klange changed the topic of #osdev to: Operating System Development || Don't ask to ask---just ask! || For 3+ LoC, use a pastebin (for example https://gist.github.com/) || Stats + Old logs: http://osdev-logs.qzx.com New Logs: https://libera.irclog.whitequark.org/osdev || Visit https://wiki.osdev.org and https://forum.osdev.org || Books: https://wiki.osdev.org/Books
<gog> no
<gog> not in windows
<gaxar77> Oh ok.
<gog> DOS did that
<gog> well, also
<morgan> linux system calls also work like that at least sometimes
<gog> system calls were sometimes implemented as one interrupt
<gaxar77> So, what part of the OS does the Window stuff then?
<gog> it depends on the OS
<geist> it's complicated
<gog> and it requires a lot of parts working together
<geist> i honestly dont know where the core windowing stuff lives in windows 10/11 anymore
<gog> dwm
<morgan> is it not still in dwm?
<geist> it was in the kernel, out of the kernel, spread across a bunch of layers
<geist> dwm.exe, in user space? i though there was still some component in kernel space
<gog> sort of
<gog> video driver is two part
<gog> kernel and user side
<geist> i thought it had a lot to do with tracking resources that back the HWNDs and whatnot and where to route them
<gaxar77> Can you recommend a book on OS development for beginners?
<morgan> yeah I guess the real answer is simply that "window stuff" does not happen in one place
<geist> i suppose that could be layered on top of a more generic IPC nowadays, instead of directly like it was in the old days
<gaxar77> I prefer a book.
<gog> iirc they split that out of kernel mode with W7
<geist> yah that's about when i stoppde paying attention
<gog> and that there is still a part of resource management there
<gog> but they wanted to make it possible to have headless servers with windows
<gog> secure situations
<gog> no unnecessary code
<gaxar77> What's that library called? MSMQ?
Matt|home has quit [Remote host closed the connection]
<gaxar77> The one for messaging between proceses?
nyah has quit [Quit: leaving]
<gaxar77> I'm reading a book I found on OS Dev
<gaxar77> I forget what it's called.
<gaxar77> But is that a good book or should I buy something from Amazon?
heat has quit [Remote host closed the connection]
heat has joined #osdev
<gaxar77> Is this a good book on here? http://539kernel.com/
<bslsk05> ​539kernel.com: A Journey in Creating an Operating System Kernel: The 539kernel Book - Home
<heat> geist, seems that linux is keeping ia64 support after all
<heat> ia-64 stronk
<gaxar77> I think I'll read that tomorrow.
<gaxar77> Or at least get started.
<gaxar77> Can a kernel load and unload modules of itself while running?
<gog> msmsqs
<gaxar77> Like drivers and all kinds of stuff like that
<gog> yes
<gog> that's a basic function of a kernel
<gaxar77> I see.
<gaxar77> Well, if that's the basic function of a kernel, why do I have to restart Windows after every update?
<nikolar> Well there are also kernel updates
<lav> windows lacks basic functionality /j
<gaxar77> LOL
<gaxar77> lol
<gaxar77> Sorry, caps was on.
<nikolar> You should also restart when you update the Linux kernel
<nikolar> Or things eventually start braking
<gaxar77> I see,
<gog> in part it's to ensure that the updated components are actually updated
<gog> another reason is that sometimes files are scheduled to be overwritten at shutdown or boot up
<gog> so as not to disrupt running applications
<gaxar77> I see.
<gaxar77> Could experience developing an OS help me get a job working on an actual OS?
<nikolar> I once had an uptime of 70 days, and after a couple of kernel updates the kernel stopped recognizing USB devices for whatever reason
<gaxar77> Tomorrow morning, I hopefully will begin reading this book on kernel development.
<gaxar77> I see.
<gog> idk my experience programming in general got me a web dev job
<gog> which i was reluctant to do but i actually enjoy it a lot
<gaxar77> cool
<gog> mostly the company i work for
<gaxar77> Well, I did work at a small startup.
<gog> the job itself is mid
<morgan> yeah I had a system with some distro that a) didn't load whatever kernel module was responsible for USB block devices on startup and b) aggressively removed old kernel module trees and so every time i updated the kernel on there USB drivers wouldn't work until i restarted it
<gaxar77> Doing .NET, SharePoint, and CRM stuff.
<gaxar77> How long does it usually take for one person to implement a fully functional OS?
<morgan> depends on how good you are at programming and what your threshold for "fully functional" is
<morgan> so somewhere between one day and eighty years
<gaxar77> Like, GUI, and TCP/UDP support, and WiFI.
<gaxar77> And everything prerequisite.
<gaxar77> And file systems.
<gaxar77> And stuff.
<morgan> "a while"
<morgan> none of those are particularly easy
<gaxar77> Oh ok.
<gaxar77> I knew that WiFi was not easy
<gaxar77> I figured I could do GUI.
<morgan> wifi might be the easiest of those, assuming you're only targeting one hardware configuration
<gaxar77> But I don't know how to do region shaped Windows and Widgets.
<gaxar77> I only have ideas about rectangular ones.
<morgan> hell writing a *good* TCP/IP stack might literally be impossible for one person to do
<gaxar77> oh
Turn_Left has joined #osdev
<gaxar77> Why don't any of you form OS Dev teams?
<gaxar77> If you did, you could build more powerful systems.
<heat> morgan, not true
<klys> gaxar77, are you part of the discord server
<gaxar77> What is that?
<heat> morgan, 1) define good tcp/ip stack
<heat> 2) sortie's stack is pretty high quality. probably not very performant, but high quality
<morgan> i'll do no such thing
Left_Turn has quit [Ping timeout: 246 seconds]
<heat> can I copy one google chrome's data to the other (other operating system)?
<heat> does it keep my preference data and logins
<gaxar77> Has anyone here written a full web browser for their OS?
<CompanionCube> lol no
<heat> i wanted to give chrome another shot but damn I do not feel like re-logging in
<gaxar77> lol
<heat> shit
<CompanionCube> heat: maybe you could try copying the profile dir
<CompanionCube> i wouldn't know, i use firefox
<heat> yeah i'm just tired of firefox
<morgan> or whatever chrome's equivalent of the profile dir is
<heat> also seems to keep working worse over time
<gaxar77> What's harder? Writing a Web Browser or writing an OS?
<morgan> web browser, 100%
<gaxar77> Why?
<heat> lately I've been having a ff bug where sometimes the text cursor just disappears randomly and I need to refocus it on the URL bar and back to where I was typing
<CompanionCube> gaxar77: consider that microsoft gave up on developing their own web browser
<gaxar77> Yeah. Why did they do that?
<klys> gaxar77, because you have to implement html5, javascript, dom, etc.
<gaxar77> They couldn't do that?
<morgan> go have a quick read through the HTML5 and JS specs for us would you
<heat> CompanionCube, the serenityOS people gave up on developing their own operating system and now almost solely develop their web browser :v
<gaxar77> I know it's large...
<CompanionCube> isn't that mostly one person?
<gaxar77> But MS should have been able to do it.
<heat> no
<morgan> keep in mind you've gotta implement all of that if you want modern websites to work. oh also you'll have to keep it up to date with standards changes if you want websites to keep working. oh also all of this has incredible performance and security concerns about it
<klys> your eyes are bigger than your stomach
* gog eyes klys
<klys> ^,,^
<heat> :3
<gaxar77> I'm not saying I can build a browser.
<heat> morgan, chrome's security record is pretty impressive
<gaxar77> But I want to learn how to build a kernel.
<heat> they are soooo goooooooood at what they do
<heat> chromium people, v8 people, etc
<gaxar77> Will there ever be something to replace HTML5, CSS3, etc?
<gog> probably
<gaxar77> When?
<morgan> yes and if they're paying less than $10,000,000 a year in salary to the people responsible for keeping that security record up i'd be astonished
<gog> but more like supersede
<klys> they'll prolly call them html6 and css4, etc.
<gog> yes
<gaxar77> oh
<gaxar77> If a bunch of you worked on one OS project what would happen?
<morgan> arguments, presumably
<gaxar77> lol
<geist> not a chance
<gaxar77> Why not?
<klys> if you thorw a bunch of noodles into a pot it doesn't form a mind easily
<geist> right there needs to be a vision, some sort of direction, and then everyone would need to be paid full time for many years
<gaxar77> I see.
<geist> but cloeset i've seen is the haiku project
<gog> herding cats
<gog> haiku is so good though
<geist> they had a vision and a direction, which was to directly reimplement beos, which i think helped immensely
<heat> tbf I think I could work with geist pretty well
<gog> i should really try to get into working on it again
<heat> although I would probably want to implement UNIX
<geist> probably yeah
<geist> yah i dont ind implementing unix
<heat> and geist thinks thats laaaaaaaaame
<geist> it's fun
<geist> gosh no. why do you think that?
<morgan> but we already have unix
<gog> do y'all wanna implement unix poorly together?
<klys> svr4?
<gaxar77> Is Minix slower than Linux?
<heat> geist, because you work on the opposite of unix :P
<geist> heat: note the phrase 'work'
<geist> ie i'm paid to do what others want me to do, so i work within those constructs
<gog> minix is an academic project not a mainstream OS
<gog> it's not as performant in a lot of ways
<gaxar77> oh
* heat nods
<heat> <gog> do y'all wanna implement unix poorly together? <-- don't steal my shtick
<gog> sorry
<gog> won't happen again
<sakasama> gog: Can it be on a mixed ternary/quaternary computer, so we can make our mistakes on an experimental architecture?
<geist> if when i ever get uninterrupted time and motivation i want to take the lkuser project to a real posix thing
<geist> because why the hell not
<gog> sakasama: where are we going to get one of those
<gog> which channel am i in wtf
<gaxar77> How can a flash drive have a boot sector?
<gaxar77> Or an SSD?
