/
|
2014-05-22 03:35:51 +07:00
|
|
|
static noinline_for_stack int writepage_delalloc(struct inode *inode,
|
|
|
|
struct page *page, struct writeback_control *wbc,
|
|
|
|
struct extent_page_data *epd,
|
|
|
|
u64 delalloc_start,
|
|
|
|
unsigned long *nr_written)
|
|
|
|
{
|
|
|
|
struct extent_io_tree *tree = epd->tree;
|
|
|
|
u64 page_end = delalloc_start + PAGE_CACHE_SIZE - 1;
|
|
|
|
u64 nr_delalloc;
|
|
|
|
u64 delalloc_to_write = 0;
|
|
|
|
u64 delalloc_end = 0;
|
|
|
|
int ret;
|
|
|
|
int page_started = 0;
|
|
|
|
|
|
|
|
if (epd->extent_locked || !tree->ops || !tree->ops->fill_delalloc)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
while (delalloc_end < page_end) {
|
|
|
|
nr_delalloc = find_lock_delalloc_range(inode, tree,
|
|
|
|
page,
|
|
|
|
&delalloc_start,
|
|
|
|
&delalloc_end,
|
|
|
|
128 * 1024 * 1024);
|
|
|
|
if (nr_delalloc == 0) {
|
|
|
|
delalloc_start = delalloc_end + 1;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
ret = tree->ops->fill_delalloc(inode, page,
|
|
|
|
delalloc_start,
|
|
|
|
delalloc_end,
|
|
|
|
&page_started,
|
|
|
|
nr_written);
|
|
|
|
/* File system has been set read-only */
|
|
|
|
if (ret) {
|
|
|
|
SetPageError(page);
|
|
|
|
/* fill_delalloc should be return < 0 for error
|
|
|
|
* but just in case, we use > 0 here meaning the
|
|
|
|
* IO is started, so we don't want to return > 0
|
|
|
|
* unless things are going well.
|
|
|
|
*/
|
|
|
|
ret = ret < 0 ? ret : -EIO;
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* delalloc_end is already one less than the total
|
|
|
|
* length, so we don't subtract one from
|
|
|
|
* PAGE_CACHE_SIZE
|
|
|
|
*/
|
|
|
|
delalloc_to_write += (delalloc_end - delalloc_start +
|
|
|
|
PAGE_CACHE_SIZE) >>
|
|
|
|
PAGE_CACHE_SHIFT;
|
|
|
|
delalloc_start = delalloc_end + 1;
|
|
|
|
}
|
|
|
|
if (wbc->nr_to_write < delalloc_to_write) {
|
|
|
|
int thresh = 8192;
|
|
|
|
|
|
|
|
if (delalloc_to_write < thresh * 2)
|
|
|
|
thresh = delalloc_to_write;
|
|
|
|
wbc->nr_to_write = min_t(u64, delalloc_to_write,
|
|
|
|
thresh);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* did the fill delalloc function already unlock and start
|
|
|
|
* the IO?
|
|
|
|
*/
|
|
|
|
if (page_started) {
|
|
|
|
/*
|
|
|
|
* we've unlocked the page, so we can't update
|
|
|
|
* the mapping's writeback index, just update
|
|
|
|
* nr_to_write.
|
|
|
|
*/
|
|
|
|
wbc->nr_to_write -= *nr_written;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = 0;
|
|
|
|
|
|
|
|
done:
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* helper for __extent_writepage. This calls the writepage start hooks,
|
|
|
|
* and does the loop to map the page into extents and bios.
|
|
|
|
*
|
|
|
|
* We return 1 if the IO is started and the page is unlocked,
|
|
|
|
* 0 if all went well (page still locked)
|
|
|
|
* < 0 if there were errors (page still locked)
|
|
|
|
*/
|
|
|
|
static noinline_for_stack int __extent_writepage_io(struct inode *inode,
|
|
|
|
struct page *page,
|
|
|
|
struct writeback_control *wbc,
|
|
|
|
struct extent_page_data *epd,
|
|
|
|
loff_t i_size,
|
|
|
|
unsigned long nr_written,
|
|
|
|
int write_flags, int *nr_ret)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct extent_io_tree *tree = epd->tree;
|
2012-12-21 16:17:45 +07:00
|
|
|
u64 start = page_offset(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
u64 page_end = start + PAGE_CACHE_SIZE - 1;
|
|
|
|
u64 end;
|
|
|
|
u64 cur = start;
|
|
|
|
u64 extent_offset;
|
|
|
|
u64 block_start;
|
|
|
|
u64 iosize;
|
|
|
|
sector_t sector;
|
2009-09-03 02:04:12 +07:00
|
|
|
struct extent_state *cached_state = NULL;
|
2008-01-25 04:13:08 +07:00
|
|
|
struct extent_map *em;
|
|
|
|
struct block_device *bdev;
|
2008-07-18 23:01:11 +07:00
|
|
|
size_t pg_offset = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
size_t blocksize;
|
2014-05-22 03:35:51 +07:00
|
|
|
int ret = 0;
|
|
|
|
int nr = 0;
|
|
|
|
bool compressed;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
|
2008-07-17 23:53:51 +07:00
|
|
|
if (tree->ops && tree->ops->writepage_start_hook) {
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
ret = tree->ops->writepage_start_hook(page, start,
|
|
|
|
page_end);
|
2012-02-15 22:23:57 +07:00
|
|
|
if (ret) {
|
|
|
|
/* Fixup worker will requeue */
|
|
|
|
if (ret == -EBUSY)
|
|
|
|
wbc->pages_skipped++;
|
|
|
|
else
|
|
|
|
redirty_page_for_writepage(wbc, page);
|
2014-05-22 03:35:51 +07:00
|
|
|
|
2009-04-21 02:50:09 +07:00
|
|
|
update_nr_written(page, wbc, nr_written);
|
2008-07-17 23:53:51 +07:00
|
|
|
unlock_page(page);
|
2014-05-22 03:35:51 +07:00
|
|
|
ret = 1;
|
2009-04-21 02:50:09 +07:00
|
|
|
goto done_unlocked;
|
2008-07-17 23:53:51 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-04-21 02:50:09 +07:00
|
|
|
/*
|
|
|
|
* we don't want to touch the inode after unlocking the page,
|
|
|
|
* so we update the mapping writeback index now
|
|
|
|
*/
|
|
|
|
update_nr_written(page, wbc, nr_written + 1);
|
2008-11-07 10:02:51 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
end = page_end;
|
2014-05-22 03:35:51 +07:00
|
|
|
if (i_size <= start) {
|
2008-07-17 23:53:50 +07:00
|
|
|
if (tree->ops && tree->ops->writepage_end_io_hook)
|
|
|
|
tree->ops->writepage_end_io_hook(page, start,
|
|
|
|
page_end, NULL, 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
|
|
|
blocksize = inode->i_sb->s_blocksize;
|
|
|
|
|
|
|
|
while (cur <= end) {
|
2014-05-22 03:35:51 +07:00
|
|
|
u64 em_end;
|
|
|
|
if (cur >= i_size) {
|
2008-07-17 23:53:50 +07:00
|
|
|
if (tree->ops && tree->ops->writepage_end_io_hook)
|
|
|
|
tree->ops->writepage_end_io_hook(page, cur,
|
|
|
|
page_end, NULL, 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
break;
|
|
|
|
}
|
2008-07-18 23:01:11 +07:00
|
|
|
em = epd->get_extent(inode, page, pg_offset, cur,
|
2008-01-25 04:13:08 +07:00
|
|
|
end - cur + 1, 1);
|
2011-04-19 23:00:01 +07:00
|
|
|
if (IS_ERR_OR_NULL(em)) {
|
2008-01-25 04:13:08 +07:00
|
|
|
SetPageError(page);
|
2014-05-09 23:17:40 +07:00
|
|
|
ret = PTR_ERR_OR_ZERO(em);
|
2008-01-25 04:13:08 +07:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
extent_offset = cur - em->start;
|
2014-05-22 03:35:51 +07:00
|
|
|
em_end = extent_map_end(em);
|
|
|
|
BUG_ON(em_end <= cur);
|
2008-01-25 04:13:08 +07:00
|
|
|
BUG_ON(end < cur);
|
2014-05-22 03:35:51 +07:00
|
|
|
iosize = min(em_end - cur, end - cur + 1);
|
2013-02-26 15:10:22 +07:00
|
|
|
iosize = ALIGN(iosize, blocksize);
|
2008-01-25 04:13:08 +07:00
|
|
|
sector = (em->block_start + extent_offset) >> 9;
|
|
|
|
bdev = em->bdev;
|
|
|
|
block_start = em->block_start;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
|
2008-01-25 04:13:08 +07:00
|
|
|
free_extent_map(em);
|
|
|
|
em = NULL;
|
|
|
|
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
/*
|
|
|
|
* compressed and inline extents are written through other
|
|
|
|
* paths in the FS
|
|
|
|
*/
|
|
|
|
if (compressed || block_start == EXTENT_MAP_HOLE ||
|
2008-01-25 04:13:08 +07:00
|
|
|
block_start == EXTENT_MAP_INLINE) {
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
/*
|
|
|
|
* end_io notification does not happen here for
|
|
|
|
* compressed extents
|
|
|
|
*/
|
|
|
|
if (!compressed && tree->ops &&
|
|
|
|
tree->ops->writepage_end_io_hook)
|
2008-07-17 23:53:50 +07:00
|
|
|
tree->ops->writepage_end_io_hook(page, cur,
|
|
|
|
cur + iosize - 1,
|
|
|
|
NULL, 1);
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
else if (compressed) {
|
|
|
|
/* we don't want to end_page_writeback on
|
|
|
|
* a compressed extent. this happens
|
|
|
|
* elsewhere
|
|
|
|
*/
|
|
|
|
nr++;
|
|
|
|
}
|
|
|
|
|
|
|
|
cur += iosize;
|
2008-07-18 23:01:11 +07:00
|
|
|
pg_offset += iosize;
|
2008-01-25 04:13:08 +07:00
|
|
|
continue;
|
|
|
|
}
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
if (tree->ops && tree->ops->writepage_io_hook) {
|
|
|
|
ret = tree->ops->writepage_io_hook(page, cur,
|
|
|
|
cur + iosize - 1);
|
|
|
|
} else {
|
|
|
|
ret = 0;
|
|
|
|
}
|
2008-05-13 00:39:03 +07:00
|
|
|
if (ret) {
|
2008-01-25 04:13:08 +07:00
|
|
|
SetPageError(page);
|
2008-05-13 00:39:03 +07:00
|
|
|
} else {
|
2014-05-22 03:35:51 +07:00
|
|
|
unsigned long max_nr = (i_size >> PAGE_CACHE_SHIFT) + 1;
|
2008-07-18 23:01:11 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
set_range_writeback(tree, cur, cur + iosize - 1);
|
|
|
|
if (!PageWriteback(page)) {
|
2013-12-20 23:37:06 +07:00
|
|
|
btrfs_err(BTRFS_I(inode)->root->fs_info,
|
|
|
|
"page %lu not writeback, cur %llu end %llu",
|
2013-08-20 18:20:07 +07:00
|
|
|
page->index, cur, end);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2009-04-21 02:50:09 +07:00
|
|
|
ret = submit_extent_page(write_flags, tree, page,
|
|
|
|
sector, iosize, pg_offset,
|
|
|
|
bdev, &epd->bio, max_nr,
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
end_bio_extent_writepage,
|
|
|
|
0, 0, 0);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (ret)
|
|
|
|
SetPageError(page);
|
|
|
|
}
|
|
|
|
cur = cur + iosize;
|
2008-07-18 23:01:11 +07:00
|
|
|
pg_offset += iosize;
|
2008-01-25 04:13:08 +07:00
|
|
|
nr++;
|
|
|
|
}
|
2014-05-22 03:35:51 +07:00
|
|
|
done:
|
|
|
|
*nr_ret = nr;
|
|
|
|
|
|
|
|
done_unlocked:
|
|
|
|
|
|
|
|
/* drop our reference on any cached states */
|
|
|
|
free_extent_state(cached_state);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* the writepage semantics are similar to regular writepage. extent
|
|
|
|
* records are inserted to lock ranges in the tree, and as dirty areas
|
|
|
|
* are found, they are marked writeback. Then the lock bits are removed
|
|
|
|
* and the end_io handler clears the writeback ranges
|
|
|
|
*/
|
|
|
|
static int __extent_writepage(struct page *page, struct writeback_control *wbc,
|
|
|
|
void *data)
|
|
|
|
{
|
|
|
|
struct inode *inode = page->mapping->host;
|
|
|
|
struct extent_page_data *epd = data;
|
|
|
|
u64 start = page_offset(page);
|
|
|
|
u64 page_end = start + PAGE_CACHE_SIZE - 1;
|
|
|
|
int ret;
|
|
|
|
int nr = 0;
|
|
|
|
size_t pg_offset = 0;
|
|
|
|
loff_t i_size = i_size_read(inode);
|
|
|
|
unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
|
|
|
|
int write_flags;
|
|
|
|
unsigned long nr_written = 0;
|
|
|
|
|
|
|
|
if (wbc->sync_mode == WB_SYNC_ALL)
|
|
|
|
write_flags = WRITE_SYNC;
|
|
|
|
else
|
|
|
|
write_flags = WRITE;
|
|
|
|
|
|
|
|
trace___extent_writepage(page, inode, wbc);
|
|
|
|
|
|
|
|
WARN_ON(!PageLocked(page));
|
|
|
|
|
|
|
|
ClearPageError(page);
|
|
|
|
|
|
|
|
pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
|
|
|
|
if (page->index > end_index ||
|
|
|
|
(page->index == end_index && !pg_offset)) {
|
|
|
|
page->mapping->a_ops->invalidatepage(page, 0, PAGE_CACHE_SIZE);
|
|
|
|
unlock_page(page);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (page->index == end_index) {
|
|
|
|
char *userpage;
|
|
|
|
|
|
|
|
userpage = kmap_atomic(page);
|
|
|
|
memset(userpage + pg_offset, 0,
|
|
|
|
PAGE_CACHE_SIZE - pg_offset);
|
|
|
|
kunmap_atomic(userpage);
|
|
|
|
flush_dcache_page(page);
|
|
|
|
}
|
|
|
|
|
|
|
|
pg_offset = 0;
|
|
|
|
|
|
|
|
set_page_extent_mapped(page);
|
|
|
|
|
|
|
|
ret = writepage_delalloc(inode, page, wbc, epd, start, &nr_written);
|
|
|
|
if (ret == 1)
|
|
|
|
goto done_unlocked;
|
|
|
|
if (ret)
|
|
|
|
goto done;
|
|
|
|
|
|
|
|
ret = __extent_writepage_io(inode, page, wbc, epd,
|
|
|
|
i_size, nr_written, write_flags, &nr);
|
|
|
|
if (ret == 1)
|
|
|
|
goto done_unlocked;
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
done:
|
|
|
|
if (nr == 0) {
|
|
|
|
/* make sure the mapping tag for page dirty gets cleared */
|
|
|
|
set_page_writeback(page);
|
|
|
|
end_page_writeback(page);
|
|
|
|
}
|
2014-05-09 23:17:40 +07:00
|
|
|
if (PageError(page)) {
|
|
|
|
ret = ret < 0 ? ret : -EIO;
|
|
|
|
end_extent_writepage(page, ret, start, page_end);
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
unlock_page(page);
|
2014-05-22 03:35:51 +07:00
|
|
|
return ret;
|
2008-11-07 10:02:51 +07:00
|
|
|
|
2009-04-21 02:50:09 +07:00
|
|
|
done_unlocked:
|
2008-01-25 04:13:08 +07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-04-25 03:41:19 +07:00
|
|
|
void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
|
2012-03-13 20:38:00 +07:00
|
|
|
{
|
sched: Remove proliferation of wait_on_bit() action functions
The current "wait_on_bit" interface requires an 'action'
function to be provided which does the actual waiting.
There are over 20 such functions, many of them identical.
Most cases can be satisfied by one of just two functions, one
which uses io_schedule() and one which just uses schedule().
So:
Rename wait_on_bit and wait_on_bit_lock to
wait_on_bit_action and wait_on_bit_lock_action
to make it explicit that they need an action function.
Introduce new wait_on_bit{,_lock} and wait_on_bit{,_lock}_io
which are *not* given an action function but implicitly use
a standard one.
The decision to error-out if a signal is pending is now made
based on the 'mode' argument rather than being encoded in the action
function.
All instances of the old wait_on_bit and wait_on_bit_lock which
can use the new version have been changed accordingly and their
action functions have been discarded.
wait_on_bit{_lock} does not return any specific error code in the
event of a signal so the caller must check for non-zero and
interpolate their own error code as appropriate.
The wait_on_bit() call in __fscache_wait_on_invalidate() was
ambiguous as it specified TASK_UNINTERRUPTIBLE but used
fscache_wait_bit_interruptible as an action function.
David Howells confirms this should be uniformly
"uninterruptible"
The main remaining user of wait_on_bit{,_lock}_action is NFS
which needs to use a freezer-aware schedule() call.
A comment in fs/gfs2/glock.c notes that having multiple 'action'
functions is useful as they display differently in the 'wchan'
field of 'ps'. (and /proc/$PID/wchan).
