Guest Post: The Curious Case of the Intel Microcode Part #2 – It Gets Better — Then Worse

This guest post by Bjørn Anders Jørgensen, Senior Systems Consultant Basefarm, first appeared on LinkedIn.

Before you start on this rather long post, have a go at part #1:


This is a long read. To get to the juicy part on how Intel potentially shipped pre-release microcode to partners, skip to section 3. The short, short version is that the official Intel microcode update contains newer microcode for Skylake-SP and Kaby Lake/Coffe Lake than what currently is shipping from VMware/HPE/DellEMC etc.

Section 1: The good

Last week I wrote about how Intel should improve their microcode update delivery mechanism and offer full disclosure on their microcode changes. Then this week events progressed in rapid succession:

  • Intel released updated microcode bundle 20180108
  • VMware released updated patches for vSphere, including microcode updates
  • Then it was discovered that the updates cause stability issues
  • Most computer vendors recalled BIOS updates for Haswell/Broadwell
  • VMware recommended not to expose VMs to the new CPU feature flag
  • I discovered that Xeon SP and Kaby Lake/Coffe Lake updates from most or all vendors is based on the pre-release Intel bundle 20171215!

First of all I have to say I feel for all the engineers and product managers taken by surprise when the news broke early on the Meltdown and Spectre vulnerabilities. Apparently the embargo was suppose to be lifted on the 9. January, and everyone was working towards this date. There must have been extreme pressure from peers and customers. Mistakes will be made, but could have been avoided if Intel was more transparent on their process and releases.

Lets start with the Intel microcode bundle. Deconstructing the bundle using MC extractor you’ll find 94 binary files containing 196 individual microcode updates. Intel has released a single bundle with all updates going back to 1998. This is good! As far as I know there has not been a single update bundle going back this far. Additionally all vSphere patches contains microcode updates for most production systems running vSphere, including an update for AMD EPYC processors as described in the security advisory. You can find all the details on my github repository.

This is progress – however there are still no release notes or details except for the updated microcode versions for each processor family:

- Updates upon 20171117 release --
IVT C0          (06-3e-04:ed) 428->42a
SKL-U/Y D0      (06-4e-03:c0) ba->c2
BDW-U/Y E/F     (06-3d-04:c0) 25->28
HSW-ULT Cx/Dx   (06-45-01:72) 20->21
Crystalwell Cx  (06-46-01:32) 17->18
BDW-H E/G       (06-47-01:22) 17->1b
HSX-EX E0       (06-3f-04:80) 0f->10
SKL-H/S R0      (06-5e-03:36) ba->c2
HSW Cx/Dx       (06-3c-03:32) 22->23
HSX C0          (06-3f-02:6f) 3a->3b
BDX-DE V0/V1    (06-56-02:10) 0f->14
BDX-DE V2       (06-56-03:10) 700000d->7000011
KBL-U/Y H0      (06-8e-09:c0) 62->80
KBL Y0 / CFL D0 (06-8e-0a:c0) 70->80
KBL-H/S B0      (06-9e-09:2a) 5e->80
CFL U0          (06-9e-0a:22) 70->80
CFL B0          (06-9e-0b:02) 72->80
SKX H0          (06-55-04:b7) 2000035->200003c
GLK B0          (06-7a-01:01) 1e->22

Note that 20171117 is Intels previous official release

Keep a mental note for version 80 for Kaby Lake/Coffee Lake and 200003c for Skylake-SP (Xeon SP). Also note the absence of Sandy Bridge updates.

Section 2: The bad

With new microcode and OS patches available, tens if not hundreds of thousands of IT professionals spend much of last week planning and executing mitigation for the Meltdown and Spectre vulnerabilities. And with Spectre variant #2 (Branch Target Injection) requiring microcode update, I will assume most also started deploying updated BIOS and Intel patches through the OS mechanism for Linux and vSphere.

With the wider deployment it did not take long for customers to discover there where issues with system crashes or “higher system reboots” as Intel marking tries to spin it. How are system administrators suppose to make a informed decision based on that? It is quite ironic that Intel CEO, Brian Krzanich at the same time makes his “Security-First Pledge” offering “Transparent and Timely Communications”. It seems like Lenovo is the only vendor with some actual details on the issues:

  1. (Kaby Lake U/Y, U23e, H/S/X) Symptom: Intermittent system hang during system sleep (S3) cycling. If you have already applied the firmware update and experience hangs during sleep/wake, please flash back to the previous BIOS/UEFI level, or disable sleep (S3) mode on your system; and then apply the improved update when it becomes available. If you have not already applied the update, please wait until the improved firmware level is available.
  2. (Broadwell E) Symptom: Intermittent blue screen during system restart. If you have already applied the update, Intel suggests continuing to use the firmware level until an improved one is available. If you have not applied the update, please wait until the improved firmware level is available.
  3. (Broadwell E, H, U/Y; Haswell standard, Core Extreme, ULT) Symptom: Intel has received reports of unexpected page faults, which they are currently investigating. Out of an abundance of caution, Intel requested Lenovo to stop distributing this firmware.