<gog> depends on the firmware the computer has
<klys> sector zero is used, and it has the 55 AA magic
<morgan> well the boot sector is just sector 0
<gog> if it's booting with standard BIOS you write to sector 0 with the correct format
<sakasama> gog: I'll design one and entrap someone to manufacture it or perhaps emulate it using neurons harvested from snails.
<gaxar77> I thought sectors were something to do with magnetic disks.
<gog> sakasama: neat
<geist> you thought wrong
<gog> a sector is the right word for a segment of a cylinder
<gog> but the terminology carried over
<klys> a track is
<gaxar77> oh ok
<gog> right
<geist> now it just generally means a unit of a block oriented device
<gog> no a sector is a segment of a track, a cylinder is the projection of all tracks
<gog> yes
<geist> and on many types of computer, the 0th sector can be loaded by firmware and booted, and thus it's the boot sector
<gog> yes
<geist> 0th sector because why not. the standard could be the 1st or 2nd, but 0th seems like simple enough
<morgan> well it's the first one
<klys> # qemu-system-x86_64 -device ide-hd,drive=drive0,lcyls=476940,lheads=64,lsecs=32 -drive file=/dev/nvme0n1,cache=none,if=none,id=drive0,media=disk,format=raw,index=1 \
<gaxar77> But is the kernel a file?
<gaxar77> Or just some bytes on the drive that the bootloader loads?
<gaxar77> file or set of files
<gog> yes
<geist> excalty. if you're building a piece of firmware you could, for exmaple, declare that the 0th sector has some sort of descriptor that points to the real boot sector, or some sort of hardware description, or whatnot
<gog> it can be any of that
<klys> in this command I specify the logical cylinders heads and sectors-per-track values for my nvme ssd
<geist> but PCs are simple devices, and PC bios just loads 0 into ram and branches into it
<geist> and that has generally stuck
<gaxar77> I see.
<heat> gog, I think i'm going to try and make sic into an EFI irc client
<heat> for shits and giggles
<geist> moder UEFI desn't use a boot sector, for example, bcause it parses soething more complicated
<gog> heat: :O
<zid> uefi has some funny class ids for the partition types afaik?
<gaxar77> Now I'm thinking about the name bootloader...
<geist> yeh and thats how it finds the BSP partition (boot something partition) and then thats formatted FAT and holds the thing it'll start, etc
<morgan> uefi is basically a whole ass operating system that exists to find your actual OS and load that
<heat> zid, yeah partitions have GUIDs
<gaxar77> I think the bootloader's job is to load the something more basic than the kernel.
<geist> yah basically what morgan said
<heat> note that they don't actually matter AFAIK
<gaxar77> Considering it's called the "boot" "loader"
<heat> they are just part of the boot path
<gog> yeah it loads your boots
<zid> guids matter they're the best
<klys> the bootloader traditionally loads the grub2 core.img file which loads a kernel via a module that recognizes that kernel's format
<zid> 21686148-6449-6E6F-744E-656564454649
<zid> that's the bios one
<gaxar77> I see.
<morgan> idk grub2 core.img is also part of the bootloader imo
<zid> '!haHdInotNeedEFI'
<morgan> bootloader-ing stops when you get to the kernel
<geist> but there are 100s or 1000s of type of machines that have existed over the years, with varying amounts of boot schemes
<zid> idk how to properly endian flip it
<geist> the PC one is particularly simple (prior to UEFI) but it's nice
<gaxar77> Can anyone think of an fs object besides files and folders?
<klys> inodes?
<gaxar77> Or is that a dangerous brain exercise?
<gaxar77> What's that?
<gog> files and folders are abstractions
<zid> 'Hah! I don't Need EFI' is the GUID for a dos partition, heat, now tell me GUIDs don't matter
<morgan> inodes are what modern filesystems actually track. files and folders are certain types of inodes
<gaxar77> oh ok
<morgan> you can also have symbolic links! there may be others
<FireFly> there's def other kinds of things in filesystems
<heat> zid, heh that's pretty funny
<gaxar77> Yeah, I read about them in the Java tutorials.
<geist> i'm always fairly foind of older machinse where the boot mechanism is some human keys in the bootloader with switches
<morgan> i've done that! it's a total pain in the ass
<geist> or the next step: the main machine is booted by another machine that loads the system into ram and then resets the main machine
<geist> lots of mainframes did it that way
<morgan> though even on those you're usually toggling in the first stage of a multi-stage bootloader
<geist> yah or you're keying in the load address for the ROM that does it
<zid> heat: make sure you write if(efi_guid_strcmp("Hah! I don't Need EFI", partition) == 0) somewhere
<geist> i have an altair machine where when you boot it you have to halt, set the address to FF00 and then release the cpu
<geist> since that's where the ROM block is that loads something off a disk
<morgan> i'm sort of surprised they couldn't just make it do that
<morgan> or y'know put that block at 0000
<geist> you could, and lots of later 8080 and z80 machines did exactly that
<geist> trouble is the 8080/z80 has fixed exception vector at 0, and the reset vector
<geist> so you eventually want to put ram there
<morgan> ah right
<geist> so that means yo need a piece of logic to move the ROM out of the way once it's booted
<gaxar77> Okay. I'm going to go for now. Thank you for the insight and conversation.
<geist> and that's probably a whole nother board, etc
<morgan> or a piece of logic to bonk the CPU over to 0xFF00 at startup
<geist> yah but you still want ram at 0 eventually because code expects to fiddle with the exception table
gaxar77 has quit [Quit: Leaving]
<morgan> oh does it end up remapping it
<geist> depends on the machine. i dont remember what a lot of them do. the altair in particular being very early does nothing of the sort
<geist> soince itwas doing good to just get a cpu and like 4K of ram in the same box
<geist> later stuff, like say a Kaypro CP/M machine had more sophisticated autostart rom that i think it did the zero remap thing
<morgan> i should fuck around on cp/m someday
<geist> iirc the kaypro does something like you hit a register and it swaps either the bottom 32k or top 32k with a ROM where a lot of the BIOS was iplement
<geist> dont remember if it was bottom or top
<geist> but normally you're running with a full 64k RAM, and CP/M is running in the top banks
<geist> so something simple like the altair just starts filling in rom from the very top, and CP/M loads to just under that, and then something like 0x100 - bottom_of_cpm is user memory basically
<geist> where 0 - 0x100 or so is the exception vectors. i think 0x100, maybe more like 0x300. there was some sort of file control block stuff down there i think
<morgan> and then someday we invented the memory management unit
<geist> cp/ms big advancment at least was you could actually write a real program that could actually run on different cp/M machines, because it actually abstracted enough of the hardware via OS and bios calls so that you didn't really need to know
<geist> obviously meant only one program could be loaded at a time realistically, but the load address was known because cp/m got out of the way
<geist> now it seems kinda obvious, but that was a big win vs a bunch of homebrew stuff that tended to pop up around one or another of the many machines floating around
SpikeHeron has quit [Quit: WeeChat 3.8]
<zid> DOS was abstract until you needed to do audio :P
<geist> yah DOS more or less copied the CP/M model
<zid> "What is the exact model of your sound card, what are all of its jumpers set to, which slot is it inserted into, what is your mother's maiden name?"
<zid> "initialization failed"
dude12312414 has joined #osdev
<morgan> audio might as well still be like that honestly
<zid> audio's worked fine on windows since forever, linux's had alsa for.. also nearly forever
gog has quit [Ping timeout: 255 seconds]
<heat> i give up
<heat> efi networking is so complex
<zid> on audio?
<heat> fucc this
<zid> oh
dude12312414 has quit [Quit: THE RAM IS TOO DAMN HIGH]
<heat> all this for a joke? nah im good
gog has joined #osdev
<heat> gog, gave up, sorry
<gog> hm?
<heat> on the efi irc client
<gog> o
<gog> why?
<heat> have you looked at the networking protocols?
<gog> not really lol
<zid> important announcement, dosh has finished his sushi belt playthrough of factorio and uploaded
<heat> the basic EFI_TCP4_PROTOCOL::Configure() takes like 30 parameters no joke
<heat> can I set it up? probably, but honestly screw that
<heat> https://github.com/tianocore/edk2-libc <-- if anyone (or future me) wants to actually do it
<bslsk05> ​tianocore/edk2-libc - Port of libc to EDK II with applications that depend on libc (44 forks/50 stargazers/NOASSERTION)
divine has joined #osdev
<gog> I'll do it tomorrow
<gog> a new toy eee
mctpyt has joined #osdev
SpikeHeron has joined #osdev
gxt has quit [Remote host closed the connection]
gxt has joined #osdev
heat has quit [Ping timeout: 246 seconds]
heat has joined #osdev
joe9 has joined #osdev
joe9 has quit [Client Quit]
slidercrank has joined #osdev
bradd has quit [Ping timeout: 255 seconds]
mctpyt has quit [Ping timeout: 252 seconds]
zaquest has joined #osdev
Arthuria has joined #osdev
Matt|home has joined #osdev
xenos1984 has quit [Read error: Connection reset by peer]
Arthuria has quit [Killed (NickServ (GHOST command used by Guest684531))]
Arthuria has joined #osdev
xenos1984 has joined #osdev
fedorafan has quit [Quit: Textual IRC Client: www.textualapp.com]
Arthuria has quit [Remote host closed the connection]
Arthuria has joined #osdev
heat has quit [Ping timeout: 252 seconds]
[itchyjunk] has quit [Ping timeout: 255 seconds]
Arthuria has quit [Remote host closed the connection]
matheusmoreira has joined #osdev
daily has joined #osdev
daily has quit [Quit: Leaving]
bradd has joined #osdev
<bslsk05> ​github.com: edk2-libc/Tcp4.c at master · tianocore/edk2-libc · GitHub
kof123 has left #osdev [#osdev]
<geist> huh. re: recent discussion about memory usage, chrome seemed to have just added an explicit 'memory saver' setting
<geist> dunno what it does over the older mechanism. maybe it's mor aggressive
<zid> yea I noticed when i was using 4GB ram that it's very very aggressive at unloading tabs
<zid> swapping tab was doing a full network refresh
<geist> yah this new setting lets you actually override that for particular sites
<zid> yea I added discord to it so I could tab back properly
<geist> i was thinking of doing precisely that
<zid> I use desktop discord normally though
<zid> I just didn't have anything installed yet
<zid> was just doing windows updates and stuff
<zid> I think I settled on which ssd to buy so that I can replicate all this suffering of reinstalling windows on another day at least
<geist> ugh trying to remember that chrome:: page i shared the other day
<geist> i always have to look uit up again
<zid> https://www.amazon.co.uk/dp/B0781VSXBP cheap SSDs are cheap now at least, my dead 850 pro was similar specs and cost like 4x that
<bslsk05> ​www.amazon.co.uk: 503 - Service Unavailable Error
<geist> chrome:://discards
<geist> ooh did your 850 pro die because of the known issue with dead 850s?
