2024-01-01 18:23:43 Happy New Year \o/ 2024-01-02 09:38:11 One of the two arm servers is now on the way with nu 2024-01-02 09:45:41 Right, the CI server 2024-01-02 09:47:02 yup 2024-01-02 09:48:34 oh no, it's che-bld-1 2024-01-02 09:53:40 why oh no :D 2024-01-02 09:58:56 I first said it was the CI host 2024-01-02 09:59:01 che-ci-1 2024-01-02 10:00:46 is that a problem? 2024-01-02 10:05:16 numono: No, it's fine 2024-01-02 10:23:45 ah, just got now what u said^^" still sleepy~ 2024-01-02 10:24:10 We did initially plan for the ci host to be sent, but it's fine 2024-01-02 12:13:41 Does this mean the ARM CI runners are now offline? i've started seeing MR pipelines fail on ARM 2024-01-02 12:14:21 celie: No, they are still available 2024-01-02 12:15:30 Ok, but they seems to be having problems reaching the DNS server? 2024-01-02 12:16:30 I'm chekcing 2024-01-02 12:18:03 Oh right, there still exists a dependency on the build server 2024-01-02 12:20:06 Ok 2024-01-02 13:03:51 celie: CI should work again 2024-01-02 13:25:07 Thanks 2024-01-02 15:16:25 its running 2024-01-02 16:00:33 it should have v6 and configurable v4 2024-01-02 16:01:21 can anyone access it, and check if something needs to be modified? 2024-01-02 16:02:37 im on the site for ~1h 2024-01-02 16:03:00 What's the IP address 2024-01-02 16:04:19 I'm visiting family, so my availability is limited 2024-01-02 16:06:47 we can continue another time^^ 2024-01-02 16:07:40 the network i can change remotely, and the only unknown is how to give you access to the mgmt port (since i didnt bring any apu-s) 2024-01-02 16:07:51 i can wire it into a vm 2024-01-02 16:08:40 Right, we can discuss that with clandmeter 2024-01-02 16:08:42 ip is 2a02:168:a077:2002:da5e:d3ff:fee6:1f48 2024-01-02 16:08:51 same as before except the prefix 2024-01-02 16:09:09 no more cofe :P 2024-01-02 21:46:20 numono: I cannot ping / ssh yet 2024-01-03 09:52:14 numono: do you have ipv4 i can connect to? 2024-01-03 09:52:19 ikke: does dmvpn work? 2024-01-03 10:02:51 it was firewalled 2024-01-03 10:03:00 it might work now 2024-01-03 10:03:48 @clandmeter: do you not have v6? v4 needs to be confed on the builder 2024-01-03 10:04:06 nope no ipv6 2024-01-03 10:05:19 i can give you a wireguard tunnel if u want (to get v6) 2024-01-03 10:07:41 what do you mean with builder? the actual box? 2024-01-03 10:08:21 yes 2024-01-03 10:08:35 ok understand 2024-01-03 10:08:47 i dont run dhcp atm 2024-01-03 10:10:41 ok it works now 2024-01-03 10:11:02 do you need me to add ipv4? 2024-01-03 10:11:38 unless you need to 2024-01-03 10:12:00 what does that mean? 2024-01-03 10:12:06 ipv6 nat/dns64 are available to reach v4 2024-01-03 10:12:46 so unless you have special usecases with legacy ip, then you dont need to configure it 2024-01-03 10:14:05 ipv6 only has caused issues before 2024-01-03 10:14:26 try ping6 github.com 2024-01-03 10:14:26 to the world it has only a A entry, but in the builder the command will work 2024-01-03 10:14:57 (if u take my dns ofc from the router advertisements :P) 2024-01-03 10:15:10 (rdnssd daemon can auto update resolv.conf) 2024-01-03 10:15:21 its trying to use the local ipv4 config 2024-01-03 10:15:38 want to give it a try if it works with v6 only? 2024-01-03 10:15:56 i dont mind ipv6 only, if it works. 2024-01-03 10:16:21 ikke bumped into issues before, thats what im echoing here now. 2024-01-03 10:16:28 lets wait for him to comment on it 2024-01-03 10:16:55 but in any case, thanks for setting this up :) 2024-01-03 10:17:06 we need to adjust the config anyways 2024-01-03 10:17:41 idk if u want to spend the time to invest into it again, if yes it would be nice, if not i can just give a v4 static gw/ip 2024-01-03 10:17:54 lmk^^ 2024-01-03 10:18:40 ikke did the network config, im not sure what he did. I think its related to the router in front of it before. 2024-01-03 10:18:44 just curious, did you use the ip i sent you earlier to access it, or was it through a dmvpn ip? 2024-01-03 10:18:54 your ipv6 2024-01-03 10:18:57 niiice 2024-01-03 10:19:16 but eth0 has an ipv4 config 2024-01-03 10:19:33 which is probably the subnet of the router at prev location 2024-01-03 10:19:45 probably static aswell :D 2024-01-03 10:20:20 ofc 2024-01-03 10:20:29 which is perfect if things dont change :p 2024-01-03 10:21:01 dw the vlan is dedicated to you 2024-01-03 10:21:13 dw? 2024-01-03 10:21:17 dont worry 2024-01-03 10:21:38 im almost 50, be gentle with slang ;-) 2024-01-03 10:23:06 i think we may need ipv4 for dmvpn iirc 2024-01-03 10:23:22 some parts of dmvpn is not ipv6 compar 2024-01-03 10:23:23 t 2024-01-03 10:23:25 would be nice to explore the v6 only path, id be motivated to resolve any issues that may arise from it 2024-01-03 10:23:25 its much simpler to maintain a single stack on the infra 2024-01-03 10:24:32 if needed the static v4 is ip: 10.159.1.2/24, gw: 10.159.1.1 2024-01-03 10:24:55 from what i understand fixing dmvpn it non trivial 2024-01-03 10:34:55 I think I have to chime in here... from what I can see is that IPv6 only is more and more the default and maintaining dual stack is indeed very cumbersome. I have heard in here about the dmvpn issues before and if it is not fixable, it might be a good route to take the IPv4 hack at nu 's location, IPv4 in place5 is also just a hack/add-on, as all other networks are IPv6 only. Mid term I would suggest to look for a replacement, either 2024-01-03 10:34:55 wireguard or openvpn based, both work fine IPv6 only 2024-01-03 11:09:19 thank you! 2024-01-03 11:10:20 i think we need ipv4. my isp is not making it easy with ipv6 for me 2024-01-03 11:11:22 ncopa: I think both nu and I can easily offer IPv6 tunnels that you can take with you to anywhere, a simple `wg-quick up wg0` will give you full IPv6 access 2024-01-03 11:11:33 Sadly many isps, docker and GitHub did not receive the message yet 2024-01-03 11:12:30 ikke: docker is ipv6 for about 3-6 months now, they dropped the ipv6. registry and enabled IPv6 on the main registry some time ago 2024-01-03 11:12:39 For github, I agree and I fight with their staff about 3 times a year 2024-01-03 11:14:04 my isp gives me a /48 prefix. But thats the only prefix I get. I have a ISP provided router (which looks nice) which grabs this prefix. I use this for the home wifi. For my work network I use ethernet and a linux router (with dmvpn). 2024-01-03 11:14:20 since the isp router takes the ipv6 prefix I cannot get any for my linux router 2024-01-03 11:14:58 and there is no way to turn that off on the ISP provided router 2024-01-03 11:15:12 that would be very strange, you should be either able to route it or to get it via dhcp 2024-01-03 11:15:33 But either way the proposed tunnel above would work, both for a single device or even for you whole home network 2024-01-03 11:15:34 the ISP provided router gets it via dhcp6 2024-01-03 11:15:53 ok. Maybe I should set that up then 2024-01-03 11:15:58 oh, then you might be able to get a sub prefix via dhcp from that router as well 2024-01-03 11:16:05 it's callled prefix-delegation 2024-01-03 11:16:20 I would assume they do that, because otherwise giving a whole /48 is pretty senseless 2024-01-03 11:16:21 i think that ISP router give a /64 prefix 2024-01-03 11:17:13 I am not fully understanding your setup, but if your ISP assigns a /48 to you, then usually the idea is that you can use /64's for your own management / own networks 2024-01-03 11:17:27 but that means that my work net would be behind the home net. I'd prefer to have it side by side, so they are completely independent 2024-01-03 11:17:58 Something needs to route the complete/64 2024-01-03 11:18:11 /48* 2024-01-03 11:18:36 i believe it will work if i put my work net behind the ISP router 2024-01-03 11:18:48 i mean, my work net would get a /64 2024-01-03 11:20:17 but I would ideally want to disable ipv6 on the ISP provided router, and use my alpine router for ipv6 and dmvpn etc. connected on the same switch. so they are separated. two different public ipv4 2024-01-03 11:20:18 ikke: Usually ("by the book") you design it like this: /48 goes to a location/customer, that should be terminated on the ISP's given equipment 2024-01-03 11:20:46 Then the ISP router should give out /64's either by routing (manual settings in the router) or by dhcp prefix delegation (if it is a more and end consumer device) 2024-01-03 11:21:15 ncopa: out of curiosity, where are you based at? 2024-01-03 11:21:23 norway. 2024-01-03 11:21:32 ISP is telenor 2024-01-03 11:21:44 ah, the country with 80% evs, pretty impressive 2024-01-03 11:21:58 yeah, maybe I should rewire the work net to go behind the ISP provided router 2024-01-03 11:22:39 Hmm, telenor looks good in what they do in general: https://stats.labs.apnic.net/ipv6/NO 2024-01-03 11:23:30 they do. except when it comes to documentation how the ipv6 works 2024-01-03 11:23:37 i kinda figured it out using tcpdump :) 2024-01-03 11:23:59 and then someone working at telenor confirmed it on a norwegian forume 2024-01-03 11:24:00 Do you know what kind of router they gave you? 2024-01-03 11:24:32 yup. this one: https://www.telenor.no/kundeservice/internett/modem-og-ruter/wifi-ruter/ 2024-01-03 11:24:45 it looks nice on the TV bench :) 2024-01-03 11:26:06 If I say that norwegian looks similar dutch, will I get kicked? 2024-01-03 11:26:22 It is actually almost readable for a German speaker (German and Dutch having similar roots) 2024-01-03 11:26:48 you wont get kicked for telling the truth :) 2024-01-03 11:26:52 "Koble opp WiFi Ruter" 2024-01-03 11:27:08 norwegian and duch is similar 2024-01-03 11:27:09 That could also pass as a Swiss dialect 2024-01-03 11:27:33 Ah, are they. I was a bit in Finnland, many years ago, which seems to be pretty much different to almost everything else 2024-01-03 11:27:53 So I was assuming Norwegian would also be far off 2024-01-03 11:28:04 there is a switch in the UI to disable ipv6. but it only prevents it from forwarding the ipv6 stuff on the LAN side 2024-01-03 11:28:18 tcpdump revealed that it still asks for the /48 prefix 2024-01-03 11:28:35 I'm actually wondering if that device is internally openwrt 2024-01-03 11:29:33 not sure. I sent a bug report/feature request to their development team. But I dont expect any answer 2024-01-03 11:47:28 My parents isp (local provider) does not support it yet, it seems 2024-01-03 11:52:29 my office isp same thing, need to purchase it separately by some special request. 2024-01-03 11:53:20 do we need ipv4 for other things then dmvpn? github can work with dns64? 2024-01-03 11:54:07 clandmeter: github works with dns64+nat64, yes. We use it everyday like that 2024-01-03 11:54:39 i think we ran into issues even with dns64 2024-01-03 11:54:58 not sure ikke can elaborate on this topic, he hit most of the issues. 2024-01-03 11:56:09 ikke: i guess the GRE issues with ipv6 are still existing? 2024-01-03 11:56:44 dns64 basically only kicking in, if you use dns 2024-01-03 11:57:04 If there are hardcoded IPv4 addresses, they will fail 2024-01-03 12:00:24 nod 2024-01-03 12:00:50 good and flawless solution for this mess doesn't exists yet and I could bet will not appear soon 2024-01-03 12:02:17 ACTION follow IPv6 from 1996 in a hope that it will be fixed soon 2024-01-03 12:05:41 Linux support for multipoint GRE which is used by dmvpn is still missing 2024-01-03 18:31:22 ikke: x86_64 CI out of space? 2024-01-03 20:23:49 ptrc: it's not completely out of space, but large builds would probably fail, will clean up 2024-01-03 20:47:18 numono: I have access to the server now 2024-01-03 20:53:13 numono: Is there some public ipv6 prefix that we can use? The containers right now receive a public ip from the previous /48 prefix 2024-01-04 10:00:24 sure, ive routed 2a02:168:a077:700::/56 to you 2024-01-04 10:19:25 numono: thanks 2024-01-04 10:19:41 np^^ 2024-01-04 17:45:18 ikke: are the ppc64le builds like the one reported here: https://sourceware.org/bugzilla/show_bug.cgi?id=30824 running on power10s? 2024-01-04 17:52:17 PureTryOut: ping 2024-01-04 17:52:25 pong 2024-01-04 17:52:31 hello 2024-01-04 17:52:49 could you help to add lv2 to pipewire 2024-01-04 17:53:24 I'm trying to debug it on #asahi-alt channel for asahi audio 2024-01-04 17:53:40 lv2? Not sure what that is πŸ€” 2024-01-04 17:53:53 some people from other distros to help there 2024-01-04 17:53:53 But probably very relevant for me since I'll be getting a M2 macbook from work soon 2024-01-04 17:54:40 lv2 plugins/filters 2024-01-04 17:54:41 any particular issue though? Is it something pipewire doesn't support yet? 2024-01-04 17:55:32 I build latest pipewire with -Dlv2=enabled, and it builds but missing some filters 2024-01-04 17:56:18 if you mind to join #asahi-alt there are people who understand these things (which I don't understand) 2024-01-04 17:56:43 not sure why you think I will understand but sure lol. What network? Just OFTC? 2024-01-04 17:56:54 (or is there a native Matrix room?) 2024-01-04 17:56:55 OFTC 2024-01-04 17:57:24 you are PW maintainter and I thought you know these things 2024-01-04 17:57:43 Annnndddd I'm banned from that channel 2024-01-04 17:57:51 I guess they don't allow bridging. SOL then 2024-01-04 17:57:59 iirc matrix doesn't work on asahi channels 2024-01-04 17:58:18 you are right, there is no bridge 2024-01-04 17:58:21 I'm not that deep into PipeWire knowledge tbh 2024-01-04 17:58:39 ah ok, and sorry then 2024-01-04 18:00:34 PureTryOut: but if you plan to run alpine or pmOS on M2 you will probably have to learn more about PW, as I'm trying now ;-) 2024-01-04 18:00:58 probably will haha 2024-01-04 18:02:06 FYI, I got working speakers on apple silicon but can't get to fix these filters 2024-01-04 18:08:46 ah, looks like I found solution 2024-01-04 18:25:55 PureTryOut: I would like to post patches for pipewire and wireplumber aports, i.e. create MR. is that ok for you? 2024-01-04 18:26:36 Yeah ofc, we can always discuss on the MR 2024-01-04 18:29:52 ok 2024-01-04 18:36:22 mick_ibm: our ppc64le server is power9 2024-01-04 18:41:12 telmich: vigir23 is unreachable again since last night 2024-01-04 20:10:07 PureTryOut: !58402 2024-01-04 20:11:45 Yeah I saw. That's easy enough 2024-01-04 20:12:27 thanks 2024-01-04 20:13:02 Btw in preparation for the incoming macbook, got any instructions on how to install Alpine on it? I can't find it on the wiki 2024-01-04 20:13:21 (easy after I learned by 'try-and-fail' method ;-) ) 2024-01-04 20:14:32 maybe we will need something for wireplumber also but will leave it to test tomorrow 2024-01-04 20:16:15 only not sure how to 'alarm' users that they must enable speakersafetyd on boot 2024-01-04 20:17:16 btw, do you know what controls 'Night colors' in KDE? 2024-01-04 20:17:29 PureTryOut: ^ 2024-01-04 20:20:24 Eh, just kwin I guess? 