mld done

include/merror.h

@@ -3,6 +3,7 @@
 #define __MERROR_H__
 
 #include <errno.h>
+#include <string.h>
 
 /* Error codes
  */

makefile.mk

@@ -13,7 +13,7 @@ H_SRC = $(shell find $(SRC) $(INCLUDE) -type f -name "*.h")
 C_SRC = $(shell find $(SRC) -type f -name "*.c")
 C_OBJ = $(patsubst %.c,$(BIN)/%.o,$(C_SRC))
 
-.PHONY: clean build run
+.PHONY: clean build run test
 
 build: $(BIN)/$(OUT)
 
@@ -27,7 +27,8 @@ run: build
 test:
 	make -C ../test $(OUT)
 	mkdir -p ../fuzz
-	afl-fuzz -i ../test/$(OUT) -o ../fuzz -- $(BIN)/$(OUT) @@
+	rm -fr ../fuzz/$(OUT)
+	afl-fuzz -i ../test/$(OUT) -o ../fuzz -M $(OUT) -- $(BIN)/$(OUT) @@
 
 $(C_OBJ): $(BIN)/%.o : %.c
 	@mkdir -p $(@D)

masm/asm.c

@@ -315,7 +315,7 @@ static int assemble_phdr(struct assembler *assembler, Elf32_Phdr **res,
 	Elf32_Phdr *phdr = malloc(sizeof(Elf32_Phdr) *
 				  assembler->sectab.len);
 	if (phdr == NULL) {
-		ERROR("cannot alloc");
+		PERROR("cannot alloc");
 		return M_ERROR;;
 	}
 
@@ -354,6 +354,11 @@ static int assemble_shdr(struct assembler *assembler, Elf32_Shdr **res,
 
 	Elf32_Shdr *shdr = malloc(sizeof(Elf32_Shdr) * max_entries);
 
+	if (shdr == NULL) {
+		PERROR("cannot alloc");
+		return M_ERROR;
+	}
+
 	size_t str_off;
 	uint32_t count = 0;
 
@@ -593,7 +598,7 @@ static int write_file(struct assembler *assembler, Elf32_Ehdr *ehdr,
 
 	if (out == NULL)
 	{
-		ERROR("cannot write '%s'", path);
+		PERROR("cannot write '%s'", path);
 		return M_ERROR;
 	}
 

masm/lex.c

@@ -309,7 +309,7 @@ int lexer_init(const char *path, struct lexer *lexer)
 {
 	FILE *file = fopen(path, "r");
 	if (file == NULL) {
-		ERROR("cannot read '%s'", path);
+		PERROR("cannot read '%s'", path);
 		return M_ERROR;
 	}
 	lexer->file = file;

masm/parse.c

@@ -1,6 +1,5 @@
 #include <mlimits.h>
 #include <merror.h>
-#include <netinet/in.h>
 #include <stdint.h>
 #include <stdio.h>
 #include <string.h>
@@ -10,6 +9,9 @@
 #include "lex.h"
 #include "mips.h"
 
+#define B16(x) (x)
+#define B32(x) (x)
+
 static int next_token(struct parser *parser, struct token *tok)
 {
 	if (parser->peek.type != TOK_EOF) {
@@ -475,7 +477,7 @@ off:
 
 	if (assert_token(parser, TOK_NUMBER, &token))
 		return M_ERROR;
-	fi->data.offset = htons(token.number);
+	fi->data.immd = B16(token.number);
 
 	if (peek_token(parser, &token))
 		return M_ERROR;
@@ -618,7 +620,7 @@ static int parse_instruction_i(struct parser *parser,
 
 	if (token.number >= MAX16)
 		return M_ERROR;
-	ins->data.immd = htons(token.number);
+	ins->data.immd = B16(token.number);
 
 	return M_SUCCESS;
 }
@@ -634,12 +636,12 @@ static int parse_instruction_offset(struct parser *parser,
 	case MAX26:
 		if (get_offset_26(parser, &n, ref))
 			return M_ERROR;
-		ins->data.offs26 = htonl(n);
+		ins->data.offs26 = B32(n);
 		break;
 	case MAX16:
 		if (get_offset(parser, &n, ref))
 			return M_ERROR;
-		ins->data.offset = htons(n);
+		ins->data.offset = B16(n);
 		break;
 	default:
 		return M_ERROR;
@@ -682,7 +684,7 @@ static int parse_instruction_branch_equal(struct parser *parser,
 	int32_t off;
 	if (get_offset(parser, &off, ref))
 		return M_ERROR;
-	ins->data.offset = htons(off);
+	ins->data.offset = B16(off);
 
 	return M_SUCCESS;
 }
@@ -703,7 +705,7 @@ static int parse_instruction_branch(struct parser *parser,
 
 	if (get_offset(parser, &n, ref))
 		return M_ERROR;
-	ins->data.offset = htons(n);
+	ins->data.offset = B16(n);
 
 	return M_SUCCESS;
 }
@@ -742,7 +744,7 @@ static int parse_instruction_sli(struct parser *parser,
 
 	if (assert_token(parser, TOK_NUMBER, &token) || token.number > MAX16)
 		return M_ERROR;
-	ins->data.immd = htons(token.number);
+	ins->data.immd = B16(token.number);
 
 	return M_SUCCESS;
 }
@@ -818,7 +820,7 @@ static int parse_pseudo_li(struct parser *parser, struct ins_expr *expr)
 	expr->ins[0] = mips_instructions[MIPS_INS_ORI];
 	expr->ins[0].data.rt = reg;
 	expr->ins[0].data.rs = MIPS_REG_ZERO;
-	expr->ins[0].data.immd = htons(immd);
+	expr->ins[0].data.immd = B16(immd);
 	expr->ref[0].type = R_MIPS_NONE;
 
 	return M_SUCCESS;
@@ -859,11 +861,11 @@ static int parse_pseudo_la(struct parser *parser, struct ins_expr *expr)
 	expr->ins_len = 2;
 	expr->ins[0] = mips_instructions[MIPS_INS_LUI];
 	expr->ins[0].data.rt = reg;
-	expr->ins[0].data.immd = htons(hi);
+	expr->ins[0].data.immd = B16(hi);
 	expr->ins[1] = mips_instructions[MIPS_INS_ORI];
 	expr->ins[1].data.rt = reg;
 	expr->ins[1].data.rs = MIPS_REG_ZERO;
-	expr->ins[1].data.immd = htons(lo);
+	expr->ins[1].data.immd = B16(lo);
 
 	return M_SUCCESS;
 }

masm/reltab.c

@@ -13,7 +13,7 @@ int reltab_init(struct relocation_table *reltab)
 	reltab->data = malloc(sizeof(Elf32_Rela) * RELTAB_INIT_LEN);
 
 	if (reltab->data == NULL) {
-		ERROR("cannot alloc");
+		PERROR("cannot alloc");
 		return M_ERROR;
 	}
 
