Switched to -Os flag and fixed issues with it

Somehow the insl function gets optimized to one instruction so I need
a gcc attribute to exclude this function from optimization. I may fix
this in the future though. Anyways, the kernel is waay smaller now! :)

4 files changed, 16 insertions, 14 deletions

diff --git a/Makefile b/Makefile
index 314c894..4b97fcc 100644
--- a/Makefile
+++ b/Makefile
@@ -22,17 +22,17 @@ LD = cross/opt/bin/i686-elf-ld
 AS = nasm
 
 # Flags to make the binary smaller TODO: Remove after indirect pointer support!
-CSFLAGS = -fno-stack-protector -fomit-frame-pointer -ffunction-sections -fdata-sections -Wl,--gc-sections -mpreferred-stack-boundary=2 -falign-functions=1 -falign-jumps=1 -falign-loops=1 -fno-unwind-tables -fno-asynchronous-unwind-tables -fno-math-errno -fno-unroll-loops -fmerge-all-constants -fno-ident -ffast-math
+CSFLAGS = -mpreferred-stack-boundary=2 -fno-asynchronous-unwind-tables -Os
 
 # TODO: Use lib as external library
-CFLAGS = $(CSFLAGS) -Wall -Wextra -nostdlib -nostdinc -ffreestanding -fno-builtin -fno-pic -mgeneral-regs-only -std=c99 -m32 -pedantic-errors -Isrc/lib/inc/ -Isrc/inc/ -c
+CFLAGS = $(CSFLAGS) -Wall -Wextra -nostdlib -nostdinc -ffreestanding -fno-builtin -fno-pic -mgeneral-regs-only -std=c99 -m32 -pedantic-errors -Isrc/lib/inc/ -Isrc/inc/
 
 ASFLAGS = -f elf32
 
 all: compile clean
 
 %.o: %.c
-	@$(CC) $(CFLAGS) $< -o $@
+	@$(CC) -c $(CFLAGS) $< -o $@
 
 %_asm.o: %.asm
 	@$(AS) $(ASFLAGS) $< -o $@
@@ -43,6 +43,7 @@ compile: kernel
 	@mkdir -p build/
 	@$(AS) -f bin src/entry.asm -o build/boot.bin
 	@$(LD) -N -emain -Ttext 0x00050000 -o build/kernel.bin $(COBJS) --oformat binary
+	@$(CC) $(CFLAGS) -emain -o build/debug.o $(COBJS)
 
 clean:
 	@find src/ -name "*.o" -type f -delete
diff --git a/run b/run
index 687ac46..3774f4d 100755
--- a/run
+++ b/run
@@ -112,8 +112,8 @@ make_debug() {
 }
 
 make_disasm() {
-    objdump -drwC -Mintel build/melvix.bin --visualize-jumps=color | less -R
-    #hexdump -C build/melvix.bin | less -R
+    objdump -drwC -Mintel build/debug.o --visualize-jumps=color | less -R
+    #hexdump -C build/kernel.bin | less -R
 }
 
 make_sync() {
diff --git a/src/drivers/cpu.c b/src/drivers/cpu.c
index d6a1760..945dc4a 100644
--- a/src/drivers/cpu.c
+++ b/src/drivers/cpu.c
@@ -24,7 +24,8 @@ u32 inl(u16 port)
 	return value;
 }
 
-void insl(u16 port, void *addr, int n)
+// TODO: Fix optimization issues with insl
+void __attribute__((optimize("O0"))) insl(u16 port, void *addr, int n)
 {
 	__asm__("cld; rep insl"
 		: "=D"(addr), "=c"(n)
diff --git a/src/features/fs.c b/src/features/fs.c
index 7795210..1eec30e 100644
--- a/src/features/fs.c
+++ b/src/features/fs.c
@@ -1,6 +1,7 @@
 // MIT License, Copyright (c) 2020 Marvin Borner
 // EXT2 based filesystem
 
+#include <assert.h>
 #include <def.h>
 #include <fs.h>
 #include <ide.h>
@@ -29,7 +30,9 @@ struct bgd *get_bgd()
 struct inode *get_inode(int i)
 {
 	struct superblock *s = get_superblock();
+	assert(s);
 	struct bgd *b = get_bgd();
+	assert(b);
 
 	int block_group = (i - 1) / s->inodes_per_group;
 	int index = (i - 1) % s->inodes_per_group;
@@ -50,28 +53,23 @@ u32 read_indirect(u32 indirect, u32 block_num)
 
 void *read_inode(struct inode *in)
 {
-	//assert(in);
+	assert(in);
 	if (!in)
 		return NULL;
 
 	int num_blocks = in->blocks / (BLOCK_SIZE / SECTOR_SIZE);
 
-	//assert(num_blocks != 0);
+	assert(num_blocks != 0);
 	if (!num_blocks)
 		return NULL;
 
 	u32 sz = BLOCK_SIZE * num_blocks;
 	void *buf = malloc(sz);
 	printf("Loading %dKiB\n", sz >> 10);
-	//assert(buf != NULL);
+	assert(buf != NULL);
 
 	int indirect;
 
-	// Singly indirect pointer
-	// TODO: Support doubly and triply pointers
-	if (num_blocks > 12)
-		indirect = in->block[12];
-
 	int blocknum = 0;
 	char *data;
 	for (int i = 0; i < num_blocks; i++) {
@@ -80,6 +78,8 @@ void *read_inode(struct inode *in)
 			data = buffer_read(blocknum);
 			memcpy((u32 *)((u32)buf + i * BLOCK_SIZE), data, BLOCK_SIZE);
 		} else {
+			// TODO: Support doubly and triply pointers
+			indirect = in->block[12];
 			blocknum = read_indirect(indirect, i - 12);
 			data = buffer_read(blocknum);
 			memcpy((u32 *)((u32)buf + (i - 1) * BLOCK_SIZE), data, BLOCK_SIZE);