WT Play 2: NAND and SD - alpha 10: network disconnecting during playback (Solved)

Quote from Raybuntu

Ask amlogic. The kernel is supplied by Amlogic and they screwed up big time many things. You can go ahead and try to backport a recent version. Just make sure to use git and keep the git history. cherry-pick might help.

Good luck.

Thanks, i'll have a go, and yeah that kernel looks a bit "patchy" i'll try and update parts.

btw for others:
I'm still compiling test builds for the WeTek Play 2, if people like to try (dont blame me if stuff breaks)
I will rename a file if its patched by me, if not, they are straight compiles from github master.

LibreELEC WeTek Play 2 test builds

GDPR-1

So i'm trying to compile with the updated driver in the kernel, however it isnt compiling...

only change I made was change package.mk to point to my github instead of the libreelec one for linux-amlogic
PKG_VERSION="e2670ec"
PKG_URL="https://github.com/jester-xbmc/linux-amlogic/archive/$PKG_VERSION.tar.gz"
PKG_SOURCE_DIR="$PKG_NAME-amlogic-$PKG_VERSION*"
PKG_PATCH_DIRS="amlogic-3.14"

however i'm getting:

mode of '/home/x/http://LibreELEC.tv/build.LibreELEC-WeTek_Play_2.aarch64-8.0-devel/toolchain/bin/lex' changed from 0664 (rw-rw-r--) to 0755 (rwxr-xr-x)
/home/x/http://LibreELEC.tv/packages/linux/package.mk: line 18: $'\r': command not found
Makefile:12: recipe for target 'image' failed
make: *** [image] Error 127

any clue ?

Quote from Raybuntu

Jester You did not attach the file but looks like you have some nonsense inside. Looks like you edited that file in an archaic windows editor because \r is a special character for carriage return. Looks like you have that in one of the variables.

I must be drunk, but no, I used vi

Ill have another look, thx

Quote from Raybuntu

You can use different modes in vi. Also it can happen if you copy paste something?

To be sure I just did a git reset and edited the package.mk again, it now downloaded from my repo, so lets see what happens.
build system is an old NUC, so have to wait 4 hours...I need to upgrade, lol

thx

well, that driver upgrade is proving to be a pain, fixed a couple of errors, but running into more....

drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c: In function 'stmmac_axi_setup':
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:115:23: error: dereferencing pointer to incomplete type 'struct stmmac_axi'
axi = kzalloc(sizeof(*axi), GFP_KERNEL);
^~~~
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:133:61: error: 'AXI_BLEN' undeclared (first use in this function)
of_property_read_u32_array(np, "snps,blen", axi->axi_blen, AXI_BLEN);
^~~~~~~~
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:133:61: note: each undeclared identifier is reported only once for each function it appears in
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c: In function 'stmmac_dt_phy':
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:170:6: error: 'struct plat_stmmacenet_data' has no member named 'phy_node'; did you mean 'phy_addr'?
plat->phy_node = of_parse_phandle(np, "phy-handle", 0);

I'll play around a bit more, but I dont want to spend too much time re-writing a driver to make it compile....
[hr]
P.S.

while waiting on compiles, I tried Raybuntu's 32bit build, no issues so far with the network cutting out (also looks like the mem leak is gone...and youtube live streams work)

if people want to try:
LibreELEC-WeTek_Play_2.arm-8.0-rb-beta1i.img.gz
(just upload it to your .update folder, and add a file called .nocompat to skip the warning as ur going from 64 to 32 bit)
[hr]

update:

Spoke too soon, same issue (network disconnect) happened on the raybuntu built....
[hr]

In the mean time after some digging:
So this eth issue seems to be a known kernel/driver problem, the guys over at the ODROID C2 forum have mentioned this too (I believe the WT Play2 and C2 both use S905) it seems S905 now has support in the mainline linux kernel (4.10)

Dev's reading this, the eth issue (at least for the C2) can be fixed by disabling eee-1000t

ODROID Forum • View topic - Odroid C2 mainline kernel support
stmmac/RTL8211F/Meson GXBB: TX throughput problems

indeed eee is enabled:
ethtool --show-eee eth0
EEE Settings for eth0:
EEE status: enabled - active
Tx LPI: disabled
Supported EEE link modes: 100baseT/Full
1000baseT/Full
Advertised EEE link modes: 100baseT/Full
1000baseT/Full
Link partner advertised EEE link modes: 100baseT/Full
1000baseT/Full

but disabling it doesnt work:
ethtool --set-eee eth0 eee off
ethtool --show-eee eth0
EEE Settings for eth0:
EEE status: enabled - active

so its hardcoded in the driver im guessing.

might patch the driver to always disable eee, but not sure if an easy fix, looks like its in here:
linux-amlogic/stmmac_main.c at amlogic-3.14.y · LibreELEC/linux-amlogic · GitHub

Quote from Raybuntu

Thanks for testing my build, I'm interested in that issue now ;). In fact thats true the memleak seems to gone in my builds.

Edit: I'll read up and throw a patch in my builds for you to test. I don't have that issue so I need testers.

Thanks, i'll test whatever you throw at me.

I was already looking into doing a patch, but im at work now so can't do much.

if you set eee_active = 0 on line 319 of linux-amlogic/stmmac_main.c at amlogic-3.14.y · LibreELEC/linux-amlogic · GitHub that might work, then again, I didnt look proplery yet.

Thanks !

I'm checking it remotely from work, lets see.