diff options
Diffstat (limited to 'kernel/fs/btrfs/inode.c')
| -rw-r--r-- | kernel/fs/btrfs/inode.c | 546 |
1 files changed, 389 insertions, 157 deletions
diff --git a/kernel/fs/btrfs/inode.c b/kernel/fs/btrfs/inode.c index 8bb013672..4bc9dbf29 100644 --- a/kernel/fs/btrfs/inode.c +++ b/kernel/fs/btrfs/inode.c @@ -310,6 +310,13 @@ static noinline int cow_file_range_inline(struct btrfs_root *root, btrfs_delalloc_release_metadata(inode, end + 1 - start); btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0); out: + /* + * Don't forget to free the reserved space, as for inlined extent + * it won't count as data extent, free them directly here. + * And at reserve time, it's always aligned to page size, so + * just free one page here. + */ + btrfs_qgroup_free_data(inode, 0, PAGE_CACHE_SIZE); btrfs_free_path(path); btrfs_end_transaction(trans, root); return ret; @@ -1096,6 +1103,9 @@ static noinline void async_cow_submit(struct btrfs_work *work) nr_pages = (async_cow->end - async_cow->start + PAGE_CACHE_SIZE) >> PAGE_CACHE_SHIFT; + /* + * atomic_sub_return implies a barrier for waitqueue_active + */ if (atomic_sub_return(nr_pages, &root->fs_info->async_delalloc_pages) < 5 * 1024 * 1024 && waitqueue_active(&root->fs_info->async_submit_wait)) @@ -1294,8 +1304,14 @@ next_slot: num_bytes = 0; btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); - if (found_key.objectid > ino || - found_key.type > BTRFS_EXTENT_DATA_KEY || + if (found_key.objectid > ino) + break; + if (WARN_ON_ONCE(found_key.objectid < ino) || + found_key.type < BTRFS_EXTENT_DATA_KEY) { + path->slots[0]++; + goto next_slot; + } + if (found_key.type > BTRFS_EXTENT_DATA_KEY || found_key.offset > end) break; @@ -1766,7 +1782,8 @@ static void btrfs_clear_bit_hook(struct inode *inode, if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID && do_list && !(state->state & EXTENT_NORESERVE)) - btrfs_free_reserved_data_space(inode, len); + btrfs_free_reserved_data_space_noquota(inode, + state->start, len); __percpu_counter_add(&root->fs_info->delalloc_bytes, -len, root->fs_info->delalloc_batch); @@ -1845,8 +1862,10 @@ static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio, int ret; ret = btrfs_map_bio(root, rw, bio, mirror_num, 1); - if (ret) - bio_endio(bio, ret); + if (ret) { + bio->bi_error = ret; + bio_endio(bio); + } return ret; } @@ -1859,15 +1878,15 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, u64 bio_offset) { struct btrfs_root *root = BTRFS_I(inode)->root; + enum btrfs_wq_endio_type metadata = BTRFS_WQ_ENDIO_DATA; int ret = 0; int skip_sum; - int metadata = 0; int async = !atomic_read(&BTRFS_I(inode)->sync_writers); skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM; if (btrfs_is_free_space_inode(inode)) - metadata = 2; + metadata = BTRFS_WQ_ENDIO_FREE_SPACE; if (!(rw & REQ_WRITE)) { ret = btrfs_bio_wq_end_io(root->fs_info, bio, metadata); @@ -1906,8 +1925,10 @@ mapit: ret = btrfs_map_bio(root, rw, bio, mirror_num, 0); out: - if (ret < 0) - bio_endio(bio, ret); + if (ret < 0) { + bio->bi_error = ret; + bio_endio(bio); + } return ret; } @@ -1985,7 +2006,8 @@ again: goto again; } - ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE); + ret = btrfs_delalloc_reserve_space(inode, page_start, + PAGE_CACHE_SIZE); if (ret) { mapping_set_error(page->mapping, ret); end_extent_writepage(page, ret, page_start, page_end); @@ -2111,7 +2133,13 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans, ins.type = BTRFS_EXTENT_ITEM_KEY; ret = btrfs_alloc_reserved_file_extent(trans, root, root->root_key.objectid, - btrfs_ino(inode), file_pos, &ins); + btrfs_ino(inode), file_pos, + ram_bytes, &ins); + /* + * Release the reserved range from inode dirty range map, as it is + * already moved into delayed_ref_head + */ + btrfs_qgroup_release_data(inode, file_pos, ram_bytes); out: btrfs_free_path(path); @@ -2569,7 +2597,7 @@ again: ret = btrfs_inc_extent_ref(trans, root, new->bytenr, new->disk_len, 0, backref->root_id, backref->inum, - new->file_pos, 0); /* start - extent_offset */ + new->file_pos); /* start - extent_offset */ if (ret) { btrfs_abort_transaction(trans, root, ret); goto out_free_path; @@ -2595,7 +2623,6 @@ static void free_sa_defrag_extent(struct new_sa_defrag_extent *new) return; list_for_each_entry_safe(old, tmp, &new->head, list) { - list_del(&old->list); kfree(old); } kfree(new); @@ -2820,6 +2847,14 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent) if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) { BUG_ON(!list_empty(&ordered_extent->list)); /* Logic error */ + + /* + * For mwrite(mmap + memset to write) case, we still reserve + * space for NOCOW range. + * As NOCOW won't cause a new delayed ref, just free the space + */ + btrfs_qgroup_free_data(inode, ordered_extent->file_offset, + ordered_extent->len); btrfs_ordered_update_i_size(inode, 0, ordered_extent); if (nolock) trans = btrfs_join_transaction_nolock(root); @@ -3014,8 +3049,6 @@ static int __readpage_endio_check(struct inode *inode, char *kaddr; u32 csum_expected; u32 csum = ~(u32)0; - static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL, - DEFAULT_RATELIMIT_BURST); csum_expected = *(((u32 *)io_bio->csum) + icsum); @@ -3028,9 +3061,8 @@ static int __readpage_endio_check(struct inode *inode, kunmap_atomic(kaddr); return 0; zeroit: - if (__ratelimit(&_rs)) - btrfs_warn(BTRFS_I(inode)->root->fs_info, - "csum failed ino %llu off %llu csum %u expected csum %u", + btrfs_warn_rl(BTRFS_I(inode)->root->fs_info, + "csum failed ino %llu off %llu csum %u expected csum %u", btrfs_ino(inode), start, csum, csum_expected); memset(kaddr + pgoff, 1, len); flush_dcache_page(page); @@ -3110,8 +3142,6 @@ void btrfs_run_delayed_iputs(struct btrfs_root *root) if (empty) return; - down_read(&fs_info->delayed_iput_sem); - spin_lock(&fs_info->delayed_iput_lock); list_splice_init(&fs_info->delayed_iputs, &list); spin_unlock(&fs_info->delayed_iput_lock); @@ -3122,8 +3152,6 @@ void btrfs_run_delayed_iputs(struct btrfs_root *root) iput(delayed->inode); kfree(delayed); } - - up_read(&root->fs_info->delayed_iput_sem); } /* @@ -3654,6 +3682,35 @@ cache_index: set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags); + /* + * We don't persist the id of the transaction where an unlink operation + * against the inode was last made. So here we assume the inode might + * have been evicted, and therefore the exact value of last_unlink_trans + * lost, and set it to last_trans to avoid metadata inconsistencies + * between the inode and its parent if the inode is fsync'ed and the log + * replayed. For example, in the scenario: + * + * touch mydir/foo + * ln mydir/foo mydir/bar + * sync + * unlink mydir/bar + * echo 2 > /proc/sys/vm/drop_caches # evicts inode + * xfs_io -c fsync mydir/foo + * <power failure> + * mount fs, triggers fsync log replay + * + * We must make sure that when we fsync our inode foo we also log its + * parent inode, otherwise after log replay the parent still has the + * dentry with the "bar" name but our inode foo has a link count of 1 + * and doesn't have an inode ref with the name "bar" anymore. + * + * Setting last_unlink_trans to last_trans is a pessimistic approach, + * but it guarantees correctness at the expense of ocassional full + * transaction commits on fsync if our inode is a directory, or if our + * inode is not a directory, logging its parent unnecessarily. + */ + BTRFS_I(inode)->last_unlink_trans = BTRFS_I(inode)->last_trans; + path->slots[0]++; if (inode->i_nlink != 1 || path->slots[0] >= btrfs_header_nritems(leaf)) @@ -3985,9 +4042,7 @@ int btrfs_unlink_inode(struct btrfs_trans_handle *trans, */ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir) { - struct btrfs_trans_handle *trans; struct btrfs_root *root = BTRFS_I(dir)->root; - int ret; /* * 1 for the possible orphan item @@ -3996,27 +4051,7 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir) * 1 for the inode ref * 1 for the inode */ - trans = btrfs_start_transaction(root, 5); - if (!IS_ERR(trans) || PTR_ERR(trans) != -ENOSPC) - return trans; - - if (PTR_ERR(trans) == -ENOSPC) { - u64 num_bytes = btrfs_calc_trans_metadata_size(root, 5); - - trans = btrfs_start_transaction(root, 0); - if (IS_ERR(trans)) - return trans; - ret = btrfs_cond_migrate_bytes(root->fs_info, - &root->fs_info->trans_block_rsv, - num_bytes, 5); - if (ret) { - btrfs_end_transaction(trans, root); - return ERR_PTR(ret); - } - trans->block_rsv = &root->fs_info->trans_block_rsv; - trans->bytes_reserved = num_bytes; - } - return trans; + return btrfs_start_transaction_fallback_global_rsv(root, 