<geist>
you mean the roundrect around the description box?
<heat>
youtube has been rectangular for ages
<heat>
yeah the "roundrect"'s everywhere
<geist>
oh yeah, that changed a week or so ago here. roundrects are the thing now mang. get with the program
<geist>
designers gotta earn their scratch
<heat>
what's this design called?
<Griwes>
a couple of weeks ago google maps lost the points of the pins on the map too
<geist>
probably some material thing
<kazinsal>
roundrects are back, death to squares
<Griwes>
now all marked locations are just disembodied floating heads that have lost any attachment to actual places on this earth
<heat>
this change seems to have rolled out today for me. curiously I saw it first on my phone, I wonder if it rolls out based on the device
<kazinsal>
all aboard for the great 2025 return of skeuomorphism
<Griwes>
so, in other words, google continues to wage a war against its users
<heat>
google.com itself had rolled this out some weeks ago for me
<geist>
i think it rolls out in stages. there are multiple buckets of features/etc that are constantly being tested
<Griwes>
no offense to any googlers present but could they just _stop_
<geist>
so you are in one of N buckets at any given point of time for all the apps
<Griwes>
just... decide on a design (ideally the previous one for both youtube and maps) and just *stop*
<heat>
hehehhehe funny Griwes
<Griwes>
you know, you can just... walk away!
<geist>
heh silly Griwes
* Griwes
wonders how far he can adapt the elden ring meme to this
<heat>
the crap they pulled with the icons was way worse
<geist>
been fiddling with moving more devices over to my iot vlan and using mdns bridging across the vlans to still get to them
<geist>
been fairly educational what does and desn't work across the boundary
<heat>
everything looks very samey now
<heat>
i appreciate the theming but it's horrible when you're trying to distinguish icons
<geist>
i hate it when anyone redesigns icons. OTOH, in my experience after a few weeks you'll forget what the old one looked like
<geist>
its rare that there's an actual loss of recognition. i'm not defending changing of icons, it's just... never as big of a deal in the long run as i think
<geist>
all this aside color is my biggest differentiator, though of course stuff has to be designed around color blindness too
<heat>
i mean, I don't have problems because I usually click things based on the position, so I'm not actively scanning for icons
<heat>
yeah, and now every google app has the google colours applied to them
<geist>
the small webicon matters a lot to me because i tend to keep all my interesting web apps pinned on the left side of my chrome tabs and thus it shows the mini icon
<Griwes>
yeah I don't rally care about icons much
<heat>
(red, green, yellow, blue)
<geist>
if they're distinct i dont care too much what they are, beause i'll just learn them after a few days
<Griwes>
and other than the 4 column limit in channel videos view, the youtube thing is just annoying and not actively harmful
<geist>
just make them distinct
<Griwes>
but I can't say the same about the maps thing
<Griwes>
they literally made markers less accurate for no reason
<Griwes>
it's infuriating
<geist>
file a bug
<Griwes>
hahaha
<Griwes>
good joke
<geist>
not saying that flippantly, i mean literally file a bug
<heat>
Griwes, can you screnshot your maps thing?
<geist>
and if you send me the bug link i can check internally what is up with it
<Griwes>
google has a user facing bug reporting for maps that humans look at?
<geist>
contrary to popular belief, folks do look at bugs, it's just always being prioritized against a bunch of other things
<geist>
but... good bugs are worth their weight in gold, and most folks file bad bugs
<heat>
gogle stupid
<Griwes>
I was questioning both parts of that question
<kazinsal>
people just tend to either a) not file bugs and just get ANGERY online instead or b) file bugs really poorly
<geist>
something that can be actioned against simply usually makes itself to the top of the pile. stuff that requires some new fundamental feature are the ones that may appear ignored
<heat>
why program look bad???
<heat>
kazinsal, dude I filed the most detailed bug report for discord and they basically flipped me off
<geist>
Griwes: you can probalby file a bug in the ui. *hopefully* it gives you some sort of ID at least
<geist>
my huge complaint right now with youtube is the stupid shorts thing. i get it, it is what the genzers want i guess
<geist>
but i wish i could just turn it off
<geist>
obviously i cant and obviously it's not going anywhere, but i use youtube for *information* to look at informational things. sometimes funny things, but i generally like long formish stuff, beause it' hard to convey *information* in 30 second shorts you can't even scrub around in
<heat>
you'll get lost in an hour of your mind fucking off
<geist>
but now even my regular stuff i subscribe to seems like are being 'forced' to generate little silyl shorts and they chew up like 1/3 of my subscriber page
<Griwes>
I _think_ the markers are now anchored by the middle of the circle
<heat>
aw that looks odd
<geist>
Griwes: hmm, so if you zoom in does it go to the right spot?
