[MIPS] Reimplement clear_page/copy_page

Fold the SB-1 specific implementation of clear_page/copy_page in the
generic version, and rewrite that one in tlbex style. The immediate
benefits:
  - It converts the compile-time workaround for SB-1 pass 1 prefetches
    to a more efficient run-time check.
  - It allows adjustment of loop unfolling, which helps to reduce the
    number of redundant cdex cache ops.
  - It fixes some esoteric cornercases (the cache line length calculations
    can go wrong, and support for 64k pages without prefetch instructions
    will overflow the addiu immediate).
  - Somewhat better guesses of "good" prefetch values.

Signed-off-by: Thiemo Seufer <ths@networkno.de>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

1 files changed, 684 insertions, 0 deletions

diff --git a/arch/mips/mm/page.c b/arch/mips/mm/page.c
new file mode 100644
index 000000000000..d827d6144369
--- /dev/null
+++ b/arch/mips/mm/page.c
@@ -0,0 +1,684 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2003, 04, 05 Ralf Baechle (ralf@linux-mips.org)
+ * Copyright (C) 2007  Maciej W. Rozycki
+ * Copyright (C) 2008  Thiemo Seufer
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/proc_fs.h>
+
+#include <asm/bugs.h>
+#include <asm/cacheops.h>
+#include <asm/inst.h>
+#include <asm/io.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/prefetch.h>
+#include <asm/system.h>
+#include <asm/bootinfo.h>
+#include <asm/mipsregs.h>
+#include <asm/mmu_context.h>
+#include <asm/cpu.h>
+#include <asm/war.h>
+
+#ifdef CONFIG_SIBYTE_DMA_PAGEOPS
+#include <asm/sibyte/sb1250.h>
+#include <asm/sibyte/sb1250_regs.h>
+#include <asm/sibyte/sb1250_dma.h>
+#endif
+
+#include "uasm.h"
+
+/* Registers used in the assembled routines. */
+#define ZERO 0
+#define AT 2
+#define A0 4
+#define A1 5
+#define A2 6
+#define T0 8
+#define T1 9
+#define T2 10
+#define T3 11
+#define T9 25
+#define RA 31
+
+/* Handle labels (which must be positive integers). */
+enum label_id {
+	label_clear_nopref = 1,
+	label_clear_pref,
+	label_copy_nopref,
+	label_copy_pref_both,
+	label_copy_pref_store,
+};
+
+UASM_L_LA(_clear_nopref)
+UASM_L_LA(_clear_pref)
+UASM_L_LA(_copy_nopref)
+UASM_L_LA(_copy_pref_both)
+UASM_L_LA(_copy_pref_store)
+
+/* We need one branch and therefore one relocation per target label. */
+static struct uasm_label __cpuinitdata labels[5];
+static struct uasm_reloc __cpuinitdata relocs[5];
+
+#define cpu_is_r4600_v1_x()	((read_c0_prid() & 0xfffffff0) == 0x00002010)
+#define cpu_is_r4600_v2_x()	((read_c0_prid() & 0xfffffff0) == 0x00002020)
+
+/*
+ * Maximum sizes:
+ *
+ * R4000 128 bytes S-cache:		0x058 bytes
+ * R4600 v1.7:				0x05c bytes
+ * R4600 v2.0:				0x060 bytes
+ * With prefetching, 16 word strides	0x120 bytes
+ */
+
+static u32 clear_page_array[0x120 / 4];
+
+#ifdef CONFIG_SIBYTE_DMA_PAGEOPS
+void clear_page_cpu(void *page) __attribute__((alias("clear_page_array")));
+#else
+void clear_page(void *page) __attribute__((alias("clear_page_array")));
+#endif
+
+EXPORT_SYMBOL(clear_page);
+
+/*
+ * Maximum sizes:
+ *
+ * R4000 128 bytes S-cache:		0x11c bytes
+ * R4600 v1.7:				0x080 bytes
+ * R4600 v2.0:				0x07c bytes
