You are not logged in.
Good news: I was able to boot without "acpi=off" and load the amdgpu in both the RX580 and Vega card. I used the following parameters:
BOOT_IMAGE=/vmlinuz-linux root=UUID=90b89d1d-e013-4458-96bf-c105d0fec91b rw zswap.enabled=0 rootfstype=ext4 loglevel=3 quiet amdgpu.aspm=0 amdgpu.rumpm=0 amdgpu.gpu_recovery=1 sysrq_always_enabled=1But the GPU isn't displaying anything atm. Might be the "amdgpu.gpu_recovery=1" parameter. I'll keep trying stuff and report back.
dmesg: https://termbin.com/44sk
journal: https://termbin.com/4p2c
It's "amdgpu.aspm=0" that it's actually letting me boot, no image from both the APU or GPU, but it lets me boot.
$ sudo modprobe amdgpu -vv
modprobe: INFO: custom logging function 0x5c4f194eb200 registered
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/acpi/video.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/display/drm_display_helper.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/drm_ttm_helper.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/drm_suballoc_helper.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/i2c/algos/i2c-algo-bit.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/drm_buddy.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/scheduler/gpu-sched.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/drm_exec.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/amd/amdxcp/amdxcp.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/amd/amdgpu/amdgpu.ko.zst aspm=0
modprobe: INFO: context 0x5c4f1a77a3c0 releasedThis is only with
BOOT_IMAGE=/vmlinuz-linux root=UUID=90b89d1d-e013-4458-96bf-c105d0fec91b rw zswap.enabled=0 rootfstype=ext4 loglevel=3 quiet amdgpu.aspm=0 sysrq_always_enabled=1as cmdline.
OK. It's gotta be this "ASPM" thing. Even with "pcie_aspm=off", the system boots (but no image). I've used this parameter in a NVIDIA-based OPTIMUS laptop. Here, the results aren't as good, sadly. I'll keep trying different parameters and see if I can come up with something.
Offline
The journal doesn't have 1002:15dd and it also doesn't show up in lspci, instead it turns into
+-08.1-[0a]--+-00.0 Advanced Micro Devices, Inc. [AMD] Zeppelin/Raven/Raven2 PCIe Dummy Function [1022:145a]what likely explains why you can't plug an output there (and kills my idea to just plug an output into both)
Can you boot w/ "pci=noacpi" or "pci=nocrs"?
I will attempt this asap.
EDIT: tried both, no luck with either.
Last edited by b0ss_ (2024-03-10 22:45:37)
Offline
mar 10 19:16:59 home kernel: amdgpu 0000:01:00.0: amdgpu: Using BACO for runtime pm
mar 10 19:16:59 home kernel: [drm] Initialized amdgpu 3.57.0 20150101 for 0000:01:00.0 on minor 1
mar 10 19:16:59 home kernel: amdgpu 0000:01:00.0: [drm] Cannot find any crtc or sizesamdgpu.aspm=0 amdgpu.bapm=0 amdgpu.runpm=0 amdgpu.audio=0 amdgpu.ppfeaturemask=0xffffbffbEdit: you had "amdgpu.rumpm=0" (I typo a lot, so I won't make it, but you know that there's a joke in that
)
Last edited by seth (2024-03-10 22:51:27)
Offline
mar 10 19:16:59 home kernel: amdgpu 0000:01:00.0: amdgpu: Using BACO for runtime pm mar 10 19:16:59 home kernel: [drm] Initialized amdgpu 3.57.0 20150101 for 0000:01:00.0 on minor 1 mar 10 19:16:59 home kernel: amdgpu 0000:01:00.0: [drm] Cannot find any crtc or sizesamdgpu.aspm=0 amdgpu.bapm=0 amdgpu.runpm=0 amdgpu.audio=0 amdgpu.ppfeaturemask=0xffffbffb
Tried yout boot parameters, let me boot yeah (thanks to amdgpu.aspm=0) but nothing else happens. Nevertheless, I discovered something else.
