chefkissinc/legacyred

Legacy AMD iGPU support kext

legacyred's Issues

Property 'model' gets overridden by the Controller kexts

Not a code issue, ATY,DeviceName gets populated

Some APUs (Mullins and potentially others) have bad connectors

FB-only is currently broken on some APUs, notably my Mullins E1-7010 APU, currently the state of other APUs are unknown, however, current debugging shows that on my device at the least, AtiAtomObjectinfoTable_V1::getAtomConnectorInfo is failing at connector '3' which does not exist, this, presumably, brings down the whole stack and prevents display from going to the LVDS connector.

Update 1/8/23:
Attempts at overriding the value from within AtiAtomObjectInfoTable_V1::getNumberOfConnectors has proven futile at the current point in time

Update 7/8/23:
Attempts to grab the Display Object Path Table have been proven to be a success, now I need to figure out how to grab the object type and compare it against GRAPH_OBJECT_TYPE_CONNECTOR

Update 10/8/23:

LegacyRed now trims off the extra objects in the Connector Info Table that don't use GRAPH_OBJECT_TYPE_CONNECTOR

Recently Came Into Possession of an A6-9225 (Stoney) Based Laptop

This is isn't an issue, but rather an inquiry: as I have an AMD A6-9225 Stoney laptop I'd be happy to test if you're interested.

HDMI is broken on some APUs when it is not the primary display

Plugging in HDMI on the Mullins/Godavari machine results in a blank display until force restart

Also observed on a Kabini machine

Current state of LegacyRed

Idk why, but I want to share the current state of this kext on my PC (and probably on all others PC running this kext)

This is probably the most useless "issue" of the entire GitHub, but nevermind.

I can't wait to see what this kext can do, although I know that currently you are mainly occupied with NootedRed, and the development of this kext can be complicated.

So, to start, below you can find the specs of my PC :

CPU : AMD A8-7600, Kaveri
MB : GA-F2A88XM-HD3 w/ BIOS FC
RAM : 2x4gb DDR3
GPU : iGPU : R7 Graphics, GCN 2-based, codename Kaveri
macOS Version : Latest Big Sur 

And now, the state of this kext is... just a Kernel Panic, not surprising
The KPanic talks about Lilu, LRed, and macOS dependencies, so it's no surprise that it's the same issue that's in the README, which makes this issue even more useless.

UPDATE 28/06 : The FB-Only works on my APU (Kaveri) like it's wrotten in the README, but the "normal mode" keeps on a black screen and macOS don't boot
I removed my dGPU since it's TeraScale and the patches for Big Sur conflitcs with LRed like it's wrotten below

Some notes :

• VGA isn't supported since it's not converted by the MotherBoard
• OCLP patches for Terascales create conflicts with LegacyRed, and I wasn't able to remove OCLP using the software, I needed to update (/reinstall) macOS.

Idk if this issue, or me, can eventually help, but if you need someone to test something about this kext, don't hesitate.

And sorry for my poor English, I've written this with GTranslate and my English knowledge.
Have a good day (and thank you for your work on NootedRed, you're incredible (even if I don't have an AMD APU anymore) !

AGDC issues on some APUs

Issue observed on a Mullins/Godavari APU, setting CFG_USE_AGDC to false makes FB-only work, no logs yet.

Primary suspects are the following:

• Link Training
• AtiDeviceControl in general

Current WA:
Setting CFG_USE_AGDC to false in Info.plist -> IOKit Personalities -> Controller (Stoney) -> aty_config or Info.plist -> IOKit Personalities -> com.ChefKissInc.LegacyRed -> Framebuffers (Legacy) -> Controller (<Carrizo or GCN 2 depending on the iGPU generation>) -> aty_config

No Acceleration on all iGPUs

Just stating the obvious, and also here for tracking purposes.

HotPlug Detection may be broken on FB-only

AMDSupport prints out: ATOM: uint32_t AtiGpioPinLutInterface_V1::getHpdLine(uint32_t): ASSERT(kIOReturnSuccess == err) on some GCN 2 systems, this, however, isn't a major problem for the systems tested that are confirmed to work with FB-only, as the drivers continue functioning like normal.

Update 1/8/23:
Debug wrap has been added and can be activated by using -lredgpiodbg, further testing is required but the issue stems from getGpioPinInfo returning a non-zero value

Using AMD9500Controller breaks DCE Clock setting in AMDSupport

Outputs in question:
[ 62.265214]: ATOM: virtual IOReturn AtiBiosParser1::setDisplayClock(const DCEClockParams *): ASSERT(NULL != m_setDceClock) [ 62.265224]: ATOM: virtual IOReturn AtiBiosParser1::setDisplayClock(const DCEClockParams *): ASSERT(NULL != m_setDceClock)

Update 31/7/23:
Switching to AMD9000Controller negates the above issue, but we want to use AMD9500Controller for Ventura+ support on Stoney ASICs, further investigation into the underlying cause is required, attempting to bypass the check triggered also gives us a kernel panic with a chain of symbols that are most likely related