GPU

[multi.flexi]

Hi,
did anybody make desktop environment work in Ubuntu? After installing GNOME and GDM3 I still am getting error and cannot start X server and have no output from GPU.

[kgbvax]

Did you make any progress?

I stumble running X with error
” (EE) open /dev/dri/card0: No such file or directory”

[b-man]

What revision of the board/graphics card are you using? Are you booting with UEFI or u-boot?
I’ve been using a Asus Radeon R7 240 graphics card with my rev 1.3 board and I’ve had consistent success booting into Gnome 3 from UEFI, however I still needed to disable DPM (Dynamic Power Management) in the radeon driver due to it crashing when accessing the ECAM area to query PCIe bus speeds (this is due to the Synopsys PCIe controller in the 8040 not being totally ECAM compliant, and requiring a workaround that Marvell engineers had implemented).
I still have a weird problem where some (x11?) apps such as glxgears and minetest crash, and TuxCart is glitchy, but I believe that this may be more of a driver related issue than anything else.

[kgbvax]

I am using the Rev 1.3 board pretty mich stock: With u-boot & ubuntu 16.04 (sd-card image).
Only kernel change is that I’ve enabled support for Radeon (AMDGPU).
The card is a Radeon R230.

I worry a bit that flashing the UEFI EDK may brick the system.

[kgbvax]

Ok, one silly mistake was to try to use AMDGPU for the older card.

Now I get:
[ 6.735200] radeon 0000:01:00.0: enabling device (0000 -> 0003)
[ 6.735489] [drm] initializing kernel modesetting (CEDAR 0x1002:0x68E1 0x1787:0x3000).
[ 6.735499] [drm] register mmio base: 0xF6100000
[ 6.735501] [drm] register mmio size: 131072
[ 6.856533] ATOM BIOS: PARK
[ 6.856606] [drm] GPU not posted. posting now…
[ 6.859838] radeon 0000:01:00.0: VRAM: 1024M 0x0000000000000000 – 0x000000003FFFFFFF (1024M used)
[ 6.859845] radeon 0000:01:00.0: GTT: 1024M 0x0000000040000000 – 0x000000007FFFFFFF
[ 6.859848] [drm] Detected VRAM RAM=1024M, BAR=0M
[ 6.859851] [drm] RAM width 64bits DDR
[ 6.859936] [TTM] Zone kernel: Available graphics memory: 2020752 kiB
[ 6.859940] [TTM] Initializing pool allocator
[ 6.859949] [TTM] Initializing DMA pool allocator
[ 6.859987] radeon 0000:01:00.0: Fatal error during GPU init

this is also described here: https://bugs.freedesktop.org/show_bug.cgi?id=102497

Can you post your dmesg output for reference?

[b-man]

Sorry for the slow response, I’ve been very busy.

Here are copies of the dmesg dumps for both acpi and device tree. In both instances I’ve been running the vanilla Fedora 28 kernel.
I do know that there has been recent work in enabling PCIe in the UEFI, perhaps the PCIe controller is initialized differently when booting with UEFI than with u-boot?

[kgbvax]

PCI is in fact initialised differently with UEFI than in u-boot. In u-Boot, the PCI memory window is much smaller.

