source: trunk/src/dnmalloc.c@ 558

/* DistriNet malloc (dnmalloc): a more secure memory allocator. 
   Copyright (C) 2005, Yves Younan, Wouter Joosen, Frank Piessens 
   and Rainer Wichmann

   The authors can be contacted by:
      Email: dnmalloc@fort-knox.org
      Address:
#if defined(WITH_TPT) 
             Yves Younan
             Celestijnenlaan 200A
             B-3001 Heverlee
             Belgium
   
   This library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   This library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with this library; if not, write to the Free Software
   Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
   
*/

/* Current version: dnmalloc 1.0 */
/* Includes arc4random from OpenBSD, which is under the BDS license     */

/* Versions:
   0.1-0.5:
   Proof of concept implementation by Hans Van den Eynden and Yves Younan
   0.6-0.7:
   Bug fixes by Yves Younan
   0.8-1.0.beta4:
   Reimplementation from scratch by Yves Younan
   1.0.beta4:
   Public release
   1.0.beta5:
   Prev_chunkinfo speeded up, was really slow because of the way we did lookups
   A freechunkinfo region is now freed when it is completely empty and 
   not the current one

   1.0 (Rainer Wichmann [support at la dash samhna dot org]):
   ---------------------

   Compiler warnings fixed
   Define REALLOC_ZERO_BYTES_FREES because it's what GNU libc does
       (and what the standard says)
   Removed unused code
   Fix       assert(aligned_OK(chunk(newp)));
         ->  assert(aligned_OK(chunk(oldp)));
   Fix statistics in sYSMALLOc
   Fix overwrite of av->top in sYSMALLOc
   Provide own assert(), glibc assert() doesn't work (calls malloc)
   Fix bug in mEMALIGn(), put remainder in hashtable before calling fREe
   Remove cfree, independent_cmalloc, independent_comalloc (untested
       public functions not covered by any standard)
   Provide posix_memalign (that one is in the standard)
   Move the malloc_state struct to mmapped memory protected by guard pages 
   Add arc4random function to initialize random canary on startup
   Implement random canary at end of (re|m)alloced/memaligned buffer,
       check at free/realloc
   Remove code conditional on !HAVE_MMAP, since mmap is required anyway.
   Use standard HAVE_foo macros (as generated by autoconf) instead of LACKS_foo

   Profiling: Reorder branches in hashtable_add, next_chunkinfo, 
                  prev_chunkinfo, hashtable_insert, mALLOc, fREe, request2size,
                  checked_request2size (gcc predicts if{} branch to be taken).
              Use UNLIKELY macro (gcc __builtin_expect()) where branch
                  reordering would make the code awkward.

   Portability: Hashtable always covers full 32bit address space to
                avoid assumptions about memory layout.
   Portability: Try hard to enforce mapping of mmapped memory into
                32bit address space, even on 64bit systems.
   Portability: Provide a dnmalloc_pthread_init() function, since
                pthread locking on HP-UX only works if initialized
                after the application has entered main().
   Portability: On *BSD, pthread_mutex_lock is unusable since it
                calls malloc, use spinlocks instead.
   Portability: Dynamically detect whether the heap is within
                32bit address range (e.g. on Linux x86_64, it isn't).
                Don't use sbrk() if the heap is mapped to an address 
                outside the 32bit range, since this doesn't work with 
                the hashtable. New macro morecore32bit.
                
   Success on: HP-UX 11.11/pthread, Linux/pthread (32/64 bit),
               FreeBSD/pthread, and Solaris 10 i386/pthread.
   Fail    on: OpenBSD/pthread (in  _thread_machdep_save_float_state),
               might be related to OpenBSD pthread internals (??).
               Non-treaded version (#undef USE_MALLOC_LOCK) 
               works on OpenBSD.
   
   further to 1.0:
   Valgrind client requests inserted (#define USE_VALGRIND)
   Fix: malloc_consolidate (nextchunk->fd, nextchunk->bck may be NULL)
   Portability: minsize = 32 bit on 64bit architecture
   Minor cleanups
   Fix: eliminate prototypes for memset, memcpy (they're in string.h)
   Fix: one more malloc_consolidate segfault

   There may be some bugs left in this version. please use with caution.
*/



/* Please read the following papers for documentation: 

   Yves Younan, Wouter Joosen, and Frank Piessens, A Methodology for Designing
   Countermeasures against Current and Future Code Injection Attacks,
   Proceedings of the Third IEEE International Information Assurance
   Workshop 2005 (IWIA2005), College Park, Maryland, U.S.A., March 2005,
   IEEE, IEEE Press.
   http://www.fort-knox.org/younany_countermeasures.pdf
   
   Yves Younan, Wouter Joosen and Frank Piessens and Hans Van den
   Eynden. Security of Memory Allocators for C and C++. Technical Report
   CW419, Departement Computerwetenschappen, Katholieke Universiteit
   Leuven, July 2005. http://www.fort-knox.org/CW419.pdf
 
 */

/* Compile:
   gcc -fPIC -rdynamic -c -Wall dnmalloc-portable.c
   "Link": 
   Dynamic:
   gcc -shared -Wl,-soname,libdnmalloc.so.0 -o libdnmalloc.so.0.0 dnmalloc-portable.o -lc
   Static:
   ar -rv libdnmalloc.a dnmalloc-portable.o
   
*/

/* 
   dnmalloc is based on dlmalloc 2.7.2 (by Doug Lea (dl@cs.oswego.edu))
   dlmalloc was released as public domain and contained the following license:
   
   "This is a version (aka dlmalloc) of malloc/free/realloc written by
   Doug Lea and released to the public domain.  Use, modify, and
   redistribute this code without permission or acknowledgement in any
   way you wish.  Send questions, comments, complaints, performance
   data, etc to dl@cs.oswego.edu
   
   * VERSION 2.7.2 Sat Aug 17 09:07:30 2002  Doug Lea  (dl at gee)
   
   Note: There may be an updated version of this malloc obtainable at
   ftp://gee.cs.oswego.edu/pub/misc/malloc.c
   Check before installing!"
   
*/

/* The following preprocessor macros are tested, 
 *   and hence should have #define directives:
 *
 *   HAVE_CONFIG_H    Define to #include "config.h" (autoconf-generated)
 *
 *   HAVE_UNISTD_H    Define to #include <unistd.h>
 *
 *   HAVE_SYS_UIO_H   Define to #include <sys/uio.h> (for writev)
 *   HAVE_WRITEV      Define if the 'writev' function is available
 *
 *   HAVE_SYS_PARAM_H Define to #include <sys/param.h> (for pagesize)
 *
 *   HAVE_MALLOC_H    Define to #include <malloc.h> (for struct mallinfo)
 *
 *   HAVE_FCNTL_H     Define to #include <fcntl.h>
 *
 *   HAVE_SYS_MMAN_H  Define to #include <sys/mman.h>
 *   HAVE_MMAP        Define if the 'mmap' function is available.
 *
 *   HAVE_SCHED_H     Define to #include <sched.h>
 *   HAVE_SCHED_YIELD Define id the 'sched_yield' function is available
 */


/*
  __STD_C should be nonzero if using ANSI-standard C compiler, a C++
  compiler, or a C compiler sufficiently close to ANSI to get away
  with it.
*/

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#ifdef USE_VALGRIND
#include <valgrind/memcheck.h>
#else
#define VALGRIND_FREELIKE_BLOCK(a,b)       ((void)0)
#define VALGRIND_MALLOCLIKE_BLOCK(a,b,c,d) ((void)0)
#define VALGRIND_CREATE_MEMPOOL(a,b,c)     ((void)0)
#define VALGRIND_MEMPOOL_ALLOC(a,b,c)      ((void)0)
#define VALGRIND_MEMPOOL_FREE(a,b)         ((void)0)
#define VALGRIND_DESTROY_MEMPOOL(a)        ((void)0)
#define VALGRIND_MAKE_MEM_DEFINED(a,b)     ((void)0)
#define VALGRIND_MAKE_MEM_UNDEFINED(a,b)   ((void)0)
#define VALGRIND_MAKE_MEM_NOACCESS(a,b)    ((void)0)
#endif

#if defined (__GNUC__) && __GNUC__ > 2
# define LIKELY(expression) (__builtin_expect(!!(expression), 1))
# define UNLIKELY(expression) (__builtin_expect(!!(expression), 0))
# define __attribute_malloc__ __attribute__ ((__malloc__))
#else
# define LIKELY(x)       (x)
# define UNLIKELY(x)     (x)
# define __attribute_malloc__ /* Ignore */
#endif

/*
  Define HAVE_MREMAP to make realloc() use mremap() to re-allocate
  large blocks.  This is currently only possible on Linux with
  kernel versions newer than 1.3.77.
*/

#ifndef HAVE_MREMAP
#ifdef linux
#define HAVE_MREMAP 1
#define _GNU_SOURCE 1
#else
#define HAVE_MREMAP 0
#endif
#endif /* HAVE_MREMAP */



#ifndef __STD_C
#if defined(__STDC__) || defined(_cplusplus)
#define __STD_C     1
#else
#define __STD_C     0
#endif 
#endif /*__STD_C*/


/*
  Void_t* is the pointer type that malloc should say it returns
*/

#ifndef Void_t
#if (__STD_C || defined(WIN32))
#define Void_t      void
#else
#define Void_t      char
#endif
#endif /*Void_t*/

#if __STD_C
#include <stddef.h>   /* for size_t */
#else
#include <sys/types.h>
#endif

#if !defined(USE_SYSTEM_MALLOC)

#ifdef __cplusplus
extern "C" {
#endif

/* define HAVE_UNISTD_H if your system has a <unistd.h>. */

#ifdef HAVE_UNISTD_H
#include <unistd.h>
#endif

#ifdef HAVE_SYS_UIO_H
#include <sys/uio.h>
#endif

#include <stdio.h>    /* needed for malloc_stats */
#include <errno.h>    /* needed for optional MALLOC_FAILURE_ACTION */

#include <string.h>
#include <stdlib.h>

#include <sys/resource.h>

extern int errno;

  /* 0: lazy, 
   * 1: medium (assertions compiled in), 
   * 2: high (guard pages at end of hash table and ciregions) 
   * 3: paranoid (guards at end of each allocated chunk, check at free) 
   */
#ifndef PARANOIA
#define PARANOIA 9
#endif

  /* Using assert() with multithreading will cause the code 
   * to deadlock since glibc __assert_fail will call malloc().
   * We need our very own assert().
   */
typedef void assert_handler_tp(const char * error, const char *file, int line);

#if  PARANOIA > 0

#ifdef NDEBUG
#undef NDEBUG
#endif

static void default_assert_handler(const char *error, 
                                   const char *file, int line)
{
#ifdef HAVE_WRITEV
  struct iovec iov[5];
  char * i1 = "assertion failed (";
  char * i3 = "): ";
  char * i5 = "\n";
  int    res = 0;
  char   ifile[128];
  char   ierr[128];

  strncpy(ifile, file, sizeof(ifile)); ifile[sizeof(ifile)-1] = '\0';
  strncpy(ierr, error, sizeof(ierr));  ierr[sizeof(ierr)-1]   = '\0';

  iov[0].iov_base = i1;     iov[0].iov_len = strlen(i1); 
  iov[1].iov_base = ifile;  iov[1].iov_len = strlen(ifile); 
  iov[2].iov_base = i3;     iov[2].iov_len = strlen(i3); 
  iov[3].iov_base = ierr;   iov[3].iov_len = strlen(ierr); 
  iov[4].iov_base = i5;     iov[4].iov_len = strlen(i5);
  do {
    res = writev(STDERR_FILENO, iov, 5);
  } while (res < 0 && errno == EINTR);  
#else
  fputs("assertion failed (", stderr);
  fputs(file, stderr);
  fputs("): ", stderr);
  fputs(error, stderr);
  fputc('\n', stderr);
#endif
  (void) line;
  abort();
}
static assert_handler_tp *assert_handler = default_assert_handler;


#define assert(x)                               \
  do {                                          \
    if (UNLIKELY(!(x))) {                       \
      assert_handler(#x, __FILE__, __LINE__);   \
    }                                           \
  } while (0)

#else

static assert_handler_tp *assert_handler = NULL;
#ifndef NDEBUG
#define NDEBUG
#endif
#define assert(x) ((void)0)

#endif

assert_handler_tp *dnmalloc_set_handler(assert_handler_tp *new)
{
  assert_handler_tp *old = assert_handler;
  assert_handler = new;
  return old;
}


#include <stdarg.h>

  /* define for debugging */
  /* #define DNMALLOC_DEBUG */ 

  /* Do some extra checks? if not, covered by assrt()s */
  /* #define DNMALLOC_CHECKS */

  /*
    The unsigned integer type used for comparing any two chunk sizes.
    This should be at least as wide as size_t, but should not be signed.
  */

#ifndef CHUNK_SIZE_T
#define CHUNK_SIZE_T unsigned long
#endif

/* 
  The unsigned integer type used to hold addresses when they are are
  manipulated as integers. Except that it is not defined on all
  systems, intptr_t would suffice.
*/
#ifndef PTR_UINT
#define PTR_UINT unsigned long
#endif


/*
  INTERNAL_SIZE_T is the word-size used for internal bookkeeping
  of chunk sizes.

  The default version is the same as size_t.

  While not strictly necessary, it is best to define this as an
  unsigned type, even if size_t is a signed type. This may avoid some
  artificial size limitations on some systems.

  On a 64-bit machine, you may be able to reduce malloc overhead by
  defining INTERNAL_SIZE_T to be a 32 bit `unsigned int' at the
  expense of not being able to handle more than 2^32 of malloced
  space. If this limitation is acceptable, you are encouraged to set
  this unless you are on a platform requiring 16byte alignments. In
  this case the alignment requirements turn out to negate any
  potential advantages of decreasing size_t word size.

  Implementors: Beware of the possible combinations of:
     - INTERNAL_SIZE_T might be signed or unsigned, might be 32 or 64 bits,
       and might be the same width as int or as long
     - size_t might have different width and signedness as INTERNAL_SIZE_T
     - int and long might be 32 or 64 bits, and might be the same width
  To deal with this, most comparisons and difference computations
  among INTERNAL_SIZE_Ts should cast them to CHUNK_SIZE_T, being
  aware of the fact that casting an unsigned int to a wider long does
  not sign-extend. (This also makes checking for negative numbers
  awkward.) Some of these casts result in harmless compiler warnings
  on some systems.
*/

#ifndef INTERNAL_SIZE_T
#define INTERNAL_SIZE_T size_t
#endif

/* The corresponding word size */
#define SIZE_SZ                (sizeof(INTERNAL_SIZE_T))



/*
  MALLOC_ALIGNMENT is the minimum alignment for malloc'ed chunks.
  It must be a power of two at least 2 * SIZE_SZ, even on machines
  for which smaller alignments would suffice. It may be defined as
  larger than this though. Note however that code and data structures
  are optimized for the case of 8-byte alignment.
*/


#ifndef MALLOC_ALIGNMENT
#define MALLOC_ALIGNMENT       (2 * SIZE_SZ)
#endif

/* The corresponding bit mask value */
#define MALLOC_ALIGN_MASK      (MALLOC_ALIGNMENT - 1)



/*
  REALLOC_ZERO_BYTES_FREES should be set if a call to
  realloc with zero bytes should be the same as a call to free.
  Some people think it should. Otherwise, since this malloc
  returns a unique pointer for malloc(0), so does realloc(p, 0).
*/

#define REALLOC_ZERO_BYTES_FREES

/*
  TRIM_FASTBINS controls whether free() of a very small chunk can
  immediately lead to trimming. Setting to true (1) can reduce memory
  footprint, but will almost always slow down programs that use a lot
  of small chunks.

  Define this only if you are willing to give up some speed to more
  aggressively reduce system-level memory footprint when releasing
  memory in programs that use many small chunks.  You can get
  essentially the same effect by setting MXFAST to 0, but this can
  lead to even greater slowdowns in programs using many small chunks.
  TRIM_FASTBINS is an in-between compile-time option, that disables
  only those chunks bordering topmost memory from being placed in
  fastbins.
*/

#ifndef TRIM_FASTBINS
#define TRIM_FASTBINS  0
#endif


/*
  USE_DL_PREFIX will prefix all public routines with the string 'dl'.
  This is necessary when you only want to use this malloc in one part 
  of a program, using your regular system malloc elsewhere.
*/

/* #define USE_DL_PREFIX */


/*
  USE_MALLOC_LOCK causes wrapper functions to surround each
  callable routine with pthread mutex lock/unlock.

  USE_MALLOC_LOCK forces USE_PUBLIC_MALLOC_WRAPPERS to be defined
*/

/* #define USE_MALLOC_LOCK */


/*
  If USE_PUBLIC_MALLOC_WRAPPERS is defined, every public routine is
  actually a wrapper function that first calls MALLOC_PREACTION, then
  calls the internal routine, and follows it with
  MALLOC_POSTACTION. This is needed for locking, but you can also use
  this, without USE_MALLOC_LOCK, for purposes of interception,
  instrumentation, etc. It is a sad fact that using wrappers often
  noticeably degrades performance of malloc-intensive programs.
*/


#ifdef USE_MALLOC_LOCK
#define USE_PUBLIC_MALLOC_WRAPPERS
#else
/* #define USE_PUBLIC_MALLOC_WRAPPERS */
#endif


/* 
   Two-phase name translation.
   All of the actual routines are given mangled names.
   When wrappers are used, they become the public callable versions.
   When DL_PREFIX is used, the callable names are prefixed.
*/

#ifndef USE_PUBLIC_MALLOC_WRAPPERS
#define cALLOc      public_cALLOc
#define fREe        public_fREe
#define mALLOc      public_mALLOc
#define mEMALIGn    public_mEMALIGn
#define posix_mEMALIGn    public_posix_mEMALIGn
#define rEALLOc     public_rEALLOc
#define vALLOc      public_vALLOc
#define pVALLOc     public_pVALLOc
#define mALLINFo    public_mALLINFo
#define mALLOPt     public_mALLOPt
#define mTRIm       public_mTRIm
#define mSTATs      public_mSTATs
#define mUSABLe     public_mUSABLe
#endif

#ifdef USE_DL_PREFIX
#define public_cALLOc    dlcalloc
#define public_fREe      dlfree
#define public_mALLOc    dlmalloc
#define public_mEMALIGn  dlmemalign
#define public_posix_mEMALIGn  dlposix_memalign
#define public_rEALLOc   dlrealloc
#define public_vALLOc    dlvalloc
#define public_pVALLOc   dlpvalloc
#define public_mALLINFo  dlmallinfo
#define public_mALLOPt   dlmallopt
#define public_mTRIm     dlmalloc_trim
#define public_mSTATs    dlmalloc_stats
#define public_mUSABLe   dlmalloc_usable_size
#else /* USE_DL_PREFIX */
#define public_cALLOc    calloc
#define public_fREe      free
#define public_mALLOc    malloc
#define public_mEMALIGn  memalign
#define public_posix_mEMALIGn  posix_memalign
#define public_rEALLOc   realloc
#define public_vALLOc    valloc
#define public_pVALLOc   pvalloc
#define public_mALLINFo  mallinfo
#define public_mALLOPt   mallopt
#define public_mTRIm     malloc_trim
#define public_mSTATs    malloc_stats
#define public_mUSABLe   malloc_usable_size
#endif /* USE_DL_PREFIX */


/*
  HAVE_MEMCPY should be defined if you are not otherwise using
  ANSI STD C, but still have memcpy and memset in your C library
  and want to use them in calloc and realloc. Otherwise simple
  macro versions are defined below.

  USE_MEMCPY should be defined as 1 if you actually want to
  have memset and memcpy called. People report that the macro
  versions are faster than libc versions on some systems.
  
  Even if USE_MEMCPY is set to 1, loops to copy/clear small chunks
  (of <= 36 bytes) are manually unrolled in realloc and calloc.
*/

#ifndef HAVE_MEMCPY
#define HAVE_MEMCPY
#endif

#ifndef USE_MEMCPY
#ifdef HAVE_MEMCPY
#define USE_MEMCPY 1
#else
#define USE_MEMCPY 0
#endif
#endif


#if (__STD_C || defined(HAVE_MEMCPY))

#ifdef WIN32
  /* On Win32 memset and memcpy are already declared in windows.h */
#else
#if __STD_C
  /* Defined in string.h */
#else
Void_t* memset();
Void_t* memcpy();
#endif
#endif
#endif

/*
  MALLOC_FAILURE_ACTION is the action to take before "return 0" when
  malloc fails to be able to return memory, either because memory is
  exhausted or because of illegal arguments.
  
  By default, sets errno if running on STD_C platform, else does nothing.  
*/

#ifndef MALLOC_FAILURE_ACTION
#if __STD_C
#define MALLOC_FAILURE_ACTION \
   errno = ENOMEM;

#else
#define MALLOC_FAILURE_ACTION
#endif
#endif

/*
  MORECORE-related declarations. By default, rely on sbrk
*/


#if !defined(HAVE_UNISTD_H)
#if !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__NetBSD__)
#if __STD_C
extern Void_t*     sbrk(ptrdiff_t);
#else
extern Void_t*     sbrk();
#endif
#endif
#endif

/*
  MORECORE_FAILURE is the value returned upon failure of MORECORE
  as well as mmap. Since it cannot be an otherwise valid memory address,
  and must reflect values of standard sys calls, you probably ought not
  try to redefine it.
*/

#ifndef MORECORE_FAILURE
#define MORECORE_FAILURE ((void*)(-1UL))
#endif

/*
  MORECORE is the name of the routine to call to obtain more memory
  from the system.  See below for general guidance on writing
  alternative MORECORE functions, as well as a version for WIN32 and a
  sample version for pre-OSX macos.
*/

#ifndef MORECORE
#define MORECORE sbrk
#endif


/*
  If MORECORE_CONTIGUOUS is true, take advantage of fact that
  consecutive calls to MORECORE with positive arguments always return
  contiguous increasing addresses.  This is true of unix sbrk.  Even
  if not defined, when regions happen to be contiguous, malloc will
  permit allocations spanning regions obtained from different
  calls. But defining this when applicable enables some stronger
  consistency checks and space efficiencies. 
*/

#ifndef MORECORE_CONTIGUOUS
#define MORECORE_CONTIGUOUS 1
#endif

/*
  Define MORECORE_CANNOT_TRIM if your version of MORECORE
  cannot release space back to the system when given negative
  arguments. This is generally necessary only if you are using
  a hand-crafted MORECORE function that cannot handle negative arguments.
*/

/* #define MORECORE_CANNOT_TRIM */


/*
  This malloc requires mmap for heap management data. It is an error
  if mmap is not available.

  Additionally, mmap will be used to satisfy large requests.
*/

#ifndef HAVE_MMAP
#  error HAVE_MMAP not defined, has your operating system mmap?
#endif

/* 
   Standard unix mmap using /dev/zero clears memory so calloc doesn't
   need to.
*/

#ifndef MMAP_CLEARS
#define MMAP_CLEARS 1
#endif


/* 
   MMAP_AS_MORECORE_SIZE is the minimum mmap size argument to use if
   sbrk fails, and mmap is used as a backup (which is done only if
   HAVE_MMAP).  The value must be a multiple of page size.  This
   backup strategy generally applies only when systems have "holes" in
   address space, so sbrk cannot perform contiguous expansion, but
   there is still space available on system.  On systems for which
   this is known to be useful (i.e. most linux kernels), this occurs
   only when programs allocate huge amounts of memory.  Between this,
   and the fact that mmap regions tend to be limited, the size should
   be large, to avoid too many mmap calls and thus avoid running out
   of kernel resources.
*/

#ifndef MMAP_AS_MORECORE_SIZE
#define MMAP_AS_MORECORE_SIZE (1024 * 1024)
#endif


/*
  The system page size. To the extent possible, this malloc manages
  memory from the system in page-size units.  Note that this value is
  cached during initialization into a field of malloc_state. So even
  if malloc_getpagesize is a function, it is only called once.

  The following mechanics for getpagesize were adapted from bsd/gnu
  getpagesize.h. If none of the system-probes here apply, a value of
  4096 is used, which should be OK: If they don't apply, then using
  the actual value probably doesn't impact performance.
*/


#ifndef malloc_getpagesize

#  ifdef _SC_PAGESIZE         /* some SVR4 systems omit an underscore */
#    ifndef _SC_PAGE_SIZE
#      define _SC_PAGE_SIZE _SC_PAGESIZE
#    endif
#  endif

#  ifdef _SC_PAGE_SIZE
#    define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
#  else
#    if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
       extern size_t getpagesize();
#      define malloc_getpagesize getpagesize()
#    else
#      ifdef WIN32 /* use supplied emulation of getpagesize */
#        define malloc_getpagesize getpagesize() 
#      else
#        if defined(HAVE_SYS_PARAM_H)
#          include <sys/param.h>
#        endif
#        ifdef EXEC_PAGESIZE
#          define malloc_getpagesize EXEC_PAGESIZE
#        else
#          ifdef NBPG
#            ifndef CLSIZE
#              define malloc_getpagesize NBPG
#            else
#              define malloc_getpagesize (NBPG * CLSIZE)
#            endif
#          else
#            ifdef NBPC
#              define malloc_getpagesize NBPC
#            else
#              ifdef PAGESIZE
#                define malloc_getpagesize PAGESIZE
#              else /* just guess */
#                define malloc_getpagesize (4096) 
#              endif
#            endif
#          endif
#        endif
#      endif
#    endif
#  endif
#endif

/*
  This version of malloc supports the standard SVID/XPG mallinfo
  routine that returns a struct containing usage properties and
  statistics. It should work on any SVID/XPG compliant system that has
  a /usr/include/malloc.h defining struct mallinfo. (If you'd like to
  install such a thing yourself, cut out the preliminary declarations
  as described above and below and save them in a malloc.h file. But
  there's no compelling reason to bother to do this.)

  The main declaration needed is the mallinfo struct that is returned
  (by-copy) by mallinfo().  The SVID/XPG malloinfo struct contains a
  bunch of fields that are not even meaningful in this version of
  malloc.  These fields are are instead filled by mallinfo() with
  other numbers that might be of interest.

  HAVE_MALLOC_H should be set if you have a
  /usr/include/malloc.h file that includes a declaration of struct
  mallinfo.  If so, it is included; else an SVID2/XPG2 compliant
  version is declared below.  These must be precisely the same for
  mallinfo() to work.  The original SVID version of this struct,
  defined on most systems with mallinfo, declares all fields as
  ints. But some others define as unsigned long. If your system
  defines the fields using a type of different width than listed here,
  you must #include your system version and #define
  HAVE_MALLOC_H.
*/

/* #define HAVE_MALLOC_H */

/* On *BSD, malloc.h is deprecated, and on some *BSD including 
 * it may actually raise an error.
 */
#if defined(HAVE_MALLOC_H) && !defined(__OpenBSD__) && !defined(__FreeBSD__) && !defined(__NetBSD__) && defined(__GLIBC__)
#include <malloc.h>
#else

/* SVID2/XPG mallinfo structure */

struct mallinfo {
  int arena;    /* non-mmapped space allocated from system */
  int ordblks;  /* number of free chunks */
  int smblks;   /* number of fastbin blocks */
  int hblks;    /* number of mmapped regions */
  int hblkhd;   /* space in mmapped regions */
  int usmblks;  /* maximum total allocated space */
  int fsmblks;  /* space available in freed fastbin blocks */
  int uordblks; /* total allocated space */
  int fordblks; /* total free space */
  int keepcost; /* top-most, releasable (via malloc_trim) space */
};

/*
  SVID/XPG defines four standard parameter numbers for mallopt,
  normally defined in malloc.h.  Only one of these (M_MXFAST) is used
  in this malloc. The others (M_NLBLKS, M_GRAIN, M_KEEP) don't apply,
  so setting them has no effect. But this malloc also supports other
  options in mallopt described below.
*/
#endif


/* ---------- description of public routines ------------ */

/*
  malloc(size_t n)
  Returns a pointer to a newly allocated chunk of at least n bytes, or null
  if no space is available. Additionally, on failure, errno is
  set to ENOMEM on ANSI C systems.

  If n is zero, malloc returns a minumum-sized chunk. (The minimum
  size is 16 bytes on most 32bit systems, and 24 or 32 bytes on 64bit
  systems.)  On most systems, size_t is an unsigned type, so calls
  with negative arguments are interpreted as requests for huge amounts
  of space, which will often fail. The maximum supported value of n
  differs across systems, but is in all cases less than the maximum
  representable value of a size_t.
*/
#if __STD_C
Void_t*  public_mALLOc(size_t) __attribute_malloc__;
#else
Void_t*  public_mALLOc();
#endif

/*
  free(Void_t* p)
  Releases the chunk of memory pointed to by p, that had been previously
  allocated using malloc or a related routine such as realloc.
  It has no effect if p is null. It can have arbitrary (i.e., bad!)
  effects if p has already been freed.

  Unless disabled (using mallopt), freeing very large spaces will
  when possible, automatically trigger operations that give
  back unused memory to the system, thus reducing program footprint.
*/
#if __STD_C
void     public_fREe(Void_t*);
#else
void     public_fREe();
#endif

/*
  calloc(size_t n_elements, size_t element_size);
  Returns a pointer to n_elements * element_size bytes, with all locations
  set to zero.
*/
#if __STD_C
Void_t*  public_cALLOc(size_t, size_t) __attribute_malloc__;
#else
Void_t*  public_cALLOc();
#endif

/*
  realloc(Void_t* p, size_t n)
  Returns a pointer to a chunk of size n that contains the same data
  as does chunk p up to the minimum of (n, p's size) bytes, or null
  if no space is available. 

  The returned pointer may or may not be the same as p. The algorithm
  prefers extending p when possible, otherwise it employs the
  equivalent of a malloc-copy-free sequence.

  If p is null, realloc is equivalent to malloc.  

  If space is not available, realloc returns null, errno is set (if on
  ANSI) and p is NOT freed.

  if n is for fewer bytes than already held by p, the newly unused
  space is lopped off and freed if possible.  Unless the #define
  REALLOC_ZERO_BYTES_FREES is set, realloc with a size argument of
  zero (re)allocates a minimum-sized chunk.

  Large chunks that were internally obtained via mmap will always
  be reallocated using malloc-copy-free sequences unless
  the system supports MREMAP (currently only linux).

  The old unix realloc convention of allowing the last-free'd chunk
  to be used as an argument to realloc is not supported.
*/
#if __STD_C
Void_t*  public_rEALLOc(Void_t*, size_t) __attribute_malloc__;
#else
Void_t*  public_rEALLOc();
#endif

/*
  memalign(size_t alignment, size_t n);
  Returns a pointer to a newly allocated chunk of n bytes, aligned
  in accord with the alignment argument.

  The alignment argument should be a power of two. If the argument is
  not a power of two, the nearest greater power is used.
  8-byte alignment is guaranteed by normal malloc calls, so don't
  bother calling memalign with an argument of 8 or less.

  Overreliance on memalign is a sure way to fragment space.
*/
#if __STD_C
Void_t*  public_mEMALIGn(size_t, size_t) __attribute_malloc__;
#else
Void_t*  public_mEMALIGn();
#endif

/*
  posix_memalign(void** memptr, size_t alignment, size_t n);
  Sets *memptr to the address of a newly allocated chunk of n bytes, aligned
  in accord with the alignment argument. Returns 0 on success, otherwise
  an error (EINVAL for incorrect alignment, ENOMEM for out of memory).

  The alignment must be a power of two, and a multiple of sizeof(void *).
*/
#if __STD_C
int public_posix_mEMALIGn(Void_t**, size_t, size_t);
#else
int public_posix_mEMALIGn();
#endif

/*
  valloc(size_t n);
  Equivalent to memalign(pagesize, n), where pagesize is the page
  size of the system. If the pagesize is unknown, 4096 is used.
*/
#if __STD_C
Void_t*  public_vALLOc(size_t) __attribute_malloc__;
#else
Void_t*  public_vALLOc();
#endif



/*
  mallopt(int parameter_number, int parameter_value)
  Sets tunable parameters The format is to provide a
  (parameter-number, parameter-value) pair.  mallopt then sets the
  corresponding parameter to the argument value if it can (i.e., so
  long as the value is meaningful), and returns 1 if successful else
  0.  SVID/XPG/ANSI defines four standard param numbers for mallopt,
  normally defined in malloc.h.  Only one of these (M_MXFAST) is used
  in this malloc. The others (M_NLBLKS, M_GRAIN, M_KEEP) don't apply,
  so setting them has no effect. But this malloc also supports four
  other options in mallopt. See below for details.  Briefly, supported
  parameters are as follows (listed defaults are for "typical"
  configurations).

  Symbol            param #   default    allowed param values
  M_MXFAST          1         64         0-80  (0 disables fastbins)
  M_TRIM_THRESHOLD -1         256*1024   any   (-1U disables trimming)
  M_TOP_PAD        -2         0          any  
  M_MMAP_THRESHOLD -3         256*1024   any   (or 0 if no MMAP support)
  M_MMAP_MAX       -4         65536      any   (0 disables use of mmap)
*/
#if __STD_C
int      public_mALLOPt(int, int);
#else
int      public_mALLOPt();
#endif


/*
  mallinfo()
  Returns (by copy) a struct containing various summary statistics:

  arena:     current total non-mmapped bytes allocated from system 
  ordblks:   the number of free chunks 
  smblks:    the number of fastbin blocks (i.e., small chunks that
               have been freed but not use resused or consolidated)
  hblks:     current number of mmapped regions 
  hblkhd:    total bytes held in mmapped regions 
  usmblks:   the maximum total allocated space. This will be greater
                than current total if trimming has occurred.
  fsmblks:   total bytes held in fastbin blocks 
  uordblks:  current total allocated space (normal or mmapped)
  fordblks:  total free space 
  keepcost:  the maximum number of bytes that could ideally be released
               back to system via malloc_trim. ("ideally" means that
               it ignores page restrictions etc.)

  Because these fields are ints, but internal bookkeeping may
  be kept as longs, the reported values may wrap around zero and 
  thus be inaccurate.
*/
#if __STD_C
struct mallinfo public_mALLINFo(void);
#else
struct mallinfo public_mALLINFo();
#endif

/*
  pvalloc(size_t n);
  Equivalent to valloc(minimum-page-that-holds(n)), that is,
  round up n to nearest pagesize.
 */
#if __STD_C
Void_t*  public_pVALLOc(size_t) __attribute_malloc__;
#else
Void_t*  public_pVALLOc();
#endif

/*
  malloc_trim(size_t pad);

  If possible, gives memory back to the system (via negative
  arguments to sbrk) if there is unused memory at the `high' end of
  the malloc pool. You can call this after freeing large blocks of
  memory to potentially reduce the system-level memory requirements
  of a program. However, it cannot guarantee to reduce memory. Under
  some allocation patterns, some large free blocks of memory will be
  locked between two used chunks, so they cannot be given back to
  the system.
  
  The `pad' argument to malloc_trim represents the amount of free
  trailing space to leave untrimmed. If this argument is zero,
  only the minimum amount of memory to maintain internal data
  structures will be left (one page or less). Non-zero arguments
  can be supplied to maintain enough trailing space to service
  future expected allocations without having to re-obtain memory
  from the system.
  
  Malloc_trim returns 1 if it actually released any memory, else 0.
  On systems that do not support "negative sbrks", it will always
  rreturn 0.
*/
#if __STD_C
int      public_mTRIm(size_t);
#else
int      public_mTRIm();
#endif

/*
  malloc_usable_size(Void_t* p);

  Returns the number of bytes you can actually use in
  an allocated chunk, which may be more than you requested (although
  often not) due to alignment and minimum size constraints.
  You can use this many bytes without worrying about
  overwriting other allocated objects. This is not a particularly great
  programming practice. malloc_usable_size can be more useful in
  debugging and assertions, for example:

  p = malloc(n);
  assert(malloc_usable_size(p) >= 256);

*/
#if __STD_C
size_t   public_mUSABLe(Void_t*);
#else
size_t   public_mUSABLe();
#endif

/*
  malloc_stats();
  Prints on stderr the amount of space obtained from the system (both
  via sbrk and mmap), the maximum amount (which may be more than
  current if malloc_trim and/or munmap got called), and the current
  number of bytes allocated via malloc (or realloc, etc) but not yet
  freed. Note that this is the number of bytes allocated, not the
  number requested. It will be larger than the number requested
  because of alignment and bookkeeping overhead. Because it includes
  alignment wastage as being in use, this figure may be greater than
  zero even when no user-level chunks are allocated.

  The reported current and maximum system memory can be inaccurate if
  a program makes other calls to system memory allocation functions
  (normally sbrk) outside of malloc.

  malloc_stats prints only the most commonly interesting statistics.
  More information can be obtained by calling mallinfo.

*/
#if __STD_C
void     public_mSTATs();
#else
void     public_mSTATs();
#endif

/* mallopt tuning options */

/*
  M_MXFAST is the maximum request size used for "fastbins", special bins
  that hold returned chunks without consolidating their spaces. This
  enables future requests for chunks of the same size to be handled
  very quickly, but can increase fragmentation, and thus increase the
  overall memory footprint of a program.

  This malloc manages fastbins very conservatively yet still
  efficiently, so fragmentation is rarely a problem for values less
  than or equal to the default.  The maximum supported value of MXFAST
  is 80. You wouldn't want it any higher than this anyway.  Fastbins
  are designed especially for use with many small structs, objects or
  strings -- the default handles structs/objects/arrays with sizes up
  to 16 4byte fields, or small strings representing words, tokens,
  etc. Using fastbins for larger objects normally worsens
  fragmentation without improving speed.

  M_MXFAST is set in REQUEST size units. It is internally used in
  chunksize units, which adds padding and alignment.  You can reduce
  M_MXFAST to 0 to disable all use of fastbins.  This causes the malloc
  algorithm to be a closer approximation of fifo-best-fit in all cases,
  not just for larger requests, but will generally cause it to be
  slower.
*/


/* M_MXFAST is a standard SVID/XPG tuning option, usually listed in malloc.h */
#ifndef M_MXFAST
#define M_MXFAST            1    
#endif

#ifndef DEFAULT_MXFAST
#define DEFAULT_MXFAST     64
#endif


/*
  M_TRIM_THRESHOLD is the maximum amount of unused top-most memory
  to keep before releasing via malloc_trim in free().

  Automatic trimming is mainly useful in long-lived programs.
  Because trimming via sbrk can be slow on some systems, and can
  sometimes be wasteful (in cases where programs immediately
  afterward allocate more large chunks) the value should be high
  enough so that your overall system performance would improve by
  releasing this much memory.

  The trim threshold and the mmap control parameters (see below)
  can be traded off with one another. Trimming and mmapping are
  two different ways of releasing unused memory back to the
  system. Between these two, it is often possible to keep
  system-level demands of a long-lived program down to a bare
  minimum. For example, in one test suite of sessions measuring
  the XF86 X server on Linux, using a trim threshold of 128K and a
  mmap threshold of 192K led to near-minimal long term resource
  consumption.

  If you are using this malloc in a long-lived program, it should
  pay to experiment with these values.  As a rough guide, you
  might set to a value close to the average size of a process
  (program) running on your system.  Releasing this much memory
  would allow such a process to run in memory.  Generally, it's
  worth it to tune for trimming rather tham memory mapping when a
  program undergoes phases where several large chunks are
  allocated and released in ways that can reuse each other's
  storage, perhaps mixed with phases where there are no such
  chunks at all.  And in well-behaved long-lived programs,
  controlling release of large blocks via trimming versus mapping
  is usually faster.

  However, in most programs, these parameters serve mainly as
  protection against the system-level effects of carrying around
  massive amounts of unneeded memory. Since frequent calls to
  sbrk, mmap, and munmap otherwise degrade performance, the default
  parameters are set to relatively high values that serve only as
  safeguards.

  The trim value must be greater than page size to have any useful
  effect.  To disable trimming completely, you can set to 
  (unsigned long)(-1)

  Trim settings interact with fastbin (MXFAST) settings: Unless
  TRIM_FASTBINS is defined, automatic trimming never takes place upon
  freeing a chunk with size less than or equal to MXFAST. Trimming is
  instead delayed until subsequent freeing of larger chunks. However,
  you can still force an attempted trim by calling malloc_trim.

  Also, trimming is not generally possible in cases where
  the main arena is obtained via mmap.

  Note that the trick some people use of mallocing a huge space and
  then freeing it at program startup, in an attempt to reserve system
  memory, doesn't have the intended effect under automatic trimming,
  since that memory will immediately be returned to the system.
*/

#define M_TRIM_THRESHOLD       -1

#ifndef DEFAULT_TRIM_THRESHOLD
#define DEFAULT_TRIM_THRESHOLD (256 * 1024)
#endif

/*
  M_TOP_PAD is the amount of extra `padding' space to allocate or
  retain whenever sbrk is called. It is used in two ways internally:

  * When sbrk is called to extend the top of the arena to satisfy
  a new malloc request, this much padding is added to the sbrk
  request.

  * When malloc_trim is called automatically from free(),
  it is used as the `pad' argument.

  In both cases, the actual amount of padding is rounded
  so that the end of the arena is always a system page boundary.

  The main reason for using padding is to avoid calling sbrk so
  often. Having even a small pad greatly reduces the likelihood
  that nearly every malloc request during program start-up (or
  after trimming) will invoke sbrk, which needlessly wastes
  time.

  Automatic rounding-up to page-size units is normally sufficient
  to avoid measurable overhead, so the default is 0.  However, in
  systems where sbrk is relatively slow, it can pay to increase
  this value, at the expense of carrying around more memory than
  the program needs.
*/

#define M_TOP_PAD              -2

#ifndef DEFAULT_TOP_PAD
#define DEFAULT_TOP_PAD        (0)
#endif

/*
  M_MMAP_THRESHOLD is the request size threshold for using mmap()
  to service a request. Requests of at least this size that cannot
  be allocated using already-existing space will be serviced via mmap.
  (If enough normal freed space already exists it is used instead.)

  Using mmap segregates relatively large chunks of memory so that
  they can be individually obtained and released from the host
  system. A request serviced through mmap is never reused by any
  other request (at least not directly; the system may just so
  happen to remap successive requests to the same locations).

  Segregating space in this way has the benefits that:

   1. Mmapped space can ALWAYS be individually released back 
      to the system, which helps keep the system level memory 
      demands of a long-lived program low. 
   2. Mapped memory can never become `locked' between
      other chunks, as can happen with normally allocated chunks, which
      means that even trimming via malloc_trim would not release them.
   3. On some systems with "holes" in address spaces, mmap can obtain
      memory that sbrk cannot.

  However, it has the disadvantages that:

   1. The space cannot be reclaimed, consolidated, and then
      used to service later requests, as happens with normal chunks.
   2. It can lead to more wastage because of mmap page alignment
      requirements
   3. It causes malloc performance to be more dependent on host
      system memory management support routines which may vary in
      implementation quality and may impose arbitrary
      limitations. Generally, servicing a request via normal
      malloc steps is faster than going through a system's mmap.

  The advantages of mmap nearly always outweigh disadvantages for
  "large" chunks, but the value of "large" varies across systems.  The
  default is an empirically derived value that works well in most
  systems.
*/

#define M_MMAP_THRESHOLD      -3

#ifndef DEFAULT_MMAP_THRESHOLD
#define DEFAULT_MMAP_THRESHOLD (256 * 1024)
#endif

/*
  M_MMAP_MAX is the maximum number of requests to simultaneously
  service using mmap. This parameter exists because
. Some systems have a limited number of internal tables for
  use by mmap, and using more than a few of them may degrade
  performance.