With all the system manufacturers removing their BIOS downloads for Haswell and Broadwell and VMware recommending to “hide the speculative-execution control mechanism for virtual machines” if patches have already applied, this cannot be described as nothing short of an industry wide recall. It is hard to say how bad it is, but if it is so bad that Intel is recalling the update and CEO and EVP posting publicly I think it is wise to await further updates and see how things pan out. Intel will release updated microcode soon and the various vendors will post new BIOS updates in the coming few weeks. Go to the VMware link and note the various processors affected in yellow, because it gets worse:

Section 3: The ugly

Disclaimer: The following analysis is made with best effort on little information. Intel is shipping old and new versions in their microcode update bundle. It may be that they are for different CPU steppings of the same processor.

While all this is unfolding I am trying to piece together what the Intel and VMware patches actually contains. Using MC extractor I was able to inspect the microcode bundles, map the updates to CPU models and cross reference different releases. The complete list is on github. In the rush to get patches out, it seems like Intel made some last minute changes that was not communicated to partners. I have been in contact with several vendors and none seems to be aware that there changes made between the unofficial pre-release bundle from 15. December that was shared under embargo. Luckily this only affect three processor generations, but combined with the Haswell/Broadwell issues, it leaves us with an awful mess and a potential open vulnerability.

To quote my one week younger self:

So with six months lead time Intel has not been able to release an official microcode bundle update. The latest official one is from 17. November and does not seem to contain any Spectre fixes. Worse, it seems like the unofficial bundle is only partially fixing variant #2/Spectre.

It is obvious that Intel made some last minute changes to microcode and failed to notify partners. What changes has been made between update bundle 20171215 and 20180108? Are we installing incomplete non-released microcode to millions of computers? Only Intel knows! It might be as simple as an effort to include some minor fixes now what everyone with an Intel processor have to update their computer, or it can be a monumental error where many computers are still partially vulnerable to Spectre even with patches installed. With the “security first” pledge from CEO Brian Krzanich promising transparency I expect and demand an explanation from Intel. How is it that VMware, HPE and DellEMC can ship pre-release microcode to millions of computers?

Intel microcode update bundle pre-release 15. December
| CPUID |         Platform         | Version |    Date    | Release |   Size  | Checksum |  Offset  | Latest |
| 50654 |  B7 [0, 1, 2, 4, 5, 7]   | 200003A | 2017-11-21 |   PRD   |  0x6C00 | C088D252 | 0x64400  |  Yes   |
| 806E9 |        C0 [6, 7]         |    7C   | 2017-12-03 |   PRD   | 0x18000 | 5C75A5FE | 0x8B000  |   No   |
| 806EA |        C0 [6, 7]         |    7C   | 2017-12-03 |   PRD   | 0x17C00 | B81BC926 | 0xA3000  |  Yes   |
| 906E9 |       2A [1, 3, 5]       |    7C   | 2017-12-03 |   PRD   | 0x18000 | 6CF72404 | 0xBAC00  |   No   |
| 906EA |        22 [1, 5]         |    7C   | 2017-12-03 |   PRD   | 0x17800 | 55695D1F | 0xD2C00  |  Yes   |
| 906EB |          02 [1]          |    7C   | 2017-12-03 |   PRD   | 0x18000 | 5046D998 | 0xEA400  |  Yes   |

Intel microcode update bundle released 8. january
| CPUID |         Platform         | Version |    Date    | Release |   Size  | Checksum |  Offset  | Latest |
| 50654 | B7 [0, 1, 2, 4, 5, 7]    | 200003C | 2017-12-08 |   PRD   | 0x6C00  | A4059069 |  0x0     |  Yes   |
| 806E9 | C0 [6, 7]                |    80   | 2018-01-04 |   PRD   | 0x18000 | 6961A256 |  0x0     |  Yes   |
| 806EA | C0 [6, 7]                |    80   | 2018-01-04 |   PRD   | 0x18000 | F6263DAE |  0x0     |  Yes   |
| 906E9 | 2A [1, 3, 5]             |    80   | 2018-01-04 |   PRD   | 0x18000 | 6AA1DE93 |  0x0     |  Yes   |
| 906EA | 22 [1, 5]                |    80   | 2018-01-04 |   PRD   | 0x17C00 | 84CABC68 |  0x0     |  Yes   |
| 906EB |  02 [1]                  |    80   | 2018-01-04 |   PRD   | 0x18000 | D24EDB7F |  0x0     |  Yes   |

