Even the next up thread should sleep in the futex

Former-commit-id: 8b76b01bfe710603bcdc101da6eb27afcee7e1b1

src/fastlock.cpp

@@ -120,13 +120,10 @@ extern "C" void fastlock_lock(struct fastlock *lock)
         __asm__ ("pause");
 #endif
         if ((++cloops % 1024*1024) == 0)
-        {
-            if (static_cast<uint16_t>(ticketT.m_active+1U) != myticket)
         {
             __atomic_fetch_or(&lock->futex, mask, __ATOMIC_ACQUIRE);
             futex(&lock->m_ticket.u, FUTEX_WAIT_BITSET_PRIVATE, ticketT.u, nullptr, mask);
             __atomic_fetch_and(&lock->futex, ~mask, __ATOMIC_RELEASE);
-            }
             ++g_longwaits;
         }
     }

src/fastlock_x64.asm

@@ -44,11 +44,7 @@ ALIGN 16
 	;	But the compiler doesn't know that we rarely hit this, and when we do we know the lock is
 	;	taking a long time to be released anyways.  We optimize for the common case of short
 	;	lock intervals.  That's why we're using a spinlock in the first place
-	inc edx
-	cmp dx, ax
-	je .LLoop
-	dec edx					; restore the current ticket
-.LFutexWait:
+	; If we get here we're going to sleep in the kernel with a futex
 	push rsi
 	push rax
 	; Setup the syscall args