<nick64>
`"A boot parameter has been added: strict-devmem=0 will disable it now."`. : They must be talking about adding the parameter to redhat specific linux build I suppose?
<nick64>
I'm guessing it won't work for off the shelf ubuntu
<geist>
gentle reminder we should probably not bring that sort of stuff here
<geist>
waiting for my intel arc gpu. delivered sometime today, but given ups's track record it'll probably be in a while
<mjg_>
geist: ok, fair
* mjg_
will ramble elsewhere :p
<geist>
didn't occur to me until just now to double check the power supply of the test machine is sufficient
<mjg_>
i have to say my last non-laptop was over 12 years ago
<mjg_>
since then i do 0 fiddling with hw
<mjg_>
for personal use anyway :p
xvmt has quit [Ping timeout: 272 seconds]
<mjg_>
i slightly miss it though
<mjg_>
i remember being stoken installing an os onto a server pxebooting from my laptop
<mjg_>
now i would be annoyed to have to set it up
<mjg_>
g/stoken/stoked/
xvmt has joined #osdev
gog has quit [Ping timeout: 276 seconds]
<nick64>
Okay, I tried recompiling my kernel with CONFIG_STRICT_DEVMEM disabled, but I am still getting error trying to `mmap` after `open("/dev/mem")`
<mjg_>
and i doubt anyone knows off hand either, it is a rather obscure thing to do
<mjg_>
it made sense to ask as a time-saving measure in case someone happens to know
<mjg_>
but ultimately if you have any legitimate reason tow rite this code, you should also have the skillset needed to find out why this fails
<mjg_>
i do note a debugger like crash is probably doing it, so a quick check over there might help
<mjg_>
i don't remember if it is mapping this rw, it definitley is doing at least read access
<mjg_>
lol, great work ubuntu, as always
<mjg_>
WARNING: machine type mismatch: crash utility: X86_64 /boot/memtest86+.elf: X86
heat has joined #osdev
<heat>
nick64, i told you to go to kernelnewbies
<heat>
most of us aren't linux experts
<heat>
you'll get way better help there
<heat>
from people that, erm, make the kernel
<heat>
do you want greg kh? you'll get gregkh
<mjg_>
no i wonder if crash actually mmaps though. newfstatat(AT_FDCWD, "/dev/mem", {st_mode=S_IFCHR|0640, st_rdev=makedev(0x1, 0x1), ...}, 0) = 0
<heat>
peter ziljstra?
<heat>
he's there
<mjg_>
openat(AT_FDCWD, "/dev/mem", O_RDONLY) = 3
<mjg_>
read(3, "\0", 1) = 1
<mjg_>
really? nice
<mjg_>
positively surprised panda over here
<heat>
lots of prominent devs on #kernelnewbies
<heat>
even the best linux dev, j`ey
<mjg_>
is that the android backdoor guy?
<heat>
no idea what you're talking about
hmmmm has joined #osdev
bauen1 has joined #osdev
<nick64>
heat: kernelnewbies looks dead on the internet
<heat>
353 people online
<mjg_>
bigger than this dead channel for sure
<heat>
shit's dead
<heat>
no flamegraphs and performance dickmeasuring no party
<sortie>
no u flamegraph
<heat>
i'm 2 busy playing football manager 2 give a shit about stupid performance
<heat>
haha me go into manager of the football
<mjg_>
you playin fantasy football mofo
<heat>
i would be playin fantasy unix if it wasnt for this
<mjg_>
Fantasy Unix is a great name
<mjg_>
FU for short
* mjg_
writes a new kernel real quick
<heat>
hey didn't linux write a fantasy unix a bunch of times then
<heat>
linus****
<mjg_>
last Onyx commit oct 24. dead os.
