*** leon-anavi has quit IRC | 00:13 | |
*** toscalix_ has quit IRC | 01:48 | |
*** Wetmelon has quit IRC | 04:10 | |
*** anjumkaiser has joined #automotive | 05:25 | |
*** vrubiolo has joined #automotive | 07:54 | |
*** walzert has joined #automotive | 08:13 | |
*** walzert has quit IRC | 09:47 | |
paulbarker | dl9pf: https://www.mail-archive.com/debian-bugs-dist@lists.debian.org/msg1743984.html | 10:14 |
---|---|---|
paulbarker | dl9pf: Maybe try dropping "cma=64M" from the kernel cmdline | 10:19 |
paulbarker | I've just tried adding that here to see if it stops audio from coming up, but instead it completely stops the kernel from booting | 10:20 |
paulbarker | Ok, so "cma=64M" prevents the SD card from being read. That won't be a problem in CI if you're net booting | 10:24 |
*** vrubiolo has quit IRC | 10:29 | |
*** vrubiolo has joined #automotive | 10:29 | |
dl9pf | but it might lead to other issues. | 10:54 |
dl9pf | ok we can take cma=64M out ... | 10:54 |
dl9pf | the same filesystem works if booted from the sdcard | 10:56 |
dl9pf | so our kernel cfg seems ok wrt modules and such | 10:56 |
paulbarker | dl9pf: Yes, it looks like it breaks MMC, audio, several other drivers which all depend on the vchiq mailboxes | 10:56 |
dl9pf | thats all due to cma ? ... heh ... that setting was just for things to work wrt gfx back then. well sure. Can remove. But might break older kernels in CI/test then. | 10:57 |
dl9pf | worth a try | 10:57 |
dl9pf | <rant mode>undocumented *****</rant mode> | 10:58 |
paulbarker | Pretty standard for rpi | 10:58 |
dl9pf | indeed | 10:59 |
paulbarker | It will be annoying if the kernel cmdline in CI needs to depend on the kernel version | 11:00 |
dl9pf | that would be 'jikes' but not impossible ... | 11:01 |
dl9pf | ok | 11:01 |
dl9pf | + aplay -l | 11:01 |
dl9pf | **** List of PLAYBACK Hardware Devices **** | 11:01 |
dl9pf | card 0: Headphones [bcm2835 Headphones], device 0: bcm2835 Headphones [bcm2835 Headphones] | 11:01 |
dl9pf | Subdevices: 8/8 | 11:01 |
dl9pf | Subdevice #0: subdevice #0 | 11:01 |
dl9pf | Subdevice #1: subdevice #1 | 11:01 |
dl9pf | Subdevice #2: subdevice #2 | 11:01 |
paulbarker | https://www.raspberrypi.org/forums/viewtopic.php?p=1674488#p1674488 may be relevant | 11:01 |
dl9pf | Subdevice #3: subdevice #3 | 11:01 |
dl9pf | Subdevice #4: subdevice #4 | 11:01 |
dl9pf | Subdevice #5: subdevice #5 | 11:01 |
dl9pf | Subdevice #6: subdevice #6 | 11:01 |
dl9pf | Subdevice #7: subdevice #7 | 11:01 |
dl9pf | + set +x | 11:01 |
dl9pf | https://lava.automotivelinux.org/scheduler/job/15850 with cma=64M off | 11:01 |
dl9pf | easy to fix | 11:02 |
dl9pf | read half through that post and so far all just half-baked and no explanation as to why | 11:04 |
dl9pf | but thats how it is | 11:04 |
paulbarker | dl9pf: For me, using "cma=64M@256M" works, just "cma=64M" breaks | 11:05 |
paulbarker | I believe the issue is where in memory that cma block gets allocated, not its size | 11:05 |
dl9pf | heh ... shouldn't there be known boundaries ... lol . | 11:06 |
dl9pf | well cma=64m@256m sound good to me then ... trying | 11:06 |
dl9pf | https://lava.automotivelinux.org/scheduler/job/15852 | 11:07 |
dl9pf | if I recall correctly cma=64 was necessary to render gfx | 11:08 |
paulbarker | It should be set in the device tree these days | 11:10 |
dl9pf | iirc there are settings in there now | 11:10 |
dl9pf | https://github.com/raspberrypi/linux/blob/88debfb15b3ac9059b72dc1ebc5b82f3394cac87/arch/arm/boot/dts/bcm2711.dtsi#L871 | 11:12 |
paulbarker | The "cma=..." cmdline argument overrides what's in the dtb | 11:14 |
dl9pf | ok, so better kill the cmdline as its now in the devicetree and let it handle that | 11:14 |
dl9pf | less to care about | 11:14 |
