| josch | you are! | 00:01 |
|---|---|---|
| josch | now you can post your a311d kernel build time yourself even :) | 00:01 |
| vkoskiv | Done! First toot | 00:08 |
| vkoskiv | Hey! Follower! | 00:10 |
| vkoskiv | It even makes a lil sound | 00:10 |
| josch | i hate it when my browser tabs make sound XD | 00:11 |
| vkoskiv | Favourite image viewers on wayland? I've used feh, but it's a bit spartan | 00:11 |
| vkoskiv | Yeah, discord is also really good at making a sound, and then not indicating anywhere where it came from | 00:11 |
| josch | i'm a feh user as well | 00:12 |
| minute | vkoskiv: sxiv (well, i guess that's X, and gthumb) | 00:12 |
| josch | when i like most about feh in contrast to other viewers is the --action option which lets me script all kinds of stuff | 00:12 |
| josch | and then always with --sort mtime --reverse --auto-rotate --recursive --fullscreen --hide-pointer --draw-filename --draw-actions | 00:13 |
| vkoskiv | Really, the thing I miss from macOS is a quick and ergonomic way to browse a tree of folders containing pics | 00:13 |
| vkoskiv | My film photos are organized (loosely) hierarchically | 00:13 |
| josch | thunar or the kde equivalent do this | 00:13 |
| - S0rin (QUIT: Ping timeout: 255 seconds) (~S0rin@user/s0rin) | 00:24 | |
| + S0rin (~S0rin@user/s0rin) | 00:25 | |
| vkoskiv | Wonder when the bot detector gets triggered when I just spam the follow button on anyone that sounds familiar :D | 00:39 |
| vkoskiv | A lot of notifications are happening now | 00:44 |
| vkoskiv | I guess the bot detector did fire, I followed a bunch of people, but it still only shows 4 | 00:46 |
| vkoskiv | Wasn't expecting *this* many notifications :D | 01:09 |
| - minute (QUIT: Ping timeout: 248 seconds) (~mntirc@softboy.mntmn.com) | 01:27 | |
| + minute (~mntirc@softboy.mntmn.com) | 01:29 | |
| + ndufresne2 (~ndufresne@apple.collaboradmins.com) | 01:51 | |
| - ndufresne (QUIT: Ping timeout: 255 seconds) (~ndufresne@apple.collaboradmins.com) | 01:52 | |
| * ndufresne2 -> ndufresne | 01:52 | |
| - Boostisbetter (QUIT: Ping timeout: 252 seconds) (4a410829d7@irc.cheogram.com) | 02:11 | |
| vkoskiv | I posted a thing: https://community.mnt.re/t/my-a311d-installation-with-pictures/1693 | 02:29 |
| - mark_ (QUIT: Ping timeout: 255 seconds) (~mjw@gnu.wildebeest.org) | 02:51 | |
| - nsc (QUIT: Ping timeout: 264 seconds) (~nicolas@201-98-142-46.pool.kielnet.net) | 03:55 | |
| + nsc (~nicolas@14-48-142-46.pool.kielnet.net) | 03:57 | |
| + Boostisbetter (4a410829d7@irc.cheogram.com) | 08:12 | |
| Boostisbetter | vkoskiv: have you tried suspend on the A311D? Is it working? | 08:26 |
| Boostisbetter | vkoskiv: I just remembered that it is CM4 based, and does not have the faculities for suspend. | 08:27 |
| Boostisbetter | I'm not sure if banana pi has RTC or sleep functionalities included. | 08:30 |
| minute | Boostisbetter: it does have the facility to suspend. it is not cm4 "based". it just has the same shape as the rpi cm4, but it's a totally different processor. | 11:04 |
| josch | Boostisbetter: so once the a311d can suspend we both upgrade to it? :D | 11:10 |
| + mark_ (~mjw@gnu.wildebeest.org) | 13:27 | |
| - S0rin (QUIT: Ping timeout: 255 seconds) (~S0rin@user/s0rin) | 13:28 | |
| + S0rin (~S0rin@user/s0rin) | 13:42 | |
| - S0rin (QUIT: Ping timeout: 245 seconds) (~S0rin@user/s0rin) | 13:55 | |
| + S0rin (~S0rin@user/s0rin) | 14:13 | |
| Boostisbetter | Not sure, needs more RAM to encourage me | 14:22 |
| - TadeusTaD (QUIT: Remote host closed the connection) (tadeustad@psifactor.pl) | 14:44 | |
| + TadeusTaD (tadeustad@psifactor.pl) | 14:50 | |
| Boostisbetter | But I do like the ability to suspend. I also think it is cool that the Banana Pi is faster than the CM4 and the IMX8 | 15:02 |
