<adder>
So... I'm pretty sure this has to do with my lack of asm skills, as in putting the code in the right place, rather than any particular line being wrong. I got it to print out 8 and then a bunch of 13s, with qemu only reporting 13s. Hence, that eight is a disguised IRQ0.
frkzoid has joined #osdev
Turn_Left has quit [Read error: Connection reset by peer]
exit70 has quit [Read error: Connection reset by peer]
zxrom has quit [Quit: Leaving]
exit70 has joined #osdev
voidah has quit [Remote host closed the connection]
voidah has joined #osdev
gog has quit [Quit: byee]
<geist>
so the irq0 thing wasn't it?
<adder>
I'm just now looking at objdump -d of the code I had prior to using .section directives. It just outputs this:
<adder>
boot/interrupt.o: file format elf32-i386
<adder>
That's all.
<adder>
So then I added some .section directives, and now I'm seeing proper output.
<geist>
yah you probably want to at least put .text at the start
<geist>
what is the deal with the .rodata.isr_table stuff?
<geist>
that doesn't seem necessary
<geist>
just start the whole file in .text and then use the macro to stamp out multiple things
<adder>
Well, don't I want my isr table to be read only?
<geist>
mine does it in a kinda different way, but functionally its the same thing
<geist>
oh wait, the table is the addresses?
<adder>
Also, this is what I see now: check_exception old: 0xffffffff new 0xd
<geist>
ah i see. so yeah i did it a bit different
<adder>
While the printout shows: int_no: 8, then a bunch of int_no: 13
<bslsk05>
github.com: lk/arch/x86/32/exceptions.S at master · littlekernel/lk · GitHub
<linearcannon>
target remote /dev/ttyU0
<linearcannon>
wrong window
navi has quit [Quit: WeeChat 4.1.2]
<geist>
but anyway the key to *all* of this is *do not sti until you're ready*
<geist>
if you dont have the PIC masked off, dont do it
<geist>
remember the other exceptions will still happen even if IRQs are disabled
<geist>
ther'es no point at all enabling irqs if you dont have the interrupt controller set up
<geist>
absolutely no value in that
lentement has joined #osdev
lentement has quit [Ping timeout: 256 seconds]
heat_ has joined #osdev
heat has quit [Read error: Connection reset by peer]
<adder>
Oh. Somehow I thought that I need to sti in order to get anything at all.
<heat_>
geist, the .rodata.isr_table thing has an explicit purpose in my code, adder just copied it
<geist>
yeah makes sense once i grokked it
<geist>
it's building the table of addresses
<heat_>
yeah, it's a linker set of isr addresses
Matt|home has quit [Remote host closed the connection]
<heat_>
how do you do that isr_stub_len thing?
<heat_>
how does it work?
<geist>
see the second link at link 85
<heat_>
oh, you pad the isr stub. yeah i guess that works too
<heat_>
yeah that's what i was looking at
<geist>
yep, they're all the same length so that it's trivial to compute their lenghts
<adder>
Can confirm I'm not getting anything when I don't sti. I tried to trigger an exception by printing 1/0, and I'm getting 0x6? Doesn't look like Division Error (0x0).
<geist>
divide by zero is difficult to write unles syou do it in asm
<geist>
since the compiler usually senses what you're doing
<heat_>
i originally had trouble with any sort of similar approach because my isr stubs just get larger from isr 0x7f -> 0x80, because of the push imm8 -> push imm32
<heat_>
adder, 0x6 is the correct exception, gcc will see your UB div by 0 and replace it with ud2
<geist>
yah i just pushl (or pushq on x86-64) so it doesn't do it
<adder>
Alright then.