5, 5); } static int btrfs_unlink(struct inode *dir, struct dentry *dentry) @@ -4184,6 +4219,47 @@ static int truncate_space_check(struct btrfs_trans_handle *trans, } +static int truncate_inline_extent(struct inode *inode, + struct btrfs_path *path, + struct btrfs_key *found_key, + const u64 item_end, + const u64 new_size) +{ + struct extent_buffer *leaf = path->nodes[0]; + int slot = path->slots[0]; + struct btrfs_file_extent_item *fi; + u32 size = (u32)(new_size - found_key->offset); + struct btrfs_root *root = BTRFS_I(inode)->root; + + fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item); + + if (btrfs_file_extent_compression(leaf, fi) != BTRFS_COMPRESS_NONE) { + loff_t offset = new_size; + loff_t page_end = ALIGN(offset, PAGE_CACHE_SIZE); + + /* + * Zero out the remaining of the last page of our inline extent, + * instead of directly truncating our inline extent here - that + * would be much more complex (decompressing all the data, then + * compressing the truncated data, which might be bigger than + * the size of the inline extent, resize the extent, etc). + * We release the path because to get the page we might need to + * read the extent item from disk (data not in the page cache). + */ + btrfs_release_path(path); + return btrfs_truncate_page(inode, offset, page_end - offset, 0); + } + + btrfs_set_file_extent_ram_bytes(leaf, fi, size); + size = btrfs_file_extent_calc_inline_size(size); + btrfs_truncate_item(root, path, size, 1); + + if (test_bit(BTRFS_ROOT_REF_COWS, &root->state)) + inode_sub_bytes(inode, item_end + 1 - new_size); + + return 0; +} + /* * this can truncate away extent items, csum items and directory items. * It starts at a high offset and removes keys until it can't find @@ -4209,7 +4285,7 @@ int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans, u64 extent_num_bytes = 0; u64 extent_offset = 0; u64 item_end = 0; - u64 last_size = (u64)-1; + u64 last_size = new_size; u32 found_type = (u8)-1; int found_extent; int del_item; @@ -4378,27 +4454,40 @@ search_again: * special encodings */ if (!del_item && - btrfs_file_extent_compression(leaf, fi) == 0 && btrfs_file_extent_encryption(leaf, fi) == 0 && btrfs_file_extent_other_encoding(leaf, fi) == 0) { - u32 size = new_size - found_key.offset; - - if (test_bit(BTRFS_ROOT_REF_COWS, &root->state)) - inode_sub_bytes(inode, item_end + 1 - - new_size); /* - * update the ram bytes to properly reflect - * the new size of our item + * Need to release path in order to truncate a + * compressed extent. So delete any accumulated + * extent items so far. */ - btrfs_set_file_extent_ram_bytes(leaf, fi, size); - size = - btrfs_file_extent_calc_inline_size(size); - btrfs_truncate_item(root, path, size, 1); + if (btrfs_file_extent_compression(leaf, fi) != + BTRFS_COMPRESS_NONE && pending_del_nr) { + err = btrfs_del_items(trans, root, path, + pending_del_slot, + pending_del_nr); + if (err) { + btrfs_abort_transaction(trans, + root, + err); + goto error; + } + pending_del_nr = 0; + } + + err = truncate_inline_extent(inode, path, + &found_key, + item_end, + new_size); + if (err) { + btrfs_abort_transaction(trans, + root, err); + goto error; + } } else if (test_bit(BTRFS_ROOT_REF_COWS, &root->state)) { - inode_sub_bytes(inode, item_end + 1 - - found_key.offset); + inode_sub_bytes(inode, item_end + 1 - new_size); } } delete: @@ -4428,7 +4517,7 @@ delete: ret = btrfs_free_extent(trans, root, extent_start, extent_num_bytes, 0, btrfs_header_owner(leaf), - ino, extent_offset, 0); + ino, extent_offset); BUG_ON(ret); if (btrfs_should_throttle_delayed_refs(trans, root)) btrfs_async_run_delayed_refs(root, @@ -4493,8 +4582,7 @@ out: btrfs_abort_transaction(trans, root, ret); } error: - if (last_size != (u64)-1 && - root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) + if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) btrfs_ordered_update_i_size(inode, last_size, NULL); btrfs_free_path(path); @@ -4543,14 +4631,17 @@ int btrfs_truncate_page(struct inode *inode, loff_t from, loff_t len, if ((offset & (blocksize - 1)) == 0 && (!len || ((len & (blocksize - 1)) == 0))) goto out; - ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE); + ret = btrfs_delalloc_reserve_space(inode, + round_down(from, PAGE_CACHE_SIZE), PAGE_CACHE_SIZE); if (ret) goto out; again: page = find_or_create_page(mapping, index, mask); if (!page) { - btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE); + btrfs_delalloc_release_space(inode, + round_down(from, PAGE_CACHE_SIZE), + PAGE_CACHE_SIZE); ret = -ENOMEM; goto out; } @@ -4618,7 +4709,8 @@ again: out_unlock: if (ret) - btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE); + btrfs_delalloc_release_space(inode, page_start, + PAGE_CACHE_SIZE); unlock_page(page); page_cache_release(page); out: @@ -4986,24 +5078,53 @@ static void evict_inode_truncate_pages(struct inode *inode) } write_unlock(&map_tree->lock); + /* + * Keep looping until we have no more ranges in the io tree. + * We can have ongoing bios started by readpages (called from readahead) + * that have their endio callback (extent_io.c:end_bio_extent_readpage) + * still in progress (unlocked the pages in the bio but did not yet + * unlocked the ranges in the io tree). Therefore this means some + * ranges can still be locked and eviction started because before + * submitting those bios, which are executed by a separate task (work + * queue kthread), inode references (inode->i_count) were not taken + * (which would be dropped in the end io callback of each bio). + * Therefore here we effectively end up waiting for those bios and + * anyone else holding locked ranges without having bumped the inode's + * reference count - if we don't do it, when they access the inode's + * io_tree to unlock a range it may be too late, leading to an + * use-after-free issue. + */ spin_lock(&io_tree->lock); while (!RB_EMPTY_ROOT(&io_tree->state)) { struct extent_state *state; struct extent_state *cached_state = NULL; + u64 start; + u64 end; node = rb_first(&io_tree->state); state = rb_entry(node, struct extent_state, rb_node); - atomic_inc(&state->refs); + start = state->start; + end = state->end; spin_unlock(&io_tree->lock); - lock_extent_bits(io_tree, state->start, state->end, - 0, &cached_state); - clear_extent_bit(io_tree, state->start, state->end, + lock_extent_bits(io_tree, start, end, 0, &cached_state); + + /* + * If still has DELALLOC flag, the extent didn't reach disk, + * and its reserved space won't be freed by delayed_ref. + * So we need to free its reserved space here. + * (Refer to comment in btrfs_invalidatepage, case 2) + * + * Note, end is the bytenr of last byte, so we need + 1 here. + */ + if (state->state & EXTENT_DELALLOC) + btrfs_qgroup_free_data(inode, start, end - start + 1); + + clear_extent_bit(io_tree, start, end, EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 1, 1, &cached_state, GFP_NOFS); - free_extent_state(state); cond_resched(); spin_lock(&io_tree->lock); @@ -5035,7 +5156,8 @@ void btrfs_evict_inode(struct inode *inode) goto no_delete; } /* do we really want it for ->i_nlink > 0 and zero btrfs_root_refs? */ - btrfs_wait_ordered_range(inode, 0, (u64)-1); + if (!special_file(inode->i_mode)) + btrfs_wait_ordered_range(inode, 0, (u64)-1); btrfs_free_io_failure_record(inode, 0, (u64)-1); @@ -5615,6 +5737,7 @@ static int btrfs_real_readdir(struct file *file, struct dir_context *ctx) char *name_ptr; int name_len; int is_curr = 0; /* ctx->pos points to the current index? */ + bool emitted; /* FIXME, use a real flag for deciding about the key type */ if (root->fs_info->tree_root == root) @@ -5643,6 +5766,7 @@ static int btrfs_real_readdir(struct file *file, struct dir_context *ctx) if (ret < 0) goto err; + emitted = false; while (1) { leaf = path->nodes[0]; slot = path->slots[0]; @@ -5722,6 +5846,7 @@ skip: if (over) goto nopos; + emitted = true; di_len = btrfs_dir_name_len(leaf, di) + btrfs_dir_data_len(leaf, di) + sizeof(*di); di_cur += di_len; @@ -5734,11 +5859,20 @@ next: if (key_type == BTRFS_DIR_INDEX_KEY) { if (is_curr) ctx->pos++; - ret = btrfs_readdir_delayed_dir_index(ctx, &ins_list); + ret = btrfs_readdir_delayed_dir_index(ctx, &ins_list, &emitted); if (ret) goto nopos; } + /* + * If we haven't emitted any dir entry, we must not touch ctx->pos as + * it was was set to the termination value in previous call. We assume + * that "." and ".." were emitted if we reach this point and set the + * termination value as well for an empty directory. + */ + if (ctx->pos > 2 && !emitted) + goto