As the new bit_wait{,_io} functions are tagged "__sched", they
will not show up at all, but something higher in the stack. So
the distinction will still be visible, only with different
function names (gds2_glock_wait versus gfs2_glock_dq_wait in the
gfs2/glock.c case).
Since first version of this patch (against 3.15) two new action
functions appeared, on in NFS and one in CIFS. CIFS also now
uses an action function that makes the same freezer aware
schedule call as NFS.
Signed-off-by: NeilBrown <neilb@suse.de>
Acked-by: David Howells <dhowells@redhat.com> (fscache, keys)
Acked-by: Steven Whitehouse <swhiteho@redhat.com> (gfs2)
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Steve French <sfrench@samba.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20140707051603.28027.72349.stgit@notabene.brown
Signed-off-by: Ingo Molnar <mingo@kernel.org>
2014-07-07 12:16:04 +07:00
|
|
|
wait_on_bit_io(&eb->bflags, EXTENT_BUFFER_WRITEBACK,
|
|
|
|
TASK_UNINTERRUPTIBLE);
|
2012-03-13 20:38:00 +07:00
|
|
|
}
|
|
|
|
|
2014-05-20 10:55:27 +07:00
|
|
|
static noinline_for_stack int
|
|
|
|
lock_extent_buffer_for_io(struct extent_buffer *eb,
|
|
|
|
struct btrfs_fs_info *fs_info,
|
|
|
|
struct extent_page_data *epd)
|
2012-03-13 20:38:00 +07:00
|
|
|
{
|
|
|
|
unsigned long i, num_pages;
|
|
|
|
int flush = 0;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
if (!btrfs_try_tree_write_lock(eb)) {
|
|
|
|
flush = 1;
|
|
|
|
flush_write_bio(epd);
|
|
|
|
btrfs_tree_lock(eb);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags)) {
|
|
|
|
btrfs_tree_unlock(eb);
|
|
|
|
if (!epd->sync_io)
|
|
|
|
return 0;
|
|
|
|
if (!flush) {
|
|
|
|
flush_write_bio(epd);
|
|
|
|
flush = 1;
|
|
|
|
}
|
2012-03-21 23:09:56 +07:00
|
|
|
while (1) {
|
|
|
|
wait_on_extent_buffer_writeback(eb);
|
|
|
|
btrfs_tree_lock(eb);
|
|
|
|
if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
|
|
|
|
break;
|
2012-03-13 20:38:00 +07:00
|
|
|
btrfs_tree_unlock(eb);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-07-21 03:25:24 +07:00
|
|
|
/*
|
|
|
|
* We need to do this to prevent races in people who check if the eb is
|
|
|
|
* under IO since we can end up having no IO bits set for a short period
|
|
|
|
* of time.
|
|
|
|
*/
|
|
|
|
spin_lock(&eb->refs_lock);
|
2012-03-13 20:38:00 +07:00
|
|
|
if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
|
|
|
|
set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
|
2012-07-21 03:25:24 +07:00
|
|
|
spin_unlock(&eb->refs_lock);
|
2012-03-13 20:38:00 +07:00
|
|
|
btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
|
2013-01-29 17:09:20 +07:00
|
|
|
__percpu_counter_add(&fs_info->dirty_metadata_bytes,
|
|
|
|
-eb->len,
|
|
|
|
fs_info->dirty_metadata_batch);
|
2012-03-13 20:38:00 +07:00
|
|
|
ret = 1;
|
2012-07-21 03:25:24 +07:00
|
|
|
} else {
|
|
|
|
spin_unlock(&eb->refs_lock);
|
2012-03-13 20:38:00 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_tree_unlock(eb);
|
|
|
|
|
|
|
|
if (!ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
num_pages = num_extent_pages(eb->start, eb->len);
|
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
struct page *p = extent_buffer_page(eb, i);
|
|
|
|
|
|
|
|
if (!trylock_page(p)) {
|
|
|
|
if (!flush) {
|
|
|
|
flush_write_bio(epd);
|
|
|
|
flush = 1;
|
|
|
|
}
|
|
|
|
lock_page(p);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void end_extent_buffer_writeback(struct extent_buffer *eb)
|
|
|
|
{
|
|
|
|
clear_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
|
2014-03-18 00:06:10 +07:00
|
|
|
smp_mb__after_atomic();
|
2012-03-13 20:38:00 +07:00
|
|
|
wake_up_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void end_bio_extent_buffer_writepage(struct bio *bio, int err)
|
|
|
|
{
|
2013-11-08 03:20:26 +07:00
|
|
|
struct bio_vec *bvec;
|
2012-03-13 20:38:00 +07:00
|
|
|
struct extent_buffer *eb;
|
2013-11-08 03:20:26 +07:00
|
|
|
int i, done;
|
2012-03-13 20:38:00 +07:00
|
|
|
|
2013-11-08 03:20:26 +07:00
|
|
|
bio_for_each_segment_all(bvec, bio, i) {
|
2012-03-13 20:38:00 +07:00
|
|
|
struct page *page = bvec->bv_page;
|
|
|
|
|
|
|
|
eb = (struct extent_buffer *)page->private;
|
|
|
|
BUG_ON(!eb);
|
|
|
|
done = atomic_dec_and_test(&eb->io_pages);
|
|
|
|
|
2013-11-08 03:20:26 +07:00
|
|
|
if (err || test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
|
2012-03-13 20:38:00 +07:00
|
|
|
set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
|
|
|
|
ClearPageUptodate(page);
|
|
|
|
SetPageError(page);
|
|
|
|
}
|
|
|
|
|
|
|
|
end_page_writeback(page);
|
|
|
|
|
|
|
|
if (!done)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
end_extent_buffer_writeback(eb);
|
2013-11-08 03:20:26 +07:00
|
|
|
}
|
2012-03-13 20:38:00 +07:00
|
|
|
|
|
|
|
bio_put(bio);
|
|
|
|
}
|
|
|
|
|
2014-05-20 10:55:27 +07:00
|
|
|
static noinline_for_stack int write_one_eb(struct extent_buffer *eb,
|
2012-03-13 20:38:00 +07:00
|
|
|
struct btrfs_fs_info *fs_info,
|
|
|
|
struct writeback_control *wbc,
|
|
|
|
struct extent_page_data *epd)
|
|
|
|
{
|
|
|
|
struct block_device *bdev = fs_info->fs_devices->latest_bdev;
|
2013-12-17 01:24:27 +07:00
|
|
|
struct extent_io_tree *tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
|
2012-03-13 20:38:00 +07:00
|
|
|
u64 offset = eb->start;
|
|
|
|
unsigned long i, num_pages;
|
2012-09-26 01:25:58 +07:00
|
|
|
unsigned long bio_flags = 0;
|
2013-04-20 06:49:09 +07:00
|
|
|
int rw = (epd->sync_io ? WRITE_SYNC : WRITE) | REQ_META;
|
2012-04-24 01:00:51 +07:00
|
|
|
int ret = 0;
|
2012-03-13 20:38:00 +07:00
|
|
|
|
|
|
|
clear_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
|
|
|
|
num_pages = num_extent_pages(eb->start, eb->len);
|
|
|
|
atomic_set(&eb->io_pages, num_pages);
|
2012-09-26 01:25:58 +07:00
|
|
|
if (btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID)
|
|
|
|
bio_flags = EXTENT_BIO_TREE_LOG;
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
struct page *p = extent_buffer_page(eb, i);
|
|
|
|
|
|
|
|
clear_page_dirty_for_io(p);
|
|
|
|
set_page_writeback(p);
|
2013-12-17 01:24:27 +07:00
|
|
|
ret = submit_extent_page(rw, tree, p, offset >> 9,
|
2012-03-13 20:38:00 +07:00
|
|
|
PAGE_CACHE_SIZE, 0, bdev, &epd->bio,
|
|
|
|
-1, end_bio_extent_buffer_writepage,
|
2012-09-26 01:25:58 +07:00
|
|
|
0, epd->bio_flags, bio_flags);
|
|
|
|
epd->bio_flags = bio_flags;
|
2012-03-13 20:38:00 +07:00
|
|
|
if (ret) {
|
|
|
|
set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
|
|
|
|
SetPageError(p);
|
|
|
|
if (atomic_sub_and_test(num_pages - i, &eb->io_pages))
|
|
|
|
end_extent_buffer_writeback(eb);
|
|
|
|
ret = -EIO;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
offset += PAGE_CACHE_SIZE;
|
|
|
|
update_nr_written(p, wbc, 1);
|
|
|
|
unlock_page(p);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (unlikely(ret)) {
|
|
|
|
for (; i < num_pages; i++) {
|
|
|
|
struct page *p = extent_buffer_page(eb, i);
|
|
|
|
unlock_page(p);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btree_write_cache_pages(struct address_space *mapping,
|
|
|
|
struct writeback_control *wbc)
|
|
|
|
{
|
|
|
|
struct extent_io_tree *tree = &BTRFS_I(mapping->host)->io_tree;
|
|
|
|
struct btrfs_fs_info *fs_info = BTRFS_I(mapping->host)->root->fs_info;
|
|
|
|
struct extent_buffer *eb, *prev_eb = NULL;
|
|
|
|
struct extent_page_data epd = {
|
|
|
|
.bio = NULL,
|
|
|
|
.tree = tree,
|
|
|
|
.extent_locked = 0,
|
|
|
|
.sync_io = wbc->sync_mode == WB_SYNC_ALL,
|
2012-09-26 01:25:58 +07:00
|
|
|
.bio_flags = 0,
|
2012-03-13 20:38:00 +07:00
|
|
|
};
|
|
|
|
int ret = 0;
|
|
|
|
int done = 0;
|
|
|
|
int nr_to_write_done = 0;
|
|
|
|
struct pagevec pvec;
|
|
|
|
int nr_pages;
|
|
|
|
pgoff_t index;
|
|
|
|
pgoff_t end; /* Inclusive */
|
|
|
|
int scanned = 0;
|
|
|
|
int tag;
|
|
|
|
|
|
|
|
pagevec_init(&pvec, 0);
|
|
|
|
if (wbc->range_cyclic) {
|
|
|
|
index = mapping->writeback_index; /* Start from prev offset */
|
|
|
|
end = -1;
|
|
|
|
} else {
|
|
|
|
index = wbc->range_start >> PAGE_CACHE_SHIFT;
|
|
|
|
end = wbc->range_end >> PAGE_CACHE_SHIFT;
|
|
|
|
scanned = 1;
|
|
|
|
}
|
|
|
|
if (wbc->sync_mode == WB_SYNC_ALL)
|
|
|
|
tag = PAGECACHE_TAG_TOWRITE;
|
|
|
|
else
|
|
|
|
tag = PAGECACHE_TAG_DIRTY;
|
|
|
|
retry:
|
|
|
|
if (wbc->sync_mode == WB_SYNC_ALL)
|
|
|
|
tag_pages_for_writeback(mapping, index, end);
|
|
|
|
while (!done && !nr_to_write_done && (index <= end) &&
|
|
|
|
(nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
|
|
|
|
min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
|
|
|
|
unsigned i;
|
|
|
|
|
|
|
|
scanned = 1;
|
|
|
|
for (i = 0; i < nr_pages; i++) {
|
|
|
|
struct page *page = pvec.pages[i];
|
|
|
|
|
|
|
|
if (!PagePrivate(page))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (!wbc->range_cyclic && page->index > end) {
|
|
|
|
done = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2012-09-15 00:43:01 +07:00
|
|
|
spin_lock(&mapping->private_lock);
|
|
|
|
if (!PagePrivate(page)) {
|
|
|
|
spin_unlock(&mapping->private_lock);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
eb = (struct extent_buffer *)page->private;
|
2012-09-15 00:43:01 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Shouldn't happen and normally this would be a BUG_ON
|
|
|
|
* but no sense in crashing the users box for something
|
|
|
|
* we can survive anyway.
|
|
|
|
*/
|
2013-10-31 12:00:08 +07:00
|
|
|
if (WARN_ON(!eb)) {
|
2012-09-15 00:43:01 +07:00
|
|
|
spin_unlock(&mapping->private_lock);
|
2012-03-13 20:38:00 +07:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2012-09-15 00:43:01 +07:00
|
|
|
if (eb == prev_eb) {
|
|
|
|
spin_unlock(&mapping->private_lock);
|
2012-03-13 20:38:00 +07:00
|
|
|
continue;
|
2012-09-15 00:43:01 +07:00
|
|
|
}
|
2012-03-13 20:38:00 +07:00
|
|
|
|
2012-09-15 00:43:01 +07:00
|
|
|
ret = atomic_inc_not_zero(&eb->refs);
|
|
|
|
spin_unlock(&mapping->private_lock);
|
|
|
|
if (!ret)
|
2012-03-13 20:38:00 +07:00
|
|
|
continue;
|
|
|
|
|
|
|
|
prev_eb = eb;
|
|
|
|
ret = lock_extent_buffer_for_io(eb, fs_info, &epd);
|
|
|
|
if (!ret) {
|
|
|
|
free_extent_buffer(eb);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = write_one_eb(eb, fs_info, wbc, &epd);
|
|
|
|
if (ret) {
|
|
|
|
done = 1;
|
|
|
|
free_extent_buffer(eb);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
free_extent_buffer(eb);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* the filesystem may choose to bump up nr_to_write.
|
|
|
|
* We have to make sure to honor the new nr_to_write
|
|
|
|
* at any time
|
|
|
|
*/
|
|
|
|
nr_to_write_done = wbc->nr_to_write <= 0;
|
|
|
|
}
|
|
|
|
pagevec_release(&pvec);
|
|
|
|
cond_resched();
|
|
|
|
}
|
|
|
|
if (!scanned && !done) {
|
|
|
|
/*
|
|
|
|
* We hit the last page and there is more work to be done: wrap
|
|
|
|
* back to the start of the file
|
|
|
|
*/
|
|
|
|
scanned = 1;
|
|
|
|
index = 0;
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
flush_write_bio(&epd);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/**
|
2008-09-08 22:18:08 +07:00
|
|
|
* write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
|
2008-01-25 04:13:08 +07:00
|
|
|
* @mapping: address space structure to write
|
|
|
|
* @wbc: subtract the number of written pages from *@wbc->nr_to_write
|
|
|
|
* @writepage: function called for each page
|
|
|
|
* @data: data passed to writepage function
|
|
|
|
*
|
|
|
|
* If a page is already under I/O, write_cache_pages() skips it, even
|
|
|
|
* if it's dirty. This is desirable behaviour for memory-cleaning writeback,
|
|
|
|
* but it is INCORRECT for data-integrity system calls such as fsync(). fsync()
|
|
|
|
* and msync() need to guarantee that all the data which was dirty at the time
|
|
|
|
* the call was made get new I/O started against them. If wbc->sync_mode is
|
|
|
|
* WB_SYNC_ALL then we were called for data integrity and we must wait for
|
|
|
|
* existing IO to complete.
|
|
|
|
*/
|
2008-12-02 21:54:17 +07:00
|
|
|
static int extent_write_cache_pages(struct extent_io_tree *tree,
|
2008-09-08 22:18:08 +07:00
|
|
|
struct address_space *mapping,
|
|
|
|
struct writeback_control *wbc,
|
2008-11-20 00:44:22 +07:00
|
|
|
writepage_t writepage, void *data,
|
|
|
|
void (*flush_fn)(void *))
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2012-06-28 04:18:41 +07:00
|
|
|
struct inode *inode = mapping->host;
|
2008-01-25 04:13:08 +07:00
|
|
|
int ret = 0;
|
|
|
|
int done = 0;
|
2014-05-09 23:17:40 +07:00
|
|
|
int err = 0;
|
2009-09-19 03:03:16 +07:00
|
|
|
int nr_to_write_done = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
struct pagevec pvec;
|
|
|
|
int nr_pages;
|
|
|
|
pgoff_t index;
|
|
|
|
pgoff_t end; /* Inclusive */
|
|
|
|
int scanned = 0;
|
2011-07-16 04:26:38 +07:00
|
|
|
int tag;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2012-06-28 04:18:41 +07:00
|
|
|
/*
|
|
|
|
* We have to hold onto the inode so that ordered extents can do their
|
|
|
|
* work when the IO finishes. The alternative to this is failing to add
|
|
|
|
* an ordered extent if the igrab() fails there and that is a huge pain
|
|
|
|
* to deal with, so instead just hold onto the inode throughout the
|
|
|
|
* writepages operation. If it fails here we are freeing up the inode
|
|
|
|
* anyway and we'd rather not waste our time writing out stuff that is
|
|
|
|
* going to be truncated anyway.
