From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mx0a-00082601.pphosted.com ([67.231.145.42]:32706 "EHLO mx0a-00082601.pphosted.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752198AbaIPUcS (ORCPT ); Tue, 16 Sep 2014 16:32:18 -0400 Received: from pps.filterd (m0004347 [127.0.0.1]) by m0004347.ppops.net (8.14.5/8.14.5) with SMTP id s8GKW8RY026582 for ; Tue, 16 Sep 2014 13:32:18 -0700 Received: from mail.thefacebook.com (mailwest.thefacebook.com [173.252.71.148]) by m0004347.ppops.net with ESMTP id 1peevq9jma-1 (version=TLSv1/SSLv3 cipher=AES128-SHA bits=128 verify=OK) for ; Tue, 16 Sep 2014 13:32:18 -0700 From: Josef Bacik To: Subject: [PATCH] Btrfs: try not to ENOSPC on log replay Date: Tue, 16 Sep 2014 16:32:15 -0400 Message-ID: <1410899535-20363-1-git-send-email-jbacik@fb.com> MIME-Version: 1.0 Content-Type: text/plain Sender: linux-btrfs-owner@vger.kernel.org List-ID: When doing log replay we may have to update inodes, which traditionally goes through our delayed inode stuff. This will try to move space over from the trans handle, but we don't reserve space in our trans handle on replay since we don't know how much we will need, so instead we try to flush. But because we have a trans handle open we won't flush anything, so if we are out of reserve space we will simply return ENOSPC. Since we know that if an operation made it into the log then we definitely had space before the box bought the farm then we don't need to worry about doing this space reservation. Just set a flag saying we are doing log replay and update the tree directly and bypass the reservation code altogether. This will keep us from ENOSP'ing during log replay. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/inode.c | 3 ++- fs/btrfs/transaction.c | 1 + fs/btrfs/transaction.h | 1 + fs/btrfs/tree-log.c | 1 + 4 files changed, 5 insertions(+), 1 deletion(-) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 28e693e..dfdc656 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -3705,7 +3705,8 @@ noinline int btrfs_update_inode(struct btrfs_trans_handle *trans, * without delay */ if (!btrfs_is_free_space_inode(inode) - && root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) { + && root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID + && !trans->log_replay) { btrfs_update_root_times(trans, root); ret = btrfs_delayed_update_inode(trans, root, inode); diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 16d0c1b..1676c5d 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -486,6 +486,7 @@ again: h->allocating_chunk = false; h->reloc_reserved = false; h->sync = false; + h->log_replay = false; INIT_LIST_HEAD(&h->qgroup_ref_list); INIT_LIST_HEAD(&h->new_bgs); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 579be51..924cd1f 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -97,6 +97,7 @@ struct btrfs_trans_handle { bool allocating_chunk; bool reloc_reserved; bool sync; + bool log_replay; unsigned int type; /* * this root is only needed to validate that the root passed to diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c index dce33b5..ba31dd9 100644 --- a/fs/btrfs/tree-log.c +++ b/fs/btrfs/tree-log.c @@ -4353,6 +4353,7 @@ int btrfs_recover_log_trees(struct btrfs_root *log_root_tree) ret = PTR_ERR(trans); goto error; } + trans->log_replay = true; wc.trans = trans; wc.pin = 1; -- 1.8.3.1