Re: [PATCH v3 3/6] ext4: cache extent hole in extent status tree for ext4_da_map_blocks()

[Zheng Liu <gnehzuil.liu@gmail.com>]

On Mon, Sep 01, 2014 at 10:43:50PM -0400, Theodore Ts'o wrote:
> On Thu, Aug 07, 2014 at 11:35:50AM +0800, Zheng Liu wrote:
> > From: Zheng Liu <wenqing.lz@taobao.com>
> > 
> > Currently extent status tree doesn't cache extent hole when a write
> > looks up in extent tree to make sure whether a block has been allocated
> > or not.  In this case, we don't put extent hole in extent cache because
> > later this extent might be removed and a new delayed extent might be
> > added back.  But it will cause a defect when we do a lot of writes.
> > If we don't put extent hole in extent cache, the following writes also
> > need to access extent tree to look at whether or not a block has been
> > allocated.  It brings a cache miss.  This commit fixes this defect.
> > Meanwhile, if an inode has no any extent, this extent hole also will
> > be cached.
> 
> Hi Zheng,
> 
> I thought the reason why we have the EXT4_GET_BLOCKS_NO_PUT_HOLE flag
> is because in ext4_da_map_blocks(), if there is a hole, we will be
> immediately following it up with a call to ext4_es_insert_extent() to
> fill in the hole with the EXTENT_STATUS_DELAYED flag.  The only time
> we don't is if we run into an ENOSPC error.
> 
> Am I missing something?

Yes, you are right.  The purpose is used to do the work like you said
above.  As the commit log described this flag brings a huge number of
cache misses when an empty file is written.  So it might be worth
putting a hole in the cache when delalloc is enabled.

Regards,
                                                - Zheng