Makefile

@@ -1,8 +1,10 @@
 SOURCES = src/sched/scheduler.c src/sched/process.c src/mem/heap/kheap.c src/mem/paging/vmm.c src/mem/paging/paging.c src/mem/paging/pmm.c src/string/string.c src/io/kbd/ps2.c src/io/serial/serial.c src/io/term/printf.c src/io/term/term.c src/idt/idt.c src/mem/gdt/gdt.c src/mem/misc/utils.c src/time/timer.c src/kmain.c
 
+PROBLEMATIC_FLAGS=-Wno-unused-parameter -Wno-unused-variable
+
 build:
 	rm -f *.o
-	x86_64-elf-gcc -g -c -Isrc $(SOURCES) -Wall -Wextra -std=gnu99 -nostdlib -ffreestanding -fno-stack-protector -fno-omit-frame-pointer -fno-stack-check -fno-PIC -ffunction-sections -fdata-sections -mcmodel=kernel
+	x86_64-elf-gcc -g -c -Isrc $(SOURCES) $(PROBLEMATIC_FLAGS) -Wall -Wextra -std=gnu99 -nostdlib -ffreestanding -fno-stack-protector -fno-omit-frame-pointer -fno-stack-check -fno-PIC -ffunction-sections -fdata-sections -mcmodel=kernel
 	objcopy -O elf64-x86-64 -B i386 -I binary zap-light16.psf zap-light16.o
 	nasm -f elf64 src/idt/idt.S -o idt_stub.o
 	x86_64-elf-ld -o pepperk -T linker.ld *.o

README.md

@@ -11,18 +11,28 @@ To run it with QEMU, `make run`
 
 The basics that I'm targeting are:
 
+### Basic utility of what we call a "kernel"
+
 - Fix terminal driver (backspace issues, scrolling) OR add Flanterm or equivalent
-- Implement paging / see what Limine does at boot with memory management
 - Implement tasks, and task switching
 - Load an executable
-- Scheduler (round-robin using the PIT timer interrupt)
+- Filesystem (TAR for read-only initfs, then maybe read-write using FAT12/16/32 or easier fs) w/ VFS layer
-- Filesystem (TAR for read-only initfs, then maybe read-write using FAT12/16/32
 - Getting to userspace (syscalls)
 - Porting musl libc or equivalent
 
-In the future, maybe?
+### Scalability/maintenance/expansion features
 
-- SMP support
+- Global config header file
+- Documentation
+- SOME error handling in functions
+- Unit tests
+- Good error codes (like Linux kernel: ENOMEM, ENOENT, ...)
+- Make the panic function work within itself without dependencies + error message (and still get cpu context?)
+
+### Optional features
+
+In the future, maybe?
+- SMP support (Limine provides functionality to make this easier)
 - Parsing the ACPI tables and using them for something
 - Replacing the PIT timer with APIC
 

src/idt/idt.c

@@ -188,12 +188,14 @@ struct cpu_status_t* interrupt_dispatch(struct cpu_status_t* context)
         case 32:
             //DEBUG("Timer Interrupt");
             ticks++;
+
             if (ticks % SCHEDULER_QUANTUM == 0)
             {
                 CLEAR_INTERRUPTS;
                 scheduler_schedule();
                 SET_INTERRUPTS;
             }
+
             // Send an EOI so that we can continue having interrupts
             outb(0x20, 0x20);
             break;

src/kmain.c

@@ -67,7 +67,7 @@ void hcf()
 
 void panic(struct cpu_status_t* ctx)
 {
-	DEBUG("\x1b[38;5;231m\x1b[48;5;196mKernel panic!!!\x1b[0m at rip=%p\nSomething went horribly wrong! vect=0x%.2x errcode=0x%x\nrax=%p rbx=%p rcx=%p rdx=%p\nrsi=%p rdi=%p  r8=%p  r9=%p\nr10=%p r11=%p r12=%p r13=%p\nr14=%p r15=%p\n\nflags=%p\nstack at rbp=%p\nHalting...",
+	DEBUG("\x1b[38;5;231m\x1b[48;5;196mKernel panic!!!\x1b[0m at rip=%p\nSomething went horribly wrong! vect=0x%.2x errcode=0x%x\n\rrax=%p rbx=%p rcx=%p rdx=%p\n\rrsi=%p rdi=%p  r8=%p  r9=%p\n\rr10=%p r11=%p r12=%p r13=%p\n\rr14=%p r15=%p\n\n\rflags=%p\n\rstack at rbp=%p\n\rHalting...",
 		ctx->iret_rip,
 		ctx->vector_number, ctx->error_code, ctx->rax, ctx->rbx, ctx->rcx, ctx->rdx, ctx->rsi, ctx->rdi,
 		ctx->r8, ctx->r9, ctx->r10, ctx->r11, ctx->r12, ctx->r13, ctx->r14, ctx->r15, ctx->iret_flags,
@@ -126,11 +126,6 @@ void kmain()
 	
