4297ff84dc
When diagnosing a slowdown of generic/224 I noticed we were not doing anything when calling into shrink_delalloc(). This is because all writes in 224 are O_DIRECT, not delalloc, and thus our delalloc_bytes counter is 0, which short circuits most of the work inside of shrink_delalloc(). However O_DIRECT writes still consume metadata resources and generate ordered extents, which we can still wait on. Fix this by tracking outstanding DIO write bytes, and use this as well as the delalloc bytes counter to decide if we need to lookup and wait on any ordered extents. If we have more DIO writes than delalloc bytes we'll go ahead and wait on any ordered extents regardless of our flush state as flushing delalloc is likely to not gain us anything. Signed-off-by: Josef Bacik <josef@toxicpanda.com> [ use dio instead of odirect in identifiers ] Signed-off-by: David Sterba <dsterba@suse.com>
981 lines
27 KiB
C
981 lines
27 KiB
C
// SPDX-License-Identifier: GPL-2.0
|
|
/*
|
|
* Copyright (C) 2007 Oracle. All rights reserved.
|
|
*/
|
|
|
|
#include <linux/slab.h>
|
|
#include <linux/blkdev.h>
|
|
#include <linux/writeback.h>
|
|
#include <linux/sched/mm.h>
|
|
#include "ctree.h"
|
|
#include "transaction.h"
|
|
#include "btrfs_inode.h"
|
|
#include "extent_io.h"
|
|
#include "disk-io.h"
|
|
#include "compression.h"
|
|
|
|
static struct kmem_cache *btrfs_ordered_extent_cache;
|
|
|
|
static u64 entry_end(struct btrfs_ordered_extent *entry)
|
|
{
|
|
if (entry->file_offset + entry->len < entry->file_offset)
|
|
return (u64)-1;
|
|
return entry->file_offset + entry->len;
|
|
}
|
|
|
|
/* returns NULL if the insertion worked, or it returns the node it did find
|
|
* in the tree
|
|
*/
|
|
static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset,
|
|
struct rb_node *node)
|
|
{
|
|
struct rb_node **p = &root->rb_node;
|
|
struct rb_node *parent = NULL;
|
|
struct btrfs_ordered_extent *entry;
|
|
|
|
while (*p) {
|
|
parent = *p;
|
|
entry = rb_entry(parent, struct btrfs_ordered_extent, rb_node);
|
|
|
|
if (file_offset < entry->file_offset)
|
|
p = &(*p)->rb_left;
|
|
else if (file_offset >= entry_end(entry))
|
|
p = &(*p)->rb_right;
|
|
else
|
|
return parent;
|
|
}
|
|
|
|
rb_link_node(node, parent, p);
|
|
rb_insert_color(node, root);
|
|
return NULL;
|
|
}
|
|
|
|
static void ordered_data_tree_panic(struct inode *inode, int errno,
|
|
u64 offset)
|
|
{
|
|
struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
|
|
btrfs_panic(fs_info, errno,
|
|
"Inconsistency in ordered tree at offset %llu", offset);
|
|
}
|
|
|
|
/*
|
|
* look for a given offset in the tree, and if it can't be found return the
|
|
* first lesser offset
|
|
*/
|
|
static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset,
|
|
struct rb_node **prev_ret)
|
|
{
|
|
struct rb_node *n = root->rb_node;
|
|
struct rb_node *prev = NULL;
|
|
struct rb_node *test;
|
|
struct btrfs_ordered_extent *entry;
|
|
struct btrfs_ordered_extent *prev_entry = NULL;
|
|
|
|
while (n) {
|
|
entry = rb_entry(n, struct btrfs_ordered_extent, rb_node);
|
|
prev = n;
|
|
prev_entry = entry;
|
|
|
|
if (file_offset < entry->file_offset)
|
|
n = n->rb_left;
|
|
else if (file_offset >= entry_end(entry))
|
|
n = n->rb_right;
|
|
else
|
|
return n;
|
|
}
|
|
if (!prev_ret)
|
|
return NULL;
|
|
|
|
while (prev && file_offset >= entry_end(prev_entry)) {
|
|
test = rb_next(prev);
|
|
if (!test)
|
|
break;
|
|
prev_entry = rb_entry(test, struct btrfs_ordered_extent,
|
|
rb_node);
|
|
if (file_offset < entry_end(prev_entry))
|
|
break;
|
|
|
|
prev = test;
|
|
}
|
|
if (prev)
|
|
prev_entry = rb_entry(prev, struct btrfs_ordered_extent,
|
|
rb_node);
|
|
while (prev && file_offset < entry_end(prev_entry)) {
|
|
test = rb_prev(prev);
|
|
if (!test)
|
|
break;
|
|
prev_entry = rb_entry(test, struct btrfs_ordered_extent,
|
|
rb_node);
|
|
prev = test;
|
|
}
|
|
*prev_ret = prev;
|
|
return NULL;
|
|
}
|
|
|
|
/*
|
|