2024-01-04 20:21:13 thanks. but too much to find how to make separate program from it 2024-01-04 20:21:37 just checked, yeah kwin 2024-01-04 20:21:42 Why would you do that...? 2024-01-04 20:21:46 FYI, Night Color works fine on apple silicon with KDE plasma 2024-01-04 20:23:18 I don't like KDE and any other big DE so wanted to see if I can make simple utility for 'barebones' install 2024-01-04 20:23:55 ah ok, better write something from scratch probably 2024-01-04 20:24:04 anyway do you have instructions somewhere to install Alpine on Apple silicon? 2024-01-04 20:25:11 yes, https://arvanta.net/alpine/install-alpine-m1/ 2024-01-04 20:25:37 but it is outdated somewhat now, have to find time to update it 2024-01-04 20:26:15 you can always ask me on IRC if you decide to install 2024-01-04 20:26:44 and https://arvanta.net/alpine/upgrade-kernel-m1/ 2024-01-05 22:12:06 could someone reset pkgrel for telegram-desktop, it is 2 for few last upgrades 2024-01-05 22:12:48 uh, sorry, wrong channel 2024-01-05 22:20:50 Fyi, resetting pkgrel after the fact has no use and even can cause issues 2024-01-05 22:22:44 I know 2024-01-07 10:23:40 numono: thanks (re ipv6 prefix) 2024-01-07 10:35:25 welcome:> 2024-01-07 10:59:15 it works :) 2024-01-07 21:26:40 Hey, it seems that alpine gitlab isn't recieving my emails anymore. I can send emails to other fine though, can someone check please? My emails all come from matthiasahouansou.cz 2024-01-07 21:26:50 matthias@ahouansou.cz 2024-01-07 21:30:08 Kladky: when was the last mail you sent? 2024-01-07 21:30:40 6-7 mins ago: 21:24 UK time 2024-01-07 21:31:02 I see, postfix accepts it 2024-01-07 21:31:23 but? 2024-01-07 21:32:07 Trying to see if I can find more logs 2024-01-07 22:00:46 Kladky: I see the mail in a queue, but no idea yet why it's not being processed 2024-01-07 22:01:33 Odd, can other people send emails to gitlab or is it just me having this issue? 2024-01-07 22:03:14 I see more e-mails in the queue 2024-01-07 22:03:38 First one is from december 24 2024-01-07 22:05:35 I first started experiencing this issue on the 28th, so it probably is that. 2024-01-07 22:10:36 Do you mind opening an issue here? https://gitlab.alpinelinux.org/alpine/infra/ 2024-01-07 22:14:38 might be related to the last gitlab upgrade 2024-01-08 11:31:20 would someone review !58442 2024-01-08 11:31:31 nmeum: ncopa: ^ 2024-01-08 14:27:30 what do you think about creating a channel for #alpine-loongarch64? https://gitlab.alpinelinux.org/alpine/tsc/-/issues/72#note_367311 2024-01-08 14:28:27 ncopa: thanks for merging eudev 2024-01-08 14:28:47 now we have speakers working on apple silicon 2024-01-08 14:30:15 ncopa: I remember that you told earlier that you think to work on asahi installer for alpine. I started to learn it few days ago and now preparing base things 2024-01-08 14:31:03 still don't understand all details but hope to have first testing on next weekend 2024-01-08 17:01:47 ikkie: done #10813 2024-01-08 17:01:58 https://gitlab.alpinelinux.org/alpine/infra/infra/-/issues/10813 2024-01-09 10:09:57 I've got asahi installer to install alpine on apple silicon but don't know how to make EFI/BOOT/BOOTAA64.EFI to find root partition 2024-01-09 10:11:12 other distros use some complicated shims and similar things but I think we don't have such things in alpine 2024-01-09 10:12:10 searched net but couldn't find any hints to lead me to solution 2024-01-09 10:12:33 do anyone of you have any idea how this can be solved 2024-01-09 10:19:03 It is usually detected with grub-install? 2024-01-09 10:21:27 I think grub .efi finds the grub.cfg where kernel and kernel options are found. Root partition is boot option for kernel and used by initramfs 2024-01-09 10:21:56 ncopa: yes, in our normal install. but on apple silicon it couldn't be run with asahi installer 2024-01-09 10:22:55 boot partition number is 'embeded' in EFI/BOOT/BOOTAA64.EFI during grub install in normal cases 2024-01-09 10:23:27 for asahi installer we have to prepare EFI/BOOT/BOOTAA64.EFI in advance 2024-01-09 15:46:54 minimal: ping 2024-01-09 15:58:14 mps: Hi 2024-01-09 15:59:01 minimal: sorry to bother you but I remember that you have good knowledge about grub and I need help 2024-01-09 15:59:17 problem after 2.12 update? 2024-01-09 15:59:54 He was talking about Asahi earlier 2024-01-09 15:59:56 ah, it is updated, didn't noticed 2024-01-09 16:00:59 minimal: I'm trying to build asahi installer for alpine and last problem I have is embeding root partition in BOOTAA64.EFI 2024-01-09 16:01:44 no matter what I try grub-install embeds '(,gpt6)/boot/grub' 2024-01-09 16:01:47 embedding? not sure what you mean 2024-01-09 16:02:38 grub-install embeds root partition in BOOTAA64.EFI, where to search /boot/grub 2024-01-09 16:02:57 you mean how to tell grub-install where to copy the modules and config? 2024-01-09 16:03:26 you are specifying "--boot-directory=/boot" for grub-install? 2024-01-09 16:03:27 no, to set proper partition number in BOOTAA64.EFI 2024-01-09 16:04:33 normally you run "grub-install --efi-directory=/boot/efi --boot-directory=/boot" and that tells grub-install what it needs to know 2024-01-09 16:05:00 the "--efi-directory" option tells it where the ESP filesystems is (for it to place EFI files) 2024-01-09 16:05:20 yes, I have this 2024-01-09 16:05:40 the "--boot-directory" option tells it where the grub.cfg is to be found and where (in a subdirectory) it should copy the Grub modules etc 2024-01-09 16:05:41 `grub-install --target=arm64-efi --efi-directory=/boot/efi --boot-directory=/boot` 2024-01-09 16:06:19 all this works fine, except '(,gpt6)/boot/grub' 2024-01-09 16:06:49 grub-install amongst other things runs grub-mkimage as part of assembling the EFI file from components like the modules that need to be "built-in" 2024-01-09 16:08:04 run grub-install with "--verbose" so see exactly what it is doing 2024-01-09 16:08:55 aha, ok 2024-01-09 16:10:51 here is complete log from my install script https://tpaste.us/vNDk 2024-01-09 16:11:32 grub-mkimage takes a "--prefix=DIR" option which seems to related to "PATH" you're talking about (i.e. the manpage for grub-mkimage states that "-m implies -p (memdisk)/boot/grub" but any "-p" option should be passed to grub-mkimage by grub-install based on the "--boot-directory" you specify AFAIK 2024-01-09 16:12:34 there you go, search for the grub-image command in that output and see grub-install passes it "--prefix '(,gpt4)/boot/grub'" 2024-01-09 16:13:40 what's the 3rd line message regarding "old partition table" about?? 2024-01-09 16:14:00 you changed partitioning in some way but the kernel isn't "seeing" the changes? 2024-01-09 16:14:12 sgdisk message 2024-01-09 16:14:43 here is script which I use to create image https://tpaste.us/QRxw 2024-01-09 16:15:36 in your output it shows "grub-install --target=arm64-efi --efi-directory=/boot/efi --removable" - you didn't specify "--boot-directory" (not sure what it defaults to) 2024-01-09 16:16:22 also you're doing this with a loop device rather than a real disk... 2024-01-09 16:18:18 oh, maybe wrong output posted 2024-01-09 16:19:03 ah this is left cruft in 'echo' 2024-01-09 16:19:38 real command is `grub-install --target=arm64-efi --efi-directory=/boot/efi --boot-directory=/boot --verbose' 2024-01-09 16:20:44 but now I see file 'EFI/alpine/grubaa64.efi', for what it is used 2024-01-09 16:21:52 it's creating grubx64.efi as you removed the "--removable" flag otherwise it would create bootx64.efi 2024-01-09 16:23:10 could this file be used instead of BOOTAA64.EFI 2024-01-09 16:24:23 I see that fedora for asahi have booth plus some shims 2024-01-09 16:25:33 both* 2024-01-09 16:26:07 hm you gave me some ideas to test 2024-01-09 16:26:19 minimal: thank you for the help 2024-01-09 16:30:34 mps: let me look at one of the grub-install logs I have for a similar UEFI/GPT grub-install on loop device 2024-01-09 16:31:07 mps: if you are creating a (intended to be) removable EFI media then you need to use "--removable" 2024-01-09 16:31:22 as otherwise it won't be bootable 2024-01-09 16:32:00 EFP/BOOT/BOOTX64.EFI is the "fallback" EFI file for booting, when there are no suitable EFI boot variables to tell it how to boot 2024-01-09 16:32:23 EFI boot variables are stored on each *machine*, not on the boot media 2024-01-09 16:34:50 correction, you need to use both "--no-nvram" and "--removable" together, the 1st to not create EFI boot vars (as they're "useless" for removable media), the 2nd to create a BOOTX64.EFI file rather than a GRUBX64.~EFI/ALPINEX86.EFI/whatever.EFI file as there's no way to find a GRUBX64.EFI file *without* boot variables, only the fallback file can be founf 2024-01-09 16:37:04 also I believe you should run grub-mkconfig BEFORE grub-install 2024-01-09 16:38:02 as I believe grub-install looks at the /boot/grub/grub.cfg file to "inform it" of what Grub modules are required to be embedded into the generated EFI file 2024-01-09 16:39:46 minimal: it is for internal nvme on apple silicon, I have long time one which runs from removables 2024-01-09 16:39:52 e.g. if you /boot is encrypted then the EFI file needs to have "luks" or "luks2" module embedded into the EFI file otherwise it has no way to access the encrypted /boot filesystem and AFAIK grub-install looks at the grub.cfg to figure out that /boot is encrypted 2024-01-09 16:40:20 no, for now I try to get simple instaleer 2024-01-09 16:40:34 I was assuming removable based on the "m1-usb.img" filename used for the loop device 2024-01-09 16:40:54 no? no what? I don't understand 2024-01-09 16:41:13 filename is 'artefact' from time when I built removables 2024-01-09 16:41:51 no luks, lvm and other complicated setups 2024-01-09 16:44:33 encryption was just an example to make my point of why I believe grub-install reads grub.cfg (not 100% sure of that as grub-install does also run things like grub-probe) and therefore why I recommended you run grub-mkconfig before it 2024-01-09 16:44:52 I run grub-mkconfig before grub-install in my own script 2024-01-09 16:47:39 grub is 'strange' land for me and overcomplicated IMO 2024-01-09 16:48:58 "it is what it is" 2024-01-09 16:50:16 heh 2024-01-09 16:51:16 any the "--prefix '(,gpt4)/boot/grub'" looks correct as partition 4 is the rootfs containing the /boot/grub directory 2024-01-09 16:51:21 why would it be gpt6? 2024-01-09 16:52:03 it shouldn't but it is 2024-01-09 16:52:17 that is what I asked about 2024-01-09 16:54:59 in the output you provided it is gpt4 as I quoted above 2024-01-09 16:55:04 only what I can tell on fedora this is very diffent 2024-01-09 16:55:40 in grubaa64.efi it is 4 2024-01-09 17:01:23 I guess I could solve this by installing grub on ESP but think it is not good option 2024-01-09 17:01:24 well that's the only output you provided me with to look at... 2024-01-09 17:02:00 IDK what else is needed 2024-01-09 17:02:03 you ARE installing grub (EFI file) on ESP, the rest of Grub however is on the rootfs 2024-01-09 17:02:14 right 2024-01-09 17:03:14 so do you have output showing exactly what you want? do you want grubaa64.efi or bootaa64.efi? do you want to use EFI boot variables? I'd assume so if you want the Mac dual-boot 2024-01-09 17:03:41 if I know I will answer 2024-01-09 17:04:01 I want this to be bootable after install 2024-01-09 17:04:40 are you creating a disk image to later copy onto the Mac? or installing *directly* onto the Mac? 2024-01-09 17:05:34 no, install over http/s 2024-01-09 17:06:05 I already have disk image for installation 2024-01-09 17:06:21 https://arvanta.net/alpine/install-alpine-m1/ 2024-01-09 17:06:44 hah, now nearly 2 years old 2024-01-09 17:11:55 minimal: thanks again for help, I must go out for about one hour so will be AFK 2024-01-09 17:12:11 mps: so you ARE using a disk image? 