  The default is set to a value that serves only as a safeguard.
  Setting to 0 disables use of mmap for servicing large requests.  If
  HAVE_MMAP is not set, the default value is 0, and attempts to set it
  to non-zero values in mallopt will fail.
*/

#define M_MMAP_MAX             -4

#ifndef DEFAULT_MMAP_MAX
#define DEFAULT_MMAP_MAX       (65536)
#endif

#ifdef __cplusplus
};  /* end of extern "C" */
#endif

/* 
  ========================================================================
  To make a fully customizable malloc.h header file, cut everything
  above this line, put into file malloc.h, edit to suit, and #include it 
  on the next line, as well as in programs that use this malloc.
  ========================================================================
*/

/* #include "malloc.h" */

/* --------------------- public wrappers ---------------------- */

#ifdef USE_PUBLIC_MALLOC_WRAPPERS

/* DL_STATIC used to make functions (deep down) consistent
 * with prototypes (otherwise the prototypes are static
 * with USE_PUBLIC_MALLOC_WRAPPERS, but the functions aren't).
 * The gcc compiler doesn't care, but the HP-UX compiler does.
 */
#define DL_STATIC static

/* Declare all routines as internal */
#if __STD_C
static Void_t*  mALLOc(size_t) __attribute_malloc__;
static void     fREe(Void_t*);
static Void_t*  rEALLOc(Void_t*, size_t) __attribute_malloc__;
static Void_t*  mEMALIGn(size_t, size_t) __attribute_malloc__;
static int      posix_mEMALIGn(Void_t**, size_t, size_t);
static Void_t*  vALLOc(size_t) __attribute_malloc__;
static Void_t*  pVALLOc(size_t) __attribute_malloc__;
static Void_t*  cALLOc(size_t, size_t) __attribute_malloc__;
static int      mTRIm(size_t);
static size_t   mUSABLe(Void_t*);
static void     mSTATs();
static int      mALLOPt(int, int);
static struct mallinfo mALLINFo(void);
#else
static Void_t*  mALLOc();
static void     fREe();
static Void_t*  rEALLOc();
static Void_t*  mEMALIGn();
static int      posix_mEMALIGn();
static Void_t*  vALLOc();
static Void_t*  pVALLOc();
static Void_t*  cALLOc();
static int      mTRIm();
static size_t   mUSABLe();
static void     mSTATs();
static int      mALLOPt();
static struct mallinfo mALLINFo();
#endif

/*
  MALLOC_PREACTION and MALLOC_POSTACTION should be
  defined to return 0 on success, and nonzero on failure.
  The return value of MALLOC_POSTACTION is currently ignored
  in wrapper functions since there is no reasonable default
  action to take on failure.
*/


#ifdef USE_MALLOC_LOCK

# ifdef WIN32

static int mALLOC_MUTEx;
#define MALLOC_PREACTION   slwait(&mALLOC_MUTEx)
#define MALLOC_POSTACTION  slrelease(&mALLOC_MUTEx)
int dnmalloc_pthread_init(void) { return 0; }

# elif defined(__NetBSD__) || defined(__OpenBSD__) || defined(__FreeBSD__)

#   if defined(__NetBSD__)
#include <reentrant.h>
extern int __isthreaded;
static mutex_t thread_lock = MUTEX_INITIALIZER;
#define _MALLOC_LOCK()   if (__isthreaded) mutex_lock(&thread_lock)
#define _MALLOC_UNLOCK() if (__isthreaded) mutex_unlock(&thread_lock)
void _malloc_prefork(void) {  _MALLOC_LOCK(); }
void _malloc_postfork(void) { _MALLOC_UNLOCK(); }
#   endif

#   if defined(__OpenBSD__)
extern int  __isthreaded;
void   _thread_malloc_lock(void);
void   _thread_malloc_unlock(void);
#define _MALLOC_LOCK()           if (__isthreaded) _thread_malloc_lock()
#define _MALLOC_UNLOCK()         if (__isthreaded) _thread_malloc_unlock()
#   endif

#   if defined(__FreeBSD__)
extern int      __isthreaded;
struct _spinlock {
        volatile long   access_lock;
        volatile long   lock_owner;
        volatile char   *fname;
        volatile int    lineno;
};
typedef struct _spinlock spinlock_t;
#define _SPINLOCK_INITIALIZER   { 0, 0, 0, 0 }
void    _spinlock(spinlock_t *);
void    _spinunlock(spinlock_t *);
/* # include "/usr/src/lib/libc/include/spinlock.h" */
static spinlock_t thread_lock   = _SPINLOCK_INITIALIZER;
spinlock_t *__malloc_lock       = &thread_lock;
#define _MALLOC_LOCK()           if (__isthreaded) _spinlock(&thread_lock)
#define _MALLOC_UNLOCK()         if (__isthreaded) _spinunlock(&thread_lock)
#   endif

/* Common for all three *BSD
 */
static int malloc_active = 0;
static int dnmalloc_mutex_lock()
{
  _MALLOC_LOCK();
  if (!malloc_active)
    {
      ++malloc_active;
      return 0;
    }
  assert(malloc_active == 0);
  _MALLOC_UNLOCK();
  errno = EDEADLK;
  return 1;
}
static int dnmalloc_mutex_unlock()
{
  --malloc_active;
  _MALLOC_UNLOCK();
  return 0;
}
#define MALLOC_PREACTION   dnmalloc_mutex_lock()
#define MALLOC_POSTACTION  dnmalloc_mutex_unlock()
int dnmalloc_pthread_init(void) { return 0; }

# else

/* Wrapping malloc with pthread_mutex_lock/pthread_mutex_unlock
 *
 * Works fine on linux (no malloc in pthread_mutex_lock)
 * Works with on HP-UX if initialized after entering main()
 */ 
#include <pthread.h>
static int malloc_active      = 0;
void dnmalloc_fork_prepare(void);
void dnmalloc_fork_parent(void);
void dnmalloc_fork_child(void);

#if !defined(__linux__)

static pthread_mutex_t mALLOC_MUTEx;
pthread_once_t dnmalloc_once_control = PTHREAD_ONCE_INIT;
static int dnmalloc_use_mutex = 0;
static void dnmalloc_pthread_init_int(void)
{
  pthread_mutexattr_t   mta;
  pthread_mutexattr_init(&mta);
  pthread_mutexattr_settype(&mta, PTHREAD_MUTEX_RECURSIVE);
  pthread_mutex_init(&(mALLOC_MUTEx), &mta);
  pthread_mutexattr_destroy(&mta);       
  pthread_atfork(dnmalloc_fork_prepare, 
                 dnmalloc_fork_parent,
                 dnmalloc_fork_child);
  dnmalloc_use_mutex = 1;
}
int dnmalloc_pthread_init(void)
{
  return pthread_once(&dnmalloc_once_control, dnmalloc_pthread_init_int);
}

#else

static pthread_mutex_t mALLOC_MUTEx = PTHREAD_MUTEX_INITIALIZER;
static int dnmalloc_use_mutex = 1;
int dnmalloc_pthread_init(void) { 
  return pthread_atfork(dnmalloc_fork_prepare, 
                        dnmalloc_fork_parent,
                        dnmalloc_fork_child); 
}
#endif /* !defined(__linux__) */

void dnmalloc_fork_prepare(void) { 
  if (dnmalloc_use_mutex) 
    pthread_mutex_lock(&mALLOC_MUTEx);
}
void dnmalloc_fork_parent(void) { 
  if (dnmalloc_use_mutex)
    pthread_mutex_unlock(&mALLOC_MUTEx); 
}
void dnmalloc_fork_child(void) {
  int rc = 0;
#ifdef __GLIBC__
  if (dnmalloc_use_mutex)
    {
      pthread_mutex_unlock (&mALLOC_MUTEx);
      pthread_mutex_destroy(&mALLOC_MUTEx);
      rc = pthread_mutex_init(&mALLOC_MUTEx, NULL);
    } 
#else
  if (dnmalloc_use_mutex)
    rc = pthread_mutex_unlock(&mALLOC_MUTEx); 
#endif
  if (rc != 0)
    {
      fputs("fork_child failed", stderr);
      _exit(EXIT_FAILURE);
    }
}
static int dnmalloc_mutex_lock(pthread_mutex_t *mutex)
{
  if (dnmalloc_use_mutex)
    {
      int rc = pthread_mutex_lock(mutex);
      if (rc == 0)
        {
          if (!malloc_active)
            {
              ++malloc_active;
              return 0;
            }
          assert(malloc_active == 0);
          (void) pthread_mutex_unlock(mutex);
          errno = EDEADLK;
          return 1;
        }
      return rc;
    }
  return 0;
}
static int dnmalloc_mutex_unlock(pthread_mutex_t *mutex)
{
  if (dnmalloc_use_mutex)
    {
      --malloc_active;
      return pthread_mutex_unlock(mutex);
    }
  return 0;
}
# define MALLOC_PREACTION   dnmalloc_mutex_lock(&mALLOC_MUTEx)
# define MALLOC_POSTACTION  dnmalloc_mutex_unlock(&mALLOC_MUTEx)

# endif

#else

/* Substitute anything you like for these */

# define MALLOC_PREACTION   (0)
# define MALLOC_POSTACTION  (0)
int dnmalloc_pthread_init(void) { return 0; }

#endif /* USE_MALLOC_LOCK */

Void_t* public_mALLOc(size_t bytes) {
  Void_t* m;
  if (MALLOC_PREACTION == 0) {
    m = mALLOc(bytes);
    (void) MALLOC_POSTACTION;
    return m;
  }
  return 0;
}

void public_fREe(Void_t* m) {
  if (MALLOC_PREACTION == 0) {
    fREe(m);
    (void) MALLOC_POSTACTION;
  }
}

Void_t* public_rEALLOc(Void_t* m, size_t bytes) {
  if (MALLOC_PREACTION == 0) {
    m = rEALLOc(m, bytes);
    (void) MALLOC_POSTACTION;
    return m;
  }
  return 0;
}

Void_t* public_mEMALIGn(size_t alignment, size_t bytes) {
  Void_t* m;
  if (MALLOC_PREACTION == 0) {
    m = mEMALIGn(alignment, bytes);
    (void) MALLOC_POSTACTION;
    return m;
  }
  return 0;
}

int public_posix_mEMALIGn(Void_t**memptr, size_t alignment, size_t bytes) {
  int m, ret;
  if ((ret = MALLOC_PREACTION) == 0) {
    m = posix_mEMALIGn(memptr, alignment, bytes);
    (void) MALLOC_POSTACTION;
    return m;
  }
  return ret;
}

Void_t* public_vALLOc(size_t bytes) {
  Void_t* m;
  if (MALLOC_PREACTION == 0) {
    m = vALLOc(bytes);
    (void) MALLOC_POSTACTION;
    return m;
  }
  return 0;
}

Void_t* public_pVALLOc(size_t bytes) {
  Void_t* m;
  if (MALLOC_PREACTION == 0) {
    m = pVALLOc(bytes);
    (void) MALLOC_POSTACTION;
    return m;
  }
  return 0;
}

Void_t* public_cALLOc(size_t n, size_t elem_size) {
  Void_t* m;
  if (MALLOC_PREACTION == 0) {
    m = cALLOc(n, elem_size);
    (void) MALLOC_POSTACTION;
    return m;
  }
  return 0;
}

int public_mTRIm(size_t s) {
  int result;
  if (MALLOC_PREACTION == 0) {
    result = mTRIm(s);
    (void) MALLOC_POSTACTION;
    return result;
  }
  return 0;
}

size_t public_mUSABLe(Void_t* m) {
  size_t result;
  if (MALLOC_PREACTION == 0) {
    result = mUSABLe(m);
    (void) MALLOC_POSTACTION;
    return result;
  }
  return 0;
}

void public_mSTATs() {
  if (MALLOC_PREACTION == 0) {
    mSTATs();
    (void) MALLOC_POSTACTION;
  }
}

struct mallinfo public_mALLINFo() {
  struct mallinfo m;
  if (MALLOC_PREACTION == 0) {
    m = mALLINFo();
    (void) MALLOC_POSTACTION;
    return m;
  } else {
    struct mallinfo nm = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
    return nm;
  }
}

int public_mALLOPt(int p, int v) {
  int result;
  if (MALLOC_PREACTION == 0) {
    result = mALLOPt(p, v);
    (void) MALLOC_POSTACTION;
    return result;
  }
  return 0;
}

#else

int dnmalloc_pthread_init(void) { return 0; }
#define DL_STATIC

#endif /* USE_PUBLIC_MALLOC_WRAPPERS */



/* ------------- Optional versions of memcopy ---------------- */


#if USE_MEMCPY

/* 
  Note: memcpy is ONLY invoked with non-overlapping regions,
  so the (usually slower) memmove is not needed.
*/

#define MALLOC_COPY(dest, src, nbytes)  memcpy(dest, src, nbytes)
#define MALLOC_ZERO(dest, nbytes)       memset(dest, 0,   nbytes)

#else /* !USE_MEMCPY */

/* Use Duff's device for good zeroing/copying performance. */

#define MALLOC_ZERO(charp, nbytes)                                            \
do {                                                                          \
  INTERNAL_SIZE_T* mzp = (INTERNAL_SIZE_T*)(charp);                           \
  CHUNK_SIZE_T  mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T);                     \
  long mcn;                                                                   \
  if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; }             \
  switch (mctmp) {                                                            \
    case 0: for(;;) { *mzp++ = 0;                                             \
    case 7:           *mzp++ = 0;                                             \
    case 6:           *mzp++ = 0;                                             \
    case 5:           *mzp++ = 0;                                             \
    case 4:           *mzp++ = 0;                                             \
    case 3:           *mzp++ = 0;                                             \
    case 2:           *mzp++ = 0;                                             \
    case 1:           *mzp++ = 0; if(mcn <= 0) break; mcn--; }                \
  }                                                                           \
} while(0)

#define MALLOC_COPY(dest,src,nbytes)                                          \
do {                                                                          \
  INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) src;                            \
  INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) dest;                           \
  CHUNK_SIZE_T  mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T);                     \
  long mcn;                                                                   \
  if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; }             \
  switch (mctmp) {                                                            \
    case 0: for(;;) { *mcdst++ = *mcsrc++;                                    \
    case 7:           *mcdst++ = *mcsrc++;                                    \
    case 6:           *mcdst++ = *mcsrc++;                                    \
    case 5:           *mcdst++ = *mcsrc++;                                    \
    case 4:           *mcdst++ = *mcsrc++;                                    \
    case 3:           *mcdst++ = *mcsrc++;                                    \
    case 2:           *mcdst++ = *mcsrc++;                                    \
    case 1:           *mcdst++ = *mcsrc++; if(mcn <= 0) break; mcn--; }       \
  }                                                                           \
} while(0)

#endif

/* ------------------ MMAP support ------------------  */


#if defined(HAVE_FCNTL_H)
#include <fcntl.h>
#endif

#if defined(HAVE_SYS_MMAN_H)
#include <sys/mman.h>
#endif

#if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
#define MAP_ANONYMOUS MAP_ANON
#endif

/* 
   Nearly all versions of mmap support MAP_ANONYMOUS, 
   so the following is unlikely to be needed, but is
   supplied just in case.
*/

#ifndef MAP_ANONYMOUS

/* rw 19.05.2008 changed to avoid cached file descriptor, untested 
 */
void * anon_mmap (void *addr, size_t length, int prot, int flags)
{
  void * retval   = NULL;
  int dev_zero_fd = -1; /* File descriptor for /dev/zero. */

  dev_zero_fd = open("/dev/zero", O_RDWR);
  if (dev_zero_fd >= 0)
    {
      retval = mmap((addr), (size), (prot), (flags), dev_zero_fd, 0);
      /* closing the file descriptor does not unmap the region */
      close(dev_zero_fd); 
    }
  return retval;
}
  
#define MMAP(addr, size, prot, flags) \
  (anon_mmap((addr), (size), (prot), (flags)))


#else /* have MAP_ANONYMOUS */

#if !defined(MAP_32BIT) && defined(MAP_ADDR32)
#define MAP_32BIT MAP_ADDR32
#endif

#if defined(MAP_32BIT)
#define MMAP(addr, size, prot, flags) \
 (mmap((addr), (size), (prot), (flags)|MAP_ANONYMOUS|MAP_32BIT, -1, 0))
#elif defined(__sun)
/* 
 * Hint an address within 32bit address space
 */
#define MMAP(addr, size, prot, flags) \
 (mmap((void*)0xC0000000, (size), (prot), (flags)|MAP_ANONYMOUS, -1, 0))
#else
/* *BSD */
#define MMAP(addr, size, prot, flags) \
 (mmap((void*)0x80000000, (size), (prot), (flags)|MAP_ANONYMOUS, -1, 0))
#endif

#endif /* have MAP_ANONYMOUS */


/*
  -----------------------  Chunk representations -----------------------
*/

typedef void * mchunkptr;

struct chunkinfo {
   INTERNAL_SIZE_T      prev_size;  /* Size of previous in bytes          */
   INTERNAL_SIZE_T      size;       /* Size in bytes, including overhead. */
   INTERNAL_SIZE_T      req;        /* Original request size, for guard.  */
   struct chunkinfo* hash_next;     /* contains a pointer to the next chunk 
                                       in the linked list if the hash
                                       value is the same as the chunk     */
   struct chunkinfo* fd;            /* double links -- used only if free. */
   struct chunkinfo* bk;
   mchunkptr chunk;
};

typedef struct chunkinfo* chunkinfoptr;

struct cireginfo {
        unsigned long position;
        unsigned long *freebitmap;
        struct cireginfo* next;
        struct chunkinfo *freelist;
        struct chunkinfo *begin; 
        unsigned long freecounter; 
};

/*
  ---------- Size and alignment checks and conversions ----------
*/

/* conversion from malloc headers to user pointers, and back */
#define chunk(p) (p->chunk)


#define chunk2mem(p)   (chunk(p))
#define mem2chunk(mem) (hashtable_lookup(mem))

/* The smallest possible chunk      */
/* #define MIN_CHUNK_SIZE        16 */
#if (SIZEOF_UNSIGNED_LONG == 8) || defined(__arch64__) || defined(__ia64__) || defined(__x86_64__) || defined(__LP64__) || defined(__64BIT__) || defined(_LP64) || defined(_M_IA64) || (defined(_MIPS_SZLONG) && (_MIPS_SZLONG == 64))
#  define MIN_CHUNK_SIZE 32
#else
#  define MIN_CHUNK_SIZE 16
#endif

/* The smallest size we can malloc is an aligned minimal chunk */

#define MINSIZE  \
  (CHUNK_SIZE_T)(((MIN_CHUNK_SIZE+MALLOC_ALIGN_MASK) & ~MALLOC_ALIGN_MASK))

/* Check if m has acceptable alignment */

#define aligned_OK(m)  (((PTR_UINT)((m)) & (MALLOC_ALIGN_MASK)) == 0)

#define GUARD_SIZE 4

/* 
   Check if a request is so large that it would wrap around zero when
   padded and aligned. To simplify some other code, the bound is made
   low enough so that adding MINSIZE will also not wrap around zero.

   Make it 4*MINSIZE.
*/

#define REQUEST_OUT_OF_RANGE(req)                                 \
  ((CHUNK_SIZE_T)(req) >=                                         \
   (CHUNK_SIZE_T)(INTERNAL_SIZE_T)(-4 * MINSIZE))    

/* pad request bytes into a usable size -- internal version */

#define request2size(req)                                         \
  (((req) + GUARD_SIZE + MALLOC_ALIGN_MASK >= MINSIZE)  ?         \
   ((req) + GUARD_SIZE + MALLOC_ALIGN_MASK) & ~MALLOC_ALIGN_MASK :\
   MINSIZE)

/*  Same, except also perform argument check */

#define checked_request2size(req, sz)                             \
  if (!REQUEST_OUT_OF_RANGE(req)) {                               \
    (sz) = request2size(req);                                     \
    assert((sz-req) >= GUARD_SIZE);                               \
  } else {                                                        \
    MALLOC_FAILURE_ACTION;                                        \
    return 0;                                                     \
  }

#if PARANOIA > 2
static char * guard_set_p;
static char * guard_set_q;

#define guard_set(guard, P, request, sz)                          \
  assert((sz-request) >= GUARD_SIZE);                             \
  guard_set_p = (char*)(chunk(P));                                \
  guard_set_p += request;                                         \
  VALGRIND_MAKE_MEM_UNDEFINED(guard_set_p,GUARD_SIZE);            \
  guard_set_q = (char*)(guard);                                   \
  *guard_set_p = *guard_set_q; ++guard_set_p; ++guard_set_q;      \
  *guard_set_p = *guard_set_q; ++guard_set_p; ++guard_set_q;      \
  *guard_set_p = *guard_set_q; ++guard_set_p; ++guard_set_q;      \
  *guard_set_p = *guard_set_q;                                    \
  VALGRIND_MAKE_MEM_NOACCESS((((char*)chunk(P))+request),GUARD_SIZE);   \
  (P)->req = request
 
#define guard_check(guard, P)                                     \
  VALGRIND_MAKE_MEM_DEFINED((((char *)chunk(P))+(P)->req), GUARD_SIZE); \
  assert(0 == memcmp((((char *)chunk(P))+(P)->req),(void*)(guard),GUARD_SIZE));\
  VALGRIND_MAKE_MEM_NOACCESS((((char *)chunk(P))+(P)->req), GUARD_SIZE);

#else
#define guard_set(guard, P, request, sz) ((void)0)
#define guard_check(guard, P) ((void)0)
#endif /* PARANOIA > 2 */

/* dnmalloc forward declarations */
static char * dnmalloc_arc4random(void);
static void dnmalloc_init (void);
static void malloc_mmap_state(void);
static void cireg_extend (void);
static chunkinfoptr cireg_getfree (void);
static void hashtable_add (chunkinfoptr ci);
static void hashtable_insert (chunkinfoptr ci_orig, chunkinfoptr ci_insert);
static void hashtable_remove (mchunkptr p);              
static void hashtable_skiprm (chunkinfoptr ci_orig, chunkinfoptr ci_todelete);
static chunkinfoptr hashtable_lookup (mchunkptr p);
static chunkinfoptr next_chunkinfo (chunkinfoptr ci);          
static chunkinfoptr prev_chunkinfo (chunkinfoptr ci);



/*
  --------------- Physical chunk operations ---------------
*/


/* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */
#define PREV_INUSE 0x1

/* extract inuse bit of previous chunk */
#define prev_inuse(p)       ((p)->size & PREV_INUSE)

/* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */
#define IS_MMAPPED 0x2

/* check for mmap()'ed chunk */
#define chunk_is_mmapped(p) ((p)->size & IS_MMAPPED)


/* size field is or'ed when the chunk is in use */
#define INUSE 0x4

/* extract inuse bit of chunk */
#define inuse(p)       ((p)->size & INUSE)

/* 
  Bits to mask off when extracting size 

  Note: IS_MMAPPED is intentionally not masked off from size field in
  macros for which mmapped chunks should never be seen. This should
  cause helpful core dumps to occur if it is tried by accident by
  people extending or adapting this malloc.
*/
#define SIZE_BITS (PREV_INUSE|IS_MMAPPED|INUSE)

/* Bits to mask off when extracting size of chunks for macros which do not use mmap */
#define SIZE_NOMMAP (PREV_INUSE|INUSE)

/* Get size, ignoring use bits */
#define chunksize(p)         ((p)->size & ~(SIZE_BITS))

/* Ptr to chunkinfo of next physical malloc_chunk. */
#define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->size & SIZE_NOMMAP) ))

/* Treat space at ptr + offset as a chunk */
#define chunk_at_offset(p, s)  ((mchunkptr)(((char*)(p)) + (s)))

/* set/clear chunk as being inuse without otherwise disturbing */
#define set_inuse(p) ((p)->size |= INUSE)

#define clear_inuse(p) ((p)->size &= ~(INUSE))

#define set_previnuse(p) ((p)->size |= PREV_INUSE)

#define clear_previnuse(p) ((p)->size &= ~(PREV_INUSE))

static void set_previnuse_next (chunkinfoptr p)
{
   chunkinfoptr q;
   q = next_chunkinfo (p);
   if (q)
      set_previnuse (q);
}

#define set_all_inuse(p) \
set_inuse(p); \
set_previnuse_next(p);


/* Set size at head, without disturbing its use bit */
#define set_head_size(p, s)  ((p)->size = (((p)->size & SIZE_NOMMAP) | (s)))

/* Set size/use field */
#define set_head(p, s)       ((p)->size = (s))

/*
  Bins

    An array of bin headers for free chunks. Each bin is doubly
    linked.  The bins are approximately proportionally (log) spaced.
    There are a lot of these bins (128). This may look excessive, but
    works very well in practice.  Most bins hold sizes that are
    unusual as malloc request sizes, but are more usual for fragments
    and consolidated sets of chunks, which is what these bins hold, so
    they can be found quickly.  All procedures maintain the invariant
    that no consolidated chunk physically borders another one, so each
    chunk in a list is known to be preceeded and followed by either
    inuse chunks or the ends of memory.

    Chunks in bins are kept in size order, with ties going to the
    approximately least recently used chunk. Ordering isn't needed
    for the small bins, which all contain the same-sized chunks, but
    facilitates best-fit allocation for larger chunks. These lists
    are just sequential. Keeping them in order almost never requires
    enough traversal to warrant using fancier ordered data
    structures.  

