OFI/MTL: bypass HMEM mem reg for CXI provider

The code in the OFI MTL to register acclerator memory doesn't work well with providers like CXI which require the libfabric
consumer to specify the requested_key field in the fi_mr_attr arg to fi_mr_regattr.  As it turns out there's a simpe workaround
though with this provider as it actually doesn't need accelerator memory to be registered for FI_MSG/FI_TAGGED type transfers.

This patch restores support for transfers to/from accelerator memory using the OFI MTL on SS11 (CXI provider).

Signed-off-by: Howard Pritchard <howardp@lanl.gov>

Author: hppritcha

ompi/mca/mtl/ofi/mtl_ofi.h

@@ -2,7 +2,7 @@
  * Copyright (c) 2013-2018 Intel, Inc. All rights reserved
  * Copyright (c) 2017      Los Alamos National Security, LLC. All rights
  *                         reserved.
- * Copyright (c) 2019-2021 Triad National Security, LLC. All rights
+ * Copyright (c) 2019-2022 Triad National Security, LLC. All rights
  *                         reserved.
  * Copyright (c) 2018-2022 Amazon.com, Inc. or its affiliates.  All Rights reserved.
  *                         reserved.
@@ -306,7 +306,7 @@ int ompi_mtl_ofi_register_buffer(struct opal_convertor_t *convertor,
         return OMPI_SUCCESS;
     }
 
-    if (convertor->flags & CONVERTOR_ACCELERATOR) {
+    if ((convertor->flags & CONVERTOR_ACCELERATOR) && ompi_mtl_ofi.hmem_needs_reg) {
         /* Register buffer */
         int ret;
         struct fi_mr_attr attr = {0};

ompi/mca/mtl/ofi/mtl_ofi_component.c

@@ -777,6 +777,19 @@ ompi_mtl_ofi_component_init(bool enable_progress_threads,
         *accelerator_support = false;
     } else {
         *accelerator_support = true;
+        ompi_mtl_ofi.hmem_needs_reg = true;
+        /*
+         * Workaround for the fact that the CXI provider actually doesn't need for accelerator memory to be registered
+         * for local buffers, but if one does do so using fi_mr_regattr, one actually needs to manage the
+         * requested_key field in the fi_mr_attr attr argument, and the OFI MTL doesn't track which requested_keys
+         * have already been registered. So just set a flag to disable local registration.  Note the OFI BTL doesn't
+         * have a problem here since it uses fi_mr_regattr only within the context of an rcache, and manages the
+         * requested_key field in this way.
+         */
+         if (!strncasecmp(prov->fabric_attr->prov_name, "cxi", 3)) {
+             ompi_mtl_ofi.hmem_needs_reg = false;
+         }
+
     }
 
     /**

ompi/mca/mtl/ofi/mtl_ofi_types.h

@@ -2,6 +2,8 @@
  * Copyright (c) 2013-2018 Intel, Inc. All rights reserved
  *
  * Copyright (c) 2015 Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2022 Triad National Security, LLC. All rights
+ *                    reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -97,6 +99,7 @@ typedef struct mca_mtl_ofi_module_t {
 
     bool is_initialized;
     bool has_posted_initial_buffer;
+    bool hmem_needs_reg;
 
 } mca_mtl_ofi_module_t;