2024-01-09 17:12:52 asahi installer is complicated thing because apple is complicated 2024-01-09 17:13:55 installer download from net some files and from macos creates partitions and copy files and 'dd' rootFS 2024-01-09 17:14:08 still doesn't change that I need a precise answer to my question before I can help further 2024-01-09 17:14:49 so you're creating the partitions directly on the Mac, not creating a partitioned disk image (like the script you showed me does)? 2024-01-09 17:15:38 question is simple, why grub doesn't set gpt number I want in BOOTAA64.EFI 2024-01-09 17:15:46 and therefore your runnnig grub-install *directly* on the Mac and not as part of any disk image prepation (which that script does) 2024-01-09 17:16:21 no, I have to prepare all these files on other machines and put them on the web 2024-01-09 17:16:40 I cannot attempt to answer the question if I don't understand precisely what you're doing and have grub-install verbose output to look at 2024-01-09 17:17:00 it grub-install runs on M1 machine it does correct things 2024-01-09 17:17:02 "prepare all these files on other machines" - including the Grub EFI file? 2024-01-09 17:17:15 yes 2024-01-09 17:17:33 so you ARE preparing a partitioned disk image to dd onto Mac? 2024-01-09 17:18:14 not disk image, file and root partition image 2024-01-09 17:18:14 or you arre preparing a partitioned disk image to create EFI file and then throwing aware the partitioned disk image once you extract the EFI file from it? 2024-01-09 17:18:22 s/aware/away/ 2024-01-09 17:19:57 in script I posted I created disk image, then copy root partition as image and files from ESP to upload in zip file 2024-01-09 17:21:34 and for now there is no even short guide how asahi installer works, I had to look and guess 2024-01-09 17:21:55 and ask on IRC 2024-01-09 17:23:53 so you are preparing a partitioned disk image to create EFI (and other) files and then throwing away the partitioned disk image once you extract the EFI file from it 2024-01-09 17:24:28 right 2024-01-09 17:24:35 why?? either (a) create a disk image like your script currently does and "dd" it onto the storage device, or (b) set up Grub directly on the Mac 2024-01-09 17:25:01 this is what done 2 years ago 2024-01-09 17:25:23 you're taking a EFI that Grub created BASED ON THE PARTITIONING LAYOUT OF THE DISK IMAGE and then putting it onto an already partitioned Mac and wondering why Grub is "confused" 2024-01-09 17:25:38 it is described here https://arvanta.net/alpine/install-alpine-m1/ 2024-01-09 17:26:40 well, not sure if proper word is confused, I doesn't set proper gpt number in BOOTAA64.EFI 2024-01-09 17:28:17 that article do not default copying any Grub EFI file onto the Mac, it shows grub-install being run on the Mac itself 2024-01-09 17:28:35 right 2024-01-09 17:28:37 s/default/detail/ 2024-01-09 17:29:13 so it is not "this is what done 2 years ago" - you didn't copy the EFI but that is what you are talking about now 2024-01-09 17:29:18 I wanted to tell that I solved problem with intall using removable image earlier 2024-01-09 17:29:21 so it is not the same 2024-01-09 17:29:34 sorry for confusion 2024-01-09 17:29:36 I'm totally confused as to what you're trying to achieve 2024-01-09 17:30:36 I'm trying to 'force' grub to embed desired gpt number in BOOTAA64.EFI 2024-01-09 17:31:10 I know this could be done by hex editor 2024-01-09 17:31:47 and I don't need all these things for me, I'm trying to make asahi installer for alpine users 2024-01-09 17:32:00 when (a) running grub-install directly on the Mac to create the BOOTAA64.EFI file? or (b) when running grub-install on a loopback device to generate BOOTAA64.EFI that you are then going to copy across onto the Mac? 2024-01-09 17:32:40 well, I want to make and publish these scripts for alpine users 2024-01-09 17:32:47 a or b? 2024-01-09 17:33:04 (b) for now 2024-01-09 17:33:26 I don't believe (a) is a valid thing to do 2024-01-09 17:33:45 oops, I meant I don't believe (b) is a valid thing to do 2024-01-09 17:34:07 as the partitioning of the loop device disk image is not 100% identical as the partitioning of the Mac 2024-01-09 17:34:32 for alpine (b) should be option in future 2024-01-09 17:34:43 for example is the Mac's NVME 512 or 4K sectors? 2024-01-09 17:35:06 asahi installer does ok partitioning 2024-01-09 17:36:34 minimal: sorry, I really must go now 2024-01-09 17:36:40 I don't know what "asahi installer" is and so I cannot comment on it 2024-01-09 17:37:08 basically it sounds like you're trying to do something that grub-install isn't really designed to handle 2024-01-09 17:37:35 and whilst they may be other alternative ways to achieve it I do not think to do so is a good idea 2024-01-09 18:35:35 minimal: sorry, I was in a hurry and forgot to tell that I very appreciate your willing to help 2024-01-09 18:37:01 I see that some other distros (debian, fedora ...) solved this problem somehow and had a hope that we also can 2024-01-09 18:37:27 I really don't understand why you do not want to run grub-install on the Mac itself as part of installation 2024-01-09 18:38:04 because it is not possible with asahi installer, at least for now 2024-01-09 18:38:13 https://github.com/AsahiLinux/asahi-installer 2024-01-09 18:39:05 also debian guide and some scripts here https://git.zerfleddert.de/cgi-bin/gitweb.cgi/m1-debian/ 2024-01-09 18:40:30 your method is creating a EFI based on the particular "setup" (i.e. partitioning) of the disk image which is different to that of the Mac 2024-01-09 18:40:44 right 2024-01-09 18:40:57 so you could try setting up the disk image to look as close to the Mac's partitioning as possible and that then may work 2024-01-09 18:41:08 I did 2024-01-09 18:41:24 your disk image has 4 partitions, the Mac appears to have at least 6 2024-01-09 18:41:39 but problem is 'embeded' gpt root partition number 2024-01-09 18:42:01 sigh, that is passed on the partitioning grub seeing when you run grub-install 2024-01-09 18:42:11 yes 2024-01-09 18:42:31 if the disk image's partitioning looks like the Mac's partitioning then the generated EFI should have whatever the "correct" value is 2024-01-09 18:42:57 so it seems like the disk image partitioning is NOT like that of the Mac 2024-01-09 18:43:02 asahi installer creates some extra partitions in container, two at least, ESP and one for rootFS 2024-01-09 18:43:15 container? 2024-01-09 18:43:27 macos term 2024-01-09 18:43:42 I don't know what that means 2024-01-09 18:43:54 container can contain partitions 2024-01-09 18:44:09 so, again, you need to make the disk image partitioning reflect the reality of the Mac partitioning 2024-01-09 18:44:33 I did 2024-01-09 18:44:37 so then why aren't you creating this "magical" container in the disk image also? 2024-01-09 18:45:05 on install installer creates ESP as partition 4 and root as 5 2024-01-09 18:45:32 and I created same on image 2024-01-09 18:45:48 whereas your script creates ESP as part 3 and rootfs as part 4 2024-01-09 18:45:58 so that is NOT the same 2024-01-09 18:46:25 I expected that grub-install will embed 5 as gpt boot partition where it have to search for /boot/grub 2024-01-09 18:46:35 and did you create this "container" also? 2024-01-09 18:46:59 minimal: this 3 and 4 were for testing, really I use 4 and 5 2024-01-09 18:47:04 why would grub-partition embed part 5 as boot partition when your script didn't crate a 5th partition? 2024-01-09 18:47:35 so can you show me the "real" script you use to create a disk image and the verbose grub-install output from it? 2024-01-09 18:47:49 no matter how I set it for image creation it always set 6 2024-01-09 18:48:32 if I can't see what I asked for then I don't know how I can help further 2024-01-09 18:48:37 here is script https://tpaste.us/KE8B 2024-01-09 18:49:37 so you are creating partitions 4 & 5.........but no partitions 1-3? 2024-01-09 18:50:00 and here is log https://tpaste.us/ogYL 2024-01-09 18:50:13 only 4 and 5 are needed 2024-01-09 18:50:30 but then thew partitioning arrangement does not reflect that of the Mac 2024-01-09 18:51:49 right, and I think they are not needed 2024-01-09 18:53:06 but ok, I will look at debian grub source when I find some free time 2024-01-09 18:53:33 obviously alpine grub can't do this 2024-01-09 18:53:51 I do not believe it is anything specific to Alpine 2024-01-09 18:54:05 but we seem to be going around in circles... 2024-01-09 18:54:28 one of debian developer told me earlier their grub works thanks to some patches 2024-01-09 18:55:23 in which case then it is NOT a case of "obviously alpine grub can't do this" but rather "GRUB cannot do this unless you add some special patches" 2024-01-09 18:56:33 better wording, I agree 2024-01-09 18:57:32 also how did you decide to use "--set-alignment=2" for sgdisk? 2024-01-09 18:58:50 yes, this is good in most cases 2024-01-09 18:59:11 how did you determine that you needed to use that? 2024-01-09 18:59:57 by creating images for riscv starfive visionfive boards 2024-01-09 19:00:23 and you checked their storage devices logical and physical sector sizes? 2024-01-09 19:00:32 without it they sometimes doesn't boot 2024-01-09 19:00:32 and you did the same for the Mac NVME? 2024-01-09 19:00:52 yes, I tested all these 2024-01-09 19:01:06 you *tested* or you check the values I mentioned? 2024-01-09 19:01:14 s/check/checked/ 2024-01-09 19:01:37 don't understand, what is difference 2024-01-09 19:02:02 testing implies you just tried different values without understanding what is the correct value 2024-01-09 19:02:16 checking is determining what is the correct value 2024-01-09 19:02:31 a "working" value is not necessarily the correct value 2024-01-09 19:03:00 storage devices are typical either 512/512, 512/4K or 4K/4K 2024-01-09 19:03:24 hm 2024-01-09 19:03:30 I know this 2024-01-09 19:03:46 partitioning should be correctly aligned based on each device's logical sector size 2024-01-09 19:04:17 in asahi installer this is not important 2024-01-09 19:04:57 only important is to make root FS with 4096 block size 2024-01-09 19:05:21 nope, it is important in general 2024-01-09 19:05:54 if you have a device with 4K logical sector size (i.e. some NVMEs) and your partitions are not 4K-aligned then it can affect performance 2024-01-09 19:05:57 hm 2024-01-09 19:07:28 yes, this parameter influence performance but not important for test if things works 2024-01-09 19:07:53 an old article: https://lwn.net/Articles/377895/ 2024-01-09 19:08:04 "But that goes badly wrong if the partition itself is not properly aligned; in this case, every carefully-arranged 4KB block will overlap two physical sectors - hardly an optimal outcome." 2024-01-09 19:18:27 anyway, your original point was "no matter what I try grub-install embeds '(,gpt6)/boot/grub" but in the last output you sent I see grub-mkimage being called with "--prefix '(,gpt5)/boot/grub'" 2024-01-09 19:20:09 also add "--no-nvram" to your script to stop the warnings about EFI variables 2024-01-09 19:20:13 right, and it embeds correct in EFI/alpine/grubaa64.efi i.e. '(,gpt5)/boot/grub' 2024-01-09 19:20:30 so then what's the issue? 2024-01-09 19:21:00 it doesn't in BOOTAA64.EFI 2024-01-09 19:21:50 where is BOOTAA64.EFI being created in the latest script you sent me and in the latest output you sent? 2024-01-09 19:22:20 EFI/boot/BOOTAA64.EFI 2024-01-09 19:22:36 where does the script create that file? 2024-01-09 19:23:04 I think grub-install create it by default 2024-01-09 19:23:07 no 2024-01-09 19:23:23 you told grub-install to create GRUBAA64.EFI, not BOOTAA64.EFI 2024-01-09 19:23:37 how then it appears on FS 2024-01-09 19:23:44 I have no idea 2024-01-09 19:24:14 if you specify "--removable" it will create BOOTAA64.EFI, if you don't it won't 2024-01-09 19:24:41 tha tis why Alpine's setup-disk *manually* copies the grub EFI file to the boot EFI file 2024-01-09 19:25:33 https://gitlab.alpinelinux.org/alpine/alpine-conf/-/blob/master/setup-disk.in?ref_type=heads#L332 2024-01-09 19:26:06 wait, if I put just grubaa64.efi on ESP will it be read by default 2024-01-09 19:26:13 no 2024-01-09 19:26:29 bootaa64.efi is the "fallback" file 2024-01-09 19:26:45 the one that is read if NO EFI boot variables can be used 2024-01-09 19:27:20 ah, and apple doesn't support efi variables 2024-01-09 19:28:02 the reason why grub-install does NOT create the boot file by default is that there can only be *ONE* fallback file in a ESP partition yet you could have multiple OSes/distros installed on the same disk 2024-01-09 19:28:30 so Windows, other Linux distros, BSDs etc all might write to that single file 2024-01-09 19:28:49 so it grub-install was to replace it that might break another OS' booting 2024-01-09 19:29:10 holly uh, too much screws 2024-01-09 19:29:13 so what does Apple do instead of EFI boot variables? 2024-01-09 19:29:31 it relies on u-boot 2024-01-09 19:29:47 how do you tell Apple to boot a specific EFI file and give it parameters for that particular file booting? 2024-01-09 19:30:45 IDK, only I know it searches ESP for BOOTAA64.EFI 2024-01-09 19:37:31 U-boot searches for that you mean? or the Mac's own bootloader? 