<geist> if so you might be able to get a refund from samsung
<zid> no idea, it just went bang
<geist> oh no that's 950s i think
<geist> er 980s and 990 pros
<geist> oh it went bang and took out a bunch of your computer?
<zid> no even fucking worse
<zid> my computer went bang
<zid> 2 days of trying to revive it and beg/borrow/stealing parts to make a new PC
<zid> finally get it POSTing, windows installing
<geist> well i mean yeah your power supply probably went bang and took out the rest of it maybe?
<geist> oh i see. but that killed your 850 too?
<zid> I go to chown C:/windows.old/
<zid> and it died half way through
<zid> losing me all my files, and the windows install I just spent 2 days getting back
<zid> I cried
<zid> god samsung's website is crappy
<geist> it's the universes gentle reminder to back up your data
<zid> most of it is, but windows isn't exactly "yea just copy /etc"
<zid> it's harder to try restore it than start fresh
<geist> yah i usually just start fresh in these situations
<geist> if possible copy out all your appdata/roaming somewhere, start clean and then try to page back in what you needed
<geist> half the tmie you only end up reinstalling half of what you had before anyway
<zid> yea I lost most of that, so I lose all my video game saves and stuff
fedorafan has joined #osdev
<geist> yah that you should back up
<geist> i should double check mie is till running it's backups
<geist> windows file backup is such a piece of shit. it seems to eventualy stumble over something that it can't backup, so it just starts generating logs somewhere that says it cant
<geist> have had this happen more than once
<geist> i'd loe for it to tell me "i failed to back up your data you ight want to look at it"
<zid> can't even find the email for buying this ssd
<zid> it's.. pretty old :D
<geist> yah would you look at that. windows file backup last ran 5 days ago
<geist> and now it's completely lost it's configuration. silent
<geist> says 'nope. no backups here!'
<geist> piece of shit!
<geist> and now i dont trust it either, because it told it to run and it returned back like 5 seconds later: "okay! done!"
<geist> like, did you really back anything up? goddamnit
<zid> let's see if the person whose amazon account I used still has the order info I guess
<geist> reminds me i shold flash the firmware on my 980 pro. it's a later version than the affected one, but still one back from mainline
<geist> looks like if you grub around on samsungs site yuo can get to the download part where you can get an iso boot disk for it
<geist> to update without running winders
<zid> neat
<bslsk05> ​semiconductor.samsung.com: Samsung Magician & SSD Tools & Software Update | Samsung Semiconductor Global
<zid> I can just samsung wizard it on windows probably, or I could
<zid> except it's gone from the bus
<zid> total paperweight
<zid> They *really* need to stop making SSD firmware that just stops the pci-e/sata/etc controller working if the flash goes bad
<zid> it seems comically normal
<geist> usually samsung is good at that. they just go read only
<zid> depends how they fail
<geist> huh looks like my windows NVME drive is at 50%
<zid> if they fail while updating the internal journalling / wear levelling etc
<geist> lower than i'd like
<zid> they fail dead
<geist> about 50TB written (2TB drive)
<zid> the controller just abort()'s
<zid> and it never turns on
<geist> hmm, this drive *should* be like 1200 TBW endurance rating. the fact that it's at 50% after only 50TBW is a bit concerning
<geist> unless the smart entry isn't read that way
<zid> 1200BTW is insane
<zid> I've only seen hundreds
<geist> yah, that's what it says at least. 970 EVO 2TB
<zid> I just saw some voodoo, apparently leaving it plugged in but with no sata cable for an hour
<zid> can make it come back to life a bit
<zid> maybe
<geist> yah i actually have an 850 pro here
<geist> it's my old root disk that now i just use as D:
<zid> I turned hotplug on on all the sata ports, we'll give it a bit then add a sata cable I guess
<zid> if it comes back I can at least grab Roaming/
<geist> well, i guess it is 5 years old, so it's had a good run. apparently i bought it back in 2018
bnchs has quit [Remote host closed the connection]
V-T60 has quit [Ping timeout: 265 seconds]
<zid> ah I think the big capacity ones just scale the TBW linearly
<geist> yah, in general yeah
<zid> so it has 8x the TBW of my 150TBW or whatever on the 256
<zid> which adds up about perfect
<geist> though it's possible that thats in perfect condition (ie, completely empty and trimmed drive)
<geist> whereas since i've had it mostly filled it's really having a lot of internal write amplification
<geist> but i dnot think they report the internal amount of writes/erases that have happened, only the amount of writes at the interface
<zid> yea that's what killed my ssd I think, installing windows again on it a second time (windows moves to windows.old)
<zid> then doing the big chown so that I could access the files
<zid> was lots of writes on a fullish drive
<geist> yah i've had mine about 80% for some time, and who knows how much windows slams stuff to the drive
<zid> windows fucking rapes ssds
<geist> i fairly recentlyish moved my temp dir over to d: and whatnot
<zid> temp barely ges touched
<geist> probably not yeah
<zid> windows just loves molesting random files
<geist> and the stupid 'run trim every night' thing i've set just doesn't do it most of the time
<geist> i can even leave the computer on, nope. i dont think i choose to run tonight
<geist> i dont know precisely what planets need to align for it to decide to actually do it
<geist> it's a frustrating mess. tis why i mostly just use windows for games n browser. no heavy lifting
<zid> my computer woke me up, some 5am scheduled task + mech drive = omg the army is invadi- oh it's just the drive
<zid> usually my drives just get 20GB .mkvs on them or whatever, not used to them going absolutely nuts :P
<geist> that's what i have a NAS for
<geist> but it can do nuts all it wants, thats its job
<zid> I'd have had to pull the nas apart to install windows on it regardless
<zid> firefox has decided I don't deserve.. fonts?
<zid> all text missing
<geist> hieroglyphics FTW!
<zid> did my hdd just fucking die now or something
<zid> maybe plugging that sata power in upset things
<zid> my drives are unsresponsive
<geist> you're not using the same power supply are you?
<geist> from the old dead computer
zid has quit [Read error: Connection reset by peer]
matheusmoreira has left #osdev [#osdev]
zid has joined #osdev
<zid> *aggressive windows startup mech drive noises*
<zid> firefox vs discord vs steam, let's see who wins
<zid> steam made its window first, but firefox loads up first, steam second, discord third
<zid> discord *still* loading, infact.
<geist> noice
genpaku has quit [Remote host closed the connection]
genpaku has joined #osdev
<zid> well, the voodoo didn't help it seems
<zid> unless my hotplug didn't work
fedorafan has quit [Ping timeout: 248 seconds]
ThinkT510 has quit [Quit: WeeChat 3.8]
fedorafan has joined #osdev
ThinkT510 has joined #osdev
elastic_dog is now known as Guest9568
Guest9568 has quit [Killed (zinc.libera.chat (Nickname regained by services))]
elastic_dog has joined #osdev
danilogondolfo has joined #osdev
<sortie> <heat> 2) sortie's stack is pretty high quality. probably not very performant, but high quality
<mjg> that's not very high quality
<sortie> ^ I wouldn't say it's unperformant though. I've been able to push 20 MiB/s downloads through it using qemu slirp on my old desktop. I got gigabit though and it'd download more like 100 MiB/s on my Linux host.
<mjg> so you are beating openbsd/ :]
<sortie> Overall that's a decent enough speed and I haven't done much big optimizations
<sortie> But hey I did try to focus more on having a good implementation and that's way more performant than none and I wouldn't want it to be badly insecure
<mjg> so how do you look like weird-shit wise
<mjg> there is no way following rfcs results in a real-world usable stack
<zid> Gotta start with something readable, then patch some dirty disgusting performance hacks in
<zid> so there's a paper trail
zxrom has joined #osdev
<sortie> mjg, what do you mean? I followed the RFCs and it worked
<sortie> I didn't implement all the features just yet but I didn't intentionally deviate
<zid> did you try talking to a windows vista machine using a cheap realtek yet
<zid> that's the true test
<sortie> I don't really recall any workarounds for stuff yet
<sortie> zid, has Windows Vista realtek talked TO ME??
<sortie> So far nobody has reported trouble connecting to my irc.sortix.org network and people have done so from a range of devices
SGautam has joined #osdev
<zid> everybody always asks WHAT is vista, but nobody asks HOW is vista
<sortie> It's running my network stack with TLS on top
<zid> you're behind the router you're behind though, you get the nice sanitized version of everything :(
<mjg> sortie: there are stupid bugs all over
<mjg> sortie: if it worked, you mean you barely use it
<zid> my desktop's internet is fucky and needs tuning hard to make torrents work which is fun
<sortie> mjg, I mean I'm running my own infrastructure natively so it is really being put to use
<sortie> Though these days everyone is running Linux
fedorafan has quit [Ping timeout: 246 seconds]
<SerHack> Does multiboot support RGB with alpha channel? Or only RGB?