@@ -33,7 +33,7 @@ int reltab_push(struct relocation_table *reltab, const Elf32_Rela rel)
 			 * reltab->size);
 
 		if (reltab->data == NULL) {
-			ERROR("cannot realloc");
+			PERROR("cannot realloc");
 			return M_ERROR;
 		}
 	}

masm/sectab.c

@@ -16,7 +16,7 @@ int sectab_init(struct section_table *sectab)
 	sectab->sections = malloc(sizeof(struct section) * SECTBL_INIT_LEN);
 
 	if (sectab->sections == NULL) {
-		ERROR("cannot alloc");
+		PERROR("cannot alloc");
 		return M_ERROR;
 	}
 
@@ -59,7 +59,7 @@ int sectab_alloc(struct section_table *sectab, struct section **res,
 			      sizeof(struct section) * sectab->size);
 
 		if (sectab->sections == NULL) {
-			ERROR("cannot realloc");
+			PERROR("cannot realloc");
 			return M_ERROR;
 		}
 	}
@@ -94,7 +94,7 @@ int sectab_alloc(struct section_table *sectab, struct section **res,
 	}
 
 	if (sec->entries == NULL) {
-		ERROR("cannot alloc");
+		PERROR("cannot alloc");
 		return M_ERROR;
 	}
 
@@ -127,7 +127,7 @@ int sec_push(struct section *section, struct section_entry entry)
 			 sizeof(struct section_entry) * section->size);
 
 		if (new == NULL) {
-			ERROR("cannot realloc");
+			PERROR("cannot realloc");
 			return M_ERROR;
 		}
 

masm/strtab.c

@@ -41,7 +41,7 @@ int strtab_init(struct str_table *strtab)
 	strtab->size = 1;
 	strtab->ptr = malloc(1);
 	if (strtab->ptr == NULL) {
-		ERROR("cannot alloc");
+		PERROR("cannot alloc");
 		return M_ERROR;
 	}
 	*strtab->ptr = '\0';

masm/symtab.c

@@ -18,7 +18,7 @@ int symtab_init(struct symbol_table *symtab)
         symtab->sections = malloc(sizeof(ssize_t) * SYMTBL_INIT_LEN);
 
         if (symtab->symbols == NULL || symtab->sections == NULL) {
-                ERROR("cannot alloc");
+                PERROR("cannot alloc");
                 return M_ERROR;
         }
 
@@ -44,7 +44,7 @@ int symtab_push(struct symbol_table *symtab, Elf32_Sym sym, ssize_t sec_idx)
                 symtab->sections = realloc(symtab->sections,
                                         sizeof(ssize_t) * symtab->size);
                 if (symtab->symbols == NULL || symtab->sections == NULL) {
-                        ERROR("cannot realloc");
+                        PERROR("cannot realloc");
                         return M_ERROR;
                 }
         }

mld/link.c

@@ -1,9 +1,17 @@
+#include <elf.h>
+#include <stddef.h>
+#include <stdint.h>
+#include <stdio.h>
 #include <stdlib.h>
 #include <merror.h>
 #include <string.h>
 #include <sys/stat.h>
+#include <melf.h>
 
 #include "link.h"
+#include "mips.h"
+
+#define SEC_ALIGN 0x1000
 
 static int load_objects(struct linker *linker)
 {
@@ -12,7 +20,7 @@ static int load_objects(struct linker *linker)
 	linker->obj_len = 0;
 
 	if (linker->objects == NULL) {
-		ERROR("cannot alloc");
+		PERROR("cannot alloc");
 		return M_ERROR;
 	}
 
