Looks good,
Reviewed-by: Christoph Hellwig <hch@xxxxxx>
Some minor comments below:
> +STATIC void
> +__xfs_efi_release(
> + struct xfs_efi_log_item *efip)
> +{
> + struct xfs_ail *ailp = efip->efi_item.li_ailp;
> +
> + if (!test_and_clear_bit(XFS_EFI_COMMITTED, &efip->efi_flags)) {
> + spin_lock(&ailp->xa_lock);
> + /* xfs_trans_ail_delete() drops the AIL lock. */
> + xfs_trans_ail_delete(ailp, (xfs_log_item_t *)efip);
The second argument should be &efip->efi_item to preserve ty;e safety.
> void
> xfs_efi_release(xfs_efi_log_item_t *efip,
> uint nextents)
> {
> + ASSERT(atomic_read(&efip->efi_next_extent) >= nextents);
> + if (!atomic_sub_and_test(nextents, &efip->efi_next_extent))
> + return;
>
> + __xfs_efi_release(efip);
Why not just:
if (atomic_sub_and_test(nextents, &efip->efi_next_extent))
__xfs_efi_release(efip);
?
_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs
