Merge pull request #1002 from igor-ivanov/pr/v1.x/issue-1336

v1.10: Move Memory Allocation Hooks usage from openib

ompi/mca/btl/openib/btl_openib.h

@@ -310,11 +310,6 @@ struct mca_btl_openib_component_t {
 #if BTL_OPENIB_FAILOVER_ENABLED
     int verbose_failover;
 #endif
-#if BTL_OPENIB_MALLOC_HOOKS_ENABLED
-    int use_memalign;
-    size_t memalign_threshold;
-    void* (*previous_malloc_hook)(size_t __size, const void*);
-#endif
 #if OPAL_CUDA_SUPPORT
     bool cuda_async_send;
     bool cuda_async_recv;

ompi/mca/btl/openib/btl_openib_component.c

@@ -11,15 +11,15 @@
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
  * Copyright (c) 2006-2013 Cisco Systems, Inc.  All rights reserved.
- * Copyright (c) 2006-2009 Mellanox Technologies. All rights reserved.
+ * Copyright (c) 2006-2016 Mellanox Technologies. All rights reserved.
  * Copyright (c) 2006-2013 Los Alamos National Security, LLC.  All rights
  *                         reserved.
  * Copyright (c) 2006-2007 Voltaire All rights reserved.
  * Copyright (c) 2009-2012 Oracle and/or its affiliates.  All rights reserved.
  * Copyright (c) 2011-2014 NVIDIA Corporation.  All rights reserved.
  * Copyright (c) 2012      Oak Ridge National Laboratory.  All rights reserved
  * Copyright (c) 2013      Intel, Inc. All rights reserved
- * Copyright (c) 2014-2015 Research Organization for Information Science
+ * Copyright (c) 2014-2016 Research Organization for Information Science
  *                         and Technology (RIST). All rights reserved.
  * Copyright (c) 2014      Bull SAS.  All rights reserved.
  * $COPYRIGHT$
@@ -42,22 +42,8 @@
 #include <fcntl.h>
 #include <stdlib.h>
 #include <stddef.h>
-#if BTL_OPENIB_MALLOC_HOOKS_ENABLED
-/*
- * The include of malloc.h below breaks abstractions in OMPI (by
- * directly including a header file from another component), but has
- * been ruled "ok" because the openib component is only supported on
- * Linux.  
- *
- * The malloc hooks in newer glibc were deprecated, including stock
- * malloc.h causes compilation warnings.  Instead, we use the internal
- * linux component malloc.h which does not cause these warnings.
- * Internally, OMPI uses the built-in ptmalloc from the linux memory
- * component anyway.
- */
-#include "opal/mca/memory/linux/malloc.h"
-#endif
 
+#include "opal/mca/memory/memory.h"
 #include "opal/mca/event/event.h"
 #include "opal/align.h"
 #include "opal/util/output.h"
@@ -129,7 +115,6 @@ static void btl_openib_handle_incoming_completion(mca_btl_base_module_t* btl,
  * Local variables
  */
 static mca_btl_openib_device_t *receive_queues_device = NULL;
-static bool malloc_hook_set = false;
 static int num_devices_intentionally_ignored = 0;
 
 mca_btl_openib_component_t mca_btl_openib_component = {
@@ -159,30 +144,6 @@ mca_btl_openib_component_t mca_btl_openib_component = {
     }
 };
 
-#if BTL_OPENIB_MALLOC_HOOKS_ENABLED
-/* This is a memory allocator hook. The purpose of this is to make
- * every malloc aligned since this speeds up IB HCA work.
- * There two basic cases here:
- *
- * 1. Memory manager for Open MPI is enabled. Then memalign below will
- * be overridden by __memalign_hook which is set to
- * opal_memory_linux_memalign_hook.  Thus, _malloc_hook is going to
- * use opal_memory_linux_memalign_hook.
- *
- * 2. No memory manager support. The memalign below is just regular glibc
- * memalign which will be called through __malloc_hook instead of malloc.
- */
-static void *btl_openib_malloc_hook(size_t sz, const void* caller)
-{
-    if (sz < mca_btl_openib_component.memalign_threshold &&
-        malloc_hook_set) {
-        return mca_btl_openib_component.previous_malloc_hook(sz, caller);
-    } else {
-        return memalign(mca_btl_openib_component.use_memalign, sz);
-    }
-}
-#endif
-
 static int btl_openib_component_register(void)
 {
     int ret;
@@ -287,16 +248,6 @@ static int btl_openib_component_close(void)
     if (NULL != mca_btl_openib_component.default_recv_qps) {
         free(mca_btl_openib_component.default_recv_qps);
     }
-
-#if BTL_OPENIB_MALLOC_HOOKS_ENABLED
-    /* Must check to see whether the malloc hook was set before
-       assigning it back because ompi_info will call _register() and
-       then _close() (which won't set the hook) */
-    if (malloc_hook_set) {
-        __malloc_hook = mca_btl_openib_component.previous_malloc_hook;
-        malloc_hook_set = false;
-    }
-#endif
 
     /* close memory registration debugging output */
     opal_output_close (mca_btl_openib_component.memory_registration_verbose);
@@ -2461,19 +2412,8 @@ btl_openib_component_init(int *num_btl_modules,
     *num_btl_modules = 0;
     num_devs = 0;
 
-#if BTL_OPENIB_MALLOC_HOOKS_ENABLED
-    /* If we got this far, then setup the memory alloc hook (because
-       we're most likely going to be using this component). The hook
-       is to be set up as early as possible in this function since we
-       want most of the allocated resources be aligned.*/
-    if (mca_btl_openib_component.use_memalign > 0 &&
-        (opal_mem_hooks_support_level() &
-            (OPAL_MEMORY_FREE_SUPPORT | OPAL_MEMORY_CHUNK_SUPPORT)) != 0) {
-        mca_btl_openib_component.previous_malloc_hook = __malloc_hook;
-        __malloc_hook = btl_openib_malloc_hook;
-        malloc_hook_set = true;
-    }
-#endif
+    opal_memory->memoryc_set_alignment(32, mca_btl_openib_module.super.btl_eager_limit);
+
     /* Currently refuse to run if MPI_THREAD_MULTIPLE is enabled */
     if (ompi_mpi_thread_multiple && !mca_btl_base_thread_multiple_override) {
         opal_output_verbose(5, ompi_btl_base_framework.framework_output,
@@ -2894,13 +2834,6 @@ btl_openib_component_init(int *num_btl_modules,
 
     mca_btl_openib_component.ib_num_btls = 0;
     btl_openib_modex_send();
-#if BTL_OPENIB_MALLOC_HOOKS_ENABLED
-    /*Unset malloc hook since the component won't start*/
-    if (malloc_hook_set) {
-        __malloc_hook = mca_btl_openib_component.previous_malloc_hook;
-        malloc_hook_set = false;
-    }
-#endif
     return NULL;
 }
 

ompi/mca/btl/openib/btl_openib_mca.c

@@ -703,26 +703,18 @@ int btl_openib_register_mca_params(void)
                   0, &mca_btl_openib_component.gid_index,
                   REGINT_GE_ZERO));
 
-#if BTL_OPENIB_MALLOC_HOOKS_ENABLED
-    CHECK(reg_int("memalign", NULL,
-                  "[64 | 32 | 0] - Enable (64bit or 32bit)/Disable(0) memory"
-                  "alignment for all malloc calls if btl openib is used.",
-                  32, &mca_btl_openib_component.use_memalign,
-                  REGINT_GE_ZERO));
-
-    mca_btl_openib_component.memalign_threshold =
-        mca_btl_openib_module.super.btl_eager_limit;
-    tmp = mca_base_component_var_register(&mca_btl_openib_component.super.btl_version,
-                                          "memalign_threshold",
-                                          "Allocating memory more than btl_openib_memalign_threshhold"
-                                          "bytes will automatically be aligned to the value of btl_openib_memalign bytes."
-                                          "memalign_threshhold defaults to the same value as mca_btl_openib_eager_limit.",
-                                          MCA_BASE_VAR_TYPE_SIZE_T, NULL, 0, 0,
-                                          OPAL_INFO_LVL_9,
-                                          MCA_BASE_VAR_SCOPE_READONLY,
-                                          &mca_btl_openib_component.memalign_threshold);
-    if (0 > tmp) ret = tmp;
-#endif
+#if MEMORY_LINUX_MALLOC_ALIGN_ENABLED
+    tmp = mca_base_var_find ("opal", "memory", "linux", "memalign");
+    if (0 <= tmp) {
+        (void) mca_base_var_register_synonym(tmp, "ompi", "btl", "openib", "memalign",
+                                             MCA_BASE_VAR_SYN_FLAG_DEPRECATED);
+    }
+    tmp = mca_base_var_find ("opal", "memory", "linux", "memalign_threshold");
+    if (0 <= tmp) {
+        (void) mca_base_var_register_synonym(tmp, "opal", "btl", "openib", "memalign_threshold",
+                                             MCA_BASE_VAR_SYN_FLAG_DEPRECATED);
+    }
+#endif /* MEMORY_LINUX_MALLOC_ALIGN_ENABLED */
 
     /* Register any MCA params for the connect pseudo-components */
     if (OMPI_SUCCESS == ret) {
@@ -816,16 +808,5 @@ int btl_openib_verify_mca_params (void)
     }
 #endif
 
-#if BTL_OPENIB_MALLOC_HOOKS_ENABLED
-    if (mca_btl_openib_component.use_memalign != 32  
-        && mca_btl_openib_component.use_memalign != 64
-        && mca_btl_openib_component.use_memalign != 0){ 
-        opal_show_help("help-mpi-btl-openib.txt", "invalid mca param value",
-                       true, "Wrong btl_openib_memalign parameter value. Allowed values: 64, 32, 0.",
-                       "btl_openib_memalign is reset to 32");
-        mca_btl_openib_component.use_memalign = 32; 
-    }
-#endif
-
     return OMPI_SUCCESS;
 }

ompi/mca/btl/openib/configure.m4

@@ -129,29 +129,6 @@ AC_DEFUN([MCA_ompi_btl_openib_CONFIG],[
                        [enable openib BTL failover])
     AM_CONDITIONAL([MCA_btl_openib_enable_failover], [test "x$btl_openib_failover_enabled" = "x1"])
 
-
-    # Check for __malloc_hook availability
-    AC_ARG_ENABLE(btl-openib-malloc-alignment,
-    	AC_HELP_STRING([--enable-btl-openib-malloc-alignment], [Enable support for allocated memory alignment. Default: enabled if supported, disabled otherwise.]))
-
-    btl_openib_malloc_hooks_enabled=0
-    AS_IF([test "$enable_btl_openib_malloc_alignment" != "no"],
-        [AC_CHECK_HEADER([malloc.h],
-             [AC_CHECK_FUNC([__malloc_hook],
-                  [AC_CHECK_FUNC([__realloc_hook],
-                       [AC_CHECK_FUNC([__free_hook],
-                            [btl_openib_malloc_hooks_enabled=1])])])])])
-
-    AS_IF([test "$enable_btl_openib_malloc_alignment" = "yes" -a "$btl_openib_malloc_hooks_enabled" = "0"],
-          [AC_MSG_ERROR([openib malloc alignment is requested but __malloc_hook is not available])])
-    AC_MSG_CHECKING([whether the openib BTL will use malloc hooks])
-    AS_IF([test "$btl_openib_malloc_hooks_enabled" = "0"],
-          [AC_MSG_RESULT([no])],
-          [AC_MSG_RESULT([yes])])
-
-    AC_DEFINE_UNQUOTED(BTL_OPENIB_MALLOC_HOOKS_ENABLED, [$btl_openib_malloc_hooks_enabled],
-                       [Whether the openib BTL malloc hooks are enabled]) 
-
     # make sure that CUDA-aware checks have been done
     AC_REQUIRE([OPAL_CHECK_CUDA])
 

ompi/runtime/ompi_mpi_init.c

@@ -103,7 +103,7 @@
    from here, since any MPI code is going to call MPI_Init... */
 OPAL_DECLSPEC void (*__malloc_initialize_hook) (void) = 
     opal_memory_linux_malloc_init_hook;
-#endif
+#endif /* defined(MEMORY_LINUX_PTMALLOC2) && MEMORY_LINUX_PTMALLOC2 */
 
 /* This is required for the boundaries of the hash tables used to store
  * the F90 types returned by the MPI_Type_create_f90_XXX functions.

opal/mca/memory/base/empty.h

@@ -1,5 +1,7 @@
 /*
- * Copyright (c) 2009 Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2009      Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2016      Research Organization for Information Science
+ *                         and Technology (RIST). All rights reserved.
  * $COPYRIGHT$
  * 
  * Additional copyrights may follow
@@ -51,6 +53,15 @@ OPAL_DECLSPEC int opal_memory_base_component_deregister_empty(void *start,
                                                               size_t len,
                                                               uint64_t cookie);
 
+/**
+ * Default (empty) implementation of the memoryc_set_alignment function
+ *
+ * See opal/mca/memory/memory.h for a description of the parameters.
+ */
+OPAL_DECLSPEC void opal_memory_base_component_set_alignment_empty(int use_memalign,
+                                                                  size_t memalign_threshold);
+
+
 END_C_DECLS
 
 #endif

opal/mca/memory/base/memory_base_empty.c

@@ -9,6 +9,8 @@
  *                         University of Stuttgart.  All rights reserved.
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
+ * Copyright (c) 2016      Research Organization for Information Science
+ *                         and Technology (RIST). All rights reserved.
  * $COPYRIGHT$
  * 
  * Additional copyrights may follow
@@ -34,3 +36,9 @@ int opal_memory_base_component_deregister_empty(void *base, size_t len,
 {
     return OPAL_SUCCESS;
 }
+
+void opal_memory_base_component_set_alignment_empty(int use_memalign,
+                                                    size_t memalign_threshold)
+{
+}
+

opal/mca/memory/base/memory_base_open.c

@@ -11,6 +11,8 @@
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
  * Copyright (c) 2009      Cisco Systems, Inc.  All rights reserved.
+ * Copyright (c) 2016      Research Organization for Information Science
+ *                         and Technology (RIST). All rights reserved.
  * $COPYRIGHT$
  * 
  * Additional copyrights may follow
@@ -54,6 +56,7 @@ static opal_memory_base_component_2_0_0_t empty_component = {
     empty_process,
     opal_memory_base_component_register_empty,
     opal_memory_base_component_deregister_empty,
+    opal_memory_base_component_set_alignment_empty,
 };
 
 

opal/mca/memory/linux/configure.m4

@@ -61,6 +61,42 @@ AC_DEFUN([MCA_opal_memory_linux_CONFIG],[
                  [memory_linux_ptmalloc2_happy=no
                   memory_linux_ummu_happy=no])])
 
