Refactor some code

- Write tests for cthreads
- Fix bugs in pe2.com tool
- Fix ASAN issue with GetDosEnviron()
- Consolidate the cthread header files
- Some code size optimizations for MODE=
- Attempted to squash a tls linker warning
- Attempted to get futexes working on FreeBSD

libc/runtime/clone.c

@@ -468,7 +468,8 @@ int sys_clone_linux(int flags, char *stk, int *ptid, int *ctid, void *tls,
  * effectively work around libc features like atfork(), so that means
  * other calls like getpid() may return incorrect values.
  *
- * @param func is your callback function
+ * @param func is your callback function, which this wrapper requires
+ *     not be null, otherwise EINVAL is raised
  * @param stk points to the bottom of a caller allocated stack, which
  *     must be allocated via mmap() using the MAP_STACK flag, or else
  *     you won't get optimal performance and it won't work on OpenBSD
@@ -532,20 +533,22 @@ int clone(int (*func)(void *), void *stk, size_t stksz, int flags, void *arg,
     __threaded = gettid();
   }
 
-  if (IsAsan() &&
-      ((stksz > PAGESIZE &&
-        !__asan_is_valid((char *)stk + PAGESIZE, stksz - PAGESIZE)) ||
-       ((flags & CLONE_SETTLS) && !__asan_is_valid(tls, tlssz)) ||
-       ((flags & CLONE_SETTLS) && !__asan_is_valid(tls, sizeof(long))) ||
-       ((flags & CLONE_PARENT_SETTID) &&
-        !__asan_is_valid(ptid, sizeof(*ptid))) ||
-       ((flags & CLONE_CHILD_SETTID) &&
-        !__asan_is_valid(ctid, sizeof(*ctid))))) {
-    rc = efault();
+  if (!func) {
+    rc = einval();
   } else if (!IsTiny() &&
              (((flags & CLONE_VM) && (stksz < PAGESIZE || (stksz & 15))) ||
               ((flags & CLONE_SETTLS) && (tlssz < 64 || (tlssz & 7))))) {
     rc = einval();
+  } else if (IsAsan() &&
+             ((stksz > PAGESIZE &&
+               !__asan_is_valid((char *)stk + PAGESIZE, stksz - PAGESIZE)) ||
+              ((flags & CLONE_SETTLS) && !__asan_is_valid(tls, tlssz)) ||
+              ((flags & CLONE_SETTLS) && !__asan_is_valid(tls, sizeof(long))) ||
+              ((flags & CLONE_PARENT_SETTID) &&
+               !__asan_is_valid(ptid, sizeof(*ptid))) ||
+              ((flags & CLONE_CHILD_SETTID) &&
+               !__asan_is_valid(ctid, sizeof(*ctid))))) {
+    rc = efault();
   } else if (IsLinux()) {
     rc =
         sys_clone_linux(flags, (char *)stk + stksz, ptid, ctid, tls, func, arg);

libc/runtime/getdosenviron.c

@@ -125,7 +125,7 @@ textwindows noasan noinstrument int GetDosEnviron(const char16_t *env,
   while (*env) {
     if (i + 1 < max) envp[i++] = buf;
     r = Recode16to8(buf, size, env);
-    if ((p = memchr(buf, '=', r.ax)) && IsAlpha(p[1]) && p[2] == ':' &&
+    if ((p = MemChr(buf, '=', r.ax)) && IsAlpha(p[1]) && p[2] == ':' &&
         (p[3] == '\\' || p[3] == '/')) {
       FixPath(p + 1);
     }

libc/runtime/pthread.c

@@ -16,29 +16,56 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
+#include "libc/bits/atomic.h"
+#include "libc/calls/calls.h"
 #include "libc/errno.h"
 #include "libc/intrin/lockcmpxchg.h"
+#include "libc/intrin/spinlock.h"
 #include "libc/runtime/runtime.h"
 
 typedef void *pthread_t;
-typedef bool pthread_once_t;
+typedef int pthread_once_t;
 typedef int pthread_mutex_t;
 
 int pthread_once(pthread_once_t *once, void init(void)) {
-  if (_lockcmpxchg(once, 0, 1)) init();
+  int x;
+  unsigned tries;
+  switch ((x = atomic_load(once))) {
+    case 0:
+      if (atomic_compare_exchange_strong(once, &x, 1)) {
+        init();
+        atomic_store(once, 2);
+        break;
+      }
+      // fallthrough
+    case 1:
+      tries = 0;
+      do {
+        if (++tries & 7) {
+          __builtin_ia32_pause();
+        } else {
+          sched_yield();
+        }
+      } while (atomic_load(once) == 1);
+      break;
+    default:
+      break;
+  }
   return 0;
 }
 
 int pthread_mutex_lock(pthread_mutex_t *mutex) {
-  return EINVAL;
+  _spinlock(mutex);
+  return 0;
 }
 
 int pthread_mutex_trylock(pthread_mutex_t *mutex) {
-  return EINVAL;
+  return _trylock(mutex);
 }
 
 int pthread_mutex_unlock(pthread_mutex_t *mutex) {
-  return EPERM;
+  _spunlock(mutex);
+  return 0;
 }
 
 int pthread_cancel(pthread_t thread) {