add vorgabe10

c_os/lib/SpinLock.cc

@@ -0,0 +1,63 @@
+/*****************************************************************************
+ *                                                                           *
+ *                              S P I N L O C K                              *
+ *                                                                           *
+ *---------------------------------------------------------------------------*
+ * Beschreibung:    Implementierung eines Spinlocks mithilfe der cmpxchg     *
+ *                  Instruktion.                                             *
+ *                                                                           *
+ * Autor:           Michael Schoettner, 2.2.2018                             *
+ *****************************************************************************/
+
+#include "lib/SpinLock.h"
+#include "kernel/Globals.h"
+
+/*****************************************************************************
+ * Methode:         CAS                                                      *
+ *---------------------------------------------------------------------------*
+ * Parameter:       *ptr    Adresse der Variable des Locks                   *
+ *                   old    Wert gegen den verglichen wird                   *
+ *                  _new    Wert der gesetzt werden soll                     *
+ *                                                                           *
+ * Beschreibung:    Semantik der Funktion CAS = Cmompare & Swap:             *
+ *                      if old == *ptr then                                  *
+ *                          *ptr := _new                                     *
+ *                      return prev                                          *
+ *****************************************************************************/
+static inline unsigned long CAS(unsigned long* ptr, unsigned long old, unsigned long _new) {
+    unsigned long prev;
+
+    /*
+        AT&T/UNIX assembly syntax
+
+        The 'volatile' keyword after 'asm' indicates that the instruction 
+        has important side-effects. GCC will not delete a volatile asm if 
+        sit is reachable.
+     */
+    asm volatile("lock;"                           // prevent race conditions with other cores
+                 "cmpxchg %1, %2;"                 // %1 = _new; %2 = *ptr
+                                                   // constraints
+                 : "=a"(prev)                      // output: =a: RAX -> prev (%0))
+                 : "r"(_new), "m"(*ptr), "a"(old)  // input = %1, %2, %3 (r=register, m=memory, a=accumlator = eax
+                 : "memory");                      // ensures assembly block will not be moved by gcc
+
+    return prev;  // return pointer instead of prev to prevent unnecessary second call
+}
+
+/*****************************************************************************
+ * Methode:         SpinLock::acquire                                        *
+ *---------------------------------------------------------------------------*
+ * Beschreibung:    Lock belegen.                                            *
+ *****************************************************************************/
+void SpinLock::acquire() {
+    while (CAS(ptr, 0, 1) != 0) {}
+}
+
+/*****************************************************************************
+ * Methode:         SpinLock::release                                        *
+ *---------------------------------------------------------------------------*
+ * Beschreibung:    Lock freigeben.                                          *
+ *****************************************************************************/
+void SpinLock::release() {
+    lock = 0;
+}