@@ -40,6 +48,627 @@ skip_obj:
 	return M_SUCCESS;
 }
 
+/**
+ * Relocates all segments with the given name
+ * (since they need to be next to eachother)
+ */
+static int relocate_segment_name(struct linker *linker, const char *name)
+{
+	for (size_t i = 0; i < linker->obj_len; i++) {
+		struct object *obj = &linker->objects[i];
+		for (size_t j = 0; j < obj->segment_len; j++) {
+			struct segment *seg = &obj->segments[j];
+
+			// check if the segment has already been relocated
+			if (seg->new_vaddr != 0)
+				continue;
+
+			// make sure the segments name matches what
+			// we are looking for
+			if (strcmp(seg->name, name) != 0)
+				continue;
+
+			if (ADDR_CHK(linker->off, seg->size, UINT32_MAX)) {
+				ERROR("linker offset overflow");
+				return M_ERROR;
+			}
+
+			// is the segment a TEXT type or DATA type??
+			if (B32(seg->phdr->p_flags) & PF_X) {
+				// TEXT
+				if (ADDR_CHK(linker->text_vaddr, seg->size,
+							DATA_VADDR_MIN)) {
+					ERROR("linker text vaddr overflow");
+					return M_ERROR;
+				}
+
+				seg->new_off = linker->off;
+				seg->new_vaddr = linker->text_vaddr;
+				linker->off += seg->size;
+				linker->text_vaddr += seg->size;
+			} else {
+				// DATA
+				if (ADDR_CHK(linker->data_vaddr, seg->size,
+							UINT32_MAX)) {
+					ERROR("linker data vaddr overflow");
+					return M_ERROR;
+				}
+
+				seg->new_off = linker->off;
+				seg->new_vaddr = linker->data_vaddr;
+				linker->off += seg->size;
+				linker->data_vaddr += seg->size;
+			}
+
+			// if this is an existing segment, append this
+			// part
+			struct segment_table_entry *ent;
+			if (segtab_get(&linker->segments, &ent, name) ==
+				M_SUCCESS) {
+				if (segtab_ent_push(ent, seg))
+					return M_ERROR;
+			} else {
+				// else create a new segment
+				if (segtab_push(&linker->segments, NULL, seg))
+					return M_ERROR;
+			}
+		}
+	}
+
+	return M_SUCCESS;
+}
+
+static int relocate_segments(struct linker *linker)
+{
+	for (size_t i = 0; i < linker->obj_len; i++) {
+		struct object *obj = &linker->objects[i];
+		for (size_t j = 0; j < obj->segment_len; j++) {
+			struct segment *seg = &obj->segments[j];
+
+			// check if the segment has already been relocated
+			if (seg->new_vaddr != 0)
+				continue;
+			if(relocate_segment_name(linker, seg->name))
+				return M_ERROR;
+		}
+	}
+
+	return M_SUCCESS;
+}
+
+static int relocate_symbol(struct linker *linker, struct object *obj,
+			   struct symbol_table *symtab, const Elf32_Sym *sym)
+{
+	size_t shndx = B16(sym->st_shndx);
+	if (shndx == 0)
+		return M_SUCCESS; // ignore this symbol
+
+	// find the given section
+	const char *name = symtab->strtab->data + B32(sym->st_name);
+
+	if (shndx >= obj->shdr_len) {
+		ERROR("shdr entry [%d] name out of bounds", shndx);
+		return M_ERROR;
+	}
+
+	if (B32(sym->st_name) >= symtab->strtab->len) {
+		ERROR("symbol name out of bounds");
+		return M_ERROR;
+	}
+
+	Elf32_Shdr const *shdr = &obj->shdr[shndx];
+	struct segment *sec = NULL;
+	for (size_t i = 0; i < obj->phdr_len; i++) {
+		Elf32_Phdr *temp = &obj->phdr[i];
+		if (shdr->sh_offset == temp->p_offset) {
+			sec = &obj->segments[i];
+			break;
+		}
+	}
+
+	if (sec == NULL) {
+		ERROR("could not locate segment for symbol '%s'", name);
+		return M_ERROR;
+	}
+
+	struct segment_table_entry *ent = NULL;
+	if (segtab_get(&linker->segments, &ent, sec->name)) {
+		ERROR("could not locate segment for symbol '%s'", name);
+		return M_ERROR;
+	}
+
+	// segments start at shindx 1
+	ptrdiff_t new_shndx = (ent - linker->segments.entries) + 1;
+
+	size_t str_off = 0;
+	if (strtab_push(linker->symtab.strtab, name, &str_off))
+		return M_ERROR;
+
+	int32_t off = sec->new_vaddr + B32(sym->st_value);
+	Elf32_Sym new = *sym;
+	new.st_name = B32(str_off);
+	new.st_value = B32(off);
+	new.st_shndx = B16(new_shndx);
+	new.st_size = 0;
+
+	if (symtab_get(&linker->symtab, NULL, name) == M_SUCCESS) {
+		ERROR("cannot link doubly defiend symbol '%s'", name);
+		return M_ERROR;
+	}
+
+	if (symtab_push(&linker->symtab, &new))
+		return M_ERROR;
+
+	return M_SUCCESS;
+}
+
+static int relocate_symbols(struct linker *linker)
+{
+	for (size_t i = 0; i < linker->obj_len; i++) {
+		struct object *obj = &linker->objects[i];
+		for (size_t j = 0; j < obj->shdr_len; j++) {
+			struct symbol_table *symtab = &obj->symtabs[j];
+			if (symtab->len < 1)
+				continue;
+
+			for (size_t k = 0; k < symtab->len; k++) {
+				const Elf32_Sym *sym = &symtab->syms[k];
+				if (relocate_symbol(linker, obj, symtab, sym))
+					return M_ERROR;
+			}
+		}
+	}
+	return M_SUCCESS;
+}
+
+static int assemble_phdr(struct linker *linker)
+{
+	Elf32_Phdr *phdr = malloc(sizeof(Elf32_Phdr) * linker->segments.len);
+
+	if (phdr == NULL) {
+		PERROR("cannot alloc");
+		return M_ERROR;
+	}
+
+	for (uint32_t i = 0; i < linker->segments.len; i++) {
+		Elf32_Phdr *hdr = &phdr[i];
+		struct segment_table_entry *ent = &linker->segments.entries[i];
+		size_t size = segtab_ent_size(ent);
+		hdr->p_type = B32(PT_LOAD);
+		hdr->p_flags = B32(
+			(ent->parts[0]->execute << 0) |
+			(ent->parts[0]->write << 1) |
+			(ent->parts[0]->read << 2));
+		hdr->p_offset = B32(ent->off);
+		hdr->p_vaddr = B32(ent->vaddr);
+		hdr->p_paddr = B32(ent->vaddr);
+		hdr->p_filesz = B32(size);
+		hdr->p_memsz = B32(size);
+		hdr->p_align = B32(SEC_ALIGN);
+	}
+
+	linker->phdr = phdr;
+	linker->phdr_len = linker->segments.len;
+	return M_SUCCESS;
+}
+
+static int assemble_shdr(struct linker *linker)
