Introduction

During the development of the Windows XP operating system, Microsoft discovered many ACPI BIOS implementations that directly access and attempt to manipulate system hardware resources from BIOS ASL code. At run time, system board resources must not be simultaneously accessed or modified by both the BIOS and the operating system, because these accesses cannot be synchronized with operating system access to the same system resources. As a result, BIOS read or write access to these resources can cause adverse effects, ranging from general instability to causing the system to stop responding.

Historically, error conditions of this nature might have been construed by the operating system as an unrecoverable error, resulting in a stop error (blue screen) and system shutdown. In an effort to improve the end user experience and increase system reliability, Windows XP has been re-designed as described in this article.

Changes in Windows XP

While developing and testing Windows XP a list of system board resources was identified that, when accessed from BIOS AML code, proved to be the most problematic. This article refers to the list of addresses associated with these resources as the "blocked ports list." Table 1 lists affected system resources and their associated I/O addresses.

The ACPI AML interpreter in the Windows XP kernel monitors all attempts by BIOS AML code to read from or write to the specific addresses on the blocked ports list. When a read or write access is detected at any of these addresses, the following actions will occur:

Windows XP BIOS Compatibility Determined Using _OSI

The specific actions taken by the operating system when read or write accesses to these resources are detected depends on the version of ACPI interface that the BIOS indicates it supports. This is determined through the BIOS use of the _OSI method as described in this article.

For complete details about the _OSI method, see How to Identify Windows Versions in ACPI Using _OSI.

The _OSI method is being introduced with the Windows XP operating system. BIOS ASL code can test for the level of features supported in the current Windows operating system by passing a string to the _OSI method. The operating system returns TRUE for any string that represents a feature set that it can support. Windows XP, for example, returns TRUE for the string "Windows 2001".

By passing the string "Windows 2001" into the _OSI method, the BIOS indicates to the operating system that the BIOS is aware of and compatible with the ACPI implementation and feature set of Windows XP. Windows XP will then reject all I/O reads or writes from BIOS ASL code to addresses on the blocked ports list and generate an error in the system event log.

Windows XP and Legacy BIOS ASL Implementations

For those BIOS implementations that are compatible with versions of Windows released before Windows XP, the operating system will allow the read or write access to succeed in most cases. Accesses to the PIC and cascaded PIC, and to the PIC Edge/Level Control Registers, are always blocked. In all cases an error will be written to the system event log.

To help ensure system stability on systems with legacy BIOS ASL implementations, Windows XP attempts to synchronize accesses to resources whenever possible. For example:

Accessing BIOS Non-volatile Memory (CMOS NVRAM)

With these changes in Windows XP, BIOS developers may be looking for alternative methods to access CMOS NVRAM. The following examples present some possible solutions.

Reading from CMOS NVRAM

This example ASL code defines an operation region in the system memory address space. During POST, BIOS code copies the data from CMOS to this memory area. ASL code can then use fields to read the CMOS data.

Each CMnn field in the example corresponds to 8 bits of CMOS. The BIOS engineer is responsible for defining these fields appropriately. Each field of this memory can be any combination of bits.

During system boot, BIOS POST code should update this memory area with the value written in the CMOS before handing off to the operating system. Once the operating system has switched to ACPI mode, any attempt to read CMOS from ASL code should be done by reading from these memory locations and not from CMOS NVRAM I/O ports.

// Declare a memory operation region for 255 bytes of CMOS. 
// First 0 to 127 bytes reflects CMOS access from IO ports 0x70 
// and 0x71 and the subsequent bytes reflect CMOS access from IO 
// ports 0x72 to 0x73.  BIOS code should properly set up the base 
// address of the beginning of this memory range at offset during 
// BIOS POST.

OperationRegion (CMRM, SystemMemory, offset, 255)  //Operation Region 

Field (CMRM, AnyAcc, NoLock, Preserve) {   //Field
    //  Memory corresponding to addresses 0x70, 0x71
    CM00,    8,
    CM01,    8,
    CM02,    8,
    .
    .
    CM7f,     8,     

    // Memory corresponding to addresses 0x72, 0x73
    CM80,    8,
    .
    .
    CMFF
    }   // End of CMOS field

The following is an example of ASL code which reads from CMOS offset 52 bits 4 to 7 and returns the value read.

// This example returns bits 4 to 7 of CMOS offset 52

Method (RDCM) {
    ShiftRight (CM52, 4, Local0)
    Return (Local0) 
    }

Writing to CMOS NVRAM

BIOS code should write to CMOS NVRAM by generating a system management interrupt (SMI). AML code can generate a SMI by writing a specific value to the SMI command port. AML code can pass the CMOS offset and value to be written through the NVRAM memory operation region. The BIOS SMI handles the writes to CMOS, and also updates the memory area pointed by the CMRM operation region to reflect the correct CMOS contents.

Affected System Resources and Addresses

Table 1 lists the system resources and associated I/O addresses that should not be directly accessed by BIOS AML code (the "blocked ports" list).

Table 1 Blocked I/O Port Addresses and System Board Resources

*Read or write accesses to these ports are always blocked, regardless of the BIOS use of the _OSI method.

