libera/#devuan/ Wednesday, 2025-02-05

golinuxHi AlexLikeRock!00:12
amarsh04thought I'd have trouble transferring a boot hard disk from an old BIOS motherboard to a new UEFI motherboard but it worked fine11:01
fsmithredamarsh04, did you already have an efi partition on the disk? Or a bios-grub partition?11:20
djphUEFI "should(tm)" gracefully fallback to MBR as well (or is it the other way around -- GPT has a graceful fallback)12:14
fsmithredgpt with bios boot requires a special partition, at least 1MB with flag ef02 (bios_grub) and no filesystem on it.12:23
fsmithredthere's also something called "protective mbr" or something like that. I'm not sure how that works.12:24
fsmithredwith gpt ^^^12:24
djphyeah, my knowledge on it is pretty fuzzy ... half of hte information just sounds like "it'll just work, trust me"12:25
fsmithredI had to understand enough to get the live installer to work with uefi. There are still some rough edges, like when there are two efi partitions (not on the same drive).12:38
fsmithredand motherboards don't all match standards, so it can be a crap shoot.12:40
freem<fsmithred> gpt with bios boot requires a special partition, at least 1MB with flag ef02 (bios_grub) and no filesystem on it.12:40
freemthat is *grub*12:40
fsmithredwhat do you mean "that is grub"?12:40
freemusing a different bootloader will make this partition non needed, it is not related to gpt12:41
fsmithredyou mean using a bootloader on a different disk?12:41
freemno12:41
freemI mean that only grub have this silly requirement for a dedicated partition12:42
fsmithredah, ok12:42
freemas the partition type says itself, too: bios_grub12:42
amarsh04fsmithred - turns out that the new motherboard still supports legacy bios booting12:43
freemwith other bootloaders, you only need to (grub needs it, too, mind you) write a small file to the 1st sector of the HDD12:43
fsmithrednow I want to try extlinux with gpt12:44
freemin syslinux's case, it can be found in `/usr/lib/SYSLINUX`, those files have a size <= 440 bytes12:44
freem./lib/syslinux/mbr/gptmbr.bin this is the one I systematically "dd if=src of=dst"  :)12:45
fsmithredamarsh04, how is you disk set up. msdos, gpt, ???12:45
amarsh04gpt12:45
freemthe only cons of gpt+syslinux I have so far is that I need to copy the vmlinuz and initrd on the EFI partition, but I'm not sure it is really worth automating12:45
freemI think it's possible to chain-boot, but I have not gone this far.12:46
fsmithredrrq was working on that I think.12:46
freemI'd love to read about that12:49
avir327Hello! Updating pam (1.5.3-7+b1 -> 1.7.0-2) just noticed this change to 'pam_limits': "do not override systemd's limits by default; add the set_all option to restore previous behavior". Does this mean, every non-systemd system needs to get adapted manually, or can/will this be handled by the Devuan devs?12:52
rrqfreem: extlinux has a bootloader for gpt partition table12:53
rrqfreem: generally syslinux does not have a "disk" concept, so everyting has to be in the boot partition (that contains "ldlinux")12:54
freemyes, for the partition table. No idea how, but it manages to find the partition it needs to, and chain into the program there, which can read the config and load other modules12:56
freems/, for the partition table//12:56
freem(ignore that)12:56
freembut it does not requires a specific parition for the payload, which can thus be ext/vfat or others, I think?12:57
rrqextlinux requires ext[234]12:58
freemI have not looked at the internals of this stuff since more than a decade, and don't remember much. I never got much in this, neither12:58
freemyes, but, the MBR one installs seems to be able to detect iso/ext/vfat variants?12:59
freems/variants/filesystems/12:59
rrqyou mean the installer?13:00
freem find /usr -ipath '*syslinux*' -iname '*mbr*' -type f | upl -> https://p.mort.coffee/i2p I don't remember using other bootloaders13:00
freemah, right, there is an installer for non-gpt13:01
freemoh... yes, old parition system just had a bootable flag13:01
freemsorry, too long not using non-efi hardware (and I don't miss it)13:01
rrqI think you're mixing things :)13:01
freemI think so too13:02
rrqthe bios boot starts with a boot loader fitted into the first 448 bytes, and you chose that depending on partition table type13:03
