CS 3710

Introduction to Cybersecurity

 

Aaron Bloomfield (aaron@virginia.edu)
@github | ↑ |

 

Virtual Machines

History

Hardware utilization

  • During the 1960s and 1970s
  • IBM VM/370 operating system
    • Hypervisor (control program) creates virtual machine environment
      • Each user has own virtual machine with guest OS, own address space, virtual devices, etc.

Decline

  • In the 1980s and 1990s, VMs decline in popularity
    • Poor performance
    • Client-server applications
    • Inexpensive PCs
    • Distributed computing
  • VMs started as mainframes (one computer for the entire company)
    • Then cheap PCs showed up

Renewed Interest

  • In the late 1990s
  • Issues/Challenges
    • Low Infrastructure Utilization
    • Physical infrastructure costs
    • IT management costs
    • Failover/disaster protection
    • High maintenance end-user desktops

Renewed Interest cont’d

  • Issues to resolve…
    • Software compatibility
    • Isolation capability
    • Encapsulation
    • Low overhead/high performance

Concepts

Virtual machine types

  • Language level
    • Example: Java Virtual Machine
  • Process level
    • Example: Wine
  • Operating System level
    • Example: Docker, BSD jails
  • System level
    • Example: VirtualBox, VMware, Hyper-V

Terminology

  • Host OS: the operating system that booted the computer
  • Guest thing: what is being run in the virtual machine
    • Here, thing could be a process, operating system, etc.

Language virtual machine

  • The Java Virtual Machine is a classic example
    • Compiler translates .java files to bytecode (.class) files
  • Any interpreter is really a language virtual machine
    • Python, Ruby, PHP, etc.
  • The interpreter / langauge VM can:
    • Handle memory issues (including garbage collection)
    • Do lots of error checking and tracing
    • Still be reasonably fast

JVM diagram

jvm From here

Process virtual machine

  • Allows one to run a single process
  • Meant to provied a platform independent environment for a process to run
    • Abstracting away things like the hardware, host OS, etc.
  • Wine (Windows emulator for Linux) is an example
    • The binary executes, but Windows system calls are translated into Linux system calls
    • A library (Winelib) fills in for Windows system calls that do not have Linux equivalents

Process VM examples

  • Multiprogramming
    • Each user process given the illusion of having complete machine to itself
      • OS supporting multiple user processes
  • Emulators and dynamic binary translators
    • Emulate one instruction set on hardware designed for different instruction set
    • Dynamic binary translation: convert blocks of source instructions to target instructions that perform equivalent functions

Process VM Examples (cont’d)

  • High-level language VMs
    • Cross-platform portability
    • Minimize hardware-specific and OS-specific features
      • Java VM
  • So one can argue that a language VM is really a process VM…
    • But we’ll differentiate them

OS virtual machine

  • The host OS shares the same kernel with all the guest OSes
  • But each guest OS has its own memory space
    • This separates the different guest OSes from each other and the host OS
  • First implementation was FreeBSD jails; Linux’s first implementation was LXC
  • Most common known example now is Docker

OS virtual machine

  • FreeBSD jails are a first-class concept in an OS
    • Meaning something the OS directly supports and implements
  • Containers are not
    • They are built from other first-class concepts (namespaces, cgroups, etc.)
    • That means you can do more, but there are more bugs and complexity
  • See jfrazelle’s blog post for more

OS VM examples

  • Docker is popular
    • Also LXC, LXD
  • FreeBSD jails
  • Solaris Zones

Image from here

System virtual machine

  • Designed to emulate “real” hardwre
  • Presents hardware, not drivers, to the guest
    • Whereas the other VMs will present network connections to the guest, a system VM will present an Ethernet device
    • Likewise, not memory allocation calls, but page tables
  • These run entire operating systems, and take the most resources
  • VirtualBox is the example we all know and love
    • VMWare is popular; Hyper-V is for Windows

System VM examples

  • Classic system
    • Virtual machine monitor (VMM) on bare hardware
    • IBM VM/370

Image from here

System VM examples

hosted vm
  • Hosted VM
    • VM runs on OS
    • VMWare, VirtualBox

Image from here

System VM examples (cont’d)

  • Whole System VMs (Emulation)
    • Complete software system (OS and applications) supported on a host system running a different ISA and OS
      • VirtualPC (Windows on Mac)
  • Codesigned VMs (Hardware optimization)
    • No native applications; VM software is part of hardware implementation
      • Transmeta Crusoe

Terminology confusion

  • Some will eliminate OS level, since that’s just process level on steroids
  • Others will combine language level and process level, since they are both for a single process

Virtualization

  • Formally:
    • Virtualization involves the construction of an isomorphism that maps a virtual guest system to a real host. (Popek and Goldberg, 1974)

