<Ermine>
and there's some userspase daemon for that, used by Adélie
<bslsk05>
github.com: security-research/pocs/cpus/entrysign/zentool/README.md at master · google/security-research · GitHub
<heat_>
lol
<heat_>
they fucked AMD microcode!!
<Ermine>
and by CachyOS
<zid>
nikolar died in his chair, rip
<zid>
heat_: Nikolar's stupid laptop didn't support 1G pages so I rewrote most of my memory initialization code to also support dropping back to 2M pages
<zid>
which required a nice refactor ofc
<heat_>
hah yes
<heat_>
1G pages date to IIRC either haswell or sandybridge
<zid>
it has pat and pse so I didn't even think to check
Goodbye_Vincent1 has quit [Remote host closed the connection]
Goodbye_Vincent1 has joined #osdev
<ring0_starr>
ooooooooh
citrons has quit [Ping timeout: 276 seconds]
citrons has joined #osdev
Lucretia has quit [Remote host closed the connection]
<ring0_starr>
i think the advisory is significantly downplaying the severity
<ring0_starr>
sure it might say "high", but only specifying that their encrypted VM extensions are at risk is a bit rich
<Ermine>
or you can avoid problems by good ole "do not do stupid shit"?
<ring0_starr>
like what's stupid here?
<Ermine>
running untrusted microcode?
<ring0_starr>
you don't have to be stupid to do that, just insecure
gog has quit [Quit: byee]
<ring0_starr>
whether or not you're insecure depends a lot on luck
<Ermine>
you can as well run that keygen-which-is-totes-not-a-trojan.exe from 10th page of google results as admin
<Ermine>
and call everything insecure
<ring0_starr>
user intervention is a last resort
<ring0_starr>
what happens when the attacker is lucky and has an exploit chain the rest of the way to root?
<Ermine>
otherwise microcode is somewhere in the end of the chain
<zid>
it's just a virus you cant kill with taskmgr, need to reboot
<ring0_starr>
and then it just gets reloaded after boot
<ring0_starr>
you do realize how microcode updates get applied right?
<Ermine>
and if attacker can get to your ring 0, you're already busted
<ring0_starr>
how do you know something got to ring 0?
<ring0_starr>
when it does get to ring 0, what happens then?
<ring0_starr>
how do you respond? do you throw all the hardware into the garbage?
<ring0_starr>
you can do a thorough audit that shows nothing, because there's an instruction-level backdoor now
<kof673>
> how do you know i would find a person with ring0 in their name and ask them
<kof673>
have to find the person who knows
<zid>
it can bust into your hypervisor which is neat
<zid>
and smm
<ring0_starr>
secure enclaves are also busted
<Ermine>
that audit would show nothing even without instruction-level backdoor
<Ermine>
because busted kernel can decieve anyone
<samis>
ring0_starr: i mean, you could get away with formatting the os disk so future boots don't load the pwned microcode
<ring0_starr>
would it?? we live in the era of virtualization-based security
<Ermine>
it would
<Ermine>
the kernel has the full and utter control over the machine
<heat_>
it can't bust into your hypervisor
<ring0_starr>
nor your EL3 monitor
<heat_>
you can't load microcode under a hypervisor
<ring0_starr>
samis: game of whack-a-mole
<heat_>
it's a very cool exploit but also requires ring0 access
<heat_>
with which you can pull off any number of fucked up things already
<Ermine>
so, if you suspect your machine to be busted, you probably acting according to some playbook defined in your organization
<Ermine>
wich probably involves isolating and assesing the damage, and also calling the police and contacting the legal team
<Ermine>
while pretending you haven't noticed anything in order to not scare attackers away
<ring0_starr>
security only seems to be considered from a corporate point of view
<Ermine>
if it's your personal server, then i gotta say
<Ermine>
LOL SKILL ISSUE
<heat_>
no one is targeting a lone individual with a handcrafted microcode exploit
<ring0_starr>
too many unknowns, if they had the capabilities to get on that far who knows if they laterally moved to a different machine you didn't isolate, where it lays dormant, specifically for survivability?
