Re: [PATCH 14/19] btrfs: redirty released extent buffers in sequential BGs

[Naohiro Aota]

On 2019/06/13 23:24, Josef Bacik wrote:
> On Fri, Jun 07, 2019 at 10:10:20PM +0900, Naohiro Aota wrote:
>> Tree manipulating operations like merging nodes often release
>> once-allocated tree nodes. Btrfs cleans such nodes so that pages in the
>> node are not uselessly written out. On HMZONED drives, however, such
>> optimization blocks the following IOs as the cancellation of the write out
>> of the freed blocks breaks the sequential write sequence expected by the
>> device.
>>
>> This patch introduces a list of clean extent buffers that have been
>> released in a transaction. Btrfs consult the list before writing out and
>> waiting for the IOs, and it redirties a buffer if 1) it's in sequential BG,
>> 2) it's in un-submit range, and 3) it's not under IO. Thus, such buffers
>> are marked for IO in btrfs_write_and_wait_transaction() to send proper bios
>> to the disk.
>>
>> Signed-off-by: Naohiro Aota <naohiro.aota@xxxxxxx>
>> ---
>>   fs/btrfs/disk-io.c     | 27 ++++++++++++++++++++++++---
>>   fs/btrfs/extent_io.c   |  1 +
>>   fs/btrfs/extent_io.h   |  2 ++
>>   fs/btrfs/transaction.c | 35 +++++++++++++++++++++++++++++++++++
>>   fs/btrfs/transaction.h |  3 +++
>>   5 files changed, 65 insertions(+), 3 deletions(-)
>>
>> diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
>> index 6651986da470..c6147fce648f 100644
>> --- a/fs/btrfs/disk-io.c
>> +++ b/fs/btrfs/disk-io.c
>> @@ -535,7 +535,9 @@ static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page)
>>   	if (csum_tree_block(eb, result))
>>   		return -EINVAL;
>>   
>> -	if (btrfs_header_level(eb))
>> +	if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags))
>> +		ret = 0;
>> +	else if (btrfs_header_level(eb))
>>   		ret = btrfs_check_node(eb);
>>   	else
>>   		ret = btrfs_check_leaf_full(eb);
>> @@ -1115,10 +1117,20 @@ struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr,
>>   void btrfs_clean_tree_block(struct extent_buffer *buf)
>>   {
>>   	struct btrfs_fs_info *fs_info = buf->fs_info;
>> -	if (btrfs_header_generation(buf) ==
>> -	    fs_info->running_transaction->transid) {
>> +	struct btrfs_transaction *cur_trans = fs_info->running_transaction;
>> +
>> +	if (btrfs_header_generation(buf) == cur_trans->transid) {
>>   		btrfs_assert_tree_locked(buf);
>>   
>> +		if (btrfs_fs_incompat(fs_info, HMZONED) &&
>> +		    list_empty(&buf->release_list)) {
>> +			atomic_inc(&buf->refs);
>> +			spin_lock(&cur_trans->releasing_ebs_lock);
>> +			list_add_tail(&buf->release_list,
>> +				      &cur_trans->releasing_ebs);
>> +			spin_unlock(&cur_trans->releasing_ebs_lock);
>> +		}
>> +
>>   		if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
>>   			percpu_counter_add_batch(&fs_info->dirty_metadata_bytes,
>>   						 -buf->len,
>> @@ -4533,6 +4545,15 @@ void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
>>   	btrfs_destroy_pinned_extent(fs_info,
>>   				    fs_info->pinned_extents);
>>   
>> +	while (!list_empty(&cur_trans->releasing_ebs)) {
>> +		struct extent_buffer *eb;
>> +
>> +		eb = list_first_entry(&cur_trans->releasing_ebs,
>> +				      struct extent_buffer, release_list);
>> +		list_del_init(&eb->release_list);
>> +		free_extent_buffer(eb);
>> +	}
>> +
>>   	cur_trans->state =TRANS_STATE_COMPLETED;
>>   	wake_up(&cur_trans->commit_wait);
>>   }
>> diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
>> index 13fca7bfc1f2..c73c69e2bef4 100644
>> --- a/fs/btrfs/extent_io.c
>> +++ b/fs/btrfs/extent_io.c