<heat_>
one of these days i need to restart my meltdown mitigation efforts
<heat_>
i tried to do it but ended up getting bored
Arthuria has quit [Ping timeout: 246 seconds]
<geist>
yeh it's pretty boring and sad
<heat_>
it's not as awful as it is hard to write in a way that doesn't make me very sad
npc has quit [Ping timeout: 255 seconds]
netbsduser has joined #osdev
netbsduser has quit [Ping timeout: 255 seconds]
netbsduser has joined #osdev
<heat_>
geist, beware: your isr_stub_len can be silently wrong if NUM_INT is more than 0x31
<geist>
oh yeah?
<heat_>
those pushl don't actually force the imm8 to imm32 afaik
<geist>
i think they do, because of the 'l' explicitly, but let me double check
<heat_>
l forces the length of the pushed data
<geist>
but anyway it doesn't matter right? because the highest number it pushes is 255
<heat_>
you kinda acknowledge it in the nop case, when replacing the pushl $0 by two nops
<geist>
NUM_INT would have to be > 255
<heat_>
it does matter, you can't encode imm8 with the high bit set, without it being interpreted as negative
<geist>
hmmm
<geist>
so how does this work...
<heat_>
so what the assembler does is for N < 0x80 it uses imm8, for >= 0x80 it uses imm32
<geist>
hmm, but this has worked for years and i know i use vectors > 0x80
<geist>
unless i dont...
<heat_>
your code happens to work because you only set up 0x31 (NUM_INT) in the loop, i think
<geist>
oh huh you're right
<geist>
i guess we expanded it in zircon to the full suite
<geist>
ah yes but the 64bit one does do the full thing
<geist>
so now i'm curious what's up
voidah has quit [Ping timeout: 260 seconds]
<geist>
okay good find. yeah it's totally broken on x86-64, i just must not ever use any vectors > 0x80
<geist>
i think the vector allocator starts low
<geist>
oh yeah duh, i forgot i never added SMP to the x86 port, so it doesn't use the lapic vectors, which i traditionally allocate to 250+
<heat_>
how about msi?
<heat_>
oh, i guess those use the low vectors
<geist>
it allocates from low to high
<heat_>
yeah
<geist>
so i've just never run a case where it allocated >= 0x80
<geist>
good catch. looking at the fuchsia code, someone had rewritten it to do the table thing you do
<heat_>
:)
<heat_>
fwiw i'm not sure if there's GAS syntax to force imm32? without explicitly encoding the insn with .byte
<adder>
Will immediates take on the smallest possible encoding by default?
<geist>
yeah
<adder>
e.g. 147 is 1 byte, while 300 is 2 bytes?
<heat_>
147 is 4 bytes because >= 0x80, 127 is 1 byte
<heat_>
there are no imm16 encodings
<adder>
(I was asking how as behaves, beyond geist's code. :))
<heat_>
well, 4 bytes of immediate, at least 1 byte for the opcode
<geist>
ah no it's fixed on x86-64, i had just added an .align 16to it
<geist>
so that each of them is 16 byte aligned, which is enough space
<heat_>
hah that also works i guess
<heat_>
probably not a horrible idea either, given the ISRs are jump targets?