+{
+	uint32_t max_entries = 0;
+	max_entries += 1; // null
+	max_entries += 1; // symtab
+	max_entries += 1; // strtab
+	max_entries += 1; // shstrtab
+	max_entries += linker->segments.len; // segments
+
+	Elf32_Shdr *shdr = malloc(sizeof(Elf32_Shdr) * max_entries);
+	if (shdr == NULL) {
+		PERROR("cannot alloc");
+		return M_ERROR;
+	}
+	linker->shdr = shdr;
+
+	size_t str_off;
+	uint32_t count = 0;
+
+	// null
+	shdr[count++] = (Elf32_Shdr) {0};
+
+	// segments
+	for (uint32_t i = 0; i < linker->segments.len; i++) {
+		struct segment_table_entry *ent = &linker->segments.entries[i];
+		if (strtab_push(&linker->shstrtab, ent->name, &str_off))
+			return M_ERROR;
+
+		shdr[count++] = (Elf32_Shdr) {
+			.sh_name = B32(str_off),
+			.sh_type = B32(SHT_PROGBITS),
+			.sh_flags = B32(
+				(ent->parts[0]->write << 0) |
+				(ent->parts[0]->execute << 2) |
+				SHF_ALLOC),
+			.sh_addr = B32(ent->vaddr),
+			.sh_offset = B32(ent->off),
+			.sh_size = B32(segtab_ent_size(ent)),
+			.sh_link = 0,
+			.sh_info = 0,
+			.sh_addralign = B32(ent->parts[0]->align),
+			.sh_entsize = 0,
+		};
+	}
+
+	// symbol table
+	if (strtab_push(&linker->shstrtab, ".symtab", &str_off))
+		return M_ERROR;
+
+	linker->symtab_shidx = count;
+	shdr[count++] = (Elf32_Shdr) {
+		.sh_name = B32(str_off),
+		.sh_type = B32(SHT_SYMTAB),
+		.sh_flags = 0,
+		.sh_addr = 0,
+		.sh_offset = 0,
+		.sh_size = 0,
+		.sh_link = 0,
+		.sh_info = 0,
+		.sh_addralign = B32(1),
+		.sh_entsize = B32(sizeof(Elf32_Sym)),
+	};
+
+	// string table
+	if (strtab_push(&linker->shstrtab, ".strtab", &str_off))
+		return M_ERROR;
+
+	linker->strtab_shidx = count;
+	shdr[count++] = (Elf32_Shdr) {
+		.sh_name = B32(str_off),
+		.sh_type = B32(SHT_STRTAB),
+		.sh_flags = B32(SHF_STRINGS),
+		.sh_addr = 0,
+		.sh_offset = 0,
+		.sh_size = 0,
+		.sh_link = 0,
+		.sh_info = 0,
+		.sh_addralign = B32(1),
+		.sh_entsize = 0,
+	};
+
+	// shstring table
+	if (strtab_push(&linker->shstrtab, ".shstrtab", &str_off))
+		return M_ERROR;
+
+	linker->shstrtab_shidx = count;
+	shdr[count++] = (Elf32_Shdr) {
+		.sh_name = B32(str_off),
+		.sh_type = B32(SHT_STRTAB),
+		.sh_flags = B32(SHF_STRINGS),
+		.sh_addr = 0,
+		.sh_offset = 0,
+		.sh_size = 0,
+		.sh_link = 0,
+		.sh_info = 0,
+		.sh_addralign = B32(1),
+		.sh_entsize = 0,
+	};
+
+	linker->shdr_len = count;
+	return M_SUCCESS;
+}
+
+static int relocate_instruction_rela(struct linker *linker,
+				     struct segment *seg,
+				     Elf32_Rela *rel)
+{
+	/// get start of the segment part in bytes and the byte offset
+	/// of this relocation in that segment part
+	uint32_t off = B32(rel->r_offset);
+	if (off > seg->size) {
+		ERROR("relocation in segment '%s' out of bounds", seg->name);
+		return M_ERROR;
+	}
+
+	if (B32(rel->r_info) == 0) {
+		WARNING("skiping empty relocation entry");
+		return M_SUCCESS;
+	}
+
+	/// read the relocation entry
+	uint8_t idx = B32(rel->r_info) >> 8;
+	uint8_t typ = B32(rel->r_info) & 0xFF;
+	int32_t add = B32(rel->r_addend);
+
+	/// read the symbol from the relocation
+	struct symbol_table *symtab = seg->reltab.symtab;
+	if (idx >= symtab->len) {
+		ERROR("relocation in segment '%s', symbol index [%d] out of "
+			"bounds", seg->name, idx);
+		return M_ERROR;
+	}
+	Elf32_Sym *sym = &symtab->syms[idx];
+	const char *symname = symtab->strtab->data + B32(sym->st_name);
+
+
+	/// get the section header that the symbol is related to
+	Elf32_Shdr *shdr = NULL;
+	if (B16(sym->st_shndx) >= seg->obj->shdr_len) {
+		ERROR("shndx index [%d] out of bounds", B16(sym->st_shndx));
+		return M_ERROR;
+	}
+	shdr = &seg->obj->shdr[B16(sym->st_shndx)];
+
+	/// get the segment that the symbol is in
+	struct segment_table_entry *ent;
+	const char *segname = seg->obj->shstrtab->data + B32(shdr->sh_name);
+	if (segtab_get(&linker->segments, &ent, segname)) {
+		ERROR("could not locate segment for relocation");
+		return M_ERROR;
+	}
+
+	uint32_t sym_vaddr = B32(sym->st_value) + ent->vaddr;
+	uint32_t *ins_raw = (uint32_t *) &seg->bytes[off];
+
+	union mips_instruction_data ins;
+	ins.raw = B32(*ins_raw);
+
+	uint32_t ins_vaddr = seg->new_vaddr + off;
+
+	uint32_t vaddr_abs = sym_vaddr + add;
+	 int32_t vaddr_rel = (vaddr_abs - ins_vaddr - 4) >> 2;
+	bool warn = false;
+
+	switch (typ) {
+	case R_MIPS_16:
+		// 16bit absolute
+		if (vaddr_abs > UINT16_MAX)
+			warn = true;
+		ins.immd = (uint16_t)vaddr_abs;
+		break;
+	case R_MIPS_PC16:
+		// 16bit relative shifted
+		if (vaddr_rel > INT16_MAX || vaddr_rel < INT16_MIN)
+			warn = true;
+		ins.offset = vaddr_rel;
+		break;
+	case R_MIPS_26:
+		// 26bit absolute shifted
+		if (vaddr_abs >= (1 << 25))
+			warn = true;
+		ins.target = (vaddr_abs & 0x0FFFFFFF) >> 2;
+		break;
+	case R_MIPS_PC26_S2:
+		// 26bit relative shifted
+		if (vaddr_rel >= (1 << 24) || -vaddr_rel > (1 << 24))
+			warn = true;
+		ins.offs26 = vaddr_rel;
+		break;
+	case R_MIPS_LO16:
+		// lo 16bit absolute
+		ins.immd = (uint16_t)(vaddr_abs & 0xFFFF);
+		break;
+	case R_MIPS_HI16:
+		// hi 16bit absolute
+		ins.immd = (uint16_t)(vaddr_abs >> 16);
+		break;
+	default:
+		ERROR("do not know how do handle relocation type [%d]", typ);
+		return M_ERROR;
+	}
+
+	*ins_raw = B32(ins.raw);
+
+	if (warn)
+		WARNING("truncating relocation for symbol '%s'", symname);
+
+	return M_SUCCESS;
+
+}
+
+static int relocate_instruction_rel(struct linker *linker,
+				    struct segment *seg,
+				    Elf32_Rel *rel)
+{
+	Elf32_Rela temp;
