LA UOSC

显卡是 R5 340 (GCN 1.0 Oland 核心，R7 240/250 同款)，具体表现为开机画面冻结（Freezing），但SSH正常，通过SSH登录后使用dmesg 查看日志发现：

[    4.625417] amdgpu 0000:07:00.0: amdgpu: (-12) failed to allocate kernel bo
[    4.625420] [drm:amdgpu_doorbell_create_kernel_doorbells [amdgpu]] *ERROR* Failed to allocate kernel doorbells, err=-12
[    4.625682] [drm:amdgpu_ttm_init [amdgpu]] *ERROR* Failed to initialize kernel doorbells.
[    4.625931] [drm:amdgpu_device_init [amdgpu]] *ERROR* sw_init of IP block <gmc_v6_0> failed -12
[    4.626179] amdgpu 0000:07:00.0: amdgpu: amdgpu_device_ip_init failed
[    4.626183] amdgpu 0000:07:00.0: amdgpu: Fatal error during GPU init

内核参数如下:

GRUB_CMDLINE_LINUX="radeon.si_support=0 amdgpu.si_support=1 amdgpu.dc=1"

根据这个log应该是 amdgpu_bo_create 返回了 -ENOMEM。这个函数在最新的内核里只有两种可能返回 -ENOMEM，一个是 amdgpu_bo_validate_size 返回 false 时, 一个是 kvzalloc 没法分配内核内存时。要知道具体是那个你得自己编译一个开了调试的内核然后在相应函数处打一些log。话说GCN1.0 radeon驱动也能够支持吧？换radeon驱动试试看？

autistic-chad

radeon驱动是可以的，但是radeon驱动在wayland下没法使用pipewire

现在使用 6.7 内核可以进入桌面了，但运行一段时间驱动会崩溃，但重启 plasmashell 后正常，可问题是这样会导致显存溢出导致画面再次冻结

dmesg 日志是这些，感觉还是内核和驱动有问题：

[  491.262271] amdgpu 0000:07:00.0: amdgpu: GPU fault detected: 147 0x020c5002
[  491.262278] amdgpu 0000:07:00.0: amdgpu:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x00041010
[  491.262280] amdgpu 0000:07:00.0: amdgpu:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0C050002
[  491.262281] amdgpu 0000:07:00.0: amdgpu: VM fault (0x02, vmid 6) at page 266256, read from '' (0x00000000) (80)
[  501.344863] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx timeout, signaled seq=12889, emitted seq=12891
[  501.345189] [drm:amdgpu_job_timedout [amdgpu]] *ERROR* Process information: process kwin_wayland pid 3009 thread kwin_wayla:cs0 pid 3053
[  501.345490] amdgpu 0000:07:00.0: amdgpu: GPU reset begin!
[  502.025082] amdgpu 0000:07:00.0: amdgpu: PCI CONFIG reset
[  502.028595] amdgpu 0000:07:00.0: amdgpu: GPU reset succeeded, trying to resume
[  502.028699] [drm] PCIE gen 3 link speeds already enabled
[  502.029987] amdgpu 0000:07:00.0: amdgpu: PCIE GART of 1024M enabled (table at 0x000000F400200000).
[  502.666952] [drm] UVD initialized successfully.
[  502.987853] amdgpu 0000:07:00.0: amdgpu: recover vram bo from shadow start
[  502.989557] amdgpu 0000:07:00.0: amdgpu: recover vram bo from shadow done
[  502.989581] [drm] Skip scheduling IBs!
[  503.026235] amdgpu 0000:07:00.0: amdgpu: GPU reset(1) succeeded!
[  503.026536] [drm:amdgpu_cs_ioctl [amdgpu]] *ERROR* Failed to initialize parser -125!

目前更新了 https://github.com/loongson/Firmware/tree/main/6000Series/PC/XA61200 的固件后正常

内核参数参考（参数只对GCN 1.0/2.0架构生效，北极星架构之后的显卡谨慎使用）：

GRUB_CMDLINE_LINUX="radeon.cik_support=0 radeon.si_support=0 amdgpu.cik_support=1 amdgpu.si_support=1 amdgpu.sg_display=0 amdgpu.runpm=0 amdgpu.ppfeaturemask=0xffffffff"

已知问题：

1. 使用该显卡播放并解码H.264视频一段时间后会强制降频到最低频率，原因不明