Here’s u-boot with mainline kernel and a Radeon R9 nano:
[ 3.463038] pci 0000:00:00.0: BAR 15: no space for [mem size 0x18000000 pref]
[ 3.470207] pci 0000:00:00.0: BAR 15: failed to assign [mem size 0x18000000 pref]
[ 3.477726] pci 0000:00:00.0: BAR 0: assigned [mem 0xf6000000-0xf60fffff 64bit]
[ 3.485073] pci 0000:00:00.0: BAR 14: assigned [mem 0xf6100000-0xf61fffff]
[ 3.491982] pci 0000:00:00.0: BAR 13: assigned [io 0x1000-0x1fff]
[ 3.498196] pci 0000:01:00.0: BAR 0: no space for [mem size 0x10000000 64bit pref]
[ 3.505800] pci 0000:01:00.0: BAR 0: failed to assign [mem size 0x10000000 64bit pref]
[ 3.513753] pci 0000:01:00.0: BAR 2: no space for [mem size 0x00200000 64bit pref]
[ 3.521357] pci 0000:01:00.0: BAR 2: failed to assign [mem size 0x00200000 64bit pref]
[ 3.529309] pci 0000:01:00.0: BAR 5: assigned [mem 0xf6100000-0xf613ffff]
[ 3.536136] pci 0000:01:00.0: BAR 6: assigned [mem 0xf6140000-0xf615ffff pref]
[ 3.543391] pci 0000:01:00.1: BAR 0: assigned [mem 0xf6160000-0xf6163fff 64bit]
[ 3.550761] pci 0000:01:00.0: BAR 4: assigned [io 0x1000-0x10ff]
[ 3.556891] pci 0000:00:00.0: PCI bridge to [bus 01-ff]
[ 3.562141] pci 0000:00:00.0: bridge window [io 0x1000-0x1fff]
[ 3.568262] pci 0000:00:00.0: bridge window [mem 0xf6100000-0xf61fffff]
[ 3.575114] pcieport 0000:00:00.0: assign IRQ: got 40
[ 3.580283] pcieport 0000:00:00.0: Signaling PME with IRQ 40
[ 3.586025] pcieport 0000:00:00.0: AER enabled with IRQ 40
[ 3.591617] amdgpu 0000:01:00.0: assign IRQ: got 40
[ 3.596784] [drm] initializing kernel modesetting (FIJI 0x1002:0x7300 0x1043:0x04A6 0xCA).
[ 3.605098] [drm] register mmio base: 0xF6100000
[ 3.609739] [drm] register mmio size: 262144
[ 3.614031] [drm] probing gen 2 caps for device 11ab:110 = 523c43/e
[ 3.620328] [drm] probing mlw for device 11ab:110 = 523c43
[ 3.625841] [drm] add ip block number 0 <vi_common>
[ 3.630742] [drm] add ip block number 1 <gmc_v8_0>
[ 3.635555] [drm] add ip block number 2 <tonga_ih>
[ 3.640368] [drm] add ip block number 3 <powerplay>
[ 3.645268] [drm] add ip block number 4 <dce_virtual>
[ 3.650343] [drm] add ip block number 5 <gfx_v8_0>
[ 3.655155] [drm] add ip block number 6 <sdma_v3_0>
[ 6.814873] random: fast init done
[ 27.722357] Doesn’t get ack from pf.
[ 27.725947] Doesn’t get ack from pf, give up
[ 27.730238] amdgpu 0000:01:00.0: Fatal error during GPU init

And the same card with UEFI (and Fedora28 stock kernel)

