Michal Kalderon has found some corner cases around device unload with active NFS mounts that I didn't have the imagination to test when xprtrdma device removal was added last year.

- The ULP device removal handler is responsible for deallocating the PD. That wasn't clear to me initially, and my own testing suggested it was not necessary, but that is incorrect.

- The transport destruction path can no longer assume that there is a valid ID.

- When destroying a transport, ensure that ib_free_cq() is not invoked on a CQ that was already released.

Reported-by: Michal Kalderon Michal.Kalderon@cavium.com Fixes: bebd031866ca ("xprtrdma: Support unplugging an HCA from ...") Signed-off-by: Chuck Lever chuck.lever@oracle.com Cc: stable@vger.kernel.org --- net/sunrpc/xprtrdma/verbs.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-)

diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c index fe518b3..5e3afed 100644 --- a/net/sunrpc/xprtrdma/verbs.c +++ b/net/sunrpc/xprtrdma/verbs.c @@ -250,7 +250,6 @@ wait_for_completion(&ia->ri_remove_done);

ia->ri_id = NULL; - ia->ri_pd = NULL; ia->ri_device = NULL; /* Return 1 to ensure the core destroys the id. */ return 1; @@ -448,7 +447,9 @@ ia->ri_id->qp = NULL; } ib_free_cq(ep->rep_attr.recv_cq); + ep->rep_attr.recv_cq = NULL; ib_free_cq(ep->rep_attr.send_cq); + ep->rep_attr.send_cq = NULL;

/* The ULP is responsible for ensuring all DMA * mappings and MRs are gone. @@ -461,6 +462,8 @@ rpcrdma_dma_unmap_regbuf(req->rl_recvbuf); } rpcrdma_mrs_destroy(buf); + ib_dealloc_pd(ia->ri_pd); + ia->ri_pd = NULL;

/* Allow waiters to continue */ complete(&ia->ri_remove_done); @@ -628,14 +631,16 @@ { cancel_delayed_work_sync(&ep->rep_connect_worker);

- if (ia->ri_id->qp) { + if (ia->ri_id && ia->ri_id->qp) { rpcrdma_ep_disconnect(ep, ia); rdma_destroy_qp(ia->ri_id); ia->ri_id->qp = NULL; }

- ib_free_cq(ep->rep_attr.recv_cq); - ib_free_cq(ep->rep_attr.send_cq); + if (ep->rep_attr.recv_cq) + ib_free_cq(ep->rep_attr.recv_cq); + if (ep->rep_attr.send_cq) + ib_free_cq(ep->rep_attr.send_cq); }

/* Re-establish a connection after a device removal event.