*** eduardas has quit IRC | 00:26 | |
*** Wetmelon has quit IRC | 00:45 | |
*** brlogger` has joined #automotive | 02:17 | |
*** bn_mobile has quit IRC | 02:57 | |
*** jdmark_ has quit IRC | 03:04 | |
*** CathyInBlue has quit IRC | 04:32 | |
*** Wetmelon has joined #automotive | 04:44 | |
*** CathyInBlue has joined #automotive | 04:45 | |
*** jdmark has joined #automotive | 05:59 | |
*** anjumkaiser has joined #automotive | 06:38 | |
*** Wetmelon has quit IRC | 06:59 | |
*** leon-anavi has joined #automotive | 07:29 | |
leon-anavi | morning | 07:31 |
---|---|---|
*** anjumkaiser has quit IRC | 08:14 | |
*** anjumkaiser has joined #automotive | 08:39 | |
paulbarker | dl9pf: I'm looking at the rpi audio issue | 08:59 |
paulbarker | dl9pf: In this comment https://gerrit.automotivelinux.org/gerrit/c/AGL/meta-agl/+/25840/3#message-0f708b610e96a6d3cedb4714d38bd3b47e32bfbc, where does that dt overlay come from? | 08:59 |
paulbarker | I can't find it in the source tree | 08:59 |
*** hch has quit IRC | 09:12 | |
*** Piraty has quit IRC | 09:12 | |
*** Piraty has joined #automotive | 09:14 | |
*** hch has joined #automotive | 09:14 | |
*** kooltux_ has joined #automotive | 10:18 | |
dl9pf | mvlad: montjoie: can you exchange/hash this out wrt screenshooter here? | 10:49 |
mvlad | qemux86-64:~# ls -al *.png || echo $XDG_RUNTIME_DIR && agl-screenshooter && ls -al *.png | 11:13 |
mvlad | ls: cannot access '*.png': No such file or directory | 11:13 |
mvlad | /run/user/0 | 11:13 |
mvlad | -rw-r--r--. 1 root root 1087468 Jan 18 11:13 agl-screenshot-2021-01-18_11-13-40.png | 11:13 |
mvlad | montjoie: could you check if XDG_RUNTIME_DIR is actually set? | 11:15 |
montjoie | I resend a job with more debug | 11:18 |
montjoie | mvlad: yes it is set https://lava.automotivelinux.org/scheduler/job/15809 | 11:27 |
montjoie | but the binary is not found, so unrelated with XDG_RUNTIME_DIR ? | 11:27 |
dl9pf | change script to use just /usr/bin/agl-screenshooter and not whereis | 11:32 |
dl9pf | might also be some HOME or TMP dir missing in the environment | 11:32 |
dl9pf | mvlad: do you depend in any other Env var bes8de XDG_RUNTIME_DIR ? | 11:33 |
dl9pf | montjoie: the kingfisher might not embed the screenshooter. bug in packagegroup. try pi4 or qemu! | 11:39 |
*** kooltux_ has quit IRC | 11:39 | |
montjoie | dl9pf: it works on qemu https://lava.automotivelinux.org/scheduler/job/15810 | 12:02 |
mvlad | dl9pf: no, not that I know of. | 12:06 |
mvlad | dl9pf: agl-screenshooter is not packge, it comes by default the compositor. | 12:09 |
mvlad | s/packge/package, or better said a yocto/OE recipe. | 12:09 |
mvlad | dl9pf: it should be available on every platform. | 12:09 |
mvlad | montjoie: it seems so yes. Need to tune it a bit, as in, to stop homescreen and re-load/re-running w/ the correct QMLs to match the ref image. | 12:11 |
*** kooltux_ has joined #automotive | 12:11 | |
mvlad | quit | 12:50 |
mvlad | oops :) | 12:50 |
*** adriano has quit IRC | 12:56 | |
*** sknebel has quit IRC | 12:56 | |
*** adriano has joined #automotive | 12:59 | |
*** sknebel has joined #automotive | 12:59 | |
montjoie | in fact XDG_RUNTIME_DIR is not mandatory to work | 13:02 |
*** gkiagia has quit IRC | 13:03 | |
*** gkiagia has joined #automotive | 13:22 | |
dl9pf | ok, so we got a step closer. will chat during the call | 13:30 |
*** gkiagia has quit IRC | 13:35 | |
dl9pf | paulbarker: we do use the u-boot cmdline ... | 13:39 |
dl9pf | hmm | 13:39 |
dl9pf | mvlad: whereis probably was the problem | 13:40 |
dl9pf | let's just either call agl-screenshooter from PATH then . | 13:40 |