* helper to check if a given offset is inside a given entry
|
|
*/
|
|
static int offset_in_entry(struct btrfs_ordered_extent *entry, u64 file_offset)
|
|
{
|
|
if (file_offset < entry->file_offset ||
|
|
entry->file_offset + entry->len <= file_offset)
|
|
return 0;
|
|
return 1;
|
|
}
|
|
|
|
static int range_overlaps(struct btrfs_ordered_extent *entry, u64 file_offset,
|
|
u64 len)
|
|
{
|
|
if (file_offset + len <= entry->file_offset ||
|
|
entry->file_offset + entry->len <= file_offset)
|
|
return 0;
|
|
return 1;
|
|
}
|
|
|
|
/*
|
|
* look find the first ordered struct that has this offset, otherwise
|
|
* the first one less than this offset
|
|
*/
|
|
static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree,
|
|
u64 file_offset)
|
|
{
|
|
struct rb_root *root = &tree->tree;
|
|
struct rb_node *prev = NULL;
|
|
struct rb_node *ret;
|
|
struct btrfs_ordered_extent *entry;
|
|
|
|
if (tree->last) {
|
|
entry = rb_entry(tree->last, struct btrfs_ordered_extent,
|
|
rb_node);
|
|
if (offset_in_entry(entry, file_offset))
|
|
return tree->last;
|
|
}
|
|
ret = __tree_search(root, file_offset, &prev);
|
|
if (!ret)
|
|
ret = prev;
|
|
if (ret)
|
|
tree->last = ret;
|
|
return ret;
|
|
}
|
|
|
|
/* allocate and add a new ordered_extent into the per-inode tree.
|
|
* file_offset is the logical offset in the file
|
|
*
|
|
* start is the disk block number of an extent already reserved in the
|
|
* extent allocation tree
|
|
*
|
|
* len is the length of the extent
|
|
*
|
|
* The tree is given a single reference on the ordered extent that was
|
|
* inserted.
|
|
*/
|
|
static int __btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
|
|
u64 start, u64 len, u64 disk_len,
|
|
int type, int dio, int compress_type)
|
|
{
|
|
struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
|
|
struct btrfs_root *root = BTRFS_I(inode)->root;
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
struct rb_node *node;
|
|
struct btrfs_ordered_extent *entry;
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
|
entry = kmem_cache_zalloc(btrfs_ordered_extent_cache, GFP_NOFS);
|
|
if (!entry)
|
|
return -ENOMEM;
|
|
|
|
entry->file_offset = file_offset;
|
|
entry->start = start;
|
|
entry->len = len;
|
|
entry->disk_len = disk_len;
|
|
entry->bytes_left = len;
|
|
entry->inode = igrab(inode);
|
|
entry->compress_type = compress_type;
|
|
entry->truncated_len = (u64)-1;
|
|
if (type != BTRFS_ORDERED_IO_DONE && type != BTRFS_ORDERED_COMPLETE)
|
|
set_bit(type, &entry->flags);
|
|
|
|
if (dio) {
|
|
percpu_counter_add_batch(&fs_info->dio_bytes, len,
|
|
fs_info->delalloc_batch);
|
|
set_bit(BTRFS_ORDERED_DIRECT, &entry->flags);
|
|
}
|
|
|
|
/* one ref for the tree */
|
|
refcount_set(&entry->refs, 1);
|
|
init_waitqueue_head(&entry->wait);
|
|
INIT_LIST_HEAD(&entry->list);
|
|
INIT_LIST_HEAD(&entry->root_extent_list);
|
|
INIT_LIST_HEAD(&entry->work_list);
|
|
init_completion(&entry->completion);
|
|
INIT_LIST_HEAD(&entry->log_list);
|
|
INIT_LIST_HEAD(&entry->trans_list);
|
|
|
|
trace_btrfs_ordered_extent_add(inode, entry);
|
|
|
|
spin_lock_irq(&tree->lock);
|
|
node = tree_insert(&tree->tree, file_offset,
|
|
&entry->rb_node);
|
|
if (node)
|
|
ordered_data_tree_panic(inode, -EEXIST, file_offset);
|
|
spin_unlock_irq(&tree->lock);
|
|
|
|
spin_lock(&root->ordered_extent_lock);
|
|
list_add_tail(&entry->root_extent_list,
|
|
&root->ordered_extents);
|
|
root->nr_ordered_extents++;
|
|
if (root->nr_ordered_extents == 1) {
|
|
spin_lock(&fs_info->ordered_root_lock);
|
|
BUG_ON(!list_empty(&root->ordered_root));
|
|
list_add_tail(&root->ordered_root, &fs_info->ordered_roots);
|
|
spin_unlock(&fs_info->ordered_root_lock);
|
|
}
|
|
spin_unlock(&root->ordered_extent_lock);
|
|
|
|
/*
|
|
* We don't need the count_max_extents here, we can assume that all of
|
|
* that work has been done at higher layers, so this is truly the
|
|
* smallest the extent is going to get.
