kmallllllllllllloc

TODO

@@ -14,7 +14,7 @@ Low-Level Hardware & Interrupts
 [x] IDT
 [x] Basic IRQs (Internal/Legacy)
 [x] Proper fault handling (Exceptions, Page Faults, #GP)
-[ ] Check for features with CPUID (NX-bit, SSE, AVX)
+[-] Check for features with CPUID (NX-bit, SSE, AVX)
 [ ] APIC driver (Local APIC & I/O APIC)
 [ ] ACPI driver (Parsing RSDP/MADT for multi-core/APIC)
 [ ] HPET driver (High precision timing)

kernel/include/kmalloc.h

@@ -15,6 +15,32 @@ void* kzalloc(size_t size);
 // zeroed-out memory. Returns a pointer to allocated memory or NULL.
 void* kcalloc(size_t count, size_t size);
 
+// 9 bytes in 8 bytes
+#define TAG_SHIFT 56
+#define TAG_MASK ((uintptr_t)0xFF << TAG_SHIFT)
+#define SIGN_BIT ((uintptr_t)1 << 55)
+
+typedef struct {
+	uintptr_t bits;
+} tagged_ptr;
+
+static inline tagged_ptr tp_make(void* p, uint8_t tag) {
+	uintptr_t addr = (uintptr_t)p & ~TAG_MASK;
+	return (tagged_ptr){addr | ((uintptr_t)tag << TAG_SHIFT)};
+}
+
+static inline uint8_t tp_tag(tagged_ptr t) {
+	return (uint8_t)(t.bits >> TAG_SHIFT);
+}
+
+static inline void* tp_ptr(tagged_ptr t) {
+	/* Bits 48..55 were untouched and still equal the canonical sign.
+	   Copy bit 55 into bits 56..63. */
+	uintptr_t addr =
+	    (t.bits & SIGN_BIT) ? (t.bits | TAG_MASK) : (t.bits & ~TAG_MASK);
+	return (void*)addr;
+}
+
 // powers of 2
 #define KMALLOC_MAX_SIZE 17
 #define LOWEST_POWER 3
@@ -28,6 +54,12 @@ struct partialy_freelist {
 	struct partialy_freelist* partialy_freelist_nextnode;
 };
 
+// this is used when object is too small to hold the big one in this case we use
+// first object as pointer storage and dont allocate it
+struct partialy_freelist_short {
+	struct partialy_freelist_short* partialy_freelist_nextnode;
+};
+
 union freelists_list {
 	struct freelist* freelists[KMALLOC_MAX_SIZE];
 	struct {

kernel/include/random.h

@@ -2,7 +2,7 @@
 #define _RANDOM_H
 
 #include <types.h>
-void rand_init();
+void rand_init(bool rdrand, bool rdseed);
 
 extern uint64_t (*get_random_number)();
 #endif

kernel/include/system.h

@@ -9,6 +9,8 @@
  * CPUID
  */
 enum {
+    CPUID_FEAT_EBX_RDSEED       = 1 << 18,
+
     CPUID_FEAT_ECX_SSE3         = 1 << 0,
     CPUID_FEAT_ECX_PCLMUL       = 1 << 1,
     CPUID_FEAT_ECX_DTES64       = 1 << 2,
@@ -87,4 +89,6 @@ void     outl(uint16_t port, uint32_t data);
 void cli(void);
 void sti(void);
 
+extern uint8_t (*popcountll)(uint64_t num);
+void detect_processor_features();
 #endif

kernel/include/vmm.h

@@ -6,4 +6,5 @@
 #define CANONICAL_MASK_USER 0x00007FFFFFFFFFF000
 
 int vmm_alloc(void** vaddr, uint64_t count, uint64_t flags);
+int vmm_free(void* vaddr);
 #endif

kernel/src/kmain.c

@@ -30,6 +30,9 @@ __noreturn void _start(void) {
 	cpuid(&eax, (void*)&s[0], (void*)&s[8], (void*)&s[4]);
 	pr_info("[kernel] CPUID_GETVENDORID: %s\n", s);
 
+	// enable processor extensions
+	detect_processor_features();
+
 	// System init
 	if ((err = gdt_init()) < 0)
 		panic(NULL, "Failed to initalize GDT. (%d)", err);
@@ -43,8 +46,7 @@ __noreturn void _start(void) {
 	if ((err = pag_init()) < 0)
 		panic(NULL, "Failed to initalize Paging. (%d)", err);
 
