search

ROCm / ROCm

AMD ROCm™ Software - GitHub Home
https://rocm.docs.amd.com
MIT License
4.62k stars 384 forks source link

ROCm can't run on Intel NUC 8 NUC8i3CYSM #939

Closed luguang closed 10 months ago

luguang commented 4 years ago

Hi guys,

I tried to install ROCm on Intel NUC 8 NUC8i3CYSM recently. After the installation, I couldn't run rocminfo or clinfo.

Hardware spec: https://www.intel.com/content/www/us/en/products/boards-kits/nuc/mini-pcs/nuc8i3cysm.html

Intel Core i3-8121U (Gen 8) RAM: 8GB PCIe 3 GPU: Radeon RX540 with 2GB RAM Disk: 128GB SSD

OS: Ubuntu 18.04.3 x64 ROCm: newest version 2.9

It's weird because lspci says it's RX550. $ lspci | grep VGA 01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Lexa PRO [Radeon RX 550/550X] (rev c3)

Before I could install ROCm on NUC, I failed once due to Intel Security check. It failed at last step to install. Then I rebooted to BIOS and disable Intel security technology, and booted to Ubuntu and apt autoremove rocm-dkms and apt install rocm-dkms again. I am not sure whether this is clean enough.

Then I followed the guide at https://rocm.github.io/ROCmInstall.html#installing-from-amd-rocm-repositories, till:

$ /opt/rocm/bin/rocminfo ROCk module is loaded username is member of video group Killed

And at the console I found some error messages

1623148944

$ /opt/rocm/opencl/bin/x86_64/clinfo

It hangs there always, then I killed it.

1769260135

At last I could not restart the box normally because clinfo hangs there.

What would be the problem?

Thanks,

Gavin

luguang commented 4 years ago

$ ./rocm_smi.py ========================ROCm System Management Interface========================

GPU Temp AvgPwr SCLK MCLK Fan Perf PwrCap VRAM% GPU%
0 34.0c 3.179W 214Mhz 300Mhz 18.82% auto 25.0W 1% 0%

