pthread_support.c

linux下建立JAVA虚拟机的源码KAFFE

/* 
 * Copyright (c) 1994 by Xerox Corporation.  All rights reserved.
 * Copyright (c) 1996 by Silicon Graphics.  All rights reserved.
 * Copyright (c) 1998 by Fergus Henderson.  All rights reserved.
 * Copyright (c) 2000-2004 by Hewlett-Packard Company.  All rights reserved.
 *
 * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY EXPRESSED
 * OR IMPLIED.  ANY USE IS AT YOUR OWN RISK.
 *
 * Permission is hereby granted to use or copy this program
 * for any purpose,  provided the above notices are retained on all copies.
 * Permission to modify the code and to distribute modified code is granted,
 * provided the above notices are retained, and a notice that the code was
 * modified is included with the above copyright notice.
 */
/*
 * Support code for LinuxThreads, the clone()-based kernel
 * thread package for Linux which is included in libc6.
 *
 * This code relies on implementation details of LinuxThreads,
 * (i.e. properties not guaranteed by the Pthread standard),
 * though this version now does less of that than the other Pthreads
 * support code.
 *
 * Note that there is a lot of code duplication between linux_threads.c
 * and thread support for some of the other Posix platforms; any changes
 * made here may need to be reflected there too.
 */
 /* DG/UX ix86 support <[e-mail]> */
/*
 * Linux_threads.c now also includes some code to support HPUX and
 * OSF1 (Compaq Tru64 Unix, really).  The OSF1 support is based on Eric Benson's
 * patch.
 *
 * Eric also suggested an alternate basis for a lock implementation in
 * his code:
 * + #elif defined(OSF1)
 * +    unsigned long GC_allocate_lock = 0;
 * +    msemaphore GC_allocate_semaphore;
 * + #  define GC_TRY_LOCK() \
 * +    ((msem_lock(&GC_allocate_semaphore, MSEM_IF_NOWAIT) == 0) \
 * +     ? (GC_allocate_lock = 1) \
 * +     : 0)
 * + #  define GC_LOCK_TAKEN GC_allocate_lock
 */

/*#define DEBUG_THREADS 1*/
/*#define GC_ASSERTIONS*/

# include "private/pthread_support.h"

# if defined(GC_PTHREADS) && !defined(GC_SOLARIS_THREADS) \
     && !defined(GC_WIN32_THREADS)

# if defined(GC_HPUX_THREADS) && !defined(USE_PTHREAD_SPECIFIC) \
     && !defined(USE_COMPILER_TLS)
#   ifdef __GNUC__
#     define USE_PTHREAD_SPECIFIC
      /* Empirically, as of gcc 3.3, USE_COMPILER_TLS doesn't work.	*/
#   else
#     define USE_COMPILER_TLS
#   endif
# endif

# if defined USE_HPUX_TLS
    --> Macro replaced by USE_COMPILER_TLS
# endif

# if (defined(GC_DGUX386_THREADS) || defined(GC_OSF1_THREADS) || \
      defined(GC_DARWIN_THREADS) || defined(GC_AIX_THREADS)) \
      && !defined(USE_PTHREAD_SPECIFIC)
#   define USE_PTHREAD_SPECIFIC
# endif

# if defined(GC_DGUX386_THREADS) && !defined(_POSIX4A_DRAFT10_SOURCE)
#   define _POSIX4A_DRAFT10_SOURCE 1
# endif

# if defined(GC_DGUX386_THREADS) && !defined(_USING_POSIX4A_DRAFT10)
#   define _USING_POSIX4A_DRAFT10 1
# endif

# ifdef THREAD_LOCAL_ALLOC
#   if !defined(USE_PTHREAD_SPECIFIC) && !defined(USE_COMPILER_TLS)
#     include "private/specific.h"
#   endif
#   if defined(USE_PTHREAD_SPECIFIC)
#     define GC_getspecific pthread_getspecific
#     define GC_setspecific pthread_setspecific
#     define GC_key_create pthread_key_create
      typedef pthread_key_t GC_key_t;
#   endif
#   if defined(USE_COMPILER_TLS)
#     define GC_getspecific(x) (x)
#     define GC_setspecific(key, v) ((key) = (v), 0)
#     define GC_key_create(key, d) 0
      typedef void * GC_key_t;
#   endif
# endif
# include <stdlib.h>
# include <pthread.h>
# include <sched.h>
# include <time.h>
# include <errno.h>
# include <unistd.h>
# include <sys/mman.h>
# include <sys/time.h>
# include <sys/types.h>
# include <sys/stat.h>
# include <fcntl.h>
# include <signal.h>