Vercas has quit [Quit: Ping timeout (120 seconds)]
bgs has joined #osdev
Vercas has joined #osdev
fedorafan has joined #osdev
GeDaMo has joined #osdev
<zid> GeDaMo: Did you get otto'd?
<GeDaMo> wassat?
<zid> some storm that hit up norf
<zid> friend's internet was dying on its arse cus of it
<GeDaMo> Oh, yeah it was a bit windy yesterday
<GeDaMo> Lots of small branches down but I didn't see any major damage
<zid> squardle resets at such a weird time
<zid> 11am
<GeDaMo> I have a similar game on my phone called Lexica, it's on F-Droid
<zid> I'll have to play weird shit like cell tower
<gog> mew
<zid> moaning gog
<GeDaMo> Do you know about https://wordhoot.com/ ? It's like wordle
<bslsk05> ​wordhoot.com: WordHoot! - Competitive Wordle
<zid> multipooper?
<GeDaMo> I've not tried the multiplayer
<bslsk05> ​wordhoot.com: WordHoot!
wootehfoot has joined #osdev
wootehfoot has quit [Remote host closed the connection]
wootehfoot has joined #osdev
<zid> I think GeDaMo is about to commit sudoku
gog has quit [Quit: byee]
gog has joined #osdev
wootehfoot has quit [Quit: Leaving]
terminalpusher has joined #osdev
terminalpusher has quit [Remote host closed the connection]
bgs has quit [Remote host closed the connection]
<zid> That was my best
<GeDaMo> Yeah, looks like guesses matter
<GeDaMo> Which is a pity but I use a lot of guesses :P
<zid> 10 seconds in 1 guess is 70 points or whatever, 10 seconds in 2 guesses is 35
<zid> so you wanna do few guesses as possible, but still be fast ish
<gog> mew
wootehfoot has joined #osdev
craigo has joined #osdev
<zid> k finished my squardle
<zid> maybe it's time for napple
Left_Turn has joined #osdev
<gog> i'm gonna write codle
<gog> i remembered what i was working on
morgan has quit [Ping timeout: 248 seconds]
Turn_Left has quit [Ping timeout: 252 seconds]
morgan has joined #osdev
bnchs has joined #osdev
bradd has quit [Ping timeout: 252 seconds]
GeDaMo has quit [Ping timeout: 252 seconds]
slidercrank has quit [Ping timeout: 255 seconds]
bnchs has quit [Remote host closed the connection]
SpikeHeron has quit [Quit: WeeChat 3.8]
SpikeHeron has joined #osdev
<SerHack> I guess no
gaxar77 has joined #osdev
<gaxar77> I wanted to begin a tutorial on OSDev today, but thought I should become familiar with the tools first.
<gaxar77> What do you think?
zxrom has quit [Ping timeout: 255 seconds]
<gaxar77> Why does the bare bones tutorial have one function taking a color as an enum, and another function taking a color as an integer?
<gog> enums in C are stored as integers
<gaxar77> I know.
<gog> well there's your answer
<gog> you can use them interchangably
<gaxar77> But that's not the point.
<gog> that is the point
<gaxar77> Should I actually follow this tutorial?
<gaxar77> As a first tutorial on OS Dev?
<gog> yes
<gog> and if it helps you can change the other function taking color as an int to an enum
<gog> it
<gog> 's the same thing
<gog> you can change these tutorials all you want and see what happens
<gog> the point is familiarizing yourself with the basics
<gog> and the tools
<gaxar77> Joke: You know what's a good software for ensuring that everyone agrees with one another? SharePoint
gildasio1 has quit [Remote host closed the connection]
gildasio1 has joined #osdev
<gaxar77> I know, but, I don't understand everything in the code samples (assembly), and I'm not sure if reading the tutorial will be enough.
<gaxar77> Yeah, doesn't this tutorial assume you already know assembly?
<gog> it assumes you have a familiarity with it
<gog> and if not that you can grasp the basics
<gaxar77> I see.
<gog> you don't need to be a veteran expert you just need to be able to learn and be ok with having to maybe go find the answers yourself
<gaxar77> ok
<gaxar77> I'd prefer a book that teaches the assembly while teaching me to write an OS.
<gog> i don't know of any such book
<gaxar77> ok
<gaxar77> Well, I feel kind of sluggish right now.
<gaxar77> It may not be the right time to do this tutorial.
<gog> then maybe have a rest and come back to it another time
<gaxar77> ok
<gaxar77> Maybe I need coffee.
<gaxar77> I ate scoops of whipped cream.
<gaxar77> Maybe that's why I'm sluggish.
<gaxar77> I mean spoon fulls.
<gaxar77> See, I think that the C code is the easy part.
<gaxar77> Even easier than building a cross-compiler.
<gaxar77> I don't understand assembly. And I don't like GAS; I like NASM.
<gaxar77> gog, what are the prerequisites for building an OS? I think they are: knowledge of assembly, linker script, high-level systems language, compiler and linker tools, Intel architecture, hardware protocols (for writing drivers), etc.
<gog> those are prerequisites that will help you, but nobody can master this knowledge in a reasonable amount of time and the only way to do it is to find some task and do it
<gog> so, you start with the tutorial. the "hello world" type thing
<gog> and then decide where you want to go next with it
<gog> maybe you want to be able to have virtual memory
<gog> maybe you want to have a timer
<gog> maybe you want to be able to switch contexts
<gog> maybe you want keyboard input
heat has joined #osdev
<gaxar77> So, do you think I should start with the bare bones tutorial, then?
<gog> i do
<gaxar77> ok.
<gaxar77> I will see if I can try it. I'll have to look up a lot of stuff, and I'll have to get the GNU tools and build a cross-compiler.
<gaxar77> I don't have GNU, neither the source nor the binaries.
<gaxar77> I don't think.
<gaxar77> I'll be back later.
gaxar77 has quit [Quit: Leaving]
<heat> gog, hello gogert langdon world renowned symbologist
<gog> hello uh
<gog> heat
<gog> wow just read a synopsis of angels & demons
<gog> sounds liek rubbish
<gog> and i saw the film version of the da vinci code
<gog> actual rubbish
<heat> the dan brown books were actually so bad they're good
* gog mews sadly
<gog> i'm a rubbish person
<heat> no
* Amorphia raises an eyebrow
<heat> u have self worth
<heat> u go queen
<Amorphia> heat: damn right
<Amorphia> gog: listen to heat
<gog> yes miss
<gog> thank you heat
<heat> np
<Amorphia> gog is good kitty
* Amorphia give kitty feeshes
* gog mew and chomp fishy
<SerHack> Can anyone confirm that the framebuffer requested with multiboot can not handle alpha channel, but only rgb?
<gog> yes
<gog> generally it's going to be 32-bit but in BGRx or RGBx format
<gog> where the x is an ignored byte
<gog> what i don't know is whether tehre are any framebuffers with VBE/GOP that also do alpha blending in hardware
<gog> they probably exist
<gog> i just don't know
<gog> s/framebuffers/devices
<Amorphia> gog: there also exist VECTOR SCREENS
<Amorphia> they're kinda rare and mega retro
<Amorphia> but they exist
<Amorphia> the Computerspiel Museum in Berlin has one
<gog> yes
GeDaMo has joined #osdev
<mrvn> SerHack: what is alpha supposed to do on a framebuffer?
<mrvn> unless you have an additional input that you could blend into it alpha simply makes no sense.
<mrvn> Amorphia: it's too bad the polygon count for those is so low.
<gog> Amorphia: have you seen techmoan
<gog> he has a vectrex that he's taken apart and fixed
<gog> documented the whole thing
<Amorphia> no :o
<gog> i think he has two vectrexes actually
<gog> vectrices??
<Amorphia> vectrex is the thingy they had at the Computespiel Museum
<Amorphia> iirc
<gog> yeh it was a popular home arcade machine
<Amorphia> there was a small one sold for home use but it wasn't very popular
<Amorphia> and didn't have that many games
<SerHack> mrvn: I do not know, I just asked
airplanemodes has joined #osdev
SpikeHeron has quit [Quit: WeeChat 3.8]
SpikeHeron has joined #osdev
SGautam has quit [Quit: Connection closed for inactivity]
<mrvn> SerHack: there is some hardware that uses RGBA mixing in a black background. So the A is basically brightness.
<mrvn> I'm sure the RPi can do that too.
<gog> i want to ask whether there are any with a hardware blit that takes a foreground and background and alpha blends but that's just EGL
<gog> heh
<mrvn> every GPU does that
<gog> yes
<gog> i'm not very smart
<gog> maybe i should go do things that don't make me feel like shit
<mrvn> that's how green screen functions
gog has quit [Quit: byee]
airplanemodes has quit [Quit: WeeChat 3.8]
gaxar77 has joined #osdev
Arthuria has joined #osdev
<gaxar77> How do I install GCC on Windows? Someone said not to use MinGW.
<gaxar77> I don't want to use WSL.
<zid> then install mingw
<gaxar77> Ok.
<gaxar77> actually, I already installed cygwin? Will that interfere?
<gaxar77> I'm installing the compilers and dev tools now.
<zid> I feel samsung's warranty process may be this obtuse intentionally to stop me trying.
<gaxar77> Can an OS be written in objective-C?
<gaxar77> a kernel I mean.
<Ermine> gog: may I pet you
<lav> she left
dude12312414 has joined #osdev
<zid> Ermine: no but you could help me do this warranty claim if you like
<gaxar77> Okay. I installed MingGW, but I don't know where it is installed, so I can't set the env variable path.
<zid> see if you can find me *anything* but a email address that doesn't respond
<Ermine> zid: I have samsung phone and I decided it would be my last samsung phone
Arthuria has quit [Ping timeout: 260 seconds]
<gaxar77> Okay. I set the path variable.
<heat> i really like samsung
<gaxar77> So now I have installed MinGW.
<gaxar77> So now I need to build a cross-compiler.
<heat> i became a samsung stan over the last few years
<heat> really like their products
<gaxar77> I shall procrastinate for a while.
<zid> okay heat you do the warranty claim
<heat> no
<zid> worst 'stan' ever
<heat> my cat will do ur warranty claim
<gaxar77> What are you doing a mobile OS?
<zid> just think, you might get to interact with a samsung rep
<zid> that's like meeting your idol
<heat> lmao
<heat> does he play foooooooootball
<zid> ask him
<gaxar77> So, to build a cross compiler, I need to download all those libraries plus the source code?
<zid> "henlo, I would like to warranty, which football team do you supporb?"
<gaxar77> libraries and programs
<heat> if he's indian, it's man united
<heat> there's no need to even ask that silly question
<zid> no no that's what JAMES the character they play in the call center likes
<zid> you need to get to know the real imran
<zid> he prefers cricket.
<gaxar77> heat, are you talking to me?
<heat> isn't that more of a scammer thing?
<zid> no all indian call centers use fake names and stuff
<zid> sometimes one per business they're doing support for
<gaxar77> I need help to understand some stuff.
<gaxar77> Do I need to download everything in the table in that article to build a cross-compiler?
<gaxar77> Bison, Make, Flex, GMP, MPC, etc?
<Ermine> gaxar77: yes, you need those packages to build gcc and binutils from source
<Ermine> Follow the wiki instructions
<gaxar77> Okay.
<gaxar77> But... this is a bit overwhelming...
foudfou has quit [Remote host closed the connection]
<gaxar77> I have to build a cross-compiler, then write a multi-boot header and linker script, two things of which I have not a full understanding of how they function.
foudfou has joined #osdev
<gaxar77> Well, I'm not sure what the boot thing is called, but the tutorial teaches how to use grub.
<mrvn> you want multiboot for x86-64
<gaxar77> So these captions are package managers?
<gaxar77> Why?
<mrvn> because it's simple, it works and you skip having to write a bootloader
<gaxar77> ok
<gaxar77> Uhm, which package manager am I supposed to use to download these things?
<gaxar77> I have cygwin installed.
<mrvn> the browser of your choice
<gaxar77> I thought it would come with gcc but apparently it didn't.
<gaxar77> So I should just download it via a link?
slidercrank has joined #osdev
<gaxar77> The non-hyperlink ones are the ones already installed with the GNU right?
<gaxar77> Like Make, etc.
<mrvn> I have no idea what you have installed. just follow the wiki.
<gaxar77> Or is make not part of GN?
<Ermine> you need to install GNU make. See cygwin docs for instructions on how to install packages in it
<gaxar77> What about the MinGW package GUI?
<gaxar77> Can I use that instead of Cygwin?
<gaxar77> I already installed MinGW from sourceforge.
<gaxar77> separately from cygwin.
<gaxar77> I did this whole thing once already a long time ago (building a cross-compiler).. I don't remember installing each one of these packages.
<gaxar77> Ermine, can I install with cygwin even though I installed MinGW not with cygwin?
<Ermine> Ah, make can be installed from mingw
<gaxar77> What about the rest of the packages?
Arthuria has joined #osdev
Arthuria has quit [Killed (NickServ (GHOST command used by Guest684531))]
Arthuria has joined #osdev
<Ermine> The same way
<gaxar77> In MinGW?
<gaxar77> Ok.
<gaxar77> I can't find flex and bison.
<gaxar77> I installed GMP and MPC, if I did it correctly.
<gaxar77> I'm taking a break.
<mrvn> isn't there a pre-build cross compiler for windows?
<Ermine> It's better to install real linux. Or WSL at least.
<gaxar77> true
<gaxar77> mrvn, I don't know.
dude12312414 has quit [Quit: THE RAM IS TOO DAMN HIGH]
<gaxar77> Where do I download this pre-built cross compiler?
<gaxar77> I give up. I'm not getting the necessary answers to my questions.
<gaxar77> Ubuntu takes up too much space.
<gaxar77> So I uninstalled it.
<Ermine> Use alpine then
<mrvn> too much space? Go to McDonalds and buy a happy meal with USB stick.
<Ermine> Also yes, compared to Windows Ubuntu is not that big
xenos1984 has quit [Ping timeout: 246 seconds]
xenos1984 has joined #osdev
<gaxar77> I don't have that much space on my SSD.
<gaxar77> Most of it is occupied.
<gaxar77> I just installed Alpine.
Arthuria has quit [Killed (NickServ (GHOST command used by Guest684531))]
Arthuria has joined #osdev
<gaxar77> Well, thanks for all the help everyone, but I think I will work on something else for now.
gaxar77 has quit [Quit: Leaving]
joe9 has joined #osdev
<bslsk05> ​www.youtube.com: Queen - Another One Bites the Dust (Official Video) - YouTube
Arthuria has quit [Killed (NickServ (GHOST command used by Guest684531))]
Arthuria has joined #osdev
xenos1984 has quit [Ping timeout: 248 seconds]
<Ermine> First time I did bare bones, I had no clue how does it work, but running it was super satisfying
gog has joined #osdev
xenos1984 has joined #osdev
bnchs has joined #osdev
Vercas2 has joined #osdev
Vercas has quit [Ping timeout: 255 seconds]
Vercas2 is now known as Vercas
SGautam has joined #osdev
danilogondolfo has quit [Remote host closed the connection]
Arthuria has quit [Remote host closed the connection]
Arthuria has joined #osdev
nyah has joined #osdev
fedorafan has quit [Ping timeout: 246 seconds]
fedorafan has joined #osdev
demindiro has joined #osdev
<demindiro> gaxar77: Have you considered getting a HDD? You should be able to get a 1TB one new for less than €50.
<demindiro> Also, when using Linux it may be useful to use something like ZFS + enable compression on it, it can save a lot of space.
<demindiro> (If only today's commonly used filesystems all had transparent compression, so many exabytes of space would be saved, but alas)
<mrvn> it's not like porn compresses all that well
Arthuria has quit [Remote host closed the connection]
dude12312414 has joined #osdev
<geist> hah great queen song
demindiro has quit [Quit: Client closed]
nj0rd has quit [Quit: WeeChat 3.8]
nj0rd has joined #osdev
demindiro has joined #osdev
<gog> it does when you have the raws
<sakasama> ... or you can convert it with libcaca first.
<gog> heh i remember doing that with aalib
Arthuria has joined #osdev
kof123 has joined #osdev
gildasio1 has quit [Ping timeout: 255 seconds]
<geist> ugh. updated my firewall software yesterday and DNS forwarding has been generally unstable since
<geist> i think it's because the newer version of unbound (the DNS forwarder/resolver) is more picky about DNSSEC
<geist> seems to talk to the upstream servers (8.8.8.8, etc) and eventually decide they're bogus and gives up
<geist> but the eventually seems to be like 10 minutes to 2 hours, so it's hard to debug
<geist> disabled DNSSEC and it seems stable now
<geist> ah there we go, right as i said that it seems to have decided that 3 out of 4 servers are 'lame'
gildasio1 has joined #osdev
slidercrank has quit [Ping timeout: 246 seconds]
dude12312414 has quit [Remote host closed the connection]
<geist> hmm, it seems to just give up on an upstream server the instant it receives a no data from it
<geist> and that seems to just sometimse happen
<geist> once all the upstream servers do, it just stops doing dns
airplanemodes has joined #osdev
<heat> lmao
<heat> that's a bit broken
<geist> well, i think that's whats going on. waiting for now to see what happens
GeDaMo has quit [Quit: That's it, you people have stood in my way long enough! I'm going to clown college!]
<geist> i can watch it send outgoing queries with tcpdump on the firewall, and it still seems to talk to the ones its decided are lame, but seems to favor the one that it hasn't
<geist> well, even that. it seems to use the other ones... so i dunno
<geist> the logs aren't useful. though earlier it was bitching that it ca't do DNSSEC, but apoarently that's a known issue with dns forwarding (vs acting as a root dns resolver)
<geist> (this is all on pfsense firewall)
<heat> good luck debugging that lol
<heat> cold take: sun microsystems was great
<gog> better than fucking oracle
<geist> well it's all educational if nothing else
<geist> yah back in the day sun was fairly well respected
<geist> they were like HP of old. a good company run by decent people. or at least in as much as a company can be good, etc
<gog> :(
<geist> i always use oracle buying sun and weaponizing existing patents as a cautionary tale of having companies you work for file patents
<geist> you never know if you're gonna get dragged into some legal battle 10 years from now because of transfer of ownership of them
<heat> sounds like you want yourself a patent for small kernels
<heat> and new operating systems
<gog> patented gog
<gog> USPTO #42069
<geist> no i dont like to be on patents at all
<geist> or at lest software patents
<gog> software patents bad anyway
<heat> software bad
<gog> yes
<gog> best take
<geist> sometimes it's hard to get out of it, since it's generally effectively part of your employment agreement
Arthuria has quit [Killed (NickServ (GHOST command used by Guest684531))]
Arthuria has joined #osdev
fedorafansuper has joined #osdev
fedorafan has quit [Ping timeout: 252 seconds]
Arthuria has quit [Killed (NickServ (GHOST command used by Guest684531))]
Arthuria has joined #osdev
Turn_Left has joined #osdev
Turn_Left has quit [Max SendQ exceeded]
Left_Turn has quit [Ping timeout: 264 seconds]
Turn_Left has joined #osdev
<heat> geist, do you have a performance tab in your chrome settings?
<heat> they seem to have added a memory saving and power saving feature that's supposed to be in that tab but I don't have it lol
<geist> i dont know why but you may have to enable it in one of the settings panels
<geist> i dunno if it's a percentage of users thing, or if it's per platform or what
SGautam has quit [Quit: Connection closed for inactivity]
<geist> that features panel is what i'm talking about, it's under efficiency mode i think
<geist> youll have to google for it, i dont have it offhand
<geist> i think sometimes those features are turned on as an experiment by percentage of users over time. though this one isn't labelled experimental
<geist> perhaps it's per platform. what OS are you on?
