mycelium-hal's interrupts trait is stupid #337
Comments
hawkw
added
area: kernel
|
in particular i would still quite like to write an async serial port driver for the LPC UART (c.f. #222, from AN ENTIRE GODDAMN YEAR AGO). i have been violently reminded that there is just “no way” to get the IRQs i need for that, because i didn’t put them in the stupid fucking trait, and then i had written a stupid and worse serial port driver that flushed by busy looping anyway because im a fucking dumb bitch. so, in the fullness of time, i have now come to believe that the best solution is to travel back in time and kill myself before i can do that. since we can't do that, though, we should just make the IDT stuff public and consign the stupid fucking interrupts trait to the trash heap of history. |
|
OH GOD. if we want to do this we are probably gonna have to update our Mycelium dep,1 which means i have to figure out what i had done with the new timer API, and confront whatever problem stopped us from updating to use it the first time. hopefully the problem was just my own laziness, but i doubt it Footnotes
|
MnemOS would like us to provide lower-level access to the IDT in `hal-x86_64` so that they can work around the nice, high level abstractions we worked so hard to provide for them (see tosc-rs/mnemos#337). This PR capitulates to mnemOS' demands by making the IDT methods for registering raw ISR functions public. I hope mnemOS doesn't shoot themselves in the foot using this :)
MnemOS would like us to provide lower-level access to the IDT in `hal-x86_64` so that they can work around the nice, high level abstractions we worked so hard to provide for them (see tosc-rs/mnemos#337). This PR capitulates to mnemOS' demands by making the IDT methods for registering raw ISR functions public. I hope mnemOS doesn't shoot themselves in the foot using this :)
|
Upstream has finally decided to unblock us by permitting raw IDT access in hawkw/mycelium#487. I've started working on #343 to pick up the latest Mycelium deps; I've got the x86 kernel working with the new Mycelium timer API, but I seem to have busted something on the D1, which I'm presently investigating. |
This updates our dependencies on Mycelium crates to hawkw/mycelium@ba56bb4. This picks up the new `maitake` timer API, as well as the upstream change hawkw/mycelium#487 to allow raw access to the IDT in the Mycelium x86_64 HAL, which we need to do an async UART driver (see #337). While updating everything to use the new timer API, I also made some tweaks to our existing timer code, including implementing 32-bit timer rollover for the D1's TIMER0 (our current timestamp source on that platform). This was an ancient TODO item since basically forever. Fixes #337 --------- Co-authored-by: Anatol Ulrich <[email protected]>
in ancient days, i was foolish and succumbed to hubris, by thinking that a cross-platform abstraction over interrupts was a cool idea and could be fun to do. so, the mycelium HAL uses a trait with a bunch of methods to represent generic interrupt handlers
it turns out that this sucks and is bad and i hate it. in particular it makes writing handlers for platform specific interrupts really painful, and makes it hard to e.g. use all the different timers that exist on x86. we should rip that thing out of the x86 MnemOS platform and actually put stuff in the IDT ourselves, like normal people.
The text was updated successfully, but these errors were encountered: