From patchwork Tue May 28 15:47:24 2013 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Jan Schmidt X-Patchwork-Id: 2625621 Return-Path: X-Original-To: patchwork-linux-btrfs@patchwork.kernel.org Delivered-To: patchwork-process-083081@patchwork1.kernel.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by patchwork1.kernel.org (Postfix) with ESMTP id AD48940077 for ; Tue, 28 May 2013 15:47:51 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934623Ab3E1Prr (ORCPT ); Tue, 28 May 2013 11:47:47 -0400 Received: from xp-ob.rzone.de ([81.169.146.140]:34986 "EHLO xp-ob.rzone.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S934599Ab3E1Prq (ORCPT ); Tue, 28 May 2013 11:47:46 -0400 X-RZG-CLASS-ID: xp Received: from pizpot.store ([192.168.43.236]) by jored.store (jored xp2) (RZmta 31.27 OK) with ESMTP id g03fe4p4S3rldU for ; Tue, 28 May 2013 17:47:44 +0200 (CEST) From: Jan Schmidt To: linux-btrfs@vger.kernel.org Subject: [PATCH 3/3] Btrfs: fix qgroup rescan resume on mount Date: Tue, 28 May 2013 17:47:24 +0200 Message-Id: <1369756044-17194-4-git-send-email-list.btrfs@jan-o-sch.net> X-Mailer: git-send-email 1.8.3 In-Reply-To: <1369756044-17194-1-git-send-email-list.btrfs@jan-o-sch.net> References: <1369756044-17194-1-git-send-email-list.btrfs@jan-o-sch.net> Sender: linux-btrfs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-btrfs@vger.kernel.org When called during mount, we cannot start the rescan worker thread until open_ctree is done. This commit restuctures the qgroup rescan internals to enable a clean deferral of the rescan resume operation. First of all, the struct qgroup_rescan is removed, saving us a malloc and some initialization synchronizations problems. Its only element (the worker struct) now lives within fs_info just as the rest of the rescan code. Then setting up a rescan worker is split into several reusable stages. Currently we have three different rescan startup scenarios: (A) rescan ioctl (B) rescan resume by mount (C) rescan by quota enable Each case needs its own combination of the four following steps: (1) set the progress [A, C: zero; B: state of umount] (2) commit the transaction [A] (3) set the counters [A, C: zero; B: state of umount] (4) start worker [A, B, C] qgroup_rescan_init does step (1). There's no extra function added to commit a transaction, we've got that already. qgroup_rescan_zero_tracking does step (3). Step (4) is nothing more than a call to the generic btrfs_queue_worker. We also get rid of a double check for the rescan progress during btrfs_qgroup_account_ref, which is no longer required due to having step 2 from the list above. As a side effect, this commit prepares to move the rescan start code from btrfs_run_qgroups (which is run during commit) to a less time critical section. Signed-off-by: Jan Schmidt --- fs/btrfs/ctree.h | 2 + fs/btrfs/disk-io.c | 2 + fs/btrfs/qgroup.c | 190 +++++++++++++++++++++++++++++++++------------------- 3 files changed, 125 insertions(+), 69 deletions(-) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index fd62aa8..8ac8d52 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1610,6 +1610,7 @@ struct btrfs_fs_info { struct btrfs_key qgroup_rescan_progress; struct btrfs_workers qgroup_rescan_workers; struct completion qgroup_rescan_completion; + struct btrfs_work qgroup_rescan_work; /* filesystem state */ unsigned long fs_state; @@ -3856,6 +3857,7 @@ int btrfs_quota_enable(struct btrfs_trans_handle *trans, int btrfs_quota_disable(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info); int btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info); +void btrfs_qgroup_rescan_resume(struct btrfs_fs_info *fs_info); int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info); int btrfs_add_qgroup_relation(struct btrfs_trans_handle *trans, struct btrfs_fs_info *fs_info, u64 src, u64 dst); diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index d7b46c6..da4a10c 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -2879,6 +2879,8 @@ retry_root_backup: return ret; } + btrfs_qgroup_rescan_resume(fs_info); + return 0; fail_qgroup: diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c index c6ce642..1280eff 100644 --- a/fs/btrfs/qgroup.c +++ b/fs/btrfs/qgroup.c @@ -98,13 +98,10 @@ struct btrfs_qgroup_list { struct btrfs_qgroup *member; }; -struct qgroup_rescan { - struct btrfs_work work; - struct btrfs_fs_info *fs_info; -}; - -static void qgroup_rescan_start(struct