The kernel NFS client specifies 'rdma' as the transport protocol for
NFS/RDMA mounts. This is a generic display token, probably not
something the kernel should lie about. The kernel uses it to form
NFSv4 client IDs, for example.
Until gssd can speak RPC-over-RDMA, convert 'rdma' to 'tcp' in
kernel gss upcalls.
This means that the Linux NFS client requires an NFS/TCP service on
an NFS/RDMA server when establishing a GSS context for proto=rdma
mount points. This is probably not an onerous requirement for the
time being.
Eventually user space support for RPC-over-RDMA might be introduced,
and gssd can use that for context establishment.
Signed-off-by: Chuck Lever <chuck.lever@xxxxxxxxxx>
---
utils/gssd/gssd.c | 16 ++++++++++++++++
1 file changed, 16 insertions(+)
diff --git a/utils/gssd/gssd.c b/utils/gssd/gssd.c
index 7ba27b1..51b2cd7 100644
--- a/utils/gssd/gssd.c
+++ b/utils/gssd/gssd.c
@@ -301,6 +301,22 @@ gssd_read_service_info(int dirfd, struct clnt_info *clp)
goto fail;
}
+ /*
+ * The user space RPC library has no support for
+ * RPC-over-RDMA at this time, so change 'rdma'
+ * to 'tcp', and '20049' to '2049'.
+ */
+ if (strcmp(protoname, "rdma") == 0) {
+ free(protoname);
+ protoname = strdup("tcp");
+ if (!protoname)
+ goto fail;
+ free(port);
+ port = strdup("2049");
+ if (!port)
+ goto fail;
+ }
+
if (!gssd_addrstr_to_sockaddr((struct sockaddr *)&clp->addr,
address, port ? port : ""))
goto fail;
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