<geist>
i'm not sure i get the issue given you're so zoomed out
<geist>
oooo i see wha tyo mean, you mean it used to be a pin, where there was a single point at the end of it?
<Griwes>
I have a marker in the arizona strip and at a moderate zoom out (that shows western US) on my phone you can't even tell which state it is in
<Griwes>
geist: yes!
<geist>
ah. so now we got to the meat of it. you dont like the new icons because they're less accurate
<Griwes>
and like this may not be the best example of it but it's the one that's annoying me a lot
<geist>
yeah that's fair
<geist>
i hadn't thoughg about it but now that you mention it
<geist>
but i usually just zoom in until it's clear what the pin is
<heat>
hm I wonder if discord's hw video acceleration is still broken
<Griwes>
also I have a ton of brain memory, that google trained me to, that the actual position is several pixels below the marker, but now it's in the center of the marker
<heat>
it was crapping itself on intel GPUs. I filled the most insightful bug I could, and besides getting to put up with annoying support people that think I'm bad with computers, some engineer said at the end that they would "look at it later"
<geist>
i mean i'm not seeing what you see, but i'm seeing these little pin like things
chartreus has joined #osdev
<geist>
heat: yeah they probably thought they were being nice too, just bad engineer UI
<Griwes>
geist: in the maps UI, under "get help", only things I'm seeing are effectively "report content or data error" or "send us feedback", neither of these sound like "report a bug and get a bug ID" :P
<geist>
yah in this case it's not a bug, that's a feedback
<Griwes>
geist: interesting, they must be rolling this not per device but per account
<geist>
i thought you literally had a bug, like you did something and it doesn't work right
<heat>
geist, like their "support" solution for "hw accel is broken" was to disable hw acceleration
<Griwes>
right, but even if I did have a bug, I am not, in fact, seeing a way to report it
* geist
nods
<Griwes>
anyway it must be per account because it flipped on my phone and my desktop at the same time
<heat>
if I got to talk with an actual engineer we could've actually fixed it
<heat>
hell, it was so frustrating I would work for them for a week for free just so I could fix it
<geist>
heat: doesn't mean it wont get listened to. but yeah that's the problem with big corporations
<geist>
or maybe not that, but large projects
<heat>
and this is just discord. discord isn't supposed to be big
chartreuse has quit [Ping timeout: 252 seconds]
<geist>
you can only hope that the bug is sitting somewhere, and the info you added means it'll eventually get looked at
<geist>
trouble is this stuff is mostly staffed with humans
<heat>
running bug reports through support engineers is great and all but at some point it should go through someone with actual technical ability
<heat>
particularly if it's a technical issue
<geist>
hmm, issuetracker.google.com maybe?
<geist>
i honeslty dont know. i think the different groups have different public facing bugs/feedback
<geist>
i'm used to chrome/android where you can file bugs i think
<heat>
monorail is pretty decent
<geist>
and fuchsia you can, of course
<heat>
although way more technically focused I think
<Griwes>
there is a lot of components popping up when you type google maps, but none of them looks like it's actually about the end user gmaps experience lol
<geist>
yeah it's pretty good. it has some annoyances that i dont like but so it goes
epony has quit [Ping timeout: 252 seconds]
epony has joined #osdev
archenoth has quit [Ping timeout: 246 seconds]
archenoth has joined #osdev
X-Scale has joined #osdev
TempNick has quit [Ping timeout: 252 seconds]
X-Scale is now known as TempNick
archenoth has quit [Ping timeout: 252 seconds]
archenoth has joined #osdev
zid has quit [Read error: Connection reset by peer]
zid has joined #osdev
zid has quit [Read error: Connection reset by peer]
zid has joined #osdev
<heat>
geist, have you tried CoW page tables?
ZombieChicken has joined #osdev
darkstarx has joined #osdev
darkstardevx has quit [Remote host closed the connection]
weinholt has quit [Ping timeout: 246 seconds]
<geist>
no
<geist>
would have some interesting TLB flushing implications on ARM, if not x86
<geist>
heh, aokay so apparently some guy just checker flagged on a nascar race by going full throttle against the outer wall and just dragging the car
<heat>
friction should be a hell of a lot slower than just erm, turning
<geist>
i think with nascar the general idea is they have a lot more horsepower than they have traction. yes yes i know it's super redneck american, lets ignore that
<geist>
it's bsaically way overpowered, barely holding on, *arguably* that's interesting in its own right. it's sort of the opposite of F1 but i think kinda interesting nonetheless, just considering the constraints
<heat>
yeah but like, friction bad turning good still
epony has quit [Ping timeout: 252 seconds]
<geist>
and yes they just drive in a circle. but on the turns they have to really roll of the throttle!