<heat>
omg its LITERALLY DEAD
<heat>
killedbyheat.com
<mjg_>
FAANGH
<heat>
people that get panic attacks over new gmail layouts will literally die when I BREAK THE ABI
pesdauskes has joined #osdev
<heat>
it was just so i got promoted anyway, not my fault
bauen1 has quit [Ping timeout: 255 seconds]
<sortie>
osdev aka unix fanfic
<heat>
you're either unix or anti-unix here
<heat>
but at the end of the day we're all very opinionated people
* \Test_User
prefers to be both at the same time
<heat>
with severe NIH syndromes
<sortie>
I laugh at your strong opinions as my OS infrastructure powers the internet
<heat>
sortix routing? :eyes:
<sortie>
There is only one route
<sortie>
* -> sortix.org
<heat>
welcome to sortixorg
<heat>
this is sortixorg
<kazinsal>
sortix bgpd when
<sortie>
the only limit is yourself
<sortie>
kazinsal you won't be laughing when I just dump in a whole bunch of scary openbsd code
<sortie>
Pretty sure they actually got a daemon I can probably port
<kazinsal>
yep, OpenBGPD
<kazinsal>
they also have an osfpd
<kazinsal>
ospfd*
<sortie>
Although at this point I'm just ranting a bit in my head at ISO 9660 and trying to come up with some sort of stateless inode id scheme (32-bit integer that lets me uniquely and efficiently look up an inode (both file and directory) even if I throw away all the caches)
<heat>
just port quagga?
wand has quit [Ping timeout: 258 seconds]
<heat>
or frrouting apparently
<sortie>
I was using the offset to the directory entry although that blew up on the . and .. entries, so through about using the LBA of the data stream, but that makes it indirect to locate the inode metadata, and files won't have a good way of looking up the metadata
<heat>
mjg_, what strong opinion do you have on openbsd httpd
<sortie>
Hmmmm yeah I think I gotta stay with using the dirent offset as the inode id for iso 9660 and then just somehow fix the. and .. entries
<heat>
you could also, yknow, just not discard caches
<sortie>
heat, oh stop it, if I tell you I have a filesystem that never drops caches, you're just doing to tell me not to do that
<sortie>
:P
<heat>
sortie, i don't drop dentries (in the dentry cache) that are live
<heat>
so... it's a possibility
<mjg_>
heat: i have to say obsd rewrote file... except it turns out be a subset of the original
<mjg_>
do they speak performance in onyx?
<kazinsal>
drop the caches... on unmount
<sortie>
heat, well, are they live if some file or dir deep in the hierarchy is open?
<sortie>
Does that keep everything inbetween live?
<sortie>
It's a reasonable thing to say yes to
<sortie>
Although if you chdir into some dir and only use relative paths, doesn't necessarily need to be yes that's an answer
<sortie>
E.g. to find out the current directory in the shell and other places, it checks if the $PWD variable gets you the right place when updating it, and maybe otherwise just trusts it
<nick64>
heat: 300+ folks online and no chatter for the last 3 hours. At least 3 hours, because that was when I asked this there
<heat>
sortie, yes they are
<mjg_>
i think you need to differentiate between holding inode/vnode/whatever you name it hostage
<mjg_>
and just namespace linkage
<heat>
do they speak performance in onyx? <-- good q
<heat>
mjg_, what's file...?
<mjg_>
same as the difference between a duck
<mjg_>
oh, actal question
<mjg_>
i meant file(1)
<mjg_>
afair it was supposed to be a drop in replacement, except it turned out to not be
<heat>
sortie, a part of the point of keeping the whole tree live is if you want to go up the tree for some reason, like VFS operations (link, unlink, rename, etc) and dentry_to_path
<heat>
it also makes the logic significantly easier
<mjg_>
you want the tree so you can make fast lookups, which happen A LOT
<sortie>
heat, yeah, and I'm actually not QUIITE sure if that is actually always alive in Sortix
<sortie>
Although the design is that it might NOT be
<mjg_>
so how bout a will-it-scale showdown between sortix and onyx
<heat>
sortix doesn't have smp
<mjg_>
we can so single threaded
<heat>
and not really any vfs caches?
<sortie>
For all the other filesystems I have, ino_t basically lets them recover any file/directory statelessly
<mjg_>
wut?
<sortie>
heat, yep, so you can spawn N Sortix VMs instead
<mjg_>
sortie: sounds very old unix-y man, and that's not a compliment :)
<heat>
there's no dentry stuff last I checked, nor inode caches
<heat>
sortie has been focusing on making his OS do stuff
* mjg_
writes a paper "sad state of hobby operating systems"
<mjg_>
be a better linux
<heat>
i've been focusing on... what have I been focusing on again?
