1 files changed, 113 insertions, 0 deletions

diff --git a/arch/arm/mm/copypage-xscale.S b/arch/arm/mm/copypage-xscale.S
new file mode 100644
index 000000000000..bb277316ef52
--- /dev/null
+++ b/arch/arm/mm/copypage-xscale.S
@@ -0,0 +1,113 @@
+/*
+ *  linux/arch/arm/lib/copypage-xscale.S
+ *
+ *  Copyright (C) 2001 Russell King
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+#include <linux/linkage.h>
+#include <linux/init.h>
+#include <asm/constants.h>
+
+/*
+ * General note:
+ *  We don't really want write-allocate cache behaviour for these functions
+ *  since that will just eat through 8K of the cache.
+ */
+
+        .text
+        .align  5
+/*
+ * XScale optimised copy_user_page
+ *  r0 = destination
+ *  r1 = source
+ *  r2 = virtual user address of ultimate destination page
+ *
+ * The source page may have some clean entries in the cache already, but we
+ * can safely ignore them - break_cow() will flush them out of the cache
+ * if we eventually end up using our copied page.
+ *
+ * What we could do is use the mini-cache to buffer reads from the source
+ * page.  We rely on the mini-cache being smaller than one page, so we'll
+ * cycle through the complete cache anyway.
+ */
+ENTRY(xscale_mc_copy_user_page)
+        stmfd   sp!, {r4, r5, lr}
+        mov     r5, r0
+        mov     r0, r1
+        bl      map_page_minicache
+        mov     r1, r5
+        mov     lr, #PAGE_SZ/64-1
+
+        /*
+         * Strangely enough, best performance is achieved
+         * when prefetching destination as well.  (NP)
+         */
+        pld     [r0, #0]
+        pld     [r0, #32]
+        pld     [r1, #0]
+        pld     [r1, #32]
+
+1:      pld     [r0, #64]
+        pld     [r0, #96]
+        pld     [r1, #64]
+        pld     [r1, #96]
+
+2:      ldrd    r2, [r0], #8
+        ldrd    r4, [r0], #8
+        mov     ip, r1
+        strd    r2, [r1], #8
+        ldrd    r2, [r0], #8
+        strd    r4, [r1], #8
+        ldrd    r4, [r0], #8
+        strd    r2, [r1], #8
+        strd    r4, [r1], #8
+        mcr     p15, 0, ip, c7, c10, 1          @ clean D line
+        ldrd    r2, [r0], #8
+        mcr     p15, 0, ip, c7, c6, 1           @ invalidate D line
+        ldrd    r4, [r0], #8
+        mov     ip, r1
+        strd    r2, [r1], #8
+        ldrd    r2, [r0], #8
+        strd    r4, [r1], #8
+        ldrd    r4, [r0], #8
+        strd    r2, [r1], #8
+        strd    r4, [r1], #8
+        mcr     p15, 0, ip, c7, c10, 1          @ clean D line
+        subs    lr, lr, #1
+        mcr     p15, 0, ip, c7, c6, 1           @ invalidate D line
+        bgt     1b
+        beq     2b
+
+        ldmfd   sp!, {r4, r5, pc}
+
+        .align  5
+/*
+ * XScale optimised clear_user_page
+ *  r0 = destination
+ *  r1 = virtual user address of ultimate destination page
+ */
+ENTRY(xscale_mc_clear_user_page)
+        mov     r1, #PAGE_SZ/32
+        mov     r2, #0
+        mov     r3, #0
+1:      mov     ip, r0
+        strd    r2, [r0], #8
+        strd    r2, [r0], #8
+        strd    r2, [r0], #8
+        strd    r2, [r0], #8
+        mcr     p15, 0, ip, c7, c10, 1          @ clean D line
+        subs    r1, r1, #1
+        mcr     p15, 0, ip, c7, c6, 1           @ invalidate D line
+        bne     1b
+        mov     pc, lr
+
+        __INITDATA
+
+        .type   xscale_mc_user_fns, #object
+ENTRY(xscale_mc_user_fns)
+        .long   xscale_mc_clear_user_page
+        .long   xscale_mc_copy_user_page
+        .size   xscale_mc_user_fns, . - xscale_mc_user_fns