diff options
author | Darrick J. Wong <darrick.wong@oracle.com> | 2020-10-25 17:14:52 -0700 |
---|---|---|
committer | Darrick J. Wong <darrick.wong@oracle.com> | 2020-10-26 18:32:20 -0700 |
commit | d483a176342d5a1c0babd6332c120ca106a5894c (patch) | |
tree | c9c46df2ced626119e6b59ab82f5adb7478192f5 | |
parent | 15b5ccb2fdcb0eb7388fe87089c935a78138dbd9 (diff) |
xfs: parallelize inode inactivation
Split the inode inactivation work into per-AG work items so that we can
take advantage of parallelization.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
-rw-r--r-- | fs/xfs/scrub/common.c | 2 | ||||
-rw-r--r-- | fs/xfs/scrub/quotacheck.c | 2 | ||||
-rw-r--r-- | fs/xfs/xfs_icache.c | 92 | ||||
-rw-r--r-- | fs/xfs/xfs_icache.h | 3 | ||||
-rw-r--r-- | fs/xfs/xfs_mount.c | 3 | ||||
-rw-r--r-- | fs/xfs/xfs_mount.h | 4 | ||||
-rw-r--r-- | fs/xfs/xfs_super.c | 5 |
7 files changed, 94 insertions, 17 deletions
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c index a0b4ec115263..b26e792bb459 100644 --- a/fs/xfs/scrub/common.c +++ b/fs/xfs/scrub/common.c @@ -922,6 +922,7 @@ xchk_stop_reaping( { sc->flags |= XCHK_REAPING_DISABLED; xfs_blockgc_stop(sc->mp); + xfs_inactive_cancel_work(sc->mp); } /* Restart background reaping of resources. */ @@ -929,6 +930,7 @@ void xchk_start_reaping( struct xfs_scrub *sc) { + xfs_inactive_schedule_now(sc->mp); xfs_blockgc_start(sc->mp); sc->flags &= ~XCHK_REAPING_DISABLED; } diff --git a/fs/xfs/scrub/quotacheck.c b/fs/xfs/scrub/quotacheck.c index c18663b005a1..b6baf3db342a 100644 --- a/fs/xfs/scrub/quotacheck.c +++ b/fs/xfs/scrub/quotacheck.c @@ -572,7 +572,7 @@ xqcheck_iwalk( * links but hasn't yet been inactivated. Try to push * it towards inactivation. */ - xfs_inactive_inodes(xqc->sc->mp, NULL); + xfs_inactive_force_ino(xqc->sc->mp, ino); /* fall through */ case -EINVAL: /* diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c index 8939fc6bf1fe..426e63c7718a 100644 --- a/fs/xfs/xfs_icache.c +++ b/fs/xfs/xfs_icache.c @@ -1962,12 +1962,12 @@ xfs_inode_clear_cowblocks_tag( /* Queue a new inode inactivation pass if there are reclaimable inodes. */ static void xfs_inactive_work_queue( - struct xfs_mount *mp) + struct xfs_perag *pag) { rcu_read_lock(); - if (radix_tree_tagged(&mp->m_perag_tree, XFS_ICI_INACTIVE_TAG)) - queue_delayed_work(mp->m_inactive_workqueue, - &mp->m_inactive_work, + if (radix_tree_tagged(&pag->pag_ici_root, XFS_ICI_INACTIVE_TAG)) + queue_delayed_work(pag->pag_mount->m_inactive_workqueue, + &pag->pag_inactive_work, msecs_to_jiffies(xfs_syncd_centisecs / 6 * 10)); rcu_read_unlock(); } @@ -1990,7 +1990,7 @@ xfs_perag_set_inactive_tag( spin_unlock(&mp->m_perag_lock); /* schedule periodic background inode inactivation */ - xfs_inactive_work_queue(mp); + xfs_inactive_work_queue(pag); trace_xfs_perag_set_inactive(mp, pag->pag_agno, -1, _RET_IP_); } @@ -2123,8 +2123,9 @@ void xfs_inactive_worker( struct work_struct *work) { - struct xfs_mount *mp = container_of(to_delayed_work(work), - struct xfs_mount, m_inactive_work); + struct xfs_perag *pag = container_of(to_delayed_work(work), + struct xfs_perag, pag_inactive_work); + struct