| josch | okay, i cannot reproduce it anymore with loglevel=7, so here with the default loglevel: https://mister-muffin.de/p/nU2S.jpg | 00:08 |
|---|---|---|
| ch | hmm. i think mine tends to hang before it finds the rootfs. will take a picture next time | 00:09 |
| ch | `make -j4 bindeb-pkg KBUILD_DEFCONFIG=pocketreform_defconfig` works surprisingly well (didnt boot yet tho) | 00:10 |
| josch | try localmodconfig for even faster builds. You can use another machine to do the build by using the LSMOD parameter to point to the lsmod output on your pocket. | 00:11 |
| josch | with that, i get a kernel built in less than 50 minutes from zero on a311d | 00:11 |
| josch | with "make bindeb-pkg -j6" that is | 00:12 |
| ch | doesnt boot | 00:15 |
| ch | "nice" | 00:15 |
| minute | josch: also yes to wayland sddm | 00:18 |
| josch | it's now part of MR 108: https://source.mnt.re/reform/reform-tools/-/commit/4743d5de050ec2516082d8e48ce9c9619c50c30f | 00:20 |
| + wielaard (~mjw@212.78.174.2) | 01:02 | |
| - mjw (QUIT: Read error: Connection reset by peer) (~mjw@212.78.174.2) | 01:03 | |
| - chomwitt (QUIT: Quit: WeeChat 3.8) (~chomwitt@2a02:587:7a14:db00:1ac0:4dff:fedb:a3f1) | 02:08 | |
| - cobra (QUIT: Ping timeout: 246 seconds) (~cobra@user/Cobra) | 02:39 | |
| midfavila | josch: are you trying to get an X server on the mq? | 03:11 |
| midfavila | just saw in the logs | 03:12 |
| midfavila | didn't read through for full disclosure :P | 03:12 |
| midfavila | but when my reform was still in action i got TinyX running with a simple one-line change in the source code | 03:12 |
| midfavila | it's not as full-featured as X.Org but for everyday stuff it works just fine | 03:12 |
| midfavila | i dont recall the source file specifically but it wasn't hard to figure out... there's a file that has a bunch of ifdefs for different architectures, modifying the line that checks for arm32 to also match against aarch64 with an or preproc statement gets it to build without trouble | 03:13 |
| - nsc (QUIT: Ping timeout: 265 seconds) (~nicolas@i5C74DD22.versanet.de) | 03:24 | |
| + nsc (~nicolas@87.122.72.126) | 03:30 | |
| - aloo_shu (QUIT: Ping timeout: 265 seconds) (~aloo_shu@90.166.98.198) | 04:05 | |
| + aloo_shu (~aloo_shu@90.166.98.198) | 04:15 | |
| + cobra (~cobra@user/Cobra) | 04:35 | |
| josch | midfavila: nah, i just care about just getting pixels to the screen -- doesn't matter whether that's via wayland or X :) | 07:22 |
| josch | minute: your soft-start charging patch turned a boolean expression around and i'm not sure whether that was a bugfix or introduced a bug. | 09:43 |
| josch | before your soft-start patch, disable_charge_current();set_discharge_bits(discharge_bits); was called if (state != ST_CHARGE || discharge_bits != 0) | 09:43 |
| josch | now those two functions get called if (state != ST_CHARGE && discharge_bits != 0) | 09:43 |
| minute | josch: hmm | 11:00 |
| * wielaard -> mjw | 11:16 | |
| - NanoCodeBug (QUIT: Ping timeout: 244 seconds) (~NanoCodeB@c-67-170-15-47.hsd1.wa.comcast.net) | 11:40 | |
| josch | you see the difference in 22d5c1b942b3d73fb2c4c968d7b38f6ee43af606 which changes !(state == ST_CHARGE && discharge_bits == 0) to (state != ST_CHARGE && discharge_bits != 0) | 12:11 |
| + gustav28 (~gustav@c-78-82-52-175.bbcust.telenor.se) | 13:02 | |
| - nocko (QUIT: Ping timeout: 248 seconds) (~nock@user/nocko) | 13:06 | |
| minute | josch: ok, will take a look later | 13:31 |
| minute | we just did an upgrade from a311d to rk3588 in pocket reform at home with anri (holo_memory) and it was quite interesting | 13:32 |
| minute | reform-flash-uboot has a few bugs | 13:32 |
| minute | it doesn't recognize the --offline argument that it says it does | 13:32 |
| minute | also the partition offset check failed in our scenario (it couldn't see the first (only) partition i made on emmc at 16mb offset) | 13:33 |
| minute | and we need a tool that, started from sd card, can help migrating from one cpu to another by setting up emmc boot partition and wiring up an existing nvme install to it | 13:33 |
