Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / crypto / memneq.c
diff --git a/kernel/crypto/memneq.c b/kernel/crypto/memneq.c
new file mode 100644 (file)
index 0000000..afed1bd
--- /dev/null
@@ -0,0 +1,168 @@
+/*
+ * Constant-time equality testing of memory regions.
+ *
+ * Authors:
+ *
+ *   James Yonan <james@openvpn.net>
+ *   Daniel Borkmann <dborkman@redhat.com>
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2013 OpenVPN Technologies, Inc. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ * The full GNU General Public License is included in this distribution
+ * in the file called LICENSE.GPL.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2013 OpenVPN Technologies, Inc. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ *   * Redistributions of source code must retain the above copyright
+ *     notice, this list of conditions and the following disclaimer.
+ *   * Redistributions in binary form must reproduce the above copyright
+ *     notice, this list of conditions and the following disclaimer in
+ *     the documentation and/or other materials provided with the
+ *     distribution.
+ *   * Neither the name of OpenVPN Technologies nor the names of its
+ *     contributors may be used to endorse or promote products derived
+ *     from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <crypto/algapi.h>
+
+#ifndef __HAVE_ARCH_CRYPTO_MEMNEQ
+
+/* Generic path for arbitrary size */
+static inline unsigned long
+__crypto_memneq_generic(const void *a, const void *b, size_t size)
+{
+       unsigned long neq = 0;
+
+#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS)
+       while (size >= sizeof(unsigned long)) {
+               neq |= *(unsigned long *)a ^ *(unsigned long *)b;
+               OPTIMIZER_HIDE_VAR(neq);
+               a += sizeof(unsigned long);
+               b += sizeof(unsigned long);
+               size -= sizeof(unsigned long);
+       }
+#endif /* CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS */
+       while (size > 0) {
+               neq |= *(unsigned char *)a ^ *(unsigned char *)b;
+               OPTIMIZER_HIDE_VAR(neq);
+               a += 1;
+               b += 1;
+               size -= 1;
+       }
+       return neq;
+}
+
+/* Loop-free fast-path for frequently used 16-byte size */
+static inline unsigned long __crypto_memneq_16(const void *a, const void *b)
+{
+       unsigned long neq = 0;
+
+#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+       if (sizeof(unsigned long) == 8) {
+               neq |= *(unsigned long *)(a)   ^ *(unsigned long *)(b);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned long *)(a+8) ^ *(unsigned long *)(b+8);
+               OPTIMIZER_HIDE_VAR(neq);
+       } else if (sizeof(unsigned int) == 4) {
+               neq |= *(unsigned int *)(a)    ^ *(unsigned int *)(b);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned int *)(a+4)  ^ *(unsigned int *)(b+4);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned int *)(a+8)  ^ *(unsigned int *)(b+8);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned int *)(a+12) ^ *(unsigned int *)(b+12);
+               OPTIMIZER_HIDE_VAR(neq);
+       } else
+#endif /* CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS */
+       {
+               neq |= *(unsigned char *)(a)    ^ *(unsigned char *)(b);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+1)  ^ *(unsigned char *)(b+1);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+2)  ^ *(unsigned char *)(b+2);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+3)  ^ *(unsigned char *)(b+3);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+4)  ^ *(unsigned char *)(b+4);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+5)  ^ *(unsigned char *)(b+5);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+6)  ^ *(unsigned char *)(b+6);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+7)  ^ *(unsigned char *)(b+7);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+8)  ^ *(unsigned char *)(b+8);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+9)  ^ *(unsigned char *)(b+9);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+10) ^ *(unsigned char *)(b+10);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+11) ^ *(unsigned char *)(b+11);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+12) ^ *(unsigned char *)(b+12);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+13) ^ *(unsigned char *)(b+13);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+14) ^ *(unsigned char *)(b+14);
+               OPTIMIZER_HIDE_VAR(neq);
+               neq |= *(unsigned char *)(a+15) ^ *(unsigned char *)(b+15);
+               OPTIMIZER_HIDE_VAR(neq);
+       }
+
+       return neq;
+}
+
+/* Compare two areas of memory without leaking timing information,
+ * and with special optimizations for common sizes.  Users should
+ * not call this function directly, but should instead use
+ * crypto_memneq defined in crypto/algapi.h.
+ */
+noinline unsigned long __crypto_memneq(const void *a, const void *b,
+                                      size_t size)
+{
+       switch (size) {
+       case 16:
+               return __crypto_memneq_16(a, b);
+       default:
+               return __crypto_memneq_generic(a, b, size);
+       }
+}
+EXPORT_SYMBOL(__crypto_memneq);
+
+#endif /* __HAVE_ARCH_CRYPTO_MEMNEQ */