| Message ID | 1302886575-2025-1-git-send-email-josef@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Excerpts from Josef Bacik's message of 2011-04-15 12:56:15 -0400: > 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, Just curious, which call stacks are currently nesting transactions? -chris > > Signed-off-by: Josef Bacik <josef@redhat.com> > --- > fs/btrfs/transaction.c | 17 +++++++++++++++++ > fs/btrfs/transaction.h | 2 ++ > 2 files changed, 19 insertions(+), 0 deletions(-) > > diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c > index 70bfb26..46f4056 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 @@ again: > 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 @@ again: > } > } > > +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 1f573f0..154314f 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 { -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 04/15/2011 01:36 PM, Chris Mason wrote: > Excerpts from Josef Bacik's message of 2011-04-15 12:56:15 -0400: >> 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, > > Just curious, which call stacks are currently nesting transactions? > The one that kept getting me in trouble was the run_delalloc_nocow which could possibly call cow_file_range. I didn't notice any others, but I didn't really look all that hard either. Thanks, Josef -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c index 70bfb26..46f4056 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 @@ again: 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 @@ again: } } +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 1f573f0..154314f 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 {
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 <josef@redhat.com> --- fs/btrfs/transaction.c | 17 +++++++++++++++++ fs/btrfs/transaction.h | 2 ++ 2 files changed, 19 insertions(+), 0 deletions(-)