Re: [PATCH] btrfs: speedup mount time with force readahead chunk tree

[Qu Wenruo <quwenruo.btrfs@gmx.com>]

[-- Attachment #1.1: Type: text/plain, Size: 3629 bytes --]



On 2020/7/6 下午2:13, Robbie Ko wrote:
> Does anyone have any suggestions?

I believe David's suggestion on using regular readahead is already good
enough for chunk tree.

Especially since chunk tree is not really the main cause for slow mount.

Thanks,
Qu

> 
> robbieko 於 2020/7/1 下午5:29 寫道:
>> From: Robbie Ko <robbieko@synology.com>
>>
>> When mounting, we always need to read the whole chunk tree,
>> when there are too many chunk items, most of the time is
>> spent on btrfs_read_chunk_tree, because we only read one
>> leaf at a time.
>>
>> We fix this by adding a new readahead mode READA_FORWARD_FORCE,
>> which reads all the leaves after the key in the node when
>> reading a level 1 node.
>>
>> Signed-off-by: Robbie Ko <robbieko@synology.com>
>> ---
>>   fs/btrfs/ctree.c   | 7 +++++--
>>   fs/btrfs/ctree.h   | 2 +-
>>   fs/btrfs/volumes.c | 1 +
>>   3 files changed, 7 insertions(+), 3 deletions(-)
>>
>> diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
>> index 3a7648bff42c..abb9108e2d7d 100644
>> --- a/fs/btrfs/ctree.c
>> +++ b/fs/btrfs/ctree.c
>> @@ -2194,7 +2194,7 @@ static void reada_for_search(struct
>> btrfs_fs_info *fs_info,
>>               if (nr == 0)
>>                   break;
>>               nr--;
>> -        } else if (path->reada == READA_FORWARD) {
>> +        } else if (path->reada == READA_FORWARD || path->reada ==
>> READA_FORWARD_FORCE) {
>>               nr++;
>>               if (nr >= nritems)
>>                   break;
>> @@ -2205,12 +2205,15 @@ static void reada_for_search(struct
>> btrfs_fs_info *fs_info,
>>                   break;
>>           }
>>           search = btrfs_node_blockptr(node, nr);
>> -        if ((search <= target && target - search <= 65536) ||
>> +        if ((path->reada == READA_FORWARD_FORCE) ||
>> +            (search <= target && target - search <= 65536) ||
>>               (search > target && search - target <= 65536)) {
>>               readahead_tree_block(fs_info, search);
>>               nread += blocksize;
>>           }
>>           nscan++;
>> +        if (path->reada == READA_FORWARD_FORCE)
>> +            continue;
>>           if ((nread > 65536 || nscan > 32))
>>               break;
>>       }
>> diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
>> index d404cce8ae40..808bcbdc9530 100644
>> --- a/fs/btrfs/ctree.h
>> +++ b/fs/btrfs/ctree.h
>> @@ -353,7 +353,7 @@ struct btrfs_node {
>>    * The slots array records the index of the item or block pointer
>>    * used while walking the tree.
>>    */
>> -enum { READA_NONE, READA_BACK, READA_FORWARD };
>> +enum { READA_NONE, READA_BACK, READA_FORWARD, READA_FORWARD_FORCE };
>>   struct btrfs_path {
>>       struct extent_buffer *nodes[BTRFS_MAX_LEVEL];
>>       int slots[BTRFS_MAX_LEVEL];
>> diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
>> index 0d6e785bcb98..78fd65abff69 100644
>> --- a/fs/btrfs/volumes.c
>> +++ b/fs/btrfs/volumes.c
>> @@ -7043,6 +7043,7 @@ int btrfs_read_chunk_tree(struct btrfs_fs_info
>> *fs_info)
>>       path = btrfs_alloc_path();
>>       if (!path)
>>           return -ENOMEM;
>> +    path->reada = READA_FORWARD_FORCE;
>>         /*
>>        * uuid_mutex is needed only if we are mounting a sprout FS


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]