[ 2.259106] PCI: OF: host bridge /cp0/pcie@f2600000 ranges:
[ 2.264722] PCI: OF: IO 0xf9000000..0xf900ffff -> 0xf9000000
[ 2.270671] PCI: OF: MEM 0xc0000000..0xdfffffff -> 0xc0000000
[ 2.276917] armada8k-pcie f2600000.pcie: link up
[ 2.281597] armada8k-pcie f2600000.pcie: PCI host bridge to bus 0000:00
[ 2.288261] pci_bus 0000:00: root bus resource [bus 00-ff]
[ 2.293771] pci_bus 0000:00: root bus resource [io 0x0000-0xffff] (bus address [0xf9000000-0xf900ffff])
[ 2.303296] pci_bus 0000:00: root bus resource [mem 0xc0000000-0xdfffffff]
[ 2.310215] pci 0000:00:00.0: [11ab:0110] type 01 class 0x060400
[ 2.310230] pci 0000:00:00.0: reg 0x10: [mem 0x00000000-0x000fffff 64bit]
[ 2.310266] pci 0000:00:00.0: supports D1 D2
[ 2.310268] pci 0000:00:00.0: PME# supported from D0 D1 D3hot
[ 2.310415] pci 0000:01:00.0: [1002:7300] type 00 class 0x030000
[ 2.310533] pci 0000:01:00.0: reg 0x10: [mem 0x800000000-0x80fffffff 64bit pref]
[ 2.310572] pci 0000:01:00.0: reg 0x18: [mem 0x810000000-0x8101fffff 64bit pref]
[ 2.310595] pci 0000:01:00.0: reg 0x20: initial BAR value 0x00000000 invalid
[ 2.317679] pci 0000:01:00.0: reg 0x20: [io size 0x0100]
[ 2.317701] pci 0000:01:00.0: reg 0x24: [mem 0xc0000000-0xc003ffff]
[ 2.317724] pci 0000:01:00.0: reg 0x30: [mem 0xfffe0000-0xffffffff pref]
[ 2.317747] pci 0000:01:00.0: enabling Extended Tags
[ 2.322997] pci 0000:01:00.0: supports D1 D2
[ 2.322999] pci 0000:01:00.0: PME# supported from D1 D2 D3hot D3cold
[ 2.323217] pci 0000:01:00.0: vgaarb: VGA device added: decodes=io+mem,owns=none,locks=none
[ 2.331697] pci 0000:01:00.1: [1002:aae8] type 00 class 0x040300
[ 2.331794] pci 0000:01:00.1: reg 0x10: [mem 0x810200000-0x810203fff 64bit]
[ 2.331919] pci 0000:01:00.1: enabling Extended Tags
[ 2.337110] pci 0000:01:00.1: supports D1 D2
[ 2.337323] pci 0000:00:00.0: BAR 15: assigned [mem 0xc0000000-0xd7ffffff 64bit pref]
[ 2.345196] pci 0000:00:00.0: BAR 0: assigned [mem 0xd8000000-0xd80fffff 64bit]
[ 2.352541] pci 0000:00:00.0: BAR 14: assigned [mem 0xd8100000-0xd81fffff]
[ 2.359450] pci 0000:00:00.0: BAR 13: assigned [io 0x1000-0x1fff]
[ 2.365661] pci 0000:01:00.0: BAR 0: assigned [mem 0xc0000000-0xcfffffff 64bit pref]
[ 2.373468] pci 0000:01:00.0: BAR 2: assigned [mem 0xd0000000-0xd01fffff 64bit pref]
[ 2.381277] pci 0000:01:00.0: BAR 5: assigned [mem 0xd8100000-0xd813ffff]
[ 2.388106] pci 0000:01:00.0: BAR 6: assigned [mem 0xd8140000-0xd815ffff pref]
[ 2.395363] pci 0000:01:00.1: BAR 0: assigned [mem 0xd8160000-0xd8163fff 64bit]
[ 2.402733] pci 0000:01:00.0: BAR 4: assigned [io 0x1000-0x10ff]
[ 2.408865] pci 0000:00:00.0: PCI bridge to [bus 01-ff]
[ 2.414113] pci 0000:00:00.0: bridge window [io 0x1000-0x1fff]
[ 2.420236] pci 0000:00:00.0: bridge window [mem 0xd8100000-0xd81fffff]
[ 2.427057] pci 0000:00:00.0: bridge window [mem 0xc0000000-0xd7ffffff 64bit pref]
[ 2.434966] pcieport 0000:00:00.0: Signaling PME with IRQ 19
[ 2.440692] pcieport 0000:00:00.0: AER enabled with IRQ 19
….
[ 4.483998] ATOM BIOS: 115-C8820000-100
[ 4.488832] [drm] GPU posting now…
…
[ 9.874591] [drm:atom_op_jump [amdgpu]] *ERROR* atombios stuck in loop for more than 5secs aborting
[ 9.883906] [drm:amdgpu_atom_execute_table_locked [amdgpu]] *ERROR* atombios stuck executing B434 (len 304, WS 4, PS 0) @ 0xB4B7
[ 9.895757] [drm:amdgpu_atom_execute_table_locked [amdgpu]] *ERROR* atombios stuck executing B0E2 (len 183, WS 0, PS 8) @ 0xB15C
[ 9.907386] amdgpu 0000:01:00.0: gpu post error!
[ 9.912024] amdgpu 0000:01:00.0: Fatal error during GPU init
[ 9.917749] [drm] amdgpu: finishing device.
[ 9.922191] amdgpu: probe of 0000:01:00.0 failed with error -22

[Mikulas]

GPU only works if you use the EFI bootloader, it doesn’t work with U-BOOT.

However, even if you boot it from EFI and it initializes, there are more problems:

The PCIe controller corrupts data when there are multiple overlapping writes: https://lkml.org/lkml/2018/8/2/742
With AMD Radeon, OpenGL rendering is corrupted. EXA seems to work.
Nouveau seems to work but locks up when playing video.

I use MacchiatoBIN with DisplayLink USB2 adapter, it is much more reliable than a GPU.

[Author]

Posts