On Mon, May 23, 2016 at 05:07:57PM +0200, Christoph Hellwig wrote:
> As it can be called under the ilock.
>
> Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> ---
> fs/xfs/libxfs/xfs_refcount_btree.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/xfs/libxfs/xfs_refcount_btree.c
> b/fs/xfs/libxfs/xfs_refcount_btree.c
> index 770aee5..3391cbd 100644
> --- a/fs/xfs/libxfs/xfs_refcount_btree.c
> +++ b/fs/xfs/libxfs/xfs_refcount_btree.c
> @@ -339,7 +339,7 @@ xfs_refcountbt_init_cursor(
>
> ASSERT(agno != NULLAGNUMBER);
> ASSERT(agno < mp->m_sb.sb_agcount);
> - cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
> + cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_NOFS);
Looks reasonable, though I suspect the rmapbt cursor needs the same
treatment, since it too can be called that way, I think.
/me has been wondering if all of those cursor allocations ought to be
KM_NOFS, though I haven't had time to investigate that.
--D
>
> cur->bc_tp = tp;
> cur->bc_mp = mp;
> --
> 2.1.4
>
|