| josch | (i cancelled the pipeline as it only makes sense to run it after reform-tools 1.25 shows up here https://mntre.com/reform-debian-repo/pool/main/r/reform-tools/ ) | 00:05 |
|---|---|---|
| - klardotsh (QUIT: Ping timeout: 246 seconds) (~klardotsh@c-67-170-115-80.hsd1.wa.comcast.net) | 00:49 | |
| - vagrantc (QUIT: Quit: leaving) (~vagrant@2600:3c01:e000:21:7:77:0:50) | 01:15 | |
| - stites (QUIT: Ping timeout: 248 seconds) (~stites@130.44.147.204) | 01:33 | |
| + stites (~stites@2607:fb91:dc9:6ccd:6a4b:9556:1e53:57eb) | 01:35 | |
| - stites (QUIT: Read error: Connection reset by peer) (~stites@2607:fb91:dc9:6ccd:6a4b:9556:1e53:57eb) | 01:47 | |
| + stites (~stites@130.44.147.204) | 01:48 | |
| - stites (QUIT: Ping timeout: 245 seconds) (~stites@130.44.147.204) | 01:52 | |
| + stites (~stites@2607:fb91:dc9:6ccd:6a4b:9556:1e53:57eb) | 01:52 | |
| - stites (QUIT: Read error: Connection reset by peer) (~stites@2607:fb91:dc9:6ccd:6a4b:9556:1e53:57eb) | 01:54 | |
| + stites (~stites@130.44.147.204) | 01:54 | |
| - eery (QUIT: Ping timeout: 240 seconds) (~eery@156.146.51.201) | 01:58 | |
| - mtm (QUIT: Ping timeout: 248 seconds) (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 02:03 | |
| - jacobk (QUIT: Ping timeout: 246 seconds) (~quassel@utdpat241106.utdallas.edu) | 02:47 | |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:50) | 03:23 | |
| - mjw (QUIT: Ping timeout: 246 seconds) (~mjw@gnu.wildebeest.org) | 03:48 | |
| - nsc (QUIT: Ping timeout: 245 seconds) (~nicolas@50-99-142-46.pool.kielnet.net) | 03:59 | |
| + nsc (~nicolas@40-99-142-46.pool.kielnet.net) | 04:01 | |
| josch | it seems quite the night shift was pulled to queue yet another reform-system-image run to be done over-night -- nice! :D | 04:08 |
| + mtm (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 04:09 | |
| - sknebel (QUIT: Server closed connection) (~quassel@v22016013254630973.happysrv.de) | 04:11 | |
| + sknebel (~quassel@v22016013254630973.happysrv.de) | 04:11 | |
| + jacobk (~quassel@utdpat241055.utdallas.edu) | 04:15 | |
| josch | minute: so it seems the last big difference of reform-system-image between a311d and imx8mq is the value of LINUX_KERNEL_CMDLINE in /etc/default/flash-kernel. It would allow for an extremely nice property if this difference could be avoided. In that case, you could ship a bit-by-bit identical root filesystem for all platforms and only have /boot differ. | 04:59 |
| josch | you could either do that by finding one value that fits all platforms (even if that is possible right now, this might not be possible with future boards) | 04:59 |
| josch | or (and i think this is better) let uboot set the correct ${bootargs} value | 05:00 |
| josch | why does the reform-boundary-uboot branch not do this? | 05:00 |
| josch | yet to be tested and thus marked as draft: https://source.mnt.re/reform/reform-boundary-uboot/-/merge_requests/19 | 05:10 |
| josch | vagrantc: ah good you are still online -- what is your opinion on that topic? | 05:12 |
| josch | i can think of a third solution: copypaste flash-kernel's bootscr.uboot-generic to bootscr.reform2-imx8mq and only have it differ from bootscr.uboot-generic in what it does with "setenv bootargs" | 05:14 |
| josch | vagrantc: what is your opinion of platform-specific bootargs? Should u-boot set those? Should flash-kernel supply them via platform specific boot.scr? | 05:15 |
| josch | the situation is even worse for u-boot-menu which does not allow for platform specific U_BOOT_PARAMETERS in /usr/share/u-boot-menu/conf.d/*.conf but seems to just rely on ${bootargs} set by u-boot | 05:17 |
| josch | so probably setting this in u-boot would be best as it would also fix u-boot-menu? | 05:17 |
