10:07 <loki666> apritzel: I guess with what ChenYu asked for my ramp-delay patch, I should create a new patch, not a V2

10:24 <apritzel> I think it would indeed be a new patch/thread, since the prefix will change and the patch will be completely different

10:27 <loki666> ok, thanks, I was also looking at the axp20x-pek driver, and I think it needs a different config for axp717 ...

10:30 <loki666> ie: it seems to read, the startup/shutdown press time from a reg (AXP20X_PEK_KEY) that doesn't seems to exist on ax717, I think it should be 0x26, not 0x36

10:32 <loki666> and I think bit5 of 0x25 should be set to 1 for irq wakeup

10:33 <hlauer> <apritzel> "1) v1.0 never needed PH23. It..." <- Yes. In the available schematics of v1.0 PH23 is nc at the R40 side. The two USB-A ports are powered at startup even without a SD card inserted (so not booting)

10:36 <hlauer> <apritzel> "2) v1.1 uses PH23 for something,..." <- v1.1 resets the system during usb setup without any error message and it starts booting again

10:38 <apritzel> does that happen even in U-Boot, for instance? So if you set PH23 on a v1.1 board using "gpio set ph23"

10:39 <hlauer> <apritzel> "3) USB VBUS is not always-on for..." <- So far I didn't manage to get any power on the two UBS-A ports - yes, most probably another GPIO is needed.

10:42 <apritzel> hlauer: so are you going to send a patch for the M2U dts then? To remove PH23? Happy to review that ...

10:52 <hlauer> Yes, even in u-boot v1.1 resets immediately without any message. Yesterday I tried 6.13-rc1 with mmc2 bus-width = <1> removed and the original 8 put back in place, which yields opses in the mmc kernel part. So there are more differences between v1.0 and v1.1

10:57 <apritzel> hlauer: so if eMMC has issues, it's mostly due to non-appropriate speed modes, like the board not being able to cope with HS-200

12:57 <machinehum> What are some of the common low cost allwinnder TV boxes with good mainline support? I'm looking to by one

12:57 <machinehum> buy* lol

13:17 <apritzel> machinehum: all the H618 ones should be very similar, and boot headless just fine. But there is no mainline graphics support and no support for the built-in PHY

13:18 <apritzel> there are out of tree patches for both, ready to apply to mainline, but none of that is merged, and it will probably take a few months

13:19 <apritzel> you could fix the Ethernet problem by getting one with Gigabit Ethernet (external PHY works), though there are only a few of them, if you can still find them even

13:20 <apritzel> machinehum: or you get one to help review the DE3.3 patches, and help getting the graphics stack support merged ;-)

13:21 <machinehum> https://www.aliexpress.com/item/1005006863899228.html?channel=twinner

13:21 <machinehum> I can help with the patches for sure

13:22 <machinehum> + testing, or whatever.

13:24 <apritzel> machinehum: yeah, that's one of those typical ones, with 100Mbit Ethernet, using the (yet unsupported) internal PHY

13:25 <machinehum> Yeah happy to help test that

13:25 <apritzel> and Tanix is slightly better than the other no name ones

13:25 <machinehum> crazy cheap

13:26 <apritzel> indeed, I'd take the 4GB version

13:29 <machinehum> Seems like there's quite a bit of info about the TX6 (H6 based) but not a lot on the TX68

13:32 <machinehum> I'm going to get one of each lol

13:53 <apritzel> well, H6 is a bit better, graphics works there, but same problem (and same potential solution) with (100MBit) Ethernet

13:53 <apritzel> H6 has (working) USB3 support, but only up to 3GB of DRAM

13:54 <apritzel> but the main problem seems to be availability on those sites: they tend to quickly shift over, and then you don't get the older generation anymore

13:55 <machinehum> hmm, so I guess my H6 TX6 advertised with 4GB is a lie

13:55 <apritzel> yes

13:55 <apritzel> if you are lucky, it's 3GB pus 1GB "virtual DRAM"

13:55 <machinehum> Not upset, expected at such

13:56 <machinehum> ;o virtual, sounds fast!

