Re: [PATCH v3 6/7] btrfs: delayed-ref: Introduce new parameter for btrfs_add_delayed_tree_ref() to reduce unnecessary qgroup tracing

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 




On 2018/9/26 下午10:40, David Sterba wrote:
> On Tue, Sep 11, 2018 at 01:38:17PM +0800, Qu Wenruo wrote:
>> For btrfs_add_delayed_tree_ref(), its ref_root parameter can be
>> different from its real root.
>> This is pretty common for reloc tree, in that case @ref_root is passed
>> as the original tree owner (source file tree).
>>
>> However btrfs_add_delayed_tree_ref() uses @ref_root to determine whether
>> we should do heavy qgroup tracing for it, so for balance this could
>> cause a lot of unnecessary extent tracing.
>>
>> This patch will introduce a parameter, @root for qgroup code to
>> determine whether we really need to do the heavy-lifting tracing work.
>>
>> This patch would reduce the qgroup dirty extent number from 'number of
>> extents being relocated' to 0 for the transaction before real tree swap.
>>
>> Signed-off-by: Qu Wenruo <wqu@xxxxxxxx>
>> ---
>>  fs/btrfs/delayed-ref.c |  5 +++--
>>  fs/btrfs/delayed-ref.h |  1 +
>>  fs/btrfs/extent-tree.c | 17 ++++++++++-------
>>  3 files changed, 14 insertions(+), 9 deletions(-)
>>
>> diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c
>> index 62ff545ba1f7..7f681dd36e7f 100644
>> --- a/fs/btrfs/delayed-ref.c
>> +++ b/fs/btrfs/delayed-ref.c
>> @@ -710,8 +710,9 @@ static void init_delayed_ref_common(struct btrfs_fs_info *fs_info,
>>   * transaction commits.
>>   */
>>  int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
>> +			       struct btrfs_root *root,
>>  			       u64 bytenr, u64 num_bytes, u64 parent,
>> -			       u64 ref_root,  int level, int action,
>> +			       u64 ref_root, int level, int action,
> 
> Unrelated change

Will avoid it in next update.

> 
>>  			       struct btrfs_delayed_extent_op *extent_op,
>>  			       int *old_ref_mod, int *new_ref_mod)
>>  {
>> @@ -737,7 +738,7 @@ int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
>>  	}
>>  
>>  	if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) &&
>> -	    is_fstree(ref_root)) {
>> +	    is_fstree(root->root_key.objectid)) {
>>  		record = kmalloc(sizeof(*record), GFP_NOFS);
>>  		if (!record) {
>>  			kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
>> diff --git a/fs/btrfs/delayed-ref.h b/fs/btrfs/delayed-ref.h
>> index d9f2a4ebd5db..a51560a3cb66 100644
>> --- a/fs/btrfs/delayed-ref.h
>> +++ b/fs/btrfs/delayed-ref.h
>> @@ -235,6 +235,7 @@ static inline void btrfs_put_delayed_ref_head(struct btrfs_delayed_ref_head *hea
>>  }
>>  
>>  int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
>> +			       struct btrfs_root *root,
>>  			       u64 bytenr, u64 num_bytes, u64 parent,
>>  			       u64 ref_root, int level, int action,
>>  			       struct btrfs_delayed_extent_op *extent_op,
>> diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
>> index 4588153f414c..7a3c650b11f8 100644
>> --- a/fs/btrfs/extent-tree.c
>> +++ b/fs/btrfs/extent-tree.c
>> @@ -2040,9 +2040,10 @@ int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
>>  			   owner, offset, BTRFS_ADD_DELAYED_REF);
>>  
>>  	if (owner < BTRFS_FIRST_FREE_OBJECTID) {
>> -		ret = btrfs_add_delayed_tree_ref(trans, bytenr,
>> +		ret = btrfs_add_delayed_tree_ref(trans, root, bytenr,
>>  						 num_bytes, parent,
>> -						 root_objectid, (int)owner,
>> +						 root_objectid,
>> +						 (int)owner,
> 
> Here as well and a few more below. The (int) typecast is actually a bug
> as it trims the owner to just 32bits, so if you want to update the
> formatting, please send a separate patch.

Will avoid the unrelated change in next update.

But for the "bug" part, it's a mess.

Since owner is smaller than BTRFS_FIRST_FREE_OBJECTID, truncating to
32bits is in fact completely OK.

However the declaration of btrfs_add_delayed_tree_ref() says that
parameter is in fact @level.
On the other hand, ref->level is passed to __btrfs_inc_extent_ref() as
@onwer (and u64) again, so it's not causing problem (or we're already
screwed up badly).

The naming and forced convert is really a mess here though.

I'll just keep that mess as is and don't touch it for the patchset.

Thanks,
Qu

> 
>>  						 BTRFS_ADD_DELAYED_REF, NULL,
>>  						 &old_ref_mod, &new_ref_mod);
>>  	} else {
>> @@ -6993,7 +6994,7 @@ void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
>>  				   root->root_key.objectid,
>>  				   btrfs_header_level(buf), 0,
>>  				   BTRFS_DROP_DELAYED_REF);
>> -		ret = btrfs_add_delayed_tree_ref(trans, buf->start,
>> +		ret = btrfs_add_delayed_tree_ref(trans, root, buf->start,
>>  						 buf->len, parent,
>>  						 root->root_key.objectid,
>>  						 btrfs_header_level(buf),
>> @@ -7072,9 +7073,10 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans,
>>  		old_ref_mod = new_ref_mod = 0;
>>  		ret = 0;
>>  	} else if (owner < BTRFS_FIRST_FREE_OBJECTID) {
>> -		ret = btrfs_add_delayed_tree_ref(trans, bytenr,
>> +		ret = btrfs_add_delayed_tree_ref(trans, root, bytenr,
>>  						 num_bytes, parent,
>> -						 root_objectid, (int)owner,
>> +						 root_objectid,
>> +						 (int)owner,
>>  						 BTRFS_DROP_DELAYED_REF, NULL,
>>  						 &old_ref_mod, &new_ref_mod);
>>  	} else {
>> @@ -8293,9 +8295,10 @@ struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
>>  		btrfs_ref_tree_mod(root, ins.objectid, ins.offset, parent,
>>  				   root_objectid, level, 0,
>>  				   BTRFS_ADD_DELAYED_EXTENT);
>> -		ret = btrfs_add_delayed_tree_ref(trans, ins.objectid,
>> +		ret = btrfs_add_delayed_tree_ref(trans, root, ins.objectid,
>>  						 ins.offset, parent,
>> -						 root_objectid, level,
>> +						 root_objectid,
>> +						 level,
>>  						 BTRFS_ADD_DELAYED_EXTENT,
>>  						 extent_op, NULL, NULL);
>>  		if (ret)
>> -- 
>> 2.18.0

Attachment: signature.asc
Description: OpenPGP digital signature


[Index of Archives]     [Linux Filesystem Development]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux