opal/memory: Move Memory Allocation Hooks usage from openib

These changes fix issue open-mpi/ompi#1336

- improve abstractions: opal/memory/linux component should be single place that opeartes with
Memory Allocation Hooks.
- avoid collisions in case dynamic component open/close: it is safe because it is linked statically.
- does not change original behaivour.

Author: igor-ivanov

ompi/runtime/ompi_mpi_init.c

@@ -92,15 +92,15 @@
 #include "ompi/dpm/dpm.h"
 #include "ompi/mpiext/mpiext.h"
 
-#if defined(MEMORY_LINUX_PTMALLOC2) && MEMORY_LINUX_PTMALLOC2
+#if MEMORY_LINUX_HAVE_MALLOC_HOOK_SUPPORT
 #include "opal/mca/memory/linux/memory_linux.h"
 /* So this sucks, but with OPAL in its own library that is brought in
    implicity from libmpi, there are times when the malloc initialize
    hook in the memory component doesn't work.  So we have to do it
    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 /* MEMORY_LINUX_HAVE_MALLOC_HOOK_SUPPORT */
 
 /* 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/btl/base/btl_base_select.c

@@ -35,6 +35,21 @@
 #include "opal/mca/btl/btl.h"
 #include "opal/mca/btl/base/btl_base_error.h"
 #include "opal/mca/btl/base/base.h"
+#if MEMORY_LINUX_MALLOC_ALIGN_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/memory_linux.h"
+#endif
 
 OBJ_CLASS_INSTANCE( mca_btl_base_selected_module_t,
                     opal_list_item_t,
@@ -107,6 +122,16 @@ int mca_btl_base_select(bool enable_progress_threads,
                                 "select: no init function; ignoring component %s",
                                 component->btl_version.mca_component_name);
         } else {
+#if MEMORY_LINUX_MALLOC_ALIGN_ENABLED
+            if (strcmp(component->btl_version.mca_component_name, "openib") == 0) {
+                /* 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.
+                 */
+                opal_memory_linux_malloc_set_alignment(32, 12288);
+            }
+#endif /* MEMORY_LINUX_MALLOC_ALIGN_ENABLED */
             modules = component->btl_init(&num_btls, enable_progress_threads,
                                           enable_mpi_threads);
 

opal/mca/btl/openib/btl_openib.h

@@ -302,11 +302,6 @@ 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;

opal/mca/btl/openib/btl_openib_component.c

@@ -42,21 +42,6 @@
 #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/event/event.h"
 #include "opal/align.h"
@@ -123,7 +108,6 @@ 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 = {
@@ -147,30 +131,6 @@ 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;
@@ -257,16 +217,6 @@ static int btl_openib_component_close(void)
         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);
 
@@ -2538,20 +2488,6 @@ 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
-
     /* Per https://svn.open-mpi.org/trac/ompi/ticket/1305, check to
        see if $sysfsdir/class/infiniband exists.  If it does not,
        assume that the RDMA hardware drivers are not loaded, and
@@ -2951,13 +2887,6 @@ 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
     if (NULL != btls) {
         free(btls);
     }

opal/mca/btl/openib/btl_openib_mca.c

@@ -703,26 +703,19 @@ 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));
+#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, "opal", "btl", "openib", "memalign",
+                                             MCA_BASE_VAR_SYN_FLAG_DEPRECATED);
+    }
 
-    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 algined 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
+    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 (OPAL_SUCCESS == ret) {
@@ -823,16 +816,5 @@ 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 OPAL_SUCCESS;
 }

opal/mca/btl/openib/configure.m4

@@ -120,28 +120,6 @@ AC_DEFUN([MCA_opal_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" && test "$btl_openib_malloc_hooks_enabled" = "0"],
-          [AC_MSG_ERROR([openib malloc alignment is requested but __malloc_hook is not available])])
-    AC_MSG_CHECKING([whether the openib BTL will use malloc hooks])
-    AS_IF([test "$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])
 

opal/mca/memory/linux/configure.m4

@@ -63,6 +63,42 @@ 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/help-opal-memory-linux.txt

@@ -27,3 +27,10 @@ alternate memory hook manager *may* be used instead (if available).
   Local host:  %s
   UMMU device: %s
   Error:       %s (%d)
+#
+[invalid mca param value]
+WARNING: An invalid MCA parameter value was found for memory/linux
+component.
+
+  Problem:    %s
+  Resolution: %s

opal/mca/memory/linux/hooks.c

@@ -33,6 +33,7 @@
 #include "opal/mca/mca.h"
 #include "opal/mca/memory/memory.h"
 #include "opal/constants.h"
+#include "opal/memoryhooks/memory.h"
 
 #include "opal/mca/memory/linux/memory_linux.h"
 
@@ -734,7 +735,10 @@ 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