diff options
Diffstat (limited to 'fs/btrfs/ordered-data.c')
| -rw-r--r-- | fs/btrfs/ordered-data.c | 1546 |
1 files changed, 891 insertions, 655 deletions
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c index 81369827e514..5df02c707aee 100644 --- a/fs/btrfs/ordered-data.c +++ b/fs/btrfs/ordered-data.c @@ -1,38 +1,33 @@ +// SPDX-License-Identifier: GPL-2.0 /* * Copyright (C) 2007 Oracle. All rights reserved. - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public - * License v2 as published by the Free Software Foundation. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * General Public License for more details. - * - * You should have received a copy of the GNU General Public - * License along with this program; if not, write to the - * Free Software Foundation, Inc., 59 Temple Place - Suite 330, - * Boston, MA 021110-1307, USA. */ #include <linux/slab.h> #include <linux/blkdev.h> #include <linux/writeback.h> -#include <linux/pagevec.h> +#include <linux/sched/mm.h> +#include "messages.h" +#include "misc.h" #include "ctree.h" #include "transaction.h" #include "btrfs_inode.h" #include "extent_io.h" #include "disk-io.h" +#include "compression.h" +#include "delalloc-space.h" +#include "qgroup.h" +#include "subpage.h" +#include "file.h" +#include "block-group.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) + if (entry->file_offset + entry->num_bytes < entry->file_offset) return (u64)-1; - return entry->file_offset + entry->len; + return entry->file_offset + entry->num_bytes; } /* returns NULL if the insertion worked, or it returns the node it did find @@ -62,14 +57,6 @@ static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset, 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\n", (unsigned long long)offset); -} - /* * look for a given offset in the tree, and if it can't be found return the * first lesser offset @@ -124,22 +111,11 @@ static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset, 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) +static int btrfs_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) + entry->file_offset + entry->num_bytes <= file_offset) return 0; return 1; } @@ -148,126 +124,198 @@ static int range_overlaps(struct btrfs_ordered_extent *entry, u64 file_offset, * 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) +static inline struct rb_node *ordered_tree_search(struct btrfs_inode *inode, + 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, + if (inode->ordered_tree_last) { + entry = rb_entry(inode->ordered_tree_last, struct btrfs_ordered_extent, rb_node); - if (offset_in_entry(entry, file_offset)) - return tree->last; + if (in_range(file_offset, entry->file_offset, entry->num_bytes)) + return inode->ordered_tree_last; } - ret = __tree_search(root, file_offset, &prev); + ret = __tree_search(&inode->ordered_tree, file_offset, &prev); if (!ret) ret = prev; if (ret) - tree->last = ret; + inode->ordered_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) +static struct btrfs_ordered_extent *alloc_ordered_extent( + struct btrfs_inode *inode, u64 file_offset, u64 num_bytes, + u64 ram_bytes, u64 disk_bytenr, u64 disk_num_bytes, + u64 offset, unsigned long flags, int compress_type) { - struct btrfs_root *root = BTRFS_I(inode)->root; - struct btrfs_ordered_inode_tree *tree; - struct rb_node *node; struct btrfs_ordered_extent *entry; + int ret; + u64 qgroup_rsv = 0; + const bool is_nocow = (flags & + ((1U << BTRFS_ORDERED_NOCOW) | (1U << BTRFS_ORDERED_PREALLOC))); + + /* + * For a NOCOW write we can free the qgroup reserve right now. For a COW + * one we transfer the reserved space from the inode's iotree into the + * ordered extent by calling btrfs_qgroup_release_data() and tracking + * the qgroup reserved amount in the ordered extent, so that later after + * completing the ordered extent, when running the data delayed ref it + * creates, we free the reserved data with btrfs_qgroup_free_refroot(). + */ + if (is_nocow) + ret = btrfs_qgroup_free_data(inode, NULL, file_offset, num_bytes, &qgroup_rsv); + else + ret = btrfs_qgroup_release_data(inode, file_offset, num_bytes, &qgroup_rsv); + + if (ret < 0) + return ERR_PTR(ret); - tree = &BTRFS_I(inode)->ordered_tree; entry = kmem_cache_zalloc(btrfs_ordered_extent_cache, GFP_NOFS); - if (!entry) - return -ENOMEM; + if (!entry) { + entry = ERR_PTR(-ENOMEM); + goto out; + } entry->file_offset = file_offset; - entry->start = start; - entry->len = len; - if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM) && - !(type == BTRFS_ORDERED_NOCOW)) - entry->csum_bytes_left = disk_len; - entry->disk_len = disk_len; - entry->bytes_left = len; - entry->inode = igrab(inode); + entry->num_bytes = num_bytes; + entry->ram_bytes = ram_bytes; + entry->disk_bytenr = disk_bytenr; + entry->disk_num_bytes = disk_num_bytes; + entry->offset = offset; + entry->bytes_left = num_bytes; + if (WARN_ON_ONCE(!igrab(&inode->vfs_inode))) { + kmem_cache_free(btrfs_ordered_extent_cache, entry); + entry = ERR_PTR(-ESTALE); + goto out; + } + entry->inode = inode; entry->compress_type = compress_type; - if (type != BTRFS_ORDERED_IO_DONE && type != BTRFS_ORDERED_COMPLETE) - set_bit(type, &entry->flags); - - if (dio) - set_bit(BTRFS_ORDERED_DIRECT, &entry->flags); - - /* one ref for the tree */ - atomic_set(&entry->refs, 1); + entry->truncated_len = (u64)-1; + entry->qgroup_rsv = qgroup_rsv; + entry->flags = flags; + refcount_set(&entry->refs, 1); init_waitqueue_head(&entry->wait); INIT_LIST_HEAD(&entry->list); + INIT_LIST_HEAD(&entry->log_list); INIT_LIST_HEAD(&entry->root_extent_list); INIT_LIST_HEAD(&entry->work_list); + INIT_LIST_HEAD(&entry->bioc_list); init_completion(&entry->completion); - INIT_LIST_HEAD(&entry->log_list); + + /* + * 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(&inode->lock); + btrfs_mod_outstanding_extents(inode, 1); + spin_unlock(&inode->lock); + +out: + if (IS_ERR(entry) && !is_nocow) + btrfs_qgroup_free_refroot(inode->root->fs_info, + btrfs_root_id(inode->root), + qgroup_rsv, BTRFS_QGROUP_RSV_DATA); + + return entry; +} + +static void insert_ordered_extent(struct btrfs_ordered_extent *entry) +{ + struct btrfs_inode *inode = entry->inode; + struct btrfs_root *root = inode->root; + struct btrfs_fs_info *fs_info = root->fs_info; + struct rb_node *node; trace_btrfs_ordered_extent_add(inode, entry); - spin_lock_irq(&tree->lock); - node = tree_insert(&tree->tree, file_offset, + percpu_counter_add_batch(&fs_info->ordered_bytes, entry->num_bytes, + fs_info->delalloc_batch); + + /* One ref for the tree. */ + refcount_inc(&entry->refs); + + spin_lock(&inode->ordered_tree_lock); + node = tree_insert(&inode->ordered_tree, entry->file_offset, &entry->rb_node); - if (node) - ordered_data_tree_panic(inode, -EEXIST, file_offset); - spin_unlock_irq(&tree->lock); + if (unlikely(node)) + btrfs_panic(fs_info, -EEXIST, + "inconsistency in ordered tree at offset %llu", + entry->file_offset); + spin_unlock(&inode->ordered_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(&root->fs_info->ordered_root_lock); + spin_lock(&fs_info->ordered_root_lock); BUG_ON(!list_empty(&root->ordered_root)); - list_add_tail(&root->ordered_root, - &root->fs_info->ordered_roots); - spin_unlock(&root->fs_info->ordered_root_lock); + list_add_tail(&root->ordered_root, &fs_info->ordered_roots); + spin_unlock(&fs_info->ordered_root_lock); } spin_unlock(&root->ordered_extent_lock); - - return 0; } -int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, - u64 start, u64 len, u64 disk_len, int type) +/* + * Add an ordered extent to the per-inode tree. + * + * @inode: Inode that this extent is for. + * @file_offset: Logical offset in file where the extent starts. + * @num_bytes: Logical length of extent in file. + * @ram_bytes: Full length of unencoded data. + * @disk_bytenr: Offset of extent on disk. + * @disk_num_bytes: Size of extent on disk. + * @offset: Offset into unencoded data where file data starts. + * @flags: Flags specifying type of extent (1U << BTRFS_ORDERED_*). + * @compress_type: Compression algorithm used for data. + * + * Most of these parameters correspond to &struct btrfs_file_extent_item. The + * tree is given a single reference on the ordered extent that was inserted, and + * the returned pointer is given a second reference. + * + * Return: the new ordered extent or error pointer. + */ +struct btrfs_ordered_extent *btrfs_alloc_ordered_extent( + struct btrfs_inode *inode, u64 file_offset, + const struct btrfs_file_extent *file_extent, unsigned long flags) { - return __btrfs_add_ordered_extent(inode, file_offset, start, len, - disk_len, type, 0, - BTRFS_COMPRESS_NONE); -} + struct btrfs_ordered_extent *entry; -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); -} + ASSERT((flags & ~BTRFS_ORDERED_TYPE_FLAGS) == 0); -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); + /* + * For regular writes, we just use the members in @file_extent. + * + * For NOCOW, we don't really care about the numbers except @start and + * file_extent->num_bytes, as we won't insert a file extent item at all. + * + * For PREALLOC, we do not use ordered extent members, but + * btrfs_mark_extent_written() handles everything. + * + * So here we always pass 0 as offset for NOCOW/PREALLOC ordered extents, + * or btrfs_split_ordered_extent() cannot handle it correctly. + */ + if (flags & ((1U << BTRFS_ORDERED_NOCOW) | (1U << BTRFS_ORDERED_PREALLOC))) + entry = alloc_ordered_extent(inode, file_offset, + file_extent->num_bytes, + file_extent->num_bytes, + file_extent->disk_bytenr + file_extent->offset, + file_extent->num_bytes, 0, flags, + file_extent->compression); + else + entry = alloc_ordered_extent(inode, file_offset, + file_extent->num_bytes, + file_extent->ram_bytes, + file_extent->disk_bytenr, + file_extent->disk_num_bytes, + file_extent->offset, flags, + file_extent->compression); + if (!IS_ERR(entry)) + insert_ordered_extent(entry); + return entry; } /* @@ -275,212 +323,293 @@ int btrfs_add_ordered_extent_compress(struct inode *inode, u64 file_offset, * 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 inode *inode, - struct btrfs_ordered_extent *entry, +void btrfs_add_ordered_sum(struct btrfs_ordered_extent *entry, struct btrfs_ordered_sum *sum) { - struct btrfs_ordered_inode_tree *tree; + struct btrfs_inode *inode = entry->inode; - tree = &BTRFS_I(inode)->ordered_tree; - spin_lock_irq(&tree->lock); + spin_lock(&inode->ordered_tree_lock); list_add_tail(&sum->list, &entry->list); - WARN_ON(entry->csum_bytes_left < sum->len); - entry->csum_bytes_left -= sum->len; - if (entry->csum_bytes_left == 0) - wake_up(&entry->wait); - spin_unlock_irq(&tree->lock); + spin_unlock(&inode->ordered_tree_lock); +} + +void btrfs_mark_ordered_extent_error(struct btrfs_ordered_extent *ordered) +{ + if (!test_and_set_bit(BTRFS_ORDERED_IOERR, &ordered->flags)) + mapping_set_error(ordered->inode->vfs_inode.i_mapping, -EIO); +} + +static void finish_ordered_fn(struct btrfs_work *work) +{ + struct btrfs_ordered_extent *ordered_extent; + + ordered_extent = container_of(work, struct btrfs_ordered_extent, work); + btrfs_finish_ordered_io(ordered_extent); +} + +static bool can_finish_ordered_extent(struct btrfs_ordered_extent *ordered, + struct folio *folio, u64 file_offset, + u64 len, bool uptodate) +{ + struct btrfs_inode *inode = ordered->inode; + struct btrfs_fs_info *fs_info = inode->root->fs_info; + + lockdep_assert_held(&inode->ordered_tree_lock); + + if (folio) { + ASSERT(folio->mapping); + ASSERT(folio_pos(folio) <= file_offset); + ASSERT(file_offset + len <= folio_next_pos(folio)); + + /* + * Ordered flag indicates whether we still have + * pending io unfinished for the ordered extent. + * + * If it's not set, we need to skip to next range. + */ + if (!btrfs_folio_test_ordered(fs_info, folio, file_offset, len)) + return false; + btrfs_folio_clear_ordered(fs_info, folio, file_offset, len); + } + + /* Now we're fine to update the accounting. */ + if (WARN_ON_ONCE(len > ordered->bytes_left)) { + btrfs_crit(fs_info, +"bad ordered extent accounting, root=%llu ino=%llu OE offset=%llu OE len=%llu to_dec=%llu left=%llu", + btrfs_root_id(inode->root), btrfs_ino(inode), + ordered->file_offset, ordered->num_bytes, + len, ordered->bytes_left); + ordered->bytes_left = 0; + } else { + ordered->bytes_left -= len; + } + + if (!uptodate) + set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); + + if (ordered->bytes_left) + return false; + + /* + * All the IO of the ordered extent is finished, we need to queue + * the finish_func to be executed. + */ + set_bit(BTRFS_ORDERED_IO_DONE, &ordered->flags); + cond_wake_up(&ordered->wait); + refcount_inc(&ordered->refs); + trace_btrfs_ordered_extent_mark_finished(inode, ordered); + return true; +} + +static void btrfs_queue_ordered_fn(struct btrfs_ordered_extent *ordered) +{ + struct btrfs_inode *inode = ordered->inode; + struct btrfs_fs_info *fs_info = inode->root->fs_info; + struct btrfs_workqueue *wq = btrfs_is_free_space_inode(inode) ? + fs_info->endio_freespace_worker : fs_info->endio_write_workers; + + btrfs_init_work(&ordered->work, finish_ordered_fn, NULL); + btrfs_queue_work(wq, &ordered->work); +} + +void