Josef Bacik
2012-Dec-18 20:51 UTC
[PATCH] Btrfs: don''t bother updating the inode when evicting
We''re deleting the stupid thing, no sense in updating the inode for the
new
size. We''re running into having 50-100 orphans left over with xfstests
83
because of ENOSPC when trying to start the transaction for the inode update.
This patch fixes this problem. Thanks,
Signed-off-by: Josef Bacik <jbacik@fusionio.com>
---
fs/btrfs/inode.c | 6 +-----
1 files changed, 1 insertions(+), 5 deletions(-)
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index f33269a..ac7f471 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -3898,7 +3898,7 @@ void btrfs_evict_inode(struct inode *inode)
goto no_delete;
}
- trans = btrfs_start_transaction_lflush(root, 1);
+ trans = btrfs_join_transaction(root);
if (IS_ERR(trans)) {
btrfs_orphan_del(NULL, inode);
btrfs_free_block_rsv(root, rsv);
@@ -3911,10 +3911,6 @@ void btrfs_evict_inode(struct inode *inode)
if (ret != -ENOSPC)
break;
- trans->block_rsv = &root->fs_info->trans_block_rsv;
- ret = btrfs_update_inode(trans, root, inode);
- BUG_ON(ret);
-
btrfs_end_transaction(trans, root);
trans = NULL;
btrfs_btree_balance_dirty(root);
--
1.7.7.6
--
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
Miao Xie
2012-Dec-19 01:58 UTC
Re: [PATCH] Btrfs: don''t bother updating the inode when evicting
On tue, 18 Dec 2012 15:51:57 -0500, Josef Bacik wrote:> We''re deleting the stupid thing, no sense in updating the inode for the new > size. We''re running into having 50-100 orphans left over with xfstests 83 > because of ENOSPC when trying to start the transaction for the inode update. > This patch fixes this problem. Thanks,This patch is wrong, it will introduce the inconsonant metadata in the snapshot tree. The reason is folloing: commit 8407aa464331556e4f6784f974030b83fc7585ed Author: Miao Xie <miaox@cn.fujitsu.com> Date: Fri Sep 7 01:43:32 2012 -0600 Btrfs: fix corrupted metadata in the snapshot When we delete a inode, we will remove all the delayed items including delayed inode update, and then truncate all the relative metadata. If there is lots of metadata, we will end the current transaction, and start a new transaction to truncate the left metadata. In this way, we will leave a inode item that its link counter is > 0, and also may leave some directory index items in fs/file tree after the current transaction ends. In other words, the metadata in this fs/file tree is inconsistent. If we create a snapshot for this tree now, we will find a inode with corrupted metadata in the new snapshot, and we won''t continue to drop the left metadata, because its link counter is not 0. We fix this problem by updating the inode item before the current transaction ends. Signed-off-by: Miao Xie <miaox@cn.fujitsu.com> I will write a new patch to fix the problem you said above. Thanks Miao> > Signed-off-by: Josef Bacik <jbacik@fusionio.com> > --- > fs/btrfs/inode.c | 6 +----- > 1 files changed, 1 insertions(+), 5 deletions(-) > > diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c > index f33269a..ac7f471 100644 > --- a/fs/btrfs/inode.c > +++ b/fs/btrfs/inode.c > @@ -3898,7 +3898,7 @@ void btrfs_evict_inode(struct inode *inode) > goto no_delete; > } > > - trans = btrfs_start_transaction_lflush(root, 1); > + trans = btrfs_join_transaction(root); > if (IS_ERR(trans)) { > btrfs_orphan_del(NULL, inode); > btrfs_free_block_rsv(root, rsv); > @@ -3911,10 +3911,6 @@ void btrfs_evict_inode(struct inode *inode) > if (ret != -ENOSPC) > break; > > - trans->block_rsv = &root->fs_info->trans_block_rsv; > - ret = btrfs_update_inode(trans, root, inode); > - BUG_ON(ret); > - > btrfs_end_transaction(trans, root); > trans = NULL; > btrfs_btree_balance_dirty(root); >-- 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
The delayed item commit code in several functions is similar, so
cleanup it.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
fs/btrfs/delayed-inode.c | 83 +++++++++++++++++++++---------------------------
1 file changed, 37 insertions(+), 46 deletions(-)
diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
index 3483603..4e204bb 100644
--- a/fs/btrfs/delayed-inode.c
+++ b/fs/btrfs/delayed-inode.c
@@ -1110,6 +1110,25 @@ static int btrfs_update_delayed_inode(struct
btrfs_trans_handle *trans,
return 0;
}
+static inline int
+__btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
+ struct btrfs_path *path,
+ struct btrfs_delayed_node *node)
+{
+ int ret;
+
+ ret = btrfs_insert_delayed_items(trans, path, node->root, node);
+ if (ret)
+ return ret;
+
+ ret = btrfs_delete_delayed_items(trans, path, node->root, node);
+ if (ret)
+ return ret;
+
+ ret = btrfs_update_delayed_inode(trans, node->root, path, node);
+ return ret;
+}
+
/*
* Called when committing the transaction.
