JSVMP执行流程分析

Security Classification: 【C-1】 | Publish Time:2024-02-21 | Category:Test Notes | Edit

Expiry Notice: The article was published three months ago. Please independently assess the validity of the technical methods and code mentioned within. :)

AI摘要:本文通过分析vmp虚拟机的解释器结构，详细探讨了字节码的执行过程。文章首先介绍了主要函数`interpreter`及其参数，接着提取了字节码和常量池，并逐步跟踪字节码的执行。通过调试，分析了每个操作码的功能，特别是变量定义、栈操作和赋值逻辑，最终实现了原始JavaScript代码的功能。在整个分析中，详细记录了各个操作码的用途和执行步骤，包括变量的定义、值的赋值、逻辑判断及算术运算等。文章总结了vmp虚拟机如何通过字节码执行JavaScript代码，并指出了其在虚拟化保护中的作用。最后，提供了原始JavaScript代码和样本代码供参考。 --- (来自模型:gpt-4o-mini-2024-07-18)

### 结构分析
观察vmp后函数的结构,样本代码见文末。（测试地址：https://rce.ink/vmp.html )

```
function interpreter(parentScope, index, stack, constantPool, bytecode, option={}, args)
```
参数释义如下：
parentScope：作用域，用于保存指令处理后的结果及属性
index:指针，初始为0
stack:存值的栈
constantPool:常量池，获取常量
bytecode:字节码数组，用于提取opcode以及相应的操作数
option:特殊状态下有效
args:特殊状态下有效

提取出字节码如下：
```
[3,3,3,4,4,5,12,4,3,0,1,17,0,0,5,3,1,4,6,54,1,12,4,4,0,1,17,0,0,5,4,1,4,7,53,1,5,3,1,4,8,52,1,51,1,13,4,9,0,0,5,9,1,4,10,50,1,7,76,14,74,5,9,1,4,11,54,1,13,4,9,0,0,16,6,76,4,12,4,13,51,1,5,9,1,51,1,12,4,14,1,1,4,15,2,1,0]
```
提取出常量池如下：
```
["@faceless","arguments","__proto__","a","b",22,2,6,4,"c",70,7,"test",100,"console","log"]
```

我们来跟进观察一下，vmp是如何实现虚拟化保护的：
自运行函数一开始定义了一个localScope空对象，定义@faceless与arguments两个属性，并保存参数中的option 与args， 判断option中的选项开关，从window对象中取值并设置到localScope中。并把localScope原型定义为window对象,这样就可以通过localScope调用到window对象中的方法。

![](https://p0.meituan.net/xianfu/2d4256e0beb990bc88975b137cb622b847632.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

定义了一个getScope的方法，目的是在给定的作用域链中查找具有指定名称的属性，并返回拥有该属性的对象。如果没有找到，则返回 null。

![](https://p0.meituan.net/xianfu/d74efdd115dde16f0816eefdd244958372967.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

后面就进入vmp的分发器Dispatcher了。在html页面19行打一个断点，这代表当前获取的指针位置。

![](https://p0.meituan.net/xianfu/ae5bd41b8ab3f25d6d999b97e38b9ad740440.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

刷新页面

![](https://p0.meituan.net/xianfu/2c25cf1929265421789e3042bec2d056213178.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

通过指针0，获取到的opcode为3。它将进入某个分支执行，在下面找到这个逻辑判断：

![](https://p1.meituan.net/xianfu/b7837063d4ae2797df70bb532d2c10f772682.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

在这个handler里，他将指针自增1，也就是1。在字节码数组中向后取一位值，也就是3，这个值将作为常量池中的索引。并将t1赋值为常量池中的a，并把localscope定义一个属性a，值为undefined。

总结一下

| opcode  |  用途 |
| ------------ | ------------ |
|  3 | 变量定义，类似于 var xx = undefind  |

去掉这里的断点，继续向下跟进。

![](https://p1.meituan.net/xianfu/6535f8ad016768b9a46dab09a12e169e83968.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

这里的opcode还是3，前面的流程做的事原始代码中的var a的部分，大胆猜测这里做的事var b的操作。

![](https://p0.meituan.net/xianfu/02d88ead445303e834cc8ed6fd82900e63956.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

指针继续自增1，取到常量池中的b。也同样赋值给localScope。

![](https://p0.meituan.net/xianfu/87c80b156880958df571211398af123183438.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