<samis>
heat_: well, state actors might, but again, lol skill issue
<ring0_starr>
heat, all backdoors are handcrafted
<ring0_starr>
well
<heat_>
not true
<ring0_starr>
imagine if you will, AI generated malware
<Ermine>
and if someone is targeting you with such backdoors, you have *much* bigger problems
<ring0_starr>
right, right, that's the common sentiment when it comes to advanced attack chains
<ring0_starr>
but keep in mind it costs zero
<ring0_starr>
once it already exists it can be used on many targets
<Ermine>
it doesn't
<Ermine>
such stuff costs shitloads of money actually
<ring0_starr>
in fact there's a strong incentive that as soon as the capability exists, to use it as much as possible since it doesn't have a very long shelf life
<heat_>
>once it already exists it can be used on many targets
<heat_>
those are generally quite generalized vulns
<Ermine>
it's not something your mad neighbour can afford
<heat_>
i can try and login to your db with login admin passwd admin
<heat_>
i can't craft a microcode exploit and somehow make you load it if i don't even know what hardware you're running
<ring0_starr>
you already got that information since you're root on that machine
<heat_>
this would possibly be a really great exploit for a state actor
<ring0_starr>
or private firms that sell to state actors for lotsamoney
<heat_>
if im root already then a microcode rootkit is just bonus
<ring0_starr>
and all other kinds of groups as well that they TOTALLY swear up and down that they absolutely do not
<heat_>
i probably already have what i want
<ring0_starr>
they make sure they respect embargos and do extensive background checks when somebody's buying their malware
<ring0_starr>
totally
<immibis>
remember that microcode signatures are something that keeps us out of our own machines
<kof673>
^ :D
<kof673>
security is keeping you out :
<kof673>
double-sided
<kof673>
*vulture
<immibis>
kof673: SIMD instructions presumably use more energy, but they also do more stuff with a better stuff per energy ratio
<ring0_starr>
they'd serve up the microcode exploit to defeat secure VMs or enclaves or whatever, but once it's already there... hey you can get super low level persistence for free now since it already exists and you know your target is supported
<ring0_starr>
I'd rather not have microcode at all. risc4life
<kof673>
microcode signature secures you!
<heat_>
uhhh yess totally risc does not have microcode haha definitely not
<ring0_starr>
except when they apparently sign it with sha1 like here
<immibis>
ring0_starr: CPUs don't store microcode in flash. it gets reloaded by the kernel every time the CPU is turned on.
<ring0_starr>
heat_: IT USED to not have microcode
<ring0_starr>
but everything's gotta get all shitty and overcomplicated
<immibis>
it may be possible to modify the firmware (which is in flash) to load bad microcode, but bad firmware can already do plenty of things without messing with microcode, so it's not obvious that also messing with microcode gives you useful new capabilities
<heat_>
maybe properly complicated, cancer research and all that
<heat_>
but you know we've talked about this before
<heat_>
immibis, additional stealth or a supply chain attack
<ring0_starr>
we should probably stop doing cancer research
<immibis>
heat_: but you could modify the firmware ("bios") and do bad stuff. in particular, note that firmware has control of system management mode.
<heat_>
yes but that's signed and checksummed
<heat_>
we can't modify that, we can modify this
<immibis>
heat_: so what's gained by modifying microcode? you mean you could MITM AMD's website so people download bad microcode updates from you?
<ring0_starr>
those updates should be signed as well
<ring0_starr>
the issue lies in "should be"
<heat_>
AMD's website, the linux-firmware repo, any number of packages in distros, etc
<ring0_starr>
there are so many links in the chain to break
<immibis>
heat_: same as the kernel, then
<ring0_starr>
and everything's been dodgy lately
<heat_>
the kernel is also signed and checksummed in proper security scenarios
<ring0_starr>
need a tighter chain tbh
<immibis>
I think having more visibility into our own CPUs will be great. at least, some bored hacker learning their about own CPU and then publishing it for the rest of us to read. The last time that happened was AMD K6?
