使用 Boost::interprocess 在共享内存中映射 <int, void*>

Question

我正在尝试在以下类型的共享内存中构建一个映射

我这样创建共享内存区域：

 managed_shared_memory segment(create_only ,"MyMap"  ,size);       

ShMemAllocator_t alloc_inst (segment.get_segment_manager());

 map =   segment.construct<MyMap_t>("MyMap")      
                             (std::less<int>() 
                             ,alloc_inst); 

地图中的值如下：

       typedef pair<MutexType, boost::interprocess::offset_ptr<void> > ValueType ; 

MutexType 本身就是一个包含读写互斥锁的结构（使用 read_lock 和 write_lock）； 定义如下：

typedef struct  mutex_struct{ 
   sharable_lock<interprocess_mutex> read_lock(interprocess_mutex, defer_lock); 
  scoped_lock<interprocess_mutex> write_lock(interprocess_mutex, defer_lock); 
} MutexType;

“size”是map的总大小（以对象计，即所有void指针指向的数据大小之和）。

如何确保这个 void* 数据也位于我创建的这个内存段中，如何在现有的共享内存区域中实例化它）。这样做的原因是我只想分配这个大缓冲区一次但重复地删除/添加对象（映射模型缓存）我还没有找到一种方法可以在同一个内存段中分配多个对象地图内。此外，寻求分配 MutexType 对会返回一个编译错误，指出没有提供“调用”运算符。

Answer 1

你基本上已经在那里了。调用您在共享内存SecondValue_t 中分配的任何对象类型。代替ShMemAllocator_t，定义一个不同的进程间分配器类型，比如SecondValueAllocator_t，用于分配SecondValue_t 对象。每当您想将ValueType 对象插入映射时，ValueType 对象的第二个值将分配给SecondValueAllocator_t 实例。

这是一个完整的例子，部分使用my answer中的代码为Interprocess reader/writer lock with Boost：

#include <cstdlib>
#include <functional>
#include <iostream>
#include <string>
#include <utility>

#include <boost/scope_exit.hpp>
#include <boost/interprocess/managed_shared_memory.hpp>
#include <boost/interprocess/allocators/allocator.hpp>
#include <boost/interprocess/allocators/private_node_allocator.hpp>
#include <boost/interprocess/containers/map.hpp>
#include <boost/interprocess/sync/interprocess_upgradable_mutex.hpp>
#include <boost/interprocess/sync/scoped_lock.hpp>
#include <boost/interprocess/sync/sharable_lock.hpp>
#include <boost/interprocess/sync/upgradable_lock.hpp>

#define SHARED_MEMORY_NAME "SO13783012-MyMap"

// https://stackoverflow.com/questions/13783012/map-of-int-void-in-shared-memory-using-boostinterprocess

using namespace boost::interprocess;

typedef int SecondValue_t;
typedef allocator<SecondValue_t, managed_shared_memory::segment_manager> SecondValueAllocator_t;

typedef struct mutex_struct {
    //...
} MutexType;

typedef std::pair<MutexType, SecondValueAllocator_t::pointer> ValueType;

typedef map<int, ValueType>::value_type MyMapValueType;
typedef allocator<MyMapValueType, managed_shared_memory::segment_manager> MyMapEntryAllocator_t;
typedef map<int, ValueType, std::less<int>, MyMapEntryAllocator_t> MyMap_t;

struct shared_data {
private:
    typedef boost::interprocess::interprocess_upgradable_mutex upgradable_mutex_type;

    mutable upgradable_mutex_type mutex;
    MyMap_t my_map;

public:
    shared_data(const MyMapEntryAllocator_t& alloc)
        : my_map(MyMap_t::key_compare(), alloc)
    {
    }