<geist>
so anyway apparently that trick works at least for a short while
<heat>
from the comments it seems that it only works at that track? *shrug*
<geist>
hmm, looking at the SMART on my main windows nvme drive it's actually chewing into the spare space fairly quickly
<geist>
ie, only a few years old and it's down to 54 percent spare free
epony has joined #osdev
<heat>
i fully appreciate how stupidly unsafe that move is
<geist>
yeah which is why i'm sure this was the one and only time it was ever done intententionally
<heat>
NASCAR HAS GONE SOFT
<heat>
DAMN LIBERALS
<heat>
they should give him the race already for dangerous driving
<geist>
well he didn't win but he got 2 spots up into the top 4
<geist>
which i guess is where you get points
<heat>
do you like nascar?
<heat>
i don't quite get the appeal. maybe it's a cultural thing
<heat>
i find rallying to be way more interesting than any other racing type
<klange>
Little wheels on the arms run against the bars on the side of the guideway.
heat has quit [Ping timeout: 276 seconds]
archenoth has joined #osdev
zid has quit [Ping timeout: 252 seconds]
zid has joined #osdev
<\Test_User>
ye
smeso has joined #osdev
ZombieChicken has quit [Ping timeout: 258 seconds]
ZombieChicken has joined #osdev
archenoth has quit [Ping timeout: 255 seconds]
Burgundy has joined #osdev
orccoin has joined #osdev
ZombieChicken has quit [Ping timeout: 258 seconds]
ZombieChicken has joined #osdev
ZombieChicken has quit [Client Quit]
darkstardev13 has joined #osdev
darkstarx has quit [Ping timeout: 252 seconds]
Burgundy has quit [Ping timeout: 248 seconds]
orccoin has quit [Ping timeout: 248 seconds]
orccoin has joined #osdev
zaquest has quit [Remote host closed the connection]
epony has quit [Ping timeout: 252 seconds]
epony has joined #osdev
ZombieChicken has joined #osdev
zaquest has joined #osdev
<geist>
Heat: oh no, not at all. not really into it, i was just trying to head off the expected 'hurr durr american racing is dumb' stuf
epony has quit [Quit: QUIT]
<zid>
It skeeves me out that he's allowed to do something like that
<zid>
just because of the danger to the other drivers
<zid>
wouldn't take much for him to kick out, pit maneuver someone, and them faceplant the wall and die
epony has joined #osdev
darkstardev13 has quit [Read error: Connection reset by peer]
epony has quit [Remote host closed the connection]
epony has joined #osdev
ZombieChicken has quit [Quit: WeeChat 3.6]
<geist>
heh that's precisely why he'll get to probably do it exactly once
darkstardevx has joined #osdev
darkstardevx has quit [Remote host closed the connection]
<kazinsal>
yeah I'm definitely not a nascar guy but when I saw that clip my jaw dropped
darkstardevx has joined #osdev
<kazinsal>
especially his post-race interview where he said something along the lines of "yeah I played a lot of nascar on the gamecube"
GeDaMo has joined #osdev
darkstardevx has quit [Remote host closed the connection]
darkstardevx has joined #osdev
orccoin has quit [Ping timeout: 252 seconds]
archenoth has joined #osdev
archenoth has quit [Ping timeout: 252 seconds]
archenoth has joined #osdev
xenos1984 has quit [Quit: Leaving.]