    Chunks of the same size are linked with the most
    recently freed at the front, and allocations are taken from the
    back.  This results in LRU (FIFO) allocation order, which tends
    to give each chunk an equal opportunity to be consolidated with
    adjacent freed chunks, resulting in larger free chunks and less
    fragmentation.

    To simplify use in double-linked lists, each bin header acts
    as a malloc_chunk. This avoids special-casing for headers.
    But to conserve space and improve locality, we allocate
    only the fd/bk pointers of bins, and then use repositioning tricks
    to treat these as the fields of a malloc_chunk*.  
*/

typedef struct chunkinfo* mbinptr;

/* addressing -- note that bin_at(0) does not exist */
#define bin_at(m, i) (&(m)->bins[i])

/* analog of ++bin */
#define next_bin(b)  (b+1)

/* Reminders about list directionality within bins */
#define first(b)     ((b)->fd)
#define last(b)      ((b)->bk)

/* Take a chunk off a bin list */
#define unlink(P, BK, FD) {                                            \
  FD = P->fd;                                                          \
  BK = P->bk;                                                          \
  FD->bk = BK;                                                         \
  BK->fd = FD;                                                         \
}

/*
  Indexing

    Bins for sizes < 512 bytes contain chunks of all the same size, spaced
    8 bytes apart. Larger bins are approximately logarithmically spaced:

    64 bins of size       8
    32 bins of size      64
    16 bins of size     512
     8 bins of size    4096
     4 bins of size   32768
     2 bins of size  262144
     1 bin  of size what's left

    The bins top out around 1MB because we expect to service large
    requests via mmap.
*/

#define NBINS              96
#define NSMALLBINS         32
#define SMALLBIN_WIDTH      8
#define MIN_LARGE_SIZE    256

#define in_smallbin_range(sz)  \
  ((CHUNK_SIZE_T)(sz) < (CHUNK_SIZE_T)MIN_LARGE_SIZE)

#define smallbin_index(sz)     (((unsigned)(sz)) >> 3)

/*
  Compute index for size. We expect this to be inlined when
  compiled with optimization, else not, which works out well.
*/
static int largebin_index(size_t sz) {

  unsigned long  xx = sz >> SMALLBIN_WIDTH; 

  if (xx < 0x10000) 
    {
      unsigned int  m;           /* bit position of highest set bit of m */
      
      /* On intel, use BSRL instruction to find highest bit */
#if defined(__GNUC__) && defined(i386) && !defined(USE_UNO)

      unsigned int  x = (unsigned int) xx;

      __asm__("bsrl %1,%0\n\t"
              : "=r" (m) 
              : "rm"  (x));

#elif defined(__GNUC__) && defined(x86_64) && !defined(USE_UNO)

      __asm__("bsrq %1,%0\n\t"
              : "=r" (m)
              : "rm"  (xx));

#else

      /* Taken from Bit Twiddling Hacks
       * http://graphics.stanford.edu/~seander/bithacks.html
       * public domain
       */
      unsigned int  v  = (unsigned int) xx;
      register unsigned int shift;
      
      m =     (v > 0xFFFF) << 4; v >>= m;
      shift = (v > 0xFF  ) << 3; v >>= shift; m |= shift;
      shift = (v > 0xF   ) << 2; v >>= shift; m |= shift;
      shift = (v > 0x3   ) << 1; v >>= shift; m |= shift;
      m |= (v >> 1);
      
#endif
      
      /* Use next 2 bits to create finer-granularity bins */
      return NSMALLBINS + (m << 2) + ((sz >> (m + 6)) & 3);
    }
  else
    {
      return NBINS-1;
    }
}

#define bin_index(sz) \
 ((in_smallbin_range(sz)) ? smallbin_index(sz) : largebin_index(sz))

/*
  FIRST_SORTED_BIN_SIZE is the chunk size corresponding to the
  first bin that is maintained in sorted order. This must
  be the smallest size corresponding to a given bin.

  Normally, this should be MIN_LARGE_SIZE. But you can weaken
  best fit guarantees to sometimes speed up malloc by increasing value.
  Doing this means that malloc may choose a chunk that is 
  non-best-fitting by up to the width of the bin.

  Some useful cutoff values:
      512 - all bins sorted
     2560 - leaves bins <=     64 bytes wide unsorted  
    12288 - leaves bins <=    512 bytes wide unsorted
    65536 - leaves bins <=   4096 bytes wide unsorted
   262144 - leaves bins <=  32768 bytes wide unsorted
       -1 - no bins sorted (not recommended!)
*/

/* #define FIRST_SORTED_BIN_SIZE 65536 */

#define FIRST_SORTED_BIN_SIZE MIN_LARGE_SIZE


/*
  Unsorted chunks

    All remainders from chunk splits, as well as all returned chunks,
    are first placed in the "unsorted" bin. They are then placed
    in regular bins after malloc gives them ONE chance to be used before
    binning. So, basically, the unsorted_chunks list acts as a queue,
    with chunks being placed on it in free (and malloc_consolidate),
    and taken off (to be either used or placed in bins) in malloc.
*/

/* The otherwise unindexable 1-bin is used to hold unsorted chunks. */
#define unsorted_chunks(M)          (bin_at(M, 1))

/*
  Top

    The top-most available chunk (i.e., the one bordering the end of
    available memory) is treated specially. It is never included in
    any bin, is used only if no other chunk is available, and is
    released back to the system if it is very large (see
    M_TRIM_THRESHOLD).  Because top initially
    points to its own bin with initial zero size, thus forcing
    extension on the first malloc request, we avoid having any special
    code in malloc to check whether it even exists yet. But we still
    need to do so when getting memory from system, so we make
    initial_top treat the bin as a legal but unusable chunk during the
    interval between initialization and the first call to
    sYSMALLOc. (This is somewhat delicate, since it relies on
    the 2 preceding words to be zero during this interval as well.)
*/

/* Conveniently, the unsorted bin can be used as dummy top on first call */
#define initial_top(M)              (unsorted_chunks(M))

/*
  Binmap

    To help compensate for the large number of bins, a one-level index
    structure is used for bin-by-bin searching.  `binmap' is a
    bitvector recording whether bins are definitely empty so they can
    be skipped over during during traversals.  The bits are NOT always
    cleared as soon as bins are empty, but instead only
    when they are noticed to be empty during traversal in malloc.
*/

/* Conservatively use 32 bits per map word, even if on 64bit system */
#define BINMAPSHIFT      5
#define BITSPERMAP       (1U << BINMAPSHIFT)
#define BINMAPSIZE       (NBINS / BITSPERMAP)

#define idx2block(i)     ((i) >> BINMAPSHIFT)
#define idx2bit(i)       ((1U << ((i) & ((1U << BINMAPSHIFT)-1))))

#define mark_bin(m,i)    ((m)->binmap[idx2block(i)] |=  idx2bit(i))
#define unmark_bin(m,i)  ((m)->binmap[idx2block(i)] &= ~(idx2bit(i)))
#define get_binmap(m,i)  ((m)->binmap[idx2block(i)] &   idx2bit(i))

/*
  Fastbins

    An array of lists holding recently freed small chunks.  Fastbins
    are not doubly linked.  It is faster to single-link them, and
    since chunks are never removed from the middles of these lists,
    double linking is not necessary. Also, unlike regular bins, they
    are not even processed in FIFO order (they use faster LIFO) since
    ordering doesn't much matter in the transient contexts in which
    fastbins are normally used.

    Chunks in fastbins keep their inuse bit set, so they cannot
    be consolidated with other free chunks. malloc_consolidate
    releases all chunks in fastbins and consolidates them with
    other free chunks. 
*/

typedef struct chunkinfo* mfastbinptr;

/* offset 2 to use otherwise unindexable first 2 bins */
#define fastbin_index(sz)        ((((unsigned int)(sz)) >> 3) - 2)

/* The maximum fastbin request size we support */
#define MAX_FAST_SIZE     80

#define NFASTBINS  (fastbin_index(request2size(MAX_FAST_SIZE))+1)

/*
  FASTBIN_CONSOLIDATION_THRESHOLD is the size of a chunk in free()
  that triggers automatic consolidation of possibly-surrounding
  fastbin chunks. This is a heuristic, so the exact value should not
  matter too much. It is defined at half the default trim threshold as a
  compromise heuristic to only attempt consolidation if it is likely
  to lead to trimming. However, it is not dynamically tunable, since
  consolidation reduces fragmentation surrounding loarge chunks even 
  if trimming is not used.
*/

#define FASTBIN_CONSOLIDATION_THRESHOLD  \
  ((unsigned long)(DEFAULT_TRIM_THRESHOLD) >> 1)

/*
  Since the lowest 2 bits in max_fast don't matter in size comparisons, 
  they are used as flags.
*/

/*
  ANYCHUNKS_BIT held in max_fast indicates that there may be any
  freed chunks at all. It is set true when entering a chunk into any
  bin.
*/

#define ANYCHUNKS_BIT        (1U)

#define have_anychunks(M)     (((M)->max_fast &  ANYCHUNKS_BIT))
#define set_anychunks(M)      ((M)->max_fast |=  ANYCHUNKS_BIT)
#define clear_anychunks(M)    ((M)->max_fast &= ~ANYCHUNKS_BIT)

/*
  FASTCHUNKS_BIT held in max_fast indicates that there are probably
  some fastbin chunks. It is set true on entering a chunk into any
  fastbin, and cleared only in malloc_consolidate.
*/

#define FASTCHUNKS_BIT        (2U)

#define have_fastchunks(M)   (((M)->max_fast &  FASTCHUNKS_BIT))
#define set_fastchunks(M)    ((M)->max_fast |=  (FASTCHUNKS_BIT|ANYCHUNKS_BIT))
#define clear_fastchunks(M)  ((M)->max_fast &= ~(FASTCHUNKS_BIT))

/* 
   Set value of max_fast. 
   Use impossibly small value if 0.
*/

#define set_max_fast(M, s) \
  (M)->max_fast = (((s) == 0)? SMALLBIN_WIDTH: request2size(s)) | \
  ((M)->max_fast &  (FASTCHUNKS_BIT|ANYCHUNKS_BIT))

#define get_max_fast(M) \
  ((M)->max_fast & ~(FASTCHUNKS_BIT | ANYCHUNKS_BIT))


/*
  morecore_properties is a status word holding dynamically discovered
  or controlled properties of the morecore function
*/

#define MORECORE_CONTIGUOUS_BIT  (1U)

#define contiguous(M) \
        (((M)->morecore_properties &  MORECORE_CONTIGUOUS_BIT))
#define noncontiguous(M) \
        (((M)->morecore_properties &  MORECORE_CONTIGUOUS_BIT) == 0)
#define set_contiguous(M) \
        ((M)->morecore_properties |=  MORECORE_CONTIGUOUS_BIT)
#define set_noncontiguous(M) \
        ((M)->morecore_properties &= ~MORECORE_CONTIGUOUS_BIT)

#define MORECORE_32BIT_BIT  (2U)

#define morecore32bit(M) \
        (((M)->morecore_properties &  MORECORE_32BIT_BIT))
#define nonmorecore32bit(M) \
        (((M)->morecore_properties &  MORECORE_32BIT_BIT) == 0)
#define set_morecore32bit(M) \
        ((M)->morecore_properties |=  MORECORE_32BIT_BIT)
#define set_nonmorecore32bit(M) \
        ((M)->morecore_properties &= ~MORECORE_32BIT_BIT)



/* ----------------- dnmalloc -------------------- */

/* size of pages */
#define PGSIZE malloc_getpagesize
/* pointer size */
#define PTRSIZE sizeof(long)



/* TODO: mmapped chunks are always multiples of pagesize -> we're wasting 
   address space: the hashtable has granularity of 16*8, set it to something 
   closer to pagesize for mmapped chunks (current waste: 32 positions/mmapped 
   page) 
*/

/* The maximum heap size that dnmalloc can operate with
 * represented in hex to avoid annoying gcc warning
 *
 * Avoid integer overflow, cover complete 32bit address 
 * space for portability. With deferred allocation, the
 * hashtable size is a non-issue.
 */
#define HEAPMAXSIZE_HALF 0x80000000UL

/* How many elements are stored in the linked list */
#define LINKEDLSTELS 8

/* Minimum size of a chunk */

#if (SIZEOF_UNSIGNED_LONG == 8) || defined(__arch64__) || defined(__ia64__) || defined(__x86_64__) || defined(__LP64__) || defined(__64BIT__) || defined(_LP64) || defined(_M_IA64) || (defined(_MIPS_SZLONG) && (_MIPS_SZLONG == 64))
#  define MINCHUNKSIZE 32
#else
#  define MINCHUNKSIZE 16
#endif


/* The amount of hashtable entries for each page:
Pagesize divded by the numer of elements in the linkedlists
divided by the minimum chunk size
*/
#define CHUNKINFOPAGE (PGSIZE / LINKEDLSTELS / MINCHUNKSIZE)

/* The amount of hashtable entries needed to manage the memory:
Maximum heap size divided by page size multiplied by the amount
of chunk info's per page 
*/
#define AMOUNTHASH ((HEAPMAXSIZE_HALF / PGSIZE) * CHUNKINFOPAGE * 2)

/* Initial size of the map for the hashtable 
Amount of entries muliplied by pointer size
*/
#define  HASHTABLESIZE (AMOUNTHASH * PTRSIZE)

/* Amount of free chunks that the system should allocate at the start */
#define NUMBER_FREE_CHUNKS 32768

/* Initial size of the chunk info region, 
also used when growing the region */
#define CIREGSIZE (NUMBER_FREE_CHUNKS * sizeof(struct chunkinfo))

/* Start address of the heap */
char *startheap;

/* pointer to the hashtable: struct chunkinfo **hashtable -> *hashtable[] */
chunkinfoptr *hashtable;

/* Current chunkinfo region */
struct cireginfo *currciinfo = 0;
struct cireginfo *firstciinfo = 0;

unsigned long totalcictr = 0;


/* Initialize the area for chunkinfos and the hashtable and protect 
 * it with non-writable pages 
 */
static void
dnmalloc_init ()
{
   void *hashtb;
   int mprot;
   int flags = MAP_PRIVATE;

   /* Allocate the malloc_state struct */
   malloc_mmap_state();

   /* Use MAP_NORESERVE if available (Solaris, HP-UX; most other
    * systems use defered allocation anyway.
    */
#ifdef MAP_NORESERVE
   flags |= MAP_NORESERVE;
#endif

   /* Always start at 0, hashtable covers whole 32bit address space
    */
#define STARTHEAP_IS_ZERO
   startheap = 0;

   /* Map space for the hashtable */
#if PARANOIA > 1
   hashtb = MMAP(0, HASHTABLESIZE+(2*PGSIZE), PROT_READ|PROT_WRITE, flags);
#else
   hashtb = MMAP(0, HASHTABLESIZE+PGSIZE, PROT_READ|PROT_WRITE, flags);
#endif

#ifdef NDEBUG
   if (hashtb == MAP_FAILED) {
      fprintf (stderr, "Couldn't mmap hashtable: %s\n", strerror (errno));
      abort ();
   }
#else
   assert(hashtb != MAP_FAILED);
#endif

   /* Protect the hashtable with non-writable pages */
   mprot = mprotect(hashtb, (size_t) PGSIZE, PROT_NONE);
#ifdef NDEBUG
   if (mprot == -1) {
     fprintf (stderr, "Couldn't mprotect first non-rw page for hashtable: %s\n",
              strerror (errno));
     abort ();
   }
#else
   assert(mprot != -1);
#endif

   /* HP-UX: Cannot do arithmetic with pointers to objects of unknown size. */
   hashtable = (chunkinfoptr *) (((char*)hashtb) + PGSIZE);

   /* Protect the hashtable with non-writable pages */
#if PARANOIA > 1
   mprot = mprotect((void*)((char*)hashtb+HASHTABLESIZE+PGSIZE), (size_t) PGSIZE, PROT_NONE);
#ifdef NDEBUG
   if (mprot == -1) {
     fprintf (stderr, "Couldn't mprotect last non-rw page for hashtable: %s\n",
              strerror (errno));
     abort ();
   }
#else
   assert(mprot != -1);
#endif
#endif
}



/* Extend the region for chunk infos by mapping more memory before the region */
static void
cireg_extend ()
{
   void *newcireg;
   int mprot;
   struct cireginfo *tempciinfo = 0;
   
#if PARANOIA > 1
   newcireg = MMAP(0, CIREGSIZE+(2*PGSIZE), PROT_READ|PROT_WRITE, MAP_PRIVATE);
#else
   newcireg = MMAP(0, CIREGSIZE+PGSIZE, PROT_READ|PROT_WRITE, MAP_PRIVATE);
#endif

#ifdef NDEBUG
   if (newcireg == MAP_FAILED)
   {
           fprintf (stderr, "Couldn't extend chunkinfo region: %s\n",
                    strerror (errno));
           abort ();
   }
#else
   assert(newcireg != MAP_FAILED);
#endif
   mprot = mprotect(newcireg, PGSIZE, PROT_NONE);
#ifdef NDEBUG
   if (mprot == -1) {
           fprintf (stderr, "Couldn't mprotect first non-rw page for extended region: %s\n",
                    strerror (errno));
           abort ();
   }
#else
   assert(mprot != -1);
#endif
   newcireg = ((char*)newcireg)+PGSIZE;
   
#if PARANOIA > 1
   mprot = mprotect((void*)((char*)newcireg+CIREGSIZE), (size_t) PGSIZE, PROT_NONE);
#ifdef NDEBUG
   if (mprot == -1) {
     fprintf (stderr, "Couldn't mprotect last non-rw page for extended region: %s\n",
              strerror (errno));
     abort ();
   }
#else
   assert(mprot != -1);
#endif
#endif

   tempciinfo = currciinfo;
   currciinfo = (struct cireginfo *) newcireg;
   if (tempciinfo)
           tempciinfo->next = currciinfo;
   currciinfo->position = 1;
   currciinfo->freecounter = NUMBER_FREE_CHUNKS;
   if (!firstciinfo)
           firstciinfo = currciinfo;
   totalcictr++;
   VALGRIND_CREATE_MEMPOOL(newcireg, 0, 0);
}


/* Get a free chunkinfo */
static chunkinfoptr
cireg_getfree ()
{
   chunkinfoptr freeci;
   chunkinfoptr freelst = 0;
   struct cireginfo *newciinfo = firstciinfo;
   
   if (newciinfo) {
        freelst = newciinfo->freelist;
    
        if (!freelst && newciinfo->next) {
          do {
            newciinfo = newciinfo->next;
            freelst = newciinfo->freelist;
          } while (!freelst && newciinfo->next);
        }
   }

   /* Check if there are any free chunkinfos on the list of free chunkinfos */
   if (freelst)
   {
     freeci = freelst;
     newciinfo->freecounter--;
     newciinfo->freelist = freelst->fd;
     
     VALGRIND_MEMPOOL_ALLOC((char*)currciinfo, (char*)freeci, 
                            sizeof(struct chunkinfo));
     
     freeci->prev_size = 0;
     freeci->size      = 0;
     freeci->req       = 0;
     freeci->hash_next = NULL;
     freeci->fd        = NULL;
     freeci->bk        = NULL;
     freeci->chunk     = NULL;
     return (freeci);
   }
   else
   {
     /* No free chunkinfos, check if chunkinfo region still has place 
      * for a chunkinfo. If not, extend the region. 
      */
     if (UNLIKELY(!currciinfo || currciinfo->position == NUMBER_FREE_CHUNKS))
       cireg_extend ();
     /* Get a chunkinfo from the chunkinfo region */
     freeci = (chunkinfoptr) currciinfo + currciinfo->position; 
     currciinfo->freecounter--;
     currciinfo->position++;

     VALGRIND_MEMPOOL_ALLOC((char*)currciinfo, (char*)freeci, 
                            sizeof(struct chunkinfo));

     return (freeci);
   }
}

static void freeciregion(struct cireginfo *freeme) {
  /* free the chunkinfo region */
  struct cireginfo *newciinfo = firstciinfo;
  struct cireginfo *prevciinfo = firstciinfo;
  void *unmapme;

  while (newciinfo && newciinfo != freeme) {
    prevciinfo = newciinfo;
    newciinfo = newciinfo->next;
  }
  assert(freeme == newciinfo); /* rw */
  assert(newciinfo != NULL);   /* rw */
  if (newciinfo)
    prevciinfo->next = newciinfo->next;
  unmapme = (void *) ((char*)freeme - PGSIZE);
  VALGRIND_DESTROY_MEMPOOL((char*)freeme);
#if PARANOIA > 1
  munmap(unmapme, CIREGSIZE+(2*PGSIZE));
#else
  munmap(unmapme, CIREGSIZE+PGSIZE);
#endif
}


static void freecilst_add(chunkinfoptr p) {

  struct cireginfo *newciinfo;

  newciinfo = currciinfo;
  if (((chunkinfoptr) newciinfo < p) && (p  <  (chunkinfoptr) (newciinfo+NUMBER_FREE_CHUNKS))) {
    p->fd = newciinfo->freelist;
    newciinfo->freelist = p;
    newciinfo->freecounter++;
    VALGRIND_MEMPOOL_FREE((char*)newciinfo, (char*)p);
    VALGRIND_MAKE_MEM_DEFINED(p,sizeof(struct chunkinfo));
    VALGRIND_MAKE_MEM_NOACCESS(p->size,     sizeof(INTERNAL_SIZE_T));
    VALGRIND_MAKE_MEM_NOACCESS(p->req,      sizeof(INTERNAL_SIZE_T));
    VALGRIND_MAKE_MEM_NOACCESS(p->bk,       sizeof(struct chunkinfo*));
    VALGRIND_MAKE_MEM_NOACCESS(p->chunk,    sizeof(mchunkptr));
  } else {
    newciinfo = firstciinfo;
    if (newciinfo) {
      do {
        if (((chunkinfoptr) newciinfo < p) && (p  <  (chunkinfoptr) (newciinfo+NUMBER_FREE_CHUNKS))) {
          p->fd = newciinfo->freelist;
          newciinfo->freelist = p;
          newciinfo->freecounter++;
          VALGRIND_MEMPOOL_FREE((char*)newciinfo, (char*)p);
          VALGRIND_MAKE_MEM_DEFINED(p,sizeof(struct chunkinfo));
          VALGRIND_MAKE_MEM_NOACCESS(p->size,     sizeof(INTERNAL_SIZE_T));
          VALGRIND_MAKE_MEM_NOACCESS(p->req,      sizeof(INTERNAL_SIZE_T));
          VALGRIND_MAKE_MEM_NOACCESS(p->bk,       sizeof(struct chunkinfo*));
          VALGRIND_MAKE_MEM_NOACCESS(p->chunk,    sizeof(mchunkptr));
          if (UNLIKELY(newciinfo->freecounter == NUMBER_FREE_CHUNKS))
            freeciregion(newciinfo);
          break;
        }
        newciinfo = newciinfo->next;
      } while (newciinfo);
    }
  }
}

