skylake/kabylake bug?

[ghanwani]

Have you guys heard of this? Any idea if it's an issue for Macs/macOS?

https://arstechnica.com/information...have-a-crash-bug-with-hyperthreading-enabled/

(To credit my original source, I learned about the issue from http://market-ticker.org/akcs-www?post=232173)

 

[fokmik]

Nope

 

[treekram]

From what's been published, the bug affects Macs with Skylake or Kabylake CPU's which with hyper-threading (some iMacs don't have hyper-threading CPU's). So MBP-wise, that would be the 2016 and 2017 MBP's. It's triggered under certain circumstances and seems rare for a typical user to encounter it. The fix is supposed to be making it's way to the system vendors. For Apple, this will probably take the form of a firmware update. The bug can be avoided by disabling hyper-threading if you have a hyper-threading CPU. On the web, it says to invoke the following command from the Terminal:

sudo nvram SMT=0 (to re-enable, do sudo nvram -d SMT). The computer has to be re-booted for this to take effect.

I haven't tried this so I can't vouch for this solution. My opinion would be to wait for Apple to issue the firmware (or other) update. I have a non-Mac Skylake computer and I plan on waiting for the BIOS update to appear.

 

[T'hain Esh Kelch]

Unless you are doing specific high end calculations or professional work, which keeps crashing, you will likely not ever encounter this bug. So just wait it out, as treekram suggests above.

 

[Ries]

Unless you are doing specific high end calculations or professional work, which keeps crashing, you will likely not ever encounter this bug. So just wait it out, as treekram suggests above.

Doesn't require high end calculations or professional work in any way to crash. Just 2 tight loops running on hyperthreading on the same core using AH/BH/CH/DH (and wide counter part) registers while the cpu is in some state were it can happen. It can happen for any program that meets the conditions, it can run 100 times fine and fail on 101 when the stars align. The microcode update seems to fix it.

 

[fokmik]

there is no problems regarding any mac...i have 2 mbp 2017 and none have experience any crushes, hyper-threading enabled and used almost all the time. imacs no need to worry. 12" Macbook does not have i think, so, no not 1 mac has this issue

 

[Ries]

there is no problems regarding any mac...i have 2 mbp 2017 and none have experience any crushes, hyper-threading enabled and used almost all the time. imacs no need to worry. 12" Macbook does not have i think, so, no not 1 mac has this issue

Yes they do.

https://caml.inria.fr/mantis/view.php?id=7452

"Two macbooks encountering the issues seems to be using the latest microcode available, cpu infos:"

 

[treekram]

All the Retina Macbooks have hyper-threading CPU's (checked only base models, I presume upgraded CPU's are also HT). 2016 and 2017 models have Skylake/Kaby Lake.

Yes they do.

https://caml.inria.fr/mantis/view.php?id=7452

"Two macbooks encountering the issues seems to be using the latest microcode available, cpu infos:"

The last quote, taken by itself, is a bit misleading. It was made in January of this year. At the very bottom of the thread, they mention they have microcode from Intel, obtained in late May and this seems to have solved the issue to the point that they've closed the issue.

 

[ghanwani]

All the Retina Macbooks have hyper-threading CPU's (checked only base models, I presume upgraded CPU's are also HT). 2016 and 2017 models have Skylake/Kaby Lake.

The last quote, taken by itself, is a bit misleading. It was made in January of this year. At the very bottom of the thread, they mention they have microcode from Intel, obtained in late May and this seems to have solved the issue to the point that they've closed the issue.

Is there some way to tell if a given macOS machine has the fix or not?

 

[treekram]

Is there some way to tell if a given macOS machine has the fix or not?

I think the people who got the microcode fix from Intel (in the link in post #7) got it because they reported it to Intel and Intel wanted them to test the fix (speculation on my part). I don't think that Apple has made the fix available to the public yet. For Macs, I would guess that the fix would come in the form of a firmware update and the way to check would be to make sure you have the latest firmware. Because of the publicity that has taken place over the last few days, the fix, when available, will probably also be widely reported.

For MBP's, Apple may decide to bundle it in the upgrade with Sierra 10.12.6 as the only MBP's affected are those that can run Sierra.

 

[Count Blah]

there is no problems regarding any mac...i have 2 mbp 2017 and none have experience any crushes, hyper-threading enabled and used almost all the time. imacs no need to worry. 12" Macbook does not have i think, so, no not 1 mac has this issue

Ahhh, the old 'I've not seen it happen/recognized it happening, so therefor the problem doesn't exist' routine.

Interesting.

