<kazinsal>
I'm glad I never got into academia or the research side of computer touching because the idea of "publish or die" is just painful to me
frkzoid has quit [Read error: Connection reset by peer]
netbsduser`` has quit [Ping timeout: 252 seconds]
johnjaye has joined #osdev
Arthuria has quit [Ping timeout: 255 seconds]
netbsduser`` has joined #osdev
Gooberpatrol66 has joined #osdev
netbsduser`` has quit [Ping timeout: 255 seconds]
<geist>
playing with proxmox again. it's pretty neat
<geist>
certainly a better solution than what i'm hacking together with screen and shell scripts
<geist>
but i'm not sure i want to turn my linux server over to it
Osmten has quit [Quit: Client closed]
<kazinsal>
I think some of the guys at work want to use it for our lab environment
<kazinsal>
so I should probably figure it out
<geist>
yeah it's really just a plain debian install with a web front end to run qemu
<geist>
but it also supports clusters and migration and snapshots and whatnot, so it's pretty slick
<kazinsal>
I'd convert my home environment over to it from vmware but I need to figure out how raw disk passthrough works
<geist>
yah worth fiddling with
<kazinsal>
because on vmware I passed my spinning rust through directly to my "NAS VM"
<kof123>
> the idea of "publish or die" is just painful to me > real artists ship -- steve job, perhaps a play on real artists steal
kfv has joined #osdev
<geist>
i fiddled a bit trying to get raw punch through of some ethernet devices and whatnot, but that seems to be more dictated by precisely how your host machine divvies up pci devices across iommu domains
<kof123>
i think "research" is the only area without such
<geist>
ie, you need a separate iommu domain for each device to properly punch it through
<kazinsal>
so if I can just pass my disks through to a VM on proxmox then the VM can go "yep, I see the right disklabels [or whatever the NT storage VM equivalent is], good to go" and immediately mount the RAID
<geist>
and my old nehalem test box dont do it
<geist>
that is a good question... can you just punch a disk through
<kazinsal>
I'm still planning on swapping out my sandy bridge xeon box that makes horrendous noises out for my previous desktop mobo, which was an 8700K
<kazinsal>
so two fewer cores/four fewer hyperthreads, but more than making up for that in IPC and almost doubling the clock speed
<geist>
oh also a thing proxmox does do right is it has the whole memory ballooning thing
<kazinsal>
I think the all-core boost clock on the xeon box is 2.6 GHz and the all-core boost on the 8700K is 4.6 GHz
<kof123>
*"research" defined as some other dept. is bringing in the quick money
<geist>
ah yeah looks like you can pass a disk directly through but looks like it's command line
<kazinsal>
doable but not exposed to the GUI is fine, considering my home firewall is a friggin openbsd VM haha
<bslsk05>
pve.proxmox.com: Passthrough Physical Disk to Virtual Machine (VM) - Proxmox VE
<geist>
and i forget where but at the end of the day each vm is really just some .conf file that you can fiddle with directly, so it's not a bunch of binary crap at the bottom
<kazinsal>
sounds similar enough to the vmware side. the .vmx is just a glorified ini
<kazinsal>
but half the stuff in it is just... not documented, because proprietary hypervisor software
<kazinsal>
I guess I should probably switch to proxmox
<kazinsal>
seems my VMUG license expired
<geist>
yah i looked at it and noticed it's only a 60 day license now. i dunno what happens when it expires, but last i fiddled with esxi which was i guess about 4 years ago you could run it perpetually in eval mode
<kazinsal>
I think now if you hit 0 days on eval, your VMs keep running, but you can't power VMs on
<kazinsal>
so if you have a power outage, you're hosed
<geist>
huh, how did you get to 49 days? did you have a year old license and you just lucked out that you're near the limit?
Osmten has joined #osdev
<kazinsal>
yeah, I had a VMware Users Group license which is a one year renewal for full feature licensing on a not-for-resale agreement
<kazinsal>
and when that expires it drops you to the 60-day eval
<geist>
i do admit i was primarily interested in ESXi just because it was a different kernel, because i'm an osdev nerd
<kazinsal>
oh yeah totally
<geist>
but once that particular sheen wore off i was not as interested
<geist>
i had netware and banyan vines nostalgia
<kazinsal>
especially that weird shim they had from like, 4.x-6.0 where you could throw linux drivers at it
<bslsk05>
'Banyan Vines 8.5 (x86) running under VirtualBox' by Digital Archaeologist (00:03:50)
<kazinsal>
I did a lot of vmware stuff at work early in my career (I say, not quite ten years in) but these days so much is either cloud or hyperconverged that it's not worth keeping up on from a certification and vendor extension perspective
<kazinsal>
hell, I've touched more hyper-v than vmware recently, and that's horrifying
<geist>
oh interesting, was mentioning the other day and it turned out it was to the wrong irc channel (#riscv) but i noticed that with WSL2 you can now access the linux fs
<geist>
it shows up in explorer
<geist>
like //wsl.local/...
