[SOLVED] Problem with AMDGPU: blank screen at boot

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

Hello again! (old user coming back).

I am writting because I am having problems with a new AMD GPU on my main desktop (It has been working fine with a Nvidia card, but I decided to replace it in order to be able to use newer kernels and free drivers).

The card I did choose is a ATI Radeon RX570 (but lspci reports it as an RX 470/480). I am using as CPU a Ryzen7, so I don't have any integrated GPU to produce video output. I followed all the AMDGPU instructions and made sure everything was set in the kernel (4.14.11-r2), bit after what would be loading KMS I only get a black screen. The computer otherwise works and I am able to access it via ssh. Typing my username/password and startx won't start my monitor and X. I don't normally use any initrd image, but after googling I decided to try it out and that won't do anything. I see that someone had the same problem years ago and it was a firmware problem. However, after installing linux-firmware and setting it into the config file I didn't make any change.

The firmware section I did set in my .config is as it says: (the wiki says that the RX570 is POLARIS10)

Chrishas · n00b Joined: 19 Jul 2015 Posts: 12

Hi vladimir1986,

Can you provide your dmesg output?

DancesWithWords · Posted: Fri Jan 05, 2018 7:23 pm Post subject:

I would forget about installing the firmware into the kernel and do the following.

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

Thank you!

about the DMESG, I get this:

[ 0.610272] [drm] amdgpu kernel modesetting enabled.
[ 0.611494] amdgpu 0000:24:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
[ 0.612155] amdgpu 0000:24:00.0: SME is active, device will require DMA bounce buffers
[ 0.612320] amdgpu 0000:24:00.0: SME is active, device will require DMA bounce buffers
[ 0.612491] amdgpu 0000:24:00.0: VRAM: 4096M 0x000000F400000000 - 0x000000F4FFFFFFFF (4096M used)
[ 0.612657] amdgpu 0000:24:00.0: GTT: 256M 0x0000000000000000 - 0x000000000FFFFFFF
[ 0.613510] [drm] amdgpu: 4096M of VRAM memory ready
[ 0.613614] [drm] amdgpu: 4096M of GTT memory ready.
[ 0.614370] amdgpu 0000:24:00.0: amdgpu: using MSI.
[ 0.614463] [drm] amdgpu: irq initialized.
[ 0.795847] amdgpu: [powerplay] amdgpu: powerplay sw initialized
[ 0.796067] [drm] AMDGPU Display Connectors
[ 0.799687] amdgpu 0000:24:00.0: fence driver on ring 0 use gpu addr 0x0000000000400040, cpu addr 0xffff975348709040
[ 0.799851] amdgpu 0000:24:00.0: fence driver on ring 1 use gpu addr 0x00000000004000c0, cpu addr 0xffff9753487090c0
[ 0.800019] amdgpu 0000:24:00.0: fence driver on ring 2 use gpu addr 0x0000000000400140, cpu addr 0xffff975348709140
[ 0.800183] amdgpu 0000:24:00.0: fence driver on ring 3 use gpu addr 0x00000000004001c0, cpu addr 0xffff9753487091c0
[ 0.800340] amdgpu 0000:24:00.0: fence driver on ring 4 use gpu addr 0x0000000000400240, cpu addr 0xffff975348709240
[ 0.800499] amdgpu 0000:24:00.0: fence driver on ring 5 use gpu addr 0x00000000004002c0, cpu addr 0xffff9753487092c0
[ 0.800658] amdgpu 0000:24:00.0: fence driver on ring 6 use gpu addr 0x0000000000400340, cpu addr 0xffff975348709340
[ 0.800814] amdgpu 0000:24:00.0: fence driver on ring 7 use gpu addr 0x00000000004003c0, cpu addr 0xffff9753487093c0
[ 0.800970] amdgpu 0000:24:00.0: fence driver on ring 8 use gpu addr 0x0000000000400440, cpu addr 0xffff975348709440
[ 0.801144] amdgpu 0000:24:00.0: fence driver on ring 9 use gpu addr 0x00000000004004e0, cpu addr 0xffff9753487094e0
[ 0.801691] amdgpu 0000:24:00.0: fence driver on ring 10 use gpu addr 0x0000000000400560, cpu addr 0xffff975348709560
[ 0.801863] amdgpu 0000:24:00.0: fence driver on ring 11 use gpu addr 0x00000000004005e0, cpu addr 0xffff9753487095e0
[ 0.802382] amdgpu 0000:24:00.0: fence driver on ring 12 use gpu addr 0x000000f4002ad420, cpu addr 0xffff975348e5a420
[ 0.802672] amdgpu 0000:24:00.0: fence driver on ring 13 use gpu addr 0x00000000004006e0, cpu addr 0xffff9753487096e0
[ 0.802829] amdgpu 0000:24:00.0: fence driver on ring 14 use gpu addr 0x0000000000400760, cpu addr 0xffff975348709760
[ 1.603540] [drm:gfx_v8_0_ring_test_ring] *ERROR* amdgpu: ring 0 test failed (scratch(0xC040)=0xCAFEDEAD)
[ 1.603692] [drm:amdgpu_device_init] *ERROR* hw_init of IP block <gfx_v8_0> failed -22
[ 1.603838] amdgpu 0000:24:00.0: amdgpu_init failed
[ 2.290390] [drm] amdgpu: ttm finalized
[ 2.290488] amdgpu 0000:24:00.0: Fatal error during GPU init
[ 2.290590] [drm] amdgpu: finishing device.
[ 2.291153] amdgpu: probe of 0000:24:00.0 failed with error -22