|
|
|
|
*/
|
|
|
|
if (!igrab(inode))
|
|
|
|
return 0;
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
pagevec_init(&pvec, 0);
|
|
|
|
if (wbc->range_cyclic) {
|
|
|
|
index = mapping->writeback_index; /* Start from prev offset */
|
|
|
|
end = -1;
|
|
|
|
} else {
|
|
|
|
index = wbc->range_start >> PAGE_CACHE_SHIFT;
|
|
|
|
end = wbc->range_end >> PAGE_CACHE_SHIFT;
|
|
|
|
scanned = 1;
|
|
|
|
}
|
2011-07-16 04:26:38 +07:00
|
|
|
if (wbc->sync_mode == WB_SYNC_ALL)
|
|
|
|
tag = PAGECACHE_TAG_TOWRITE;
|
|
|
|
else
|
|
|
|
tag = PAGECACHE_TAG_DIRTY;
|
2008-01-25 04:13:08 +07:00
|
|
|
retry:
|
2011-07-16 04:26:38 +07:00
|
|
|
if (wbc->sync_mode == WB_SYNC_ALL)
|
|
|
|
tag_pages_for_writeback(mapping, index, end);
|
2009-09-19 03:03:16 +07:00
|
|
|
while (!done && !nr_to_write_done && (index <= end) &&
|
2011-07-16 04:26:38 +07:00
|
|
|
(nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
|
|
|
|
min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
|
2008-01-25 04:13:08 +07:00
|
|
|
unsigned i;
|
|
|
|
|
|
|
|
scanned = 1;
|
|
|
|
for (i = 0; i < nr_pages; i++) {
|
|
|
|
struct page *page = pvec.pages[i];
|
|
|
|
|
|
|
|
/*
|
|
|
|
* At this point we hold neither mapping->tree_lock nor
|
|
|
|
* lock on the page itself: the page may be truncated or
|
|
|
|
* invalidated (changing page->mapping to NULL), or even
|
|
|
|
* swizzled back from swapper_space to tmpfs file
|
|
|
|
* mapping
|
|
|
|
*/
|
2013-02-11 23:33:00 +07:00
|
|
|
if (!trylock_page(page)) {
|
|
|
|
flush_fn(data);
|
|
|
|
lock_page(page);
|
2011-11-01 21:08:06 +07:00
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
if (unlikely(page->mapping != mapping)) {
|
|
|
|
unlock_page(page);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!wbc->range_cyclic && page->index > end) {
|
|
|
|
done = 1;
|
|
|
|
unlock_page(page);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2008-11-20 00:44:22 +07:00
|
|
|
if (wbc->sync_mode != WB_SYNC_NONE) {
|
2008-11-20 22:46:35 +07:00
|
|
|
if (PageWriteback(page))
|
|
|
|
flush_fn(data);
|
2008-01-25 04:13:08 +07:00
|
|
|
wait_on_page_writeback(page);
|
2008-11-20 00:44:22 +07:00
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
if (PageWriteback(page) ||
|
|
|
|
!clear_page_dirty_for_io(page)) {
|
|
|
|
unlock_page(page);
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = (*writepage)(page, wbc, data);
|
|
|
|
|
|
|
|
if (unlikely(ret == AOP_WRITEPAGE_ACTIVATE)) {
|
|
|
|
unlock_page(page);
|
|
|
|
ret = 0;
|
|
|
|
}
|
2014-05-09 23:17:40 +07:00
|
|
|
if (!err && ret < 0)
|
|
|
|
err = ret;
|
2009-09-19 03:03:16 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* the filesystem may choose to bump up nr_to_write.
|
|
|
|
* We have to make sure to honor the new nr_to_write
|
|
|
|
* at any time
|
|
|
|
*/
|
|
|
|
nr_to_write_done = wbc->nr_to_write <= 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
pagevec_release(&pvec);
|
|
|
|
cond_resched();
|
|
|
|
}
|
2014-05-09 23:17:40 +07:00
|
|
|
if (!scanned && !done && !err) {
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* We hit the last page and there is more work to be done: wrap
|
|
|
|
* back to the start of the file
|
|
|
|
*/
|
|
|
|
scanned = 1;
|
|
|
|
index = 0;
|
|
|
|
goto retry;
|
|
|
|
}
|
2012-06-28 04:18:41 +07:00
|
|
|
btrfs_add_delayed_iput(inode);
|
2014-05-09 23:17:40 +07:00
|
|
|
return err;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2009-04-21 02:50:09 +07:00
|
|
|
static void flush_epd_write_bio(struct extent_page_data *epd)
|
2008-11-20 00:44:22 +07:00
|
|
|
{
|
|
|
|
if (epd->bio) {
|
2011-10-04 10:23:14 +07:00
|
|
|
int rw = WRITE;
|
|
|
|
int ret;
|
|
|
|
|
2009-04-21 02:50:09 +07:00
|
|
|
if (epd->sync_io)
|
2011-10-04 10:23:14 +07:00
|
|
|
rw = WRITE_SYNC;
|
|
|
|
|
2012-09-26 01:25:58 +07:00
|
|
|
ret = submit_one_bio(rw, epd->bio, 0, epd->bio_flags);
|
2012-03-12 22:03:00 +07:00
|
|
|
BUG_ON(ret < 0); /* -ENOMEM */
|
2008-11-20 00:44:22 +07:00
|
|
|
epd->bio = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-04-21 02:50:09 +07:00
|
|
|
static noinline void flush_write_bio(void *data)
|
|
|
|
{
|
|
|
|
struct extent_page_data *epd = data;
|
|
|
|
flush_epd_write_bio(epd);
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
|
|
|
|
get_extent_t *get_extent,
|
|
|
|
struct writeback_control *wbc)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct extent_page_data epd = {
|
|
|
|
.bio = NULL,
|
|
|
|
.tree = tree,
|
|
|
|
.get_extent = get_extent,
|
2008-11-07 10:02:51 +07:00
|
|
|
.extent_locked = 0,
|
2009-04-21 02:50:09 +07:00
|
|
|
.sync_io = wbc->sync_mode == WB_SYNC_ALL,
|
2012-09-26 01:25:58 +07:00
|
|
|
.bio_flags = 0,
|
2008-01-25 04:13:08 +07:00
|
|
|
};
|
|
|
|
|
|
|
|
ret = __extent_writepage(page, wbc, &epd);
|
|
|
|
|
2009-04-21 02:50:09 +07:00
|
|
|
flush_epd_write_bio(&epd);
|
2008-01-25 04:13:08 +07:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-11-07 10:02:51 +07:00
|
|
|
int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode,
|
|
|
|
u64 start, u64 end, get_extent_t *get_extent,
|
|
|
|
int mode)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
struct address_space *mapping = inode->i_mapping;
|
|
|
|
struct page *page;
|
|
|
|
unsigned long nr_pages = (end - start + PAGE_CACHE_SIZE) >>
|
|
|
|
PAGE_CACHE_SHIFT;
|
|
|
|
|
|
|
|
struct extent_page_data epd = {
|
|
|
|
.bio = NULL,
|
|
|
|
.tree = tree,
|
|
|
|
.get_extent = get_extent,
|
|
|
|
.extent_locked = 1,
|
2009-04-21 02:50:09 +07:00
|
|
|
.sync_io = mode == WB_SYNC_ALL,
|
2012-09-26 01:25:58 +07:00
|
|
|
.bio_flags = 0,
|
2008-11-07 10:02:51 +07:00
|
|
|
};
|
|
|
|
struct writeback_control wbc_writepages = {
|
|
|
|
.sync_mode = mode,
|
|
|
|
.nr_to_write = nr_pages * 2,
|
|
|
|
.range_start = start,
|
|
|
|
.range_end = end + 1,
|
|
|
|
};
|
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (start <= end) {
|
2008-11-07 10:02:51 +07:00
|
|
|
page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT);
|
|
|
|
if (clear_page_dirty_for_io(page))
|
|
|
|
ret = __extent_writepage(page, &wbc_writepages, &epd);
|
|
|
|
else {
|
|
|
|
if (tree->ops && tree->ops->writepage_end_io_hook)
|
|
|
|
tree->ops->writepage_end_io_hook(page, start,
|
|
|
|
start + PAGE_CACHE_SIZE - 1,
|
|
|
|
NULL, 1);
|
|
|
|
unlock_page(page);
|
|
|
|
}
|
|
|
|
page_cache_release(page);
|
|
|
|
start += PAGE_CACHE_SIZE;
|
|
|
|
}
|
|
|
|
|
2009-04-21 02:50:09 +07:00
|
|
|
flush_epd_write_bio(&epd);
|
2008-11-07 10:02:51 +07:00
|
|
|
return ret;
|
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
int extent_writepages(struct extent_io_tree *tree,
|
|
|
|
struct address_space *mapping,
|
|
|
|
get_extent_t *get_extent,
|
|
|
|
struct writeback_control *wbc)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
struct extent_page_data epd = {
|
|
|
|
.bio = NULL,
|
|
|
|
.tree = tree,
|
|
|
|
.get_extent = get_extent,
|
2008-11-07 10:02:51 +07:00
|
|
|
.extent_locked = 0,
|
2009-04-21 02:50:09 +07:00
|
|
|
.sync_io = wbc->sync_mode == WB_SYNC_ALL,
|
2012-09-26 01:25:58 +07:00
|
|
|
.bio_flags = 0,
|
2008-01-25 04:13:08 +07:00
|
|
|
};
|
|
|
|
|
2008-09-08 22:18:08 +07:00
|
|
|
ret = extent_write_cache_pages(tree, mapping, wbc,
|
2008-11-20 00:44:22 +07:00
|
|
|
__extent_writepage, &epd,
|
|
|
|
flush_write_bio);
|
2009-04-21 02:50:09 +07:00
|
|
|
flush_epd_write_bio(&epd);
|
2008-01-25 04:13:08 +07:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int extent_readpages(struct extent_io_tree *tree,
|
|
|
|
struct address_space *mapping,
|
|
|
|
struct list_head *pages, unsigned nr_pages,
|
|
|
|
get_extent_t get_extent)
|
|
|
|
{
|
|
|
|
struct bio *bio = NULL;
|
|
|
|
unsigned page_idx;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
unsigned long bio_flags = 0;
|
Btrfs: improve multi-thread buffer read
While testing with my buffer read fio jobs[1], I find that btrfs does not
perform well enough.
Here is a scenario in fio jobs:
We have 4 threads, "t1 t2 t3 t4", starting to buffer read a same file,
and all of them will race on add_to_page_cache_lru(), and if one thread
successfully puts its page into the page cache, it takes the responsibility
to read the page's data.
And what's more, reading a page needs a period of time to finish, in which
other threads can slide in and process rest pages:
t1 t2 t3 t4
add Page1
read Page1 add Page2
| read Page2 add Page3
| | read Page3 add Page4
| | | read Page4
-----|------------|-----------|-----------|--------
v v v v
bio bio bio bio
Now we have four bios, each of which holds only one page since we need to
maintain consecutive pages in bio. Thus, we can end up with far more bios
than we need.
Here we're going to
a) delay the real read-page section and
b) try to put more pages into page cache.
With that said, we can make each bio hold more pages and reduce the number
of bios we need.
Here is some numbers taken from fio results:
w/o patch w patch
------------- -------- ---------------
READ: 745MB/s +25% 934MB/s
[1]:
[global]
group_reporting
thread
numjobs=4
bs=32k
rw=read
ioengine=sync
directory=/mnt/btrfs/
[READ]
filename=foobar
size=2000M
invalidate=1
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
2012-07-21 10:43:09 +07:00
|
|
|
struct page *pagepool[16];
|
|
|
|
struct page *page;
|
2013-07-25 18:22:37 +07:00
|
|
|
struct extent_map *em_cached = NULL;
|
Btrfs: improve multi-thread buffer read
While testing with my buffer read fio jobs[1], I find that btrfs does not
perform well enough.
Here is a scenario in fio jobs:
We have 4 threads, "t1 t2 t3 t4", starting to buffer read a same file,
and all of them will race on add_to_page_cache_lru(), and if one thread
successfully puts its page into the page cache, it takes the responsibility
to read the page's data.
And what's more, reading a page needs a period of time to finish, in which
other threads can slide in and process rest pages:
t1 t2 t3 t4
add Page1
read Page1 add Page2
| read Page2 add Page3
| | read Page3 add Page4
| | | read Page4
-----|------------|-----------|-----------|--------
v v v v
bio bio bio bio
Now we have four bios, each of which holds only one page since we need to
maintain consecutive pages in bio. Thus, we can end up with far more bios
than we need.
Here we're going to
a) delay the real read-page section and
b) try to put more pages into page cache.
With that said, we can make each bio hold more pages and reduce the number
of bios we need.
Here is some numbers taken from fio results:
w/o patch w patch
------------- -------- ---------------
READ: 745MB/s +25% 934MB/s
[1]:
[global]
group_reporting
thread
numjobs=4
bs=32k
rw=read
ioengine=sync
directory=/mnt/btrfs/
[READ]
filename=foobar
size=2000M
invalidate=1
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
2012-07-21 10:43:09 +07:00
|
|
|
int nr = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
for (page_idx = 0; page_idx < nr_pages; page_idx++) {
|
Btrfs: improve multi-thread buffer read
While testing with my buffer read fio jobs[1], I find that btrfs does not
perform well enough.
Here is a scenario in fio jobs:
We have 4 threads, "t1 t2 t3 t4", starting to buffer read a same file,
and all of them will race on add_to_page_cache_lru(), and if one thread
successfully puts its page into the page cache, it takes the responsibility
to read the page's data.
And what's more, reading a page needs a period of time to finish, in which
other threads can slide in and process rest pages:
t1 t2 t3 t4
add Page1
read Page1 add Page2
| read Page2 add Page3
| | read Page3 add Page4
| | | read Page4
-----|------------|-----------|-----------|--------
v v v v
bio bio bio bio
Now we have four bios, each of which holds only one page since we need to
maintain consecutive pages in bio. Thus, we can end up with far more bios
than we need.
Here we're going to
a) delay the real read-page section and
b) try to put more pages into page cache.
With that said, we can make each bio hold more pages and reduce the number
of bios we need.
Here is some numbers taken from fio results:
w/o patch w patch
------------- -------- ---------------
READ: 745MB/s +25% 934MB/s
[1]:
[global]
group_reporting
thread
numjobs=4
bs=32k
rw=read
ioengine=sync
directory=/mnt/btrfs/
[READ]
filename=foobar
size=2000M
invalidate=1
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
2012-07-21 10:43:09 +07:00
|
|
|
page = list_entry(pages->prev, struct page, lru);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
prefetchw(&page->flags);
|
|
|
|
list_del(&page->lru);
|
Btrfs: improve multi-thread buffer read
While testing with my buffer read fio jobs[1], I find that btrfs does not
perform well enough.
Here is a scenario in fio jobs:
We have 4 threads, "t1 t2 t3 t4", starting to buffer read a same file,
and all of them will race on add_to_page_cache_lru(), and if one thread
successfully puts its page into the page cache, it takes the responsibility
to read the page's data.
And what's more, reading a page needs a period of time to finish, in which
other threads can slide in and process rest pages:
t1 t2 t3 t4
add Page1
read Page1 add Page2
| read Page2 add Page3
| | read Page3 add Page4
| | | read Page4
-----|------------|-----------|-----------|--------
v v v v
bio bio bio bio
Now we have four bios, each of which holds only one page since we need to
maintain consecutive pages in bio. Thus, we can end up with far more bios
than we need.
Here we're going to
a) delay the real read-page section and
b) try to put more pages into page cache.
With that said, we can make each bio hold more pages and reduce the number
of bios we need.
Here is some numbers taken from fio results:
w/o patch w patch
------------- -------- ---------------
READ: 745MB/s +25% 934MB/s
[1]:
[global]
group_reporting
thread
numjobs=4
bs=32k
rw=read
ioengine=sync
directory=/mnt/btrfs/
[READ]
filename=foobar
size=2000M
invalidate=1
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
2012-07-21 10:43:09 +07:00
|
|
|
if (add_to_page_cache_lru(page, mapping,
|
2011-04-26 06:43:51 +07:00
|
|
|
page->index, GFP_NOFS)) {
|
Btrfs: improve multi-thread buffer read
While testing with my buffer read fio jobs[1], I find that btrfs does not
perform well enough.
Here is a scenario in fio jobs:
We have 4 threads, "t1 t2 t3 t4", starting to buffer read a same file,
and all of them will race on add_to_page_cache_lru(), and if one thread
successfully puts its page into the page cache, it takes the responsibility
to read the page's data.
And what's more, reading a page needs a period of time to finish, in which
other threads can slide in and process rest pages:
t1 t2 t3 t4
add Page1
read Page1 add Page2
| read Page2 add Page3
| | read Page3 add Page4
| | | read Page4
-----|------------|-----------|-----------|--------
v v v v
bio bio bio bio
Now we have four bios, each of which holds only one page since we need to
maintain consecutive pages in bio. Thus, we can end up with far more bios
than we need.
Here we're going to
a) delay the real read-page section and
b) try to put more pages into page cache.
With that said, we can make each bio hold more pages and reduce the number
of bios we need.
Here is some numbers taken from fio results:
w/o patch w patch
------------- -------- ---------------
READ: 745MB/s +25% 934MB/s
[1]:
[global]
group_reporting
thread
numjobs=4
bs=32k
rw=read
ioengine=sync
directory=/mnt/btrfs/
[READ]
filename=foobar
size=2000M
invalidate=1
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
2012-07-21 10:43:09 +07:00
|
|
|
page_cache_release(page);
|
|
|
|
continue;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
Btrfs: improve multi-thread buffer read
While testing with my buffer read fio jobs[1], I find that btrfs does not
perform well enough.
Here is a scenario in fio jobs:
We have 4 threads, "t1 t2 t3 t4", starting to buffer read a same file,
and all of them will race on add_to_page_cache_lru(), and if one thread
successfully puts its page into the page cache, it takes the responsibility
to read the page's data.