If I boot without "pcie_aspm.policy=performance/powersave", the system boots. Without a display that is, but it boots WITH aspm. Now, I'm not getting a display out from any cable (tried both HDMI and DP, sadly I don't have a DVI connector).
journal with your params: https://termbin.com/p5ej
dmesg w your params: https://termbin.com/2b1c
[b0ss] @ 192.168.1.168 ~ 08:03
$ sudo modprobe -vvr amdgpu
modprobe: INFO: custom logging function 0x593edbb54200 registered
rmmod amdgpu
rmmod video
rmmod drm_display_helper
rmmod drm_ttm_helper
rmmod drm_suballoc_helper
rmmod i2c_algo_bit
rmmod drm_buddy
rmmod gpu_sched
rmmod drm_exec
rmmod amdxcp
modprobe: INFO: context 0x593edd2553c0 released
[b0ss] @ 192.168.1.168 ~ 08:03
$ sudo modprobe -vv amdgpu
modprobe: INFO: custom logging function 0x5c9310a43200 registered
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/acpi/video.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/display/drm_display_helper.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/drm_ttm_helper.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/drm_suballoc_helper.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/i2c/algos/i2c-algo-bit.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/drm_buddy.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/scheduler/gpu-sched.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/drm_exec.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/amd/amdxcp/amdxcp.ko.zst
insmod /lib/modules/6.7.9-arch1-1/kernel/drivers/gpu/drm/amd/amdgpu/amdgpu.ko.zst aspm=0 bapm=0 runpm=0 ppfeaturemask=0xffffbffb
modprobe: INFO: context 0x5c9310a8e3c0 released
$ cat /proc/cmdline
BOOT_IMAGE=/vmlinuz-linux root=UUID=90b89d1d-e013-4458-96bf-c105d0fec91b rw zswap.enabled=0 rootfstype=ext4 loglevel=9 amdgpu.aspm=0 amdgpu.bapm=0 amdgpu.runpm=0 aamdgpu.audio=0 amdgpu.ppfeaturemask=0xffffbffb sysrq_always_enabled=1And of course it doesn't load WMI now. Hell, I tried to remove that module AND EVEN THEN! IT DIDN'T LOAD IT BY DEFAULT ALONGSIDE AMDGPU!
Dang it. At least some progress is being made lol
Edit: you had "amdgpu.rumpm=0" (I typo a lot, so I won't make it, but you know that there's a joke in that wink)
hehe my finger slipped lol but i tried it again and it didn't change much
Offline
I'VE MADE IT!
Since I realized that the issue was within ASPM and not ACPI, I started to toy around with the parameters. What worked for me, and doesn't make much sense if you ask me, are the following parameters: "pcie_aspm=1 pcie_aspm.policy=performance amdgpu.aspm=1". I don't know why it works, but it works lol. Per documentation, both this options should be enabled by default. Why they were failing before, I do not know. I don't understand it either.
I'll mark this thread as solved, for now at least.
Offline
pcie_aspm and amdgpu.aspm default to whether your HW supports it.
The canonical settings for "pcie_aspm" are actually on/off, not 1/0
Have you tried "pcie_aspm.policy=performance" in isolation?
It should™ be equivalent to "pcie_aspm=off" (and at this point amdgpu.aspm should™ become moot, there's no global aspm, but oc. it might hit a bug)
Offline
pcie_aspm and amdgpu.aspm default to whether your HW supports it.
The canonical settings for "pcie_aspm" are actually on/off, not 1/0
Makes sense. Nevertheless, in my BIOS options I do have a "ASPM Support" option which is set to Enabled. I tried it with and without the feature enabled and it didn't work, at least not without the explicit kernel parameters I passed it. As long as the canonical settings you mention, GRUB didn't say something like invalid option like it usually does when you pass a wrong argument to a parameter. Don't know why, maybe it's intentional?