<samis>
immibis: also the VIA C3?
m5_ has quit [Read error: Connection reset by peer]
<ring0_starr>
maybe in some alternate universe we'll have a TCB built upon really dead simple, formally verified designs
<samis>
at least i think it was the C3, might have been a different VIA chip
<ring0_starr>
so any reasonable person can look at it and say
<ring0_starr>
yeah this is definitely going to work as designed, i can 100% trust it to not get fucked at some point by somebody with near-infinite resources
<Ermine>
immibis: you shoud use transistor-based machines, since transistors are human visible and you can poke them with multimeters
<m5>
lol
<ring0_starr>
cause right now the game is very much open to anybody with sufficient resources
<m5>
i feel you immibis
<immibis>
Ermine: i actually have half of a transistor CPU, which i just got out of storage last month and have remembered enough of how it worked, but not poked it yet
<immibis>
i *think* the next thing to implement is a memory module, which i may prototype with an arduino or something, for convenient debugging
<Ermine>
immibis: have fun witg it
<Ermine>
wait wat
<Ermine>
arduino is silicon based
<Ermine>
bad tech
<m5>
is it fair to say that arm has "less" microcode?
<Ermine>
complex
<Ermine>
unpokeable
<immibis>
Ermine: the arduino would replace the memory subsystem until the other part that isn't memory is known to work
<ring0_starr>
and ik heat is going to say wE TaLkEd AbOuT ThIs AlReAdY but the secure thing doesn't need to be -O4 --fomg-optimized super fast race car, client endpoints can be all that. but something trusted 100% ought to be as slow as it needs to be as long as it can be readily verifiable by independent auditors on the design, or mathematical models, or whatever
<immibis>
then it can be substituted for actual memory, in the form of relatively trustworthy sram chips
<immibis>
that reminds me i also want to get a tillitis tkey at some point
<m5>
ring0_starr: there's a project called VST (verified software toolchain)
<heat_>
Ermine, how do you know the multimeter isn't compromised as well?
<heat_>
checkmate atheist
<immibis>
build your own multimeter
<ring0_starr>
soldering irons have modifiable firmware.
<ring0_starr>
let that sink in
<ring0_starr>
soldering irons have modifiable firmware.
<m5>
lol
<ring0_starr>
you think you're being sarcastic but it's really a valid thing to worry about now
<immibis>
my soldering iron has modifiable firmware
<immibis>
what if it's remotely updating my transistors?
<ring0_starr>
instead of piling more shit on top let's step back and fix the tech problems created by tech
<heat_>
see, stallman is so upset right now
<immibis>
i have another one without modifiable firmware, but i like the small one because it heats up very quickly and can run on batteries
<m5>
ring0_starr: i hate propriety firmware like everybody else, what's the mystery :-)
<heat_>
he only advocates for NON-MODIFIABLE FIRMWARE
<ring0_starr>
it's not even about proprietaryness
<ring0_starr>
it's just the fact that
<immibis>
(how big of a battery do you think you need for soldering? well, i planned to bring a lead-acid in my backpack...)
<ring0_starr>
holy shit, my soldering iron doesn't need to run IronOS
<ring0_starr>
it needs to get hot
<m5>
i see the point, i was thinking about uefi
<immibis>
ring0_starr: it's pretty basic stuff though. it's like, temperature control, auto sleep, and thermal runaway protection, and that's about it. somewhere in the menus you can see the input voltage too.
<ring0_starr>
true
<immibis>
because it's software, all those things are configurable in a menu
<ring0_starr>
actually the input voltage is displayed on the default screen
<immibis>
you COULD have knobs for all those things, and analog electronics, but it's probably not a great design.