之后opcode就变成4了，向下跟进发现它先指针自增取字节码，然后从常量池中取这个索引的值并压入栈中。

![](https://p0.meituan.net/xianfu/edca00df5c3922424f217b47e20e5db150835.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

| opcode  |  用途 |
| ------------ | ------------ |
|  3 | 变量定义，类似于 var xx = undefind  |
|4 |根据操作数，从常量池取值，压入栈|

继续往下，发现opcode是12了

![](https://p0.meituan.net/xianfu/213e8975e982a00825c4233de973c8fa129769.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

找到对应的handler

![](https://p0.meituan.net/xianfu/f8e994a0a56522280bf067c0d4bcaf8219935.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

发现是把当前的本地作用域放入栈中

| opcode  |  用途 |
| ------------ | ------------ |
|  3 | 变量定义，类似于 var xx = undefind  |
|4 |根据操作数，从常量池取值，压入栈|
|12 |将当前的localScope压入栈|

接着往下跟进，opcode又为4
![](https://p0.meituan.net/xianfu/56f7b05dadf516e6070d8985373f8bad68605.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)
往后取操作数,是用这个值作为常量池数组索引取值

![](https://p0.meituan.net/xianfu/97d45841bbdbecb61d85ae5735fd2d6f55206.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

可以发现此时栈中存在22、window、a三个值。我们可以整理归纳一下，前面字节码做了哪些事：

3,3   定义变量a
3,4   定义变量b
4,5   取了原始代码中的a的值 放入栈中
12    将localScope放入栈中
4,3,  取了a这个变量名 放入栈中

这里其实有点感觉了，三个值放入栈中，一个值为window对象，一个属性名一个属性值，大概率后面的opcode要做值出栈，赋值的作用。类似弹出a和window，然后做类似window.a  = 22 。印证一下接着往下吧：

![](https://p0.meituan.net/xianfu/708ae137a5e35c089a88f6c45116dca874402.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

![](https://p1.meituan.net/xianfu/96ac09b08580854b0d54843b5fb0a1db54095.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

这里的逻辑就很清晰了，弹出三个元素，分别为属性名a、window对象、值22，然后向下找一个操作数，此时为1。通过getScope函数获取存在属性a的作用域，因为前面localScope定义过a的属性（通过opecode=3的第一次操作），所以会返回作用域localScope，并把这个作用域的属性a赋值22。由于操作数t4为1，需要把这个值压回栈内。

![](https://p0.meituan.net/xianfu/cb5c5a322a04126a069ef03a4e342d6b122057.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

3,3   定义变量a
3,4   定义变量b
4,5   取了原始代码中的a的值 放入栈中
12    将localScope放入栈中
4,3,  取了a这个变量名 放入栈中
0,1  localScope 赋值，将赋的值放入栈中

此时通过6条指令，12个字节码，我们实现了原始代码中的var a =22的效果。

接着往下看别的opcode

![](https://p0.meituan.net/xianfu/243ed710beb6b51e8ac3f15a3880ee2350246.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

取出栈顶元素，往下取操作数为0，将此值作为弹出元素的数量，对栈顶进行清理的操作。再往下取操作数t5，判断是否要把前面栈顶的值放回。

![](https://p0.meituan.net/xianfu/d3397176f8f241d6ff99d21badfadfcb74482.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

可以看到栈此时被清空了。

3,3   定义变量a
3,4   定义变量b
4,5   取了原始代码中的a的值 放入栈中
12    将localScope放入栈中
4,3,  取了a这个变量名 放入栈中
0,1  localScope 赋值，将赋的值放入栈中
17,0,0 弹出栈顶，清空0个栈上元素，不把弹出的值压回栈内

| opcode  |  用途 |
| ------------ | ------------ |
|  3 | 变量定义，类似于 var xx = undefind  |
|4 |根据操作数，从常量池取值，压入栈|
|12 |将当前的localScope压入栈|
|0 |弹出栈顶的三个元素，找到存在栈顶元素名属性的作用域，设置这个属性的值为第三个元素的值，最后把值压回栈内|
|17 |弹出栈顶元素，根据第一个操作数判断需要清空的栈内元素的个数，第二个操作数判断是否把值压回栈内|

opcode为0时，赋值结束不确定后续是否要从栈获取这个值，所以会将这个值重新压入栈中。
为17时，根据操作数是否清空栈。或者仅保留栈顶元素。

向下跟进，opcode为5。

![](https://p0.meituan.net/xianfu/d577e47d445183c968143d952e1b81e897097.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

