Fix ELF parsing

src/compile.c

@@ -495,23 +495,153 @@ static void _child_cleanup(asmsh_asmc_child_t *child)
 }
 
 
+static int text_section_informations(const int fd,
+		const unsigned long int strtab_section_offset,
+		const unsigned long int section_header_off,
+		const unsigned short section_header_count,
+		const unsigned short section_header_sz,
+		unsigned long int *text_off, unsigned long int *text_sz)
+{
+	const int off_section_off = 0x18;
+
+	off_t off;
+	ssize_t rret;
+	int err;
+	unsigned long int strtab_offset, strtab_sz;
+
+	if((off = lseek(fd, strtab_section_offset + off_section_off, SEEK_SET)) != \
+			strtab_section_offset + off_section_off)
+	{
+		err=errno;
+		asmsh_log_perror("Unable to seek at strtab offset");
+		errno = err;
+		return -1;
+	}
+	if((rret = read(fd, &strtab_offset, 8)) != 8)
+	{
+		err=errno;
+		asmsh_log_perror("Unable to read strtab offset");
+		errno = err;
+		return -1;
+	}
+	if((rret = read(fd, &strtab_sz, 8)) != 8)
+	{
+		err=errno;
+		asmsh_log_perror("Unable to read strtab size");
+		errno = err;
+		return -1;
+	}
+
+	char strtab[strtab_sz];
+	if((off = lseek(fd, strtab_offset, SEEK_SET)) != strtab_offset)
+	{
+		err=errno;
+		asmsh_log_perror("Unable to seek at strtab");
+		errno = err;
+		return -1;
+	}
+	if((rret = read(fd, strtab, strtab_sz)) != strtab_sz)
+	{
+		err=errno;
+		asmsh_log_perror("Unable to read strtab");
+		errno = err;
+		return -1;
+	}
+
+
+	for(unsigned short i=0; i<section_header_count; i++)
+	{
+		unsigned long int cur_off = section_header_off + i*section_header_sz;
+		if((off = lseek(fd, cur_off, SEEK_SET)) != cur_off)
+		{
+			err=errno;
+			asmsh_log_perror("Unable to seek section");
+			errno = err;
+			return -1;
+		}
+		unsigned int name_off;
+		if((rret = read(fd, &name_off, 4)) != 4)
+		{
+			err=errno;
+			asmsh_log_perror("Unable to read section's name offset");
+			errno = err;
+			return -1;
+		}
+		if(name_off > strtab_sz)
+		{
+			asmsh_log_error("Invalid section's name offset");
+			errno=EINVAL;
+			return -1;
+		}
+		if(name_off == 0)
+		{
+			continue;
+		}
+		char *name = strtab + name_off;
+		if(strcmp(name, ".text") != 0)
+		{
+			continue;
+		}
+		unsigned int type;
+		if((rret = read(fd, &type, 4)) != 4)
+		{
+			err=errno;
+			asmsh_log_perror("Unable to read section type");
+			errno = err;
+			return -1;
+		}
+		if(type != 1)
+		{
+			asmsh_log_error("Section .text is not of type SHT_PROGBITS");
+			errno=EINVAL;
+			return -1;
+		}
+
+		if((off = lseek(fd, cur_off + off_section_off, SEEK_SET)) != \
+				cur_off + off_section_off)
+		{
+			err=errno;
+			asmsh_log_perror("Unable to seek at .text section offset");
+			errno = err;
+			return -1;
+		}
+
+		if((rret = read(fd, text_off, 8)) != 8)
+		{
+			err=errno;
+			asmsh_log_perror("Unable to read .text section offset");
+			errno = err;
+			return -1;
+		}
+		if((rret = read(fd, text_sz, 8)) != 8)
+		{
+			err=errno;
+			asmsh_log_perror("Unable to read .text section size");
+			errno = err;
+			return -1;
+		}
+		return 0;
+	}
+
+	asmsh_log_error("Cannot find .text section");
+	return -1;
+}
+
 int asmh_asmc_bytecode_from_obj(const char *objfile, asmsh_bytecode_t *bytecode)
 {
-	/*TODO parse ELF header instead of harcoded addr ?
-	  	at least check if the object file has the expected
-	  	size for the whole file.
-	   Header parsing is better in case the ELF organisation is
-	   platform/version dependent ?
-	*/
+	/*TODO check if the object file has the expected version etc ? */
+
 	// ELF 64 constants
-	const int bytecode_sz_off = 0x138;
-	const int bytecode_sz_sz = 4;
-	const int bytecode_off = 0x40;
+	const int off_section_header_off = 0x28;
+	const int off_section_header_sz = 0x3A;
+	const int off_section_header_count = 0x3C;
+	const int off_section_header_str = 0x3E;
 
 	int fd, err;
 	off_t off;
 	ssize_t rret;
-	unsigned int bytecode_sz;
+	unsigned long int section_header_off;
+	unsigned short section_header_sz, section_header_count, section_header_str;
 
 	if((fd = open(objfile, O_RDONLY)) < 0)
 	{
@@ -519,70 +649,90 @@ int asmh_asmc_bytecode_from_obj(const char *objfile, asmsh_bytecode_t *bytecode)
 		return -1;
 	}
 
