Bug 209159

Summary:	AMD Vega 20 framebuffer switch fails on 5.9rc2+
Product:	Drivers	Reporter:	ryan
Component:	Video(DRI - non Intel)	Assignee:	drivers_video-dri
Status:	RESOLVED INVALID
Severity:	normal	CC:	alexdeucher
Priority:	P1
Hardware:	x86-64
OS:	Linux
Kernel Version:	5.9-rc2/3	Subsystem:
Regression:	No	Bisected commit-id:
Attachments:	lspci dmesg

Description ryan 2020-09-05 10:19:22 UTC

Created attachment 292349 [details]
lspci

Vega 20 (Radeon VII) on X99 platform - boot freezes during FB switch. Booting without quiet and with earlyprintk=efi,keep shows a stall immediately after

...
[    1.941238] AMD-Vi: AMD IOMMUv2 driver by Joerg Roedel <jroedel@suse.de>
[    1.941239] AMD-Vi: AMD IOMMUv2 functionality not available on this system
[    1.948454] nvme nvme0: 20/0/0 default/read/poll queues
[    1.953176]  nvme0n1: p1 p2 p3 p4 p5 p6 p7
[    1.956313] usb 1-14: new full-speed USB device number 5 using xhci_hcd
[    1.989790] [drm] amdgpu kernel modesetting enabled.
[    1.989849] CRAT table not found
[    1.989850] Virtual CRAT table created for CPU
[    1.989857] amdgpu: Topology: Add CPU node
[    2.006244] checking generic (c0000000 300000) vs hw (c0000000 10000000)
[    2.006246] fb0: switching to amdgpudrmfb from EFI VGA


On -rc1 boot continues normally with
...
[    2.006315] amdgpu 0000:67:00.0: vgaarb: deactivate vga console
[    2.006345] amdgpu 0000:67:00.0: enabling device (0106 -> 0107)
[    2.006408] [drm] initializing kernel modesetting (VEGA20 0x1002:0x66AF 0x1002:0x081E 0xC1).
...

Comment 1 Alex Deucher 2020-09-05 15:31:42 UTC

Please attach your full dmesg output.  Can you bisect?

Comment 2 ryan 2020-09-06 02:00:30 UTC

Created attachment 292367 [details]
dmesg

Comment 3 ryan 2020-09-06 02:01:32 UTC

Thanks, sorry had to bypass LUKS to get the whole thing. This looks bad, looks like the firmware upload to the card is failing. It is also happening on all my installed kernels sorry, including the Fedora provided 5.8 series. Bad hardware?

Comment 4 ryan 2020-09-06 02:42:36 UTC

Whoops seems this was PEBKAC, I'd previously overridden (and forgot about overriding) my dracut.conf to load a specific firmware version and when that fell off the bottom of the kernel rotation, obviously it didn't load into my initrd and failed. Cleared that and all well, sorry for the noise.