btrfs_finish_ordered_extent(struct btrfs_ordered_extent *ordered, + struct folio *folio, u64 file_offset, u64 len, + bool uptodate) +{ + struct btrfs_inode *inode = ordered->inode; + bool ret; + + trace_btrfs_finish_ordered_extent(inode, file_offset, len, uptodate); + + spin_lock(&inode->ordered_tree_lock); + ret = can_finish_ordered_extent(ordered, folio, file_offset, len, + uptodate); + spin_unlock(&inode->ordered_tree_lock); + + /* + * If this is a COW write it means we created new extent maps for the + * range and they point to unwritten locations if we got an error either + * before submitting a bio or during IO. + * + * We have marked the ordered extent with BTRFS_ORDERED_IOERR, and we + * are queuing its completion below. During completion, at + * btrfs_finish_one_ordered(), we will drop the extent maps for the + * unwritten extents. + * + * However because completion runs in a work queue we can end up having + * a fast fsync running before that. In the case of direct IO, once we + * unlock the inode the fsync might start, and we queue the completion + * before unlocking the inode. In the case of buffered IO when writeback + * finishes (end_bbio_data_write()) we queue the completion, so if the + * writeback was triggered by a fast fsync, the fsync might start + * logging before ordered extent completion runs in the work queue. + * + * The fast fsync will log file extent items based on the extent maps it + * finds, so if by the time it collects extent maps the ordered extent + * completion didn't happen yet, it will log file extent items that + * point to unwritten extents, resulting in a corruption if a crash + * happens and the log tree is replayed. Note that a fast fsync does not + * wait for completion of ordered extents in order to reduce latency. + * + * Set a flag in the inode so that the next fast fsync will wait for + * ordered extents to complete before starting to log. + */ + if (!uptodate && !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags)) + set_bit(BTRFS_INODE_COW_WRITE_ERROR, &inode->runtime_flags); + + if (ret) + btrfs_queue_ordered_fn(ordered); } /* - * 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. + * Mark all ordered extents io inside the specified range finished. * - * 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. + * @folio: The involved folio for the operation. + * For uncompressed buffered IO, the folio status also needs to be + * updated to indicate whether the pending ordered io is finished. + * Can be NULL for direct IO and compressed write. + * For these cases, callers are ensured they won't execute the + * endio function twice. * - * file_offset is updated to one byte past the range that is recorded as - * complete. This allows you to walk forward in the file. + * This function is called for endio, thus the range must have ordered + * extent(s) covering it. */ -int btrfs_dec_test_first_ordered_pending(struct inode *inode, - struct btrfs_ordered_extent **cached, - u64 *file_offset, u64 io_size, int uptodate) +void btrfs_mark_ordered_io_finished(struct btrfs_inode *inode, + struct folio *folio, u64 file_offset, + u64 num_bytes, bool uptodate) { - 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; - } + u64 cur = file_offset; + const u64 end = file_offset + num_bytes; - entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); - if (!offset_in_entry(entry, *file_offset)) { - ret = 1; - goto out; - } + trace_btrfs_writepage_end_io_hook(inode, file_offset, end - 1, uptodate); - 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) { - printk(KERN_CRIT "bad ordering dec_start %llu end %llu\n", - (unsigned long long)dec_start, - (unsigned long long)dec_end); - } - to_dec = dec_end - dec_start; - if (to_dec > entry->bytes_left) { - printk(KERN_CRIT "bad ordered accounting left %llu size %llu\n", - (unsigned long long)entry->bytes_left, - (unsigned long long)to_dec); - } - entry->bytes_left -= to_dec; - if (!uptodate) - set_bit(BTRFS_ORDERED_IOERR, &entry->flags); + spin_lock(&inode->ordered_tree_lock); + while (cur < end) { + u64 entry_end; + u64 this_end; + u64 len; - if (entry->bytes_left == 0) - ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags); - else - ret = 1; -out: - if (!ret && cached && entry) { - *cached = entry; - atomic_inc(&entry->refs); + node = ordered_tree_search(inode, cur); + /* No ordered extents at all */ + if (!node) + break; + + entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); + entry_end = entry->file_offset + entry->num_bytes; + /* + * |<-- OE --->| | + * cur + * Go to next OE. + */ + if (cur >= entry_end) { + node = rb_next(node); + /* No more ordered extents, exit */ + if (!node) + break; + entry = rb_entry(node, struct btrfs_ordered_extent, + rb_node); + + /* Go to next ordered extent and continue */ + cur = entry->file_offset; + continue; + } + /* + * | |<--- OE --->| + * cur + * Go to the start of OE. + */ + if (cur < entry->file_offset) { + cur = entry->file_offset; + continue; + } + + /* + * Now we are definitely inside one ordered extent. + * + * |<--- OE --->| + * | + * cur + */ + this_end = min(entry_end, end); + len = this_end - cur; + ASSERT(len < U32_MAX); + + if (can_finish_ordered_extent(entry, folio, cur, len, uptodate)) { + spin_unlock(&inode->ordered_tree_lock); + btrfs_queue_ordered_fn(entry); + spin_lock(&inode->ordered_tree_lock); + } + cur += len; } - spin_unlock_irqrestore(&tree->lock, flags); - return ret == 0; + spin_unlock(&inode->ordered_tree_lock); } /* - * 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. + * Finish IO for one ordered extent across a given range. The range can only + * contain one ordered extent. + * + * @cached: The cached ordered extent. If not NULL, we can skip the tree + * search and use the ordered extent directly. + * Will be also used to store the finished ordered extent. + * @file_offset: File offset for the finished IO + * @io_size: Length of the finish IO range * - * 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. + * Return true if the ordered extent is finished in the range, and update + * @cached. + * Return false otherwise. + * + * NOTE: The range can NOT cross multiple ordered extents. + * Thus caller should ensure the range doesn't cross ordered extents. */ -int btrfs_dec_test_ordered_pending(struct inode *inode, - struct btrfs_ordered_extent **cached, - u64 file_offset, u64 io_size, int uptodate) +bool btrfs_dec_test_ordered_pending(struct btrfs_inode *inode, + struct btrfs_ordered_extent **cached, + u64 file_offset, u64 io_size) { - struct btrfs_ordered_inode_tree *tree; struct rb_node *node; struct btrfs_ordered_extent *entry = NULL; - unsigned long flags; - int ret; + bool finished = false; - tree = &BTRFS_I(inode)->ordered_tree; - spin_lock_irqsave(&tree->lock, flags); + spin_lock(&inode->ordered_tree_lock); if (cached && *cached) { entry = *cached; goto have_entry; } - node = tree_search(tree, file_offset); - if (!node) { - ret = 1; + node = ordered_tree_search(inode, file_offset); + if (!node) goto out; - } entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); have_entry: - if (!offset_in_entry(entry, file_offset)) { - ret = 1; + if (!in_range(file_offset, entry->file_offset, entry->num_bytes)) goto out; - } - if (io_size > entry->bytes_left) { - printk(KERN_CRIT "bad ordered accounting left %llu size %llu\n", - (unsigned long long)entry->bytes_left, - (unsigned long long)io_size); - } + if (io_size > entry->bytes_left) + btrfs_crit(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); - else - ret = 1; + if (entry->bytes_left == 0) { + /* + * Ensure only one caller can set the flag and finished_ret + * accordingly + */ + finished = !test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags); + /* test_and_set_bit implies a barrier */ + cond_wake_up_nomb(&entry->wait); + } out: - if (!ret && cached && entry) { + if (finished && cached && entry) { *cached = entry; - atomic_inc(&entry->refs); + refcount_inc(&entry->refs); + trace_btrfs_ordered_extent_dec_test_pending(inode, entry); } - spin_unlock_irqrestore(&tree->lock, flags); - return ret == 0; -} - -/* Needs to either be called under a log transaction or the log_mutex */ -void btrfs_get_logged_extents(struct btrfs_root *log, struct inode *inode) -{ - struct btrfs_ordered_inode_tree *tree; - struct btrfs_ordered_extent *ordered; - struct rb_node *n; - int index = log->log_transid % 2; - - tree = &BTRFS_I(inode)->ordered_tree; - spin_lock_irq(&tree->lock); - for (n = rb_first(&tree->tree); n; n = rb_next(n)) { - ordered = rb_entry(n, struct btrfs_ordered_extent, rb_node); - spin_lock(&log->log_extents_lock[index]); - if (list_empty(&ordered->log_list)) { - list_add_tail(&ordered->log_list, &log->logged_list[index]); - atomic_inc(&ordered->refs); - } - spin_unlock(&log->log_extents_lock[index]); - } - spin_unlock_irq(&tree->lock); -} - -void btrfs_wait_logged_extents(struct btrfs_root *log, u64 transid) -{ - struct btrfs_ordered_extent *ordered; - int index = transid % 2; - - spin_lock_irq(&log->log_extents_lock[index]); - while (!list_empty(&log->logged_list[index])) { - ordered = list_first_entry(&log->logged_list[index], - struct btrfs_ordered_extent, - log_list); - list_del_init(&ordered->log_list); - spin_unlock_irq(&log->log_extents_lock[index]); - wait_event(ordered->wait, test_bit(BTRFS_ORDERED_IO_DONE, - &ordered->flags)); - btrfs_put_ordered_extent(ordered); - spin_lock_irq(&log->log_extents_lock[index]); - } - spin_unlock_irq(&log->log_extents_lock[index]); -} - -void btrfs_free_logged_extents(struct btrfs_root *log, u64 transid) -{ - struct btrfs_ordered_extent *ordered; - int index = transid % 2; - - spin_lock_irq(&log->log_extents_lock[index]); - while (!list_empty(&log->logged_list[index])) { - ordered = list_first_entry(&log->logged_list[index], - struct btrfs_ordered_extent, - log_list); - list_del_init(&ordered->log_list); - spin_unlock_irq(&log->log_extents_lock[index]); - btrfs_put_ordered_extent(ordered); - spin_lock_irq(&log->log_extents_lock[index]); - } - spin_unlock_irq(&log->log_extents_lock[index]); + spin_unlock(&inode->ordered_tree_lock); + return finished; } /* @@ -489,20 +618,18 @@ void btrfs_free_logged_extents(struct btrfs_root *log, u64 transid) */ 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 (atomic_dec_and_test(&entry->refs)) { - 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); - kfree(sum); - } + if (refcount_dec_and_test(&entry->refs)) { + struct btrfs_ordered_sum *sum; + struct btrfs_ordered_sum *tmp; + + ASSERT(list_empty(&entry->root_extent_list)); + ASSERT(list_empty(&entry->log_list)); + ASSERT(RB_EMPTY_NODE(&entry->rb_node)); + btrfs_add_delayed_iput(entry->inode); + list_for_each_entry_safe(sum, tmp, &entry->list, list) + kvfree(sum); kmem_cache_free(btrfs_ordered_extent_cache, entry); } } @@ -511,45 +638,96 @@ void btrfs_put_ordered_extent(struct btrfs_ordered_extent *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, +void btrfs_remove_ordered_extent(struct btrfs_inode *btrfs_inode, struct btrfs_ordered_extent *entry) { - struct btrfs_ordered_inode_tree *tree; - struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_root *root = btrfs_inode->root; + struct btrfs_fs_info *fs_info = root->fs_info; struct rb_node *node; + bool pending; + bool freespace_inode; + + /* + * If this is a free space inode the thread has not acquired the ordered + * extents lockdep map. + */ + freespace_inode = btrfs_is_free_space_inode(btrfs_inode); + + btrfs_lockdep_acquire(fs_info, btrfs_trans_pending_ordered); + /* This is paired with alloc_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) { + u64 release; + + if (test_bit(BTRFS_ORDERED_ENCODED, &entry->flags)) + release = entry->disk_num_bytes; + else + release = entry->num_bytes; + btrfs_delalloc_release_metadata(btrfs_inode, release, + test_bit(BTRFS_ORDERED_IOERR, + &entry->flags)); + } + + percpu_counter_add_batch(&fs_info->ordered_bytes, -entry->num_bytes, + fs_info->delalloc_batch); - tree = &BTRFS_I(inode)->ordered_tree; - spin_lock_irq(&tree->lock); + spin_lock(&btrfs_inode->ordered_tree_lock); node = &entry->rb_node; - rb_erase(node, &tree->tree); - tree->last = NULL; + rb_erase(node, &btrfs_inode->ordered_tree); + RB_CLEAR_NODE(node); + if (btrfs_inode->ordered_tree_last == node) + btrfs_inode->ordered_tree_last = NULL; set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags); - spin_unlock_irq(&tree->lock); + pending = test_and_clear_bit(BTRFS_ORDERED_PENDING, &entry->flags); + spin_unlock(&btrfs_inode->ordered_tree_lock); + + /* + * The current running transaction is waiting on us, we need to let it + * know that we're complete and wake it up. + */ + if (pending) { + struct btrfs_transaction *trans; + + /* + * The checks for trans are just a formality, it should be set, + * but if it isn't we don't want to deref/assert under the spin + * lock, so be nice and check if trans is set, but ASSERT() so + * if it isn't set a developer will notice. + */ + spin_lock(&fs_info->trans_lock); + trans = fs_info->running_transaction; + if (trans) + refcount_inc(&trans->use_count); + spin_unlock(&fs_info->trans_lock); + + ASSERT(trans || BTRFS_FS_ERROR(fs_info)); + if (trans) { + if (atomic_dec_and_test(&trans->pending_ordered)) + wake_up(&trans->pending_wait); + btrfs_put_transaction(trans); + } + } + + btrfs_lockdep_release(fs_info, btrfs_trans_pending_ordered); spin_lock(&root->ordered_extent_lock); list_del_init(&entry->root_extent_list); root->nr_ordered_extents--; - trace_btrfs_ordered_extent_remove(inode, entry); - - /* - * we have no more ordered extents for this inode and - * no dirty pages. We can safely remove it from the - * list of ordered extents - */ - if (RB_EMPTY_ROOT(&tree->tree) && - !mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY)) { - list_del_init(&BTRFS_I(inode)->ordered_operations); - } + trace_btrfs_ordered_extent_remove(btrfs_inode, entry); if (!root->nr_ordered_extents) { - spin_lock(&root->fs_info->ordered_root_lock); + spin_lock(&fs_info->ordered_root_lock); BUG_ON(list_empty(&root->ordered_root)); list_del_init(&root->ordered_root); - spin_unlock(&root->fs_info->ordered_root_lock); + spin_unlock(&fs_info->ordered_root_lock); } spin_unlock(&root->ordered_extent_lock); wake_up(&entry->wait); + if (!freespace_inode) + btrfs_lockdep_release(fs_info, btrfs_ordered_extent); } static void btrfs_run_ordered_extent_work(struct btrfs_work *work) @@ -557,241 +735,197 @@ 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); + btrfs_start_ordered_extent(ordered); complete(&ordered->completion); } /* - * wait for all the ordered extents in a root. This is done when balancing - * space between drives. + * Wait for all the ordered extents in a root. Use @bg as range or do whole + * range if it's NULL. */ -void btrfs_wait_ordered_extents(struct btrfs_root *root, int delay_iput) +u64 btrfs_wait_ordered_extents(struct btrfs_root *root, u64 nr, + const struct btrfs_block_group *bg) { - struct list_head splice, works; + struct btrfs_fs_info *fs_info = root->fs_info; + LIST_HEAD(splice); + LIST_HEAD(skipped); + LIST_HEAD(works); struct btrfs_ordered_extent *ordered, *next; - struct inode *inode; + u64 count = 0; + u64 range_start, range_len; + u64 range_end; - INIT_LIST_HEAD(&splice); - INIT_LIST_HEAD(&works); + if (bg) { + range_start = bg->start; + range_len = bg->length; + } else { + range_start = 0; + range_len = U64_MAX; + } + range_end = range_start + range_len; - mutex_lock(&root->fs_info->ordered_operations_mutex); + mutex_lock(&root->ordered_extent_mutex); spin_lock(&root->ordered_extent_lock); list_splice_init(&root->ordered_extents, &splice); - while (!list_empty(&splice)) { + while (!list_empty(&splice) && nr) { ordered = list_first_entry(&splice, struct btrfs_ordered_extent, root_extent_list); - list_move_tail(&ordered->root_extent_list, - &root->ordered_extents); - /* - * the inode may be getting freed (in sys_unlink path). - */ - inode = igrab(ordered->inode); - if (!inode) { + + if (range_end <= ordered->disk_bytenr || + ordered->disk_bytenr + ordered->disk_num_bytes <= range_start) { + list_move_tail(&ordered->root_extent_list, &skipped); cond_resched_lock(&root->ordered_extent_lock); continue; } - atomic_inc(&ordered->refs); + list_move_tail(&ordered->root_extent_list, + &root->ordered_extents); + refcount_inc(&ordered->refs); spin_unlock(&root->ordered_extent_lock); - ordered->flush_work.func = btrfs_run_ordered_extent_work; + btrfs_init_work(&ordered->flush_work, + btrfs_run_ordered_extent_work, NULL); list_add_tail(&ordered->work_list, &works); - btrfs_queue_worker(&root->fs_info->flush_workers, - &ordered->flush_work); + btrfs_queue_work(fs_info->flush_workers, &ordered->flush_work); cond_resched(); + if (nr != U64_MAX) + nr--; + count++; spin_lock(&root->ordered_extent_lock); } + 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); - - inode = ordered->inode; btrfs_put_ordered_extent(ordered); - if (delay_iput) - btrfs_add_delayed_iput(inode); - else - iput(inode); - cond_resched(); } - mutex_unlock(&root->fs_info->ordered_operations_mutex); + mutex_unlock(&root->ordered_extent_mutex); + + return count; } -void btrfs_wait_all_ordered_extents(struct btrfs_fs_info *fs_info, - int delay_iput) +/* + * Wait for @nr ordered extents that intersect the @bg, or the whole range of + * the filesystem if @bg is NULL. + */ +void btrfs_wait_ordered_roots(struct btrfs_fs_info *fs_info, u64 nr, + const struct btrfs_block_group *bg) { struct btrfs_root *root; - struct list_head splice; - - INIT_LIST_HEAD(&splice); + LIST_HEAD(splice); + u64 done; + 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)) { + while (!list_empty(&splice) && nr) { root = list_first_entry(&splice, struct btrfs_root, ordered_root); - root = btrfs_grab_fs_root(root); + root = btrfs_grab_root(root); BUG_ON(!root); list_move_tail(&root->ordered_root, &fs_info->ordered_roots); spin_unlock(&fs_info->ordered_root_lock); - btrfs_wait_ordered_extents(root, delay_iput); - btrfs_put_fs_root(root); + done = btrfs_wait_ordered_extents(root, nr, bg); + btrfs_put_root(root); + + if (nr != U64_MAX) + nr -= done; spin_lock(&fs_info->ordered_root_lock); } + list_splice_tail(&splice, &fs_info->ordered_roots); spin_unlock(&fs_info->ordered_root_lock); + mutex_unlock(&fs_info->ordered_operations_mutex); } /* - * this is used during transaction commit to write all the inodes - * added to the ordered operation list. These files must be fully on - * disk before the transaction commits. + * Start IO and wait for a given ordered extent to finish. * - * we have two modes here, one is to just start the IO via filemap_flush - * and the other is to wait for all the io. When we wait, we have an - * extra check to make sure the ordered operation list really is empty - * before we return + * Wait on page writeback for all the pages in the extent but not in + * [@nowriteback_start, @nowriteback_start + @nowriteback_len) and the + * IO completion code to insert metadata into the btree corresponding to the extent. */ -int btrfs_run_ordered_operations(struct btrfs_trans_handle *trans, - struct btrfs_root *root, int wait) -{ - struct btrfs_inode *btrfs_inode; - struct inode *inode; - struct btrfs_transaction *cur_trans = trans->transaction; - struct list_head splice; - struct list_head works; - struct btrfs_delalloc_work *work, *next; - int ret = 0; - - INIT_LIST_HEAD(&splice); - INIT_LIST_HEAD(&works); - - mutex_lock(&root->fs_info->ordered_operations_mutex); - spin_lock(&root->fs_info->ordered_root_lock); - list_splice_init(&cur_trans->ordered_operations, &splice); - while (!list_empty(&splice)) { - btrfs_inode = list_entry(splice.next, struct btrfs_inode, - ordered_operations); - inode = &btrfs_inode->vfs_inode; - - list_del_init(&btrfs_inode->ordered_operations); - - /* - * the inode may be getting freed (in sys_unlink path). - */ - inode = igrab(inode); - if (!inode) - continue; - - if (!wait) - list_add_tail(&BTRFS_I(inode)->ordered_operations, - &cur_trans->ordered_operations); - spin_unlock(&root->fs_info->ordered_root_lock); - - work = btrfs_alloc_delalloc_work(inode, wait, 1); - if (!work) { - spin_lock(&root->fs_info->ordered_root_lock); - if (list_empty(&BTRFS_I(inode)->ordered_operations)) - list_add_tail(&btrfs_inode->ordered_operations, - &splice); - list_splice_tail(&splice, - &cur_trans->ordered_operations); - spin_unlock(&root->fs_info->ordered_root_lock); - ret = -ENOMEM; - goto out; - } - list_add_tail(&work->list, &works); - btrfs_queue_worker(&root->fs_info->flush_workers, - &work->work); - - cond_resched(); - spin_lock(&root->fs_info->ordered_root_lock); - } - spin_unlock(&root->fs_info->ordered_root_lock); -out: - list_for_each_entry_safe(work, next, &works, list) { - list_del_init(&work->list); - btrfs_wait_and_free_delalloc_work(work); - } - mutex_unlock(&root->fs_info->ordered_operations_mutex); - return ret; -} - -/* - * 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) +void btrfs_start_ordered_extent_nowriteback(struct btrfs_ordered_extent *entry, + u64 nowriteback_start, u32 nowriteback_len) { u64 start = entry->file_offset; - u64 end = start + entry->len - 1; + u64 end = start + entry->num_bytes - 1; + struct btrfs_inode *inode = entry->inode; + bool freespace_inode; trace_btrfs_ordered_extent_start(inode, entry); /* + * If this is a free space inode do not take the ordered extents lockdep + * map. + */ + freespace_inode = btrfs_is_free_space_inode(inode); + + /* * 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)); + if (!test_bit(BTRFS_ORDERED_DIRECT, &entry->flags)) { + if (!nowriteback_len) { + filemap_fdatawrite_range(inode->vfs_inode.i_mapping, start, end); + } else { + if (start < nowriteback_start) + filemap_fdatawrite_range(inode->vfs_inode.i_mapping, start, + nowriteback_start - 1); + if (nowriteback_start + nowriteback_len < end) + filemap_fdatawrite_range(inode->vfs_inode.i_mapping, + nowriteback_start + nowriteback_len, + end); + } } + + if (!freespace_inode) + btrfs_might_wait_for_event(inode->root->fs_info, btrfs_ordered_extent); + wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE, &entry->flags)); } /* * Used to wait on ordered extents across a large range of bytes. */ -void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) +int btrfs_wait_ordered_range(struct btrfs_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); + orig_end = OFFSET_MAX; } else { orig_end = start + len - 1; - if (orig_end > INT_LIMIT(loff_t)) - orig_end = INT_LIMIT(loff_t); + if (orig_end > OFFSET_MAX) + orig_end = OFFSET_MAX; } /* start IO across the range first to instantiate any delalloc * extents */ - filemap_fdatawrite_range(inode->i_mapping, start, orig_end); + ret = btrfs_fdatawrite_range(inode, start, orig_end); + if (ret) + return ret; /* - * So with compression we will find and lock a dirty page and clear the - * first one as dirty, setup an async extent, and immediately return - * with the entire range locked but with nobody actually marked with - * writeback. So we can't just filemap_write_and_wait_range() and - * expect it to work since it will just kick off a thread to do the - * actual work. So we need to call filemap_fdatawrite_range _again_ - * since it will wait on the page lock, which won't be unlocked until - * after the pages have been marked as writeback and so we're good to go - * from there. We have to do this otherwise we'll miss the ordered - * extents and that results in badness. Please Josef, do not think you - * know better and pull this out at some point in the future, it is - * right and you are wrong. + * 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. */ - if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT, - &BTRFS_I(inode)->runtime_flags)) - filemap_fdatawrite_range(inode->i_mapping, start, orig_end); - - filemap_fdatawait_range(inode->i_mapping, start, orig_end); + ret_wb = filemap_fdatawait_range(inode->vfs_inode.i_mapping, start, orig_end); end = orig_end; while (1) { @@ -802,69 +936,74 @@ void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) btrfs_put_ordered_extent(ordered); break; } - if (ordered->file_offset + ordered->len < start) { + if (ordered->file_offset + ordered->num_bytes <= start) { btrfs_put_ordered_extent(ordered); break; } - btrfs_start_ordered_extent(inode, ordered, 1); + btrfs_start_ordered_extent(ordered); end = ordered->file_offset; + /* + * If the ordered extent had an error save the error but don't + * exit without waiting first for all other ordered extents in + * the range to complete. + */ + if (test_bit(BTRFS_ORDERED_IOERR, &ordered->flags)) + ret = -EIO; btrfs_put_ordered_extent(ordered); if (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, +struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct btrfs_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); + spin_lock(&inode->ordered_tree_lock); + node = ordered_tree_search(inode, file_offset); if (!node) goto out; entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); - if (!offset_in_entry(entry, file_offset)) + if (!in_range(file_offset, entry->file_offset, entry->num_bytes)) entry = NULL; - if (entry) - atomic_inc(&entry->refs); + if (entry) { + refcount_inc(&entry->refs); + trace_btrfs_ordered_extent_lookup(inode, entry); + } out: - spin_unlock_irq(&tree->lock); + spin_unlock(&inode->ordered_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 inode *inode, - u64 file_offset, - u64 len) +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 = &BTRFS_I(inode)->ordered_tree; - spin_lock_irq(&tree->lock); - node = tree_search(tree, file_offset); + spin_lock(&inode->ordered_tree_lock); + node = ordered_tree_search(inode, file_offset); if (!node) { - node = tree_search(tree, file_offset + len); + node = ordered_tree_search(inode, 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)) + if (btrfs_range_overlaps(entry, file_offset, len)) break; if (entry->file_offset >= file_offset + len) { @@ -877,246 +1016,343 @@ struct btrfs_ordered_extent *btrfs_lookup_ordered_range(struct inode *inode, break; } out: - if (entry) - atomic_inc(&entry->refs); - spin_unlock_irq(&tree->lock); + if (entry) { + refcount_inc(&entry->refs); + trace_btrfs_ordered_extent_lookup_range(inode, entry); + } + spin_unlock(&inode->ordered_tree_lock); return entry; } /* + * Adds all ordered extents to the given list. The list ends up sorted by the + * file_offset of the ordered extents. + */ +void btrfs_get_ordered_extents_for_logging(struct btrfs_inode *inode, + struct list_head *list) +{ + struct rb_node *n; + + btrfs_assert_inode_locked(inode); + + spin_lock(&inode->ordered_tree_lock); + for (n = rb_first(&inode->ordered_tree); n; n = rb_next(n)) { + struct btrfs_ordered_extent *ordered; + + ordered = rb_entry(n, struct btrfs_ordered_extent, rb_node); + + if (test_bit(BTRFS_ORDERED_LOGGED, &ordered->flags)) + continue; + + ASSERT(list_empty(&ordered->log_list)); + list_add_tail(&ordered->log_list, list); + refcount_inc(&ordered->refs); + trace_btrfs_ordered_extent_lookup_for_logging(inode, ordered); + } + spin_unlock(&inode->ordered_tree_lock); +} + +/* * 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) +btrfs_lookup_first_ordered_extent(struct btrfs_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); + spin_lock(&inode->ordered_tree_lock); + node = ordered_tree_search(inode, file_offset); if (!node) goto out; entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); - atomic_inc(&entry->refs); + refcount_inc(&entry->refs); + trace_btrfs_ordered_extent_lookup_first(inode, entry); out: - spin_unlock_irq(&tree->lock); + spin_unlock(&inode->ordered_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. + * Lookup the first ordered extent that overlaps the range + * [@file_offset, @file_offset + @len). + * + * The difference between this and btrfs_lookup_first_ordered_extent() is + * that this one won't return any ordered extent that does not overlap the range. + * And the difference against btrfs_lookup_ordered_extent() is, this function + * ensures the first ordered extent gets returned. */ -int btrfs_ordered_update_i_size(struct inode *inode, u64 offset, - struct btrfs_ordered_extent *ordered) +struct btrfs_ordered_extent *btrfs_lookup_first_ordered_range( + struct btrfs_inode *inode, u64 file_offset, u64 len) { - 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; - - if (ordered) - offset = entry_end(ordered); - else - offset = ALIGN(offset, BTRFS_I(inode)->root->sectorsize); - - spin_lock_irq(&tree->lock); - disk_i_size = BTRFS_I(inode)->disk_i_size; - - /* truncate file */ - if (disk_i_size > i_size) { - BTRFS_I(inode)->disk_i_size = i_size; - 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; + struct rb_node *cur; + struct rb_node *prev; + struct rb_node *next; + struct btrfs_ordered_extent *entry = NULL; + spin_lock(&inode->ordered_tree_lock); + node = inode->ordered_tree.rb_node; /* - * 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 + * Here we don't want to use tree_search() which will use tree->last + * and screw up the search order. + * And __tree_search() can't return the adjacent ordered extents + * either, thus here we do our own search. */ - 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); + while (node) { + entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); - /* We treat this entry as if it doesnt exist */ - if (test_bit(BTRFS_ORDERED_UPDATED_ISIZE, &test->flags)) - continue; - if (test->file_offset + test->len <= disk_i_size) - break; - if (test->file_offset >= i_size) - break; - if (entry_end(test) > disk_i_size) { + if (file_offset < entry->file_offset) { + node = node->rb_left; + } else if (file_offset >= entry_end(entry)) { + node = node->rb_right; + } else { /* - * we don't update disk_i_size now, so record this - * undealt i_size. Or we will not know the real - * i_size. + * Direct hit, got an ordered extent that starts at + * @file_offset */ - 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); + if (!entry) { + /* Empty tree */ + goto out; + } - /* - * 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; + cur = &entry->rb_node; + /* We got an entry around @file_offset, check adjacent entries */ + if (entry->file_offset < file_offset) { + prev = cur; + next = rb_next(cur); + } else { + prev = rb_prev(cur); + next = cur; + } + if (prev) { + entry = rb_entry(prev, struct btrfs_ordered_extent, rb_node); + if (btrfs_range_overlaps(entry, file_offset, len)) + goto out; + } + if (next) { + entry = rb_entry(next, struct btrfs_ordered_extent, rb_node); + if (btrfs_range_overlaps(entry, file_offset, len)) + goto out; + } + /* No ordered extent in the range */ + entry = NULL; 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; + if (entry) { + refcount_inc(&entry->refs); + trace_btrfs_ordered_extent_lookup_first_range(inode, entry); + } + + spin_unlock(&inode->ordered_tree_lock); + return entry; } /* - * 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 + * Lock the passed range and ensures all pending ordered extents in it are run + * to completion. + * + * @inode: Inode whose ordered tree is to be searched + * @start: Beginning of range to flush + * @end: Last byte of range to lock + * @cached_state: If passed, will return the extent state responsible for the + * locked range. It's the caller's responsibility to free the + * cached state. + * + * Always return with the given range locked, ensuring after it's called no + * order extent can be pending. */ -int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr, - u32 *sum, int len) +void btrfs_lock_and_flush_ordered_range(struct btrfs_inode *inode, u64 start, + u64 end, + struct extent_state **cached_state) { - 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_I(inode)->root->sectorsize; - int index = 0; + struct extent_state *cache = NULL; + struct extent_state **cachedp = &cache; - ordered = btrfs_lookup_ordered_extent(inode, offset); - if (!ordered) - return 0; + if (cached_state) + cachedp = cached_state; - 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; + while (1) { + btrfs_lock_extent(&inode->io_tree, start, end, cachedp); + ordered = btrfs_lookup_ordered_range(inode, start, + end - start + 1); + if (!ordered) { + /* + * If no external cached_state has been passed then + * decrement the extra ref taken for cachedp since we + * aren't exposing it outside of this function + */ + if (!cached_state) + refcount_dec(&cache->refs); + break; } + btrfs_unlock_extent(&inode->io_tree, start, end, cachedp); + btrfs_start_ordered_extent(ordered); + btrfs_put_ordered_extent(ordered); } -out: - spin_unlock_irq(&tree->lock); - btrfs_put_ordered_extent(ordered); - return index; } - /* - * add a given inode to the list of inodes that must be fully on - * disk