2024-01-09 19:49:00 no, u-boot for asahi 2024-01-09 19:50:23 booting linux on apple silicon is few step, apple Iboot -> separate partition with m1n1 -> u-boot+m1n1 -> grub 2024-01-09 19:51:03 ok so then why don't you configure u-boot to use the grubaa64.efi file for booting? 2024-01-09 19:52:04 because u-boot search for this to find root partition 2024-01-09 19:52:26 I don't understand 2024-01-09 19:52:53 TBH, also I don't understand all details 2024-01-09 19:52:54 as there are no EFI variables then I assume u-boot must use some static config file to tell it how to UEFI boot 2024-01-09 19:53:25 u-boot have EFI inside it 2024-01-09 19:53:27 the *standard* method for UEFI to boot is via boot variable (wth fallback as last resort) 2024-01-09 19:53:35 so are their EFI variables "inside" u-boot? 2024-01-09 19:54:43 it is described here in more details https://social.treehouse.systems/@marcan/111686102134837867 2024-01-09 19:57:07 so copy the generated grubaa64.efi to the bootaa64.efi file which that post says is run by u-boot 2024-01-09 19:57:39 seems so 2024-01-09 19:58:03 so it uses the fallback mode 2024-01-09 19:58:23 exactly like the Alpine setup-disk for EFI 2024-01-09 19:59:52 alpine puts grubaa64.efi in EFI/alpine dir 2024-01-09 20:02:35 yes and copy it to EFI/boot/bootaa64.efi then 2024-01-09 20:02:43 just like setup-disk does 2024-01-09 20:03:02 EFI/boot/bootaa64.efi is the fallback location for aarch64 2024-01-09 20:03:03 let me try with grubaa64.efi in EFI/boot, which will take about hour or more 2024-01-09 20:03:23 nope, won't work, you NEED to name it bootaa64.efi in the right directory 2024-01-09 20:03:59 unless you meant copying it as bootaa64.efi... 2024-01-09 20:04:40 so, EFI/boot/bootaa64.efi will not work? 2024-01-09 20:05:06 *that* will work 2024-01-09 20:05:15 ok, lets try 2024-01-09 20:05:23 thanks again 2024-01-09 20:05:39 you said "try with grubaa64.efi in EFI/boot" which I took to mean as "cp EFI/alpine/grubaa64.efi EFI/boot/" 2024-01-09 20:06:50 yes, I meant this 2024-01-09 20:40:45 minimal: this is waste of our time 2024-01-09 20:41:06 thank for trying to help but I give up 2024-01-09 20:41:17 thanks* 2024-01-09 20:41:27 did bootaa64.efi not work? 2024-01-09 20:41:52 will left this to someone which knows details of grub 2024-01-09 20:42:07 did u-boot load the bootaa64.efi? 2024-01-09 20:42:08 no, bootaa64.efi also didn't worked 2024-01-09 20:42:36 I think it doesn't 2024-01-09 20:42:43 so then it is not a Grub issue 2024-01-09 20:42:49 it is a u-boot issue 2024-01-09 20:43:00 it says something 'cannot find Boot000' 2024-01-09 20:43:23 that seems to be a reference to a EFI boot variable 2024-01-09 20:43:51 typically each boot entry is a variable Boot0001, Boot0002, etc 2024-01-09 20:44:04 same u-boot works fine from usb ssd and from nvme already installed 2024-01-09 20:44:30 but that posting you mentioned said u-boot DO NOT support EFI boot variables 2024-01-09 20:44:43 did it mean it (a) has no support, or (b) has broken support? 2024-01-09 20:45:04 I meant not persistent variables 2024-01-09 20:45:09 if it has no support then why would it complain about being unable to find Boot000? 2024-01-09 20:45:29 they are not needed for fallback to work 2024-01-09 20:45:31 I have no idea about this 2024-01-09 20:45:45 indeed if they exist (and are valid) then fallback cannot work 2024-01-09 20:45:59 but it is NOT a Grub issue based on what you have described 2024-01-09 20:46:23 Grub hasn't been loaded yet (from BOOTAA64.EFI) so how can it be at fault? 2024-01-09 20:47:54 this happened with bootaa64.efi, with BOOTAA64.EFI I've got grub rescue prompt 2024-01-09 20:49:08 eh? they're the same file - ESP is a vfat filesystem so uppercase==lowercase 2024-01-09 20:49:39 unless you're using (from memory) some non-standard u-boot ext2 ESP filesystem? 2024-01-09 20:51:11 both EFI/BOOT/BOOTAA64.EFI and EFI/boot/bootaa64.efi *are* the same file on a VFAT filesystem so u-boot should not be behaving differently 2024-01-09 20:52:48 it is VFAT 2024-01-09 20:54:01 maybe to rename bootaa64.efi to BOOTAA64.EFI could help 2024-01-09 20:56:22 it shouldn't make any difference at all as it is a VFAT filesystem 2024-01-09 20:58:01 you said you got a grub rescue prompt with BOOTAA64.EFI, what error did it show? 2024-01-09 20:58:54 I forgot and didn't made photo 2024-01-09 21:02:35 minimal: copying bootaa64.efi to BOOTAA64.EFI helped 2024-01-09 21:03:15 now kernel boots 2024-01-09 21:05:11 and have to fix 'root' in append line 2024-01-09 21:05:22 so we got it 2024-01-09 21:09:20 so means that u-boot is "fussy" about case despite vfat being case independent 2024-01-09 21:09:46 that's in "boot" or "BOOT" directory? 2024-01-09 21:10:58 minimal: look here, https://github.com/AsahiLinux/asahi-alarm-builder/blob/main/build.sh#L116 2024-01-09 21:12:45 yes that is using all uppercase but I was asking about *testing*, do you have a EFI/BOOT directory or a EFI/boot directory in the ESP? 2024-01-09 21:13:58 EFI/boot 2024-01-09 21:19:04 ok, so u-boot does NOT care about the case of the subdirectory ("boot" vs "BOOT") but does care about the case of the file (only handles "BOOTAA64.EFI" and not "bootaa64.efi")...that's crap 2024-01-09 21:21:28 no, maybe it works with bootaa64.efi, I didn't tried this. previously I simply copied bootaa64.efi to EFI/boot and not renamed to bootaa64.efi 2024-01-09 21:22:09 no, copied grubaa64.efi 2024-01-09 21:22:56 but enough for today, at least for me 2024-01-09 21:23:22 I'm already too tired 2024-01-09 22:00:35 fallback booting doesn't work for the file called grubaa64.efi 2024-01-09 22:00:48 it has to be called bootaa64.efi and be in the boot directory 2024-01-09 22:03:40 minimal: uh oh, I made stupid mistake, I copied wrong BOOTAA64.EFI to my zip file for asahi installer 2024-01-09 22:04:10 now I fixed it and it contains proper gpt root partition number 2024-01-09 22:08:41 last thing, how to force grub-mkconfig to set FS UUID for root=UUID= 2024-01-09 23:01:59 in /etc/default/grub you'd add "GRUB_DISABLE_LINUX_UUID=false" before running grub-mkconfig 2024-01-10 11:15:41 ncopa: on usa-bld-1, edge builders, I notice that when I start mqtt-exec.aports-build, the aports-build output is shown on stdout and the rc-service command hangs 2024-01-10 11:18:33 ikke: probably due to the recent change in mqtt-exec init.d 2024-01-10 11:19:00 2b9f3eec187012e6199ed90d920c9683f8a48b13 2024-01-10 11:19:35 +# Run with process supervisor. If you want to disable it, comment it out. 2024-01-10 11:19:35 +supervisor=supervise-daemon 2024-01-10 11:19:54 i think we can add `supervisor=supervise-daemon` to conf.d 2024-01-10 11:20:47 but i woudl kinda expect it to work without supervise-daemon 2024-01-10 11:25:54 ncopa: if I check the init.d file, it still appears to have the old contents 2024-01-10 11:26:13 supervisor=supervise-daemon is there 2024-01-10 11:54:09 ikke: my lxc containers are not accessible 2024-01-10 11:54:31 arm ones I mean 2024-01-10 11:55:31 Yeah, che-bld-1 is unreachable 2024-01-10 11:56:02 again /o\ 2024-01-10 11:57:07 ACTION would like to forget about this provider 2024-01-10 11:57:25 is che-bld-1 the machine we moved to new location? 2024-01-10 11:57:33 no 2024-01-10 11:57:44 oh yeah 2024-01-10 11:57:46 it is 2024-01-10 11:57:54 yea 2024-01-10 11:58:24 so the IPs changed 2024-01-10 11:58:25 so we need new IPs and routes? 2024-01-10 11:59:15 2a02:168:a077:701:fc7f:6dff:fe4e:ee06 2024-01-10 11:59:19 2a02:168:a077:701:fc07:9bff:fe47:d403 2024-01-10 11:59:23 Only the prefix changedf 2024-01-10 11:59:46 and IPs for containers 2024-01-10 12:00:27 Those are the container ips 2024-01-10 12:01:46 no IPv4? 2024-01-10 12:02:39 Never had 2024-01-10 12:02:50 Not publically anyway 2024-01-10 12:03:13 Or do you dmvpn IPs? Because dmvpn is not working due to lack of ipv4 2024-01-10 12:03:21 ikke: I had IPv4 addresses for my containers 2024-01-10 12:03:44 yes, IPv4 over WG 2024-01-10 12:15:39 ikke: I can ping 172.104.203.112, is it dmvpn IP iirc 2024-01-10 12:17:55 and I'm connected to x86_64 lxc over WG 2024-01-10 12:18:43 Yes, it just does not work for che-bld-1 2024-01-10 12:20:46 aha, but you wrote above che-bld-1 is moved somewhere 2024-01-10 12:35:24 ikke: are these two IPv6 addresses above my lxc addresses 2024-01-10 12:35:57 I can ping them over proxyhost but can't connect with ssh 2024-01-10 12:36:15 Yes 2024-01-10 12:36:31 I might need to update the firewall rules 2024-01-10 12:37:19 are they started? I remember that clandmeter had to start them when hosts reboots 2024-01-10 12:39:11 Yes, they are 2024-01-10 12:39:25 Otherwise you couldn't ping them 2024-01-10 12:42:32 aha, ssh is blocked I guess 2024-01-10 16:20:03 Good morning, I'm getting a no spacel left error on aarch64 ci: https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/49831 2024-01-10 16:20:30 ayakael0: yeah, it's running atm on a system with limited space 2024-01-10 16:20:39 Our main CI host is unreachable atm 2024-01-10 16:20:43 (for ARM) 2024-01-10 16:32:23 did something change with lxc containers, can't connect 2024-01-10 16:33:08 The IP address has changed: 2a02:168:a077:701:216:3eff:fec1:a4eb 2024-01-10 16:34:17 mps: I've updated the firewall rules (it was the NAT rule that was pointing to the old IP) 2024-01-10 16:34:36 pj: does that work? 2024-01-10 16:34:57 it responds 2024-01-10 16:35:23 doesn't necessarily want to auth me 2024-01-10 16:36:07 since my key wasn't loaded in ssh agent 2024-01-10 16:36:09 :) 2024-01-10 16:36:11 it works now 2024-01-10 16:36:16 I just saw the success in the logs 2024-01-10 16:36:49 I'm using ssh key via ssh agent via gpg agent via yubikey 2024-01-10 16:37:00 it's been a bit of pain lol 2024-01-10 16:41:35 should I be able to dump core file in the container? 2024-01-10 16:42:05 not sure what the config is and since /proc is ro 2024-01-10 16:46:47 4AF9DF7:error:0A000086:SSL routines:tls_post_process_server_certificate:certificate verify failed:ssl/statem/statem_clnt.c:1889: 2024-01-10 16:47:36 pj: I think so. Would sys_ptrace be required for that? 2024-01-10 16:49:09 sys_ptrace and ulimit? 2024-01-10 16:49:26 to be fair, I don't know since I don't debug much in linux 2024-01-10 16:49:36 sys_ptrace is limited 2024-01-10 16:49:45 did you already run ulimit -c unlimited? 2024-01-10 16:49:50 I know that ulimit and eventual /proc/sys/kernel/core_* thingy 2024-01-10 16:49:56 yes 2024-01-10 16:52:23 "Sat Jun 24 19:41:02 UTC 2023" 2024-01-10 16:52:29 i guess that would explain ssl error 2024-01-10 16:52:33 hmm, that doesn't seem right 2024-01-10 16:54:05 is the date on host correct? 2024-01-10 16:54:32 no 2024-01-10 16:54:46 :) 2024-01-10 16:54:56 There is ntpd running 2024-01-10 16:55:44 pool.ntp.org has no AAAA record 2024-01-10 16:58:35 2.pool.ntp.org has some 2024-01-10 16:59:47 I found ntp.time.nl as well 2024-01-10 16:59:51 pj: date is fixed 2024-01-10 16:59:55 I'm using pool.ntp.org as well but I do have IPv4 on out 2024-01-10 17:00:23 ikke: thanks <3 2024-01-10 17:00:55 as for the coredump, I guess we can try with sys_ptrace and will see how that goes 2024-01-10 17:01:12 pj: the container runs with sys_ptrace drops atm 2024-01-10 17:05:55 pj: I don't think sys_ptrace is required for coredumsp 2024-01-10 17:08:26 well, it's one way to do coredumps I think? it's either ulimit + sysctl or ulimit + sys_ptrace 2024-01-10 17:08:57 -rw------- 1 pj pj 87.1M Jan 10 17:07 src/rustc-1.73.0-src/core 2024-01-10 17:09:13 or maybe I was just blind and didn't notice it 2024-01-10 17:09:26 (the first time I tried to debug) 2024-01-10 17:11:15 ikke: thanks, it works now 2024-01-10 17:11:48 not so simple as with IPv4 but at least accessible 2024-01-10 18:05:23 ncopa: adding command_foreground=true fixes the mqtt-exec service 2024-01-10 18:05:37 but it's anoying that we now have to adjust all builders if we want supervision again 2024-01-10 18:07:30 I meant command_background=true 2024-01-10 18:07:35 mps: your Asahi booting/installer issues all sorted now? 2024-01-10 18:15:46 ikke: maybe we should do soemthing like: if [Β -n "$supervisor" ]; then command_foreground=true; fi 2024-01-10 18:16:06 ikke: Maybe create an issue about it with the details and ping jirutka 2024-01-10 18:16:12 ncopa: I just tested it, I think it works if you have it set while setting supervisor 2024-01-10 18:16:24 ncopa: already did: #15663 2024-01-10 18:16:50 "It seems, however, that I forgot to add command_background=yes." 2024-01-10 18:17:23 Yes 2024-01-10 18:18:59 alright, i think i have an approximate idea why chromium segfaults 2024-01-10 18:25:35 chromium has a DnsConfigService which monitors and reads dns config on system. 