From patchwork Wed Jul 27 05:50:21 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Xiaoguang Wang X-Patchwork-Id: 9249269 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id 39C62607F0 for ; Wed, 27 Jul 2016 05:54:57 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 2B5D227BF8 for ; Wed, 27 Jul 2016 05:54:57 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 1F33927BFC; Wed, 27 Jul 2016 05:54:57 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-6.9 required=2.0 tests=BAYES_00,RCVD_IN_DNSWL_HI autolearn=ham version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 0158E27BF8 for ; Wed, 27 Jul 2016 05:54:55 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756736AbcG0Fyn (ORCPT ); Wed, 27 Jul 2016 01:54:43 -0400 Received: from cn.fujitsu.com ([222.73.24.84]:8360 "EHLO song.cn.fujitsu.com" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1751627AbcG0Fym (ORCPT ); Wed, 27 Jul 2016 01:54:42 -0400 X-IronPort-AV: E=Sophos;i="5.20,367,1444665600"; d="scan'208";a="693585" Received: from unknown (HELO cn.fujitsu.com) ([10.167.250.3]) by song.cn.fujitsu.com with ESMTP; 27 Jul 2016 13:53:13 +0800 Received: from localhost.localdomain (unknown [10.167.226.107]) by cn.fujitsu.com (Postfix) with ESMTP id B8EEC4056403; Wed, 27 Jul 2016 13:53:12 +0800 (CST) From: Wang Xiaoguang To: linux-btrfs@vger.kernel.org Cc: jbacik@fb.com, dsterba@suse.cz Subject: [PATCH v4] btrfs: should block unused block groups deletion work when allocating data space Date: Wed, 27 Jul 2016 13:50:21 +0800 Message-Id: <20160727055021.18888-1-wangxg.fnst@cn.fujitsu.com> X-Mailer: git-send-email 2.9.0 In-Reply-To: <57980DBE.9050702@cn.fujitsu.com> References: <57980DBE.9050702@cn.fujitsu.com> MIME-Version: 1.0 X-yoursite-MailScanner-ID: B8EEC4056403.A7566 X-yoursite-MailScanner: Found to be clean X-yoursite-MailScanner-From: wangxg.fnst@cn.fujitsu.com Sender: linux-btrfs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-btrfs@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP cleaner_kthread() may run at any time, in which it'll call btrfs_delete_unused_bgs() to delete unused block groups. Because this work is asynchronous, it may also result in false ENOSPC error. Please see below race window: CPU1 | CPU2 | |-> btrfs_alloc_data_chunk_ondemand() |-> cleaner_kthread() |-> do_chunk_alloc() | | | assume it returns ENOSPC, which means | | | btrfs_space_info is full and have free| | | space to satisfy data request. | | | | |- > btrfs_delete_unused_bgs() | | | it will decrease btrfs_space_info | | | total_bytes and make | | | btrfs_space_info is not full. | | | In this case, we may get ENOSPC error, but btrfs_space_info is not full. To fix this issue, in btrfs_alloc_data_chunk_ondemand(), if we need to call do_chunk_alloc() to allocating new chunk, we should block btrfs_delete_unused_bgs(). Here we introduce a new struct rw_semaphore bg_delete_sem to do this job. Indeed there is already a "struct mutex delete_unused_bgs_mutex", but it's mutex, we can not use it for this purpose. Of course, we can re-define it to be struct rw_semaphore, then use it in btrfs_alloc_data_chunk_ondemand(). Either method will work. But given that delete_unused_bgs_mutex's name length is longer than bg_delete_sem, I choose the first method, to create a new struct rw_semaphore bg_delete_sem and delete delete_unused_bgs_mutex :) Signed-off-by: Wang Xiaoguang --- v3: improve one code logic suggested by Josef, thanks. v4: replace delete_unused_bgs_mutex