Note that all updates are labeled PRD for production. If we take the VMware matrix and mark the recalled microcode in yellow and the potential pre-release in red, we get a pretty depressing chart:

The same versions was also shipped in Dell BIOS updates. Thanks to Dell engineering for actually listing the microcode updates, I have not seen the same transparency from other vendors. HPE take note!


Updated the Intel Xeon Processor Scalable Family Processor Microcode to version 0x3A.


Updated the Xeon Processor E5-2600 v3 Product Family Processor Microcode to version 0x3B
Updated the Intel Xeon Processor E5-2600 v4 Product Family Processor Microcode to version 0x0b000025


Updated the Xeon Processor E5-2600 v3 Product Family Processor Microcode to version 0x3B
Updated the Intel Xeon Processor E5-2600 v4 Product Family Processor Microcode to version 0x0b000025


Updated the Intel Xeon processor E7-4800/8800 v3 Product Family Processor Microcode to version 0x10
Updated the Intel Xeon processor E7-4800/8800 v4 Product Family Processor Microcode to version 0x0b000025

The Rx30 updates have now been removed from the Dell web page due to the aforementioned issues, but the R740 update is still there. Should it be recalled as well? Hard to tell, I think an update from Intel is in order.

I have not been able to attempt to load the new microcode on a server to confirm that it actually would apply, but I tested on my Precision 5520 laptop that use a Xeon E3-1505M v6 processor with the help of the VMware microcode update driver for Windows and the microcode will update from 7C to 80:

Some advice

So what is a system administrator to do? My advice is to wait it out. The situation is still fluid and changes almost on a daily basis. Let the dust settle and hope for some more clarity in the coming week. Remember that microcode updates are only for Spectre variant #2 vulnerability which is the hardest to exploit. If you already started patching vSphere, consider a rollback or implement the workaround from VMware.

Focus on patching Meltdown on soft targets such as laptops. These devices are much more likely to be running untrusted code. Dont forget applications, especially browsers. Researchers has been able to exploit meltdown and read non-cached data. The hackers can not be far behind. Combining Meltdown with a remote exploit could be disastrous.

As everything except Rasberry Pi is vulnerable to Spectre, focus on mobiles and tablets. Map you vulnerable devices and vendor updates. Pace yourself, we will be fighting this for years to come. There will be more side-channel vulnerabilities in the future and this is not the last round of patching.

Keep in mind that there are no microcode updates for Sandy Bridge processors yet. This is why Dell and HPE is holding back updates for Poweredge G12 servers and Proliant Gen8 servers. If you have a lot of Xeon 4600/2600/1600/1400/1200 V1 servers, you will have to wait on Intel and server vendors if you want full mitigation for Spectre.

Guest Post: The Curious Case of the Intel Microcode

This guest post by Bjørn Anders Jørgensen, Senior Systems Consultant Basefarm, first appeared on LinkedIn.

Disclaimer: This is a report based on current development as of 7. January, the situation is changing by the hour so read this opinion piece with that in hindsight.








Unless you have been living under a rock for the last week you will know by now that there is a universal design flaw in most modern microprocessors, leaving them vulnerable to a serious information disclosure problem that requires updates to all operating systems and processors.

If you are not familiar with the issue, start here: Meltdown and Spectre

Jan Wildeboer also has a comprehensive timeline: How we got to #Spectre and #Meltdown A Timeline

The issue has been known by Intel at least since June, and has been under embargo while everyone has been hammering out code to mitigate the threat and be ready when the embargo was lifted.

So we have three vulnerabilities, one that requires microcode update:

CVE-2017-5715 (variant #2/Spectre) aka branch target injection

Microcode is a small piece of code that can be loaded by the processor at boot time, either from the BIOS or from the operating system. Intel has been using this method for more than a decade to solve bugs and issues. It is a well known process. All HW vendors ships regular BIOS updates containing microcode updates and most modern operating systems has facilities to distribute and deploy microcode updates. This includes Linux, Windows and ESXi server. While most Linux distros keep an updated microcode repository, Microsoft and VMware has occasionally added microcode updates for serious issues.