+
+    ######################################################################
+    # if memory hook available
+    ######################################################################
+    memory_hook_found=1
+    AS_IF([test "$memory_hook_found" -eq 1],
+        [memory_hook_found=0 AC_CHECK_HEADER([malloc.h],
+             [AC_CHECK_FUNC([__malloc_initialize_hook],
+                 [AC_CHECK_FUNC([__malloc_hook],
+                     [AC_CHECK_FUNC([__realloc_hook],
+                         [AC_CHECK_FUNC([__free_hook],
+                            [memory_hook_found=1])])])])])])
+    AC_MSG_CHECKING([whether the system can use malloc hooks])
+    AS_IF([test "$memory_hook_found" = "0"],
+          [AC_MSG_RESULT([no])],
+          [AC_MSG_RESULT([yes])])
+    AC_DEFINE_UNQUOTED([MEMORY_LINUX_HAVE_MALLOC_HOOK_SUPPORT], [$memory_hook_found],
+                   	   [Whether the system has Memory Allocation Hooks])
+
+    AC_ARG_ENABLE(memory-linux-malloc-alignment,
+        AC_HELP_STRING([--enable-memory-linux-malloc-alignment], [Enable support for allocated memory alignment. Default: enabled if supported, disabled otherwise.]))
+
+    malloc_align_enabled=0
+    AS_IF([test "$enable_memory_linux_malloc_alignment" != "no"],
+        [malloc_align_enabled=$memory_hook_found])
+
+    AS_IF([test "$enable_memory_linux_malloc_alignment" = "yes" && test "$malloc_align_enabled" = "0"],
+          [AC_MSG_ERROR([memory linux malloc alignment is requested but __malloc_hook is not available])])
+    AC_MSG_CHECKING([whether the memory linux will use malloc alignment])
+    AS_IF([test "$malloc_align_enabled" = "0"],
+          [AC_MSG_RESULT([no])],
+          [AC_MSG_RESULT([yes])])
+
+    AC_DEFINE_UNQUOTED(MEMORY_LINUX_MALLOC_ALIGN_ENABLED, [$malloc_align_enabled],
+                       [Whether the memory linux malloc alignment is enabled])
+
     ######################################################################
     # ptmalloc2
     ######################################################################

opal/mca/memory/linux/hooks.c

@@ -34,6 +34,7 @@
 #include "opal/mca/memory/memory.h"
 #include "opal/util/show_help.h"
 #include "opal/constants.h"
+#include "opal/memoryhooks/memory.h"
 
 #include "opal/mca/memory/linux/memory_linux.h"
 
@@ -735,7 +736,10 @@ static check_result_t check(const char *name)
     }
 }
 
-/* OMPI's init function */
+
+/* This function is called on loading libmpi in case system has Memory Allocation Hooks
+ * (see ompi/runtime/ompi_mpi_init.c for details)
+ */
 void opal_memory_linux_malloc_init_hook(void)
 {
     check_result_t r1, lp, lpp;

opal/mca/memory/linux/memory_linux.h

@@ -31,6 +31,11 @@ typedef struct opal_memory_linux_component_t {
     int ummunotify_fd;
 #endif
 
+#if MEMORY_LINUX_MALLOC_ALIGN_ENABLED
+    int use_memalign;
+    size_t memalign_threshold;
+#endif
+
 #if MEMORY_LINUX_PTMALLOC2
     /* Ptmalloc2-specific data. Note that these variables are all marked as volatile.
      * This is needed because of what may be a buggy optimization in the GCC 4.9.2
@@ -64,13 +69,20 @@ int opal_memory_linux_ummunotify_close(void);
 /* memory_linux_ptmalloc2.c */
 int opal_memory_linux_ptmalloc2_open(void);
 int opal_memory_linux_ptmalloc2_close(void);
-OPAL_DECLSPEC void opal_memory_linux_malloc_init_hook(void);
 
 /* memory_linux_munmap.c */
 OPAL_DECLSPEC int opal_memory_linux_free_ptmalloc2_munmap(void *start, size_t length, int from_alloc);
 OPAL_DECLSPEC int munmap(void* addr, size_t len);
 #endif /* !MEMORY_LINUX_PTMALLOC2 */
 
+#if MEMORY_LINUX_HAVE_MALLOC_HOOK_SUPPORT
+OPAL_DECLSPEC void opal_memory_linux_malloc_init_hook(void);
+#endif /* MEMORY_LINUX_HAVE_MALLOC_HOOK_SUPPORT */
+
+#if MEMORY_LINUX_MALLOC_ALIGN_ENABLED
+OPAL_DECLSPEC void opal_memory_linux_malloc_set_alignment(int use_memalign, size_t memalign_threshold);
+#endif /* MEMORY_LINUX_MALLOC_ALIGN_ENABLED */
+
 END_C_DECLS
 
 #endif