Call to Action

BIOS developers should design their BIOS code according to the guidelines in this article:

1） RSDP位于F段，用于OSPM搜索ACPI Table，RSDP可以定位其他所有ACPI Table

2） FACS位于ACPI NVS内存，用于系统进行S3保存的恢复指针，内存为NV Store

3） 剩下所有ACPI Table都位于ACPI Reclaim内存，进入OS后，内存可以释放

           ACPI Table根据版本又分为1.0B，2.0，3.0，4.0。2.0以后，支持了64-bit的地址空间，因此几个重要的Table会不大一样，比如：RSDP，RSDT，FADT，FACS。简单的列举一下不同版本的ACPI Table：

1） ACPI 1.0B：RSDP1，RSDT，FADT1，FACS1，DSDT，MADT，SSDT，HPET，MCFG等

2） ACPI 3.0 ：RSDP3，RSDT，XSDT，FADT3，FACS3，DSDT，MADT，HPET，MCFG，SSDT等

以系统支持ACPI3.0为例子，说明系统中ACPI table之间的关系如图：

 其中绿色代表在内存F段，蓝色是ACPI Reclaim内存，红色是NV store内存

1．函数指针定义

函数类型 （*指针变量名）(形参列表)；

“函数类型”说明函数的返回类型，由于“()”的优先级高于“*”,所以指针变量名外的括号必不可少，后面的“形参列表”表示指针变量指向的函数所带的参数列表。

例如：

int (*f)(int x);

double (*ptr)(double x);

在定义函数指针时请注意：
   
函数指针和它指向的函数的参数个数和类型都应该是—致的；

函数指针的类型和函数的返回值类型也必须是一致的。

2．函数指针的赋值

函数名和数组名一样代表了函数代码的首地址，因此在赋值时，直接将函数指针指向函数名就行了。

例如，

int func(int x);   /* 声明一个函数 */

int (*f) (int x);    /* 声明一个函数指针 */

f=func;            /* 将func函数的首地址赋给指针f */

赋值时函数func不带括号，也不带参数，由于func代表函数的首地址，因此经过赋值以后，指针f就指向函数func(x)的代码的首地址。

3．通过函数指针调用函数

函数指针是通过函数名及有关参数进行调用的。

与其他指针变量相类似，如果指针变量pi是指向某整型变量i的指针，则*p等于它所指的变量i；如果pf是指向某浮点型变量f的指针，则*pf就等价于它所指的变量f。同样地，*f是指向函数func(x)的指针，则*f就代表它所指向的函数func。所以在执行了f=func;之后，(*f)和func代表同一函数。

由于函数指针指向存储区中的某个函数，因此可以通过函数指针调用相应的函数。现在我们就讨论如何用函数指针调用函数，它应执行下面三步：

首先，要说明函数指针变量。

例如：int (*f)(int x);

其次，要对函数指针变量赋值。

例如： f=func;    (func(x)必须先要有定义)

最后，要用 (*指针变量)(参数表);调用函数。

例如：    (*f)(x);(x必须先赋值)

【例】任意输入n个数，找出其中最大数，并且输出最大数值。

main()

{

        int f();

        int i，a，b;

        int (*p)();    /* 定义函数指针 */

        scanf("%d"，&a);

        p=f;            /* 给函数指针p赋值，使它指向函数f */

        for(i＝1;i<9;i++)

        {
                scanf("%d"，&b);

                a＝(*p)(a，b);    /* 通过指针p调用函数f */

        }

        printf("The Max Number is:%d"，a)

}

f(int x，int y)

{

    int z;

    z＝(x>y)?x:y;

    return(z);

}

运行结果为：

343 -45 4389 4235 1 -534 988 555 789↙

The Max Number is：4389

【指针函数】

一个函数不仅可以带回一个整型数据的值，字符类型值和实型类型的值，还可以带回指针类型的数据，使其指向某个地址单元。

        返回指针的函数，一般定义格式为：

        类型标识符    *函数名(参数表)

int *f(x，y);

其中x，y是形式参数，f是函数名，调用后返回一个指向整型数据的地址指针。f(x，y)是函数，其值是指针。

如：char *ch();表示的就是一个返回字符型指针的函数，请看下面的例题：

【例】将字符串1(str1)复制到字符串2(str2)，并输出字符串2.

#include "stdio.h"

main()

{

    char *ch(char *，char *);

    char str1[]="I am glad to meet you!";

    char str2[]="Welcom to study C!";

    printf("%s"，ch(str1，str2));

}

char *ch(char *str1，char *str2)

{

    int i;

    char *p;

    p=str2
   
    if(*str2==NULL) exit(-1);

    do

    {

        *str2=*str1;

        str1++;

        str2++;

    }while(*str1!=NULL);

    return(p);

}

通过分析可得

函数指针是一个指向函数的指针，而指针函数只是说明他是一个返回值为指针的函数，

函数指针可以用来指向一个函数。

The following map varies in size and locus

	07C0:0		Boot code is loaded here at startup (31k mark)
A000:0		EGA/VGA RAM for graphics display mode 0Dh & above
B000:0		MDA RAM, Hercules graphics display RAM
B800:0		CGA display RAM
C000:0		EGA/VGA BIOS ROM (thru C7FF)
C400:0		Video adapter ROM space
C600:0 256bytes PGA communication area
C800:0	 16K	Hard disk adapter BIOS ROM
C800:5		XT Hard disk ROM format, AH=Drive, AL=Interleave
D000:0	 32K	Cluster adapter BIOS ROM
D800:0		PCjr conventionalsoftware cartridge address
E000:0	 64K	Expansion ROM space (hardwired on AT)
128K	PS/2 System ROM (thru F000)
F000:0		System monitor ROM
PCjr: software cartridge override address
F400:0		System expansion ROMs
F600:0		IBM ROM BASIC (AT)
F800:0		PCjr software cartridge override address
FC00:0		BIOS ROM
FF00:0		System ROM
FFA6:E		ROM graphics character table
FFFF:0		ROM bootstrap code
FFFF:5 8 bytes	ROM date (not applicable for all clones)
FFFF:E	byte	ROM machine id	(see MACHINE ID)