|
|
*/
|
|
spin_lock(&BTRFS_I(inode)->lock);
|
|
btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
|
|
spin_unlock(&BTRFS_I(inode)->lock);
|
|
|
|
return 0;
|
|
}
|
|
|
|
int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
|
|
u64 start, u64 len, u64 disk_len, int type)
|
|
{
|
|
return __btrfs_add_ordered_extent(inode, file_offset, start, len,
|
|
disk_len, type, 0,
|
|
BTRFS_COMPRESS_NONE);
|
|
}
|
|
|
|
int btrfs_add_ordered_extent_dio(struct inode *inode, u64 file_offset,
|
|
u64 start, u64 len, u64 disk_len, int type)
|
|
{
|
|
return __btrfs_add_ordered_extent(inode, file_offset, start, len,
|
|
disk_len, type, 1,
|
|
BTRFS_COMPRESS_NONE);
|
|
}
|
|
|
|
int btrfs_add_ordered_extent_compress(struct inode *inode, u64 file_offset,
|
|
u64 start, u64 len, u64 disk_len,
|
|
int type, int compress_type)
|
|
{
|
|
return __btrfs_add_ordered_extent(inode, file_offset, start, len,
|
|
disk_len, type, 0,
|
|
compress_type);
|
|
}
|
|
|
|
/*
|
|
* Add a struct btrfs_ordered_sum into the list of checksums to be inserted
|
|
* when an ordered extent is finished. If the list covers more than one
|
|
* ordered extent, it is split across multiples.
|
|
*/
|
|
void btrfs_add_ordered_sum(struct btrfs_ordered_extent *entry,
|
|
struct btrfs_ordered_sum *sum)
|
|
{
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
|
|
tree = &BTRFS_I(entry->inode)->ordered_tree;
|
|
spin_lock_irq(&tree->lock);
|
|
list_add_tail(&sum->list, &entry->list);
|
|
spin_unlock_irq(&tree->lock);
|
|
}
|
|
|
|
/*
|
|
* this is used to account for finished IO across a given range
|
|
* of the file. The IO may span ordered extents. If
|
|
* a given ordered_extent is completely done, 1 is returned, otherwise
|
|
* 0.
|
|
*
|
|
* test_and_set_bit on a flag in the struct btrfs_ordered_extent is used
|
|
* to make sure this function only returns 1 once for a given ordered extent.
|
|
*
|
|
* file_offset is updated to one byte past the range that is recorded as
|
|
* complete. This allows you to walk forward in the file.
|
|
*/
|
|
int btrfs_dec_test_first_ordered_pending(struct inode *inode,
|
|
struct btrfs_ordered_extent **cached,
|
|
u64 *file_offset, u64 io_size, int uptodate)
|
|
{
|
|
struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
struct rb_node *node;
|
|
struct btrfs_ordered_extent *entry = NULL;
|
|
int ret;
|
|
unsigned long flags;
|
|
u64 dec_end;
|
|
u64 dec_start;
|
|
u64 to_dec;
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
|
spin_lock_irqsave(&tree->lock, flags);
|
|
node = tree_search(tree, *file_offset);
|
|
if (!node) {
|
|
ret = 1;
|
|
goto out;
|
|
}
|
|
|
|
entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
|
if (!offset_in_entry(entry, *file_offset)) {
|
|
ret = 1;
|
|
goto out;
|
|
}
|
|
|
|
dec_start = max(*file_offset, entry->file_offset);
|
|
dec_end = min(*file_offset + io_size, entry->file_offset +
|
|
entry->len);
|
|
*file_offset = dec_end;
|
|
if (dec_start > dec_end) {
|
|
btrfs_crit(fs_info, "bad ordering dec_start %llu end %llu",
|
|
dec_start, dec_end);
|
|
}
|
|
to_dec = dec_end - dec_start;
|
|
if (to_dec > entry->bytes_left) {
|
|
btrfs_crit(fs_info,
|
|
"bad ordered accounting left %llu size %llu",
|
|
entry->bytes_left, to_dec);
|
|
}
|
|
entry->bytes_left -= to_dec;
|
|
if (!uptodate)
|
|
set_bit(BTRFS_ORDERED_IOERR, &entry->flags);
|
|
|
|
if (entry->bytes_left == 0) {
|
|
ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
|
|
/* test_and_set_bit implies a barrier */
|
|
cond_wake_up_nomb(&entry->wait);
|
|
} else {
|
|
ret = 1;
|
|
}
|
|
out:
|
|
if (!ret && cached && entry) {
|
|
*cached = entry;
|
|
refcount_inc(&entry->refs);
|
|
}
|
|
spin_unlock_irqrestore(&tree->lock, flags);
|
|
return ret == 0;
|
|
}
|
|
|
|
/*
|
|
* this is used to account for finished IO across a given range
|
|
* of the file. The IO should not span ordered extents. If
|
|
* a given ordered_extent is completely done, 1 is returned, otherwise
|
|
* 0.
