Report about Amlogic S905X4 HK1 RBOX X4

Help & Support Hardware
465

Yes, much better now. Speed ~780Mbit.
But once per 5-10 min:
[ 710.199172@0] meson8b-dwmac fdc00000.ethernet eth0: Link is Down
[ 710.199346@0] [dhd] CFG80211-ERROR) wl_cfg80211_netdev_notifier_call : wdev null. Do nothing
[ 713.271959@0] meson8b-dwmac fdc00000.ethernet eth0: Link is Up - 1Gbps/Full - flow control off
[ 713.272001@0] [dhd] CFG80211-ERROR) wl_cfg80211_netdev_notifier_call : wdev null. Do nothing

will you make changes for ce19 ?

466

Dump your Android dtb and shared it please, info is in Wiki.

Looks like there exist a few HK1 variations:
internal PHY: 100MBit
external PHY: RTL8211f
external PHY: JL2xx1
external PHY: JL2???

Itā€™s more worse than the other Xā€¦ devices on market.

Itā€™s definitely not recommended when you want to use CoreELEC to buy such ā€œuselessā€ hardware!!

467

Here is the one I posted earlier Report about Amlogic S905X4 HK1 RBOX X4 - #76 by emveepee I am running a newer Android build now and will dump it later.

Note that the device and driver itself it ā€œuselessā€ since in Android I can watch live TV for hours, but I agree the vendor support that CE needs is certainly lacking.

468

I need it raw, not decompiled

I mean useless as we spend and will spend days for getting this item running.
Other, supported hardware by CE just works our of the box and bring quality in the houseā€¦

I donā€™t know how long I am anymore interested at all to get this fixed. When not it will stay at 100mbit or WiFi onlyā€¦

469

Ok sorry here you go. dtb.img.gz (18.7 KB)

When the new kernel source is available I would also try to get it going.

470

I do agree that the HK1 box sucks. Get the X96 x4.

471

Your hardware is sc2_s905x4_ah212, the hardware where itā€™s working is sc2_s905x4_ah218ā€¦

So what should I say now? Search for a good place in the round storage below the table?

Ask your vendor to provide kernel sources, or at least the changes he have done in kernel.
You see now the big difference between supported or not supported hardware by CoreELEC?

You can hook up the device available by SSH so we can remote attach to it and debug it. I see no other option.

472

Since I was able to force 100Mb in the old kernel with this patch LubanCatWrt/602-fix-kernel-stmmac-drivers-for-jl2101.patch at master Ā· LubanCat/LubanCatWrt Ā· GitHub, could you make a build with it or post the kernel source and I can try various things.

473

These changes arenā€™t included. I can make you a debug build including it tomorrow.

474

Thanks I will be happy to test that.

475

@emveepee please try with this image:

And make a dmesg | paste if something changed

476

That works thanks, http://ix.io/4bIR it still doesnā€™t find the MAC but it makes the device much better.

477

It have now ā€œfullā€ speed and it get an IP?
eth0: Link is Up - 1Gbps/Full

I will make another debug image where I will need dmesg | paste again please.

478

@emveepee please update with this tar:

You can just do it by SSH into your device:

cd .update
megadl https://mega.nz...
sync
reboot

Or just place the tar file in update folder and reboot.

479

Yes the previous version seemed to be 1000Mb as does this new one http://ix.io/4bJe

CoreELEChkk:~ # ./iperf -c 172.16.3.105
------------------------------------------------------------
Client connecting to 172.16.3.105, TCP port 5001
TCP window size: 85.0 KByte (default)
------------------------------------------------------------
[  3] local 172.16.3.180 port 54682 connected with 172.16.3.105 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3] 0.0000-10.0094 sec  1.03 GBytes   883 Mbits/sec

Martin

480

Ok, the problem is in stmmac_mdio_register the PHY can not be detected.
So here is another image where this code is not included, only the sleep:

Please update and after update pull DC plug for a few minutes (that the PHY chip registers get reset).
Then try again if it works please. If it does not work again then we will need a user option in CE settings to be able to enable this PHY if assembled.

481

Yes that still works http://ix.io/4bJp and ifconfig eth0 up/down still works too which didnā€™t work on 4.9 kernel with sleep only

482

On 4.9 there is still something additional missing. I will take a look tomorrow about it.

When itā€™s only the sleep needed itā€™s doable as it can be enclosed by matching phy_id.

483

So here a new Amlogic-ne test version. It should have now a ā€œusableā€ MAC address for this JL2xx1 eth PHY:
CoreELEC-Amlogic-ne.arm-20.0-Nexus_devel_20220929082110-Generic.img.gz

484

I still get the same ā€œbadā€ MAC http://ix.io/4bNI