| Boostisbetter | Heck who am I kidding. That thing is super appealing. I do wonder if regular CM4 ships would work in there as well? It seems like 8GB CM4 chips might start be possible to get again. | 15:05 |
| - mjw (QUIT: Killed (NickServ (GHOST command used by mark_!~mjw@gnu.wildebeest.org))) (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 15:08 | |
| * mark_ -> mjw | 15:08 | |
| + mark_ (~mjw@2001:1c06:2488:1400:4fd:39a7:74ac:7bae) | 15:08 | |
| josch | Boostisbetter: when you say "CM4" you actuaully mean "raspi CM4", right? Becauuse CM4 is just the form factor+connectors. | 15:19 |
| jfred | I plan on switching to the a311d module when mine comes in but likewise also like the ability to suspend. It hasn't yet really worked for me on the original CPU module either though | 15:24 |
| jfred | USB 2 is the only noteworthy downgrade for me I think but I rarely need USB 3 speeds | 15:24 |
| jfred | (I might eat my words the next time I need to write a new live USB 😆) | 15:25 |
| Boostisbetter | josch: yes I mean Raspberry Pi. | 15:36 |
| minute | yes, you can also use rpi cm4 with 8gb. | 15:56 |
| minute | there's a caveat though, we don't have a system image for that | 15:56 |
| minute | https://source.mnt.re/reform/mnt-reform-raspberry-pi-cm4-som#2-working-with-raspberry-pi-cm4 | 15:57 |
| - stites (QUIT: Read error: Connection reset by peer) (~stites@130.44.147.204) | 16:37 | |
| + stites (~stites@130.44.147.204) | 16:38 | |
| - stites (QUIT: Ping timeout: 255 seconds) (~stites@130.44.147.204) | 16:49 | |
| + stites (~stites@2607:fb90:ade3:cc85:9fb2:d897:5057:9250) | 16:49 | |
| Boostisbetter | minute: thanks! Impressive work as usual! | 17:29 |
| Boostisbetter | I wonder if the the new RPI5 will eventually have compute models as well. | 17:30 |
| Boostisbetter | Since it is using cusotm silicon (RP1) I wonder if that would be possible. | 17:30 |
| Boostisbetter | Also sense the Pi 5 now has sleep functionality it would be something very nice to possibly be able to get for a CPU upgrade in the Reform or Pocket. Although the active cooling might be a problem. | 17:30 |
| vkoskiv | Boostisbetter: I've tried suspend, it doesn't work at the moment | 17:58 |
| vkoskiv | I do have UART set up to investigate, but haven't gotten anywhere yet | 17:59 |
| Boostisbetter | vkoskiv: fingers crossed! Hope you have luck! | 18:04 |
| josch | Boostisbetter: in before the rpi5 compute module is completely incompatible with CM4 XD | 18:09 |
| josch | vkoskiv: do you have some time of testing custom u-boot on your a311d reform? :) | 18:10 |
| josch | *for testing | 18:10 |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:20) | 18:17 | |
| + IchikaZou (~IchikaZou@36-231-116-153.dynamic-ip.hinet.net) | 19:00 | |
| - stites (QUIT: Read error: Connection reset by peer) (~stites@2607:fb90:ade3:cc85:9fb2:d897:5057:9250) | 19:02 | |
| + stites (~stites@130.44.147.204) | 19:02 | |
| - stites (QUIT: Remote host closed the connection) (~stites@130.44.147.204) | 20:07 | |
| vkoskiv | Yeah, what are you working on> | 20:20 |
| vkoskiv | s/>/? | 20:20 |
| josch | vkoskiv: bit-by-bit reproducible u-boot for a311d | 20:28 |
| josch | currently, there are a few random bytes in it | 20:29 |
| josch | vkoskiv: as you maybe have seen in the conversation over in #linux-amlogic, those are supposedly *random* bytes so maybe we can just overwrite them with zeroes and everything keeps working | 20:29 |
| vkoskiv | Cool! Side note, I don't have video in u-boot with the a311d, is that expected? | 20:29 |
| josch | yes, display support is not part of upstream u-boot yet | 20:30 |
| vkoskiv | Okay | 20:32 |
| vkoskiv | Meanwhile, I'm reading Azoteq documentation, the linux multitouch hid driver code, and mnt trackpad firmware code to try and make sense of it all :D | 20:32 |
| vkoskiv | I'm fairly confident I can make some improvements to the firmware eventually | 20:33 |