I recompiled as recommended and had the same result

DancesWithWords · Posted: Fri Jan 05, 2018 8:32 pm Post subject:

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

Mhhh I already tried to do the same and avoid loading the firmware, but I still simply get a black screen. I am going to try today a couple of new things and see if that helps: I noticed there is a 4.14.12 kernel, I might try to remove the card and check the connectors (the case is very tight in space and maybe the card isn't properly connected). Hope any of that will help

Marlo · Veteran Joined: 26 Jul 2003 Posts: 1591

Try this and pay attention to the USE flags

Ant P. · Watchman Joined: 18 Apr 2009 Posts: 6920

@ above: This is failing in the kernel before X11 even runs. Try and pay attention to what you're replying to next time.

Marlo · Veteran Joined: 26 Jul 2003 Posts: 1591

If it's just a kernel problem, as Ant P. says, then that's the relevant part of the .config:

Chrishas · n00b Joined: 19 Jul 2015 Posts: 12

I had a similar error with my Vega 64 and the latest 4.15-rc kernel which went away when I disabled AMD Secure Memory Encryption (SME) support under Processor Type and Features, although could be completely unrelated to your issue.

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

Hey, thank you for the replies!

I tried the new .config, in case I mispelled something in the firmware, and it still doesn't work... (Yeah, I did set the video card as in the make.conf, but I need to have any video output before attacking X).

I did get some new error messages on Dmesg tho, or maybe I didn't see it before. Is that failing of amdgpu.

Chrishas · n00b Joined: 19 Jul 2015 Posts: 12

Can you provide your kernel command line?[/code]

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

I must add that I tried with AMDGPU both as a module and into the Kernel, with no changes. It seems this is quite a difficult problem....

I will hold into the hope that it is some sort of driver problem and knowing that the 4.15 kernel will include some improvements for AMD gpu, I'll just wait till they release it. Otherwise I'll try the closed drivers though I'll rather avoid them.

The other solutions go through using my previous Nvidia card, and that is the very last thing I want.

I will keep this updated, in case I can mark it as solved.

chithanh · Posted: Wed Jan 10, 2018 4:25 pm Post subject:

The problem is that the amdgpu driver cannot initialize the hardware correctly. As you have already ruled out firmware problems, this is most likely a kernel bug. Some workarounds like messing with IOMMU configuration etc. might be possible, but in general you have to report to the developers and then wait for a bugfix.

In the meantime you can use unaccelerated graphics if you enable

NeddySeagoon · Posted: Wed Jan 10, 2018 4:46 pm Post subject:

vladimir1986,

Its worth trying the new amdgpu driver in the 4.15.0 kernel.
That's still a -rc version,
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

NeddySeagoon · Posted: Thu Jan 11, 2018 12:05 pm Post subject:

vladimir1986,

The 4.15 kernel requires more firmware. I found that out the hard way. :)
Its in linux-firmware if you get a new enough one.

Did you find the extra kernel option?
The old driver is still there too. I think the new one builds on top of it.

Check your firnware in the kernel. From linux firmware, I have

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

I have all those... I did install the latest linux-firmware from the repository. At the time I am writting this it is 20180103-r1
Which extra kernel option?

I am going to recompile after putting all that into the kernel conf file again, let's see if that works or gives something new to work with

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

ok, I missed that post about disabling SME support. Saddly that didn't do anything.

However, adding the firmware location into the kernel prompted a different error: I wish I knew what "error -2" means. Now it fails in a couple of firmware modules, so maybe it is a firmware problem after all...

I do have the possibility of having unacelerated graphics as suggested, or downgrading to a Nvidia card if I really needed to, but having my laptop in working order I'll like to track down this problem and learn how to fix it.