scoobydoob has joined #osdev
scoobydoo has quit [Ping timeout: 246 seconds]
scoobydoob is now known as scoobydoo
xenos1984 has joined #osdev
archenoth has quit [Ping timeout: 255 seconds]
archenoth has joined #osdev
archenoth has quit [Ping timeout: 252 seconds]
archenoth has joined #osdev
archenoth has quit [Ping timeout: 255 seconds]
archenoth has joined #osdev
archenoth has quit [Ping timeout: 248 seconds]
Jari-- has joined #osdev
sahibatko has joined #osdev
archenoth has joined #osdev
sahibatko has quit [Quit: bye]
amine8 has joined #osdev
sahibatko has joined #osdev
amine has quit [Ping timeout: 250 seconds]
amine8 is now known as amine
darkstardevx has quit [Remote host closed the connection]
archenoth has quit [Ping timeout: 255 seconds]
darkstardevx has joined #osdev
darkstardevx has quit [Remote host closed the connection]
darkstardevx has joined #osdev
darkstardevx has quit [Remote host closed the connection]
darkstardevx has joined #osdev
nyah has joined #osdev
darkstardevx has quit [Remote host closed the connection]
darkstardevx has joined #osdev
archenoth has joined #osdev
darkstardevx has quit [Remote host closed the connection]
darkstardevx has joined #osdev
darkstardevx has quit [Remote host closed the connection]
darkstardevx has joined #osdev
darkstardevx has quit [Remote host closed the connection]
darkstardevx has joined #osdev
darkstardevx has quit [Remote host closed the connection]
darkstardevx has joined #osdev
darkstardevx has quit [Client Quit]
Jari-- has quit [Ping timeout: 272 seconds]
archenoth has quit [Ping timeout: 255 seconds]
zaquest has quit [Ping timeout: 255 seconds]
archenoth has joined #osdev
weinholt has joined #osdev
noxim has joined #osdev
archenoth has quit [Ping timeout: 255 seconds]
TempNick is now known as X-Scale
Burgundy has joined #osdev
grange_c has joined #osdev
weinholt has quit [Ping timeout: 248 seconds]
eau has joined #osdev
outfox has quit [Ping timeout: 264 seconds]
outfox has joined #osdev
Burgundy has quit [Ping timeout: 248 seconds]
noxim has quit [Ping timeout: 252 seconds]
isaacwoods has joined #osdev
archenoth has joined #osdev
dzwdz has quit [Ping timeout: 260 seconds]
dzwdz has joined #osdev
grange_c0 has joined #osdev
grange_c has quit [Ping timeout: 252 seconds]
grange_c0 is now known as grange_c
dude12312414 has joined #osdev
eau has quit [Ping timeout: 252 seconds]
eau has joined #osdev
zaquest has joined #osdev
Jari-- has joined #osdev
elastic_dog has quit [Read error: Connection reset by peer]
<zid>
physics is over now right, we're just collecting decimal places, I think once heat teaches us all a bit more about osdev we'll be in the same spot
elastic_dog has joined #osdev
wand has quit [Ping timeout: 258 seconds]
wand has joined #osdev
archenoth has quit [Ping timeout: 255 seconds]
archenoth has joined #osdev
Ali_A has joined #osdev
Terlisimo has quit [Quit: Connection reset by beer]
Terlisimo has joined #osdev
romzx has joined #osdev
archenoth has quit [Ping timeout: 248 seconds]
dude12312414 has quit [Quit: THE RAM IS TOO DAMN HIGH]
<Jari-->
What is the opcode, e.g. hexadecimals for CPU idle, a standard way ?
<Jari-->
I am using VMWare_IDLE I picked somewhere, forgot where
<heat>
wdym standard way? what cpu?
<Jari-->
heat: IA-32
<heat>
f4
<Jari-->
heat: I assume that also works on a native system, not only VMware or VirtualBox?
<Jari-->
the name was just misleading first
<heat>
that works everywhere
<heat>
it's hlt
<Jari-->
yeah
<heat>
and yes I knew that by heart lol
<Jari-->
Google says it might lower the core temperature or power consumption
<heat>
it should yeah
<heat>
this is not the optimal way, but the simplest way
<heat>
hlt blocks until an interrupt arrives
<heat>
so to be used effectively you need something like 1: hlt jmp 1b
<heat>
you can cli; hlt if you want to fully hlt the machine but you can still can NMIs
Jari-- has quit [Ping timeout: 252 seconds]
zaquest has joined #osdev
archenoth has quit [Ping timeout: 255 seconds]
archenoth has joined #osdev
Burgundy has joined #osdev
xenos1984 has quit [Ping timeout: 255 seconds]
xenos1984 has joined #osdev
wootehfoot has joined #osdev
moberg has joined #osdev
moberg1 has quit [Ping timeout: 246 seconds]
* geist
yawns
<heat>
this is the worst
<heat>
i was expecting a flashy vulnerability with a badass codename and everything
<heat>
... turns out nobody gives a shit and no one is updating it immediately
<heat>
in fact, they retracted the CRITICAL
<heat>
a lot of shit stirring for not too much shit
<geist>
oh oh the vuln
gog has joined #osdev
krom has joined #osdev
xenos1984 has quit [Ping timeout: 246 seconds]
heat has quit [Remote host closed the connection]
heat has joined #osdev
<clever>
heat: ah, potential RCE, but only if a CA signed an already malicious cert, kinda tricky to rate that
<zid>
if it can be solved by policy and there are no extant examples
<zid>
I'd rate it very low
xenos1984 has joined #osdev
<clever>
it can also occur if you continue despite the cert not being signed by a valid CA
<heat>
in which case, you did a self-goof
<sortie>
Oh no, think of all those curl -k ... | ssh invocations that'll now result in RCE :(
<zid>
well.. result in a 4 byte overflow
<zid>
which is damn near impossible to RCE from
<heat>
never say never
<heat>
you can do it
* sortie