btrfs_fs_info *fs_info, - struct qgroup_rescan *qscan); +static int +qgroup_rescan_init(struct btrfs_fs_info *fs_info, u64 progress_objectid, + int init_flags); +static void qgroup_rescan_zero_tracking(struct btrfs_fs_info *fs_info); /* must be called with qgroup_ioctl_lock held */ static struct btrfs_qgroup *find_qgroup_rb(struct btrfs_fs_info *fs_info, @@ -255,6 +252,7 @@ int btrfs_read_qgroup_config(struct btrfs_fs_info *fs_info) int slot; int ret = 0; u64 flags = 0; + u64 rescan_progress = 0; if (!fs_info->quota_enabled) return 0; @@ -312,20 +310,7 @@ int btrfs_read_qgroup_config(struct btrfs_fs_info *fs_info) } fs_info->qgroup_flags = btrfs_qgroup_status_flags(l, ptr); - fs_info->qgroup_rescan_progress.objectid = - btrfs_qgroup_status_rescan(l, ptr); - if (fs_info->qgroup_flags & - BTRFS_QGROUP_STATUS_FLAG_RESCAN) { - struct qgroup_rescan *qscan = - kmalloc(sizeof(*qscan), GFP_NOFS); - if (!qscan) { - ret = -ENOMEM; - goto out; - } - fs_info->qgroup_rescan_progress.type = 0; - fs_info->qgroup_rescan_progress.offset = 0; - qgroup_rescan_start(fs_info, qscan); - } + rescan_progress = btrfs_qgroup_status_rescan(l, ptr); goto next1; } @@ -427,12 +412,16 @@ out: if (!(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_ON)) { fs_info->quota_enabled = 0; fs_info->pending_quota_state = 0; + } else if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN && + ret >= 0) { + ret = qgroup_rescan_init(fs_info, rescan_progress, 0); } btrfs_free_path(path); if (ret < 0) { ulist_free(fs_info->qgroup_ulist); fs_info->qgroup_ulist = NULL; + fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN; } return ret < 0 ? ret : 0; @@ -1449,14 +1438,7 @@ int btrfs_qgroup_account_ref(struct btrfs_trans_handle *trans, if (ret < 0) return ret; - mutex_lock(&fs_info->qgroup_rescan_lock); spin_lock(&fs_info->qgroup_lock); - if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) { - if (fs_info->qgroup_rescan_progress.objectid <= node->bytenr) { - ret = 0; - goto unlock; - } - } quota_root = fs_info->quota_root; if (!quota_root) @@ -1496,7 +1478,6 @@ int btrfs_qgroup_account_ref(struct btrfs_trans_handle *trans, unlock: spin_unlock(&fs_info->qgroup_lock); - mutex_unlock(&fs_info->qgroup_rescan_lock); ulist_free(roots); return ret; @@ -1544,9 +1525,12 @@ int btrfs_run_qgroups(struct btrfs_trans_handle *trans, fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_INCONSISTENT; if (!ret && start_rescan_worker) { - ret = btrfs_qgroup_rescan(fs_info); - if (ret) - pr_err("btrfs: start rescan quota failed: %d\n", ret); + ret = qgroup_rescan_init(fs_info, 0, 1); + if (!ret) { + qgroup_rescan_zero_tracking(fs_info); + btrfs_queue_worker(&fs_info->qgroup_rescan_workers, + &fs_info->qgroup_rescan_work); + } ret = 0; } @@ -1880,12 +1864,11 @@ void assert_qgroups_uptodate(struct btrfs_trans_handle *trans) * returns 1 when done, 2 when done and FLAG_INCONSISTENT was cleared. */ static int -qgroup_rescan_leaf(struct qgroup_rescan *qscan, struct btrfs_path *path, +qgroup_rescan_leaf(struct btrfs_fs_info *fs_info, struct btrfs_path *path, struct btrfs_trans_handle *trans, struct ulist *tmp, struct extent_buffer *scratch_leaf) { struct btrfs_key found; - struct btrfs_fs_info *fs_info = qscan->fs_info; struct ulist *roots = NULL; struct ulist_node *unode; struct ulist_iterator uiter; @@ -2013,11 +1996,10 @@ out: static void btrfs_qgroup_rescan_worker(struct btrfs_work *work) { - struct qgroup_rescan *qscan = container_of(work, struct qgroup_rescan, - work); + struct btrfs_fs_info *fs_info = container_of(work, struct btrfs_fs_info, + qgroup_rescan_work); struct btrfs_path *path; struct btrfs_trans_handle *trans = NULL; - struct btrfs_fs_info *fs_info = qscan->fs_info; struct ulist *tmp = NULL; struct extent_buffer *scratch_leaf = NULL; int err = -ENOMEM; @@ -2042,7 +2024,7 @@ static void btrfs_qgroup_rescan_worker(struct btrfs_work *work) if (!fs_info->quota_enabled) { err = -EINTR; } else { - err = qgroup_rescan_leaf(qscan, path, trans, + err = qgroup_rescan_leaf(fs_info, path, trans, tmp, scratch_leaf); } if (err > 0) @@ -2055,7 +2037,6 @@ out: kfree(scratch_leaf); ulist_free(tmp); btrfs_free_path(path); - kfree(qscan); mutex_lock(&fs_info->qgroup_rescan_lock); fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN; @@ -2078,46 +2059,70 @@ out: complete_all(&fs_info->qgroup_rescan_completion); } -static void -qgroup_rescan_start(struct btrfs_fs_info *fs_info, struct