Posts by mtk1

Quote from escalade

mtk1

Looks like Sway triggers a bug in the i915 kernel driver. Doesn't happen here on the latest image, give it a try.

slacker666

I have no idea, so go with generic.

A slight improvement I would say. But, eventually FF still hangs and eventually crashes.

This problem was gone (for me) with the -rc6 kernel. More patches went in before the final release.

So, maybe it's a kernel issue ?

Or, did you configure FF differently at the time of the -rc6-based builds (10...14 days ago) ?

dmesg:

[ 1979.399923] Asynchronous wait on fence i915:sway[1357]:48b0 timed out (hint:intel_atomic_commit_ready+0x0/0x54)

[ 1983.473520] i915 0000:00:02.0: GPU HANG: ecode 8:1:0x85dffffb, in sway [1357], stopped heartbeat on rcs0

[ 1983.473524] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.

[ 1983.473525] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel

[ 1983.473526] drm/i915 developers can then reassign to the right component if it's not a kernel issue.

[ 1983.473527] The GPU crash dump is required to analyze GPU hangs, so please always attach it.

[ 1983.473528] GPU crash dump saved to /sys/class/drm/card0/error

[ 1983.576026] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 2078.791318] perf: interrupt took too long (2520 > 2500), lowering kernel.perf_event_max_sample_rate to 79200

[ 2287.443598] Asynchronous wait on fence i915:sway[1357]:875a timed out (hint:intel_atomic_commit_ready+0x0/0x54)

[ 2292.665644] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 2303.443252] Asynchronous wait on fence i915:sway[1357]:8764 timed out (hint:intel_atomic_commit_ready+0x0/0x54)

[ 2307.599778] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 2343.643548] docker0: port 1(veth7fa588b) entered disabled state

[ 2343.643827] veth7ec5664: renamed from eth0

[ 2343.708970] docker0: port 1(veth7fa588b) entered disabled state

[ 2343.723557] device veth7fa588b left promiscuous mode

[ 2343.723565] docker0: port 1(veth7fa588b) entered disabled state

[ 2346.535583] Asynchronous wait on fence i915:sway[1357]:8cf2 timed out (hint:intel_atomic_commit_ready+0x0/0x54)

[ 2350.478511] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 2366.751954] docker0: port 1(veth687e866) entered blocking state

[ 2366.751958] docker0: port 1(veth687e866) entered disabled state

[ 2366.752741] device veth687e866 entered promiscuous mode

[ 2367.249717] eth0: renamed from veth4daef5d

[ 2367.263889] IPv6: ADDRCONF(NETDEV_CHANGE): veth687e866: link becomes ready

[ 2367.264018] docker0: port 1(veth687e866) entered blocking state

[ 2367.264025] docker0: port 1(veth687e866) entered forwarding state

[ 2437.514826] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 2483.594441] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 2507.218543] perf: interrupt took too long (3170 > 3150), lowering kernel.perf_event_max_sample_rate to 63000

[ 2591.515987] r8169 0000:03:00.0: invalid short VPD tag 00 at offset 1

installed the latest generic build (from the 29th). Installation from scratch worked. Kodi works fine.

Firefox starts up, but as soon as I play a video from youtube, it crashed within 10 to 30 seconds.

Sometime the display freezes for seconds. Same stuff I reported a few days ago.

Those problem seemed to have been fixed about a week ago.

Logfiles attached. Please let me know, if I can provide any further info.

Quote from Liquidmercury

tried installing the latest generic build it appears to be uploaded this morning but it gives a kernel panic not syncing

same here.

Quote from escalade

mtk1

Can't reproduce here, tested update and install on VMware fusion no problems. Also updated my NUC with the same build.

RomMon

Fixed in the next release, the problem was caused by libtiff compiled with zstd support.

Hm, I reflashed the USB stick and a fresh install completed ok.

Firefox starts up ok. But, now I do have those 'freezing' effects again. The whole screen freezes for seconds incl. the cursor.
After a few of those 'freezing' events, Firefox crashes and Kodi comes back up.

The kernel logs shows this again:

[ 476.365777] Asynchronous wait on fence i915:sway[991]:974 timed out (hint:intel_atomic_commit_ready+0x0/0x54)

[ 480.433214] i915 0000:00:02.0: GPU HANG: ecode 8:1:0x85dffffb, in sway [991], stopped heartbeat on rcs0

[ 480.433218] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.

[ 480.433219] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel

[ 480.433220] drm/i915 developers can then reassign to the right component if it's not a kernel issue.

[ 480.433220] The GPU crash dump is required to analyze GPU hangs, so please always attach it.

