From 74b2107543da4ed9607ec484f63c42362dc9fca6 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 13 Apr 2011 12:02:53 -0400 Subject: [PATCH 01/25] Btrfs: make sure to use the delalloc reserve when filling delalloc In the prealloc filling code and compressed code we don't set trans->block_rsv to the delalloc block reserve properly, which is going to make us use metadata from the wrong pool, this patch fixes that. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/inode.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 7cd8ab0ef04d..3b9f1643aa57 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -619,6 +619,7 @@ static noinline int submit_compressed_extents(struct inode *inode, trans = btrfs_join_transaction(root, 1); BUG_ON(IS_ERR(trans)); + trans->block_rsv = &root->fs_info->delalloc_block_rsv; ret = btrfs_reserve_extent(trans, root, async_extent->compressed_size, async_extent->compressed_size, @@ -1060,6 +1061,7 @@ static noinline int run_delalloc_nocow(struct inode *inode, trans = btrfs_join_transaction(root, 1); } BUG_ON(IS_ERR(trans)); + trans->block_rsv = &root->fs_info->delalloc_block_rsv; cow_start = (u64)-1; cur_offset = start; From 7a7eaa40a39bde4eefc91aadeb1ce3dc4e6a1252 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 13 Apr 2011 12:54:33 -0400 Subject: [PATCH 02/25] Btrfs: take away the num_items argument from btrfs_join_transaction I keep forgetting that btrfs_join_transaction() just ignores the num_items argument, which leads me to sending pointless patches and looking stupid :). So just kill the num_items argument from btrfs_join_transaction and btrfs_start_ioctl_transaction, since neither of them use it. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/disk-io.c | 6 +++--- fs/btrfs/extent-tree.c | 12 ++++++------ fs/btrfs/inode.c | 34 +++++++++++++++++----------------- fs/btrfs/ioctl.c | 4 ++-- fs/btrfs/relocation.c | 12 ++++++------ fs/btrfs/transaction.c | 13 +++++-------- fs/btrfs/transaction.h | 9 +++------ 7 files changed, 42 insertions(+), 48 deletions(-) diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 228cf36ece83..9d6c9e332ca3 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1568,7 +1568,7 @@ static int transaction_kthread(void *arg) transid = cur->transid; spin_unlock(&root->fs_info->new_trans_lock); - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); if (transid == trans->transid) { ret = btrfs_commit_transaction(trans, root); @@ -2495,13 +2495,13 @@ int btrfs_commit_super(struct btrfs_root *root) down_write(&root->fs_info->cleanup_work_sem); up_write(&root->fs_info->cleanup_work_sem); - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) return PTR_ERR(trans); ret = btrfs_commit_transaction(trans, root); BUG_ON(ret); /* run commit again to drop the original snapshot */ - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) return PTR_ERR(trans); btrfs_commit_transaction(trans, root); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 9ee6bd55e16c..941b28e78931 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3174,7 +3174,7 @@ int btrfs_check_data_free_space(struct inode *inode, u64 bytes) spin_unlock(&data_sinfo->lock); alloc: alloc_target = btrfs_get_alloc_profile(root, 1); - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) return PTR_ERR(trans); @@ -3202,7 +3202,7 @@ int btrfs_check_data_free_space(struct inode *inode, u64 bytes) commit_trans: if (!committed && !root->fs_info->open_ioctl_trans) { committed = 1; - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) return PTR_ERR(trans); ret = btrfs_commit_transaction(trans, root); @@ -3589,7 +3589,7 @@ static int reserve_metadata_bytes(struct btrfs_trans_handle *trans, goto out; ret = -ENOSPC; - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) goto out; ret = btrfs_commit_transaction(trans, root); @@ -3816,7 +3816,7 @@ int btrfs_block_rsv_check(struct btrfs_trans_handle *trans, if (trans) return -EAGAIN; - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); ret = btrfs_commit_transaction(trans, root); return 0; @@ -7649,7 +7649,7 @@ int btrfs_drop_dead_reloc_roots(struct btrfs_root *root) BUG_ON(reloc_root->commit_root != NULL); while (1) { - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); mutex_lock(&root->fs_info->drop_mutex); @@ -8176,7 +8176,7 @@ int btrfs_set_block_group_ro(struct btrfs_root *root, BUG_ON(cache->ro); - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); alloc_flags = update_block_group_flags(root, cache->flags); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 3b9f1643aa57..e47bdf0fb75a 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -420,7 +420,7 @@ static noinline int compress_file_range(struct inode *inode, } } if (start == 0) { - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); btrfs_set_trans_block_group(trans, inode); trans->block_rsv = &root->fs_info->delalloc_block_rsv; @@ -617,7 +617,7 @@ static noinline int submit_compressed_extents(struct inode *inode, async_extent->start + async_extent->ram_size - 1, GFP_NOFS); - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); trans->block_rsv = &root->fs_info->delalloc_block_rsv; ret = btrfs_reserve_extent(trans, root, @@ -779,7 +779,7 @@ static noinline int cow_file_range(struct inode *inode, int ret = 0; BUG_ON(root == root->fs_info->tree_root); - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); btrfs_set_trans_block_group(trans, inode); trans->block_rsv = &root->fs_info->delalloc_block_rsv; @@ -1056,9 +1056,9 @@ static noinline int run_delalloc_nocow(struct inode *inode, BUG_ON(!path); if (root == root->fs_info->tree_root) { nolock = true; - trans = btrfs_join_transaction_nolock(root, 1); + trans = btrfs_join_transaction_nolock(root); } else { - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); } BUG_ON(IS_ERR(trans)); trans->block_rsv = &root->fs_info->delalloc_block_rsv; @@ -1718,9 +1718,9 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) ret = btrfs_ordered_update_i_size(inode, 0, ordered_extent); if (!ret) { if (nolock) - trans = btrfs_join_transaction_nolock(root, 1); + trans = btrfs_join_transaction_nolock(root); else - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); btrfs_set_trans_block_group(trans, inode); trans->block_rsv = &root->fs_info->delalloc_block_rsv; @@ -1735,9 +1735,9 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) 0, &cached_state, GFP_NOFS); if (nolock) - trans = btrfs_join_transaction_nolock(root, 1); + trans = btrfs_join_transaction_nolock(root); else - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); btrfs_set_trans_block_group(trans, inode); trans->block_rsv = &root->fs_info->delalloc_block_rsv; @@ -2415,7 +2415,7 @@ int btrfs_orphan_cleanup(struct btrfs_root *root) (u64)-1); if (root->orphan_block_rsv || root->orphan_item_inserted) { - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); if (!IS_ERR(trans)) btrfs_end_transaction(trans, root); } @@ -4378,9 +4378,9 @@ int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc) if (wbc->sync_mode == WB_SYNC_ALL) { if (nolock) - trans = btrfs_join_transaction_nolock(root, 1); + trans = btrfs_join_transaction_nolock(root); else - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) return PTR_ERR(trans); btrfs_set_trans_block_group(trans, inode); @@ -4407,7 +4407,7 @@ void btrfs_dirty_inode(struct inode *inode) if (BTRFS_I(inode)->dummy_inode) return; - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); btrfs_set_trans_block_group(trans, inode); @@ -5226,7 +5226,7 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, free_extent_map(em); em = NULL; btrfs_release_path(root, path); - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) return ERR_CAST(trans); goto again; @@ -5470,7 +5470,7 @@ static struct extent_map *btrfs_new_extent_direct(struct inode *inode, btrfs_drop_extent_cache(inode, start, start + len - 1, 0); } - trans = btrfs_join_transaction(root, 0); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) return ERR_CAST(trans); @@ -5703,7 +5703,7 @@ static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock, * to make sure the current transaction stays open * while we look for nocow cross refs */ - trans = btrfs_join_transaction(root, 0); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) goto must_cow; @@ -5841,7 +5841,7 @@ static void btrfs_endio_direct_write(struct bio *bio, int err) BUG_ON(!ordered); - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) { err = -ENOMEM; goto out; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 2616f7ed4799..908c3d4b48c6 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -242,7 +242,7 @@ static int btrfs_ioctl_setflags(struct file *file, void __user *arg) ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS); } - trans = btrfs_join_transaction(root, 1); + trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); ret = btrfs_update_inode(trans, root, inode); @@ -2182,7 +2182,7 @@ static long btrfs_ioctl_trans_start(struct file *file) mutex_unlock(&root->fs_info->trans_mutex); ret = -ENOMEM; - trans = btrfs_start_ioctl_transaction(root, 0); + trans = btrfs_start_ioctl_transaction(root); if (IS_ERR(trans)) goto out_drop; diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 199a80134312..8bb256667f2d 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -2149,7 +2149,7 @@ int prepare_to_merge(struct reloc_control *rc, int err) err = ret; } - trans = btrfs_join_transaction(rc->extent_root, 1); + trans = btrfs_join_transaction(rc->extent_root); if (IS_ERR(trans)) { if (!err) btrfs_block_rsv_release(rc->extent_root, @@ -3233,7 +3233,7 @@ static int delete_block_group_cache(struct btrfs_fs_info *fs_info, goto out; } - trans = btrfs_join_transaction(root, 0); + trans = btrfs_join_transaction(root); if (IS_ERR(trans)) { btrfs_free_path(path); ret = PTR_ERR(trans); @@ -3642,7 +3642,7 @@ int prepare_to_relocate(struct reloc_control *rc) rc->create_reloc_tree = 1; set_reloc_control(rc); - trans = btrfs_join_transaction(rc->extent_root, 1); + trans = btrfs_join_transaction(rc->extent_root); BUG_ON(IS_ERR(trans)); btrfs_commit_transaction(trans, rc->extent_root); return 0; @@ -3831,7 +3831,7 @@ static noinline_for_stack int relocate_block_group(struct reloc_control *rc) btrfs_block_rsv_release(rc->extent_root, rc->block_rsv, (u64)-1); /* get rid of pinned extents */ - trans = btrfs_join_transaction(rc->extent_root, 1); + trans = btrfs_join_transaction(rc->extent_root); if (IS_ERR(trans)) err = PTR_ERR(trans); else @@ -4156,7 +4156,7 @@ int btrfs_recover_relocation(struct btrfs_root *root) set_reloc_control(rc); - trans = btrfs_join_transaction(rc->extent_root, 1); + trans = btrfs_join_transaction(rc->extent_root); if (IS_ERR(trans)) { unset_reloc_control(rc); err = PTR_ERR(trans); @@ -4190,7 +4190,7 @@ int btrfs_recover_relocation(struct btrfs_root *root) unset_reloc_control(rc); - trans = btrfs_join_transaction(rc->extent_root, 1); + trans = btrfs_join_transaction(rc->extent_root); if (IS_ERR(trans)) err = PTR_ERR(trans); else diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index c571734d5e5a..70bfb26df967 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -257,22 +257,19 @@ struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, { return start_transaction(root, num_items, TRANS_START); } -struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root, - int num_blocks) +struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root) { return start_transaction(root, 0, TRANS_JOIN); } -struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root, - int num_blocks) +struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root) { return start_transaction(root, 0, TRANS_JOIN_NOLOCK); } -struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r, - int num_blocks) +struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *root) { - return start_transaction(r, 0, TRANS_USERSPACE); + return start_transaction(root, 0, TRANS_USERSPACE); } /* wait for a transaction commit to be fully complete */ @@ -1171,7 +1168,7 @@ int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans, INIT_DELAYED_WORK(&ac->work, do_async_commit); ac->root = root; - ac->newtrans = btrfs_join_transaction(root, 0); + ac->newtrans = btrfs_join_transaction(root); if (IS_ERR(ac->newtrans)) { int err = PTR_ERR(ac->newtrans); kfree(ac); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index e441acc6c584..1f573f09dba2 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -92,12 +92,9 @@ int btrfs_end_transaction_nolock(struct btrfs_trans_handle *trans, struct btrfs_root *root); struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, int num_items); -struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root, - int num_blocks); -struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root, - int num_blocks); -struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r, - int num_blocks); +struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root); +struct btrfs_trans_handle *btrfs_join_transaction_nolock(struct btrfs_root *root); +struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *root); int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid); int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, struct btrfs_root *root); From 2a1eb4614d984d5cd4c928784e9afcf5c07f93be Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 13 Apr 2011 15:15:59 -0400 Subject: [PATCH 03/25] Btrfs: if we've already started a trans handle, use that one We currently track trans handles in current->journal_info, but we don't actually use it. This patch fixes it. This will cover the case where we have multiple people starting transactions down the call chain. This keeps us from having to allocate a new handle and all of that, we just increase the use count of the current handle, save the old block_rsv, and return. I tested this with xfstests and it worked out fine. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/transaction.c | 17 +++++++++++++++++ fs/btrfs/transaction.h | 2 ++ 2 files changed, 19 insertions(+) diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 70bfb26df967..46f40564c168 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -184,6 +184,15 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) return ERR_PTR(-EROFS); + + if (current->journal_info) { + WARN_ON(type != TRANS_JOIN && type != TRANS_JOIN_NOLOCK); + h = current->journal_info; + h->use_count++; + h->orig_rsv = h->block_rsv; + h->block_rsv = NULL; + goto got_it; + } again: h = kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS); if (!h) @@ -213,7 +222,9 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, h->block_group = 0; h->bytes_reserved = 0; h->delayed_ref_updates = 0; + h->use_count = 1; h->block_rsv = NULL; + h->orig_rsv = NULL; smp_mb(); if (cur_trans->blocked && may_wait_transaction(root, type)) { @@ -241,6 +252,7 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, } } +got_it: if (type != TRANS_JOIN_NOLOCK) mutex_lock(&root->fs_info->trans_mutex); record_root_in_trans(h, root); @@ -428,6 +440,11 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, struct btrfs_fs_info *info = root->fs_info; int count = 0; + if (--trans->use_count) { + trans->block_rsv = trans->orig_rsv; + return 0; + } + while (count < 4) { unsigned long cur = trans->delayed_ref_updates; trans->delayed_ref_updates = 0; diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 1f573f09dba2..154314f80f8d 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -47,11 +47,13 @@ struct btrfs_trans_handle { u64 transid; u64 block_group; u64 bytes_reserved; + unsigned long use_count; unsigned long blocks_reserved; unsigned long blocks_used; unsigned long delayed_ref_updates; struct btrfs_transaction *transaction; struct btrfs_block_rsv *block_rsv; + struct btrfs_block_rsv *orig_rsv; }; struct btrfs_pending_snapshot { From a4abeea41adfa3c143c289045f4625dfaeba2212 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Mon, 11 Apr 2011 17:25:13 -0400 Subject: [PATCH 04/25] Btrfs: kill trans_mutex We use trans_mutex for lots of things, here's a basic list 1) To serialize trans_handles joining the currently running transaction 2) To make sure that no new trans handles are started while we are committing 3) To protect the dead_roots list and the transaction lists Really the serializing trans_handles joining is not too hard, and can really get bogged down in acquiring a reference to the transaction. So replace the trans_mutex with a trans_lock spinlock and use it to do the following 1) Protect fs_info->running_transaction. All trans handles have to do is check this, and then take a reference of the transaction and keep on going. 2) Protect the fs_info->trans_list. This doesn't get used too much, basically it just holds the current transactions, which will usually just be the currently committing transaction and the currently running transaction at most. 3) Protect the dead roots list. This is only ever processed by splicing the list so this is relatively simple. 4) Protect the fs_info->reloc_ctl stuff. This is very lightweight and was using the trans_mutex before, so this is a pretty straightforward change. 5) Protect fs_info->no_trans_join. Because we don't hold the trans_lock over the entirety of the commit we need to have a way to block new people from creating a new transaction while we're doing our work. So we set no_trans_join and in join_transaction we test to see if that is set, and if it is we do a wait_on_commit. 6) Make the transaction use count atomic so we don't need to take locks to modify it when we're dropping references. 7) Add a commit_lock to the transaction to make sure multiple people trying to commit the same transaction don't race and commit at the same time. 8) Make open_ioctl_trans an atomic so we don't have to take any locks for ioctl trans. I have tested this with xfstests, but obviously it is a pretty hairy change so lots of testing is greatly appreciated. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/ctree.h | 6 +- fs/btrfs/disk-io.c | 30 ++--- fs/btrfs/extent-tree.c | 3 +- fs/btrfs/file.c | 4 +- fs/btrfs/ioctl.c | 12 +- fs/btrfs/relocation.c | 16 +-- fs/btrfs/transaction.c | 271 +++++++++++++++++++++-------------------- fs/btrfs/transaction.h | 4 +- 8 files changed, 177 insertions(+), 169 deletions(-) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 8f4b81de3ae2..522a39b0033d 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -919,7 +919,6 @@ struct btrfs_fs_info { * is required instead of the faster short fsync log commits */ u64 last_trans_log_full_commit; - u64 open_ioctl_trans; unsigned long mount_opt:20; unsigned long compress_type:4; u64 max_inline; @@ -936,7 +935,6 @@ struct btrfs_fs_info { struct super_block *sb; struct inode *btree_inode; struct backing_dev_info bdi; - struct mutex trans_mutex; struct mutex tree_log_mutex; struct mutex transaction_kthread_mutex; struct mutex cleaner_mutex; @@ -957,6 +955,7 @@ struct btrfs_fs_info { struct rw_semaphore subvol_sem; struct srcu_struct subvol_srcu; + spinlock_t trans_lock; struct list_head trans_list; struct list_head hashers; struct list_head dead_roots; @@ -969,6 +968,7 @@ struct btrfs_fs_info { atomic_t async_submit_draining; atomic_t nr_async_bios; atomic_t async_delalloc_pages; + atomic_t open_ioctl_trans; /* * this is used by the balancing code to wait for all the pending @@ -1032,6 +1032,7 @@ struct btrfs_fs_info { int closing; int log_root_recovering; int enospc_unlink; + int trans_no_join; u64 total_pinned; @@ -1053,7 +1054,6 @@ struct btrfs_fs_info { struct reloc_control *reloc_ctl; spinlock_t delalloc_lock; - spinlock_t new_trans_lock; u64 delalloc_bytes; /* data_alloc_cluster is only used in ssd mode */ diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 9d6c9e332ca3..93ef254ec432 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1551,22 +1551,22 @@ static int transaction_kthread(void *arg) vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); mutex_lock(&root->fs_info->transaction_kthread_mutex); - spin_lock(&root->fs_info->new_trans_lock); + spin_lock(&root->fs_info->trans_lock); cur = root->fs_info->running_transaction; if (!cur) { - spin_unlock(&root->fs_info->new_trans_lock); + spin_unlock(&root->fs_info->trans_lock); goto sleep; } now = get_seconds(); if (!cur->blocked && (now < cur->start_time || now - cur->start_time < 30)) { - spin_unlock(&root->fs_info->new_trans_lock); + spin_unlock(&root->fs_info->trans_lock); delay = HZ * 5; goto sleep; } transid = cur->transid; - spin_unlock(&root->fs_info->new_trans_lock); + spin_unlock(&root->fs_info->trans_lock); trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); @@ -1658,7 +1658,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, INIT_LIST_HEAD(&fs_info->ordered_operations); INIT_LIST_HEAD(&fs_info->caching_block_groups); spin_lock_init(&fs_info->delalloc_lock); - spin_lock_init(&fs_info->new_trans_lock); + spin_lock_init(&fs_info->trans_lock); spin_lock_init(&fs_info->ref_cache_lock); spin_lock_init(&fs_info->fs_roots_radix_lock); spin_lock_init(&fs_info->delayed_iput_lock); @@ -1687,6 +1687,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->sb = sb; fs_info->max_inline = 8192 * 1024; fs_info->metadata_ratio = 0; + fs_info->trans_no_join = 0; fs_info->thread_pool_size = min_t(unsigned long, num_online_cpus() + 2, 8); @@ -1735,7 +1736,6 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->do_barriers = 1; - mutex_init(&fs_info->trans_mutex); mutex_init(&fs_info->ordered_operations_mutex); mutex_init(&fs_info->tree_log_mutex); mutex_init(&fs_info->chunk_mutex); @@ -3006,10 +3006,13 @@ static int btrfs_cleanup_transaction(struct btrfs_root *root) WARN_ON(1); - mutex_lock(&root->fs_info->trans_mutex); mutex_lock(&root->fs_info->transaction_kthread_mutex); + spin_lock(&root->fs_info->trans_lock); list_splice_init(&root->fs_info->trans_list, &list); + root->fs_info->trans_no_join = 1; + spin_unlock(&root->fs_info->trans_lock); + while (!list_empty(&list)) { t = list_entry(list.next, struct btrfs_transaction, list); if (!t) @@ -3034,23 +3037,18 @@ static int btrfs_cleanup_transaction(struct btrfs_root *root) t->blocked = 0; if (waitqueue_active(&root->fs_info->transaction_wait)) wake_up(&root->fs_info->transaction_wait); - mutex_unlock(&root->fs_info->trans_mutex); - mutex_lock(&root->fs_info->trans_mutex); t->commit_done = 1; if (waitqueue_active(&t->commit_wait)) wake_up(&t->commit_wait); - mutex_unlock(&root->fs_info->trans_mutex); - - mutex_lock(&root->fs_info->trans_mutex); btrfs_destroy_pending_snapshots(t); btrfs_destroy_delalloc_inodes(root); - spin_lock(&root->fs_info->new_trans_lock); + spin_lock(&root->fs_info->trans_lock); root->fs_info->running_transaction = NULL; - spin_unlock(&root->fs_info->new_trans_lock); + spin_unlock(&root->fs_info->trans_lock); btrfs_destroy_marked_extents(root, &t->dirty_pages, EXTENT_DIRTY); @@ -3064,8 +3062,10 @@ static int btrfs_cleanup_transaction(struct btrfs_root *root) kmem_cache_free(btrfs_transaction_cachep, t); } + spin_lock(&root->fs_info->trans_lock); + root->fs_info->trans_no_join = 0; + spin_unlock(&root->fs_info->trans_lock); mutex_unlock(&root->fs_info->transaction_kthread_mutex); - mutex_unlock(&root->fs_info->trans_mutex); return 0; } diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 941b28e78931..ca599654ce19 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3200,7 +3200,8 @@ int btrfs_check_data_free_space(struct inode *inode, u64 bytes) /* commit the current transaction and try again */ commit_trans: - if (!committed && !root->fs_info->open_ioctl_trans) { + if (!committed && + !atomic_read(&root->fs_info->open_ioctl_trans)) { committed = 1; trans = btrfs_join_transaction(root); if (IS_ERR(trans)) diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 75899a01dded..cd5e82e500cf 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1222,14 +1222,12 @@ int btrfs_sync_file(struct file *file, int datasync) * the current transaction, we can bail out now without any * syncing */ - mutex_lock(&root->fs_info->trans_mutex); + smp_mb(); if (BTRFS_I(inode)->last_trans <= root->fs_info->last_trans_committed) { BTRFS_I(inode)->last_trans = 0; - mutex_unlock(&root->fs_info->trans_mutex); goto out; } - mutex_unlock(&root->fs_info->trans_mutex); /* * ok we haven't committed the transaction yet, lets do a commit diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 908c3d4b48c6..a578620e06a8 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -2177,9 +2177,7 @@ static long btrfs_ioctl_trans_start(struct file *file) if (ret) goto out; - mutex_lock(&root->fs_info->trans_mutex); - root->fs_info->open_ioctl_trans++; - mutex_unlock(&root->fs_info->trans_mutex); + atomic_inc(&root->fs_info->open_ioctl_trans); ret = -ENOMEM; trans = btrfs_start_ioctl_transaction(root); @@ -2190,9 +2188,7 @@ static long btrfs_ioctl_trans_start(struct file *file) return 0; out_drop: - mutex_lock(&root->fs_info->trans_mutex); - root->fs_info->open_ioctl_trans--; - mutex_unlock(&root->fs_info->trans_mutex); + atomic_dec(&root->fs_info->open_ioctl_trans); mnt_drop_write(file->f_path.mnt); out: return ret; @@ -2426,9 +2422,7 @@ long btrfs_ioctl_trans_end(struct file *file) btrfs_end_transaction(trans, root); - mutex_lock(&root->fs_info->trans_mutex); - root->fs_info->open_ioctl_trans--; - mutex_unlock(&root->fs_info->trans_mutex); + atomic_dec(&root->fs_info->open_ioctl_trans); mnt_drop_write(file->f_path.mnt); return 0; diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 8bb256667f2d..09c30d37d43e 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -2136,10 +2136,10 @@ int prepare_to_merge(struct reloc_control *rc, int err) u64 num_bytes = 0; int ret; - mutex_lock(&root->fs_info->trans_mutex); + spin_lock(&root->fs_info->trans_lock); rc->merging_rsv_size += root->nodesize * (BTRFS_MAX_LEVEL - 1) * 2; rc->merging_rsv_size += rc->nodes_relocated * 2; - mutex_unlock(&root->fs_info->trans_mutex); + spin_unlock(&root->fs_info->trans_lock); again: if (!err) { num_bytes = rc->merging_rsv_size; @@ -2208,9 +2208,9 @@ int merge_reloc_roots(struct reloc_control *rc) int ret; again: root = rc->extent_root; - mutex_lock(&root->fs_info->trans_mutex); + spin_lock(&root->fs_info->trans_lock); list_splice_init(&rc->reloc_roots, &reloc_roots); - mutex_unlock(&root->fs_info->trans_mutex); + spin_unlock(&root->fs_info->trans_lock); while (!list_empty(&reloc_roots)) { found = 1; @@ -3583,17 +3583,17 @@ int find_next_extent(struct btrfs_trans_handle *trans, static void set_reloc_control(struct reloc_control *rc) { struct btrfs_fs_info *fs_info = rc->extent_root->fs_info; - mutex_lock(&fs_info->trans_mutex); + spin_lock(&fs_info->trans_lock); fs_info->reloc_ctl = rc; - mutex_unlock(&fs_info->trans_mutex); + spin_unlock(&fs_info->trans_lock); } static void unset_reloc_control(struct reloc_control *rc) { struct btrfs_fs_info *fs_info = rc->extent_root->fs_info; - mutex_lock(&fs_info->trans_mutex); + spin_lock(&fs_info->trans_lock); fs_info->reloc_ctl = NULL; - mutex_unlock(&fs_info->trans_mutex); + spin_unlock(&fs_info->trans_lock); } static int check_extent_flags(u64 flags) diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 46f40564c168..43816f8b23e7 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -34,6 +34,7 @@ static noinline void put_transaction(struct btrfs_transaction *transaction) { WARN_ON(atomic_read(&transaction->use_count) == 0); if (atomic_dec_and_test(&transaction->use_count)) { + BUG_ON(!list_empty(&transaction->list)); memset(transaction, 0, sizeof(*transaction)); kmem_cache_free(btrfs_transaction_cachep, transaction); } @@ -48,47 +49,73 @@ static noinline void switch_commit_root(struct btrfs_root *root) /* * either allocate a new transaction or hop into the existing one */ -static noinline int join_transaction(struct btrfs_root *root) +static noinline int join_transaction(struct btrfs_root *root, int nofail) { struct btrfs_transaction *cur_trans; + + spin_lock(&root->fs_info->trans_lock); + if (root->fs_info->trans_no_join) { + if (!nofail) { + spin_unlock(&root->fs_info->trans_lock); + return -EBUSY; + } + } + cur_trans = root->fs_info->running_transaction; - if (!cur_trans) { - cur_trans = kmem_cache_alloc(btrfs_transaction_cachep, - GFP_NOFS); - if (!cur_trans) - return -ENOMEM; - root->fs_info->generation++; - atomic_set(&cur_trans->num_writers, 1); - cur_trans->num_joined = 0; - cur_trans->transid = root->fs_info->generation; - init_waitqueue_head(&cur_trans->writer_wait); - init_waitqueue_head(&cur_trans->commit_wait); - cur_trans->in_commit = 