Posts by Hauke

Hi CvH,

I've just seen that the new release is there, and that CrazyCat drivers are in there in the newest version. Thanks a lot for that! Just tested on Le Potato: It is certainly the best experience I yet had with Kodi on LePotato regarding the Hauppauge DualHD card, but it is still not good. When watching a second channel, I still get too many errors to call the experience enjoyable.

In the meanwhile I tested with Raspberry Pi and the 4.18 kernel, since Hauppauge claims native support for the dualHD stick starting with kernel 4.17, and I can confirm that the DVB-T2 reception is flawless with 4.18, even watching two channels simultaneaously. Raspberry, even 3B+, ist not powerful enough to watch even a single channel in HEVC (German DVB-T2 uses it), but using TVheadend on Raspberry and a Windows machine to watch the channels works totally fine.

So, I suppose I need to stay patient until LibreELEC on Le Potato is on mainline kernel... According to a tweet by the Le Potato team, they already did a showcase with LibreELEC and mainline recently, so it seems to be close to reality anyhow...

I really appreciate your efforts! Thanks again!

Hauke

OK, understood. Thanks for taking care with the next release!

Dear all,

is there a comprehensive guide how to update the CrazyCat TBS drivers? I am using

• @kszaq's image for amlogic (on Le Potato)
• a WinTV Hauppauge dualHD DVB-T2 USB stick,
• The CrazyCat TBS drivers
• TVheadend 4.3

and both tuners are properly recognized, but Tuner #1 works nearly fine (once in a while a continuity error) while Tuner #2 only gives jerky video and loads of continuity errors. I seem to understand that the drivers have improved, but for all Googling I seem to be to stupid to get a up-to-date package. Can you help me?

Side remark: Yes, I am aware of CoreELEC and of the experimental support of Le Potato in the official LibreELEC build, but CoreELEC has very old CrazyCat drivers (by design/on purpose) and only TVheadend 4.2, and the official does not even recognize my TV card and also only has TVheadend 4.2. For DVB-T2 I need TVheadend 4.3, which luckily is in kszaq's image.

Thanks for help!

Hauke

Quote from Hauke

I'm even no longer sure that it is the amlogic network driver after all

I'm now rather sure that it is not the network driver. I kicked the ethernet from the device tree, it is no longer recognized at boot, and no eth0 shows up at all. So to my understanding the code for it should be completely dormant. Still: Crashes

I've got the feeling that I have to wait until LePotato mainline kernel includes hardware accel and someone ports LibreELEC to mainline then...

Or is there any developer here willing to dubug? Would assist by providing whatever logs required, UART output, whatever.

adamg's latest devel build reports "not compatible with LePotato hardware" - cannot tell if the bug's there.

I have been working quite a bit on the problem of Le Potato crashing on huge file copy/write operations (see my earlier posts). What I did:

• I worked my way into compiling LibreELEC - used the official LebreELEC repo and
  PROJECT=Amlogic DEVICE=LePotato ARCH=arm make image
• I identified the code parts that correspond to the Armbian patches that the Armbian team claims solved the stability issues (all network driver code)
• I managed to adopt the code to go into the LibreELEC kernel (which wasn't too difficult after all)

To no avail The crashes still happen. I tried a bit around with a few variations of the code, but I'm afraid my understanding of the depth of network drivers is not sufficient to do more but fumble around. So I've to give up here - my time budget does not allow to learn the intricacies of linux network drivers.

I also did a DEBUG compile, and read out crash information from the debug UART, but this information is totally inconclusive, the crash practically never happening at the same place - it looks like randomly parts of the OS go down - very confusing and not helpful.

I'm even no longer sure that it is the amlogic network driver after all: I remebered that in some drawer I had lying around an USB2 Gigabit network adapter (smsc75xx based), which was immediatly recognised and works. But also with this dongle the crash happens, even when I disable the eth0 interface by ifconfig eth0 down. However, my shell script did not involve network directly also, so the picture is inconclusive. And Armbian team confirms that the mentioned two patches (and only these, no other) did the trick for them, and I can confirm that with Armbian the bug is gone.

