From d7b6df5808e7bef5930b61a82e880699a9f9e208 Mon Sep 17 00:00:00 2001
From: Khem Raj <raj.khem@gmail.com>
Date: Thu, 29 Jun 2017 15:39:19 -0700
Subject: [PATCH] implement 64bit atomic for mips

GCC does not provide 64bit atomics for mips32

Signed-off-by: Khem Raj <raj.khem@gmail.com>
---
 src/Makefile.am |   1 +
 src/atomic64.c  | 228 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 229 insertions(+)
 create mode 100644 src/atomic64.c

diff --git a/src/Makefile.am b/src/Makefile.am
index 99aaace0..cbbbbee9 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -27,6 +27,7 @@ libtorrent_la_LIBADD = \
 	utils/libsub_utils.la
 
 libtorrent_la_SOURCES = \
+	atomic64.c \
 	globals.cc \
 	globals.h \
 	manager.cc \
diff --git a/src/atomic64.c b/src/atomic64.c
new file mode 100644
index 00000000..f841b39b
--- /dev/null
+++ b/src/atomic64.c
@@ -0,0 +1,228 @@
+/*===----- atomic64.c - Support functions for 64-bit atomic operations.-----===
+ *
+ *                     The LLVM Compiler Infrastructure
+ *
+ * This file is dual licensed under the MIT and the University of Illinois Open
+ * Source Licenses. See LICENSE.TXT for details.
+ *
+ *===-----------------------------------------------------------------------===
+ *
+ *  atomic64.c defines a set of functions for performing atomic accesses on
+ *  64-bit memory locations. It also implements spinlock synchronization
+ *  operations.
+ *
+ *===-----------------------------------------------------------------------===
+ */
+
+#include <stdint.h>
+#include <stdbool.h>
+
+/*
+ * only need these on MIPS, since it lacks hardware 64-bit atomics,
+ * unlike x86 and ARM.
+ */
+#if defined(__mips__) || defined(__mipsel__)
+
+static void __spin_lock(volatile int *lock) {
+  while (__sync_lock_test_and_set(lock, 1))
+    while (*lock) {}
+}
+
+static void __spin_unlock(volatile int *lock) {
+  __sync_lock_release(lock);
+}
+
+/*
+ * Make sure the lock is on its own cache line to prevent false sharing.
+ * Put it inside a struct that is aligned and padded to the typical MIPS
+ * cacheline which is 32 bytes.
+ */
+static struct {
+  int lock;
+  char pad[32 - sizeof(int)];
+} __attribute__((aligned (32))) lock = { 0 };
+
+
+uint64_t __sync_fetch_and_add_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr;
+  *ptr = ret + val;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_fetch_and_sub_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr;
+  *ptr = ret - val;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_fetch_and_and_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr;
+  *ptr = ret & val;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_fetch_and_or_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr;
+  *ptr = ret | val;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_fetch_and_xor_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr;
+  *ptr = ret ^ val;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_add_and_fetch_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr + val;
+  *ptr = ret;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_sub_and_fetch_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr - val;
+  *ptr = ret;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_and_and_fetch_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr & val;
+  *ptr = ret;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_or_and_fetch_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr | val;
+  *ptr = ret;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_xor_and_fetch_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr ^ val;
+  *ptr = ret;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+bool __sync_bool_compare_and_swap_8(volatile uint64_t *ptr,
+                                         uint64_t oldval, uint64_t newval) {
+  bool ret = false;
+
+  __spin_lock(&lock.lock);
+
+  if (*ptr == oldval) {
+    *ptr = newval;
+    ret = true;
+  }
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_val_compare_and_swap_8(volatile uint64_t *ptr,
+                                            uint64_t oldval, uint64_t newval) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr;
+  if (ret == oldval)
+    *ptr = newval;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+uint64_t __sync_lock_test_and_set_8(volatile uint64_t *ptr, uint64_t val) {
+  uint64_t ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr;
+  *ptr = val;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+void __sync_lock_release_8(volatile uint64_t *ptr) {
+  __spin_lock(&lock.lock);
+
+  *ptr = 0;
+
+  __spin_unlock(&lock.lock);
+}
+
+#endif
-- 
2.13.2