[ 480.433221] GPU crash dump saved to /sys/class/drm/card0/error

This seemed fixed in previous builds. Anything you changed here ?

Quote from escalade

jaykass

Hm, I fixed that issue but looks like maybe I didn't upload the new image. New RPi4 image with the latest and greatest on the way in about 15 min

EDIT: New image ready, let me know how it works

mtk1

Found the firefox issue, can you try the generic image under "test"? If you look at /usr/bin/firefox.sh it tries to launch "firefox-developeredition", it should just be "firefox". I used the Arch dev package a while but now FF 72 is released and it works great on Wayland from my minimum of testing.

* Update via Update folder failed (stuck at RetroElec Bootup Screen)

* Install from USB stick (installer) failed with mount_flash not being able to mount UUID=xxxxx

Any info you need from the debug console ?

Quote from escalade

Provide the logfile as requested on page 1 and mentioned a million times lately.

sorry. I rebooted the box. Started Firefox. Same results: sway terminates and Kodi is started. Logs attached.

I'm on the latest generic build (from the 23rd). Kodi and ES startup fine, but Firefox does not. I see a blank 'sway' background and then (I assume when Firefox should populate the screen) the screen goes blank and Kodi comes up again.

I can not find any logfile that mentions Firefox. Is there any particular log file that you need here ?

not as far as I can see. libreelec/x86/generic is empty.

I've extracted the img from the *ova (virtual). I assume it's the same image ?

there is no generic build in the gdrive folder...

Quote from escalade

New image uploaded

Hi, any chance you'll be posting a non-skylake build ? i.e. silvermont ? I have a Braswell CPU. The skylake kernel does not boot

Quick update: The latest x86 build (presumably the 5.5-rc6 kernel) seems to fix the DRM issues I had seen when using Firefox. I tested it with media-heavy websites and video streaming. No issues, so far.

Quote from elecuser

Great LibreELEC spin! The combination with ES/Libretro is a winner.

My NUC8 is on your latest release. There is this issue when I power down via the Kodi menu (power off) or with the TV remote (HDMI-CEC) the box goes to sleep. In the LibreELEC settings "Shutdown function" is set to "Shutdown". The issue arises when the setup is switched back on but there is no output on the display (the TV stays black). I can hear the sounds of Kodi and when I ssh into the box all looks fine. To resolve the issue I have to hard power down the whole setup manually.

I would prefer a real "power off"... the system boots of an SSD and is plenty fast -- any ideas?

This seems related to my issue when switching between Kodi and RetroArch. I assume, that Kodi is 'shut down' in this case which might trigger some action via CEC. I thought I had disabled any CEC triggers towards my TV or receiver. Let me check.

Update: crackling sound in RA seems to depend on ROM or maybe CPU load.

oh. and once kodi is in this 'no display' state, restarting it (via kill) does not help. kodi comes back up (gui sounds are there and I can see it burn CPU cycles in 'top') but there's still no display. Is the display somehow 'suspended' ?

Installed the new image - fresh install, no update.

* Kodi works out of the box incl. PassThrough. I needed to change the PassThrough device to HMDI (from SPDIF) to enable 7.1 PassThrough

* I can now select ES from within Kodi and ES comes up. I could only switch back to Kodi once (the first time). After that, I always get stuck with a running kodi (incl. gui sounds), but no display - just a blank screen.

* I can now start n64 games from within ES, but the sound seems crackling - using the default 'openal' driver in RA. (Previously, I had to use 'alsa' - and there was no crackling). Choosing the 'alsa' driver now results in no sound at all. Maybe I have to specify a specific device ? But as fas as I understand the goal is to use 'openal', correct ?

* Firefox still cause GPU hangs from time to time. I can not say if the kernel update helped or not. I will file a bug report with the DRI guys.

kodi.log and related dmesg output is attached.

If you need any other log files, etc. please let me know.

Thx!

Hi,

I can confirm that HD passthrough in Kodi now works out of the box for me.

But, ES would not start (I see the sway background) and RA would hang as soon as I tried to start a ROM. 'strace' shows that the process is looping around a usleep(). After switching the audio driver from 'openal' to 'alsa' in RA, the ROM would start (I tried N64 games).

ES is mostly still stuck - I got it to start up once or twice....

Which one is the 'preferred' audio drive to use here ? Any other settings (other than the defaults) that are recommended ?

Sometimes, when I quit RA, Kodi gets started, but I only see a text console (not the actual Kodi screen), although Kodi is running and I can hear Kodi's GUI sounds when switching though the menus. May there's a race condition handing off/over the display ? I can not switch to the Kodi screen with ALT-F* or CTRL-ALT-F*.