#if defined(GC_DARWIN_THREADS)
# include "private/darwin_semaphore.h"
#else
# include <semaphore.h>
#endif /* !GC_DARWIN_THREADS */

#if defined(GC_DARWIN_THREADS) || defined(GC_FREEBSD_THREADS)
# include <sys/sysctl.h>
#endif /* GC_DARWIN_THREADS */



#if defined(GC_DGUX386_THREADS)
# include <sys/dg_sys_info.h>
# include <sys/_int_psem.h>
  /* sem_t is an uint in DG/UX */
  typedef unsigned int  sem_t;
#endif /* GC_DGUX386_THREADS */

#ifndef __GNUC__
#   define __inline__
#endif

#ifdef GC_USE_LD_WRAP
#   define WRAP_FUNC(f) __wrap_##f
#   define REAL_FUNC(f) __real_##f
#else
#   define WRAP_FUNC(f) GC_##f
#   if !defined(GC_DGUX386_THREADS)
#     define REAL_FUNC(f) f
#   else /* GC_DGUX386_THREADS */
#     define REAL_FUNC(f) __d10_##f
#   endif /* GC_DGUX386_THREADS */
#   undef pthread_create
#   if !defined(GC_DARWIN_THREADS)
#     undef pthread_sigmask
#   endif
#   undef pthread_join
#   undef pthread_detach
#   if defined(GC_OSF1_THREADS) && defined(_PTHREAD_USE_MANGLED_NAMES_) \
       && !defined(_PTHREAD_USE_PTDNAM_)
/* Restore the original mangled names on Tru64 UNIX.  */
#     define pthread_create __pthread_create
#     define pthread_join __pthread_join
#     define pthread_detach __pthread_detach
#   endif
#endif

void GC_thr_init();

static GC_bool parallel_initialized = FALSE;

void GC_init_parallel();

# if defined(THREAD_LOCAL_ALLOC) && !defined(DBG_HDRS_ALL)

/* We don't really support thread-local allocation with DBG_HDRS_ALL */

#ifdef USE_COMPILER_TLS
  __thread
#endif
GC_key_t GC_thread_key;

static GC_bool keys_initialized;

/* Recover the contents of the freelist array fl into the global one gfl.*/
/* Note that the indexing scheme differs, in that gfl has finer size	*/
/* resolution, even if not all entries are used.			*/
/* We hold the allocator lock.						*/
static void return_freelists(ptr_t *fl, ptr_t *gfl)
{
    int i;
    ptr_t q, *qptr;
    size_t nwords;

    for (i = 1; i < NFREELISTS; ++i) {
	nwords = i * (GRANULARITY/sizeof(word));
        qptr = fl + i;	
	q = *qptr;
	if ((word)q >= HBLKSIZE) {
	  if (gfl[nwords] == 0) {
	    gfl[nwords] = q;
	  } else {
	    /* Concatenate: */
	    for (; (word)q >= HBLKSIZE; qptr = &(obj_link(q)), q = *qptr);
	    GC_ASSERT(0 == q);
	    *qptr = gfl[nwords];
	    gfl[nwords] = fl[i];
	  }
	}
	/* Clear fl[i], since the thread structure may hang around.	*/
	/* Do it in a way that is likely to trap if we access it.	*/
	fl[i] = (ptr_t)HBLKSIZE;
    }
}

/* We statically allocate a single "size 0" object. It is linked to	*/
/* itself, and is thus repeatedly reused for all size 0 allocation	*/
/* requests.  (Size 0 gcj allocation requests are incorrect, and	*/
/* we arrange for those to fault asap.)					*/
static ptr_t size_zero_object = (ptr_t)(&size_zero_object);