| josch | oh my uboot patch is probably wrong. It should be CONFIG_USE_BOOTARGS=y together with CONFIG_BOOTARGS right? | 05:23 |
| vagrantc | oh, i am glad u-boot-menu actually uses bootargs :) | 05:24 |
| vagrantc | i was worried it does not at all ... | 05:25 |
| vagrantc | but yes, seems like the bootloader is really the main part that actually differs ... may as well set it there. | 05:25 |
| josch | wait... it does not? | 05:25 |
| josch | i didn't test whether it does | 05:25 |
| vagrantc | josch: oh ... well ... test that then :) | 05:25 |
| vagrantc | you are also copying all the appropriate .dtb files somehow? | 05:27 |
| vagrantc | what are the differing boot arguments needed? | 05:27 |
| josch | vagrantc: https://paste.debian.net/1291412/ | 05:29 |
| vagrantc | crazy idea ... are some of them possible to tweak from /sys or /proc from initrd ? | 05:31 |
| vagrantc | obviously the serial console not so much ... | 05:31 |
| vagrantc | oh, you can probably safely specify both serial consoles ... it might complain about one not existing ... but should work | 05:32 |
| vagrantc | ugly, ye. | 05:32 |
| josch | lets first try if setting ${bootargs} in u-boot works | 05:32 |
| josch | yes, concatenating all serial consoles should be fine | 05:32 |
| vagrantc | you could also make a u-boot bootscript that checks and sets bootargs differently | 05:33 |
| vagrantc | and serial console too, for that matter | 05:34 |
| josch | vagrantc: yes, that's what i meant by copypasting bootscr.uboot-generic above | 05:34 |
| vagrantc | and then /etc/flash-kernel/machine to "crazy all the MNT Reforms" | 05:34 |
| josch | vagrantc: but such a change would need to end up in flash-kernel itself | 05:34 |
| vagrantc | eventually ... but while you're prototyping it, you can just specify a custom bootscript | 05:35 |
| vagrantc | at some point, you might fall down and not be able to add new ones ... especially with the CM4 module opening many possibilities | 05:36 |
| josch | vagrantc: you'd really ship a reform-specific boot.scr in flash-kernel? | 05:36 |
| josch | evef if the problem can be solved at the uboot level properly forever? | 05:36 |
| vagrantc | josch: just exploring ideas :) | 05:36 |
| josch | okay | 05:36 |
| josch | lets try ${bootargs} first, then | 05:37 |
| vagrantc | obviously, if you can teach u-boot to do make available the right things, that would be nicer. | 05:37 |
| josch | okay, it's already good to know that that is the proper way forward | 05:38 |
| vagrantc | i seem to recall a reluctance with updating u-boot once it made it into the field, though? | 05:38 |
| josch | yes, there is no auto-updating yet | 05:38 |
| josch | but at least new platforms like a311d would benefit from the start | 05:39 |
| vagrantc | josch: in unrelated news, is metasnap.debian.net still operating? :) | 05:39 |
| josch | vagrantc: yes, why? | 05:39 |
| vagrantc | josch: exploring some things that might need it ... at least for the moment | 05:39 |
| josch | vagrantc: keep me updated about them :) | 05:40 |
| josch | vagrantc: it is lagging behind several days because snapshot.d.o is borked :( | 05:40 |
| vagrantc | that's what i was afraid of | 05:40 |
| josch | nothing metasnap can do about that i'm afraid | 05:41 |
| josch | snapshot.d.o just returns 404 if you don't wait long enough | 05:41 |
| vagrantc | more follow-up on #reproducible-builds | 05:42 |
| josch | back to uboot | 05:44 |
| josch | minute: here i built a u-boot for a311d that sets ${bootargs} with the a311d specific stuff: https://source.mnt.re/josch/reform-system-image/-/jobs/2020 | 05:45 |
