net/rds: WARNING: at net/rds/recv.c:222 rds_recv_hs_exthdrs+0xf8/0x1e0

net/rds: WARNING: at net/rds/recv.c:222 rds_recv_hs_exthdrs+0xf8/0x1e0

The stack trace looks as follows:

WARNING: at net/rds/recv.c:222 rds_recv_hs_exthdrs+0xf8/0x1e0 [rds]()
Call Trace:
dump_stack+0x63/0x81
warn_slowpath_common+0x8a/0xc0
warn_slowpath_null+0x1a/0x20
rds_recv_hs_exthdrs+0xf8/0x1e0 [rds]
rds_recv_local.isra.7+0x396/0x440 [rds]
rds_recv_incoming+0x2d8/0x3c0 [rds]
rds_ib_recv_cqe_handler+0x44f/0x6d0 [rds_rdma]
poll_rcq+0x7a/0xa0 [rds_rdma]
rds_ib_rx+0xa4/0x220 [rds_rdma]
rds_ib_tasklet_fn_recv+0x30/0x40 [rds_rdma]
...

commit 041dc3e4d3
("Backport multipath RDS from upstream to UEK4") treats an
incoming rds ping or rds pong differently if the local (in case of pong) or
sender's port (in case of ping) is 1 (RDS_FLAG_PROBE_PORT).

There is nothing stopping rds-ping from picking this port for it's local side
since it does wildcard socket bind.

The fix is to check for t_mp_capable transport.

Orabug: 29201779

Reviewed-by: Haakon Bugge <haakon.bugge@oracle.com>
Reviewed-by: Rama Nichanamatlu <rama.nichanamatlu@oracle.com>
Signed-off-by: Venkat Venkatsubra <venkat.x.venkatsubra@oracle.com>
Signed-off-by: Brian Maly <brian.maly@oracle.com>