==============================End of ROCm SMI Log ==============================

Part of syslog after rocminfo

Nov 19 15:59:33 ecl-amd kernel: [ 309.879633] BUG: unable to handle kernel NULL pointer dereference at 00000000000001fc Nov 19 15:59:33 ecl-amd kernel: [ 309.879769] IP: kfd_create_process+0x466/0x540 [amdgpu] Nov 19 15:59:33 ecl-amd kernel: [ 309.879808] PGD 800000024d7f6067 P4D 800000024d7f6067 PUD 254ff0067 PMD 0 Nov 19 15:59:33 ecl-amd kernel: [ 309.879863] Oops: 0002 [#1] SMP PTI Nov 19 15:59:33 ecl-amd kernel: [ 309.879889] Modules linked in: ccm nls_iso8859_1 arc4 8250_dw snd_soc_skl snd_soc_skl_ipc snd_hda_ext_core iwlmvm s nd_soc_sst_dsp snd_soc_sst_ipc mac80211 snd_soc_acpi rtsx_pci_ms memstick snd_soc_core snd_hda_codec_realtek snd_hda_codec_hdmi snd_hda_codec_generic snd_compress ac97_bus snd_pcm_dmaengine intel_rapl x86_pkg_temp_thermal idma64 intel_powerclamp virt_dma coretemp kvm_intel kvm irqbypass snd_hda_inte l snd_hda_codec snd_hda_core snd_hwdep btusb btrtl btbcm intel_wmi_thunderbolt wmi_bmof btintel snd_pcm iwlwifi snd_timer snd bluetooth soundcore cfg8 0211 ecdh_generic shpchp intel_pch_thermal mei_me mei intel_lpss_pci intel_lpss mac_hid acpi_pad sch_fq_codel ib_iser rdma_cm iw_cm ib_cm ib_core iscs i_tcp libiscsi_tcp libiscsi scsi_transport_iscsi ip_tables x_tables autofs4 btrfs zstd_compress Nov 19 15:59:33 ecl-amd kernel: [ 309.880481] raid10 raid456 async_raid6_recov async_memcpy async_pq async_xor async_tx xor raid6_pq libcrc32c raid1 raid0 multipath linear amdgpu(OE) amdttm(OE) crct10dif_pclmul crc32_pclmul ghash_clmulni_intel pcbc amdkcl(OE) amd_sched(OE) i2c_algo_bit amd_iommu_v 2 drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops aesni_intel e1000e aes_x86_64 crypto_simd glue_helper rtsx_pci_sdmmc cryptd drm ptp pps _core ahci rtsx_pci libahci sdhci_pci sdhci wmi video Nov 19 15:59:33 ecl-amd kernel: [ 309.880969] CPU: 0 PID: 1345 Comm: rocminfo Tainted: G OE 4.15.0-55-generic #60-Ubuntu Nov 19 15:59:33 ecl-amd kernel: [ 309.880992] Hardware name: Intel(R) Client Systems NUC8i3CYS/NUC8CYB, BIOS CYCNLi35.86A.0036.2018.0801.1506 08/01/2 018 Nov 19 15:59:33 ecl-amd kernel: [ 309.881075] RIP: 0010:kfd_create_process+0x466/0x540 [amdgpu] Nov 19 15:59:33 ecl-amd kernel: [ 309.881092] RSP: 0018:ffffacac01a6fb90 EFLAGS: 00010286 Nov 19 15:59:33 ecl-amd kernel: [ 309.881107] RAX: ffffa0ab4d6bdc00 RBX: fffffffffffffff4 RCX: 0000000000000000 Nov 19 15:59:33 ecl-amd kernel: [ 309.881126] RDX: 0000000000000040 RSI: 0000000000000000 RDI: ffffa0ab57003800 Nov 19 15:59:33 ecl-amd kernel: [ 309.881145] RBP: ffffacac01a6fbc8 R08: ffffa0ab5fc270a0 R09: ffffa0ab4d6bdc00 Nov 19 15:59:33 ecl-amd kernel: [ 309.881164] R10: ffffecf78935a000 R11: 666f72500064666b R12: ffffa0ab4e72e100 Nov 19 15:59:33 ecl-amd kernel: [ 309.881183] R13: ffffa0ab54890000 R14: ffffa0ab4e72e100 R15: ffffa0ab4d681c00 Nov 19 15:59:33 ecl-amd kernel: [ 309.881203] FS: 00007f130c4d5780(0000) GS:ffffa0ab5fc00000(0000) knlGS:0000000000000000 Nov 19 15:59:33 ecl-amd kernel: [ 309.881224] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 Nov 19 15:59:33 ecl-amd kernel: [ 309.881240] CR2: 00000000000001fc CR3: 000000024d2a8002 CR4: 0000000000760ef0 Nov 19 