On Thu, 2025-02-13 at 10:47 -0500, Olga Kornievskaia wrote:
> Don't ignore return code of adding rdma listener. If nfs.conf has asked
> for "rdma=y" but adding the listener fails, don't ignore the failure.
> Note in soft-rdma-provider environment (such as soft iwarp, soft roce),
> when no address-constraints are used, an "any" listener is created and
> rdma-enabling is done independently.
>
> Fixes: e3b72007ab31 ("nfs-utils: nfsdctl: cleanup listeners if some failed")
> Signed-off-by: Olga Kornievskaia <okorniev@xxxxxxxxxx>
> ---
> utils/nfsdctl/nfsdctl.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/utils/nfsdctl/nfsdctl.c b/utils/nfsdctl/nfsdctl.c
> index 05fecc71..244910ef 100644
> --- a/utils/nfsdctl/nfsdctl.c
> +++ b/utils/nfsdctl/nfsdctl.c
> @@ -1388,7 +1388,7 @@ static int configure_listeners(void)
> if (tcp)
> ret = add_listener("tcp", n->field, port);
> if (rdma)
> - add_listener("rdma", n->field, rdma_port);
> + ret = add_listener("rdma", n->field, rdma_port);
> if (ret)
> return ret;
> }
> @@ -1398,7 +1398,7 @@ static int configure_listeners(void)
> if (tcp)
> ret = add_listener("tcp", "", port);
> if (rdma)
> - add_listener("rdma", "", rdma_port);
> + ret = add_listener("rdma", "", rdma_port);
> }
> return ret;
> }
add_listener() is just adding the given listener to the nfsd_sockets
array in memory. It's not consulting the kernel at that point. That
won't happen until set_listeners() is called.
It does make sense to check the return code though, just in case we
pass it some bogus arguments somehow. So...
Reviewed-by: Jeff Layton <jlayton@xxxxxxxxxx>