|
|
*
|
|
* test_and_set_bit on a flag in the struct btrfs_ordered_extent is used
|
|
* to make sure this function only returns 1 once for a given ordered extent.
|
|
*/
|
|
int btrfs_dec_test_ordered_pending(struct inode *inode,
|
|
struct btrfs_ordered_extent **cached,
|
|
u64 file_offset, u64 io_size, int uptodate)
|
|
{
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
struct rb_node *node;
|
|
struct btrfs_ordered_extent *entry = NULL;
|
|
unsigned long flags;
|
|
int ret;
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
|
spin_lock_irqsave(&tree->lock, flags);
|
|
if (cached && *cached) {
|
|
entry = *cached;
|
|
goto have_entry;
|
|
}
|
|
|
|
node = tree_search(tree, file_offset);
|
|
if (!node) {
|
|
ret = 1;
|
|
goto out;
|
|
}
|
|
|
|
entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
|
have_entry:
|
|
if (!offset_in_entry(entry, file_offset)) {
|
|
ret = 1;
|
|
goto out;
|
|
}
|
|
|
|
if (io_size > entry->bytes_left) {
|
|
btrfs_crit(BTRFS_I(inode)->root->fs_info,
|
|
"bad ordered accounting left %llu size %llu",
|
|
entry->bytes_left, io_size);
|
|
}
|
|
entry->bytes_left -= io_size;
|
|
if (!uptodate)
|
|
set_bit(BTRFS_ORDERED_IOERR, &entry->flags);
|
|
|
|
if (entry->bytes_left == 0) {
|
|
ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
|
|
/* test_and_set_bit implies a barrier */
|
|
cond_wake_up_nomb(&entry->wait);
|
|
} else {
|
|
ret = 1;
|
|
}
|
|
out:
|
|
if (!ret && cached && entry) {
|
|
*cached = entry;
|
|
refcount_inc(&entry->refs);
|
|
}
|
|
spin_unlock_irqrestore(&tree->lock, flags);
|
|
return ret == 0;
|
|
}
|
|
|
|
/*
|
|
* used to drop a reference on an ordered extent. This will free
|
|
* the extent if the last reference is dropped
|
|
*/
|
|
void btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry)
|
|
{
|
|
struct list_head *cur;
|
|
struct btrfs_ordered_sum *sum;
|
|
|
|
trace_btrfs_ordered_extent_put(entry->inode, entry);
|
|
|
|
if (refcount_dec_and_test(&entry->refs)) {
|
|
ASSERT(list_empty(&entry->log_list));
|
|
ASSERT(list_empty(&entry->trans_list));
|
|
ASSERT(list_empty(&entry->root_extent_list));
|
|
ASSERT(RB_EMPTY_NODE(&entry->rb_node));
|
|
if (entry->inode)
|
|
btrfs_add_delayed_iput(entry->inode);
|
|
while (!list_empty(&entry->list)) {
|
|
cur = entry->list.next;
|
|
sum = list_entry(cur, struct btrfs_ordered_sum, list);
|
|
list_del(&sum->list);
|
|
kvfree(sum);
|
|
}
|
|
kmem_cache_free(btrfs_ordered_extent_cache, entry);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* remove an ordered extent from the tree. No references are dropped
|
|
* and waiters are woken up.
|
|
*/
|
|
void btrfs_remove_ordered_extent(struct inode *inode,
|
|
struct btrfs_ordered_extent *entry)
|
|
{
|
|
struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
|
|
struct btrfs_root *root = btrfs_inode->root;
|
|
struct rb_node *node;
|
|
|
|
/* This is paired with btrfs_add_ordered_extent. */
|
|
spin_lock(&btrfs_inode->lock);
|
|
btrfs_mod_outstanding_extents(btrfs_inode, -1);
|
|
spin_unlock(&btrfs_inode->lock);
|
|
if (root != fs_info->tree_root)
|
|
btrfs_delalloc_release_metadata(btrfs_inode, entry->len, false);
|
|
|
|
if (test_bit(BTRFS_ORDERED_DIRECT, &entry->flags))
|
|
percpu_counter_add_batch(&fs_info->dio_bytes, -entry->len,
|
|
fs_info->delalloc_batch);
|
|
|
|
tree = &btrfs_inode->ordered_tree;
|
|
spin_lock_irq(&tree->lock);
|
|
node = &entry->rb_node;
|
|
rb_erase(node, &tree->tree);
|
|
RB_CLEAR_NODE(node);
|
|
if (tree->last == node)
|
|
tree->last = NULL;
|
|
set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags);
|
|
spin_unlock_irq(&tree->lock);
|
|
|
|
spin_lock(&root->ordered_extent_lock);
|
|
list_del_init(&entry->root_extent_list);
|
|
root->nr_ordered_extents--;
|
|
|
|
trace_btrfs_ordered_extent_remove(inode, entry);
|
|
|
|
if (!root->nr_ordered_extents) {
|
|
spin_lock(&fs_info->ordered_root_lock);
|
|
BUG_ON(list_empty(&root->ordered_root));
|
|
list_del_init(&root->ordered_root);
|
|
spin_unlock(&fs_info->ordered_root_lock);
|
|
}
|
|
spin_unlock(&root->ordered_extent_lock);
|
|
wake_up(&entry->wait);
|
|
}
|
|
|
|
static void btrfs_run_ordered_extent_work(struct btrfs_work *work)
|
|
{
|
|
struct btrfs_ordered_extent *ordered;
|
|
|
|
ordered = container_of(work, struct btrfs_ordered_extent, flush_work);
|
|
btrfs_start_ordered_extent(ordered->inode, ordered, 1);
|
|
complete(&ordered->completion);
|
|
}
|
|
|
|
/*
|
|
* wait for all the ordered extents in a root. This is done when balancing
|
|
* space between drives.
