Armbian doesnt seem to see sata harddrives.

[DavidFajardo]

I am just posting so I can keep track of this thread.

[Werner]

3 hours ago, DavidFajardo said:

am just posting so I can keep track of this thread.

I suggest to use the "Follow" button instead.

[Maurycy]

Finally working with Rolling release.

 

[Tomogo]

Hi, I have still problems also with rolling release so I'm using one older kernel with vmlinuz-6.1.75-vendor-rk35xx.

[Maurycy]

Hi,

 

I was too optimistic, the problem persist with rolling release. I rearranged power supply for the drives so now they are not supplied from one line- it didn't solve the issue.

 

[Tomogo]

I had the same.

[Igor]

2 hours ago, Maurycy said:

the problem persist with rolling release.

 

Release is not a warranty that all bugs get fixed. That process takes several years and new bugs are keep getting introduced ... proper maitainances keeps this low.

Here https://github.com/armbian/linux-rockchip/commits/rk-6.1-rkr4.1/ you can monitor what is getting merged. Rolling update just ships this few hours after merge is made in automatic way.

[DavidFajardo]

Ok sir, I will keep it in mind next time.

[Tomogo]

I don't know if this can help, but here are differences between not working and working kernel config config-6.1.75-vendor-rk35xx:

 

config-6.1.75-vendor-rk35xx.diff.txt

[Igor]

1 hour ago, Tomogo said:

I don't know if this can help, but here are differences between not working and working kernel config config-6.1.75-vendor-rk35xx:

Sadly the problem is a lot more complex  Somewhere around PCI bus signaling / initialization / devices comm (tl;dr;)

[Tomogo]

OK, so if anybody want working kernel it can be downloaded on my web pages:

https://files.tomogo.cz/rock/linux.tar.gz

Edited December 12, 2024 by Tomogo

[Maurycy]

Thanks, I'll try it soon.

[Maurycy]

 

 @Tomogo: SATA works great with the kernel provided. I really hope that it will be accesible in normal way. Now switching kernel and installing ZFS is doable but painful.

 

My concusion: If someone want's OMV, I recommend DELL SFF Intel 6100T based unit. It is cheaper (used), uses less power (5W with the same amount of discs) and works. I had been using it for some time.

 

Thanks!

[Krishna0]

Hi,

    I am using the same board but I am planning to connect a RAID card which is LSI Broadcom SAS 9300-8i. When I connect the card through the NVME slot, the board is not getting recognized. Should I install any drivers for the card?

[Maurycy]

Afaik adding something to arm device is quite complicated as there is no bios like solution. Device first needs to be in device tree overlay to be seen by the kernel.

[buhtux]

On 12/26/2024 at 5:33 PM, Krishna0 said:

Hi,

    I am using the same board but I am planning to connect a RAID card which is LSI Broadcom SAS 9300-8i. When I connect the card through the NVME slot, the board is not getting recognized. Should I install any drivers for the card?

 

Hi, based on my experience you will need to flash the 9300-8i card into IT-Mode (Link). Please use x86 or amd64 system for the flashing process. IT mode works like HBA, it pass all connected drives directly to the system. Afterwards you can create RAID volumes with mdadm, zfs, btrfs.

Edited January 5 by buhtux

[Tomogo]

Hi,

has anybody info if the new official kerenl 6.1.84-vendor-rk35xx works fine with SATA disks?

Thanks...

[ypopovych]

I have the same problem on the latest rolling release kernel. Could be that this is not a PCI problem? I have a second controller in the m.2 e-key port with JMB582 and it works fine. Also I can see ASM1164 in the lspci output but without "ahci" driver loaded for it.

 

I can enable disks if I do:

echo "1" > /sys/bus/pci/devices/0001\:11\:00.0/remove
sleep 1
echo "1" > /sys/bus/pci/rescan

 

What can I do to help debug the problem?