<geist>
exactly, i now vaguely remember this
<geist>
ah no actually someone from intel pushed the change
<geist>
but they had just only fixed it on x86-64
<geist>
but what is silly (and i should have caught it) is they didn't remove the nops which are useless
<heat_>
i distinctly remember doug16k had his ISR stubs grouped by (speculated) hotness and the really hot ones were better aligned
<Mondenkind>
i meeeeaaaaaaannnn
<Mondenkind>
(it is appropriately doug16kish, though)
netbsduser has quit [Ping timeout: 260 seconds]
<heat_>
yep. i wonder if it has any measurable effect
<geist>
wonder what doug16k is up to
<geist>
also just verified. the l in push does have a measured effect, in that on an x86-32 compile it ensures it's just the default push
<bslsk05>
doug65536/funrender - Fast multithreaded SIMD software renderer with Z-buffering and perspective correct textures (0 forks/0 stargazers)
<geist>
but if you push b or w it adds the prefix, etc
sbalmos has quit [Ping timeout: 246 seconds]
<heat_>
question: does it matter more if you cacheline-align things or if you pack more things into one cache line
<geist>
so it's optional in that case, though i think i like to use the size to be explicit what's going on
<heat_>
say, for the ISR case, one could .align 64 every isr stub, or one could try to pack as many isr stubs in a single cache line
sbalmos has joined #osdev
<heat_>
in the function case, you're not really getting false sharing, so in that case it probably only matters if the CPU likes the "jump target" alignment
<geist>
yah i'd assume the point is to keep the sequence of instructions from crossing a cache line
Matt|home has joined #osdev
<geist>
the other thing that changes in that logic is the distance to the interrupt common changes, and eventually the assembler switches to a smaller jmp
<geist>
so i had some terrible hack to add 256 bytes between them to avoid it
<geist>
also fixing by just aligning things
<heat_>
the intel optimizing manual just says "direct branches that are mostly taken should have all their insn bytes in a single cache line, and nearer the end of the cache line. targets should be at or near the beginning of a cache line"
<heat_>
no explanation :/
<geist>
the beginning part kinda makes sense, but the nearer the end for the bytes is weird
<heat_>
yeah. here's another fun one "Do not put more than four branches in a 16-byte chunk"
<heat_>
which i assume is related to the CPU's branch prediction
<geist>
yeah that has something to do with the way the branch predictor indexes stuff
<geist>
various cpus over the years have had various limitations like that
justache is now known as justache_Test
justache_Test is now known as justache
exit70_ has joined #osdev
exit70 has quit [Read error: Connection reset by peer]
heat_ has quit [Ping timeout: 268 seconds]
netbsduser has joined #osdev
<geist>
there, updated and cleaned it up a bit
lentement has joined #osdev
netbsduser has quit [Ping timeout: 252 seconds]
\Test_User has quit [Quit: e]
\Test_User has joined #osdev
lentement has quit [Remote host closed the connection]
<sham1>
In papers we do write by way of "we", since it's the author(s) and the reader(s)
simpl_e has quit [Ping timeout: 255 seconds]
<zid>
sham1: yus
<zid>
> because it goes with 'Let us consider what happens blah blah'
<zid>
> Yea, it's not the royal we per se it's more like, 'The reader and I'
<zid>
Maybe we should use the germanic bi pronoun
<zid>
I've seen 'menyou' proposed as a modern dual pronoun, disgusting :P
<zid>
and mitu, yutu and themtu in certain pigeons
zetef has quit [Ping timeout: 260 seconds]
<nortti>
can you say which ones? those are reminiscent of austronesian creoles' (e.g. tok pisin, bislama) pronouns but not quite
<zid>
torres strait english
<zid>
thursday island
<pounce>
im honestly of the opinion that all singular pronouns in english should die
<pounce>
because it has been on this trajectory already for about a thousand years
<zid>
eww
<pounce>
:p
<zid>
I could understand removing articles
<zid>
but pronouns? fuck that
<pounce>
but we already have (see what i did here)
<zid>
japanese sounds so awks when they constantly call themselves their name
<pounce>
nobody does 'thou' anymore. it's only plural 'you'.
<zid>
"I'm"
<zid>
"it"
<zid>
you literally used two
<pounce>
and now singular they is really common
<pounce>
singular 'we' is really common
<pounce>
why keep the singular ones at all
<zid>
it's not singular we, it's audience-author we
<zid>
and royal we
<nikolar>
Singular they is wrong
<zid>
singular they is completely fine
<zid>
and has been in use for hundreds of years
<pounce>
it's been used for hundreds of years, it's fine. it outdates singular 'you'
<pounce>
if anything i think singular 'you' is more contraversial, but that argument is long gone
<nikolar>
Not as a replacement for normal third person pronouns
<zid>
How the fuck else are you supposed to say "Where did they go?"
<sham1>
We can use "we", we already use "you" and "they" is singular
<zid>
"Where did the person we were just both refering to go?"
