Messages

1

General Discussion / Re: Thanks WebKit... [amdgpu crash]

« on: April 17, 2023, 10:45:47 pm »

It's a Radeon 580, and has been working fine otherwise. I haven't played any games, but it seems OpenGL works at least. I've played around with the fancy desktop effects, and I'm using dual monitors. It was a second-hand purchase, so maybe it does have some issues.

Speaking of Chromium, it worked rather well when I was using Void Linux. I could share my desktop on Google Meet meetings. Even when tinkering with some of Chromium's advanced options on Debian 11, the desktop sharing fails with a message that it's not supported. In both cases with Void and Debian I was using a Linux 6.1 kernel.

And for whatever it's worth, I was able to follow the instructions on the Wiki to get the GPU graphics to work at boot by flashing the squashfs firmware image.

2

General Discussion / Thanks WebKit... [amdgpu crash]

« on: April 17, 2023, 09:25:11 am »

I'm surprised this happened:

Code: [Select]

[86046.568669] amdgpu 0000:01:00.0: amdgpu: IH ring buffer overflow (0x0008A700, 0x0000F9B0, 0x0000A710)
[86046.568709] amdgpu 0000:01:00.0: amdgpu: GPU fault detected: 147 0x02e0c802 for process WebKitWebProces pid 132102 thread WebKitWebP:cs0 pid 132125
[86046.568740] amdgpu 0000:01:00.0: amdgpu:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x00101400
[86046.568753] amdgpu 0000:01:00.0: amdgpu:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x06048002
[86046.568776] amdgpu 0000:01:00.0: amdgpu: VM fault (0x02, vmid 3, pasid 32773) at page 1053696, read from 'TC4' (0x54433400) (72)

I'm using Debian 11.6 and figured I'd try Chromium. After a few pages loaded and having it running for ~12 hours, the applications seem to have frozen. The music in the background continued to play and the cursor would move, but the applications are all seemingly non-responsive. SSHing in and running `dmesg` showed that error. Kernel is 6.1.0-7. Anyone else see this before? I noticed that Chromium on Debian also reports it's x86_64, that should have been reason enough to go back to Firefox!

3

GPU Compute / Accelerators / GPU & disk encryption?

« on: January 07, 2023, 12:09:01 pm »

I flashed the AMD firmware to the BOOTKERNFW, and can see the Petitboot screen via the AMD GPU! However, when the OS boots, it goes dark for a few seconds, then a cursor appears. Only via the serial display via the BMC web interface am I able to see the prompt to unlock the disk to resume booting. Trying to enter the password on the keyboard doesn't work. Is there a way around this, so I don't need another computer to boot the encrypted disk? I used Debian's auto formatter, so /boot is unencrypted, and LVM has the encrypted swap and root partition. I wouldn't be opposed to having a way to inform Petitboot or BMC of the password somehow.

4

General Discussion / Re: Raptor-provided SATA card issues

« on: January 07, 2023, 12:03:05 pm »

It was recognised by Petitboot, and able to boot an OS installed on a SATA HDD. I'll keep investigating.

5

General Discussion / Raptor-provided SATA card issues

« on: December 27, 2022, 09:50:33 am »

I'm using the SATA card provided by Raptor when I bought the Talos II. lsusb identifies it as: SATA controller: Marvell Technology Group Ltd. 88SE9215 PCIe 2.0 x1 4-port SATA 6 Gb/s Controller (rev ff).

Void Linux can see the optical drive as /dev/cdrom which is a link to /dev/sr0. However, I cannot mount a disc, nor can fdisk open it, both with the same error: "no medium found". I was trying some other distros, and after a few warm reboots, the OS no longer sees the optical drive. I attached a hard drive as well, and the distros I tried couldn't see the hard drive at all (though the hard drive works via USB-SATA cable).

dmesg | grep ata has this output:

Code: [Select]

[    2.522439] libata version 3.00 loaded.
[    2.547068] ata1: SATA max UDMA/133 abar m2048@0x620c080040000 port 0x620c080040100 irq 128
[    2.547072] ata2: SATA max UDMA/133 abar m2048@0x620c080040000 port 0x620c080040180 irq 128
[    2.547075] ata3: SATA max UDMA/133 abar m2048@0x620c080040000 port 0x620c080040200 irq 128
[    2.547078] ata4: SATA max UDMA/133 abar m2048@0x620c080040000 port 0x620c080040280 irq 128
[    2.862532] ata1: SATA link down (SStatus 0 SControl 300)
[    2.862564] ata2: SATA link down (SStatus 0 SControl 300)
[    2.863697] ata4: SATA link down (SStatus 0 SControl 300)
[   12.544009] ata3: softreset failed (1st FIS failed)
[   22.544008] ata3: softreset failed (1st FIS failed)
[   57.544575] ata3: softreset failed (1st FIS failed)
[   62.544519] ata3: softreset failed (1st FIS failed)
[   62.544542] ata3: reset failed, giving up
[   62.598911] EXT4-fs (nvme0n1p1): mounted filesystem with ordered data mode. Quota mode: none.
[   62.671972] EXT4-fs (nvme0n1p1): mounted filesystem with ordered data mode. Quota mode: none.
[   65.356285] cfg80211: Loading compiled-in X.509 certificates for regulatory database
[60247.550077] PHB4 PHB#49 Diag-data (Version: 1)
[60247.572032] ata1: failed to stop engine (-19)
[60247.572050] ata2: failed to stop engine (-19)
[60247.572087] ata3: failed to stop engine (-19)
[60247.572103] ata4: failed to stop engine (-19)

So I'm wondering, is there a missing firmware file for this? Should I buy another SATA adapter?

Kernel: 6.0.13_1. Didn't work on older kernels either.

6

General Discussion / Re: AMD RX 6900 XT from Sapphire NITRO

« on: November 11, 2022, 09:03:51 am »

Good news folks, 6900 XT runs fine under 6.1.0rc3!

If you know anyone in Victoria, Australia that might be able to lend me 6800 or 6700 card for few hours, please reach out to me via private message

Working, and with 64k pages too if I remember correctly. This is good news. Does that mean other AMD GPUs also work under 64k pages?

7

User Zone / Re: Void Linux thread

« on: November 02, 2022, 09:37:43 am »

Wow, that did it! Thank you!

8

User Zone / Re: Void Linux thread

« on: October 31, 2022, 12:00:15 pm »

Regarding Void's (and Chimera's) use of 4k pages, how is this used with Qemu & virt-manager? Virt-manager is nice since it creates the XML for libvirt, but I run into the issue where Qemu complains about wanting 64k pages, but not being able to do so on 4k pages.

The Void PPC doc has an entry regarding this: https://docs.voidlinux-ppc.org/configuration/virtualization.html
Which specifies to use -machine pseries,cap-hpt-max-page-size=4096.

How is this specified in the virt-manager or libvirt XML? Looking at https://libvirt.org/drvqemu.html, the closest I could find is using <qemu:commandline>, but that doesn't help, nor does adding the page size flag to machine="pseries".

9

Operating Systems and Porting / Re: I am looking for a dnetc to count OGR-28 on PPC64LE

« on: October 21, 2022, 08:15:23 am »

Maybe try my Power9 fork? https://github.com/rjzak/dnetc-client-base/tree/power9. I was able to get some things working, still needs some improvements.
If I can get it to a working state, I can see if the maintainer will take it and make a ppc64le release, as per their instruction https://www.distributed.net/Download_clients (very bottom of the page).

10

Applications and Porting / Re: Interest in Porting Wasmtime & Cranelift

« on: October 08, 2022, 01:18:29 pm »

There's no work on Cranelift yet because I haven't gotten wasm fully working. It passes the tests but some apps inexplicably fail, and I suspect it's not handling int32s correctly which will require a non-trivial refactor.

What aspect of Wasm are you working on? Something in the browser? Another WebAssembly library?

One thing I was thinking of doing with Wasmtime & Cranelift was to start with Wasmtime FIber. It has assembly code for context switching, and of course lacks ppc64le support. I also figured it would be easier for me to get started on, since it's smaller and presumably has unit tests so I can know if I did it right. This is the code for x86_64: https://github.com/bytecodealliance/wasmtime/blob/main/crates/fiber/src/unix/x86_64.rs

11

Applications and Porting / Re: Interest in Porting Wasmtime & Cranelift

« on: October 07, 2022, 01:55:10 pm »

I don't see cranelift in there, or anywhere under the chmeeedalf user name.

Cranelift: https://github.com/bytecodealliance/wasmtime/tree/main/cranelift

Specifically, where the ppc64le code would be, if it existed: https://github.com/bytecodealliance/wasmtime/tree/main/cranelift/codegen/meta/src/isa

12

Applications and Porting / Interest in Porting Wasmtime & Cranelift

« on: October 05, 2022, 11:13:25 am »

I'm interested in getting Wasmtime (WebAssembly Wasi runtime library and program) and Cranelift (JIT used by Wasmtime, others) working on ppc64le. One thing that's interesting is that Power seems to be the only modern platform missing (RISC-V support was merged in recently, and they also support s390x, ARM, and of course, x86_64). As visible in the RISC-V PR, the amount of code required for such an undertaking is massive, and requires extensive knowledge of assembly. I'm interested in doing this work, and have been reading IBM's ISA and ELF ABI documentation, but could use a hand.

It would benefit Firefox, since Cranelift is used by Firefox for WebAssembly JIT compilation, I believe. It would help with some random apps, such as the text editor Lapce, which use Cranelift & Wasmtime for plugins.

Also relevant: https://forums.raptorcs.com/index.php/topic,27.msg3125.html#msg3125

13

Talos II / Re: Damaged motherboard?

« on: October 05, 2022, 10:21:42 am »

I think it's still a pretty big risk to try and use it.
With that big of a solder splash, there could be other smaller and harder to see ones anywhere else on the motherboard.
Or even loose solder balls rolling around waiting to short out some pins.

Turns out, it was fine using the damaged motherboard. I did receive the new one ~2 weeks ago. So all is good!

14

Talos II / Re: Damaged motherboard?

« on: September 12, 2022, 06:15:34 pm »

It's the third week, and I'm still waiting for the replacement motherboard. It's frustrating that "early next week" has become week three and no response from Support or Sales, and I've emailed them a few times last week and the week before. I also tried reaching out on Twitter last week, and the message is unread yet they're tweeting. It's not fun being ignored waiting for weeks for something which was supposed to be quick, and I'm trying to be patient since I'm sure they've got a million things going on.

In the meantime, what does the community think about using it with just CPU1, since the damage is to a memory slot for CPU 2? It seems that the RAM slot connects only to the CPU, and it not being there wouldn't damage anything if the solder blob is connecting wires that shouldn't be connected. Thoughts?

15

Talos II / Re: Damaged motherboard?

« on: August 27, 2022, 08:46:29 am »

They are sending a new board early next week, and covering shipping (and return shipping of the solder blob board). They said this hasn't happened before. I do appreciate their quick response (and on a Saturday too!). Also, the paint on the chips indicates the version of the firmware at shipping, which is neat.