fastmalloc.cpp

linux下开源浏览器WebKit的源码,市面上的很多商用浏览器都是移植自WebKit

// Copyright (c) 2005, 2007, Google Inc.
// All rights reserved.
// Copyright (C) 2005, 2006, 2007, 2008 Apple Inc. All rights reserved.
// 
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
// 
//     * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
//     * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
//     * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
// 
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

// ---
// Author: Sanjay Ghemawat <[e-mail]>
//
// A malloc that uses a per-thread cache to satisfy small malloc requests.
// (The time for malloc/free of a small object drops from 300 ns to 50 ns.)
//
// See doc/tcmalloc.html for a high-level
// description of how this malloc works.
//
// SYNCHRONIZATION
//  1. The thread-specific lists are accessed without acquiring any locks.
//     This is safe because each such list is only accessed by one thread.
//  2. We have a lock per central free-list, and hold it while manipulating
//     the central free list for a particular size.
//  3. The central page allocator is protected by "pageheap_lock".
//  4. The pagemap (which maps from page-number to descriptor),
//     can be read without holding any locks, and written while holding
//     the "pageheap_lock".
//  5. To improve performance, a subset of the information one can get
//     from the pagemap is cached in a data structure, pagemap_cache_,
//     that atomically reads and writes its entries.  This cache can be
//     read and written without locking.
//
//     This multi-threaded access to the pagemap is safe for fairly
//     subtle reasons.  We basically assume that when an object X is
//     allocated by thread A and deallocated by thread B, there must
//     have been appropriate synchronization in the handoff of object
//     X from thread A to thread B.  The same logic applies to pagemap_cache_.
//
// THE PAGEID-TO-SIZECLASS CACHE
// Hot PageID-to-sizeclass mappings are held by pagemap_cache_.  If this cache
// returns 0 for a particular PageID then that means "no information," not that
// the sizeclass is 0.  The cache may have stale information for pages that do
// not hold the beginning of any free()'able object.  Staleness is eliminated
// in Populate() for pages with sizeclass > 0 objects, and in do_malloc() and
// do_memalign() for all other relevant pages.
//
// TODO: Bias reclamation to larger addresses
// TODO: implement mallinfo/mallopt
// TODO: Better testing
//
// 9/28/2003 (new page-level allocator replaces ptmalloc2):
// * malloc/free of small objects goes from ~300 ns to ~50 ns.
// * allocation of a reasonably complicated struct
//   goes from about 1100 ns to about 300 ns.

#include "config.h"
#include "FastMalloc.h"

#include "Assertions.h"
#if ENABLE(JSC_MULTIPLE_THREADS)
#include <pthread.h>
#endif

#ifndef NO_TCMALLOC_SAMPLES
#ifdef WTF_CHANGES
#define NO_TCMALLOC_SAMPLES
#endif
#endif

#if !defined(USE_SYSTEM_MALLOC) && defined(NDEBUG)
#define FORCE_SYSTEM_MALLOC 0
#else
#define FORCE_SYSTEM_MALLOC 1
#endif

#define TCMALLOC_TRACK_DECOMMITED_SPANS (HAVE(VIRTUALALLOC))

#ifndef NDEBUG
namespace WTF {

#if ENABLE(JSC_MULTIPLE_THREADS)
static pthread_key_t isForbiddenKey;
static pthread_once_t isForbiddenKeyOnce = PTHREAD_ONCE_INIT;
static void initializeIsForbiddenKey()
{
  pthread_key_create(&isForbiddenKey, 0);
}

static bool isForbidden()
{
    pthread_once(&isForbiddenKeyOnce, initializeIsForbiddenKey);
    return !!pthread_getspecific(isForbiddenKey);
}

void fastMallocForbid()
{
    pthread_once(&isForbiddenKeyOnce, initializeIsForbiddenKey);
    pthread_setspecific(isForbiddenKey, &isForbiddenKey);
}

void fastMallocAllow()
{
    pthread_once(&isForbiddenKeyOnce, initializeIsForbiddenKey);
    pthread_setspecific(isForbiddenKey, 0);
}

#else

static bool staticIsForbidden;
static bool isForbidden()
{
    return staticIsForbidden;
}

void fastMallocForbid()
{
    staticIsForbidden = true;
}

void fastMallocAllow()
{
    staticIsForbidden = false;
}
#endif // ENABLE(JSC_MULTIPLE_THREADS)