第一个操作数用于在常量池获取属性名，第二个操作数用于判断是否压入栈内。此时将作用域中的a的值压入了栈内。

3,3   定义变量a
3,4   定义变量b
4,5   取了原始代码中的a的值 放入栈中
12    将localScope放入栈中
4,3,  取了a这个变量名 放入栈中
0,1  localScope 赋值，将赋的值放入栈中
17,0,0 弹出栈顶，清空0个栈上元素，不把弹出的值压回栈内
5,4,1 取常量池中第四个值作为属性名，获取作用域中该属性的值并压入栈内

后面到了opcode为4的位置，将2的值压入栈内。此时栈分布

![](https://p1.meituan.net/xianfu/bc9639b1bc16f58ec4373fde4e546a8010433.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

再到opcode为54的操作了，这里稍微麻烦点：

先进入一个指令范围的判断，把栈内两个元素弹出。（可以想到这段范围的opcode实现了对栈内两个元素的各类运算）

![](https://p0.meituan.net/xianfu/a244244d6aad0c36343664a4571277c232175.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

再进入这个分支进行除运算

![](https://p0.meituan.net/xianfu/110ae171161ff47698dc9bb31e05228814159.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

这不就是原始代码中的 a/2的部分吗。再取操作数判断是否要压回栈内,最终把除后的结果压入栈内：

![](https://p0.meituan.net/xianfu/ce99ad13d9bc643a2955db414b27933537439.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

后面又到了opcode为12，把localscope压入栈，略过。
又到了opcode为4，跟据操作数把b压入栈内。
随后就是0了，此时栈内分布：

![](https://p0.meituan.net/xianfu/456c423a2bd04f5b342023d04c1769bd10367.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

很显然做了同样的步骤，把localscope中的b属性赋值为11。预测下一条指令必定是17以清空栈。

![](https://p0.meituan.net/xianfu/dea8a98bdc9b8b9e85c439a0aa4e466175394.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

结果真如预想的一样。后面的步骤，进行简化：

5 -> 将b的值压入栈
4 -> 将常量池6压入栈
53 -> 弹出两个元素，进行乘法操作，结果压入栈
5 -> 将a的值 压入栈
4 -> 将常量4压入栈
52 -> 弹出栈顶两个元素 进行减法操作，结果压入栈

（stop，此时栈内存放着b*6的值与a-4的值，不出所料后面必定是进行加操作）
![](https://p0.meituan.net/xianfu/e2ebacdd944572d1701a3f262efc0d1964919.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

此时栈内存放着 (b*6)+(a-4)的结果值，接着是需要赋值给c的。但是和ab有区别，他并不是var定义的值，而是直接赋值。在js的赋值操作中，他是直接成为window的一个属性的：
![](https://p0.meituan.net/xianfu/ea4a028f885a4c89836bfc07d9fa1ce812279.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

我们来预测一下指令吧，先把window作用域放到栈内，再从常量池取c这个属性名放入栈内，最后进行赋值操作，顺序应该是：

![](https://p0.meituan.net/xianfu/d36fb076b5986972d77587058114b20e20625.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

13（通过代码阅读直接找到opcode）
4
0

看下结果吧。

![](https://p0.meituan.net/xianfu/ba0667d78b4a422788e7f270ea4e32f163735.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

![](https://p0.meituan.net/xianfu/a1900ad206c57f4c453e5e140672376564164.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

![](https://p0.meituan.net/xianfu/1a4f5eccfa40bca6c6f619bed62277d376551.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

完全正确。此时已经赋值了c在window属性中，下一条指令的opcode为5，根据操作数拿到了c这个值并压入栈内。（虽然c定义在window中，但是localScope的原型指向了window，所以按照查找机制也可以通过localScope.c取到c的值。这也就是一开始初始化的时候为什么需要那样定义。）

之后的opcode为4，根据表推测，将常量池中的70压入栈内，即将与c的值做比较。
之后的opcode为50，是一个比较运算

![](https://p0.meituan.net/xianfu/0c6e25ff584658a460f66970da7081ba25264.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

运算结束后，栈内存在一个布尔值

![](https://p1.meituan.net/xianfu/a7e9be38469a7e407fe5e91c972383548689.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

往后的操作码为7，如果栈内元素为false，则修改索引值为第一个操作数。这里为true，所以没有跳转。有没有有点像，如果是false进入else分支，跳过if逻辑直接进入else的分支起始指针？