drinks another gulp of the LibreSSL shipped with Sortix and laughs at your OpenSSL 3
<sortie>
heat, yeah think I even one time saw an exploit successfully use series of fixed bytes (like the '.' char CVE) to subvert
<zid>
it'd be nice if libressl was better too
<sortie>
libressl does improve
<zid>
so does stepping in poop
<heat>
libressl has no CVEs
<heat>
i wonder why that is
<zid>
because nobody uses it? :P
<sortie>
“Only two remote holes in the default install, in a heck of a long time!“
<zid>
That sounds sexual
<sortie>
OpenBSDM is all about the pleasure
<heat>
no pleasure in open
<heat>
FreeBSDM is better
gxt has quit [Remote host closed the connection]
gxt has joined #osdev
<Ermine>
BSDM-based MacOS when
<geist>
oh that reminds me, saw some vid where a guy mentioned some sort of vm container linux product thing. meant to check it out
<geist>
crap what ws it called
<heat>
d o c k e r
<geist>
proxmox
<geist>
(had to look up the vid i saw it in)
zaquest has quit [Ping timeout: 252 seconds]
<mjg_>
burp
<gog>
burp
<mjg_>
^5
darkstardevx has joined #osdev
darkstardevx has quit [Remote host closed the connection]
darkstardevx has joined #osdev
<geist>
hmm, wonder if it's a good idea to add google and cloudflare's dns servers side by side on your dns resolver
<geist>
i guess so, since both should return the same results
<geist>
then the question is how many upstream servers is enough? i have 8, which is probably so super overkill (both google and cloudflares 4 dns servers)
<heat>
for glibc it will just take longer to fail a resolve no?
<heat>
do NOT do that on musl systems though
<heat>
musl fires off N DNS requests for the N resolv.conf hosts (although I think it caps N at 3 or 4)
<geist>
yah this is on my firewall (pfsense) which is running unbound and forwarding dns requests on the lan to them
<geist>
since google and cloudflare both support SSL dns it's even better, since all outgoing dns is using that
darkstardevx has left #osdev [Leaving]
<heat>
actually, I don't know if having the "secondary IPs" makes any difference in practice
<geist>
from tcpdumping it it seems that unbound basically holds 8 sockets open (port 853 because SSL) and seems to be generally round robining between them, so i guess that's about as good as i can expect
zaquest has joined #osdev
<heat>
idk about google but cloudflare uses regular old anycast
<heat>
if the service is down, i'm willing to bet it's down on 1.1.1.1 and 1.0.0.1
<geist>
far as i can tell cloudflare dns and google dns behave exactly the same
<geist>
the cloudflares seem to be a teensy bit slower (RTT/RTO times are a little higher) but very very close. probably just based on where their local boxes are located
<heat>
odd
<heat>
there's a seattle colo
<geist>
yeah they seem to be pretty much the same ping time. about 15ms from here
<geist>
there's a stats page on pfsense that shwos the RTT/RTO of each of the servers
<geist>
and it seems the cloudflares are in general a little higher
<geist>
it seems to vary quite a bit when i refresh it so i have very little faith that those stats are particularly useful
<heat>
are they for queries or ICMP?
<geist>
queries
<geist>
icmp all of the servers seem to be a nicely 17ms +/- 2
<heat>
64 bytes from 1.1.1.1: icmp_seq=1 ttl=59 time=5.59 ms
<geist>
also those queries are over TLS so maybe that adds a bit. i honeslty dunno how fast a dns query *should* take
<heat>
64 bytes from 8.8.8.8: icmp_seq=1 ttl=116 time=15.0 ms
<heat>
which is *very fucking odd*
<geist>
5ms is damn close. i'm pretty sure my cable modem adds about that much
<geist>
or probalby just getting off the island. i can't get anything below about 12 to anything
<heat>
yeah, I even know where the datacenter is :D
<heat>
it's very very close by
<geist>
but in my case both cloudflare and google dns servers are about the same icmp distance it seems
<heat>
well, 40 minutes by public transporation away
<geist>
slightly differnet routes but wouldn't be surprised if they're not in the same colocated building
<heat>
which makes the ping very impressive
<geist>
that RTT/RTO stuff must be the actual time taken to return the query at the highest layer
<geist>
through TLS, etc
<geist>
so i guess each of the servers takes 200ms or so to return the data
<geist>
(re ping times, seems the fist hop out of my house is literally 12ms. so that's just overhead of dealing with comcast's cable network)
<heat>
oh I know why google DNS is slow here
<heat>
it's going all the way to belgium
<heat>
-_-
<geist>
ah see now my unbound server has for some reason decided to fall back to a single server. i've seen this more than once. it seems to just silently forget about 7 servers after a while
<heat>
wait, not quite, wtf
<geist>
dunno if thats expected behavior (like it decides after a while that this is the 'best' server and closes sockets to everything else)
isaacwoods has quit [Quit: WeeChat 3.7.1]
<geist>
ah the others came back. okay,so there's some sort of behavior of unbound that is non obvious
<geist>
maybe it's based on holding open sockets to the dns servers because of TLS. perhaps there's some sort of load balancing the dns servers do where they kick you off after a while because you're holding a socket open
* geist
shrugs
<geist>
not really important, everything is working fine, just curious
<heat>
yeah
<einkoder>
what does MMU means?
