My ATI Radeon 5000 GPU won’t work on the Milk-V Jupiter.
I have a working openSUSE Tumbleweed 20240825 on my StarFive VisionFive 2 and an ATI Redeon 5000. (Custom kernel).
So I took the Milk-V Jupiter Ubuntu image and added a copy of my openSUSE Tumbleweed root partition. Made my own kernel from your repo and it works (almost).
I have text output from the ATI Radeon 5000 GPU, but when I want to go to X, it fails.
Ignore the hostname of the machine, that is due to the partition copy.
This is my u-boot
setenv dtb_addr 0x16000000
setenv ramdisk_addr 0x16100000
load nvme 0:7 ${kernel_addr_r} /home/opvolger/Image.gz
load nvme 0:5 ${dtb_addr} /spacemit/6.1.15/k1-x_milkv-jupiter.dtb
load nvme 0:5 ${ramdisk_addr} /initrd.img-6.1.15
setenv bootargs 'root=/dev/nvme0n1p7 rw radeon.modeset=1 radeon.dpm=0 radeon.pcie_gen2=0 swiotlb=131072 console=tty0 console=ttyS0,115200 earlycon rootwait stmmaceth=chain_mode:1 selinux=0'
booti $kernel_addr_r $ramdisk_addr:$filesize $dtb_addr
After booting, I have text and a “working” driver
starfive:~ # lspci -k
0001:00:00.0 PCI bridge: ASR Microelectronics Device 3003 (rev 01)
Kernel driver in use: pcieport
lspci: Unable to load libkmod resources: error -2
0001:01:00.0 Non-Volatile memory controller: Silicon Motion, Inc. SM2263EN/SM2263XT (DRAM-less) NVMe SSD Controllers (rev 03)
Subsystem: Silicon Motion, Inc. SM2263EN/SM2263XT (DRAM-less) NVMe SSD Controllers
Kernel driver in use: nvme
0002:00:00.0 PCI bridge: ASR Microelectronics Device 3003 (rev 01)
Kernel driver in use: pcieport
0002:01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Cedar [Radeon HD 5000/6000/7350/8350 Series]
Subsystem: ASUSTeK Computer Inc. Device 0366
Kernel driver in use: radeon
0002:01:00.1 Audio device: Advanced Micro Devices, Inc. [AMD/ATI] Cedar HDMI Audio [Radeon HD 5400/6300/7300 Series]
Subsystem: ASUSTeK Computer Inc. Device aa68
Kernel driver in use: snd_hda_intel
Error in dmesg:
starfive:~ # dmesg | grep radeon
[ 0.000000] Kernel command line: root=/dev/nvme0n1p7 rw amdgpu.pcie_gen2=0 radeon.pcie_gen2=0 swiotlb=131072 console=tty0 console=ttyS0,115200 earlycon rootwait stmmaceth=chain_mode:1 selinux=0
[ 2.253291] [drm] radeon kernel modesetting enabled.
[ 2.266646] radeon 0002:01:00.0: enabling device (0000 -> 0003)
[ 2.410109] radeon 0002:01:00.0: VRAM: 1024M 0x0000000000000000 - 0x000000003FFFFFFF (1024M used)
[ 2.419148] radeon 0002:01:00.0: GTT: 1024M 0x0000000040000000 - 0x000000007FFFFFFF
[ 2.435820] [drm] radeon: 1024M of VRAM memory ready
[ 2.440873] [drm] radeon: 1024M of GTT memory ready.
[ 2.466048] [drm] radeon: dpm initialized
[ 2.562715] radeon 0002:01:00.0: WB enabled
[ 2.566995] radeon 0002:01:00.0: fence driver on ring 0 use gpu addr 0x0000000040000c00
[ 2.575157] radeon 0002:01:00.0: fence driver on ring 3 use gpu addr 0x0000000040000c0c
[ 2.594296] radeon 0002:01:00.0: fence driver on ring 5 use gpu addr 0x000000000005c418
[ 2.603199] radeon 0002:01:00.0: radeon: MSI limited to 32-bit
[ 2.609361] radeon 0002:01:00.0: radeon: using MSI.
[ 2.614400] [drm] radeon: irq initialized.
[ 2.872558] [drm:r600_ring_test] *ERROR* radeon: ring 0 test failed (scratch(0x8504)=0xCAFEDEAD)
[ 2.881527] radeon 0002:01:00.0: disabling GPU acceleration
[ 3.362376] radeon 0002:01:00.0: [drm] fb0: radeondrmfb frame buffer device
[ 3.370356] [drm] Initialized radeon 2.50.0 20080528 for 0002:01:00.0 on minor 0
[ 27.044595] [drm:radeon_bo_move] *ERROR* failed to bind 1 pages at 0x00000000
[ 27.051938] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (4096, 2, 4096, -22)
[ 27.061430] [drm:radeon_bo_move] *ERROR* failed to bind 1 pages at 0x00000000
[ 27.068777] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (4096, 2, 4096, -22)
[ 27.706885] [drm:radeon_bo_move] *ERROR* failed to bind 6 pages at 0x00000000
[ 27.714312] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (24576, 2, 4096, -22)
[ 27.723996] [drm:radeon_bo_move] *ERROR* failed to bind 6 pages at 0x00000000
[ 27.731384] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (24576, 2, 4096, -22)
[ 27.748715] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 27.756762] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
[ 27.767290] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 27.775291] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
[ 27.787727] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 27.795795] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
[ 27.806444] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 27.814471] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
... etc etc etc
But if it boots into X Display Manager and used the DRM-driver
[ OK ] Reached target Network.