 

[ghanwani]

Because of the publicity that has taken place over the last few days, the fix, when available, will probably also be widely reported.

Was the bug ever discussed on MR on the main page? I don't recall seeing it, but I must admit, sometimes a few days go by between checking MR.

 

[Ries]

All the Retina Macbooks have hyper-threading CPU's (checked only base models, I presume upgraded CPU's are also HT). 2016 and 2017 models have Skylake/Kaby Lake.



The last quote, taken by itself, is a bit misleading. It was made in January of this year. At the very bottom of the thread, they mention they have microcode from Intel, obtained in late May and this seems to have solved the issue to the point that they've closed the issue.

And until Apple releases a firmware update including that microcode, the MacBooks with HT + skylake/kabylake are affected. Just like any computer out there without it.

 

[fokmik]

Yes they do.

https://caml.inria.fr/mantis/view.php?id=7452

"Two macbooks encountering the issues seems to be using the latest microcode available, cpu infos:"

Status closed Resolution no change required

 

[peraspera]

The original advisory was posted on Debian on Sun, 25 Jun 2017 09:19:36 -0300, https://lists.debian.org/debian-devel/2017/06/msg00308.html

The Register picked up the story, http://www.theregister.co.uk/2017/06/25/intel_skylake_kaby_lake_hyperthreading/.
"Symptoms can include “application and system misbehaviour, data corruption, and data loss”.

Henrique de Moraes Holschuh, who authored the advisory, notes that all operating systems, not only Linux, are subject to the bug.

"Intel's errata note (for example in this document), describes the issue like this:

Under complex micro-architectural conditions, short loops of less than 64 instructions that use AH, BH, CH or DH registers as well as their corresponding wider register (eg RAX, EAX or AX for AH) may cause unpredictable system behaviour. This can only happen when both logical processors on the same physical processor are active."

I also heard Steve Gibson mention on his Security Now podcast that this bug also poses a currently unexploited security vulnerability. When the recorded podcast goes up it will be on Steve's site, https://www.grc.com/securitynow.htm.

 

[S8ER01Z]

My microcode_version updated with todays Beta release, hopefully this stops the random bizarre lockups I've been seeing.

2016 MBP with i7 6660u
machdep.cpu.model: 78
machdep.cpu.stepping: 3
machdep.cpu.microcode_version: 186 (It was 174 prior)

 

[treekram]

My microcode_version updated with todays Beta release, hopefully this stops the random bizarre lockups I've been seeing.

2016 MBP with i7 6660u
machdep.cpu.model: 78
machdep.cpu.stepping: 3
machdep.cpu.microcode_version: 186 (It was 174 prior)

From what's known about this bug, it can be consistently triggered by a program if it fits a certain profile. Part of that profile is that it makes heavy use of the CPU and uses hyper-threading. How long before the CPU encounters the error varies but it doesn't happen immediately. For people wondering if the bug affects them (or affected them since Apple likely has applied the patch in recent updates), if the problems they see always involves the same program and if the CPU virtual cores (4 in the case of the i7 6660U) are at/near 100% utilization when the problem occurs, then it's possible, otherwise, it's very unlikely that this particular bug is involved. Personally, I think that the coding technique used that uncovered this bug is unusual and it may very well be that the program which uncovered this bug is the only one which will trigger it.

 

[S8ER01Z]

I agree with you, I just run a heavily utilized system with Windows 10 in a VMWare instance. The system has not locked without the VMWare instance running but if I am running the system near max for periods of time the system would randomly lock/quit responding with the mouse moving. (It's not completely frozen though, hard to explain). So far it has not occurred after the latest update, I'm not definitively saying this was the cause but it would explain some things. It could have also been something in the last release of 10.13 as well so I have no way to know.

 

[treekram]

I agree with you, I just run a heavily utilized system with Windows 10 in a VMWare instance. The system has not locked without the VMWare instance running but if I am running the system near max for periods of time the system would randomly lock/quit responding with the mouse moving. (It's not completely frozen though, hard to explain). So far it has not occurred after the latest update, I'm not definitively saying this was the cause but it would explain some things. It could have also been something in the last release of 10.13 as well so I have no way to know.

Well, considering that you have a VM instance running a beta version of an OS, I would think that would be a more likely candidate for your past issues. You could have easily tested whether this bug was the likely cause of your issues at an earlier point by turning off hyper-threading. You wouldn't have suffered that major a performance hit (15% maybe, probably less). But at this point, unless you want to take the trouble to revert back your computer to the pre-update state or find a similar computer that doesn't have the likely firmware fix and run a clone of your computer there, I don't know how one can say with any conviction that the Intel bug was causing your issues.