$ dmesg | grep drm
[ 0.801110] [drm] amdgpu kernel modesetting enabled.
[ 0.801328] [drm] initializing kernel modesetting (POLARIS10 0x1002:0x67DF 0x1DA2:0xE366 0xEF).
[ 0.801498] [drm] register mmio base: 0xFE900000
[ 0.801599] [drm] register mmio size: 262144
[ 0.801706] [drm] probing gen 2 caps for device 1022:1453 = 733903/e
[ 0.801812] [drm] probing mlw for device 1022:1453 = 733903
[ 0.801921] [drm] UVD is enabled in VM mode
[ 0.802036] [drm] UVD ENC is enabled in VM mode
[ 0.802139] [drm] VCE enabled in VM mode
[ 0.802696] [drm] GPU post is not needed
[ 0.802890] [drm] vm size is 64 GB, block size is 13-bit, fragment size is 9-bit
[ 0.803407] [drm] Detected VRAM RAM=4096M, BAR=256M
[ 0.803509] [drm] RAM width 256bits GDDR5
[ 0.805584] [drm] amdgpu: 4096M of VRAM memory ready
[ 0.805687] [drm] amdgpu: 4096M of GTT memory ready.
[ 0.805797] [drm] GART: num cpu pages 65536, num gpu pages 65536
[ 0.806044] [drm] PCIE GART of 256M enabled (table at 0x000000F400040000).
[ 0.806245] [drm] Supports vblank timestamp caching Rev 2 (21.10.2013).
[ 0.806328] [drm] Driver supports precise vblank timestamp query.
[ 0.806424] [drm] amdgpu: irq initialized.
[ 0.806700] [drm] AMDGPU Display Connectors
[ 0.806780] [drm] Connector 0:
[ 0.806855] [drm] DP-1
[ 0.806929] [drm] HPD6
[ 0.807011] [drm] DDC: 0x4868 0x4868 0x4869 0x4869 0x486a 0x486a 0x486b 0x486b
[ 0.807127] [drm] Encoders:
[ 0.807202] [drm] DFP1: INTERNAL_UNIPHY2
[ 0.807280] [drm] Connector 1:
[ 0.807355] [drm] DP-2
[ 0.807429] [drm] HPD4
[ 0.807503] [drm] DDC: 0x4870 0x4870 0x4871 0x4871 0x4872 0x4872 0x4873 0x4873
[ 0.807619] [drm] Encoders:
[ 0.807694] [drm] DFP2: INTERNAL_UNIPHY2
[ 0.807772] [drm] Connector 2:
[ 0.807847] [drm] HDMI-A-1
[ 0.807921] [drm] HPD1
[ 0.807995] [drm] DDC: 0x486c 0x486c 0x486d 0x486d 0x486e 0x486e 0x486f 0x486f
[ 0.808126] [drm] Encoders:
[ 0.808210] [drm] DFP3: INTERNAL_UNIPHY1
[ 0.808288] [drm] Connector 3:
[ 0.808363] [drm] HDMI-A-2
[ 0.808437] [drm] HPD5
[ 0.808512] [drm] DDC: 0x4874 0x4874 0x4875 0x4875 0x4876 0x4876 0x4877 0x4877
[ 0.808628] [drm] Encoders:
[ 0.808702] [drm] DFP4: INTERNAL_UNIPHY1
[ 0.808780] [drm] Connector 4:
[ 0.808855] [drm] DVI-D-1
[ 0.808929] [drm] HPD3
[ 0.809014] [drm] DDC: 0x487c 0x487c 0x487d 0x487d 0x487e 0x487e 0x487f 0x487f
[ 0.809132] [drm] Encoders:
[ 0.809208] [drm] DFP5: INTERNAL_UNIPHY
[ 0.812286] [drm] Found UVD firmware Version: 1.79 Family ID: 16
[ 0.813098] [drm] Found VCE firmware Version: 52.4 Binary ID: 3
[ 0.813800] [drm:amdgpu_cgs_get_firmware_info] *ERROR* Failed to request firmware
[ 0.895268] [drm:amdgpu_device_init] *ERROR* hw_init of IP block <amdgpu_powerplay> failed -22
[ 0.961735] WARNING: CPU: 2 PID: 1 at drivers/gpu/drm/drm_mm.c:897 drm_mm_takedown+0x1b/0x20
[ 0.962143] RIP: 0010:drm_mm_takedown+0x1b/0x20
[ 0.962297] drm_dev_register+0x15e/0x1f0
[ 0.965906] WARNING: CPU: 2 PID: 1 at drivers/gpu/drm/drm_mm.c:897 drm_mm_takedown+0x1b/0x20
[ 0.966350] RIP: 0010:drm_mm_takedown+0x1b/0x20
[ 0.966738] drm_dev_register+0x15e/0x1f0
[ 0.971383] [drm] amdgpu: ttm finalized
[ 0.971543] [drm] amdgpu: finishing device.