rrqit includes an index for which partition to boot from, and that partition has the next stage bootloader, wchih interprets a me nu config etc13:04
freemrrq: to me, the bootloader is the part that gpt+EFI puts into the special boot partition, and what dos partition systems (and their limit of 4 "real" partitions) had in the 1st sector of the partition, which means less than 512 bytes. Grub2 needs a partition in gpt mode because they used to abuse a convention to not use the first 8 mebibytes or something of the partition13:04
freemyes, you are more accurate than me13:05
rrqEFI bios is different from legacy bios13:05
freemyes, there is a partition with a specific ID13:05
freemor part of ID13:05
freemand the bootloader is no longer limited to less than 500 bytes :D13:05
rrqyes, EFI bios boots by picking a bootloader file in its FAT filesystem13:06
freemdo you think it's possible with syslinux & friends to chain-boot from the FAT filesystem to the other partitions? That's something I have not done myself, but wuold ease my life quite a bit13:07
rrqno, syslinux doesn't include code do check other partition(s)13:07
rrqit can only enumerate the boot partition13:08
freemyes, but the bootloader-partition could get those written manually13:08
freemin the config I mean13:08
rrqthere's no syntax for partitions13:08
rrqsyslinux doesn;t have code for enumerating partitions13:09
freemso far (and since years) I copy the kernel files on the fat-partition, but it's not exactly convenient13:09
freembut then I don't update kernels everyday neither13:10
fsmithredisolinux boot menu can select partitions to boot13:11
fsmithredI haven't tried it lately13:11
rrqdo you have doc pointer for that?13:13
rrqafaik there's no syntax for partitions (above /)13:14
rrqand I don;t think the code includes any table of partitions13:15
rrqit does however include a lua intepreter so one could imagine someone implementing a lua table of partitions ...13:16
rrqthough it'd need to interface both efi and legacy partition enumerations13:17
fsmithredchain.c32 hd0,013:18
fsmithredso not booting directly. Also the old syntax was: localboot 0x8013:18
fsmithredboth are in the isolinux boot menu from refractasnapshot13:19
rrqnot available for efi bios... maybe they work for extlinux?13:22
rrq(0x80 is legacy bios refrence to "first harddisk", and localboot runs the legacy boot procedure stamped on that)13:23
* freem wonders if those stuff are still maintained13:23
rrqlikewise for chain32; it runs the legacy boot of the targeted partition13:24
freemsyslinux.org indicates last news in 201413:24
freembut can't find a link toward a forge, so maybe there are changes under the hood?13:24
fsmithredI'm gonna test13:25
rrqI believe last packaged version by devuan is hte most recent13:25
freemI think so too. This is not stuff that needs many changes neither13:26
fsmithredshhh!13:26
fsmithreddon't let them know.13:27
rrqhttps://git.devuan.org/devuan/syslinux13:27
freemfsmithred: lol don't worry, I heard systemd already have a bootloader13:27
freemthx rrq13:27
rrqbranch suites/unstable13:28
freemhm... I'm curious, what is the name of the installer package, if there is such? So that I cna find the repo, I'd like to peek at how it's done13:29
rrqdevuan/installer-iso13:30
freemah, thx13:30
rrqbranch main13:31
freemyou mean master?13:31
rrqthe installer includes 3 different and intermixed boot equipments13:31
rrqyes13:31
freemiso, mbr and gpt?13:32
freemor iso, mbr/gpt, and nfs?13:32
freem(or maybe I'm completely off)13:32
rrqit's a hybrid ISO13:33
rrqwith DOS partition table13:33
freemfun.13:33
rrqit's made to handle efi or legacy with media as disk or cdrom, with syslinux/isolinux and a single boot menu configurtaion13:34
rrqthe hybrid setup is El Torrito13:35
rrqcalled "El Torrito"13:35
freemrings a bell yes13:35
rrqwhich allows for an embeded FAT/EFI filesystem less than 32MB in size13:36
freemis that normally impossible?13:36
freemI always have assumed FAT to be damn small... 0.7MB or close... since I remember the values of my childhood 1"44 floppy disks13:37
freemhuh... 1"4, well, 1.44MiB raw, and 1.38 avail on the fat13:37
* freem reading https://wiki.osdev.org/El-Torito now13:38