/* Each thread structure must be initialized.	*/
/* This call must be made from the new thread.	*/
/* Caller holds allocation lock.		*/
void GC_init_thread_local(GC_thread p)
{
    int i;

    if (!keys_initialized) {
	if (0 != GC_key_create(&GC_thread_key, 0)) {
	    ABORT("Failed to create key for local allocator");
        }
	keys_initialized = TRUE;
    }
    if (0 != GC_setspecific(GC_thread_key, p)) {
	ABORT("Failed to set thread specific allocation pointers");
    }
    for (i = 1; i < NFREELISTS; ++i) {
	p -> ptrfree_freelists[i] = (ptr_t)1;
	p -> normal_freelists[i] = (ptr_t)1;
#	ifdef GC_GCJ_SUPPORT
	  p -> gcj_freelists[i] = (ptr_t)1;
#	endif
    }   
    /* Set up the size 0 free lists.	*/
    p -> ptrfree_freelists[0] = (ptr_t)(&size_zero_object);
    p -> normal_freelists[0] = (ptr_t)(&size_zero_object);
#   ifdef GC_GCJ_SUPPORT
        p -> gcj_freelists[0] = (ptr_t)(-1);
#   endif
}

#ifdef GC_GCJ_SUPPORT
  extern ptr_t * GC_gcjobjfreelist;
#endif

/* We hold the allocator lock.	*/
void GC_destroy_thread_local(GC_thread p)
{
    /* We currently only do this from the thread itself or from	*/
    /* the fork handler for a child process.			*/
#   ifndef HANDLE_FORK
      GC_ASSERT(GC_getspecific(GC_thread_key) == (void *)p);
#   endif
    return_freelists(p -> ptrfree_freelists, GC_aobjfreelist);
    return_freelists(p -> normal_freelists, GC_objfreelist);
#   ifdef GC_GCJ_SUPPORT
   	return_freelists(p -> gcj_freelists, GC_gcjobjfreelist);
#   endif
}

extern GC_PTR GC_generic_malloc_many();

GC_PTR GC_local_malloc(size_t bytes)
{
    if (EXPECT(!SMALL_ENOUGH(bytes),0)) {
        return(GC_malloc(bytes));
    } else {
	int index = INDEX_FROM_BYTES(bytes);
	ptr_t * my_fl;
	ptr_t my_entry;
#	if defined(REDIRECT_MALLOC) && !defined(USE_PTHREAD_SPECIFIC)
	GC_key_t k = GC_thread_key;
#	endif
	void * tsd;

#	if defined(REDIRECT_MALLOC) && !defined(USE_PTHREAD_SPECIFIC)
	    if (EXPECT(0 == k, 0)) {
		/* This can happen if we get called when the world is	*/
		/* being initialized.  Whether we can actually complete	*/
		/* the initialization then is unclear.			*/
		GC_init_parallel();
		k = GC_thread_key;
	    }
#	endif
	tsd = GC_getspecific(GC_thread_key);
#	ifdef GC_ASSERTIONS
	  LOCK();
	  GC_ASSERT(tsd == (void *)GC_lookup_thread(pthread_self()));
	  UNLOCK();
#	endif
	my_fl = ((GC_thread)tsd) -> normal_freelists + index;
	my_entry = *my_fl;
	if (EXPECT((word)my_entry >= HBLKSIZE, 1)) {
	    ptr_t next = obj_link(my_entry);
	    GC_PTR result = (GC_PTR)my_entry;
	    *my_fl = next;
	    obj_link(my_entry) = 0;
	    PREFETCH_FOR_WRITE(next);
	    return result;
	} else if ((word)my_entry - 1 < DIRECT_GRANULES) {
	    *my_fl = my_entry + index + 1;
            return GC_malloc(bytes);
	} else {
	    GC_generic_malloc_many(BYTES_FROM_INDEX(index), NORMAL, my_fl);
	    if (*my_fl == 0) return GC_oom_fn(bytes);
	    return GC_local_malloc(bytes);
	}
    }
}

