LE has no use-case for video encoding so it's not something I can comment upon. Video decoding is generally supported.
Posts by chewitt
-
-
Code
2024-07-29 15:26:39 up 10 min 538352 2024-07-29 15:27:39 up 11 min 543008 2024-07-29 15:28:39 up 12 min 542848 2024-07-29 15:29:39 up 13 min 544544 ... 2024-08-03 11:15:11 up 4 days 803584 2024-08-03 11:16:11 up 4 days 799680 2024-08-03 11:17:11 up 4 days 799888 2024-08-03 11:18:11 up 4 days 810112
Defnitely an increase in RAM consumption over time, but figuring out what is beyond my Linux skillset.
-
The debug log doesn't show anything being played, so not much use
-
I believe you are using rockchip BSP.
LE runs upstream mainline kernels only (inevitably with some patches, but we try to limit that). LE 11.x uses Linux 6.1, LE 12.x uses Linux 6.6 and LE13 (master/dev) currently usues 6.6, 6.9 or 6.10 depending on the SoC platform. All kernel and ffmpeg sources are in the buildsystem.
BSP kernels are universally horrible and need to die in fire
-
Diff
--- a/sound/pci/hda/patch_hdmi.c +++ b/sound/pci/hda/patch_hdmi.c @@ -1996,6 +1996,7 @@ static const struct snd_pci_quirk force_connect_list[] = { SND_PCI_QUIRK(0x1043, 0x86ae, "ASUS", 1), /* Z170 PRO */ SND_PCI_QUIRK(0x1043, 0x86c7, "ASUS", 1), /* Z170M PLUS */ SND_PCI_QUIRK(0x1462, 0xec94, "MS-7C94", 1), + SND_PCI_QUIRK(0x8086, 0x06c8, "HP", 1), /* Elite Mini 800 G6 */ SND_PCI_QUIRK(0x8086, 0x2060, "Intel NUC5CPYB", 1), SND_PCI_QUIRK(0x8086, 0x2081, "Intel NUC 10", 1), {}
If anyone can build a test image (I'm not able to build things where I am) .. that ^ kernel patch should do it?
If you can confirm yes, I'll upstream the patch to the kernel.
-
-
-
-
Either a whole show or 60 seconds of a show if that's easier .. enough to test.
-
LE version? .. Debug Logs? .. Media Sample for testing?
-
Code
Display MoreRPi5:~ # cat .config/autostart.sh #!/bin/sh ( L=/storage/memtest.log while sleep 60 do P=$(pidof kodi.bin) test -z "$P" && continue echo $(date "+%Y-%m-%d ") $(uptime | cut -f1,1 -d',') $((16*$(cut -f 2 -d ' ' </proc/$P/statm))) >>$L done )& RPi5:~ # cat memtest.log 2024-07-29 15:17:39 up 1 min 384368 2024-07-29 15:18:39 up 2 min 384368 2024-07-29 15:19:39 up 3 min 384368 2024-07-29 15:20:39 up 4 min 384368 2024-07-29 15:21:39 up 5 min 384368 2024-07-29 15:22:39 up 6 min 384144 2024-07-29 15:23:39 up 7 min 384144 2024-07-29 15:24:39 up 8 min 384144 2024-07-29 15:25:39 up 9 min 536416 <= start playing tvheadend stream 2024-07-29 15:26:39 up 10 min 538352 2024-07-29 15:27:39 up 11 min 543008 2024-07-29 15:28:39 up 12 min 542848 2024-07-29 15:29:39 up 13 min 544544
^ minor changes to log uptime and background commands else they block boot. I do have a large media library so scrolling around will consume some RAM with caching and kids do play old SNES games somtimes (although they've been sent to grandparents for the next two weeks). The rest of our family usage is rather ordinary: watching media from a NAS in the LAN and streams from a remote Tvheadend instance.
-
Enable SSH in the LE settings add-on and login remotely. No need for the textmode console.
-
Code
RPi5:~ # uptime 17:00:37 up 12 days, 20:36, load average: 2.62, 2.65, 2.67 RPi5:~ # free -m total used free shared buff/cache available Mem: 8052 2177 4711 201 1461 5875 Swap: 0 0 0
This is LE13 (master) not LE12 but it's the same Kodi version in the image (not Piers, yet). Remind me in another couple of days and let's see if memory has substantially increased.
-
I've seen past reports of Windows tablets running LibreELEC, but to run the Generic image they MUST be using Intel x86_64 CPUs. If the devices are using some kind of ARM SoC (and many do these days) we have no arm64 image.
-
-
I've some some Google searching on the error messages but didn't find anything conclusive. The card is using the unmtaintained Broadcom "wl" driver and there's been a few issues reported recently that just seem to point to kernel drift as a cause. In short, the kernel keeps moving forwards and this (not in-kernel) vendor driver has been in a static state for some years (at least a decade, maybe longer) and the phrase "if you're not moving forwards, you're going backwards" applies. Eventually the kernel has moved forward far enough that random stuff just starts breaking, and that seems to be the pain zone where we are now. The main option for LE will be to drop "wl" and swap to the in-kernel "b43" driver, but that's not brilliant (not really maintained since forever too) and lacks 5GHz support which will break some people's setups. It's probably the lesser of evils though.
If you can swap the card for something else, it might time to investigate doing so. Most of the devices I've seen this cards in use some kind of PCIe card or mini-PCIe daughter-card so they can be pulled and inexpensively replaced with something better.
-
-
Please share a debug log that demonstrates the problem. Thanks.