<geist>
memory management unit
<heat>
geist, wait doesn't TLS need a domain name?
<geist>
i dunno, good question. reminds me, i should fill those in
<geist>
though i dunno if secure dns is TLS or SSL or what version is what
<einkoder>
I watched a ccc video about 'writing a firmware from scratch' and the talk explains that interactions with peripheral is done via memory addresses? how is that possible?
<GeDaMo>
Memory mapped I/O
<gog>
yeh the physical address space is more than just RAM
vdamewood has quit [Read error: Connection reset by peer]
<heat>
it's also ✨magic✨
<gog>
indeed
vdamewood has joined #osdev
<heat>
have I mentioned how intel x86 CPUs lock up if they have to internally write to ROM
<heat>
(as writeback of bits, etc)
<gog>
interstng
<heat>
which is why x86 firmware when setting up the GDT needs to write the accessed bit on each entry
<geist>
einkoder: so say for exampe you have a peripheral that has say 4 registers of 32 bits of size each
<geist>
the hardware could, for example, 'memory map' the 4 registers as address 0x1000, 0x1004, 0x1008, 0x100c
eck has quit [Quit: PIRCH98:WIN 95/98/WIN NT:1.0 (build 1.0.1.1190)]
<geist>
and if the cpu accesses any of those 'memory' addresses instead of reading/writing to memory, it actually accesses the hardware
eck has joined #osdev
<geist>
most hardware and anything modern is memory mapped like this, io mapping is a legacy x86 thing that is only around because x86 is still around
<gog>
yes, PCI devices have one or more base address registers which sets where their resources exist in physical address space for example
<gog>
before ISAPnP this was configured wiht jumpers on the card
<geist>
heads up! we got a jumper!
<heat>
nooooo dont jump
<heat>
its not worth it
<geist>
<insert van halen reference>
<heat>
intel chipsets in fact do have ways to remap memory
<heat>
you don't lose memory by having MMIO
<heat>
qemu does it on its own
<geist>
side note, i'm seeing that newer 11th and 12th gen stuff seems to generally not bother including as much lower (<4GB) memory)
<geist>
the alder lake machine i have here stops at 2GB before swtiching to IO
<geist>
one of the nuc11s at work literally only maps 1GB before switching to IO
<gog>
i have no idea what my chip does
<heat>
heh that's cool
<geist>
i guess they just gave up pretending that folks run 32bit stuff anymore
<heat>
I think the standard is 0xE0000000 or 0xC0000000?
<geist>
i can see the endgame here: just include some basic amount of low memory, move everything above it. maybe 1GB? maybe 128MB?
<heat>
I can't remember
<gog>
i wonder how many memory regions can exist like that
<gog>
surely chipset dependent
<heat>
QEMU 0xB0000000
<geist>
well the standard is 'whatever you didn't otherwise need for io' and it's configurable via MSRs and PCI device 0:0
<heat>
TOLUD BABY
<geist>
you can configure it. the chipset docs and/or AMD BKDG actually show you the knobs you can turn. and yeah its TOLUD on intel
<heat>
i'll spew some loose possibly not quite correct facts, TOLUD marks the point at which accesses stop being decoded by the memory controller and start being decoded by IIRC DMI
<geist>
but you can imagine back when folks still ran a lot of 32bit stuff without PAE there was an incentive to try to minimize the amout of lost memory <4GB, but that seems to be generally not a concern much anymore, which makes sense
<geist>
heat: yeah basically. and then there's a TOLUD2 which is the same thing >4GB. it's the point at which the 64bit bank of ram stops and you can start mapping more pci
<geist>
there's some more lines in there with 'stolen graphics memory' too, but in general it's fairly straightforward with a diagram
<heat>
fun fact: PAE doesn't solve much in linux
<heat>
they say it's only "officially supported" up to 8GB of ram
<geist>
yah i'm generally thinking about the juggernaut that folks still run: Windows XP 32bit
<gog>
how big is struct page on x86? 64 bytes?