<immibis>
my non-digital soldering iron just has a temperature knob, which is fine
<ring0_starr>
that said i feel like the art of electronics has taken a nosedive
<ring0_starr>
everything is just a microcontroller with some periperials
<Ermine>
heat_: oh indeed, we need to use only good old analog ammmeters and voltmeters
<immibis>
it has, but there's a good reason for that, namely: microcontrollers are very effective at everything and convenient
<ring0_starr>
not even any buttons anymore, it's just a touchscreen cause it's cheap/easy
<Ermine>
unless physic laws are also exploited
<immibis>
touchscreens suck
<heat_>
even the voltmeter could be sabotaged
<m5>
ring0_starr: why's that bad? we don't control the microchip industry afterall
<heat_>
i have an idea: rocks and sticks
<heat_>
new techology: fire
<Ermine>
or just accept that the matrix pwned us all already
<heat_>
the technology people would call it fire.js or something
<ring0_starr>
there is no spoone
<heat_>
Ermine, hmmm no i don't use matrix, im on irc after all
<kof673>
<sees gog walk by, glitching>
<m5>
Ermine: the simulation hypothesis isn't true, i guess
<Ermine>
heat_: oh, how does real world look like?
<heat_>
it looks like a bunch of text
<heat_>
it's kind of depressing, i recommend matrix
<immibis>
matrix has to be the worst chat protocol because of its insistence on distributed consensus
<heat_>
you could maybe add some funny gifs or something after all
<m5>
immibis: matrix the movie
<Ermine>
btw matrix.org is banned here
<m5>
nice
<heat_>
good, the great mother state is protecting you from bad chat protocols
<immibis>
...hungary?
<heat_>
no im not hungry
<m5>
you guys hate the intel management engine?
<Ermine>
no
<heat_>
middle management in general is usually a bit of a pain
<kof673>
that's the old view, i don't know who that person is :D “Water is Life's Matter and Matrix, Mother and Medium. There is no Life without Water.” — Albert Szent-Gyorgyi
<heat_>
but i dont hate anyone
<Ermine>
i just don't care
<heat_>
actually im lying, im a big middle management fan
<m5>
kof673: maybe the quantum snake oil will get us an answer
<heat_>
they think "mm" stands for memory management, but it's actually middle management
<kof673>
i just meant one mother protecting you from another mother :D, re: "the great mother state" :D
<m5>
heat_: mommy milkers is the right answer, what are you even saying
<immibis>
/kick m5
<m5>
it worked
<immibis>
hey the strangest thing in this microcode presentation is that "in al, dx" resets microcode
<immibis>
although maybe it's a certain register designed to do that. looks like they didn't try to figure out what the address was.
<m5>
maybe just dump the microcode somehow
<m5>
idk i'm dumb af
m5_ has joined #osdev
<m5>
i'm sorry guys, peace
m5 has left #osdev [#osdev]
m5_ has left #osdev [#osdev]
edr has quit [Quit: Leaving]
m5 has joined #osdev
<zid>
heat_: nikolar ruined my mommy milkers, then fell asleep in the middle :(
<bslsk05>
liblisa.nl: libLISA - Instruction Discovery and Analysis on x86-64
<zid>
yea I've seen the instruction search tool
<m5>
:-)
<zid>
they found some good stuff on VIA
<zid>
not much on real intels
<heat_>
imagine falling asleep lol
<heat_>
i would never
<geist>
i was just thinking about VIA the other day. supposedly they were working on a 64bit core in the early 2010s but i guess that either never came into existance or it was so specialized you never see it
<geist>
last one of theirs i had i think was a VIA c7 in the early 2000s
<geist>
it was slow then, but i was using it for a NAS computer i had put together and it was fine for that
<zid>
via bios was supposed to write to an MSR to disable the system management instructions but.. neglected to, afaik
<bslsk05>
www.ebay.com: DEC RAINBOW 100 Model B3 - Digital Equipment Corp. from 1982 - CP/M MS-DOS VR201 | eBay
m5_ has joined #osdev
<klys>
I understand the monitor also worked with the decmate ii pc278
m5_ has quit [Read error: Connection reset by peer]
m5 is now known as d1rg3
d1rg3_ has joined #osdev
Arthuria has joined #osdev
d1rg3_ has quit [Read error: Connection reset by peer]
eluks has quit [Remote host closed the connection]
eluks has joined #osdev
craigo has quit [Read error: Connection reset by peer]
craigo has joined #osdev
guideX has quit [Read error: Connection reset by peer]
d1rg3_ has joined #osdev
<mpetch>
I remember getting FidoNet BBS software to run on one of those Dec Rainbows in the mid 80s. The BIOS wasn't 100% IBM compatible so made running some software a little more difficult.