dl9pf | https://lava.automotivelinux.org/scheduler/job/15852#L2446 works, thus let me deploy that change and retrigger the 5.4 job | 11:15 |
dl9pf | nasty little bugger | 11:15 |
paulbarker | https://github.com/raspberrypi/linux/blob/88debfb15b3ac9059b72dc1ebc5b82f3394cac87/arch/arm/boot/dts/bcm283x.dtsi#L40 | 11:16 |
paulbarker | That's the current default size | 11:16 |
paulbarker | 64M | 11:16 |
dl9pf | ok, so should be enough space by default. | 11:17 |
dl9pf | fine | 11:17 |
dl9pf | let's remove the direct setting and be done | 11:17 |
paulbarker | Sounds good | 11:17 |
paulbarker | The error you posted late last night put me onto the right track | 11:18 |
dl9pf | only left now: | 11:18 |
dl9pf | extra_kernel_args: "coherent_pool=1M 8250.nr_uarts=1 rw fstab=no " | 11:18 |
dl9pf | cool - glad it worked out. | 11:18 |
paulbarker | The firmware adds more options by default but of course none of them are documented | 11:18 |
paulbarker | I think we just have to say that if the tests pass we go with what we've got | 11:19 |
dl9pf | was already thinking 'omg /me old mouldwarp, that can't be rocket science' | 11:20 |
dl9pf | w/o the additional options we also see: | 11:22 |
dl9pf | https://www.irccloud.com/pastebin/qxo2vCD9/ | 11:22 |
*** leon-anavi has joined #automotive | 11:22 | |
leon-anavi | hi | 11:23 |
dl9pf | whats the jira again ? | 11:25 |
dl9pf | will use SPEC-3702 | 11:30 |
paulbarker | dl9pf: Yes that's the one | 11:36 |
*** AlisonChaiken has quit IRC | 11:46 | |
*** kooltux has quit IRC | 11:57 | |
*** kooltux has joined #automotive | 11:58 | |
*** leon-anavi has quit IRC | 11:58 | |
dl9pf | ok one down | 12:02 |
*** anjumkaiser has quit IRC | 13:10 | |
*** anjumkaiser has joined #automotive | 13:11 | |
*** Newami has joined #automotive | 14:26 | |
*** Newami has quit IRC | 14:27 | |
*** eduardas has joined #automotive | 15:06 | |
*** anjumkaiser has quit IRC | 15:22 | |
*** anjumkaiser has joined #automotive | 16:23 | |
dl9pf | mvlad: ping | 16:42 |
dl9pf | mvlad: See https://lava.automotivelinux.org/scheduler/job/15868#L5105 | 16:42 |
dl9pf | mvlad: any idea what | 16:42 |
dl9pf | [ 887.275053] vmap allocation for size 8298496 failed: use vmalloc=<size> to increase size | 16:42 |
dl9pf | means we need to do ? | 16:42 |
dl9pf | this is qemu-system-arm | 16:42 |
*** eduardas has quit IRC | 16:48 | |
*** eduardas has joined #automotive | 16:48 | |
*** eduardas has quit IRC | 16:49 | |
*** vrubiolo has quit IRC | 16:58 | |
dl9pf | montjoie: try this: | 17:53 |
dl9pf | image_arg: '-kernel {kernel} -append "console=ttyAMA0,115200 root=/dev/vda verbose systemd.log_color=false mem=2048M vmalloc=256M"' | 17:53 |
dl9pf | see last 2 options added | 17:53 |
dl9pf | seems slow, need to check | 17:53 |
montjoie | dl9pf: I have started one with memory 4096 https://lava.automotivelinux.org/scheduler/job/15869 the message goes away | 17:56 |
montjoie | the screenshot is still bad | 17:57 |
*** toscalix has joined #automotive | 18:17 | |
dl9pf | yes | 18:30 |
dl9pf | you need the vmalloc= kernel cmdline ... figuring out the value | 18:31 |
dl9pf | mvlad: https://lava.automotivelinux.org/scheduler/job/15883#L2489 | 18:47 |
dl9pf | "Usable area:" does not show on qemu-system-arm | 18:48 |
*** eduardas has joined #automotive | 19:03 | |
*** eduardas has quit IRC | 19:11 | |
*** Wetmelon has joined #automotive | 21:12 | |
*** anjumkaiser has quit IRC | 21:22 | |
*** Wetmelon has quit IRC | 21:51 | |
*** Net147 has quit IRC | 22:05 | |
*** Net147 has joined #automotive | 22:06 | |
*** bn_mobile has joined #automotive | 22:13 | |
*** Wetmelon has joined #automotive | 22:42 | |
*** toscalix has quit IRC | 22:49 | |
*** Net147 has quit IRC | 23:15 | |
*** Net147 has joined #automotive | 23:15 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!