+	temp.r_info = rel->r_info;
+	temp.r_offset = rel->r_offset;
+	temp.r_addend = 0;
+
+	return relocate_instruction_rela(linker, seg, &temp);
+}
+
+static int relocate_segment_instructions(struct linker *linker,
+					 struct segment *seg)
+{
+	for (uint32_t i = 0; i < seg->reltab.len; i++) {
+		int res = M_SUCCESS;
+		if (seg->reltab.type == SHT_RELA)
+			res = relocate_instruction_rela(linker, seg,
+				   &seg->reltab.rela[i]);
+		else if (seg->reltab.type == SHT_REL)
+			res = relocate_instruction_rel(linker, seg,
+				   &seg->reltab.rel[i]);
+		else {
+			ERROR("unknown reltab type");
+			return M_ERROR;
+		}
+		if (res)
+			return M_ERROR;
+	}
+	return M_SUCCESS;
+}
+
+static int relocate_instructions(struct linker *linker)
+{
+	for (uint32_t i = 0; i < linker->segments.len; i++) {
+		struct segment_table_entry *ent = &linker->segments.entries[i];
+		for (uint32_t j = 0; j < ent->len; j++) {
+			struct segment *seg = ent->parts[j];
+			if (relocate_segment_instructions(linker, seg))
+				return M_ERROR;
+		}
+	}
+	return M_SUCCESS;
+}
+
+static void update_offsets(struct linker *linker)
+{
+	uint32_t ptr = 0;
+
+	// we must now correct offsets and sizes in side the ehdr, phdr,
+	// and shdr
+	ptr += sizeof(Elf32_Ehdr);
+
+	// phdr
+	linker->ehdr.e_phoff = B32(ptr);
+	ptr += linker->phdr_len * sizeof(Elf32_Phdr);
+
+	// section padding
+	{
+		uint32_t mod = ptr % SEC_ALIGN;
+		if (mod != 0)
+			linker->secalign = (SEC_ALIGN - mod);
+		else
+			 linker->secalign = 0;
+		ptr += linker->secalign;
+	}
+
+	// sections
+	for (uint32_t i = 0; i < linker->segments.len; i++) {
+		struct segment_table_entry *ent = &linker->segments.entries[i];
+		uint32_t idx = i + 1;
+		uint32_t size = segtab_ent_size(ent);
+		linker->phdr[i].p_offset = B32(ptr);
+		linker->shdr[idx].sh_offset = linker->phdr[i].p_offset;
+		ptr += size;
+	}
+
+	// symtab
+	Elf32_Shdr *symtab = &linker->shdr[linker->symtab_shidx];
+	symtab->sh_offset = B32(ptr);
+	symtab->sh_link = B32(linker->strtab_shidx);
+	symtab->sh_size = B32(linker->symtab.len * sizeof(Elf32_Sym));
+	ptr += B32(symtab->sh_size);
+
+	// strtab
+	Elf32_Shdr *strtab = &linker->shdr[linker->strtab_shidx];
+	strtab->sh_offset = B32(ptr);
+	strtab->sh_size = B32(linker->strtab.len);
+	ptr += linker->strtab.len;
+
+	// shstrtab
+	Elf32_Shdr *shstrtab = &linker->shdr[linker->shstrtab_shidx];
+	shstrtab->sh_offset = B32(ptr);
+	shstrtab->sh_size = B32(linker->shstrtab.len);
+	ptr += linker->shstrtab.len;
+
+	// shdr
+	linker->ehdr.e_shoff = B32(ptr);
+}
+
+static int write_file(struct linker *linker)
+{
+	extern char *current_file;
+	current_file = linker->args->out_file;
+
+	FILE *out = fopen(linker->args->out_file, "w");
+	if (out == NULL) {
+		PERROR("cannot write");
+		return M_ERROR;
+	}
+
+	int res = 0;
+
+	// ehdr
+	res |= fwrite(&linker->ehdr, sizeof(Elf32_Ehdr), 1, out);
+
+	// phdr
+	res |= fwrite(linker->phdr, sizeof(Elf32_Phdr), linker->phdr_len, out);
+
+	// section padding
+	for (uint32_t i = 0; i < linker->secalign; i++) {
+		uint8_t zero = 0;
+		res |= fwrite(&zero, 1, 1, out);
+	}
+
+	// sections
+	for (uint32_t i = 0; i < linker->segments.len; i++) {
+		struct segment_table_entry *ent = &linker->segments.entries[i];
+		for (uint32_t j = 0; j < ent->len; j++) {
+			struct segment *seg = ent->parts[j];
+			res |= fwrite(seg->bytes, 1, seg->size, out);
+		}
+	}
+
+	// sym tbl
+	res |= fwrite(linker->symtab.syms, sizeof(Elf32_Sym), linker->symtab.len, out);
+
+	// str tbl
+	res |= fwrite(linker->strtab.data, 1, linker->strtab.len, out);
+
+	// shstr tbl
+	res |= fwrite(linker->shstrtab.data, 1, linker->shstrtab.len, out);
+
+	// shdr
+	res |= fwrite(linker->shdr, sizeof(Elf32_Shdr), linker->shdr_len, out);
+
+	if (res < 0) {
+		ERROR("cannot write data");
+		return M_ERROR;
+	}
+
+	fclose(out);
+
+	return M_SUCCESS;
+}
+
+static int link_executable(struct linker *linker)
+{
+	Elf32_Ehdr *ehdr = &linker->ehdr;
+	*ehdr = MIPS_ELF_EHDR;
+
+	if (assemble_phdr(linker))
+		return M_ERROR;
+
+	if (assemble_shdr(linker))
+		return M_ERROR;
+
+	ehdr->e_type = B16(ET_EXEC);
+	ehdr->e_phnum = B16(linker->phdr_len);
+	ehdr->e_shnum = B16(linker->shdr_len);
+	ehdr->e_shstrndx = B16(linker->shstrtab_shidx);
+
+	update_offsets(linker);
+
+	if (write_file(linker))
+		return M_ERROR;
+
+	return M_SUCCESS;
+}
+
+static int linker_init(struct linker *linker, struct linker_arguments *args)
+{
+	linker->args = args;
+	linker->off = 0;
+	linker->text_vaddr = TEXT_VADDR_MIN;
+	linker->data_vaddr = DATA_VADDR_MIN;
+	linker->objects = NULL;
+	linker->segments.size = 0;
+	linker->symtab.syms = NULL;
+	linker->shstrtab.data = NULL;
+	linker->strtab.data = NULL;
+	linker->shdr = NULL;
+	linker->phdr = NULL;
+	if (segtab_init(&linker->segments))
+		return M_ERROR;
+	if (strtab_init(&linker->shstrtab))
+		return M_ERROR;
+	if (strtab_init(&linker->strtab))
+		return M_ERROR;
+	if (symtab_init(&linker->symtab))
+		return M_ERROR;
+	linker->symtab.strtab = &linker->strtab;
+	return M_SUCCESS;
+}
+
 static void linker_free(struct linker *linker)
 {
 	if (linker->objects != NULL) {
@@ -47,15 +676,32 @@ static void linker_free(struct linker *linker)
 			object_free(&linker->objects[i]);
 		free(linker->objects);
 	}
+	if (linker->shdr != NULL)
+		free(linker->shdr);
+	if (linker->phdr != NULL)
+		free(linker->phdr);
+	segtab_free(&linker->segments);
+	strtab_free(&linker->shstrtab);
+	strtab_free(&linker->strtab);
+	symtab_free(&linker->symtab);
 }
 