<klys>
woah fidonet? I remember fidonet too
<mpetch>
lol
<mpetch>
The guy who ended up running that Dec Rainbow called his system "Not Quite a Fell Dec BBS". Those were the days haha
<mpetch>
Full
<zid>
mpetch: Now tell us the one about the onions you used to wear on your belt
<zid>
Three bees for a nickel you used to tell us
<mpetch>
Haha I'm old and never watched the Simpsons so I actually had to Google that pop culture reference lol
<zid>
five bees for a quarter anyway, dang
<zid>
God, that aired in 1993
<zid>
that was before heat was born
<mpetch>
LOLOLOL
<d1rg3>
i came into existence in 2003, am i cooked? and got into computers in 2019 :-(
<d1rg3>
y'all are experts
<Mutabah>
Everyone starts somewhere, don't worry
<ring0_starr>
cooked
<ring0_starr>
there's still some hope for you if you get LOCKED IN
bauen1 has quit [Ping timeout: 252 seconds]
<ring0_starr>
cooked at 350 degrees in an oven
<zid>
I have no idea how you get into computers anymore
<zid>
google's fucked and hardware is incredibly obtuse and complicated
<d1rg3>
yeah :-(
<ring0_starr>
just shut up and flick the magical black rectangle to see skibbidy toilets
<ring0_starr>
you do not need to know anything
<Mutabah>
On the other hand... I have some not-too-old dell mini PCs, that use Realtek 8168 ethernet controllers.
<Mutabah>
Damn I had forgotten how nice Realtek's docs and hardware was
<d1rg3>
apparently i'm on heat's ignore :P
<zid>
depends, some of it is super cheap and janky
<zid>
8139too being the notorious example
<Mutabah>
8139 = Amazing to program, 8168 (and friends) = Almost as easy
<d1rg3>
zid i remember you telling me once to start with a 6502, that looked easy
<zid>
I doubt it, I probably said gameboy
<Mutabah>
... well, excluding that damn 0x10 offset on the 8139
<d1rg3>
:-)
<zid>
Feature stripped and cheap makes the docs a lot easier to read at least :P
hwpplayer1 has joined #osdev
<zid>
I have made a huge mistake
<zid>
I went downstairs to get a drink, and came back with a snack instead
<Mutabah>
Mistake?
<zid>
I'm gagging for a drink even more now, but I've already gotten up once
<d1rg3>
never tried wine/whiskey but i'm 22 already
<d1rg3>
guess i should, someday
<zid>
Either that was an incredible non-sequitor, or there's a ghost somewhere I can't hear
<d1rg3>
how's that a non-sequitor :D
<d1rg3>
anyway, get a drink zid
<zid>
what was it related to?
