06:36 <jbrunet> "it seems audio support is on a mission to be a pain in the arse for distros" ... nice

06:37 <jbrunet> FWIW, alsa provides aslactl which allows to save and restore the cards settings based on the card name

11:51 <chewitt> jbrunet: apologies, I am mostly frustrated at my lack of alsa knowledge

13:26 <ChrisBillington> Hello, I am working on testing mainline on an S905X3 'TV Box', aka X96 Air. I have written a dts file and am testing. It has some 'interesting' issues not least the Ethernet (RTL8211F) only works at 100M not 1000M as expected.

13:28 <ChrisBillington> Please can you advise me on what branch on the 'official' kernel repo I should use for testing, so I can give you useful testing info? And the same for u-boot, because I am having to chainload from the vendor u-boot in order to get round display corruption (wrong colours) when the HDMI is initialised by the vendor bootloader.

13:46 <ChrisBillington> Current kernel used is test build of Armbian by a forum user but the location of sources is unclear, so I want to build a known kernel. Ethernet traffic seems to stop on the RX side in gigE mode, as TX packet count advances in ifconfig but RX doesn't.

14:02 <chewitt> ChrisBillington: this contains some useful fixes, so would be a good base to work from

14:02 <chewitt> https://github.com/chewitt/linux/commits/amlogic

14:03 <chewitt> I'm also interested in the dts that you're creatting

14:04 <chewitt> if the armbian user is balbes150, i'm the upstream source for his patches

14:05 <chewitt> afk for a bit..

14:26 <ChrisBillington> Yes it's one of balbes150 builds, of 5.7.0-rc3. You are welcome to the .dts once it's working right, currently having fun trying to understand how the sound card section is supposed to work and get SPDIF/Analogue configured. Also there is zilch thermal management in the sei610 file I used as a base that needs adding too.

14:28 <ChrisBillington> I'll try with a kernel from your repo. What's best repo for u-boot? Eventually I want to build a u-boot with ext2/4 support so I don't have to have a fatxx boot partition.

15:01 <chewitt> ChrisBillington: have a look at https://github.com/chewitt/linux/commit/0a17eb855f40ed3257ccbe1d0e2622a11f5dd514 for audio

15:02 <chewitt> and for u-boot https://github.com/chewitt/u-boot/tree/2020.04-patches has everything Amlogic related since 2020.04 shipped

15:04 <chewitt> for the fip sources, have a look at the c4/vim3l/sei610/u200 folders in here https://github.com/chewitt/amlogic-boot-fip/tree/master

15:05 <chewitt> I'd make sure you back-up the contents of emmc first as there's no guarantee they will work on your box

15:05 <chewitt> I have pre-compiled versions for C4/VIM3L here https://chewitt.libreelec.tv/testing/9.80/u-boot/

15:07 <chewitt> btw, are you sure the box is supposed to have 1Gb ethernet?

15:08 <chewitt> most of the 'air' devices are marketed as low(er) end streaming devices, and this usually means "rmii" not rgmii and 10/100

15:20 <ChrisBillington> Thanks! Yes it's an RT8211F External phy chip and works at 1G in Android. The basic Air version is 100M but this is a 4/32.

15:43 <ChrisBillington> making the dts is WIP, because with no schematic it's based on physical exam of the board, educated guesswork and extracting snippets of info from the decompiled Android dtb

15:45 <ChrisBillington> Android dtb hints it also has a t9015 device for analogue audio out.

16:06 <ChrisBillington> Client: HexChat 2.14.3 • OS: ArchLinux • CPU: Intel(R) Xeon(R) CPU W3565 @ 3.20GHz (1.60GHz) • Memory: Physical: 11.4 GiB Total (8.6 GiB Free) Swap: 7.8 GiB Total (7.8 GiB Free) • Storage: 273.7 GB / 1.9 TB (1.7 TB Free) • VGA: NVIDIA Corporation G94GL [Quadro FX 1800] @ Intel Corporation Xeon 5500/Core i7 Integrated Memory Controller Channel 2 Address Registers • Uptime: 4d 1h 12m 4s

16:07 <ChrisBillington> woops

16:08 <ChrisBillington> Sorry, that was a 'what does this button do' moment

19:40 <unlord> Hi, I have a Nintendo Switch I just booted into L4T. Should I be seeing more than 4 cores?

20:30 <xdarklight> unlord: you may want to ask in a nvidia tegra related channel, I'm not sure how many of us Amlogic guys have nvidia hardware

20:56 <unlord> xdarklight: I asked around, it is only 4 cores

20:56 <unlord> thanks for the tip though

21:17 <ChrisBillington> <chewitt> I tested X96 Air S905X3/sm1 with 5.7.0-rc6-164932-gae3accb7412b-dirty from your repo, GigE is still not working. I need to check it's being correctly set up by u-boot. Wondering if the lack of RX direction is connected to the 'rx-delay-ns' parapmeter recently added by xdarklight for meson8/8m2. Could this patch be required for G12 and SM1 Soc too?

21:19 <chewitt> can you pastebin the dts and dmesg somewhere?

