Kernel: Fix 32 bit target compile and runtime

Apparently I have to reload stack in the fork trampoline. Not sure why
or why not on x86_64. Also sse builtins did not compile

kernel/arch/i686/Syscall.S

@@ -63,7 +63,7 @@ sys_fork_trampoline:
 
 	call read_ip
 	testl %eax, %eax
-	jz .done
+	jz .reload_stack
 
 	movl %esp, %ebx
 
@@ -79,3 +79,9 @@ sys_fork_trampoline:
 	popl %ebx
 	popl %ebp
 	ret
+
+ .reload_stack:
+	call get_thread_start_sp
+	movl %eax, %esp
+	xorl %eax, %eax
+	jmp .done

kernel/arch/i686/interrupts.S

@@ -6,7 +6,6 @@
 	movw %ax, %ds
 	movw %ax, %es
 	movw %ax, %fs
-
 	movw $0x28, %ax
 	movw %ax, %gs
 1:

kernel/kernel/Processor.cpp

@@ -102,7 +102,7 @@ namespace Kernel
 			asm volatile("wrmsr" :: "d"(val_hi), "a"(val_lo), "c"(MSR_IA32_GS_BASE));
 		}
 #elif ARCH(i686)
-		asm volatile("movw %0, %%gs" :: "r"(0x28));
+		asm volatile("movw %0, %%gs" :: "r"(static_cast<uint16_t>(0x28)));
 #endif
 
 #if ARCH(x86_64)

kernel/kernel/Thread.cpp

@@ -910,7 +910,8 @@ namespace Kernel
 #if ARCH(x86_64)
 		__builtin_ia32_fxsave64(m_sse_storage);
 #elif ARCH(i686)
-		__builtin_ia32_fxsave(m_sse_storage);
+		// no idea why the builtin don't work
+		asm volatile("fxsave %0" :: "m"(m_sse_storage));
 #else
 #error
 #endif
@@ -921,7 +922,8 @@ namespace Kernel
 #if ARCH(x86_64)
 		__builtin_ia32_fxrstor64(m_sse_storage);
 #elif ARCH(i686)
-		__builtin_ia32_fxrstor(m_sse_storage);
+		// no idea why the builtin don't work
+		asm volatile("fxrstor %0" :: "m"(m_sse_storage));
 #else
 #error
 #endif