IRC channel logs

2023-04-23.log

back to list of logs

<luna>ACTION is watching the talk about your stuff at LAS 2023 atm
<doras>:)
<gforce_de1977>LAS 2023 = lean, agile, scrum? which talk? (https://lean-agile-scrum.ch)
<fossy>Mikaku: nice, thanks for your work! I'm not sure why keyboard doesn't work, but that's not specific to Fiwix kexec - I can repro that
<fossy>i -think- i had keyboard before, but that was a few days ago, i'm uncertain hwat configuration that was working in
<stikonas[m]>gforce_de1977: no, LAS is Linux App Summit
<gforce_de1977>ok, that was: https://conf.linuxappsummit.org/event/5/contributions/177/ ("Bootstrappable Freedesktop SDK")
<roconnor>heh, naming is hard: mes-m2 is a mes executable compiled with m2; mes is a mes executable compiled with mescc; mes-tcc is not a mes executable; it a tcc executable compiled with mes.
<stikonas>roconnor: mes-tcc can be renamed
<stikonas>it's entirely up to us
<stikonas>could be tcc-mes
<stikonas>oriansj, rickmasters: I've now pushed new GAS style defines for cc_x86
<stikonas>so I think we have now fully migrated away from them in stage0-posix
<roconnor>apparently coreutils printf is looking for lib/charset.alias
<roconnor>more specifically looking for lib/mes/charset.alias. :)
<roconnor>It's harmless though because it just goes on without it.
<rickmasters>fossy: Unfortunately, my attempts to speed up writing a large ext2 partition from builder-hex0 have failed so far.
<rickmasters>I spent a long time on this due to the difficult of working in hex assembly and silly mistakes I made.
<rickmasters>Ultimately, writing more sectors per BIOS request resulted in no significant speed up.
<rickmasters>It still takes 22 minutes to write the image.
<rickmasters>There are other techniques I can try which have been discussed here,
<rickmasters>but I've decided to take a break from working on that for now.
<rickmasters>.
<rickmasters>Yesterday I finally started exploring how to boot Linux.
<rickmasters>Unfortunately, the Linux kernel we build is neither ELF nor multiboot compliant.
<rickmasters>So, the Fiwix kexec mechanism is pretty far away from directly booting it.
<rickmasters>But your attempt to launch grub seems more within reach and maybe more practical.
<rickmasters>I've reproduced booting the image you provided and the keyboard problem.
<rickmasters>Are you saying you also had keyboard issues outside of Fiwix?
<rickmasters>Would you mind sharing which version of grub you built,
<rickmasters>how you built it and a brief description of your build environment?
<rickmasters>.
<rickmasters>One idea I've been toying with is to prepare a bootable hard drive (sysc)
<rickmasters>with grub on it, loading the MBR manually and jump to it in real mode,
<rickmasters>like the BIOS, so grub just boots normally.
<rickmasters>The benefit of this approach is that live-bootstrap could produce a
<rickmasters>a somewhat normal operating system that can be rebooted and evolved further.
<rickmasters>I suppose I could try to write the sysb image to disk and get grub to load it
<rickmasters>as the initramfs. But if sysc becomes a bootable drive then I'm not sure of the
<rickmasters>need for sysb. I'm wondering what was the rationale for using sysb to populate
<rickmasters>sysc rather than preparing/populating sysc directly from sysa?
<rickmasters>Was there a problem with kexec'ing Linux directly to sysc (without a sysb ramdrive in between)?
<rickmasters>Anyway, using Grub seems promising although it assumes we have what we need to build it, which remains to be seen...
<rickmasters>The End.
<muurkha>rickmasters: sounds reasonable
<gforce_de1977>roconnor: I also stumpled upon the naming scheme while explaining it in out hackerspace. yes: tcc-mes + mes-cc (maybe)
<stikonas[m]>rickmasters: sysb was added to remove assumption of hard disk support in bootstrap kernel
<stikonas[m]>Even Fiwix only supports a fraction of storage media that Linux does...
<rickmasters>stikonas: Yes, but we've been doing a lot of work recently on builder-hex0 and Fiwix to put hard drive support into the bootstrap kernels.
<rickmasters>stikonas: So that seems inconsistent, TBH.
<stikonas>rickmasters: I was thinking e.g. about USB stick, as far as I understand, it wouldn't work on Fiwix
<rickmasters>Not sure I understand. None of the bootstrap kernels are assuming USB support so that's consistent.
<stikonas>if BIOS supports USB, then builder-hex0 would just work with it
<stikonas>and then if fiwix runs from ram it would just work
<rickmasters>Sorry that I'm not understanding. I think I need to ask a few more questions.
<Mikaku>fossy: I've been working this weekend on fixing the problem of hard coded entry address and I succeeded, I'll push the commit tomorrow
<rickmasters>stikonas: Is there are requirement for the type of storage device for sysc?
<rickmasters>are/a
<rickmasters>stikonas: I'm trying to find where the sysc drive is mounted by sysb but I'm not finding it...
<Mikaku>doras: I've been watching your talk this morning, it was nice, congrats! :-)
<stikonas>rickmasters: well, ideally sysc is anything that Linux supports, but of course it's not a hard requirement
<stikonas>better to get at least something working
<stikonas>rickmasters: mounting is done here: https://github.com/fosslinux/live-bootstrap/blob/master/sysb/run.sh#L85
<doras>Mikaku: thanks! :)
<rickmasters>doras: Stupid question: how do I watch? Do I need to register. I wasn't able to click on anything that lead to a video.
<rickmasters>stikonas: Ok, I guess what I
<rickmasters>am getting at is that we currently have a lot of flexibility
<rickmasters>But the recent work on builder-hex0 and Fiwix will limit the sysa storage medium to only that which Fiwix supports.
<rickmasters>So maybe that's not a good idea and we should leave it in ram? (Also, disk is adding over 30minutes of time)
<Mikaku>rickmasters: you can watch the doras talk here <https://www.youtube.com/watch?v=jwMEDI4WsAE>
<rickmasters>Mikaku: Thanks!
<Mikaku>:-)
<stikonas>rickmasters: well, depending on whether we can make multiboot linux image...
<rickmasters>stikonas: Could you elaborate? Are you referring to that fact that we also need to kexec Linux from Fiwix to avoid writing a boot disk from Fiwix?
<stikonas[m]>I think so
<stikonas[m]>but like I said, that's not a strict requirement, let's see how it goes...
<rickmasters>Well I haven't done anything yet. I'm curious what fossy thinks.
<rickmasters>Staying in RAM until Linux boots is still possible but it will make saving memory a lot more difficult.
<rickmasters>I can try to make kexec to Linux work without involving a disk/grub. It's been my default plan but I thought I'd explore alternatives.
<rickmasters>I was assuming we'd boot Linux as-is and make Fiwix do what is necessary on the kexec / boot loading side.
<rickmasters>That's not multiboot. It's the Linux boot protocol https://www.kernel.org/doc/html/v5.10/x86/boot.html#