flsystem.h

（1）本人基于MPC860的vxworks bsp的程序 （2）实现了FEC 100M和 SCC 10M的网口功能 （3）实现了nor flash的TFFS文件系统 （4）实现了系统的自检 （

/*
 * $Log:   V:/flsystem.h_v  $

      Rev 1.0   13 Jan 1999 22:18:07   Yogu
   added vxLib.h and macro for fault safe write in isRAM.

      Rev 1.0   08 Jan 1998 17:20:00   Hdei
   changed MALLOC FREE to MALLOC_TFFS FREE_TFFS.

      Rev 1.0   24 Jul 1997 18:13:06   amirban
   Initial revision.
 */

/************************************************************************/
/*                                                                      */
/*		FAT-FTL Lite Software Development Kit			*/
/*		Copyright (C) M-Systems Ltd. 1995-1996			*/
/*									*/
/************************************************************************/


#ifndef FLSYSTEM_H
#define FLSYSTEM_H

#include <vxWorks.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <ctype.h>
#include <tickLib.h>
#include <wdLib.h>
#include <sysLib.h>
#include <excLib.h>
#include <semLib.h>
#include <ioLib.h>
#include <iosLib.h>
#include <memLib.h>
#include <errnoLib.h>
#include <vxLib.h>


/*
 * 			signed/unsigned char
 *
 * It is assumed that 'char' is signed. If this is not your compiler
 * default, use compiler switches, or insert a #pragma here to define this.
 *
 */

/* char is signed by default in GNU C */


/* 			CPU target
 *
 * Use compiler switches or insert a #pragma here to select the CPU type
 * you are targeting.
 *
 * If the target is an Intel 80386 or above, also uncomment the CPU_i386
 * definition.
 */

/* defined in VxWorks makefile */

/* 			NULL constant
 *
 * Some compilers require a different definition for the NULL pointer
 */

/* #include <_null.h> */


/* 			Little-endian/big-endian
 *
 * FAT and FTL structures use the little-endian (Intel) format for integers.
 * If your machine uses the big-endian (Motorola) format, uncomment the
 * following line.
 * This option needs a C++ compiler.
 * Note that even on big-endian machines you may omit the BIG_ENDIAN
 * definition for smaller code size and better performance, but your media
 * will not be compatible with standard FAT and FTL.
 */

/* we are using VxWorks #define _BYTE_ORDER here */
#ifndef _BYTE_ORDER
#error "byte order is not defined"
#else  /* _BYTE_ORDER */
#if (_BYTE_ORDER == _BIG_ENDIAN)
#define BIG_ENDIAN
#else  /* (_BYTE_ORDER == _BIG_ENDIAN) */
#undef BIG_ENDIAN
#endif /* (_BYTE_ORDER == _BIG_ENDIAN) */
#endif /* _BYTE_ORDER */


/* 			Far pointers
 *
 * Specify here which pointers may be far, if any.
 * Far pointers are usually relevant only to 80x86 architectures.
 *
 * Specify FAR_LEVEL:
 *   0 -	if using a flat memory model or having no far pointers.
 *   1 -        if only the socket window may be far
 *   2 -	if only the socket window and caller's read/write buffers
 *		may be far.
 *   3 -	if socket window, caller's read/write buffers and the
 *		caller's I/O request packet may be far
 */

#define FAR_LEVEL       0


/* 			Memory routines
 *
 * You need to supply library routines to copy, set and compare blocks of
 * memory, internally and to/from callers. The code uses the names 'tffscpy',
 * 'tffsset' and 'tffscmp' with parameters as in the standard 'memcpy',
 * 'memset' and 'memcmp' C library routines.
 */

#if FAR_LEVEL > 0
#define tffscpy _fmemcpy
#define tffsset  _fmemset
#define tffscmp  _fmemcmp
#else
/* XXX
#define tffscpy  sysTffsCpy
#define tffsset  sysTffsSet
*/
#define tffscpy  memcpy
#define tffsset  memset
#define tffscmp  memcmp
#endif

#define tffscpyWords(dest,src,nbytes)     bcopyWords((char *)(src), \
                                                  (char *)(dest), (nbytes)/2)
extern int tffscmpWords(void *buf1, void *buf2, int nbytes);


/* 			Pointer arithmetic
 *
 * The following macros define machine- and compiler-dependent macros for
 * handling pointers to physical window addresses. The definitions below are
 * for PC real-mode Borland-C.
 *
 * 'physicalToPointer' translates a physical flat address to a (far) pointer.
 * Note that if when your processor uses virtual memory, the code should
 * map the physical address to virtual memory, and return a pointer to that
 * memory (the size parameter tells how much memory should be mapped).
 *
 * 'addToFarPointer' adds an increment to a pointer and returns a new
 * pointer. The increment may be as large as your window size. The code
 * below assumes that the increment may be larger than 64 KB and so performs
 * huge pointer arithmetic.
 */