kfv has joined #osdev
<geist>
unclear precosely how that works
<kazinsal>
huh
<kazinsal>
haven't tried that, I should poke at that on my work box
<geist>
clearly it requires some sort of linux kernel stuff i guess, because the first time you do it it starts the VM image and the kernel will keep running as long as you're accessing it
<kazinsal>
I have hyper-v on that but not on my home machine
<geist>
so there must be some sort of daemon or kernel thread doing it in linux, but i dont see it in top, so dunno what's up
<kazinsal>
yeah, I wonder if that's being done by the in-kernel smb server
<geist>
yeah could be
<kazinsal>
I remember that being a thing that people were going "oh no why would you do that when samba exists"
<geist>
so now it's basically bidirectional: you can access windows drives with /mnt/c, etc and you can now access the linux disk from windows
<kazinsal>
if it works well I might even consider swapping from esxi to hyper-v...
<kazinsal>
my biggest use case for linux VM access is a build VM
<geist>
aaaaaand i noticed over in /mnt/lib/wsl/lib there are a bunch of nvidia cuda and whatnot utilities, so you can actually get to the vid card from inside linux inside the hyper-v
<kazinsal>
and right now I've got the horrendous hackery that is NFS on windows and it's just... bad
<kazinsal>
like, you've gotta write an /etc/passwd in your windows client that sorta matches the unix machine you're mounting a share from, but referencing the windows user accounts on the client machine
<geist>
that's another thing proxmox does that's pretty nice: you can set up a bunch of network fses at the cluster level. nfs, ceph, iscsi, etc. that lets you migrate vms across as long as their storage is shared
<kazinsal>
yeah, that seems handy for that migration setup
<geist>
i guess that's not really that exciting coming from esxi or whatnot, but its still pretty functional
<kazinsal>
I think for my own environment I'm going to have to go to best buy or whatever and buy a 4 TB USB hard drive, format it with ext4, plug it into the ESXi box, and move all the VMDKs over to it
<kazinsal>
then proxmox should be able to import those
<geist>
yah can you export from esxi to the precanned vm format?
<kazinsal>
because the ~3 TB of local disk storage on the ESXi environment uses VMFS6
<kazinsal>
yeah thankfully OVF/OVA uses vmware's disk format
<kazinsal>
it's the underlying filesystem on the physical disks that's the thing to work around
duderonomy has quit [Read error: Connection reset by peer]
<geist>
if you're not going to recycle the machine you can probably ftp or ssh it over
<geist>
scp
<kazinsal>
yeah, might check on the prices of some new M.2 SSDs and go from there
<kazinsal>
old machine is just SATA all the way through but the "new" one has a couple M.2 slots
<kazinsal>
so maybe once the last two weeks' overtime worth of paycheque comes in I might grab a couple 2 TB M.2s
duderonomy has joined #osdev
<geist>
looks like it might only support ovf, not ova
<geist>
but the way to bring it in is apparently `qm importovf ...`
<kazinsal>
ah, yeah, that's fine. ova is just an ovf + vmdk in a zip
<geist>
oh i see yeah you untar the ova and then import it
<kazinsal>
yeah
<kazinsal>
probably will end up moving all my VMs over to the proxmox environment except for the firewall VM, then just create a new one with opnsense or something
<kazinsal>
unless I can get work to cough up some "lab money" for a physical firewall for me
<geist>
yah i was fiddling with exactly that, running one of the VMs as a pfsense thing to expose a virtual bridge that the other ones used as their network
<kazinsal>
I'm trying to get them to buy me a juniper SRX320
<geist>
works fine, and you can set some rules to start that VM first, wait a few minutes, then start the other ones
<kazinsal>
4 gbps bidirectional firewall throughput is more than enough
<kazinsal>
I've only got a symmetrical gig here anyways
<kazinsal>
which I only ever hit when steam decides to update things
<geist>
yah i have a medium sized netgate pfsense one for my firewall
<geist>
i'd like to replace it but i can't justify it becaus eit runs just fine
<kazinsal>
totally
<kazinsal>
pipe dream is to one day have my own osdev hackery running my home firewall but y'know
<kazinsal>
first I need to actually get back into doing osdev hackery
<geist>
heh yeah i have this arm box right here with 3 2.5gb ports, that i should totally hack codes onto
<geist>
it'd be a perfect little router
<geist>
some RKblahblahbla thing
<kazinsal>
yeah this 4x1GbE realtek nic I impulse purchased is sitting here taunting me
<kazinsal>
or maybe it's a broadcom, can't remember
<geist>
the weirder the better
<kazinsal>
it was one of those "ooh, twenty bucks on ebay buy it now? sure, why not"
<geist>
esoteric ethernet chipset for the win!