And what's more, reading a page needs a period of time to finish, in which
other threads can slide in and process rest pages:
t1 t2 t3 t4
add Page1
read Page1 add Page2
| read Page2 add Page3
| | read Page3 add Page4
| | | read Page4
-----|------------|-----------|-----------|--------
v v v v
bio bio bio bio
Now we have four bios, each of which holds only one page since we need to
maintain consecutive pages in bio. Thus, we can end up with far more bios
than we need.
Here we're going to
a) delay the real read-page section and
b) try to put more pages into page cache.
With that said, we can make each bio hold more pages and reduce the number
of bios we need.
Here is some numbers taken from fio results:
w/o patch w patch
------------- -------- ---------------
READ: 745MB/s +25% 934MB/s
[1]:
[global]
group_reporting
thread
numjobs=4
bs=32k
rw=read
ioengine=sync
directory=/mnt/btrfs/
[READ]
filename=foobar
size=2000M
invalidate=1
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
2012-07-21 10:43:09 +07:00
|
|
|
|
|
|
|
pagepool[nr++] = page;
|
|
|
|
if (nr < ARRAY_SIZE(pagepool))
|
|
|
|
continue;
|
2013-07-25 18:22:37 +07:00
|
|
|
__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
|
2013-07-25 18:22:36 +07:00
|
|
|
&bio, 0, &bio_flags, READ);
|
Btrfs: improve multi-thread buffer read
While testing with my buffer read fio jobs[1], I find that btrfs does not
perform well enough.
Here is a scenario in fio jobs:
We have 4 threads, "t1 t2 t3 t4", starting to buffer read a same file,
and all of them will race on add_to_page_cache_lru(), and if one thread
successfully puts its page into the page cache, it takes the responsibility
to read the page's data.
And what's more, reading a page needs a period of time to finish, in which
other threads can slide in and process rest pages:
t1 t2 t3 t4
add Page1
read Page1 add Page2
| read Page2 add Page3
| | read Page3 add Page4
| | | read Page4
-----|------------|-----------|-----------|--------
v v v v
bio bio bio bio
Now we have four bios, each of which holds only one page since we need to
maintain consecutive pages in bio. Thus, we can end up with far more bios
than we need.
Here we're going to
a) delay the real read-page section and
b) try to put more pages into page cache.
With that said, we can make each bio hold more pages and reduce the number
of bios we need.
Here is some numbers taken from fio results:
w/o patch w patch
------------- -------- ---------------
READ: 745MB/s +25% 934MB/s
[1]:
[global]
group_reporting
thread
numjobs=4
bs=32k
rw=read
ioengine=sync
directory=/mnt/btrfs/
[READ]
filename=foobar
size=2000M
invalidate=1
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
2012-07-21 10:43:09 +07:00
|
|
|
nr = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
2013-07-25 18:22:36 +07:00
|
|
|
if (nr)
|
2013-07-25 18:22:37 +07:00
|
|
|
__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
|
2013-07-25 18:22:36 +07:00
|
|
|
&bio, 0, &bio_flags, READ);
|
Btrfs: improve multi-thread buffer read
While testing with my buffer read fio jobs[1], I find that btrfs does not
perform well enough.
Here is a scenario in fio jobs:
We have 4 threads, "t1 t2 t3 t4", starting to buffer read a same file,
and all of them will race on add_to_page_cache_lru(), and if one thread
successfully puts its page into the page cache, it takes the responsibility
to read the page's data.
And what's more, reading a page needs a period of time to finish, in which
other threads can slide in and process rest pages:
t1 t2 t3 t4
add Page1
read Page1 add Page2
| read Page2 add Page3
| | read Page3 add Page4
| | | read Page4
-----|------------|-----------|-----------|--------
v v v v
bio bio bio bio
Now we have four bios, each of which holds only one page since we need to
maintain consecutive pages in bio. Thus, we can end up with far more bios
than we need.
Here we're going to
a) delay the real read-page section and
b) try to put more pages into page cache.
With that said, we can make each bio hold more pages and reduce the number
of bios we need.
Here is some numbers taken from fio results:
w/o patch w patch
------------- -------- ---------------
READ: 745MB/s +25% 934MB/s
[1]:
[global]
group_reporting
thread
numjobs=4
bs=32k
rw=read
ioengine=sync
directory=/mnt/btrfs/
[READ]
filename=foobar
size=2000M
invalidate=1
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
2012-07-21 10:43:09 +07:00
|
|
|
|
2013-07-25 18:22:37 +07:00
|
|
|
if (em_cached)
|
|
|
|
free_extent_map(em_cached);
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
BUG_ON(!list_empty(pages));
|
|
|
|
if (bio)
|
2012-03-12 22:03:00 +07:00
|
|
|
return submit_one_bio(READ, bio, 0, bio_flags);
|
2008-01-25 04:13:08 +07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* basic invalidatepage code, this waits on any locked or writeback
|
|
|
|
* ranges corresponding to the page, and then deletes any extent state
|
|
|
|
* records from the tree
|
|
|
|
*/
|
|
|
|
int extent_invalidatepage(struct extent_io_tree *tree,
|
|
|
|
struct page *page, unsigned long offset)
|
|
|
|
{
|
2010-02-04 02:33:23 +07:00
|
|
|
struct extent_state *cached_state = NULL;
|
2012-12-21 16:17:45 +07:00
|
|
|
u64 start = page_offset(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
u64 end = start + PAGE_CACHE_SIZE - 1;
|
|
|
|
size_t blocksize = page->mapping->host->i_sb->s_blocksize;
|
|
|
|
|
2013-02-26 15:10:22 +07:00
|
|
|
start += ALIGN(offset, blocksize);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (start > end)
|
|
|
|
return 0;
|
|
|
|
|
2012-03-01 20:57:19 +07:00
|
|
|
lock_extent_bits(tree, start, end, 0, &cached_state);
|
2009-09-03 00:24:36 +07:00
|
|
|
wait_on_page_writeback(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
clear_extent_bit(tree, start, end,
|
2009-10-09 00:34:05 +07:00
|
|
|
EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
|
|
|
|
EXTENT_DO_ACCOUNTING,
|
2010-02-04 02:33:23 +07:00
|
|
|
1, 1, &cached_state, GFP_NOFS);
|
2008-01-25 04:13:08 +07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-04-18 21:29:50 +07:00
|
|
|
/*
|
|
|
|
* a helper for releasepage, this tests for areas of the page that
|
|
|
|
* are locked or under IO and drops the related state bits if it is safe
|
|
|
|
* to drop the page.
|
|
|
|
*/
|
2013-04-26 03:41:01 +07:00
|
|
|
static int try_release_extent_state(struct extent_map_tree *map,
|
|
|
|
struct extent_io_tree *tree,
|
|
|
|
struct page *page, gfp_t mask)
|
2008-04-18 21:29:50 +07:00
|
|
|
{
|
2012-12-21 16:17:45 +07:00
|
|
|
u64 start = page_offset(page);
|
2008-04-18 21:29:50 +07:00
|
|
|
u64 end = start + PAGE_CACHE_SIZE - 1;
|
|
|
|
int ret = 1;
|
|
|
|
|
2008-07-18 22:56:15 +07:00
|
|
|
if (test_range_bit(tree, start, end,
|
2009-09-03 03:53:46 +07:00
|
|
|
EXTENT_IOBITS, 0, NULL))
|
2008-04-18 21:29:50 +07:00
|
|
|
ret = 0;
|
|
|
|
else {
|
|
|
|
if ((mask & GFP_NOFS) == GFP_NOFS)
|
|
|
|
mask = GFP_NOFS;
|
2009-09-24 07:28:46 +07:00
|
|
|
/*
|
|
|
|
* at this point we can safely clear everything except the
|
|
|
|
* locked bit and the nodatasum bit
|
|
|
|
*/
|
2011-02-15 00:52:08 +07:00
|
|
|
ret = clear_extent_bit(tree, start, end,
|
2009-09-24 07:28:46 +07:00
|
|
|
~(EXTENT_LOCKED | EXTENT_NODATASUM),
|
|
|
|
0, 0, NULL, mask);
|
2011-02-15 00:52:08 +07:00
|
|
|
|
|
|
|
/* if clear_extent_bit failed for enomem reasons,
|
|
|
|
* we can't allow the release to continue.
|
|
|
|
*/
|
|
|
|
if (ret < 0)
|
|
|
|
ret = 0;
|
|
|
|
else
|
|
|
|
ret = 1;
|
2008-04-18 21:29:50 +07:00
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
/*
|
|
|
|
* a helper for releasepage. As long as there are no locked extents
|
|
|
|
* in the range corresponding to the page, both state records and extent
|
|
|
|
* map records are removed
|
|
|
|
*/
|
|
|
|
int try_release_extent_mapping(struct extent_map_tree *map,
|
2008-01-29 21:59:12 +07:00
|
|
|
struct extent_io_tree *tree, struct page *page,
|
|
|
|
gfp_t mask)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct extent_map *em;
|
2012-12-21 16:17:45 +07:00
|
|
|
u64 start = page_offset(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
u64 end = start + PAGE_CACHE_SIZE - 1;
|
2008-04-18 21:29:50 +07:00
|
|
|
|
2008-01-29 21:59:12 +07:00
|
|
|
if ((mask & __GFP_WAIT) &&
|
|
|
|
page->mapping->host->i_size > 16 * 1024 * 1024) {
|
2008-02-15 22:40:50 +07:00
|
|
|
u64 len;
|
2008-01-29 21:59:12 +07:00
|
|
|
while (start <= end) {
|
2008-02-15 22:40:50 +07:00
|
|
|
len = end - start + 1;
|
2009-09-03 03:24:52 +07:00
|
|
|
write_lock(&map->lock);
|
2008-02-15 22:40:50 +07:00
|
|
|
em = lookup_extent_mapping(map, start, len);
|
2012-02-16 14:23:58 +07:00
|
|
|
if (!em) {
|
2009-09-03 03:24:52 +07:00
|
|
|
write_unlock(&map->lock);
|
2008-01-29 21:59:12 +07:00
|
|
|
break;
|
|
|
|
}
|
2008-07-18 23:01:11 +07:00
|
|
|
if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
|
|
|
|
em->start != start) {
|
2009-09-03 03:24:52 +07:00
|
|
|
write_unlock(&map->lock);
|
2008-01-29 21:59:12 +07:00
|
|
|
free_extent_map(em);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (!test_range_bit(tree, em->start,
|
|
|
|
extent_map_end(em) - 1,
|
2009-09-03 03:53:46 +07:00
|
|
|
EXTENT_LOCKED | EXTENT_WRITEBACK,
|
2009-09-03 02:22:30 +07:00
|
|
|
0, NULL)) {
|
2008-01-29 21:59:12 +07:00
|
|
|
remove_extent_mapping(map, em);
|
|
|
|
/* once for the rb tree */
|
|
|
|
free_extent_map(em);
|
|
|
|
}
|
|
|
|
start = extent_map_end(em);
|
2009-09-03 03:24:52 +07:00
|
|
|
write_unlock(&map->lock);
|
2008-01-29 21:59:12 +07:00
|
|
|
|
|
|
|
/* once for us */
|
2008-01-25 04:13:08 +07:00
|
|
|
free_extent_map(em);
|
|
|
|
}
|
|
|
|
}
|
2008-04-18 21:29:50 +07:00
|
|
|
return try_release_extent_state(map, tree, page, mask);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2011-02-24 04:23:20 +07:00
|
|
|
/*
|
|
|
|
* helper function for fiemap, which doesn't want to see any holes.
|
|
|
|
* This maps until we find something past 'last'
|
|
|
|
*/
|
|
|
|
static struct extent_map *get_extent_skip_holes(struct inode *inode,
|
|
|
|
u64 offset,
|
|
|
|
u64 last,
|
|
|
|
get_extent_t *get_extent)
|
|
|
|
{
|
|
|
|
u64 sectorsize = BTRFS_I(inode)->root->sectorsize;
|
|
|
|
struct extent_map *em;
|
|
|
|
u64 len;
|
|
|
|
|
|
|
|
if (offset >= last)
|
|
|
|
return NULL;
|
|
|
|
|
2013-10-31 12:03:04 +07:00
|
|
|
while (1) {
|
2011-02-24 04:23:20 +07:00
|
|
|
len = last - offset;
|
|
|
|
if (len == 0)
|
|
|
|
break;
|
2013-02-26 15:10:22 +07:00
|
|
|
len = ALIGN(len, sectorsize);
|
2011-02-24 04:23:20 +07:00
|
|
|
em = get_extent(inode, NULL, 0, offset, len, 0);
|
2011-04-19 23:00:01 +07:00
|
|
|
if (IS_ERR_OR_NULL(em))
|
2011-02-24 04:23:20 +07:00
|
|
|
return em;
|
|
|
|
|
|
|
|
/* if this isn't a hole return it */
|
|
|
|
if (!test_bit(EXTENT_FLAG_VACANCY, &em->flags) &&
|
|
|
|
em->block_start != EXTENT_MAP_HOLE) {
|
|
|
|
return em;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* this is a hole, advance to the next extent */
|
|
|
|
offset = extent_map_end(em);
|
|
|
|
free_extent_map(em);
|
|
|
|
if (offset >= last)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2009-01-22 02:39:14 +07:00
|
|
|
int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
|
|
|
|
__u64 start, __u64 len, get_extent_t *get_extent)
|
|
|
|
{
|
2010-11-24 02:36:57 +07:00
|
|
|
int ret = 0;
|
2009-01-22 02:39:14 +07:00
|
|
|
u64 off = start;
|
|
|
|
u64 max = start + len;
|
|
|
|
u32 flags = 0;
|
2010-11-24 02:36:57 +07:00
|
|
|
u32 found_type;
|
|
|
|
u64 last;
|
2011-02-24 04:23:20 +07:00
|
|
|
u64 last_for_get_extent = 0;
|
2009-01-22 02:39:14 +07:00
|
|
|
u64 disko = 0;
|
2011-02-24 04:23:20 +07:00
|
|
|
u64 isize = i_size_read(inode);
|
2010-11-24 02:36:57 +07:00
|
|
|
struct btrfs_key found_key;
|
2009-01-22 02:39:14 +07:00
|
|
|
struct extent_map *em = NULL;
|
2010-02-04 02:33:23 +07:00
|
|
|
struct extent_state *cached_state = NULL;
|
2010-11-24 02:36:57 +07:00
|
|
|
struct btrfs_path *path;
|
2014-09-11 03:20:45 +07:00
|
|
|
struct btrfs_root *root = BTRFS_I(inode)->root;
|
2009-01-22 02:39:14 +07:00
|
|
|
int end = 0;
|
2011-02-24 04:23:20 +07:00
|
|
|
u64 em_start = 0;
|
|
|
|
u64 em_len = 0;
|
|
|
|
u64 em_end = 0;
|
2009-01-22 02:39:14 +07:00
|
|
|
|
|
|
|
if (len == 0)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2010-11-24 02:36:57 +07:00
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
path->leave_spinning = 1;
|
|
|
|
|
2014-07-18 08:55:43 +07:00
|
|
|
start = round_down(start, BTRFS_I(inode)->root->sectorsize);
|
|
|
|
len = round_up(max, BTRFS_I(inode)->root->sectorsize) - start;
|
2011-11-17 23:34:31 +07:00
|
|
|
|
2011-02-24 04:23:20 +07:00
|
|
|
/*
|
|
|
|
* lookup the last file extent. We're not using i_size here
|
|
|
|
* because there might be preallocation past i_size
|
|
|
|
*/
|
2014-09-11 03:20:45 +07:00
|
|
|
ret = btrfs_lookup_file_extent(NULL, root, path, btrfs_ino(inode), -1,
|
|
|
|
0);
|
2010-11-24 02:36:57 +07:00
|
|
|
if (ret < 0) {
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
WARN_ON(!ret);
|
|
|
|
path->slots[0]--;
|
|
|
|
btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
|
2014-06-04 23:41:45 +07:00
|
|
|
found_type = found_key.type;
|
2010-11-24 02:36:57 +07:00
|
|
|
|
2011-02-24 04:23:20 +07:00
|
|
|
/* No extents, but there might be delalloc bits */
|
2011-04-20 09:31:50 +07:00
|
|
|
if (found_key.objectid != btrfs_ino(inode) ||
|
2010-11-24 02:36:57 +07:00
|
|
|
found_type != BTRFS_EXTENT_DATA_KEY) {
|
2011-02-24 04:23:20 +07:00
|
|
|
/* have to trust i_size as the end */
|
|
|
|
last = (u64)-1;
|
|
|
|
last_for_get_extent = isize;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* remember the start of the last extent. There are a
|
|
|
|
* bunch of different factors that go into the length of the
|
|
|
|
* extent, so its much less complex to remember where it started
|
|
|
|
*/
|
|
|
|
last = found_key.offset;
|
|
|
|
last_for_get_extent = last + 1;
|
2010-11-24 02:36:57 +07:00
|
|
|
}
|
2013-09-22 11:54:23 +07:00
|
|
|
btrfs_release_path(path);
|
2010-11-24 02:36:57 +07:00
|
|
|
|
2011-02-24 04:23:20 +07:00
|
|
|
/*
|
|
|
|
* we might have some extents allocated but more delalloc past those
|
|
|
|
* extents. so, we trust isize unless the start of the last extent is
|
|
|
|
* beyond isize
|
|
|
|
*/
|
|
|
|
if (last < isize) {
|
|
|
|
last = (u64)-1;
|
|
|
|
last_for_get_extent = isize;
|
|
|
|
}
|
|
|
|
|
2013-05-01 23:23:41 +07:00
|
|
|
lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1, 0,
|
2012-03-01 20:57:19 +07:00
|
|
|
&cached_state);
|
2011-02-24 04:23:20 +07:00
|
|
|
|
2011-11-17 23:34:31 +07:00
|
|
|
em = get_extent_skip_holes(inode, start, last_for_get_extent,
|
2011-02-24 04:23:20 +07:00
|
|
|
get_extent);
|
2009-01-22 02:39:14 +07:00
|
|
|
if (!em)
|
|
|
|
goto out;
|
|
|
|
if (IS_ERR(em)) {
|
|
|
|
ret = PTR_ERR(em);
|
|
|
|
goto out;
|
|
|
|
}
|
2010-11-24 02:36:57 +07:00
|
|
|
|
2009-01-22 02:39:14 +07:00
|
|
|
while (!end) {
|
2013-07-06 00:52:51 +07:00
|
|
|
u64 offset_in_extent = 0;
|
2011-03-08 23:54:40 +07:00
|
|
|
|
|
|
|
/* break if the extent we found is outside the range */
|
|
|
|
if (em->start >= max || extent_map_end(em) < off)
|
|
|
|
break;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* get_extent may return an extent that starts before our
|
|
|
|
* requested range. We have to make sure the ranges
|
|
|
|
* we return to fiemap always move forward and don't
|
|
|
|
* overlap, so adjust the offsets here
|
|
|
|
*/
|
|
|
|
em_start = max(em->start, off);
|
2009-01-22 02:39:14 +07:00
|
|
|
|
2011-03-08 23:54:40 +07:00
|
|
|
/*
|
|
|
|
* record the offset from the start of the extent
|
2013-07-06 00:52:51 +07:00
|
|
|
* for adjusting the disk offset below. Only do this if the
|
|
|
|
* extent isn't compressed since our in ram offset may be past
|
|
|
|
* what we have actually allocated on disk.