| ch | --offline must be the first arg, then it should work (at least it used to) | 13:36 |
| josch | i'm interested in the partition offset bug -- do you have more details? | 13:39 |
| josch | specifically the output of this would be interesting: parted --script --machine /dev/mmc_yourdev unit B print free | 13:41 |
| + glu_ (~glu@user/glu) | 13:45 | |
| - glu (QUIT: Ping timeout: 252 seconds) (~glu@user/glu) | 13:46 | |
| * glu_ -> glu | 13:46 | |
| minute | josch: i don't have the device here anymore but i remember the script is grepping for something like 1024B, which wasn't there. there were some other numbers in the output | 13:59 |
| minute | weird, the following: sudo parted --script --machine /dev/mmcblk0 unit B print free | 14:01 |
| minute | outputs > Error: /dev/mmcblk0: unrecognised disk label | 14:01 |
| minute | but then it outputs something anyway: | 14:01 |
| minute | > /dev/mmcblk0:250148290560B:sd/mmc:512:512:unknown:MMC Y0S256:; | 14:01 |
| minute | but i don't currently have a partition on my system there | 14:01 |
| minute | josch: maybe you assume a specific partition layout? | 14:02 |
| minute | i had created a first partition manually at offset 16MB | 14:02 |
| josch | minute: what you pasted is the normal output when there is no partition table at all | 14:04 |
| minute | ok sure... that's on my device now | 14:05 |
| minute | i can create the same partition here | 14:05 |
| minute | and delete it after | 14:05 |
| minute | (parted) mklabel gpt | 14:06 |
| minute | (parted) mkpart primary ext4 16MB 1GB | 14:06 |
| minute | josch: ok, here's my output http://dump.mntmn.com/parted.txt | 14:07 |
| minute | josch: ah, for me the free space starts at 17408B | 14:08 |
| minute | i think that's what we're after? | 14:08 |
| josch | uh funny | 14:08 |
| josch | yes | 14:08 |
| minute | btw gnome-disks is not able to partition this disk at all | 14:08 |
| minute | but parted can, without any issues | 14:08 |
| josch | o0 | 14:08 |
| josch | what does gnome-disks have to say about it? | 14:09 |
| minute | ah well now i have a gpt table so now gnome-disks is ok... let my try to kill it | 14:09 |
| josch | maybe the difference is "mklabel gpt" versus "mklabel msdos"? gpt needs a bit more space at the beginning | 14:09 |
| josch | and currently, the sd-card images built by gitlab CI use msdos table, so i didn't try the script with gpt yet, i guess? | 14:10 |
| minute | josch: yes. wouldn't it make sense to grep for the ":free" type and sort? | 14:10 |
| josch | yes, lets ignore the 1024 | 14:10 |
| minute | hmm maybe not, if there's no free space at the beginning at all | 14:10 |
| josch | are you able to create such a situation? | 14:11 |
| minute | basically, free needs to be the first entry | 14:11 |
| josch | i tried and wasn't able to | 14:11 |
| minute | ah | 14:11 |
| minute | josch: yes, with gpt i don't have free space | 14:12 |
| josch | without the 1024B, the script needs to make sure that there are no other partitions in front of it | 14:12 |
| minute | josch: http://dump.mntmn.com/parted2.txt | 14:13 |
| minute | josch: wouldn't it make sense to look at just the first entry and make sure it is of type "free" and that it has the minimum size? | 14:13 |
| josch | or differently: look at the first line that is *not* free, which is the first partition, and make sure that it starts late enough | 14:14 |
| minute | also good | 14:14 |
| josch | while making sure that reform-flash-uboot does not error out if there is *no* partition table -- it's okay to flash u-boot to an empty disk | 14:15 |
| josch | and that reform-flash-uboot works if there are no partitions at all | 14:15 |
| minute | josch: correct | 14:16 |
| josch | thank you for your parted output -- that will help me test this | 14:16 |
| minute | cool :3 | 14:25 |
| josch | oh and indeed --offline is broken! | 16:04 |
| josch | minute: about a migration tool: this will be difficult i think unless you make a bunch of assumptions. Do you have a plan for what you'd like to see? | 16:05 |