dl9pf | most of the time we use 'which' or 'which | head -1' in such cases | 13:41 |
mvlad | ok | 13:57 |
*** gkiagia has joined #automotive | 14:03 | |
*** anjumkaiser has quit IRC | 14:04 | |
*** anjumkaiser has joined #automotive | 14:26 | |
dl9pf | paulbarker: See https://lava.automotivelinux.org/scheduler/job/15819 | 14:46 |
dl9pf | paulbarker: the driver seems to be built as "=M" module, let's try if it is a built-in. | 14:46 |
paulbarker | dl9pf: Sounds good. I'm looking into the kernel cmdline to see if the diff between my local pi and what you've got there in lava is important | 14:47 |
paulbarker | I keep running into boot failures though | 14:48 |
paulbarker | Something very sketchy about this. It works for several boots then I start getting failure to load the fdt. After that no more successful boots until I re-write the SD card image | 14:48 |
dl9pf | huch ??? wrt boots ... | 14:49 |
paulbarker | I should buy another rpi4. I got mine for free, I believe they're production samples or early revisions | 14:49 |
dl9pf | https://lava.automotivelinux.org/scheduler/job/15819#L790 shows no alsa sound cards, but this is kernel bootup right before userspace and the module load is of course later when nbd is mounted and systemd init is running. | 14:50 |
paulbarker | No alsa sound cards is expected at that point I believe | 14:51 |
dl9pf | but the module load does not seem to enable audio then lateron | 14:51 |
dl9pf | aplay -l: https://lava.automotivelinux.org/scheduler/job/15819#L2439 | 14:52 |
dl9pf | probably we're missing something | 14:53 |
paulbarker | dl9pf: Do you have a HDMI display attached? | 14:59 |
paulbarker | I get very different `aplay -l` output with and without HDMI connected, makes me worry a little | 15:00 |
dl9pf | the DSI (official touchscreen) is attached | 15:01 |
dl9pf | audio should be using the default 3,5mm jack in our case | 15:01 |
paulbarker | dl9pf: I should look what `aplay -l` prints under 4.19 | 15:03 |
* paulbarker goes looking in lava | 15:04 | |
paulbarker | Ok, so that command isn't normally in the test | 15:05 |
paulbarker | dl9pf: At least I figured out the problem with my pi. It's saving junk into uboot.env, removing that makes it boot again. So it's not corrupting the SD card | 15:09 |
*** kooltux_ has quit IRC | 16:08 | |
dl9pf | hmm, do you have a dump of the runtime devicetree for audio=on vs audio=off in dtparams ? | 16:37 |
*** CathyInBlue has quit IRC | 16:45 | |
*** CathyInBlue has joined #automotive | 16:46 | |
paulbarker | dl9pf: I can't get the pi to boot at all when loading the dtb manually in u-boot | 16:59 |
paulbarker | So I can't see if the firmware is doing anything magic to the fdt after it is loaded | 16:59 |
paulbarker | I'm trying to replicate what the lava job does but using `fatload` instead of `tftp` and setting `root=/dev/mmcblk0p2` on the kernel cmdline | 17:00 |
dl9pf | i mean not with u-boot | 17:02 |
dl9pf | what does the devicetree in /sys show on a raspbian if booted with dtparam=audio=on and w/o ? | 17:03 |
dl9pf | we could diff to see if it sets another option | 17:03 |
paulbarker | dl9pf: I can try that. The thing is, with the AGL images I do see soundcards without `dtparam=audio=on` if I just set the audio status to okay in the fdt from u-boot | 17:07 |
paulbarker | So what I'd want to compare against is the case when the firmware hasn't touched the fdt at all (i.e. when loading a dtb manually in u-boot using tftp or fatload) | 17:08 |
paulbarker | dl9pf: Tried everything I can think of there, going to look at Raspbian now | 17:13 |