#if FAR_LEVEL > 0
#include <dos.h>

#define physicalToPointer(physical,size,drive)		\
	MK_FP((int) ((physical) >> 4),(int) (physical) & 0xF)

#define addToFarPointer(base,increment)		\
	MK_FP(FP_SEG(base) +			\
		((unsigned short) ((FP_OFF(base) + (increment)) >> 16) << 12), \
	      FP_OFF(base) + (int) (increment))
#else
#define physicalToPointer(physical,size,drive)          \
	((void *) (physical))

#define addToFarPointer(base,increment)		\
	((void *) ((unsigned char *) (base) + (increment)))
#endif


/* 			Default calling convention
 *
 * C compilers usually use the C calling convention to routines (cdecl), but
 * often can also use the pascal calling convention, which is somewhat more
 * economical in code size. Some compilers also have specialized calling
 * conventions which may be suitable. Use compiler switches or insert a
 * #pragma here to select your favorite calling convention.
 */

/* use GNU C default calling convention */



/* 			Mutex type
 *
 * If you intend to access the FLite API in a multi-tasking environment,
 * you may need to implement some resource management and mutual-exclusion
 * of FLite with mutex & semaphore services that are available to you. In
 * this case, define here the Mutex type you will use, and provide your own
 * implementation of the Mutex functions incustom.c
 *
 * By default, a Mutex is defined as a simple counter, and the Mutex
 * functions in custom.c implement locking and unlocking by incrementing
 * and decrementing the counter. This will work well on all single-tasking
 * environment, as well as on many multi-tasking environments.
 */

typedef SEM_ID FLMutex;

#define flStartCriticalSection(mutexPtr)  flTakeMutex(mutexPtr,1)
#define flEndCriticalSection(mutexPtr)    flFreeMutex(mutexPtr)



/* 			Memory allocation
 *
 * The translation layers (e.g. FTL) need to allocate memory to handle
 * Flash media. The size needed depends on the media being handled.
 *
 * You may choose to use the standard 'malloc' and 'free' to handle such
 * memory allocations, provide your own equivalent routines, or you may
 * choose not to define any memory allocation routine. In this case, the
 * memory will be allocated statically at compile-time on the assumption of
 * the largest media configuration you need to support. This is the simplest
 * choice, but may cause your RAM requirements to be larger than you
 * actually need.
 *
 * If you define routines other than malloc & free, they should have the
 * same parameters and return types as malloc & free. You should either code
 * these routines in custom.c or include them when you link your application.
 */

#define MALLOC_TFFS malloc
#define FREE_TFFS free

/*			isRAM fault safe write
 * The function isRAM() does a direct write to verify if the location 
 * specified is RAM. It is possible that hardware is tailered to fault 
 * in such situations. 
 */

#ifdef ORIG
#define ISRAM_WRITE(src,dst) dst = src
#else /* !ORG */
#define ISRAM_WRITE(src,dst) \
	vxMemProbe ((char *)dst, VX_WRITE, 4, (char *)&src)
#endif /* ORIG */

/*                      Pointer arithmetic
 *
 * The following macros define machine- and compiler-dependent macros for
 * handling pointers to physical window addresses. The definitions below are
 * for PC real-mode Borland-C.
 *
 * 'physicalToPointer' translates a physical flat address to a (far) pointer.
 * Note that if when your processor uses virtual memory, the code should
 * map the physical address to virtual memory, and return a pointer to that
 * memory (the size parameter tells how much memory should be mapped).
 *
 * 'addToFarPointer' adds an increment to a pointer and returns a new
 * pointer. The increment may be as large as your window size. The code
 * below assumes that the increment may be larger than 64 KB and so performs
 * huge pointer arithmetic.
 */

#if FAR_LEVEL > 0
#include <dos.h>

#define physicalToPointer(physical,size,drive)          \
        MK_FP((int) ((physical) >> 4),(int) (physical) & 0xF)

/*NEW*/ #define pointerToPhysical(ptr)                  \
        (((unsigned long) FP_SEG(ptr) << 4) + FP_OFF(ptr))

#define addToFarPointer(base,increment)         \
        MK_FP(FP_SEG(base) +                    \
                ((unsigned short) ((FP_OFF(base) + (increment)) >> 16) << 12), \
        FP_OFF(base) + (int) (increment))
#else
#define physicalToPointer(physical,size,drive)          \
        ((void *) (physical))

/*NEW*/ #define pointerToPhysical(ptr)  ((unsigned long)(ptr))

#define addToFarPointer(base,increment)         \
        ((void *) ((unsigned char *) (base) + (increment)))
#endif


#endif