before a transaction commit finishes. - * - * This basically gives us the ext3 style data=ordered mode, and it is mostly - * used to make sure renamed files are fully on disk. - * - * It is a noop if the inode is already fully on disk. + * Lock the passed range and ensure all pending ordered extents in it are run + * to completion in nowait mode. * - * If trans is not null, we'll do a friendly check for a transaction that - * is already flushing things and force the IO down ourselves. + * Return true if btrfs_lock_ordered_range does not return any extents, + * otherwise false. */ -void btrfs_add_ordered_operation(struct btrfs_trans_handle *trans, - struct btrfs_root *root, struct inode *inode) +bool btrfs_try_lock_ordered_range(struct btrfs_inode *inode, u64 start, u64 end, + struct extent_state **cached_state) +{ + struct btrfs_ordered_extent *ordered; + + if (!btrfs_try_lock_extent(&inode->io_tree, start, end, cached_state)) + return false; + + ordered = btrfs_lookup_ordered_range(inode, start, end - start + 1); + if (!ordered) + return true; + + btrfs_put_ordered_extent(ordered); + btrfs_unlock_extent(&inode->io_tree, start, end, cached_state); + + return false; +} + +/* Split out a new ordered extent for this first @len bytes of @ordered. */ +struct btrfs_ordered_extent *btrfs_split_ordered_extent( + struct btrfs_ordered_extent *ordered, u64 len) { - struct btrfs_transaction *cur_trans = trans->transaction; - u64 last_mod; + struct btrfs_inode *inode = ordered->inode; + struct btrfs_root *root = inode->root; + struct btrfs_fs_info *fs_info = root->fs_info; + u64 file_offset = ordered->file_offset; + u64 disk_bytenr = ordered->disk_bytenr; + unsigned long flags = ordered->flags; + struct btrfs_ordered_sum *sum, *tmpsum; + struct btrfs_ordered_extent *new; + struct rb_node *node; + u64 offset = 0; - last_mod = max(BTRFS_I(inode)->generation, BTRFS_I(inode)->last_trans); + trace_btrfs_ordered_extent_split(inode, ordered); + + ASSERT(!(flags & (1U << BTRFS_ORDERED_COMPRESSED))); /* - * if this file hasn't been changed since the last transaction - * commit, we can safely return without doing anything + * The entire bio must be covered by the ordered extent, but we can't + * reduce the original extent to a zero length either. */ - if (last_mod < root->fs_info->last_trans_committed) - return; + if (WARN_ON_ONCE(len >= ordered->num_bytes)) + return ERR_PTR(-EINVAL); + /* + * If our ordered extent had an error there's no point in continuing. + * The error may have come from a transaction abort done either by this + * task or some other concurrent task, and the transaction abort path + * iterates over all existing ordered extents and sets the flag + * BTRFS_ORDERED_IOERR on them. + */ + if (unlikely(flags & (1U << BTRFS_ORDERED_IOERR))) { + const int fs_error = BTRFS_FS_ERROR(fs_info); - spin_lock(&root->fs_info->ordered_root_lock); - if (list_empty(&BTRFS_I(inode)->ordered_operations)) { - list_add_tail(&BTRFS_I(inode)->ordered_operations, - &cur_trans->ordered_operations); + return fs_error ? ERR_PTR(fs_error) : ERR_PTR(-EIO); + } + /* We cannot split partially completed ordered extents. */ + if (ordered->bytes_left) { + ASSERT(!(flags & ~BTRFS_ORDERED_TYPE_FLAGS)); + if (WARN_ON_ONCE(ordered->bytes_left != ordered->disk_num_bytes)) + return ERR_PTR(-EINVAL); } - spin_unlock(&root->fs_info->ordered_root_lock); + /* We cannot split a compressed ordered extent. */ + if (WARN_ON_ONCE(ordered->disk_num_bytes != ordered->num_bytes)) + return ERR_PTR(-EINVAL); + + new = alloc_ordered_extent(inode, file_offset, len, len, disk_bytenr, + len, 0, flags, ordered->compress_type); + if (IS_ERR(new)) + return new; + + /* One ref for the tree. */ + refcount_inc(&new->refs); + + /* + * Take the root's ordered_extent_lock to avoid a race with + * btrfs_wait_ordered_extents() when updating the disk_bytenr and + * disk_num_bytes fields of the ordered extent below. + * + * There's no concern about a previous caller of + * btrfs_wait_ordered_extents() getting the trimmed ordered extent + * before we insert the new one, because even if it gets the ordered + * extent before it's trimmed and the new one inserted, right before it + * uses it or during its use, the ordered extent might have been + * trimmed in the meanwhile, and it missed the new ordered extent. + * There's no way around this and it's harmless for current use cases, + * so we take the root's ordered_extent_lock to fix that race during + * trimming and silence tools like KCSAN. + */ + spin_lock_irq(&root->ordered_extent_lock); + spin_lock(&inode->ordered_tree_lock); + + /* + * We don't have overlapping ordered extents (that would imply double + * allocation of extents) and we checked above that the split length + * does not cross the ordered extent's num_bytes field, so there's + * no need to remove it and re-insert it in the tree. + */ + ordered->file_offset += len; + ordered->disk_bytenr += len; + ordered->num_bytes -= len; + ordered->disk_num_bytes -= len; + ordered->ram_bytes -= len; + + if (test_bit(BTRFS_ORDERED_IO_DONE, &ordered->flags)) { + ASSERT(ordered->bytes_left == 0); + new->bytes_left = 0; + } else { + ordered->bytes_left -= len; + } + + if (test_bit(BTRFS_ORDERED_TRUNCATED, &ordered->flags)) { + if (ordered->truncated_len > len) { + ordered->truncated_len -= len; + } else { + new->truncated_len = ordered->truncated_len; + ordered->truncated_len = 0; + } + } + + list_for_each_entry_safe(sum, tmpsum, &ordered->list, list) { + if (offset == len) + break; + list_move_tail(&sum->list, &new->list); + offset += sum->len; + } + + node = tree_insert(&inode->ordered_tree, new->file_offset, &new->rb_node); + if (unlikely(node)) + btrfs_panic(fs_info, -EEXIST, + "inconsistency in ordered tree at offset %llu after split", + new->file_offset); + spin_unlock(&inode->ordered_tree_lock); + + list_add_tail(&new->root_extent_list, &root->ordered_extents); + root->nr_ordered_extents++; + spin_unlock_irq(&root->ordered_extent_lock); + return new; } int __init ordered_data_init(void) { - btrfs_ordered_extent_cache = kmem_cache_create("btrfs_ordered_extent", - sizeof(struct btrfs_ordered_extent), 0, - SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, - NULL); + btrfs_ordered_extent_cache = KMEM_CACHE(btrfs_ordered_extent, 0); if (!btrfs_ordered_extent_cache) return -ENOMEM; return 0; } -void ordered_data_exit(void) +void __cold ordered_data_exit(void) { - if (btrfs_ordered_extent_cache) - kmem_cache_destroy(btrfs_ordered_extent_cache); + kmem_cache_destroy(btrfs_ordered_extent_cache); } |