xfs_mount *mp = pag->pag_mount; int error; /* @@ -2139,12 +2140,32 @@ xfs_inactive_worker( if (!sb_start_write_trylock(mp->m_super)) return; - error = xfs_inactive_inodes(mp, NULL); + error = xfs_inode_walk_ag(pag, XFS_INODE_WALK_INACTIVE, + xfs_inactive_inode, NULL, XFS_ICI_INACTIVE_TAG); if (error && error != -EAGAIN) xfs_err(mp, "inode inactivation failed, error %d", error); sb_end_write(mp->m_super); - xfs_inactive_work_queue(mp); + xfs_inactive_work_queue(pag); +} + +/* Wait for all background inactivation work to finish. */ +static void +xfs_inactive_flush( + struct xfs_mount *mp) +{ + struct xfs_perag *pag; + xfs_agnumber_t agno; + + for_each_perag_tag(mp, agno, pag, XFS_ICI_INACTIVE_TAG) { + bool flush; + + spin_lock(&pag->pag_ici_lock); + flush = pag->pag_ici_inactive > 0; + spin_unlock(&pag->pag_ici_lock); + if (flush) + flush_delayed_work(&pag->pag_inactive_work); + } } /* Flush all inode inactivation work that might be queued. */ @@ -2152,8 +2173,8 @@ void xfs_inactive_force( struct xfs_mount *mp) { - queue_delayed_work(mp->m_inactive_workqueue, &mp->m_inactive_work, 0); - flush_delayed_work(&mp->m_inactive_work); + xfs_inactive_schedule_now(mp); + xfs_inactive_flush(mp); } /* @@ -2165,9 +2186,54 @@ void xfs_inactive_shutdown( struct xfs_mount *mp) { - cancel_delayed_work_sync(&mp->m_inactive_work); - flush_workqueue(mp->m_inactive_workqueue); + xfs_inactive_cancel_work(mp); xfs_inactive_inodes(mp, NULL); cancel_delayed_work_sync(&mp->m_reclaim_work); xfs_reclaim_inodes(mp); } + +/* Force the inactivation of this inode. */ +void +xfs_inactive_force_ino( + struct xfs_mount *mp, + xfs_ino_t ino) +{ + struct xfs_perag *pag; + + pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ino)); + xfs_inode_walk_ag(pag, XFS_INODE_WALK_INACTIVE, xfs_inactive_inode, + NULL, XFS_ICI_INACTIVE_TAG); + xfs_perag_put(pag); +} + +/* Cancel all queued inactivation work. */ +void +xfs_inactive_cancel_work( + struct xfs_mount *mp) +{ + struct xfs_perag *pag; + xfs_agnumber_t agno; + + for_each_perag_tag(mp, agno, pag, XFS_ICI_INACTIVE_TAG) + cancel_delayed_work_sync(&pag->pag_inactive_work); + flush_workqueue(mp->m_inactive_workqueue); +} + +/* Cancel all pending deferred inactivation work and reschedule it now. */ +void +xfs_inactive_schedule_now( + struct xfs_mount *mp) +{ + struct xfs_perag *pag; + xfs_agnumber_t agno; + + for_each_perag_tag(mp, agno, pag, XFS_ICI_INACTIVE_TAG) { + spin_lock(&pag->pag_ici_lock); + if (pag->pag_ici_inactive) { + cancel_delayed_work(&pag->pag_inactive_work); + queue_delayed_work(mp->m_inactive_workqueue, + &pag->pag_inactive_work, 0); + } + spin_unlock(&pag->pag_ici_lock); + } +} diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h index 90425c7f2be4..b633b358685a 100644 --- a/fs/xfs/xfs_icache.h +++ b/fs/xfs/xfs_icache.h @@ -77,6 +77,9 @@ void xfs_blockgc_start(struct xfs_mount *mp); void xfs_inactive_worker(struct work_struct *work); int xfs_inactive_inodes(struct xfs_mount *mp, struct xfs_eofblocks *eofb); void xfs_inactive_force(struct xfs_mount *mp); +void xfs_inactive_force_ino(struct xfs_mount *mp, xfs_ino_t ino); void