+ * With prefetching, 16 word strides	0x540 bytes
+ */
+static u32 copy_page_array[0x540 / 4];
+
+#ifdef CONFIG_SIBYTE_DMA_PAGEOPS
+void
+copy_page_cpu(void *to, void *from) __attribute__((alias("copy_page_array")));
+#else
+void copy_page(void *to, void *from) __attribute__((alias("copy_page_array")));
+#endif
+
+EXPORT_SYMBOL(copy_page);
+
+
+static int pref_bias_clear_store __cpuinitdata;
+static int pref_bias_copy_load __cpuinitdata;
+static int pref_bias_copy_store __cpuinitdata;
+
+static u32 pref_src_mode __cpuinitdata;
+static u32 pref_dst_mode __cpuinitdata;
+
+static int clear_word_size __cpuinitdata;
+static int copy_word_size __cpuinitdata;
+
+static int half_clear_loop_size __cpuinitdata;
+static int half_copy_loop_size __cpuinitdata;
+
+static int cache_line_size __cpuinitdata;
+#define cache_line_mask() (cache_line_size - 1)
+
+static inline void __cpuinit
+pg_addiu(u32 **buf, unsigned int reg1, unsigned int reg2, unsigned int off)
+{
+	if (cpu_has_64bit_gp_regs && DADDI_WAR && r4k_daddiu_bug()) {
+		if (off > 0x7fff) {
+			uasm_i_lui(buf, T9, uasm_rel_hi(off));
+			uasm_i_addiu(buf, T9, T9, uasm_rel_lo(off));
+		} else
+			uasm_i_addiu(buf, T9, ZERO, off);
+		uasm_i_daddu(buf, reg1, reg2, T9);
+	} else {
+		if (off > 0x7fff) {
+			uasm_i_lui(buf, T9, uasm_rel_hi(off));
+			uasm_i_addiu(buf, T9, T9, uasm_rel_lo(off));
+			UASM_i_ADDU(buf, reg1, reg2, T9);
+		} else
+			UASM_i_ADDIU(buf, reg1, reg2, off);
+	}
+}
+
+static void __cpuinit set_prefetch_parameters(void)
+{
+	if (cpu_has_64bit_gp_regs || cpu_has_64bit_zero_reg)
+		clear_word_size = 8;
+	else
+		clear_word_size = 4;
+
+	if (cpu_has_64bit_gp_regs)
+		copy_word_size = 8;
+	else
+		copy_word_size = 4;
+
+	/*
+	 * The pref's used here are using "streaming" hints, which cause the
+	 * copied data to be kicked out of the cache sooner.  A page copy often
+	 * ends up copying a lot more data than is commonly used, so this seems
+	 * to make sense in terms of reducing cache pollution, but I've no real
+	 * performance data to back this up.
+	 */
+	if (cpu_has_prefetch) {
+		/*
+		 * XXX: Most prefetch bias values in here are based on
+		 * guesswork.
+		 */
+		cache_line_size = cpu_dcache_line_size();
+		switch (current_cpu_type()) {
+		case CPU_TX49XX:
+			/* TX49 supports only Pref_Load */
+			pref_bias_copy_load = 256;
+			break;
+
+		case CPU_RM9000:
+			/*
+			 * As a workaround for erratum G105 which make the
+			 * PrepareForStore hint unusable we fall back to
+			 * StoreRetained on the RM9000.  Once it is known which
+			 * versions of the RM9000 we'll be able to condition-
+			 * alize this.
+			 */
+
+		case CPU_R10000:
+		case CPU_R12000:
+		case CPU_R14000:
+			/*
+			 * Those values have been experimentally tuned for an
+			 * Origin 200.
+			 */
+			pref_bias_clear_store = 512;
+			pref_bias_copy_load = 256;
+			pref_bias_copy_store = 256;
+			pref_src_mode = Pref_LoadStreamed;
+			pref_dst_mode = Pref_StoreStreamed;
+			break;
+
+		case CPU_SB1:
+		case CPU_SB1A:
+			pref_bias_clear_store = 128;
+			pref_bias_copy_load = 128;
+			pref_bias_copy_store = 128;
+			/*
+			 * SB1 pass1 Pref_LoadStreamed/Pref_StoreStreamed
+			 * hints are broken.
