Jump to content

Help installing Mojave on Xeon W-2175 and Asus WS C422 mobo

obus

By obus,
in OSx86 10.14 (Mojave)

 Share
852 posts in this topic

Recommended Posts

eritius
eritius
Posted
Posted
On 6/14/2019 at 4:13 PM, Balamut said:

 

The chipset will be different, c621 instead of c422. I'm very curious to see if Xeon Golds work on the MacOS.

 

 

 

Me too! I'm getting a setup with a C621 chipset and dual Xeon Golds up and running at the moment. I've encountered a lot of the issues mentioned here. Unfortunately, the same CPUID patch (0x0506E3) causes all sorts of issues with my setup. For some reason, it causes random memory corruption in the kernel which results in kernel panics in all sorts of extensions and messages like: 

"Zone cache element was used after free!"

and 

"a freed zone element has been modified"

So far, the only workaround I've found is to use a less compatible CPUID 0x040674. In the handful of cases where I got luck and the memory scribbling didn't land on anything critical, the dual processors showed up as Xeon Ws in About This Mac. Geek Bench scores were not where I would expect them to be, though, so there are some issues to sort out even if that becomes stable.

 

As well, as of Catalina beta 5, the C621 chipset doesn't appear to be entirely support. USB all work out of the box which is great, but there's no driver for the X722 ethernet ports, or at least not one that comes up automatically for the 1Gb ports on this motherboard.

 

If anyone has any suggestions for things to try, I'm all ears!

Link to comment
Share on other sites
obus
obus
Posted
  • Author
Posted
6 hours ago, eritius said:

 

Me too! I'm getting a setup with a C621 chipset and dual Xeon Golds up and running at the moment. I've encountered a lot of the issues mentioned here. Unfortunately, the same CPUID patch (0x0506E3) causes all sorts of issues with my setup. For some reason, it causes random memory corruption in the kernel which results in kernel panics in all sorts of extensions and messages like: 


"Zone cache element was used after free!"

and 


"a freed zone element has been modified"

So far, the only workaround I've found is to use a less compatible CPUID 0x040674. In the handful of cases where I got luck and the memory scribbling didn't land on anything critical, the dual processors showed up as Xeon Ws in About This Mac. Geek Bench scores were not where I would expect them to be, though, so there are some issues to sort out even if that becomes stable.

 

As well, as of Catalina beta 5, the C621 chipset doesn't appear to be entirely support. USB all work out of the box which is great, but there's no driver for the X722 ethernet ports, or at least not one that comes up automatically for the 1Gb ports on this motherboard.

 

If anyone has any suggestions for things to try, I'm all ears!

Are you running clover? And if that's the case may I see your EFI folder.

 

Link to comment
Share on other sites
obus
obus
Posted
  • Author
Posted (edited)

Yes send me your old Clover EFI and the OC EFI you use for the moment.

By the way do you have CFG lock setting in bios set to disabled?

 

Edited by obus
Link to comment
Share on other sites
obus
obus
Posted
  • Author
Posted (edited)

Which can means your MsrE2 register is locked. You can check more details about this here:https://github.com/acidanthera/AppleSupportPkg#verifymsre2

Try embedded config.plist and remove  USBInjectAll.kext from your kext-folder. In embedded config.plist I have checked XhciPortLimit instead. (seeC onfiguration page 14 chapter 6.7.10) The best way is to make a XHCI.kext.zip with a 15 ports limitation (see embedded example).

I'am not using TSCAdjustReset.kext for my rig. Try without and see if that make any difference in performance.

You need SSDT-PLUG.aml to get native powermanagment. (Embedded you find an example from one of my rigs.)

What kind of graphics are you running? iGPU only or???

If you have a HiDPI screen that value should be 02 instead of 01.

Why npci=0x2000?? Test with that bootflag removed.

Set csr-active-config with csr utility enabled respectively disabled in recovery partition instead of using csr-active-config -> <e7030000>. (I changed that to <00000000> in embedded config.plist). To get into recovery mode you have to swap VBoxHfs.efi too embedded HFSPlus.efi. (I changed that in config.plist)

If you use ConsoleBehaviourOs and/or ConsoleBehaviourUi you need to set ConsoleControl to YES. (I've changed that in embedded config.plist)

 

Don't forget to put back serials and mlb and so on in platform info.

config.plist

Edited by obus
Link to comment
Share on other sites
eritius
eritius
Posted
Posted

Thanks for taking a look at all this! For what it's worth, both configurations do boot and are stable on my system. Performance isn't where it should be and I get warnings about both "Unknown CPU" and "Unknown PCH". But most everything works without issue other than the built-in ethernet. I haven't gotten the built-in audio working yet either but it's not high on my list right now.

