Bananymous/banan-os
1
1
Fork 1
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add back x86_32 support #5

Merged
Bananymous merged 46 commits from x86_32 into main 2024-04-02 23:36:29 +00:00
Conversation 0 Commits 46 Files Changed 76 +2596 -1815
Bananymous commented 2024-04-02 23:35:45 +00:00
Owner
Copy Link

32 bit has not been supported in a long time but it is now back and working. There are still some panics that don't exist in x86_64 probably partially since the whole scheduler was rewritten. Threads now use processor specific stack and forced interrupt while rescheduling. This fixes the old hacky context saving which depended on compiler stack usage and code generation. This approach did not work for 32 bit which uses stack more heavily.

32 bit has not been supported in a long time but it is now back and working. There are still some panics that don't exist in x86_64 probably partially since the whole scheduler was rewritten. Threads now use processor specific stack and forced interrupt while rescheduling. This fixes the old hacky context saving which depended on compiler stack usage and code generation. This approach did not work for 32 bit which uses stack more heavily.
Bananymous added 46 commits 2024-04-02 23:35:46 +00:00
BAN: Implement make_signed and make_unsigned 3c62be3f5d
Kenrel: Ext2 fix signed-unsigned comparisons 9ecd156622
Kernel: E1000 fix physical address on 32 bit target aa2e53c4f8
Kernel: Define KERNEL_OFFSET for i386 target 7ce0370b6a
Kernel: Remove old GDT, IDT and MMU code from i386 26ed689d30 
It will be easier to just rewrite them
Kernel: Move sys_fork trampolines to kernel/arch/ directory 45cea14165
Kernel: RDRAND on i386 is called twice with 32 bit register d920785256
Kernel: Scheduler/Thread add inline assembly for i386 c0dff5e203
Kernel: i386 has 14 indirect blocks in TmpInode instead of 2 2352c86048 
This allows keeping size of TmpInodeInfo as 128
Kernel: Only compile lai for x86_64 targets 0424082e7b 
I will be dropping lai entirely soon. Once I get to writing AML
interpreter.
Kernel: Allow Processor compilation for i386 targets 7f029b2713 
This is achieved by rewriting some inline assembly and changing
ProcessorID to be 32 bit value. For some reason if processor id
is 8 bits gcc runs out of 8 bit registers on i386.
Kernel: Specify template paramenters where they cannot be deduced b5aae34d86
Kernel: Fix AHCI device physical address writing on i386 target 3e4d410646
Kernel: Rename rsp->sp and rip->ip fe17958b9f 
This makes more sense if we support i386
Kernel: Add PageTable stub to progress linking 1ab2722850
Kernel: Remove old i386 spinlock code 0d92719433
Kernel: Implement signal trampoline for i386 26585bb1d9
Kernel: Make i386 thread tramplines crash 2dd0bfdece
Kernel: Implement dummy IDT and GDT for i386 097d9a6479
Kernel: Rewrite i386 boot code + linker script fc7e96fa66
Kernel: NVMe Queue max simultaneous commands is dependent on arch a9db4dd9a3 
This allows mask to be atomic on 32 bit architectures
Kernel/LibC: Write cxx abi with proper locking fbef90f7cb
Kernel: Process signal mask is now 2 32 bit values 93975fdc45 
This allows signal mask to be atomic on 32 bit target
Kernel: Replace i386 with i686 99e30a4d7d 
I don't really want to be working with i386 since it doesn't support
compare exchange instruction
Userspace: Update printf formats to compile on 32 bit 5dbe51a52e
Bootloader: Fix GDRT pointer size to 32 bits ca23360d07
BuildSystem: Allow running qemu with i686 target 84ef2161a1
Kernel: Fix boot code for x86_32 af050cc729 
Boot assembly now initializes processor and jumps to kernel
Kernel: Unify IDT and GDT code between x86_64 and x86_32 1943c3e7a1 
The code is pretty much the same, so there are just couple macros
differiating initialization.
Kernel: Write isr handler for x86_32 and cleanup x86_64 d7bf34ecd0
Kernel: Fix physical address size for x86_32 4d1f0e77f2 
Having 32 bit address space does not mean physical address space
is also only 32 bits...
Kernel: Implement PageTable for x86_32 c12d1e9bd9 
This is mostly copied from x86_64 with necessary modifications
Kernel: Implement Thread trampolines for x86_32 96babec22a
Kernel: Fix ISR error code formatting to 32 bit 7c2933aae1
Kernel: Rename thread stacks to more appropriate names 1b65f850ee
Kernel: Rewrite whole scheduler 5050047cef 
Current context saving was very hacky and dependant on compiler
behaviour that was not consistent. Now we always use iret for
context saving. This makes everything more clean.
Kernel: Rework scheduler/processor stacks. 2106a9e373
Bootloader allow installation when BANAN_ARCH=i686 b35cad0c2e
Bootloader: Implement loading for 32 bit ELF files. 6ac3681604
Kernel: Fix i686 page table global mappings d9b8391968
Kernel: Save segment registers on all interrupts on i686 4d70322eab
Kernel: Implement thread start trampoline for userspace fe386fa819 
This is needed on i686 to set segment registers.
Kernel: Set ss in i686 tss c95a271821
Kernel: Add offset for interrupt stack in Scheduler::yield() 9e073e9fa0 
This allows accessing (garbage) sp and ss in interrupt stack.
Kernel: Implement syscalls for i686 and cleanup x86_64 0dd74e3c9d 
This actually allows i686 to boot properly!
Kernel: Allow booting with multiple processors on i686 d2df55b1ac 
Also remove unnecessary ds clearing in x86_64.
Bananymous merged commit 731330c6b5 into main 2024-04-02 23:36:29 +00:00
Bananymous deleted branch x86_32 2024-04-02 23:36:29 +00:00
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
No items
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
Bananymous
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Bananymous/banan-os#5
Delete Branch "x86_32"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?