<geist>
i still have a place in my heart for this wonky gigabit one i bought years ago back when gigabit was new
<geist>
and even started to write a driver for it. totally unheard of. i already forget the chipset
<bslsk05>
github.com: newos/kernel/addons/dev/net/ns83820 at master · travisg/newos · GitHub
<geist>
came and went, but you could get docs at the time
<kazinsal>
broadcom holds a special place in my heart out of respect for them just building switching ASICs that were good enough that the big network companies started adopting them because it was a tenth the price of doing custom silicon
<geist>
also omg trippy that github literally says '20 years ago' cripes
<kazinsal>
haha
<kazinsal>
that's kinda personally humbling as 20 years ago I was still a solid six or seven years out from writing code worth learning source control about
Nixkernal has joined #osdev
<kazinsal>
I shudder to think about what code of mine people will be seeing when it has the label "20 years ago"
<geist>
yeah
<kazinsal>
I think my oldest semi-public repo right now was a cheap riff on the concept of "microblogging" when that term came about, so it'll be 20 years old around the same time twitter is
<geist>
yeah tghat newos repo is definitely not git original. was originally perforce, then imported into svn, then imported into git
<Ermine>
I feel like revisiting my bootloader, but I have stuff to do
Griwes has joined #osdev
xenos1984 has quit [Ping timeout: 240 seconds]
duderonomy has joined #osdev
Ellenor has joined #osdev
xenos1984 has joined #osdev
<Ermine>
So do I guess correctly that kernel relocation boils down to putting image at the right place? So I need to allocate right pages or set virtual memory mapping?
heat_ has quit [Read error: Connection reset by peer]
heat has joined #osdev
<gog>
it could also do a self-reloc
<gog>
or be position-independent
<Ermine>
gog: multiboot2 allows to specify preferred loading address
<bslsk05>
github.com: SAMEthing/src/core/src/private/core.c at main · mcroddev/SAMEthing · GitHub
<mcrod>
and i can't possibly see why
craigo has joined #osdev
<mcrod>
SAMETHING_CORE_SEQ_STATE_AFSK_HEADER_FIRST doesn't actually fully finish generating, although it comes close, before moving on to SAMETHING_CORE_SEQ_STATE_AFSK_HEADER_SECOND
<mcrod>
ctx->afsk is stuck at data_pos=57 (correct), bit_pos=6 (wrong), and sample_num=18
<geist>
mcrod: re: perforce, well back in 1999 it was fantastic
craigo has quit [Client Quit]
<geist>
it's a big aged now, but i used it at various companies and personally through most of the 2000s until subversion started to replace it and then git shortly afterwards
<geist>
like you said it deals with binary assets quite well, especially with a local perforce proxy
<mcrod>
at work we still use perforce but we've been told company wide we're going to git
<geist>
also the merging capability is quite good
<mcrod>
since people legitimately struggle to use perforce
<geist>
but the 'you must check it out to merge it' is pretty annoying. svn at least is the same general model but lets you do more stuff on the server
<geist>
oh hah, if they strugged with p4, have fun with git
<mcrod>
well there's more resources for git on average
<mcrod>
not saying you can't read perforce's manual
<geist>
yeah nowadays for sure
<mcrod>
but often times you can just google "how do I do x with git" and get 5000 answers
<geist>
exactly
<formerly-twitter>
git allows you to make an absolute mess from a repo, but basic usage is pretty is pretty simple
<geist>
5000 *different* answers
<formerly-twitter>
provided you RTFM
<geist>
formerly-twitter: can you maybe, like, stick with one handle maybe?
<geist>
ktnhxbye
<formerly-twitter>
is that a microaggressin
<formerly-twitter>
<formerly-twitter>
i chose this one
<geist>
it's more annoying, took me a bit to figure out who you are
<formerly-twitter>
so i am not *that* predictable after all
<formerly-twitter>
you may notice i'm saying something positive about git
<formerly-twitter>
why are you chagning the subject
<formerly-twitter>
:X
<geist>
mcrod: last time i used p4 was at a game company, and back in 2007 we had like 50GB of assets to check out
<geist>
and that worked nicely
<mcrod>
yeah perforce still dominates the gaming industry
<geist>
well, okay, worked nicely enough that it worked, vs being a total disaster
<mcrod>
although from what I hear it seems like people just use p4 for assets and git for code
<geist>
yah that would make a lot of sense
<formerly-twitter>
interesting, i thought perforce died off completely
<mcrod>
no not at all
* geist
heads to the orifice
<mcrod>
it's definitely more 'niche' these days
goliath has joined #osdev
admiral_frost has quit [Quit: It's time]
<formerly-twitter>
at this point svn is probably the biggest use of svn
<formerly-twitter>
user
<Ermine>
googled vsyscall, found docs for some mail server that requires vsyscall. el_risitas.mp4
<Ermine>
onlyoffice one actually
craigo has joined #osdev
Thedarkb has joined #osdev
gog has quit [Quit: Konversation terminated!]
