ff952fa2e4
Confirm alignment when calculating an offset to split at
2022-11-20 18:04:29 +01:00
29defdf54d
Add a debug function to dump heap usage
2022-11-20 16:33:54 +01:00
d54c882c63
Add new/delete-like functions but with Results
2022-11-20 15:48:08 +01:00
caf6d1563c
Kernel: Add a heap memory allocator
...
This is the first time I've actually dropped liballoc in favor of writing my own implementation.
Usually, malloc() and such looked so complicated that I preferred to let a nice external library do the job.
But I've decided to try writing my own allocator, and now we have heap memory without any 3rd party code!
2022-11-20 15:15:26 +01:00
cb88630d86
Bugfix: Make alloc_at return the start of the virtual memory range instead of the end
2022-11-20 15:11:53 +01:00
af267772f0
Rename MemoryManager::map to map_frames_at
...
To make it clearer that this API is only intented for mapping contiguous frames of physical memory (for MMIO perhaps?)
2022-11-19 22:32:48 +01:00
1b41a3e9cf
Introduce a few helpers to allocate/map/unmap several pages, which wrap around the MMU functionality
2022-11-19 22:28:45 +01:00
c886669d56
Break down operations to make protect_kernel_sections() more readable
2022-11-19 22:27:59 +01:00
d96cb73995
Rename physical "pages" to "frames"
...
AFAIK, this is the proper naming scheme. "Pages" are virtual, and "frames" physical.
2022-11-19 22:27:08 +01:00
847f2b4f4c
Replace more usage of stdint.h types with Luna's Types.h
2022-11-19 22:25:03 +01:00
047f445651
Kernel: Add helpers to validate memory addresses
2022-11-19 18:38:47 +01:00
70c6b78e35
Organize source tree
2022-11-19 17:59:49 +01:00
cf758fdfdc
Initial commit :)
2022-11-13 10:09:09 +01:00
37bb3273ce
Kernel: Add a C interface to the logging system
2022-11-12 18:12:18 +01:00
58fb422161
Kernel: Return a Result in MemoryManager and strdup_from_user
2022-11-12 11:56:40 +01:00
662afad426
Kernel: Add a Result class
2022-11-12 11:30:28 +01:00
1ea216dfd1
Kernel: Rename assert.h to ensure.h
...
The macro is named "ensure", the header file should reflect that.
2022-11-09 15:28:35 +01:00
98da473fdc
Run clang-format
2022-11-08 19:07:43 +01:00
6f72f92493
Kernel: Start preparing for recursive paging
2022-11-08 17:12:59 +01:00
b78f6f269d
Kernel: Omit displaying KernelHeap's buffer location in the log
2022-11-05 11:59:00 +01:00
8f2308c80d
Kernel: Implement mmap-able device files (regular files are not mmap-able yet)
2022-11-02 20:24:07 +01:00
c604c074a1
Kernel: Rename ASSERT() to ensure()
...
Doesn't get stripped on release builds, so it shouldn't be named assert.
2022-11-02 19:38:15 +01:00
e5cf32c7b3
Kernel: Introduce page ownership
...
Some pages, such as framebuffer pages, are not physical memory frames reserved for the current process.
Some, such as the framebuffer, may be shared between all processes.
Yet, on exit() or on exec(), a process frees all frames mapped into its address spaces.
And on fork(), it copies all data between frames. So how could we map framebuffers.
Simple: we use one of the bits in page table entries which are available to the OS, and mark whether that page is owned by the current process.
If it is owned, it will be:
- Freed on address space destruction
- Its data will be copied to a new page owned by the child process on fork()
If it is not owned, it will be:
- Left alone on address space destruction
- On fork(), the child's virtual page will be mapped to the same physical frame as the parent
This still needs a bit more work, such as keeping a reference of how many processes use a page to free it when all processes using it exit/exec.
This should be done for MAP_SHARED mappings, for example, since they are not permanent forever,
unlike the framebuffer for example.
2022-11-02 19:32:28 +01:00
22740e69bf
Kernel: Add support for the NX bit
...
Not support, actually. We now REQUIRE it.
2022-11-02 18:34:57 +01:00
712f4f5e51
KernelHeap: Add more debug logging
2022-10-20 18:49:33 +02:00
9d0dfbaedf
PMM: Log invalid frees
2022-10-20 18:49:12 +02:00
7e9744419e
sh: Improve it
2022-10-19 20:16:21 +02:00
b035795eb3
Kernel: Move errno.h and (k)assert.h out of the main include directory
...
This is mostly so IDEs don't pick them up instead of the userspace headers :)
2022-10-19 17:41:23 +02:00
755242719c
VMM: Add a few memsets
...
This seems to fix exec() making weird page tables!!
