内存共享－－救命阿！

我用DriverWorks开发驱动,想在内核和应用之间共享内存,先在内核分配内存块,供DMA缓存数据，应用程序取得该内存的首指针，再利用这个指针读出该块的数据。现在初步的实验是回传指针，生成框架之后，我只改了kernel的READ（）和app的doRead()函数，kernel里的READ()函数如下：
NTSTATUS Pointer2Device::Read(KIrp I)
{
t << "Entering Pointer2Device::Read, " << I << EOL;

KMemory Mem(I.Mdl()); //应用程序的MDL；

PULONG pBuffer = (PULONG) Mem.MapToSystemSpace();


ULONG BufferSize = 128;

PVOID pImageMemory = ExAllocatePoolWithTag(NonPagedPool,
                                   BufferSize,
                                   'MpaM');

    if(!pImageMemory)
{
        t << "status = STATUS_UNSUCCESSFUL" << EOL;
    }

KMemory ImageMdl = KMemory(pImageMemory,BufferSize);

PUCHAR pp = (PUCHAR)ImageMdl.VirtualAddress();

t << "ImageMdl.VirtualAddress = " << (ULONG)pp << EOL;

for(int i = 0; i < 16; i++) //写数0、2、4、6...
*pp++  = 2 * i;

if(NT_SUCCESS(ImageMdl.LockPages(IoReadAccess,UserMode))) //锁定
{
t << "lockpages sucessed" << EOL;
}
else
{
t << "lockpages unsucessed" << EOL;
}

ImageMdl.SetPageArray();

PVOID m_UserSpace = (PVOID)ImageMdl.MapToUserSpace();

t << "ImageMdl.PointerOfUserSpace = " << (ULONG)m_UserSpace << EOL;

*pBuffer = (ULONG)(m_UserSpace);

I.Information() = 4;

return I.PnpComplete(this, STATUS_SUCCESS);
}


应用程序doRead()如下：
void doRead(int n)
{
unsigned long *buf;
ULONG nRead;
ULONG i;

buf = (unsigned long *)malloc(n);
if (buf == NULL)
{
printf("Failed to allocate buffer for read");
Exit(1);
}

// Read data from driver
printf("Reading from device - ");
ReadFile(hDevice, buf, n, &nRead, NULL);
printf("%d bytes read from device (%d requested).n", nRead, n);

// Print what was read

printf("the pointer form kernel is :n");

ULONG pToKernelBuffer;

pToKernelBuffer = *buf;
printf("pToKernelBuffer = %xn",pToKernelBuffer);

for(i = 0; i < 16; i++)
{
printf("%u ",*(unsigned char *)(pToKernelBuffer++));
}  

printf("n");

free(buf);
}
doRead()能够读到m_UserSpace = (PVOID)ImageMdl.MapToUserSpace()的指针，也能够把写入缓冲区的0、2、4、6、、、读出来。但DriverMonitor锁定缓冲页时显示lockpages unsucessed，那我没有锁定缓冲区就MapToUserSpace是不是有错误呢.而最严重的是如果我把生成MDL的程序放到其他例程，例如把PVOID pImageMemory = ExAllocatePoolWithTag(NonPagedPool,
                                   BufferSize,
                                   'MpaM');
KMemory ImageMdl = KMemory(pImageMemory,BufferSize)放到onstartdivece ()时就会篮平（该在那里生成呢？onstartdevice（）还是driverentry（）），我肯定是要在其他地方生存MDL,所以很郁闷，请指点下这关于内核分配内存块、回传指针的编程，谢谢！告诉我怎么样给分，多多都给！

另外如果READ（）函数这样生成MDL时显示能够生成MDL,但是执行MapToUserSpace会篮平。
pImageMemory = ExAllocatePoolWithTag(NonPagedPool,
                                   128,
                                   'MpaM');
 
    if(!pImageMemory)
{
        status = STATUS_UNSUCCESSFUL;
return status;
    }

int MdlSize = KMemory::RequiredMdlStorage(pImageMemory, 128);

p = (PMDL) new (NonPagedPool) char[MdlSize];
if ( p == NULL)
{
ExFreePool(pImageMemory);
pImageMemory = NULL;
status = STATUS_UNSUCCESSFUL;
return status;
}

ImageMdl = KMemory(pImageMemory,128,p);

if ((PMDL)ImageMdl == NULL)
{
ExFreePool(pImageMemory);
delete p;
pImageMemory = NULL;
status = STATUS_UNSUCCESSFUL;
return status;
}
t << "ImageMdl = KMemory(pImageMemory,128,p) successn";

PUCHAR pp = (PUCHAR)ImageMdl.VirtualAddress();

t << "ImageMdl.VirtualAddress = " << (ULONG)pp << EOL;

for(int i = 0; i < 16; i++) //写数0、2、4、6...
*pp++  = 2 * i;

if(NT_SUCCESS(ImageMdl.LockPages(IoReadAccess,UserMode))) //锁定
{
t << "lockpages sucessed" << EOL;
}
else
{
t << "lockpages unsucessed" << EOL;
}

ImageMdl.SetPageArray();

PVOID m_UserSpace = (PVOID)ImageMdl.MapToUserSpace();

t << "ImageMdl.PointerOfUserSpace = " << (ULONG)m_UserSpace << EOL;

*pBuffer = (ULONG)(m_UserSpace);

I.Information() = 4;

return I.PnpComplete(this, STATUS_SUCCESS);
}

谢谢！


[编辑 -  7/1/04 by  chylml]

ImageMdl.LockPages(IoReadAccess,UserMode)？把内核的内存锁到应用层？

to楼上：是啊，我是想锁定，再回传指针给app啊。你的意思是怎么样啊。

这样的问题我至少回答了3次，在论坛中搜索一下