Also, Firefox sometimes still freezes with those GPU hanging errors, I reported earlier. Looking in the kernel 5.5-rc3 diffs, it seems that there's active development going on. So, hopefully, this will be fixed soon.

I really like the combination of Kodi, ES/RA and Firefox.

Let me know if there is anything I can do to help debug or test.

Thx.

Quote from slacker666

Lastly, I couldn't go back to Kodi once I launched ES, if I chose Shutdown Emulationstation it would just start again.

Same here, the only option to 'leave' ES is to choose restart/reboot, but that reboots the whole system...

Quote from escalade

The RPi image is severely outdated, I've removed it for now. Will move to vc4 and wayland for RPi later.

New generic image is out:

• Linux 5.5-rc2
• Qt 5.14
• Updated Mesa/LLVM, RPCS3 and libretro cores
• PPSSPP does not work with Wayland (or KMS) at the moment, solved it by running in Xwayland which runs great
• Updated xfstools, f2fs-tools and btrfs-progs

Some initial feedback (Intel i915 platform):

* Kodi seems to work quite well

* Emulations station and RetroArch work. Audio on RetroArch is distorted.

* Firefox hangs while browsing and sometimes crashes when watching videos / movies.

'dmesg' shows the following:

[ 253.717593] i915 0000:00:02.0: GPU HANG: ecode 8:1:0x85dffffb, in sway [979], stopped heartbeat on rcs0

[ 253.717597] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.

[ 253.717598] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel

[ 253.717598] drm/i915 developers can then reassign to the right component if it's not a kernel issue.

[ 253.717599] The GPU crash dump is required to analyze GPU hangs, so please always attach it.

[ 253.717600] GPU crash dump saved to /sys/class/drm/card0/error

[ 253.818910] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 277.700806] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 305.328406] docker0: port 1(veth9015086) entered disabled state

[ 305.328580] vethd92864a: renamed from eth0

[ 305.406963] docker0: port 1(veth9015086) entered disabled state

[ 305.418372] device veth9015086 left promiscuous mode

[ 305.418379] docker0: port 1(veth9015086) entered disabled state

[ 310.665831] Asynchronous wait on fence i915:sway[979]:291c timed out (hint:intel_atomic_commit_ready+0x0/0x54)

[ 315.674310] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 375.530248] docker0: port 1(veth9dbe76f) entered blocking state

[ 375.530253] docker0: port 1(veth9dbe76f) entered disabled state

[ 375.530353] device veth9dbe76f entered promiscuous mode

[ 375.920760] eth0: renamed from vethe38bd60

[ 375.934393] IPv6: ADDRCONF(NETDEV_CHANGE): veth9dbe76f: link becomes ready

[ 375.934454] docker0: port 1(veth9dbe76f) entered blocking state

[ 375.934457] docker0: port 1(veth9dbe76f) entered forwarding state

[ 713.673405] perf: interrupt took too long (2521 > 2500), lowering kernel.perf_event_max_sample_rate to 79200

[ 867.629023] perf: interrupt took too long (3164 > 3151), lowering kernel.perf_event_max_sample_rate to 63000

[ 1121.589470] perf: interrupt took too long (3979 > 3955), lowering kernel.perf_event_max_sample_rate to 50100

[ 1296.659249] Asynchronous wait on fence i915:sway[1589]:10b6a timed out (hint:intel_atomic_commit_ready+0x0/0x54)

[ 1300.813647] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 1365.662484] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 1521.139933] perf: interrupt took too long (5051 > 4973), lowering kernel.perf_event_max_sample_rate to 39600

[ 1934.711125] Asynchronous wait on fence i915:sway[1589]:1b8a0 timed out (hint:intel_atomic_commit_ready+0x0/0x54)

[ 1939.719734] i915 0000:00:02.0: Resetting rcs0 for stopped heartbeat on rcs0

[ 1939.946970] vethe38bd60: renamed from eth0

[ 1939.987282] docker0: port 1(veth9dbe76f) entered disabled state

[ 1940.005109] docker0: port 1(veth9dbe76f) entered disabled state

[ 1940.017128] device veth9dbe76f left promiscuous mode

[ 1940.017134] docker0: port 1(veth9dbe76f) entered disabled state

[ 3640.563563] perf: interrupt took too long (6363 > 6313), lowering kernel.perf_event_max_sample_rate to 31200

[ 8388.785593] perf: interrupt took too long (9279 > 7953), lowering kernel.perf_event_max_sample_rate to 21300

card0-error.txt