IRC channel logs

2022-01-02.log

back to list of logs

<damo22>youpi: i submitted my proposal before the deadline, but the date for acceptance has passed for the microkernel devroom and i havent heard anything, who should i contact?
<youpi>I'd say contact the person who pinged us
<youpi>I wouldn't be surprised that they are late
<damo22>ok i might just wait another day or so and then make contact
***duds-_ is now known as duds-
<luckyluke>youpi: would it make sense to change VM_MIN_KERNEL_ADDRESS to be over 4G, so at least 0x100000000 on 64 bit, and leave the 4G range for user-space?
<youpi>with a 64bit kernel, yes
<youpi>4G is still very tight by nowadays standards
<luckyluke>I meant only to run 32-bit code in user space, for a 64 bit user space maybe it makes more sense to go to the last 2G
<luckyluke>however it seems that some more changes are needed for that; for example, by linking a 64bit kernel in the last 2G it seems I need to change some asm files (e.g. ebx->rbx) and constants
<luckyluke>what would be the desired vm layout on 64bit? kernel in upper 2G and user below, or is there a simpler layout (at least for 32-bit userland)?
<youpi>what do you call "upper 2G" ?
<youpi>upper part of the lower 4G, or upper part of the large 64bit addressing space ?
<youpi>that being said I'd say it does not matter that much what we choose now
<youpi>we can change if ever need
<luckyluke>I meant upper 2G of the 64-bit addressing space, so the whole lower 4G would be usable for user-space
<youpi>damo22: uh, the new rumpkernel source is huge, is that all needed?
<youpi>(I know you have already excluded gcc+gdb+binutils)
<youpi>take care of patch fuzz, some tools reject it
<youpi>you can use quilt push -a --fuzz=0 for instance to check for that
<youpi>I guess there are large parts of external/ that aren't needed,
<youpi>possibly a whitelist would help here
<youpi>Robert had included only byacc, flex, mdocml, osnet and nawk
<youpi>I wonder whether crypto is needed either
<youpi>we also probably don't need all sys/arch :)
<youpi>also we'd want to remove non-free software
<Pellescours>damo I don’t know what changed but now I get SEGV_ACCERR (Invalid permissions for mapped object) when I try to print a partition in parted for a ahcisata disk
***Server sets mode: +nt
***valeriusN is now known as valerius
***valerius_ is now known as valerius
***biblio_ is now known as biblio
<youpi1>damo22: I uploaded the latest version to debian-ports. I added items in the changelog, this is however missing the "why" parts, each patch should have a little header telling why we have this patch, whether it'd be worth pushing it upstream, etc.
<youpi1>(and I queued the rebuild of the hurd package, it'd be good to fix the use of the .so files)
<Pellescours>there is a regression with rump, I’m not able to boot on it anymore, and everytime I try do do something with wd0 or a partition there is a lot of crashes
<youpi1>Pellescours: maybe wait for the rebuild with the newer librump
<youpi1>the use of the /dev/rwd version was posing problems with the previous librump iirc
<Pellescours>yes but I’m using a local rump I built
<youpi1>ah, then I don't know
<Pellescours>and it’s hard to debug a translator, when I try to use gdb I lost my VM internet due to shared irq
<youpi1>you could try to use the ne2k_isa qemu model and the gnumach ne2k driver