Sleep Kernel Panic

[jzrodriguez98]

@0xFireWolf First of all, congratulations for being successful here and thank you for making this kext solution available to the community. I'm not a programmer and I've been hackintoshing since 2012. I got a BCM943602CS card to test this kext and I was curious about the sleep kernel panic issues you mention that can happen while using this kext. I'm not a fan of AI and I use it knowing its limitations but I tested the waters asking ChatGPT about sleep panic issues found in real Macs using this card when loading Linux and then asked to come up with a solution adapted to macOS. Below is a link to what I got:

https://chatgpt.com/share/68f646c6-1bdc-800d-8bc4-f3e7235b4830

Based on your experience developing and testing the kext, do you think this AI generated proposed solution can really be feasible and adaptable to macOS?

[markovicdenis]

well, seems like he doesn't expose the device properties to disable sleep states on the chip. and the last solution offered by the AI would probably work, but that one you'd add yourself to your own machine. Sleepwatcher that is.

If i remember correctly, I might have used sleepwatcher myself on some older macos setups whre usb wifi broke my sleep.

I think it even worked back then, don't know about now 😀

[andreszerocross]

Same here, Kernel panic when put computer to sleep. @0xFireWolf

[0xFireWolf]

Thanks for your interest in this project and for joining the beta testing.

The kernel panic (that happens when you put your computer to sleep) has nothing to do with the device properties, and I knew there would be a kernel panic before I even tried it while writing the Known Issues section.

I haven't yet looked into the power management routines, because I usually prioritize getting all necessary functionality working before tackling sleep/wake issues.

To give you a bit of technical detail, the Wi-Fi driver relies on another driver AppleOLYHALMac.kext, which requires a couple of ACPI methods to manipulate the GPIO signals. Instead of asking you all to add another SSDT with all the required dummy ACPI methods implemented, I decided to zero out some initialization routines in AppleOLYHALMac, and as a result, left some objects uninitialized (null pointers). When you put your computer to sleep, the Wi-Fi driver invokes the interface exposed by AppleOLYHALMac to change some power settings (because there is real platform device called AMFM on a real Mac). However, since the objects are null, a page fault would occur, as indicated by the kernel panic report uploaded by @andreszerocross (by the way, thanks for uploading the report -- this is exactly what I need for solving a problem).

I understand that you all want sleep/wake to work properly, but there is a lot of work that needs to be done behind the scenes. Yes, I can easily patch the OLY driver to avoid the page fault. However, the Wi-Fi chip will probably be powered off while your computer is sleeping, so when your computer wakes up, the driver needs to power it back on and download the firmware to the chip again. I need to analyze and understand the control flow and make sure that each step is compatible with legacy Wi-Fi chips, which requires significant reverse engineering effort and takes time. As such, I would appreciate your patience as I work through this.

[0xFireWolf]

@armincheng Please add the boot argument keepsyms=1 that keeps the symbols during kernel panics and update the kernel panic report. Thank you.

[usr-sse2]

I understand that you all want sleep/wake to work properly, but there is a lot of work that needs to be done behind the scenes. Yes, I can easily patch the OLY driver to avoid the page fault. However, the Wi-Fi chip will probably be powered off while your computer is sleeping, so when your computer wakes up, the driver needs to power it back on and download the firmware to the chip again.

Waking without network is much better than kernel panic, it at least gives a chance to save work and reboot correctly without possible file system corruption if the machine was accidentally put to sleep, so patching it would be great.