|
2011-03-08 23:54:40 +07:00
|
|
|
*/
|
2013-07-06 00:52:51 +07:00
|
|
|
if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
|
|
|
|
offset_in_extent = em_start - em->start;
|
2011-02-24 04:23:20 +07:00
|
|
|
em_end = extent_map_end(em);
|
2011-03-08 23:54:40 +07:00
|
|
|
em_len = em_end - em_start;
|
2009-01-22 02:39:14 +07:00
|
|
|
disko = 0;
|
|
|
|
flags = 0;
|
|
|
|
|
2011-03-08 23:54:40 +07:00
|
|
|
/*
|
|
|
|
* bump off for our next call to get_extent
|
|
|
|
*/
|
|
|
|
off = extent_map_end(em);
|
|
|
|
if (off >= max)
|
|
|
|
end = 1;
|
|
|
|
|
2009-04-03 21:33:45 +07:00
|
|
|
if (em->block_start == EXTENT_MAP_LAST_BYTE) {
|
2009-01-22 02:39:14 +07:00
|
|
|
end = 1;
|
|
|
|
flags |= FIEMAP_EXTENT_LAST;
|
2009-04-03 21:33:45 +07:00
|
|
|
} else if (em->block_start == EXTENT_MAP_INLINE) {
|
2009-01-22 02:39:14 +07:00
|
|
|
flags |= (FIEMAP_EXTENT_DATA_INLINE |
|
|
|
|
FIEMAP_EXTENT_NOT_ALIGNED);
|
2009-04-03 21:33:45 +07:00
|
|
|
} else if (em->block_start == EXTENT_MAP_DELALLOC) {
|
2009-01-22 02:39:14 +07:00
|
|
|
flags |= (FIEMAP_EXTENT_DELALLOC |
|
|
|
|
FIEMAP_EXTENT_UNKNOWN);
|
2014-09-11 03:20:45 +07:00
|
|
|
} else if (fieinfo->fi_extents_max) {
|
|
|
|
u64 bytenr = em->block_start -
|
|
|
|
(em->start - em->orig_start);
|
2013-09-22 11:54:23 +07:00
|
|
|
|
2011-03-08 23:54:40 +07:00
|
|
|
disko = em->block_start + offset_in_extent;
|
2013-09-22 11:54:23 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* As btrfs supports shared space, this information
|
|
|
|
* can be exported to userspace tools via
|
2014-09-11 03:20:45 +07:00
|
|
|
* flag FIEMAP_EXTENT_SHARED. If fi_extents_max == 0
|
|
|
|
* then we're just getting a count and we can skip the
|
|
|
|
* lookup stuff.
|
2013-09-22 11:54:23 +07:00
|
|
|
*/
|
2014-09-11 03:20:45 +07:00
|
|
|
ret = btrfs_check_shared(NULL, root->fs_info,
|
|
|
|
root->objectid,
|
|
|
|
btrfs_ino(inode), bytenr);
|
|
|
|
if (ret < 0)
|
2013-09-22 11:54:23 +07:00
|
|
|
goto out_free;
|
2014-09-11 03:20:45 +07:00
|
|
|
if (ret)
|
2013-09-22 11:54:23 +07:00
|
|
|
flags |= FIEMAP_EXTENT_SHARED;
|
2014-09-11 03:20:45 +07:00
|
|
|
ret = 0;
|
2009-01-22 02:39:14 +07:00
|
|
|
}
|
|
|
|
if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
|
|
|
|
flags |= FIEMAP_EXTENT_ENCODED;
|
|
|
|
|
|
|
|
free_extent_map(em);
|
|
|
|
em = NULL;
|
2011-02-24 04:23:20 +07:00
|
|
|
if ((em_start >= last) || em_len == (u64)-1 ||
|
|
|
|
(last == (u64)-1 && isize <= em_end)) {
|
2009-01-22 02:39:14 +07:00
|
|
|
flags |= FIEMAP_EXTENT_LAST;
|
|
|
|
end = 1;
|
|
|
|
}
|
|
|
|
|
2011-02-24 04:23:20 +07:00
|
|
|
/* now scan forward to see if this is really the last extent. */
|
|
|
|
em = get_extent_skip_holes(inode, off, last_for_get_extent,
|
|
|
|
get_extent);
|
|
|
|
if (IS_ERR(em)) {
|
|
|
|
ret = PTR_ERR(em);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
if (!em) {
|
2010-11-24 02:36:57 +07:00
|
|
|
flags |= FIEMAP_EXTENT_LAST;
|
|
|
|
end = 1;
|
|
|
|
}
|
2011-02-24 04:23:20 +07:00
|
|
|
ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
|
|
|
|
em_len, flags);
|
|
|
|
if (ret)
|
|
|
|
goto out_free;
|
2009-01-22 02:39:14 +07:00
|
|
|
}
|
|
|
|
out_free:
|
|
|
|
free_extent_map(em);
|
|
|
|
out:
|
2013-09-22 11:54:23 +07:00
|
|
|
btrfs_free_path(path);
|
2013-05-01 23:23:41 +07:00
|
|
|
unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
|
2010-02-04 02:33:23 +07:00
|
|
|
&cached_state, GFP_NOFS);
|
2009-01-22 02:39:14 +07:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2010-08-07 00:21:20 +07:00
|
|
|
static void __free_extent_buffer(struct extent_buffer *eb)
|
|
|
|
{
|
2013-04-22 23:12:31 +07:00
|
|
|
btrfs_leak_debug_del(&eb->leak_list);
|
2010-08-07 00:21:20 +07:00
|
|
|
kmem_cache_free(extent_buffer_cache, eb);
|
|
|
|
}
|
|
|
|
|
2014-03-29 04:07:27 +07:00
|
|
|
int extent_buffer_under_io(struct extent_buffer *eb)
|
2013-08-08 01:54:37 +07:00
|
|
|
{
|
|
|
|
return (atomic_read(&eb->io_pages) ||
|
|
|
|
test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags) ||
|
|
|
|
test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Helper for releasing extent buffer page.
|
|
|
|
*/
|
|
|
|
static void btrfs_release_extent_buffer_page(struct extent_buffer *eb,
|
|
|
|
unsigned long start_idx)
|
|
|
|
{
|
|
|
|
unsigned long index;
|
|
|
|
unsigned long num_pages;
|
|
|
|
struct page *page;
|
|
|
|
int mapped = !test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);
|
|
|
|
|
|
|
|
BUG_ON(extent_buffer_under_io(eb));
|
|
|
|
|
|
|
|
num_pages = num_extent_pages(eb->start, eb->len);
|
|
|
|
index = start_idx + num_pages;
|
|
|
|
if (start_idx >= index)
|
|
|
|
return;
|
|
|
|
|
|
|
|
do {
|
|
|
|
index--;
|
|
|
|
page = extent_buffer_page(eb, index);
|
|
|
|
if (page && mapped) {
|
|
|
|
spin_lock(&page->mapping->private_lock);
|
|
|
|
/*
|
|
|
|
* We do this since we'll remove the pages after we've
|
|
|
|
* removed the eb from the radix tree, so we could race
|
|
|
|
* and have this page now attached to the new eb. So
|
|
|
|
* only clear page_private if it's still connected to
|
|
|
|
* this eb.
|
|
|
|
*/
|
|
|
|
if (PagePrivate(page) &&
|
|
|
|
page->private == (unsigned long)eb) {
|
|
|
|
BUG_ON(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
|
|
|
|
BUG_ON(PageDirty(page));
|
|
|
|
BUG_ON(PageWriteback(page));
|
|
|
|
/*
|
|
|
|
* We need to make sure we haven't be attached
|
|
|
|
* to a new eb.
|
|
|
|
*/
|
|
|
|
ClearPagePrivate(page);
|
|
|
|
set_page_private(page, 0);
|
|
|
|
/* One for the page private */
|
|
|
|
page_cache_release(page);
|
|
|
|
}
|
|
|
|
spin_unlock(&page->mapping->private_lock);
|
|
|
|
|
|
|
|
}
|
|
|
|
if (page) {
|
|
|
|
/* One for when we alloced the page */
|
|
|
|
page_cache_release(page);
|
|
|
|
}
|
|
|
|
} while (index != start_idx);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Helper for releasing the extent buffer.
|
|
|
|
*/
|
|
|
|
static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
|
|
|
|
{
|
|
|
|
btrfs_release_extent_buffer_page(eb, 0);
|
|
|
|
__free_extent_buffer(eb);
|
|
|
|
}
|
|
|
|
|
2013-12-17 01:24:27 +07:00
|
|
|
static struct extent_buffer *
|
|
|
|
__alloc_extent_buffer(struct btrfs_fs_info *fs_info, u64 start,
|
|
|
|
unsigned long len, gfp_t mask)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
struct extent_buffer *eb = NULL;
|
|
|
|
|
|
|
|
eb = kmem_cache_zalloc(extent_buffer_cache, mask);
|
2011-01-05 09:32:22 +07:00
|
|
|
if (eb == NULL)
|
|
|
|
return NULL;
|
2008-01-25 04:13:08 +07:00
|
|
|
eb->start = start;
|
|
|
|
eb->len = len;
|
2013-12-17 01:24:27 +07:00
|
|
|
eb->fs_info = fs_info;
|
2012-05-16 22:00:02 +07:00
|
|
|
eb->bflags = 0;
|
2011-07-17 02:23:14 +07:00
|
|
|
rwlock_init(&eb->lock);
|
|
|
|
atomic_set(&eb->write_locks, 0);
|
|
|
|
atomic_set(&eb->read_locks, 0);
|
|
|
|
atomic_set(&eb->blocking_readers, 0);
|
|
|
|
atomic_set(&eb->blocking_writers, 0);
|
|
|
|
atomic_set(&eb->spinning_readers, 0);
|
|
|
|
atomic_set(&eb->spinning_writers, 0);
|
2011-09-13 15:55:48 +07:00
|
|
|
eb->lock_nested = 0;
|
2011-07-17 02:23:14 +07:00
|
|
|
init_waitqueue_head(&eb->write_lock_wq);
|
|
|
|
init_waitqueue_head(&eb->read_lock_wq);
|
Btrfs: Change btree locking to use explicit blocking points
Most of the btrfs metadata operations can be protected by a spinlock,
but some operations still need to schedule.
So far, btrfs has been using a mutex along with a trylock loop,
most of the time it is able to avoid going for the full mutex, so
the trylock loop is a big performance gain.
This commit is step one for getting rid of the blocking locks entirely.
btrfs_tree_lock takes a spinlock, and the code explicitly switches
to a blocking lock when it starts an operation that can schedule.
We'll be able get rid of the blocking locks in smaller pieces over time.
Tracing allows us to find the most common cause of blocking, so we
can start with the hot spots first.
The basic idea is:
btrfs_tree_lock() returns with the spin lock held
btrfs_set_lock_blocking() sets the EXTENT_BUFFER_BLOCKING bit in
the extent buffer flags, and then drops the spin lock. The buffer is
still considered locked by all of the btrfs code.
If btrfs_tree_lock gets the spinlock but finds the blocking bit set, it drops
the spin lock and waits on a wait queue for the blocking bit to go away.
Much of the code that needs to set the blocking bit finishes without actually
blocking a good percentage of the time. So, an adaptive spin is still
used against the blocking bit to avoid very high context switch rates.
btrfs_clear_lock_blocking() clears the blocking bit and returns
with the spinlock held again.
btrfs_tree_unlock() can be called on either blocking or spinning locks,
it does the right thing based on the blocking bit.
ctree.c has a helper function to set/clear all the locked buffers in a
path as blocking.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-02-04 21:25:08 +07:00
|
|
|
|
2013-04-22 23:12:31 +07:00
|
|
|
btrfs_leak_debug_add(&eb->leak_list, &buffers);
|
|
|
|
|
2012-03-10 04:01:49 +07:00
|
|
|
spin_lock_init(&eb->refs_lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
atomic_set(&eb->refs, 1);
|
2012-03-13 20:38:00 +07:00
|
|
|
atomic_set(&eb->io_pages, 0);
|
2010-08-07 00:21:20 +07:00
|
|
|
|
2013-02-28 21:54:18 +07:00
|
|
|
/*
|
|
|
|
* Sanity checks, currently the maximum is 64k covered by 16x 4k pages
|
|
|
|
*/
|
|
|
|
BUILD_BUG_ON(BTRFS_MAX_METADATA_BLOCKSIZE
|
|
|
|
> MAX_INLINE_EXTENT_BUFFER_SIZE);
|
|
|
|
BUG_ON(len > MAX_INLINE_EXTENT_BUFFER_SIZE);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
return eb;
|
|
|
|
}
|
|
|
|
|
2012-05-16 22:00:02 +07:00
|
|
|
struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src)
|
|
|
|
{
|
|
|
|
unsigned long i;
|
|
|
|
struct page *p;
|
|
|
|
struct extent_buffer *new;
|
|
|
|
unsigned long num_pages = num_extent_pages(src->start, src->len);
|
|
|
|
|
2013-08-08 03:57:23 +07:00
|
|
|
new = __alloc_extent_buffer(NULL, src->start, src->len, GFP_NOFS);
|
2012-05-16 22:00:02 +07:00
|
|
|
if (new == NULL)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
for (i = 0; i < num_pages; i++) {
|
2013-08-08 03:57:23 +07:00
|
|
|
p = alloc_page(GFP_NOFS);
|
2013-08-08 01:54:37 +07:00
|
|
|
if (!p) {
|
|
|
|
btrfs_release_extent_buffer(new);
|
|
|
|
return NULL;
|
|
|
|
}
|
2012-05-16 22:00:02 +07:00
|
|
|
attach_extent_buffer_page(new, p);
|
|
|
|
WARN_ON(PageDirty(p));
|
|
|
|
SetPageUptodate(p);
|
|
|
|
new->pages[i] = p;
|
|
|
|
}
|
|
|
|
|
|
|
|
copy_extent_buffer(new, src, 0, 0, src->len);
|
|
|
|
set_bit(EXTENT_BUFFER_UPTODATE, &new->bflags);
|
|
|
|
set_bit(EXTENT_BUFFER_DUMMY, &new->bflags);
|
|
|
|
|
|
|
|
return new;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct extent_buffer *alloc_dummy_extent_buffer(u64 start, unsigned long len)
|
|
|
|
{
|
|
|
|
struct extent_buffer *eb;
|
|
|
|
unsigned long num_pages = num_extent_pages(0, len);
|
|
|
|
unsigned long i;
|
|
|
|
|
2013-08-08 03:57:23 +07:00
|
|
|
eb = __alloc_extent_buffer(NULL, start, len, GFP_NOFS);
|
2012-05-16 22:00:02 +07:00
|
|
|
if (!eb)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
for (i = 0; i < num_pages; i++) {
|
2013-08-08 03:57:23 +07:00
|
|
|
eb->pages[i] = alloc_page(GFP_NOFS);
|
2012-05-16 22:00:02 +07:00
|
|
|
if (!eb->pages[i])
|
|
|
|
goto err;
|
|
|
|
}
|
|
|
|
set_extent_buffer_uptodate(eb);
|
|
|
|
btrfs_set_header_nritems(eb, 0);
|
|
|
|
set_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);
|
|
|
|
|
|
|
|
return eb;
|
|
|
|
err:
|
2012-10-11 20:25:16 +07:00
|
|
|
for (; i > 0; i--)
|
|
|
|
__free_page(eb->pages[i - 1]);
|
2012-05-16 22:00:02 +07:00
|
|
|
__free_extent_buffer(eb);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
static void check_buffer_tree_ref(struct extent_buffer *eb)
|
|
|
|
{
|
2013-01-30 05:49:37 +07:00
|
|
|
int refs;
|
2012-03-13 20:38:00 +07:00
|
|
|
/* the ref bit is tricky. We have to make sure it is set
|
|
|
|
* if we have the buffer dirty. Otherwise the
|
|
|
|
* code to free a buffer can end up dropping a dirty
|
|
|
|
* page
|
|
|
|
*
|
|
|
|
* Once the ref bit is set, it won't go away while the
|
|
|
|
* buffer is dirty or in writeback, and it also won't
|
|
|
|
* go away while we have the reference count on the
|
|
|
|
* eb bumped.
