Fix sticking at KernelCache validating on Tahoe #126

laobamac · 2025-10-29T19:14:32Z

Fix "WEG Causes Boot Hang in macOS Tahoe Installer (Auqa Session issue), Works After Installation" acidanthera/bugtracker#2509

averycblack · 2025-10-31T15:55:23Z

WhateverGreen/kern_rad.cpp

+	// This step is unnecessary and will cause macOS Tahoe recovery mode to freeze when booting.
+	// Committed by laobamac
+	if (getKernelVersion() >= KernelVersion::Tahoe) {
+		if (kextRadeonServiceManager.loadIndex != index) {


ifNeedOverrideConnector is only ever called when index == kextRadeonSupport.loadIndex. Does this not have the affect of removing all connector patches in all Tahoe versions?

Removed references to AMDRadeonServiceManager and updated connector override logic.

laobamac · 2025-10-31T16:15:59Z

ifNeedOverrideConnector is only ever called when index == kextRadeonSupport.loadIndex. Does this not have the affect of removing all connector patches in all Tahoe versions?

Yes, because RadeonServiceManager will load normally when macOS starts normally.

We can change it to a simpler way, see the latest commit.

Added check for 'com.apple.recoveryosd' service in isNormalSys function.

Added fileExistsAtPath function to check for file existence and updated isNormalSys logic to handle new conditions.

laobamac · 2025-11-01T10:58:58Z

Now it works perfectly on macOS Tahoe, supporting installer/recovery mode/OTA.

laobamac · 2025-11-02T01:02:02Z

It turns out that it works very stably on macOS Tahoe and has been tested a lot. The following is the feedback from a small number of users.
https://www.insanelymac.com/forum/topic/361713-pre-release-macos-tahoe-261/page/10/#findComment-2843066

https://www.insanelymac.com/forum/topic/361713-pre-release-macos-tahoe-261/page/10/#findComment-2843098

https://www.insanelymac.com/forum/topic/361713-pre-release-macos-tahoe-261/page/9/#findComment-2843040

https://www.insanelymac.com/forum/topic/361713-pre-release-macos-tahoe-261/page/9/#findComment-2843051

pyquick · 2025-11-02T04:03:16Z

Yes, it could work well on Tahoe&rec

vit9696 · 2025-11-02T15:01:29Z

WhateverGreen/kern_rad.cpp

+	// Because in the second stage of the OTA (when the macOS Installer finishes booting and the volume
+	// label returns to Macintosh HD), it still uses BaseSystemKernelExtensions.kc, but Lilu mistakenly determines
+	// it as a normal boot. This causes it to still freeze during the KC verification stage.
+	if (fileExistsAtPath("/System/Library/KernelCollections/BaseSystemKernelExtensions.kc")) {


I believe we want this to be fixed on Lilu side.

I believe we want this to be fixed on Lilu side.

As of now, only this patch has encountered issues on macOS Tahoe. Perhaps modifying WhateverGreen is already simple enough, there is no need to modify Lilu temporarily

vit9696 · 2025-11-02T15:07:22Z

WhateverGreen/kern_rad.cpp

 	if (kextRadeonSupport.loadIndex == index) {
-		processConnectorOverrides(patcher, address, size, true);
+		// The abnormal start-up system does not cover the connector on Tahoe.
+		if (ifNeedOverrideConnector(patcher, index, address, size) != false) {


After we fix Lilu to correctly detect normal mode I believe the change here should look as follows:

if (getKernelVersion() < KernelVersion::Tahoe || checkKernelArgument("-radconnector") || lilu.getRunMode() == LiluAPI::RunningNormal) { processConnectorOverrides(patcher, address, size, true); DBGLOG("rad", "processing override connector"); } else { DBGLOG("rad", "skipping override connector"); }

No other changes should be necessary.

After we fix Lilu to correctly detect normal mode I believe the change here should look as follows:

if (getKernelVersion() < KernelVersion::Tahoe || checkKernelArgument("-radconnector") || lilu.getRunMode() == LiluAPI::RunningNormal) { processConnectorOverrides(patcher, address, size, true); DBGLOG("rad", "processing override connector"); } else { DBGLOG("rad", "skipping override connector"); }

No other changes should be necessary.

I understand, do you mean to put the method that recognizes BaseSystemKernelExtendes.kc into Lilu's getRunMode?

Well, we could go this way if no better is found. Please make sure we guard it with macOS 26, so that we definitely do not regress on older operating systems. Ideally we want a lighter approach.

Well, we could go this way if no better is found. Please make sure we guard it with macOS 26, so that we definitely do not regress on older operating systems. Ideally we want a lighter approach.

But when Lilu.kext checks the running status, the root file system may not have been loaded yet, and using vfs_lookup at this time will cause a panic.

In this case I think we should find some other approach. Do you have an ioreg dump? Ideally both IODeviceTree and IOService planes.

In this case I think we should find some other approach. Do you have an ioreg dump? Ideally both IODeviceTree and IOService planes.