>> @@ -4816,6 +4816,7 @@ __alloc_extent_buffer(struct btrfs_fs_info *fs_info, u64 start,
>>   	init_waitqueue_head(&eb->read_lock_wq);
>>   
>>   	btrfs_leak_debug_add(&eb->leak_list, &buffers);
>> +	INIT_LIST_HEAD(&eb->release_list);
>>   
>>   	spin_lock_init(&eb->refs_lock);
>>   	atomic_set(&eb->refs, 1);
>> diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
>> index aa18a16a6ed7..2987a01f84f9 100644
>> --- a/fs/btrfs/extent_io.h
>> +++ b/fs/btrfs/extent_io.h
>> @@ -58,6 +58,7 @@ enum {
>>   	EXTENT_BUFFER_IN_TREE,
>>   	/* write IO error */
>>   	EXTENT_BUFFER_WRITE_ERR,
>> +	EXTENT_BUFFER_NO_CHECK,
>>   };
>>   
>>   /* these are flags for __process_pages_contig */
>> @@ -186,6 +187,7 @@ struct extent_buffer {
>>   	 */
>>   	wait_queue_head_t read_lock_wq;
>>   	struct page *pages[INLINE_EXTENT_BUFFER_PAGES];
>> +	struct list_head release_list;
>>   #ifdef CONFIG_BTRFS_DEBUG
>>   	atomic_t spinning_writers;
>>   	atomic_t spinning_readers;
>> diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
>> index 3f6811cdf803..ded40ad75419 100644
>> --- a/fs/btrfs/transaction.c
>> +++ b/fs/btrfs/transaction.c
>> @@ -236,6 +236,8 @@ static noinline int join_transaction(struct btrfs_fs_info *fs_info,
>>   	spin_lock_init(&cur_trans->dirty_bgs_lock);
>>   	INIT_LIST_HEAD(&cur_trans->deleted_bgs);
>>   	spin_lock_init(&cur_trans->dropped_roots_lock);
>> +	INIT_LIST_HEAD(&cur_trans->releasing_ebs);
>> +	spin_lock_init(&cur_trans->releasing_ebs_lock);
>>   	list_add_tail(&cur_trans->list, &fs_info->trans_list);
>>   	extent_io_tree_init(fs_info, &cur_trans->dirty_pages,
>>   			IO_TREE_TRANS_DIRTY_PAGES, fs_info->btree_inode);
>> @@ -2219,7 +2221,31 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans)
>>   
>>   	wake_up(&fs_info->transaction_wait);
>>   
>> +	if (btrfs_fs_incompat(fs_info, HMZONED)) {
>> +		struct extent_buffer *eb;
>> +
>> +		list_for_each_entry(eb, &cur_trans->releasing_ebs,
>> +				    release_list) {
>> +			struct btrfs_block_group_cache *cache;
>> +
>> +			cache = btrfs_lookup_block_group(fs_info, eb->start);
>> +			if (!cache)
>> +				continue;
>> +			mutex_lock(&cache->submit_lock);
>> +			if (cache->alloc_type == BTRFS_ALLOC_SEQ &&
>> +			    cache->submit_offset <= eb->start &&
>> +			    !extent_buffer_under_io(eb)) {
>> +				set_extent_buffer_dirty(eb);
>> +				cache->space_info->bytes_readonly += eb->len;
> 
> Huh?
> 

I'm tracking once allocated then freed region in "space_info->bytes_readonly".
As I wrote in the other reply, I can add and use "space_info->bytes_zone_unavailable" instead.

>> +				set_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags);
>> +			}
>> +			mutex_unlock(&cache->submit_lock);
>> +			btrfs_put_block_group(cache);
>> +		}
>> +	}
>> +
> 
> Helper here please.
>>   	ret = btrfs_write_and_wait_transaction(trans);
>> +
>>   	if (ret) {
>>   		btrfs_handle_fs_error(fs_info, ret,
>>   				      "Error while writing out transaction");
>> @@ -2227,6 +2253,15 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans)
>>   		goto scrub_continue;
>>   	}
>>   
>> +	while (!list_empty(&cur_trans->releasing_ebs)) {
>> +		struct extent_buffer *eb;
>> +
>> +		eb = list_first_entry(&cur_trans->releasing_ebs,
>> +				      struct extent_buffer, release_list);
>> +		list_del_init(&eb->release_list);
>> +		free_extent_buffer(eb);
>> +	}
>> +
> 
> Another helper, and also can't we release eb's above that we didn't need to
> re-mark dirty?  Thanks,
> 
> Josef
> 

hm, we can do so. I'll change the code in the next version.
Thanks,