* Returns 0 on success.
@@ -1119,7 +1138,6 @@ static int btrfs_update_delayed_inode(struct
btrfs_trans_handle *trans,
static int __btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
struct btrfs_root *root, int nr)
{
- struct btrfs_root *curr_root = root;
struct btrfs_delayed_root *delayed_root;
struct btrfs_delayed_node *curr_node, *prev_node;
struct btrfs_path *path;
@@ -1142,15 +1160,8 @@ static int __btrfs_run_delayed_items(struct
btrfs_trans_handle *trans,
curr_node = btrfs_first_delayed_node(delayed_root);
while (curr_node && (!count || (count && nr--))) {
- curr_root = curr_node->root;
- ret = btrfs_insert_delayed_items(trans, path, curr_root,
- curr_node);
- if (!ret)
- ret = btrfs_delete_delayed_items(trans, path,
- curr_root, curr_node);
- if (!ret)
- ret = btrfs_update_delayed_inode(trans, curr_root,
- path, curr_node);
+ ret = __btrfs_commit_inode_delayed_items(trans, path,
+ curr_node);
if (ret) {
btrfs_release_delayed_node(curr_node);
curr_node = NULL;
@@ -1183,36 +1194,12 @@ int btrfs_run_delayed_items_nr(struct btrfs_trans_handle
*trans,
return __btrfs_run_delayed_items(trans, root, nr);
}
-static int __btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
- struct btrfs_delayed_node *node)
-{
- struct btrfs_path *path;
- struct btrfs_block_rsv *block_rsv;
- int ret;
-
- path = btrfs_alloc_path();
- if (!path)
- return -ENOMEM;
- path->leave_spinning = 1;
-
- block_rsv = trans->block_rsv;
- trans->block_rsv = &node->root->fs_info->delayed_block_rsv;
-
- ret = btrfs_insert_delayed_items(trans, path, node->root, node);
- if (!ret)
- ret = btrfs_delete_delayed_items(trans, path, node->root, node);
- if (!ret)
- ret = btrfs_update_delayed_inode(trans, node->root, path, node);
- btrfs_free_path(path);
-
- trans->block_rsv = block_rsv;
- return ret;
-}
-
int btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
struct inode *inode)
{
struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
+ struct btrfs_path *path;
+ struct btrfs_block_rsv *block_rsv;
int ret;
if (!delayed_node)
@@ -1226,8 +1213,20 @@ int btrfs_commit_inode_delayed_items(struct
btrfs_trans_handle *trans,
}
mutex_unlock(&delayed_node->mutex);
- ret = __btrfs_commit_inode_delayed_items(trans, delayed_node);
+ path = btrfs_alloc_path();
+ if (!path)
+ return -ENOMEM;
+ path->leave_spinning = 1;
+
+ block_rsv = trans->block_rsv;
+ trans->block_rsv =
&delayed_node->root->fs_info->delayed_block_rsv;
+
+ ret = __btrfs_commit_inode_delayed_items(trans, path, delayed_node);
+
btrfs_release_delayed_node(delayed_node);
+ btrfs_free_path(path);
+ trans->block_rsv = block_rsv;
+
return ret;
}
@@ -1258,7 +1257,6 @@ static void btrfs_async_run_delayed_node_done(struct
btrfs_work *work)
struct btrfs_root *root;
struct btrfs_block_rsv *block_rsv;
int need_requeue = 0;
- int ret;
async_node = container_of(work, struct btrfs_async_delayed_node, work);
@@ -1277,14 +1275,7 @@ static void btrfs_async_run_delayed_node_done(struct
btrfs_work *work)
block_rsv = trans->block_rsv;
trans->block_rsv = &root->fs_info->delayed_block_rsv;
- ret = btrfs_insert_delayed_items(trans, path, root, delayed_node);
- if (!ret)
- ret = btrfs_delete_delayed_items(trans, path, root,
- delayed_node);
-
- if (!ret)
- btrfs_update_delayed_inode(trans, root, path, delayed_node);