![](https://p0.meituan.net/xianfu/85763aed1429c9e1d13ccde45fb51e3031733.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

opcode为14, 这应该是所有指令中最难理解的了。

这条指令分成几部分：

|  opcode | 操作数1  | body  | 操作数2  |
| ------------ | ------------ | ------------ | ------------ |
| 14  | 跳转位置   |  指令集 | 16  |

进入14时，将指令集用分发器再执行一遍，并把结果存储到作用域。直到遇到opcode16跳出。
在抽象语法树中，按照深度优先的逻辑，遇到ifstatement时，会执行push（14）->对子块继续做字节码转换->执行push(16)结束这个statement。

![](https://p0.meituan.net/xianfu/0b5b1cfd99b0e94010cd7fb2d75b707976626.png%40watermark=1&&object=L3dkY2Zsb3cvN2RiN2M4NTFjYmVjZDg4MTM1OTZjMTYzOWE2MzQ4MDM0MjY0LnBuZw==&p=8&t=90&x=10&y=10)

至此 对jsvmp有基本的认知即可。后续会继续学习相应的开发以及对抗vmp的一些探索。

### 原始代码
```javascript
var a = 22
var b =  a/2
c = (b*6)+(a-4)
if(c>70){
  c = c/7
}
console.log("test"+100+c)
```
### 样本代码
````javascript
(function interpreter(parentScope, index, stack, constantPool, bytecode, option = {}, args) {
  let localScope = {};
  localScope[constantPool[0]] = option.t || this;
  localScope[constantPool[1]] = args;
  if (option.n) localScope[option.n] = parentScope[option.f];
  localScope.__proto__ = parentScope;
  function getScope(scope, name) {
    if (!scope || name == constantPool[2]) return null;
    if (scope.hasOwnProperty(name)) return scope;
    return getScope(Object.getPrototypeOf(scope), name);
  }
  if (option.e) {
    localScope[constantPool[bytecode[index + 1]]] = option.e;
  }
  while (index < bytecode.length) {
    let t0, t1, t2, t3, t4, t5;
    let instruction = bytecode[index++];
    if (instruction == 16) {
      return;
    }
    if (instruction == 21) {
      stack.push(`${stack.pop()}`);
    }
    if (instruction == 58) {
      t0 = stack.pop();
      t1 = !t0;
      t5 = bytecode[index++];
      if (t5) stack.push(t1);
    }
    if (instruction == undefined) {
      throw new Error("当前指令不存在，请检查指令集!");
    }
    if (instruction == 24) {
      debugger;
    }
    if (instruction == 9) {
      t0 = stack.pop();
      t1 = bytecode[index++];
      t2 = bytecode[index++];
      t3 = stack[stack.length - 1];
      t4 = function () {
        return interpreter(localScope, t1, stack, constantPool, bytecode, {
          t: this,
          n: t0,
          f: t0 || t3,
          r: 1
        }, arguments);
      };
      if (t2) {
        stack.push(t4);
      } else {
        localScope[t0] = t4;
      }
    }
    if (instruction == 30) {
      t0 = bytecode[index++];
      t1 = [];
      for (t2 = 0; t2 < t0; t2++) {
        t3 = stack.pop();
        t1.unshift(t3);
      }
      t5 = bytecode[index++];
      if (t5) stack.push(t1);
    }
    if (instruction == 23) {
      t0 = stack.pop();
      t1 = stack.pop();
      t3 = new RegExp(t1, t0);
      stack.push(t3);
    }
    if (instruction == 15) {
      t0 = bytecode[index++];
      t1 = interpreter(localScope, index, stack, constantPool, bytecode, {
        t: localScope[constantPool[0]]
      });
      if (!t1) {
        index = t0;
      } else {
        if (t1 === 1) {
          return;
        } else {
          if (option.r) {
            return stack.pop();
          }
          return t1;
        }
      }
    }
    if (instruction == 14) {
      t0 = bytecode[index++];
      t1 = interpreter(localScope, index, stack, constantPool, bytecode, {
        t: localScope[constantPool[0]]
      });
      if (t1 === undefined) {
        index = t0;
      } else {
        if (option.r) {
          return stack.pop();
        }
        return t1;
      }
    }
    if (33 < instruction && instruction < 57) {
      t0 = stack.pop();
      t1 = stack.pop();
      if (instruction == 40) {
        t2 = t1 || t0;
      }
      if (instruction == 43) {
        t2 = t1 >= t0;
      }
      if (instruction == 53) {
        t2 = t1 * t0;