2024-01-10 18:25:46 so that chromium can use its own dns implementation 2024-01-10 18:27:30 this dns config reader appears to use __res_state to find the dns servers 2024-01-10 18:28:21 that obviously does not work with musl, since it does not use res_* at at all (it uses a statless implementation) 2024-01-10 18:29:00 i believe that chromium falls back to getaddrinfo 2024-01-10 18:29:05 which is why it works 2024-01-10 18:29:32 however, when the linux dns service is refreshed, it crashes occationally 2024-01-10 18:29:59 chromium apparently makes assumptions of the __res_state 2024-01-10 18:30:14 and it goes boom 2024-01-10 18:30:49 there are different dns config service implementations for different OSes, and I have tested the Posix fallback 2024-01-10 18:31:11 (i believe it does not care about NSS, which linux implementation does) 2024-01-10 18:32:09 but it still uses __res_state and it still crashes (I have tested it) 2024-01-10 18:33:22 I observed that iOS just returns a nullptr, so no dns config service there 2024-01-10 18:33:42 also Fuchsia is very bare bones and might work for alpine 2024-01-10 18:33:58 https://github.com/chromium/chromium/blob/main/net/dns/dns_config_service_fuchsia.cc 2024-01-10 18:36:49 oh, im in wrong channel 2024-01-10 18:37:04 maybe i should try #musl-distros 2024-01-10 18:53:40 minimal: yes 2024-01-10 18:54:49 problem with UUID in grub.cfg is in that grub-mkconfig will not add it if it doesn't find initramfs 2024-01-10 18:55:22 now alpine installer for asahi linux works \o/ 2024-01-10 18:55:55 yyck indeed 2024-01-10 18:55:55 brave ones could install it from the net if they want 2024-01-10 18:56:01 wrong channel 2024-01-10 18:56:19 mps: strange, an initramfs isn't necessarily required for booting Linux. Also the UUID is that of the rootfs isn't it? 2024-01-10 18:56:30 yes 2024-01-10 18:56:59 but without initramfs root=UUID=... will not work anyway 2024-01-10 18:57:57 UUID partition is searched in initramfs when it is loaded and started init in it 2024-01-10 18:59:26 kernel don't 'know' how to process root=UUID 2024-01-10 19:00:09 also root=LABEL=... doesn't work without initramfs 2024-01-10 19:03:33 right, build-in kernel root handling is limited to device name (I think) basically 2024-01-10 19:04:34 PARTUUID and PARTNROFF works in kernel 2024-01-10 19:05:13 had a quote look and I see major/minor mentioned, e.g. root=0x803 2024-01-10 19:05:20 s/quote/quick/ 2024-01-10 19:06:49 interesting rootFS for 'root=' option is 'root=PARTUUID=%U/PARTNROFF=1' 2024-01-10 19:09:41 yeah I'm looking at block/early-lookup.c currently 2024-01-10 19:10:15 it mentions support for matchnig my partition label 2024-01-10 19:11:06 yeah, there's a list of 9 formats in that file 2024-01-10 19:11:52 as you're using GPT then you could use PARTLABEL 2024-01-10 19:13:10 so you're booting Alpine on M1 without any initramfs? 2024-01-10 19:14:21 no, initramfs is used in installer I finished today, though now more than two years I run alpine on m1 without initramfs 2024-01-10 19:14:50 I've built needed fs driver as =y in kernel 2024-01-10 19:15:02 drivers* 2024-01-10 19:15:47 minimal: do you have apple silicon machine? 2024-01-10 19:16:22 nope, no Apple computers here at all 2024-01-10 19:18:49 good for you, you didn't made mistake as I did 2024-01-10 19:20:40 last time I used an Apple computer frequently was an System 7 days (approx 1992), and before that Apple II lol 2024-01-10 19:21:09 though Jobs did draw me in with NeXT machines (which became the basis for MacOS X) 2024-01-11 11:12:19 fatal error: error in backend: IO failure on output stream: No space left on device 2024-01-11 11:12:33 we are running out of diskspace on the arm machine 2024-01-11 11:17:20 ncopa: CI or builders? 2024-01-11 11:20:53 Because che-ci-1 is down, I've enabled the runners on usa-bld-1, but they are constrained in space 2024-01-11 11:31:10 CI 2024-01-11 11:31:25 ok. I suppose I can just merge the MR then 2024-01-11 11:31:52 which MR? 2024-01-11 11:32:38 https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/58730 2024-01-11 11:32:47 it fails due to disk space outage 2024-01-11 11:33:02 but it passes on all other arches 2024-01-11 11:33:50 right, chromium is huge 2024-01-12 06:07:00 https://about.gitlab.com/releases/2024/01/11/critical-security-release-gitlab-16-7-2-released/ 2024-01-13 08:19:09 https://ptrc.gay/uXHAHQRm 2024-01-13 08:19:12 .~. 2024-01-13 08:32:29 ptrc: I marked you as trusted user, can you try again? 2024-01-13 08:32:42 i added the word "Alpine" to my issue and that did the trick :p 2024-01-13 17:13:16 'curl: (7) Failed to connect to github.com port 443 after 3040 ms: Couldn't connect to server' on my arm dev lxc 2024-01-13 17:15:16 Host github.com. has no AAAA record 2024-01-13 17:22:47 they still cant enable it in 2024 (which is a shame), but there is ipv6.github.com 2024-01-13 17:24:30 so we have to change 'source' in all aports for which source is on github 2024-01-13 17:25:58 or override in /etc/hosts (or elsewhere) to resolve github.com to the ipv6.github.com AAAA entries 2024-01-13 17:26:35 i just tested it and the ipv6 webserver serves for github.com too 2024-01-13 17:27:28 doesn't work in my case 2024-01-13 17:29:01 what do you get? 2024-01-13 17:32:53 yeah, cloning doesnt seem to work:/ and nor from the ipv6 domain, wth 2024-01-13 17:33:01 (curl works) 2024-01-13 17:36:45 sad https://github.com/orgs/community/discussions/10539 2024-01-13 17:59:59 if u are blocked i could offer you a v6 ip which is natted into github's v4 address 2024-01-13 18:15:15 nu[m]: lxc has IPv6 address 2a02:168:a077:701:fc7f:6dff:fe4e:ee06 2024-01-13 18:18:23 i meant that i could give u an ipv6 address that can be used as a fake destination address for github.com 2024-01-13 18:18:44 I can ping it's ipv6 2024-01-13 18:19:06 you cant ping github.com though 2024-01-13 18:19:14 right 2024-01-13 18:26:52 actually this should already work if you would be using my dns resolver, but i would understand if its not possible for e.g. security reasons 2024-01-13 18:27:58 however if you run your own bind you could fake these addresses for a prefix i or you would provide, which also would solve these kind of issues 2024-01-13 18:27:58 I use resolver which ikke set 2024-01-13 18:28:21 don't want to use other 2024-01-13 18:28:57 all work there are alpine related and I don't want hacks there 2024-01-13 18:30:30 yeah, manual override would be a hack 2024-01-13 18:39:48 *manual dns override per ip 2024-01-15 08:26:20 morning! seems like aarch64 CI is missing? 2024-01-15 08:38:59 The vm had crashed 2024-01-15 08:39:23 I need to improve monitoring for that 2024-01-15 08:45:41 I think we should move https://gitlab.alpinelinux.org/Leo/atools to alpine namespace 2024-01-15 08:49:26 ncopa: I'm almost ready with a reimplantation in go, which can do static analysis as dynamic analysis as well. 2024-01-15 08:49:59 https://gitlab.alpinelinux.org/kdaudt/atools-go 2024-01-15 08:51:04 It solves things like not warning about ` in comments 2024-01-15 08:52:39 oh! nice! 2024-01-15 08:52:47 what is missing there? 2024-01-15 08:54:14 Not much anymore, just some liters 2024-01-15 08:54:25 linters 2024-01-15 09:00:39 invalid-arch, bad-version 2024-01-15 09:18:30 the LoongArch devs would like to set up an edge builder: https://gitlab.alpinelinux.org/alpine/tsc/-/issues/72 2024-01-15 09:18:51 what's the best way to provide support for them? 2024-01-15 09:34:34 kunkku: if you want i can reply them and create the needed comm channels 2024-01-15 13:34:28 clandmeter: please do, thanks! 2024-01-15 16:08:52 Found this spam account while looking through https://gitlab.alpinelinux.org/explore: https://gitlab.alpinelinux.org/EscLondon 2024-01-15 16:12:01 celie: Yeah, there are more, I regularly clean them up. Thanks! 2024-01-15 16:12:10 You're welcome 2024-01-17 14:45:47 mps, heyo 2024-01-17 14:46:41 KREYREN_oftc: \o 2024-01-17 14:46:51 https://dpaste.org/uPi80 2024-01-17 14:46:56 updated it a bit more 2024-01-17 14:47:04 untested, need checksum regenerated 2024-01-17 14:47:35 this will not work, but I already prepared correct one 2024-01-17 14:49:33 oke 2024-01-17 14:49:50 mps, btw i also have a gift https://dpaste.org/FPzeo save this as .shellcheckrc in the root of the repo 2024-01-17 14:50:27 will look later 2024-01-17 14:50:43 https://imgur.com/a/uyQMqdh have fun :p 2024-01-17 14:53:18 for now only option for alpine is to move u-boot to community if we want to build it aarch64 with SCP 2024-01-17 14:53:46 lets do it then 2024-01-17 14:53:46 else you will have to build it locally 2024-01-17 14:54:29 ye local building is the exact thing i am trying to avoid with pmbootstrap 2024-01-17 14:54:41 I don't think it is good idea to move u-boot 2024-01-17 14:55:02 i don't think it's a good idea to keep crust in testing tbh 2024-01-17 14:55:21 simple software easy to audit seems mature 2024-01-17 14:55:41 and alpine's the only distro that keeps that in testing atm afaik 2024-01-17 14:56:05 solution could be to move gcc-cross-embedded from community to main 2024-01-17 14:57:48 mps, i dunno what that implicates, but if i get crust in u-boot sure 2024-01-17 14:57:50 TBH I'm not sure how to solve all these dependencies 2024-01-17 14:57:56 do you still have that optimized linux kernel config btw 2024-01-17 14:58:02 i am working on a tiny and defconfig atm 2024-01-17 14:58:38 if you mean linux-edge it is not optimized anymore, too much is added in last year 2024-01-17 14:59:19 mps, makedepend="crust arm-trusted-firmware"; export SCP=/usr/share/crust/teres_i/scp/scp.bin; export BL31=/usr/share/arm-trusted-firmware/sun50i_a64/bl31.bin; build() { make teres_i_defconfig; make ;} 2024-01-17 14:59:33 but it is still optimized to some degree 2024-01-17 14:59:49 mps, i just shorten it to like 20 lines of config code and working on adjusting the depends in linux kernel for it to work atm 2024-01-17 14:59:58 so teres should be able to build it within 20 min 2024-01-17 15:00:05 KREYREN_oftc: I know how to build it with SCP 2024-01-17 15:00:06 and future-proof 2024-01-17 15:00:22 mps, i dunno what you meant by not knowing how to solve the dependencies then 2024-01-17 15:00:32 just the release thing? 2024-01-17 15:00:43 this is alpine 'problem' 2024-01-17 15:01:05 or like you can just keep the crust, atf and u-boot where they are and i puzzle them together in pmos 2024-01-17 15:01:27 well that would suck 2024-01-17 15:01:31 alpine is not pmOS, yet ;p 2024-01-17 15:01:41 rather pmos is not alpine yet xD 2024-01-17 15:02:12 seems like it would be great addition to alpine's device compatibility overall 2024-01-17 15:02:40 you are right, but I will not fight this battle 2024-01-17 15:03:14 well fix the invidual packages and ideally try to handle the alpine release thing if you fail then let me know and i try to do stuff 2024-01-17 15:04:24 in the meantime i am importing a u-boot from armbian in my testing pmos thing and go work on the linux kernel optimizations while checking on thermal devs every 30 min to see if i have to rush in their room with a broom and tell them to either figure out how to handle trips in a sane way and revert the 8 months of work they did bcs it sucks ass 2024-01-17 15:04:33 (causes teres to not boot on never kernels atm) 2024-01-17 15:23:38 *newer 2024-01-17 17:58:19 upgraded deu5-dev1 to alpine 3.19 2024-01-17 19:02:18 mps, using your u-boot-sunxi package from pmos i get black screen btw 2024-01-17 19:02:25 so it's probably some issue with intialization 2024-01-17 19:03:22 i just tested this with using armbian vs freshly compiled u-boot-sunxi btw armbian's u-boot works perfectly and u-boot-sunxi no display 2024-01-17 19:03:28 so same rootfs for both cases 2024-01-17 19:04:15 The issue seems to be within the BL31 as i don't get the notices of it setting up the PMIC etc.. 2024-01-17 19:04:35 but compiling u-boot-sunxi using BL31.bin from nixos also doesn't work so no idea 2024-01-17 19:16:38 KREYREN_oftc: at least we know bug is not related to SCP/crust 2024-01-17 19:17:11 probably armbian have some patches to fix it 2024-01-17 21:50:47 mps, i am the one who maintains armbian and i am not aware of any relevant patches 2024-01-17 21:51:13 nixos, arch, parabola, debian, ubuntu, devuan all use the same packaging instructions as alpine and it works there 2024-01-17 21:53:25 mps, if you have a working u-boot for alpine then plz gib it would help me with pmos development 2024-01-17 21:58:30 KREYREN_oftc: hm, if armbian works and other distros doesn't then I have no idea what could be problem 2024-01-17 21:58:51 could you point me to armbian build script 2024-01-17 21:59:51 could be something related to musl 2024-01-17 22:00:19 but it shouldn't be 2024-01-17 22:01:35 maybe BL31 on alpine is broken 2024-01-17 22:02:44 maybe build with BL31 from armbian 2024-01-17 22:03:14 mps, all other distros that i maintain work it's only alpine that doesn't 2024-01-17 22:04:00 > maybe BL31 on alpine is broken < i built it with BL31.bin from nixos and it had the same issue but try 2024-01-17 22:04:12 > 18<mps18> could you point me to armbian build script < on it 2024-01-17 22:04:18 KREYREN_oftc: please give me url of debian pkg 2024-01-17 22:04:44 https://github.com/armbian/build/tree/main/lib/functions/compilation refer t atf, crust, kernel and u-boot files 2024-01-17 22:04:59 https://github.com/armbian/build/blob/main/lib/functions/compilation/uboot.sh#L73 2024-01-17 22:05:03 u-boot compilation command 2024-01-17 22:05:39 and which u-boot version you use on these distros 2024-01-17 22:05:41 if you need something found there then let me known i am quite familiar with the codebase ^-^ 2024-01-17 22:05:57 armbian uses 2024.01-rc4 2024-01-17 22:06:08 why not stable 2024-01-17 22:06:20 mps, because the codebase is spaghetti and nothing is stable there 2024-01-17 22:06:39 aha 2024-01-17 22:06:47 i just do my beyond best to keep the whole thing together tbh 2024-01-17 22:06:51 it's hell 2024-01-17 22:07:03 thus why i am looking into postmarketOS :D 2024-01-17 22:07:53 ok, I will look at debian when I found time 2024-01-17 22:08:17 because they use stable 2024-01-17 22:08:34 talk to jc if needs be he maintain debian 2024-01-17 22:08:49 like debian without using armbian framework 2024-01-17 22:09:08 https://github.com/jcstaudt 2024-01-17 22:09:32 in case you need other maintainers see https://wiki.postmarketos.org/wiki/OLIMEX_Teres_i i tries to make a list for users to find 2024-01-17 22:10:17 *tried 2024-01-17 22:11:31 mps, btw what linux package you use on alpine for teres? 