$ dmesg | grep amdgpu
[ 0.801110] [drm] amdgpu kernel modesetting enabled.
[ 0.802412] amdgpu 0000:24:00.0: Invalid PCI ROM header signature: expecting 0xaa55, got 0xffff
[ 0.803079] amdgpu 0000:24:00.0: VRAM: 4096M 0x000000F400000000 - 0x000000F4FFFFFFFF (4096M used)
[ 0.803243] amdgpu 0000:24:00.0: GTT: 256M 0x0000000000000000 - 0x000000000FFFFFFF
[ 0.805584] [drm] amdgpu: 4096M of VRAM memory ready
[ 0.805687] [drm] amdgpu: 4096M of GTT memory ready.
[ 0.806165] amdgpu 0000:24:00.0: amdgpu: using MSI.
[ 0.806424] [drm] amdgpu: irq initialized.
[ 0.806514] amdgpu: [powerplay] amdgpu: powerplay sw initialized
[ 0.809296] amdgpu 0000:24:00.0: Direct firmware load for amdgpu/polaris10_pfp_2.bin failed with error -2
[ 0.809421] amdgpu 0000:24:00.0: Direct firmware load for amdgpu/polaris10_me_2.bin failed with error -2
[ 0.809546] amdgpu 0000:24:00.0: Direct firmware load for amdgpu/polaris10_ce_2.bin failed with error -2
[ 0.809673] amdgpu 0000:24:00.0: Direct firmware load for amdgpu/polaris10_mec_2.bin failed with error -2
[ 0.809798] amdgpu 0000:24:00.0: Direct firmware load for amdgpu/polaris10_mec2_2.bin failed with error -2
[ 0.810141] amdgpu 0000:24:00.0: fence driver on ring 0 use gpu addr 0x0000000000400040, cpu addr 0x0000000009df9e11
[ 0.810300] amdgpu 0000:24:00.0: fence driver on ring 1 use gpu addr 0x00000000004000c0, cpu addr 0x00000000594f905f
[ 0.810452] amdgpu 0000:24:00.0: fence driver on ring 2 use gpu addr 0x0000000000400140, cpu addr 0x000000000bceb655
[ 0.810604] amdgpu 0000:24:00.0: fence driver on ring 3 use gpu addr 0x00000000004001c0, cpu addr 0x000000002aab8f24
[ 0.810754] amdgpu 0000:24:00.0: fence driver on ring 4 use gpu addr 0x0000000000400240, cpu addr 0x00000000c2a29eba
[ 0.810905] amdgpu 0000:24:00.0: fence driver on ring 5 use gpu addr 0x00000000004002c0, cpu addr 0x00000000bc2d43a7
[ 0.811075] amdgpu 0000:24:00.0: fence driver on ring 6 use gpu addr 0x0000000000400340, cpu addr 0x000000001b605875
[ 0.811227] amdgpu 0000:24:00.0: fence driver on ring 7 use gpu addr 0x00000000004003c0, cpu addr 0x00000000370abf85
[ 0.811378] amdgpu 0000:24:00.0: fence driver on ring 8 use gpu addr 0x0000000000400440, cpu addr 0x00000000e2350542
[ 0.811516] amdgpu 0000:24:00.0: fence driver on ring 9 use gpu addr 0x00000000004004e0, cpu addr 0x00000000cde88522
[ 0.811976] amdgpu 0000:24:00.0: fence driver on ring 10 use gpu addr 0x0000000000400560, cpu addr 0x00000000d46a316e
[ 0.812155] amdgpu 0000:24:00.0: fence driver on ring 11 use gpu addr 0x00000000004005e0, cpu addr 0x000000002cd3ccde
[ 0.812651] amdgpu 0000:24:00.0: fence driver on ring 12 use gpu addr 0x000000f4001e6420, cpu addr 0x000000004041f004
[ 0.812799] amdgpu 0000:24:00.0: fence driver on ring 13 use gpu addr 0x00000000004006e0, cpu addr 0x000000005cd50b01
[ 0.812945] amdgpu 0000:24:00.0: fence driver on ring 14 use gpu addr 0x0000000000400760, cpu addr 0x00000000740451a5
[ 0.813248] amdgpu 0000:24:00.0: fence driver on ring 15 use gpu addr 0x00000000004007e0, cpu addr 0x00000000b145e47e
[ 0.813395] amdgpu 0000:24:00.0: fence driver on ring 16 use gpu addr 0x0000000000400860, cpu addr 0x000000004238527b
[ 0.813676] amdgpu 0000:24:00.0: Direct firmware load for amdgpu/polaris10_k_smc.bin failed with error -2
[ 0.813800] [drm:amdgpu_cgs_get_firmware_info] *ERROR* Failed to request firmware
[ 0.851225] amdgpu: [powerplay] SMU Firmware start failed!
[ 0.895015] amdgpu: [powerplay] SMU Firmware start failed!
[ 0.895099] amdgpu: [powerplay] Failed to load SMU ucode.
[ 0.895180] amdgpu: [powerplay] smc start failed
[ 0.895268] [drm:amdgpu_device_init] *ERROR* hw_init of IP block <amdgpu_powerplay> failed -22
[ 0.895386] amdgpu 0000:24:00.0: amdgpu_init failed
[ 0.962297] amdgpu_vram_mgr_fini+0x28/0x50
[ 0.962297] amdgpu_ttm_fini+0xdb/0x210
[ 0.962297] amdgpu_bo_fini+0x9/0x30
[ 0.962297] amdgpu_fini+0x1ee/0x300
[ 0.962297] amdgpu_device_init+0xd88/0x1530
[ 0.962297] amdgpu_driver_load_kms+0x51/0x210
[ 0.962297] amdgpu_pci_probe+0x11d/0x150
[ 0.966738] amdgpu_gtt_mgr_fini+0x1c/0x50
[ 0.966738] amdgpu_ttm_fini+0xe8/0x210
[ 0.966738] amdgpu_bo_fini+0x9/0x30
[ 0.966738] amdgpu_fini+0x1ee/0x300
[ 0.966738] amdgpu_device_init+0xd88/0x1530
[ 0.966738] amdgpu_driver_load_kms+0x51/0x210
[ 0.966738] amdgpu_pci_probe+0x11d/0x150
[ 0.971383] [drm] amdgpu: ttm finalized
[ 0.971461] amdgpu 0000:24:00.0: Fatal error during GPU init
[ 0.971543] [drm] amdgpu: finishing device.
[ 0.971849] amdgpu: probe of 0000:24:00.0 failed with error -22

Thank you very much for all the help so far

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

It is working at last!

In the end, what I did is to compile amdgpu as a module, not into the kernel and leaving all the firmware into the config file:

CONFIG_EXTRA_FIRMWARE="amdgpu/polaris10_ce.bin amdgpu/polaris10_mc.bin amdgpu/polaris10_me.bin amdgpu/polaris10_mec2.bin amdgpu/polaris10_mec.bin amdgpu/polaris10_pfp.bin amdgpu/polaris10_rlc.bin amdgpu/polaris10_sdma1.bin amdgpu/polaris10_sdma.bin amdgpu/polaris10_smc.bin amdgpu/polaris10_smc_sk.bin amdgpu/polaris10_uvd.bin amdgpu/polaris10_vce.bin"
CONFIG_EXTRA_FIRMWARE_DIR="/lib/firmware"

When the stable 4.15 comes out I'll try to remove that line just in case, as I am not sure you'll need to manually specify each firmware blob, but now I need a bit of a break :lol:

As a sad note, Doom4 won't work anymore with a AMD card, but Doom 1 & 2 does, plus now I have full framebuffer support, so it makes it up for that!

Thank you all for the help!

NeddySeagoon · Posted: Thu Jan 11, 2018 3:39 pm Post subject:

vladimir1986,

The rules for kernel modules that need to load firmware are the same for all modules.

The firmware must be available when the module is initialised.
That means ..
if the module is <*> the firmware must be listed in CONFIG_EXTRA_FIRMWARE
if the module is <M> the firmware must be in /lib/firmware
The other two combinations don't work.

When built in modules are initialised, root is not yet mounted, so /lib/firmware cannot be read.
_________________
Regards,

NeddySeagoon

Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail.

firephoto · Posted: Fri Jan 12, 2018 4:38 pm Post subject:

vladimir1986 · n00b Joined: 09 Dec 2010 Posts: 74

NeddySeagoon · Posted: Sat Jan 13, 2018 12:02 pm Post subject:

vladimir1986,

I admit I cheated. I put some rubbish in CONFIG_EXTRA_FIRMWARE= using menuconfig, just as a placeholder.
Then I used $EDITOR on make.conf to change the rubbish to all the files given by