paulbarker | dl9pf: https://pastebin.com/XrrSE3SS | 17:19 |
paulbarker | kaslr-seed changes on every boot so we can ignore that | 17:24 |
dl9pf | heh ... then I have zero idea ... | 17:25 |
dl9pf | we do set status to okay in u-boot cmdline already | 17:26 |
dl9pf | but no sound device shows up | 17:26 |
dl9pf | latest try : | 17:26 |
dl9pf | https://lava.automotivelinux.org/scheduler/job/15827 | 17:26 |
paulbarker | dl9pf: I've seen several completely different outputs from `aplay -l` in my testing today | 17:26 |
dl9pf | so i do not have hdmi connected here ... only dsi touchscreen, thus it should default to bcm2835 using the jack | 17:27 |
dl9pf | well we do not care actually if jack or hdmi in the test atm anyway | 17:27 |
*** Wetmelon has joined #automotive | 17:28 | |
paulbarker | dl9pf: Are you using a combined dtb for this testing? | 17:29 |
dl9pf | our stock combined-dtb as built with 'agl-demo agl-devel agl-netboot' | 17:30 |
dl9pf | hmm sec | 17:31 |
dl9pf | no | 17:31 |
dl9pf | https://lava.automotivelinux.org/scheduler/job/15825 is with stock dtb as generated by us | 17:32 |
dl9pf | https://lava.automotivelinux.org/scheduler/job/15828 with our stock 'combined' dtb | 17:33 |
paulbarker | dl9pf: I'm going to wrap up for the day and take another look tomorrow | 17:47 |
paulbarker | Hopefully will spot something with a fresh mind | 17:48 |
dl9pf | paulbarker: roger. at least the difference seems to be only that 'okay' - which we do set. | 17:58 |
paulbarker | dl9pf: My wonder is what else does the firmware unconditionally do. I think I can check that tomorrow though | 17:59 |
dl9pf | https://hydrus.org.uk/journal/rpi-sound.html seems interesting | 18:00 |
dl9pf | snd_bcm2835.enable_compat_alsa=0 snd_bcm2835.enable_hdmi=1 snd_bcm2835.enable_headphones=1 | 18:00 |
dl9pf | they seem to have put an incompatible change into late >4.19.xxx kernels | 18:00 |
dl9pf | hmm, no these options are quite old actually ... hmmm | 18:13 |
*** Wetmelon has quit IRC | 18:29 | |
*** Wetmelon has joined #automotive | 18:35 | |
*** Wetmelon has quit IRC | 19:10 | |
*** jdmark has quit IRC | 20:12 | |
*** jdmark_ has joined #automotive | 20:12 | |
*** jdmark_ has quit IRC | 20:19 | |
*** toscalix has joined #automotive | 20:22 | |
*** toscalix_ has joined #automotive | 20:44 | |
*** toscalix has quit IRC | 20:45 | |
*** leon-anavi has quit IRC | 21:27 | |
*** leon-anavi has joined #automotive | 21:27 | |
*** Wetmelon has joined #automotive | 22:09 | |
dl9pf | paulbarker: got the same result | 22:34 |
dl9pf | https://www.irccloud.com/pastebin/2al1RnwW/ | 22:34 |
dl9pf | bootargs = "coherent_pool=1M 8250.nr_uarts=1 snd_bcm2835.enable_compat_alsa=0 snd_bcm2835.enable_hdmi=1 video=HDMI-A-1:1920x1080M@60 smsc95xx.macaddr=DC:A6:32:B1:17:44 vc_mem.mem_base=0x3ec00000 vc_mem.mem_size=0x40000000 dwc_otg.lpm_enable=0 root=/dev/mmcblk0p2 rootfstype=ext4 rootwait usbhid.mousepoll=0"; | 22:34 |
dl9pf | this if the fdt delivered by the firmware | 22:34 |
dl9pf | so the snd_bcm2835 are new, vc_mem.* is new compare to the CI setup | 22:35 |
dl9pf | paulbarker: i think it is this: | 22:44 |
dl9pf | [ 1.159438] bcm2835_vchiq fe00b840.mailbox: failed to set channelbase | 22:44 |
dl9pf | [ 1.159450] vchiq: could not load vchiq | 22:44 |
dl9pf | [ 1.159523] bcm2835_vchiq: probe of fe00b840.mailbox failed with error -22 | 22:44 |
dl9pf | sound is part of the mailbox | 22:44 |
*** anjumkaiser has quit IRC | 22:46 | |
*** Newami has joined #automotive | 22:49 | |
*** Newami has quit IRC | 23:07 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!