 	kheap_init();
 
-	void* ptr = kmalloc(10); DEBUG("(KMALLOC TEST) Allocated 10 bytes at 0x%p", ptr);
-	void* ptr2 = kmalloc(200); DEBUG("(KMALLOC TEST) Allocated 200 bytes at 0x%p", ptr2);
-	kfree(ptr);
-	void* ptr3 = kmalloc(5); DEBUG("(KMALLOC TEST) Allocated 5 bytes at 0x%p", ptr3);
-
 	vmm_init();
 	
 	struct process_t* pedicel = process_create("pedicel", (void*)pedicel_main, 0);

src/mem/heap/kheap.c

@@ -16,12 +16,24 @@ static uintptr_t end;
 // Kernel root table (level 4)
 extern uint64_t *kernel_pml4;
 
-static void kheap_map_page()
+static void kheap_grow(size_t size)
+{
+    size_t pages = ALIGN_UP(size + sizeof(struct heap_block_t), PAGE_SIZE) / PAGE_SIZE;
+    
+    if (pages == 0) pages = 1;
+    
+    for (size_t i = 0; i < pages; i++)
+    {
+        kheap_map_page();
+    }
+}
+
+void kheap_map_page()
 {
     uintptr_t phys = pmm_alloc();
     paging_map_page(kernel_pml4, end, phys, PTE_PRESENT | PTE_WRITABLE | PTE_NOEXEC);
     end += PAGE_SIZE;
-    DEBUG("Mapped first kheap page");
+    //DEBUG("Mapped first kheap page");
 }
 
 void kheap_init()
@@ -44,6 +56,7 @@ void* kmalloc(size_t size)
 {
     // No size, no memory allocated!
     if (!size) return NULL;
+    size = ALIGN(size);
 
     struct heap_block_t* curr = head;
 
@@ -53,17 +66,16 @@ void* kmalloc(size_t size)
         if (curr->free && curr->size >= size)
         {
             // We split the block if it is big enough
-            if (curr->size > size + sizeof(struct heap_block_t))
+            if (curr->size >= size + BLOCK_MIN_SIZE)
             {
-                struct heap_block_t* new_block = (struct heap_block_t*)((uintptr_t)curr + sizeof(struct heap_block_t) + size);
+                //struct heap_block_t* new_block = (struct heap_block_t*)((uintptr_t)curr + sizeof(struct heap_block_t) + size);
-                // We have to subtract the size of our block struct
+                struct heap_block_t* split = (struct heap_block_t*)((uintptr_t)curr + sizeof(*curr) + size);
-                new_block->size = curr->size - size - sizeof(struct heap_block_t);
-                new_block->free = true;
                 
-                // Then we chain up the block in the list
+                split->size = curr->size - size - sizeof(*curr);
-                new_block->next = curr->next;
+                split->free = true;
-                curr->next = new_block;
+                split->next = curr->next;
 
+                curr->next = split;
                 curr->size = size;
             }
 
@@ -75,14 +87,14 @@ void* kmalloc(size_t size)
         curr = curr->next;
     }
 
-    // If we're hear it means we didn't have enough memory
+    // If we're here it means we didn't have enough memory
     // for the block allocation. So we will allocate more..
     uintptr_t old_end = end;
-    kheap_map_page();
+    kheap_grow(size + sizeof(struct heap_block_t));
 
     struct heap_block_t* block = (struct heap_block_t*)old_end;
-    block->size = PAGE_SIZE - sizeof(struct heap_block_t);
+    block->size = ALIGN_UP(end - old_end - sizeof(struct heap_block_t), 16);
-    block->free = false;
+    block->free = true;
     block->next = NULL;
 