13:56 <apritzel> well, many boxes named "TX6" are actually H616 as well, check the details

13:56 <apritzel> then the 4GB would be legit

13:57 <machinehum> Yeah I noticed that, it was listed as H6, but again, who knows \o/

13:59 <apritzel> what Mali does it list? That's typically a good indicator, G31 would be H616, T720 is H6

14:00 <machinehum> Mali-T720

14:00 <machinehum> So yeah, likely H6

14:31 <loki666> apritzel: looks like bsp is using another pek driver indeed

14:32 <loki666> I guess another patch incoming

14:33 <loki666> need to figure out, wtf is going on, and if I can make a quirk to existing driver, or write a new one

14:54 <apritzel> loki666: driver design in the BSP is really bad, so don't copy anything from there. I think you should be able to integrate it in the existing driver

15:06 <hlauer> Ok, just tried a openwrt image with 6.6 kernel - resets the v1.1 also. How can I lower the eMMC speed from HS-200?

16:14 <apritzel> hlauer: you would need to explicitly advertise HS200, also that only works with 1.8V, whereas you seem to have 3.3V on vqmmc-supply

16:14 <apritzel> to limit the frequency, use: max-frequency = <100000000>;

16:15 <apritzel> and maybe add: mmc-ddr-3_3v;

16:23 <apritzel> hlauer: try lower frequencies as well (50 or even 40 MHz). Also worth trying 1.8V vqmmc, maybe it is using this voltage?

16:24 <apritzel> hlauer: also can you post a kernel log somewhere, with those oopses? Curious to see what the MMC controller negotiated, and why there are oopses at all

17:21 <hlauer> http://paste.debian.net/1338540/ Omitted all the waiting statements of systemd, which waits for sd access

17:31 <warpme> guys: i returned to hacking with h728 tvbox and i'm stuck at stage when kernel starts to execute userspace init. and just hangs. Uart log is like this: https://gist.github.com/warpme/c43af39b9adfc9afa66ef98afedd2723 My dts is very minimal (no any cpufreq, etc): https://gist.github.com/warpme/a6d8164e88e10407456e7d130f934c28 As kernel just hand - without any message -i'm a bit out of ideas how to move forward with this....Does anybody

17:31 <warpme> have any idea here?

17:37 <hlauer> setting bus-width = <1>; for &mmc2 instead of 8 make it working again

17:40 <apritzel> warpme: I think your culprit is here: "Warning: unable to open an initial console."

17:40 <apritzel> warpme: ... and looking further up: there is no ttyAS0 in mainline, that's a pure Allwinner invention.

17:42 <apritzel> (it's just ttyS0, like any proper 8250 compatible UART always has been)

17:43 <apritzel> hlauer: it indeed smells like a speed mode issue: normally DDR50 should work everywhere .... until it doesn't :-(

17:44 <apritzel> hlauer: read this: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/mmc/host/sunxi-mmc.c#n1432

17:45 <apritzel> hlauer: as a hack you might want to duplicate the H5 check to test for the R40 too

17:46 <apritzel> or you lower the max-frequency below 50 MHz, I think this should discard DDR50 as well

17:47 <warpme> apritzel : qll: [ 4.836710] udevd[112]: starting version 3.2.14

17:47 <warpme> [ 6.191026] sunxi-wdt 2050000.watchdog: Watchdog enabled (timeout=16 sec, nowayout=0)

17:47 <warpme> [ 5.700246] random: crng init done

17:47 <warpme> [ 5.758420] udevd[113]: starting eudev-3.2.14

17:48 <warpme> i must run now... will report how it goes later. thx for help!

20:24 <hlauer> <apritzel> "hlix: so are you going to send a..." <- patch for PH23 removal is out. Hopefully the format will do as my machine crashed when cloning the git repo.

20:28 <hlauer> for the moment I removed mmc2 (soldered on board) in DT which stops the mmc driver crashing the kernel for the inserted mmc0 SD card. Thanks!

20:36 <Soupborsh> I am having fun, I found UART address and I can output to UART in FEL mode by using sunxi-fel writel 0x1C28000 0x00100053 with 2 last hex numbers replaced with hex ascii character representation.

20:40 <Soupborsh> I found addresses from https://raw.githubusercontent.com/Soupborsh/uboot_b288/refs/heads/main/include/configs/sun8iw10p1.h and https://raw.githubusercontent.com/Soupborsh/uboot_b288/refs/heads/main/arch/arm/include/asm/arch-sun8iw10p1/platform.h. There is a lot of insteresting information in the U-Boot repo and it contains Boot0 source and maybe BROM too(it says SBROM).

23:12 <apritzel> Soupborsh: so that looks a lot like the H3, but with some changes, for instance a 4th MMC, like in the A20/R40