/* Calculate the hash table entry for a chunk */
#ifdef STARTHEAP_IS_ZERO
#define hash(p)  (((unsigned long) p) >> 7)
#else
#define hash(p)  (((unsigned long) p - (unsigned long) startheap) >> 7)
#endif

static void
hashtable_add (chunkinfoptr ci)
{
   chunkinfoptr temp, next;
   unsigned long hashval;
   mchunkptr cic = chunk (ci);
   
   hashval = hash (cic);

   if (hashval < AMOUNTHASH) {

     temp = hashtable[hashval];

#ifdef DNMALLOC_DEBUG
     fprintf(stderr, "hashtable_add: %p, %lu\n", chunk(ci), hashval);
#endif

     /* If no pointer to a chunk info list is stored at this location 
      * in the hashtable or if the chunk's address is smaller than the 
      * one present, add the chunk to the front of the linked list
      */
     if (temp == 0 || chunk (temp) > cic)
       {
         ci->hash_next = temp;
         hashtable[hashval] = ci;
         if (!temp) /* more likely case */
           goto out;
         temp->prev_size = chunksize(ci);
         return;
       }
     else
       {
         /* We must place the chunk in the linked list for this hashentry
          * Loop to end of list or to a position where temp's chunk's address 
          * is larger than the new chunkinfo's chunk's address
          */
         if (!temp->hash_next || (chunk (temp->hash_next) > cic))
           {
             ci->hash_next = temp->hash_next;
             temp->hash_next = ci;
           }
         else
           {
             while ((temp->hash_next != 0) && (chunk (temp->hash_next) < cic))
               {
                 temp = temp->hash_next;
               }
             /* Place in linked list if not already there */
             if (!temp->hash_next || !(chunk (temp->hash_next) == cic))
               {
                 ci->hash_next = temp->hash_next;
                 temp->hash_next = ci;
               }
           }
       }
   }
   else {
#ifdef DNMALLOC_CHECKS
     if (hashval >= AMOUNTHASH) {
       fprintf(stderr, "Dnmalloc error: trying to write outside of the bounds of the hashtable, this is definitely a bug, please email dnmalloc@fort-knox.org (hashval: %lu, AMOUNTHASH: %lu, HEAPMAXSIZE_HALF %lu PGSIZE %ld CHUNKINFOPAGE %ld chunk: %p, chunkinfo: %p, startheap: %p).\n", hashval, AMOUNTHASH, HEAPMAXSIZE_HALF, PGSIZE, CHUNKINFOPAGE, chunk(ci), ci, startheap);
           abort();
     }
#else
     assert(hashval < AMOUNTHASH);
#endif
   }

 out:
   next = next_chunkinfo(ci);
   if (!next)
     return;
   next->prev_size = chunksize(ci);
}

static void
hashtable_insert (chunkinfoptr ci_orig, chunkinfoptr ci_insert)
{
   chunkinfoptr next;

#ifdef DNMALLOC_DEBUG
   fprintf(stderr, "hashtable_ins: %p, %lu\n", chunk(ci_insert), 
           (unsigned long)hash(chunk(ci_insert));
#endif

   if (hash(chunk(ci_orig)) != hash(chunk(ci_insert))) {
      hashtable_add(ci_insert);  
   }
   else {

      ci_insert->hash_next = ci_orig->hash_next;
      ci_orig->hash_next = ci_insert;

      /* added for prevsize */
      if (!(ci_insert->hash_next))
              next = next_chunkinfo(ci_insert);
      else
              next = ci_insert->hash_next;

      if (!next)
        {
          ci_insert->prev_size = chunksize(ci_orig);
        }
      else
        {
          next->prev_size = chunksize(ci_insert);
          ci_insert->prev_size = chunksize(ci_orig);
        }
   }
}

static void
hashtable_remove (mchunkptr p) 
{
  chunkinfoptr prevtemp, temp;
  unsigned long hashval;
  
  hashval = hash (p);
#ifdef DNMALLOC_DEBUG
  fprintf(stderr, "hashtable_rem: %p, %lu\n", p, hashval);
#endif
  assert(hashval < AMOUNTHASH); /* rw */
  prevtemp = temp = hashtable[hashval];
  if (chunk (temp) == p) {
    hashtable[hashval] = temp->hash_next;
  } 
  else
    {
      if (temp && chunk (temp) != p) {
        do
          {
            prevtemp = temp;
            temp = temp->hash_next;
          } while (temp && chunk (temp) != p);
      }
#ifdef DNMALLOC_CHECKS
      if (!temp) {
        fprintf (stderr,
                 "Dnmalloc error (hash_rm): could not find a chunkinfo for the chunk %p in the hashtable at entry %lu\n This is definitely a bug, please report it to dnmalloc@fort-knox.org.\n",
                 p, hashval);
        abort();
      }
#else
      assert(temp != NULL);
#endif
      if (temp) prevtemp->hash_next = temp->hash_next;
    }
}

/* mmapped chunks are multiples of pagesize, no hash_nexts, 
 * just remove from the hashtable 
 */
#define hashtable_remove_mmapped(p) hashtable[hash(p)] = 0;

static void
hashtable_skiprm (chunkinfoptr ci_orig, chunkinfoptr ci_todelete)
{
   unsigned long hashval;
   chunkinfoptr next;
   
#ifdef DNMALLOC_DEBUG
   fprintf(stderr, "hashtable_skiprm: %p, %lu\n", chunk(ci_todelete), hash(chunk(ci_todelete)));
#endif

   if (ci_orig->hash_next != ci_todelete) {
     hashval = hash(chunk(ci_todelete));
     assert(hashval < AMOUNTHASH); /* rw */
#ifdef DNMALLOC_CHECKS
     if (hashtable[hashval] != ci_todelete ) {
           fprintf(stderr, "Dnmalloc error: trying to delete wrong value (hash: %lu): ci_todelete: %p (%p), hashtable[hashval]: %p (%p)\n This is definitely a bug, please report it to dnmalloc@fort-knox.org.\n", hashval, ci_todelete, chunk(ci_todelete), hashtable[hashval], chunk(hashtable[hashval]));
     }
#else
     assert(hashtable[hashval] == ci_todelete);
#endif
     hashtable[hashval] = ci_todelete->hash_next;
   }

   else {
     ci_orig->hash_next = ci_todelete->hash_next;
     if (!ci_orig->hash_next) {
       next = next_chunkinfo(ci_orig);
     } else {
       next = ci_orig->hash_next;
     }
     if (next)
       next->prev_size = chunksize(ci_orig);

   } 
}


static chunkinfoptr
hashtable_lookup (mchunkptr p)
{
   chunkinfoptr ci;
   unsigned long hashval;
   
   /* if we were called wrongly
    * if ((char *) p < startheap) return 0;
    */
   if ((char *) p >= startheap)
     {
       hashval = hash (p);
       assert(hashval < AMOUNTHASH); /* rw */
       ci = hashtable[hashval];
       if (ci && chunk (ci) == p)
         return ci;

       if (ci) {
         do {
           ci = ci->hash_next;
         } while (ci && chunk (ci) != p);
       }
#ifdef DNMALLOC_CHECKS
       /* This should never occur but if it does, we'd like to know */
       if (!ci) {
         fprintf (stderr,
                  "Dnmalloc error: could not find a chunkinfo for the chunk %p in the hashtable at entry %lu\n This is definitely a bug, please report it to dnmalloc@fort-knox.org.\n",
                  p, hashval);
         abort();
       }
#else
       assert(ci != NULL);
#endif
       return ci;
     }
   return 0;
}



/*
   ----------- Internal state representation and initialization -----------
*/

struct malloc_state {

  /* The maximum chunk size to be eligible for fastbin */
  INTERNAL_SIZE_T  max_fast;   /* low 2 bits used as flags */

  /* Fastbins */
  mfastbinptr      fastbins[NFASTBINS];

  /* Base of the topmost chunk -- not otherwise kept in a bin */
  chunkinfoptr        top;

  /* The remainder from the most recent split of a small request */
  chunkinfoptr        last_remainder;

  /* Normal bins */
  struct chunkinfo bins[NBINS];

  /* Bitmap of bins. Trailing zero map handles cases of largest binned size */
  unsigned int     binmap[BINMAPSIZE+1];

  /* Tunable parameters */
  CHUNK_SIZE_T     trim_threshold;
  INTERNAL_SIZE_T  top_pad;
  INTERNAL_SIZE_T  mmap_threshold;

  /* Memory map support */
  int              n_mmaps;
  int              n_mmaps_max;
  int              max_n_mmaps;

  /* Cache malloc_getpagesize */
  unsigned int     pagesize;  

  /* Canary */
  char             guard_stored[GUARD_SIZE];

  /* Track properties of MORECORE */
  unsigned int     morecore_properties;

  /* Statistics */
  INTERNAL_SIZE_T  mmapped_mem;
  INTERNAL_SIZE_T  sbrked_mem;
  INTERNAL_SIZE_T  max_sbrked_mem;
  INTERNAL_SIZE_T  max_mmapped_mem;
  INTERNAL_SIZE_T  max_total_mem;
};

typedef struct malloc_state *mstate;

/* 
   There is exactly one instance of this struct in this malloc.
   If you are adapting this malloc in a way that does NOT use a static
   malloc_state, you MUST explicitly zero-fill it before using. This
   malloc relies on the property that malloc_state is initialized to
   all zeroes (as is true of C statics).
*/

static struct malloc_state * av_ = NULL;  /* never directly referenced */

/*
   All uses of av_ are via get_malloc_state().
   At most one "call" to get_malloc_state is made per invocation of
   the public versions of malloc and free, but other routines
   that in turn invoke malloc and/or free may call more then once. 
   Also, it is called in check* routines if DEBUG is set.
*/

#define get_malloc_state() (av_)

/*
  Initialize a malloc_state struct.

  This is called only from within malloc_consolidate, which needs
  be called in the same contexts anyway.  It is never called directly
  outside of malloc_consolidate because some optimizing compilers try
  to inline it at all call points, which turns out not to be an
  optimization at all. (Inlining it in malloc_consolidate is fine though.)
*/

#if __STD_C
static void malloc_mmap_state(void)
#else
static void malloc_mmap_state()
#endif
{
  int mprot;
  unsigned long pagesize = malloc_getpagesize;
  size_t size = (sizeof(struct malloc_state) + pagesize - 1) & ~(pagesize - 1);

  void * foo = MMAP(0, size+(2*pagesize), PROT_READ|PROT_WRITE, MAP_PRIVATE);


#ifdef NDEBUG
   if (foo == MAP_FAILED) {
      fprintf (stderr, "Couldn't mmap struct malloc_state: %s\n", strerror (errno));
      abort ();
   }
#else
   assert(foo != MAP_FAILED);
#endif

   mprot = mprotect(foo, pagesize, PROT_NONE);
#ifdef NDEBUG
   if (mprot == -1) {
     fprintf (stderr, "Couldn't mprotect first non-rw page for struct malloc_state: %s\n",
              strerror (errno));
     abort ();
   }
#else
   assert(mprot != -1);
#endif

   av_ = (struct malloc_state *) ((char*)foo + pagesize);

   MALLOC_ZERO(av_, sizeof(struct malloc_state));

   mprot = mprotect((void*)((char*)foo + size + pagesize), (size_t) pagesize, PROT_NONE);
#ifdef NDEBUG
   if (mprot == -1) {
     fprintf (stderr, 
              "Couldn't mprotect last non-rw page for struct malloc_state: %s\n",
              strerror (errno));
     abort ();
   }
#else
   assert(mprot != -1);
#endif
}

#if __STD_C
static void malloc_init_state(mstate av)
#else
static void malloc_init_state(av) mstate av;
#endif
{
  int     i;
  mbinptr bin;

  void *  morecore_test = MORECORE(0);
  unsigned long hashval;

  /* Test morecore function 
   */
  set_morecore32bit(av);

  if (morecore_test == MORECORE_FAILURE)
    {
      set_nonmorecore32bit(av);
    }
  else
    {
      /* On 64bit systems, the heap may be located above the
       * 32bit address space. Since mmap() probably still can be
       * convinced to map within 32bit, we don't use sbrk().
       */
      hashval = hash (morecore_test);
      if (hashval >= AMOUNTHASH) 
        {
          set_nonmorecore32bit(av);
        }
    }

  
  /* Establish circular links for normal bins */
  for (i = 1; i < NBINS; ++i) { 
    bin = bin_at(av,i);
    bin->fd = bin->bk = bin;
  }

  av->top_pad        = DEFAULT_TOP_PAD;
  av->n_mmaps_max    = DEFAULT_MMAP_MAX;
  av->mmap_threshold = DEFAULT_MMAP_THRESHOLD;
  av->trim_threshold = DEFAULT_TRIM_THRESHOLD;

#if MORECORE_CONTIGUOUS
  set_contiguous(av);
#else
  set_noncontiguous(av);
#endif

  set_max_fast(av, DEFAULT_MXFAST);

  av->top = cireg_getfree ();
  av->top->chunk     = (mchunkptr) startheap;
  av->top->size      = 0;
  set_previnuse(av->top);
  clear_inuse(av->top);
  hashtable[0]       = av->top;
  av->pagesize       = malloc_getpagesize;

  memcpy(av->guard_stored, dnmalloc_arc4random(), GUARD_SIZE);

}

/* 
   Other internal utilities operating on mstates
*/

#if __STD_C
static Void_t*  sYSMALLOc(INTERNAL_SIZE_T, mstate);
static int      sYSTRIm(size_t, mstate);
static void     malloc_consolidate(mstate);
#else
static Void_t*  sYSMALLOc();
static int      sYSTRIm();
static void     malloc_consolidate();
#endif

/* dnmalloc functions */
/* needs mstate so moved here */

static chunkinfoptr
next_chunkinfo (chunkinfoptr ci)
{
   mchunkptr nextp;
   unsigned long hashval;
   chunkinfoptr cinfonextp;
   mstate av = get_malloc_state();
   
   /* ci is not the last element in the linked list, just 
      return the next chunkinfo from the list 
   */
   if (!ci->hash_next)
     {
       /* ci is the last element, find the next chunkinfo by 
        * looking up the chunkinfo for the chunk that is after p's chunk 
        */
       nextp = (mchunkptr) (((char *) (ci->chunk)) + chunksize (ci));

       if (!(nextp == av->top->chunk)) 
         {
           hashval = hash (nextp);
           /* assert(hashval < AMOUNTHASH); *//* major bottleneck */
           cinfonextp = hashtable[hashval];
           if (cinfonextp && chunk (cinfonextp) == nextp)
             return cinfonextp; 
           
#ifdef DNMALLOC_CHECKS_EXTRA
           /* This seems bogus; a chunkinfo may legally have no nextp if
            * it's the last one allocated (?)
            */
           else {
             if (cinfonextp)
               fprintf (stderr,
                        "Dnmalloc error: could not find a next chunkinfo for the chunk %p in the hashtable at entry %lu, cinfonextp: %p, chunk(cinfonextp): %p, nextp: %p\n This is definitely a bug, please report it to dnmalloc@fort-knox.org.\n",
                        chunk(ci), hashval, cinfonextp, chunk(cinfonextp), nextp);
             else
               fprintf (stderr,
                        "Dnmalloc error: could not find a next chunkinfo for the chunk %p in the hashtable at entry %lu, cinfonextp: %s, chunk(cinfonextp): %s, nextp: %p\n This is definitely a bug, please report it to dnmalloc@fort-knox.org.\n",
                        chunk(ci), hashval, "null", "null", nextp);
           }
#endif
          
           return NULL;
         }
       else
         {
           return av->top;
         }

     }
   else
     {
       return (ci->hash_next);
     }
}

static int is_next_chunk(chunkinfoptr oldp, chunkinfoptr newp) {
        mchunkptr nextp;
        if (oldp->hash_next == newp)
                return 1;
        nextp = (mchunkptr) (((char *) (oldp->chunk)) + chunksize (oldp));
        if (nextp == chunk(newp))
                return 1;
        return 0;
}



/* Get the chunkinfo of the physically previous chunk */
/* Since we disposed of prev_size, we need this function to find the previous */

static chunkinfoptr
prev_chunkinfo (chunkinfoptr ci)
{
   unsigned int i;
   chunkinfoptr prev;
   mchunkptr prevchunk = 0;
   /* chunkinfoptr temp; */
   
   /* Get the hashtable location of the chunkinfo */
   i = hash (chunk (ci));
   assert(i < AMOUNTHASH); /* rw */
      
   /* Get the first element of the linked list of chunkinfo's that contains p */
   prev = hashtable[i];
   
   if (ci == prev) {
     prevchunk = (mchunkptr) (((char *) (ci->chunk)) - (ci->prev_size));
     i = hash(prevchunk);
     assert(i < AMOUNTHASH); /* rw */
     /* Loop over the linked list until we reach the last element */
     for (prev = hashtable[i]; prev->hash_next != 0; prev = prev->hash_next) ;
   } else {
     /* p is not the first element in the linked list, we can just 
        loop over the list and return the previous 
     */
     for (prev = hashtable[i]; prev->hash_next != ci; prev = prev->hash_next);
   }

   return prev;  
}


/*
  Debugging support
  Dnmalloc broke dlmallocs debugging functions, should fix them some 
  time in the future, for now leave them undefined.
*/

#define check_chunk(P)
#define check_free_chunk(P)
#define check_inuse_chunk(P)
#define check_remalloced_chunk(P,N)
#define check_malloced_chunk(P,N)
#define check_malloc_state()


/* ----------- Routines dealing with system allocation -------------- */

/*
  sysmalloc handles malloc cases requiring more memory from the system.
  On entry, it is assumed that av->top does not have enough
  space to service request for nb bytes, thus requiring that av->top
  be extended or replaced.
*/

#if __STD_C
static Void_t* sYSMALLOc(INTERNAL_SIZE_T nb, mstate av)
#else
static Void_t* sYSMALLOc(nb, av) INTERNAL_SIZE_T nb; mstate av;
#endif
{
  chunkinfoptr    old_top;        /* incoming value of av->top */
  INTERNAL_SIZE_T old_size;       /* its size */
  char*           old_end;        /* its end address */

  long            size;           /* arg to first MORECORE or mmap call */
  char*           brk;            /* return value from MORECORE */

  long            correction;     /* arg to 2nd MORECORE call */
  char*           snd_brk;        /* 2nd return val */

  INTERNAL_SIZE_T front_misalign; /* unusable bytes at front of new space */
  INTERNAL_SIZE_T end_misalign;   /* partial page left at end of new space */
  char*           aligned_brk;    /* aligned offset into brk */

  chunkinfoptr    p;              /* the allocated/returned chunk */
  chunkinfoptr    remainder;      /* remainder from allocation */
  chunkinfoptr    fencepost;      /* fencepost */
  CHUNK_SIZE_T    remainder_size; /* its size */

  CHUNK_SIZE_T    sum;            /* for updating stats */

  size_t          pagemask  = av->pagesize - 1;

#ifdef DNMALLOC_DEBUG
  fprintf(stderr, "Enter sysmalloc\n");
#endif
  /*
    If there is space available in fastbins, consolidate and retry
    malloc from scratch rather than getting memory from system.  This
    can occur only if nb is in smallbin range so we didn't consolidate
    upon entry to malloc. It is much easier to handle this case here
    than in malloc proper.
  */


  if (have_fastchunks(av)) {
    Void_t * retval;
    assert(in_smallbin_range(nb));
    malloc_consolidate(av);
#ifdef DNMALLOC_DEBUG
    fprintf(stderr, "Return sysmalloc have_fastchunks\n");
#endif
    retval = mALLOc(nb - MALLOC_ALIGN_MASK);
    VALGRIND_FREELIKE_BLOCK(retval, 0);
    return retval;
  }


  /*
    If have mmap, and the request size meets the mmap threshold, and
    the system supports mmap, and there are few enough currently
    allocated mmapped regions, try to directly map this request
    rather than expanding top.
  */

  if (UNLIKELY((CHUNK_SIZE_T)(nb) >= (CHUNK_SIZE_T)(av->mmap_threshold) &&
               (av->n_mmaps < av->n_mmaps_max))) {

    char* mm;             /* return value from mmap call*/

    /*
      Round up size to nearest page.  For mmapped chunks, the overhead
      is one SIZE_SZ unit larger than for normal chunks, because there
      is no following chunk whose prev_size field could be used.
    */
    size = (nb + MALLOC_ALIGN_MASK + pagemask) & ~pagemask;

    /* Don't try if size wraps around 0 */
    if ((CHUNK_SIZE_T)(size) > (CHUNK_SIZE_T)(nb)) {
            

      mm = (char*)(MMAP(0, size, PROT_READ|PROT_WRITE, MAP_PRIVATE));
      
      if (mm != (char*)(MORECORE_FAILURE)) {
        
        VALGRIND_MAKE_MEM_NOACCESS(mm,size);

        /*
          The offset to the start of the mmapped region is stored
          in the prev_size field of the chunk. This allows us to adjust
          returned start address to meet alignment requirements here 
          and in memalign(), and still be able to compute proper
          address argument for later munmap in free() and realloc().
        */
        
        front_misalign = (INTERNAL_SIZE_T) mm & MALLOC_ALIGN_MASK;
        p = cireg_getfree();
        
        if (front_misalign > 0) {
          correction = MALLOC_ALIGNMENT - front_misalign;
          p->chunk = (mchunkptr)(mm + correction);
          p->hash_next = (chunkinfoptr) correction;
          set_head(p, (size - correction) |INUSE|IS_MMAPPED);
        }
        else {
          p->chunk = (mchunkptr)mm;
          p->hash_next = 0;
          set_head(p, size|INUSE|IS_MMAPPED);
        }
        hashtable_add(p);
        /* update statistics */
        
        if (++av->n_mmaps > av->max_n_mmaps) 
          av->max_n_mmaps = av->n_mmaps;
        
        sum = av->mmapped_mem += size;
        if (sum > (CHUNK_SIZE_T)(av->max_mmapped_mem)) 
          av->max_mmapped_mem = sum;
        sum += av->sbrked_mem;
        if (sum > (CHUNK_SIZE_T)(av->max_total_mem)) 
          av->max_total_mem = sum;

        check_chunk(p);
        
#ifdef DNMALLOC_DEBUG
        fprintf(stderr, "Return mmapped (%lu, total %lu)\n", 
                size, (unsigned long)/* size_t */av->max_total_mem );
#endif
        return chunk(p);
      }
    }
  }

  /* Record incoming configuration of top */

  old_top  = av->top;
  old_size = chunksize(old_top);
  old_end  = (char*)(chunk_at_offset(chunk(old_top), old_size));

  brk = snd_brk = (char*)(MORECORE_FAILURE); 

  /* 
     If not the first time through, we require old_size to be
     at least MINSIZE and to have prev_inuse set.
  */

  /* assert((old_top == initial_top(av) && old_size == 0) || 
         ((CHUNK_SIZE_T) (old_size) >= MINSIZE &&
         prev_inuse(old_top))); */