rrqwell, one pair of kernel+initrd is some 40++ MB nowadays13:39
freemoh, ok, it's not because of EFI or FAT13:39
freemunderstood, then13:39
freemthanks for clarification :)13:39
freemso there is a really tiny OS installed on the iso, which boots the system sitting in a GPT/DOS partition table?13:40
freemsorry, i'm a bit tired, maybe I just fail to understand. I'll look myself, will be less noisy this way13:41
fsmithredfrom a live-usb, chain.c32 hd1,0  worked (booted the hard disk)13:42
freemthat stuff interests me because I have my own little scripts to build my own installer, and it could definitely be improved13:42
rrqthere is a preamble initrd that has modules to handle drives, so it can find the actual installer (an 84MB initrd), unpack ath switch to it13:42
rrqath=that13:43
freemnotably I got surprised yesterday that OVH no longer allows me to boot from an iso I'd send them, so now I need a way to remote install from a debian rescue environment...13:43
rrqno ath=and13:43
freemso I could not use the devuan iso yesterday :'(13:43
rrqcheck out https://git.devuan.org/rrq/softboot.git :)13:44
freemthey also put everything in one partition, so can't even tinker from the default system they install :'( but that is off-topic /whine13:44
freemI was curious to know if the installer still requires a constant presence13:45
rrqsoftboot can warp off a systemd filesystem onto another init filesystem13:45
freemneat13:45
fsmithredthe live installer might work in that situation. It uses rsync to copy the running system.13:45
freemI will definitely read that code13:45
fsmithredyou'd need to set the target to your remote disk partition13:46
rrqthe installer requires access to the ISO to fluff itself up with more modules than in the 84MB initrd13:46
freemwell, I just removed dust from my old scripts and memory, and tinkered my way. When I start from a debian rescue, as long as I have network access, I can do everything I want.13:46
freemI think I'll redo my scripts to rely on drist, for this kind of situation. I already have a suboptimal template system ,just need to improve on this13:47
freemheh, nice die() variant13:48
freemyes, I need to read that when feeling better, I'm really curious at this "involuntary transfer of the current PID 1" thing :D13:54
freemI thught killing PID1 would make kernel quite unhappy13:55
rrqthe signal is handled by the process, which reruns "itself" via its "$0 pathname"13:56
freemheh. Neat.13:56
rrqso if you make that pathname be something else, then the process will exec that13:57
freemI love that trick13:57
rrqit's useful :)13:58
freemcan most init do that?13:58
rrqworks similarly with sysvinit; I don;t think runit does a re-run though.13:59
rrqI've only looked at sysvinit and "that other one"14:01
freemsounds like a useful feature in general, I'll check if runit have some code for this14:01
rrqI believe sysvinit included re-execution as a "safety guard" for possble memory leaks or potential resource hogging...14:04
freemwould make sense imo14:05
rrqand it also handles "live upgrade"14:05
al1r4dfresh kde install and i got 100% plasma shell :(16:02
al1r4dis it normal?16:02
al1r4dhttps://ibb.co.com/qYfwzc7y16:02
gnarfaceon daedalus? i didn't look at the screenshot but i think so...16:06
al1r4dgnarface: excalibur ceres16:14
al1r4dweird, i cant turn on fn keypad thinkpad16:14
al1r4dmaybe i should switch to gnome16:15
gnarfacefirst thing i would do is verify you've got the correct keymap16:15
nemoI'm a MATE fan myself. gave gnome3 a try for a while, but, I like the pretty good featureset, and traditional desktop layout of MATE16:16
nemoI do use the fancy searchable menu sometimes16:16
nemoand optional acceleration is handy too16:17
al1r4dgnarface: i never have this problem before16:17
al1r4di just migrated to kde from xfce16:17
al1r4dxfce, i dont have a problem16:17
gnarfaceal1r4d: some window managers have their own separate setting for keymaps and some don't16:18
gnarfacebut if you're on excalibur, regressions are likely too16:19
al1r4dgnarface: thx for explanation16:36
al1r4dnow i'm on gnome16:36
gnarfaceal1r4d: well you probably could have just changed the setting in kde too, but whatever works...16:54

Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!