Add patchAPIC transition guide

[khronokernel]

As mentioned here: acidanthera/bugtracker#1070

OpenCore does not have an equivalent of patchAPIC, might be a good idea to explain the process of patching the table manually and reinjecting it. Seems to be mainly relevant for X58 and X79(and their server chipset counterpart) but may be helpful for other high core count systems.

@BrushXue Would it be possible to provide and ACPI dump? SysReport set to true will dump it(note you need DEBUG builds of OpenCorePkg)

[BrushXue]

Here you are.

opencore-2020-08-02-192854.txt

ACPI.zip

[khronokernel]

@BrushXue Got around to working on the APIC guide, can you test this table?:

• APIC.zip

Should be modified correctly to match your ACPI tables, main issue is you have a dual socket system which complicates things on where the values are pulled from. And don't forget to add this table both to your EFI and your config.plist

You will also need this under ACPI -> Delete:

Key	Type	Value
All	Boolean	YES
Comment	String	Delete APIC Table
Enabled	Boolean	YES
TableLength	Number	0
TableSignature	Data	41504943

I have an initial draft here, though it's a bit unpolished so hasn't been merged yet:

• Add patchAPIC conversion guide Getting-Started-With-ACPI#26

[BrushXue]

Thanks for your response. I have added the APIC you provided. Now the system can pass the previous point, but then freezes at
Darwin Image4 Validator Version...... without a kernel panic.
I've disabled my CPU PM and GPU spoof but same result.
Now my bootargs are npci=0x2000 -no_compat_check -xcpm -v keepsyms=1 -cpus=1
Here's the updated log:
opencore-2020-08-04-080553.txt

[khronokernel]

Have a feeling I know where the stall is, pretty sure your first 2 threads of each CPU are getting an incorrect or missing Local APIC NMI table. Can you also provide an image? Just want to verify my suspicions.

I also looked through your log, and saw some interesting things:

• -xcpm was removed in Sierra
• -cpus=1 is incorrect, it should be cpus=1. General rule with arg is it will either have a - or = but never both
• -no_compat_check What SMBIOS are you using? MacPro6,1 should work from Mavericks through Big Sur, not sure why you'd be using this arg
• NullCPUPowerManagement.kext why? Ivy Bridge-E has native CPU power management, and then this clashes with your initial -xcpm boot-arg(which doesn't work anyways) but I don't really understand the logic of the pairing

[BrushXue]

Actually I'm not surprising because I couldn't find new tutorials for x79. These are from very outdated x79 tutorials and I finally found this combination that works with Clover.
As for the Nullcpupm, I'm adding it just for debugging purpose.
Another question is the EC. I don't have a dummy EC in my clover setup but it still works. Should I add one?
So I've changed my bootargs to npci=0x2000 -v keepsyms=1 but got the same result.

[BrushXue]

I just tested my Clover setup. Removing xcpm and no_compat_check is OK. But I still need to inject CPU freq. from ssdtPrGen. So it'll be the same in OpenCore. Am I right?

[khronokernel]

Shouldn't need ssdt-PRgen, your hardware supports XCPM. SSDT-PLUG's all you should need
Oh if you have a clover setup booting, you could simply dump the patched APIC file with F5 in the picker. If you could send me this patched file I'd appreciate it

[BrushXue]

I'm sending the whole ACPI folder you can find my own SSDTs as well.
ACPI.zip

[amadeusex]

I'm sending the whole ACPI folder you can find my own SSDTs as well.
ACPI.zip

Hi there.

I'm trying to get Catalina installed on my dual E5-2687w v2 HP z820.
Opencore stops with a kernel panic and the x86_validate topology() 32 threads but 33 registered .... message.

Did you manage to overcome the issues ?

[128keaton]

I am currently running into the same issue as well @amadeusex looks like we have to manually patch things, wish the guide was updated tho

[Allubz]

I can chime in, got a X299 Extreme4 here with i9-10900X that does the same (20threads but 21 registered). Clover with APIC Patch is doable, but I'm having trouble installing Big Sur (stuck on phase 2, "DiscX is write protected", regardless of fresh install or 10.15 -> 10.16 update); probably unrelated.

It's related to what is discussed here, including a guide on how to patch APIC. That, with the above table provided by khronokernel should fix it. Honestly, I didn't finish trying yet, so I'll report back.

I remember fixing a 9820X (10-core, but cut/lasered down by Intel, hence the weird APIC, was the thought), 9900X, the thread speaks of 9920X as well, and now this 10900X does the same.

UPDATE:
What worked for me is
1/ Make an ACPI dump with Clover (F4 on boot menu)
2/ Open APIC.aml and basically only edit the first core its processor id (my default was 10, rather than 00), as such:

[02Ch 0044   1]                Subtable Type : 00 [Processor Local APIC]
[02Dh 0045   1]                       Length : 08
[02Eh 0046   1]                 Processor ID : 00
[02Fh 0047   1]                Local Apic ID : 10
[030h 0048   4]        Flags (decoded below) : 00000001
                           Processor Enabled : 1

I suppose you need to make sure no other core is seen as "00", but I didn't need to change any other entries (as suggested in the Tony guide)
3/ Add to EFI/OC/ACPI and config.plist
4/ dropping APIC in ACPI/Delete in OC. Using OCC (I know, I know) setting the TableSignature to APIC (equivalent to 41504943)

[128keaton]

@Allubz awesome thank you for this information. Where is this guide you mention? I tried looking when I originally commented on this issue but came up empty handed. Unfortunately, for some reason my APIC table has three processor IDs with 00. Not sure why. Following the guide on the branch in the PR for this issue (which is a semi-complete patching guide), I could not follow it to completion since I have a few extra entries that need IDs that I do not have in the extracted DSDT.

[Allubz]

@128keaton Welcome! Did you read the "discussed here" link in my post? It's to Tony's. Specifically page 5 post # 42 there.

I think just changed the first ID to 00, and changing any other 00's to available numbers should fix it.

[gbabook]

Open APIC.aml and basically only edit the first core its processor id

I did one more thing, removed the "Unknown ... TYPE" code like:
START WITH:
[228h 0552 1] Subtable Type : 7F [Unknown Subtable Type]
[229h 0553 1] Length : 0C

**** Unknown MADT subtable type 0x7F

[36Ch 0876 1] Subtable Type : 7F [Unknown Subtable Type]
[36Dh 0877 1] Length : 0C

more and more....

**** Unknown MADT subtable type 0x7F
END

just only remove these codes, keep the others which are looked like normal.

compile until no error, in my case, OC / Clover boot successful.