-	// random number generation
-	rand_init();
+
 	// test random number generation
 	printk("Random number generation test:\n");
 	for (uint8_t i = 0; i <= 16; i++) {

kernel/src/kmalloc.c

@@ -1,8 +1,12 @@
+#include "errno.h"
 #include "kernel.h"
+#include "paging.h"
 #include "types.h"
+#include "vmm.h"
 #include <kmalloc.h>
 #include <limits.h>
 #include <string.h>
+#include <system.h>
 // this file contains SLUB inspired HEAP allocator
 
 // TODO this will be an allocated array later when we have APIC MADT
@@ -54,8 +58,9 @@ void* kmalloc(size_t size) {
 
 	size_t alloc_index;
 	{
+		// TODO remove buildin
 		uint8_t leading_zeros = __builtin_clzll(size);
-		if (/*__builtin_popcountll(size) == */1) {
+		if (popcountll(size) == 1) {
 			alloc_index = leading_zeros;
 		} else {
 			alloc_index = sizeof(size) * CHAR_BIT - leading_zeros;
@@ -69,6 +74,8 @@ void* kmalloc(size_t size) {
 		}
 		alloc_index -= LOWEST_POWER;
 	}
+	size_t object_size = (1 << (alloc_index + LOWEST_POWER));
+
 	struct freelist* selected_freelist =
 	    freelists_all_cpus[core_id].freelists[alloc_index];
 	if (selected_freelist != NULL) {
@@ -96,14 +103,16 @@ void* kmalloc(size_t size) {
 				// only one free object available
 				if (selected_partialy_freelist
 					->partialy_freelist_nextnode != NULL) {
-					// next partialy free list available
+					// next partialy free list
+					// available
 					partial_free_lists[core_id]
 					    .partialy_freelists[alloc_index] =
 					    selected_partialy_freelist
 						->partialy_freelist_nextnode;
 
 				} else {
-					// no other partialy free list available
+					// no other partialy free list
+					// available
 					partial_free_lists[core_id]
 					    .partialy_freelists[alloc_index] =
 					    NULL;
@@ -125,6 +134,44 @@ void* kmalloc(size_t size) {
 
 		} else {
 			// slowest path
+
+			uint8_t page_count =
+			    object_size < PAGE_SIZE
+				? 1
+				: (object_size * 2) / PAGE_SIZE;
+
+			void** alloc_addr = 0;
+			{
+				int ret;
+				ret = vmm_alloc((void**)&alloc_addr, page_count,
+						0);
+				if (ret != SUCCESS) {
+					return NULL;
+				}
+			}
+
+			// construct the linked list
+			// start at offset if object too small for partial free
+			// list pointer
+			for (size_t i =
+				 object_size < sizeof(struct partialy_freelist)
+				     ? object_size
+				     : 0;
+			     i < page_count * PAGE_SIZE; i += object_size) {
+				if ((i + 1) * object_size >=
+				    page_count * PAGE_SIZE) {
+					alloc_addr[i * object_size] = NULL;
+				} else {
+					alloc_addr[i * object_size] =
+					    alloc_addr +
+					    ((i + 1) * object_size);
+				}
+			}
+
+			freelists_all_cpus[core_id].freelists[alloc_index] =
+			    (struct freelist*)(alloc_addr + object_size);
+			return alloc_addr;
 		}
 	}
 }
+void* kfree(void* p) {}

kernel/src/random.c

@@ -25,18 +25,16 @@ uint64_t rand_software() {
 	return LFSR;
 }
 
-void rand_init() {
-	uint32_t eax = 1, ebx = 0, ecx = 0, edx = 0;
-	cpuid(&eax, &ebx, &ecx, &edx);
+void rand_init(bool rdrand, bool rdseed) {
 
 	// check for RDRAND
-	if ((ecx >> 30) & 1) {
+	if (rdrand) {
 		get_random_number = rand_rdrand;
 	} else {
 		printk("[\e[33mWARNING\e[0m] RDRAND not available falling back "
 		       "to software\n");
 		// check for RDSEED
-		if ((ebx >> 18) & 1) {
+		if (rdseed) {
 			asm(".intel_syntax noprefix\n"
 			    "rdseed %0\n"
 			    ".att_syntax prefix\n"

kernel/src/system.c

@@ -1,52 +1,93 @@
+#include "types.h"
 #include <kernel.h>
+#include <random.h>
 #include <system.h>
-
 void cpuid(uint32_t* eax, uint32_t* ebx, uint32_t* ecx, uint32_t* edx) {
-	__asm__ volatile ("cpuid"
-		: "=a"(*eax), "=b"(*ebx), "=c"(*ecx), "=d"(*edx)
-		: "a"(*eax)
-		: "memory");
+	__asm__ volatile("cpuid"
+			 : "=a"(*eax), "=b"(*ebx), "=c"(*ecx), "=d"(*edx)
+			 : "a"(*eax)
+			 : "memory");
 }
 