} // namespace WTF
#endif // NDEBUG

#include <string.h>

namespace WTF {

void* fastZeroedMalloc(size_t n) 
{
    void* result = fastMalloc(n);
    memset(result, 0, n);
    return result;
}
    
void* tryFastZeroedMalloc(size_t n) 
{
    void* result = tryFastMalloc(n);
    if (!result)
        return 0;
    memset(result, 0, n);
    return result;
}

} // namespace WTF

#if FORCE_SYSTEM_MALLOC

#include <stdlib.h>
#if !PLATFORM(WIN_OS)
    #include <pthread.h>
#else
    #include "windows.h"
#endif

namespace WTF {

void* tryFastMalloc(size_t n) 
{
    ASSERT(!isForbidden());
    return malloc(n);
}

void* fastMalloc(size_t n) 
{
    ASSERT(!isForbidden());
    void* result = malloc(n);
    if (!result)
        CRASH();
    return result;
}

void* tryFastCalloc(size_t n_elements, size_t element_size)
{
    ASSERT(!isForbidden());
    return calloc(n_elements, element_size);
}

void* fastCalloc(size_t n_elements, size_t element_size)
{
    ASSERT(!isForbidden());
    void* result = calloc(n_elements, element_size);
    if (!result)
        CRASH();
    return result;
}

void fastFree(void* p)
{
    ASSERT(!isForbidden());
    free(p);
}

void* tryFastRealloc(void* p, size_t n)
{
    ASSERT(!isForbidden());
    return realloc(p, n);
}

void* fastRealloc(void* p, size_t n)
{
    ASSERT(!isForbidden());
    void* result = realloc(p, n);
    if (!result)
        CRASH();
    return result;
}

void releaseFastMallocFreeMemory() { }
    
FastMallocStatistics fastMallocStatistics()
{
    FastMallocStatistics statistics = { 0, 0, 0, 0 };
    return statistics;
}

} // namespace WTF

#if PLATFORM(DARWIN)
// This symbol is present in the JavaScriptCore exports file even when FastMalloc is disabled.
// It will never be used in this case, so it's type and value are less interesting than its presence.
extern "C" const int jscore_fastmalloc_introspection = 0;
#endif

#else // FORCE_SYSTEM_MALLOC

#if HAVE(STDINT_H)
#include <stdint.h>
#elif HAVE(INTTYPES_H)
#include <inttypes.h>
#else
#include <sys/types.h>
#endif

#include "AlwaysInline.h"
#include "Assertions.h"
#include "TCPackedCache.h"
#include "TCPageMap.h"
#include "TCSpinLock.h"
#include "TCSystemAlloc.h"
#include <algorithm>
#include <errno.h>
#include <new>
#include <pthread.h>
#include <stdarg.h>
#include <stddef.h>
#include <stdio.h>
#if COMPILER(MSVC)
#ifndef WIN32_LEAN_AND_MEAN
#define WIN32_LEAN_AND_MEAN
#endif
#include <windows.h>
#endif

#if WTF_CHANGES

#if PLATFORM(DARWIN)
#include "MallocZoneSupport.h"
#include <wtf/HashSet.h>
#endif

#ifndef PRIuS
#define PRIuS "zu"
#endif

// Calling pthread_getspecific through a global function pointer is faster than a normal
// call to the function on Mac OS X, and it's used in performance-critical code. So we
// use a function pointer. But that's not necessarily faster on other platforms, and we had
// problems with this technique on Windows, so we'll do this only on Mac OS X.
#if PLATFORM(DARWIN)
static void* (*pthread_getspecific_function_pointer)(pthread_key_t) = pthread_getspecific;
#define pthread_getspecific(key) pthread_getspecific_function_pointer(key)
#endif

#define DEFINE_VARIABLE(type, name, value, meaning) \
  namespace FLAG__namespace_do_not_use_directly_use_DECLARE_##type##_instead {  \
  type FLAGS_##name(value);                                \
  char FLAGS_no##name;                                                        \
  }                                                                           \
  using FLAG__namespace_do_not_use_directly_use_DECLARE_##type##_instead::FLAGS_##name
  
#define DEFINE_int64(name, value, meaning) \
  DEFINE_VARIABLE(int64_t, name, value, meaning)
  