with our new bg_delete_sem, also suggested by Josef. --- fs/btrfs/ctree.h | 2 +- fs/btrfs/disk-io.c | 13 ++++++------- fs/btrfs/extent-tree.c | 42 ++++++++++++++++++++++++++++++++++-------- fs/btrfs/volumes.c | 42 +++++++++++++++++++++--------------------- 4 files changed, 62 insertions(+), 37 deletions(-) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 7eb2913..90041a2 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -800,6 +800,7 @@ struct btrfs_fs_info { struct mutex cleaner_mutex; struct mutex chunk_mutex; struct mutex volume_mutex; + struct rw_semaphore bg_delete_sem; /* * this is taken to make sure we don't set block groups ro after @@ -1079,7 +1080,6 @@ struct btrfs_fs_info { spinlock_t unused_bgs_lock; struct list_head unused_bgs; struct mutex unused_bg_unpin_mutex; - struct mutex delete_unused_bgs_mutex; /* For btrfs to record security options */ struct security_mnt_opts security_opts; diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 60ce119..86cad9a 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1839,12 +1839,11 @@ static int cleaner_kthread(void *arg) btrfs_run_defrag_inodes(root->fs_info); /* - * Acquires fs_info->delete_unused_bgs_mutex to avoid racing - * with relocation (btrfs_relocate_chunk) and relocation - * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) - * after acquiring fs_info->delete_unused_bgs_mutex. So we - * can't hold, nor need to, fs_info->cleaner_mutex when deleting - * unused block groups. + * Acquires fs_info->bg_delete_sem to avoid racing with + * relocation (btrfs_relocate_chunk) and relocation acquires + * fs_info->cleaner_mutex (btrfs_relocate_block_group) after + * acquiring fs_info->bg_delete_sem. So we can't hold, nor need + * to, fs_info->cleaner_mutex when deleting unused block groups. */ btrfs_delete_unused_bgs(root->fs_info); sleep: @@ -2595,7 +2594,6 @@ int open_ctree(struct super_block *sb, spin_lock_init(&fs_info->unused_bgs_lock); rwlock_init(&fs_info->tree_mod_log_lock); mutex_init(&fs_info->unused_bg_unpin_mutex); - mutex_init(&fs_info->delete_unused_bgs_mutex); mutex_init(&fs_info->reloc_mutex); mutex_init(&fs_info->delalloc_root_mutex); mutex_init(&fs_info->cleaner_delayed_iput_mutex); @@ -2683,6 +2681,7 @@ int open_ctree(struct super_block *sb, init_rwsem(&fs_info->commit_root_sem); init_rwsem(&fs_info->cleanup_work_sem); init_rwsem(&fs_info->subvol_sem); + init_rwsem(&fs_info->bg_delete_sem); sema_init(&fs_info->uuid_tree_rescan_sem, 1); btrfs_init_dev_replace_locks(fs_info); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index df8d756..19d4bb1 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -4111,6 +4111,7 @@ int btrfs_alloc_data_chunk_ondemand(struct inode *inode, u64 bytes) int ret = 0; int need_commit = 2; int have_pinned_space; + int have_bg_delete_sem = 0; /* make sure bytes are sectorsize aligned */ bytes = ALIGN(bytes, root->sectorsize); @@ -4121,8 +4122,11 @@ int btrfs_alloc_data_chunk_ondemand(struct inode *inode, u64 bytes) } data_sinfo = fs_info->data_sinfo; - if (!data_sinfo) + if (!data_sinfo) { + down_read(&root->fs_info->bg_delete_sem); + have_bg_delete_sem = 1; goto alloc; + } again: /* make sure we have enough space to handle the data first */ @@ -4135,6 +4139,17 @@ again: struct btrfs_trans_handle *trans; /* + * We may need to allocate new chunk, so we should block + * btrfs_delete_unused_bgs() + */ + if (!have_bg_delete_sem) { + spin_unlock(&data_sinfo->lock); + down_read(&root->fs_info->bg_delete_sem); + have_bg_delete_sem = 1; + goto again; + } + + /* * if we don't have enough free bytes in this space then we need * to alloc a new chunk. */ @@ -4156,17 +4171,20 @@ alloc: * the fs. */ trans = btrfs_join_transaction(root); - if (IS_ERR(trans)) + if (IS_ERR(trans)) { + up_read(&root->fs_info->bg_delete_sem); return PTR_ERR(trans); + } ret = do_chunk_alloc(trans, root->fs_info->extent_root, alloc_target, CHUNK_ALLOC_NO_FORCE); btrfs_end_transaction(trans, root); if (ret < 0) { - if (ret != -ENOSPC) + if (ret != -ENOSPC) { + up_read(&root->fs_info->bg_delete_sem); return ret; - else { + } else { have_pinned_space = 1; goto commit_trans; } @@ -4200,15 +4218,19 @@ commit_trans: } trans = btrfs_join_transaction(root); - if (IS_ERR(trans)) + if (IS_ERR(trans)) { + up_read(&root->fs_info->bg_delete_sem); return PTR_ERR(trans); + } if (have_pinned_space >= 0 || test_bit(BTRFS_TRANS_HAVE_FREE_BGS, &trans->transaction->flags) || need_commit > 0) { ret = btrfs_commit_transaction(trans, root); - if (ret) + if (ret) { + up_read(&root->fs_info->bg_delete_sem); return ret; + } /* * The cleaner kthread might still be doing iput * operations. Wait for it to finish so that @@ -4225,6 +4247,7 @@ commit_trans: trace_btrfs_space_reservation(root->fs_info, "space_info:enospc", data_sinfo->flags, bytes, 1); + up_read(&root->fs_info->bg_delete_sem); return -ENOSPC; } data_sinfo->bytes_may_use += bytes; @@ -4232,6 +4255,9 @@ commit_trans: data_sinfo->flags, bytes, 1); spin_unlock(&data_sinfo->lock); + if (have_bg_delete_sem) + up_read(&root->fs_info->bg_delete_sem); + return ret; } @@ -10593,7 +10619,7 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) } spin_unlock(&fs_info->unused_bgs_lock); - mutex_lock(&fs_info->delete_unused_bgs_mutex); + down_write(&root->fs_info->bg_delete_sem); /* Don't want to race with allocators so take the groups_sem */ down_write(&space_info->groups_sem); @@ -10721,7 +10747,7 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info) end_trans: btrfs_end_transaction(trans, root); next: - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + up_write(&root->fs_info->bg_delete_sem); btrfs_put_block_group(block_group); spin_lock(&fs_info->unused_bgs_lock); } diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 589f128..7dbf70a 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -2876,7 +2876,7 @@ static int btrfs_relocate_chunk(struct btrfs_root *root, u64 chunk_offset) * we release the path used to search the chunk/dev tree and before * the current task acquires this mutex and calls us. */ - ASSERT(mutex_is_locked(&root->fs_info->delete_unused_bgs_mutex)); + ASSERT(rwsem_is_locked(&root->fs_info->bg_delete_sem)); ret = btrfs_can_relocate(extent_root, chunk_offset); if (ret) @@ -2929,10 +2929,10 @@ again: key.type = BTRFS_CHUNK_ITEM_KEY; while (1) { - mutex_lock(&root->fs_info->delete_unused_bgs_mutex); + down_read(&root->fs_info->bg_delete_sem); ret = btrfs_search_slot(NULL, chunk_root, &key, path, 0, 0); if (ret < 0) { - mutex_unlock(&root->fs_info->delete_unused_bgs_mutex); + up_read(&root->fs_info->bg_delete_sem); goto error; } BUG_ON(ret == 0); /* Corruption */ @@ -2940,7 +2940,7 @@ again: ret = btrfs_previous_item(chunk_root, path, key.objectid, key.type); if (ret) - mutex_unlock(&root->fs_info->delete_unused_bgs_mutex); + up_read(&root->fs_info->bg_delete_sem); if (ret < 0) goto error; if (ret > 0) @@ -2962,7 +2962,7 @@ again: else BUG_ON(ret); } - mutex_unlock(&root->fs_info->delete_unused_bgs_mutex); + up_read(&root->fs_info->bg_delete_sem); if (found_key.offset == 0) break; @@ -3498,10 +3498,10 @@ again: goto error; } - mutex_lock(&fs_info->delete_unused_bgs_mutex); + down_read(&fs_info->bg_delete_sem); ret = btrfs_search_slot(NULL, chunk_root, &key, path, 0, 0); if (ret < 0) { - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + up_read(&fs_info->bg_delete_sem); goto error; } @@ -3515,7 +3515,7 @@ again: ret = btrfs_previous_item(chunk_root, path, 0, BTRFS_CHUNK_ITEM_KEY); if (ret) { - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + up_read(&fs_info->bg_delete_sem); ret = 0; break; } @@ -3525,7 +3525,7 @@ again: btrfs_item_key_to_cpu(leaf, &found_key, slot); if (found_key.objectid != key.objectid) { - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + up_read(&fs_info->bg_delete_sem); break; } @@ -3543,12 +3543,12 @@ again: btrfs_release_path(path); if (!ret) { - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + up_read(&fs_info->bg_delete_sem); goto loop; } if (counting) { - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + up_read(&fs_info->bg_delete_sem); spin_lock(&fs_info->balance_lock); bctl->stat.expected++; spin_unlock(&fs_info->balance_lock); @@ -3573,7 +3573,7 @@ again: count_meta < bctl->meta.limit_min) || ((chunk_type & BTRFS_BLOCK_GROUP_SYSTEM) && count_sys < bctl->sys.limit_min)) { - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + up_read(&fs_info->bg_delete_sem); goto loop; } @@ -3586,7 +3586,7 @@ again: !chunk_reserved && !bytes_used) { trans = btrfs_start_transaction(chunk_root, 0); if (IS_ERR(trans)) { - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + up_read(&fs_info->bg_delete_sem); ret = PTR_ERR(trans); goto error; } @@ -3595,7 +3595,7 @@ again: BTRFS_BLOCK_GROUP_DATA); btrfs_end_transaction(trans, chunk_root); if (ret < 0) { - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + up_read(&fs_info->bg_delete_sem); goto error; } chunk_reserved = 1; @@ -3603,7 +3603,7 @@ again: ret = btrfs_relocate_chunk(chunk_root, found_key.offset); - mutex_unlock(&fs_info->delete_unused_bgs_mutex); + up_read(&fs_info->bg_delete_sem); if (ret && ret != -ENOSPC) goto error; if (ret == -ENOSPC) { @@ -4330,16 +4330,16 @@ again: key.type = BTRFS_DEV_EXTENT_KEY; do { - mutex_lock(&root->fs_info->delete_unused_bgs_mutex); + down_read(&root->fs_info->bg_delete_sem); ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) { - mutex_unlock(&root->fs_info->delete_unused_bgs_mutex); + up_read(&root->fs_info->bg_delete_sem); goto done; } ret = btrfs_previous_item(root, path, 0, key.type); if (ret) - mutex_unlock(&root->fs_info->delete_unused_bgs_mutex); + up_read(&root->fs_info->bg_delete_sem); if (ret < 0) goto done; if (ret) { @@ -4353,7 +4353,7 @@ again: btrfs_item_key_to_cpu(l, &key, path->slots[0]); if (key.objectid != device->devid) { - mutex_unlock(&root->fs_info->delete_unused_bgs_mutex); + up_read(&root->fs_info->bg_delete_sem); btrfs_release_path(path); break; } @@ -4362,7 +4362,7 @@ again: length = btrfs_dev_extent_length(l, dev_extent); if (key.offset + length <= new_size) { - mutex_unlock(&root->fs_info->delete_unused_bgs_mutex); + up_read(&root->fs_info->bg_delete_sem); btrfs_release_path(path); break; } @@ -4371,7 +4371,7 @@ again: btrfs_release_path(path); ret = btrfs_relocate_chunk(root, chunk_offset); - mutex_unlock(&root->fs_info->delete_unused_bgs_mutex); + up_read(&root->fs_info->bg_delete_sem); if (ret && ret != -ENOSPC) goto done; if (ret == -ENOSPC)