cpu: change _cpu()->thread calls to current_thread() to fix smp race

kernel-src/include/kernel/scheduler.h

@@ -132,7 +132,7 @@ typedef struct proc_t {
 
 #include <arch/cpu.h>
 
-#define UMASK(mode) ((mode) & ~_cpu()->thread->proc->umask)
+#define UMASK(mode) ((mode) & ~current_thread()->proc->umask)
 #define PROC_HOLD(v) __atomic_add_fetch(&(v)->refcount, 1, __ATOMIC_SEQ_CST)
 // the acquire of the pid table mutex is done here to prevent a nasty race condition where sched_inactiveproc could be called multiple times
 // due to the pid table itself not holding any references to the process

kernel-src/include/logging.h

@@ -16,7 +16,7 @@ extern mutex_t printf_mutex;
 #endif
 
 #define printf(...) { \
-	if (_cpu()->thread) {\
+	if (current_thread()) {\
 		MUTEX_ACQUIRE(&printf_mutex, false); \
 	} else { \
 		while (MUTEX_TRY(&printf_mutex) == false) CPU_PAUSE(); \

kernel-src/include/x86-64/arch/cpu.h

@@ -56,12 +56,18 @@ static inline uint16_t be_to_cpu_w(uint16_t w) {
 
 cpu_t *_bsp(void);
 
-static inline cpu_t *_cpu() {
+static inline cpu_t *_cpu(void) {
 	cpu_t *cpu;
 	asm volatile ("mov %%gs:8, %%rax" : "=a"(cpu) : : "memory");
 	return cpu;
 }
 
+static inline thread_t *current_thread(void) {
+	thread_t *thread;
+	asm volatile ("mov %%gs:0, %%rax" : "=a"(thread) : : "memory");
+	return thread;
+}
+
 static inline void cpu_set(cpu_t *ptr) {
 	ptr->self = ptr;
 	wrmsr(MSR_GSBASE, (uint64_t)ptr);