Merge git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile

* git://git.kernel.org/pub/scm/linux/kernel/git/cmetcalf/linux-tile:
  arch/tile: check kmalloc() result
  arch/tile: catch up on various minor cleanups.
  arch/tile: avoid erroneous error return for PTRACE_POKEUSR.
  tile: set ARCH_KMALLOC_MINALIGN
  tile: remove homegrown L1_CACHE_ALIGN macro
  arch/tile: Miscellaneous cleanup changes.
  arch/tile: Split the icache flush code off to a generic <arch> header.
  arch/tile: Fix bug in support for atomic64_xx() ops.
  arch/tile: Shrink the tile-opcode files considerably.
  arch/tile: Add driver to enable access to the user dynamic network.
  arch/tile: Enable more sophisticated IRQ model for 32-bit chips.
  Move list types from <linux/list.h> to <linux/types.h>.
  Add wait4() back to the set of <asm-generic/unistd.h> syscalls.
  Revert adding some arch-specific signal syscalls to <linux/syscalls.h>.
  arch/tile: Do not use GFP_KERNEL for dma_alloc_coherent(). Feedback from fujita.tomonori@lab.ntt.co.jp.
  arch/tile: core support for Tilera 32-bit chips.
  Fix up the "generic" unistd.h ABI to be more useful.

1 files changed, 63 insertions, 0 deletions

diff --git a/arch/tile/lib/memmove_32.c b/arch/tile/lib/memmove_32.c
new file mode 100644
index 0000000..fd615ae
--- /dev/null
+++ b/arch/tile/lib/memmove_32.c
@@ -0,0 +1,63 @@
+/*
+ * Copyright 2010 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   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, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ */
+
+#include <linux/types.h>
+#include <linux/string.h>
+#include <linux/module.h>
+
+void *memmove(void *dest, const void *src, size_t n)
+{
+	if ((const char *)src >= (char *)dest + n
+	    || (char *)dest >= (const char *)src + n) {
+		/* We found no overlap, so let memcpy do all the heavy
+		 * lifting (prefetching, etc.)
+		 */
+		return memcpy(dest, src, n);
+	}
+
+	if (n != 0) {
+		const uint8_t *in;
+		uint8_t x;
+		uint8_t *out;
+		int stride;
+
+		if (src < dest) {
+			/* copy backwards */
+			in = (const uint8_t *)src + n - 1;
+			out = (uint8_t *)dest + n - 1;
+			stride = -1;
+		} else {
+			/* copy forwards */
+			in = (const uint8_t *)src;
+			out = (uint8_t *)dest;
+			stride = 1;
+		}
+
+		/* Manually software-pipeline this loop. */
+		x = *in;
+		in += stride;
+
+		while (--n != 0) {
+			*out = x;
+			out += stride;
+			x = *in;
+			in += stride;
+		}
+
+		*out = x;
+	}
+
+	return dest;
+}
+EXPORT_SYMBOL(memmove);