<pounce>
sham1: exactly!
<pounce>
what english really needs is inclusive/exclusive 'we'
<pounce>
and then you can do away with 'I', just use inclusive 'we'
<zid>
us and them, we and they
<zid>
it has that
<zid>
already
<sham1>
We can literally just operate with epicene pronouns and context
<pounce>
yess
<zid>
I like the idea of the inclusive/exclusive 'us' though
<zid>
us (but not you) and us (including you)
<nortti>
reminds me, spoken finnish has kinda collapsed 1st person plural and passive together
<sham1>
Just leave "it" as the inanimate third-person and we'll be all set
<zid>
but then you lose animate/inanimate distinction
<nortti>
(meanwhile karelian collapsed 3rd person plural and passive)
<sham1>
I wonder where that comes from. Can't be Swedish influence since Swedish has a proper passive voice
<pounce>
'they' aready doesn't have an animate/inanimate distinction
<pounce>
we lost gendered plural third person pronouns
<zid>
singular they + it does
<zid>
it = inanimate, they = animate
<sham1>
We can work with inanimate it and animate they
<zid>
If you call someone an 'it' that's presently, rude af
<pounce>
yeah but you can call a bunch of radishes 'they'
<zid>
that's plural they
<pounce>
yeah but im saying we don't need a distinction
<zid>
so technically 'they' is 'singular animate' and 'animate/inanimate plural'
<sham1>
For example, in Finnish we have both the epicene third person singular "hän" and the non-human "se" (well, my dialect uses "se" for people and inanimate stuff and "hän" for pets and such, it's odd)
<sham1>
You can clearly have both
<zid>
You can pry my animate distinction from my cold dead hands
<zid>
sham1: yea I'm just calling that non-human vs human inanimate vs animate, there are.. levels of politeless and shit involved
<zid>
you can usually call a dog 'they'
<zid>
but not always
<sham1>
Yeah, it's just more accurate
<zid>
it is acceptable for a dog, but not a person
<sham1>
Could always retain "I" as the emphatic
<zid>
English should add pronouns
<pounce>
im calling it though, no singular in english in 600 years
<zid>
given how much people use shitty mics to talk to each other these days
<zid>
They've done tests, if you ask german speakers to say "The cat, the dog, those cats" etc and ask english speakers to say it, then play the audio to people with various levels of corruption
<zid>
the germans are more accurate
<sham1>
Because of grammatical gender, yeah
<zid>
because of the redundant info in the pronouns
<zid>
Optimizing a language for characters used is what robots do, not natural speakers of a language
<sham1>
You'd have to reintroduce grammatical gender into English. That'd probably go over like a lead balloon
<zid>
we like things like redundancy, funny rules etc
<zid>
sham1: I'll ask my dominatrix what she thinks.
<sham1>
I first wrote that as a lewd balloon, which is the exact opposite
<nortti>
sham1: hän vs se is generally a pretty weird case, since you have like three systems for that that are completely different: the old spoken system (hän is the same as se but used within quotation as a reflexive pronoun), the standardized system (hän is humans, se is non-humans), and the contemporary spoken system (you can use se in any place, using hän over se for ppl indicates politeness, formality, or
<nortti>
scorn, using hän over se for non-people indicates informality)
<sham1>
Mm
<sham1>
Of course a funny thing about hän is that it was borrowed originally from Swedish, and now they've borrowed it back again because they realised they needed an epicene third person singular pronoun
<acidx>
oh, osdev is langdev again <3
<sham1>
So it's a doublet of an existing word, where the etymology goes back to itself
<sham1>
OS nerds is an isomorphic with linguistics nerds
<nortti>
"a funny thing about hän is that it was borrowed originally from Swedish" no it wasn't? it's old uralic strata
<sham1>
s/ an//
<sham1>
Oh, hm
<sham1>
That it seems
<sham1>
I could have sworn that there was some relation to the pronoun "han"
<zid>
acidx: I'm re-adding all the gendered nouns back to english this week.