Long story short: Can someone tell me how I completely disable the network interface, beyond ifconfig eth0 down? It seems it is not a module, so just blacklisting it does not work. When I really "blocked" the driver, I'll again test with the USB dongle.

Thanks!

P.S.: Problem persists in 8.90.5 - but that's expected...

P.P.S.: If someone wants to follow up: The patches of Armbian go into stmmac_main.c and phy_device.c

P.P.P.S.: The patches of Armbian address issues with EEE and Autonegotiation functions. Disabling EEE and Auto neg. on the switchport did not solve the issue either, but showed at least one bug in the network driver: The interface went to half duplex, while the switch offered only full duplex. ethtool was needed to change to full duplex. Still, crash happens.

Short update: With DEBUG=1 as mentioned it compiles through, however, VALGRIND=yes is not supported for LePotato/AMlogic it seems. Leaving VALGRIND option out and it's OK.

OK, inserted into

/LibreELEC.tv/build.LibreELEC-LePotato.arm-9.0-devel/u-boot-a43076c/Makefile

at top the line

DEBUG=1

So now it eval's as boolean if needed...hope that does not break other DEBUG flags. At least, it currently continues compiling.

The offending file seems to be the

/LibreELEC.tv/build.LibreELEC-LePotato.arm-9.0-devel/u-boot-a43076c/arch/arm/cpu/armv8/gxl/firmware/acs/Makefile

but somehow I could not get this debugged - my understanding of Makefiles is not good enough.

@whoever maintains this repo: Perhaps have a look at some point...

Dear all,

I try to complie LibreELEC with the following setup:

PROJECT=Amlogic DEVICE=LePotato ARCH=arm DEBUG=yes VALGRIND=yes make image

After several hours of compilation, this happens:

OBJCOPY u-boot.hex
Makefile:156: *** DEBUG must be boolean.  Stop.
make[2]: *** [/home/lux/official/LibreELEC.tv/build.LibreELEC-LePotato.arm-9.0-devel/u-boot-a43076c/Makefile:918: acs.bin] Error 2
make[1]: *** [Makefile:147: sub-make] Error 2
make[1]: Leaving directory '/home/lux/official/LibreELEC.tv/build.LibreELEC-LePotato.arm-9.0-devel/u-boot-a43076c'
Makefile:12: recipe for target 'image' failed
make: *** [image] Error 2

According to Compile [LibreELEC.wiki] DEBUG=yes is right, but seems to be a problem. Any help?

Btw.: Compiling without DEBUG/VALGRIND went smoothly beforehand. Same code base.

Thanks!

Hauke

EDIT: Stupid me - there is a path to the Makefile - will have a look myself, but it's gone right now, since I started over with make clean... Argh, think before you write/delete... Long day...

Quote from adamg

[...]new devel version for you to try here.[...]

Not that I expected something different, but the network-crash-on-large-file-writes-bug is in this version also. And I just realized I can't ignore the bug by just not writing large files: It occurs even when I do a longer recording from the DVB-T2 card attached to USB... Any chance that the problem will be addressed? While I understand that the patches I quoted are for mainline kernel and thus not directly applicable, they look rather "portable" to me, but of course I've close to no idea about kernel develeopment...

Btw.: I tried kszaq 's latest Le Potato version I could find (LibreELEC-S905.arm-8.2-8.2.3.1-LePotato), and it still suffers from the problem.

Thanks for the information. Obviously I did not test with the latest build by you, since I had the problem also with your image. Good to hear that this is fixed.

Quote from Hauke

I've narrowed it down finally: It's the network that crashes. [...]

I had some exchange with the very helpful Armbian team - they had the same issue and it is solved. Here are the mainline kernel patches that adressed the issue:

[net] net: stmmac: enable EEE in MII, GMII or RGMII only - Patchwork

[net,v3] net: phy: meson-gxl: detect LPA corruption - Patchwork

kszaq & GDPR-2: Is it possible to somehow incorporate this into the LibreELEC kernel?