0; - cur_trans->blocked = 0; - atomic_set(&cur_trans->use_count, 1); - cur_trans->commit_done = 0; - cur_trans->start_time = get_seconds(); - - cur_trans->delayed_refs.root = RB_ROOT; - cur_trans->delayed_refs.num_entries = 0; - cur_trans->delayed_refs.num_heads_ready = 0; - cur_trans->delayed_refs.num_heads = 0; - cur_trans->delayed_refs.flushing = 0; - cur_trans->delayed_refs.run_delayed_start = 0; - spin_lock_init(&cur_trans->delayed_refs.lock); - - INIT_LIST_HEAD(&cur_trans->pending_snapshots); - list_add_tail(&cur_trans->list, &root->fs_info->trans_list); - extent_io_tree_init(&cur_trans->dirty_pages, - root->fs_info->btree_inode->i_mapping, - GFP_NOFS); - spin_lock(&root->fs_info->new_trans_lock); - root->fs_info->running_transaction = cur_trans; - spin_unlock(&root->fs_info->new_trans_lock); - } else { + if (cur_trans) { + atomic_inc(&cur_trans->use_count); atomic_inc(&cur_trans->num_writers); cur_trans->num_joined++; + spin_unlock(&root->fs_info->trans_lock); + return 0; } + spin_unlock(&root->fs_info->trans_lock); + + cur_trans = kmem_cache_alloc(btrfs_transaction_cachep, GFP_NOFS); + if (!cur_trans) + return -ENOMEM; + spin_lock(&root->fs_info->trans_lock); + if (root->fs_info->running_transaction) { + kmem_cache_free(btrfs_transaction_cachep, cur_trans); + cur_trans = root->fs_info->running_transaction; + atomic_inc(&cur_trans->use_count); + atomic_inc(&cur_trans->num_writers); + cur_trans->num_joined++; + spin_unlock(&root->fs_info->trans_lock); + return 0; + } + atomic_set(&cur_trans->num_writers, 1); + cur_trans->num_joined = 0; + init_waitqueue_head(&cur_trans->writer_wait); + init_waitqueue_head(&cur_trans->commit_wait); + cur_trans->in_commit = 0; + cur_trans->blocked = 0; + /* + * One for this trans handle, one so it will live on until we + * commit the transaction. + */ + atomic_set(&cur_trans->use_count, 2); + cur_trans->commit_done = 0; + cur_trans->start_time = get_seconds(); + + cur_trans->delayed_refs.root = RB_ROOT; + cur_trans->delayed_refs.num_entries = 0; + cur_trans->delayed_refs.num_heads_ready = 0; + cur_trans->delayed_refs.num_heads = 0; + cur_trans->delayed_refs.flushing = 0; + cur_trans->delayed_refs.run_delayed_start = 0; + spin_lock_init(&cur_trans->commit_lock); + spin_lock_init(&cur_trans->delayed_refs.lock); + + INIT_LIST_HEAD(&cur_trans->pending_snapshots); + list_add_tail(&cur_trans->list, &root->fs_info->trans_list); + extent_io_tree_init(&cur_trans->dirty_pages, + root->fs_info->btree_inode->i_mapping, + GFP_NOFS); + root->fs_info->generation++; + cur_trans->transid = root->fs_info->generation; + root->fs_info->running_transaction = cur_trans; + spin_unlock(&root->fs_info->trans_lock); return 0; } @@ -99,39 +126,28 @@ static noinline int join_transaction(struct btrfs_root *root) * to make sure the old root from before we joined the transaction is deleted * when the transaction commits */ -static noinline int record_root_in_trans(struct btrfs_trans_handle *trans, - struct btrfs_root *root) +int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans, + struct btrfs_root *root) { if (root->ref_cows && root->last_trans < trans->transid) { WARN_ON(root == root->fs_info->extent_root); WARN_ON(root->commit_root != root->node); + spin_lock(&root->fs_info->fs_roots_radix_lock); + if (root->last_trans == trans->transid) { + spin_unlock(&root->fs_info->fs_roots_radix_lock); + return 0; + } + root->last_trans = trans->transid; radix_tree_tag_set(&root->fs_info->fs_roots_radix, (unsigned long)root->root_key.objectid, BTRFS_ROOT_TRANS_TAG); - root->last_trans = trans->transid; + spin_unlock(&root->fs_info->fs_roots_radix_lock); btrfs_init_reloc_root(trans, root); } return 0; } -int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans, - struct btrfs_root *root) -{ - if (!root->ref_cows) - return 0; - - mutex_lock(&root->fs_info->trans_mutex); - if (root->last_trans == trans->transid) { - mutex_unlock(&root->fs_info->trans_mutex); - return 0; - } - - record_root_in_trans(trans, root); - mutex_unlock(&root->fs_info->trans_mutex); - return 0; -} - /* wait for commit against the current transaction to become unblocked * when this is done, it is safe to start a new transaction, but the current * transaction might not be fully on disk. @@ -140,21 +156,23 @@ static void wait_current_trans(struct btrfs_root *root) { struct btrfs_transaction *cur_trans; + spin_lock(&root->fs_info->trans_lock); cur_trans = root->fs_info->running_transaction; if (cur_trans && cur_trans->blocked) { DEFINE_WAIT(wait); atomic_inc(&cur_trans->use_count); + spin_unlock(&root->fs_info->trans_lock); while (1) { prepare_to_wait(&root->fs_info->transaction_wait, &wait, TASK_UNINTERRUPTIBLE); if (!cur_trans->blocked) break; - mutex_unlock(&root->fs_info->trans_mutex); schedule(); - mutex_lock(&root->fs_info->trans_mutex); } finish_wait(&root->fs_info->transaction_wait, &wait); put_transaction(cur_trans); + } else { + spin_unlock(&root->fs_info->trans_lock); } } @@ -167,10 +185,16 @@ enum btrfs_trans_type { static int may_wait_transaction(struct btrfs_root *root, int type) { - if (!root->fs_info->log_root_recovering && - ((type == TRANS_START && !root->fs_info->open_ioctl_trans) || - type == TRANS_USERSPACE)) + if (root->fs_info->log_root_recovering) + return 0; + + if (type == TRANS_USERSPACE) return 1; + + if (type == TRANS_START && + !atomic_read(&root->fs_info->open_ioctl_trans)) + return 1; + return 0; } @@ -198,23 +222,21 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, if (!h) return ERR_PTR(-ENOMEM); - if (type != TRANS_JOIN_NOLOCK) - mutex_lock(&root->fs_info->trans_mutex); if (may_wait_transaction(root, type)) wait_current_trans(root); - ret = join_transaction(root); + do { + ret = join_transaction(root, type == TRANS_JOIN_NOLOCK); + if (ret == -EBUSY) + wait_current_trans(root); + } while (ret == -EBUSY); + if (ret < 0) { kmem_cache_free(btrfs_trans_handle_cachep, h); - if (type != TRANS_JOIN_NOLOCK) - mutex_unlock(&root->fs_info->trans_mutex); return ERR_PTR(ret); } cur_trans = root->fs_info->running_transaction; - atomic_inc(&cur_trans->use_count); - if (type != TRANS_JOIN_NOLOCK) - mutex_unlock(&root->fs_info->trans_mutex); h->transid = cur_trans->transid; h->transaction = cur_trans; @@ -253,11 +275,7 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, } got_it: - if (type != TRANS_JOIN_NOLOCK) - mutex_lock(&root->fs_info->trans_mutex); - record_root_in_trans(h, root); - if (type != TRANS_JOIN_NOLOCK) - mutex_unlock(&root->fs_info->trans_mutex); + btrfs_record_root_in_trans(h, root); if (!current->journal_info && type != TRANS_USERSPACE) current->journal_info = h; @@ -289,17 +307,13 @@ static noinline int wait_for_commit(struct btrfs_root *root, struct btrfs_transaction *commit) { DEFINE_WAIT(wait); - mutex_lock(&root->fs_info->trans_mutex); while (!commit->commit_done) { prepare_to_wait(&commit->commit_wait, &wait, TASK_UNINTERRUPTIBLE); if (commit->commit_done) break; - mutex_unlock(&root->fs_info->trans_mutex); schedule(); - mutex_lock(&root->fs_info->trans_mutex); } - mutex_unlock(&root->fs_info->trans_mutex); finish_wait(&commit->commit_wait, &wait); return 0; } @@ -309,50 +323,49 @@ int btrfs_wait_for_commit(struct btrfs_root *root, u64 transid) struct btrfs_transaction *cur_trans = NULL, *t; int ret; - mutex_lock(&root->fs_info->trans_mutex); - ret = 0; if (transid) { if (transid <= root->fs_info->last_trans_committed) - goto out_unlock; + goto out; /* find specified transaction */ + spin_lock(&root->fs_info->trans_lock); list_for_each_entry(t, &root->fs_info->trans_list, list) { if (t->transid == transid) { cur_trans = t; + atomic_inc(&cur_trans->use_count); break; } if (t->transid > transid) break; } + spin_unlock(&root->fs_info->trans_lock); ret = -EINVAL; if (!cur_trans) - goto out_unlock; /* bad transid */ + goto out; /* bad transid */ } else { /* find newest transaction that is committing | committed */ + spin_lock(&root->fs_info->trans_lock); list_for_each_entry_reverse(t, &root->fs_info->trans_list, list) { if (t->in_commit) { if (t->commit_done) - goto out_unlock; + goto out; cur_trans = t; + atomic_inc(&cur_trans->use_count); break; } } + spin_unlock(&root->fs_info->trans_lock); if (!cur_trans) - goto out_unlock; /* nothing committing|committed */ + goto out; /* nothing committing|committed */ } - atomic_inc(&cur_trans->use_count); - mutex_unlock(&root->fs_info->trans_mutex); - wait_for_commit(root, cur_trans); - mutex_lock(&root->fs_info->trans_mutex); put_transaction(cur_trans); ret = 0; -out_unlock: - mutex_unlock(&root->fs_info->trans_mutex); +out: return ret; } @@ -401,10 +414,8 @@ static void throttle_on_drops(struct btrfs_root *root) void btrfs_throttle(struct btrfs_root *root) { - mutex_lock(&root->fs_info->trans_mutex); - if (!root->fs_info->open_ioctl_trans) + if (!atomic_read(&root->fs_info->open_ioctl_trans)) wait_current_trans(root); - mutex_unlock(&root->fs_info->trans_mutex); } static int should_end_transaction(struct btrfs_trans_handle *trans, @@ -422,6 +433,7 @@ int btrfs_should_end_transaction(struct btrfs_trans_handle *trans, struct btrfs_transaction *cur_trans = trans->transaction; int updates; + smp_mb(); if (cur_trans->blocked || cur_trans->delayed_refs.flushing) return 1; @@ -467,9 +479,11 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans, btrfs_trans_release_metadata(trans, root); - if (lock && !root->fs_info->open_ioctl_trans && - should_end_transaction(trans, root)) + if (lock && !atomic_read(&root->fs_info->open_ioctl_trans) && + should_end_transaction(trans, root)) { trans->transaction->blocked = 1; + smp_wmb(); + } if (lock && cur_trans->blocked && !cur_trans->in_commit) { if (throttle) @@ -739,9 +753,9 @@ static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans, */ int btrfs_add_dead_root(struct btrfs_root *root) { - mutex_lock(&root->fs_info->trans_mutex); + spin_lock(&root->fs_info->trans_lock); list_add(&root->root_list, &root->fs_info->dead_roots); - mutex_unlock(&root->fs_info->trans_mutex); + spin_unlock(&root->fs_info->trans_lock); return 0; } @@ -757,6 +771,7 @@ static noinline int commit_fs_roots(struct btrfs_trans_handle *trans, int ret; int err = 0; + spin_lock(&fs_info->fs_roots_radix_lock); while (1) { ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix, (void **)gang, 0, @@ -769,6 +784,7 @@ static noinline int commit_fs_roots(struct btrfs_trans_handle *trans, radix_tree_tag_clear(&fs_info->fs_roots_radix, (unsigned long)root->root_key.objectid, BTRFS_ROOT_TRANS_TAG); + spin_unlock(&fs_info->fs_roots_radix_lock); btrfs_free_log(trans, root); btrfs_update_reloc_root(trans, root); @@ -783,10 +799,12 @@ static noinline int commit_fs_roots(struct btrfs_trans_handle *trans, err = btrfs_update_root(trans, fs_info->tree_root, &root->root_key, &root->root_item); + spin_lock(&fs_info->fs_roots_radix_lock); if (err) break; } } + spin_unlock(&fs_info->fs_roots_radix_lock); return err; } @@ -972,7 +990,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, parent = dget_parent(dentry); parent_inode = parent->d_inode; parent_root = BTRFS_I(parent_inode)->root; - record_root_in_trans(trans, parent_root); + btrfs_record_root_in_trans(trans, parent_root); /* * insert the directory item @@ -990,7 +1008,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans, ret = btrfs_update_inode(trans, parent_root, parent_inode); BUG_ON(ret); - record_root_in_trans(trans, root); + btrfs_record_root_in_trans(trans, root); btrfs_set_root_last_snapshot(&root->root_item, trans->transid); memcpy(new_root_item, &root->root_item, sizeof(*new_root_item)); btrfs_check_and_init_root_item(new_root_item); @@ -1080,20 +1098,20 @@ static void update_super_roots(struct btrfs_root *root) int btrfs_transaction_in_commit(struct btrfs_fs_info *info) { int ret = 0; - spin_lock(&info->new_trans_lock); + spin_lock(&info->trans_lock); if (info->running_transaction) ret = info->running_transaction->in_commit; - spin_unlock(&info->new_trans_lock); + spin_unlock(&info->trans_lock); return ret; } int btrfs_transaction_blocked(struct btrfs_fs_info *info) { int ret = 0; - spin_lock(&info->new_trans_lock); + spin_lock(&info->trans_lock); if (info->running_transaction) ret = info->running_transaction->blocked; - spin_unlock(&info->new_trans_lock); + spin_unlock(&info->trans_lock); return ret; } @@ -1117,9 +1135,7 @@ static void wait_current_trans_commit_start(struct btrfs_root *root, &wait); break; } - mutex_unlock(&root->fs_info->trans_mutex); schedule(); - mutex_lock(&root->fs_info->trans_mutex); finish_wait(&root->fs_info->transaction_blocked_wait, &wait); } } @@ -1145,9 +1161,7 @@ static void wait_current_trans_commit_start_and_unblock(struct btrfs_root *root, &wait); break; } - mutex_unlock(&root->fs_info->trans_mutex); schedule(); - mutex_lock(&root->fs_info->trans_mutex); finish_wait(&root->fs_info->transaction_wait, &wait); } @@ -1193,22 +1207,18 @@ int btrfs_commit_transaction_async(struct btrfs_trans_handle *trans, } /* take transaction reference */ - mutex_lock(&root->fs_info->trans_mutex); cur_trans = trans->transaction; atomic_inc(&cur_trans->use_count); - mutex_unlock(&root->fs_info->trans_mutex); btrfs_end_transaction(trans, root); schedule_delayed_work(&ac->work, 0); /* wait for transaction to start and unblock */ - mutex_lock(&root->fs_info->trans_mutex); if (wait_for_unblock) wait_current_trans_commit_start_and_unblock(root, cur_trans); else wait_current_trans_commit_start(root, cur_trans); put_transaction(cur_trans); - mutex_unlock(&root->fs_info->trans_mutex); return 0; } @@ -1252,38 +1262,41 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, ret = btrfs_run_delayed_refs(trans, root, 0); BUG_ON(ret); - mutex_lock(&root->fs_info->trans_mutex); + spin_lock(&cur_trans->commit_lock); if (cur_trans->in_commit) { + spin_unlock(&cur_trans->commit_lock); atomic_inc(&cur_trans->use_count); - mutex_unlock(&root->fs_info->trans_mutex); btrfs_end_transaction(trans, root); ret = wait_for_commit(root, cur_trans); BUG_ON(ret); - mutex_lock(&root->fs_info->trans_mutex); put_transaction(cur_trans); - mutex_unlock(&root->fs_info->trans_mutex); return 0; } trans->transaction->in_commit = 1; trans->transaction->blocked = 1; + spin_unlock(&cur_trans->commit_lock); wake_up(&root->fs_info->transaction_blocked_wait); + spin_lock(&root->fs_info->trans_lock); if (cur_trans->list.prev != &root->fs_info->trans_list) { prev_trans = list_entry(cur_trans->list.prev, struct btrfs_transaction, list); if (!prev_trans->commit_done) { atomic_inc(&prev_trans->use_count); - mutex_unlock(&root->fs_info->trans_mutex); + spin_unlock(&root->fs_info->trans_lock); wait_for_commit(root, prev_trans); - mutex_lock(&root->fs_info->trans_mutex); put_transaction(prev_trans); + } else { + spin_unlock(&root->fs_info->trans_lock); } + } else { + spin_unlock(&root->fs_info->trans_lock); } if (now < cur_trans->start_time || now - cur_trans->start_time < 1) @@ -1291,12 +1304,12 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, do { int snap_pending = 0; + joined = cur_trans->num_joined; if (!list_empty(&trans->transaction->pending_snapshots)) snap_pending = 1; WARN_ON(cur_trans != trans->transaction); - mutex_unlock(&root->fs_info->trans_mutex); if (flush_on_commit || snap_pending) { btrfs_start_delalloc_inodes(root, 1); @@ -1316,14 +1329,15 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, prepare_to_wait(&cur_trans->writer_wait, &wait, TASK_UNINTERRUPTIBLE); - smp_mb(); if (atomic_read(&cur_trans->num_writers) > 1) schedule_timeout(MAX_SCHEDULE_TIMEOUT); else if (should_grow) schedule_timeout(1); - mutex_lock(&root->fs_info->trans_mutex); finish_wait(&cur_trans->writer_wait, &wait); + spin_lock(&root->fs_info->trans_lock); + root->fs_info->trans_no_join = 1; + spin_unlock(&root->fs_info->trans_lock); } while (atomic_read(&cur_trans->num_writers) > 1 || (should_grow && cur_trans->num_joined != joined)); @@ -1364,9 +1378,6 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_prepare_extent_commit(trans, root); cur_trans = root->fs_info->running_transaction; - spin_lock(&root->fs_info->new_trans_lock); - root->fs_info->running_transaction = NULL; - spin_unlock(&root->fs_info->new_trans_lock); btrfs_set_root_node(&root->fs_info->tree_root->root_item, root->fs_info->tree_root->node); @@ -1387,10 +1398,13 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, sizeof(root->fs_info->super_copy)); trans->transaction->blocked = 0; + spin_lock(&root->fs_info->trans_lock); + root->fs_info->running_transaction = NULL; + root->fs_info->trans_no_join = 0; + spin_unlock(&root->fs_info->trans_lock); wake_up(&root->fs_info->transaction_wait); - mutex_unlock(&root->fs_info->trans_mutex); ret = btrfs_write_and_wait_transaction(trans, root); BUG_ON(ret); write_ctree_super(trans, root, 0); @@ -1403,22 +1417,21 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans, btrfs_finish_extent_commit(trans, root); - mutex_lock(&root->fs_info->trans_mutex); - cur_trans->commit_done = 1; root->fs_info->last_trans_committed = cur_trans->transid; wake_up(&cur_trans->commit_wait); + spin_lock(&root->fs_info->trans_lock); list_del_init(&cur_trans->list); + spin_unlock(&root->fs_info->trans_lock); + put_transaction(cur_trans); put_transaction(cur_trans); trace_btrfs_transaction_commit(root); - mutex_unlock(&root->fs_info->trans_mutex); - if (current->journal_info == trans) current->journal_info = NULL; @@ -1438,9 +1451,9 @@ int btrfs_clean_old_snapshots(struct btrfs_root *root) LIST_HEAD(list); struct btrfs_fs_info *fs_info = root->fs_info; - mutex_lock(&fs_info->trans_mutex); + spin_lock(&fs_info->trans_lock); list_splice_init(&fs_info->dead_roots, &list); - mutex_unlock(&fs_info->trans_mutex); + spin_unlock(&fs_info->trans_lock); while (!list_empty(&list)) { root = list_entry(list.next, struct btrfs_root, root_list); diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 154314f80f8d..11c6efcd4ed2 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -28,10 +28,12 @@ struct btrfs_transaction { * transaction can end */ atomic_t num_writers; + atomic_t use_count; unsigned long num_joined; + + spinlock_t commit_lock; int in_commit; - atomic_t use_count; int commit_done; int blocked; struct list_head list; From fcb80c2affd63237cff5b34cba5756be7c976a5a Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Tue, 3 May 2011 10:40:22 -0400 Subject: [PATCH 05/25] Btrfs: fix how we do space reservation for truncate The ceph guys keep running into problems where we have space reserved in our orphan block rsv when freeing it up. This is because they tend to do snapshots alot, so their truncates tend to use a bunch of space, so when we go to do things like update the inode we have to steal reservation space in order to make the reservation happen. This happens because truncate can use as much space as it freaking feels like, but we still have to hold space for removing the orphan item and updating the inode, which will definitely always happen. So in order to fix this we need to split all of the reservation stuf up. So with this patch we have 1) The orphan block reserve which only holds the space for deleting our orphan item when everything is over. 2) The truncate block reserve which gets allocated and used specifically for the space that the truncate will use on a per truncate basis. 3) The transaction will always have 1 item's worth of data reserved so we can update the inode normally. Hopefully this will make the ceph problem go away. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/ctree.h | 3 ++ fs/btrfs/extent-tree.c | 46 +++++++++++++---- fs/btrfs/inode.c | 111 +++++++++++++++++++++++++++++++---------- 3 files changed, 123 insertions(+), 37 deletions(-) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 522a39b0033d..f31aed7fedd9 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -2224,6 +2224,9 @@ int btrfs_block_rsv_migrate(struct btrfs_block_rsv *src_rsv, void btrfs_block_rsv_release(struct btrfs_root *root, struct btrfs_block_rsv *block_rsv, u64 num_bytes); +int btrfs_truncate_reserve_metadata(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_block_rsv *rsv); int btrfs_set_block_group_ro(struct btrfs_root *root, struct btrfs_block_group_cache *cache); int btrfs_set_block_group_rw(struct btrfs_root *root, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index ca599654ce19..a2ca561c70f0 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3980,6 +3980,37 @@ static u64 calc_trans_metadata_size(struct btrfs_root *root, int num_items) 3 * num_items; } +int btrfs_truncate_reserve_metadata(struct btrfs_trans_handle *trans, + struct btrfs_root *root, + struct btrfs_block_rsv *rsv) +{ + struct btrfs_block_rsv *trans_rsv = &root->fs_info->trans_block_rsv; + u64 num_bytes; + int ret; + + /* + * Truncate should be freeing data, but give us 2 items just in case it + * needs to use some space. We may want to be smarter about this in the + * future. + */ + num_bytes = calc_trans_metadata_size(root, 2); + + /* We already have enough bytes, just return */ + if (rsv->reserved >= num_bytes) + return 0; + + num_bytes -= rsv->reserved; + + /* + * You should have reserved enough space before hand to do this, so this + * should not fail. + */ + ret = block_rsv_migrate_bytes(trans_rsv, rsv, num_bytes); + BUG_ON(ret); + + return 0; +} + int btrfs_trans_reserve_metadata(struct btrfs_trans_handle *trans, struct btrfs_root *root, int num_items) @@ -4020,23 +4051,18 @@ int btrfs_orphan_reserve_metadata(struct btrfs_trans_handle *trans, struct btrfs_block_rsv *dst_rsv = root->orphan_block_rsv; /* - * one for deleting orphan item, one for updating inode and - * two for calling btrfs_truncate_inode_items. - * - * btrfs_truncate_inode_items is a delete operation, it frees - * more space than it uses in most cases. So two units of - * metadata space should be enough for calling it many times. - * If all of the metadata space is used, we can commit - * transaction and use space it freed. + * We need to hold space in order to delete our orphan item once we've + * added it, so this takes the reservation so we can release it later + * when we are truly done with the orphan item. */ - u64 num_bytes = calc_trans_metadata_size(root, 4); + u64 num_bytes = calc_trans_metadata_size(root, 1); return block_rsv_migrate_bytes(src_rsv, dst_rsv, num_bytes); } void btrfs_orphan_release_metadata(struct inode *inode) { struct btrfs_root *root = BTRFS_I(inode)->root; - u64 num_bytes = calc_trans_metadata_size(root, 4); + u64 num_bytes = calc_trans_metadata_size(root, 1); btrfs_block_rsv_release(root, root->orphan_block_rsv, num_bytes); } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index e47bdf0fb75a..bc12ba23db5f 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -6591,6 +6591,7 @@ int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) static int btrfs_truncate(struct inode *inode) { struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_block_rsv *rsv; int ret; int err = 0; struct btrfs_trans_handle *trans; @@ -6604,28 +6605,83 @@ static int btrfs_truncate(struct inode *inode) btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1); btrfs_ordered_update_i_size(inode, inode->i_size, NULL); - trans = btrfs_start_transaction(root, 5); - if (IS_ERR(trans)) - return PTR_ERR(trans); + /* + * Yes ladies and gentelment, this is indeed ugly. The fact is we have + * 3 things going on here + * + * 1) We need to reserve space for our orphan item and the space to + * delete our orphan item. Lord knows we don't want to have a dangling + * orphan item because we didn't reserve space to remove it. + * + * 2) We need to reserve space to update our inode. + * + * 3) We need to have something to cache all the space that is going to + * be free'd up by the truncate operation, but also have some slack + * space reserved in case it uses space during the truncate (thank you + * very much snapshotting). + * + * And we need these to all be seperate. The fact is we can use alot of + * space doing the truncate, and we have no earthly idea how much space + * we will use, so we need the truncate reservation to be seperate so it + * doesn't end up using space reserved for updating the inode or + * removing the orphan item. We also need to be able to stop the + * transaction and start a new one, which means we need to be able to + * update the inode several times, and we have no idea of knowing how + * many times that will be, so we can't just reserve 1 item for the + * entirety of the opration, so that has to be done seperately as well. + * Then there is the orphan item, which does indeed need to be held on + * to for the whole operation, and we need nobody to touch this reserved + * space except the orphan code. + * + * So that leaves us with + * + * 1) root->orphan_block_rsv - for the orphan deletion. + * 2) rsv - for the truncate reservation, which we will steal from the + * transaction reservation. + * 3) fs_info->trans_block_rsv - this will have 1 items worth left for + * updating the inode. + */ + rsv = btrfs_alloc_block_rsv(root); + if (!rsv) + return -ENOMEM; + btrfs_add_durable_block_rsv(root->fs_info, rsv); + + trans = btrfs_start_transaction(root, 4); + if (IS_ERR(trans)) { + err = PTR_ERR(trans); + goto out; + } btrfs_set_trans_block_group(trans, inode); + /* + * Reserve space for the truncate process. Truncate should be adding + * space, but if there are snapshots it may end up using space. + */ + ret = btrfs_truncate_reserve_metadata(trans, root, rsv); + BUG_ON(ret); + ret = btrfs_orphan_add(trans, inode); if (ret) { btrfs_end_transaction(trans, root); - return ret; + goto out; } nr = trans->blocks_used; btrfs_end_transaction(trans, root); btrfs_btree_balance_dirty(root, nr); - /* Now start a transaction for the truncate */ - trans = btrfs_start_transaction(root, 0); - if (IS_ERR(trans)) - return PTR_ERR(trans); + /* + * Ok so we've already migrated our bytes over for the truncate, so here + * just reserve the one slot we need for updating the inode. + */ + trans = btrfs_start_transaction(root, 1); + if (IS_ERR(trans)) { + err = PTR_ERR(trans); + goto out; + } btrfs_set_trans_block_group(trans, inode); - trans->block_rsv = root->orphan_block_rsv; + trans->block_rsv = rsv; /* * setattr is responsible for setting the ordered_data_close flag, @@ -6649,24 +6705,18 @@ static int btrfs_truncate(struct inode *inode) while (1) { if (!trans) { - trans = btrfs_start_transaction(root, 0); - if (IS_ERR(trans)) - return PTR_ERR(trans); - btrfs_set_trans_block_group(trans, inode); - trans->block_rsv = root->orphan_block_rsv; - } + trans = btrfs_start_transaction(root, 3); + if (IS_ERR(trans)) { + err = PTR_ERR(trans); + goto out; + } - ret = btrfs_block_rsv_check(trans, root, - root->orphan_block_rsv, 0, 5); - if (ret == -EAGAIN) { - ret = btrfs_commit_transaction(trans, root); - if (ret) - return ret; - trans = NULL; - continue; - } else if (ret) { - err = ret; - break; + ret = btrfs_truncate_reserve_metadata(trans, root, + rsv); + BUG_ON(ret); + + btrfs_set_trans_block_group(trans, inode); + trans->block_rsv = rsv; } ret = btrfs_truncate_inode_items(trans, root, inode, @@ -6677,6 +6727,7 @@ static int btrfs_truncate(struct inode *inode) break; } + trans->block_rsv = &root->fs_info->trans_block_rsv; ret = btrfs_update_inode(trans, root, inode); if (ret) { err = ret; @@ -6690,6 +6741,7 @@ static int btrfs_truncate(struct inode *inode) } if (ret == 0 && inode->i_nlink > 0) { + trans->block_rsv = root->orphan_block_rsv; ret = btrfs_orphan_del(trans, inode); if (ret) err = ret; @@ -6701,15 +6753,20 @@ static int btrfs_truncate(struct inode *inode) ret = btrfs_orphan_del(NULL, inode); } + trans->block_rsv = &root->fs_info->trans_block_rsv; ret = btrfs_update_inode(trans, root, inode); if (ret && !err) err = ret; nr = trans->blocks_used; ret = btrfs_end_transaction_throttle(trans, root); + btrfs_btree_balance_dirty(root, nr); + +out: + btrfs_free_block_rsv(root, rsv); + if (ret && !err) err = ret; - btrfs_btree_balance_dirty(root, nr); return err; } From af60bed24eb0e3b6d93eaa6bb395a5721e6c09a8 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 4 May 2011 11:11:17 -0400 Subject: [PATCH 06/25] Btrfs: set range_start to the right start in count_range_bits In count_range_bits we are adjusting total_bytes based on the range we are searching for, but we don't adjust the range start according to the range we are searching for, which makes for weird results. For example, if the range [0-8192] is set DELALLOC, but I search for 4096-8192, I will get back 4096 for the number of bytes found, but the range_start will be 0, which makes it look like the range is [0-4096]. So instead set range_start = max(cur_start, state->start). This makes everything come out right. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/extent_io.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index ba41da59e31b..b5f6f227a97c 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -1480,7 +1480,7 @@ u64 count_range_bits(struct extent_io_tree *tree, if (total_bytes >= max_bytes) break; if (!found) { - *start = state->start; + *start = max(cur_start, state->start); found = 1; } last = state->end; From cb25c2ea6a79702ab7895b873c6c43e0d3bc3c72 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 11 May 2011 12:17:34 -0400 Subject: [PATCH 07/25] Btrfs: map the node block when looking for readahead targets If we have particularly full nodes, we could call btrfs_node_blockptr up to 32 times, which is 32 pairs of kmap/kunmap, which _sucks_. So go ahead and map the extent buffer while we look for readahead targets. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/ctree.c | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-) diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 84d7ca1fe0ba..009bcf7f1e4b 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1229,6 +1229,7 @@ static void reada_for_search(struct btrfs_root *root, u64 search; u64 target; u64 nread = 0; + u64 gen; int direction = path->reada; struct extent_buffer *eb; u32 nr; @@ -1256,6 +1257,15 @@ static void reada_for_search(struct btrfs_root *root, nritems = btrfs_header_nritems(node); nr = slot; while (1) { + if (!node->map_token) { + unsigned long offset = btrfs_node_key_ptr_offset(nr); + map_private_extent_buffer(node, offset, + sizeof(struct btrfs_key_ptr), + &node->map_token, + &node->kaddr, + &node->map_start, + &node->map_len, KM_USER1); + } if (direction < 0) { if (nr == 0) break; @@ -1273,14 +1283,23 @@ static void reada_for_search(struct btrfs_root *root, search = btrfs_node_blockptr(node, nr); if ((search <= target && target - search <= 65536) || (search > target && search - target <= 65536)) { - readahead_tree_block(root, search, blocksize, - btrfs_node_ptr_generation(node, nr)); + gen = btrfs_node_ptr_generation(node, nr); + if (node->map_token) { + unmap_extent_buffer(node, node->map_token, + KM_USER1); + node->map_token = NULL; + } + readahead_tree_block(root, search, blocksize, gen); nread += blocksize; } nscan++; if ((nread > 65536 || nscan > 32)) break; } + if (node->map_token) { + unmap_extent_buffer(node, node->map_token, KM_USER1); + node->map_token = NULL; + } } /* From 7e2355ba1a11649f0b212a29fdb9f47476f1248e Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 11 May 2011 12:25:37 -0400 Subject: [PATCH 08/25] Btrfs: don't look at the extent buffer level 3 times in a row We have a bit of debugging in btrfs_search_slot to make sure the level of the cow block is the same as the original block we were cow'ing. I don't think I've ever seen this tripped, so kill it. This saves us 2 kmap's per level in our search. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/ctree.c | 3 --- 1 file changed, 3 deletions(-) diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 009bcf7f1e4b..f7a0a64b868f 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -1672,9 +1672,6 @@ int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root } cow_done: BUG_ON(!cow && ins_len); - if (level != btrfs_header_level(b)) - WARN_ON(1); - level = btrfs_header_level(b); p->nodes[level] = b; if (!p->skip_locking) From d82a6f1d7e8b61ed5996334d0db66651bb43641d Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 11 May 2011 15:26:06 -0400 Subject: [PATCH 09/25] Btrfs: kill BTRFS_I(inode)->block_group Originally this was going to be used as a way to give hints to the allocator, but frankly we can get much better hints elsewhere and it's not even used at all for anything usefull. In addition to be completely useless, when we initialize an inode we try and find a freeish block group to set as the inodes block group, and with a completely full 40gb fs this takes _forever_, so I imagine with say 1tb fs this is just unbearable. So just axe the thing altoghether, we don't need it and it saves us 8 bytes in the inode and saves us 500 microseconds per inode lookup in my testcase. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/btrfs_inode.h | 3 -- fs/btrfs/ctree.h | 3 +- fs/btrfs/extent-tree.c | 10 +---- fs/btrfs/inode.c | 87 +++++------------------------------------- fs/btrfs/ioctl.c | 3 +- fs/btrfs/transaction.c | 1 - fs/btrfs/transaction.h | 14 ------- fs/btrfs/xattr.c | 2 - 8 files changed, 13 insertions(+), 110 deletions(-) diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h index 57c3bb2884ce..4bc852d3b83d 100644 --- a/fs/btrfs/btrfs_inode.h +++ b/fs/btrfs/btrfs_inode.h @@ -120,9 +120,6 @@ struct btrfs_inode { */ u64 index_cnt; - /* the start of block group preferred for allocations. */ - u64 block_group; - /* the fsync log has some corner cases that mean we have to check * directories to see if any unlinks have been done before * the directory was logged. See tree-log.c for all the diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index f31aed7fedd9..0f8c489bcc02 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -2512,8 +2512,7 @@ int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end, int btrfs_writepages(struct address_space *mapping, struct writeback_control *wbc); int btrfs_create_subvol_root(struct btrfs_trans_handle *trans, - struct btrfs_root *new_root, - u64 new_dirid, u64 alloc_hint); + struct btrfs_root *new_root, u64 new_dirid); int btrfs_merge_bio_hook(struct page *page, unsigned long offset, size_t size, struct bio *bio, unsigned long bio_flags); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index a2ca561c70f0..9f0a4e3bd8a9 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -5319,6 +5319,7 @@ static noinline int find_free_extent(struct btrfs_trans_handle *trans, btrfs_add_free_space(block_group, offset, search_start - offset); BUG_ON(offset > search_start); + btrfs_put_block_group(block_group); break; loop: failed_cluster_refill = false; @@ -5411,14 +5412,7 @@ static noinline int find_free_extent(struct btrfs_trans_handle *trans, ret = -ENOSPC; } else if (!ins->objectid) { ret = -ENOSPC; - } - - /* we found what we needed */ - if (ins->objectid) { - if (!(data & BTRFS_BLOCK_GROUP_DATA)) - trans->block_group = block_group->key.objectid; - - btrfs_put_block_group(block_group); + } else if (ins->objectid) { ret = 0; } diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index bc12ba23db5f..dd5938a7de21 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -136,7 +136,6 @@ static noinline int insert_inline_extent(struct btrfs_trans_handle *trans, return -ENOMEM; path->leave_spinning = 1; - btrfs_set_trans_block_group(trans, inode); key.objectid = inode->i_ino; key.offset = start; @@ -422,7 +421,6 @@ static noinline int compress_file_range(struct inode *inode, if (start == 0) { trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); - btrfs_set_trans_block_group(trans, inode); trans->block_rsv = &root->fs_info->delalloc_block_rsv; /* lets try to make an inline extent */ @@ -781,7 +779,6 @@ static noinline int cow_file_range(struct inode *inode, BUG_ON(root == root->fs_info->tree_root); trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); - btrfs_set_trans_block_group(trans, inode); trans->block_rsv = &root->fs_info->delalloc_block_rsv; num_bytes = (end - start + blocksize) & ~(blocksize - 1); @@ -1502,8 +1499,6 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans, { struct btrfs_ordered_sum *sum; - btrfs_set_trans_block_group(trans, inode); - list_for_each_entry(sum, list, list) { btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root->fs_info->csum_root, sum); @@ -1722,7 +1717,6 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) else trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); - btrfs_set_trans_block_group(trans, inode); trans->block_rsv = &root->fs_info->delalloc_block_rsv; ret = btrfs_update_inode(trans, root, inode); BUG_ON(ret); @@ -1739,7 +1733,6 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end) else trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); - btrfs_set_trans_block_group(trans, inode); trans->block_rsv = &root->fs_info->delalloc_block_rsv; if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags)) @@ -2495,7 +2488,6 @@ static void btrfs_read_locked_inode(struct inode *inode) struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_key location; int maybe_acls; - u64 alloc_group_block; u32 rdev; int ret; @@ -2539,8 +2531,6 @@ static void btrfs_read_locked_inode(struct inode *inode) BTRFS_I(inode)->index_cnt = (u64)-1; BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item); - alloc_group_block = btrfs_inode_block_group(leaf, inode_item); - /* * try to precache a NULL acl entry for files that don't have * any xattrs or acls @@ -2549,8 +2539,6 @@ static void btrfs_read_locked_inode(struct inode *inode) if (!maybe_acls) cache_no_acl(inode); - BTRFS_I(inode)->block_group = btrfs_find_block_group(root, 0, - alloc_group_block, 0); btrfs_free_path(path); inode_item = NULL; @@ -2630,7 +2618,7 @@ static void fill_inode_item(struct btrfs_trans_handle *trans, btrfs_set_inode_transid(leaf, item, trans->transid); btrfs_set_inode_rdev(leaf, item, inode->i_rdev); btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags); - btrfs_set_inode_block_group(leaf, item, BTRFS_I(inode)->block_group); + btrfs_set_inode_block_group(leaf, item, 0); if (leaf->map_token) { unmap_extent_buffer(leaf, leaf->map_token, KM_USER1); @@ -2971,8 +2959,6 @@ static int btrfs_unlink(struct inode *dir, struct dentry *dentry) if (IS_ERR(trans)) return PTR_ERR(trans); - btrfs_set_trans_block_group(trans, dir); - btrfs_record_unlink_dir(trans, dir, dentry->d_inode, 0); ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode, @@ -3068,8 +3054,6 @@ static int btrfs_rmdir(struct inode *dir, struct dentry *dentry) if (IS_ERR(trans)) return PTR_ERR(trans); - btrfs_set_trans_block_group(trans, dir); - if (unlikely(inode->i_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) { err = btrfs_unlink_subvol(trans, root, dir, BTRFS_I(inode)->location.objectid, @@ -3649,7 +3633,6 @@ int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size) err = PTR_ERR(trans); break; } - btrfs_set_trans_block_group(trans, inode); err = btrfs_drop_extents(trans, inode, cur_offset, cur_offset + hole_size, @@ -3785,7 +3768,6 @@ void btrfs_evict_inode(struct inode *inode) while (1) { trans = btrfs_start_transaction(root, 0); BUG_ON(IS_ERR(trans)); - btrfs_set_trans_block_group(trans, inode); trans->block_rsv = root->orphan_block_rsv; ret = btrfs_block_rsv_check(trans, root, @@ -4383,7 +4365,6 @@ int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc) trans = btrfs_join_transaction(root); if (IS_ERR(trans)) return PTR_ERR(trans); - btrfs_set_trans_block_group(trans, inode); if (nolock) ret = btrfs_end_transaction_nolock(trans, root); else @@ -4409,7 +4390,6 @@ void btrfs_dirty_inode(struct inode *inode) trans = btrfs_join_transaction(root); BUG_ON(IS_ERR(trans)); - btrfs_set_trans_block_group(trans, inode); ret = btrfs_update_inode(trans, root, inode); if (ret && ret == -ENOSPC) { @@ -4424,7 +4404,6 @@ void btrfs_dirty_inode(struct inode *inode) } return; } - btrfs_set_trans_block_group(trans, inode); ret = btrfs_update_inode(trans, root, inode); if (ret) { @@ -4519,8 +4498,8 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *dir, const char *name, int name_len, - u64 ref_objectid, u64 objectid, - u64 alloc_hint, int mode, u64 *index) + u64 ref_objectid, u64 objectid, int mode, + u64 *index) { struct inode *inode; struct btrfs_inode_item *inode_item; @@ -4567,8 +4546,6 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans, owner = 0; else owner = 1; - BTRFS_I(inode)->block_group = - btrfs_find_block_group(root, 0, alloc_hint, owner); key[0].objectid = objectid; btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY); @@ -4729,11 +4706,9 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, if (IS_ERR(trans)) return PTR_ERR(trans); - btrfs_set_trans_block_group(trans, dir); - inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dir->i_ino, objectid, - BTRFS_I(dir)->block_group, mode, &index); + mode, &index); if (IS_ERR(inode)) { err = PTR_ERR(inode); goto out_unlock; @@ -4745,7 +4720,6 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, goto out_unlock; } - btrfs_set_trans_block_group(trans, inode); err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index); if (err) drop_inode = 1; @@ -4754,8 +4728,6 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, init_special_inode(inode, inode->i_mode, rdev); btrfs_update_inode(trans, root, inode); } - btrfs_update_inode_block_group(trans, inode); - btrfs_update_inode_block_group(trans, dir); out_unlock: nr = trans->blocks_used; btrfs_end_transaction_throttle(trans, root); @@ -4791,11 +4763,9 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, if (IS_ERR(trans)) return PTR_ERR(trans); - btrfs_set_trans_block_group(trans, dir); - inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dir->i_ino, objectid, - BTRFS_I(dir)->block_group, mode, &index); + mode, &index); if (IS_ERR(inode)) { err = PTR_ERR(inode); goto out_unlock; @@ -4807,7 +4777,6 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, goto out_unlock; } - btrfs_set_trans_block_group(trans, inode); err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index); if (err) drop_inode = 1; @@ -4818,8 +4787,6 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_file_inode_operations; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } - btrfs_update_inode_block_group(trans, inode); - btrfs_update_inode_block_group(trans, dir); out_unlock: nr = trans->blocks_used; btrfs_end_transaction_throttle(trans, root); @@ -4866,8 +4833,6 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, btrfs_inc_nlink(inode); inode->i_ctime = CURRENT_TIME; - - btrfs_set_trans_block_group(trans, dir); ihold(inode); err = btrfs_add_nondir(trans, dir, dentry, inode, 1, index); @@ -4876,7 +4841,6 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, drop_inode = 1; } else { struct dentry *parent = dget_parent(dentry); - btrfs_update_inode_block_group(trans, dir); err = btrfs_update_inode(trans, root, inode); BUG_ON(err); btrfs_log_new_name(trans, inode, NULL, parent); @@ -4917,12 +4881,10 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) trans = btrfs_start_transaction(root, 5); if (IS_ERR(trans)) return PTR_ERR(trans); - btrfs_set_trans_block_group(trans, dir); inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dir->i_ino, objectid, - BTRFS_I(dir)->block_group, S_IFDIR | mode, - &index); + S_IFDIR | mode, &index); if (IS_ERR(inode)) { err = PTR_ERR(inode); goto out_fail; @@ -4936,7 +4898,6 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) inode->i_op = &btrfs_dir_inode_operations; inode->i_fop = &btrfs_dir_file_operations; - btrfs_set_trans_block_group(trans, inode); btrfs_i_size_write(inode, 0); err = btrfs_update_inode(trans, root, inode); @@ -4950,8 +4911,6 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) d_instantiate(dentry, inode); drop_on_err = 0; - btrfs_update_inode_block_group(trans, inode); - btrfs_update_inode_block_group(trans, dir); out_fail: nr = trans->blocks_used; @@ -6652,8 +6611,6 @@ static int btrfs_truncate(struct inode *inode) goto out; } - btrfs_set_trans_block_group(trans, inode); - /* * Reserve space for the truncate process. Truncate should be adding * space, but if there are snapshots it may end up using space. @@ -6680,7 +6637,6 @@ static int btrfs_truncate(struct inode *inode) err = PTR_ERR(trans); goto out; } - btrfs_set_trans_block_group(trans, inode); trans->block_rsv = rsv; /* @@ -6715,7 +6671,6 @@ static int btrfs_truncate(struct inode *inode) rsv); BUG_ON(ret); - btrfs_set_trans_block_group(trans, inode); trans->block_rsv = rsv; } @@ -6775,15 +6730,14 @@ static int btrfs_truncate(struct inode *inode) * create a new subvolume directory/inode (helper for the ioctl). */ int btrfs_create_subvol_root(struct btrfs_trans_handle *trans, - struct btrfs_root *new_root, - u64 new_dirid, u64 alloc_hint) + struct btrfs_root *new_root, u64 new_dirid) { struct inode *inode; int err; u64 index = 0; inode = btrfs_new_inode(trans, new_root, NULL, "..", 2, new_dirid, - new_dirid, alloc_hint, S_IFDIR | 0700, &index); + new_dirid, S_IFDIR | 0700, &index); if (IS_ERR(inode)) return PTR_ERR(inode); inode->i_op = &btrfs_dir_inode_operations; @@ -6893,21 +6847,6 @@ void btrfs_destroy_inode(struct inode *inode) spin_unlock(&root->fs_info->ordered_extent_lock); } - if (root == root->fs_info->tree_root) { - struct btrfs_block_group_cache *block_group; - - block_group = btrfs_lookup_block_group(root->fs_info, - BTRFS_I(inode)->block_group); - if (block_group && block_group->inode == inode) { - spin_lock(&block_group->lock); - block_group->inode = NULL; - spin_unlock(&block_group->lock); - btrfs_put_block_group(block_group); - } else if (block_group) { - btrfs_put_block_group(block_group); - } - } - spin_lock(&root->orphan_lock); if (!list_empty(&BTRFS_I(inode)->i_orphan)) { printk(KERN_INFO "BTRFS: inode %lu still on the orphan list\n", @@ -7091,8 +7030,6 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry, goto out_notrans; } - btrfs_set_trans_block_group(trans, new_dir); - if (dest != root) btrfs_record_root_in_trans(trans, dest); @@ -7331,12 +7268,9 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, if (IS_ERR(trans)) return PTR_ERR(trans); - btrfs_set_trans_block_group(trans, dir); - inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, dentry->d_name.len, dir->i_ino, objectid, - BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO, - &index); + S_IFLNK|S_IRWXUGO, &index); if (IS_ERR(inode)) { err = PTR_ERR(inode); goto out_unlock; @@ -7348,7 +7282,6 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, goto out_unlock; } - btrfs_set_trans_block_group(trans, inode); err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index); if (err) drop_inode = 1; @@ -7359,8 +7292,6 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_file_inode_operations; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } - btrfs_update_inode_block_group(trans, inode); - btrfs_update_inode_block_group(trans, dir); if (drop_inode) goto out_unlock; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index a578620e06a8..8e90ccf4b76a 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -413,8 +413,7 @@ static noinline int create_subvol(struct btrfs_root *root, btrfs_record_root_in_trans(trans, new_root); - ret = btrfs_create_subvol_root(trans, new_root, new_dirid, - BTRFS_I(dir)->block_group); + ret = btrfs_create_subvol_root(trans, new_root, new_dirid); /* * insert the directory item */ diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 43816f8b23e7..f4ea695325b2 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -241,7 +241,6 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root, h->transid = cur_trans->transid; h->transaction = cur_trans; h->blocks_used = 0; - h->block_group = 0; h->bytes_reserved = 0; h->delayed_ref_updates = 0; h->use_count = 1; diff --git a/fs/btrfs/transaction.h b/fs/btrfs/transaction.h index 11c6efcd4ed2..da7289e06a82 100644 --- a/fs/btrfs/transaction.h +++ b/fs/btrfs/transaction.h @@ -47,7 +47,6 @@ struct btrfs_transaction { struct btrfs_trans_handle { u64 transid; - u64 block_group; u64 bytes_reserved; unsigned long use_count; unsigned long blocks_reserved; @@ -70,19 +69,6 @@ struct btrfs_pending_snapshot { struct list_head list; }; -static inline void btrfs_set_trans_block_group(struct btrfs_trans_handle *trans, - struct inode *inode) -{ - trans->block_group = BTRFS_I(inode)->block_group; -} - -static inline void btrfs_update_inode_block_group( - struct btrfs_trans_handle *trans, - struct inode *inode) -{ - BTRFS_I(inode)->block_group = trans->block_group; -} - static inline void btrfs_set_inode_last_trans(struct btrfs_trans_handle *trans, struct inode *inode) { diff --git a/fs/btrfs/xattr.c b/fs/btrfs/xattr.c index cfd660550ded..72ab0295ca74 100644 --- a/fs/btrfs/xattr.c +++ b/fs/btrfs/xattr.c @@ -158,8 +158,6 @@ int __btrfs_setxattr(struct btrfs_trans_handle *trans, if (IS_ERR(trans)) return PTR_ERR(trans); - btrfs_set_trans_block_group(trans, inode); - ret = do_setxattr(trans, inode, name, value, size, flags); if (ret) goto out; From 589d8ade83f07c0f11c8191c0ca309f34d7a2c14 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Wed, 11 May 2011 17:30:53 -0400 Subject: [PATCH 10/25] Btrfs: try not to sleep as much when doing slow caching When the fs is super full and we unmount the fs, we could get stuck in this thing where unmount is waiting for the caching kthread to make progress and the caching kthread keeps scheduling because we're in the middle of a commit. So instead just let the caching kthread keep going and only yeild if need_resched(). This makes my horrible umount case go from taking up to 10 minutes to taking less than 20 seconds. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/extent-tree.c | 19 +++++++++++-------- 1 file changed, 11 insertions(+), 8 deletions(-) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 9f0a4e3bd8a9..96be62450318 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -378,15 +378,18 @@ static int caching_kthread(void *data) if (ret) break; - caching_ctl->progress = last; - btrfs_release_path(extent_root, path); - up_read(&fs_info->extent_commit_sem); - mutex_unlock(&caching_ctl->mutex); - if (btrfs_transaction_in_commit(fs_info)) - schedule_timeout(1); - else + if (need_resched() || + btrfs_next_leaf(extent_root, path)) { + caching_ctl->progress = last; + btrfs_release_path(extent_root, path); + up_read(&fs_info->extent_commit_sem); + mutex_unlock(&caching_ctl->mutex); cond_resched(); - goto again; + goto again; + } + leaf = path->nodes[0]; + nritems = btrfs_header_nritems(leaf); + continue; } if (key.objectid < block_group->key.objectid) { From 026fd317828500524cdc7e5ff9e8e7923abb2868 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Fri, 13 May 2011 10:32:11 -0400 Subject: [PATCH 11/25] Btrfs: don't always do readahead Our readahead is sort of sloppy, and really isn't always needed. For example if ls is doing a stating ls (which is the default) it's going to stat in non-disk order, so if say you have a directory with a stupid amount of files, readahead is going to do nothing but waste time in the case of doing the stat. Taking the unconditional readahead out made my test go from 57 minutes to 36 minutes. This means that everywhere we do loop through the tree we want to make sure we do set path->reada properly, so I went through and found all of the places where we loop through the path and set reada to 1. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/ctree.c | 2 -- fs/btrfs/extent-tree.c | 3 ++- fs/btrfs/inode.c | 14 ++++++++++++-- fs/btrfs/relocation.c | 6 ++++++ 4 files changed, 20 insertions(+), 5 deletions(-) diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index f7a0a64b868f..f61c16c1481a 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c @@ -48,8 +48,6 @@ struct btrfs_path *btrfs_alloc_path(void) { struct btrfs_path *path; path = kmem_cache_zalloc(btrfs_path_cachep, GFP_NOFS); - if (path) - path->reada = 1; return path; } diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 96be62450318..1ba2cc58eab5 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -347,7 +347,7 @@ static int caching_kthread(void *data) */ path->skip_locking = 1; path->search_commit_root = 1; - path->reada = 2; + path->reada = 1; key.objectid = last; key.offset = 0; @@ -8556,6 +8556,7 @@ int btrfs_read_block_groups(struct btrfs_root *root) path = btrfs_alloc_path(); if (!path) return -ENOMEM; + path->reada = 1; cache_gen = btrfs_super_cache_generation(&root->fs_info->super_copy); if (cache_gen != 0 && diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index dd5938a7de21..6228a304b547 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -4242,7 +4242,9 @@ static int btrfs_real_readdir(struct file *filp, void *dirent, filp->f_pos = 2; } path = btrfs_alloc_path(); - path->reada = 2; + if (!path) + return -ENOMEM; + path->reada = 1; btrfs_set_key_type(&key, key_type); key.offset = filp->f_pos; @@ -5043,7 +5045,15 @@ struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page, if (!path) { path = btrfs_alloc_path(); - BUG_ON(!path); + if (!path) { + err = -ENOMEM; + goto out; + } + /* + * Chances are we'll be called again, so go ahead and do + * readahead + */ + path->reada = 1; } ret = btrfs_lookup_file_extent(trans, root, path, diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 09c30d37d43e..5872b41581f4 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -676,6 +676,8 @@ struct backref_node *build_backref_tree(struct reloc_control *rc, err = -ENOMEM; goto out; } + path1->reada = 1; + path2->reada = 2; node = alloc_backref_node(cache); if (!node) { @@ -1996,6 +1998,7 @@ static noinline_for_stack int merge_reloc_root(struct reloc_control *rc, path = btrfs_alloc_path(); if (!path) return -ENOMEM; + path->reada = 1; reloc_root = root->reloc_root; root_item = &reloc_root->root_item; @@ -3297,6 +3300,7 @@ static int find_data_references(struct reloc_control *rc, path = btrfs_alloc_path(); if (!path) return -ENOMEM; + path->reada = 1; root = read_fs_root(rc->extent_root->fs_info, ref_root); if (IS_ERR(root)) { @@ -3665,6 +3669,7 @@ static noinline_for_stack int relocate_block_group(struct reloc_control *rc) path = btrfs_alloc_path(); if (!path) return -ENOMEM; + path->reada = 1; ret = prepare_to_relocate(rc); if (ret) { @@ -4090,6 +4095,7 @@ int btrfs_recover_relocation(struct btrfs_root *root) path = btrfs_alloc_path(); if (!path) return -ENOMEM; + path->reada = -1; key.objectid = BTRFS_TREE_RELOC_OBJECTID; key.type = BTRFS_ROOT_ITEM_KEY; From cca1c81f43e26ab60c0d1090fb90992358d69bdf Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Fri, 13 May 2011 11:07:12 -0400 Subject: [PATCH 12/25] Btrfs: don't try to allocate from a block group that doesn't have enough space If we have a very large filesystem, we can spend a lot of time in find_free_extent just trying to allocate from empty block groups. So instead check to see if the block group even has enough space for the allocation, and if not go on to the next block group. Signed-off-by: Josef Bacik --- fs/btrfs/extent-tree.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 1ba2cc58eab5..c8c318494dee 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -5159,6 +5159,14 @@ static noinline int find_free_extent(struct btrfs_trans_handle *trans, if (unlikely(block_group->ro)) goto loop; + spin_lock(&block_group->tree_lock); + if (cached && + block_group->free_space < num_bytes + empty_size) { + spin_unlock(&block_group->tree_lock); + goto loop; + } + spin_unlock(&block_group->tree_lock); + /* * Ok we want to try and use the cluster allocator, so lets look * there, unless we are on LOOP_NO_EMPTY_SIZE, since we will From 207dde8289d9b005b665cb9d8d2bb9464256101d Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Fri, 13 May 2011 14:49:23 -0400 Subject: [PATCH 13/25] Btrfs: check for duplicate entries in the free space cache If there are duplicate entries in the free space cache, discard the entire cache and load it the old fashioned way. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/free-space-cache.c | 27 ++++++++++++++++++++++++--- 1 file changed, 24 insertions(+), 3 deletions(-) diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index 63731a1fb0a1..d634a7e42207 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -420,7 +420,14 @@ int load_free_space_cache(struct btrfs_fs_info *fs_info, spin_lock(&block_group->tree_lock); ret = link_free_space(block_group, e); spin_unlock(&block_group->tree_lock); - BUG_ON(ret); + if (ret) { + printk(KERN_ERR "Duplicate entries in " + "free space cache, dumping\n"); + kunmap(page); + unlock_page(page); + page_cache_release(page); + goto free_cache; + } } else { e->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS); if (!e->bitmap) { @@ -437,6 +444,14 @@ int load_free_space_cache(struct btrfs_fs_info *fs_info, recalculate_thresholds(block_group); spin_unlock(&block_group->tree_lock); list_add_tail(&e->list, &bitmaps); + if (ret) { + printk(KERN_ERR "Duplicate entries in " + "free space cache, dumping\n"); + kunmap(page); + unlock_page(page); + page_cache_release(page); + goto free_cache; + } } num_entries--; @@ -909,10 +924,16 @@ static int tree_insert_offset(struct rb_root *root, u64 offset, * logically. */ if (bitmap) { - WARN_ON(info->bitmap); + if (info->bitmap) { + WARN_ON_ONCE(1); + return -EEXIST; + } p = &(*p)->rb_right; } else { - WARN_ON(!info->bitmap); + if (!info->bitmap) { + WARN_ON_ONCE(1); + return -EEXIST; + } p = &(*p)->rb_left; } } From d90c732122a1f6d0efe388a8a204f67f144b2eb3 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Tue, 17 May 2011 09:50:54 -0400 Subject: [PATCH 14/25] Btrfs: leave spinning on lookup and map the leaf On lookup we only want to read the inode item, so leave the path spinning. Also we're just wholesale reading the leaf off, so map the leaf so we don't do a bunch of kmap/kunmaps. Thanks, Signed-off-by: Josef Bacik --- fs/btrfs/inode.