 

24 minutes ago, obus said:

Which means your MsrE2 register is locked. You can check more details about this here:https://github.com/acidanthera/AppleSupportPkg#verifymsre2

Try embedded config.plist and remove  USBInjectAll.kext from your kext-folder. In embedded config.plist I have checked XhciPortLimit instead. (seeC onfiguration page 14 chapter 6.7.10) The best way is to make a XHCI.kext.zip with a 15 ports limitation (see embedded example). 

This all is a work-in-progress. I don't believe USBInjectAll is even loaded in the version I sent you. I'm working to make a kext.

 

26 minutes ago, obus said:

I'am not using TSCAdjustReset.kext for my rig. Try without and see if that make any difference in performance. 

I haven't found it makes a difference either way as far as booting at least. It seems to run fine without it but I haven't done a benchmark with it yet. I'll do that.

 

26 minutes ago, obus said:

You need SSDT-PLUG.aml to get native powermanagment. (Embedded you find an example from one of my rigs.) 

Bigger problem I suspect is I can't boot with my native CPUID (0x050654) or any of the alternates you all are using. 0x0506E3 causes random memory corruption in the kernel. It boots maybe 1 in 20 times if I get lucky and the memory overwriting doesn't scribble over something important. More likely, it causes kernel panics in all sorts of random locations.

 

28 minutes ago, obus said:

What kind of graphics are you running? iGPU only or??? 

Not sure why my signature isn't showing up. I'll have to get that sorted. The hardware configuration is:

Motherboard: Supermicro X11DAi-N

CPU: 2x Xeon Gold 6136 3.0 Ghz 12-core

GPU: Sapphire Pulse Radeon RX 580 8GB

RAM: 128 GB (4x32GB)

 

30 minutes ago, obus said:

Why npci=0x2000?? Test with that bootflag removed. 

 

It doesn't seem to make a difference with or without. People questioned why I didn't have it elsewhere so I put it back. It works fine either way.

 

31 minutes ago, obus said:

Set csr-active-config with csr utility enabled respectively disabled in recovery partition instead of using csr-active-config -> <e7030000>. (I changed that to <00000000> in embedded config.plist). To get into recovery mode you have to swap VBoxHfs.efi too embedded HFSPlus.efi. (I changed that in config.plist) 

It's been set to 0 up until about 3 hours ago when I was just doing some testing and I sent this to you. For what it's worth, I was able to boot into recovery just fine with VBoxHfs.efi, though it took a hot minute and I thought it had stalled at first.

 

33 minutes ago, obus said:

If you use ConsoleBehaviourOs and/or ConsoleBehaviourUi you need to set ConsoleControl to YES. (I've changed that in embedded config.plist) 

I'll take a look in a bit to see if this makes any difference. My reading of the documentation implies that you only need ConsoleControl when using the Force* options.

 

Link to comment
Share on other sites
eritius
eritius
Posted
Posted
41 minutes ago, obus said:

Which means your MsrE2 register is locked. You can check more details about this here:https://github.com/acidanthera/AppleSupportPkg#verifymsre2

 

So strangely, when I run that tool, it says all my cores are unlocked which I assume is good. That said, I seem to need to apply a bunch of fixes for what I assume are related, e.g. AppleXcpmCfgLock and AppleXcpmExtraMsrs in OpenCore or the xcpm_pkg_scope_msrs kernel patch in Clover.

Link to comment
Share on other sites
obus
obus
Posted
  • Author
Posted
3 hours ago, eritius said:

So strangely, when I run that tool, it says all my cores are unlocked which I assume is good. That said, I seem to need to apply a bunch of fixes for what I assume are related, e.g. AppleXcpmCfgLock and AppleXcpmExtraMsrs in OpenCore or the xcpm_pkg_scope_msrs kernel patch in Clover.

Good news then.

Try this setup in your config.plist

config-1.plist

Link to comment
Share on other sites
obus
obus
Posted
  • Author
Posted (edited)


 

4 hours ago, eritius said:

GPU: Sapphire Pulse Radeon RX 580 8G

Native. Works OOB.

4 hours ago, eritius said:

It's been set to 0 up until about 3 hours ago when I was just doing some testing and I sent this to you. For what it's worth, I was able to boot into recovery just fine with VBoxHfs.efi, though it took a hot minute and I thought it had stalled at first.

HFSPlus.efi is booting into recovery immediately VBoxHfs.efi can take up to 1 minute before you get into the recovery environment.

4 hours ago, eritius said:

Bigger problem I suspect is I can't boot with my native CPUID (0x050654) or any of the alternates you all are using. 0x0506E3 causes random memory corruption in the kernel. It boots maybe 1 in 20 times if I get lucky and the memory overwriting doesn't scribble over something important. More likely, it causes kernel panics in all sorts of random locations.

My rig is booting perfect on this Broadwell 0x040674 CPUID too. Use that until you have fixed your SSDT-PLUG.aml (native PM).

When you have this in IOReg everything is cool. 

 

269121196_Screenshot2019-08-05at21_40_10.thumb.png.59c94729ea5ccd9daa7b1c90a4abd70e.png

Our processors don't boot with the "Apple native" CPUID either.

 

Send me a copy of your AppleIntelInfo.rtf and a copy of terminal output from command:

sysctl -a | grep xcpm

Edited by obus
Link to comment
Share on other sites
eritius
eritius
Posted
Posted
53 minutes ago, obus said:

Send me a copy of your AppleIntelInfo.rtf

 

 

Alright, took me a sec to figure out what you were talking about but here you go. Also, I found out how to patch the EFI variables holding the CFG Lock setting that is hidden in my BIOS. I'm working to get that disabled on both CPUs at the moment.

 

AppleIntelInfo.dat

Link to comment
Share on other sites
obus
obus
Posted
  • Author
Posted
6 minutes ago, eritius said:

 

Alright, took me a sec to figure out what you were talking about but here you go. Also, I found out how to patch the EFI variables holding the CFG Lock setting that is hidden in my BIOS. I'm working to get that disabled on both CPUs at the moment.

 

AppleIntelInfo.dat

Sorry but it looks like your MSR0xE2 register is locked.

 

2076733011_Intel(R)Xeon(R)Gold6136.thumb.png.e1a3a850e9eb821fe5e2b19f2f840f46.png

 

This is mine:

 

1466858122_Intel(R)Xeon(R)W-2175.thumb.png.f425407ad0d21b21109ef77bf9ef46aa.png

Did you test my last config?

Link to comment
Share on other sites
eritius
eritius
Posted
Posted
2 minutes ago, obus said:

Sorry but it looks like your MSR0xE2 register is locked.

 

2076733011_Intel(R)Xeon(R)Gold6136.thumb.png.e1a3a850e9eb821fe5e2b19f2f840f46.png

 

This is mine:

 

1466858122_Intel(R)Xeon(R)W-2175.thumb.png.f425407ad0d21b21109ef77bf9ef46aa.png

Did you test my last config?

 

Not yet. As I said, I'm working on rewriting a hidden BIOS variable for CFG Lock that should keep it unlocked.

  • Like 1
Link to comment
Share on other sites
obus
obus
Posted
  • Author
Posted
1 minute ago, eritius said:

 

Not yet. As I said, I'm working on rewriting a hidden BIOS variable for CFG Lock that should keep it unlocked.

OK I See. Could you send me a copy of your IOReg. 

Link to comment
Share on other sites
eritius
eritius
Posted
Posted (edited)
6 minutes ago, obus said:

OK I See. Could you send me a copy of your IOReg. 

 

Yup, here you go!

 

(The BIOS thing is a bit of a pain... I needed to modify https://github.com/datasone/grub-mod-setup_var to look at a different EFI variable for my motherboard, but I can't get it to compile under macOS so now I need to make an Ubuntu boot just to compile this EFI so I can go in and flip two bits. Sorry, just ranting hahaha)

 

If you have any advice on getting the GC-Titan Ridge Thunderbolt card working, I'm all ears as well!

 

 

eritius-ioreg.ioreg

Edited by eritius
Specified which TB3 card
Link to comment
Share on other sites
obus
obus
Posted
  • Author
Posted (edited)

Have to take a nap. My SSDT for TB looks to be the same as you have in your Clover ACPI-folder.