<nikolar>
That's a good idea zid
<pounce>
how many genders
<zid>
all three
<nikolar>
All of them
<acidx>
pounce: false, true, filenotfound
<zid>
Cat, dog, neuter
<pounce>
kiki, bouba
<zid>
or spayed
<acidx>
zid: as a speaker of a gendered language, I approve
<acidx>
please add diminutives and aumentatives too, kthx
<zid>
I already have diminutives you little pissant
<zid>
oh wrong type?
<acidx>
ah, but it's not as widely used as the ones in romance languages
<nortti>
masculine; feminine, fire, dangerous animals; other exceptional animals; other
<zid>
lol
<zid>
masuline; feminine; tomboy; catgirl; fish; mineral; vegetable; other imo
<sham1>
nortti: I still find the development of "hen" fun and how it was probably at least influenced by "hän"
<acidx>
little coffee doesn't have the same oomph as cafézinho <3
<zid>
Nip down to the bottleo then
<pounce>
ooh fun, hán is the gender neutral singular in icelandic
<pounce>
(han/hún/hán/það)
<nortti>
oh nvm, apparently the third noun class in dyirbal was "edible plants", exceptional animals go with feminine
<zid>
hmm what do we do about femboy fish though
<zid>
and what pronoun does a male whale use
<kof673>
i'm just ...gonna... throw https://0x0.st/s/QPEkZ1OsgjuQ-2VnHj7Bhw/Hh_J.mp4 zelda 3 snes tapir/flute boy into this conversation, and you can find all the uses of tapir in various stories, or magic flutes, or tapir used to make pure earth, or brought rain, or large magellen cloud ... on your own lol this is just to say...the necessity of ruthlessly cross-referencing :D
<kof673>
he went to death mountain looking for gold apparently lol
<pounce>
anything more than 4 genders is a classifier system imo
<pounce>
at that point you just have different descriptors for long/short/round/femboy/etc
* kof673
vanishes to zzzzzzzzzzzzz land
<zid>
time for kanji?
<zid>
"Has anybody seen pounce男"
lentement has joined #osdev
<pounce>
at least, if you're doing it on visible properties it's a classifier system, if you're doing it on made up magic it's a gender system
<sham1>
zid: straight to emoji
<gog>
✅unsubscribe from gender
<zid>
I like my women like I like my coffe, unsubscribed from gender
<sham1>
Remove gender, keep the sex and only when it's required
<GeDaMo>
Solution: stop talking about people all together :P
<zid>
can I talk about genitals still
<sham1>
Users are, sadly, people
Arthuria has joined #osdev
voidah has joined #osdev
<pounce>
you have users??
<gog>
sadly yes
<sham1>
^
<sham1>
It's the webdev grind
<gog>
but also the users do things that make our business work
<gog>
like, give us money
<gog>
or are intermediaries for getting us mony
<gog>
there's also a class of users that we pay when they sell things that have commission
<sham1>
You at least get money from that. We have a bunch of people in academia doing studies and schools that take part in the studies
<nortti>
< zid> "Has anybody seen pounce男" ← are we inventing egyptian-style determinatives again?
<sham1>
So our money doesn't even come from the users, but separate research grants and projects. It also means that the "incentives" of the users make things quite difficult
vinc_ has joined #osdev
voidah has quit [Ping timeout: 240 seconds]
nur has quit [Read error: Connection reset by peer]
vinc_ is now known as vinc
zetef has joined #osdev
GeDaMo has quit [Ping timeout: 252 seconds]
polezaivsani has left #osdev [leaving]
lentement has quit [Remote host closed the connection]
voidah has joined #osdev
gog has quit [Quit: Konversation terminated!]