<heat> i bing'd it just for a joke
<heat> but man BING SUCKS
<heat> arch linux using the flatpak
<gog> AB testing software dang
<geist> so my sample size is 2: my windows box had it enabled, my mac didn't
<heat> gog, reject AB testing, just 🚢 it
<geist> also it was added in 108 i think, so you need a chrome newer than that
<heat> yeah im on 110
foudfou has quit [Quit: Bye]
<heat> what is "it" here?
<heat> the performance settings tab?
<geist> the performance settings
<heat> ah
foudfou has joined #osdev
<geist> checking on my mint linux install
<geist> yah not enabled here by default either
<geist> sample size 3 now
<geist> so may be only enabled on windows for now
<heat> probably
<heat> thanks
<geist> probably needs the help more
<geist> sample size 4, both of my macs dont have it enabled
<heat> yeah linux doesn't need the help, the OOM killer and the Great Swapping take care of it
<geist> macs in particular swap and compress like you wouldn't believe
<geist> their VM is tuned to run hard into swap and it likes it
<heat> my linux can swap pretty effectively
<heat> but then you have some out of touch people like ingo molnar saying "who tf uses swap?"
<heat> which is a big :rolls_eyes: moment as someone with freakin 8GB of ram
<clever> i got units mixed up when configuring my latest laptop, it has 40mb of swap, lol
<geist> yah on most of my machines i have a rc.local that sets swappiness to 100
<heat> lol
<heat> oh are you an openrc/sysvinit person?
<geist> my VM machine is also fairly overcommittd and i'm relying on KSM and swap to deal with it
<geist> no, but even with systemd there's a rc.local thing you can enable
joe9 has quit [Quit: leaving]
<heat> ah
<heat> i think there's also a "native" sysctl unit you can use to set sysctls at boot
<clever> /etc/sysctl.conf and /etc/sysctl.d exist even on my ancient gentoo machine
<heat> also something cute: booting VMs does trigger khugepaged pretty effectively
<clever> /etc/init.d/sysctl deals with it on that system
<geist> oh i'm sure there's almost certainly some cleaner way to set that
<geist> rc.local FTW!
<kof123> re: macs "their VM is tuned to run hard into swap and it likes it" that is something freebsd used to claim (or "lore") that it was "graceful" under load versus other systems. i'm assuming mac is unrelated, but dont know how the details/etc.
* kof123 ducks to avoid heat projectile
<geist> agreed. back in the early 2000s i personally experienced this with my freebsd server and starting to piddle with trying to use linux as a replacement
<geist> linux was a toy at the time under load. if the disk or swap was getting wailed on the console would literally just stop responding
<geist> keypresses would take a minute,e tc
<heat> kof123, why
<clever> geist: ive once had a linux machine enter swap hell for 48 hours
<heat> i'm literally writing a patch for freebsd
<geist> freebsd would very gracefully handle it. that changed over time at least, but back then it was definitely not just lore
<heat> my hater game only applies to openbsd
<clever> i wasnt home, and it took 2 days just for ssh to respond enough to kill things
<gog> heater game
vdamewood has quit [Quit: Life beckons]
<geist> also gotta remember the early unix svrs and BSDs would run on machines that were very frequently overcommitted by a huge factor. swap was there so that you could let 100 people log into a machine that could only really deal with 10 at a time, because hardware was expensive
<bslsk05> ​www.youtube.com: Chappelle's Show - The Playa Haters' Ball (ft. Ice T and Patrice O'Neal) - YouTube
<geist> of course back then if your shell took a second to respond that was probably okay too. now folks would consider that insane
<geist> i remember i worked at a company very briefly in like 2001 that for some reason, even then, had a single 'massive' linux box that you were expected to shell in and do all your work
<geist> you h ad these little Dell thin client machines on your desk that used Exceed X windows app (on windows) to log into the single box, running remote X
<geist> it was strange. also predictably it ran like shit
CryptoDavid has joined #osdev
<geist> it was a 'big' box (ie, quad something processors, maybe GB ram, etc)
<geist> but even then it was already somewhat oversubscribed
<geist> i think they had some bug up their ass about security or whatnot and didn't want to deal with letting anything outside of their core network get access to software, etc
<geist> might have had some node locked licensed stuff too
<geist> even then it was a linux machine which was strange, considering the rest of the infrastructure was freebsd and solaris based. i dont remember why that was the case
<gog> mew
* geist pets gog
* gog prr
ilovethinking has joined #osdev
<ilovethinking> do these PT entries look normal
<geist> hello to you too ilovethinking
<geist> nice to drop by to have a nice chat
<ilovethinking> hi
<ilovethinking> sorry
<gog> looks fine
<gog> but without further context
<ilovethinking> im very fucking pissed because idk what to do
<gog> idk
<ilovethinking> okok so basically
<geist> yeah no real context but, i guess it's fie?
<ilovethinking> yeah
<ilovethinking> hm
<geist> or not? i honstly can't really tell you anything because no context
<ilovethinking> what context is needed
<ilovethinking> 0x1000 pages
<geist> like, is that x86? x86-64?
<ilovethinking> x86-64
<geist> what do you think that's supposed to look like?
<heat> i have little idea what that actually is
<heat> addresses?
<ilovethinking> pagetable entries
<geist> ie, what are you trying to map? is that the entry itself? at the leaf node?
<heat> then they do not look normal
<ilovethinking> shit
<heat> they look straight up wrong
foudfou has quit [Remote host closed the connection]
<geist> the ones with the leading fffs is definitely not a valid entry
<geist> since that'd be a crazy high physical address
foudfou has joined #osdev
<ilovethinking> i don't know wtf is wrong because ive been sitting here for the past something hours and i have no clue what is wrong with my vmm
<geist> well, the page table entries probably be wrong is a guess
<geist> also if you're in qemu, i suggest getting familiar with `info mem` and `info tlb`
<ilovethinking> wait hold up
<heat> actually are you sure these aren't addresses?
<geist> nice to debug this sort of thing, sinc eyou can see from qemus point of view what it thinks your page tables are doing
<ilovethinking> wait
<heat> none of these have any permission bits
* heat is confused
<ilovethinking> heat: i masked the bits off
<geist> they could be physical addresses just below 2GB. except the ones with leading fs
<ilovethinking> basically i think i figured out what's wrong
<heat> ok then they are straight up completely wrong
<geist> that sounds like some sort of signed vs unsigned math issue
<ilovethinking> im trying to map my framebuffer
<ilovethinking> and i have framebuffer base
<ilovethinking> and now that i think about it i am pretty sure the base is virtual
<geist> and then yeah there are some bits in the bottom 12 that have to be set to make a valid page table entry
<ilovethinking> and i have been treating it as physical
<geist> if nothing else bit 0 is the V (valid) bit
<geist> otherwise the cpu just straight up ignores it
terminalpusher has joined #osdev
<heat> the base you are given should be physical unless your bootloader is stupid
<heat> aren't you using limine? could very well be the 2nd
<ilovethinking> i am using limine
<ilovethinking> i am so fucking lost man
<ilovethinking> idk what's wrong
<ilovethinking> no clue
<geist> well, i think we just pointed out that those page table entries are not valid
<ilovethinking> yeah i know
<geist> if those are indeed page table entries
<ilovethinking> but my mapping function is okay
<geist> so that's a pretty good clue as to what to look into
<geist> i dont think you're lost at all, you have a known issue right in front of you
<geist> so now you need to debug it
<mrvn> ilovethinking: start by only mapping 16MB.
<mrvn> or maybe just a few K depending on how big your kernel is
<geist> honestly i dont think your mapping function is okay, if thats what its producing
<mrvn> get it down to a number of entries that you can validate each one by hand.
<geist> one of the first mistakes to make when doing bring up like this is to assume something is 'done' and works properly and never look back
<ilovethinking> mrvn: like the first 16mb?
<geist> never trust even your own code when just getting started
<geist> verify constantly, even stuff you think is good
<gog> it's funny this is the second time ilovethinking and i are working on the same topic :P
<zid> oh gog came back higgog
<gog> but i'm modifying a page thing that already works
<gog> hi zid
<ilovethinking> what's 16mb in hex?
<mrvn> ilovethinking: yes. Start with just 2MB. That's one page directory worth of 4k pages and should cover your kernel. + the FB.
<gog> 0x1000000
<ilovethinking> mrvn: what? wdym hold on
<geist> mrvn: probably not a good idea to confuse with higher level page sies at whatnot right now
<geist> i get the sense that they only have a very tenuous grasp of this stuff
<ilovethinking> what i do rn is map out text, data and rodata and then the fb
<ilovethinking> geist: idk how i wrote the vmm but i still don't understand shit about it
mctpyt has joined #osdev
<geist> then that is also a good task to take on right now
<zid> something tells me my vmware won't boot any more
<geist> since yo udont know how you did it do *not* assume it's correct at all
<mrvn> ilovethinking: then what would be 7f2e000? That's way to high for your kernel.
<geist> go back and verify and learn
<zid> considering I have less ram than the VM uses
<geist> mrvn: they said they were mapping the framebuffer. that may be where EFI/etc says it is
<bslsk05> ​gist.github.com: thing.c · GitHub
<geist> ilovethinking: is that where you got that 7fb... address from?
<ilovethinking> i got it from limine
<ilovethinking> check the gist
<ilovethinking> it;s 17 lines
<geist> okay. so clearly the vmap() routine isn't correct then, because the entry is not valid
<mrvn> ilovethinking: that jump from 7fb2e000 to fffffffb3f000 looks verry wrong.
<geist> so i'd start there and debug it
<geist> yes, i see two things: one the PRESENT flag isn't set, the second is the 0xffff
<geist> my guess is it's a sign extension problem with PAGEFLAG_NX
<ilovethinking> because i mask off the flags when printing
<geist> since that sets bit 63 and you're probably not using the right types in C
<geist> and thus it's converting it to a sign extended number. i've had this exact bug before (even recently! in the intel committed MMU code in LK)
<mrvn> ilovethinking: you should print the 4 page table indexes, the virtual address, the physical address and the flags.
<zid> My recent bug was pretty good wrt 'yea that'll do for a type'
<mrvn> geist: integer promotion is a pain for this
<geist> yep. gotta be really careful about the types you're using here. even how PAGEFLAG_PRESENT is defined may be an issue
<bslsk05> ​github.com: mem.c: Fix incorrect type in page_table() · zid/bootstrap@7127698 · GitHub
<geist> because it may be the synthesys of PAGEFLAG_PRESENT | PAGEFLAG_RW that is getting converted to a signed int64 with sign extension
<geist> depending on how those are defined
<zid> u32 big enough, but totally not the right type for the math >_<
<geist> yup
<geist> a mistake i've seen a lot of and try to get better is to define a lot of those bit things as (1 << 55), etc
<geist> because the '1' is an int, and depending on whatever it might not shift it right
<geist> so usually yo uneed to be more aggressive there with something like
<geist> (1UL << 55) or whatnot
<zid> 1ULLLL
<zid> technically we have those.. macros now?
<zid> UINT16_C(1) or whatever? I forget.
Left_Turn has joined #osdev
<geist> yah if you'e in C++ you can do it better
<heat> yeah there are horrible macros for it
<geist> yah, manually i think ((uint64_t)1 << 55) is sufficient
<zid> nod
mctpyt has quit [Ping timeout: 252 seconds]
<heat> I just assume unsigned long = word size
<zid> I always want 55ULL to work but I don't think it does
<geist> deends on if you want to make it arch neutral or not. if it's not arch neutral you can just know that UL == 64bit
<zid> x<<1ULL would be nicer than having to cast x
<geist> heat: yeah trouble is when it doesn't, etc. but yeah i general most sane architectures that's valid
<zid> so I just change the type of x ahead of type.. except that code where I clearlyf orot
<mrvn> What does "uxth" do?
Left_Turn has quit [Max SendQ exceeded]
<zid> ahead of time*
<zid> forgot*
<geist> mrvn: hmm, i forget. it's one of the bit twidding ones. you'll have to look it up
<mrvn> zid: uint16 is a bad type, that gets promoted to int
<geist> it's a hard instruction to grok
Left_Turn has joined #osdev
<geist> i thnk it's a bit shift and mask instruction that showed up in armv6 or v7
Turn_Left has quit [Ping timeout: 246 seconds]
<mrvn> heat: unsigned long is 32bit under windows
<geist> also linux i386
<ilovethinking> osdev might not be for me because i have no fucking clue what i'm doing
<mrvn> -linux
<ilovethinking> imma just go to sleep
<bslsk05> ​developer.arm.com: Documentation – Arm Developer
<ilovethinking> thanks geist and mrvn
<geist> zero extend halfword. simpler
ilovethinking has left #osdev [i probably locked my mac and i got disconnected]
<geist> well. that was a general waste of time
<mrvn> ahh, uxth is cast to uint16
<geist> to note: when pushed they just give up. shall remember that next time
<geist> yah there's a sxth i think to sign extend
<bslsk05> ​developer.arm.com: Documentation – Arm Developer
<geist> or at least i always have to look up every time i see it becaus ei forget what the immediates mean
<bslsk05> ​godbolt.org: Compiler Explorer
<gog> if they love thinking so much why didn't they do more
<geist> you go clang!
<mrvn> Such simple code but it's UB.
<geist> oh hang on: -march=armv7-a you want to add to clang
<gog> oh limine is a bootloader
<geist> i suspect it's compiling for some old ass default machine
<zid> gog: yea the first time he came in he said limine and I said "the fuck is that"
<mrvn> geist: no uxth on older arms I guess
<zid> I've now *heard* of it at least, no idea what it actually is though
<gog> he's copying code with no clue
ilovethinking has joined #osdev
<mrvn> geist: armv6 has it
<geist> that's correct. its v6 or v7 i thik
<geist> then the deafult is probably armv5 or something
<zid> gog: yea but my dog is better
<geist> never trust what godbolt gives you unless you specify the -march or -mcpu, as a general rule
<geist> for comparison purposes at least, because you dont know what the compiler defaults to
<geist> so its likely unfair
<mrvn> geist: It's intersting to see that gcc honors the integer promotion to int while clang ignores it
<geist> how you figure?
<mrvn> geist: smulbb vs. mul
<heat> geist, gcc and clang always default to march=generic no?
<gog> zid: your dog is cute
<geist> heat: not a good assumption, no
<geist> the target triple may set a different baseline
<heat> I've never seen any target that does that
<heat> does sound crazy
<geist> well, i can only tell you what i've seen. this was an example right there, where clang defaulted to something clearly less
<zid> gog: What do icelandic dogs pee up?
<mrvn> heat: if you build a compiler for i686-* then why should it generate 486 code?
<gog> moss-covered lava fields
<mrvn> zid: gras
<gog> same place i pee
<geist> the ARM default cpu/arch is always been all over the place. more so than even x86
<zid> do they go upside down so they're still cocking a leg
<geist> tis why i always override it every time i compile something cross platform like that
<mrvn> My ARM builds tend to use -march=armv6j4747-something that's RPi1
<ilovethinking> i think the flags are correct
<ilovethinking> present is set
<mrvn> ilovethinking: those last 2 are still totaly screwed
<ilovethinking> yeah...
<mrvn> ilovethinking: did you zero out the page tables before use? Do you zero the .bss?
<heat> we're telling you those addresses are bullcrap
<geist> heat: might not be, remember they said they're mapping the framebuffer
<geist> it could quite legitimately be up there
<heat> there's no framebuffer at ~-2GB
<mrvn> geist: fffffffff... is not a physical address
<heat> I can guarantee you that
<geist> sure there is, with an integrated GPU that's a perfectly reasonably place to be
<ilovethinking> lol even if i don't map the fb these addresses are stil there
<ilovethinking> funny
<geist> i've seen those kinda address
<mrvn> fdb are not the right flaghs either.
<heat> geist, at minus 2GB??
<heat> notice the *-*
<geist> sure, if you have a 2GB machine that would be there the 'stolen graphics memory' is
<heat> I mean -2GB in 64-bit space
<gog> it can't be there on x86
<geist> yes yes, i'm talking abot the 7fb... part
<gog> oh
<geist> the leading ffff is clearky shit
<gog> lmfao
<geist> as i pointe dout re: sign extension
<geist> that's almost certainly because of the FLAGS_NX becaus ei've personally had that problem before
<heat> geist, the 0x7fb... should be real memory
<geist> ilovethinking: suggestion: remove the NX part in your code and see what it does
<ilovethinking> alright
<geist> since yo dont strictly need it
<mrvn> Yeah, having the FB at near 2GB (7fb2b000) is possible.
<geist> then that'll probaby prove that it's a sign extension issue
<geist> ilovethinking: what hardware are you running on?
<heat> what BIOS are you using on qemu? assuming this is QEMU
<ilovethinking> how do i find that out
<heat> wait, no, better question
<heat> how much RAM are you giving it
<geist> if it's qemu yeah dosen't make much sense, but a cpu with an integraded gpu i'd totally expect the framebuffer to be <sizeof(ram) - some number of MB>
<zid> I have intel GMA atm, I've never had that before
<zid> is there anything fun I can do with it
<heat> probably not
<zid> rip
<geist> you can hold a door open
<heat> i mean, you can try and program it
<heat> that sounds fun
<ilovethinking> heat: 2g
<geist> is this real hardware or qemu?
<ilovethinking> qemu
<heat> yeah ok so the 0x7fb... addresses are fine
<heat> the other ones aren't
<geist> to double check, print those addresses before trying to map them
<heat> geist, actually, fun trivia: qemu has had like 3 different ways to place ram and PCI space on x86
<geist> huh interesting. someone at work told me to switch the order of my riscv CSR inline asm constraints from "rK" to "Kr" so it favors using an immediate first
<geist> which makes sens,e but apparently that's no longer a thing either compiler cares about
<geist> ie, the relative order of the constraints dont seem to be some sort of priority, or at least in that case
<heat> clang can't do "rm"
<ilovethinking> geist: guess what
<ilovethinking> you are right
<heat> and i'm still pissed
<geist> ilovethinking: yah i've personally had exactly that bug before
<geist> which is why it sticks out
<mrvn> geist: that part gets lost in the optimizer because of how the register allocation stuff works. Known issue.
<geist> mrvn: no my point is it actually works in ehtier direction
<geist> perfectly fine, exactly as i expect
<bslsk05> ​imgur.com: Imgur: The magic of the Internet
<ilovethinking> shit sorry
<ilovethinking> wait
<ilovethinking> there
<mrvn> geist: the priority gets lost. It's picks one or the other before it even knows what it needs.
<ilovethinking> the flags on the bottom 2 are still weird
<geist> now, those last teo r... yeah they're fucked up
<ilovethinking> ...
<geist> mrvn: yeah, though in this case it seems to WAI
<mrvn> geist: it's basically chance what you get. Sometimes the optimizing step hits and destroyes the choice.
* geist shrugs. good
<heat> the flags look fine
<geist> for the last two?
<heat> actually, maybe not
<heat> yeah probably not
<mrvn> ilovethinking: you should use "typedef struct { uint64_t phys; } PhysAddr;" for physical addresses.
<geist> i'm sure it's a coincidence, but ef5b is i think the magic in a ext2 partition?
<geist> something similar at least
<heat> close
<mrvn> geist: my guess is the page table isn't zeroed.
<heat> ef53
<geist> ah close
<geist> mrvn: yeah but that's odd because they should have overwritten it
<ilovethinking> how do i zero a page table
<heat> memset
<geist> unless they're orring in these values instead of writing over the entire entry
<ilovethinking> when do i do it
<mrvn> geist: not if the FB stops before the last 2 entries
<geist> can you pastebin your vmap routine?