Starting NTP client/server...
Starting CUPS Scheduler...
Starting OpenSSH Daemon...
Starting Permit User Sessions...
[ OK ] Finished Permit User Sessions.
Starting X Display Manager...
Starting Hold until boot process finishes up...
[ 22.723121] [drm:radeon_bo_move] *ERROR* failed to bind 1 pages at 0x00000000
[ 22.730482] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (4096, 2, 4096, -22)
[ 22.739998] [drm:radeon_bo_move] *ERROR* failed to bind 1 pages at 0x00000000
[ 22.747279] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (4096, 2, 4096, -22)
[ 23.368008] [drm:radeon_bo_move] *ERROR* failed to bind 6 pages at 0x00000000
[ 23.375368] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (24576, 2, 4096, -22)
[ 23.385074] [drm:radeon_bo_move] *ERROR* failed to bind 6 pages at 0x00000000
[ 23.392446] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (24576, 2, 4096, -22)
[ 23.410720] radeon 0002:01:00.0: swiotlb buffer is full (sz: 1048576 bytes), total 65536 (slots), used 9 (slots)
[ 23.421155] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -14)
[ 23.431899] radeon 0002:01:00.0: swiotlb buffer is full (sz: 1048576 bytes), total 65536 (slots), used 9 (slots)
[ 23.442677] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -14)
[ 23.455180] radeon 0002:01:00.0: swiotlb buffer is full (sz: 1048576 bytes), total 65536 (slots), used 9 (slots)
[ 23.465573] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -14)
[ 23.476297] radeon 0002:01:00.0: swiotlb buffer is full (sz: 1048576 bytes), total 65536 (slots), used 9 (slots)
[ 23.486696] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -14)
[ 23.497436] radeon 0002:01:00.0: swiotlb buffer is full (sz: 1048576 bytes), total 65536 (slots), used 9 (slots)
[ 23.507921] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -14)
[ 23.518674] radeon 0002:01:00.0: swiotlb buffer is full (sz: 1048576 bytes), total 65536 (slots), used 9 (slots)
[ 23.529086] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -14)
[ 25.601817] k1x_emac cac81000.ethernet end1: Link is Up - 1Gbps/Full - flow control off
[ 25.601877] IPv6: ADDRCONF(NETDEV_CHANGE): end1: link becomes read
I have “fixxed” the “swiotlb buffer is full” with this patch: (Even more memory when I use this soc, this was already an adjustment for the K1, but apparently I’m running out of memory there too. + double the kernel argument)
diff --git a/include/linux/swiotlb.h b/include/linux/swiotlb.h
index eac643ef4..cfa9c9552 100644
--- a/include/linux/swiotlb.h
+++ b/include/linux/swiotlb.h
@@ -23,7 +23,7 @@ struct scatterlist;
* The complexity of {map,unmap}_single is linearly dependent on this value.
*/
#if defined(CONFIG_SOC_SPACEMIT_K1X)
-#define IO_TLB_SEGSIZE 256
+#define IO_TLB_SEGSIZE 512
#else
#define IO_TLB_SEGSIZE 128
#endif
Now i have still this errors (with more debug kernel options):
[ OK ] Reached target Network.
Starting NTP client/server...
Starting CUPS Scheduler...
Starting OpenSSH Daemon...
Starting Permit User Sessions...
[ OK ] Finished Permit User Sessions.
Starting X Display Manager...
Starting Hold until boot process finishes up...
[ 26.209756] vgaarb: vga_arb_open
[ 26.213140] vgaarb: client 0x0000000082e35b0d called 'target'
[ 26.219004] vgaarb: PCI:0002:01:00.0 ==> 0002:01:00.0 pdev 00000000b16d9eeb
[ 26.226071] vgaarb: vgadev 0000000030221129
[ 26.629622] [drm:radeon_bo_move] *ERROR* failed to bind 1 pages at 0x00000000
[ 26.637004] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (4096, 2, 4096, -22)
[ 26.647264] [drm:radeon_bo_move] *ERROR* failed to bind 1 pages at 0x00000000
[ 26.654573] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (4096, 2, 4096, -22)
[ 26.937275] vgaarb: client 0x0000000082e35b0d called 'target'
[ 26.943180] vgaarb: PCI:0002:01:00.0 ==> 0002:01:00.0 pdev 00000000b16d9eeb
[ 26.950257] vgaarb: vgadev 0000000030221129
[ 27.311882] [drm:radeon_bo_move] *ERROR* failed to bind 6 pages at 0x00000000
[ 27.319257] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (24576, 2, 4096, -22)
[ 27.328921] [drm:radeon_bo_move] *ERROR* failed to bind 6 pages at 0x00000000
[ 27.336247] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (24576, 2, 4096, -22)
[ 27.353595] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 27.361592] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
[ 27.372118] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 27.380118] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
[ 27.392171] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 27.400167] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
[ 27.410669] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 27.418735] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
[ 27.429235] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 27.437270] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
[ 27.447764] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 27.455843] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
I also tried the newer 6.6 kernel (with bigger swiotlb)
Tried to set in radeon_device the dma_bits to 32 (hardcoded, maybe 40-bit dma was a problem…)
[ 39.529290] trying to bind memory to uninitialized GART !