<mjg_>
you have been focusing?
<heat>
ah yes, i remember now
<heat>
football manager, that was it
* sortie
laughs at your cute performant hobby operating systems that can't be installed and run real workloads
<mjg_>
sortie: it can run a real microbenchmark
<mjg_>
good enough for my taste
wand has joined #osdev
<heat>
running microbenchmarks and getting flamegraphs is 3/4 of mjg_'s life
<heat>
and onyx feels that need perfectly
<sortie>
muhaha
<heat>
fills*
<mjg_>
also, is it an operating system if it is slower than obsd?
<mjg_>
real questions here
<sortie>
If it quacks like an OpenBSD
<mjg_>
sortie: is your first name theo?
<sortie>
I assure you SortixBSD is very legit
<mjg_>
if not you should probably work on it
<heat>
theodore d. sortie
<mjg_>
B Sortix Distribution
<sortie>
Very genuine discount Scandinavian “I can't believe it's NOT UNix”
<heat>
Be Sortix Distribution
<mjg_>
Better Sortix Distribution
<heat>
BeOS comeback confirmed!!1!!1!!1
<mjg_>
Behold Solaris Sibling
<mjg_>
shit
<sortie>
I have a vague plan to go to the Berkeley campus with a stack of Sortix CDs and hand them out to random people in an incident that will later be referred to as ‘The Berkeley Sortix Distribution‘
<mjg_>
Behold Solaris Descendandt
<heat>
.bss moment
<mjg_>
... except not
<heat>
sortie, if you figure out time travel you can become the BSD
<heat>
destroyer of worlds
<sortie>
I have git rebase close enough
<heat>
as in it literally gives you depression
<mjg_>
what we can do is perhaps a stability showdown?
<mjg_>
anyone with rename can run dirconc for example
<heat>
oh, I'll lose
<heat>
gg
<mjg_>
what you crash later than other systems
<mjg_>
if*
<mjg_>
then you win mofo
<heat>
no, I'll lose but later
<mjg_>
i'm telling ya man this used to *crash* openbsd
<heat>
I 100% need fuzzing
<mjg_>
i don't know what happens now
<mjg_>
user-induced kernel crash == s3cur1ty
<mjg_>
i think the yhave been told at some point
<mjg_>
penny for my thoughts, so 1 penny please
<heat>
my ext4 driver has been getting fuzzed for a few days now and they've already found like 6 or 7 crashes
<heat>
and that code is code I would rate as "Pretty Good" and as something I'm "Proud" of
<heat>
(and just 5kloc)
<mjg_>
serious comment: filesystems notoriously crash when fuzzed, so...
<mjg_>
it can be in a shape where "survives practical workloads" and the above will keep crashing it a lot
<mjg_>
not claiming you got that far :p
<heat>
but a practical workload in this case is trivially survivable
<heat>
it's uefi firmware
<mjg_>
so to be clear, you fuzz out an image
<mjg_>
and then mount it?