 int link_files(struct linker_arguments args) {
 	struct linker linker;
-	linker.args = &args;
-	int res = M_SUCCESS;
 
+	int res = M_SUCCESS;
+	if (res == M_SUCCESS)
+		res = linker_init(&linker, &args);
 	if (res == M_SUCCESS)
 		res = load_objects(&linker);
+	if (res == M_SUCCESS)
+		res = relocate_segments(&linker);
+	if (res == M_SUCCESS)
+		res = relocate_symbols(&linker);
+	if (res == M_SUCCESS)
+		res = relocate_instructions(&linker);
+	if (res == M_SUCCESS)
+		res = link_executable(&linker);
 
 	linker_free(&linker);
 	return res;

mld/link.h

@@ -9,7 +9,6 @@
 #include <merror.h>
 #include <stdint.h>
 #include <elf.h>
-#include <stdio.h>
 
 
 // when mapping porinters, we need to bounds check to
@@ -21,6 +20,15 @@
 #define BOUND_CHK(obj, len, off) \
 	(off > UINT32_MAX - len || off + len > obj->mapped_size)
 
+// when relocating segments, we need to bounds check to
+// make sure it wont overflow the addresses past the 32bit
+// ELF file
+#define ADDR_CHK(lnk_f, seg_f, max) \
+	((lnk_f) > max - (seg_f) || (lnk_f) + (seg_f) > max)
+
+// start addresses for each tyoe of segment
+#define TEXT_VADDR_MIN 0x00400000
+#define DATA_VADDR_MIN 0x10000000
 
 // pre define
 struct linker;
@@ -51,6 +59,12 @@ struct string_table {
 	size_t len;
 };
 
+int strtab_init(struct string_table *strtab);
+void strtab_free(struct string_table *strtab);
+
+int strtab_push(struct string_table *strtab, const char *str, size_t *res);
+int strtab_get(struct string_table *strtab, const char *str, size_t *res);
+
 ///
 /// symbol table
 ///
@@ -59,14 +73,42 @@ struct symbol_table {
 	struct string_table *strtab;
 	Elf32_Sym *syms;
 	size_t len;
+	size_t size;
 };
 
+int symtab_init(struct symbol_table *symtab);
+void symtab_free(struct symbol_table *symtab);
+
+int symtab_push(struct symbol_table *symtab, const Elf32_Sym *sym);
+int symtab_get(struct symbol_table *symtab, Elf32_Sym **sym, const char *name);
+
 ///
 /// segment
 ///
 
 /* a loadable program segment */
 struct segment {
+	// segment data
+	char *name;
+	unsigned char *bytes;
+
+	// current loc
+	uint32_t off;
+	uint32_t vaddr;
+
+	// new loc
+	uint32_t new_off;
+	uint32_t new_vaddr;
+
+	// meta
+	bool read;
+	bool write;
+	bool execute;
+	uint32_t align;
+
+	// size
+	uint32_t size;
+
 	// phdr
 	Elf32_Phdr *phdr;
 	uint32_t phdr_idx;
@@ -75,9 +117,8 @@ struct segment {
 	Elf32_Shdr *shdr;
 	uint32_t shdr_idx;
 
-	// segment data
+	// object im related to
-	char *name;
+	struct object *obj;
-	unsigned char *bytes;
 
 	// relocation table
 	struct relocation_table reltab;
@@ -85,6 +126,46 @@ struct segment {
 
 int segment_load(struct object *object, struct segment *segment, size_t index);
 
+///
+/// segment table
+///
+
+struct segment_table_entry {
+	char *name;
+	uint32_t len;
+	uint32_t size;
+	uint32_t off;
+	uint32_t vaddr;
+	// weak segment pointers. we do not own these!!!
+	struct segment **parts;
+};
+
+int segtab_ent_init(struct segment_table_entry *ent);
+void segtab_ent_free(struct segment_table_entry *ent);
+
+int segtab_ent_push(struct segment_table_entry *ent, struct segment *seg);
+uint32_t segtab_ent_size(struct segment_table_entry *ent);
+
+// holds each segment by name
+// and all the segment parts from each of the
+// object files
+struct segment_table {
+	uint32_t len;
+	uint32_t size;
+	struct segment_table_entry *entries;
+};
+
+int segtab_init(struct segment_table *segtab);
+void segtab_free(struct segment_table *segtab);
+
+/* create a new entry with <seg> as its first segment part */
+int segtab_push(struct segment_table *segtab, struct segment_table_entry **ent,
+		struct segment *seg);
+
+/* find a segment table entry with a given name */
+int segtab_get(struct segment_table *segtab, struct segment_table_entry **ent,
+	       const char *name);
+
 ///
 /// object file
 ///
@@ -136,6 +217,34 @@ struct linker {
 	struct object *objects;
 
 	struct linker_arguments *args;
+
+	// current pointers to relocate
+	// sections
+	uint32_t off;
+	uint32_t text_vaddr;
+	uint32_t data_vaddr;
+
+	// elf tables
+	struct string_table shstrtab;
+	struct string_table strtab;
+	struct symbol_table symtab;
+
+	// output elf
+	Elf32_Ehdr ehdr;
+	Elf32_Phdr *phdr;
+	uint32_t phdr_len;
+	Elf32_Shdr *shdr;
+	uint32_t shdr_len;
+
+	uint32_t symtab_shidx;
+	uint32_t strtab_shidx;
+	uint32_t shstrtab_shidx;
+
+	// section alignment after phdr bytes
+	uint32_t secalign;
+
+	// all segments
+	struct segment_table segments;
 };
 