| minute | josch: it would detect an existing install on nvme (encrypted, unencrypted, lvm, no lvm). mount it. chroot into it. mount /boot from emmc or sd. update /boot entry in fstab. apt update and install latest kernel package | 19:04 |
| josch | okay, so it would require internet access. Meaning that the rescue system on sd-card has to have the needed drivers/firmware installed for the wifi card if the user doesn't have ethernet connected. Yes, that would work. Complexity would probably result from all the heterogeneous setups. | 19:07 |
| minute | yeah the sd card system is just the system image, right... | 20:08 |
| minute | it mainly needs to support upgrading from imx8mp -> a311d, imx8mp -> rk3588 and from a311d to rk3588 | 20:09 |
| minute | ah, also s/imx8mp/imx8mq as well... | 20:09 |
| + NanoCodeBug (~NanoCodeB@c-67-170-15-47.hsd1.wa.comcast.net) | 20:14 | |
| josch | minute: another option would be to require the user to run "apt install --download-only" of their kernel before swapping out the module | 20:26 |
| josch | minute: reform-tools MR 108 has the fixes to reform-flash-uboot you reported earlier today -- you can try it out if you find some time :) | 20:26 |
| josch | this should support no partition table, partition table but no partitions, partition table but no free space etc | 20:27 |
| minute | josch: i wouldn't like to require that because most people won't read the instructions before just ripping out the module :D | 20:36 |
| josch | yes, makes sense | 20:37 |
| - BoostisBetter (QUIT: Remote host closed the connection) (4a410829d7@irc.cheogram.com) | 20:37 | |
| - SavagePeanut (QUIT: Remote host closed the connection) (59eaa45ac7@irc.cheogram.com) | 20:37 | |
| - Twodisbetter (QUIT: Remote host closed the connection) (2cc0e4ea1c@irc.cheogram.com) | 20:37 | |
| josch | i wonder if it makes sense to design the "swap your SoM" tool around a slightly different use-case, the "i screwed up my /boot plz repair" tool | 20:43 |
| josch | it would allow people to dd if=/dev/zero of=/dev/mmcblk0 and then still rescue their system afterwards with ease | 20:44 |
| josch | because essentially upgrading the som is similar to the scenario of "your /boot got wiped" | 20:45 |
| josch | just re-read this thread which has a script that "just" needs to be made pretty: https://community.mnt.re/t/migration-from-i-mx8mplus-to-a311d/2317 | 20:51 |
| josch | *I just re-read... | 20:52 |
| josch | or, thinking about the assumptions the tool has to make | 22:06 |
| josch | maybe it should be even broader | 22:06 |
| josch | instead of just a "please repair my /boot" tool, maybe it should be a "please repair my emmc, assuming it *should* contain /boot" tool? | 22:06 |
| josch | minute: can you confirm that the tool should assume /boot to be on emmc and that the previous content of the emmc should be wiped? | 22:07 |
| josch | and that the tool should assume / to be on the ssd, meaning that this is the setup: emmc: boot, ssd: / (maybe inside luks/lvm2 or plain), sd-card: the rescue system from which the migration is attempted | 22:13 |
| - gustav28 (QUIT: Quit: Quit) (~gustav@c-78-82-52-175.bbcust.telenor.se) | 22:15 | |
| minute | josch: correct | 22:15 |
| + SavagePeanut (59eaa45ac7@irc.cheogram.com) | 22:33 | |
| - mjw (QUIT: Ping timeout: 252 seconds) (~mjw@212.78.174.2) | 22:54 | |
| + BoostisBetter (4a410829d7@irc.cheogram.com) | 22:56 | |
| + mjw (~mjw@212.78.174.2) | 22:56 | |
| BoostisBetter | merry christmas / happy holidays everyone! | 22:56 |
| BoostisBetter | minute: if the lpc is not being responsive, is there some other way to get it to respond without power cycling the system? | 22:57 |
| - mjw (QUIT: Remote host closed the connection) (~mjw@212.78.174.2) | 23:39 | |
| + mjw (~mjw@212.78.174.2) | 23:40 | |
| + wielaard (~mjw@212.78.174.2) | 23:52 | |
| - mjw (QUIT: Read error: Connection reset by peer) (~mjw@212.78.174.2) | 23:52 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