      }
      if (instruction == 36) {
        t2 = t1 >>> t0;
      }
      if (instruction == 46) {
        t2 = t1 != t0;
      }
      if (instruction == 44) {
        t2 = t1 <= t0;
      }
      if (instruction == 48) {
        t2 = t1 !== t0;
      }
      if (instruction == 55) {
        t2 = t1 % t0;
      }
      if (instruction == 35) {
        t2 = t1 in t0;
      }
      if (instruction == 39) {
        t2 = t1 | t0;
      }
      if (instruction == 42) {
        t2 = t1 && t0;
      }
      if (instruction == 51) {
        t2 = t1 + t0;
      }
      if (instruction == 38) {
        t2 = t1 << t0;
      }
      if (instruction == 52) {
        t2 = t1 - t0;
      }
      if (instruction == 56) {
        t2 = t1 ^ t0;
      }
      if (instruction == 49) {
        t2 = t1 < t0;
      }
      if (instruction == 50) {
        t2 = t1 > t0;
      }
      if (instruction == 34) {
        t2 = t1 instanceof t0;
      }
      if (instruction == 37) {
        t2 = t1 >> t0;
      }
      if (instruction == 54) {
        t2 = t1 / t0;
      }
      if (instruction == 45) {
        t2 = t1 == t0;
      }
      if (instruction == 41) {
        t2 = t1 & t0;
      }
      if (instruction == 47) {
        t2 = t1 === t0;
      }
      t3 = bytecode[index++];
      if (t3) stack.push(t2);
    }
    if (instruction == 57) {
      t0 = stack.pop();
      t1 = ~t0;
      t5 = bytecode[index++];
      if (t5) stack.push(t1);
    }
    if (instruction == 20) {
      throw stack.pop();
    }
    if (instruction == 2) {
      t0 = bytecode[index++];
      t1 = stack.pop();
      t2 = stack.pop();
      args = [];
      for (t3 = 0; t3 < t0; t3++) args.unshift(stack.pop());
      if (t2 === 0) {
        t4 = t1.apply(localScope, args);
      } else {
        t4 = t2[t1].apply(t2, args);
      }
      t5 = bytecode[index++];
      if (t5) stack.push(t4);
    }
    if (instruction == 0) {
      t0 = stack.pop();
      t1 = stack.pop();
      t2 = stack.pop();
      t4 = bytecode[index++];
      t3 = getScope(t1, t0) || t1;
      t3[t0] = t2;
      if (t4) stack.push(t2);
    }
    if (instruction == 29) {
      t0 = stack.pop();
      t1 = stack.pop();
      t2 = stack.pop();
      t2[t1] = t0;
      t5 = bytecode[index++];
      if (t5) stack.push(t2);
    }
    if (instruction == 32) {
      t0 = stack.pop();
      t1 = void t0;
      t5 = bytecode[index++];
      if (t5) stack.push(t1);
    }
    if (instruction == 3) {
      t0 = bytecode[index++];
      t1 = constantPool[t0];
      localScope[t1] = undefined;
    }
    if (instruction == 17) {
      t0 = bytecode[index++];
      t1 = stack.pop();
      for (t2 = 0; t2 < t0; t2++) stack.pop();
      t5 = bytecode[index++];
      if (t5) stack.push(t1);
    }
    if (instruction == 33) {
      t0 = stack.pop();
      t1 = stack.pop();
      t2 = delete t1[t0];
      t5 = bytecode[index++];
      if (t5) stack.push(t2);
    }
    if (instruction == 8) {
      t0 = stack.pop();
      t1 = stack.pop();
      t2 = bytecode[index++];
      if (t1 == -1) t1 = t2;
      t1 = t0[t1];
      index = t1;
    }
    if (instruction == 18) {
      t0 = stack.pop();
      t0 = Number(t0);
      if (t0 == NaN) t0 = -1;
      stack.push(t0);
    }
    if (instruction == 6) {
      t0 = bytecode[index++];
      index = t0;
    }
    if (instruction == 26) {
      t0 = bytecode[index++];
      t1 = stack.pop();
      t2 = stack.pop();
      t3 = getScope(t2, t1) || t2;
      t4 = !t0 ? t3[t1]-- : --t3[t1];
      t5 = bytecode[index++];
      if (t5) stack.push(t4);
    }
    if (instruction == 22) {
      t0 = bytecode[index++];
      t2 = "";
      for (t1 = 0; t1 < t0; t1++) {
        t2 += stack.pop();
      }
      t5 = bytecode[index++];