<geist>
gog: yeah i was thinking what you were thinking
<geist>
at some point the needed-to-be-mapped pmm bits would chew up most of the kernel aspace
<gog>
yeh
<geist>
i'm gonna guess something between 32 and 64 bytes
<geist>
in that range
<gog>
i think it depends on kernel features yeh
<heat>
funny you mention that
<geist>
this is a problem i haven't bothered to solve on LK (lots of memory on 32bit systems with mmus)
<geist>
i basically jsut kneecap ram to something like 1GB and say get a real cpu
<bslsk05>
www.kernel.org: High Memory Handling — The Linux Kernel documentation
krom has quit [Remote host closed the connection]
<geist>
not really worth the trouble right now building up an entire infrastructure to access memory that you can't just map into the kernel
<geist>
(IMO)
<heat>
kmap go brrr
<geist>
the problem i have bumped into is ACPI tables being mapped really high. my code can't find them because they're not otherwise mapped at boot time
<geist>
a TO-DO item
<geist>
some machines put them 3GB+, most seem to put them lower
<heat>
why do you touch them before mapping stuff?
<geist>
because the VM hadn't been initialized yet
<gog>
we've had amd64 for 20 years now so anybody still running classic 686 is doing so for a special purpose or is a retro enthusiast
<heat>
for me acpi is a solid "after the kernel is solidly initialized"
<geist>
and its early in the boot process
<geist>
but yeah, that may also be the solution
<gog>
ooh retro enthusiast OS
<gog>
or just use windows xp idk
<geist>
i just generally bring up ACPI initial detection very early, but it's probalby a bad idea for exactly this reason
<heat>
particularly, ACPICA and my acpi subsystem require memory allocation
<geist>
the main thing that deps on it is bringing up secondary cpus, interrupt controllers, and PCI
<geist>
yeah, i'm just thinking about the initial find the RSDT stuff
<heat>
yeah that's all late game here
<heat>
friendly note that you can't search for the RSDT in EFI systems
<geist>
yes i know, in that case it skips the search logic (if/when i get around to it)
<geist>
EFI boot support that is
<geist>
the anchor is of course always in 640k hole, but lots of times it points up 'high'
<geist>
yeah i'll fix this next time i sit down, hadn't thought abou tit in a while
<geist>
but alas gotta work in the code mines for a bit
<heat>
you've been sitting up ever since?
<heat>
man, great glutes
<geist>
i have a standing desk, actually use it most of the time
<heat>
:DD
<gog>
i have a standing desk but i do not stand
<heat>
i have a non-standing desk where I non-stand
<geist>
it's electric, so you can switch back and forth. actually really does help with posture
<heat>
annoyingly sometimes I feel the effects of 2MUCH COMPUTER DESK
<heat>
walking in particular, running is totally fine
netbsduser has joined #osdev
<heat>
man i should go for something else
<heat>
computers are for nerds
<einkoder>
geist: what are alternative to io mapping? what does risc-v does?
tacco has joined #osdev
<geist>
memory mapped io
<geist>
that's what basically everything uses (except some legacy x86 stuff and older, non modern architectures)
<geist>
riscv especially is 100% mmio
<einkoder>
what do you mean by 100%
<einkoder>
?
<geist>
the alternative to mmio is to have a separate io address space that can only be acceseed via special instructions (see in/out instructions on x86)
<heat>
io ports aren't a thing in riscv
<geist>
100% as in there is no other mechanism but mmio on riscv
<einkoder>
ok
<einkoder>
thanks :)
<heat>
(except spooky hardware registers and spooky sbi calls)
<geist>
heat: yeah i know i'm trying not to confuse the subject
<einkoder>
say, I request to write a block of RAM into a disk, how does the disk ack that the block is written?
<heat>
interrupt
<einkoder>
disk == ssd or hdd
<heat>
or status register at least
<einkoder>
interrupt: is that what select, kqueue, io uring use?