-	if((off = lseek(fd, bytecode_sz_off, SEEK_SET)) != bytecode_sz_off)
+	if((off = lseek(fd, off_section_header_off, SEEK_SET)) != off_section_header_off)
 	{
-		/// TODO check off
-		perror("Unable to seek at bytecode size offset");
 		err = errno;
+		asmsh_log_perror("Unable to seek at section header offset");
 		goto err;
 	}
-
-	if((rret = read(fd, &bytecode_sz, bytecode_sz_sz)) != bytecode_sz_sz)
+	if((rret = read(fd, &section_header_off, 8)) != 8)
 	{
 		err = errno;
-		if(rret < 0)
-		{
-			perror("Unable to read bytecode size");
-		}
-		else
-		{
-			dprintf(2, "Unable to read the 4 bytes of the bytecode size %ld read instead\n", rret);
-			err = ENODATA;
-		}
+		asmsh_log_perror("Unable to read section header offset");
 		goto err;
 	}
 
-	if(bytecode_sz == 0)
+	if((off = lseek(fd, off_section_header_sz, SEEK_SET)) != off_section_header_sz)
+	{
+		err = errno;
+		asmsh_log_perror("Unable to seek at section header size");
+		goto err;
+	}
+	if((rret = read(fd, &section_header_sz, 2)) != 2)
 	{
-		asmsh_log_error("Null bytecode size");
-		err = ENODATA;
+		err = errno;
+		asmsh_log_perror("Unable to read section header size");
 		goto err;
 	}
-	if(bytecode_sz > sizeof(bytecode->bytes))
+
+	if((off = lseek(fd, off_section_header_count, SEEK_SET)) != off_section_header_count)
 	{
-		dprintf(2, "Bytecode size invalid, too many bytes (%d)\n",
-				bytecode_sz);
-		err = ENOBUFS;
+		err=errno;
+		asmsh_log_perror("Unable to seek at section header count");
 		goto err;
 	}
-	if((off = lseek(fd, bytecode_off, SEEK_SET)) != bytecode_off)
+	if((rret = read(fd, &section_header_count, 2)) != 2)
 	{
-		/// TODO check off
-		perror("Unable to seek at bytecode offset");
 		err = errno;
+		asmsh_log_perror("Unable to read section header count");
 		goto err;
 	}
 
-	bytecode->size = bytecode_sz;
-	if((rret = read(fd, bytecode->bytes, bytecode_sz)) != bytecode_sz)
+	if((off = lseek(fd, off_section_header_str, SEEK_SET)) != off_section_header_str)
 	{
-		err = errno;
-		if(rret < 0)
-		{
-			perror("Unable to read bytecode");
-		}
-		else
-		{
-			dprintf(2, "Unable to read the %d bytes of bytecode\n",
-					bytecode_sz);
-			err = ENODATA;
-		}
+		err=errno;
+		asmsh_log_perror("Unable to seek at section header str");
+		goto err;
+	}
+	if((rret = read(fd, &section_header_str, 2)) != 2)
+	{
+		err=errno;
+		asmsh_log_perror("Unable to read section header str");
 		goto err;
 	}
 
+	const unsigned long int strtab_section_offset = section_header_off + \
+							section_header_str * section_header_sz;
+	unsigned long int text_off, text_sz;
+	if(text_section_informations(fd, strtab_section_offset, \
+				section_header_off, section_header_count, \
+				section_header_sz, &text_off, &text_sz) < 0)
+	{
+		err=errno;
+		goto err;
+	}
+
+	if(text_sz > sizeof(bytecode->bytes))
+	{
+		asmsh_log_error("Too many bytes (%d) for an instruction", text_sz);
+		err=EINVAL;
+		goto err;
+	}
+	if((off = lseek(fd, text_off, SEEK_SET)) != text_off)
+	{
+		err=errno;
+		asmsh_log_perror("Unable to seek at text section");
+		goto err;
+	}
+	if((rret = read(fd, &bytecode->bytes, text_sz)) != text_sz)
+	{
+		err=errno;
+		asmsh_log_perror("Unable to read text section");
+		goto err;
+	}
+	bytecode->size = text_sz;
 	close(fd);
 	return 0;
-
 err:
 	close(fd);
 	errno = err;

tests/tests_compile.c

@@ -157,6 +157,7 @@ static const compile_sample_t comp_samples[] = {
 	{{"\x31\xc0\0", 2}, "xor %rax, %rax"},
 	{{"\x53\0", 1}, "push %rbx"},
 	{{"\x0f\x05\0", 2}, "syscall"},
+	{{"\x48\xb8\x78\x56\x34\x12\x00\x00\x00\x00", 10}, "movabs $0x12345678, %rax"},
 };
 
 static const int comp_samples_sz = sizeof(comp_samples)/sizeof(*comp_samples);