Alex Suykov
b1d4b246c6
rework makefiles / thin lib.a
|
3 years ago | |
|---|---|---|
| .. | ||
| bits | 5 years ago | |
| Makefile | 3 years ago | |
| README | 7 years ago | |
| _start.s | 7 years ago | |
| errnames.h | 7 years ago | |
| sigreturn.s | 7 years ago | |
| syscall.h | 6 years ago | |
| syscall.s | 7 years ago | |
README
This implementation does int 0x80 style syscalls, which work on any
i?86 but may be slow on some.
The official recommended way of doing syscalls on unknown i?86 is to
locate and use syscall trampoline. Doing so in minibase however is tricky.
For all other architectures, syscall is something self-contained.
Trampoline means syscall code depends on a global, runtime-set variable.
Which in minibase context means either a wasted rw page, or some register
tricks -- and x86 is notoriously short on those.
It's not decided yet how to handle this problem well, and 32-bit x86 support
in minibase is more of a backward-compatibility kludge anyway, so it stays
like this for now. It should work on pretty much any x86 hardware, and whoever
may be running x86 in 2018 is clearly not concerned with performance anyway.
i?86 but may be slow on some.
The official recommended way of doing syscalls on unknown i?86 is to
locate and use syscall trampoline. Doing so in minibase however is tricky.
For all other architectures, syscall is something self-contained.
Trampoline means syscall code depends on a global, runtime-set variable.
Which in minibase context means either a wasted rw page, or some register
tricks -- and x86 is notoriously short on those.
It's not decided yet how to handle this problem well, and 32-bit x86 support
in minibase is more of a backward-compatibility kludge anyway, so it stays
like this for now. It should work on pretty much any x86 hardware, and whoever
may be running x86 in 2018 is clearly not concerned with performance anyway.