2022-10-19 07:56:08 +02:00
671f2a2de3
Kernel, libc: Implement waitpid()
...
FIXME: exec() is now doing weird page table stuff. But at least it works, no panics :)
2022-10-18 21:30:52 +02:00
34fc6996b0
UserHeap: allocate needed memory in request_virtual_pages
2022-10-17 18:52:11 +02:00
92634048fc
UserHeap: some nice improvements
2022-10-17 18:49:19 +02:00
64f5078494
Kernel, libc: Implement fork()
...
This time for real.
Also, add a new per-user-task virtual address allocator (UserHeap), so that mmap'ed pages are in user range and can be copied.
2022-10-17 18:43:35 +02:00
b334e1cd50
VMM: check if the entry is already present
2022-10-17 17:26:16 +02:00
682be58d97
AddressSpace: copy instead of linking
2022-10-17 17:24:33 +02:00
68403dc029
Kernel: Make AddressSpaces reference-counted
2022-10-15 17:40:33 +02:00
62d631f1b4
Kernel: Rename assert.h to kassert.h so IDEs pickup the kernel header instead of the userspace one
2022-10-15 12:56:48 +02:00
e0aa552fae
Kernel: Add a move() function
...
The standard C++ move() function.
2022-10-14 19:04:56 +02:00
f82fbbe60c
Make address spaces not cloned by default
2022-10-14 18:23:04 +02:00
1c3377fc98
Prepare for cloning address spaces, not there yet
2022-10-14 18:17:57 +02:00
e43777bd31
Apparently, it just works now.
2022-10-14 18:00:33 +02:00
97a8a4a4a2
Solve rebase
2022-10-14 17:39:24 +02:00
bb7887a29d
Add basic address space infrastructure
2022-10-14 17:34:29 +02:00
5abd8814e3
Kernel: Continue moving utilities to a separate subdirectory
2022-10-14 17:33:06 +02:00
e21b608af4
Utilities: Start moving utilities into specific headers in a utils/ subdirectory
2022-10-14 17:31:47 +02:00
26211bd49f
It (almost) works now
...
The only thing doing weird stuff is exec(), so that's commented out and throws ENOSYS right now.
But we have two user tasks running in parallel, isolated from each other!
2022-10-14 16:46:00 +02:00
24272c57ef
Almost there!
2022-10-13 21:55:51 +02:00
ee712432bd
Some more multiple address space stuff
...
This page-faults. This is because the memory where the ELF should be is all zeroes, which the CPU tries to interpret.
2022-10-13 21:14:39 +02:00
229b06c63b
Add basic address space infrastructure
2022-10-13 19:19:51 +02:00
83e6b8cd21
VMM: Fix naming convention
2022-10-13 18:42:53 +02:00
57482e4e93
VMM: Make it even nicer
2022-10-13 18:15:52 +02:00
b360307f41
VMM: Make it so much gooder
...
There are still some fixes to be made, but I think this is already way cleaner than before.
2022-10-13 17:58:13 +02:00
52944ba5d8
Kernel/VMM: Add support for larger pages to getFlags()
2022-10-12 20:05:27 +02:00
69a9f7f06a
Kernel: Move VMM from a class to a namespace
...
Also, rename the ugly Paging::VirtualMemoryManager name to just 'VMM'. Which is now used instead of kernelVMM.
2022-10-12 20:02:25 +02:00
3ac9fed23a
ELFLoader: Add check_elf_image() and check_elf_image_from_filesystem()
...
These two functions validate an image, without actually loading it. Very useful for exec!
2022-10-12 18:37:00 +02:00
bbd9f1d187
VMM: Add FIXME
2022-10-12 14:35:34 +02:00
15f340dbbe
VMM: Do not map recursively
2022-10-12 14:34:12 +02:00
eaea4603c6
MemoryManager: Fix logging when built with debug logging
...
That is to say, -DMM_DEBUG
2022-10-12 14:29:30 +02:00
4021cb3ac0
KernelHeap: do not crash the entire system
...
Previously, calling free_virtual_page(s) would cause an assertion fail if the address was not in the kernel heap range.
Now, we just return.
2022-10-12 14:28:48 +02:00
ad9c7af0bf
VMM: add FIXME
2022-10-12 14:27:47 +02:00
950f4ef608
VMM: Add support for larger pages
...
getPhysical() now stops at a larger page, unmap() can unmap a larger page, but map() just transforms it into a normal page.
getFlags() larger pages support is still pending.
At least now we don't page fault because we're trying to free a larger page.
