c80c7bbc6e
core+init: Instead of exporting an "arch" symbol for platform-specific code, use "usingnamespace".
...
This makes imports of platform-specific files way cleaner.
2025-02-22 15:49:14 +01:00
cc87bf18c9
core+init: Move syscall code from init to the system library + add a setEventQueue() syscall
2025-02-22 15:41:11 +01:00
ee7185bb83
core: Add an "event queue" to threads using a ring buffer updated by the kernel
2025-02-22 15:39:21 +01:00
daea80cea6
core: Fix alignment in lookupThreadById()
2025-02-22 15:38:23 +01:00
6a441b9b6d
core+system: Add a new "system library" with a shared memory ring buffer implementation
2025-02-22 15:37:22 +01:00
9d26113ca9
core: Separate thread.arch from other imports
2025-02-21 19:44:54 +01:00
c1956748f3
core: Add a struct named "platform.Registers" to cover different architectures
...
This replaces the architecture-independent usage of InterruptStackFrame, which is an x86_64 implementation thing.
2025-02-21 19:40:46 +01:00
7665e4cbb6
core: Add documentation to all scheduler functions + global thread list
2025-02-21 19:37:13 +01:00
50e7b6cca7
core: Check the stack is in user memory before changing it to kernel space
...
Otherwise, we're just adding the physical base to an address that's already in the higher half.
2025-02-21 18:47:24 +01:00
9acbc45a80
core: Make thread.enterTask() automatically set the thread state to Running
2025-02-21 18:40:29 +01:00
c722341d3f
core: Change naming for virtual memory structures to reflect x86_64 terms
2025-02-21 18:33:36 +01:00
5cb9689d10
core: Don't set the second thread argument to zero
...
This defeated the point of passing the page tables to init, since rsi would be set to zero anyway.
2025-02-20 23:01:28 +01:00
2ecd24df82
core: Add some annotations to managed thread variables
...
Some thread variables don't need to be set manually,
since they are updated whenever needed by some functions.
Add comments to make that clear.
2025-02-20 17:59:27 +01:00
8ea5df2e4e
core: Give init a pointer to its own page tables
2025-02-20 17:57:50 +01:00
0782ff5fe7
core/x86_64: Fix function naming style
2025-02-20 17:55:56 +01:00
b14c473a29
core: Make some functions in thread.zig private
2025-02-18 20:10:14 +01:00
0ee587a142
core: Add sleep support to the scheduler
2025-02-18 20:06:40 +01:00
e27ef97682
core: Set new threads' priority to 127
...
This is halfway through the priority range, so user processes can either lower or raise their priority from the default.
2025-02-18 19:30:46 +01:00
9cdf9aa71b
core/x86_64: Make in/out wrappers inline
2025-02-17 23:04:58 +01:00
c455e8ea0b
core: Fix a small error in thread.zig
2025-02-17 23:04:43 +01:00
f0d5b60b83
core: Add priority-based scheduling
2025-02-17 20:15:30 +01:00
0edcae7aae
core: Store the physical page directory alongside the virtual pointer to it
2025-02-16 13:20:02 +01:00
46a0c8c800
core: Add memory and scheduler related system calls
2025-02-16 11:58:12 +01:00
5eabb6bdf0
core: Add PIT timer code to preempt the scheduler
2025-02-16 11:58:01 +01:00
c86481f2e7
core: Allocate a stack for each core's idle thread
2025-02-16 11:57:27 +01:00
6a2323c16b
core: Add thread states
2025-02-16 11:57:04 +01:00
77cee5aeb0
core: Add global FrameAllocator with locking
2025-02-16 11:56:14 +01:00
2fd618173f
core: Add spinlock code
2025-02-16 11:52:45 +01:00
09fb2b6554
core: Fix a few bugs in threading code
...
g_threads is processor-specific now, so it's unused, and setPageDirectory() needs the physical address.
2025-02-15 22:58:34 +01:00
614ed80980
core: Add a simple ELF loader
...
We only need the ELF loader for init, and possibly a few other modules. The bulk of loading executables is going to be done by userspace, so we only need the basics here.
2025-02-15 22:46:28 +01:00
985d8fec0a
core: Add basic threading code
2025-02-15 22:45:32 +01:00
14578a0fee
core/x86_64: Set up a second stack in the TSS to handle system calls
2025-02-15 22:40:48 +01:00
654b228f10
core: Fix a couple of bugs in VMM and add a few utility functions
2025-02-15 22:23:00 +01:00
ffa9b78bf7
core/x86_64: Enable the NX bit before using it
...
Spent an embarrassing amount of time trying to figure out why there was a page fault with "reserved bits set". Turns out, I had set the NX bit without enabling it.
2025-02-15 22:22:04 +01:00
da5335410d
core/x86_64: Add general protection fault handler
2025-02-15 22:19:08 +01:00
c4bb857c4e
core: Reserve physical page 0
...
In the kernel's early stages we access physical memory identity-mapped, so it's not very good for us to allocate that frame and then try to use it.
2025-02-15 22:18:49 +01:00
baf3d39e98
core: Make findMultibootTags take a context argument
...
Will be useful later.
2025-02-15 22:18:05 +01:00
d8236bc8f7
core: Add a page fault handler
2025-02-15 15:18:40 +01:00
18ec598394
core: Enable interrupts in platformEndInit()
...
This way, each architecture can re-enable interrupts whenever necessary
2025-02-15 15:18:24 +01:00
1ec239147a
core: Add workaround for QEMU-specific triple-faults without KVM
...
Without KVM, qemu inserts an invalid 12-GiB reserved entry at the end of the physical address space, making the kernel reserve way more space than necessary for the page bitmap (and overwriting the page directory pages).
2025-02-15 15:16:50 +01:00
fce8a58cf6
Ready. Set. Go!
...
Microkernel development in Zig, should be fun! =]
2025-02-13 22:39:48 +01:00