 /* defines arguments to the linker */

mld/obj.c

@@ -171,6 +171,7 @@ static int load_symtabs(struct object *object)
 
 		if (B32(hdr->sh_type) != SHT_SYMTAB) {
 			symtab->len = 0;
+			symtab->size = 0;
 			continue;
 		}
 
@@ -190,7 +191,8 @@ static int load_symtabs(struct object *object)
 		}
 
 		symtab->strtab = strtab;
-		symtab->len = len;
+		symtab->len = len / sizeof(Elf32_Sym);
+		symtab->size = len / sizeof(Elf32_Sym);
 		symtab->syms = (Elf32_Sym *) (object->mapped + off);
 
 		if (BOUND_CHK(object, len, off)) {
@@ -263,8 +265,8 @@ static int map_file(struct object *obj, char *path)
 	}
 
 	obj->mapped_size = st.st_size;
-	obj->mapped = mmap(NULL, st.st_size, PROT_READ, MAP_SHARED,
+	obj->mapped = mmap(NULL, st.st_size, PROT_READ | PROT_WRITE,
-				obj->fd, 0);
+		    MAP_PRIVATE, obj->fd, 0);
 
 	if (obj->mapped == MAP_FAILED) {
 		PERROR("cannot map");

mld/seg.c

@@ -82,6 +82,13 @@ static int load_reltab(struct object *obj, struct segment *seg)
 		if (type != SHT_REL && type != SHT_RELA)
 			continue;
 
+		if ((type == SHT_REL && B32(hdr->sh_entsize) !=
+			sizeof(Elf32_Rel)) || (type == SHT_RELA &&
+			B32(hdr->sh_entsize) != sizeof(Elf32_Rela))) {
+			ERROR("reltab [%d] has invalid entry size", i);
+			return M_ERROR;
+		}
+
 		if (B32(hdr->sh_info) != seg->shdr_idx)
 			continue;
 
@@ -101,9 +108,8 @@ static int load_reltab(struct object *obj, struct segment *seg)
 		uint32_t len = B32(hdr->sh_size);
 		uint32_t off = B32(hdr->sh_offset);
 
-
 		seg->reltab.symtab = symtab;
-		seg->reltab.len = len;
+		seg->reltab.len = len / B32(hdr->sh_entsize);
 		seg->reltab.raw = obj->mapped + off;
 		seg->reltab.type = type;
 
@@ -131,5 +137,18 @@ int segment_load(struct object *obj, struct segment *seg, size_t index)
 	if (load_reltab(obj, seg))
 		return M_ERROR;
 
+	seg->off = B32(seg->phdr->p_offset);
+	seg->vaddr = B32(seg->phdr->p_vaddr);
+	seg->size = B32(seg->phdr->p_filesz);
+	seg->new_off = 0;
+	seg->new_vaddr = 0;
+
+	seg->read = B32(seg->phdr->p_flags) & PF_R;
+	seg->write = B32(seg->phdr->p_flags) & PF_W;
+	seg->execute = B32(seg->phdr->p_flags) & PF_X;
+	seg->align = B32(seg->phdr->p_align);
+
+	seg->obj = obj;
+
 	return M_SUCCESS;
 }

mld/segtab.c

@@ -0,0 +1,147 @@
+#include <stdlib.h>
+#include <merror.h>
+
+#include "link.h"
+
+#define SEGTAB_INIT_SIZE 8
+
+int segtab_init(struct segment_table *segtab)
+{
+	segtab->len = 0;
+	segtab->size = SEGTAB_INIT_SIZE;
+	segtab->entries = malloc(sizeof(struct segment_table_entry) *
+				SEGTAB_INIT_SIZE);
+
+	if (segtab->entries == NULL) {
+		PERROR("cannot alloc");
+		return M_ERROR;
+	}
+
+	return M_SUCCESS;
+}
+
+void segtab_free(struct segment_table *segtab)
+{
+	for (uint32_t i = 0; i < segtab->len; i++) {
+		segtab_ent_free(&segtab->entries[i]);
+	}
+	free(segtab->entries);
+}
+
+/* create a new entry with <seg> as its first segment part */
+int segtab_push(struct segment_table *segtab, struct segment_table_entry **res,
+		struct segment *seg)
+{
+	if (segtab->len >= segtab->size) {
+		uint32_t size = segtab->size * 2;
+		void *new = realloc(segtab->entries,
+		      sizeof(struct segment_table_entry) * size);
+		if (new == NULL) {
+			PERROR("cannot relloc");
+			return M_ERROR;
+		}
+		segtab->size = size;
+		segtab->entries = new;
+	}
+
+	struct segment_table_entry ent;
+	if (segtab_ent_init(&ent))
+		return M_ERROR;
+	ent.name = seg->name;
+	ent.vaddr = seg->vaddr;
+	ent.off = seg->off;
+
+	if (segtab_ent_push(&ent, seg)) {
+		segtab_ent_free(&ent);
+		return M_ERROR;
+	}
+
+	segtab->entries[segtab->len] = ent;
+
+	if (res != NULL)
+		*res = &segtab->entries[segtab->len];
+
+	segtab->len++;
+
+	return M_SUCCESS;
+}
+
+/* find a segment table entry with a given name */
+int segtab_get(struct segment_table *segtab, struct segment_table_entry **ent,
+	       const char *name)
+{
+	for (uint32_t i = 0; i < segtab->len; i++) {
+		const char *segname = segtab->entries[i].name;
+		if (strcmp(name, segname) != 0)
+			continue;
+
+		*ent = &segtab->entries[i];
+		return M_SUCCESS;
+	}
+
+	return M_ERROR;
+}
+
+int segtab_ent_init(struct segment_table_entry *ent)
+{
+	ent->len = 0;
+	ent->size = SEGTAB_INIT_SIZE;
+	ent->parts = malloc(sizeof(struct segment *) *
+				SEGTAB_INIT_SIZE);
+
+	if (ent->parts == NULL) {
+		PERROR("cannot alloc");
+		return M_ERROR;
+	}
+
+	return M_SUCCESS;
+}
+
+void segtab_ent_free(struct segment_table_entry *ent)
+{
+	free(ent->parts);
+}
+
+int segtab_ent_push(struct segment_table_entry *ent, struct segment *seg)
+{
+	if (ent->len >= ent->size) {
+		uint32_t size = ent->size * 2;
+		void *new = realloc(ent->parts,
+		      sizeof(struct segment *) * size);
+		if (new == NULL) {
+			PERROR("cannot relloc");
+			return M_ERROR;
+		}
+		ent->size = size;
+		ent->parts = new;
+	}
+
+	if (ent->len > 0) {
+		struct segment *first = ent->parts[0];
+		if (first->align != seg->align) {
+			ERROR("segment '%s' doest not have matching alignment",
+				ent->name);
+		}
+		if (first->read != seg->read ||
+			first->write != seg->write ||
+			first->execute != seg->execute) {
+			ERROR("segment '%s' doest not have matching RWX",
+				ent->name);
+		}
+	} else {
+		ent->off = seg->new_off;
+		ent->vaddr = seg->new_vaddr;
+	}
+
+	ent->parts[ent->len++] = seg;
+	return M_SUCCESS;
+}
+
+uint32_t segtab_ent_size(struct segment_table_entry *ent)
+{
+	uint32_t size = 0;
+	for (uint32_t i = 0; i < ent->len; i++) {
+		size += ent->parts[i]->size;
+	}
+	return size;
+}