xfs_inactive_shutdown(struct xfs_mount *mp); +void xfs_inactive_cancel_work(struct xfs_mount *mp); +void xfs_inactive_schedule_now(struct xfs_mount *mp); #endif diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c index 573990c0f30e..01a05ec822a5 100644 --- a/fs/xfs/xfs_mount.c +++ b/fs/xfs/xfs_mount.c @@ -127,6 +127,7 @@ __xfs_free_perag( struct xfs_perag *pag = container_of(head, struct xfs_perag, rcu_head); ASSERT(!delayed_work_pending(&pag->pag_blockgc_work)); + ASSERT(!delayed_work_pending(&pag->pag_inactive_work)); ASSERT(atomic_read(&pag->pag_ref) == 0); kmem_free(pag); } @@ -148,6 +149,7 @@ xfs_free_perag( ASSERT(pag); ASSERT(atomic_read(&pag->pag_ref) == 0); cancel_delayed_work_sync(&pag->pag_blockgc_work); + cancel_delayed_work_sync(&pag->pag_inactive_work); xfs_iunlink_destroy(pag); xfs_buf_hash_destroy(pag); call_rcu(&pag->rcu_head, __xfs_free_perag); @@ -202,6 +204,7 @@ xfs_initialize_perag( pag->pag_mount = mp; spin_lock_init(&pag->pag_ici_lock); INIT_DELAYED_WORK(&pag->pag_blockgc_work, xfs_blockgc_worker); + INIT_DELAYED_WORK(&pag->pag_inactive_work, xfs_inactive_worker); INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC); if (xfs_buf_hash_init(pag)) goto out_free_pag; diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h index 7c4d07d5995c..409d4d438244 100644 --- a/fs/xfs/xfs_mount.h +++ b/fs/xfs/xfs_mount.h @@ -185,7 +185,6 @@ typedef struct xfs_mount { uint64_t m_resblks_avail;/* available reserved blocks */ uint64_t m_resblks_save; /* reserved blks @ remount,ro */ struct delayed_work m_reclaim_work; /* background inode reclaim */ - struct delayed_work m_inactive_work; /* background inode inactive */ struct xfs_kobj m_kobj; struct xfs_kobj m_error_kobj; struct xfs_kobj m_error_meta_kobj; @@ -378,6 +377,9 @@ typedef struct xfs_perag { /* background prealloc block trimming */ struct delayed_work pag_blockgc_work; + /* background inode inactivation */ + struct delayed_work pag_inactive_work; + /* reference count */ uint8_t pagf_refcount_level; diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c index 7308763eee78..42fc4e8b0a33 100644 --- a/fs/xfs/xfs_super.c +++ b/fs/xfs/xfs_super.c @@ -525,7 +525,8 @@ xfs_init_mount_workqueues( goto out_destroy_eofb; mp->m_inactive_workqueue = alloc_workqueue("xfs-inactive/%s", - WQ_MEM_RECLAIM | WQ_FREEZABLE, 0, mp->m_super->s_id); + WQ_UNBOUND | WQ_MEM_RECLAIM | WQ_FREEZABLE, 0, + mp->m_super->s_id); if (!mp->m_inactive_workqueue) goto out_destroy_sync; @@ -1463,6 +1464,7 @@ xfs_configure_background_workqueues( max_active = min_t(unsigned int, threads, WQ_UNBOUND_MAX_ACTIVE); workqueue_set_max_active(mp->m_blockgc_workqueue, max_active); + workqueue_set_max_active(mp->m_inactive_workqueue, max_active); } static int @@ -1920,7 +1922,6 @@ static int xfs_init_fs_context( mutex_init(&mp->m_growlock); INIT_WORK(&mp->m_flush_inodes_work, xfs_flush_inodes_worker); INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker); - INIT_DELAYED_WORK(&mp->m_inactive_work, xfs_inactive_worker); mp->m_kobj.kobject.kset = xfs_kset; /* * We don't create the finobt per-ag space reservation until after log |