+			 */
+			if (current_cpu_type() == CPU_SB1 &&
+			    (current_cpu_data.processor_id & 0xff) < 0x02) {
+				pref_src_mode = Pref_Load;
+				pref_dst_mode = Pref_Store;
+			} else {
+				pref_src_mode = Pref_LoadStreamed;
+				pref_dst_mode = Pref_StoreStreamed;
+			}
+			break;
+
+		default:
+			pref_bias_clear_store = 128;
+			pref_bias_copy_load = 256;
+			pref_bias_copy_store = 128;
+			pref_src_mode = Pref_LoadStreamed;
+			pref_dst_mode = Pref_PrepareForStore;
+			break;
+		}
+	} else {
+		if (cpu_has_cache_cdex_s)
+			cache_line_size = cpu_scache_line_size();
+		else if (cpu_has_cache_cdex_p)
+			cache_line_size = cpu_dcache_line_size();
+	}
+	/*
+	 * Too much unrolling will overflow the available space in
+	 * clear_space_array / copy_page_array. 8 words sounds generous,
+	 * but a R4000 with 128 byte L2 line length can exceed even that.
+	 */
+	half_clear_loop_size = min(8 * clear_word_size,
+				   max(cache_line_size >> 1,
+				       4 * clear_word_size));
+	half_copy_loop_size = min(8 * copy_word_size,
+				  max(cache_line_size >> 1,
+				      4 * copy_word_size));
+}
+
+static void __cpuinit build_clear_store(u32 **buf, int off)
+{
+	if (cpu_has_64bit_gp_regs || cpu_has_64bit_zero_reg) {
+		uasm_i_sd(buf, ZERO, off, A0);
+	} else {
+		uasm_i_sw(buf, ZERO, off, A0);
+	}
+}
+
+static inline void __cpuinit build_clear_pref(u32 **buf, int off)
+{
+	if (off & cache_line_mask())
+		return;
+
+	if (pref_bias_clear_store) {
+		uasm_i_pref(buf, pref_dst_mode, pref_bias_clear_store + off,
+			    A0);
+	} else if (cpu_has_cache_cdex_s) {
+		uasm_i_cache(buf, Create_Dirty_Excl_SD, off, A0);
+	} else if (cpu_has_cache_cdex_p) {
+		if (R4600_V1_HIT_CACHEOP_WAR && cpu_is_r4600_v1_x()) {
+			uasm_i_nop(buf);
+			uasm_i_nop(buf);
+			uasm_i_nop(buf);
+			uasm_i_nop(buf);
+		}
+
+		if (R4600_V2_HIT_CACHEOP_WAR && cpu_is_r4600_v2_x())
+			uasm_i_lw(buf, ZERO, ZERO, AT);
+
+		uasm_i_cache(buf, Create_Dirty_Excl_D, off, A0);
+	}
+}
+
+void __cpuinit build_clear_page(void)
+{
+	int off;
+	u32 *buf = (u32 *)&clear_page_array;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+	int i;
+
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	set_prefetch_parameters();
+
+	/*
+	 * This algorithm makes the following assumptions:
+	 *   - The prefetch bias is a multiple of 2 words.
+	 *   - The prefetch bias is less than one page.