  /* Precondition: not enough current space to satisfy nb request */
  assert((CHUNK_SIZE_T)(old_size) < (CHUNK_SIZE_T)(nb + MINSIZE));

  /* Precondition: all fastbins are consolidated */
  assert(!have_fastchunks(av));

  /* Request enough space for nb + pad + overhead */
  size = nb + av->top_pad + MINSIZE;

  /*
    If contiguous, we can subtract out existing space that we hope to
    combine with new space. We add it back later only if
    we don't actually get contiguous space.
  */
  if (contiguous(av))
    size -= old_size;

  /*
    Round to a multiple of page size.
    If MORECORE is not contiguous, this ensures that we only call it
    with whole-page arguments.  And if MORECORE is contiguous and
    this is not first time through, this preserves page-alignment of
    previous calls. Otherwise, we correct to page-align below.
  */

  size = (size + pagemask) & ~pagemask;

  /*
    Don't try to call MORECORE if argument is so big as to appear
    negative. Note that since mmap takes size_t arg, it may succeed
    below even if we cannot call MORECORE.
  */
  if (size > 0 && morecore32bit(av)) 
    brk = (char*)(MORECORE(size));

  /*
    If have mmap, try using it as a backup when MORECORE fails or
    cannot be used. This is worth doing on systems that have "holes" in
    address space, so sbrk cannot extend to give contiguous space, but
    space is available elsewhere.  Note that we ignore mmap max count
    and threshold limits, since the space will not be used as a
    segregated mmap region.
  */
  if (brk != (char*)(MORECORE_FAILURE)) {
    av->sbrked_mem += size;
    VALGRIND_MAKE_MEM_NOACCESS(brk,size);
  }

  else {

#ifdef DNMALLOC_DEBUG
    fprintf(stderr, "Morecore failure in sysmalloc\n");
#endif

    /* Cannot merge with old top, so add its size back in */
    if (contiguous(av))
      size = (size + old_size + pagemask) & ~pagemask;

    /* If we are relying on mmap as backup, then use larger units */
    if ((CHUNK_SIZE_T)(size) < (CHUNK_SIZE_T)(MMAP_AS_MORECORE_SIZE))
      size = MMAP_AS_MORECORE_SIZE;

    /* Don't try if size wraps around 0 */
    if ((CHUNK_SIZE_T)(size) > (CHUNK_SIZE_T)(nb)) {

#ifdef DNMALLOC_DEBUG
      fprintf(stderr, "Try mmap in sysmalloc\n");
#endif
      brk = (char*)(MMAP(0, size, PROT_READ|PROT_WRITE, MAP_PRIVATE));
      
      if (brk != (char*)(MORECORE_FAILURE)) {
        
        VALGRIND_MAKE_MEM_NOACCESS(brk,size);

        av->mmapped_mem += size;
#ifdef DNMALLOC_DEBUG
        fprintf(stderr, "Mmapped successfully in sysmalloc %p\n", brk);
#endif

        /* We do not need, and cannot use, another sbrk call to find end */
        snd_brk = brk + size;
        
        /* 
           Record that we no longer have a contiguous sbrk region. 
           After the first time mmap is used as backup, we do not
           ever rely on contiguous space since this could incorrectly
           bridge regions.
        */
        set_noncontiguous(av);
      }
    }
  }

  if (brk != (char*)(MORECORE_FAILURE)) {
#ifdef DNMALLOC_DEBUG
    fprintf(stderr, "Success path %lu allocated, sbrked %lu\n", 
            size, (unsigned long)av->sbrked_mem);
#endif
    /* av->sbrked_mem += size; moved up */

    /*
      If MORECORE extends previous space, we can likewise extend top size.
    */
    
    if (brk == old_end && snd_brk == (char*)(MORECORE_FAILURE)) {
      set_head(old_top, (size + old_size) | PREV_INUSE);
#ifdef DNMALLOC_DEBUG
      fprintf(stderr, "Previous space extended\n");
#endif
    }

    /*
      Otherwise, make adjustments:
      
      * If the first time through or noncontiguous, we need to call sbrk
        just to find out where the end of memory lies.

      * We need to ensure that all returned chunks from malloc will meet
        MALLOC_ALIGNMENT

      * If there was an intervening foreign sbrk, we need to adjust sbrk
        request size to account for fact that we will not be able to
        combine new space with existing space in old_top.

      * Almost all systems internally allocate whole pages at a time, in
        which case we might as well use the whole last page of request.
        So we allocate enough more memory to hit a page boundary now,
        which in turn causes future contiguous calls to page-align.
    */
    
    else {
      /* front_misalign = 0; *//*superfluous */
      /* end_misalign = 0; *//*superfluous */
      correction = 0;
      aligned_brk = brk;

      /*
        If MORECORE returns an address lower than we have seen before,
        we know it isn't really contiguous.  This and some subsequent
        checks help cope with non-conforming MORECORE functions and
        the presence of "foreign" calls to MORECORE from outside of
        malloc or by other threads.  We cannot guarantee to detect
        these in all cases, but cope with the ones we do detect.
      */
      if (contiguous(av) && old_size != 0 && brk < old_end) {
        set_noncontiguous(av);
      }
      
      /* handle contiguous cases */
      if (contiguous(av)) { 

#ifdef DNMALLOC_DEBUG
        fprintf(stderr, "Handle contiguous cases\n");
#endif
        /* 
           We can tolerate forward non-contiguities here (usually due
           to foreign calls) but treat them as part of our space for
           stats reporting.
        */
        if (old_size != 0) 
          av->sbrked_mem += brk - old_end;
        
        /* Guarantee alignment of first new chunk made from this space */

        front_misalign = (INTERNAL_SIZE_T) brk & MALLOC_ALIGN_MASK;
        if (front_misalign > 0) {

          /*
            Skip over some bytes to arrive at an aligned position.
            We don't need to specially mark these wasted front bytes.
            They will never be accessed anyway because
            prev_inuse of av->top (and any chunk created from its start)
            is always true after initialization.
          */

          correction = MALLOC_ALIGNMENT - front_misalign;
          aligned_brk += correction;
        }
        
        /*
          If this isn't adjacent to existing space, then we will not
          be able to merge with old_top space, so must add to 2nd request.
        */
        
        correction += old_size;
        
        /* Extend the end address to hit a page boundary */
        end_misalign = (INTERNAL_SIZE_T)(brk + size + correction);
        correction += ((end_misalign + pagemask) & ~pagemask) - end_misalign;
        
        assert(correction >= 0);
        snd_brk = (char*)(MORECORE(correction));
        
        if (snd_brk == (char*)(MORECORE_FAILURE)) {
          /*
            If can't allocate correction, try to at least find out current
            brk.  It might be enough to proceed without failing.
          */
          correction = 0;
          snd_brk = (char*)(MORECORE(0));
        }
        else if (snd_brk < brk) {
          /*
            If the second call gives noncontiguous space even though
            it says it won't, the only course of action is to ignore
            results of second call, and conservatively estimate where
            the first call left us. Also set noncontiguous, so this
            won't happen again, leaving at most one hole.
            
            Note that this check is intrinsically incomplete.  Because
            MORECORE is allowed to give more space than we ask for,
            there is no reliable way to detect a noncontiguity
            producing a forward gap for the second call.
          */
          snd_brk = brk + size;
          correction = 0;
          set_noncontiguous(av);
        }
        else {
          VALGRIND_MAKE_MEM_NOACCESS(snd_brk,correction);
        }

      }
      
      /* handle non-contiguous cases */
      else { 

#ifdef DNMALLOC_DEBUG
        fprintf(stderr, "Handle non-contiguous cases\n");
#endif

        /* MORECORE/mmap must correctly align */
        assert(aligned_OK(brk));
        
        /* Find out current end of memory */
        if (snd_brk == (char*)(MORECORE_FAILURE)) {
          snd_brk = (char*)(MORECORE(0));
          av->sbrked_mem += snd_brk - brk - size;
        }
#ifdef DNMALLOC_DEBUG
        fprintf(stderr, "Sbrked now %lu\n", (unsigned long)av->sbrked_mem);
#endif
      }
      
      /* Adjust top based on results of second sbrk.
       *
       * If mmap() has been used as backup for failed morecore(),
       * we end up in this branch as well.
       */
      if (snd_brk != (char*)(MORECORE_FAILURE)) {
#ifdef DNMALLOC_DEBUG
        fprintf(stderr, "Adjust top, correction %lu\n", correction);
#endif
        /* hashtable_remove(chunk(av->top)); *//* rw 19.05.2008 removed */
        av->top =  cireg_getfree();
        av->top->chunk = (mchunkptr)aligned_brk;
        set_head(av->top, (snd_brk - aligned_brk + correction) | PREV_INUSE);
#ifdef DNMALLOC_DEBUG
        fprintf(stderr, "Adjust top, top %p size %lu\n", 
                av->top, (unsigned long)chunksize(av->top));
#endif
        hashtable_add(av->top);
        av->sbrked_mem += correction;
     
        /*
          If not the first time through, we either have a
          gap due to foreign sbrk or a non-contiguous region.  Insert a
          double fencepost at old_top to prevent consolidation with space
          we don't own. These fenceposts are artificial chunks that are
          marked as inuse. Original dlmalloc had two of these but too 
          small to use. To ensure that the linked lists contain a maximum 
          of 8 elements we only use 1. Inuse is determined by the 
          current rather than the next chunk anyway.
        */
   
        if (old_size != 0) {
#ifdef DNMALLOC_DEBUG
          fprintf(stderr, "Shrink old_top to insert fenceposts\n");
#endif
          /* 
             Shrink old_top to insert fenceposts, keeping size a
             multiple of MALLOC_ALIGNMENT. We know there is at least
             enough space in old_top to do this.
          */
#ifdef DNMALLOC_DEBUG
          fprintf(stderr, "Adjust top, old_top %p old_size before %lu\n", 
                  old_top, (unsigned long)old_size);
#endif
          old_size = (old_size - 4*SIZE_SZ) & ~MALLOC_ALIGN_MASK;
          set_head(old_top, old_size | PREV_INUSE);
#ifdef DNMALLOC_DEBUG
          fprintf(stderr, "Adjust top, old_size after %lu\n", 
                  (unsigned long)old_size);
#endif
          
          /*
            Note that the following assignments completely overwrite
            old_top when old_size was previously MINSIZE.  This is
            intentional. We need the fencepost, even if old_top otherwise gets
            lost.
          */
          /* dnmalloc, we need the fencepost to be 16 bytes, however since 
             it's marked inuse it will never be coalesced 
          */
          fencepost = cireg_getfree();
          fencepost->chunk = (mchunkptr) chunk_at_offset(chunk(old_top), 
                                                         old_size);
          fencepost->size = 16|INUSE|PREV_INUSE;
          hashtable_add(fencepost);
          /* 
             If possible, release the rest, suppressing trimming.
          */
          if (old_size >= MINSIZE) {
            INTERNAL_SIZE_T tt = av->trim_threshold;
#ifdef DNMALLOC_DEBUG
            fprintf(stderr, "Release\n");
#endif
            av->trim_threshold = (INTERNAL_SIZE_T)(-1);
            set_head(old_top, old_size | PREV_INUSE | INUSE);
            guard_set(av->guard_stored, old_top, 0, old_size);
            VALGRIND_MALLOCLIKE_BLOCK(chunk(old_top), old_size, 0, 0);
            fREe(chunk(old_top));
            av->trim_threshold = tt;
#ifdef DNMALLOC_DEBUG
            fprintf(stderr, "Release done\n");
#endif
          }

#ifdef DNMALLOC_DEBUG
          fprintf(stderr, "Adjust top, size %lu\n", 
                  (unsigned long)chunksize(av->top));
#endif

        } /* fenceposts */
      } /* adjust top */
    } /* not extended previous region */
    
    /* Update statistics */
    sum = av->sbrked_mem;
    if (sum > (CHUNK_SIZE_T)(av->max_sbrked_mem))
      av->max_sbrked_mem = sum;
    
    sum += av->mmapped_mem;
    if (sum > (CHUNK_SIZE_T)(av->max_total_mem))
      av->max_total_mem = sum;

    check_malloc_state();
    
    /* finally, do the allocation */

    p = av->top;
    size = chunksize(p);
    
#ifdef DNMALLOC_DEBUG
    fprintf(stderr, "Size: %lu  nb+MINSIZE: %lu\n", 
            (CHUNK_SIZE_T)(size), (CHUNK_SIZE_T)(nb + MINSIZE));
#endif

    /* check that one of the above allocation paths succeeded */
    if ((CHUNK_SIZE_T)(size) >= (CHUNK_SIZE_T)(nb + MINSIZE)) {
      remainder_size = size - nb;
      remainder = cireg_getfree();
      remainder->chunk = chunk_at_offset(chunk(p), nb);
      av->top = remainder;
      set_head(p, nb | PREV_INUSE | INUSE);
      set_head(remainder, remainder_size | PREV_INUSE);
      hashtable_insert (p, av->top);
      check_malloced_chunk(p, nb);
#ifdef DNMALLOC_DEBUG
      fprintf(stderr, "Return any (total %lu)\n", 
              (unsigned long)/* size_t */av->max_total_mem );
#endif
      return chunk(p);
    }

  }

#ifdef DNMALLOC_DEBUG
  fprintf(stderr, "Return failed (total %lu)\n", 
          (unsigned long)/* size_t */av->max_total_mem );
#endif

  /* catch all failure paths */
  MALLOC_FAILURE_ACTION;
  return 0;
}




/*
  sYSTRIm is an inverse of sorts to sYSMALLOc.  It gives memory back
  to the system (via negative arguments to sbrk) if there is unused
  memory at the `high' end of the malloc pool. It is called
  automatically by free() when top space exceeds the trim
  threshold. It is also called by the public malloc_trim routine.  It
  returns 1 if it actually released any memory, else 0.
*/

#if __STD_C
static int sYSTRIm(size_t pad, mstate av)
#else
static int sYSTRIm(pad, av) size_t pad; mstate av;
#endif
{
  long  top_size;        /* Amount of top-most memory */
  long  extra;           /* Amount to release */
  long  released;        /* Amount actually released */
  char* current_brk;     /* address returned by pre-check sbrk call */
  char* new_brk;         /* address returned by post-check sbrk call */
  size_t pagesz;

  pagesz = av->pagesize;
  top_size = chunksize(av->top);
  
  /* Release in pagesize units, keeping at least one page */
  extra = ((top_size - pad - MINSIZE + (pagesz-1)) / pagesz - 1) * pagesz;
  
  if (extra > 0) {
    
    /*
      Only proceed if end of memory is where we last set it.
      This avoids problems if there were foreign sbrk calls.
    */
    current_brk = (char*)(MORECORE(0));
    if (current_brk == (char*)(av->top) + top_size) {
      
      /*
        Attempt to release memory. We ignore MORECORE return value,
        and instead call again to find out where new end of memory is.
        This avoids problems if first call releases less than we asked,
        of if failure somehow altered brk value. (We could still
        encounter problems if it altered brk in some very bad way,
        but the only thing we can do is adjust anyway, which will cause
        some downstream failure.)
      */
      
      MORECORE(-extra);
      new_brk = (char*)(MORECORE(0));
      
      if (new_brk != (char*)MORECORE_FAILURE) {
        released = (long)(current_brk - new_brk);
        
        if (released != 0) {
          /* Success. Adjust top. */
          av->sbrked_mem -= released;
          set_head(av->top, (top_size - released) | PREV_INUSE);
          check_malloc_state();
          return 1;
        }
      }
    }
  }
  return 0;
}

/*
  ------------------------------ malloc ------------------------------
*/


#if __STD_C
DL_STATIC Void_t* mALLOc(size_t bytes)
#else
DL_STATIC   Void_t* mALLOc(bytes) size_t bytes;
#endif
{
  mstate av = get_malloc_state();

  INTERNAL_SIZE_T nb;               /* normalized request size */
  unsigned int    idx;              /* associated bin index */
  mbinptr         bin;              /* associated bin */
  mfastbinptr*    fb;               /* associated fastbin */

  chunkinfoptr       victim;           /* inspected/selected chunk */
  INTERNAL_SIZE_T size;             /* its size */
  int             victim_index;     /* its bin index */

  chunkinfoptr       remainder;        /* remainder from a split */
  CHUNK_SIZE_T    remainder_size;   /* its size */

  unsigned int    block;            /* bit map traverser */
  unsigned int    bit;              /* bit map traverser */
  unsigned int    map;              /* current word of binmap */

  chunkinfoptr       fwd;              /* misc temp for linking */
  chunkinfoptr       bck;              /* misc temp for linking */
  
  Void_t*         retval;

  /* chunkinfoptr         next; */
 

  /*
    Convert request size to internal form by adding SIZE_SZ bytes
    overhead plus possibly more to obtain necessary alignment and/or
    to obtain a size of at least MINSIZE, the smallest allocatable
    size. Also, checked_request2size traps (returning 0) request sizes
    that are so large that they wrap around zero when padded and
    aligned.
  */
#if defined(SH_CUTEST)
  extern int malloc_count;
  ++malloc_count;
#endif

  checked_request2size(bytes, nb);

  /*
    Bypass search if no frees yet
   */
  if (av && have_anychunks(av)) {
    goto av_initialized;
  }
  else {
    if (!av || av->max_fast == 0) { /* initialization check */
      malloc_consolidate(av);
      av = get_malloc_state();
    }
    goto use_top;
  }

 av_initialized:

  /*
    If the size qualifies as a fastbin, first check corresponding bin.
  */
  if ((CHUNK_SIZE_T)(nb) <= (CHUNK_SIZE_T)(av->max_fast)) {
    fb = &(av->fastbins[(fastbin_index(nb))]);
    if ( (victim = *fb) != 0) {
      *fb = victim->fd;
      check_remalloced_chunk(victim, nb);
      guard_set(av->guard_stored, victim, bytes, nb);
      VALGRIND_MALLOCLIKE_BLOCK(chunk(victim), bytes, 0, 0);
      return chunk(victim);
    }
  }

  /*
    If a small request, check regular bin.  Since these "smallbins"
    hold one size each, no searching within bins is necessary.
    (For a large request, we need to wait until unsorted chunks are
    processed to find best fit. But for small ones, fits are exact
    anyway, so we can check now, which is faster.)
  */

  if (in_smallbin_range(nb)) {
    idx = smallbin_index(nb);
    bin = bin_at(av,idx);

    if ((victim = last(bin)) != bin) {
      bck = victim->bk;
      bin->bk = bck;
      bck->fd = bin;

      set_all_inuse(victim);
            
      check_malloced_chunk(victim, nb);
      guard_set(av->guard_stored, victim, bytes, nb);
      VALGRIND_MALLOCLIKE_BLOCK(chunk(victim), bytes, 0, 0);
      return chunk(victim);
    }
  }

  /* 
     If this is a large request, consolidate fastbins before continuing.
     While it might look excessive to kill all fastbins before
     even seeing if there is space available, this avoids
     fragmentation problems normally associated with fastbins.
     Also, in practice, programs tend to have runs of either small or
     large requests, but less often mixtures, so consolidation is not 
     invoked all that often in most programs. And the programs that
     it is called frequently in otherwise tend to fragment.
  */

  else {
    idx = largebin_index(nb);
    if (have_fastchunks(av)) 
      malloc_consolidate(av);
  }

  /*
    Process recently freed or remaindered chunks, taking one only if
    it is exact fit, or, if this a small request, the chunk is remainder from
    the most recent non-exact fit.  Place other traversed chunks in
    bins.  Note that this step is the only place in any routine where
    chunks are placed in bins.
  */
    
  while ( (victim = unsorted_chunks(av)->bk) != unsorted_chunks(av)) {
    bck = victim->bk;
    size = chunksize(victim);
    
    /* 
       If a small request, try to use last remainder if it is the
       only chunk in unsorted bin.  This helps promote locality for
       runs of consecutive small requests. This is the only
       exception to best-fit, and applies only when there is
       no exact fit for a small chunk.
    */
    
    if (UNLIKELY(in_smallbin_range(nb) && 
                 bck == unsorted_chunks(av) &&
                 victim == av->last_remainder &&
                 (CHUNK_SIZE_T)(size) > (CHUNK_SIZE_T)(nb + MINSIZE))) {
      
      /* split and reattach remainder */
      remainder_size = size - nb;
      remainder = cireg_getfree();
      remainder->chunk = chunk_at_offset(chunk(victim), nb);
      unsorted_chunks(av)->bk = unsorted_chunks(av)->fd = remainder;
      av->last_remainder = remainder; 
      remainder->bk = remainder->fd = unsorted_chunks(av);
      
      set_head(victim, nb | PREV_INUSE|INUSE);
      set_head(remainder, remainder_size | PREV_INUSE);      
      hashtable_insert(victim, remainder);

      check_malloced_chunk(victim, nb);
      guard_set(av->guard_stored, victim, bytes, nb);
      VALGRIND_MALLOCLIKE_BLOCK(chunk(victim), bytes, 0, 0);
      return chunk(victim);
    }
    
    /* remove from unsorted list */
    unsorted_chunks(av)->bk = bck;
    bck->fd = unsorted_chunks(av);
    
    /* Take now instead of binning if exact fit */
    
    if (UNLIKELY(size == nb)) {
      set_all_inuse(victim)
      check_malloced_chunk(victim, nb);
      guard_set(av->guard_stored, victim, bytes, nb);
      VALGRIND_MALLOCLIKE_BLOCK(chunk(victim), bytes, 0, 0);
      return chunk(victim);
    }
    
    /* place chunk in bin */
    
    if (in_smallbin_range(size)) {

      victim_index = smallbin_index(size);
      bck = bin_at(av, victim_index);
      fwd = bck->fd;
    }
    else {
      victim_index = largebin_index(size);
      bck = bin_at(av, victim_index);
      fwd = bck->fd;
      
      if (UNLIKELY(fwd != bck)) {
        /* if smaller than smallest, place first */
        if ((CHUNK_SIZE_T)(size) < (CHUNK_SIZE_T)(bck->bk->size)) {
          fwd = bck;
          bck = bck->bk;
        }
        else if ((CHUNK_SIZE_T)(size) >= 
                 (CHUNK_SIZE_T)(FIRST_SORTED_BIN_SIZE)) {
          
          /* maintain large bins in sorted order */
          size |= PREV_INUSE|INUSE; /* Or with inuse bits to speed comparisons */
          while ((CHUNK_SIZE_T)(size) < (CHUNK_SIZE_T)(fwd->size)) 
            fwd = fwd->fd;
          bck = fwd->bk;
        }
      }
    }

    mark_bin(av, victim_index);
    victim->bk = bck;
    victim->fd = fwd;
    fwd->bk = victim;
    bck->fd = victim;
  }
  
  /*
    If a large request, scan through the chunks of current bin to
    find one that fits.  (This will be the smallest that fits unless
    FIRST_SORTED_BIN_SIZE has been changed from default.)  This is
    the only step where an unbounded number of chunks might be
    scanned without doing anything useful with them. However the
    lists tend to be short.
  */

  if (!in_smallbin_range(nb)) {
    bin = bin_at(av, idx);
    
    victim = last(bin);

    if (UNLIKELY(victim != bin)) {

      do {
        size = chunksize(victim);
      
        if ((CHUNK_SIZE_T)(size) >= (CHUNK_SIZE_T)(nb)) {
          remainder_size = size - nb;
          unlink(victim, bck, fwd);
        
          /* Split */
          if (remainder_size >= MINSIZE) {
            remainder = cireg_getfree();
            remainder->chunk = chunk_at_offset(chunk(victim), nb);
            unsorted_chunks(av)->bk = unsorted_chunks(av)->fd = remainder;
            remainder->bk = remainder->fd = unsorted_chunks(av);
            set_head(victim, nb | PREV_INUSE | INUSE);
            set_head(remainder, remainder_size | PREV_INUSE);
            hashtable_insert(victim, remainder);
            check_malloced_chunk(victim, nb);
            guard_set(av->guard_stored, victim, bytes, nb);
            VALGRIND_MALLOCLIKE_BLOCK(chunk(victim), bytes, 0, 0);
            return chunk(victim);
          } 
          /* Exhaust */
          else  {
            set_all_inuse(victim);
            check_malloced_chunk(victim, nb);
            guard_set(av->guard_stored, victim, bytes, nb);
            VALGRIND_MALLOCLIKE_BLOCK(chunk(victim), bytes, 0, 0);
            return chunk(victim);
          }
        }
        victim = victim->bk;
      } while(victim != bin);
    }
  }

  /*
    Search for a chunk by scanning bins, starting with next largest
    bin. This search is strictly by best-fit; i.e., the smallest
    (with ties going to approximately the least recently used) chunk
    that fits is selected.
    
