2007-12-18 Joel Sherrill <joel.sherrill@oarcorp.com>

	* libcsupport/Makefile.am, libcsupport/preinstall.am,
	libcsupport/src/malloc.c, libcsupport/src/mallocinfo.c,
	libmisc/Makefile.am, libmisc/shell/main_mallocinfo.c,
	libmisc/shell/shellconfig.h: Split malloc.c into multiple files with
	one function per file. Also split out statistics into a separate file
	which can be plugged in dynamically. Right now, it is always in. I
	suspect that splitting the file removed more code than leaving
	statistics in. I tinkered with malloc information command in the
	shell. I resurrected the malloc arena code as malloc boundary. This
	code is now compiled all the time even though it does not appear to
	work.
	* libcsupport/include/rtems/malloc.h, libcsupport/src/_calloc_r.c,
	libcsupport/src/_free_r.c, libcsupport/src/_malloc_r.c,
	libcsupport/src/_realloc_r.c, libcsupport/src/calloc.c,
	libcsupport/src/free.c, libcsupport/src/malloc_boundary.c,
	libcsupport/src/malloc_get_statistics.c,
	libcsupport/src/malloc_initialize.c, libcsupport/src/malloc_p.h,
	libcsupport/src/malloc_report_statistics.c,
	libcsupport/src/malloc_report_statistics_plugin.c,
	libcsupport/src/malloc_statistics_helpers.c,
	libcsupport/src/malloc_walk.c, libcsupport/src/realloc.c,
	libmisc/shell/main_perioduse.c: New files.

1 files changed, 128 insertions, 0 deletions

diff --git a/cpukit/libcsupport/src/malloc_initialize.c b/cpukit/libcsupport/src/malloc_initialize.c
new file mode 100644
index 0000000000..08fc351f7f
--- /dev/null
+++ b/cpukit/libcsupport/src/malloc_initialize.c
@@ -0,0 +1,128 @@
+/*
+ *  RTEMS Malloc Family Implementation --Initialization
+ *
+ *
+ *  COPYRIGHT (c) 1989-2007.
+ *  On-Line Applications Research Corporation (OAR).
+ *
+ *  The license and distribution terms for this file may be
+ *  found in the file LICENSE in this distribution or at
+ *  http://www.rtems.com/license/LICENSE.
+ *
+ *  $Id$
+ */
+
+#if HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include <rtems.h>
+#include <rtems/malloc.h>
+#include "malloc_p.h"
+
+Heap_Control              RTEMS_Malloc_Heap;
+Chain_Control             RTEMS_Malloc_GC_list;
+size_t                    RTEMS_Malloc_Sbrk_amount;
+rtems_malloc_statistics_t rtems_malloc_statistics;
+
+void RTEMS_Malloc_Initialize(
+  void   *start,
+  size_t  length,
+  size_t  sbrk_amount
+)
+{
+  uint32_t      status;
+  void         *starting_address;
+  uintptr_t     old_address;
+  uintptr_t     uaddress;
+
+  #if defined(RTEMS_MALLOC_BOUNDARY_HELPERS)
+    /*
+     *  If configured, initialize the boundary support
+     */
+    if (rtems_malloc_boundary_helpers)
+      (*rtems_malloc_boundary_helpers->initialize)();
+  #endif
+
+  /*
+   *  If configured, initialize the statistics support
+   */
+  if ( rtems_malloc_statistics_helpers )
+    (*rtems_malloc_statistics_helpers->initialize)();
+
+  /*
+   *  Initialize the garbage collection list to start with nothing on it.
+   */
+  Chain_Initialize_empty(&RTEMS_Malloc_GC_list);
+
+  /*
+   * If the starting address is 0 then we are to attempt to
+   * get length worth of memory using sbrk. Make sure we
+   * align the address that we get back.
+   */
+
+  starting_address = start;
+  RTEMS_Malloc_Sbrk_amount = sbrk_amount;
+
+  if (!starting_address) {
+    uaddress = (uintptr_t)sbrk(length);
+
+    if (uaddress == (uintptr_t) -1) {
+      rtems_fatal_error_occurred( RTEMS_NO_MEMORY );
+      /* DOES NOT RETURN!!! */
+    }
+
+    if (uaddress & (CPU_HEAP_ALIGNMENT-1)) {
+      old_address = uaddress;
+      uaddress = (uaddress + CPU_HEAP_ALIGNMENT) & ~(CPU_HEAP_ALIGNMENT-1);
+
+      /*
+       * adjust the length by whatever we aligned by
+       */
+      length -= uaddress - old_address;
+    }
+
+    starting_address = (void *)uaddress;
+  }
+
+  /*
+   *  If the BSP is not clearing out the workspace, then it is most likely
+   *  not clearing out the initial memory for the heap.  There is no
+   *  standard supporting zeroing out the heap memory.  But much code
+   *  with UNIX history seems to assume that memory malloc'ed during
+   *  initialization (before any free's) is zero'ed.  This is true most
+   *  of the time under UNIX because zero'ing memory when it is first
+   *  given to a process eliminates the chance of a process seeing data
+   *  left over from another process.  This would be a security violation.
+   */
+
+  if ( rtems_configuration_get_do_zero_of_workspace() )
+     memset( starting_address, 0, length );
+
+  /*
+   *  Unfortunately we cannot use assert if this fails because if this
+   *  has failed we do not have a heap and if we do not have a heap
+   *  STDIO cannot work because there will be no buffers.
+   */
+
+  status = _Protected_heap_Initialize( 
+    &RTEMS_Malloc_Heap,
+    starting_address,
+    length,
+    CPU_HEAP_ALIGNMENT
+  );
+  if ( !status )
+    rtems_fatal_error_occurred( status );
+
+  #if defined(RTEMS_HEAP_DEBUG)
+    if ( _Protected_heap_Walk( &RTEMS_Malloc_Heap, 0, FALSE ) ) {
+      printk( "Malloc heap not initialized correctly\n" );
+      rtems_print_buffer( start, 32 );
+      printk( "\n" );
+      rtems_print_buffer( (start + length) - 48, 48 );
+      rtems_fatal_error_occurred( RTEMS_NO_MEMORY );
+    }
+  #endif
+
+  MSBUMP(space_available, length);
+}