2024-01-17 22:11:39 pmos has this weird allwineer variant that is full of shit 2024-01-17 22:12:36 what you mean? 2024-01-17 22:13:34 to explain I bought teres-i just to test alpine on it, not for usage 2024-01-17 22:14:00 I had a hope that olimex is better 2024-01-17 22:14:13 like what kernel are you using on it 2024-01-17 22:14:27 > I had a hope that olimex is better < hm? 2024-01-17 22:14:31 ah, mainline only 2024-01-17 22:14:43 ACTION is developing teres-2 atm feedback appreciated to be addressed in new design 2024-01-17 22:14:58 > 18<mps18> ah, mainline only < did that worked for you? 2024-01-17 22:15:09 not sure, my efforts are now on riscv 2024-01-17 22:15:33 yes, mainline kernel works fine 2024-01-17 22:15:39 riscv? TH1520? 2024-01-17 22:15:49 > 18<mps18> yes, mainline kernel works fine < linux-lts package? 2024-01-17 22:15:54 riscv in general 2024-01-17 22:16:41 do you have something better than TH1520 that i should look at? I was thinking about turning teres into a SOM platform that could have that chip on a module that can be swapped 2024-01-17 22:16:46 no, I build trimed down linux-sunxi pkg locally which you downloaded by using my script to build image for teres-i 2024-01-17 22:17:35 oh oke 2024-01-17 22:17:48 for now I only have JH7100 and JH7110 boards of riscv 2024-01-17 22:18:48 my linux-sunxi is now 6.6 series 2024-01-17 22:20:06 KREYREN_oftc: sorry, I have to go to bed now, good night 2024-01-17 22:20:38 mps, gn <3 2024-01-17 22:21:04 i was messing with JH7110 but they didn't want to give me the docs so i told them to fuck off kinda did that get any better? 2024-01-17 22:21:53 and TH1520 seems to be a better platform in comparison but has minor(?) issue with underfloat exception 2024-01-17 22:22:59 KREYREN_oftc: forgot to tell, there is #linux-cros-arm on OFTC where are some people from pmOS and some other who works with arms 2024-01-17 22:23:38 oke 2024-01-18 10:37:00 any ipv6 experts here? I am trying to get routable addresses with dnsmasq on my lxcbr0 interfaces 2024-01-18 10:37:17 the lxc containers gets link local ipv6, but no ULA 2024-01-18 10:38:02 the lxcbr0 iface does have a prefix 2024-01-18 10:38:07 inet6 2001:4646:fb05:4:6000::1/68 scope global dynamic noprefixroute 2024-01-18 10:38:07 valid_lft 770sec preferred_lft 770sec 2024-01-18 10:40:34 dnsmasq has this in the .conf: dhcp-range=::1,constructor:lxcbr0,ra-names,slaac,12h 2024-01-18 11:02:04 salut ncopa ! I would run `radvdump` on the machine to see whether or not dnsmasq is distributing RAs on the lxbr0 interface 2024-01-18 11:02:43 I haven't used dnsmasq for RAs yet (mostly radvd and bird), but debugging using radvdump is rather straight forward. 2024-01-18 11:04:52 ncopa: what is a bit atypical is the /68 prefix. There has been some discussion on whether or not that is "valid", as in: a lot of software assumed the smallest prefix is /64, especially in the context of RAs. In theory a smaller prefix can work, but usually the logic for SLAAC is as follows: 2024-01-18 11:05:44 - take the /64 prefix... (full message at ) 2024-01-18 11:16:49 I've used corerad for sending RAs 2024-01-18 11:16:53 but corerad hasn't been very stable 2024-01-18 11:17:40 For docker, I have used smaller prefixes, like /80 2024-01-18 11:49:25 oh... it works now 2024-01-18 11:50:02 i reconfigured the dhcpcd and now it appears to assign /64 to lxcbr0 and now it works 2024-01-18 11:50:34 yay 2024-01-18 11:55:08 except that it does not work 2024-01-18 11:55:12 ncopa-3-18-x86-64 [~/aports]$ wget https://vg.no 2024-01-18 11:55:12 Connecting to vg.no ([2001:67c:21e0::16]:443) 2024-01-18 11:55:12 wget: can't connect to remote host: Network unreachable 2024-01-18 11:55:31 maybe its iptables? 2024-01-18 11:56:23 If forwarding enabled? 2024-01-18 11:56:27 Is* 2024-01-18 11:56:56 yes, i can ping ipv4 2024-01-18 11:57:13 ipv6 forwarding is a different setting 2024-01-18 11:57:25 but it seems like ipv6 is broken on my host 2024-01-18 11:57:27 PING 2001:67c:21e0::16 (2001:67c:21e0::16): 56 data bytes 2024-01-18 11:57:27 ^C 2024-01-18 11:57:27 --- 2001:67c:21e0::16 ping statistics --- 2024-01-18 11:57:27 ncopa-desktop:~$ ping6 2001:67c:21e0::16 2024-01-18 11:57:27 2 packets transmitted, 0 packets received, 100% packet loss 2024-01-18 11:57:45 What address is that that you are pinging? 2024-01-18 11:58:30 Connecting to vg.no ([2001:67c:21e0::16]:443) 2024-01-18 11:58:32 vg.no 2024-01-18 11:59:20 Do you have a default gw for ipv6? 2024-01-18 11:59:48 ncopa-desktop:~$ cat /proc/sys/net/ipv6/conf/all/forwarding 2024-01-18 11:59:48 1 2024-01-18 11:59:55 ok, that's good 2024-01-18 12:00:14 from the host: default via fe80::9a77:e7ff:fee6:75d7 dev wlan0 proto ra metric 3005 mtu 1472 pref high 2024-01-18 12:00:34 i get the ipv6 from the isp wifi router 2024-01-18 12:00:42 can you ping -I wlan0 fe80::9a77:e7ff:fee6:75d7 2024-01-18 12:00:59 PING fe80::9a77:e7ff:fee6:75d7 (fe80::9a77:e7ff:fee6:75d7): 56 data bytes 2024-01-18 12:00:59 64 bytes from fe80::9a77:e7ff:fee6:75d7: seq=1 ttl=64 time=2.919 ms 2024-01-18 12:00:59 64 bytes from fe80::9a77:e7ff:fee6:75d7: seq=0 ttl=64 time=249.984 ms 2024-01-18 12:01:01 yup 2024-01-18 12:01:24 Does your router have an ipv6 address? 2024-01-18 12:01:29 ncopa-desktop:~$ traceroute6 2001:67c:21e0::16 2024-01-18 12:01:29 1 2001:4646:fb05:0:bf51:c5a8:7161:167c (2001:4646:fb05:0:bf51:c5a8:7161:167c) 0.014 ms !N 0.014 ms !N 0.007 ms !N 2024-01-18 12:01:29 traceroute to 2001:67c:21e0::16 (2001:67c:21e0::16), 30 hops max, 72 byte packets 2024-01-18 12:01:29 ncopa-desktop:~$ 2024-01-18 12:02:08 i assuem it does 2024-01-18 12:02:23 hmm, half the packets drop? 2024-01-18 12:02:26 it handed out an ipv6 prefix 2024-01-18 12:03:30 ncopa-desktop:~$ curl 'https://[2001:67c:21e0::16]' 2024-01-18 12:03:30 curl: (7) Failed to connect to 2001:67c:21e0::16 port 443 after 1013 ms: Couldn't connect to server 2024-01-18 12:06:47 from my macbooc i can curl to the same address 2024-01-18 12:06:54 using same wifi 2024-01-18 12:07:31 it complains about the cert, but at least it connects to it over ipv6 2024-01-18 12:07:43 ncopa-desktop:~$ ip addr show dev wlan0 | tpaste 2024-01-18 12:07:43 https://tpaste.us/a1VN 2024-01-18 12:08:35 Can you compare the routes between those 2 machines? 2024-01-18 12:08:52 the relevant dhcpcd.conf part: 2024-01-18 12:08:52 ia_pd 1/::56 lxcbr0/0 lxdbr0/1 2024-01-18 12:08:52 interface wlan0 2024-01-18 12:08:52 ipv6rs 2024-01-18 12:10:24 ncopa-desktop:~$ ip -6 route | tpaste 2024-01-18 12:10:24 https://tpaste.us/qE1z 2024-01-18 12:12:08 hum... now docker does not work anymore either. it worked the other day 2024-01-18 12:13:12 the default gw is the same for macbook and ncopa-desktop 2024-01-18 12:37:17 argh... i dont have time for this ipv6 blah 2024-01-18 12:37:36 it worked the other day and now nothing ipv6 appears to work 2024-01-18 12:51:08 what is the "unreachable" routes? https://tpaste.us/qE1z 2024-01-18 12:51:14 and what created them 2024-01-18 12:53:38 and the default via lo? 2024-01-18 12:54:07 That is suspect 2024-01-18 12:54:17 removing those and ipv6 works again 2024-01-18 12:57:25 Those unreachable routes remind of the issue we had with ppc64 or s390x, but you fixed that 2024-01-18 12:57:40 (caused by strongswan or something like that) 2024-01-18 13:03:55 could be any of: dhcpcd, dnsmasq, lxd 2024-01-18 13:20:55 unreachable routes sounds like bird or similar, something than is a router daemon and injecting them into the kernel 2024-01-18 13:21:44 Ahh, I think I know what you are running into, ncopa 2024-01-18 13:21:49 There is a catch with IPv6 and RAs 2024-01-18 13:22:04 The setting accept_ra in the kernel is usually set to 1 2024-01-18 13:22:23 Which means: accept_ras from upstream, if there is NO ipv6 forwarding enabled 2024-01-18 13:22:33 Which means that as soon as you trigger forwarding-> 1 -> loss of routes 2024-01-18 13:22:52 The fix for that is to set accept_ra=2 on the interface that you receive the uptream RA from 2024-01-18 14:18:58 ncopa-desktop:~$ cat /proc/sys/net/ipv6/conf/wlan0/accept_ra 2024-01-18 14:18:59 0 2024-01-18 14:20:34 i suppose this is something dhcpcd should set? 2024-01-18 14:20:51 it will act when the wlan0 iface appears/disapears 2024-01-18 14:28:21 but i think thamight be what happens. when I restart dhcpcd it sill stop dnsmasq.lxcbr0, lxd etc, and after dhcpcd is restarted it will start those again. dnsmasq.lxcbr0 does enable forwarding 2024-01-18 15:04:10 but yeah, something is adding a default route to lo 2024-01-18 15:04:26 but now it works, and it works nice 2024-01-18 15:23:54 ncopa: accept_ra=0 is strange from my side, but I haven't checked out the code of dhcpcd, so maybe the assumption in it is that it manages receiving RAs and setting values correctly, even though the kernel can do well on its own 2024-01-18 15:27:04 simply stopping the dnsmasq.lxcbr0 makes something re-add the default route to lo 2024-01-18 15:27:19 i have to manually delete that route every time 2024-01-18 15:28:12 That's a strange behaviour 2024-01-18 15:28:48 I would assume it's not dnsmasq itself 2024-01-18 15:28:56 Does it also happen if you kill -9 dnsmasq? 2024-01-18 15:43:14 mps, i have a working pmos device port only lacking is the u-boot do you need help or something? 2024-01-18 15:44:52 KREYREN_: what you mean by 'working pmos'? You have pmOS u-boot video console? 2024-01-18 15:45:15 mps, yep using armbian's u-boot 2024-01-18 15:45:26 and i just solved all problems in kernel and userland to be mergable 2024-01-18 15:45:31 aha, no I don't need this 2024-01-18 15:46:14 If you have u-boot fix then I will look if it could be added to alpine 2024-01-18 15:48:04 mps, i kinda need that u-boot fix tbh so was asking if you figured out something or if you need me to look at it 2024-01-18 15:48:38 no, not yet, don't have time these days to work on this 2024-01-18 15:49:03 git bisecting will take time 2024-01-18 15:49:30 i know that's why i wanted to throw that on you and move to figure out thermal devs breaking kernels:D 2024-01-18 18:05:41 x86_64 builders are down, right? 2024-01-18 18:38:25 KREYREN_: I found problem, alpine ATF is outdated, rebuilt with debian ATF and it works 2024-01-18 18:38:53 will look if I can upgrade ATF on alpine 2024-01-18 18:39:30 mps, i rebuilt that with nixos ATF which is up to date and had the same issue 2024-01-18 18:39:43 can you send me the file so that i can test it? 2024-01-18 18:39:49 upload.disroot.org for example 2024-01-18 18:40:01 with debian ATF it works 2024-01-18 18:40:07 weird 2024-01-18 18:40:43 I can upload it but you can rebuild with latest debian 2024-01-18 18:41:24 i want to test your version instead 2024-01-18 18:44:34 https://arvanta.net/2x3d1kfg/u-boot-sunxi-with-spl.bin 2024-01-18 18:45:04 <3 2024-01-18 18:48:11 mps, https://www.youtube.com/watch?v=6TdHYkWDHV0 2024-01-18 18:49:23 now have to find fixes for ATF 2.10 2024-01-18 18:50:45 ACTION just fixes stuff upstream 2024-01-18 19:01:09 KREYREN_: btw, do you know what should be offset for dd-ing u-boot on internal emmc on teres-i? I had it noted but deleted by mistake FS where I kept notes for this (and not only this) 2024-01-18 19:02:44 mps, i though it's same as for sdcard considering the emmc being wired to be interpreted by the SoC as an sdcard 2024-01-18 19:03:39 ok, I thought so but was not sure. anyway I can try and see 2024-01-18 19:10:01 mps, seems to be same based on armbian's scripts 2024-01-18 19:11:37 aha, ok 2024-01-18 19:13:11 and another good news, just upgraded locally ATF for alpine to 2.8.14 and rebuilt u-boot and it works 2024-01-18 19:13:29 so fix is simple, upgrade ATF and rebuild u-boot 2024-01-18 19:13:39 \o/ 2024-01-18 19:14:10 will do this later this evening if PureTryOut is not against upgrading ATF 2024-01-18 19:14:26 if it's an LTS release like the current one sure 2024-01-18 19:14:41 PureTryOut: yes, it is LTS 2024-01-18 19:14:55 then go ahead πŸ‘οΈ 2024-01-18 19:15:11 ACTION is going to stage postmarketos contrib in the meantime 2024-01-18 19:17:36 PureTryOut: !59176 2024-01-18 19:19:10 merged 2024-01-18 19:23:18 PureTryOut: FYI, I created asahi installer for alpine, not perfect but works 2024-01-18 19:46:42 Nice! 2024-01-18 19:51:45 KREYREN_: u-boot 2024.01-r2 is built on builders, in a few minutes you will be able to use proper alpine package. please test and report to me 2024-01-18 19:52:17 also it should fix problem on some other arm64 machines 2024-01-18 20:00:33 mps, can i just `apk update && apk add u-boot` on pmos? 2024-01-18 20:00:36 using edge 2024-01-18 20:35:38 yes 2024-01-18 20:36:01 apk add -u u-boot 2024-01-18 20:36:35 if pmOS syncs with alpine packages 2024-01-18 20:37:17 They use alpine repositories with their own on top 2024-01-18 20:38:48 aha, then this should work 2024-01-18 21:58:57 KREYREN_: looks like it is not solved yet. When I have serial console attached then video console works but when it unplugged then video console doesn't work 2024-01-18 21:59:29 I did my tests with serial console plugged in and didn't noticed this problem 2024-01-18 22:01:19 mps, weird 2024-01-18 22:01:23 what board revision do you have 2024-01-18 22:02:57 what file have this info 2024-01-18 22:03:38 the one you open with screwdriver and look at the white letters on the red board :D 2024-01-18 22:03:53 hah, I forgot 2024-01-18 22:04:39 I had a hope that /proc/device-tree could help 2024-01-18 22:05:28 they all use the same device tree afaik just board A and B has some issues that might be relevant 2024-01-18 22:05:58 didn't knew this 2024-01-18 22:06:38 I told you that I don't use this machine for anything, except for testing A64 2024-01-18 22:08:04 I expected that it have good support in mainline 'things' because hardware is very open, but looks like it is not 2024-01-18 22:08:29 anything missing? 