-
+ __btrfs_commit_inode_delayed_items(trans, path, delayed_node);
/*
* Maybe new delayed items have been inserted, so we need requeue
* the work. Besides that, we must dequeue the empty delayed nodes
--
1.7.11.7
--
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
Miao Xie
2012-Dec-19 06:59 UTC
[PATCH 2/2] Btrfs: fix lots of orphan inodes when the space is not enough
We''re running into having 50-100 orphans left over with xfstests 83
because of ENOSPC when trying to start the transaction for the inode update.
But in fact, it makes no sense in updating the inode for the new size while
we''re deleting the stupid thing. This patch fixes this problem.
Reported-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
fs/btrfs/delayed-inode.c | 90 +++++++++++++++++++++++++++++++++++++++++-------
fs/btrfs/delayed-inode.h | 1 +
fs/btrfs/inode.c | 11 +++---
3 files changed, 85 insertions(+), 17 deletions(-)
diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
index 4e204bb..c7e7506 100644
--- a/fs/btrfs/delayed-inode.c
+++ b/fs/btrfs/delayed-inode.c
@@ -1065,32 +1065,25 @@ static void btrfs_release_delayed_inode(struct
btrfs_delayed_node *delayed_node)
}
}
-static int btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct btrfs_path *path,
- struct btrfs_delayed_node *node)
+static int __btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path,
+ struct btrfs_delayed_node *node)
{
struct btrfs_key key;
struct btrfs_inode_item *inode_item;
struct extent_buffer *leaf;
int ret;
- mutex_lock(&node->mutex);
- if (!node->inode_dirty) {
- mutex_unlock(&node->mutex);
- return 0;
- }
-
key.objectid = node->inode_id;
btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
key.offset = 0;
+
ret = btrfs_lookup_inode(trans, root, path, &key, 1);
if (ret > 0) {
btrfs_release_path(path);
- mutex_unlock(&node->mutex);
return -ENOENT;
} else if (ret < 0) {
- mutex_unlock(&node->mutex);
return ret;
}
@@ -1105,11 +1098,28 @@ static int btrfs_update_delayed_inode(struct
btrfs_trans_handle *trans,
btrfs_delayed_inode_release_metadata(root, node);
btrfs_release_delayed_inode(node);
- mutex_unlock(&node->mutex);
return 0;
}
+static inline int btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root,
+ struct btrfs_path *path,
+ struct btrfs_delayed_node *node)
+{
+ int ret;
+
+ mutex_lock(&node->mutex);
+ if (!node->inode_dirty) {
+ mutex_unlock(&node->mutex);
+ return 0;
+ }
+
+ ret = __btrfs_update_delayed_inode(trans, root, path, node);
+ mutex_unlock(&node->mutex);
+ return ret;
+}
+
static inline int
__btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
struct btrfs_path *path,
@@ -1230,6 +1240,60 @@ int btrfs_commit_inode_delayed_items(struct
btrfs_trans_handle *trans,
return ret;
}
+int btrfs_commit_inode_delayed_inode(struct inode *inode)
+{
+ struct btrfs_trans_handle *trans;
+ struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
+ struct btrfs_path *path;
+ struct btrfs_block_rsv *block_rsv;
+ int ret;
+
+ if (!delayed_node)
+ return 0;
+
+ mutex_lock(&delayed_node->mutex);
+ if (!delayed_node->inode_dirty) {
+ mutex_unlock(&delayed_node->mutex);
+ btrfs_release_delayed_node(delayed_node);
+ return 0;
+ }
+ mutex_unlock(&delayed_node->mutex);
+
+ trans = btrfs_join_transaction(delayed_node->root);