    The bitmap avoids needing to check that most blocks are nonempty.
  */
    

  ++idx;
  bin = bin_at(av,idx);
  block = idx2block(idx);
  map = av->binmap[block];
  bit = idx2bit(idx);
  
  for (;;) {
    
    /* Skip rest of block if there are no more set bits in this block.  */
    if (bit > map || bit == 0) {
      do {
        if (++block >= BINMAPSIZE)  /* out of bins */
          goto use_top;
      } while ( (map = av->binmap[block]) == 0);
      
      bin = bin_at(av, (block << BINMAPSHIFT));
      bit = 1;
    }
    
    /* Advance to bin with set bit. There must be one. */
    while ((bit & map) == 0) {
      bin = next_bin(bin);
      bit <<= 1;
      assert(bit != 0);
    }
    
    /* Inspect the bin. It is likely to be non-empty */
    victim = last(bin);
    
    /*  If a false alarm (empty bin), clear the bit. */
    if (victim == bin) {
      av->binmap[block] = map &= ~bit; /* Write through */
      bin = next_bin(bin);
      bit <<= 1;
    }
    
    else {
      size = chunksize(victim);
      
      /*  We know the first chunk in this bin is big enough to use. */
      assert((CHUNK_SIZE_T)(size) >= (CHUNK_SIZE_T)(nb));
      
      remainder_size = size - nb;
      
      /* unlink */
      bck = victim->bk;
      bin->bk = bck;
      bck->fd = bin;
      
      /* Split */
      if (remainder_size >= MINSIZE) {
        remainder = cireg_getfree();
        remainder->chunk = chunk_at_offset(chunk(victim), nb);
        
        unsorted_chunks(av)->bk = unsorted_chunks(av)->fd = remainder;
        remainder->bk = remainder->fd = unsorted_chunks(av);
        /* advertise as last remainder */
        if (in_smallbin_range(nb))
          av->last_remainder = remainder;
        
        set_head(victim, nb | PREV_INUSE | INUSE);
        set_head(remainder, remainder_size | PREV_INUSE);
        hashtable_insert(victim, remainder);
        check_malloced_chunk(victim, nb);
        guard_set(av->guard_stored, victim, bytes, nb);
        VALGRIND_MALLOCLIKE_BLOCK(chunk(victim), bytes, 0, 0);
        return chunk(victim);
      }
      /* Exhaust */
      else {
        set_all_inuse(victim);
        check_malloced_chunk(victim, nb);
        guard_set(av->guard_stored, victim, bytes, nb);
        VALGRIND_MALLOCLIKE_BLOCK(chunk(victim), bytes, 0, 0);
        return chunk(victim);
      }
      
    }
  }

  use_top:
   

  /*
    If large enough, split off the chunk bordering the end of memory
    (held in av->top). Note that this is in accord with the best-fit
    search rule.  In effect, av->top is treated as larger (and thus
    less well fitting) than any other available chunk since it can
    be extended to be as large as necessary (up to system
    limitations).
    
    We require that av->top always exists (i.e., has size >=
    MINSIZE) after initialization, so if it would otherwise be
    exhuasted by current request, it is replenished. (The main
    reason for ensuring it exists is that we may need MINSIZE space
    to put in fenceposts in sysmalloc.)
  */
  
  victim = av->top;
  size = chunksize(victim);
  
  if ((CHUNK_SIZE_T)(size) >= (CHUNK_SIZE_T)(nb + MINSIZE)) {
    remainder = cireg_getfree();
    remainder_size = size - nb;
    remainder->chunk = chunk_at_offset(chunk(victim), nb);
    av->top = remainder;
    set_head(victim, nb | PREV_INUSE | INUSE);
    set_head(remainder, remainder_size | PREV_INUSE);
    hashtable_insert(victim, remainder);
    check_malloced_chunk(victim, nb);
    guard_set(av->guard_stored, victim, bytes, nb);
    VALGRIND_MALLOCLIKE_BLOCK(chunk(victim), bytes, 0, 0);
    return chunk(victim);
  }
  
  /* 
     If no space in top, relay to handle system-dependent cases 
  */
  retval = sYSMALLOc(nb, av);
  if (retval) {
#if PARANOIA > 2
    victim = mem2chunk(retval); /* is used in guard_set macro */
#endif
    guard_set(av->guard_stored, victim, bytes, nb);
  }
  VALGRIND_MALLOCLIKE_BLOCK(retval, bytes, 0, 0);
  return retval;
}

/*
  ------------------------------ free ------------------------------
*/

#if __STD_C
DL_STATIC void fREe(Void_t* mem)
#else
DL_STATIC void fREe(mem) Void_t* mem;
#endif
{
  mstate av = get_malloc_state();

  chunkinfoptr       p;           /* chunk corresponding to mem */
  INTERNAL_SIZE_T size;        /* its size */
  mfastbinptr*    fb;          /* associated fastbin */
  chunkinfoptr       prevchunk;   /* previous physical chunk */
  chunkinfoptr       nextchunk;   /* next contiguous chunk */
  INTERNAL_SIZE_T nextsize;    /* its size */
  INTERNAL_SIZE_T prevsize;    /* size of previous contiguous chunk */
  chunkinfoptr       bck;         /* misc temp for linking */
  chunkinfoptr       fwd;         /* misc temp for linking */
  chunkinfoptr       next;
#if defined(SH_CUTEST)
  extern int malloc_count;
  --malloc_count;
#endif

  /* free(0) has no effect */
  if (mem != 0) {
    p = hashtable_lookup(mem);
    /* check that memory is managed by us 
     * and is inuse 
     */
    if (UNLIKELY(!p || !inuse(p))) 
      {
#ifdef DNMALLOC_CHECKS
        if (p) {
          fprintf(stderr, "Attempt to free memory not in use\n");
          abort();
        } else {
          fprintf(stderr, "Attempt to free memory not allocated\n");
          abort();
        }
#endif
        assert(p && inuse(p));
        return;
      }

    VALGRIND_FREELIKE_BLOCK(mem, 0);
 
    guard_check(av->guard_stored, p);
    
    size = chunksize(p);

    check_inuse_chunk(p);

    /*
      If eligible, place chunk on a fastbin so it can be found
      and used quickly in malloc.
    */

    if ((CHUNK_SIZE_T)(size) <= (CHUNK_SIZE_T)(av->max_fast)

#if TRIM_FASTBINS
        /* 
           If TRIM_FASTBINS set, don't place chunks
           bordering top into fastbins
        */
        && (chunk_at_offset(chunk(p), size) != av->top)
#endif
        ) {

      set_fastchunks(av);
      fb = &(av->fastbins[fastbin_index(size)]);
      p->fd = *fb;
      *fb = p;
    }

    /*
       Consolidate other non-mmapped chunks as they arrive.
    */

    else if (!chunk_is_mmapped(p)) {
      set_anychunks(av);

      nextchunk = next_chunkinfo(p);
      if (nextchunk)
        nextsize = chunksize(nextchunk);
      else
        nextsize = 0;/* gcc doesn't notice that it's only used if (nextchunk)*/

      /* consolidate backward */
      if (UNLIKELY(!prev_inuse(p))) {
        prevchunk = prev_chunkinfo(p);
        prevsize = chunksize(prevchunk);
#ifdef DNMALLOC_CHECKS
        if (inuse(prevchunk)) {
                fprintf(stderr, "Dnmalloc error: trying to unlink an inuse chunk: %p (chunk: %p)\n This is definitely a bug, please report it to dnmalloc@fort-knox.org.\n", prevchunk, chunk(prevchunk));
                abort();
        }
#else
        assert(!inuse(prevchunk));
#endif
        size += prevsize;
        unlink(prevchunk, bck, fwd);
        set_head(p, size | PREV_INUSE);
        hashtable_skiprm(prevchunk,p);
        /* This chunk no longer exists in any form: release the chunkinfoptr 
         */
        freecilst_add(p);
        p = prevchunk;
      }

      if (nextchunk) {
        if (nextchunk != av->top) {
          /* get and clear inuse bit */
          clear_previnuse(nextchunk);
          
          /* consolidate forward */
          if (!inuse(nextchunk)) {
            /* if mmap is used instead of sbrk, we may have a
             * chunk with !nextchunk->fd && !nextchunk->bk
             */
            if (nextchunk->fd && nextchunk->fd)
              {
                unlink(nextchunk, bck, fwd);
                size += nextsize;
                set_head(p, size | PREV_INUSE);
                hashtable_skiprm(p, nextchunk);
                freecilst_add (nextchunk);
              }
          }
          
          set_head(p, size | PREV_INUSE);
          next = next_chunkinfo(p);
          if (next)
            next->prev_size = size;
          
          /*
            Place the chunk in unsorted chunk list. Chunks are
            not placed into regular bins until after they have
            been given one chance to be used in malloc.
          */
          
          bck = unsorted_chunks(av);
          fwd = bck->fd;
          p->bk = bck;
          p->fd = fwd;
          bck->fd = p;
          fwd->bk = p;
          
          nextchunk = next_chunkinfo(p);
          if (nextchunk)
            nextchunk->prev_size = chunksize(p);        
          
          check_free_chunk(p);
        }
        
        /*
          If the chunk borders the current high end of memory,
          consolidate into top
        */
        
        else {
          size += nextsize;
          set_head(p, size | PREV_INUSE);
          hashtable_remove(chunk(av->top));
          freecilst_add(av->top);
          av->top = p;
          check_chunk(p);
        }
      } /* if (nextchunk) */

      /*
        If freeing a large space, consolidate possibly-surrounding
        chunks. Then, if the total unused topmost memory exceeds trim
        threshold, ask malloc_trim to reduce top.

        Unless max_fast is 0, we don't know if there are fastbins
        bordering top, so we cannot tell for sure whether threshold
        has been reached unless fastbins are consolidated.  But we
        don't want to consolidate on each free.  As a compromise,
        consolidation is performed if FASTBIN_CONSOLIDATION_THRESHOLD
        is reached.
      */

      if (UNLIKELY((CHUNK_SIZE_T)(size) >= FASTBIN_CONSOLIDATION_THRESHOLD)) { 
        if (have_fastchunks(av)) 
          malloc_consolidate(av);

#ifndef MORECORE_CANNOT_TRIM        
        if ((CHUNK_SIZE_T)(chunksize(av->top)) >= 
            (CHUNK_SIZE_T)(av->trim_threshold))
          {
            if (morecore32bit(av))
              {
#ifdef DNMALLOC_DEBUG
                fprintf(stderr, "Calling systrim from free()\n");
#endif
                sYSTRIm(av->top_pad, av);
#ifdef DNMALLOC_DEBUG
                fprintf(stderr, "Systrim done\n");
#endif
              }
          }
#endif
      }

    }
    /*
      If the chunk was allocated via mmap, release via munmap()
      Note that if HAVE_MMAP is false but chunk_is_mmapped is
      true, then user must have overwritten memory. There's nothing
      we can do to catch this error unless DEBUG is set, in which case
      check_inuse_chunk (above) will have triggered error.
    */

    else {
#if  PARANOIA > 0
      int ret;
#endif
      INTERNAL_SIZE_T offset = (INTERNAL_SIZE_T) p->hash_next;
      av->n_mmaps--;
      av->mmapped_mem -= (size + offset);
#if  PARANOIA > 0
      ret = munmap((char*) chunk(p) - offset, size + offset);
#else
      munmap((char*) chunk(p) - offset, size + offset);
#endif
      hashtable_remove_mmapped(chunk(p));
      freecilst_add(p);
      /* munmap returns non-zero on failure */
      assert(ret == 0);
    }
  }
}

/*
  ------------------------- malloc_consolidate -------------------------

  malloc_consolidate is a specialized version of free() that tears
  down chunks held in fastbins.  Free itself cannot be used for this
  purpose since, among other things, it might place chunks back onto
  fastbins.  So, instead, we need to use a minor variant of the same
  code.
  
  Also, because this routine needs to be called the first time through
  malloc anyway, it turns out to be the perfect place to trigger
  initialization code.
*/

#if __STD_C
static void malloc_consolidate(mstate av)
#else
static void malloc_consolidate(av) mstate av;
#endif
{
  mfastbinptr*    fb;                 /* current fastbin being consolidated */
  mfastbinptr*    maxfb;              /* last fastbin (for loop control) */
  chunkinfoptr       p;                  /* current chunk being consolidated */
  chunkinfoptr       nextp;              /* next chunk to consolidate */
  chunkinfoptr       prevp;
  chunkinfoptr       unsorted_bin;       /* bin header */
  chunkinfoptr       first_unsorted;     /* chunk to link to */

  /* These have same use as in free() */
  chunkinfoptr       nextchunk;
  INTERNAL_SIZE_T size;
  INTERNAL_SIZE_T nextsize;
  INTERNAL_SIZE_T prevsize;
  chunkinfoptr       bck;
  chunkinfoptr       fwd;
  chunkinfoptr       next;
 
  /*
    If max_fast is 0, we know that av hasn't
    yet been initialized, in which case do so below
  */
  if (av && av->max_fast != 0) {

    clear_fastchunks(av);

    unsorted_bin = unsorted_chunks(av);

    /*
      Remove each chunk from fast bin and consolidate it, placing it
      then in unsorted bin. Among other reasons for doing this,
      placing in unsorted bin avoids needing to calculate actual bins
      until malloc is sure that chunks aren't immediately going to be
      reused anyway.
    */
    
    maxfb = &(av->fastbins[fastbin_index(av->max_fast)]);
    fb = &(av->fastbins[0]);
    do {
      if ( UNLIKELY((p = *fb) != 0)) {
        *fb = 0;
        do {
          check_inuse_chunk(p);
          nextp = p->fd;
          
          /*
           * Slightly streamlined version of consolidation code in free() 
           */

          size = chunksize(p);
          nextchunk = next_chunkinfo(p);

          /* gcc doesn't notice that it's only used if (nextchunk) */
          if (nextchunk)
            nextsize = chunksize(nextchunk);
          else
            nextsize = 0; 
          
          if (!prev_inuse(p)) {
             prevp = prev_chunkinfo(p);
             prevsize = chunksize(prevp);
             size += prevsize;
#ifdef DNMALLOC_CHECKS
             if (inuse(prevp)) {
                fprintf(stderr, "Dnmalloc error: trying to unlink an inuse chunk (2): %p (chunk: %p)\n This is definitely a bug, please report it to dnmalloc@fort-knox.org.\n", prevp, chunk(prevp));
                     abort();
             }
#else
             assert(!inuse(prevp));
#endif
             unlink(prevp, bck, fwd);
             set_head(p, size | PREV_INUSE);         
             hashtable_skiprm(prevp,p);
             freecilst_add(p);
             p=prevp;
          }
          
          if (nextchunk) {
            if (nextchunk != av->top) {

              clear_previnuse(nextchunk);
            
              /* if mmap is used instead of sbrk, we may have a
               * chunk with !nextchunk->fd && !nextchunk->bk
               */
              if (!inuse(nextchunk)) {
                if( nextchunk->fd && nextchunk->bk) {
                  size += nextsize;
                  unlink(nextchunk, bck, fwd);
                  set_head(p, size | PREV_INUSE);
                  hashtable_skiprm(p,nextchunk);
                  freecilst_add(nextchunk);
                }
              }
              
              first_unsorted = unsorted_bin->fd;
              unsorted_bin->fd = p;
              first_unsorted->bk = p;
              
              set_head(p, size | PREV_INUSE);
              p->bk = unsorted_bin;
              p->fd = first_unsorted;
              next = next_chunkinfo(p);
              if (next)
                next->prev_size = size;

            
            }
          
            else if (nextchunk == av->top) {
              size += nextsize;
              set_head(p, size | PREV_INUSE);
              hashtable_remove(chunk(av->top));
              freecilst_add(av->top);
              av->top = p;
            }
          } /* if (nextchunk) */
          
        } while ( (p = nextp) != 0);
        
      }
    } while (fb++ != maxfb);
  }
  else {
    /* Initialize dnmalloc */
    dnmalloc_init();
    malloc_init_state(get_malloc_state());
    check_malloc_state();
  }
}

/*
  ------------------------------ realloc ------------------------------
*/


#if __STD_C
DL_STATIC Void_t* rEALLOc(Void_t* oldmem, size_t bytes)
#else
DL_STATIC Void_t* rEALLOc(oldmem, bytes) Void_t* oldmem; size_t bytes;
#endif
{
  mstate av = get_malloc_state();

  INTERNAL_SIZE_T  nb;              /* padded request size */

  chunkinfoptr        oldp;            /* chunk corresponding to oldmem */
  INTERNAL_SIZE_T  oldsize;         /* its size */

  chunkinfoptr        newp;            /* chunk to return */
  INTERNAL_SIZE_T  newsize;         /* its size */
  Void_t*          newmem;          /* corresponding user mem */

  chunkinfoptr        next;            /* next contiguous chunk after oldp */

  chunkinfoptr        remainder;       /* extra space at end of newp */
  CHUNK_SIZE_T     remainder_size;  /* its size */

  chunkinfoptr        bck;             /* misc temp for linking */
  chunkinfoptr        fwd;             /* misc temp for linking */

  CHUNK_SIZE_T     copysize;        /* bytes to copy */
  unsigned int     ncopies;         /* INTERNAL_SIZE_T words to copy */
  INTERNAL_SIZE_T* s;               /* copy source */ 
  INTERNAL_SIZE_T* d;               /* copy destination */

  
#ifdef REALLOC_ZERO_BYTES_FREES
  if (UNLIKELY(bytes == 0)) {
    fREe(oldmem);
    return 0;
  }
#endif

  if (UNLIKELY(!av || av->max_fast == 0)) {
    malloc_consolidate(av);
    av = get_malloc_state();
  }

  /* realloc of null is supposed to be same as malloc */
  if (UNLIKELY(oldmem == 0)) 
    return mALLOc(bytes);

  checked_request2size(bytes, nb);

  oldp    = hashtable_lookup(oldmem);
  
  if (UNLIKELY(!oldp || !inuse(oldp))){ 
     /* attempt to either realloc memory not managed by us 
      * or memory that is not in use 
      */
#ifdef DNMALLOC_CHECKS
    if (oldp) {
      fprintf(stderr, "Attempt to free memory not in use\n");
      abort();
    } else {
      fprintf(stderr, "Attempt to free memory not allocated\n");
      abort();
    }
#endif
    assert(oldp && inuse(oldp));
    return 0;     
  }

  VALGRIND_FREELIKE_BLOCK(oldmem, 0);
  guard_check(av->guard_stored, oldp);

  oldsize = chunksize(oldp);

  check_inuse_chunk(oldp);

  if (!chunk_is_mmapped(oldp)) {

    if (UNLIKELY((CHUNK_SIZE_T)(oldsize) >= (CHUNK_SIZE_T)(nb))) {
      /* already big enough; split below */
      newp    = oldp;
      newsize = oldsize;
    }

    else {
      next = next_chunkinfo(oldp);
      if (next)
        next->prev_size = oldsize;
      /* Try to expand forward into top */
      if (next && next == av->top &&
          (CHUNK_SIZE_T)(newsize = oldsize + chunksize(next)) >=
          (CHUNK_SIZE_T)(nb + MINSIZE)) {
         set_head_size(oldp, nb);
         hashtable_remove(chunk(av->top));
         av->top->chunk = chunk_at_offset(chunk(oldp), nb);
         set_head(av->top, (newsize - nb) | PREV_INUSE);
         /* av->top->chunk has been moved move in hashtable */
         hashtable_insert(oldp, av->top);
         guard_set(av->guard_stored, oldp, bytes, nb);
         VALGRIND_MALLOCLIKE_BLOCK(chunk(oldp), bytes, 0, 0); 
         return chunk(oldp);
      }
      
      /* Try to expand forward into next chunk;  split off remainder below */
      else if (next && next != av->top && 
               !inuse(next) &&
               (CHUNK_SIZE_T)(newsize = oldsize + chunksize(next)) >=
               (CHUNK_SIZE_T)(nb)) {
        newp = oldp;
        unlink(next, bck, fwd);
        hashtable_remove(chunk(next));
        freecilst_add(next);
        next = next_chunkinfo(oldp);
        if (next)
          next->prev_size = newsize;
      }

      /* allocate, copy, free */
      else {

        newmem = mALLOc(nb - MALLOC_ALIGN_MASK);
         if (newmem == 0)
          return 0; /* propagate failure */

        newp = hashtable_lookup(newmem);
        newsize = chunksize(newp);
        
 
        /* next = next_chunkinfo(oldp); *//* 'next' never used rw 19.05.2008 */
        /*
          Avoid copy if newp is next chunk after oldp.
        */
        if (UNLIKELY(is_next_chunk(oldp, newp))) {
          newsize += oldsize;
          set_head_size(oldp, newsize);
          hashtable_skiprm(oldp, newp);
          freecilst_add(newp);    
          newp = oldp;
        }
        else {
          /*
            Unroll copy of <= 40 bytes (80 if 8byte sizes)
            We know that contents have an even number of
            INTERNAL_SIZE_T-sized words; minimally 4 (2 on amd64).
          */
          