mld/strtab.c

@@ -0,0 +1,56 @@
+#include <merror.h>
+#include <stdlib.h>
+
+#include "link.h"
+
+int strtab_init(struct string_table *strtab)
+{
+	strtab->len = 1;
+	strtab->data = malloc(1);
+
+	if (strtab->data == NULL) {
+		PERROR("cannot alloc");
+		return M_ERROR;
+	}
+
+	strtab->data[0] = '\0';
+	return M_SUCCESS;
+}
+
+void strtab_free(struct string_table *strtab)
+{
+	free(strtab->data);
+}
+
+int strtab_push(struct string_table *strtab, const char *str, size_t *res)
+{
+	if (strtab_get(strtab, str, res) == M_SUCCESS)
+		return M_SUCCESS;
+
+	size_t len = strlen(str);
+	char *new = realloc(strtab->data, strtab->len + len + 1);
+	if (new == NULL) {
+		PERROR("cannot realloc");
+		return M_ERROR;
+	}
+	strtab->data = new;
+	memcpy(strtab->data + strtab->len, str, len + 1);
+
+	if (res != NULL)
+		*res = strtab->len;
+	strtab->len += len + 1;
+
+	return M_SUCCESS;
+}
+
+int strtab_get(struct string_table *strtab, const char *str, size_t *res)
+{
+	for (size_t i = 0; i < strtab->len; i++) {
+		if (strcmp(strtab->data + i, str) == 0) {
+			if (res != NULL)
+				*res = i;
+			return M_SUCCESS;
+		}
+	}
+	return M_ERROR;
+}

mld/symtab.c

@@ -0,0 +1,60 @@
+#include <elf.h>
+#include <merror.h>
+#include <stdlib.h>
+#include <melf.h>
+
+#include "link.h"
+
+#define SYMTAB_INIT_LEN 8
+
+int symtab_init(struct symbol_table *symtab)
+{
+	symtab->len = 1;
+	symtab->size = SYMTAB_INIT_LEN;
+	symtab->syms = malloc(sizeof(Elf32_Sym) * SYMTAB_INIT_LEN);
+
+	if (symtab->syms == NULL) {
+		PERROR("cannot alloc");
+		return M_ERROR;
+	}
+
+	symtab->syms[0] = (Elf32_Sym){0};
+
+	return M_SUCCESS;
+}
+
+void symtab_free(struct symbol_table *symtab)
+{
+	free(symtab->syms);
+}
+
+int symtab_push(struct symbol_table *symtab, const Elf32_Sym *sym)
+{
+	if (symtab->len >= symtab->size) {
+		size_t size = symtab->size *= 2;
+		void *new = realloc(symtab->syms, sizeof(Elf32_Sym) * size);
+		if (new == NULL) {
+			PERROR("cannot realloc");
+			return M_ERROR;
+		}
+		symtab->size = size;
+		symtab->syms = new;
+	}
+
+	symtab->syms[symtab->len++] = *sym;
+	return M_SUCCESS;
+}
+
+int symtab_get(struct symbol_table *symtab, Elf32_Sym **res, const char *name)
+{
+	for (size_t i = 0; i < symtab->len; i++) {
+		Elf32_Sym *sym = &symtab->syms[i];
+		const char *symname = symtab->strtab->data + B32(sym->st_name);
+		if (strcmp(name, symname) == 0) {
+			if (res != NULL)
+				*res = sym;
+			return M_SUCCESS;
+		}
+	}
+	return M_ERROR;
+}

test/Makefile

@@ -3,18 +3,15 @@ LD=afl-cc
 
 BIN=../bin
 
-none:
+.PHONY: masm mld msim
 
-clean:
+masm:
-	rm -fr $(BIN)
+	make -C ../masm clean build CC=$(CC) LD=$(LD)
-
-masm: clean
-	make -C ../masm build CC=$(CC) LD=$(LD)
 	$(BIN)/masm/masm -o ./mld/test.o ./masm/test.asm
 
-mld: clean
+mld:
-	make -C ../mld build CC=$(CC) LD=$(LD)
+	make -C ../mld clean build CC=$(CC) LD=$(LD)
 	$(BIN)/mld/mld -o ./msim/test ./mld/test.o
 
-msim: clean
+msim:
-	make -C ../msim build CC=$(CC) LD=$(LD)
+	make -C ../msim clean build CC=$(CC) LD=$(LD)