Re: [f2fs-dev] [RFC v3] f2fs: extent cache: support unaligned extent

[Jaegeuk Kim]

On 08/05, Chao Yu wrote:
> On 2021/8/5 7:45, Jaegeuk Kim wrote:
> > Chao,
> > 
> > How about this?
> > https://github.com/jaegeuk/f2fs/commit/d6bbe121bc24dfabfedc07ba7cb6e921fb70ece0
> > 
> > I'm digging one bug in __insert_extent_tree w/ the patch tho.
> > 
> > On 08/04, Jaegeuk Kim wrote:
> > > On 08/04, Chao Yu wrote:
> > > > Compressed inode may suffer read performance issue due to it can not
> > > > use extent cache, so I propose to add this unaligned extent support
> > > > to improve it.
> > > > 
> > > > Currently, it only works in readonly format f2fs image.
> > > > 
> > > > Unaligned extent: in one compressed cluster, physical block number
> > > > will be less than logical block number, so we add an extra physical
> > > > block length in extent info in order to indicate such extent status.
> > > > 
> > > > The idea is if one whole cluster blocks are contiguous physically,
> > > > once its mapping info was readed at first time, we will cache an
> > > > unaligned (or aligned) extent info entry in extent cache, it expects
> > > > that the mapping info will be hitted when rereading cluster.
> > > > 
> > > > Merge policy:
> > > > - Aligned extents can be merged.
> > > > - Aligned extent and unaligned extent can not be merged.
> > > > 
> > > > Signed-off-by: Chao Yu <chao@xxxxxxxxxx>
> > > > ---
> > > > v3:
> > > > - avoid CONFIG_F2FS_FS_COMPRESSION as much as possible
> > > > - clean up codes
> > > >   fs/f2fs/compress.c     | 24 ++++++++++++
> > > >   fs/f2fs/data.c         | 28 +++++++++++---
> > > >   fs/f2fs/extent_cache.c | 88 +++++++++++++++++++++++++++++++++++++-----
> > > >   fs/f2fs/f2fs.h         | 42 +++++++++++++++++---
> > > >   fs/f2fs/node.c         | 18 +++++++++
> > > >   5 files changed, 179 insertions(+), 21 deletions(-)
> > > > 
> > > > diff --git a/fs/f2fs/compress.c b/fs/f2fs/compress.c
> > > > index 4aa166d3d9bf..296ff37d4b08 100644
> > > > --- a/fs/f2fs/compress.c
> > > > +++ b/fs/f2fs/compress.c
> > > > @@ -1719,6 +1719,30 @@ void f2fs_put_page_dic(struct page *page)
> > > >   	f2fs_put_dic(dic);
> > > >   }
> > > > +/*
> > > > + * check whether cluster blocks are contiguous, and add extent cache entry
> > > > + * only if cluster blocks are logically and physically contiguous.
> > > > + */
> > > > +int f2fs_cluster_blocks_are_contiguous(struct dnode_of_data *dn)
> > > > +{
> > > > +	bool compressed = f2fs_data_blkaddr(dn) == COMPRESS_ADDR;
> > > > +	int i = compressed ? 1 : 0;
> > > > +	block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_page,
> > > > +						dn->ofs_in_node + i);
> > > > +
> > > > +	for (i += 1; i < F2FS_I(dn->inode)->i_cluster_size; i++) {
> > > > +		block_t blkaddr = data_blkaddr(dn->inode, dn->node_page,
> > > > +						dn->ofs_in_node + i);
> > > > +
> > > > +		if (!__is_valid_data_blkaddr(blkaddr))
> > > > +			break;
> > > > +		if (first_blkaddr + i - 1 != blkaddr)
> > > > +			return 0;
> 
> The merge condition looks wrong, shouldn't be:
> 
> if (first_blkaddr + i - compressed ? 1 : 0 != blkaddr)
> 	return 0;

Great. This works. I've queued up the patch with this fix.

Thanks,

> 
> Thanks,