<geist>
right, the disk controller would interrupt the cpu with an interrupt, the cpu runs an interrupt handler (part of the disk driver) which goes and reads a status register, the status register says 'the last block you told me to write was written'
<geist>
(using a simplistic model of a single read/write at a time, etc)
<geist>
einkoder: interrupts are lower level than those software constructs
<geist>
a hardware interrupt is a mechanism for peripherals like disk controllers and timers to 'interrupt' what the cpu is doing to run a piece of code that the OS installs
<gog>
heat: i go to the gym to prevent the too much computer desk problem
<gog>
you should see my biceps :D
<heat>
gog, yeah, I do a boat load of physical activity but I don't do much walking anymore
<geist>
in this case it's running the disk driver 'i have something to tell you' routine
<GeDaMo>
Walking is pretty much my only exercise :|
<einkoder>
I need to read on hardware interrupts
<heat>
which is why running feels mostly always fine, sometimes I feel oddly tired by walking. it's a bit concerning
<geist>
yah, you should. it's a foundational thing that kernels and low level firmware uses to build higher level things like select, kqueue, etc on top of
<heat>
i assume I don't exercise those walking muscles too much
<geist>
anyway, i gotta go for a bit
<einkoder>
bbl ;)
<heat>
einkoder, basically a hardware interrupt will make your CPU run a special piece of code
<heat>
which is how eventually it goes down to poll()
<einkoder>
yeah, I just wonder how the CPU can pause, and restart without having to mess with values inside registers
<GeDaMo>
Everything gets saved then restored afterwards
<einkoder>
what does the saving / restore thing?
<heat>
it saves some, you save the rest, InterruptHandler(), you restore what you saved, CPU restores what you saved
<heat>
depends on the arch
<heat>
sillicon + some instructions
eck has quit [Quit: PIRCH98:WIN 95/98/WIN NT:1.0 (build 1.0.1.1190)]
eck has joined #osdev
archenoth has quit [Ping timeout: 255 seconds]
archenoth has joined #osdev
CryptoDavid has joined #osdev
archenoth has quit [Ping timeout: 255 seconds]
gog` has joined #osdev
gog has quit [Ping timeout: 252 seconds]
GeDaMo has quit [Quit: You are becoming what we French call 'Le Fruitcake'.]
dude12312414 has joined #osdev
<mrvn>
A separate IO address space basically just says you don't have enough address bits.
<mrvn>
The time where the cost of address decoder being a deciding factor has long been past.
<mrvn>
einkoder: For example on x86 the CPU saves a bunch of crap on the stack but on ARM the CPU flips a internal bit and some registers gets switched around to banked versions.
<einkoder>
I want to explore firmware or os dev without an actual hardware, but something simpler but similar to x86 et al.
<einkoder>
And hopefully, some hardware that is FLOSS.
<mrvn>
einkoder: bochs, qemu, other VMs are great for that
<einkoder>
Does something exists in that intersection?
<mrvn>
simpler but similar to x86 is a bit of a contradiction
<mrvn>
I can only recommend ARM/ARM64 because it's rather clean and if you do later want hardware there is tons of cheap hardware for it.
<einkoder>
ok
<einkoder>
I am looking at interim os paper, I really like the single language approach using a high level language
<einkoder>
it is targeting an rpi
<mrvn>
for FLOSS hardware I think risc-v is the only thing out there
<mrvn>
einkoder: You only need a very little asm and maybe some C code to create the glue for any language.
<mrvn>
namely bootstraping the CPU and interrupts. Some inline asm or builtins for cache handling and page tables and you are good.
<mrvn>
Modern C++ now has the concept of memory models builtin (in the STL). Not sure how many high level languages even have a concept for that.
kori has quit [Quit: zzz]
zaquest has quit [Ping timeout: 252 seconds]
epony has quit [Quit: QUIT]
epony has joined #osdev
wootehfoot has quit [Read error: Connection reset by peer]
<bslsk05>
'Can a Moon Orbit in a Square? Yes! #shorts' by The Science Asylum (00:00:53)
zaquest has joined #osdev
kori has joined #osdev
<einkoder>
fun
<einkoder>
I read the risc-v is not that FLOSS and some IP shit is happening
Gooberpatrol66 has quit [Quit: Leaving]
<Griwes>
linky?
netbsduser has quit [Quit: Leaving]
<einkoder>
linky?
<mrvn>
where did you read that?
* mrvn
doesn't quite see the need for a FOOS SoC. It's not like I can edit the sources and make my own CPU chips. We are still a bit away from someone offering a chip making service for something like 10-100 copies.
Gooberpatrol66 has joined #osdev
lkurusa has quit [Quit: I probably fell asleep (or went out). Who will ever know.]
dude12312414 has quit [Remote host closed the connection]
vai has joined #osdev
vai is now known as Jari--
<Griwes>
mrvn, I mean it'd be nice for risc-v to be usable without the posibility dramas like what's going on between arm and qualcomm
CryptoDavid has quit [Quit: Connection closed for inactivity]