     // Put the block at the end of the list
@@ -93,7 +105,7 @@ void* kmalloc(size_t size)
     }
     curr->next = block;
 
-    return (void*)((uintptr_t)block + sizeof(struct heap_block_t));
+    return kmalloc(size);
 }
 
 void kfree(void* ptr)
@@ -104,12 +116,25 @@ void kfree(void* ptr)
     // Set it free!
     struct heap_block_t* block = (struct heap_block_t*)((uintptr_t)ptr - sizeof(struct heap_block_t));
     block->free = true;
+
+    // merge adjacent free blocks (coalescing)
+    struct heap_block_t* curr = head;
+    while (curr && curr->next)
+    {
+        if (curr->free && curr->next->free)
+        {
+            curr->size += sizeof(*curr) + curr->next->size;
+            curr->next = curr->next->next;
+            continue;
+        }
+        curr = curr->next;
+    }
 }
 
 // Should alloc enough for a stack (at least 64kb) to be used for a process.
 // Should return a pointer to top of the stack (as stack grows DOWNWARDS)
 void* kalloc_stack()
 {
-    void* ptr = kmalloc(PROCESS_STACK_SIZE);
+    uint8_t* ptr = kmalloc(PROCESS_STACK_SIZE);
-    return ptr+PROCESS_STACK_SIZE;
+    return ptr ? ptr+PROCESS_STACK_SIZE : NULL;
 }

src/mem/heap/kheap.h

@@ -23,5 +23,6 @@ void kheap_init();
 void* kmalloc(size_t size);
 void kfree(void* ptr);
 void* kalloc_stack();
+void kheap_map_page();
 
 #endif

src/mem/paging/paging.c

@@ -69,7 +69,7 @@ void paging_map_page(uint64_t* root_table, uint64_t virt, uint64_t phys, uint64_
         root_table[pml4_i] = VIRT_TO_PHYS(pdpt) | PTE_PRESENT | PTE_WRITABLE;
     }
     else {
-        pdpt = (uint64_t *)PHYS_TO_VIRT(root_table[pml4_i] & ~0xFFFULL);
+        pdpt = (uint64_t *)PHYS_TO_VIRT(root_table[pml4_i] & PTE_ADDR_MASK);
     }
 
     // PDPT: same here
@@ -79,7 +79,7 @@ void paging_map_page(uint64_t* root_table, uint64_t virt, uint64_t phys, uint64_
         pdpt[pdpt_i] = VIRT_TO_PHYS(pd) | PTE_PRESENT | PTE_WRITABLE;
     } 
     else {
-        pd = (uint64_t *)PHYS_TO_VIRT(pdpt[pdpt_i] & ~0xFFFULL);
+        pd = (uint64_t *)PHYS_TO_VIRT(pdpt[pdpt_i] & PTE_ADDR_MASK);
     }
 
     // PD: and here
@@ -89,7 +89,7 @@ void paging_map_page(uint64_t* root_table, uint64_t virt, uint64_t phys, uint64_
         pd[pd_i] = VIRT_TO_PHYS(pt) | PTE_PRESENT | PTE_WRITABLE;
     }
     else {
-        pt = (uint64_t *)PHYS_TO_VIRT(pd[pd_i] & ~0xFFFULL);
+        pt = (uint64_t *)PHYS_TO_VIRT(pd[pd_i] & PTE_ADDR_MASK);
     }
 
     // PT: finally, populate the page table entry
@@ -102,13 +102,16 @@ void paging_map_page(uint64_t* root_table, uint64_t virt, uint64_t phys, uint64_
 uint64_t kernel_phys_base;
 uint64_t kernel_virt_base;
 
-void paging_init(struct limine_kernel_address_response* kaddr, struct limine_framebuffer* fb)
+extern struct boot_context boot_ctx;
+
+void paging_init()
 {
     // We should map the kernel, GDT, IDT, stack, framebuffer.
     // Optionally we could map ACPI tables (we can find them in the Limine memmap)
 
