kmallllllllllllloc

improve type clarity

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/errno.h

@@ -30,6 +30,7 @@ enum {
 	EILLSEQ,   // Illegal byte sequence
 	EOVERFLOW, //
 	EALIGN,	   // Not aligned
+	EEXIST,	   // already exists
 };
 
 #endif

kernel/include/kmalloc.h

@@ -15,48 +15,90 @@ 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
 
-union per_cpu_freelist_cache {
+struct freelist {
-	void* freelists[KMALLOC_MAX_SIZE];
+	struct freelist* nextnode;
+};
+
+struct partialy_freelist {
+	struct freelist* freelist_nextnode;
+	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 {
-		void* kmalloc_8;
+		struct freelist* kmalloc_8;
-		void* kmalloc_16;
+		struct freelist* kmalloc_16;
-		void* kmalloc_32;
+		struct freelist* kmalloc_32;
-		void* kmalloc_64;
+		struct freelist* kmalloc_64;
-		void* kmalloc_128;
+		struct freelist* kmalloc_128;
-		void* kmalloc_256;
+		struct freelist* kmalloc_256;
-		void* kmalloc_512;
+		struct freelist* kmalloc_512;
-		void* kmalloc_1024;
+		struct freelist* kmalloc_1024;
-		void* kmalloc_2048;
+		struct freelist* kmalloc_2048;
-		void* kmalloc_4096;
+		struct freelist* kmalloc_4096;
-		void* kmalloc_8192;
+		struct freelist* kmalloc_8192;
-		void* kmalloc_16384;
+		struct freelist* kmalloc_16384;
-		void* kmalloc_32768;
+		struct freelist* kmalloc_32768;
-		void* kmalloc_65536;
+		struct freelist* kmalloc_65536;
-		void* kmalloc_131072;
+		struct freelist* kmalloc_131072;
 	};
 };
 
-union partial_list {
+union partialy_freelists_list {
-	void* freelists[KMALLOC_MAX_SIZE];
+	struct partialy_freelist* partialy_freelists[KMALLOC_MAX_SIZE];
 	struct {
-		void* kmalloc_8;
+		struct partialy_freelist* kmalloc_8;
-		void* kmalloc_16;
+		struct partialy_freelist* kmalloc_16;
-		void* kmalloc_32;
+		struct partialy_freelist* kmalloc_32;
-		void* kmalloc_64;
+		struct partialy_freelist* kmalloc_64;
-		void* kmalloc_128;
+		struct partialy_freelist* kmalloc_128;
-		void* kmalloc_256;
+		struct partialy_freelist* kmalloc_256;
-		void* kmalloc_512;
+		struct partialy_freelist* kmalloc_512;
-		void* kmalloc_1024;
+		struct partialy_freelist* kmalloc_1024;
-		void* kmalloc_2048;
+		struct partialy_freelist* kmalloc_2048;
-		void* kmalloc_4096;
+		struct partialy_freelist* kmalloc_4096;
-		void* kmalloc_8192;
+		struct partialy_freelist* kmalloc_8192;
-		void* kmalloc_16384;
+		struct partialy_freelist* kmalloc_16384;
-		void* kmalloc_32768;
+		struct partialy_freelist* kmalloc_32768;
-		void* kmalloc_65536;
+		struct partialy_freelist* kmalloc_65536;
-		void* kmalloc_131072;
+		struct partialy_freelist* kmalloc_131072;
 	};
 };
 