|
|
|
|
*
|
|
|
|
* We can't just set the ref bit without bumping the
|
|
|
|
* ref on the eb because free_extent_buffer might
|
|
|
|
* see the ref bit and try to clear it. If this happens
|
|
|
|
* free_extent_buffer might end up dropping our original
|
|
|
|
* ref by mistake and freeing the page before we are able
|
|
|
|
* to add one more ref.
|
|
|
|
*
|
|
|
|
* So bump the ref count first, then set the bit. If someone
|
|
|
|
* beat us to it, drop the ref we added.
|
|
|
|
*/
|
2013-01-30 05:49:37 +07:00
|
|
|
refs = atomic_read(&eb->refs);
|
|
|
|
if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
|
|
|
|
return;
|
|
|
|
|
2012-07-21 03:11:08 +07:00
|
|
|
spin_lock(&eb->refs_lock);
|
|
|
|
if (!test_and_set_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
|
2012-03-13 20:38:00 +07:00
|
|
|
atomic_inc(&eb->refs);
|
2012-07-21 03:11:08 +07:00
|
|
|
spin_unlock(&eb->refs_lock);
|
2012-03-13 20:38:00 +07:00
|
|
|
}
|
|
|
|
|
2014-06-05 06:10:31 +07:00
|
|
|
static void mark_extent_buffer_accessed(struct extent_buffer *eb,
|
|
|
|
struct page *accessed)
|
2012-03-16 05:24:42 +07:00
|
|
|
{
|
|
|
|
unsigned long num_pages, i;
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
check_buffer_tree_ref(eb);
|
|
|
|
|
2012-03-16 05:24:42 +07:00
|
|
|
num_pages = num_extent_pages(eb->start, eb->len);
|
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
struct page *p = extent_buffer_page(eb, i);
|
2014-06-05 06:10:31 +07:00
|
|
|
if (p != accessed)
|
|
|
|
mark_page_accessed(p);
|
2012-03-16 05:24:42 +07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-12-17 01:24:27 +07:00
|
|
|
struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
|
|
|
|
u64 start)
|
2013-10-07 22:45:25 +07:00
|
|
|
{
|
|
|
|
struct extent_buffer *eb;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
2013-12-17 01:24:27 +07:00
|
|
|
eb = radix_tree_lookup(&fs_info->buffer_radix,
|
|
|
|
start >> PAGE_CACHE_SHIFT);
|
2013-10-07 22:45:25 +07:00
|
|
|
if (eb && atomic_inc_not_zero(&eb->refs)) {
|
|
|
|
rcu_read_unlock();
|
2014-06-05 06:10:31 +07:00
|
|
|
mark_extent_buffer_accessed(eb, NULL);
|
2013-10-07 22:45:25 +07:00
|
|
|
return eb;
|
|
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2014-05-08 04:06:09 +07:00
|
|
|
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
|
|
|
|
struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
|
|
|
|
u64 start, unsigned long len)
|
|
|
|
{
|
|
|
|
struct extent_buffer *eb, *exists = NULL;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
eb = find_extent_buffer(fs_info, start);
|
|
|
|
if (eb)
|
|
|
|
return eb;
|
|
|
|
eb = alloc_dummy_extent_buffer(start, len);
|
|
|
|
if (!eb)
|
|
|
|
return NULL;
|
|
|
|
eb->fs_info = fs_info;
|
|
|
|
again:
|
|
|
|
ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
|
|
|
|
if (ret)
|
|
|
|
goto free_eb;
|
|
|
|
spin_lock(&fs_info->buffer_lock);
|
|
|
|
ret = radix_tree_insert(&fs_info->buffer_radix,
|
|
|
|
start >> PAGE_CACHE_SHIFT, eb);
|
|
|
|
spin_unlock(&fs_info->buffer_lock);
|
|
|
|
radix_tree_preload_end();
|
|
|
|
if (ret == -EEXIST) {
|
|
|
|
exists = find_extent_buffer(fs_info, start);
|
|
|
|
if (exists)
|
|
|
|
goto free_eb;
|
|
|
|
else
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
check_buffer_tree_ref(eb);
|
|
|
|
set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We will free dummy extent buffer's if they come into
|
|
|
|
* free_extent_buffer with a ref count of 2, but if we are using this we
|
|
|
|
* want the buffers to stay in memory until we're done with them, so
|
|
|
|
* bump the ref count again.
|
|
|
|
*/
|
|
|
|
atomic_inc(&eb->refs);
|
|
|
|
return eb;
|
|
|
|
free_eb:
|
|
|
|
btrfs_release_extent_buffer(eb);
|
|
|
|
return exists;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2013-12-17 01:24:27 +07:00
|
|
|
struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
|
2010-08-07 00:21:20 +07:00
|
|
|
u64 start, unsigned long len)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
unsigned long num_pages = num_extent_pages(start, len);
|
|
|
|
unsigned long i;
|
|
|
|
unsigned long index = start >> PAGE_CACHE_SHIFT;
|
|
|
|
struct extent_buffer *eb;
|
2008-07-22 22:18:07 +07:00
|
|
|
struct extent_buffer *exists = NULL;
|
2008-01-25 04:13:08 +07:00
|
|
|
struct page *p;
|
2013-12-17 01:24:27 +07:00
|
|
|
struct address_space *mapping = fs_info->btree_inode->i_mapping;
|
2008-01-25 04:13:08 +07:00
|
|
|
int uptodate = 1;
|
2010-10-27 07:57:29 +07:00
|
|
|
int ret;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2013-12-17 01:24:27 +07:00
|
|
|
eb = find_extent_buffer(fs_info, start);
|
2013-10-07 22:45:25 +07:00
|
|
|
if (eb)
|
2008-07-22 22:18:07 +07:00
|
|
|
return eb;
|
|
|
|
|
2013-12-17 01:24:27 +07:00
|
|
|
eb = __alloc_extent_buffer(fs_info, start, len, GFP_NOFS);
|
2008-04-01 22:21:40 +07:00
|
|
|
if (!eb)
|
2008-01-25 04:13:08 +07:00
|
|
|
return NULL;
|
|
|
|
|
2010-08-07 00:21:20 +07:00
|
|
|
for (i = 0; i < num_pages; i++, index++) {
|
2011-07-19 23:04:14 +07:00
|
|
|
p = find_or_create_page(mapping, index, GFP_NOFS);
|
2012-10-06 03:43:45 +07:00
|
|
|
if (!p)
|
2008-07-22 22:18:07 +07:00
|
|
|
goto free_eb;
|
2012-03-08 04:20:05 +07:00
|
|
|
|
|
|
|
spin_lock(&mapping->private_lock);
|
|
|
|
if (PagePrivate(p)) {
|
|
|
|
/*
|
|
|
|
* We could have already allocated an eb for this page
|
|
|
|
* and attached one so lets see if we can get a ref on
|
|
|
|
* the existing eb, and if we can we know it's good and
|
|
|
|
* we can just return that one, else we know we can just
|
|
|
|
* overwrite page->private.
|
|
|
|
*/
|
|
|
|
exists = (struct extent_buffer *)p->private;
|
|
|
|
if (atomic_inc_not_zero(&exists->refs)) {
|
|
|
|
spin_unlock(&mapping->private_lock);
|
|
|
|
unlock_page(p);
|
2012-05-05 02:16:06 +07:00
|
|
|
page_cache_release(p);
|
2014-06-05 06:10:31 +07:00
|
|
|
mark_extent_buffer_accessed(exists, p);
|
2012-03-08 04:20:05 +07:00
|
|
|
goto free_eb;
|
|
|
|
}
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
/*
|
2012-03-08 04:20:05 +07:00
|
|
|
* Do this so attach doesn't complain and we need to
|
|
|
|
* drop the ref the old guy had.
|
|
|
|
*/
|
|
|
|
ClearPagePrivate(p);
|
2012-03-13 20:38:00 +07:00
|
|
|
WARN_ON(PageDirty(p));
|
2012-03-08 04:20:05 +07:00
|
|
|
page_cache_release(p);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
2012-03-08 04:20:05 +07:00
|
|
|
attach_extent_buffer_page(eb, p);
|
|
|
|
spin_unlock(&mapping->private_lock);
|
2012-03-13 20:38:00 +07:00
|
|
|
WARN_ON(PageDirty(p));
|
2010-08-07 00:21:20 +07:00
|
|
|
eb->pages[i] = p;
|
2008-01-25 04:13:08 +07:00
|
|
|
if (!PageUptodate(p))
|
|
|
|
uptodate = 0;
|
2011-02-11 00:35:00 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* see below about how we avoid a nasty race with release page
|
|
|
|
* and why we unlock later
|
|
|
|
*/
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
if (uptodate)
|
Btrfs: Change btree locking to use explicit blocking points
Most of the btrfs metadata operations can be protected by a spinlock,
but some operations still need to schedule.
So far, btrfs has been using a mutex along with a trylock loop,
most of the time it is able to avoid going for the full mutex, so
the trylock loop is a big performance gain.
This commit is step one for getting rid of the blocking locks entirely.
btrfs_tree_lock takes a spinlock, and the code explicitly switches
to a blocking lock when it starts an operation that can schedule.
We'll be able get rid of the blocking locks in smaller pieces over time.
Tracing allows us to find the most common cause of blocking, so we
can start with the hot spots first.
The basic idea is:
btrfs_tree_lock() returns with the spin lock held
btrfs_set_lock_blocking() sets the EXTENT_BUFFER_BLOCKING bit in
the extent buffer flags, and then drops the spin lock. The buffer is
still considered locked by all of the btrfs code.
If btrfs_tree_lock gets the spinlock but finds the blocking bit set, it drops
the spin lock and waits on a wait queue for the blocking bit to go away.
Much of the code that needs to set the blocking bit finishes without actually
blocking a good percentage of the time. So, an adaptive spin is still
used against the blocking bit to avoid very high context switch rates.
btrfs_clear_lock_blocking() clears the blocking bit and returns
with the spinlock held again.
btrfs_tree_unlock() can be called on either blocking or spinning locks,
it does the right thing based on the blocking bit.
ctree.c has a helper function to set/clear all the locked buffers in a
path as blocking.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-02-04 21:25:08 +07:00
|
|
|
set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
|
2012-03-09 21:51:43 +07:00
|
|
|
again:
|
2010-10-27 07:57:29 +07:00
|
|
|
ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
|
|
|
|
if (ret)
|
|
|
|
goto free_eb;
|
|
|
|
|
2013-12-17 01:24:27 +07:00
|
|
|
spin_lock(&fs_info->buffer_lock);
|
|
|
|
ret = radix_tree_insert(&fs_info->buffer_radix,
|
|
|
|
start >> PAGE_CACHE_SHIFT, eb);
|
|
|
|
spin_unlock(&fs_info->buffer_lock);
|
2013-10-07 22:45:25 +07:00
|
|
|
radix_tree_preload_end();
|
2010-10-27 07:57:29 +07:00
|
|
|
if (ret == -EEXIST) {
|
2013-12-17 01:24:27 +07:00
|
|
|
exists = find_extent_buffer(fs_info, start);
|
2013-10-07 22:45:25 +07:00
|
|
|
if (exists)
|
|
|
|
goto free_eb;
|
|
|
|
else
|
2012-03-09 21:51:43 +07:00
|
|
|
goto again;
|
2008-07-22 22:18:07 +07:00
|
|
|
}
|
|
|
|
/* add one reference for the tree */
|
2012-03-13 20:38:00 +07:00
|
|
|
check_buffer_tree_ref(eb);
|
2013-12-13 22:41:51 +07:00
|
|
|
set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
|
2011-02-11 00:35:00 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* there is a race where release page may have
|
|
|
|
* tried to find this extent buffer in the radix
|
|
|
|
* but failed. It will tell the VM it is safe to
|
|
|
|
* reclaim the, and it will clear the page private bit.
|
|
|
|
* We must make sure to set the page private bit properly
|
|
|
|
* after the extent buffer is in the radix tree so
|
|
|
|
* it doesn't get lost
|
|
|
|
*/
|
2010-08-07 00:21:20 +07:00
|
|
|
SetPageChecked(eb->pages[0]);
|
|
|
|
for (i = 1; i < num_pages; i++) {
|
|
|
|
p = extent_buffer_page(eb, i);
|
|
|
|
ClearPageChecked(p);
|
|
|
|
unlock_page(p);
|
|
|
|
}
|
|
|
|
unlock_page(eb->pages[0]);
|
2008-01-25 04:13:08 +07:00
|
|
|
return eb;
|
|
|
|
|
2008-07-22 22:18:07 +07:00
|
|
|
free_eb:
|
2010-08-07 00:21:20 +07:00
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
if (eb->pages[i])
|
|
|
|
unlock_page(eb->pages[i]);
|
|
|
|
}
|
2011-02-11 00:35:00 +07:00
|
|
|
|
2012-05-05 02:16:06 +07:00
|
|
|
WARN_ON(!atomic_dec_and_test(&eb->refs));
|
2010-10-27 07:57:29 +07:00
|
|
|
btrfs_release_extent_buffer(eb);
|
2008-07-22 22:18:07 +07:00
|
|
|
return exists;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2012-03-10 04:01:49 +07:00
|
|
|
static inline void btrfs_release_extent_buffer_rcu(struct rcu_head *head)
|
|
|
|
{
|
|
|
|
struct extent_buffer *eb =
|
|
|
|
container_of(head, struct extent_buffer, rcu_head);
|
|
|
|
|
|
|
|
__free_extent_buffer(eb);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Expects to have eb->eb_lock already held */
|
2013-04-26 21:56:29 +07:00
|
|
|
static int release_extent_buffer(struct extent_buffer *eb)
|
2012-03-10 04:01:49 +07:00
|
|
|
{
|
|
|
|
WARN_ON(atomic_read(&eb->refs) == 0);
|
|
|
|
if (atomic_dec_and_test(&eb->refs)) {
|
2013-12-13 22:41:51 +07:00
|
|
|
if (test_and_clear_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags)) {
|
2013-12-17 01:24:27 +07:00
|
|
|
struct btrfs_fs_info *fs_info = eb->fs_info;
|
2012-03-10 04:01:49 +07:00
|
|
|
|
2012-05-16 22:00:02 +07:00
|
|
|
spin_unlock(&eb->refs_lock);
|
2012-03-10 04:01:49 +07:00
|
|
|
|
2013-12-17 01:24:27 +07:00
|
|
|
spin_lock(&fs_info->buffer_lock);
|
|
|
|
radix_tree_delete(&fs_info->buffer_radix,
|
2012-05-16 22:00:02 +07:00
|
|
|
eb->start >> PAGE_CACHE_SHIFT);
|
2013-12-17 01:24:27 +07:00
|
|
|
spin_unlock(&fs_info->buffer_lock);
|
2013-12-13 22:41:51 +07:00
|
|
|
} else {
|
|
|
|
spin_unlock(&eb->refs_lock);
|
2012-05-16 22:00:02 +07:00
|
|
|
}
|
2012-03-10 04:01:49 +07:00
|
|
|
|
|
|
|
/* Should be safe to release our pages at this point */
|
|
|
|
btrfs_release_extent_buffer_page(eb, 0);
|
|
|
|
call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
|
2012-07-21 03:05:36 +07:00
|
|
|
return 1;
|
2012-03-10 04:01:49 +07:00
|
|
|
}
|
|
|
|
spin_unlock(&eb->refs_lock);
|
2012-07-21 03:05:36 +07:00
|
|
|
|
|
|
|
return 0;
|
2012-03-10 04:01:49 +07:00
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
void free_extent_buffer(struct extent_buffer *eb)
|
|
|
|
{
|
2013-01-30 05:49:37 +07:00
|
|
|
int refs;
|
|
|
|
int old;
|
2008-01-25 04:13:08 +07:00
|
|
|
if (!eb)
|
|
|
|
return;
|
|
|
|
|
2013-01-30 05:49:37 +07:00
|
|
|
while (1) {
|
|
|
|
refs = atomic_read(&eb->refs);
|
|
|
|
if (refs <= 3)
|
|
|
|
break;
|
|
|
|
old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
|
|
|
|
if (old == refs)
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2012-03-10 04:01:49 +07:00
|
|
|
spin_lock(&eb->refs_lock);
|
2012-05-16 22:00:02 +07:00
|
|
|
if (atomic_read(&eb->refs) == 2 &&
|
|
|
|
test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
|
|
|
|
atomic_dec(&eb->refs);
|
|
|
|
|
2012-03-10 04:01:49 +07:00
|
|
|
if (atomic_read(&eb->refs) == 2 &&
|
|
|
|
test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
|
2012-03-13 20:38:00 +07:00
|
|
|
!extent_buffer_under_io(eb) &&
|
2012-03-10 04:01:49 +07:00
|
|
|
test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
|
|
|
|
atomic_dec(&eb->refs);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* I know this is terrible, but it's temporary until we stop tracking
|
|
|
|
* the uptodate bits and such for the extent buffers.
