@kas1e

I've filed a bug report two years ago. Given the RadeonRX.chip.debug output, it was determined that it was the sam440 u-boot x86 emulator shortcoming and that Hans couldn't do anything about it on driver side.

@sailor

My Polaris was recognized in a sam44-flex. But the x86 emulator wasn't able to correctly init the card. This resulted in a RX driver error which in turn made the sam440 hang during boot.

@sailor

My Polaris was recognized in a sam44-flex. But the x86 emulator wasn't able to correctly init the card. This resulted in a RX driver error which in turn made the sam440 hang during boot.

@joerg

Sam440 has no PCIe either. Yet, a Radeon R9 270x works just fine with a PCI to PCIe bridge. No conversion of bars to 32bit with scripts needed.

It looks like the firmware plays the decisive role and not the kernel. From a kernel pov, PCIe and PCI are similar. The additional features for PCIe like msi and msi-x are not supported anyways by amigaos4

@balaton

kas1e is testing on a real Pegasos 2. I am testing on QEMU with pass-through

Tried to pass-through the host bridge but that doesn't work. I cannot attach a VFIO kernel driver to the host bridge. So it looks like that you can only pass through endpoints. I do not have regular PCI slots in my system. So an External PCI to PCIe bridge like kas1e is using is also not possible for me.
And cascading bridges (pcie to pci + pci to pcie) is a bit too much.

If 64bit bars were the issue then he should have seen at least the two 32bit bars. Like I do on QEMU with amigaboot.of and no forth script.
Since he sees only the bridge within amigaos4, it looks like the first issue is a scanning issue. (or information is not passed on to amigaos4)

@kas1e

On QEMU, every VFIO device is passed through without any bridge that might be in the host system. So the VFIO GPU is directly attached to a host bridge within of the emulated Pegasos2 as you can see in the screenshot above.

On a real Pegasos2, a bridge in between is seen by the system. So if the kernel has an issue with scanning behind bridges then you are out of luck.

But let's not draw any conclusions until you have tried a power inserter and the pericom bridge.

@kas1e

I was only talking about host bridges.

Everything starting with 00 is part of the PCI bus.

Everything starting with 02 is part of the AGP bus.

Somehow the AGP bus on QEMU starts with 01. But I guess this doesn't matter.

Here's a screenshot of Ranger on QEMU with the RX560 connected to the "AGP" bus.

@kas1e

amigaboot.of:
- RX 560 is detected by Ranger
- Only BAR4 and BAR5 are visible in Ranger.
- RadeonRX driver was removed from kicklayout because driver will fail due to missing BARs


bboot 0.4:
- RX 560 is detected by Ranger
- BAR0, BAR2, BAR4 and BAR5 are visible in Ranger.

@kas1e

I think that Ranger just lists the PCI devices which are reported by the system.
The pegasos has two bridges. That results in two busses. Ranger reports my RX 560 just fine behind the AGP bridge. With the assumption that you already use the PCIe power inserter, it looks like the kernel is not able to scan behind your 8112. Maybe you have better luck with the pericom bridge. This was also the key to get those HD cards working on a sam440.

@kas1e

In case you didn't know: 1002,AAB0 is the hdaudio soundcard endpoint for audio over HDMI or DP.

@balaton

One thing to note is that I do not experience any freeze. And I don't have that patch applied. Maybe because I'm running on an Intel platform.

It's just slow.

@MartinW

Quote:

I have my RX on the AGP bus. As far as I can tell it has its own interrupt.

@MartinW

The performance is slow on your RX cards as well?

I didn't apply the interrupt patch. So just a clean build of the QEMU master sources

@batalon

There was no bboot output. Just those two lines.

But I've upgraded ubuntu to 23.04. Installed the dependancies and rebuild qemu 8.1.2 rc0 again.

And guess what? Error gone at first and QEMU boots with bboot 0.4.

RX 560 shows an image on my monitor. Everything runs stable. With and without interrupt. With and without compositing.
Only spoiler is that the GUI is dead slow. But it works and that's a start

I've attached the RX 560 to AGP bus. So bus=pci.0
Interrupt line is 0x0B Pin A. Interrupt number is 27.

Unfortunately the "Assertion `next <= PCIE_CONFIG_SPACE_SIZE - 8' failed" error returns randomly after a restart of QEMU. So it looks related to missing reset of the RX 560 GPU.

Edit1: GPU doesn't work on pci.1. So that must be the shared interrupt bug in the RX driver.

@balaton

Rebuild QEMU 8.1.0 rc0 without patches.
Used bboot 0.4
This is what I get with a RX560

@balaton

Qemu 8.1.0 rc with bboot 0.2 doesn't work for me with vfio.

I've applied the first two "patches" in post #787 and that seems to solve the crash of the RadeonHD driver. RadeonHD driver loads but no display. Do I need to blindly apply all diffs in that thread of post #787? Because that thread is a bit of a mess.

Results of a quick test with 8.1.0 rc.

Radeon 4850 results in a crash of the RadeonHD driver right after the workbench screen shows up on the display.
Radeon RX 560 results in a crash of QEMU.
Will post more details tomorrow.

ok, I'll try the release candidate.

One thing that I do not understand in the Pegasos2 interrupts discussion is the whole edge sensitivity. PCI interrupts are level sensitive. Therefore the emulated legacy interrupts consist of a begin and end interrupt message.

@MartinW

Do I also need to apply this interrupt patch? Or is building from the latest sources enough.

I haven't followed this thread for a while. Can someone give a summary of where we are right now? bboot will replace pegasos2.com + BAR script? Or is a script still necessary?