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 6228a304b547..dc8fb2b3a145 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2493,6 +2493,7 @@ static void btrfs_read_locked_inode(struct inode *inode) path = btrfs_alloc_path(); BUG_ON(!path); + path->leave_spinning = 1; memcpy(&location, &BTRFS_I(inode)->location, sizeof(location)); ret = btrfs_lookup_inode(NULL, root, path, &location, 0); @@ -2502,6 +2503,12 @@ static void btrfs_read_locked_inode(struct inode *inode) leaf = path->nodes[0]; inode_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item); + if (!leaf->map_token) + map_private_extent_buffer(leaf, (unsigned long)inode_item, + sizeof(struct btrfs_inode_item), + &leaf->map_token, &leaf->kaddr, + &leaf->map_start, &leaf->map_len, + KM_USER1); inode->i_mode = btrfs_inode_mode(leaf, inode_item); inode->i_nlink = btrfs_inode_nlink(leaf, inode_item); @@ -2539,6 +2546,11 @@ static void btrfs_read_locked_inode(struct inode *inode) if (!maybe_acls) cache_no_acl(inode); + if (leaf->map_token) { + unmap_extent_buffer(leaf, leaf->map_token, KM_USER1); + leaf->map_token = NULL; + } + btrfs_free_path(path); inode_item = NULL; From 1bc8779349d6278e2713a1ff94418c2a6746a791 Mon Sep 17 00:00:00 2001 From: Arne Jansen Date: Sat, 28 May 2011 21:57:55 +0200 Subject: [PATCH 15/25] btrfs: scrub: don't reuse bios and pages The current scrub implementation reuses bios and pages as often as possible, allocating them only on start and releasing them when finished. This leads to more problems with the block layer than it's worth. The elevator gets confused when there are more pages added to the bio than bi_size suggests. This patch completely rips out the reuse of bios and pages and allocates them freshly for each submit. Signed-off-by: Arne Jansen Signed-off-by: Chris Maosn --- fs/btrfs/scrub.c | 114 +++++++++++++++++++++++++++-------------------- 1 file changed, 65 insertions(+), 49 deletions(-) diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c index 6dfed0c27ac3..2d1f8909a8e1 100644 --- a/fs/btrfs/scrub.c +++ b/fs/btrfs/scrub.c @@ -117,33 +117,37 @@ static void scrub_free_csums(struct scrub_dev *sdev) } } +static void scrub_free_bio(struct bio *bio) +{ + int i; + struct page *last_page = NULL; + + if (!bio) + return; + + for (i = 0; i < bio->bi_vcnt; ++i) { + if (bio->bi_io_vec[i].bv_page == last_page) + continue; + last_page = bio->bi_io_vec[i].bv_page; + __free_page(last_page); + } + bio_put(bio); +} + static noinline_for_stack void scrub_free_dev(struct scrub_dev *sdev) { int i; - int j; - struct page *last_page; if (!sdev) return; for (i = 0; i < SCRUB_BIOS_PER_DEV; ++i) { struct scrub_bio *sbio = sdev->bios[i]; - struct bio *bio; if (!sbio) break; - bio = sbio->bio; - if (bio) { - last_page = NULL; - for (j = 0; j < bio->bi_vcnt; ++j) { - if (bio->bi_io_vec[j].bv_page == last_page) - continue; - last_page = bio->bi_io_vec[j].bv_page; - __free_page(last_page); - } - bio_put(bio); - } + scrub_free_bio(sbio->bio); kfree(sbio); } @@ -156,8 +160,6 @@ struct scrub_dev *scrub_setup_dev(struct btrfs_device *dev) { struct scrub_dev *sdev; int i; - int j; - int ret; struct btrfs_fs_info *fs_info = dev->dev_root->fs_info; sdev = kzalloc(sizeof(*sdev), GFP_NOFS); @@ -165,7 +167,6 @@ struct scrub_dev *scrub_setup_dev(struct btrfs_device *dev) goto nomem; sdev->dev = dev; for (i = 0; i < SCRUB_BIOS_PER_DEV; ++i) { - struct bio *bio; struct scrub_bio *sbio; sbio = kzalloc(sizeof(*sbio), GFP_NOFS); @@ -173,32 +174,10 @@ struct scrub_dev *scrub_setup_dev(struct btrfs_device *dev) goto nomem; sdev->bios[i] = sbio; - bio = bio_kmalloc(GFP_NOFS, SCRUB_PAGES_PER_BIO); - if (!bio) - goto nomem; - sbio->index = i; sbio->sdev = sdev; - sbio->bio = bio; sbio->count = 0; sbio->work.func = scrub_checksum; - bio->bi_private = sdev->bios[i]; - bio->bi_end_io = scrub_bio_end_io; - bio->bi_sector = 0; - bio->bi_bdev = dev->bdev; - bio->bi_size = 0; - - for (j = 0; j < SCRUB_PAGES_PER_BIO; ++j) { - struct page *page; - page = alloc_page(GFP_NOFS); - if (!page) - goto nomem; - - ret = bio_add_page(bio, page, PAGE_SIZE, 0); - if (!ret) - goto nomem; - } - WARN_ON(bio->bi_vcnt != SCRUB_PAGES_PER_BIO); if (i != SCRUB_BIOS_PER_DEV-1) sdev->bios[i]->next_free = i + 1; @@ -394,6 +373,7 @@ static void scrub_bio_end_io(struct bio *bio, int err) struct btrfs_fs_info *fs_info = sdev->dev->dev_root->fs_info; sbio->err = err; + sbio->bio = bio; btrfs_queue_worker(&fs_info->scrub_workers, &sbio->work); } @@ -453,6 +433,8 @@ static void scrub_checksum(struct btrfs_work *work) } out: + scrub_free_bio(sbio->bio); + sbio->bio = NULL; spin_lock(&sdev->list_lock); sbio->next_free = sdev->first_free; sdev->first_free = sbio->index; @@ -583,25 +565,50 @@ static int scrub_checksum_super(struct scrub_bio *sbio, void *buffer) static int scrub_submit(struct scrub_dev *sdev) { struct scrub_bio *sbio; + struct bio *bio; + int i; if (sdev->curr == -1) return 0; sbio = sdev->bios[sdev->curr]; - sbio->bio->bi_sector = sbio->physical >> 9; - sbio->bio->bi_size = sbio->count * PAGE_SIZE; - sbio->bio->bi_next = NULL; - sbio->bio->bi_flags |= 1 << BIO_UPTODATE; - sbio->bio->bi_comp_cpu = -1; - sbio->bio->bi_bdev = sdev->dev->bdev; + bio = bio_alloc(GFP_NOFS, sbio->count); + if (!bio) + goto nomem; + + bio->bi_private = sbio; + bio->bi_end_io = scrub_bio_end_io; + bio->bi_bdev = sdev->dev->bdev; + bio->bi_sector = sbio->physical >> 9; + + for (i = 0; i < sbio->count; ++i) { + struct page *page; + int ret; + + page = alloc_page(GFP_NOFS); + if (!page) + goto nomem; + + ret = bio_add_page(bio, page, PAGE_SIZE, 0); + if (!ret) { + __free_page(page); + goto nomem; + } + } + sbio->err = 0; sdev->curr = -1; atomic_inc(&sdev->in_flight); - submit_bio(0, sbio->bio); + submit_bio(READ, bio); return 0; + +nomem: + scrub_free_bio(bio); + + return -ENOMEM; } static int scrub_page(struct scrub_dev *sdev, u64 logical, u64 len, @@ -633,7 +640,11 @@ static int scrub_page(struct scrub_dev *sdev, u64 logical, u64 len, sbio->logical = logical; } else if (sbio->physical + sbio->count * PAGE_SIZE != physical || sbio->logical + sbio->count * PAGE_SIZE != logical) { - scrub_submit(sdev); + int ret; + + ret = scrub_submit(sdev); + if (ret) + return ret; goto again; } sbio->spag[sbio->count].flags = flags; @@ -645,8 +656,13 @@ static int scrub_page(struct scrub_dev *sdev, u64 logical, u64 len, memcpy(sbio->spag[sbio->count].csum, csum, sdev->csum_size); } ++sbio->count; - if (sbio->count == SCRUB_PAGES_PER_BIO || force) - scrub_submit(sdev); + if (sbio->count == SCRUB_PAGES_PER_BIO || force) { + int ret; + + ret = scrub_submit(sdev); + if (ret) + return ret; + } return 0; } From 17aca1c987cff89dc4279371857035da902c8854 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 3 Jun 2011 01:13:45 -0400 Subject: [PATCH 16/25] Btrfs: fix uninit variable in the delayed inode code The nitems counter needs to start at zero Signed-off-by: Chris Mason --- fs/btrfs/delayed-inode.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c index b46d94d1dea8..c61c32cf0f71 100644 --- a/fs/btrfs/delayed-inode.c +++ b/fs/btrfs/delayed-inode.c @@ -678,6 +678,7 @@ static int btrfs_batch_insert_items(struct btrfs_trans_handle *trans, INIT_LIST_HEAD(&head); next = item; + nitems = 0; /* * count the number of the continuous items that we can insert in batch From 211f96c24f117fcc6e9e2431e40d92f4de22625e Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 3 Jun 2011 01:26:53 -0400 Subject: [PATCH 17/25] Btrfs: make sure we don't overflow the free space cache crc page The free space cache uses only one page for crcs right now, which means we can't have a cache file bigger than the crcs we can fit in the first page. This adds a check to enforce that restriction. Signed-off-by: Chris Mason --- fs/btrfs/free-space-cache.c | 27 +++++++++++++++++++-------- 1 file changed, 19 insertions(+), 8 deletions(-) diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index dd38d4c3a599..1cb72394498c 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -590,10 +590,25 @@ int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode, num_pages = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; + + /* Since the first page has all of our checksums and our generation we + * need to calculate the offset into the page that we can start writing + * our entries. + */ + first_page_offset = (sizeof(u32) * num_pages) + sizeof(u64); + filemap_write_and_wait(inode->i_mapping); btrfs_wait_ordered_range(inode, inode->i_size & ~(root->sectorsize - 1), (u64)-1); + /* make sure we don't overflow that first page */ + if (first_page_offset + sizeof(struct btrfs_free_space_entry) >= PAGE_CACHE_SIZE) { + /* this is really the same as running out of space, where we also return 0 */ + printk(KERN_CRIT "Btrfs: free space cache was too big for the crc page\n"); + ret = 0; + goto out_update; + } + /* We need a checksum per page. */ crc = checksums = kzalloc(sizeof(u32) * num_pages, GFP_NOFS); if (!crc) @@ -605,12 +620,6 @@ int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode, return -1; } - /* Since the first page has all of our checksums and our generation we - * need to calculate the offset into the page that we can start writing - * our entries. - */ - first_page_offset = (sizeof(u32) * num_pages) + sizeof(u64); - /* Get the cluster for this block_group if it exists */ if (block_group && !list_empty(&block_group->cluster_list)) cluster = list_entry(block_group->cluster_list.next, @@ -872,12 +881,14 @@ int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode, ret = 1; out_free: + kfree(checksums); + kfree(pages); + +out_update: if (ret != 1) { invalidate_inode_pages2_range(inode->i_mapping, 0, index); BTRFS_I(inode)->generation = 0; } - kfree(checksums); - kfree(pages); btrfs_update_inode(trans, root, inode); return ret; } From ca456ae280c0646e1e571c3b9a3834c55e90adfe Mon Sep 17 00:00:00 2001 From: liubo Date: Wed, 1 Jun 2011 09:42:49 +0000 Subject: [PATCH 18/25] Btrfs: don't save the inode cache in non-FS roots This adds extra checks to make sure the inode map we are caching really belongs to a FS root instead of a special relocation tree. It prevents crashes during balancing operations. Signed-off-by: Liu Bo Signed-off-by: Chris Mason --- fs/btrfs/inode-map.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 3262cd17a12f..04f7199facb4 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -388,6 +388,12 @@ int btrfs_save_ino_cache(struct btrfs_root *root, int prealloc; bool retry = false; + /* only fs tree and subvol/snap needs ino cache */ + if (root->root_key.objectid != BTRFS_FS_TREE_OBJECTID && + (root->root_key.objectid < BTRFS_FIRST_FREE_OBJECTID || + root->root_key.objectid > BTRFS_LAST_FREE_OBJECTID)) + return 0; + path = btrfs_alloc_path(); if (!path) return -ENOMEM; From 5f3f302a6f4cb74906c05fad1d03fc5e95c7e5af Mon Sep 17 00:00:00 2001 From: Arne Jansen Date: Mon, 30 May 2011 08:36:16 +0000 Subject: [PATCH 19/25] btrfs: false BUG_ON when degraded In degraded mode the struct btrfs_device of missing devs don't have device->name set. A kstrdup of NULL correctly returns NULL. Don't BUG in this case. Signed-off-by: Arne Jansen Signed-off-by: Chris Mason --- fs/btrfs/volumes.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index c48214ef5c09..da541dfca2e3 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -504,7 +504,7 @@ static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices) BUG_ON(!new_device); memcpy(new_device, device, sizeof(*new_device)); new_device->name = kstrdup(device->name, GFP_NOFS); - BUG_ON(!new_device->name); + BUG_ON(device->name && !new_device->name); new_device->bdev = NULL; new_device->writeable = 0; new_device->in_fs_metadata = 0; From d132a538d258f8f52fd0cd8b5017755f4e915386 Mon Sep 17 00:00:00 2001 From: Josef Bacik Date: Tue, 31 May 2011 19:33:33 +0000 Subject: [PATCH 20/25] Btrfs: don't save the inode cache if we are deleting this root With xfstest 254 I can panic the box every time with the inode number caching stuff on. This is because we clean the inodes out when we delete the subvolume, but then we write out the inode cache which adds an inode to the subvolume inode tree, and then when it gets evicted again the root gets added back on the dead roots list and is deleted again, so we have a double free. To stop this from happening just return 0 if refs is 0 (and we're not the tree root since tree root always has refs of 0). With this fix 254 no longer panics. Thanks, Signed-off-by: Josef Bacik Tested-by: David Sterba Signed-off-by: Chris Mason --- fs/btrfs/inode-map.