2024-01-18 22:09:00 kernel 6.6.x doesn't work for my test 2024-01-18 22:09:14 trips node not found? 2024-01-18 22:09:20 6.6.7 works for me on pmos 2024-01-18 22:09:28 no video console 2024-01-18 22:09:49 that's alpine-induced issue with the u-boot or no? 2024-01-18 22:10:14 no, u-boot works but kernel doesn't 2024-01-18 22:10:26 what's video console? 2024-01-18 22:10:34 TTY? Getty? 2024-01-18 22:10:46 FB 2024-01-18 22:10:55 FB works on my end 2024-01-18 22:11:22 maybe I didn't carefully tweaked at upgrade kernel 2024-01-18 22:11:46 like i use my teres as a daily with gnome and do freecad engineering on it :D 2024-01-18 22:12:59 ahm, this is very interesting. must say I didn't expected this with this machine 2024-01-18 22:13:28 works really good too last time in hackerspace it was programming CNC milling :D 2024-01-18 22:13:57 so we need to give more effort to make it working good with alpine 2024-01-18 22:14:08 yep yep 2024-01-18 22:14:18 and then i need to fix nixos bcs it needs better kernel 2024-01-18 22:14:37 ok, I will continue to try make it better 2024-01-18 22:15:07 do you have the new u-boot file with the described console issue? 2024-01-18 22:15:07 does latest u-boot works for you without serial console 2024-01-18 22:15:49 mps, i can't easily update my pmbootstrap atm so would prefer you sending me the file for testing 2024-01-18 22:16:15 hm 2024-01-18 22:16:41 ACTION has lot of patches staged for contribution and is waiting for compilation atm 2024-01-18 22:19:26 ACTION is tired too much for today 2024-01-18 22:19:50 ok it should finish soon i try to get the u-boot from alpine repos 2024-01-18 22:19:52 gn <3 2024-01-19 12:35:57 mps, i tried the new u-boot and i get no screen 2024-01-19 12:37:33 KREYREN_: yes, I know. did you tried with serial console 2024-01-19 12:37:54 yes 2024-01-19 12:38:05 and? 2024-01-19 12:38:09 black screen with and without serial console 2024-01-19 12:38:13 rev.c board 2024-01-19 12:38:40 uses lts-v2.8.14(release): 2024-01-19 12:38:42 hm, I always have video console when I plug serial cable 2024-01-19 12:38:43 BL31 2024-01-19 12:40:39 mps, looking through the kicad i can't see anything that would point to why that's the case 2024-01-19 12:41:13 was suspecting that the 3v3 voltage from the serial console did something but unsure 2024-01-19 12:41:29 hmm it might actually be 5V unsure now 2024-01-19 12:42:23 it's 3.3V 2024-01-19 12:42:32 mps, any chance you use 5V serial console? 2024-01-19 12:43:11 no, I use one from olimex for teres-i 2024-01-19 12:43:32 hmmm 2024-01-19 12:46:16 this is because I didn't noticed problem earlier, I use this machine with serial console for testing 2024-01-19 12:47:05 the previous image you gave me worked without issues though what changed? 2024-01-19 12:47:22 mps18> https://arvanta.net/2x3d1kfg/u-boot-sunxi-with-spl.bin 2024-01-19 12:47:24 this 2024-01-19 12:47:59 this is built with debian ATF 2024-01-19 12:49:49 is debian doing something special with the ATF? 2024-01-19 12:50:13 or maybe it's because of the SUNXI env vars that are there in alpine now? 2024-01-19 12:50:17 I don't think so, some CC flags are different 2024-01-19 12:50:21 afaik they are not needed and might mess up the PMIC 2024-01-19 12:51:32 https://git.alpinelinux.org/aports/tree/main/arm-trusted-firmware/APKBUILD#n41 2024-01-19 12:52:14 SUNXI_SETUP_REGULATORS=0 should be left as default (1) 2024-01-19 12:52:14 see https://trustedfirmware-a.readthedocs.io/en/v2.10/plat/allwinner.html 2024-01-19 12:52:37 (consider referencning the provided docs in the code) 2024-01-19 12:53:54 and if that hypothesis is correct then that would explain why i wasn't able to fix it with nixos's BL31.bin before 2024-01-19 12:55:23 mps, ^ 2024-01-19 13:01:11 KREYREN_: sorry, I'm busy, will look at it later 2024-01-19 13:08:43 oke 2024-01-19 13:14:09 mps, i will test that and fix it if it's the case so you can take a break 2024-01-19 13:14:57 I'm building right now, but have to clean FS on mmc 2024-01-19 13:16:14 I hope will have it in 15-30 minutes 2024-01-19 13:17:07 oke 2024-01-19 13:24:01 give me some more time 2024-01-19 13:25:37 building complete u-boot even on fast machine takes time 2024-01-19 13:26:05 and I feel that is lunch time 2024-01-19 13:37:11 KREYREN_: you found it 2024-01-19 13:37:31 mps, that fixes it? 2024-01-19 13:37:45 now video console is active without serial cable 2024-01-19 13:37:50 yay :D 2024-01-19 13:38:06 weird that on your end the serial cable changes the behavior i assume you have the rev.B 2024-01-19 13:38:22 yes, ATF build with SUNXI_SETUP_REGULATORS=1 2024-01-19 13:38:28 btw `# Contributor: Jacob Hrbek ` add me in the u-boot file plz 2024-01-19 13:39:48 no, Contributor is added when something substantial is done, and for someone who is consistent with help for time 2024-01-19 13:40:40 but you can create merge request and then your name will author in git commit 2024-01-19 13:41:13 if you don't want to create MR I will for sure add you in commit message 2024-01-19 13:41:53 it's fine you can make it yourself i plan on refactoring the files later to add myself 2024-01-19 13:41:54 you are Polak, btw? 2024-01-19 13:42:02 nah Moravak (czech republic) :D 2024-01-19 13:42:27 ah, similar names and lastnames 2024-01-19 13:42:33 yep 2024-01-19 13:42:44 but we have a cool blue triangle in the flag and don't have lisp 2024-01-19 13:43:03 :) 2024-01-19 13:43:33 (visegrad jokes https://www.reddit.com/r/2visegrad4you/comments/15z9630/acceptable/) 2024-01-19 13:44:48 KREYREN_: we already discussed earlier to remove 'Contributor' field from APKBUILD but we still didn't and I think this should be done 2024-01-19 13:45:21 hmm i like crediting people for work bcs i can then win arguments in hackerspace 2024-01-19 13:46:12 q66: you suck at computer science krey! i am a creator and maintainer of my own distro! .. krey: i maintain all major distributions that doesn't suck which is why i never heard about yours 2024-01-19 13:46:13 etc.. 2024-01-19 13:46:14 :P 2024-01-19 13:46:28 we are not against crediting and we also do and like this, but keeping this in APKBUILD is bad, IMO 2024-01-19 13:47:05 5.55 KiB vs 5.59 KiB 2024-01-19 13:47:17 would agree that the space it takes it not good 2024-01-19 13:47:55 would probably do something like nixos where maintainers get declared in one file and then used as a funciton for package manager to output and for CI to manage 2024-01-19 13:48:19 worse is that field is often missleading 2024-01-19 13:48:52 ok, now is really time for lunch 2024-01-19 13:49:14 and for me to do school 2024-01-19 16:56:06 clandmeter: ping 2024-01-19 16:56:41 For gitlab 16.7, I need to make a choice. There is this prometheus-mmap gem that now requires a rust compiler and llvm toolchain to compile 2024-01-19 16:57:44 I have found an unofficial patch to patch out the usage of that Gem, I still need to test for any issues, or installing the necessary build-time dependencies in the ruby container during build 2024-01-19 16:58:09 (increasing build-time, space usage, etc) 2024-01-19 17:55:44 I'm faced with the same dillemma on my gitlab-foss aports. Could you share the unofficial patch? 2024-01-19 17:56:54 ayakael: I found the patch here: https://gitlab.com/gitlab-org/gitlab/-/issues/437903 2024-01-19 17:57:29 https://tpaste.us/yYJd This is the patch I use 2024-01-19 18:04:16 Thanks! For now I'm going with removal route. I'll tell you if I encounter any errors. 2024-01-19 18:04:24 ayakael: thanks 2024-01-19 18:22:21 ikke: it looks like 16.8.0, the version that I'm upgrading to, has other gems that need rustc (glfm_markdown). 2024-01-19 18:22:48 oh, fun 2024-01-19 18:23:30 So it seems avoiding it not possible then 2024-01-19 18:27:41 Doesn't seem so. Gitlab seems to now be on the rust train. 2024-01-19 18:28:52 I really hope I won't have to maintain specific rust versions to build gitlab. I already do that for ruby. At least with gitlab 16.8, they now use ruby 3.2. 2024-01-19 18:29:03 oh, they are 2024-01-19 18:29:13 Last time I checked, the were still working on 3.1 2024-01-19 18:29:32 Hmm, well 16.8.0's .ruby-version file now says 3.2.2 2024-01-19 18:29:33 and I still have Gemfile.spec patches to make it work with 3.1 2024-01-19 18:29:40 16.7.0 was using 3.1 2024-01-19 18:29:46 they* 2024-01-19 18:30:21 Ok, maybe I can remove those patches then 2024-01-19 18:30:50 My instance is running 3.0 though, I'll spawn an edge version of it and see if anything breaks with 3.2 2024-01-19 19:43:57 When switching to 3.1, I had an issue with outbound network requests failing because of dns binding protection 2024-01-19 19:44:17 For some reason it failed, preventing things like webhooks to work 2024-01-19 20:19:36 KREYREN_: 2024-01-19 20:19:36 !59249 2024-01-19 20:29:17 I remember something along those lines when I upgraded to 3.1 with 16.4 release with some patching, but that was when upstream still had .ruby-version at 3.0. 2024-01-19 22:17:38 mps, noted i will be using the armbian u-boot for now and then flash that on the device and use it as a daily for few days before merging in pmos 2024-01-19 22:18:46 I will merge it tomorrow if maintainer miss it 2024-01-19 22:19:58 I'm sure it fixes at least one A64 machine 2024-01-19 22:20:49 s/machine/board/ 2024-01-19 22:24:01 mps, btw you said 6.7.0 doesn't work? 2024-01-19 22:24:10 and 6.6.0 ? 2024-01-19 22:24:23 bcs they do on my end 2024-01-19 22:25:01 last kernel which works for me on teres-i (A64 SoC) is 6.5.x series 2024-01-19 22:25:50 linux-lts didn't work for me either but linux-postmarketos-linux works on my end for these versions 2024-01-19 22:26:48 I could put APKBUILD and config somewhere tomorrow is you or anyone else want to test alpine on it with kernel I've built 2024-01-19 22:27:02 i will be happy to 2024-01-19 22:27:33 s/is/if/ 2024-01-19 22:28:33 yep yep 2024-01-19 22:29:27 I don't like to put my local git repo in public because it is full of mess 2024-01-19 22:30:03 heh i have it backwards i have all the mess on my profile and clean code in organizations :D 2024-01-19 22:31:10 KREYREN_: thank you for helping alpine to be better on arm 2024-01-19 22:32:04 npnpnp thank you for helping me ^-^ 2024-01-19 22:32:31 find me reference designs and plans to the riscv you work on and i might make it better on riscv too xD 2024-01-19 22:32:53 bcs my nixos setup expects a fallback distro and armbian is too unstable and alpine/pmos so far great great replacement 2024-01-19 22:33:14 we have #alpine-riscv64 on oftc 2024-01-19 22:33:47 joined~ 2024-01-20 05:00:36 ikke: getting weird errors when trying to build those new gems on gitlab 16.8.0: 'error: error writing dependencies to [...]' then I get 'No such file or directory (os error 2)'. I can't reproduce it out of CI environment (log: https://lab.ilot.io/ayakael/user-aports/-/jobs/7612 ) 2024-01-20 11:14:23 KREYREN_oftc: just tested upgraded ATF and u-boot from alpine CDN and it works fine 2024-01-20 11:15:03 you can inform pmOS people so they could test on other machines which need this fix 2024-01-20 15:20:13 ikke: we are missing keys in mksite? 2024-01-20 15:20:23 looks like one of the rv keys is not added 2024-01-20 15:21:23 What keys? 2024-01-20 15:21:36 alpinelinux.org/keys 2024-01-20 15:22:51 Yes, i meant what keys are missing? 2024-01-20 15:23:06 We only have one builder, right? 2024-01-20 15:23:14 but we have 2 rv keys 2024-01-20 15:23:30 i dont know where the seconds came from 2024-01-20 15:23:38 some change in apk? 2024-01-20 15:23:43 or libc 2024-01-20 15:36:53 ikke: https://gitlab.alpinelinux.org/alpine/aports/-/commit/aab68f8c9ab434a46710de8e12fb3206e2930a59 2024-01-20 15:45:03 Ah ok, after rotating them 2024-01-20 17:09:38 mps, checking 2024-01-20 17:38:20 KREYREN_oftc: no hurries. btw did you (or someone else) posted posted url to patch for teres-i DTS to fix display in newer kernels 2024-01-20 17:38:48 mps, i posted the patch 2024-01-20 17:39:09 https://dpaste.org/pqGvV 2024-01-20 17:39:36 testing use only, avoid submitting that to downstream or upstream 2024-01-20 17:39:44 bcs waiting for thermal devs to figure out how they want to proceed with things 2024-01-20 17:40:42 possible risk of hardware damage due to wrong critical temperatures configured so be aware 2024-01-20 17:40:57 ah, this is about temperature. I though it is for fixing blank display for 6.6 and 6.7 kernels 2024-01-20 17:41:27 the display is not known to be broken on those releases if it is then it's likely a configuration issue 2024-01-20 17:41:58 display doesn't work for me on 6.6 and 6.7 2024-01-20 17:42:00 and both kernels tested on postmarketos without display issue as long as sane u-boot is provided 2024-01-20 17:42:10 dmesg? 