My thunderbolt works perfect with HotPlug and My Buffalo disk. I haven't tried with Thunderbolt Display but according to people on the forum it should work.

Prio one for you is to unlock and after that fix a working SSDT-PLUG.aml. so you get native PM, you need to get XPCM to work. Use the 0x040674 CPUID it works on my rig without any difference in performance compared with 0x0506E3 so it should work for you too.

Keep in touch tomorrow:thumbsup_anim:

 

SSDT-C422-TB3HP-RP05-TTR.aml

Edited by obus
Link to comment
Share on other sites
eritius
eritius
Posted
Posted

It's been a long and eventful day.

 

1 hour ago, skyflying5 said:

the supermicro board can modify the bios file to enable the CFG LOCK option visible

 

I all ready got mine worked with C422 chipset。

 

BUT YOU have to take the risk of Flashing the BIOS。

 

it's up to you。

I went ahead and did this what was probably the hard way. I updated grub-mod-setup_var to modify the appropriate EFI variables I found by disassembling the BIOS. (Had to be compiled under an old version of Ubuntu too... ugh). I first switched off CFG Lock for both CPU sockets which turned out to not disable MSR E2. I then found another variable called MSR Lock Control which did seem to work, or perhaps the two together did.

 

In doing this, I found a couple of other things. With so many cores, VerifyMsrE2 actually was timing out which is what gave me the false impression it was fine. I modified it so it could enumerate through to the end and submitted a pull request with the change to the main repo.

 

I also noticed that Open Core was misidentifying the Xeon Gold as a much older Nehalem Xeon. I'm not sure if this affects anything really, but I submitted a pull request too, which pretends the new Xeon Scalable line is the same as a Xeon W, which it's roughly similar to.

 

At this point, with MSR unlocked (and verified), I went ahead and made the SSDT for XCPM (plugin-type = 1). It's ever so slightly more involved because of the dual processors:

SSDT-XCPM.dsl

 

This in turn worked to enable the power management within 10.15:

AppleIntelInfo.dat

sysctl-xcpm.txt

 

So that's good at least! Now, onto the funky:

  • The processor now lists as 2.88 Ghz chip in About this Mac. It used to say 2.99 before I did any of this. It's still marked Unknown, because...
  • The Kernel patches you suggested cause kernel panics on my machine in com.apple.driver.pmtelemetry:
    227460253_pmtelemetrykernelpanic.thumb.jpg.afd4cbcfbcc73128836ffe648e33d6e3.jpg
  • CPUID 0x506e3 still causes panics.
  • I haven't fully experimented, but the most stable configuration for me still is 0x40674, AppleXcpmCfgLock=YES and AppleXcpmExtraMsrs=YES. I think I can maybe turn off AppleXcpmExtraMsrs and use the _xcpm_pkg_scope_msrs_PMheart_10.15 patch, but AppleXcpmExtraMsrs seems to encompass that patch and more. I haven't tested just turning off AppleXcpmCfgLock yet.

For now, I've bailed on getting the Thunderbolt working. I realized I'm not good enough at following the DSDT to patch the X299 version to work for my board. Would still love to figure out how to make that work though and would love some help getting it going, though it's secondary to getting the performance numbers to where they should be.

 

Lastly, I'm not sure if getting XCPM going was supposed to result in performance improvements, but the Geek Bench score is unchanged after all this.

 

  • Like 1
Link to comment
Share on other sites
eritius
eritius
Posted
Posted
1 hour ago, skyflying5 said:

the supermicro board can modify the bios file to enable the CFG LOCK option visible

 

I all ready got mine worked with C422 chipset。

 

BUT YOU have to take the risk of Flashing the BIOS。

 

it's up to you。

706030229_QQ20190806115247.png.1726419b0d4ec4a8d5808dbedc800d04.png

X11DAi-N9.503-CFG-LOCK-DISABLED.rar

 

Thanks for looking into this, by the way! I should add that to complicate matters, I'm stuck on an older beta BIOS that supports Thunderbolt. They removed the support in subsequent updates and scrubbed any mention of it from their site and documentation. The only visible remnant is the Thunderbolt header on the motherboard, still labeled in their diagrams too.

 

Link to comment
Share on other sites
 Share
Go to topic listing
×
×
  • Create New...