On Tue, May 21, 2024 at 06:51:07AM -0700, Christoph Hellwig wrote:
> On Mon, May 20, 2024 at 06:03:38PM -0700, Darrick J. Wong wrote:
> > + tres->tr_logcount = XFS_ATTRSET_LOG_COUNT;
> > tres->tr_logflags = XFS_TRANS_PERM_LOG_RES;
> > + return args->total;
>
> Seems like indentation is off for the XFS_TRANS_PERM_LOG_RES
> assignment?
Oops. I think I could shorten this to:
/* Initialize transaction reservation for an xattr set/replace/upsert */
inline struct xfs_trans_res
xfs_attr_set_resv(
const struct xfs_da_args *args)
{
struct xfs_mount *mp = args->dp->i_mount;
struct xfs_trans_res ret = {
.tr_logres = M_RES(mp)->tr_attrsetm.tr_logres +
M_RES(mp)->tr_attrsetrt.tr_logres * args->total,
.tr_logcount = XFS_ATTRSET_LOG_COUNT,
.tr_logflags = XFS_TRANS_PERM_LOG_RES,
};
return ret;
}
> Also wju does this need to return args->total vs just handling it
> in the caller?
I'm not sure, @total for the remove action used to be in xfs_attr_set
prior to the creation of xfs_attr_recover_work.
> > +/* Initialize transaction reservation for an xattr remove */
> > +unsigned int
> > +xfs_attr_init_remove_trans(
> > + struct xfs_da_args *args,
> > + struct xfs_trans_res *tres)
> > +{
> > + struct xfs_mount *mp = args->dp->i_mount;
> > +
> > + *tres = M_RES(mp)->tr_attrrm;
> > + return XFS_ATTRRM_SPACE_RES(mp);
>
> And do we even need this helper vs open coding it like we do in
> most transaction allocations?
Not really.
--D