qgroup_rescan *qscan) -{ - memset(&qscan->work, 0, sizeof(qscan->work)); - qscan->work.func = btrfs_qgroup_rescan_worker; - qscan->fs_info = fs_info; - - pr_info("btrfs: qgroup scan started\n"); - btrfs_queue_worker(&fs_info->qgroup_rescan_workers, &qscan->work); -} - -int -btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info) +/* + * Checks that (a) no rescan is running and (b) quota is enabled. Allocates all + * memory required for the rescan context. + */ +static int +qgroup_rescan_init(struct btrfs_fs_info *fs_info, u64 progress_objectid, + int init_flags) { int ret = 0; - struct rb_node *n; - struct btrfs_qgroup *qgroup; - struct qgroup_rescan *qscan = kmalloc(sizeof(*qscan), GFP_NOFS); - if (!qscan) - return -ENOMEM; + if (!init_flags && + (!(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) || + !(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_ON))) { + ret = -EINVAL; + goto err; + } mutex_lock(&fs_info->qgroup_rescan_lock); spin_lock(&fs_info->qgroup_lock); - if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) - ret = -EINPROGRESS; - else if (!(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_ON)) - ret = -EINVAL; - if (ret) { - spin_unlock(&fs_info->qgroup_lock); - mutex_unlock(&fs_info->qgroup_rescan_lock); - kfree(qscan); - return ret; + + if (init_flags) { + if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) + ret = -EINPROGRESS; + else if (!(fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_ON)) + ret = -EINVAL; + + if (ret) { + spin_unlock(&fs_info->qgroup_lock); + mutex_unlock(&fs_info->qgroup_rescan_lock); + goto err; + } + + fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_RESCAN; } - fs_info->qgroup_flags |= BTRFS_QGROUP_STATUS_FLAG_RESCAN; memset(&fs_info->qgroup_rescan_progress, 0, sizeof(fs_info->qgroup_rescan_progress)); + fs_info->qgroup_rescan_progress.objectid = progress_objectid; + + spin_unlock(&fs_info->qgroup_lock); + mutex_unlock(&fs_info->qgroup_rescan_lock); + init_completion(&fs_info->qgroup_rescan_completion); + memset(&fs_info->qgroup_rescan_work, 0, + sizeof(fs_info->qgroup_rescan_work)); + fs_info->qgroup_rescan_work.func = btrfs_qgroup_rescan_worker; + + if (ret) { +err: + pr_info("btrfs: qgroup_rescan_init failed with %d\n", ret); + return ret; + } + + return 0; +} + +static void +qgroup_rescan_zero_tracking(struct btrfs_fs_info *fs_info) +{ + struct rb_node *n; + struct btrfs_qgroup *qgroup; + + spin_lock(&fs_info->qgroup_lock); /* clear all current qgroup tracking information */ for (n = rb_first(&fs_info->qgroup_tree); n; n = rb_next(n)) { qgroup = rb_entry(n, struct btrfs_qgroup, node); @@ -2127,9 +2132,44 @@ btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info) qgroup->excl_cmpr = 0; } spin_unlock(&fs_info->qgroup_lock); - mutex_unlock(&fs_info->qgroup_rescan_lock); +} - qgroup_rescan_start(fs_info, qscan); +int +btrfs_qgroup_rescan(struct btrfs_fs_info *fs_info) +{ + int ret = 0; + struct btrfs_trans_handle *trans; + + ret = qgroup_rescan_init(fs_info, 0, 1); + if (ret) + return ret; + + /* + * We have set the rescan_progress to 0, which means no more + * delayed refs will be accounted by btrfs_qgroup_account_ref. + * However, btrfs_qgroup_account_ref may be right after its call + * to btrfs_find_all_roots, in which case it would still do the + * accounting. + * To solve this, we're committing the transaction, which will + * ensure we run all delayed refs and only after that, we are + * going to clear all tracking information for a clean start. + */ + + trans = btrfs_join_transaction(fs_info->fs_root); + if (IS_ERR(trans)) { + fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN; + return PTR_ERR(trans); + } + ret = btrfs_commit_transaction(trans, fs_info->fs_root); + if (ret) { + fs_info->qgroup_flags &= ~BTRFS_QGROUP_STATUS_FLAG_RESCAN; + return ret; + } + + qgroup_rescan_zero_tracking(fs_info); + + btrfs_queue_worker(&fs_info->qgroup_rescan_workers, + &fs_info->qgroup_rescan_work); return 0; } @@ -2151,3 +2191,15 @@ int btrfs_qgroup_wait_for_completion(struct btrfs_fs_info *fs_info) return ret; } + +/* + * this is only called from open_ctree where we're still single threaded, thus + * locking is omitted here. + */ +void +btrfs_qgroup_rescan_resume(struct btrfs_fs_info *fs_info) +{ + if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) + btrfs_queue_worker(&fs_info->qgroup_rescan_workers, + &fs_info->qgroup_rescan_work); +}