heat_ has joined #osdev
GeDaMo has joined #osdev
lentement has joined #osdev
lentement has quit [Ping timeout: 256 seconds]
nur has joined #osdev
Turn_Left has joined #osdev
bauen1 has quit [Ping timeout: 246 seconds]
Left_Turn has quit [Ping timeout: 255 seconds]
gog has joined #osdev
xenos1984 has quit [Ping timeout: 260 seconds]
xenos1984 has joined #osdev
navi has quit [Read error: Connection reset by peer]
navi has joined #osdev
zetef has quit [Remote host closed the connection]
Arthuria has quit [Ping timeout: 268 seconds]
linearcannon has joined #osdev
linear_cannon has quit [Ping timeout: 272 seconds]
voidah has quit [Quit: Leaving]
voidah has joined #osdev
spareproject has quit [Remote host closed the connection]
bauen1 has joined #osdev
voidah has quit [Ping timeout: 256 seconds]
vinc has quit [Quit: Lost terminal]
navi has quit [Quit: WeeChat 4.1.2]
netbsduser has joined #osdev
lentement has joined #osdev
xenos1984 has quit [Ping timeout: 256 seconds]
spareproject has joined #osdev
lentement has quit [Ping timeout: 268 seconds]
xenos1984 has joined #osdev
\Test_User has quit [Ping timeout: 264 seconds]
PublicWiFi is now known as cryptic4
cryptic4 is now known as PublicWiFi
goliath has joined #osdev
carbonfiber has joined #osdev
gbowne1 has joined #osdev
\Test_User has joined #osdev
\Test_User has quit [Remote host closed the connection]
\Test_User has joined #osdev
zetef has joined #osdev
butchster has joined #osdev
gorgonical has joined #osdev
\Test_User has quit [Quit: .]
\Test_User has joined #osdev
<gorgonical>
Porting is so fatiguing
<gorgonical>
It's all the hard work of programming without any of the fun problem solving part
<nikolapdp>
what are you porting gorgonical
<gorgonical>
an iommu driver and then an image processor driver
<nikolapdp>
i had fun porting this client to pdp to be honst
<nikolapdp>
oh yeah that doesn't sound fun
<gorgonical>
especially since linux's iommu/dma system is whack
<gorgonical>
It has to be to be flexible enough, but holy moly dude
<nikolapdp>
lol
<gorgonical>
And unfortunately I have to port these things to get my phd so I don't have a choice lmao
GeDaMo has quit [Quit: That's it, you people have stood in my way long enough! I'm going to clown college!]
<gorgonical>
I whinge too much. Porting them is the proof that my research idea was right in the first place, so I literally argued to be allowed to do this to graduate about a year ago
<nikolapdp>
well good luck
<gorgonical>
thank you
<nikolapdp>
you're welcome
<gog>
meow
<nikolapdp>
hello gog
<gog>
hello nikolapdp
Vercas9 has joined #osdev
<nikolapdp>
how's it going
Arthuria has joined #osdev
Arthuria has quit [Killed (NickServ (GHOST command used by Guest684531))]
Arthuria has joined #osdev
<gog>
it's going how are you?
<nikolapdp>
had a 2 and a half hour meeting at work today :)
<gog>
:)
<nikolapdp>
we're moving from gitflow to trunk based dev if you're curious
<nikolapdp>
trade secrets right there ^
xenos1984 has quit [Ping timeout: 256 seconds]
<heat_>
gorgonical, porting is fun
<heat_>
i'm yet to port drm to onyx
<gorgonical>
it can be, I guess
<gorgonical>
I have fun porting to new archs because there's an opportunity to solve the same problem with new constraints often
<gorgonical>
Bootloaders and such
<gorgonical>
I'm not just porting in whole but trying to unwind a lot of abstractions in the ported code which means some degree of reverse engineering
<heat_>
right
<heat_>
if you don't really know what you're looking at, it's much harder
<gorgonical>
And turns out drilling through an entire abstraction layer that covers all sorts of architectures and edge cases is kind of complicated
<gorgonical>
the swiotlb code is wild in how much its infected random parts of the dma and iommu code