22:00 <ChrisBillington> dts x96air here: https://pastebin.com/xc3PwAG8

22:06 <ChrisBillington> console bootlog w/dmesg here https://pastebin.com/BANATk3H

22:06 <ChrisBillington> audio isn't working with this dts, have not worked out why yet

22:07 <ChrisBillington> eth_phy interrupts I thnk are correct, 100M only works with GPIOZ_14/15 this way round.

22:18 <xdarklight> ChrisBillington: so /proc/interrupts shows that the PHY interrupt count increases when you unplug/plug the Ethernet cable? if that's not the case then something is still wrong with the IRQ, but it *shouldn't* break 1Gbit/s speeds

22:20 <xdarklight> ChrisBillington: also check the debug prints in rtl8211f_config_init drivers/net/phy/realtek.c - enable those to see what the bootloader does on the PHY side

22:22 <xdarklight> ChrisBillington: also you can try phy-mode = "rgmii-txid"; that way you don't need amlogic,tx-delay-ns anymore. and if you want to know whether my RX delays patches will do something on your board then I need to know the value of the PRG_ETH0 register in meson8b_init_prg_eth. to get that, add: dev_err(dwmac->dev, "PRG_ETH0 = 0x%08x\n", readl(dwmac->regs + PRG_ETH0)); at the beginning of that function

22:23 <chewitt> ChrisBillington have you tried booting with the VIM3L or C4 dts in my branch?

22:24 <xdarklight> ChrisBillington: that's all I can say from looking at the .dts. maybe I'm too tired but your console boot log doesn't show *anything* about Ethernet (no dwmac-meson8b or stmmac output in there)

22:28 <ChrisBillington> sorry, my mistake, the console log doesn't capture all of dmesg for some reason. I'll post that separately

22:29 <ChrisBillington> Will try tomorrow alternative dts and the tests of /proc/interrupts

22:30 <ChrisBillington> How do I enable debug prints in rtl8211f_config_init drivers/net/phy/realtek.c? Will also have a go at getting PRG-ETH0 register value

22:32 <chewitt> add "earlyprintk" to boot params

22:33 <xdarklight> ChrisBillington: for debug prints see https://www.kernel.org/doc/html/v5.6/admin-guide/dynamic-debug-howto.html - or if you're recompiling anyways you can change the level from dev_dbg to dev_err

22:33 <chewitt> (not a response to previous comment, no idea about debug prints in drivers)

22:34 <chewitt> ChrisBillington: random Q .. what wireless chip is inside the box?

22:35 <chewitt> another 'air' device (A95X-F3) has some random realtek chip, not the usual broadcom/ampak module

22:37 <ChrisBillington> thanks - dmesg is here https://pastebin.com/3SCU0gLH

22:39 <ChrisBillington> Wifi is an RTL8822CS (sdio), I believe not supported yet but there may be an out-of-tree driver

22:40 <chewitt> I don't think rtw88 has support for sdio moduels yet, only pcie

22:41 <chewitt> so it would require the vendor binary driver

22:41 <chewitt> the BT bits are maybe already upstream

22:42 <ChrisBillington> I wasn't able to ID the regulator chips, these have markings that don't match any I saw before: "S1589QC" and "LPS 1C901". SO the regul;ators section of dts is guesswork

22:42 <ChrisBillington> BT seems to be just hanging off a UART

22:43 <ChrisBillington> This is an 'X96 Air P3' 4/32 btw. There seem to be multiple variants knocking about including clones.

22:43 <chewitt> uart_a normally

22:44 <chewitt> yup, clones are an issue with this type of box

22:45 <ChrisBillington> There are a lot of Realtek wifi modules in the china boxes at the moment. Must be 10 cents cheaper than Broadcom. This one has no screening can on the wifi, so it's a mystery how it got CE/FCC cert

22:45 <chewitt> I think CE marking means "China Edition" :)

22:46 <chewitt> it just happens to use the same logo as the other CE mark

22:52 <ChrisBillington> btw the spdifout audio-controller@480/680 nodes don't exist in the included dtsi files for sm1. Maybe they should be added in one to save putting them in every board dts with SPDIF? Also the 'toacodec'audio-controller@740 which I believe is for analogue audio out. Most TV boxes have AV and SPDIF outputs.

22:53 <ChrisBillington> The vendor represents them as being CE-approved, but, you know, China is creative!

22:54 <ChrisBillington> I see my attempt to make analogue audio out work fails with [ 11.668707] t9015-codec audio-codec-1: failed to get core clock

22:54 <ChrisBillington> [ 11.668725] t9015-codec: probe of audio-codec-1 failed with error -2

22:55 <ChrisBillington> but a t9015 node is in the android dtb

23:03 <ChrisBillington> btw Martin/xdarklight, you mentioned your 8m2 box had died. I have an old one (M8S) you are welcome to have for testing if you are interested

23:04 <ChrisBillington> midnight here, 'night!

23:10 <lvrp16> wow they're reserving 768MB of CMA on S905X3?

23:13 <chewitt> lvrp16 currently as a workaround for not-optimised vdec things

23:13 <chewitt> on the +ve it's been reduced from 896MB :)