Hello again,
I am struggling to install armbian on my H96Max (blue PCB RK3318_V1.4, 4G/64G).
The Armbian_22.11.0-trunk_Rk3318-box_bullseye_edge_5.19.15_minimal.img image works OK (at least for the display).
All the new images fail to initialise the display (a 4K HDMI monitor), either using *_minimal.img or *_gnome_desktop.img
I managed to install the latest (Armbian_community_25.8.0-trunk.8_Rk3318-box_noble_current_6.12.28_gnome_desktop.img) image using the serial console. Kernel is 6.12.28
I can now connect using ssh on ethernet (the WIFI does not work yet...)
I get in syslog and dmesg a LOT of the following errors:
[ 2683.281403] ------------[ cut here ]------------
[ 2683.281441] [CRTC:38:crtc-0] vblank wait timed out
[ 2683.281570] WARNING: CPU: 1 PID: 50 at drivers/gpu/drm/drm_atomic_helper.c:1682 drm_atomic_helper_wait_for_vblanks.part.0+0x238/0x25c
[ 2683.281629] Modules linked in: cfg80211 rfcomm algif_hash algif_skcipher af_alg bnep sunrpc snd_soc_hdmi_codec hci_uart btqca hantro_vpu btrtl btintel crct10dif_ce rockchip_vdec(C) v4l2_jpeg v4l2_vp9 btbcm rockchip_rga v4l2_h264 bluetooth videobuf2_dma_contig snd_soc_spdif_tx gpio_ir_recv videobuf2_dma_sg rc_core v4l2_mem2mem rfkill videobuf2_memops videobuf2_v4l2 dw_hdmi_i2s_audio videodev dw_hdmi_cec lima snd_soc_simple_card videobuf2_common snd_soc_rockchip_spdif snd_soc_rockchip_i2s snd_soc_rk3328 snd_soc_simple_card_utils gpu_sched drm_shmem_helper mc snd_soc_core rk_crypto snd_compress rng_core snd_pcm_dmaengine snd_pcm snd_timer snd soundcore cpufreq_dt joydev zram zsmalloc binfmt_misc sch_fq_codel dm_mod autofs4 dwmac_rk stmmac_platform stmmac pcs_xpcs nvmem_rockchip_efuse rockchipdrm dw_hdmi_qp gpio_syscon dw_hdmi cec dw_mipi_dsi analogix_dp drm_display_helper adc_keys
[ 2683.282340] CPU: 1 UID: 0 PID: 50 Comm: kworker/u16:2 Tainted: G B WC 6.12.28-current-rockchip64 #1
[ 2683.282375] Tainted: [B]=BAD_PAGE, [W]=WARN, [C]=CRAP
[ 2683.282389] Hardware name: Rockchip RK3318 BOX (DT)
[ 2683.282409] Workqueue: events_unbound commit_work
[ 2683.282446] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[ 2683.282471] pc : drm_atomic_helper_wait_for_vblanks.part.0+0x238/0x25c
[ 2683.282499] lr : drm_atomic_helper_wait_for_vblanks.part.0+0x238/0x25c
[ 2683.282531] sp : ffff8000809dbcb0
[ 2683.282542] x29: ffff8000809dbce0 x28: 00000000000007d5 x27: 0000000000000000
[ 2683.282580] x26: 0000000000000001 x25: 0000000000000000 x24: ffff6d33cdf68000
[ 2683.282616] x23: 0000000000000001 x22: 0000000000000000 x21: ffff6d33cc6f6280
[ 2683.282654] x20: ffff6d33cdf70080 x19: 0000000000000000 x18: ffffffffffffffff
[ 2683.282691] x17: 000000040044ffff x16: 00500072b5503510 x15: ffff8000809db860
[ 2683.282727] x14: 00000000ffffffea x13: ffff8000809db9f8 x12: ffffd4f64aa4ac10
[ 2683.282768] x11: 0000000000032030 x10: 0000000000031fe0 x9 : 0000000000024330
[ 2683.282805] x8 : c0000000fffff822 x7 : ffffd4f64a99aa18 x6 : 0000000000084bb0
[ 2683.282842] x5 : ffff6d34be7424c8 x4 : 0000000000000000 x3 : 0000000000000027
[ 2683.282878] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff6d33c4c4c600
[ 2683.282918] Call trace:
[ 2683.282932] drm_atomic_helper_wait_for_vblanks.part.0+0x238/0x25c
[ 2683.282962] drm_atomic_helper_commit_tail_rpm+0x84/0xa0
[ 2683.282992] commit_tail+0xa4/0x18c
[ 2683.283019] commit_work+0x14/0x20
[ 2683.283043] process_one_work+0x16c/0x3e4
[ 2683.283073] worker_thread+0x2d0/0x3d8
[ 2683.283097] kthread+0x10c/0x110
[ 2683.283123] ret_from_fork+0x10/0x20
[ 2683.283151] ---[ end trace 0000000000000000 ]---
This seems to come from the accelerated graphic driver. The corresponding function can be seen there: drm_atomic_helper.c
This function waits for receiving a vertical blanking pulse that never comes, so the warning at line 1679 is triggered.
I don't intend to do hard-core graphics so I don't care about graphic performance. In fact this lovely box will be used for hosting a headless server.
Is there a way to use a legacy or un-accelerated graphic driver?
If not, how could I remove entirely the graphic subsystem and use only character mode?
Thanks!
Michel
EDIT: In fact the first warning occurs when switching to the colour frame buffer device:
...
rk_gmac-dwmac ff550000.ethernet: Enable RX Mitigation via HW Watchdog Timer
May 29 19:00:08 h96max kernel: rk_gmac-dwmac ff550000.ethernet: device MAC address de:10:75:60:c0:cb
May 29 19:00:08 h96max kernel: Console: switching to colour frame buffer device 160x45
May 29 19:00:08 h96max kernel: ------------[ cut here ]------------
May 29 19:00:08 h96max kernel: [CRTC:38:crtc-0] vblank wait timed out
May 29 19:00:08 h96max kernel: WARNING: CPU: 0 PID: 74 at drivers/gpu/drm/drm_atomic_helper.c:1682 drm_atomic_helper_wait_for_vblanks.part.0+0x238/0x25c
May 29 19:00:08 h96max kernel: Modules linked in: dwmac_rk stmmac_platform stmmac pcs_xpcs nvmem_rockchip_efuse gpio_syscon rockchipdrm dw_hdmi_qp dw_hdmi cec dw_mipi_dsi analogix_dp drm_display_helper adc_keys
May 29 19:00:08 h96max kernel: CPU: 0 UID: 0 PID: 74 Comm: kworker/u16:5 Tainted: G B 6.12.28-current-rockchip64 #1
May 29 19:00:08 h96max kernel: Tainted: [B]=BAD_PAGE
...
So I suspect that not switching to a FB could solve my problem