      if (t5) stack.push(t2);
    }
    if (instruction == 13) {
      stack.push(window);
    }
    if (instruction == 27) {
      t0 = stack.pop();
      t1 = typeof t0;
      t5 = bytecode[index++];
      if (t5) stack.push(t1);
    }
    if (instruction == 12) {
      stack.push(localScope);
    }
    if (instruction == 4) {
      t0 = bytecode[index++];
      stack.push(constantPool[t0]);
    }
    if (instruction == 25) {
      t0 = bytecode[index++];
      t1 = stack.pop();
      t2 = stack.pop();
      t3 = getScope(t2, t1) || t2;
      t4 = !t0 ? t3[t1]++ : ++t3[t1];
      t5 = bytecode[index++];
      if (t5) stack.push(t4);
    }
    if (instruction == 28) {
      stack.push({});
    }
    if (instruction == 19) {
      t2 = bytecode[index++];
      t3 = bytecode[index++];
      t4 = bytecode[index++];
      t5 = bytecode[index++];
      try {
        t1 = interpreter(localScope, t3, stack, constantPool, bytecode, {
          t: localScope[constantPool[0]]
        });
        if (t1 > 0) {
          if (option.r) {
            return stack.pop();
          }
          return t1;
        }
      } catch (e) {
        t1 = interpreter(localScope, t4, stack, constantPool, bytecode, {
          t: localScope[constantPool[0]],
          e: e
        });
        if (t1 > 0) {
          if (option.r) {
            return stack.pop();
          }
          return t1;
        }
      } finally {
        t1 = interpreter(localScope, t5, stack, constantPool, bytecode, {
          t: localScope[constantPool[0]]
        });
        if (t1 > 0) {
          if (option.r) {
            return stack.pop();
          }
          return t1;
        }
      }
      index = t2;
    }
    if (instruction == 5) {
      t0 = constantPool[bytecode[index++]];
      t1 = localScope;
      t2 = t1[t0];
      t5 = bytecode[index++];
      if (t5) stack.push(t2);
    }
    if (instruction == 31) {
      t0 = stack.pop();
      t1 = stack.pop();
      t2 = bytecode[index++];
      args = [];
      for (t3 = 0; t3 < t2; t3++) args.unshift(stack.pop());
      t4 = new t1[t0](...args);
      t5 = bytecode[index++];
      if (t5) stack.push(t4);
    }
    if (instruction == 1) {
      t0 = stack.pop();
      t1 = stack.pop();
      t5 = bytecode[index++];
      if (t5) stack.push(t1[t0]);
    }
    if (instruction == 7) {
      t0 = stack.pop();
      t1 = bytecode[index++];
      if (!t0) index = t1;
    }
    if (instruction == 11) {
      if (option.r) {
        return stack.pop();
      }
      return bytecode[index++];
    }
    if (instruction == 10) {
      t0 = bytecode[index++];
      if (!args) args = [].concat(stack);
      for (t1 = t0; t1 >= 0; t1--) {
        t2 = stack.pop();
        localScope[t2] = args[t1];
      }
    }
  }
})(typeof window !== 'undefined' ? window : (window = global, window), 0, [],["@faceless","arguments","__proto__","a","b",22,2,6,4,"c",70,7,"test",100,"console","log"], [3,3,3,4,4,5,12,4,3,0,1,17,0,0,5,3,1,4,6,54,1,12,4,4,0,1,17,0,0,5,4,1,4,7,53,1,5,3,1,4,8,52,1,51,1,13,4,9,0,0,5,9,1,4,10,50,1,7,76,14,74,5,9,1,4,11,54,1,13,4,9,0,0,16,6,76,4,12,4,13,51,1,5,9,1,51,1,12,4,14,1,1,4,15,2,1,0])
````

Comment List

Comment

© Copyright: This article is an original work and the copyright belongs to the Depy's docs unless marked as Reproduced

Please contact the blogger for authorization to reprint

JSVMP执行流程分析

© Copyright: This article is an original work and the copyright belongs to the Depy's docs unless marked as ReproducedPlease contact the blogger for authorization to reprint

© Copyright: This article is an original work and the copyright belongs to the Depy's docs unless marked as Reproduced

Please contact the blogger for authorization to reprint