<ilovethinking> yes
<ilovethinking> sec
<geist> mrvn: yeah that's true
<ilovethinking> does gist work
<geist> sure
<geist> pastebin to me is a standin for 'anything similar to it'
<mrvn> notice how the lower part of the address isn't contigous.
<geist> mrvn: yeah probably
<geist> possibly it's printing more than it needs to, and it wasn't zeroed
<mrvn> b2d -> b2e, b2c is skipped.
<bslsk05> ​gist.github.com: virtual.c · GitHub
<mrvn> So my guess would be the loop only goes up to 7fb2d001
<geist> okay, so it's completely overwriting it
<geist> so that's something
<geist> when printing these entries later, are you sure they're witin the same page table? ie, not the last 3 entries and then printing whatever garbage is after it?
<geist> also line 4 you want to expand that to 0xfff
<geist> not 0x1ff. since it's a 12 bit field you're masking off
frkazoid333 has joined #osdev
<geist> ah yeah that'd caus eyou to incorrectly compute the last index
<geist> so depending on where it lines up it might wrap around and not set what you think
<ilovethinking> let me fix that
<geist> also assumptoin: alloc() always returns 4K aligned addresses?
<ilovethinking> wait i should change all to fff? or the pt_offset one
<ilovethinking> geist: eys
<ilovethinking> yes
<geist> no. just that one. see why? width of those entries
<geist> the 0x1ff and 0xfff is a mask that corresponds to 9 and 12 bits
<ilovethinking> oH
<ilovethinking> sorry
<ilovethinking> yeah
<ilovethinking> still get the junk
terminalpusher has quit [Remote host closed the connection]
<geist> you dont have to apologize to me about it
<zid> 1ff is 9 bit, if you think about it kind of obvious, 0xFF being 8.
<geist> suggestion: print the entry as you set it
<geist> not after the fact
<geist> ie 'printf("setting entry at index w❌y:z to NNN");
<ilovethinking> that would be a lot of prints
<mrvn> another WTF opcode: https://godbolt.org/z/xeYh3vx9q What is itttt?
<bslsk05> ​godbolt.org: Compiler Explorer
<geist> and?
<ilovethinking> ok sure give me a sec
<geist> i mean it's one printf
<mrvn> The IT (If-Then) instruction makes a single following instruction (the IT block) conditional. The conditional instruction must be from a restricted set of 16-bit instructions.
<geist> it'll generate a lot, but then you can validate at the point at which it makes the change
<geist> look i'm not going to debug it for you, i'm just giving you pointers to how to track it down
<geist> so you'll quit just givnig up and feeling overwhemled
<heat> funny irc client there lol
<geist> yah i saw that too
<bnchs> hi osdev
<ilovethinking> got the same thing
<geist> mrvn: that's because it's generating thumb
<mrvn> why is gcc suddenly doing that?
<geist> because you didn't override it to not
<geist> so it's still defaulting to whatever the default mode is
<geist> which is probably v7 + thumb
<geist> also i should point out: why are you still screwing with arm32?
<geist> it's dead, jim
<geist> (with the obvious caveat that it totally isn't because cortex-m)
<mrvn> geist: because I have hardware that isn't 64bit.
<geist> yeah i know
<heat> it's dying jimothy
<geist> anyway it clearly decided to use a conditional code path like that, probably based on whatever cpu it thinks it's compiling for
<geist> i'd set an even pkaying field like.... `-mcpu=cortex-a7`
<geist> though it still decides to use thumb2 conditional on gcc and not on the other
<bslsk05> ​godbolt.org: Compiler Explorer
<mrvn> Still doing the conditional thing.
<geist> yep. isn't that what you want? that was one of the cool arm32 features you were complaoining that they removed in arm64
<mrvn> It's interesting to see the different codegen the compilers produce
<geist> yup.
craigo has quit [Ping timeout: 255 seconds]
<mrvn> I like the clang armv7-a code most.
<mrvn> best
<ilovethinking> im just using whatever the linker gives me
<ilovethinking> and it seems like data and rodata is junk given by the linker
<geist> cool, so now you jknow precisely when it tried to set the garbage
<geist> now you can reverse back to where it goes wrong
<mrvn> ilovethinking: your .rodata isn't aligned to a page
<moon-child> mrvn: oh yeah we
<moon-child> ll
<moon-child> I like you the best
<ilovethinking> geist: reverse?
<geist> never mind. read mrvn's comment
<ilovethinking> wdym now i know that my linker is fucked up
<mrvn> ilovethinking: when you map a physical address you should mask out the lower 12 bit, or assert they are 0.
<geist> well, no you're expecting it to be aligned in a way it isn't
<mrvn> ilovethinking: as is your linker script lacks the 4k alignment for .rodata
<mrvn> which by the way is broken in the wiki
<geist> all this aside i'd highly recommed not worrying about these fine grained permissions right now
bnchs has quit [Remote host closed the connection]
<geist> because that involves making sure the liker script, etc. you dont awnna mess with that right now
<geist> it's a PITA even if you know what you're doing
<geist> just map the whole thing as RWX and move on for now
<ilovethinking> physical_address &= ~(0xFFF);
<ilovethinking> that;s the way to align it?
<geist> you can go back and tighten it up later
<mrvn> ilovethinking: missing LLU
<geist> yeah but you dont want to fiddle with that right now, because you'll end up mapping two things in the same page with different permissions
<geist> that's just another bug unless the linker script is truly aligning them
<geist> the last page will be shared between two permissions
<heat> we did go through that already
<geist> yeah but i dont think they get it
<geist> so it's really not worth messing with right now
<heat> my OG mission for them was "make sure writing to a string doesn't crash"
<ilovethinking> mrvn: LLU?
<heat> for an easy response to "why do I need paging?"
<mrvn> 0xFFF is an integer but you are working with 64bit addresses
<ilovethinking> geist: yeah but then there is no explanation why i get a bogus address from the linker
<geist> well, okay... i mean, this is a complicated task (making sur the kernel is mapped right with tight permissions) for a newbie to be fiddling with
<geist> it's not bogus, it's just not aligned
<geist> most likely that's because the linker script isbn't aligning them
<geist> which requires fiddling with the linker script
<ilovethinking> so do i cast the 0xfff to llu
<ilovethinking> lol
<geist> that's a valid binary, just not set up the way your code is expect
<geist> LU yes
<mrvn> ilovethinking: 0xFFFLLU is 64bit
<mrvn> geist: pleae, no LU
<geist> only need LU in this case, because x86-64
<geist> no, LU is what it is on x86-64. not the same on all cpus
<ilovethinking> wait i didn;t know you could do the same with hex
<mrvn> geist: wrong, on windows LU is 32bit.
<geist> yes, but they're not on windows
<ilovethinking> like ik u can do 0LLU
<geist> whatever, either way.
<mrvn> geist: LLU is always 64bit on 64bit cpus.
<geist> too many cooks.
<geist> mrvn: yes i know this i'm trying to avoid a problem later on that i dont want to explain right now
<ilovethinking> yeah now they look more sane
<ilovethinking> hooray
<geist> if yo ualign that it might be just making a more subtle bug... sigh
<mrvn> it's just good practice to forget that "long" exists at all.
<geist> mrvn: it is 100% *not* but i really dont want to argue about it
<ilovethinking> so turns out my FB lives at 0xFD000003
<geist> you really need to fix the linker script first before you try to do this
<gog> 3 is the page flags
<ilovethinking> yes i know
<mrvn> ilovethinking: highly unlikely. Why isn't that aligned to at least 4 byte?
* geist steps out and lets the other cooks cook
<gog> nope
<gog> i'm nto cooking
<ilovethinking> mrvn: cuz i print the pageflag
<mrvn> ilovethinking: then say that
<mrvn> 2 is writable?
<ilovethinking> 1 << 2 is RW right
<ilovethinking> cuz i set PRESENT and RW
<mrvn> ilovethinking: 1 << 2 is 4 so it should be 0xFD000005
<ilovethinking> no sorry rw is 1 << 1
<mrvn> better
<ilovethinking> well it still pagefaults on print either way
<ilovethinking> lol
<mrvn> so does printf work now?
<ilovethinking> no error
<ilovethinking> no
<ilovethinking> it pagefaults when i try to print
<mrvn> does your printf try to use the FPU?
<mrvn> or sse
<mrvn> wheck where it fails
<mrvn> check
<ilovethinking> on the printf
<ilovethinking> printf("Virtual Memory Manager has been successfully initialized!\n");
<mrvn> Note that if your printf supports %f or %g then it will use the FPU on entry.
<mrvn> part of how VAARGS work
<ilovethinking> well my printf just uses the limine terminal
<mrvn> How so? Does limine have a callback to output text?
foudfou has quit [Remote host closed the connection]
<ilovethinking> yes
<ilovethinking> limine has very neat tools
<mrvn> Are you keeping limine mapped?
<bslsk05> ​github.com: limine/PROTOCOL.md at trunk · limine-bootloader/limine · GitHub
<ilovethinking> wdym "keeping limine mapped"
foudfou has joined #osdev
<mrvn> ilovethinking: if your page tables doesn't have entries to where limine has it's code and data with the right permissions then calling limine will fail.
<ilovethinking> yeah makes sense
<ilovethinking> anyways almost 2am
<ilovethinking> forgot the it's
<ilovethinking> lol
<ilovethinking> anyways
<mrvn> Which basically means you can't make your own page tables. You don't knwo where limine is unless it tells you somehow.
<mrvn> ilovethinking: look into outputing text to the serial port to debug stuff
<ilovethinking> yeah i really should
<ilovethinking> it's getting annoying with limine
wootehfoot has quit [Quit: Leaving]
ilovethinking has quit [Quit: i probably locked my mac and i got disconnected]