[ 39.534764] WARNING: CPU: 2 PID: 1157 at drivers/gpu/drm/radeon/radeon_gart.c:294 radeon_gart_bind+0xd2/0xd8
[ 39.534782] Modules linked in:
[ 39.534789] CPU: 2 PID: 1157 Comm: Xorg.bin Tainted: G W 6.6.36+ #5
[ 39.534795] Hardware name: Milk-V Jupiter (DT)
[ 39.534799] epc : radeon_gart_bind+0xd2/0xd8
[ 39.534805] ra : radeon_gart_bind+0xd2/0xd8
[ 39.534811] epc : ffffffff8090cd44 ra : ffffffff8090cd44 sp : ffffffc800aa3910
[ 39.534815] gp : ffffffff82b41f78 tp : ffffffd902899780 t0 : 7420676e69797274
[ 39.534818] t1 : 0000000000000074 t2 : 6f7420676e697972 s0 : ffffffc800aa3970
[ 39.534823] s1 : ffffffd9049ab478 a0 : 000000000000002d a1 : ffffffda75edb608
[ 39.534827] a2 : ffffffda75ee79a8 a3 : 0000000000000010 a4 : 0000000000000000
[ 39.534830] a5 : 0000000000000000 a6 : 0000000000000000 a7 : 0000000000000038
[ 39.534834] s2 : ffffffd900650000 s3 : ffffffd90d360700 s4 : ffffffc800aa3b00
[ 39.534838] s5 : 0000000000000000 s6 : ffffffc800aa3a78 s7 : 0000000000000000
[ 39.534841] s8 : ffffffd9006506e0 s9 : ffffffd9006506e0 s10: ffffffd9049ab478
[ 39.534845] s11: ffffffd9049ab400 t3 : ffffffff82b5ba9f t4 : ffffffff82b5ba9f
[ 39.534849] t5 : ffffffff82b5baa0 t6 : ffffffc800aa3718
[ 39.534851] status: 0000000200000120 badaddr: 0000000000000000 cause: 0000000000000003
[ 39.534855] [<ffffffff8090cd44>] radeon_gart_bind+0xd2/0xd8
[ 39.534863] [<ffffffff8090b03c>] radeon_bo_move+0x88/0x3ce
[ 39.534870] [<ffffffff808ead98>] ttm_bo_handle_move_mem+0x8a/0x102
[ 39.534879] [<ffffffff808eb666>] ttm_bo_validate+0x78/0xee
[ 39.534886] [<ffffffff808eb7ca>] ttm_bo_init_reserved+0xee/0x126
[ 39.534893] [<ffffffff808eb83e>] ttm_bo_init_validate+0x3c/0xaa
[ 39.534901] [<ffffffff8090baac>] radeon_bo_create+0x108/0x1aa
[ 39.534908] [<ffffffff8091db08>] radeon_gem_object_create+0x80/0x146
[ 39.534915] [<ffffffff8091dca2>] radeon_gem_create_ioctl+0x5c/0x134
[ 39.534921] [<ffffffff808b6736>] drm_ioctl_kernel+0xa0/0x122
[ 39.534932] [<ffffffff808b6968>] drm_ioctl+0x1b0/0x3c6
[ 39.534938] [<ffffffff808f3060>] radeon_drm_ioctl+0x3c/0x82
[ 39.534946] [<ffffffff80209d24>] __riscv_sys_ioctl+0x8e/0xa0
[ 39.534955] [<ffffffff815205a0>] do_trap_ecall_u+0x11a/0x12c
[ 39.534964] [<ffffffff81529b72>] ret_from_exception+0x0/0x6e
[ 39.534973] ---[ end trace 0000000000000000 ]---
[ 39.534977] [drm:radeon_bo_move] *ERROR* failed to bind 256 pages at 0x00000000
[ 39.754093] [drm:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (1048576, 2, 4096, -22)
With the error of the kernel 6.6 i have the idea that the GART is not “ready/correct” yet in the PCIe driver of this SoC? GART I am no kernel developer, it is a hobby to get stuff working in Linux. Here my knowledge of Linux kernels ends. I hope you can find a solution or are already working on it. It is also possible that this is still on the backlog, then I will have to wait a little longer.
Is the PCIe kernel driver not 100% complete? are there still missing stuff or is it not working correct (yet)?
But i will try it now (patch it on my 6.6 kernel)