Raven1/Picasso gets an immediate compute crash then reboot since latest firmware logic

[chefkissinc]

Raven1/Picasso gets an immediate compute crash then reboot since latest firmware logic

[VisualEhrmanntraut]

We don't support Olarila. Don't use anything from them for your own good.

However, there is indeed a compute hang issue on Raven.

[MixedVictor]

Same here after downloading the latest version, I had to disable autologin to see what was happening. After trying to log in, it would just freeze and crash the system, I got a kernel panic log if that will help anyone. Also, I'm using a Ryzen 7 3700U.

[jalavoui]

latest github also crashing at boot

2023-09-06 13:57:47.393981+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) Accelerator successfully registered with controller.
2023-09-06 13:57:47.846839+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0][GPUCAP] refresh() --- Family: 142, Device: 0x15d8, revNo: 1, pciRevNo: c9, emuRevNo: 66.
2023-09-06 13:57:47.846844+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0][GPUCAP] refresh() --- Mem Size: FB: 2048 MB, Aper: 256 MB, Reg Aper: 512 KB.
2023-09-06 13:57:47.846846+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0][GPUCAP] refresh() --- Mem Config: Width: 128, Type: DDR4.
2023-09-06 13:57:47.846848+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0][GPUCAP] refresh() --- FB Base: 0xf400000000, Top: 0xf47fffffff, Offset: 0x1af000000.
2023-09-06 13:57:47.846849+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0][GPUCAP] refresh() --- Branding - family: "(null)"; device: "(null)"; model: "AMD Radeon Vega 8".
2023-09-06 13:57:47.846854+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0] [PPLIB] checkTtlErrorMessage() !!! TTL ERROR MESSAGE: {8743463585:[7:0:0] Error SW_IP_CLIENT_ID__GC: event_id=0xc00c0401 event_info:type=3 hw_id=0 event_specific_tag=0x0 pData=0 data_size=0 }{8838003130:[7:0:0] Error SW_IP_CLIENT_ID__GC: event_id=0xc0100201 event_info:type=3 hw_id=0 event_specific_tag=0x0 pData=0 data_size=0 }.
2023-09-06 13:57:47.849674+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0] [PPLIB] initializePowerTuneParameters() !!! HW TDP Limit is 0, use a default value.
2023-09-06 13:57:47.875230+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0] [PPLIB] checkTtlErrorMessage() !!! TTL ERROR MESSAGE: {8743463585:[7:0:0] Error SW_IP_CLIENT_ID__GC: event_id=0xc00c0401 event_info:type=3 hw_id=0 event_specific_tag=0x0 pData=0 data_size=0 }{8838003130:[7:0:0] Error SW_IP_CLIENT_ID__GC: event_id=0xc0100201 event_info:type=3 hw_id=0 event_specific_tag=0x0 pData=0 data_size=0 }.
2023-09-06 13:57:47.906247+0100 localhost kernel[0]: (AMDRadeonX5000) [7:0:0]: Controller is enabled, finish initialization
2023-09-06 13:57:47.906252+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0] [AGDC] AMDRadeonX6000_AmdAgdcServices::vendor_doDeviceAttribute() ??? Controller is NOT enabled so SKIP. cmd:[ 3].
2023-09-06 13:57:47.907345+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0] [FB:0] AmdRadeonFramebuffer::setCursorImage() !!! Driver is offline.
2023-09-06 13:57:47.907351+0100 localhost kernel[0]: (AMDRadeonX6000Framebuffer) [7:0:0] [FB:1] AmdRadeonFramebuffer::setCursorImage() !!! Driver is offline.
2023-09-06 13:57:58.517908+0100 localhost kernel[0]: (AMDRadeonX5000) [7:0:0]: channel 10 computeUQ7 event timeout; pending stamps [2..2]
2023-09-06 13:57:58.517916+0100 localhost kernel[0]: (AMDRadeonX5000) [7:0:0]: HW Channel 10 computeUQ7 is occupied by channel 10 stamp 2

i'll try figure out why

[VisualEhrmanntraut]

@jalavoui That’s the issue, yes

[VisualEhrmanntraut]

Commenting out the MEC firmware injection works around the issue. But that's not a solution.

[jalavoui]

@VisualEhrmanntraut i got it working - but not stable as it only ocasionaly loads
i'll update here if i can make it stable

[VisualEhrmanntraut]

I look forward to it. Provided it's not just disabling MEC or something

[jalavoui]

do you think the header info that exists in X5000HWLibs can cause this firmware bugs ? i kinda doubt as works fine on renoir

as for picasso can you try inject picasso_rlc_am4.bin ?

is there's a replacement for atidmcub_instruction_dcn21.bin ? i remenber some linux bin is used for raven/picasso

i'm still getting random sucess boots...

[VisualEhrmanntraut]

@jalavoui All the new firmware is taken straight from Windows.

[jalavoui]

the only thing that works is revert to gc 9.4 and disable /case kPSPCommandLoadIPFW/ for raven/picasso
but i bet you don't wanna do that
i'm kinda upset cause i like the new firmware logic you just did

[VisualEhrmanntraut]

@jalavoui Eh, don’t worry about it.