          VALGRIND_MALLOCLIKE_BLOCK(chunk(oldp), chunksize(oldp), 0, 0);

          copysize = oldsize;
          s = (INTERNAL_SIZE_T*)(oldmem);
          d = (INTERNAL_SIZE_T*)(newmem);
          ncopies = copysize / sizeof(INTERNAL_SIZE_T);
          assert(ncopies >= 2);
          
          if (ncopies > 10)
            MALLOC_COPY(d, s, copysize);
          
          else {
            *(d+0) = *(s+0);
            *(d+1) = *(s+1);
            if (ncopies > 2) {
              *(d+2) = *(s+2);
              *(d+3) = *(s+3);
              if (ncopies > 4) {
                *(d+4) = *(s+4);
                *(d+5) = *(s+5);
                if (ncopies > 6) {
                  *(d+6) = *(s+6);
                  *(d+7) = *(s+7);
                  if (ncopies > 8) {
                    *(d+8) = *(s+8);
                    *(d+9) = *(s+9);
                  }
                }
              }
            }
          }
          
          fREe(oldmem);
          check_inuse_chunk(newp);
          guard_set(av->guard_stored, newp, bytes, nb);
          return chunk(newp);
        }
      }
    }

    /* If possible, free extra space in old or extended chunk */

    assert((CHUNK_SIZE_T)(newsize) >= (CHUNK_SIZE_T)(nb));

    remainder_size = newsize - nb;

    if (remainder_size >= MINSIZE) { /* split remainder */
      remainder = cireg_getfree();
      remainder->chunk = chunk_at_offset(chunk(newp), nb);
      set_head_size(newp, nb);
      set_head(remainder, remainder_size | PREV_INUSE | INUSE);
      remainder->prev_size = nb;
      hashtable_insert(newp, remainder);
      /* Mark remainder as inuse so free() won't complain */
      set_all_inuse(remainder);
      guard_set(av->guard_stored, remainder, 0, remainder_size);
      VALGRIND_MALLOCLIKE_BLOCK(chunk(remainder), remainder_size, 0, 0);
      fREe(chunk(remainder)); 
    }
    else { /* not enough extra to split off */
      set_head_size(newp, newsize);
      set_all_inuse(newp);
    }

    check_inuse_chunk(newp);
    guard_set(av->guard_stored, newp, bytes, nb);
    VALGRIND_MALLOCLIKE_BLOCK(chunk(newp), bytes, 0, 0);
    return chunk(newp);
  }

  /*
    Handle mmap cases
  */

  else {

#if HAVE_MREMAP
    INTERNAL_SIZE_T offset = (INTERNAL_SIZE_T) oldp->hash_next;
    size_t pagemask = av->pagesize - 1;
    char *cp;
    CHUNK_SIZE_T  sum;
    

    /* Note the extra SIZE_SZ overhead */
    /* newsize = (nb + offset + SIZE_SZ + pagemask) & ~pagemask; */
    newsize = (nb + offset + pagemask) & ~pagemask;

    /* don't need to remap if still within same page */
    if (oldsize == newsize - offset)
      {
        guard_set(av->guard_stored, oldp, bytes, nb);
        VALGRIND_FREELIKE_BLOCK(oldmem, 0);
        VALGRIND_MALLOCLIKE_BLOCK(oldmem, bytes, 0, 0);
        return oldmem;
      }

    cp = (char*)mremap((char*)chunk(oldp) - offset, oldsize + offset, newsize, 1);
    
    if (cp != (char*)MORECORE_FAILURE) {
       
      hashtable_remove_mmapped(chunk(oldp));
       
      oldp->chunk = (mchunkptr)(cp + offset);
      set_head(oldp, (newsize - offset)|IS_MMAPPED|INUSE);
      
      hashtable_add(oldp);
      
      assert(aligned_OK(chunk(oldp))); /* rw fix: newp -> oldp */
      assert(( ((INTERNAL_SIZE_T) oldp->hash_next) == offset));
      
      /* update statistics */
      sum = av->mmapped_mem += newsize - oldsize;
      if (sum > (CHUNK_SIZE_T)(av->max_mmapped_mem)) 
        av->max_mmapped_mem = sum;
      sum += av->sbrked_mem;
      if (sum > (CHUNK_SIZE_T)(av->max_total_mem)) 
        av->max_total_mem = sum;
      
      guard_set(av->guard_stored, oldp, bytes, nb);
      VALGRIND_FREELIKE_BLOCK(oldmem, 0);
      VALGRIND_MALLOCLIKE_BLOCK(chunk(oldp), bytes, 0, 0);
      return chunk(oldp);
    }
#endif /* have MREMAP */

    /* Note the extra SIZE_SZ overhead. */
    if ((CHUNK_SIZE_T)(oldsize) >= (CHUNK_SIZE_T)(nb + SIZE_SZ)) 
      newmem = oldmem; /* do nothing */
    else {
      /* Must alloc, copy, free. */
      newmem = mALLOc(nb - MALLOC_ALIGN_MASK);
      if (newmem != 0) {
        MALLOC_COPY(newmem, oldmem, oldsize);
        fREe(oldmem);
      }
    }
    VALGRIND_MALLOCLIKE_BLOCK(newmem, bytes, 0, 0);
    guard_set(av->guard_stored, mem2chunk(newmem), bytes, nb);
    return newmem;
  }
}

/*
  ---------------------------posix_memalign ----------------------------
*/

#if __STD_C
DL_STATIC int posix_mEMALIGn(Void_t** memptr, size_t alignment, size_t bytes)
#else
DL_STATIC int posix_mEMALIGn(memptr, alignment, bytes) Void_t** memptr; size_t alignment; size_t bytes;
#endif
{
  mstate av;

  if (alignment % sizeof(void *) != 0)
    return EINVAL;
  if ((alignment & (alignment - 1)) != 0)
    return EINVAL;

  av = get_malloc_state();
  if (!av || av->max_fast == 0) malloc_consolidate(av);
  *memptr =  mEMALIGn(alignment, bytes);

  return (*memptr != NULL ? 0 : ENOMEM);
}

/*
  ------------------------------ memalign ------------------------------
*/

#if __STD_C
DL_STATIC Void_t* mEMALIGn(size_t alignment, size_t bytes)
#else
DL_STATIC Void_t* mEMALIGn(alignment, bytes) size_t alignment; size_t bytes;
#endif
{
  INTERNAL_SIZE_T nb;             /* padded  request size */
  char*           m;              /* memory returned by malloc call */
  chunkinfoptr       p;              /* corresponding chunk */
  char*           brk;            /* alignment point within p */
  chunkinfoptr       newp;           /* chunk to return */
  INTERNAL_SIZE_T newsize;        /* its size */
  INTERNAL_SIZE_T leadsize;       /* leading space before alignment point */
  chunkinfoptr       remainder;      /* spare room at end to split off */
  CHUNK_SIZE_T    remainder_size; /* its size */
  INTERNAL_SIZE_T size;
#if PARANOIA > 2
  mstate          av;
#endif

  /* If need less alignment than we give anyway, just relay to malloc */

  if (UNLIKELY(alignment <= MALLOC_ALIGNMENT)) return mALLOc(bytes);

  /* Otherwise, ensure that it is at least a minimum chunk size */

  if (alignment <  MINSIZE) alignment = MINSIZE;

  /* Make sure alignment is power of 2 (in case MINSIZE is not).  */
  if (UNLIKELY((alignment & (alignment - 1)) != 0)) {
    size_t a = MALLOC_ALIGNMENT * 2;
    while ((CHUNK_SIZE_T)a < (CHUNK_SIZE_T)alignment) a <<= 1;
    alignment = a;
  }

  checked_request2size(bytes, nb);

  /*
    Strategy: find a spot within that chunk that meets the alignment
    request, and then possibly free the leading and trailing space.
  */


  /* Call malloc with worst case padding to hit alignment. */

  m  = (char*)(mALLOc(nb + alignment + MINSIZE));

  if (m == 0) return 0; /* propagate failure */

#if PARANOIA > 2
  av = get_malloc_state();
#endif

  p = hashtable_lookup((mchunkptr) m);

  if ((((PTR_UINT)(m)) % alignment) != 0) { /* misaligned */

    /*
      Find an aligned spot inside chunk.  Since we need to give back
      leading space in a chunk of at least MINSIZE, if the first
      calculation places us at a spot with less than MINSIZE leader,
      we can move to the next aligned spot -- we've allocated enough
      total room so that this is always possible.
    */

    brk = (char*) ((PTR_UINT)(((PTR_UINT)(m + alignment - 1)) &
                           -((signed long) alignment)));
    if ((CHUNK_SIZE_T)(brk - (char*)(chunk(p))) < MINSIZE)
      brk += alignment;

    newp = cireg_getfree();
    newp->chunk = (mchunkptr)brk;
    leadsize = brk - (char*)(chunk(p));
    newsize = chunksize(p) - leadsize;

    /* For mmapped chunks, just adjust offset */
    if (UNLIKELY(chunk_is_mmapped(p))) {
      newp->hash_next = (chunkinfoptr) (((INTERNAL_SIZE_T) p->hash_next) + leadsize);
      set_head(newp, newsize|IS_MMAPPED|INUSE);
      hashtable_remove_mmapped(chunk(p));
      freecilst_add(p);
      hashtable_add(newp);
      guard_set(av->guard_stored, newp, bytes, nb);
      return chunk(newp);
    }

    /* Otherwise, give back leader, use the rest */
    set_head(newp, newsize | PREV_INUSE | INUSE);
    set_head_size(p, leadsize);
    set_all_inuse(newp);
    hashtable_add(newp); /* 20.05.2008 rw */
    guard_set(av->guard_stored, p, 0, leadsize);
    fREe(chunk(p));
    p = newp;

    assert (newsize >= nb &&
            (((PTR_UINT)(chunk(p))) % alignment) == 0);
  }

  /* Also give back spare room at the end */
  if (!chunk_is_mmapped(p)) {
    size = chunksize(p);
    if ((CHUNK_SIZE_T)(size) > (CHUNK_SIZE_T)(nb + MINSIZE)) {
       remainder = cireg_getfree();
       remainder_size = size - nb;
       remainder->chunk = chunk_at_offset(chunk(p), nb);
       set_head(remainder, remainder_size | PREV_INUSE | INUSE);
       set_head_size(p, nb);
       hashtable_add(remainder); /* 20.05.2008 rw */
       guard_set(av->guard_stored, remainder, 0, remainder_size);
       fREe(chunk(remainder));
    }
  }

  check_inuse_chunk(p);
  guard_set(av->guard_stored, p, bytes, nb);
  return chunk(p);
}

/*
  ------------------------------ calloc ------------------------------
*/

#if __STD_C
DL_STATIC Void_t* cALLOc(size_t n_elements, size_t elem_size)
#else
DL_STATIC Void_t* cALLOc(n_elements, elem_size) size_t n_elements; size_t elem_size;
#endif
{
  chunkinfoptr p;
  CHUNK_SIZE_T  clearsize;
  CHUNK_SIZE_T  nclears;
  INTERNAL_SIZE_T* d;
  Void_t* mem;
 
  
  mem = mALLOc(n_elements * elem_size);

  if (mem != 0) {
    p = hashtable_lookup(mem);

    if (!chunk_is_mmapped(p))
    {  
      /*
        Unroll clear of <= 40 bytes (80 if 8byte sizes)
        We know that contents have an even number of
        INTERNAL_SIZE_T-sized words; minimally 4 (2 on amd64).
      */

      d = (INTERNAL_SIZE_T*)mem;
      clearsize = chunksize(p);
      nclears = clearsize / sizeof(INTERNAL_SIZE_T);
      assert(nclears >= 2);

      if (nclears > 10) {
        MALLOC_ZERO(d, clearsize);
      }

      else {
        *(d+0) = 0;
        *(d+1) = 0;
        if (nclears > 2) {
          *(d+2) = 0;
          *(d+3) = 0;
          if (nclears > 4) {
            *(d+4) = 0;
            *(d+5) = 0;
            if (nclears > 6) {
              *(d+6) = 0;
              *(d+7) = 0;
              if (nclears > 8) {
                *(d+8) = 0;
                *(d+9) = 0;
              }
            }
          }
        }
      }
    }
#if ! MMAP_CLEARS
    else
    {
      d = (INTERNAL_SIZE_T*)mem;
      clearsize = chunksize(p);
      MALLOC_ZERO(d, clearsize);
    }
#endif
    /* Set guard again, since we just cleared it
     */
    guard_set(get_malloc_state()->guard_stored, p, (n_elements * elem_size), p->size);
  }

  return mem;
}

/*
  ------------------------------ valloc ------------------------------
*/

#if __STD_C
DL_STATIC Void_t* vALLOc(size_t bytes)
#else
DL_STATIC Void_t* vALLOc(bytes) size_t bytes;
#endif
{
  /* Ensure initialization */
  mstate av = get_malloc_state();
  if (!av || av->max_fast == 0) {
    malloc_consolidate(av);
    av = get_malloc_state();
  }
  return mEMALIGn(av->pagesize, bytes);
}

/*
  ------------------------------ pvalloc ------------------------------
*/


#if __STD_C
DL_STATIC Void_t* pVALLOc(size_t bytes)
#else
DL_STATIC Void_t* pVALLOc(bytes) size_t bytes;
#endif
{
  mstate av = get_malloc_state();
  size_t pagesz;

  /* Ensure initialization */
  if (!av || av->max_fast == 0) {
    malloc_consolidate(av);
    av = get_malloc_state();
  }
  pagesz = av->pagesize;
  return mEMALIGn(pagesz, (bytes + pagesz - 1) & ~(pagesz - 1));
}
   

/*
  ------------------------------ malloc_trim ------------------------------
*/

#if __STD_C
DL_STATIC int mTRIm(size_t pad)
#else
DL_STATIC int mTRIm(pad) size_t pad;
#endif
{
  mstate av = get_malloc_state();
  /* Ensure initialization/consolidation */
  malloc_consolidate(av);
  av = get_malloc_state();
#ifndef MORECORE_CANNOT_TRIM
  if (morecore32bit(av))
    return sYSTRIm(pad, av);
  else
    return 0;
#else
  return 0;
#endif
}



/*
  ------------------------- malloc_usable_size -------------------------
*/

#if __STD_C
DL_STATIC size_t mUSABLe(Void_t* mem)
#else
DL_STATIC size_t mUSABLe(mem) Void_t* mem;
#endif
{
  chunkinfoptr p;
  if (mem != 0) {
    p = hashtable_lookup(mem);
    if (p && inuse(p)) return chunksize(p);
  }
  return 0;
}

/*
  ------------------------------ mallinfo ------------------------------
*/

DL_STATIC struct mallinfo mALLINFo()
{
  mstate av = get_malloc_state();
  static struct mallinfo mi;
  unsigned int i;
  mbinptr b;
  chunkinfoptr p;
  INTERNAL_SIZE_T avail;
  INTERNAL_SIZE_T fastavail;
  int nblocks;
  int nfastblocks;

  /* Ensure initialization */
  if (!av || av->top == 0) {
    malloc_consolidate(av);
    av = get_malloc_state();
  }
  check_malloc_state();

  if (!av || av->top == 0) {
    return mi;
  }
  
  /* Account for top */
  avail = chunksize(av->top);
  nblocks = 1;  /* top always exists */

  /* traverse fastbins */
  nfastblocks = 0;
  fastavail = 0;

  for (i = 0; i < NFASTBINS; ++i) {
    for (p = av->fastbins[i]; p != 0; p = p->fd) {
      ++nfastblocks;
      fastavail += chunksize(p);
    }
  }

  avail += fastavail;

  /* traverse regular bins */
  for (i = 1; i < NBINS; ++i) {
    b = bin_at(av, i);
    for (p = last(b); p != b; p = p->bk) {
      ++nblocks;
      avail += chunksize(p);
    }
  }

  mi.smblks = nfastblocks;
  mi.ordblks = nblocks;
  mi.fordblks = avail;
  mi.uordblks = av->sbrked_mem - avail;
  mi.arena = av->sbrked_mem;
  mi.hblks = av->n_mmaps;
  mi.hblkhd = av->mmapped_mem;
  mi.fsmblks = fastavail;
  mi.keepcost = chunksize(av->top);
  mi.usmblks = av->max_total_mem;
  return mi;
}

/*
  ------------------------------ malloc_stats ------------------------------
*/

DL_STATIC void mSTATs()
{
  struct mallinfo mi = mALLINFo();

  fprintf(stderr, "hashtable = %10lu MB\n", 
          (CHUNK_SIZE_T)(HASHTABLESIZE / (1024*1024)));
  fprintf(stderr, "max system bytes = %10lu\n",
          (CHUNK_SIZE_T)(mi.usmblks));
  fprintf(stderr, "system bytes     = %10lu  (%10lu sbrked, %10lu mmaped)\n",
          (CHUNK_SIZE_T)(mi.arena + mi.hblkhd),
          (CHUNK_SIZE_T)(mi.arena),
          (CHUNK_SIZE_T)(mi.hblkhd));
  fprintf(stderr, "in use bytes     = %10lu\n",
          (CHUNK_SIZE_T)(mi.uordblks + mi.hblkhd));
}


/*
  ------------------------------ mallopt ------------------------------
*/

#if __STD_C
DL_STATIC int mALLOPt(int param_number, int value)
#else
DL_STATIC int mALLOPt(param_number, value) int param_number; int value;
#endif
{
  mstate av = get_malloc_state();
  /* Ensure initialization/consolidation */
  malloc_consolidate(av);
  av = get_malloc_state();

  switch(param_number) {
  case M_MXFAST:
    if (value >= 0 && value <= MAX_FAST_SIZE) {
      set_max_fast(av, value);
      return 1;
    }
    else
      return 0;

  case M_TRIM_THRESHOLD:
    av->trim_threshold = value;
    return 1;

  case M_TOP_PAD:
    av->top_pad = value;
    return 1;

  case M_MMAP_THRESHOLD:
    av->mmap_threshold = value;
    return 1;

  case M_MMAP_MAX:
    if (value != 0)
      return 0;
    av->n_mmaps_max = value;
    return 1;

  default:
    return 0;
  }
}


/*      $OpenBSD: arc4random.c,v 1.19 2008/06/04 00:50:23 djm Exp $     */

/*
 * Copyright (c) 1996, David Mazieres <dm@uun.org>
 * Copyright (c) 2008, Damien Miller <djm@openbsd.org>
 *
 * Permission to use, copy, modify, and distribute this software for any
 * purpose with or without fee is hereby granted, provided that the above
 * copyright notice and this permission notice appear in all copies.
 *
 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 */

/*
 * Arc4 random number generator for OpenBSD.
 *
 * This code is derived from section 17.1 of Applied Cryptography,
 * second edition, which describes a stream cipher allegedly
 * compatible with RSA Labs "RC4" cipher (the actual description of
 * which is a trade secret).  The same algorithm is used as a stream
 * cipher called "arcfour" in Tatu Ylonen's ssh package.
 *
 * Here the stream cipher has been modified always to include the time
 * when initializing the state.  That makes it impossible to
 * regenerate the same random sequence twice, so this can't be used
 * for encryption, but will generate good random numbers.
 *
 * RC4 is a registered trademark of RSA Laboratories.
 */

/* Moved u_int8_t -> unsigned char (portability)
 * Eliminated unneeded functions, added read from /dev/urandom taken from:
 $MirOS: contrib/code/Snippets/arc4random.c,v 1.3 2008-03-04 22:53:14 tg Exp $
 * Modified by Robert Connolly from OpenBSD lib/libc/crypt/arc4random.c v1.11.
 * This is arc4random(3) using urandom.
 */

#include <fcntl.h>
#include <limits.h>
#include <stdlib.h>
#include <sys/param.h>
#include <sys/time.h>

struct arc4_stream {
        unsigned char i;
        unsigned char j;
        unsigned char s[256];
};

static int rs_initialized;
static struct arc4_stream rs;
static pid_t arc4_stir_pid;
static int arc4_count;

static unsigned char arc4_getbyte(void);

static void
arc4_init(void)
{
        int     n;

        for (n = 0; n < 256; n++)
                rs.s[n] = n;
        rs.i = 0;
        rs.j = 0;
}

static inline void
arc4_addrandom(unsigned char *dat, int datlen)
{
        int     n;
        unsigned char si;

        rs.i--;
        for (n = 0; n < 256; n++) {
                rs.i = (rs.i + 1);
                si = rs.s[rs.i];
                rs.j = (rs.j + si + dat[n % datlen]);
                rs.s[rs.i] = rs.s[rs.j];
                rs.s[rs.j] = si;
        }
        rs.j = rs.i;
}

#ifdef HAVE_SCHED_H
#include <sched.h>
#endif

static void
arc4_stir(void)
{
        int     i;
        struct {
                struct timeval tv1;
                struct timeval tv2;
                unsigned char rnd[(128 - 2*sizeof(struct timeval)) / sizeof(unsigned char)];
        } rdat;
#if !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__NetBSD__)
        ssize_t sz = 0;
        int    fd;
#endif
 
        gettimeofday(&rdat.tv1, NULL);


        if (!rs_initialized) {
                arc4_init();
                rs_initialized = 1;
        }

#if !defined(__FreeBSD__) && !defined(__OpenBSD__) && !defined(__NetBSD__)

#ifdef HAVE_SCHED_YIELD
        /* Yield the processor to introduce some random delay. */
        (void) sched_yield();
#endif

        /*
         * Pthread problem in multithreaded code on *BSD.
         */
        fd = open("/dev/urandom", O_RDONLY);
        if (fd != -1) {
                sz = (size_t)read(fd, rdat.rnd, sizeof (rdat.rnd));
                /* 
                 * gcc complains if we ignore the return value of read(), and
                 * the llvm/clang analyzer complains if we don't use it...
                 */
                if (sz > (-256)) /* always true */
                  close(fd);
        }
        /*
        if (sz > sizeof (rdat.rnd))
                sz = 0;
        */
 #endif

        arc4_stir_pid = getpid();
        gettimeofday(&rdat.tv2, NULL);

        arc4_addrandom((void *)&rdat, sizeof(rdat));

        /*
         * Discard early keystream, as per recommendations in:
         * http://www.wisdom.weizmann.ac.il/~itsik/RC4/Papers/Rc4_ksa.ps
         */
        for (i = 0; i < 256; i++)
                (void)arc4_getbyte();
        arc4_count = 1600000;
}

static unsigned char
arc4_getbyte(void)
{
        unsigned char si, sj;

        rs.i = (rs.i + 1);
        si = rs.s[rs.i];
        rs.j = (rs.j + si);
        sj = rs.s[rs.j];
        rs.s[rs.i] = sj;
        rs.s[rs.j] = si;
        return (rs.s[(si + sj) & 0xff]);
}


 /* Changed to return char* */
static char *
dnmalloc_arc4random(void)
{
        static char val[4];
        
        /* We only call this once, hence no need for locking. */

        /* _ARC4_LOCK(); */
        arc4_count -= 4;
        if (arc4_count <= 0 || !rs_initialized || arc4_stir_pid != getpid())
                arc4_stir();

        val[0] = (char) arc4_getbyte();
        val[1] = (char) arc4_getbyte();
        val[2] = (char) arc4_getbyte();
        val[3] = (char) arc4_getbyte();

        arc4_stir();
        /* _ARC4_UNLOCK(); */
        return val;
}

#else
int dnmalloc_pthread_init() { return 0; }
#endif /* ! USE_SYSTEM_MALLOC */