| josch | minute: you could give it a test-run once you find some time | 05:45 |
| + klardotsh (~klardotsh@c-67-170-115-80.hsd1.wa.comcast.net) | 05:49 | |
| josch | vagrantc: it also happened, that MNT got shipped a new revision (AB) of the nitrogen imx8mq som which can no longer boot from sd-card -- so flashing a new u-boot now becomes more scary because if you do something wrong you soft-brick the board... | 05:53 |
| vagrantc | right. | 05:54 |
| - jacobk (QUIT: Ping timeout: 248 seconds) (~quassel@utdpat241055.utdallas.edu) | 06:26 | |
| + jacobk (~quassel@utdpat241055.utdallas.edu) | 06:30 | |
| - vagrantc (QUIT: Quit: leaving) (~vagrant@2600:3c01:e000:21:7:77:0:50) | 06:30 | |
| - klardotsh (QUIT: Ping timeout: 250 seconds) (~klardotsh@c-67-170-115-80.hsd1.wa.comcast.net) | 07:28 | |
| - jacobk (QUIT: Ping timeout: 248 seconds) (~quassel@utdpat241055.utdallas.edu) | 07:29 | |
| + reformer (~reformer@softboy.mntmn.com) | 08:08 | |
| + jacobk (~quassel@64.189.201.150) | 08:25 | |
| - ggoes (QUIT: Ping timeout: 255 seconds) (~gregf@fsf/staff/ggoes) | 08:59 | |
| - GNUmoon (QUIT: Remote host closed the connection) (~GNUmoon@gateway/tor-sasl/gnumoon) | 09:05 | |
| + GNUmoon (~GNUmoon@gateway/tor-sasl/gnumoon) | 09:09 | |
| - jjbliss (QUIT: Ping timeout: 250 seconds) (~jjbliss@1464766-static.elnsmiaa.metronetinc.net) | 09:10 | |
| + jjbliss (~jjbliss@1464766-static.elnsmiaa.metronetinc.net) | 09:24 | |
| josch | okay, u-boot-menu cannot currently work with the same contents on the / partition for different platforms | 09:36 |
| josch | so we could either: | 09:36 |
| josch | a) give up on common content for the root partition (but that would be sad because the kernel cmdline is the only blocker) | 09:37 |
| josch | b) find a common set of kernel cmdline that fits all SoMs and hope that this also works for future SoMs | 09:37 |
| josch | c) fix u-boot-menu | 09:37 |
| josch | d) never ship u-boot-menu with the system image and continue relying on the legacy boot.scr instead | 09:37 |
| josch | note for vagrantc for later: the problem is, that to fill the "append" line in extlinux.conf, u-boot-update looks at /etc/default/u-boot, /usr/share/u-boot-menu/conf.d/, /etc/u-boot-menu/conf.d/ for a good value of U_BOOT_PARAMETERS | 09:40 |
| josch | vagrantc: if it cannot find any (i.e. U_BOOT_PARAMETERS is empty) it defaults to /etc/kernel/cmdline | 09:40 |
| josch | i misread that last bit thinking it would default to /proc/cmdline which would be nice because then generating extlinux.conf on an already running system would use a cmdline that already proved itself to be working | 09:41 |
| josch | vagrantc: i think a possible fix would be to add another fallback to /proc/cmdline if /etc/kernel/cmdline doesn't exist -- what do you think? | 09:42 |
| josch | maybe accompanied by another option in /etc/default/u-boot which allows taking cmdline from /proc/cmdline as the final fallback? | 10:09 |
| - mhoye (QUIT: Server closed connection) (~mhoye@li319-32.members.linode.com) | 10:30 | |
| + mhoye (~mhoye@li319-32.members.linode.com) | 10:30 | |
| - S0rin (QUIT: Remote host closed the connection) (~S0rin@user/s0rin) | 11:29 | |
| - doppler (QUIT: Server closed connection) (~doppler@user/doppler) | 11:57 | |
| + doppler (~doppler@user/doppler) | 11:57 | |
| - yankcrime (QUIT: Server closed connection) (~nick@193.16.42.83) | 12:19 | |
| + yankcrime (~nick@193.16.42.83) | 12:20 | |
| - XYZ (QUIT: Remote host closed the connection) (~XYZ@89-24-35-94.nat.epc.tmcz.cz) | 13:51 | |
| + mjw (~mjw@gnu.wildebeest.org) | 14:00 | |
| - mtm (QUIT: Ping timeout: 240 seconds) (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 14:03 | |