|
|
*/
|
|
u64 btrfs_wait_ordered_extents(struct btrfs_root *root, u64 nr,
|
|
const u64 range_start, const u64 range_len)
|
|
{
|
|
struct btrfs_fs_info *fs_info = root->fs_info;
|
|
LIST_HEAD(splice);
|
|
LIST_HEAD(skipped);
|
|
LIST_HEAD(works);
|
|
struct btrfs_ordered_extent *ordered, *next;
|
|
u64 count = 0;
|
|
const u64 range_end = range_start + range_len;
|
|
|
|
mutex_lock(&root->ordered_extent_mutex);
|
|
spin_lock(&root->ordered_extent_lock);
|
|
list_splice_init(&root->ordered_extents, &splice);
|
|
while (!list_empty(&splice) && nr) {
|
|
ordered = list_first_entry(&splice, struct btrfs_ordered_extent,
|
|
root_extent_list);
|
|
|
|
if (range_end <= ordered->start ||
|
|
ordered->start + ordered->disk_len <= range_start) {
|
|
list_move_tail(&ordered->root_extent_list, &skipped);
|
|
cond_resched_lock(&root->ordered_extent_lock);
|
|
continue;
|
|
}
|
|
|
|
list_move_tail(&ordered->root_extent_list,
|
|
&root->ordered_extents);
|
|
refcount_inc(&ordered->refs);
|
|
spin_unlock(&root->ordered_extent_lock);
|
|
|
|
btrfs_init_work(&ordered->flush_work,
|
|
btrfs_flush_delalloc_helper,
|
|
btrfs_run_ordered_extent_work, NULL, NULL);
|
|
list_add_tail(&ordered->work_list, &works);
|
|
btrfs_queue_work(fs_info->flush_workers, &ordered->flush_work);
|
|
|
|
cond_resched();
|
|
spin_lock(&root->ordered_extent_lock);
|
|
if (nr != U64_MAX)
|
|
nr--;
|
|
count++;
|
|
}
|
|
list_splice_tail(&skipped, &root->ordered_extents);
|
|
list_splice_tail(&splice, &root->ordered_extents);
|
|
spin_unlock(&root->ordered_extent_lock);
|
|
|
|
list_for_each_entry_safe(ordered, next, &works, work_list) {
|
|
list_del_init(&ordered->work_list);
|
|
wait_for_completion(&ordered->completion);
|
|
btrfs_put_ordered_extent(ordered);
|
|
cond_resched();
|
|
}
|
|
mutex_unlock(&root->ordered_extent_mutex);
|
|
|
|
return count;
|
|
}
|
|
|
|
u64 btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, u64 nr,
|
|
const u64 range_start, const u64 range_len)
|
|
{
|
|
struct btrfs_root *root;
|
|
struct list_head splice;
|
|
u64 total_done = 0;
|
|
u64 done;
|
|
|
|
INIT_LIST_HEAD(&splice);
|
|
|
|
mutex_lock(&fs_info->ordered_operations_mutex);
|
|
spin_lock(&fs_info->ordered_root_lock);
|
|
list_splice_init(&fs_info->ordered_roots, &splice);
|
|
while (!list_empty(&splice) && nr) {
|
|
root = list_first_entry(&splice, struct btrfs_root,
|
|
ordered_root);
|
|
root = btrfs_grab_fs_root(root);
|
|
BUG_ON(!root);
|
|
list_move_tail(&root->ordered_root,
|
|
&fs_info->ordered_roots);
|
|
spin_unlock(&fs_info->ordered_root_lock);
|
|
|
|
done = btrfs_wait_ordered_extents(root, nr,
|
|
range_start, range_len);
|
|
btrfs_put_fs_root(root);
|
|
total_done += done;
|
|
|
|
spin_lock(&fs_info->ordered_root_lock);
|
|
if (nr != U64_MAX) {
|
|
nr -= done;
|
|
}
|
|
}
|
|
list_splice_tail(&splice, &fs_info->ordered_roots);
|
|
spin_unlock(&fs_info->ordered_root_lock);
|
|
mutex_unlock(&fs_info->ordered_operations_mutex);
|
|
|
|
return total_done;
|
|
}
|
|
|
|
/*
|
|
* Used to start IO or wait for a given ordered extent to finish.
