On Wed, Sep 25, 2019 at 02:40:59PM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
>
> (Ab)use the btheight command to calculate the geometry of the incore
> extent tree.
>
> Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> ---
> db/btheight.c | 87 +++++++++++++++++++++++++++++++++++++++------------------
> 1 file changed, 60 insertions(+), 27 deletions(-)
>
>
> diff --git a/db/btheight.c b/db/btheight.c
> index e2c9759f..be604ebc 100644
> --- a/db/btheight.c
> +++ b/db/btheight.c
> @@ -22,18 +22,37 @@ static int rmap_maxrecs(struct xfs_mount *mp, int blocklen, int leaf)
> return libxfs_rmapbt_maxrecs(blocklen, leaf);
> }
>
> +static int iext_maxrecs(struct xfs_mount *mp, int blocklen, int leaf)
> +{
> + blocklen -= 2 * sizeof(void *);
> +
> + return blocklen / sizeof(struct xfs_bmbt_rec);
> +}
This isn't correct for the iext nodes. They hold 16 key/ptr pairs,
not 15.
I suspect you should be lifting the iext btree format definitions
like this one:
enum {
NODE_SIZE = 256,
KEYS_PER_NODE = NODE_SIZE / (sizeof(uint64_t) + sizeof(void *)),
RECS_PER_LEAF = (NODE_SIZE - (2 * sizeof(struct xfs_iext_leaf *))) /
sizeof(struct xfs_iext_rec),
};
from libxfs/xfs_iext_tree.c to a libxfs header file and then using
KEYS_PER_NODE and RECS_PER_LEAF here. See the patch below, lifted
from a varaint of my range locking prototypes...
However, these are not on-disk values and so are subject to change,
hence it may be that a warning might be needed when xfs_db is used
to calculate the height of this tree.
> +static int disk_blocksize(struct xfs_mount *mp)
> +{
> + return mp->m_sb.sb_blocksize;
> +}
> +
> +static int iext_blocksize(struct xfs_mount *mp)
> +{
> + return 256;
> +}
NODE_SIZE....
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
