[sheepdog] [PATCH v3 3/5] replace structure of inode->data_vdi_id[] from array to btree

[MORITA Kazutaka]

At Thu, 24 Oct 2013 17:46:20 +0800,
Robin Dong wrote:
> 
> 1. add sd_extent_header to manage meta-data in data_vdi_id[] or middle-node
> 2. add new type of object: B-tree object as middle-node
> 
> Signed-off-by: Robin Dong <sanbai at taobao.com>
> ---
>  dog/vdi.c                |    2 +-
>  include/sheepdog_proto.h |   43 ++++-
>  lib/sd_inode.c           |  519 +++++++++++++++++++++++++++++++++++++++++++++-
>  sheep/vdi.c              |    1 +
>  sheepfs/volume.c         |    2 +-
>  5 files changed, 558 insertions(+), 9 deletions(-)
> 
> diff --git a/dog/vdi.c b/dog/vdi.c
> index faf06f0..960e2a0 100644
> --- a/dog/vdi.c
> +++ b/dog/vdi.c
> @@ -558,7 +558,7 @@ static int vdi_create(int argc, char **argv)
>  			goto out;
>  		}
>  
> -		sd_inode_set_vdi(inode, idx, vid);
> +		INODE_SET_VDI(inode, idx, vid);
>  		ret = sd_write_object(vid_to_vdi_oid(vid), 0, &vid, sizeof(vid),
>  				      SD_INODE_HEADER_SIZE + sizeof(vid) * idx,
>  				      0, inode->nr_copies, inode->copy_policy,
> diff --git a/include/sheepdog_proto.h b/include/sheepdog_proto.h
> index 30ff397..c338efa 100644
> --- a/include/sheepdog_proto.h
> +++ b/include/sheepdog_proto.h
> @@ -74,6 +74,9 @@
>  #define SD_RES_JOIN_FAILED   0x18 /* Target node had failed to join sheepdog */
>  #define SD_RES_HALT          0x19 /* Sheepdog is stopped doing IO */
>  #define SD_RES_READONLY      0x1A /* Object is read-only */
> +#define SD_RES_BTREE_NOT_FOUND	0x1B /* Cannot found node in btree */
> +#define SD_RES_BTREE_FOUND   0x1C /* Found node in btree */
> +#define SD_RES_BTREE_REPEAT  0x1D /* Should repeat op in btree */
>  
>  /* errors above 0x80 are sheepdog-internal */
>  
> @@ -92,8 +95,9 @@
>  #define VDI_BIT (UINT64_C(1) << 63)
>  #define VMSTATE_BIT (UINT64_C(1) << 62)
>  #define VDI_ATTR_BIT (UINT64_C(1) << 61)
> +#define VDI_BTREE_BIT (UINT64_C(1) << 60)
>  #define MAX_DATA_OBJS (1ULL << 20)
> -#define MAX_CHILDREN 1024U
> +#define MAX_CHILDREN (1024U - 1) /* we use the last uint32_t as btree_counter */
>  #define SD_MAX_VDI_LEN 256U
>  #define SD_MAX_VDI_TAG_LEN 256U
>  #define SD_MAX_VDI_ATTR_KEY_LEN 256U
> @@ -104,8 +108,8 @@
>  #define SD_MAX_VDI_SIZE (SD_DATA_OBJ_SIZE * MAX_DATA_OBJS)
>  
>  #define SD_INODE_SIZE (sizeof(struct sd_inode))
> -#define SD_INODE_HEADER_SIZE (sizeof(struct sd_inode) - \
> -			      sizeof(uint32_t) * MAX_DATA_OBJS)
> +#define SD_INODE_INDEX_SIZE (sizeof(uint32_t) * MAX_DATA_OBJS)
> +#define SD_INODE_HEADER_SIZE (sizeof(struct sd_inode) - SD_INODE_INDEX_SIZE)
>  #define SD_ATTR_OBJ_SIZE (sizeof(struct sheepdog_vdi_attr))
>  #define CURRENT_VDI_ID 0
>  
> @@ -215,16 +219,35 @@ struct sd_inode {
>  	uint64_t vdi_size;
>  	uint64_t vm_state_size;
>  	uint8_t  copy_policy;
> -	uint8_t  reserved;
> +	uint8_t  store_policy;

I wonder if we need another field.  Can we share copy_policy and
store_policy?

>  	uint8_t  nr_copies;
>  	uint8_t  block_size_shift;
>  	uint32_t snap_id;
>  	uint32_t vdi_id;
>  	uint32_t parent_vdi_id;
>  	uint32_t child_vdi_id[MAX_CHILDREN];
> +	uint32_t btree_counter;

This breaks backward compatibility.  Please put the field at least the
end of this structure (after data_vdi_id).

Thanks,

Kazutaka