2024-01-20 17:42:36 https://tpaste.us/vNkk 2024-01-20 17:43:44 mps, what kernel package is that 2024-01-20 17:43:57 do you get display init during u-boot phase 2024-01-20 17:44:02 local build 2024-01-20 17:44:17 show config 2024-01-20 17:44:30 yes, display works in u-boot 2024-01-20 17:45:08 [ 0.209280] thermal_sys: Failed to find 'trips' node 2024-01-20 17:45:08 [ 0.209323] thermal_sys: Failed to find trip points for thermal-sensor id=1 2024-01-20 17:45:17 weird that it boots 2024-01-20 17:46:03 mps, the issue seems to be with wrong DRM config 2024-01-20 17:46:18 around 8.040686 2024-01-20 17:46:21 could we discuss this in private, to not pollute this channel 2024-01-20 17:46:39 would be probably better for this to be logged bcs it's useful info 2024-01-20 17:46:59 not for #alpine-infra 2024-01-20 17:47:13 where then 2024-01-20 17:47:31 #alpine-arm ? 2024-01-20 17:47:53 it exists? 2024-01-20 17:48:07 nope but it should 2024-01-20 20:31:19 ACTION stares at clandmeter  2024-01-21 12:34:35 mps, I will submit patches to make the linux-postmarketos-allwinner to support EFI which seems to work without an issue on my end including linux 6.7.0 2024-01-21 20:25:51 nu: what would it take to connect to github.com (IPv4-only) from Alpine infra (which from what I understand is IPv6-only)? :> 2024-01-21 20:26:44 or maybe I could use different source for that container and it will work fine in CI.. 2024-01-21 22:00:30 isnt there already v4? according to the last discussion with mps there should be, but for some reason it didn't work at the time 2024-01-21 22:08:41 nu[m]: should we receive an ip via dhcp? the interface was still set for a static address 2024-01-21 22:09:17 or do you mean dns64? 2024-01-21 22:13:49 no, i wasnt thinking about the network i provide, but the container networking you manage 2024-01-21 22:14:30 i thought you get v4 via vpn or nat my nat further ^^ 2024-01-21 22:15:04 We had that for a brief time, until we received an actual public IP there 2024-01-21 22:15:29 Right now there is no IPv4 functionality at all 2024-01-21 22:15:43 how can i help? dns64 would be a solution, dhcp could be another 2024-01-21 22:16:15 Even NAT would work for us 2024-01-21 22:16:32 nat64 or nat44? 2024-01-21 22:16:38 nat44 2024-01-21 22:16:58 you already have that with a static v4 config 2024-01-21 22:17:06 There is something not working with mpgre over ipv6 2024-01-21 22:17:18 you can freely reuse any ip from that /24 2024-01-21 22:17:37 nu[m]: I have not received any details about that /24 2024-01-21 22:21:20 static v4 config: ip: 10.159.1.2/24, gw: 10.159.1.1 2024-01-21 22:22:15 mby ill create a wikipage for these kinda infos 2024-01-21 22:22:24 nu[m]: thanks 2024-01-21 22:22:30 np 2024-01-21 22:22:35 I also record these things in our netbox instance 2024-01-21 22:28:00 nice, i want to record some details about the mpgre issue, and see if i could help decomissioning legacy ip 2024-01-21 22:32:14 It works, can ping ipv4 now 2024-01-21 22:32:28 and dmvpn is working as well 2024-01-21 22:33:42 pj: ^ 2024-01-21 23:32:01 thanks, you can delete the container though, I'm not going to be able to debug the issues anymore 2024-01-22 00:52:23 ikke: when upgrading to 16.8.0, you may encounter an issue with glfm_markdown not finding its rust library. This commit is supposed to fix this: https://gitlab.com/gitlab-org/ruby/gems/gitlab-glfm-markdown/-/commit/a43ce9d9c5363829e56167a542daac8d1636c57a I've yet to test it out as I got another workaround going. Patching to version 0.0.11(16.8.0 2024-01-22 00:52:23 uses 0.0.10) picks up this commit. 2024-01-22 00:53:20 My workaround is this: https://lab.ilot.io/ayakael/user-aports/-/merge_requests/338/diffs?commit_id=4654506667c2d7c229b298f88749932254643a4c 2024-01-22 02:27:59 mps, the u-boot-24.01-r3 doesn't have a display on my end 2024-01-22 02:37:02 seems like it's not doing the PMIC init maybe it didn't grab the updated dependency? 2024-01-22 05:00:12 mps, changing the u-boot pkgrel on 3 seems to fix that.. i guess the alpine package is cached with the old arm-trusted-firmware 2024-01-22 12:44:49 oh you set that on 3 already 2024-01-22 12:44:51 ehh 2024-01-22 15:27:20 mps, i had outdated repo nwm 2024-01-22 17:00:02 ayakael: thanks for the heads-up 2024-01-22 17:23:25 mps: You should be able to reach your containers over the vpn again 2024-01-22 20:19:19 clandmeter: going to move some builders from usa-bld-1 to che-bld-1 2024-01-22 20:19:25 ncopa: ^ 2024-01-22 20:19:36 ok 2024-01-22 20:19:55 i guess it makes sense to move all builders away from the usa one 2024-01-22 20:19:59 and use it just for ci 2024-01-22 20:20:28 until we need to give it back 2024-01-22 20:20:32 yes 2024-01-22 20:21:09 When we have the space, I can increase the size of the CI vms 2024-01-22 21:57:51 sounds good! 2024-01-23 01:30:31 mps: https://wiki.postmarketos.org/wiki/OLIMEX_Teres_i 2024-01-23 01:30:46 what do you think 2024-01-23 11:41:45 ncopa: build-3-17-* have been movede to bld-che-1 2024-01-23 15:56:01 Good! Thanks! 2024-01-23 15:56:51 I'll follow up with the others 2024-01-23 17:46:53 ikke: could you tell me IP addesses of my containers on new arm machine 2024-01-23 17:50:50 They should've kept the same 2024-01-23 17:52:28 172.16.27.110 / 111 2024-01-23 17:54:37 ah, yes. works. thanks ikke 2024-01-24 20:36:22 ncopa: all 3.18 arm builders have been moved as well 2024-01-24 21:44:27 awesome! thank you sir! 2024-01-25 21:20:59 im not able to figure out the ip address of build-3-17-armv7. what is its dns name? 2024-01-25 21:21:04 or ip addr 2024-01-25 21:21:53 I wonder if it would make sense to have CNAME build-*.alpin.pw or something, but it is some work to keep them up to date 2024-01-25 21:22:02 when moving from one site to another 2024-01-25 21:23:13 ncopa: I have had the same idea before 2024-01-25 21:23:41 im trying to figure out the exact LDFLAGS we are using for build-3-17-armv7 2024-01-25 21:23:46 ncopa: build-3-17-armv7.che-bld-1.alpin.pw 2024-01-25 21:23:56 thanks! 2024-01-25 21:26:55 and build-3-18-armv7? 2024-01-25 21:26:59 should be same 2024-01-25 21:27:07 ssh: Could not resolve hostname build-3-18-armv7.che-bld-1.alpin.pw: Name does not resolve 2024-01-25 21:27:40 build-3-18-armv7.che-bld-1.alpin.pw has address 172.16.27.137 2024-01-25 21:28:45 thanks 2024-01-25 21:29:06 ncopa: would be nice if we could have DNS updated from netbox 2024-01-25 21:29:14 I keep netbox up-to-date 2024-01-26 03:32:05 ikke: Giltab 16.8.1 fixes the issues with glfm_markdown 2024-01-26 04:45:02 πŸ‘ 2024-01-26 08:24:56 ayakael: thanks for providing helpful info reg gitlab, appreciated! 2024-01-26 20:34:16 ikke: when you get a chance, can you please post a toot for alpine 3.19.1? 2024-01-26 20:34:26 sure can 2024-01-26 20:36:26 https://fosstodon.org/deck/@alpinelinux/111824167284670677 2024-01-26 20:36:39 https://fosstodon.org/@alpinelinux/111824167284670677 2024-01-26 21:14:43 seems like mqtt-exec.aports-build on build-3-17-aarch64 was stopped 2024-01-26 21:14:55 i have started it so it can build 3.17.7 2024-01-26 21:15:17 thanks 2024-01-26 21:15:49 I may have forgotten to enable the service again after moving it, so it would not auto start 2024-01-27 18:23:30 mps: you are using alpine on apple arm macs, right? is there any write-up/documentation on how do setup it up? 2024-01-27 18:32:34 nmeum_: yes, https://arvanta.net/alpine/install-alpine-m1/ 2024-01-27 18:33:44 but it is outdated, these days I'm preparing alpine installer and here is preliminary video https://www.arvanta.net/asahi/alpine-installer-1.mp4 2024-01-27 18:37:42 cool 2024-01-27 18:39:58 nmeum_: I'm not sure it works on all M2 machines 2024-01-27 18:40:28 on macbooks it should work 2024-01-27 18:43:26 I'm preparing video camera to record last 2 steps 2024-01-27 18:44:12 probably will finish this evening 2024-01-27 21:18:40 nmeum_: sorry, forgot to mention, when you set size for root FS installer usually don't set exact at size which is written on console but somewhat smaller, about 20% less (I didn't calculated exact value) 2024-01-28 11:49:43 ikke: whats up with the aarch64 shell runner? 2024-01-28 11:53:23 Don't know, seems to be running 2024-01-28 11:53:38 no its not active 2024-01-28 11:53:56 oh last contact 2024-01-28 11:54:08 is it because its using ipv6 2024-01-28 11:54:09 The container and gitlab-runner service are both running, but I don't see a logfile 2024-01-28 11:54:35 https://gitlab.alpinelinux.org/clandmeter/alpine-disk-image/-/jobs/1256824 2024-01-28 11:55:07 which host is it running on? 2024-01-28 11:55:13 che-bld-1 2024-01-28 11:57:43 i rebooted it but gitlab ci still says its offline 2024-01-28 11:59:10 the runner was paused? 2024-01-28 11:59:31 hmm weird 2024-01-28 11:59:51 looks like it runs now 2024-01-28 12:00:55 ok, I've updated the init.d file to write output to a log file 2024-01-28 12:01:31 interesting, now it fails to init qemu 2024-01-28 12:02:36 ok i need to go 2024-01-28 12:02:42 check it later 2024-01-28 12:07:51 clandmeter: seems to run fine? 2024-01-28 12:07:57 job succeeds 2024-01-28 13:04:40 clandmeter: when you have time, could we look at the e-mail to gitlab integration? For some reason gitlab is not processing these mails anymore 2024-01-28 18:47:03 ikke: sure, what is the problem? 2024-01-28 18:47:34 gitlab is not reading the imap server? 2024-01-28 18:48:01 correct 2024-01-28 18:48:20 I see e-mails in /var/vmail/gitlab.alpinelinux.org/incoming/mail/cur, but they don't seem to be processed 2024-01-28 18:48:23 but imap is running ok i guess? 2024-01-28 18:48:32 that's dovecot, right? 2024-01-28 18:48:44 uhm could be 2024-01-28 18:48:51 dont remember which imap server i used 2024-01-28 18:48:58 there are 2 options 2024-01-28 18:49:35 gitlab-inbound-email has 2 services, postgres and dovecot, both are running 2024-01-28 18:49:48 right 2024-01-28 18:49:50 checking it now 2024-01-28 18:50:41 I see the emails in the dovecot spool 2024-01-28 18:51:43 The reply by mail service has a green checkmark in the gitlab admin dashboard 2024-01-28 18:52:56 since when did it stop? 2024-01-28 18:53:07 mid december 2024-01-28 18:53:15 Probably after an upgrade 2024-01-28 18:54:30 docker container does show connections 2024-01-28 18:55:04 Just found the imap settingsd 2024-01-28 18:55:46 I do see in the mailbox logs that it's logging in 2024-01-28 18:57:19 where are the settings? 2024-01-28 18:57:44 its not like things are easier to find now with this new interface 2024-01-28 18:58:00 Those are in the config file 2024-01-28 18:58:05 gitlab.yml 2024-01-28 18:58:50 I run this: bundle exec rake gitlab:incoming_email:check RAILS_ENV=production 2024-01-28 18:58:56 and it says that it is not enabled 2024-01-28 19:00:30 which version are we on? 2024-01-28 19:00:44 https://docs.gitlab.com/ee/administration/incoming_email.html#email-ingestion-doesnt-work-in-1660 2024-01-28 19:00:52 16.6 2024-01-28 19:00:57 :) 2024-01-28 19:00:59 aha 2024-01-28 19:01:10 thanks :) 2024-01-28 19:02:12 So best is to upgrade to 16.7 2024-01-28 19:02:31 i guess that would solve this 2024-01-28 19:02:39 who knows what else gets broken :D 2024-01-28 19:03:54 We'll see when we get there :P 2024-01-28 19:04:09 a wise man ones said :) 2024-01-29 05:52:24 I would like to consolidate https://lab.ilot.io/ayakael/dotnet-stage0 on Alpine's GitLab to allow for broad collaboration. Every bootstrap release is about 300-500mb of artifacts, maybe totalling to about 3-5 gb. Is this something that would be allowed? 2024-01-29 05:53:28 I would like to eventually get an equivalent ghc-stage0 going for easier management of cross-arch bootstraps, but that's looking more like a summer projet right now. 2024-01-29 05:54:11 ikke: I think this is a question more towards you 2024-01-29 20:47:56 ayakael: I'd like to avoid storing large amounts of data on our gitlab instance 2024-01-30 21:19:29 telmich: seems like vigir23 is out again 2024-01-31 10:58:15 "nicoπŸ‡¨πŸ‡­: seems like vigir23 is..." <- I'll check it out later today and if needed replace it with a new vigir/router, if so it might take until end of week 2024-01-31 15:55:24 Note to myself: vigir23 is working "perfectly", it just has a broken VPN connection for IPv6 2024-01-31 15:56:08 ikke: Question, the server behind vigir23, it is still reachable, because dm-vpn works over IPv4, correct? 2024-01-31 15:59:54 2nd note to myself: the vpn is broken due to internal/mismatch of route with one upstream, resulting into blackhole of 194.5.220.0/24 -> to be fixed post dinner 2024-01-31 16:09:38 telmich: I didn't setup dmvpn yet on the remaining server (it used to piggybag on the other server) 2024-01-31 16:10:03 But the fact that our gitlab runners are offline means that both ipv4 and ipv6 are not working