CoreELEC 22.0-Piers_alpha3 Discussion

Perri · 21 May 2026 10:42

Can anyone get SMB share to work ?

Portisch · 21 May 2026 10:42

atirage21 · 21 May 2026 19:56

It is going, but on ce22no device and shared drive is needing set smb v1 for succes identification and pairing.

freddy · 22 May 2026 03:10

No issues, since long time.I have older dedicated laptop for smb share Win 10 ,so ihave to set minimum protocol 2 max 3

and dont forget WS discovery..

on clean install maybe this is disabled/ default

Andrey · 3 June 2026 11:31

Title: [S928X-K] Kernel Panic (DMC SEC ERROR) during VC-1 playback (amcodec)

Body:

The film is Mars Attack!

Device: Ugoos SK1 (Amlogic S928X-K) OS: CoreELEC Amlogic-no (Kernel 5.15.170)

Issue Description: I’ve tracked down a consistent hard reboot (SError / Kernel Panic) issue on the S928X platform. The crash occurs exactly 10-20 minutes into playing SDR HD content encoded in the legacy VC-1 codec.

Since recent Kodi builds removed the specific GUI toggle to disable hardware acceleration for VC-1, the stream is automatically fed to amcodec hardware decoder. If I completely disable global amcodec hardware acceleration in settings, the software decoding handles it flawlessly with no crashes.

This points directly to a memory leak or buffer mishandling inside the amvdec hardware decoder or op-tee memory allocation when dealing with VC-1 streams on the new 5.15 kernel.

Crash Details: According to the pstore dump, the TEE subsystem gets corrupted, throws a NULL pointer, outputs garbage, and then the DMC catches a memory security violation caused by the VideoPlayer process:

<3>[ 2692.735198] [TEE] WARNING: text(0000000049e3305e) is NULL <0>[ 2693.893524] DMC2 SEC READ CHECK ERROR: addr:0x6800100 <0>[ 2693.893960] FROM IRQ -- DMC1 SEC WRITE CHECK ERROR: addr:0x68002c0 <0>[ 2694.011312] Kernel panic - not syncing: Asynchronous SError Interrupt

I have attached the full dmesg-ramoops-0 log. Could you please check the hardware decoder pathways for legacy codecs on the S928X?

I can’t attach images and links.

vpeter · 3 June 2026 11:34

As usually: test first on nightly build.

theeye · 4 June 2026 10:09

I am running alpha3 on an X96 X10 Pro (S928X-J). In addition to the master profile, I have set up a second profile with its own database. I have now updated to the nightly build 22.0-Piers_nightly_20260602. Everything looks fine in the master profile. However, no database migration took place in the second profile. Furthermore, upon logging into this profile, no MyVideos146.db file is created; MyVideos143.db remains intact. The “Movies” and “TV Shows” sections display the standard screen for an empty library. If I open a folder containing video files while in File Mode, I am met with nothing but a black screen. After approximately 30 seconds, I am returned to the login screen with the profile selection. However, the device does not appear to be restarting; the display continuously shows the time, and the SAMBA connection to my PC remains uninterrupted.

https://paste.coreelec.org/VitaminTimeless

freddy · 4 June 2026 22:09

Here is link from old Kodi thread

https://forum.kodi.tv/showthread.php?tid=350660

Maybe will help with Master Profile loads first/also DB has changed from v3 to nightly

Vasco · 5 June 2026 08:06

There’s something failing in profiles and is being handled by TK. My profiles are also broken. Here’s the link:

github.com/xbmc/xbmc

[python][profiles][dbwrappers] Fix hard crash when switching to a newly created profile

master ← sasha-spb-ru:fix-profile-switch-crash

opened 01:11PM - 01 Jun 26 UTC

sasha-spb-ru

+48 -32

## Description When switching to a profile that has never been used before,… Kodi crashes with a segmentation fault. The crash occurs in two places: - dbiplus::DbErrors receives a NULL error string from sqlite3_errmsg() and passes it directly to vsnprintf, causing an immediate segfault. - The Python invoker crashes inside Py_EndInterpreter because add-on classes are left registered while the sub-interpreter is being torn down. - Additionally, the texture cache database was not being properly closed and re‑opened during a profile switch, which left background jobs running with an invalid database handle and flooded the log with SQLITE_MISUSE errors. This PR fixes all three issues: - Hardens DbErrors::DbErrors to never pass a NULL format string to vsnprintf. - Forces a full garbage collection and module dictionary clearing in CPythonInvoker::onExecutionDone() before destroying the sub‑interpreter. - Adds calls to CTextureCache::Deinitialize() and Initialize() in CProfileManager::LoadProfile() so that the texture database is properly cycled during a profile switch. ## Motivation and context The reported issue #28202 (reproduced) described a hard segfault when switching to a freshly created profile. Investigation revealed that the profile switch did not stop background texture caching jobs before the old database was closed, and two independent code paths would crash on the resulting SQLITE_MISUSE error or on Python interpreter shutdown. The changes ensure that: - Profile switching gracefully handles the texture cache lifecycle. - Any database error is logged safely instead of crashing Kodi. - Python service add‑on shutdowns no longer cause a segfault during profile switching. ## How has this been tested? Tested on a Linux Kodi build. Steps: - Launched Kodi with the master profile. - Created a new profile (tried both “Copy default” and “Start fresh”). - Selected “Load profile” for the newly created profile. - Verified that the switch completes successfully and the new profile's home screen appears without a crash. - Repeated the process several times and also switched back to the master profile. - Verified that exiting Kodi no longer crashes when service add‑ons are stopped during a profile switch. A minor observation (likely intended behavior): deleting a profile that is not currently active shows a confirmation dialog to delete its files. When deleting the currently active profile, no such prompt appears - the profile is simply removed and the profile’s files are not deleted. This does not affect stability. ## What is the effect on users? Users can now create and switch to new profiles without Kodi crashing. Profile switches are now reliable, even when the destination profile has never been used before. The fix also prevents a potential crash on exit when Python service add-ons are stopped, improving overall stability. Resolves #28202 ## Screenshots (if appropriate): ## Types of change - [X] **Bug fix** (non-breaking change which fixes an issue) - [ ] **Clean up** (non-breaking change which removes non-working, unmaintained functionality) - [ ] **Improvement** (non-breaking change which improves existing functionality) - [ ] **New feature** (non-breaking change which adds functionality) - [ ] **Breaking change** (fix or feature that will cause existing functionality to change) - [ ] **Cosmetic change** (non-breaking change that doesn't touch code) - [ ] **Student submission** (PR was done for educational purposes and will be treated as such) - [ ] **None of the above** (please explain below) ## Checklist: - [X] My code follows the **[Code Guidelines](https://github.com/xbmc/xbmc/blob/master/docs/CODE_GUIDELINES.md)** of this project - [ ] My change requires a change to the documentation, either Doxygen or wiki - [ ] I have updated the documentation accordingly - [X] I have read the **[Contributing](https://github.com/xbmc/xbmc/blob/master/docs/CONTRIBUTING.md)** document - [ ] I have added tests to cover my change - [ ] All new and existing tests passed

