RDMA/core: Fix corrupted SL on passive side
authorHåkon Bugge <haakon.bugge@oracle.com>
Mon, 22 Mar 2021 13:35:32 +0000 (14:35 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 14 May 2021 07:50:32 +0000 (09:50 +0200)
commita16f02187d9d1f351b1b43d6cbd4f297b6518c2c
tree96a12d80a2a656ef7095ce8d04b626c3c750494c
parent556e75a0ae034e8850adaa12f305de615439b724
RDMA/core: Fix corrupted SL on passive side

[ Upstream commit 194f64a3cad3ab9e381e996a13089de3215d1887 ]

On RoCE systems, a CM REQ contains a Primary Hop Limit > 1 and Primary
Subnet Local is zero.

In cm_req_handler(), the cm_process_routed_req() function is called. Since
the Primary Subnet Local value is zero in the request, and since this is
RoCE (Primary Local LID is permissive), the following statement will be
executed:

      IBA_SET(CM_REQ_PRIMARY_SL, req_msg, wc->sl);

This corrupts SL in req_msg if it was different from zero. In other words,
a request to setup a connection using an SL != zero, will not be honored,
and a connection using SL zero will be created instead.

Fixed by not calling cm_process_routed_req() on RoCE systems, the
cm_process_route_req() is only for IB anyhow.

Fixes: 3971c9f6dbf2 ("IB/cm: Add interim support for routed paths")
Link: https://lore.kernel.org/r/1616420132-31005-1-git-send-email-haakon.bugge@oracle.com
Signed-off-by: Håkon Bugge <haakon.bugge@oracle.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
drivers/infiniband/core/cm.c