|
|
*
|
|
* If wait is one, this effectively waits on page writeback for all the pages
|
|
* in the extent, and it waits on the io completion code to insert
|
|
* metadata into the btree corresponding to the extent
|
|
*/
|
|
void btrfs_start_ordered_extent(struct inode *inode,
|
|
struct btrfs_ordered_extent *entry,
|
|
int wait)
|
|
{
|
|
u64 start = entry->file_offset;
|
|
u64 end = start + entry->len - 1;
|
|
|
|
trace_btrfs_ordered_extent_start(inode, entry);
|
|
|
|
/*
|
|
* pages in the range can be dirty, clean or writeback. We
|
|
* start IO on any dirty ones so the wait doesn't stall waiting
|
|
* for the flusher thread to find them
|
|
*/
|
|
if (!test_bit(BTRFS_ORDERED_DIRECT, &entry->flags))
|
|
filemap_fdatawrite_range(inode->i_mapping, start, end);
|
|
if (wait) {
|
|
wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE,
|
|
&entry->flags));
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Used to wait on ordered extents across a large range of bytes.
|
|
*/
|
|
int btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len)
|
|
{
|
|
int ret = 0;
|
|
int ret_wb = 0;
|
|
u64 end;
|
|
u64 orig_end;
|
|
struct btrfs_ordered_extent *ordered;
|
|
|
|
if (start + len < start) {
|
|
orig_end = INT_LIMIT(loff_t);
|
|
} else {
|
|
orig_end = start + len - 1;
|
|
if (orig_end > INT_LIMIT(loff_t))
|
|
orig_end = INT_LIMIT(loff_t);
|
|
}
|
|
|
|
/* start IO across the range first to instantiate any delalloc
|
|
* extents
|
|
*/
|
|
ret = btrfs_fdatawrite_range(inode, start, orig_end);
|
|
if (ret)
|
|
return ret;
|
|
|
|
/*
|
|
* If we have a writeback error don't return immediately. Wait first
|
|
* for any ordered extents that haven't completed yet. This is to make
|
|
* sure no one can dirty the same page ranges and call writepages()
|
|
* before the ordered extents complete - to avoid failures (-EEXIST)
|
|
* when adding the new ordered extents to the ordered tree.
|
|
*/
|
|
ret_wb = filemap_fdatawait_range(inode->i_mapping, start, orig_end);
|
|
|
|
end = orig_end;
|
|
while (1) {
|
|
ordered = btrfs_lookup_first_ordered_extent(inode, end);
|
|
if (!ordered)
|
|
break;
|
|
if (ordered->file_offset > orig_end) {
|
|
btrfs_put_ordered_extent(ordered);
|
|
break;
|
|
}
|
|
if (ordered->file_offset + ordered->len <= start) {
|
|
btrfs_put_ordered_extent(ordered);
|
|
break;
|
|
}
|
|
btrfs_start_ordered_extent(inode, ordered, 1);
|
|
end = ordered->file_offset;
|
|
if (test_bit(BTRFS_ORDERED_IOERR, &ordered->flags))
|
|
ret = -EIO;
|
|
btrfs_put_ordered_extent(ordered);
|
|
if (ret || end == 0 || end == start)
|
|
break;
|
|
end--;
|
|
}
|
|
return ret_wb ? ret_wb : ret;
|
|
}
|
|
|
|
/*
|
|
* find an ordered extent corresponding to file_offset. return NULL if
|
|
* nothing is found, otherwise take a reference on the extent and return it
|
|
*/
|
|
struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode,
|
|
u64 file_offset)
|
|
{
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
struct rb_node *node;
|
|
struct btrfs_ordered_extent *entry = NULL;
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
|
spin_lock_irq(&tree->lock);
|
|
node = tree_search(tree, file_offset);
|
|
if (!node)
|
|
goto out;
|
|
|
|
entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
|
if (!offset_in_entry(entry, file_offset))
|
|
entry = NULL;
|
|
if (entry)
|
|
refcount_inc(&entry->refs);
|
|
out:
|
|
spin_unlock_irq(&tree->lock);
|
|
return entry;
|
|
}
|
|
|
|
/* Since the DIO code tries to lock a wide area we need to look for any ordered
|
|
* extents that exist in the range, rather than just the start of the range.