+ if (IS_ERR(trans)) {
+ ret = PTR_ERR(trans);
+ goto out;
+ }
+
+ path = btrfs_alloc_path();
+ if (!path) {
+ ret = -ENOMEM;
+ goto trans_out;
+ }
+ path->leave_spinning = 1;
+
+ block_rsv = trans->block_rsv;
+ trans->block_rsv =
&delayed_node->root->fs_info->delayed_block_rsv;
+
+ mutex_lock(&delayed_node->mutex);
+ if (delayed_node->inode_dirty)
+ ret = __btrfs_update_delayed_inode(trans, delayed_node->root,
+ path, delayed_node);
+ else
+ ret = 0;
+ mutex_unlock(&delayed_node->mutex);
+
+ btrfs_free_path(path);
+ trans->block_rsv = block_rsv;
+trans_out:
+ btrfs_end_transaction(trans, delayed_node->root);
+ btrfs_btree_balance_dirty(delayed_node->root);
+out:
+ btrfs_release_delayed_node(delayed_node);
+
+ return ret;
+}
+
void btrfs_remove_delayed_node(struct inode *inode)
{
struct btrfs_delayed_node *delayed_node;
diff --git a/fs/btrfs/delayed-inode.h b/fs/btrfs/delayed-inode.h
index 4f808e1..78b6ad0 100644
--- a/fs/btrfs/delayed-inode.h
+++ b/fs/btrfs/delayed-inode.h
@@ -117,6 +117,7 @@ int btrfs_commit_inode_delayed_items(struct
btrfs_trans_handle *trans,
/* Used for evicting the inode. */
void btrfs_remove_delayed_node(struct inode *inode);
void btrfs_kill_delayed_inode_items(struct inode *inode);
+int btrfs_commit_inode_delayed_inode(struct inode *inode);
int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans,
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 67ed24a..2a2b5e1 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -3855,6 +3855,12 @@ void btrfs_evict_inode(struct inode *inode)
goto no_delete;
}
+ ret = btrfs_commit_inode_delayed_inode(inode);
+ if (ret) {
+ btrfs_orphan_del(NULL, inode);
+ goto no_delete;
+ }
+
rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
if (!rsv) {
btrfs_orphan_del(NULL, inode);
@@ -3892,7 +3898,7 @@ void btrfs_evict_inode(struct inode *inode)
goto no_delete;
}
- trans = btrfs_start_transaction_lflush(root, 1);
+ trans = btrfs_join_transaction(root);
if (IS_ERR(trans)) {
btrfs_orphan_del(NULL, inode);
btrfs_free_block_rsv(root, rsv);
@@ -3906,9 +3912,6 @@ void btrfs_evict_inode(struct inode *inode)
break;
trans->block_rsv = &root->fs_info->trans_block_rsv;
- ret = btrfs_update_inode(trans, root, inode);
- BUG_ON(ret);
-
btrfs_end_transaction(trans, root);
trans = NULL;
btrfs_btree_balance_dirty(root);
--
1.7.11.7
--
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
Josef Bacik
2012-Dec-19 15:02 UTC
Re: [PATCH] Btrfs: don''t bother updating the inode when evicting
On Tue, Dec 18, 2012 at 06:58:33PM -0700, Miao Xie wrote:> On tue, 18 Dec 2012 15:51:57 -0500, Josef Bacik wrote: > > We''re deleting the stupid thing, no sense in updating the inode for the new > > size. We''re running into having 50-100 orphans left over with xfstests 83 > > because of ENOSPC when trying to start the transaction for the inode update. > > This patch fixes this problem. Thanks, > > This patch is wrong, it will introduce the inconsonant metadata in the snapshot > tree. The reason is folloing: > > commit 8407aa464331556e4f6784f974030b83fc7585ed > Author: Miao Xie <miaox@cn.fujitsu.com> > Date: Fri Sep 7 01:43:32 2012 -0600 > > Btrfs: fix corrupted metadata in the snapshot > > When we delete a inode, we will remove all the delayed items including delayed > inode update, and then