<Mutabah>
I assume zid meant water, not alocohol
<zid>
even if I did say that, it still wouldn't make much sense lol
<d1rg3>
i'm cooked
<d1rg3>
gn
<Mutabah>
Someone thinks they're too old for programming, someone is wrong
<zid>
getting into *programming* is easier than ever, imo
<zid>
when most of us started, compilers were industrial products that had seat licence pricing attached
<Mutabah>
Well... when I started, gcc was just starting to be decent (I started off osdev with bran's tutorial)
<zid>
yea same
<zid>
gcc was usable by the time I started in earnest
<zid>
msvc was still a paid product for another decade re windows though
<Mutabah>
Yeah, point there
<zid>
plus you get python instead of.. tcl
<zid>
and computers now means "the web" anyway, so you can just get infinitely lost in flavour of the week javascript frameworks woot
dormito has quit [Ping timeout: 268 seconds]
Arthuria has quit [Remote host closed the connection]
<bslsk05>
github.com: security-research/pocs/cpus/entrysign/zentool/README.md at master · google/security-research · GitHub
gildasio has joined #osdev
<nikolar>
Does that mean we're getting AMDs me equivalent patched out :P
hwpplayer1 has joined #osdev
LainExperiments has joined #osdev
xenos1984 has quit [Ping timeout: 272 seconds]
<heat_>
no? the management engine isn't in microcode
<nikolar>
I know it's not but I vaguely remember microcode patches being used to remove me from some older processors
xenos1984 has joined #osdev
<Ermine>
at best you can eliminate calls to the me, but it's a separate device
<nikolar>
As I said, I know it's separate
<heat_>
me_cleaner barfs all over the firmware
<heat_>
in AMD's case the security processor has modules inside the regular PC firmware (ARM64 modules, so they stick out)
<heat_>
you also can't fully disable it since the security processor is the one that starts the x86 core and trains memory
<nikolar>
The issue with them isn't that they are running, but that they are running unknown code and can do whatever they want, so if you can somehow make them only train the memory and initialize the main cores, that's fine
<Ermine>
i'm sure most of the code you run right now is unknown to you
<\Test_User>
they could initialize your memory with unknown code to run on said main cores that they're also initializing anyways
<\Test_User>
yeah, sure, it'd be an improvement to get rid of the later actions if you could, but still not a fix to the problem
<nikolar>
Ermine most of the code I'm running can't read arbitrary memory and then send that to a remote server
<Ermine>
linux can
<nikolar>
At least I *can* read Linux source if I want to
<nikolar>
I can't read me code
<Ermine>
good luck reading linux code
<nikolar>
I don't get what point you're making ermine
<nikolar>
Here's another thing I can do with Linux, patch it
<nikolar>
I can't patch me
<Ermine>
i'm freaking over gpu drivers code, and that's only a minor part of a kernel
<nikolar>
Sure
<nikolar>
At least you have a chance to read it
<nikolar>
And that's probably the messiest part of the kernel
<Ermine>
i even fail to comprehend Onyx so far
<nikolar>
Because GPUs suck
<nikolar>
Stare at it harder
<Ermine>
GPUs are just hard
<nikolar>
That's what I said
<Ermine>
no
<Ermine>
connotation is different
<Ermine>
but anyway
<nikolar>
Are you trying to tell me that working with GPUs doesn't suck
gildasio has quit [Remote host closed the connection]
<Ermine>
possibility to read/patch linux doesn't mean you can successfully audit it
gildasio has joined #osdev
<Ermine>
so it differs from me situation only in theory
<nikolar>
You can certainly more successfully audit it than me
<zid>
Ermine: baby, bathwater
<zid>
"You can't do it so nobody can and it isn't worth it to ever try"
<Ermine>
also, if me sent stuff to remote server, someone would already noticed it by intercepting traffic
<Ermine>
zid: i stronly believe that auditing large codebases like linux is beyond single person's capabilities, unless that's a genius
<zid>
so you're just repeating what you already said
<Ermine>
yes
<zid>
rather than responding to what *I* said
<Ermine>
it's a belief
<zid>
which is that it doesn't fucking matter, you're saying because you can't do it, it shouldn't be a thing anybody should want
<nikolar>
Ermine: I'm not saying it's happening
<nikolar>
I'm saying it could happen at any point
<zid>
nobody should want open source code
<nikolar>
And you can't do anything about it
<zid>
because they might make so much open source code that you can't read it all
<Ermine>
i'm not saying that nobody should want open source code
<zid>
well, you did
<Ermine>
i didn't
<zid>
then repeated it
<Ermine>
i didn't repeat it
netbsduser has quit [Ping timeout: 248 seconds]
gog has joined #osdev
hwpplayer1 has quit [Read error: Connection reset by peer]
xenos1984 has quit [Ping timeout: 272 seconds]
LainExperiments has quit [Ping timeout: 240 seconds]
LainExperiments7 has joined #osdev
LainExperiments has joined #osdev
LainExperiments4 has joined #osdev
LainExperiments7 has quit [Ping timeout: 240 seconds]
xenos1984 has joined #osdev
LainExperiments has quit [Ping timeout: 240 seconds]
<heat_>
linux kernel
vdamewood has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<zid>
heat_ nikolar is too busy hacking rdrand
<zid>
to make his laptop work
<zid>
so no kernel for me
<heat_>
hacking rdrand?