| - jacobk (QUIT: Ping timeout: 240 seconds) (~quassel@64.189.201.150) | 14:20 | |
| + XYZ (~XYZ@89-24-35-94.nat.epc.tmcz.cz) | 14:29 | |
| minute | josch: what kind of config file does u-boot-menu use for its entries etc? is it plaintext on the boot partition, or some binary format? asking because i'm thinking about how to modify it during debugging etc | 14:35 |
| minute | josch: so far i used mkimage to update boot.scr manually | 14:35 |
| minute | preliminary changelog for system image v4, i'm sure i forgot something, but will polish more on monday https://source.mnt.re/reform/reform-system-image/-/merge_requests/75 | 14:37 |
| josch | minute: you have u-boot-menu installed? In that case, the auto-generated configuration that is parsed by u-boot is located at /boot/extlinux/extlinux.conf | 15:10 |
| josch | one advantage of using extlinux.conf over boot.scr is, that it is declarative instead of being a turing-complete shell script. Another advantage is, that you can edit extlinux.conf with a texteditor instead of requiring a pass through mkimage. | 15:11 |
| josch | you can write your own manual /boot/extlinux/extlinux.conf but users usually do not edit it but use the u-boot-menu package to let it automatically generate entries for all installed kernels | 15:12 |
| josch | if extlinux.conf exists, u-boot will prefer it over boot.scr | 15:12 |
| - robin (QUIT: Ping timeout: 246 seconds) (~robin@user/terpri) | 15:12 | |
| + robin_ (~robin@user/terpri) | 15:12 | |
| minute | josch: ok, very nice | 15:13 |
| minute | thanks for the explainer. i will look into it in the coming week then | 15:14 |
| josch | minute: i don't think there is much to look into because of the problem that extlinux.conf ignores ${bootargs} set by u-boot :( | 15:15 |
| minute | josch: :/ | 15:15 |
| josch | i think more important would be looking at my ${bootargs} merge requests and merging them so that we can soon [tm] not set them anymore via something in /etc | 15:15 |
| josch | (as discussed with vagrantc earlier) | 15:15 |
| josch | oh cool a311d-v2 was merged | 15:17 |
| minute | josch: yes! | 15:18 |
| josch | having the ${bootargs} stuff in u-boot for a311d might already be enough, because then reform-system-image can set imx8mq specific kernel cmdline but that can be overwritten by a311d u-boot's ${bootargs} if needed | 15:21 |
| josch | (as we probably never can reliably update user's u-boot...) | 15:21 |
| minute | problem: i flashed over 30 new imx8m modules yesterday with our current uboot and those can only ever boot from emmc | 15:25 |
| minute | so can there be a fallback in case uboot didn't set bootargs? | 15:25 |
| minute | (these modules have already been built back into reforms also) | 15:26 |
| josch | oh :( | 15:29 |
| josch | well, it is as it is | 15:29 |
| josch | yes, instead of using flash-kernel's provided boot.scr we can create our own | 15:30 |
| josch | that was one of vagrant's ideas as well | 15:30 |
| minute | a boot.scr that could also then forward to extlinux.conf, or how do you mean? | 15:31 |
| josch | no | 15:31 |
| josch | a boot.scr which does different things depending on a) the reform platform and/or b) the existance of ${bootargs} | 15:31 |
| josch | vagrant hinted, that they see it as a possible solution to the problem to add yet another boot.scr template to the flash-kernel packaging | 15:32 |
| - stites (QUIT: Ping timeout: 246 seconds) (~stites@130.44.147.204) | 15:34 | |
| josch | but again, this is only a fallback solution -- it can be avoided if this works: https://source.mnt.re/josch/reform-system-image/-/commit/6577e915d6bd13c18e7e4d025322b85dcb94e842 | 15:35 |
| + stites (~stites@2607:fb91:dc6:4800:fbd5:4720:a3b4:a0d2) | 15:35 | |