theeye · 6 June 2026 21:02

So, it seems there is already a working solution. Will this be implemented in one of the upcoming nightly builds? Or will it only arrive with beta 1? By the way: The second profile works for me in the original alpha 3. But as soon as I update to a nightly build, only the master profile works. The second profile’s databases are not migrated.

Andrey · 7 June 2026 18:30

Update: The issue persists on the latest Nightly build.

I have updated my Ugoos SK1 to the latest CoreELEC nightly (Kernel 5.15.196). Unfortunately, the exact same hardware crash (Asynchronous SError) happens after 10-20 minutes of playing the legacy VC-1 codec.

The new pstore dump caught an amazing Call Trace. It shows exactly when the SError triggers:

[TEE] WARNING: text(000000000e289b0b) is NULL (TEE corrupts the memory during VC-1 hardware decoding)

And right after that, the VideoPlayer process tries to allocate a new page for network I/O (nfs_file_read), hits the corrupted memory zone, and the DMC instantly kills the system:

Plaintext

clear_page+0x14/0x50
get_page_from_freelist+0x13ec/0x1434
...
filemap_read+0x1b0/0x984
nfs_file_read+0x80/0x104
vfs_read+0x26c/0x304

Full crash log: https://paste.coreelec.org/CreamySommers

Andrey · 7 June 2026 19:36

Correction on the Local Playback Test:

I need to correct my previous statement. Playing the VC-1/H.264 files locally from a USB drive does not fully fix the issue, but it changes the failure mode.

NFS Playback: Results in an Asynchronous SError / TEE Kernel Panic (likely due to CMA starvation/fragmentation exacerbating the decoder bug).
Local USB Playback: The video plays, but eventually freezes at specific points. The system does not reboot, but the picture hangs. If I rewind, it plays a bit further, but eventually, the video player just crashes and drops me back to the Kodi menu.

Conclusion: The amvdec hardware decoder is fundamentally failing to parse/decode these VC-1 and H.264 streams on the S928X. When memory is fragmented via NFS, this decoder failure escalates into a fatal TEE memory corruption and SError. When played locally, it just results in a standard buffer freeze and player crash.

We really need a way to blacklist VC-1 (and problematic H.264 profiles) from amcodec on this SoC so it falls back to software decoding.

Portisch · 7 June 2026 21:11

Really? I never get issues. Maybe a sample would help? Read forum!

Wolverine59500 · 10 June 2026 09:47

Hello everyone, thank you for the incredible work you do for the community. I currently own two Ugoos boxes, an Am6b+ and an Am9 Pro. On both of them, FEL works without any issues, but I can’t install any theme other than Estuary. I can download the themes (I’ve tried Arctic Zephyr Reloaded, Taejo, and Nox)—they install without any problems—but I get a message saying that the menu icons can’t be created. I’m really new to this, so I’m reaching out for your help.

Oh, sorry—the issue is only with the AM9 running coreelec 22. No problems with the AM6B+ running 21.3.

vortex · 10 June 2026 11:10

For answers and more information, please contact the theme authors on the Kodi forum or in their GitHub repository.

boxr4k · 10 June 2026 15:39

Hi @darioml I don’t know if CoreELEC 22.0-Piers_alpha3 Discussion - #68 by darioml is still applying to you, I mean the need to have a small delay after a screen refresh change, but know that at least on the 20260609 nightly this is fixed and I have it set to 3 seconds without any issue (SDR, HDR, 2K, 4K, whatever)

Magician1981 · 10 June 2026 20:55

Is there an ETA for the beta release?

vortex · 10 June 2026 21:07

It all depends on when critical bugs are discovered in Kodi, LibreELEC, and the current Coreelec nightly builds. There’s no exact date. Help the community find them, and it will speed up the process.

neqwdtkukwjaanti · 10 June 2026 21:25

check milestones on github - Milestones · CoreELEC/CoreELEC · GitHub
since they don’t exist, you can be reasonably sure that CE is just going to follow Kodi

so, check milestones on Kodi - Milestones · xbmc/xbmc · GitHub
since you can see there “No due date” then you can be reasonably sure there is no ETA

since there is no ETA you can surmise that you’ll get it when you get it
and you will now never have to ask again, you know how to check for yourself

vpeter · 11 June 2026 11:12

A post was merged into an existing topic: Nightly builds (NEW)