|
|
|
|
*/
|
2013-04-26 21:56:29 +07:00
|
|
|
release_extent_buffer(eb);
|
2012-03-10 04:01:49 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
void free_extent_buffer_stale(struct extent_buffer *eb)
|
|
|
|
{
|
|
|
|
if (!eb)
|
2008-01-25 04:13:08 +07:00
|
|
|
return;
|
|
|
|
|
2012-03-10 04:01:49 +07:00
|
|
|
spin_lock(&eb->refs_lock);
|
|
|
|
set_bit(EXTENT_BUFFER_STALE, &eb->bflags);
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
|
2012-03-10 04:01:49 +07:00
|
|
|
test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
|
|
|
|
atomic_dec(&eb->refs);
|
2013-04-26 21:56:29 +07:00
|
|
|
release_extent_buffer(eb);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2012-03-29 07:31:37 +07:00
|
|
|
void clear_extent_buffer_dirty(struct extent_buffer *eb)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
unsigned long i;
|
|
|
|
unsigned long num_pages;
|
|
|
|
struct page *page;
|
|
|
|
|
|
|
|
num_pages = num_extent_pages(eb->start, eb->len);
|
|
|
|
|
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
page = extent_buffer_page(eb, i);
|
2009-03-13 22:00:37 +07:00
|
|
|
if (!PageDirty(page))
|
2008-11-20 00:44:22 +07:00
|
|
|
continue;
|
|
|
|
|
2008-07-22 22:18:08 +07:00
|
|
|
lock_page(page);
|
2011-02-11 00:35:00 +07:00
|
|
|
WARN_ON(!PagePrivate(page));
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
clear_page_dirty_for_io(page);
|
2008-07-31 03:54:26 +07:00
|
|
|
spin_lock_irq(&page->mapping->tree_lock);
|
2008-01-25 04:13:08 +07:00
|
|
|
if (!PageDirty(page)) {
|
|
|
|
radix_tree_tag_clear(&page->mapping->page_tree,
|
|
|
|
page_index(page),
|
|
|
|
PAGECACHE_TAG_DIRTY);
|
|
|
|
}
|
2008-07-31 03:54:26 +07:00
|
|
|
spin_unlock_irq(&page->mapping->tree_lock);
|
2011-11-04 23:29:37 +07:00
|
|
|
ClearPageError(page);
|
2008-07-22 22:18:08 +07:00
|
|
|
unlock_page(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
2012-03-13 20:38:00 +07:00
|
|
|
WARN_ON(atomic_read(&eb->refs) == 0);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
int set_extent_buffer_dirty(struct extent_buffer *eb)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
unsigned long i;
|
|
|
|
unsigned long num_pages;
|
2009-03-13 22:00:37 +07:00
|
|
|
int was_dirty = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
check_buffer_tree_ref(eb);
|
|
|
|
|
2009-03-13 22:00:37 +07:00
|
|
|
was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
|
2012-03-13 20:38:00 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
num_pages = num_extent_pages(eb->start, eb->len);
|
2012-03-10 04:01:49 +07:00
|
|
|
WARN_ON(atomic_read(&eb->refs) == 0);
|
2012-03-13 20:38:00 +07:00
|
|
|
WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));
|
|
|
|
|
2009-03-13 22:00:37 +07:00
|
|
|
for (i = 0; i < num_pages; i++)
|
2012-03-13 20:38:00 +07:00
|
|
|
set_page_dirty(extent_buffer_page(eb, i));
|
2009-03-13 22:00:37 +07:00
|
|
|
return was_dirty;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
int clear_extent_buffer_uptodate(struct extent_buffer *eb)
|
2008-05-13 00:39:03 +07:00
|
|
|
{
|
|
|
|
unsigned long i;
|
|
|
|
struct page *page;
|
|
|
|
unsigned long num_pages;
|
|
|
|
|
Btrfs: Change btree locking to use explicit blocking points
Most of the btrfs metadata operations can be protected by a spinlock,
but some operations still need to schedule.
So far, btrfs has been using a mutex along with a trylock loop,
most of the time it is able to avoid going for the full mutex, so
the trylock loop is a big performance gain.
This commit is step one for getting rid of the blocking locks entirely.
btrfs_tree_lock takes a spinlock, and the code explicitly switches
to a blocking lock when it starts an operation that can schedule.
We'll be able get rid of the blocking locks in smaller pieces over time.
Tracing allows us to find the most common cause of blocking, so we
can start with the hot spots first.
The basic idea is:
btrfs_tree_lock() returns with the spin lock held
btrfs_set_lock_blocking() sets the EXTENT_BUFFER_BLOCKING bit in
the extent buffer flags, and then drops the spin lock. The buffer is
still considered locked by all of the btrfs code.
If btrfs_tree_lock gets the spinlock but finds the blocking bit set, it drops
the spin lock and waits on a wait queue for the blocking bit to go away.
Much of the code that needs to set the blocking bit finishes without actually
blocking a good percentage of the time. So, an adaptive spin is still
used against the blocking bit to avoid very high context switch rates.
btrfs_clear_lock_blocking() clears the blocking bit and returns
with the spinlock held again.
btrfs_tree_unlock() can be called on either blocking or spinning locks,
it does the right thing based on the blocking bit.
ctree.c has a helper function to set/clear all the locked buffers in a
path as blocking.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-02-04 21:25:08 +07:00
|
|
|
clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
|
2012-03-13 20:38:00 +07:00
|
|
|
num_pages = num_extent_pages(eb->start, eb->len);
|
2008-05-13 00:39:03 +07:00
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
page = extent_buffer_page(eb, i);
|
2008-07-30 21:29:12 +07:00
|
|
|
if (page)
|
|
|
|
ClearPageUptodate(page);
|
2008-05-13 00:39:03 +07:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
int set_extent_buffer_uptodate(struct extent_buffer *eb)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
unsigned long i;
|
|
|
|
struct page *page;
|
|
|
|
unsigned long num_pages;
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
|
2008-01-25 04:13:08 +07:00
|
|
|
num_pages = num_extent_pages(eb->start, eb->len);
|
|
|
|
for (i = 0; i < num_pages; i++) {
|
|
|
|
page = extent_buffer_page(eb, i);
|
|
|
|
SetPageUptodate(page);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2012-03-13 20:38:00 +07:00
|
|
|
int extent_buffer_uptodate(struct extent_buffer *eb)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
2012-03-13 20:38:00 +07:00
|
|
|
return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
int read_extent_buffer_pages(struct extent_io_tree *tree,
|
2011-06-10 19:06:53 +07:00
|
|
|
struct extent_buffer *eb, u64 start, int wait,
|
2008-04-10 03:28:12 +07:00
|
|
|
get_extent_t *get_extent, int mirror_num)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
unsigned long i;
|
|
|
|
unsigned long start_i;
|
|
|
|
struct page *page;
|
|
|
|
int err;
|
|
|
|
int ret = 0;
|
2008-04-10 03:28:12 +07:00
|
|
|
int locked_pages = 0;
|
|
|
|
int all_uptodate = 1;
|
2008-01-25 04:13:08 +07:00
|
|
|
unsigned long num_pages;
|
2010-08-07 00:21:20 +07:00
|
|
|
unsigned long num_reads = 0;
|
2008-02-07 22:50:54 +07:00
|
|
|
struct bio *bio = NULL;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-30 01:49:59 +07:00
|
|
|
unsigned long bio_flags = 0;
|
2008-02-07 22:50:54 +07:00
|
|
|
|
Btrfs: Change btree locking to use explicit blocking points
Most of the btrfs metadata operations can be protected by a spinlock,
but some operations still need to schedule.
So far, btrfs has been using a mutex along with a trylock loop,
most of the time it is able to avoid going for the full mutex, so
the trylock loop is a big performance gain.
This commit is step one for getting rid of the blocking locks entirely.
btrfs_tree_lock takes a spinlock, and the code explicitly switches
to a blocking lock when it starts an operation that can schedule.
We'll be able get rid of the blocking locks in smaller pieces over time.
Tracing allows us to find the most common cause of blocking, so we
can start with the hot spots first.
The basic idea is:
btrfs_tree_lock() returns with the spin lock held
btrfs_set_lock_blocking() sets the EXTENT_BUFFER_BLOCKING bit in
the extent buffer flags, and then drops the spin lock. The buffer is
still considered locked by all of the btrfs code.
If btrfs_tree_lock gets the spinlock but finds the blocking bit set, it drops
the spin lock and waits on a wait queue for the blocking bit to go away.
Much of the code that needs to set the blocking bit finishes without actually
blocking a good percentage of the time. So, an adaptive spin is still
used against the blocking bit to avoid very high context switch rates.
btrfs_clear_lock_blocking() clears the blocking bit and returns
with the spinlock held again.
btrfs_tree_unlock() can be called on either blocking or spinning locks,
it does the right thing based on the blocking bit.
ctree.c has a helper function to set/clear all the locked buffers in a
path as blocking.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-02-04 21:25:08 +07:00
|
|
|
if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
|
2008-01-25 04:13:08 +07:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (start) {
|
|
|
|
WARN_ON(start < eb->start);
|
|
|
|
start_i = (start >> PAGE_CACHE_SHIFT) -
|
|
|
|
(eb->start >> PAGE_CACHE_SHIFT);
|
|
|
|
} else {
|
|
|
|
start_i = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
num_pages = num_extent_pages(eb->start, eb->len);
|
|
|
|
for (i = start_i; i < num_pages; i++) {
|
|
|
|
page = extent_buffer_page(eb, i);
|
2011-06-10 19:06:53 +07:00
|
|
|
if (wait == WAIT_NONE) {
|
2008-08-07 22:19:43 +07:00
|
|
|
if (!trylock_page(page))
|
2008-04-10 03:28:12 +07:00
|
|
|
goto unlock_exit;
|
2008-01-25 04:13:08 +07:00
|
|
|
} else {
|
|
|
|
lock_page(page);
|
|
|
|
}
|
2008-04-10 03:28:12 +07:00
|
|
|
locked_pages++;
|
2010-08-07 00:21:20 +07:00
|
|
|
if (!PageUptodate(page)) {
|
|
|
|
num_reads++;
|
2008-04-10 03:28:12 +07:00
|
|
|
all_uptodate = 0;
|
2010-08-07 00:21:20 +07:00
|
|
|
}
|
2008-04-10 03:28:12 +07:00
|
|
|
}
|
|
|
|
if (all_uptodate) {
|
|
|
|
if (start_i == 0)
|
Btrfs: Change btree locking to use explicit blocking points
Most of the btrfs metadata operations can be protected by a spinlock,
but some operations still need to schedule.
So far, btrfs has been using a mutex along with a trylock loop,
most of the time it is able to avoid going for the full mutex, so
the trylock loop is a big performance gain.
This commit is step one for getting rid of the blocking locks entirely.
btrfs_tree_lock takes a spinlock, and the code explicitly switches
to a blocking lock when it starts an operation that can schedule.
We'll be able get rid of the blocking locks in smaller pieces over time.
Tracing allows us to find the most common cause of blocking, so we
can start with the hot spots first.
The basic idea is:
btrfs_tree_lock() returns with the spin lock held
btrfs_set_lock_blocking() sets the EXTENT_BUFFER_BLOCKING bit in
the extent buffer flags, and then drops the spin lock. The buffer is
still considered locked by all of the btrfs code.
If btrfs_tree_lock gets the spinlock but finds the blocking bit set, it drops
the spin lock and waits on a wait queue for the blocking bit to go away.
Much of the code that needs to set the blocking bit finishes without actually
blocking a good percentage of the time. So, an adaptive spin is still
used against the blocking bit to avoid very high context switch rates.
btrfs_clear_lock_blocking() clears the blocking bit and returns
with the spinlock held again.
btrfs_tree_unlock() can be called on either blocking or spinning locks,
it does the right thing based on the blocking bit.