+	 */
+	BUG_ON(pref_bias_clear_store % (2 * clear_word_size));
+	BUG_ON(PAGE_SIZE < pref_bias_clear_store);
+
+	off = PAGE_SIZE - pref_bias_clear_store;
+	if (off > 0xffff || !pref_bias_clear_store)
+		pg_addiu(&buf, A2, A0, off);
+	else
+		uasm_i_ori(&buf, A2, A0, off);
+
+	if (R4600_V2_HIT_CACHEOP_WAR && cpu_is_r4600_v2_x())
+		uasm_i_lui(&buf, AT, 0xa000);
+
+	off = min(8, pref_bias_clear_store / cache_line_size) *
+	      cache_line_size;
+	while (off) {
+		build_clear_pref(&buf, -off);
+		off -= cache_line_size;
+	}
+	uasm_l_clear_pref(&l, buf);
+	do {
+		build_clear_pref(&buf, off);
+		build_clear_store(&buf, off);
+		off += clear_word_size;
+	} while (off < half_clear_loop_size);
+	pg_addiu(&buf, A0, A0, 2 * off);
+	off = -off;
+	do {
+		build_clear_pref(&buf, off);
+		if (off == -clear_word_size)
+			uasm_il_bne(&buf, &r, A0, A2, label_clear_pref);
+		build_clear_store(&buf, off);
+		off += clear_word_size;
+	} while (off < 0);
+
+	if (pref_bias_clear_store) {
+		pg_addiu(&buf, A2, A0, pref_bias_clear_store);
+		uasm_l_clear_nopref(&l, buf);
+		off = 0;
+		do {
+			build_clear_store(&buf, off);
+			off += clear_word_size;
+		} while (off < half_clear_loop_size);
+		pg_addiu(&buf, A0, A0, 2 * off);
+		off = -off;
+		do {
+			if (off == -clear_word_size)
+				uasm_il_bne(&buf, &r, A0, A2,
+					    label_clear_nopref);
+			build_clear_store(&buf, off);
+			off += clear_word_size;
+		} while (off < 0);
+	}
+
+	uasm_i_jr(&buf, RA);
+	uasm_i_nop(&buf);
+
+	BUG_ON(buf > clear_page_array + ARRAY_SIZE(clear_page_array));
+
+	uasm_resolve_relocs(relocs, labels);
+
+	pr_debug("Synthesized clear page handler (%u instructions).\n",
+		 (u32)(buf - clear_page_array));
+
+	pr_debug("\t.set push\n");
+	pr_debug("\t.set noreorder\n");
+	for (i = 0; i < (buf - clear_page_array); i++)
+		pr_debug("\t.word 0x%08x\n", clear_page_array[i]);
+	pr_debug("\t.set pop\n");
+}
+
+static void __cpuinit build_copy_load(u32 **buf, int reg, int off)
+{
+	if (cpu_has_64bit_gp_regs) {
+		uasm_i_ld(buf, reg, off, A1);
+	} else {
+		uasm_i_lw(buf, reg, off, A1);
+	}
+}
+
+static void __cpuinit build_copy_store(u32 **buf, int reg, int off)
+{
+	if (cpu_has_64bit_gp_regs) {
+		uasm_i_sd(buf, reg, off, A0);
+	} else {
+		uasm_i_sw(buf, reg, off, A0);
+	}
+}
+
+static inline void build_copy_load_pref(u32 **buf, int off)
+{
+	if (off & cache_line_mask())
+		return;
+
+	if (pref_bias_copy_load)
+		uasm_i_pref(buf, pref_src_mode, pref_bias_copy_load + off, A1);
+}
+
+static inline void build_copy_store_pref(u32 **buf, int off)
+{
+	if (off & cache_line_mask())
+		return;
+
+	if (pref_bias_copy_store) {
+		uasm_i_pref(buf, pref_dst_mode, pref_bias_copy_store + off,
+			    A0);
+	} else if (cpu_has_cache_cdex_s) {
+		uasm_i_cache(buf, Create_Dirty_Excl_SD, off, A0);
+	} else if (cpu_has_cache_cdex_p) {
+		if (R4600_V1_HIT_CACHEOP_WAR && cpu_is_r4600_v1_x()) {
+			uasm_i_nop(buf);
+			uasm_i_nop(buf);
+			uasm_i_nop(buf);
+			uasm_i_nop(buf);
+		}
+
+		if (R4600_V2_HIT_CACHEOP_WAR && cpu_is_r4600_v2_x())
+			uasm_i_lw(buf, ZERO, ZERO, AT);
+
+		uasm_i_cache(buf, Create_Dirty_Excl_D, off, A0);
+	}
+}
+
+void __cpuinit build_copy_page(void)
+{
+	int off;
+	u32 *buf = (u32 *)&copy_page_array;
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+	int i;
+
+	memset(labels, 0, sizeof(labels));
+	memset(relocs, 0, sizeof(relocs));
+
+	set_prefetch_parameters();
+
+	/*
+	 * This algorithm makes the following assumptions:
+	 *   - All prefetch biases are multiples of 8 words.