kernel/include/limits.h

@@ -0,0 +1,31 @@
+#ifndef _LIMIT_H
+#define _LIMIT_H
+
+#define CHAR_BIT 8
+
+#define INT8_WIDTH 8
+#define INT16_WIDTH 16
+#define INT32_WIDTH 32
+#define INT64_WIDTH 64
+
+#define UINT8_WIDTH 8
+#define UINT16_WIDTH 16
+#define UINT32_WIDTH 32
+#define UINT64_WIDTH 64
+
+#define INT8_MAX 127
+#define INT16_MAX 32767
+#define INT32_MAX 2147483647
+#define INT64_MAX 9223372036854775807LL
+
+#define INT8_MIN (-128)
+#define INT16_MIN (-32768)
+#define INT32_MIN (-2147483647 - 1)
+#define INT64_MIN (-9223372036854775807LL - 1LL)
+
+#define UINT8_MAX 255U
+#define UINT16_MAX 65535U
+#define UINT32_MAX 4294967295U
+#define UINT64_MAX 18446744073709551615ULL
+
+#endif

kernel/include/random.h

@@ -0,0 +1,8 @@
+#ifndef _RANDOM_H
+#define _RANDOM_H
+
+#include <types.h>
+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

@@ -0,0 +1,10 @@
+#ifndef _VMM_H
+#define _VMM_H
+#include "types.h"
+
+#define CANONICAL_MASK_KERNEL 0xFFFF800000000000
+#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

@@ -1,22 +1,28 @@
+#include "errno.h"
+#include "types.h"
 #include <bootloader.h>
-#include <kernel.h>
-#include <paging.h>
-#include <serial.h>
-#include <splash.h>
-#include <system.h>
 #include <gdt.h>
 #include <isr.h>
+#include <kernel.h>
+#include <paging.h>
 #include <pmm.h>
+#include <random.h>
+#include <serial.h>
+#include <splash.h>
+#include <string.h>
+#include <system.h>
+#include <vmm.h>
 
-__noreturn
+__noreturn void _start(void) {
-void _start(void) {
 	int err = 0;
 
 	// Serial console for debugging
-	if ((err = serial_init(COM1)) < 0) panic(NULL, "Failed to initialize Serial driver. (%d)", err);
+	if ((err = serial_init(COM1)) < 0)
+		panic(NULL, "Failed to initialize Serial driver. (%d)", err);
 
 	// Load structures from bootloader
-	if ((err = boot_init()) < 0) panic(NULL, "Failed to parse bootloader structures. (%d)", err);
+	if ((err = boot_init()) < 0)
+		panic(NULL, "Failed to parse bootloader structures. (%d)", err);
 
 	// CPUID thingy :p
 	uint32_t eax = 0;
@@ -24,18 +30,33 @@ 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);
+	if ((err = gdt_init()) < 0)
-	if ((err = isr_init()) < 0) panic(NULL, "Failed to initalize ISRs. (%d)", err);
+		panic(NULL, "Failed to initalize GDT. (%d)", err);
+	if ((err = isr_init()) < 0)
+		panic(NULL, "Failed to initalize ISRs. (%d)", err);
 
 	// Memory management
-	if ((err = pmm_init()) < 0) panic(NULL, "Failed to initalize Physical Memory Manager. (%d)", err);
+	if ((err = pmm_init()) < 0)
-	if ((err = pag_init()) < 0) panic(NULL, "Failed to initalize Paging. (%d)", err);
+		panic(NULL, "Failed to initalize Physical Memory Manager. (%d)",
+		      err);
+	if ((err = pag_init()) < 0)
+		panic(NULL, "Failed to initalize Paging. (%d)", err);
+
+
+	// test random number generation
+	printk("Random number generation test:\n");
+	for (uint8_t i = 0; i <= 16; i++) {
+		printk("%d: %lu\n", i, get_random_number());
+	}
 
 	// The most important function in the world.
 	splash();
 
 	// Patiently wait for things to do :)
-	while (1);
+	while (1)
+		;
 };
-

kernel/src/kmalloc.c

@@ -1,10 +1,17 @@
+#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
-union per_cpu_freelist_cache freelists_all_cpus[1];
+union freelists_list freelists_all_cpus[1];
+union partialy_freelists_list partial_free_lists[1];
 
 void* kcalloc(size_t count, size_t size) {
 	size_t size_to_alloc;
@@ -46,7 +53,125 @@ void* kmalloc(size_t size) {
 	    : "eax", "edx");
 