Edited January 15 by ypopovych
added how to enable disks after boot

[eselarm]

I had to look it up: https://docs.radxa.com/img/rock5itx/rock5itx-system-block-diagram.webp as I was not sure if it was on-chip (RK35xx) or extra. The latter is the case. On a Rock3A I also had success with m.2 e-key port with JMB582, but not used now, instead I use on-chip SATA via overlay. That makes make think that maybe for the Rock5ITX the ASM1164 kernel code is not compiled as module but in-kernel and so maybe there are timing issues in initialization at boot-up. There should not be, but that is what I would look at first as this is PCIe and we know there have been issues with it. If you do not need vendor kernel specifically, you might try 6.12 current/mainline. Also see/check what U-Boot the board boots with.

 

[ypopovych]

I tried to install OS to the SATA SSD and connect it to the ASM1164 port to test. U-Boot detects SATA SSD, loads initrd, and then initrd fails to detect SSD and boot fails. It boots after 3-7 restarts and then works fine. Seems as U-Boot can initialize SATA controller properly and read data on every boot and kernel only one boot from five.

[blood]

On 1/15/2025 at 12:54 AM, Tomogo said:

has anybody info if the new official kerenl 6.1.84-vendor-rk35xx works fine with SATA disks?

 

From what I'm seeing with the latest rk35xx vendor kernel, SATA is mostly broken once again (the SATA controller PCIe device is not detected and thus no drives attached to it). It seems to be a timing issue as occasionally they have showed up though.

 

Also, zfs-dkms results in an actually broken system when used with this kernel; kernel oops on boot.

 

I of course don't blame Armbian for this, but the rk35xx vendor kernel branch seems to be a hot mess. Too bad it's effectively required to use the VPU...

[Werner]

10 hours ago, blood said:

but the rk35xx vendor kernel branch seems to be a hot mess.

It is. We made the best possible support from it. Each new sdk version brings fixes but also regressions.

 

Though you actially have to give credit to rockchip for their rk35xx sdk. We had way worse in the past, not necessarily from rk but others like aml or aw. In comparison its surprisingly good

 

Anyway it will take another few years until it is close to feature-complete in mainline. I think rk3399 still hasn't all its features mainlined but overall it is well matured now. You always have to keep in mind that some smart person have to sit down and write drivers for sometimes poorly, sometimes not at all, documented features of a soc. Either being paid to do so or being an enthusiast where latter are very very rare people and their efforts are often strongly under-appreciated.

[sdyspb]

I fixed my Rock 5 ITX by unsoldering R29 pull-down resistor. There was wrong HW configuration of ASM1164. Now all kernels work as well as vendor's one, including the last images from Armbian - Debian12 (Bookworm) + OMV. The more details in Radxa forum.

[ypopovych]

@sdyspb Wow! I will try to do it in the couple of days and will tell result

[sdyspb]

You really need a good soldering iron and tip because of big ground plane on bottom

PS It's funny to recommend soldering iron in a software chat to fix the kernel problem

[Tomogo]

Hi, I've tried it and it works great! Many thanks!

 

But my eyes almost didn't see it. I had to use my glasses and magnifying glass together 😉

[ypopovych]

I can be wrong, but seems as hardware mod can be avoided if armbian adds this fixes to the DTS and clk-gpio.c for our board:

 

https://lore.kernel.org/all/20240906082511.2963890-6-heiko@sntech.de/

 

Seems as board has proper clock for SRIS mode to work, but it should be enabled on boot. And because PCIe3.0 is split between two devices kernel starts it only for M2 slot, so if M2 slot is initialized before SATA controller it will work, if SATA controller is initialized before M2 it will not get clock signal and will hang. I checked current DTS in armbian and it doesn't have this clock logic.

 

When we remove a resistor it simply avoids this bug and SATA controller works in a different mode (which could be slower).

Edited Thursday at 10:32 AM by ypopovych
clk-gpio.c should be patched too

[Igor]

1 hour ago, ypopovych said:

I can be wrong, but seems as hardware mod can be avoided if armbian adds this fixes to the DTS for our board:

 

@prahal Can you look into this?

[ypopovych]

i checked latest mainline kernel sources and they have this patch merged and accepted. I think it was merged in version 6.13.

[Igor]

1 hour ago, ypopovych said:

I think it was merged in version 6.13.

We will be with stable on 6.12.y for some time. Perhaps add it here:
https://github.com/armbian/build/tree/main/patch/kernel/archive/rockchip64-6.12