So with everything blowing up this week, you’d think Intel would prepare microcode updates and ensure that OS patches contained the updated microcode so that you were completely secure from the know vulnerabilities. (more on the unknowns later)

No! We all have to wait for the HW vendors to release updated BIOS for all our computers. They are of course completely overloaded and will prioritize new models. Currently (7. Jan) on the server side, Dell has pretty good coverage for 13G and 14G servers and HPE has pretty good coverage for 9G and 10G servers. Anything older than 3 years are still vulnerable even after patching the OS/hypervisor. On the consumer side, it is all over the place.

Disclaimer: The Register claims that variant #2/Spectre only requires microcode updates for Skylake and newer microarchitecture, in which case the tier 1 vendors already have good coverage, but the tier 2 and tier 3 vendors will lag for months. Even Cisco is not planning BIOS updates until 14. February!

Intel is as normal very, very quiet regarding microcode updates. They usually come with no or little release notes, and is basically treated like a black box, even though it becomes an integral part, changing critical functions in the CPU. The only content I’ve found was from an updated Debian bug report:

New upstream microcodes to partially address CVE-2017-5715
     + Updated Microcodes:
       sig 0x000306c3, pf_mask 0x32, 2017-11-20, rev 0x0023, size 23552
       sig 0x000306d4, pf_mask 0xc0, 2017-11-17, rev 0x0028, size 18432
       sig 0x000306f2, pf_mask 0x6f, 2017-11-17, rev 0x003b, size 33792
       sig 0x00040651, pf_mask 0x72, 2017-11-20, rev 0x0021, size 22528
       sig 0x000406e3, pf_mask 0xc0, 2017-11-16, rev 0x00c2, size 99328
       sig 0x000406f1, pf_mask 0xef, 2017-11-18, rev 0xb000025, size 27648
       sig 0x00050654, pf_mask 0xb7, 2017-11-21, rev 0x200003a, size 27648
       sig 0x000506c9, pf_mask 0x03, 2017-11-22, rev 0x002e, size 16384
       sig 0x000806e9, pf_mask 0xc0, 2017-12-03, rev 0x007c, size 98304
       sig 0x000906e9, pf_mask 0x2a, 2017-12-03, rev 0x007c, size 98304
   * Implements IBRS and IBPB support via new MSR (Spectre variant 2
     mitigation, indirect branches).  Support is exposed through cpuid(7).EDX.
   * LFENCE terminates all previous instructions (Spectre variant 2
     mitigation, conditional branches).

Note the use of “partially”. We’ll get back to that in a minute. Breaking open the package, the changelog says:

"unofficial bundle with CVE-2017-5715 mitigation"

So with six months lead time Intel has not been able to release an official microcode bundle update. The latest official one is from 17. November and does not seem to contain any Spectre fixes. Worse, it seems like the unofficial bundle is only partially fixing variant #2/Spectre.

Why is it that Intel is not using the tried and tested OS microcode update method to resolve what is arguably one of the most serious IT security issues in history? It could easily be included with the patch sets for the major OS vendors, like for Debian and other Linux distros. It has been done for VMware ESXi server and Microsoft Windows for less serious issues. And with the microcode revision that “partially address CVE-2017-5715” matching what the server vendors are releasing, do we have more updates coming?

Intel really needs to get a grip on this. These kind of side channel attacks are a completely new vulnerability category, and now that researchers are focusing on this, there will be more coming. As Anders Fogh said in a early blog post:

“While I did set out to read kernel mode without privileges and that produced a negative result, I do feel like I opened a Pandora’s box. “

So we have possibly a partial patch currently and most likely new vulnerabilities coming down the road, I absolutely implore Intel to get a grip and prepare for a quick microcode update and distribution. There is no way that the server vendors will be able to release another BIOS update for a huge amount of servers. The only sensible way is to use the OS update mechanism, even though that is not persistent and could potentially be tampered with. And please, we need full disclosure. I suspect a lot of people believe they have resolved the vulnerability, while in fact they have not.

For virtual machines there are more steps required for the VM to enable the patch. More on that in a follow up post.

Note: We do have some unofficial Intel microcode repositories by BIOS modders and enthusiasts. While I applaud the effort and work, it really should not be necessary, and while the microcode is signed by Intel, modifying the inner workings of the CPU opens for absolutely undetectable persistent hacks and should only come from an official source.