| josch | then reform-system-image would set the bootargs for imx8mq and all other platforms would override them like in this commit | 15:35 |
| josch | (override them if needed) | 15:35 |
| josch | for example, imx8mq sets pci=nomsi but u-boot for a311d overrides that with pci=pcie_bus_perf (last option wins) | 15:37 |
| josch | both share the cma=512M option | 15:37 |
| josch | and no_console_suspend probably doesn't hurt on a311d | 15:37 |
| minute | josch: i see! | 15:37 |
| josch | so this might just work but somebody with an a311d would have to try it out :D | 15:37 |
| josch | *wink* *wink* | 15:38 |
| josch | :) | 15:38 |
| minute | hehe | 15:39 |
| minute | well, that will be me! | 15:39 |
| - qbit (QUIT: Ping timeout: 246 seconds) (~qbit@h.suah.dev) | 15:41 | |
| josch | you can do so by applying this patch and building a new a311d system image: https://paste.debian.net/1291449/ | 15:41 |
| + qbit (~qbit@h.suah.dev) | 15:42 | |
| josch | and then put this flash.bin onto that image: https://source.mnt.re/josch/reform-system-image/-/jobs/2020/artifacts/browse | 15:42 |
| josch | i'd be interested whether that system boots and what the resulting /proc/cmdline is | 15:43 |
| + mtm (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 16:09 | |
| minute | josch: ok, writing in my todo for monday | 16:11 |
| josch | sweet :) | 16:12 |
| josch | another blocker for me is, that i would like a project in the reform namespace that builds the a311d u-boot, maybe named reform-a311d-uboot similar to reform-boundary-uboot | 16:13 |
| josch | once i have that, i can unify the reform-system-image scripts into a single one | 16:14 |
| + ggoes (~gregf@fsf/staff/ggoes) | 16:17 | |
| - xktr (QUIT: Quit: leaving) (~xktr@2602:fe3d:c01:10ca:1050:1ace:0:b) | 16:40 | |
| + xktr (~xktr@2602:fe3d:c01:10ca:1050:1ace:0:b) | 16:48 | |
| - stites (QUIT: Read error: Connection reset by peer) (~stites@2607:fb91:dc6:4800:fbd5:4720:a3b4:a0d2) | 17:01 | |
| - cwebber (QUIT: Remote host closed the connection) (~user@user/cwebber) | 17:01 | |
| + stites (~stites@130.44.147.204) | 17:01 | |
| minute | josch: ah yes, sorry for letting you wait for that | 17:20 |
| josch | it's okay, you had so much other stuff to do! :) | 17:23 |
| josch | while writing the unification code for reform-system-image i also thought of yet another way to avoid hardcoding the kernel cmdline in reform-system-image: by doing the platform detection in /usr/share/flash-kernel/ubootenv.d/00reform2_ubootenv and letting that snippet modify ${bootargs} accordingly if it detects imx8mq | 17:24 |
| Boostisbetter | josch: what kind of battery life do you see on your Refrom? About 4 hours? | 17:56 |
| josch | yup, around that | 17:58 |
| josch | though it's getting less | 18:00 |
| josch | right now, my reform dies at around 10-12% remaining battery according to the oled | 18:00 |
| Boostisbetter | hmmm, I have never checked to see myself. I only let it run down once, and that was just to reset the lpc after getting the 2000 mah batteries | 18:35 |
| + jacobk (~quassel@2603-8080-b200-7b02-0000-0000-0000-0b71.res6.spectrum.com) | 18:39 | |
| - mtm (QUIT: Ping timeout: 255 seconds) (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 19:18 | |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:50) | 20:34 | |
| josch | vagrantc: so... we have multiple problems and some solutions :D | 21:23 |
| josch | vagrantc: https://mntre.com/reform-irc-logs/2023-09-09.log.html#t09:36:29 | 21:23 |
| + mtm (~mtm@c-71-228-84-213.hsd1.fl.comcast.net) | 21:24 | |
| josch | that link goes to my thoughts for u-boot-menu | 21:24 |