| josch | vkoskiv: once you have some time, could you please test if this u-boot.bin works on a311d for you: https://source.mnt.re/reform/reform-a311d-uboot/-/jobs/2343/artifacts/raw/flash.bin | 20:33 |
| vkoskiv | azoteq doesn't document the default values for the touch controller parameters, I think I'll just have to dump it out to look at it | 20:34 |
| vkoskiv | What's the offset for that binary? | 20:35 |
| josch | you flash it to your sd-card like this: | 20:35 |
| josch | dd if=./flash.bin of="/dev/yoursd" conv=notrunc bs=512 skip=1 seek=1 | 20:35 |
| josch | this will probably break so please have a backup sd prepared XD | 20:36 |
| vkoskiv | Yeah, writing it on my other card | 20:36 |
| vkoskiv | josch: I see linux booting, seems like you got it right? | 20:40 |
| josch | that was the unmodified version to make sure that the base of my changes works :) | 20:41 |
| josch | here is the modified flash.bin: | 20:41 |
| vkoskiv | Aha | 20:41 |
| josch | https://mister-muffin.de/reform/flash.bin | 20:42 |
| josch | sha1sum should be c1b9d8e43ece1b2d29ff64d57277674ec133495e | 20:42 |
| josch | you can diff it to the flash.bin you had before to make sure it replaces the first 512 bytes with zeroes and then some random bytes with MNTREFORMAMLBOOT | 20:43 |
| vkoskiv | It is, yes | 20:43 |
| vkoskiv | Diff looks okay, flashing | 20:44 |
| vkoskiv | Still boots! | 20:46 |
| josch | wow | 20:46 |
| josch | so those proprietary statically linked x86 binaries really write useless random bytes to flash.bin | 20:46 |
| josch | that was too easy | 20:46 |
| vkoskiv | Maybe they are receiving money from Big Compression? :O | 20:46 |
| josch | minute: do you agree to overwrite the random bytes in our build process to make flash.bin bit-by-bit reproducible if the input is the same? | 20:47 |
| minute | josch: suee | 21:01 |
| minute | sure | 21:02 |
| - bkeys (QUIT: Remote host closed the connection) (~Thunderbi@199.226-24.cm.ptn.tftn.dynamic.friendlycity.net) | 21:23 | |
| vkoskiv | The 'failed to send h2c command' spam in dmesg is quite fast sometimes, it seems | 22:22 |
| vkoskiv | I get a few of them every 2 seconds or so now | 22:22 |
| vkoskiv | And the wifi is not sending any packets in this state | 22:23 |
| vkoskiv | There seems to be some issue that causes the wifi to not work after a while, reconnecting fixes it | 22:23 |
| minute | vkoskiv: hmm ok so for me, what happens is that the wifi stops working for a bunch of seconds and then just magically works again | 22:24 |
| minute | i'm mostly on ethernet atm though | 22:24 |
| vkoskiv | reconnecting also silences the spam of that in the log, at least for a bit | 22:24 |
| minute | hm | 22:24 |
| vkoskiv | For me, it might do that also, but most of the times I've noticed it, it's down until I reconnect | 22:24 |
| vkoskiv | I've also mostly been on ethernet | 22:24 |
| minute | so from previous debugging what i think happens is that the module sends some garbage data to the host / they get out of sync | 22:25 |
| minute | while this was in development the driver would then crash and take parts of the system with it | 22:25 |
| vkoskiv | About 60sec after reassociating with base station, I get the first error in dmesg | 22:25 |
| minute | i worked with the driver maintainer(s) to fix that though | 22:25 |
| vkoskiv | Yeah, I haven't taken a look at the driver yet | 22:25 |
| minute | the big Q is, why doesn't it happen for vendor linux | 22:25 |
| vkoskiv | Is that fix in the patches? | 22:25 |
| minute | afaik it only happens in mainline | 22:25 |
| vkoskiv | Ah. Could we compare mainline vs vendor drivers to check? | 22:26 |
| vkoskiv | Okay, I have my first attempt to resolve the sleepy trackpad thing, about to flash it to test | 22:26 |
| vkoskiv | I'm just disabling the active mode timeout entirely for now | 22:27 |
| minute | the person who knows most about the driver is xdarklight in #linux-amlogic btw | 22:27 |