Virtualization

  • Mapping of virtual resources or state to real resources on underlying machine
    • E.g., registers, memory, files
  • Emulation of the virtual machine ABI or ISA
    • ABI: Application Binary Interface: how an executable interacts with a library or the OS
    • Use of real machine instructions and/or system calls to carry out actions specified by virtual machine instructions and/or system calls

ABI vs API

abi vs api From wikipedia

VM Operations

OS refresher

  • User programs run in user mode
    • Where a crash (segfault) will kill the process, but not the entire OS
    • Programs are restricted in what they can do
      • They can’t directly access hardware; they have to go through the OS
  • The OS itself runs in kernel mode
    • Where a crash will, in fact, crash the entire OS

User mode vs Kernel mode

  • Boundary is defined by:
    • System calls
      • Called by user-mode programs, but the function body is completed in kernel mode
    • Device drivers
      • Code written by somebody other than the OS writer, but runs in kernel mode
      • (whether they are in kernel mode fully or not varies by OS)

User mode vs Kernel mode

  • If a guest OS tries a priviledged operation:
    • The VM forwards to the VMM
      • VMM: Virtual Machine Monitor
    • The VMM handles it by a system call if possible
      • But it’s in kernel mode, so it can do it directly

VMs and kernel mode

  • VMs tend to run the guest OS in user mode
    • But the OSs think they are running on hardware, so they will attempt to do “priviledged” things (like access hardware directly)
  • The Virtual Machine Monitor (VMM) handles this
    • It is tyipcally a device driver running in kernel mode
  • Exceptions from the OS are passed, through the VMM, back to the VM

VMM diagram

Image from here

Abstraction

Levels of Abstraction

  • Allow implementation details at lower levels of a design to be ignored or simplified
  • Arranged in a hierarchy
    • Lower levels - Hardware
      • Physical components with real properties
    • Higher levels - Software
    • Logical components

Levels of Abstraction: HW/SW Interface

  • High level language (HLL) -> Assembly language
    • Compiler (translation)
  • Assembly Language -> Operating System
    • Assembler (translation/mapping to ISA)
  • Operating System -> Instruction Set Architecture (ISA)
    • Partial Interpretation (device drivers, etc)
  • Instruction Set Architecture -> Digital logic
    • Hardware
  • Digital logic -> Electronics (transistors, etc.)
    • Hardware

Levels of Abstraction: Java

  • Java (HLL) -> Java Byte Code (ISA)
    • Java Compiler
  • Java Byte Code (ISA) -> Java Virtual Machine (JVM)
    • Interpreter and JIT compiler
  • Java Virtual Machine (JVM) -> OS
    • Machine ISA

Well-defined Interfaces

  • Decouple design tasks
    • Instruction set
      • Intel and AMD
        • IA-32 instruction set
      • Compilers
        • Map high level language to ISA
    • Operating system interface
      • Application development

Well-defined interfaces: Challenges

  • Reduced interoperability
    • Processors support limited instruction sets
      • IA-32 vs. PowerPC
    • Different operating systems
      • Windows vs. Linux
    • Application binaries
      • Dependent on OS and instruction set
  • Hardware resource considerations
    • OS

VM Examples

Hosted VM: VMWare

VMWare Workstation

  • Virtual machine software suite for x86 and x86-64 computers
  • Supported host OS’s
    • Windows and Linux
  • Supported guest OS’s
    • Over 200 supported
    • Windows, Linux, BSD variants
  • Allows one physical machine to run multiple operating systems simultaneously

VMWare Workstation

  • Uses bridging to provide mapping to
    • Network adaptors
    • CD-ROMs
    • Hard disk drives
    • USB devices
  • Simulation capability for some hardware
    • Mounting ISO as CD-ROM
    • Mounting .vmdk files as hard disks

VMWare Workstation

  • Snapshot functionality
    • Enables rollback to saved status
  • VMWare Tools
  • Package with drivers and other software that can be installed in guest operating systems to increase their performance
    • Drivers for emulated hardware
    • Drag and drop file support
    • Clipboard sharing between host and guest

VMs and Malware

Why do we care about VMs?

  • We can run malware within them to ‘sandbox’ them
    • This allows pausing the execution at various points
  • We can simulate other architectures / software systems to study malware
  • If we are trying to “break” things, that “breaking” is contained
    • Experimenting with fork() for example – we don’t want to fork bomb our host machine

But are you in a VM?

  • Meta-question: are you living in a computer simulation
    • We’ll leave that question for the philosophy classes…
  • In Linux, you can install and run virt-what, which will determine if you are in a VM or not

virt-what

VM Security

  • Docker is known to be rather insecure
    • Containers use the same OS, so a kernel bug will exist in both
  • There have been “break-outs” of system VMs as well
    • One from March 2017

VM Break-out

vm break out Article is here

https://aaronbloomfield.github.io/ics