) my Setup is : 

**Board**: Helios64 (RK3399) - **Kernel**: `6.18.44-current-rockchip64` - **Armbian**: 26.8.3 (upgraded from 25.11.2 via apt) - **Boot**: currently booting from SD card (rootfs on mmcblk0);

 

b) symptoms :

After upgrading to 26.8.3, my drives are no longer detected — but only on a "warm reboot": - ASM1061 PCIe-SATA controller (bays 1-2) completely missing from `lspci` - JMS561 USB-SATA bridges (bays 3-5) missing from `lsusb` - Other USB devices (VIA hubs, Realtek USB NIC) enumerate fine - No `sdX` block devices at all in `lsblk`

 

The only fix is a "cold boot": `poweroff`, unplug the power cable, wait 1-2 minutes, power back on. After a cold boot, all 4 drives (4x 5.5TB) come back perfectly. Every warm reboot reproduces the failure.

 

 c) dmesg output (warm reboot, drives missing)

[ 2.445661] rockchip-pcie f8000000.pcie: host bridge /pcie@f8000000 ranges: [ 2.475639] rockchip-pcie f8000000.pcie: MEM 0x00fa000000..0x00fbdfffff -> 0x00fa000000 [ 2.475650] rockchip-pcie f8000000.pcie: IO 0x00fbe00000..0x00fbefffff -> 0x00fbe00000 [ 2.476138] rockchip-pcie f8000000.pcie: no bus scan delay, default to 0 ms [ 3.182900] rockchip-pcie f8000000.pcie: PCIe link training gen1 timeout! [ 3.182983] rockchip-pcie f8000000.pcie: probe with driver rockchip-pcie failed with error -110

 

d)This looks very similar to previously reported issues: -

-

The difference in my case: the failure is 100% reproducible and strictly tied to **warm reboot vs cold boot**, which points to the PCIe PHY not being properly reset during a soft reboot with 6.18.44. 

 

=>  Is this a known regression in 6.18.x on Helios64? Is there a fix planned?

 

Thks a lot all 

Hello,

 

To my side the only way to back to a stable helios64 system is to freeze at kernel version 6.18.10 and use the specific DTB file patch.

This patch was include in armbian-config program. I was happy but…

 

Since the kernel version upper that 6.18.10… the helios64 nightmare come back 😞

 

Keep in touch