nopos; + /* Reached end of directory/root. Bump pos past the last item. */ ctx->pos++; @@ -6218,9 +6352,6 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, u64 objectid; u64 index = 0; - if (!new_valid_dev(rdev)) - return -EINVAL; - /* * 2 for inode item and ref * 2 for dir items @@ -6358,7 +6489,7 @@ out_unlock_inode: static int btrfs_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) { - struct btrfs_trans_handle *trans; + struct btrfs_trans_handle *trans = NULL; struct btrfs_root *root = BTRFS_I(dir)->root; struct inode *inode = d_inode(old_dentry); u64 index; @@ -6384,6 +6515,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, trans = btrfs_start_transaction(root, 5); if (IS_ERR(trans)) { err = PTR_ERR(trans); + trans = NULL; goto fail; } @@ -6417,9 +6549,10 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, btrfs_log_new_name(trans, inode, NULL, parent); } - btrfs_end_transaction(trans, root); btrfs_balance_delayed_items(root); fail: + if (trans) + btrfs_end_transaction(trans, root); if (drop_inode) { inode_dec_link_count(inode); iput(inode); @@ -6860,8 +6993,7 @@ out: trace_btrfs_get_extent(root, em); - if (path) - btrfs_free_path(path); + btrfs_free_path(path); if (trans) { ret = btrfs_end_transaction(trans, root); if (!err) @@ -7360,6 +7492,32 @@ static struct extent_map *create_pinned_em(struct inode *inode, u64 start, return em; } +struct btrfs_dio_data { + u64 outstanding_extents; + u64 reserve; +}; + +static void adjust_dio_outstanding_extents(struct inode *inode, + struct btrfs_dio_data *dio_data, + const u64 len) +{ + unsigned num_extents; + + num_extents = (unsigned) div64_u64(len + BTRFS_MAX_EXTENT_SIZE - 1, + BTRFS_MAX_EXTENT_SIZE); + /* + * If we have an outstanding_extents count still set then we're + * within our reservation, otherwise we need to adjust our inode + * counter appropriately. + */ + if (dio_data->outstanding_extents) { + dio_data->outstanding_extents -= num_extents; + } else { + spin_lock(&BTRFS_I(inode)->lock); + BTRFS_I(inode)->outstanding_extents += num_extents; + spin_unlock(&BTRFS_I(inode)->lock); + } +} static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create) @@ -7367,10 +7525,10 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock, struct extent_map *em; struct btrfs_root *root = BTRFS_I(inode)->root; struct extent_state *cached_state = NULL; + struct btrfs_dio_data *dio_data = NULL; u64 start = iblock << inode->i_blkbits; u64 lockstart, lockend; u64 len = bh_result->b_size; - u64 *outstanding_extents = NULL; int unlock_bits = EXTENT_LOCKED; int ret = 0; @@ -7388,7 +7546,7 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock, * that anything that needs to check if there's a transction doesn't get * confused. */ - outstanding_extents = current->journal_info; + dio_data = current->journal_info; current->journal_info = NULL; } @@ -7396,8 +7554,11 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock, * If this errors out it's because we couldn't invalidate pagecache for * this range and we need to fallback to buffered. */ - if (lock_extent_direct(inode, lockstart, lockend, &cached_state, create)) - return -ENOTBLK; + if (lock_extent_direct(inode, lockstart, lockend, &cached_state, + create)) { + ret = -ENOTBLK; + goto err; + } em = btrfs_get_extent(inode, NULL, 0, start, len, 0); if (IS_ERR(em)) { @@ -7515,21 +7676,11 @@ unlock: if (start + len > i_size_read(inode)) i_size_write(inode, start + len); - /* - * If we have an outstanding_extents count still set then we're - * within our reservation, otherwise we need to adjust our inode - * counter appropriately. - */ - if (*outstanding_extents) { - (*outstanding_extents)--; - } else { - spin_lock(&BTRFS_I(inode)->lock); - BTRFS_I(inode)->outstanding_extents++; - spin_unlock(&BTRFS_I(inode)->lock); - } - - current->journal_info = outstanding_extents; - btrfs_free_reserved_data_space(inode, len); + adjust_dio_outstanding_extents(inode, dio_data, len); + btrfs_free_reserved_data_space(inode, start, len); + WARN_ON(dio_data->reserve < len); + dio_data->reserve -= len; + current->journal_info = dio_data; } /* @@ -7552,8 +7703,17 @@ unlock: unlock_err: clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend, unlock_bits, 1, 0, &cached_state, GFP_NOFS); - if (outstanding_extents) - current->journal_info = outstanding_extents; +err: + if (dio_data) + current->journal_info = dio_data; + /* + * Compensate the delalloc release we do in btrfs_direct_IO() when we + * write less data then expected, so that we don't underflow our inode's + * outstanding extents counter. + */ + if (create && dio_data) + adjust_dio_outstanding_extents(inode, dio_data, len); + return ret; } @@ -7671,13 +7831,13 @@ struct btrfs_retry_complete { int uptodate; }; -static void btrfs_retry_endio_nocsum(struct bio *bio, int err) +static void btrfs_retry_endio_nocsum(struct bio *bio) { struct btrfs_retry_complete *done = bio->bi_private; struct bio_vec *bvec; int i; - if (err) + if (bio->bi_error) goto end; done->uptodate = 1; @@ -7726,7 +7886,7 @@ try_again: return 0; } -static void btrfs_retry_endio(struct bio *bio, int err) +static void btrfs_retry_endio(struct bio *bio) { struct btrfs_retry_complete *done = bio->bi_private; struct btrfs_io_bio *io_bio = btrfs_io_bio(bio); @@ -7735,7 +7895,7 @@ static void btrfs_retry_endio(struct bio *bio, int err) int ret; int i; - if (err) + if (bio->bi_error) goto end; uptodate = 1; @@ -7818,12 +7978,13 @@ static int btrfs_subio_endio_read(struct inode *inode, } } -static void btrfs_endio_direct_read(struct bio *bio, int err) +static void btrfs_endio_direct_read(struct bio *bio) { struct btrfs_dio_private *dip = bio->bi_private; struct inode *inode = dip->inode; struct bio *dio_bio; struct btrfs_io_bio *io_bio = btrfs_io_bio(bio); + int err = bio->bi_error; if (dip->flags & BTRFS_DIO_ORIG_BIO_SUBMITTED) err = btrfs_subio_endio_read(inode, io_bio, err); @@ -7834,17 +7995,15 @@ static void btrfs_endio_direct_read(struct bio *bio, int err) kfree(dip); - /* If we had a csum failure make sure to clear the uptodate flag */ - if (err) - clear_bit(BIO_UPTODATE, &dio_bio->bi_flags); - dio_end_io(dio_bio, err); + dio_bio->bi_error = bio->bi_error; + dio_end_io(dio_bio, bio->bi_error); if (io_bio->end_io) io_bio->end_io(io_bio, err); bio_put(bio); } -static void btrfs_endio_direct_write(struct bio *bio, int err) +static void btrfs_endio_direct_write(struct bio *bio) { struct btrfs_dio_private *dip = bio->bi_private; struct inode *inode = dip->inode; @@ -7855,12 +8014,11 @@ static void btrfs_endio_direct_write(struct bio *bio, int err) struct bio *dio_bio; int ret; - if (err) - goto out_done; again: ret = btrfs_dec_test_first_ordered_pending(inode, &ordered, &ordered_offset, - ordered_bytes, !err); + ordered_bytes, + !bio->bi_error); if (!ret) goto out_test; @@ -7879,15 +8037,12 @@ out_test: ordered = NULL; goto again; } -out_done: dio_bio = dip->dio_bio; kfree(dip); - /* If we had an error make sure to clear the uptodate flag */ - if (err) - clear_bit(BIO_UPTODATE, &dio_bio->bi_flags); - dio_end_io(dio_bio, err); + dio_bio->bi_error = bio->bi_error; + dio_end_io(dio_bio, bio->bi_error); bio_put(bio); } @@ -7902,9 +8057,10 @@ static int __btrfs_submit_bio_start_direct_io(struct inode *inode, int rw, return 0; } -static void btrfs_end_dio_bio(struct bio *bio, int err) +static void btrfs_end_dio_bio(struct bio *bio) { struct btrfs_dio_private *dip = bio->bi_private; + int err = bio->bi_error; if (err) btrfs_warn(BTRFS_I(dip->inode)->root->fs_info, @@ -7933,8 +8089,8 @@ static void btrfs_end_dio_bio(struct bio *bio, int err) if (dip->errors) { bio_io_error(dip->orig_bio); } else { - set_bit(BIO_UPTODATE, &dip->dio_bio->bi_flags); - bio_endio(dip->orig_bio, 0); + dip->dio_bio->bi_error = 0; + bio_endio(dip->orig_bio); } out: bio_put(bio); @@ -7943,8 +8099,11 @@ out: static struct bio *btrfs_dio_bio_alloc(struct block_device *bdev, u64 first_sector, gfp_t gfp_flags) { - int nr_vecs = bio_get_nr_vecs(bdev); - return btrfs_bio_alloc(bdev, first_sector, nr_vecs, gfp_flags); + struct bio *bio; + bio = btrfs_bio_alloc(bdev, first_sector, BIO_MAX_PAGES, gfp_flags); + if (bio) + bio_associate_current(bio); + return bio; } static inline int btrfs_lookup_and_bind_dio_csum(struct btrfs_root *root, @@ -8147,9 +8306,8 @@ out_err: static void btrfs_submit_direct(int rw, struct bio *dio_bio, struct inode *inode, loff_t file_offset) { - struct btrfs_root *root = BTRFS_I(inode)->root; - struct btrfs_dio_private *dip; - struct bio *io_bio; + struct btrfs_dio_private *dip = NULL; + struct bio *io_bio = NULL; struct btrfs_io_bio *btrfs_bio; int skip_sum; int write = rw & REQ_WRITE; @@ -8166,7 +8324,7 @@ static void btrfs_submit_direct(int rw, struct bio *dio_bio, dip = kzalloc(sizeof(*dip), GFP_NOFS); if (!dip) { ret = -ENOMEM; - goto free_io_bio; + goto free_ordered; } dip->private = dio_bio->bi_private; @@ -8194,25 +8352,56 @@ static void btrfs_submit_direct(int rw, struct bio *dio_bio, if (btrfs_bio->end_io) btrfs_bio->end_io(btrfs_bio, ret); -free_io_bio: - bio_put(io_bio); free_ordered: /* - * If this is a write, we need to clean up the reserved space and kill - * the ordered extent. + * If we arrived here it means either we failed to submit the dip + * or we either failed to clone the dio_bio or failed to allocate the + * dip. If we cloned the dio_bio and allocated the dip, we can just + * call bio_endio against our io_bio so that we get proper resource + * cleanup if we fail to submit the dip, otherwise, we must do the + * same as btrfs_endio_direct_[write|read] because we can't call these + * callbacks - they require an allocated dip and a clone of dio_bio. */ - if (write) { - struct btrfs_ordered_extent *ordered; - ordered = btrfs_lookup_ordered_extent(inode, file_offset); - if (!test_bit(BTRFS_ORDERED_PREALLOC, &ordered->flags) && - !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags)) - btrfs_free_reserved_extent(root, ordered->start, - ordered->disk_len, 1); - btrfs_put_ordered_extent(ordered); - btrfs_put_ordered_extent(ordered); + if (io_bio && dip) { + io_bio->bi_error = -EIO; + bio_endio(io_bio); + /* + * The end io callbacks free our dip, do the final put on io_bio + * and all the cleanup and final put for dio_bio (through + * dio_end_io()). + */ + dip = NULL; + io_bio = NULL; + } else { + if (write) { + struct btrfs_ordered_extent *ordered; + + ordered = btrfs_lookup_ordered_extent(inode, + file_offset); + set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); + /* + * Decrements our ref on the ordered extent and removes + * the ordered extent from the inode's ordered tree, + * doing all the proper resource cleanup such as for the + * reserved space and waking up any waiters for this + * ordered extent (through btrfs_remove_ordered_extent). + */ + btrfs_finish_ordered_io(ordered); + } else { + unlock_extent(&BTRFS_I(inode)->io_tree, file_offset, + file_offset + dio_bio->bi_iter.bi_size - 1); + } + dio_bio->bi_error = -EIO; + /* + * Releases and cleans up our dio_bio, no need to bio_put() + * nor bio_endio()/bio_io_error() against dio_bio. + */ + dio_end_io(dio_bio, ret); } - bio_endio(dio_bio, ret); + if (io_bio) + bio_put(io_bio); + kfree(dip); } static ssize_t check_direct_IO(struct btrfs_root *root, struct kiocb *iocb, @@ -8253,7 +8442,8 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter, { struct file *file = iocb->ki_filp; struct inode *inode = file->f_mapping->host; - u64 outstanding_extents = 0; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_dio_data dio_data = { 0 }; size_t count = 0; int flags = 0; bool wakeup = true; @@ -8288,10 +8478,10 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter, mutex_unlock(&inode->i_mutex); relock = true; } - ret = btrfs_delalloc_reserve_space(inode, count); + ret = btrfs_delalloc_reserve_space(inode, offset, count); if (ret) goto out; - outstanding_extents = div64_u64(count + + dio_data.outstanding_extents = div64_u64(count + BTRFS_MAX_EXTENT_SIZE - 1, BTRFS_MAX_EXTENT_SIZE); @@ -8300,7 +8490,8 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter, * do the accounting properly if we go over the number we * originally calculated. Abuse current->journal_info for this. */ - current->journal_info = &outstanding_extents; + dio_data.reserve = round_up(count, root->sectorsize); + current->journal_info = &dio_data; } else if (test_bit(BTRFS_INODE_READDIO_NEED_LOCK, &BTRFS_I(inode)->runtime_flags)) { inode_dio_end(inode); @@ -8314,10 +8505,12 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter, btrfs_submit_direct, flags); if (iov_iter_rw(iter) == WRITE) { current->journal_info = NULL; - if (ret < 0 && ret != -EIOCBQUEUED) - btrfs_delalloc_release_space(inode, count); - else if (ret >= 0 && (size_t)ret < count) - btrfs_delalloc_release_space(inode, + if (ret < 0 && ret != -EIOCBQUEUED) { + if (dio_data.reserve) + btrfs_delalloc_release_space(inode, offset, + dio_data.reserve); + } else if (ret >= 0 && (size_t)ret < count) + btrfs_delalloc_release_space(inode, offset, count - (size_t)ret); } out: @@ -8353,15 +8546,28 @@ int btrfs_readpage(struct file *file, struct page *page) static int btrfs_writepage(struct page *page, struct writeback_control *wbc) { struct extent_io_tree *tree; - + struct inode *inode = page->mapping->host; + int ret; if (current->flags & PF_MEMALLOC) { redirty_page_for_writepage(wbc, page); unlock_page(page); return 0; } + + /* + * If we are under memory pressure we will call this directly from the + * VM, we need to make sure we have the inode referenced for the ordered + * extent. If not just return like we didn't do anything. + */ + if (!igrab(inode)) { + redirty_page_for_writepage(wbc, page); + return AOP_WRITEPAGE_ACTIVATE; + } tree = &BTRFS_I(page->mapping->host)->io_tree; - return extent_write_full_page(tree, page, btrfs_get_extent, wbc); + ret = extent_write_full_page(tree, page, btrfs_get_extent, wbc); + btrfs_add_delayed_iput(inode); + return ret; } static int btrfs_writepages(struct address_space *mapping, @@ -8476,6 +8682,18 @@ static void btrfs_invalidatepage(struct page *page, unsigned int offset, } } + /* + * Qgroup reserved space handler + * Page here will be either + * 1) Already written to disk + * In this case, its reserved space is released from data rsv map + * and will be freed by delayed_ref handler finally. + * So even we call qgroup_free_data(), it won't decrease reserved + * space. + * 2) Not written to disk + * This means the reserved space should be freed here. + */ + btrfs_qgroup_free_data(inode, page_start, PAGE_CACHE_SIZE); if (!inode_evicting) { clear_extent_bit(tree, page_start, page_end, EXTENT_LOCKED | EXTENT_DIRTY | @@ -8526,7 +8744,11 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) u64 page_end; sb_start_pagefault(inode->i_sb); - ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE); + page_start = page_offset(page); + page_end = page_start + PAGE_CACHE_SIZE - 1; + + ret = btrfs_delalloc_reserve_space(inode, page_start, + PAGE_CACHE_SIZE); if (!ret) { ret = file_update_time(vma->vm_file); reserved = 1; @@ -8545,8 +8767,6 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) again: lock_page(page); size = i_size_read(inode); - page_start = page_offset(page); - page_end = page_start + PAGE_CACHE_SIZE - 1; if ((page->mapping != inode->i_mapping) || (page_start >= size)) { @@ -8623,7 +8843,7 @@ out_unlock: } unlock_page(page); out: - btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE); + btrfs_delalloc_release_space(inode, page_start, PAGE_CACHE_SIZE); out_noreserve: sb_end_pagefault(inode->i_sb); return ret; @@ -8912,6 +9132,7 @@ void btrfs_destroy_inode(struct inode *inode) btrfs_put_ordered_extent(ordered); } } + btrfs_qgroup_check_reserved_leak(inode); inode_tree_del(inode); btrfs_drop_extent_cache(inode, 0, (u64)-1, 0); free: @@ -9440,9 +9661,11 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, /* * 2 items for inode item and ref * 2 items for dir items + * 1 item for updating parent inode item + * 1 item for the inline extent item * 1 item for xattr if selinux is on */ - trans = btrfs_start_transaction(root, 5); + trans = btrfs_start_transaction(root, 7); if (IS_ERR(trans)) return PTR_ERR(trans); @@ -9548,6 +9771,7 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode, u64 cur_offset = start; u64 i_size; u64 cur_bytes; + u64 last_alloc = (u64)-1; int ret = 0; bool own_trans = true; @@ -9564,6 +9788,13 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode, cur_bytes = min(num_bytes, 256ULL * 1024 * 1024); cur_bytes = max(cur_bytes, min_size); + /* + * If we are severely fragmented we could end up with really + * small allocations, so if the allocator is returning small + * chunks lets make its job easier by only searching for those + * sized chunks. + */ + cur_bytes = min(cur_bytes, last_alloc); ret = btrfs_reserve_extent(root, cur_bytes, min_size, 0, *alloc_hint, &ins, 1, 0); if (ret) { @@ -9572,6 +9803,7 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode, break; } + last_alloc = ins.offset; ret = insert_reserved_file_extent(trans, inode, cur_offset, ins.objectid, ins.offset, ins.offset, |