    // Tries to get the mapped value for the given key `k'. If successful, the mapped value is
    // copied into `out' and `true' is returned. Otherwise, returns `false' and does not modify
    // `out'.
    bool try_get(MyMap_t::mapped_type& out, MyMap_t::key_type k) const {
        boost::interprocess::sharable_lock<upgradable_mutex_type> lock(mutex);
        MyMap_t::const_iterator pos = my_map.find(k);
        if (pos != my_map.end()) {
            out = pos->second;
            return true;
        }
        return false;
    }

    void put(MyMap_t::key_type k, MyMap_t::mapped_type v) {
        boost::interprocess::scoped_lock<upgradable_mutex_type> lock(mutex);
        my_map.insert(MyMap_t::value_type(my_map.size(), v));
    }
};

int main(int argc, char *argv[])
{
    if (argc != 2) {
        std::cerr << "Usage: " << argv[0] << " WHICH\n";
        return EXIT_FAILURE;
    }

    const std::string which = argv[1];

    if (which == "parent") {
        shared_memory_object::remove(SHARED_MEMORY_NAME);
        BOOST_SCOPE_EXIT(argc) {
            shared_memory_object::remove(SHARED_MEMORY_NAME);
        } BOOST_SCOPE_EXIT_END;
        managed_shared_memory shm(create_only, SHARED_MEMORY_NAME, 65536);

        MyMapEntryAllocator_t entry_alloc(shm.get_segment_manager());
        shared_data& d = *shm.construct<shared_data>("theSharedData")(entry_alloc);

        SecondValueAllocator_t second_value_alloc(shm.get_segment_manager());

        // Insert some test data.
        SecondValueAllocator_t::pointer p;
        p = second_value_alloc.allocate(1);
        second_value_alloc.construct(p, -3);
        d.put(0, std::make_pair(MutexType(), p));
        p = second_value_alloc.allocate(1);
        second_value_alloc.construct(p, 70);
        d.put(1, std::make_pair(MutexType(), p));
        p = second_value_alloc.allocate(1);
        second_value_alloc.construct(p, -18);
        d.put(2, std::make_pair(MutexType(), p));
        p = second_value_alloc.allocate(1);
        second_value_alloc.construct(p, 44);
        d.put(3, std::make_pair(MutexType(), p));
        p = second_value_alloc.allocate(1);
        second_value_alloc.construct(p, 0);
        d.put(4, std::make_pair(MutexType(), p));

        // Go to sleep for a minute - gives us a chance to start a child process.
        sleep(60);
    } else {
        managed_shared_memory shm(open_only, SHARED_MEMORY_NAME);
        std::pair<shared_data *, std::size_t> find_res = shm.find<shared_data>("theSharedData");
        if (!find_res.first) {
            std::cerr << "Failed to find `theSharedData'.\n";
            return EXIT_FAILURE;
        }
        shared_data& d = *find_res.first;

        MyMap_t::mapped_type v;
        int i = 0;
        for (; d.try_get(v, i); ++i) {
            std::cout << i << ": " << *v.second << '\n';
        }

        // Add an entry.
        srand(time(NULL));
        SecondValueAllocator_t second_value_alloc(shm.get_segment_manager());
        SecondValueAllocator_t::pointer p = second_value_alloc.allocate(1);
        second_value_alloc.construct(p, (rand() % 200) - 100);
        d.put(i, v = std::make_pair(MutexType(), p));
        std::cout << "placed " << *v.second << " into the map.\n";
    }

    return EXIT_SUCCESS;
}

先启动父进程来测试一下：

./SO13783012 父级

然后是一些孩子：

./SO13783012 孩子

样本输出：

> ./SO13783012 孩子 0：-3 1：70 2：-18 3：44 4：0 将 5:-63 放入地图。 > ./SO13783012 孩子 0：-3 1：70 2：-18 3：44 4：0 5：-63 将6：-42放入地图。 > ./SO13783012 孩子 0：-3 1：70 2：-18 3：44 4：0 5：-63 6：-42 将7：-28放入地图。

Answer 2

你可以在分配器中使用重新绑定

//    convert an allocator<T> to allocator<U>
template<typename U>
struct rebind {
    typedef Allocator<U> other;
};

制作一个

typedef std::allocator< char, managed_shared_memory::segment_manager > char_alloc;

然后像这样分配你的内存

char_alloc char_alloc_obj;
char * ptr = new (char_alloc_obj.allocate(size)) char[size];
void * vptr = (void *) ptr;

默认的 std::allocator 是无状态的。共享内存分配器将具有状态，因此您需要弄清楚如何在不同类型的分配器之间复制状态。