| josch | yesterday we also talked about a custom boot.scr for the reform. I don't think that's necessary after all because i can just customize bootscr.uboot-generic using a script in /usr/share/flash-kernel/ubootenv.d/ if i want to | 21:25 |
| josch | that script could then do some magic with ${bootargs} depending on the SoM if needed | 21:25 |
| josch | but maybe all our problems go away (for now) because it might be possible to give flash-kernel the bootargs for the imx8mq and then let u-boot for a311d (where we can make sure we do not do the same mistake again) overwrite whatever it needs with its own ${bootargs} | 21:26 |
| josch | minute: i now have a proof-of-concept for a unified reform-system-image. Creating the root-partition takes 45 minutes on my reform and customizing that takes another 7 minutes for each board we want to support. So instead of O(45*N) we get O(45+7*N) which is still linear but not as steep :) | 21:28 |
| minute | josch: oh cool! | 21:58 |
| - _nrb_ (QUIT: Server closed connection) (~nrbnrb@2a01:4f8:172:299c:1::29) | 22:04 | |
| + _nrb_ (~nrbnrb@2a01:4f8:172:299c:1::29) | 22:04 | |
| vagrantc | josch: sounds promising :) | 22:32 |
| josch | vagrantc: except for u-boot-menu :( | 22:34 |
| - jacobk (QUIT: Ping timeout: 248 seconds) (~quassel@2603-8080-b200-7b02-0000-0000-0000-0b71.res6.spectrum.com) | 22:34 | |
| vagrantc | ACTION nods | 22:35 |
| vagrantc | welllll.... | 22:35 |
| vagrantc | i think some platforms have some sort of pre-boot.scr ... | 22:36 |
| josch | vagrantc: but distroboot will by default prefer extlinux.conf over boot.cr | 22:36 |
| josch | *boot.scr | 22:36 |
| vagrantc | josch: this happens before distroboot | 22:36 |
| josch | oh? | 22:36 |
| josch | i never heard of pre-boot.scr | 22:37 |
| vagrantc | another workaround i have done on some systems is to load a boot.scr from a higher priority boot device, and but not actually use it to boot, and then it falls back to extlinux.conf on the next boot device | 22:37 |
| vagrantc | josch: it might not be widely supported | 22:37 |
| vagrantc | and not finding any hits on that exact match in u-boot | 22:38 |
| vagrantc | there is one platform that supports /u-boot-preboot.scr ... but one platform does not make for a compelling default | 22:39 |
| vagrantc | also, the shiny new hotness is defaultboot ... distroboot is now the old thing :) | 22:40 |
| vagrantc | defaultboot? gah. no, stdboot | 22:40 |
| vagrantc | or standard boot ... | 22:40 |
| vagrantc | it is the classic case of "sure, it is better, but now we have yet another boot path to support." | 22:40 |
| josch | nice, i love throwing out all the established and stable stuff to replace it with the hotness of tomorrow :D | 22:41 |
| vagrantc | it does have fallbacks for distroboot-like behavior ... and better option for selecting boot media | 22:41 |
| josch | ah i found it in the u-boot git: doc/develop/bootstd.rst | 22:41 |
| - stites (QUIT: Ping timeout: 248 seconds) (~stites@130.44.147.204) | 23:20 | |
| + stites (~stites@2607:fb91:dcf:4896:22a5:6d79:f786:b171) | 23:21 | |
| + jacobk (~quassel@2603-8080-b200-7b02-0000-0000-0000-0b71.res6.spectrum.com) | 23:22 | |
| - jacobk (QUIT: Ping timeout: 240 seconds) (~quassel@2603-8080-b200-7b02-0000-0000-0000-0b71.res6.spectrum.com) | 23:27 | |
| * robin_ -> robin | 23:46 | |
| - ethulhu (QUIT: Quit: connection reset by purr) (ethulhu@nora.ethulhu.co.uk) | 23:53 | |
| + ethulhu (ethulhu@nora.ethulhu.co.uk) | 23:53 | |
| - stites (QUIT: Read error: Connection reset by peer) (~stites@2607:fb91:dcf:4896:22a5:6d79:f786:b171) | 23:57 | |
| + stites (~stites@2603:3005:b69:4100:afa0:f7be:6ecd:53f2) | 23:57 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!