Quote from Hauke

[...]I've a reproducible stability issue with both the 8.90.4 adamg image as well with kszaq's image on Le Potato: Copying a huge amount of data crashes the device. [...]

I've narrowed it down finally: It's the network that crashes. When I run a huge file operation and ssh session disconnects, the file operation contniues and is successfull in the end. Tearing down and restarting the network brings the box up'n'running again... Btw. - did this not with LibreELEC, but with Armbian, which has teh same problem.

Quote from meadrocks

I'd try scp DVD.iso MyHost:/dev/null, [...]

Good idea - tried, and no crash! So it is media access...

Could someone else with a Le Potato try something similar? If my box is really a lemon, I'd need to return it... Wouldn't like to.

EDIT: I decided to look for CPU temp and opened the System information from Kodi on the TV. Started scp (this time pscp istead of WinSCP), and it crashed:

DVD.iso               | 852260 kB | 4984.0 kB/s | ETA: 00:21:59 |  11%Fatal: Received unexpected end-of-file from server

Before it, you could see the upload speed constantly going down. Temperature went to moderate 53°C (The original heatsink is on the device). Now the funny thing is: The status display was still alive, Temperature started dropping and CPU load was displayed! So no complete crash. Kodi Web-Interface worked partially, but all other services, like SSH, Samba, tvheadend, were not accessible, so I could not extract any logs. While trying, the box died completely, eboot necessary.

Any idea?

And still would be curious if I'm the only one having this problem...

EDIT 2: To be really sure, I measured voltage: It stays at a very comfortable 5.38 V while copying the file...

Hi all,

I've a reproducible stability issue with both the 8.90.4 adamg image as well with kszaq's image on Le Potato: Copying a huge amount of data crashes the device. I tried - and all crash:

cp /var/media/USB1/DVD.iso /var/media/USB2

cp /var/media/USB1/DVD.iso /storage

cp /var/media/USB1/DVD.iso /dev/null

From a Windows PC via network:

DVD.iso via Samba --> \\MyIP\Videos

scp DVD.iso MyHost:/storage

scp DVD.iso MyHost:/var/media/USB

For those methods that show a progress bar, the copy job crashes after a few 100 MB, and it's not always at the same amount of data. DVD.iso has about 7GB. Since even copy to /dev/null fails, I'd say it's something with the general load on some bus or so, not just the write operation. It can't be just the read process, since network also fails. Network itself is robust as far as I can tell by putting some load to it. Tried 3 PSUs, all working well even with demanding Raspberry Pi tasks, where Voltage never drops below 5V.

No idea how to collect troubleshooting info for this - any hints?

Hauke

For me it works rock solid as of now - so thanks for the great work! I'll see how stable it is long term with doing recordings etc.

Quick update: I finally figured out that picking the "hauppauge" drivers is not the best idea - I picked now "DVB drivers for TBS (CrazyCat)", and both tuners are recognized now - happy me! Thought "TBS" is for "TurboSight" DVB-cards - a little confusing to me.

Just running a scan, lets see how good it works.

EDIT: Scan's through, watching TV works nicely and stable, both tuners can be used - perfect! Thanks to crazycat for this!

OK, figured it out finally: The "LibreELEC Module Drivers" it is, and there crazycat is "TBS" - thought that meant "TirboSight" DVB cards... That's not really intuitive or I'm just too stupid. However, now the card is recognized, even both tuners *yay*! Just running a scan, and then will see how stable it is.

Thanks!

/me again: Today I tried with adamg's image and with kszaq's image to copy a DVD iso file of 7.something GB size via network to the Le Potato. Both with Samba as well as with SCP the copy job crashes after ~500-700 MB, not always at the same point. I tried SD card and USB storage as target, both the same. Is there any log worth extracting for troubleshooting?