-void iowait(void) {
-	outb(0x80, 0);
-}
+void iowait(void) { outb(0x80, 0); }
 
 uint8_t inb(uint16_t port) {
 	uint8_t r;
-	__asm__ volatile ("inb %1, %0" : "=a"(r) : "dN"(port));
+	__asm__ volatile("inb %1, %0" : "=a"(r) : "dN"(port));
 	return r;
 }
 
 void outb(uint16_t port, uint8_t data) {
-	__asm__ volatile ("outb %1, %0" :: "dN"(port), "a"(data));
+	__asm__ volatile("outb %1, %0" ::"dN"(port), "a"(data));
 }
 
 uint16_t inw(uint16_t port) {
 	uint16_t r;
-	__asm__ volatile ("inw %1, %0" : "=a"(r) : "dN"(port));
+	__asm__ volatile("inw %1, %0" : "=a"(r) : "dN"(port));
 	return r;
 }
 
 void outw(uint16_t port, uint16_t data) {
-	__asm__ volatile ("outw %1, %0" :: "dN"(port), "a"(data));
+	__asm__ volatile("outw %1, %0" ::"dN"(port), "a"(data));
 }
 
 uint32_t inl(uint16_t port) {
 	uint32_t r;
-	__asm__ volatile ("inl %1, %0" : "=a"(r) : "dN"(port));
+	__asm__ volatile("inl %1, %0" : "=a"(r) : "dN"(port));
 	return r;
 }
 
 void outl(uint16_t port, uint32_t data) {
-	__asm__ volatile ("outl %1, %0" :: "dN"(port), "a"(data));
+	__asm__ volatile("outl %1, %0" ::"dN"(port), "a"(data));
 }
 
-void cli(void) {
-	__asm__ volatile ("cli");
+void cli(void) { __asm__ volatile("cli"); }
+
+void sti(void) { __asm__ volatile("sti"); }
+
+uint8_t (*popcountll)(uint64_t num);
+
+extern uint8_t popcountll_native(uint64_t num);
+asm(".intel_syntax noprefix\n"
+    ".global popcountll_native\n"
+    "popcountll_native:"
+    "   popcnt rax, rdi\n"
+    "   ret\n"
+    ".att_syntax prefix\n");
+
+// I have no idea how this works but it is apperently called SWAR
+uint8_t popcountll_software(uint64_t num) {
+	num = num - ((num >> 1) & 0x5555555555555555ULL);
+	num = (num & 0x3333333333333333ULL) +
+	      ((num >> 2) & 0x3333333333333333ULL);
+	num = (num + (num >> 4)) & 0x0f0f0f0f0f0f0f0fULL;
+	return (int)((num * 0x0101010101010101ULL) >> 56);
 }
 
-void sti(void) {
-	__asm__ volatile ("sti");
-}
+void detect_processor_features() {
+	// leaf 1
+	uint32_t eax = 1, ebx = 0, ecx = 0, edx = 0;
+	cpuid(&eax, &ebx, &ecx, &edx);
 
+	bool rdrand = false;
+	if (ecx & CPUID_FEAT_ECX_RDRAND) {
+		rdrand = true;
+	}
+
+	if (ecx & CPUID_FEAT_ECX_POPCNT) {
+		popcountll = popcountll_native;
+	} else {
+		popcountll = popcountll_software;
+	}
+
+	// leaf 7
+	eax = 7, ebx = 0, ecx = 0, edx = 0;
+	cpuid(&eax, &ebx, &ecx, &edx);
+	bool rdseed = false;
+	if (ebx & CPUID_FEAT_EBX_RDSEED) {
+		rdseed = true;
+	}
+
+	// init rng
+	rand_init(rdrand, rdseed);
+}

kernel/src/vmm.c

@@ -87,3 +87,8 @@ int vmm_alloc(void** vaddr, uint64_t count, uint64_t flags) {
 	}
 	return SUCCESS;
 }
+
+int vmm_free(void* vaddr) {
+	uint64_t* pml4 = pag_get_pml4();
+	return pag_unmap(pml4, vaddr);
+}