On Wed, Feb 26, 2020 at 11:21:09AM +1100, Dave Chinner wrote:
> > /* subtract away the unused acl entries */
> > - len -= sizeof(struct xfs_acl_entry) *
> > + args.valuelen -= sizeof(struct xfs_acl_entry) *
> > (XFS_ACL_MAX_ENTRIES(ip->i_mount) - acl->a_count);
>
> why do we allocate a maximally sized buffer for the attribute data
> (64kB for v5 filesystems) when we already know the size of the
> data we are about to format into it? Why isn't this just:
>
> if (acl) {
> args.valuelen = sizeof(struct xfs_acl_entry) * acl->a_count;
> args.value = kmem_zalloc_large(args.valuelen, 0);
> if (!args.value)
> return -ENOMEM;
>
> xfs_acl_to_disk(args.value, acl);
> }
We need to also account for the size of the acl header, and have the
XFS_ACL_SIZE macro to do so. But with that it makes sense, and I've
added a patch to do so to the end of the series.