|
|
*/
|
|
struct btrfs_ordered_extent *btrfs_lookup_ordered_range(
|
|
struct btrfs_inode *inode, u64 file_offset, u64 len)
|
|
{
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
struct rb_node *node;
|
|
struct btrfs_ordered_extent *entry = NULL;
|
|
|
|
tree = &inode->ordered_tree;
|
|
spin_lock_irq(&tree->lock);
|
|
node = tree_search(tree, file_offset);
|
|
if (!node) {
|
|
node = tree_search(tree, file_offset + len);
|
|
if (!node)
|
|
goto out;
|
|
}
|
|
|
|
while (1) {
|
|
entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
|
if (range_overlaps(entry, file_offset, len))
|
|
break;
|
|
|
|
if (entry->file_offset >= file_offset + len) {
|
|
entry = NULL;
|
|
break;
|
|
}
|
|
entry = NULL;
|
|
node = rb_next(node);
|
|
if (!node)
|
|
break;
|
|
}
|
|
out:
|
|
if (entry)
|
|
refcount_inc(&entry->refs);
|
|
spin_unlock_irq(&tree->lock);
|
|
return entry;
|
|
}
|
|
|
|
/*
|
|
* lookup and return any extent before 'file_offset'. NULL is returned
|
|
* if none is found
|
|
*/
|
|
struct btrfs_ordered_extent *
|
|
btrfs_lookup_first_ordered_extent(struct inode *inode, u64 file_offset)
|
|
{
|
|
struct btrfs_ordered_inode_tree *tree;
|
|
struct rb_node *node;
|
|
struct btrfs_ordered_extent *entry = NULL;
|
|
|
|
tree = &BTRFS_I(inode)->ordered_tree;
|
|
spin_lock_irq(&tree->lock);
|
|
node = tree_search(tree, file_offset);
|
|
if (!node)
|
|
goto out;
|
|
|
|
entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
|
refcount_inc(&entry->refs);
|
|
out:
|
|
spin_unlock_irq(&tree->lock);
|
|
return entry;
|
|
}
|
|
|
|
/*
|
|
* After an extent is done, call this to conditionally update the on disk
|
|
* i_size. i_size is updated to cover any fully written part of the file.
|
|
*/
|
|
int btrfs_ordered_update_i_size(struct inode *inode, u64 offset,
|
|
struct btrfs_ordered_extent *ordered)
|
|
{
|
|
struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree;
|
|
u64 disk_i_size;
|
|
u64 new_i_size;
|
|
u64 i_size = i_size_read(inode);
|
|
struct rb_node *node;
|
|
struct rb_node *prev = NULL;
|
|
struct btrfs_ordered_extent *test;
|
|
int ret = 1;
|
|
u64 orig_offset = offset;
|
|
|
|
spin_lock_irq(&tree->lock);
|
|
if (ordered) {
|
|
offset = entry_end(ordered);
|
|
if (test_bit(BTRFS_ORDERED_TRUNCATED, &ordered->flags))
|
|
offset = min(offset,
|
|
ordered->file_offset +
|
|
ordered->truncated_len);
|
|
} else {
|
|
offset = ALIGN(offset, btrfs_inode_sectorsize(inode));
|
|
}
|
|
disk_i_size = BTRFS_I(inode)->disk_i_size;
|
|
|
|
/*
|
|
* truncate file.
|
|
* If ordered is not NULL, then this is called from endio and
|
|
* disk_i_size will be updated by either truncate itself or any
|
|
* in-flight IOs which are inside the disk_i_size.
|
|
*
|
|
* Because btrfs_setsize() may set i_size with disk_i_size if truncate
|
|
* fails somehow, we need to make sure we have a precise disk_i_size by
|
|
* updating it as usual.
|
|
*
|
|
*/
|
|
if (!ordered && disk_i_size > i_size) {
|
|
BTRFS_I(inode)->disk_i_size = orig_offset;
|
|
ret = 0;
|
|
goto out;
|
|
}
|
|
|
|
/*
|
|
* if the disk i_size is already at the inode->i_size, or
|
|
* this ordered extent is inside the disk i_size, we're done
|
|
*/
|
|
if (disk_i_size == i_size)
|
|
goto out;
|
|
|
|
/*
|
|
* We still need to update disk_i_size if outstanding_isize is greater
|
|
* than disk_i_size.
|
|
*/
|
|
if (offset <= disk_i_size &&
|
|
(!ordered || ordered->outstanding_isize <= disk_i_size))
|
|
goto out;
|
|
|
|
/*
|
|
* walk backward from this ordered extent to disk_i_size.