<zid>
zenbleedtoolthing
<heat_>
ah cool
<heat_>
fun fact the zen rdrand takes the carry bit as entropy
<zid>
yea I posted the bug report
<zid>
I think they just
<zid>
fucked up and used adc in the microops
<zid>
not add
LainExperiments4 has quit [Ping timeout: 240 seconds]
<zid>
so it was returning 5 if carry was set
netbsduser has joined #osdev
<nikolar>
heat_: should we make a fib instruction
<nikolar>
Sounds more useful than fpatan
<heat_>
fib can't work as an instruction
<heat_>
some terms would take so long to compute that you'd need for a way to keep track of progress, which is possibly workeable
<heat_>
but the main thing is that you'd want to catch interrupts rep movsb style
<nikolar>
How about nextfib
<nikolar>
(unfortunately I don't think you can get jumps though, so no fib anyway)
<zid>
you can absolutely jump, I just don't think we know how :P
<nikolar>
That's what I meant kek
Left_Turn has joined #osdev
Turn_Left has quit [Ping timeout: 252 seconds]
Turn_Left has joined #osdev
Left_Turn has quit [Ping timeout: 265 seconds]
alpha2023 has quit [Read error: Connection reset by peer]
alpha2023 has joined #osdev
TkTech has quit [Ping timeout: 244 seconds]
demindiro has joined #osdev
<demindiro>
I have an idea that's either great or terrible
<heat_>
shoot it
<demindiro>
I'm going to write (yet another) OS in Rust and include whatever I find on crates.io
<demindiro>
It may end up a mess
<heat_>
terrible. next!
<demindiro>
But OTOH I might also get something working very quick
<heat_>
you can get something working very quickly without taking kernalover69's shitty x86 mmu crate that's subtly wrong in many ways
<demindiro>
MMU I'll do myself, since I have some unconventional ideas about it
<demindiro>
But stuff like UEFI? crates.io will do, I'm sure
<heat_>
lord
<demindiro>
And PCI and virtio and and and
<heat_>
UEFI, the brittlest little thing ever? that's bound to work really well
<demindiro>
We'll find out soon enough
<nikolar>
heat_: subtly wrong how
<nikolar>
Just curious
<heat_>
there are many ways mmu code can be subtly wrong and i wouldn't trust a rando on this
<heat_>
like: doing TLB invalidations wrong, not invalidating at all, not invalidating when you remove a page table, doing the change-perms dance wrongly (on e.g arm64), doing the huge-page-to-page-table dance wrongly/not at all
<nikolar>
Yeah fair enough
user71 has quit [Quit: Leaving]
kof673 has quit [*.net *.split]
duckworld has quit [*.net *.split]
kof673 has joined #osdev
duckworld has joined #osdev
<demindiro>
OVMF doesn't support GOP?
<demindiro>
Looks like it should, hm
<heat_>
yes it does
<demindiro>
I was silly and passing the wrong handle
<demindiro>
(handle of loaded app instead of first quering for protocols, herpaderp)
<demindiro>
Rainbow! That works at least
GeDaMo has quit [Quit: 0wt 0f v0w3ls.]
demindiro has quit [Quit: Client closed]
Affliction has quit [*.net *.split]
vdamewood has joined #osdev
<vdamewood>
Hi, people and cats.
<vdamewood>
... and cat people.
alexander has quit [Remote host closed the connection]
netbsduser has quit [Remote host closed the connection]
netbsduser has joined #osdev
eddof13 has joined #osdev
ThinkT510 has quit [Ping timeout: 260 seconds]
alexander has joined #osdev
Affliction has joined #osdev
netbsduser has quit [Ping timeout: 252 seconds]
eddof13 has quit [Quit: eddof13]
Turn_Left has quit [Read error: Connection reset by peer]
eddof13 has joined #osdev
Lucretia has quit [Remote host closed the connection]