 	// freelists_all_cpus[core_id];
-	// TODO enable when we actualy know how many CPUs we have
+	// TODO disable when we actualy know how many CPUs we have
-	if () {
+	core_id = 0;
+
+	size_t alloc_index;
+	{
+		// TODO remove buildin
+		uint8_t leading_zeros = __builtin_clzll(size);
+		if (popcountll(size) == 1) {
+			alloc_index = leading_zeros;
+		} else {
+			alloc_index = sizeof(size) * CHAR_BIT - leading_zeros;
+		}
+		if (alloc_index > KMALLOC_MAX_SIZE) {
+			// NOT IMPLEMENTED
+			printk("Cannot allocate [%lu] bytes: \e[1mNOT "
+			       "IMPLEMENTED!\e[0m\n",
+			       size);
+			return NULL;
+		}
+		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) {
+		// fast path
+		if (selected_freelist->nextnode == NULL) {
+			freelists_all_cpus[core_id].freelists[alloc_index] =
+			    NULL;
+		} else {
+			freelists_all_cpus[core_id].freelists[alloc_index] =
+			    selected_freelist->nextnode;
+		}
+		return selected_freelist;
+
+	} else {
+		// slower path
+		if (partial_free_lists[core_id]
+			.partialy_freelists[alloc_index] != NULL) {
+			// partialy freelist available
+			struct partialy_freelist* selected_partialy_freelist =
+			    partial_free_lists[core_id]
+				.partialy_freelists[alloc_index];
+
+			if (selected_partialy_freelist->freelist_nextnode ==
+			    NULL) {
+				// only one free object available
+				if (selected_partialy_freelist
+					->partialy_freelist_nextnode != NULL) {
+					// 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
+					partial_free_lists[core_id]
+					    .partialy_freelists[alloc_index] =
+					    NULL;
+				}
+
+			} else {
+				// multiple free objects available
+				partial_free_lists[core_id]
+				    .partialy_freelists[alloc_index] =
+				    selected_partialy_freelist
+					->partialy_freelist_nextnode;
+
+				freelists_all_cpus[core_id]
+				    .freelists[alloc_index] =
+				    selected_partialy_freelist
+					->freelist_nextnode;
+			}
+			return selected_partialy_freelist;
+
+		} 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

@@ -0,0 +1,50 @@
+#include "kernel.h"
+#include <system.h>
+#include <types.h>
+
+uint64_t (*get_random_number)();
+
+// this starting value it truly random i have thrown a dice few times
+uint64_t LFSR = 61113534451;
+
+// RDSEED may be backdoored by NSA
+extern uint64_t rand_rdrand();
+asm(".intel_syntax noprefix\n"
+    ".global rand_rdrand\n"
+    "rand_rdrand:"
+    "   rdrand rax\n"
+    "   ret\n"
+    ".att_syntax prefix\n");
+
+uint64_t rand_software() {
+	uint64_t lsb = LFSR & 1;
+	LFSR >>= 1;
+	if (lsb) {
+		LFSR ^= 0x800000000000000DULL;
+	}
+	return LFSR;
+}
+
+void rand_init(bool rdrand, bool rdseed) {
+
+	// check for RDRAND
+	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 (rdseed) {
+			asm(".intel_syntax noprefix\n"
+			    "rdseed %0\n"
+			    ".att_syntax prefix\n"
+			    : "=r"(LFSR));
+		} else {
+			printk("[\e[\e[33mWARNING\e[0m] RDSEED not available "
+			       "falling back "
+			       "to developers dice trows\n");
+		}
+		get_random_number = rand_software;
+	}
+    printk("[\e[32mINFO\e[0m] RNG initialized!\n");
+}

kernel/src/system.c

@@ -1,6 +1,7 @@
+#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)
@@ -8,9 +9,7 @@ void cpuid(uint32_t* eax, uint32_t* ebx, uint32_t* ecx, uint32_t* edx) {
 			 : "memory");
 }
 