-    kernel_phys_base = kaddr->physical_base;
+    kernel_phys_base = boot_ctx.kaddr->physical_base;
-    kernel_virt_base = kaddr->virtual_base;
+    kernel_virt_base = boot_ctx.kaddr->virtual_base;
+    struct limine_framebuffer* fb = boot_ctx.fb;
 
     DEBUG("Kernel lives at virt=0x%p phys=0x%p", kernel_virt_base, kernel_phys_base);
 
@@ -117,14 +120,38 @@ void paging_init(struct limine_kernel_address_response* kaddr, struct limine_fra
     // for debug
     uint64_t page_count = 0;
 
-    // HHDM map first 1 GB using given offset
+    // Find max physical address from limine memmap
-    for (uint64_t i=0; i<0x40000000; i += PAGE_SIZE)
+    uint64_t max_phys = 0;
+    for (uint64_t i=0; i<boot_ctx.mmap->entry_count; i++)
+    {
+        struct limine_memmap_entry* entry = boot_ctx.mmap->entries[i];
+        if (entry->length == 0)
+        {
+            continue;
+        }
+        uint64_t top = entry->base + entry->length;
+        if (top > max_phys)
+        {
+            max_phys = top;
+        }
+        //DEBUG("max_phys=0x%p", max_phys);
+    }
+
+    // 4GB
+    if (max_phys > 0x100000000)
+    {
+        DEBUG("WARNING: max_phys capped to 4GB (0x100000000) (from max_phys=%p)", max_phys);
+        max_phys = 0x100000000;
+    }
+
+    // HHDM map up to max_phys or 4GB, whichever is smaller, using given offset
+    for (uint64_t i=0; i<max_phys; i += PAGE_SIZE)
     {
         //paging_kmap_page(i+hhdm_off, i, PTE_WRITABLE);
-        paging_map_page(kernel_pml4, i+hhdm_off, i, PTE_WRITABLE);
+        paging_map_page(kernel_pml4, i+hhdm_off, i, PTE_WRITABLE | PTE_PRESENT);
         page_count++;
     }
-    DEBUG("Mapped %u pages for first 1GB (HHDM)", page_count); page_count = 0;
+    DEBUG("Mapped %u pages up to 0x%p (HHDM)", page_count, max_phys); page_count = 0;
 
     // Map the kernel (according to virt/phys_base given by Limine)
     // SOME DAY when we want a safer kernel we should map .text as Read/Exec

src/mem/paging/paging.h

@@ -6,8 +6,9 @@
 
 #include <stdint.h>
 #include <limine.h>
+#include "mem/heap/kheap.h"
 
-void paging_init(struct limine_kernel_address_response* kaddr, struct limine_framebuffer* fb);
+void paging_init();
 void paging_map_page(uint64_t* root_table, uint64_t virt, uint64_t phys, uint64_t flags);
 
 extern uint64_t hhdm_off;
@@ -15,10 +16,14 @@ extern uint64_t hhdm_off;
 #define PHYS_TO_VIRT(x) ((void*)((uintptr_t)(x) + hhdm_off))
 #define VIRT_TO_PHYS(x) ((uintptr_t)(x) - hhdm_off)
 
+#define PTE_ADDR_MASK 0x000FFFFFFFFFF000
 // Stole it
 #define ALIGN_UP(x, align)   (((x) + ((align) - 1)) & ~((align) - 1))
 #define ALIGN_DOWN(x, align) ((x) & ~((align) - 1))
-#define PAGE_ALIGN_DOWN(x) ((x) & ~0xFFFULL)
+#define PAGE_ALIGN_DOWN(x) ((x) & PTE_ADDR_MASK)
+
+#define ALIGN(size) ALIGN_UP(size, 16)
+#define BLOCK_MIN_SIZE (sizeof(struct heap_block_t) + 16)
 
 #define PML4_INDEX(x) (((x) >> 39) & 0x1FF)
 #define PDPT_INDEX(x) (((x) >> 30) & 0x1FF)

src/sched/process.h

@@ -11,7 +11,7 @@ typedef enum
 } status_t;
 
 #define PROCESS_NAME_MAX 64
-#define PROCESS_STACK_SIZE 0x100 // 64kb
+#define PROCESS_STACK_SIZE 0x10000 // 64kb
 
 struct process_t
 {