|
|
* if we find an ordered extent then we can't update disk i_size
|
|
* yet
|
|
*/
|
|
if (ordered) {
|
|
node = rb_prev(&ordered->rb_node);
|
|
} else {
|
|
prev = tree_search(tree, offset);
|
|
/*
|
|
* we insert file extents without involving ordered struct,
|
|
* so there should be no ordered struct cover this offset
|
|
*/
|
|
if (prev) {
|
|
test = rb_entry(prev, struct btrfs_ordered_extent,
|
|
rb_node);
|
|
BUG_ON(offset_in_entry(test, offset));
|
|
}
|
|
node = prev;
|
|
}
|
|
for (; node; node = rb_prev(node)) {
|
|
test = rb_entry(node, struct btrfs_ordered_extent, rb_node);
|
|
|
|
/* We treat this entry as if it doesn't exist */
|
|
if (test_bit(BTRFS_ORDERED_UPDATED_ISIZE, &test->flags))
|
|
continue;
|
|
|
|
if (entry_end(test) <= disk_i_size)
|
|
break;
|
|
if (test->file_offset >= i_size)
|
|
break;
|
|
|
|
/*
|
|
* We don't update disk_i_size now, so record this undealt
|
|
* i_size. Or we will not know the real i_size.
|
|
*/
|
|
if (test->outstanding_isize < offset)
|
|
test->outstanding_isize = offset;
|
|
if (ordered &&
|
|
ordered->outstanding_isize > test->outstanding_isize)
|
|
test->outstanding_isize = ordered->outstanding_isize;
|
|
goto out;
|
|
}
|
|
new_i_size = min_t(u64, offset, i_size);
|
|
|
|
/*
|
|
* Some ordered extents may completed before the current one, and
|
|
* we hold the real i_size in ->outstanding_isize.
|
|
*/
|
|
if (ordered && ordered->outstanding_isize > new_i_size)
|
|
new_i_size = min_t(u64, ordered->outstanding_isize, i_size);
|
|
BTRFS_I(inode)->disk_i_size = new_i_size;
|
|
ret = 0;
|
|
out:
|
|
/*
|
|
* We need to do this because we can't remove ordered extents until
|
|
* after the i_disk_size has been updated and then the inode has been
|
|
* updated to reflect the change, so we need to tell anybody who finds
|
|
* this ordered extent that we've already done all the real work, we
|
|
* just haven't completed all the other work.
|
|
*/
|
|
if (ordered)
|
|
set_bit(BTRFS_ORDERED_UPDATED_ISIZE, &ordered->flags);
|
|
spin_unlock_irq(&tree->lock);
|
|
return ret;
|
|
}
|
|
|
|
/*
|
|
* search the ordered extents for one corresponding to 'offset' and
|
|
* try to find a checksum. This is used because we allow pages to
|
|
* be reclaimed before their checksum is actually put into the btree
|
|
*/
|
|
int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr,
|
|
u32 *sum, int len)
|
|
{
|
|
struct btrfs_ordered_sum *ordered_sum;
|
|
struct btrfs_ordered_extent *ordered;
|
|
struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree;
|
|
unsigned long num_sectors;
|
|
unsigned long i;
|
|
u32 sectorsize = btrfs_inode_sectorsize(inode);
|
|
int index = 0;
|
|
|
|
ordered = btrfs_lookup_ordered_extent(inode, offset);
|
|
if (!ordered)
|
|
return 0;
|
|
|
|
spin_lock_irq(&tree->lock);
|
|
list_for_each_entry_reverse(ordered_sum, &ordered->list, list) {
|
|
if (disk_bytenr >= ordered_sum->bytenr &&
|
|
disk_bytenr < ordered_sum->bytenr + ordered_sum->len) {
|
|
i = (disk_bytenr - ordered_sum->bytenr) >>
|
|
inode->i_sb->s_blocksize_bits;
|
|
num_sectors = ordered_sum->len >>
|
|
inode->i_sb->s_blocksize_bits;
|
|
num_sectors = min_t(int, len - index, num_sectors - i);
|
|
memcpy(sum + index, ordered_sum->sums + i,
|
|
num_sectors);
|
|
|
|
index += (int)num_sectors;
|
|
if (index == len)
|
|
goto out;
|
|
disk_bytenr += num_sectors * sectorsize;
|
|
}
|
|
}
|
|
out:
|
|
spin_unlock_irq(&tree->lock);
|
|
btrfs_put_ordered_extent(ordered);
|
|
return index;
|
|
}
|
|
|
|
int __init ordered_data_init(void)
|
|
{
|
|
btrfs_ordered_extent_cache = kmem_cache_create("btrfs_ordered_extent",
|
|
sizeof(struct btrfs_ordered_extent), 0,
|
|
SLAB_MEM_SPREAD,
|
|
NULL);
|
|
if (!btrfs_ordered_extent_cache)
|
|
return -ENOMEM;
|
|
|
|
return 0;
|
|
}
|
|
|
|
void __cold ordered_data_exit(void)
|
|
{
|
|
kmem_cache_destroy(btrfs_ordered_extent_cache);
|
|
}
|