GC_PTR GC_local_malloc_atomic(size_t bytes)
{
    if (EXPECT(!SMALL_ENOUGH(bytes), 0)) {
        return(GC_malloc_atomic(bytes));
    } else {
	int index = INDEX_FROM_BYTES(bytes);
	ptr_t * my_fl = ((GC_thread)GC_getspecific(GC_thread_key))
		        -> ptrfree_freelists + index;
	ptr_t my_entry = *my_fl;
    
	if (EXPECT((word)my_entry >= HBLKSIZE, 1)) {
	    GC_PTR result = (GC_PTR)my_entry;
	    *my_fl = obj_link(my_entry);
	    return result;
	} else if ((word)my_entry - 1 < DIRECT_GRANULES) {
	    *my_fl = my_entry + index + 1;
        return GC_malloc_atomic(bytes);
	} else {
	    GC_generic_malloc_many(BYTES_FROM_INDEX(index), PTRFREE, my_fl);
	    /* *my_fl is updated while the collector is excluded;	*/
	    /* the free list is always visible to the collector as 	*/
	    /* such.							*/
	    if (*my_fl == 0) return GC_oom_fn(bytes);
	    return GC_local_malloc_atomic(bytes);
	}
    }
}

#ifdef GC_GCJ_SUPPORT

#include "include/gc_gcj.h"

#ifdef GC_ASSERTIONS
  extern GC_bool GC_gcj_malloc_initialized;
#endif

extern int GC_gcj_kind;

GC_PTR GC_local_gcj_malloc(size_t bytes,
			   void * ptr_to_struct_containing_descr)
{
    GC_ASSERT(GC_gcj_malloc_initialized);
    if (EXPECT(!SMALL_ENOUGH(bytes), 0)) {
        return GC_gcj_malloc(bytes, ptr_to_struct_containing_descr);
    } else {
	int index = INDEX_FROM_BYTES(bytes);
	ptr_t * my_fl = ((GC_thread)GC_getspecific(GC_thread_key))
	                -> gcj_freelists + index;
	ptr_t my_entry = *my_fl;
	if (EXPECT((word)my_entry >= HBLKSIZE, 1)) {
	    GC_PTR result = (GC_PTR)my_entry;
	    GC_ASSERT(!GC_incremental);
	    /* We assert that any concurrent marker will stop us.	*/
	    /* Thus it is impossible for a mark procedure to see the 	*/
	    /* allocation of the next object, but to see this object 	*/
	    /* still containing a free list pointer.  Otherwise the 	*/
	    /* marker might find a random "mark descriptor".		*/
	    *(volatile ptr_t *)my_fl = obj_link(my_entry);
	    /* We must update the freelist before we store the pointer.	*/
	    /* Otherwise a GC at this point would see a corrupted	*/
	    /* free list.						*/
	    /* A memory barrier is probably never needed, since the 	*/
	    /* action of stopping this thread will cause prior writes	*/
	    /* to complete.						*/
	    GC_ASSERT(((void * volatile *)result)[1] == 0); 
	    *(void * volatile *)result = ptr_to_struct_containing_descr; 
	    return result;
	} else if ((word)my_entry - 1 < DIRECT_GRANULES) {
	    if (!GC_incremental) *my_fl = my_entry + index + 1;
	    	/* In the incremental case, we always have to take this */
	    	/* path.  Thus we leave the counter alone.		*/
            return GC_gcj_malloc(bytes, ptr_to_struct_containing_descr);
	} else {
	    GC_generic_malloc_many(BYTES_FROM_INDEX(index), GC_gcj_kind, my_fl);
	    if (*my_fl == 0) return GC_oom_fn(bytes);
	    return GC_local_gcj_malloc(bytes, ptr_to_struct_containing_descr);
	}
    }
}

#endif /* GC_GCJ_SUPPORT */

# else  /* !THREAD_LOCAL_ALLOC  && !DBG_HDRS_ALL */

#   define GC_destroy_thread_local(t)