#define DEFINE_double(name, value, meaning) \
  DEFINE_VARIABLE(double, name, value, meaning)

namespace WTF {

#define malloc fastMalloc
#define calloc fastCalloc
#define free fastFree
#define realloc fastRealloc

#define MESSAGE LOG_ERROR
#define CHECK_CONDITION ASSERT

#if PLATFORM(DARWIN)
class Span;
class TCMalloc_Central_FreeListPadded;
class TCMalloc_PageHeap;
class TCMalloc_ThreadCache;
template <typename T> class PageHeapAllocator;

class FastMallocZone {
public:
    static void init();

    static kern_return_t enumerate(task_t, void*, unsigned typeMmask, vm_address_t zoneAddress, memory_reader_t, vm_range_recorder_t);
    static size_t goodSize(malloc_zone_t*, size_t size) { return size; }
    static boolean_t check(malloc_zone_t*) { return true; }
    static void  print(malloc_zone_t*, boolean_t) { }
    static void log(malloc_zone_t*, void*) { }
    static void forceLock(malloc_zone_t*) { }
    static void forceUnlock(malloc_zone_t*) { }
    static void statistics(malloc_zone_t*, malloc_statistics_t* stats) { memset(stats, 0, sizeof(malloc_statistics_t)); }

private:
    FastMallocZone(TCMalloc_PageHeap*, TCMalloc_ThreadCache**, TCMalloc_Central_FreeListPadded*, PageHeapAllocator<Span>*, PageHeapAllocator<TCMalloc_ThreadCache>*);
    static size_t size(malloc_zone_t*, const void*);
    static void* zoneMalloc(malloc_zone_t*, size_t);
    static void* zoneCalloc(malloc_zone_t*, size_t numItems, size_t size);
    static void zoneFree(malloc_zone_t*, void*);
    static void* zoneRealloc(malloc_zone_t*, void*, size_t);
    static void* zoneValloc(malloc_zone_t*, size_t) { LOG_ERROR("valloc is not supported"); return 0; }
    static void zoneDestroy(malloc_zone_t*) { }

    malloc_zone_t m_zone;
    TCMalloc_PageHeap* m_pageHeap;
    TCMalloc_ThreadCache** m_threadHeaps;
    TCMalloc_Central_FreeListPadded* m_centralCaches;
    PageHeapAllocator<Span>* m_spanAllocator;
    PageHeapAllocator<TCMalloc_ThreadCache>* m_pageHeapAllocator;
};

#endif

#endif

#ifndef WTF_CHANGES
// This #ifdef should almost never be set.  Set NO_TCMALLOC_SAMPLES if
// you're porting to a system where you really can't get a stacktrace.
#ifdef NO_TCMALLOC_SAMPLES
// We use #define so code compiles even if you #include stacktrace.h somehow.
# define GetStackTrace(stack, depth, skip)  (0)
#else
# include <google/stacktrace.h>
#endif
#endif

// Even if we have support for thread-local storage in the compiler
// and linker, the OS may not support it.  We need to check that at
// runtime.  Right now, we have to keep a manual set of "bad" OSes.
#if defined(HAVE_TLS)
  static bool kernel_supports_tls = false;      // be conservative
  static inline bool KernelSupportsTLS() {
    return kernel_supports_tls;
  }
# if !HAVE_DECL_UNAME   // if too old for uname, probably too old for TLS
    static void CheckIfKernelSupportsTLS() {
      kernel_supports_tls = false;
    }
# else
#   include <sys/utsname.h>    // DECL_UNAME checked for <sys/utsname.h> too
    static void CheckIfKernelSupportsTLS() {
      struct utsname buf;
      if (uname(&buf) != 0) {   // should be impossible
        MESSAGE("uname failed assuming no TLS support (errno=%d)\n", errno);
        kernel_supports_tls = false;
      } else if (strcasecmp(buf.sysname, "linux") == 0) {
        // The linux case: the first kernel to support TLS was 2.6.0
        if (buf.release[0] < '2' && buf.release[1] == '.')    // 0.x or 1.x
          kernel_supports_tls = false;
        else if (buf.release[0] == '2' && buf.release[1] == '.' &&
                 buf.release[2] >= '0' && buf.release[2] < '6' &&
                 buf.release[3] == '.')                       // 2.0 - 2.5
          kernel_supports_tls = false;
        else