自旋一奇怪问题！！

以前在实验从保护模式进入实模式进行系统调用的时候，一点儿代码调试了长达一周时间。后来发现确实一些细小的问题，例如某个寄存器高16位有数据，在虚拟机下没问题，在真机实模式下就无法正常运行了。





同样，这次试验的是vm86，在bochs和qemu都通过了，在vpc和物理机却发生各种奇妙的错误。。。





莫非。。。








下面给出VM86的主要代码：


// VM86 General Protection Fault ...


int gpf_handler( int err_code, I386_REGISTERS* r)


{


	t_8* ip;


	t_16* stack, *ivt, *stack32;


	THREAD* thr;


	ARCH_THREAD* arch;


	int is_data32, is_addr32;


	thr = current_thread();


	// X86相关线程信息


	arch = &thr->arch;


	// 非vm86模式下的异常不进行处理


	if( !arch->in_vm86 )


		return 0;


	// 得到ip的线性地址


	ip = (t_8*)FARPTR_TO_LINEAR(r->cs, r->eip);


	// 得到堆栈的线性地址


	stack = (t_16*)FARPTR_TO_LINEAR(r->ss, r->esp);


	stack32 = stack;


	// 默认的bios中断向量在地址0处


	ivt = (t_16*)0;


	// 是否32位操作


	is_data32 = is_addr32 = 0;


	for(;;){


		// Check the Opcode


		switch( ip[0] ){


		case 0x66:	//32bit data


			is_data32 = 1;


			ip ++;


			r->eip = (t_16) (r->eip + 1);


			break;


		case 0x67:	//32bit address


			is_addr32 = 1;


			ip ++;


			r->eip = (t_16) (r->eip + 1);


			PERROR("##not supported not.");


			break;


		case 0xCD:	//int n


			switch( ip[1] ){


			case 0xFB:


			case 0x03: //debug


				PERROR("[%d]VM86 thread exit with code:0x%X", thr->tid, r->eax);


				thread_terminate( thr, r->eax );


				return 0;


				//never return here


			case 0xFC: //delay 5 seconds


				thread_wait(r->eax);


				r->eip = (t_16) (r->eip + 1);


				return 1;


			default:


//				kprintf("vm86 interrupt 0x%x at 0x%X\n", ip[1], r->eip );


				stack -= 3;


				r->esp = ((r->esp & 0xffff) - 6) & 0xffff;


				stack[0] = (t_16) (r->eip + 2);


				stack[1] = r->cs;


				stack[2] = (t_16) r->eflags;


				//!!check the IF if necessary


				if( arch->vmflag_if )


					stack[2] |= EFLAG_IF;


				else


					stack[2] &= ~EFLAG_IF;


				r->cs = ivt[ip[1] * 2 + 1];


				r->eip = ivt[ip[1] * 2];


				return 1;


			}


		case 0x9c:	//pushf


			if( is_data32 ){


				r->esp = ((r->esp & 0xffff) - 4) & 0xffff;


				stack32--;


				stack32[0] = r->eflags;


				if( arch->vmflag_if )


					stack32[0] |= EFLAG_IF;


				else


					stack32[0] &= ~EFLAG_IF;


			}else{


				r->esp = ((r->esp & 0xffff) - 2) & 0xffff;


				stack--;


				stack[0] = (t_16) r->eflags;


				if( arch->vmflag_if )


					stack[0] |= EFLAG_IF;


				else


					stack[0] &= ~EFLAG_IF;


			}


			r->eip = (t_16) (r->eip + 1);


			return 1;


		case 0x9d:	//popf


			if( is_data32 ){


				r->eflags = stack32[0]|EFLAG_VM|EFLAG_IF;


				r->esp = ((r->esp & 0xffff) + 4) & 0xffff;


				arch->vmflag_if = (stack32[0]&EFLAG_IF)!=0;


			}else{


				r->eflags = stack[0]|EFLAG_VM|EFLAG_IF;


				r->esp = ((r->esp & 0xffff) + 2) & 0xffff;


				arch->vmflag_if = (stack[0]&EFLAG_IF)!=0;


			}


			r->eip = (t_16) (r->eip + 1);


			return 1;


		case 0xEE:	//out byte


			out_byte(r->edx, r->eax&0xFF);


			r->eip = (t_16) (r->eip + 1);


			return 1;


		case 0xEF:	//out word


			/* 我看的资料上竟然没有说在0xEF前加0x66要输出32位值，


			 * 在VMware上无法改变分辨率。后来发现此处最可疑，想不到


			 * 改成dword后竟然成功了。我真聪明:-)


			 */


			if( is_data32 ){


				out_dword(r->edx, r->eax);


			}else{


				out_word(r->edx, r->eax&0xFFFF);


			}


			r->eip = (t_16) (r->eip + 1);


			return 1;


		case 0xEC:	//in byte


			r->eax = (r->eax&0xFFFFFF00)|in_byte(r->edx);


			r->eip = (t_16) (r->eip + 1);


			return 1;


		case 0xED:	//in word


			if( is_data32 ){


				r->eax = in_dword(r->edx);


			}else{


				r->eax = (r->eax&0xFFFF0000)|in_word(r->edx);


			}


			r->eip = (t_16) (r->eip + 1);


			return 1;


		case 0xCF:	//iret


			if( is_data32 ){


				r->eip = stack32[0];


				r->cs = stack32[1];


				r->eflags = EFLAG_VM|EFLAG_IF|stack32[2];


				arch->vmflag_if = (stack32[2]&EFLAG_IF)!=0;


				r->esp = ((r->esp & 0xffff) + 12) & 0xffff;


			}else{


				r->eip = stack[0];


				r->cs = stack[1];


				r->eflags = EFLAG_VM|EFLAG_IF|stack[2];


				arch->vmflag_if = (stack[2]&EFLAG_IF)!=0;


				r->esp = ((r->esp & 0xffff) + 6) & 0xffff;


			}


			return 1;


		case 0xFA:	//cli


			arch->vmflag_if = 0;


			r->eip = (t_16) (r->eip + 1);


			return 1;


		case 0xFB:	//sti


			arch->vmflag_if = 1;


			r->eip = (t_16) (r->eip + 1);


			return 1;


		default:


			PERROR("Unhandled Opcode: 0x%X at 0x%X", ip[0], ip );


			return 0;


		}


	}


	return 0; //kill the thread


}