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 04f7199facb4..2d0d50067a7b 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -394,6 +394,11 @@ int btrfs_save_ino_cache(struct btrfs_root *root, root->root_key.objectid > BTRFS_LAST_FREE_OBJECTID)) return 0; + /* Don't save inode cache if we are deleting this root */ + if (btrfs_root_refs(&root->root_item) == 0 && + root != root->fs_info->tree_root) + return 0; + path = btrfs_alloc_path(); if (!path) return -ENOMEM; From a4689d2bd3b00dcf5c4320f06e0ab88810fbff9c Mon Sep 17 00:00:00 2001 From: David Sterba Date: Tue, 31 May 2011 17:08:14 +0000 Subject: [PATCH 21/25] btrfs: use btrfs_ino to access inode number commit 4cb5300bc ("Btrfs: add mount -o auto_defrag") accesses inode number directly while it should use the helper with the new inode number allocator. Signed-off-by: David Sterba Signed-off-by: Chris Mason --- fs/btrfs/file.c | 2 +- fs/btrfs/ioctl.c | 7 ++++--- 2 files changed, 5 insertions(+), 4 deletions(-) diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index e3a1b0c2394c..982b5ea9762f 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -144,7 +144,7 @@ int btrfs_add_inode_defrag(struct btrfs_trans_handle *trans, if (!defrag) return -ENOMEM; - defrag->ino = inode->i_ino; + defrag->ino = btrfs_ino(inode); defrag->transid = transid; defrag->root = root->root_key.objectid; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 74c80595d707..ac37040e426a 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -706,16 +706,17 @@ static int find_new_extents(struct btrfs_root *root, struct btrfs_file_extent_item *extent; int type; int ret; + u64 ino = btrfs_ino(inode); path = btrfs_alloc_path(); if (!path) return -ENOMEM; - min_key.objectid = inode->i_ino; + min_key.objectid = ino; min_key.type = BTRFS_EXTENT_DATA_KEY; min_key.offset = *off; - max_key.objectid = inode->i_ino; + max_key.objectid = ino; max_key.type = (u8)-1; max_key.offset = (u64)-1; @@ -726,7 +727,7 @@ static int find_new_extents(struct btrfs_root *root, path, 0, newer_than); if (ret != 0) goto none; - if (min_key.objectid != inode->i_ino) + if (min_key.objectid != ino) goto none; if (min_key.type != BTRFS_EXTENT_DATA_KEY) goto none; From e7786c3ae517b2c433edc91714e86be770e9f1ce Mon Sep 17 00:00:00 2001 From: Arne Jansen Date: Sat, 28 May 2011 20:58:38 +0000 Subject: [PATCH 22/25] btrfs: scrub: add explicit plugging With the removal of the implicit plugging scrub ends up doing more and smaller I/O than necessary. This patch adds explicit plugging per chunk. Signed-off-by: Arne Jansen Signed-off-by: Chris Mason --- fs/btrfs/scrub.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c index 2d1f8909a8e1..1204eab94028 100644 --- a/fs/btrfs/scrub.c +++ b/fs/btrfs/scrub.c @@ -348,9 +348,6 @@ static int scrub_fixup_io(int rw, struct block_device *bdev, sector_t sector, int ret; DECLARE_COMPLETION_ONSTACK(complete); - /* we are going to wait on this IO */ - rw |= REQ_SYNC; - bio = bio_alloc(GFP_NOFS, 1); bio->bi_bdev = bdev; bio->bi_sector = sector; @@ -359,6 +356,7 @@ static int scrub_fixup_io(int rw, struct block_device *bdev, sector_t sector, bio->bi_private = &complete; submit_bio(rw, bio); + /* this will also unplug the queue */ wait_for_completion(&complete); ret = !test_bit(BIO_UPTODATE, &bio->bi_flags); @@ -743,6 +741,7 @@ static noinline_for_stack int scrub_stripe(struct scrub_dev *sdev, struct btrfs_root *root = fs_info->extent_root; struct btrfs_root *csum_root = fs_info->csum_root; struct btrfs_extent_item *extent; + struct blk_plug plug; u64 flags; int ret; int slot; @@ -847,6 +846,7 @@ static noinline_for_stack int scrub_stripe(struct scrub_dev *sdev, * the scrub. This might currently (crc32) end up to be about 1MB */ start_stripe = 0; + blk_start_plug(&plug); again: logical = base + offset + start_stripe * increment; for (i = start_stripe; i < nstripes; ++i) { @@ -988,6 +988,7 @@ static noinline_for_stack int scrub_stripe(struct scrub_dev *sdev, scrub_submit(sdev); out: + blk_finish_plug(&plug); btrfs_free_path(path); return ret < 0 ? ret : 0; } From 4b9465cb9e3859186eefa1ca3b990a5849386320 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Fri, 3 Jun 2011 09:36:29 -0400 Subject: [PATCH 23/25] Btrfs: add mount -o inode_cache This makes the inode map cache default to off until we fix the overflow problem when the free space crcs don't fit inside a single page. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/free-space-cache.c | 6 ++++++ fs/btrfs/inode-map.c | 20 ++++++++++++++++++++ fs/btrfs/super.c | 8 +++++++- 4 files changed, 34 insertions(+), 1 deletion(-) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 8f98c2005715..4958ef5417d6 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1340,6 +1340,7 @@ struct btrfs_ioctl_defrag_range_args { #define BTRFS_MOUNT_USER_SUBVOL_RM_ALLOWED (1 << 14) #define BTRFS_MOUNT_ENOSPC_DEBUG (1 << 15) #define BTRFS_MOUNT_AUTO_DEFRAG (1 << 16) +#define BTRFS_MOUNT_INODE_MAP_CACHE (1 << 17) #define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt) #define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index 1cb72394498c..bffa5c4a633b 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -2536,6 +2536,9 @@ int load_free_ino_cache(struct btrfs_fs_info *fs_info, struct btrfs_root *root) int ret = 0; u64 root_gen = btrfs_root_generation(&root->root_item); + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return 0; + /* * If we're unmounting then just return, since this does a search on the * normal root and not the commit root and we could deadlock. @@ -2575,6 +2578,9 @@ int btrfs_write_out_ino_cache(struct btrfs_root *root, struct inode *inode; int ret; + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return 0; + inode = lookup_free_ino_inode(root, path); if (IS_ERR(inode)) return 0; diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 2d0d50067a7b..cb79b8975c9f 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -38,6 +38,9 @@ static int caching_kthread(void *data) int slot; int ret; + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return 0; + path = btrfs_alloc_path(); if (!path) return -ENOMEM; @@ -141,6 +144,9 @@ static void start_caching(struct btrfs_root *root) int ret; u64 objectid; + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return; + spin_lock(&root->cache_lock); if (root->cached != BTRFS_CACHE_NO) { spin_unlock(&root->cache_lock); @@ -178,6 +184,9 @@ static void start_caching(struct btrfs_root *root) int btrfs_find_free_ino(struct btrfs_root *root, u64 *objectid) { + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return btrfs_find_free_objectid(root, objectid); + again: *objectid = btrfs_find_ino_for_alloc(root); @@ -201,6 +210,10 @@ void btrfs_return_ino(struct btrfs_root *root, u64 objectid) { struct btrfs_free_space_ctl *ctl = root->free_ino_ctl; struct btrfs_free_space_ctl *pinned = root->free_ino_pinned; + + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return; + again: if (root->cached == BTRFS_CACHE_FINISHED) { __btrfs_add_free_space(ctl, objectid, 1); @@ -250,6 +263,9 @@ void btrfs_unpin_free_ino(struct btrfs_root *root) struct rb_node *n; u64 count; + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return; + while (1) { n = rb_first(rbroot); if (!n) @@ -399,9 +415,13 @@ int btrfs_save_ino_cache(struct btrfs_root *root, root != root->fs_info->tree_root) return 0; + if (!btrfs_test_opt(root, INODE_MAP_CACHE)) + return 0; + path = btrfs_alloc_path(); if (!path) return -ENOMEM; + again: inode = lookup_free_ino_inode(root, path); if (IS_ERR(inode) && PTR_ERR(inode) != -ENOENT) { diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 28e3cb2607ff..3559d0b3518a 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -160,7 +160,8 @@ enum { Opt_compress_type, Opt_compress_force, Opt_compress_force_type, Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard, Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed, - Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_err, + Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, + Opt_inode_cache, Opt_err, }; static match_table_t tokens = { @@ -192,6 +193,7 @@ static match_table_t tokens = { {Opt_enospc_debug, "enospc_debug"}, {Opt_subvolrootid, "subvolrootid=%d"}, {Opt_defrag, "autodefrag"}, + {Opt_inode_cache, "inode_cache"}, {Opt_err, NULL}, }; @@ -360,6 +362,10 @@ int btrfs_parse_options(struct btrfs_root *root, char *options) printk(KERN_INFO "btrfs: enabling disk space caching\n"); btrfs_set_opt(info->mount_opt, SPACE_CACHE); break; + case Opt_inode_cache: + printk(KERN_INFO "btrfs: enabling inode map caching\n"); + btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE); + break; case Opt_clear_cache: printk(KERN_INFO "btrfs: force clearing of disk cache\n"); btrfs_set_opt(info->mount_opt, CLEAR_CACHE); From 7841cb2898f66a73062c64d0ef5733dde7279e46 Mon Sep 17 00:00:00 2001 From: David Sterba Date: Tue, 31 May 2011 18:07:27 +0200 Subject: [PATCH 24/25] btrfs: add helper for fs_info->closing wrap checking of filesystem 'closing' flag and fix a few missing memory barriers. Signed-off-by: David Sterba --- fs/btrfs/ctree.h | 9 +++++++++ fs/btrfs/extent-tree.c | 3 +-- fs/btrfs/file.c | 4 ++-- fs/btrfs/free-space-cache.c | 10 ++++------ fs/btrfs/inode-map.c | 3 +-- fs/btrfs/inode.c | 3 +-- fs/btrfs/scrub.c | 2 +- fs/btrfs/transaction.c | 2 +- 8 files changed, 20 insertions(+), 16 deletions(-) diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 4958ef5417d6..8490ee063709 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -2354,6 +2354,15 @@ int btrfs_drop_subtree(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct extent_buffer *node, struct extent_buffer *parent); +static inline int btrfs_fs_closing(struct btrfs_fs_info *fs_info) +{ + /* + * Get synced with close_ctree() + */ + smp_mb(); + return fs_info->closing; +} + /* root-item.c */ int btrfs_find_root_ref(struct btrfs_root *tree_root, struct btrfs_path *path, diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index c9173a7827b0..5b9b6b6df242 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -366,8 +366,7 @@ static int caching_kthread(void *data) nritems = btrfs_header_nritems(leaf); while (1) { - smp_mb(); - if (fs_info->closing > 1) { + if (btrfs_fs_closing(fs_info) > 1) { last = (u64)-1; break; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 982b5ea9762f..fa4ef18b66b1 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -129,7 +129,7 @@ int btrfs_add_inode_defrag(struct btrfs_trans_handle *trans, if (!btrfs_test_opt(root, AUTO_DEFRAG)) return 0; - if (root->fs_info->closing) + if (btrfs_fs_closing(root->fs_info)) return 0; if (BTRFS_I(inode)->in_defrag) @@ -229,7 +229,7 @@ int btrfs_run_defrag_inodes(struct btrfs_fs_info *fs_info) first_ino = defrag->ino + 1; rb_erase(&defrag->rb_node, &fs_info->defrag_inodes); - if (fs_info->closing) + if (btrfs_fs_closing(fs_info)) goto next_free; spin_unlock(&fs_info->defrag_inodes_lock); diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index bffa5c4a633b..ad144736a5fd 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -98,7 +98,7 @@ struct inode *lookup_free_space_inode(struct btrfs_root *root, return inode; spin_lock(&block_group->lock); - if (!root->fs_info->closing) { + if (!btrfs_fs_closing(root->fs_info)) { block_group->inode = igrab(inode); block_group->iref = 1; } @@ -493,8 +493,7 @@ int load_free_space_cache(struct btrfs_fs_info *fs_info, * If we're unmounting then just return, since this does a search on the * normal root and not the commit root and we could deadlock. */ - smp_mb(); - if (fs_info->closing) + if (btrfs_fs_closing(fs_info)) return 0; /* @@ -2513,7 +2512,7 @@ struct inode *lookup_free_ino_inode(struct btrfs_root *root, return inode; spin_lock(&root->cache_lock); - if (!root->fs_info->closing) + if (!btrfs_fs_closing(root->fs_info)) root->cache_inode = igrab(inode); spin_unlock(&root->cache_lock); @@ -2543,8 +2542,7 @@ int load_free_ino_cache(struct btrfs_fs_info *fs_info, struct btrfs_root *root) * If we're unmounting then just return, since this does a search on the * normal root and not the commit root and we could deadlock. */ - smp_mb(); - if (fs_info->closing) + if (btrfs_fs_closing(fs_info)) return 0; path = btrfs_alloc_path(); diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index cb79b8975c9f..b4087e0fa871 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -62,8 +62,7 @@ static int caching_kthread(void *data) goto out; while (1) { - smp_mb(); - if (fs_info->closing) + if (btrfs_fs_closing(fs_info)) goto out; leaf = path->nodes[0]; diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index a83e44bf3206..02ff4a1b968b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -4266,8 +4266,7 @@ int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc) if (BTRFS_I(inode)->dummy_inode) return 0; - smp_mb(); - if (root->fs_info->closing && is_free_space_inode(root, inode)) + if (btrfs_fs_closing(root->fs_info) && is_free_space_inode(root, inode)) nolock = true; if (wbc->sync_mode == WB_SYNC_ALL) { diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c index 1204eab94028..df50fd1eca8f 100644 --- a/fs/btrfs/scrub.c +++ b/fs/btrfs/scrub.c @@ -1183,7 +1183,7 @@ int btrfs_scrub_dev(struct btrfs_root *root, u64 devid, u64 start, u64 end, int ret; struct btrfs_device *dev; - if (root->fs_info->closing) + if (btrfs_fs_closing(root->fs_info)) return -EINVAL; /* diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 2d5c6d2aa4e4..dd719662340e 100644 --- a/fs/btrfs/transaction.c +++ b/fs/btrfs/transaction.c @@ -817,7 +817,7 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) btrfs_btree_balance_dirty(info->tree_root, nr); cond_resched(); - if (root->fs_info->closing || ret != -EAGAIN) + if (btrfs_fs_closing(root->fs_info) || ret != -EAGAIN) break; } root->defrag_running = 0; From aa0467d8d2a00e75b2bb6a56a4ee6d70c5d1928f Mon Sep 17 00:00:00 2001 From: David Sterba Date: Fri, 3 Jun 2011 16:29:08 +0200 Subject: [PATCH 25/25] btrfs: fix uninitialized variable warning With Linus' tree, today's linux-next build (powercp ppc64_defconfig) produced this warning: fs/btrfs/delayed-inode.c: In function 'btrfs_delayed_update_inode': fs/btrfs/delayed-inode.c:1598:6: warning: 'ret' may be used uninitialized in this function Introduced by commit 16cdcec736cd ("btrfs: implement delayed inode items operation"). This fixes a bug in btrfs_update_inode(): if the returned value from btrfs_delayed_update_inode is a nonzero garbage, inode stat data are not updated and several call paths may hit a BUG_ON or fail with strange code. Reported-by: Stephen Rothwell Signed-off-by: David Sterba --- fs/btrfs/delayed-inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c index c61c32cf0f71..6462c29d2d37 100644 --- a/fs/btrfs/delayed-inode.c +++ b/fs/btrfs/delayed-inode.c @@ -1595,7 +1595,7 @@ int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans, struct btrfs_root *root, struct inode *inode) { struct btrfs_delayed_node *delayed_node; - int ret; + int ret = 0; delayed_node = btrfs_get_or_create_delayed_node(inode); if (IS_ERR(delayed_node))