-void iowait(void) {
+void iowait(void) { outb(0x80, 0); }
-	outb(0x80, 0);
-}
 
 uint8_t inb(uint16_t port) {
 	uint8_t r;
@@ -42,11 +41,53 @@ void outl(uint16_t port, uint32_t data) {
 	__asm__ volatile("outl %1, %0" ::"dN"(port), "a"(data));
 }
 
-void cli(void) {
+void cli(void) { __asm__ volatile("cli"); }
-	__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) {
+void detect_processor_features() {
-	__asm__ volatile ("sti");
+	// 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

@@ -0,0 +1,94 @@
+#include "vmm.h"
+#include "kernel.h"
+#include <errno.h>
+#include <isr.h>
+#include <limits.h>
+#include <paging.h>
+#include <pmm.h>
+#include <random.h>
+#include <types.h>
+
+// allocates set number of pages at address provided
+// if address is NULL address is chosen internaly
+int vmm_alloc(void** vaddr, uint64_t count, uint64_t flags) {
+	if (flags == 0) {
+		flags = PTE_FLAG_PRESENT | PTE_FLAG_WRITE | PTE_FLAG_NOEXEC;
+	}
+	bool random_vaddr = *vaddr ? false : true;
+
+	uint64_t* pml4 = pag_get_pml4();
+	{
+		uint8_t limit_counter = 0;
+		while (true) {
+			if (random_vaddr) {
+			retry_selection:
+				*vaddr = (void*)ALIGN_DOWN(get_random_number(),
+							   PAGE_SIZE);
+
+				*vaddr = (void*)((uint64_t)(*vaddr) &
+						 CANONICAL_MASK_USER);
+				if (!(flags & PTE_FLAG_USER)) {
+					*vaddr = (void*)((uint64_t)(*vaddr) |
+							 CANONICAL_MASK_KERNEL);
+				}
+			} else {
+				if (flags & PTE_FLAG_USER) {
+					if ((void*)((uint64_t)(*vaddr) &
+						    CANONICAL_MASK_KERNEL)) {
+						return -EINVAL;
+					}
+				} else {
+					if (!((void*)((uint64_t)(*vaddr) &
+						      CANONICAL_MASK_KERNEL))) {
+						return -EINVAL;
+					}
+				}
+			}
+			for (uint64_t i = 0; i < count; i++) {
+				uint64_t* pte =
+				    pag_get_pte(pml4, *vaddr + (i * PAGE_SIZE));
+				if (!IS_ERR(pte) && (*pte & PTE_FLAG_PRESENT)) {
+					if (!random_vaddr) {
+						return -EEXIST;
+					} else {
+						if (limit_counter < UINT8_MAX) {
+							goto retry_selection;
+						} else {
+							return -EAGAIN;
+						}
+					}
+				}
+			}
+			break;
+		}
+	}
+
+	for (uint64_t i = 0; i < count; i++) {
+		void* paddr = pmm_alloc();
+		int pag_map_ret =
+		    pag_map(pml4, *vaddr + (i * PAGE_SIZE), paddr, flags);
+		if (pag_map_ret != SUCCESS) {
+			if (i != 0) {
+				for (i -= 1; i != UINT64_MAX; i--) {
+					if (pag_unmap(pml4,
+						      *vaddr +
+							  (i * PAGE_SIZE)) !=
+					    SUCCESS) {
+						panic(NULL,
+						      "failed to unmap "
+						      "previously mapped "
+						      "page. %s:%d",
+						      __FILE_NAME__, __LINE__);
+					}
+				}
+			}
+			return pag_map_ret;
+		}
+	}
+	return SUCCESS;
+}
+
+int vmm_free(void* vaddr) {
+	uint64_t* pml4 = pag_get_pml4();
+	return pag_unmap(pml4, vaddr);
+}