+	 *   - The prefetch biases are less than one page.
+	 *   - The store prefetch bias isn't greater than the load
+	 *     prefetch bias.
+	 */
+	BUG_ON(pref_bias_copy_load % (8 * copy_word_size));
+	BUG_ON(pref_bias_copy_store % (8 * copy_word_size));
+	BUG_ON(PAGE_SIZE < pref_bias_copy_load);
+	BUG_ON(pref_bias_copy_store > pref_bias_copy_load);
+
+	off = PAGE_SIZE - pref_bias_copy_load;
+	if (off > 0xffff || !pref_bias_copy_load)
+		pg_addiu(&buf, A2, A0, off);
+	else
+		uasm_i_ori(&buf, A2, A0, off);
+
+	if (R4600_V2_HIT_CACHEOP_WAR && cpu_is_r4600_v2_x())
+		uasm_i_lui(&buf, AT, 0xa000);
+
+	off = min(8, pref_bias_copy_load / cache_line_size) * cache_line_size;
+	while (off) {
+		build_copy_load_pref(&buf, -off);
+		off -= cache_line_size;
+	}
+	off = min(8, pref_bias_copy_store / cache_line_size) * cache_line_size;
+	while (off) {
+		build_copy_store_pref(&buf, -off);
+		off -= cache_line_size;
+	}
+	uasm_l_copy_pref_both(&l, buf);
+	do {
+		build_copy_load_pref(&buf, off);
+		build_copy_load(&buf, T0, off);
+		build_copy_load_pref(&buf, off + copy_word_size);
+		build_copy_load(&buf, T1, off + copy_word_size);
+		build_copy_load_pref(&buf, off + 2 * copy_word_size);
+		build_copy_load(&buf, T2, off + 2 * copy_word_size);
+		build_copy_load_pref(&buf, off + 3 * copy_word_size);
+		build_copy_load(&buf, T3, off + 3 * copy_word_size);
+		build_copy_store_pref(&buf, off);
+		build_copy_store(&buf, T0, off);
+		build_copy_store_pref(&buf, off + copy_word_size);
+		build_copy_store(&buf, T1, off + copy_word_size);
+		build_copy_store_pref(&buf, off + 2 * copy_word_size);
+		build_copy_store(&buf, T2, off + 2 * copy_word_size);
+		build_copy_store_pref(&buf, off + 3 * copy_word_size);
+		build_copy_store(&buf, T3, off + 3 * copy_word_size);
+		off += 4 * copy_word_size;
+	} while (off < half_copy_loop_size);
+	pg_addiu(&buf, A1, A1, 2 * off);
+	pg_addiu(&buf, A0, A0, 2 * off);
+	off = -off;
+	do {
+		build_copy_load_pref(&buf, off);
+		build_copy_load(&buf, T0, off);
+		build_copy_load_pref(&buf, off + copy_word_size);
+		build_copy_load(&buf, T1, off + copy_word_size);
+		build_copy_load_pref(&buf, off + 2 * copy_word_size);
+		build_copy_load(&buf, T2, off + 2 * copy_word_size);
+		build_copy_load_pref(&buf, off + 3 * copy_word_size);
+		build_copy_load(&buf, T3, off + 3 * copy_word_size);
+		build_copy_store_pref(&buf, off);
+		build_copy_store(&buf, T0, off);
+		build_copy_store_pref(&buf, off + copy_word_size);
+		build_copy_store(&buf, T1, off + copy_word_size);
+		build_copy_store_pref(&buf, off + 2 * copy_word_size);
+		build_copy_store(&buf, T2, off + 2 * copy_word_size);
+		build_copy_store_pref(&buf, off + 3 * copy_word_size);
+		if (off == -(4 * copy_word_size))