<mjg_>
and perhaps do some opso n it
<heat>
if I worked on a real IBV the bar for "Ship it" would be "works sometimes? wonderful"
<heat>
yeah I think that's what they're doing
<heat>
I haven't looked into it yet, just given out some patch reviews and helped out with a few things
<mjg_>
i guess fuzzingi to ut and mere find $mountpoint would already trip on top of stuff
<mjg_>
and probably crash your fsck too
<heat>
fuzzing is cool and great
<heat>
if I magically turned it on for my OS I would find soo many crashes
<heat>
which is why I want that
<heat>
hell, running complex things is already a good sign things are stableish
<heat>
which is why I'm confident sortix would beat onyx very easily in stability
<sortie>
The craziest thing I do is the nightly job that does a million-process self-hosting recompile of everything, that's a god damn stability test
<mjg_>
ok, i give you the silver medal in this 2 player competition
Matt|home has quit [Remote host closed the connection]
bauen1 has quit [Ping timeout: 272 seconds]
bauen1 has joined #osdev
<heat>
it broke so hard sortie is offline
<sortie>
:)
<heat>
it broke so hard sortie can only talk in smilies
<sortie>
:))
<mjg_>
maybe it renamed a test smile over his dictionary
<mjg_>
which means it worked at least
<mjg_>
albeit not the way it should have
moatx_ has joined #osdev
einkoder has quit [Quit: WeeChat 3.5]
heather has joined #osdev
dutch has quit [Quit: WeeChat 3.7]
moatx_ has quit [Quit: Leaving]
<sortie>
Alright neat my .. entries work properly in my ISO 9660 filesystem driver now, it did have to search the parent directory for its . record, but at least it's cached
<heat>
good shit
<sortie>
The cool new thing is that I can put my sysroot on an .iso and my driver on a normal live system can now mount the sysroot iso
<sortie>
Next up let's see if I can chain init the sysroot on an .iso
<heat>
:P
<mjg_>
sortie: now i'm genuninlycurious about dirconc on sortix. when you expect to run it?
<sortie>
mjg_, will do it tomorrow :)
<sortie>
It's 01:30 so yeah just got enough brain cells for my own misguided project
<mjg_>
buy insurance
<heat>
don't buy mjg_'s snake oil
spikeheron has joined #osdev
<heat>
you'll run dirconc and 2 hours later you're optimizing path walking with flamegraphs
<mjg_>
and your life is better for it!
<mjg_>
i'm tempted to bench haiku
<mjg_>
downloading now, but my shiternet means 15 mins
<mjg_>
... now 8
<heat>
be glad you've never went through downloading 30GB of football manager skins and faces
<heat>
the zips legitimately make windows explorer's zip extraction go boom
<mjg_>
why are you memeing so hard about this?
<mjg_>
do you genuinely intend to play?
<mjg_>
is there some drama surrounding it
<mjg_>
or what
<heat>
i'm playing right now
<heat>
im enjoying not osdeving for once
<klange>
Impossible.
<klange>
Joy can not be had from things that are not osdev.
<mjg_>
you either sad or miserable, the true osdev way
<sortie>
Alright I could chroot -d /cdrom /sbin/init
<sortie>
Time for the natural next stage, replacing all the .iso grub bootloading of all the loose files, instead just directly mounting the iso as /
<klange>
Using an ISO9660 filesystem as a root? That's disturbing.
CryptoDavid has quit [Quit: Connection closed for inactivity]
Vercas6 has quit [Remote host closed the connection]
<sortie>
Read-only of course :) (Although the plan is to basically mount a copy-on-write layer on top of it)
<mjg_>
welp, haiku does not install for me
heather has quit [Remote host closed the connection]
<sortie>
psst need installable OS??
<sortie>
Basically experimenting with just how damn fast I can boot my OS
Vercas6 has joined #osdev
<mjg_>
is it laced with unix?
<sortie>
May contain trace elements of Unix
<mjg_>
the installer wants me to create partitions, but the tool to do it does not spawn
<mjg_>
newest snap they have, dated 2021 though
<klange>
~3s to desktop here
<CompanionCube>
hey kid, wanna buy some multics?
heather has joined #osdev
<sortie>
klange, how you count that?
<sortie>
From pressing enter on the bootloader?
<klange>
With a stopwatch?
<klange>
Yes
<sortie>
I think I can beat that although it'll be close
<heat>
stopwatch? bleh
<heat>
just count to 3 and space them out by more or less a second
vdameboo has quit [Read error: Connection reset by peer]
<mjg_>
speaking of optimizing boot times
<mjg_>
did you know flamegraphs are of use there too
vdamewood has joined #osdev
<CompanionCube>
doesn't systemd have a cool boot graph thing
<klange>
My boot sequence is kinda dumb.
<klange>
And it has things like "check for updates from the remote package manager" in it, so I'm sure it could be a lot faster...
<klange>
1s loading the ramdisk in the bootloader, 1s of unpacking the ramdisk tarball into a writable tmpfs (can turn that off), spinning up device drivers, checking for packages...
<mjg_>
got haiku installed! even comes with a compiler