15:59:33 ecl-amd kernel: [ 309.881259] PKRU: 55555554 Nov 19 15:59:33 ecl-amd kernel: [ 309.881267] Call Trace: Nov 19 15:59:33 ecl-amd kernel: [ 309.881280] ? cdev_purge+0x70/0x70 Nov 19 15:59:33 ecl-amd kernel: [ 309.881329] kfd_open+0x3b/0xd0 [amdgpu] Nov 19 15:59:33 ecl-amd kernel: [ 309.881343] chrdev_open+0xc4/0x1b0 Nov 19 15:59:33 ecl-amd kernel: [ 309.881355] do_dentry_open+0x1c2/0x310 Nov 19 15:59:33 ecl-amd kernel: [ 309.881367] ? inode_permission+0x5b/0x160 Nov 19 15:59:33 ecl-amd kernel: [ 309.881380] ? cdev_put.part.3+0x20/0x20 Nov 19 15:59:33 ecl-amd kernel: [ 309.881393] vfs_open+0x4f/0x80 Nov 19 15:59:33 ecl-amd kernel: [ 309.881403] path_openat+0x66e/0x1770 Nov 19 15:59:33 ecl-amd kernel: [ 309.881416] ? filemap_map_pages+0x36c/0x390 Nov 19 15:59:33 ecl-amd kernel: [ 309.881430] do_filp_open+0x9b/0x110 Nov 19 15:59:33 ecl-amd kernel: [ 309.881442] ? check_object_size+0xaf/0x1b0 Nov 19 15:59:33 ecl-amd kernel: [ 309.881456] ? __alloc_fd+0x46/0x170 Nov 19 15:59:33 ecl-amd kernel: [ 309.881467] do_sys_open+0x1bb/0x2c0 Nov 19 15:59:33 ecl-amd kernel: [ 309.881479] ? do_sys_open+0x1bb/0x2c0 Nov 19 15:59:33 ecl-amd kernel: [ 309.881491] SyS_openat+0x14/0x20 Nov 19 15:59:33 ecl-amd kernel: [ 309.881503] do_syscall_64+0x73/0x130 Nov 19 15:59:33 ecl-amd kernel: [ 309.881517] entry_SYSCALL_64_after_hwframe+0x3d/0xa2 Nov 19 15:59:33 ecl-amd kernel: [ 309.881531] RIP: 0033:0x7f130b99dc8e Nov 19 15:59:33 ecl-amd kernel: [ 309.881542] RSP: 002b:00007ffc516b41a0 EFLAGS: 00000246 ORIG_RAX: 0000000000000101 Nov 19 15:59:33 ecl-amd kernel: [ 309.881562] RAX: ffffffffffffffda RBX: 00007f130b675900 RCX: 00007f130b99dc8e Nov 19 15:59:33 ecl-amd kernel: [ 309.881581] RDX: 0000000000080002 RSI: 00007f130b466bc8 RDI: 00000000ffffff9c Nov 19 15:59:33 ecl-amd kernel: [ 309.881622] RBP: 00007f130b6759a8 R08: 00007f130b675900 R09: 0000000000004d44 Nov 19 15:59:33 ecl-amd kernel: [ 309.881641] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000 Nov 19 15:59:33 ecl-amd kernel: [ 309.881659] R13: 00007ffc516b4400 R14: 0000000000000000 R15: 0000000000000000 Nov 19 15:59:33 ecl-amd kernel: [ 309.881677] Code: 91 7d 8a fb 48 83 3d 59 a5 22 00 00 0f 84 2d fc ff ff 48 8b 3d 5c 0f 0c fd ba 40 00 00 00 be c0 8 0 40 01 e8 3d 90 8a fb 48 85 c0 <48> 89 83 08 02 00 00 0f 84 b8 00 00 00 48 8b 53 60 48 c7 c1 7d Nov 19 15:59:33 ecl-amd kernel: [ 309.881782] RIP: kfd_create_process+0x466/0x540 [amdgpu] RSP: ffffacac01a6fb90 Nov 19 15:59:33 ecl-amd kernel: [ 309.881801] CR2: 00000000000001fc Nov 19 15:59:33 ecl-amd kernel: [ 309.881811] ---[ end trace 41383f71c50d817e ]---

luguang commented 4 years ago

$ dmesg | grep kfd [ 13.155802] kfd kfd: skipped device 1002:699f, PCI rejects atomics

It turns out that NUC may not support atomics. I will ask at NUC community.

nartmada commented 11 months ago

Hi @luguang , please check latest ROCm Documentation and ROCm 5.7.1 to see if your query has been resolved. If resolved, please close the ticket. Thanks.

nartmada commented 10 months ago

Original ticket is more than a year old and the person that opened the ticket has not responded to the latest request. If this is still an issue, please file a new ticket and we will investigate. Thanks!