2022-10-12 14:27:26 +02:00
525d567af6
VMM: When unmapping a page, invalidate the TLB for that page
2022-10-12 14:24:34 +02:00
c9ebe89899
Kernel/KernelHeap: Add MODULE #define
2022-10-12 14:04:41 +02:00
97eacc027e
Kernel: Use PAGE_SIZE in more places
2022-10-12 13:05:57 +02:00
d5f59b666a
Kernel/Memory: Use %p in printf
2022-10-08 18:21:02 +02:00
533b7c9e71
Refactor ACPI::get_rsdt_or_xsdt()
...
Much better now.
Also, remove a FIXME in PMM.cpp, since we do map the page bitmap to virtual memory now.
2022-10-08 15:00:42 +02:00
abcf1b6118
Define PAGE_SIZE as 4096 and use it everywhere instead of using 4096 as a magic number
2022-10-08 14:52:28 +02:00
028a1b1a3c
libc: Enable even more warnings
2022-10-07 18:19:06 +02:00
b7ee746da3
Kernel: Enable even more warnings
2022-10-07 18:10:20 +02:00
594d79143e
Kernel: enable -Wconversion
2022-10-06 17:13:34 +02:00
952d8fa294
Be more strict with warnings
2022-10-05 17:34:22 +02:00
8f310dd307
PMM: Map the page bitmap to kernel heap once the PMM and VMM are both initialized
2022-10-04 18:36:09 +02:00
e769cb45a7
MemoryManager: add get_page_at and get_pages_at functions to map any physical memory at a fixed virtual address (inverse of get_mapping and get_mappings)
2022-10-01 12:12:50 +02:00
522d74b65d
Add is_user_address and is_kernel_address functions
2022-09-29 19:30:42 +02:00
f1a7138568
User mode (with a few syscalls)
...
IT ACTUALLY WORKS NOW.
Why wasn't it working? Oh, because I was not setting already present page tables's permissions to user mode. Just a little bug. THAT I SPENT DAYS TRYING TO FIND
Anyways, it works now. Such a relief...
2022-09-25 20:35:05 +02:00
a078a11dde
Move PMM and VMM initialization into MemoryManager::init
2022-09-25 17:38:17 +02:00
704a23d0ad
Sanity checks
2022-09-24 23:09:39 +02:00
6bd3529f32
Port liballoc to get proper kmalloc/kcalloc/krealloc/kfree functions.
...
Yes, that's not completely-from-scratch.
But let's be honest, am I going to do everything from scratch? Probably not. I'm not making my own bootloader.
And making a proper smaller-than-4-KB allocator is not something I want to do.
Plus, liballoc works perfectly in this rewrite, seeing as the MM code actually works, instead of leaking all your poor memory
And liballoc_{lock, unlock} can be actually defined, since we have spinlocks here!
2022-09-24 22:40:59 +02:00
2511b7d7a1
Add optional (compile-opt-in) debug logging to the MemoryManager (which helped catch the bug that got fixed in the previous commit)
2022-09-24 22:38:07 +02:00
a0af8fa432
QUICK BUGFIX: Shift PDE.Address to the left by 12 bits in VMM::getPhysical, to return the actual physical address
2022-09-24 22:35:19 +02:00
3891d0c52e
Rename KernelMemoryManager to MemoryManager
...
Kind of a more catchy name, isn't it?
2022-09-24 21:45:13 +02:00
46b7dab847
Remove RangeAllocator and make a PMM namespace
2022-09-24 21:27:45 +02:00
c3e5251687
Some more userland and font failing
2022-09-23 16:41:43 +02:00
11dd165a8e
Scheduler: add proper support for user tasks (still faults though)
2022-09-22 08:14:04 +02:00
543fe2885f
Add support for user pages
2022-09-22 07:57:30 +02:00
db4c8e96b8
commit
2022-09-14 17:55:24 +02:00
0cb59ee17f
please work please
2022-09-11 08:23:32 +02:00
929962efc9
Fixed size counting in RangeAllocator
...
Which was displaying VERY wrong sizes beforehand. (Number of regions * 4096 instead of number of bytes)
2022-09-07 20:01:44 +02:00
b4484e951d
LOTS MORE LOGGING. which is great, of course.
2022-09-07 19:41:08 +02:00
1820286d8b
stuff
2022-09-07 10:33:22 +02:00
bd0a24097f
Add a KernelMemoryManager namespace to wrap PMM + VMM
2022-09-06 18:08:15 +02:00
0fbc68ca88
Finish VMM
2022-09-06 13:49:17 +02:00
8bff2ee0f5
Add a RangeAllocator which serves as a physical memory allocator
2022-09-06 13:21:54 +02:00
6a6be3292d
Memory::get_system and Memory::get_usable
2022-09-05 17:13:12 +02:00
1b727a66ea
Ready. Set. Go!
2022-09-05 16:13:51 +02:00