| vkoskiv | I'll ask them if they need help debugging, perhaps? | 22:31 |
| vkoskiv | Assuming they have the hw though, if they are developing the thing | 22:31 |
| vkoskiv | So far, my little fix seems to work. I no longer get a brief moment of no tracking when mousing after some idle time | 22:32 |
| vkoskiv | Hmm, maybe it just happeed... | 22:33 |
| vkoskiv | Testing involves waiting ~30s and moving the cursor | 22:33 |
| vkoskiv | I don't know the default value, docs don't say, I think the vendor configures that during manufacture | 22:33 |
| vkoskiv | I could read it and dump it out | 22:33 |
| vkoskiv | Okay, yeah, the issue persists | 22:34 |
| vkoskiv | Whoops | 22:34 |
| minute | vkoskiv: xdarklight does not have this exact hw combo unfortunately | 22:35 |
| vkoskiv | Oh, maybe it's the idle touch mode timeout, actually | 22:37 |
| vkoskiv | minute: I'll offer to test things on my hw, then? | 22:37 |
| vkoskiv | Maybe a bit difficult though | 22:37 |
| vkoskiv | Many variables | 22:38 |
| minute | vkoskiv: that's kind of how i did it, yeah | 22:38 |
| minute | and i also tried debugging it myself | 22:38 |
| vkoskiv | I'll read the driver code myself and pretend to understand it. | 22:38 |
| minute | see also https://lore.kernel.org/lkml/20230709195712.603200-1-martin.blumenstingl@googlemail.com/T/ | 22:38 |
| minute | in the end i figured out that the patch was not completely right, but that in case of errors, one had to read the whole packet from the card | 22:40 |
| minute | otherwise things would go out of step | 22:40 |
| minute | what i would suggest at first is to turn of rx aggregation in the driver, see if the problem goes away | 22:40 |
| minute | ah yeah, so we carry the fixed patch https://source.mnt.re/reform/reform-debian-packages/-/blob/main/linux/patches/meson-g12b-bananapi-cm4-mnt-reform2/0019-wifi-rtw88-sdio-Honor-the-host-max_req_size-in-the-R.patch?ref_type=heads | 22:42 |
| minute | but if one turns of rx aggregation there's no read loop at all | 22:43 |
| minute | off | 22:43 |
| minute | vkoskiv: you can try to modify this function to just return at the top https://elixir.bootlin.com/linux/latest/source/drivers/net/wireless/realtek/rtw88/sdio.c#L655 | 22:44 |
| vkoskiv | I found another error in the kernel log, firmware failed to ack driver for entering Deep Power Mode | 22:46 |
| minute | maybe this can be turned off? | 22:47 |
| vkoskiv | Yeah, not sure if it's related, I'll investigate | 22:47 |
| vkoskiv | realtek/rtw88/ps.c:106 | 22:48 |
| josch | minute: all three u-boot builds are now bit-by-bit reproducible. Should I let reform-system-image be built with those new u-boot versions? | 22:53 |
| minute | josch: sounds good! | 22:53 |
| josch | i tested that it works on imx and ls1028a | 22:53 |
| josch | but it should not break anything as i only added SOURCE_DATE_EPOCH and vkoskiv tested a311d | 22:53 |
| josch | but vkoskiv only tested the manually patched binary and not what the CI pipeline produced but it should be identical | 22:54 |
| minute | josch: you have my blessing | 23:02 |
| josch | minute: cool! the next big thing is the reform-tools MR which adds support for ls1028a and a311d | 23:03 |
| josch | https://source.mnt.re/reform/reform-tools/-/merge_requests/58 | 23:03 |
| josch | i marked it as draft because it should be tested more before people's setups break again | 23:03 |
| minute | josch: awesome! i'll write down to review it | 23:04 |
| josch | thank you! | 23:05 |
| josch | i tested it on ls1028a only | 23:05 |
| josch | because i cannot thrash my imx XD | 23:05 |
| minute | understandable! | 23:08 |
| - GNUmoon (QUIT: Remote host closed the connection) (~GNUmoon@gateway/tor-sasl/gnumoon) | 23:36 | |
| + GNUmoon (~GNUmoon@gateway/tor-sasl/gnumoon) | 23:37 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!