Unfortunately, ioreg only recorded BootKernelExtensions.kc. And there is no other obvious difference except BaseSystemKernelExtendes.kc.

Hmmm, could you upload one for me to think about it as well? If it contains private data, you could also e-mail me to vit9696 at pm dot me ^^

Hmmm, could you upload one for me to think about it as well? If it contains private data, you could also e-mail me to vit9696 at pm dot me ^^

OK， Wait for me a moment, I will restart into Recovery and save one

laobamac · 2025-11-02T17:01:11Z

normal.zip

rec.txt.zip

@vit9696

laobamac · 2025-11-02T17:11:26Z

The system in the second stage of OTA is unable to capture IOReg. But in fact, we just want to know how to implement the OTA phase😂

vit9696 · 2025-11-02T17:38:01Z

Hmm, that is problematic, SSH also does not work?

laobamac · 2025-11-02T17:39:12Z

Hmm, that is problematic, SSH also does not work?

Yes, the entire system crashed directly.
It's not ruled out that it's a problem with my machine, maybe you can also try OTA.

vit9696 · 2025-11-02T17:40:42Z

Also, your recovery dump does not include IODeviceTree plane, thus e.g. chosen is missing. The command should be ioreg -l -p IODeviceTree.

laobamac · 2025-11-02T17:48:27Z

rec2.txt.zip

here

vit9696 · 2025-11-02T18:00:00Z

From this I see several candidates for OTA test:

We could try to check whether IODeviceTree:/chosen/booter-name contains bootbase.efi or boot.efi.
We could try to check for IODeviceTree:/chosen/root-snapshot-name presence.
We could try to check whether IODeviceTree:/chosen/boot-file contains boot\System\Library\KernelCollections\BootKernelExtensions.kc or something else.

Ideally we need an IOReg dump from OTA v2 stage though, hard to be sure otherwise.

laobamac · 2025-11-02T18:05:46Z

From this I see several candidates for OTA test:

We could try to check whether IODeviceTree:/chosen/booter-name contains bootbase.efi or boot.efi.

We could try to check for IODeviceTree:/chosen/root-snapshot-name presence.

We could try to check whether IODeviceTree:/chosen/boot-file contains boot\System\Library\KernelCollections\BootKernelExtensions.kc
or something else.

Ideally we need an IOReg dump from OTA v2 stage though, hard to be sure otherwise.

These are all unavailable, I have already tried them here. Because the second phase of OTA has already used boot.efi and BootSystemExtendes.kc, BaseSystemKernelExtendes.kc will only be loaded just before initializing the user interface.

This is my attempt from last month, and my thoughts are the same as yours. But it is not available.kern_extra.zip

laobamac · 2025-11-02T18:12:34Z

My friend, I will come back tomorrow to see this issue. It's already 2:10 in the morning in China, I'll go rest first.

bye

Fix sticking at KernelCache validating on Tahoe

0a855d3

laobamac mentioned this pull request Oct 29, 2025

WEG Causes Boot Hang in macOS Tahoe Installer (Auqa Session issue), Works After Installation acidanthera/bugtracker#2509

Open

Add ifNeedOverrideConnector() method

4094786

averycblack reviewed Oct 31, 2025

View reviewed changes

laobamac added 2 commits November 1, 2025 00:11

Add isNormalSys function

cdad012

Remove RadeonServiceManager and refine connector logic

660e9f6

Removed references to AMDRadeonServiceManager and updated connector override logic.

laobamac added 2 commits November 1, 2025 01:33

Enhance isNormalSys to check for recovery service

dcebbbe

Added check for 'com.apple.recoveryosd' service in isNormalSys function.

Fixed the problem of freezing in the second phase of OTA

1d74eed

Added fileExistsAtPath function to check for file existence and updated isNormalSys logic to handle new conditions.

vit9696 reviewed Nov 2, 2025

View reviewed changes

Fix sticking at KernelCache validating on Tahoe #126

Are you sure you want to change the base?

Fix sticking at KernelCache validating on Tahoe #126

Conversation

laobamac commented Oct 29, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

laobamac commented Oct 31, 2025

Uh oh!

laobamac commented Nov 1, 2025

Uh oh!

laobamac commented Nov 2, 2025

Uh oh!

pyquick commented Nov 2, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

vit9696 Nov 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

laobamac commented Nov 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

laobamac commented Nov 2, 2025

Uh oh!

vit9696 commented Nov 2, 2025

Uh oh!

laobamac commented Nov 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

vit9696 commented Nov 2, 2025

Uh oh!

laobamac commented Nov 2, 2025

Uh oh!

vit9696 commented Nov 2, 2025

Uh oh!

laobamac commented Nov 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

laobamac commented Nov 2, 2025

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

4 participants

vit9696 Nov 2, 2025 •

edited

Loading

laobamac commented Nov 2, 2025 •

edited

Loading

laobamac commented Nov 2, 2025 •

edited

Loading

laobamac commented Nov 2, 2025 •

edited

Loading