truncate all the relative metadata. If there is lots of > metadata, we will end the current transaction, and start a new transaction to > truncate the left metadata. In this way, we will leave a inode item that its > link counter is > 0, and also may leave some directory index items in fs/file tree > after the current transaction ends. In other words, the metadata in this fs/file tree > is inconsistent. If we create a snapshot for this tree now, we will find a inode with > corrupted metadata in the new snapshot, and we won''t continue to drop the left metadata, > because its link counter is not 0. > > We fix this problem by updating the inode item before the current transaction ends. > > Signed-off-by: Miao Xie <miaox@cn.fujitsu.com> >So why don''t we fix unlink to call btrfs_update_inode_item so that the nlink counter is set to 0? The orphan item will be carried over into the snapshot if we don''t actually evict the inode before we do the snapshot and then the orphan cleanup will take care of the rest? 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
Miao Xie
2012-Dec-20 03:04 UTC
Re: [PATCH] Btrfs: don''t bother updating the inode when evicting
On wed, 19 Dec 2012 10:02:59 -0500, Josef Bacik wrote:> On Tue, Dec 18, 2012 at 06:58:33PM -0700, Miao Xie wrote: >> On tue, 18 Dec 2012 15:51:57 -0500, Josef Bacik wrote: >>> We''re deleting the stupid thing, no sense in updating the inode for the new >>> size. We''re running into having 50-100 orphans left over with xfstests 83 >>> because of ENOSPC when trying to start the transaction for the inode update. >>> This patch fixes this problem. Thanks, >> >> This patch is wrong, it will introduce the inconsonant metadata in the snapshot >> tree. The reason is folloing: >> >> commit 8407aa464331556e4f6784f974030b83fc7585ed >> Author: Miao Xie <miaox@cn.fujitsu.com> >> Date: Fri Sep 7 01:43:32 2012 -0600 >> >> Btrfs: fix corrupted metadata in the snapshot >> >> When we delete a inode, we will remove all the delayed items including delayed >> inode update, and then truncate all the relative metadata. If there is lots of >> metadata, we will end the current transaction, and start a new transaction to >> truncate the left metadata. In this way, we will leave a inode item that its >> link counter is > 0, and also may leave some directory index items in fs/file tree >> after the current transaction ends. In other words, the metadata in this fs/file tree >> is inconsistent. If we create a snapshot for this tree now, we will find a inode with >> corrupted metadata in the new snapshot, and we won''t continue to drop the left metadata, >> because its link counter is not 0. >> >> We fix this problem by updating the inode item before the current transaction ends. >> >> Signed-off-by: Miao Xie <miaox@cn.fujitsu.com> >> > > So why don''t we fix unlink to call btrfs_update_inode_item so that the nlink > counter is set to 0? The orphan item will be carried over into the snapshot if > we don''t actually evict the inode before we do the snapshot and then the orphan > cleanup will take care of the rest? Thanks,But it would make the file deletion performance down. Thanks Miao> > 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 >-- 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