Have you tried "pcie_aspm.policy=performance" in isolation?
Yes I have. The system didn't boot, it acted just as it did without "acpi=off".
It should™ be equivalent to "pcie_aspm=off" (and at this point amdgpu.aspm should™ become moot, there's no global aspm, but oc. it might hit a bug)
You *might* be right. Nevertheless, the system boots lol. This solution shouldn't work, but by some miracle it does. I think I'll never really understand why it's working now. Hell, I even played a bit of Yakuza Kiwami 2 (on ultra for the first time in my lifeeeeee YES) and it worked flawlessly.
I guess we'll probably never know why. At least I can play some Yakuza lol
Offline
The bool option parser is likely tolerant enough to support variaous syntax here.
The constellation you have (forcefully) enables pcie_aspm (can you omit that one) and (forcefully) enables amdgpu.aspm but w/ "pcie_aspm.policy=performance" makes sure that the device never gets powered down.
The hypothesis would be that the driver needs to enter some aspm codepath to not run into a bug, but aspm itself causes fatal underpowering.
This might be very specific to your MoBo, but should probably show up somewhere in https://wiki.archlinux.org/title/AMDGPU#Troubleshooting
(Ideally w/ as much data as possible, your board supports aspm and pcie_aspm should™ therefore not be necessary - if it is, that would have to be pointed out)
Did you file a bug against https://gitlab.freedesktop.org/drm/amd/-/issues/ ?
Last edited by seth (2024-03-11 15:59:26)
Offline
The bool option parser is likely tolerant enough to support variaous syntax here.
The constellation you have (forcefully) enables pcie_aspm (can you omit that one) and (forcefully) enables amdgpu.aspm but w/ "pcie_aspm.policy=performance" makes sure that the device never gets powered down.
The hypothesis would be that the driver needs to enter some aspm codepath to not run into a bug, but aspm itself causes fatal underpowering.
Now that I think about it, I remember changing the display output to HDMI, just to make sure, and it worked. As soon as the new monitor arrives, I'll try unsetting the "pcie_aspm" parameter and see what happens.
This might be very specific to your MoBo, but should probably show up somewhere in https://wiki.archlinux.org/title/AMDGPU#Troubleshooting
(Ideally w/ as much data as possible, your board supports aspm and pcie_aspm should™ therefore not be necessary - if it is, that would have to be pointed out)
Now I know that wiki page as the palm of my hand lol. No, my issue I have is nowhere to be seen there. I could add a section detailing issues with Lenovo mobos, but that's a task for another day. And honestly, I think you're correct. It most probably is a problem with this proprietary motherboard.
Did you file a bug against https://gitlab.freedesktop.org/drm/amd/-/issues/ ?
Nah, and honestly, I don't think it's worth the trouble. How many people are combining an old Lenovo M715s with a chinese RX580? Uhm, several probably, but I haven't read anywhere online about someone with my same exact issue. If the problem comes back, I will file a bug. But for now, I can rest, for a little while.
Thanks for your help seth, you've been an amazing help in this silly little troubleshooting journey.
Offline
Don't know if anyone is still looking at this issue but I'm having the same issue except - I am unable to get the APU to actually post on my 7945HX. My 7600M is alive and active but its like my system skips over even looking for the APU.
If anyone is still looking at this forum and can help i'd welcome anything. This is a fairly new system. Atomman G7 PT from Minisforum. Tore through bios. Nothing appears to be able to dedicate it to just the APU. Smart shift looks like its always enabled or on 'Auto' which still doesn't help but I know its suppose to be there. This system has never had windows on it and was bare bones. Any and all help appreciated. Let me know what you need.
Offline
How do you enter the BIOS if the system doesn't POST? External monitor?
My 7600M is alive and active but its like my system skips over even looking for the APU.
Post a system journal and lspci output from that configuration.
Offline