ctree.c has a helper function to set/clear all the locked buffers in a
path as blocking.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-02-04 21:25:08 +07:00
|
|
|
set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
|
2008-04-10 03:28:12 +07:00
|
|
|
goto unlock_exit;
|
|
|
|
}
|
|
|
|
|
2012-03-27 08:57:36 +07:00
|
|
|
clear_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
|
2012-04-16 20:42:26 +07:00
|
|
|
eb->read_mirror = 0;
|
2012-03-13 20:38:00 +07:00
|
|
|
atomic_set(&eb->io_pages, num_reads);
|
2008-04-10 03:28:12 +07:00
|
|
|
for (i = start_i; i < num_pages; i++) {
|
|
|
|
page = extent_buffer_page(eb, i);
|
|
|
|
if (!PageUptodate(page)) {
|
2008-04-10 03:28:12 +07:00
|
|
|
ClearPageError(page);
|
2008-02-07 22:50:54 +07:00
|
|
|
err = __extent_read_full_page(tree, page,
|
2008-04-10 03:28:12 +07:00
|
|
|
get_extent, &bio,
|
2013-04-20 06:49:09 +07:00
|
|
|
mirror_num, &bio_flags,
|
|
|
|
READ | REQ_META);
|
2009-01-06 09:25:51 +07:00
|
|
|
if (err)
|
2008-01-25 04:13:08 +07:00
|
|
|
ret = err;
|
|
|
|
} else {
|
|
|
|
unlock_page(page);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-10-04 10:23:14 +07:00
|
|
|
if (bio) {
|
2013-04-20 06:49:09 +07:00
|
|
|
err = submit_one_bio(READ | REQ_META, bio, mirror_num,
|
|
|
|
bio_flags);
|
2012-03-12 22:03:00 +07:00
|
|
|
if (err)
|
|
|
|
return err;
|
2011-10-04 10:23:14 +07:00
|
|
|
}
|
2008-02-07 22:50:54 +07:00
|
|
|
|
2011-06-10 19:06:53 +07:00
|
|
|
if (ret || wait != WAIT_COMPLETE)
|
2008-01-25 04:13:08 +07:00
|
|
|
return ret;
|
2009-01-06 09:25:51 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
for (i = start_i; i < num_pages; i++) {
|
|
|
|
page = extent_buffer_page(eb, i);
|
|
|
|
wait_on_page_locked(page);
|
2009-01-06 09:25:51 +07:00
|
|
|
if (!PageUptodate(page))
|
2008-01-25 04:13:08 +07:00
|
|
|
ret = -EIO;
|
|
|
|
}
|
2009-01-06 09:25:51 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
return ret;
|
2008-04-10 03:28:12 +07:00
|
|
|
|
|
|
|
unlock_exit:
|
|
|
|
i = start_i;
|
2009-01-06 09:25:51 +07:00
|
|
|
while (locked_pages > 0) {
|
2008-04-10 03:28:12 +07:00
|
|
|
page = extent_buffer_page(eb, i);
|
|
|
|
i++;
|
|
|
|
unlock_page(page);
|
|
|
|
locked_pages--;
|
|
|
|
}
|
|
|
|
return ret;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
void read_extent_buffer(struct extent_buffer *eb, void *dstv,
|
|
|
|
unsigned long start,
|
|
|
|
unsigned long len)
|
|
|
|
{
|
|
|
|
size_t cur;
|
|
|
|
size_t offset;
|
|
|
|
struct page *page;
|
|
|
|
char *kaddr;
|
|
|
|
char *dst = (char *)dstv;
|
|
|
|
size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
|
|
|
|
unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
|
|
|
|
|
|
|
|
WARN_ON(start > eb->len);
|
|
|
|
WARN_ON(start + len > eb->start + eb->len);
|
|
|
|
|
2013-08-20 18:20:16 +07:00
|
|
|
offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (len > 0) {
|
2008-01-25 04:13:08 +07:00
|
|
|
page = extent_buffer_page(eb, i);
|
|
|
|
|
|
|
|
cur = min(len, (PAGE_CACHE_SIZE - offset));
|
2011-07-19 23:04:14 +07:00
|
|
|
kaddr = page_address(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
memcpy(dst, kaddr + offset, cur);
|
|
|
|
|
|
|
|
dst += cur;
|
|
|
|
len -= cur;
|
|
|
|
offset = 0;
|
|
|
|
i++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-01-30 22:24:01 +07:00
|
|
|
int read_extent_buffer_to_user(struct extent_buffer *eb, void __user *dstv,
|
|
|
|
unsigned long start,
|
|
|
|
unsigned long len)
|
|
|
|
{
|
|
|
|
size_t cur;
|
|
|
|
size_t offset;
|
|
|
|
struct page *page;
|
|
|
|
char *kaddr;
|
|
|
|
char __user *dst = (char __user *)dstv;
|
|
|
|
size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
|
|
|
|
unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
WARN_ON(start > eb->len);
|
|
|
|
WARN_ON(start + len > eb->start + eb->len);
|
|
|
|
|
|
|
|
offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
|
|
|
|
|
|
|
|
while (len > 0) {
|
|
|
|
page = extent_buffer_page(eb, i);
|
|
|
|
|
|
|
|
cur = min(len, (PAGE_CACHE_SIZE - offset));
|
|
|
|
kaddr = page_address(page);
|
|
|
|
if (copy_to_user(dst, kaddr + offset, cur)) {
|
|
|
|
ret = -EFAULT;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
dst += cur;
|
|
|
|
len -= cur;
|
|
|
|
offset = 0;
|
|
|
|
i++;
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
|
2011-07-19 23:04:14 +07:00
|
|
|
unsigned long min_len, char **map,
|
2008-01-25 04:13:08 +07:00
|
|
|
unsigned long *map_start,
|
2011-07-19 23:04:14 +07:00
|
|
|
unsigned long *map_len)
|
2008-01-25 04:13:08 +07:00
|
|
|
{
|
|
|
|
size_t offset = start & (PAGE_CACHE_SIZE - 1);
|
|
|
|
char *kaddr;
|
|
|
|
struct page *p;
|
|
|
|
size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
|
|
|
|
unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
|
|
|
|
unsigned long end_i = (start_offset + start + min_len - 1) >>
|
|
|
|
PAGE_CACHE_SHIFT;
|
|
|
|
|
|
|
|
if (i != end_i)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
if (i == 0) {
|
|
|
|
offset = start_offset;
|
|
|
|
*map_start = 0;
|
|
|
|
} else {
|
|
|
|
offset = 0;
|
|
|
|
*map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset;
|
|
|
|
}
|
2009-01-06 09:25:51 +07:00
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
if (start + min_len > eb->len) {
|
2012-11-03 17:58:34 +07:00
|
|
|
WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
|
2013-08-20 18:20:07 +07:00
|
|
|
"wanted %lu %lu\n",
|
|
|
|
eb->start, eb->len, start, min_len);
|
2011-03-16 01:52:12 +07:00
|
|
|
return -EINVAL;
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
p = extent_buffer_page(eb, i);
|
2011-07-19 23:04:14 +07:00
|
|
|
kaddr = page_address(p);
|
2008-01-25 04:13:08 +07:00
|
|
|
*map = kaddr + offset;
|
|
|
|
*map_len = PAGE_CACHE_SIZE - offset;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
|
|
|
|
unsigned long start,
|
|
|
|
unsigned long len)
|
|
|
|
{
|
|
|
|
size_t cur;
|
|
|
|
size_t offset;
|
|
|
|
struct page *page;
|
|
|
|
char *kaddr;
|
|
|
|
char *ptr = (char *)ptrv;
|
|
|
|
size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
|
|
|
|
unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
WARN_ON(start > eb->len);
|
|
|
|
WARN_ON(start + len > eb->start + eb->len);
|
|
|
|
|
2013-08-20 18:20:16 +07:00
|
|
|
offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (len > 0) {
|
2008-01-25 04:13:08 +07:00
|
|
|
page = extent_buffer_page(eb, i);
|
|
|
|
|
|
|
|
cur = min(len, (PAGE_CACHE_SIZE - offset));
|
|
|
|
|
2011-07-19 23:04:14 +07:00
|
|
|
kaddr = page_address(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
ret = memcmp(ptr, kaddr + offset, cur);
|
|
|
|
if (ret)
|
|
|
|
break;
|
|
|
|
|
|
|
|
ptr += cur;
|
|
|
|
len -= cur;
|
|
|
|
offset = 0;
|
|
|
|
i++;
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
void write_extent_buffer(struct extent_buffer *eb, const void *srcv,
|
|
|
|
unsigned long start, unsigned long len)
|
|
|
|
{
|
|
|
|
size_t cur;
|
|
|
|
size_t offset;
|
|
|
|
struct page *page;
|
|
|
|
char *kaddr;
|
|
|
|
char *src = (char *)srcv;
|
|
|
|
size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
|
|
|
|
unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
|
|
|
|
|
|
|
|
WARN_ON(start > eb->len);
|
|
|
|
WARN_ON(start + len > eb->start + eb->len);
|
|
|
|
|
2013-08-20 18:20:16 +07:00
|
|
|
offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (len > 0) {
|
2008-01-25 04:13:08 +07:00
|
|
|
page = extent_buffer_page(eb, i);
|
|
|
|
WARN_ON(!PageUptodate(page));
|
|
|
|
|
|
|
|
cur = min(len, PAGE_CACHE_SIZE - offset);
|
2011-07-19 23:04:14 +07:00
|
|
|
kaddr = page_address(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
memcpy(kaddr + offset, src, cur);
|
|
|
|
|
|
|
|
src += cur;
|
|
|
|
len -= cur;
|
|
|
|
offset = 0;
|
|
|
|
i++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void memset_extent_buffer(struct extent_buffer *eb, char c,
|
|
|
|
unsigned long start, unsigned long len)
|
|
|
|
{
|
|
|
|
size_t cur;
|
|
|
|
size_t offset;
|
|
|
|
struct page *page;
|
|
|
|
char *kaddr;
|
|
|
|
size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
|
|
|
|
unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
|
|
|
|
|
|
|
|
WARN_ON(start > eb->len);
|
|
|
|
WARN_ON(start + len > eb->start + eb->len);
|
|
|
|
|
2013-08-20 18:20:16 +07:00
|
|
|
offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (len > 0) {
|
2008-01-25 04:13:08 +07:00
|
|
|
page = extent_buffer_page(eb, i);
|
|
|
|
WARN_ON(!PageUptodate(page));
|
|
|
|
|
|
|
|
cur = min(len, PAGE_CACHE_SIZE - offset);
|
2011-07-19 23:04:14 +07:00
|
|
|
kaddr = page_address(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
memset(kaddr + offset, c, cur);
|
|
|
|
|
|
|
|
len -= cur;
|
|
|
|
offset = 0;
|
|
|
|
i++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
|
|
|
|
unsigned long dst_offset, unsigned long src_offset,
|
|
|
|
unsigned long len)
|
|
|
|
{
|
|
|
|
u64 dst_len = dst->len;
|
|
|
|
size_t cur;
|
|
|
|
size_t offset;
|
|
|
|
struct page *page;
|
|
|
|
char *kaddr;
|
|
|
|
size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
|
|
|
|
unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
|
|
|
|
|
|
|
|
WARN_ON(src->len != dst_len);
|
|
|
|
|
|
|
|
offset = (start_offset + dst_offset) &
|
2013-08-20 18:20:16 +07:00
|
|
|
(PAGE_CACHE_SIZE - 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (len > 0) {
|
2008-01-25 04:13:08 +07:00
|
|
|
page = extent_buffer_page(dst, i);
|
|
|
|
WARN_ON(!PageUptodate(page));
|
|
|
|
|
|
|
|
cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset));
|
|
|
|
|
2011-07-19 23:04:14 +07:00
|
|
|
kaddr = page_address(page);
|
2008-01-25 04:13:08 +07:00
|
|
|
read_extent_buffer(src, kaddr + offset, src_offset, cur);
|
|
|
|
|
|
|
|
src_offset += cur;
|
|
|
|
len -= cur;
|
|
|
|
offset = 0;
|
|
|
|
i++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-04-12 04:52:52 +07:00
|
|
|
static inline bool areas_overlap(unsigned long src, unsigned long dst, unsigned long len)
|
|
|
|
{
|
|
|
|
unsigned long distance = (src > dst) ? src - dst : dst - src;
|
|
|
|
return distance < len;
|
|
|
|
}
|
|
|
|
|
2008-01-25 04:13:08 +07:00
|
|
|
static void copy_pages(struct page *dst_page, struct page *src_page,
|
|
|
|
unsigned long dst_off, unsigned long src_off,
|
|
|
|
unsigned long len)
|
|
|
|
{
|
2011-07-19 23:04:14 +07:00
|
|
|
char *dst_kaddr = page_address(dst_page);
|
2008-01-25 04:13:08 +07:00
|
|
|
char *src_kaddr;
|
2010-08-07 00:21:20 +07:00
|
|
|
int must_memmove = 0;
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2011-04-12 04:52:52 +07:00
|
|
|
if (dst_page != src_page) {
|
2011-07-19 23:04:14 +07:00
|
|
|
src_kaddr = page_address(src_page);
|
2011-04-12 04:52:52 +07:00
|
|
|
} else {
|
2008-01-25 04:13:08 +07:00
|
|
|
src_kaddr = dst_kaddr;
|
2010-08-07 00:21:20 +07:00
|
|
|
if (areas_overlap(src_off, dst_off, len))
|
|
|
|
must_memmove = 1;
|
2011-04-12 04:52:52 +07:00
|
|
|
}
|
2008-01-25 04:13:08 +07:00
|
|
|
|
2010-08-07 00:21:20 +07:00
|
|
|
if (must_memmove)
|
|
|
|
memmove(dst_kaddr + dst_off, src_kaddr + src_off, len);
|
|
|
|
else
|
|
|
|
memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
|
2008-01-25 04:13:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
|
|
|
|
unsigned long src_offset, unsigned long len)
|
|
|
|
{
|
|
|
|
size_t cur;
|
|
|
|
size_t dst_off_in_page;
|
|
|
|
size_t src_off_in_page;
|
|
|
|
size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
|
|
|
|
unsigned long dst_i;
|
|
|
|
unsigned long src_i;
|
|
|
|
|
|
|
|
if (src_offset + len > dst->len) {
|
2013-12-20 23:37:06 +07:00
|
|
|
printk(KERN_ERR "BTRFS: memmove bogus src_offset %lu move "
|
2009-01-06 09:25:51 +07:00
|
|
|
"len %lu dst len %lu\n", src_offset, len, dst->len);
|
2008-01-25 04:13:08 +07:00
|
|
|
BUG_ON(1);
|
|
|
|
}
|
|
|
|
if (dst_offset + len > dst->len) {
|
2013-12-20 23:37:06 +07:00
|
|
|
printk(KERN_ERR "BTRFS: memmove bogus dst_offset %lu move "
|
2009-01-06 09:25:51 +07:00
|
|
|
"len %lu dst len %lu\n", dst_offset, len, dst->len);
|
2008-01-25 04:13:08 +07:00
|
|
|
BUG_ON(1);
|
|
|
|
}
|
|
|
|
|
2009-01-06 09:25:51 +07:00
|
|
|
while (len > 0) {
|
2008-01-25 04:13:08 +07:00
|
|
|
dst_off_in_page = (start_offset + dst_offset) &
|
2013-08-20 18:20:16 +07:00
|
|
|
(PAGE_CACHE_SIZE - 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
src_off_in_page = (start_offset + src_offset) &
|
2013-08-20 18:20:16 +07:00
|
|
|
(PAGE_CACHE_SIZE - 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
|
|
|
|
src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT;
|
|
|
|
|
|
|
|
cur = min(len, (unsigned long)(PAGE_CACHE_SIZE -
|
|
|
|
src_off_in_page));
|
|
|
|
cur = min_t(unsigned long, cur,
|
|
|
|
(unsigned long)(PAGE_CACHE_SIZE - dst_off_in_page));
|
|
|
|
|
|
|
|
copy_pages(extent_buffer_page(dst, dst_i),
|
|
|
|
extent_buffer_page(dst, src_i),
|
|
|
|
dst_off_in_page, src_off_in_page, cur);
|
|
|
|
|
|
|
|
src_offset += cur;
|
|
|
|
dst_offset += cur;
|
|
|
|
len -= cur;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
|
|
|
|
unsigned long src_offset, unsigned long len)
|
|
|
|
{
|
|
|
|
size_t cur;
|
|
|
|
size_t dst_off_in_page;
|
|
|
|
size_t src_off_in_page;
|
|
|
|
unsigned long dst_end = dst_offset + len - 1;
|
|
|
|
unsigned long src_end = src_offset + len - 1;
|
|
|
|
size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
|
|
|
|
unsigned long dst_i;
|
|
|
|
unsigned long src_i;
|
|
|
|
|
|
|
|
if (src_offset + len > dst->len) {
|
2013-12-20 23:37:06 +07:00
|
|
|
printk(KERN_ERR "BTRFS: memmove bogus src_offset %lu move "
|
2009-01-06 09:25:51 +07:00
|
|
|
"len %lu len %lu\n", src_offset, len, dst->len);
|
2008-01-25 04:13:08 +07:00
|
|
|
BUG_ON(1);
|
|
|
|
}
|
|
|
|
if (dst_offset + len > dst->len) {
|
2013-12-20 23:37:06 +07:00
|
|
|
printk(KERN_ERR "BTRFS: memmove bogus dst_offset %lu move "
|
2009-01-06 09:25:51 +07:00
|
|
|
"len %lu len %lu\n", dst_offset, len, dst->len);
|
2008-01-25 04:13:08 +07:00
|
|
|
BUG_ON(1);
|
|
|
|
}
|
2010-08-07 00:21:20 +07:00
|
|
|
if (dst_offset < src_offset) {
|
2008-01-25 04:13:08 +07:00
|
|
|
memcpy_extent_buffer(dst, dst_offset, src_offset, len);
|
|
|
|
return;
|
|
|
|
}
|
2009-01-06 09:25:51 +07:00
|
|
|
while (len > 0) {
|
2008-01-25 04:13:08 +07:00
|
|
|
dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT;
|
|
|
|
src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT;
|
|
|
|
|
|
|
|
dst_off_in_page = (start_offset + dst_end) &
|
2013-08-20 18:20:16 +07:00
|
|
|
(PAGE_CACHE_SIZE - 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
src_off_in_page = (start_offset + src_end) &
|
2013-08-20 18:20:16 +07:00
|
|
|
(PAGE_CACHE_SIZE - 1);
|
2008-01-25 04:13:08 +07:00
|
|
|
|
|
|
|
cur = min_t(unsigned long, len, src_off_in_page + 1);
|
|
|
|
cur = min(cur, dst_off_in_page + 1);
|
2013-10-17 02:10:33 +07:00
|
|
|
copy_pages(extent_buffer_page(dst, dst_i),
|
2008-01-25 04:13:08 +07:00
|
|
|
extent_buffer_page(dst, src_i),
|
|
|
|
dst_off_in_page - cur + 1,
|
|
|
|
src_off_in_page - cur + 1, cur);
|
|
|
|
|
|
|
|
dst_end -= cur;
|
|
|
|
src_end -= cur;
|
|
|
|
len -= cur;
|
|
|
|
}
|
|
|
|
}
|
2008-07-22 22:18:07 +07:00
|
|
|
|
2013-04-26 21:56:29 +07:00
|
|
|
int try_release_extent_buffer(struct page *page)
|
2010-10-27 07:57:29 +07:00
|
|
|
{
|
2008-07-22 22:18:07 +07:00
|
|
|
struct extent_buffer *eb;
|
|
|
|
|
2012-03-10 04:01:49 +07:00
|
|
|
/*
|
|
|
|
* We need to make sure noboody is attaching this page to an eb right
|
|
|
|
* now.
|
|
|
|
*/
|
|
|
|
spin_lock(&page->mapping->private_lock);
|
|
|
|
if (!PagePrivate(page)) {
|
|
|
|
spin_unlock(&page->mapping->private_lock);
|
2012-03-08 04:20:05 +07:00
|
|
|
return 1;
|
2010-11-22 10:27:44 +07:00
|
|
|
}
|
2008-07-22 22:18:07 +07:00
|
|
|
|
2012-03-10 04:01:49 +07:00
|
|
|
eb = (struct extent_buffer *)page->private;
|
|
|
|
BUG_ON(!eb);
|
2010-10-27 07:57:29 +07:00
|
|
|
|
|
|
|
/*
|
2012-03-10 04:01:49 +07:00
|
|
|
* This is a little awful but should be ok, we need to make sure that
|
|
|
|
* the eb doesn't disappear out from under us while we're looking at
|
|
|
|
* this page.
|
2010-10-27 07:57:29 +07:00
|
|
|
*/
|
2012-03-10 04:01:49 +07:00
|
|
|
spin_lock(&eb->refs_lock);
|
2012-03-13 20:38:00 +07:00
|
|
|
if (atomic_read(&eb->refs) != 1 || extent_buffer_under_io(eb)) {
|
2012-03-10 04:01:49 +07:00
|
|
|
spin_unlock(&eb->refs_lock);
|
|
|
|
spin_unlock(&page->mapping->private_lock);
|
|
|
|
return 0;
|
2009-03-13 22:00:37 +07:00
|
|
|
}
|
2012-03-10 04:01:49 +07:00
|
|
|
spin_unlock(&page->mapping->private_lock);
|
2010-10-27 07:57:29 +07:00
|
|
|
|
2010-10-27 07:57:29 +07:00
|
|
|
/*
|
2012-03-10 04:01:49 +07:00
|
|
|
* If tree ref isn't set then we know the ref on this eb is a real ref,
|
|
|
|
* so just return, this page will likely be freed soon anyway.
|
2010-10-27 07:57:29 +07:00
|
|
|
*/
|
2012-03-10 04:01:49 +07:00
|
|
|
if (!test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags)) {
|
|
|
|
spin_unlock(&eb->refs_lock);
|
|
|
|
return 0;
|
2009-03-13 22:00:37 +07:00
|
|
|
}
|
2010-10-27 07:57:29 +07:00
|
|
|
|
2013-04-26 21:56:29 +07:00
|
|
|
return release_extent_buffer(eb);
|
2008-07-22 22:18:07 +07:00
|
|
|
}
|