<formerly-twitter>
mutex_enter(&dn->dn_mtx);
<formerly-twitter>
if (zfs_refcount_is_zero(&dn->dn_holds)) {
<bslsk05>
wiki.osdev.org: Windows NT - OSDev Wiki
<netbsduser``>
formerly-twitter: it's the job of the virtual memory system
<netbsduser``>
so a paper about a virtual memory system would be helpful
bauen1 has quit [Ping timeout: 255 seconds]
<netbsduser``>
the principle all modern OSes are built on was succinctly put in the sunos vm paper like this: "the OS deals with files, and main memory is nothing but the cache of files". the two sorts of file are regular files and pagefiles, the latter backing anonymous memory
<netbsduser``>
so dealing with conditions where main memory is running low consists in putting back the cached contents of files to their homes on disk
bauen1 has joined #osdev
<Ermine>
netbsduser``: thank you
<Ermine>
I see that windows has dtrace
<heat>
yep
bauen1 has quit [Ping timeout: 240 seconds]
<heat>
i have tried porting dtrace and i must admit it's hell on earth
<heat>
the great UNIX ENGINEERS looked at dtrace and thought "hmm, how do we port this to CrapBSD? Oh, I know, let's provide various bastardized versions of solaris headers"
<Ermine>
unix wizards
<heat>
the only magic is that their systems continue to work
<heat>
nowhere in 20 years did they think it'd be a good idea to add things like *abstraction layers*
<heat>
no! let's pretend we're solaris, also add some ifdefs around the code boss!
<Ermine>
Will you port it to unix given the need for abstraction layer?
<heat>
what?
<Ermine>
oh, to onyx*
<heat>
i mean... its an idea, but it's seriously horrendous
<heat>
the kernel bits aren't all that interesting honestly, easily reimplementable
<heat>
dtrace bytecode isn't even JITted, just interpreted...
<heat>
the bit i really wouldn't enjoy doing is the dtrace userspace crap, with the compiler and all
<Ermine>
compilers are oof
vai has joined #osdev
<heat>
tbf dtrace isn't really a language, it's just some awk-ish thing
<vai>
hows development guys... lots of planning too it must be as well?
pg12 has quit [Ping timeout: 255 seconds]
pg12 has joined #osdev
leon has quit [Ping timeout: 255 seconds]
GeDaMo has quit [Quit: That's it, you people have stood in my way long enough! I'm going to clown college!]
leon has joined #osdev
danilogondolfo has quit [Remote host closed the connection]
divine has joined #osdev
gbowne1 has joined #osdev
zxrom has joined #osdev
mahk has joined #osdev
Left_Turn has quit [Remote host closed the connection]
Left_Turn has joined #osdev
Left_Turn has quit [Remote host closed the connection]
heat has quit [Read error: Connection reset by peer]
heat has joined #osdev
Left_Turn has joined #osdev
Matt|home has joined #osdev
<mcrod>
ok I fixed all of my problems except for one
<mcrod>
story of our lives
zxrom has quit [Ping timeout: 255 seconds]
Left_Turn has quit [Ping timeout: 240 seconds]
Left_Turn has joined #osdev
graphitemaster has quit [Ping timeout: 272 seconds]
heat_ has joined #osdev
heat has quit [Read error: Connection reset by peer]
frkazoid333 has joined #osdev
heat_ is now known as heat
goliath has quit [Quit: SIGSEGV]
graphitemaster has joined #osdev
[itchyjunk] has joined #osdev
Left_Turn has quit [Ping timeout: 240 seconds]
Left_Turn has joined #osdev
<heat>
formerly-twitter, what trick does freebsd use for the u-area now?
<heat>
you can't keep remapping that shit right?
Burgundy has quit [Ping timeout: 255 seconds]
duderonomy has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
duderonomy has joined #osdev
<kof123>
i feel now i should say niceperson should go with ex-twitter akin to ex-parrot. i said nothing because that sounds like taking a side, but this will lock down a nick
<kof123>
re: memory, the dumb answer might be just undercommit?
<kof123>
i.e. pre-reserve some scratch space
duderonomy has quit [Ping timeout: 252 seconds]
duderonomy has joined #osdev
agent314 has joined #osdev
agent314 has quit [Read error: Connection reset by peer]
Left_Turn has quit [Read error: Connection reset by peer]