On Wed, Jun 20, 2012 at 03:27:10AM -0400, Christoph Hellwig wrote:
> > +struct _map_info {
> > + xfs_bmbt_irec_t *map; /* map vector for blocks */
> > + xfs_extlen_t map_blocks; /* number of fsbs in map */
> > + xfs_dablk_t map_off; /* last mapped file offset */
> > + int map_size; /* total entries in *map */
> > + int map_valid; /* valid entries in *map */
> > + int nmap; /* mappings to ask xfs_bmapi */
> > + xfs_dir2_db_t curdb; /* db for current block */
> > +};
> > +
> > +struct _ra_info {
>
> Can you give these structure names xfs_dir2_leaf prefixes, please?
Sure.
> Also any reason the ra_info structure is kept entirely separate?
Logical grouping, easy to validate.
> While
> I see a bit of a point to have a logical grouping, it still semes more
> useful to just embedd it into the map_info.
Ok, I can do that.
> > + map_info.map_size = howmany(bufsize + mp->m_dirblksize,
> > + mp->m_sb.sb_blocksize);
> > + map_info.map = kmem_zalloc(map_info.map_size *
> > + sizeof(struct xfs_bmbt_irec), KM_SLEEP);
>
> I'd be tempted to say that the map field in the map_info should be a
> variable sized array the end, and the whole structure should be
> dynamically allocated to get a couple more variables off the stack.
Makes sense, even though we aren't usually under stack pressure
here.
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
|