+			uasm_il_bne(&buf, &r, A2, A0, label_copy_pref_both);
+		build_copy_store(&buf, T3, off + 3 * copy_word_size);
+		off += 4 * copy_word_size;
+	} while (off < 0);
+
+	if (pref_bias_copy_load - pref_bias_copy_store) {
+		pg_addiu(&buf, A2, A0,
+			 pref_bias_copy_load - pref_bias_copy_store);
+		uasm_l_copy_pref_store(&l, buf);
+		off = 0;
+		do {
+			build_copy_load(&buf, T0, off);
+			build_copy_load(&buf, T1, off + copy_word_size);
+			build_copy_load(&buf, T2, off + 2 * copy_word_size);
+			build_copy_load(&buf, T3, off + 3 * copy_word_size);
+			build_copy_store_pref(&buf, off);
+			build_copy_store(&buf, T0, off);
+			build_copy_store_pref(&buf, off + copy_word_size);
+			build_copy_store(&buf, T1, off + copy_word_size);
+			build_copy_store_pref(&buf, off + 2 * copy_word_size);
+			build_copy_store(&buf, T2, off + 2 * copy_word_size);
+			build_copy_store_pref(&buf, off + 3 * copy_word_size);
+			build_copy_store(&buf, T3, off + 3 * copy_word_size);
+			off += 4 * copy_word_size;
+		} while (off < half_copy_loop_size);
+		pg_addiu(&buf, A1, A1, 2 * off);
+		pg_addiu(&buf, A0, A0, 2 * off);
+		off = -off;
+		do {
+			build_copy_load(&buf, T0, off);
+			build_copy_load(&buf, T1, off + copy_word_size);
+			build_copy_load(&buf, T2, off + 2 * copy_word_size);
+			build_copy_load(&buf, T3, off + 3 * copy_word_size);
+			build_copy_store_pref(&buf, off);
+			build_copy_store(&buf, T0, off);
+			build_copy_store_pref(&buf, off + copy_word_size);
+			build_copy_store(&buf, T1, off + copy_word_size);
+			build_copy_store_pref(&buf, off + 2 * copy_word_size);
+			build_copy_store(&buf, T2, off + 2 * copy_word_size);
+			build_copy_store_pref(&buf, off + 3 * copy_word_size);
+			if (off == -(4 * copy_word_size))
+				uasm_il_bne(&buf, &r, A2, A0,
+					    label_copy_pref_store);
+			build_copy_store(&buf, T3, off + 3 * copy_word_size);
+			off += 4 * copy_word_size;
+		} while (off < 0);
+	}
+
+	if (pref_bias_copy_store) {
+		pg_addiu(&buf, A2, A0, pref_bias_copy_store);
+		uasm_l_copy_nopref(&l, buf);
+		off = 0;
+		do {
+			build_copy_load(&buf, T0, off);
+			build_copy_load(&buf, T1, off + copy_word_size);
+			build_copy_load(&buf, T2, off + 2 * copy_word_size);
+			build_copy_load(&buf, T3, off + 3 * copy_word_size);
+			build_copy_store(&buf, T0, off);
+			build_copy_store(&buf, T1, off + copy_word_size);
+			build_copy_store(&buf, T2, off + 2 * copy_word_size);
+			build_copy_store(&buf, T3, off + 3 * copy_word_size);
+			off += 4 * copy_word_size;
+		} while (off < half_copy_loop_size);
+		pg_addiu(&buf, A1, A1, 2 * off);
+		pg_addiu(&buf, A0, A0, 2 * off);
+		off = -off;
+		do {
+			build_copy_load(&buf, T0, off);
+			build_copy_load(&buf, T1, off + copy_word_size);
+			build_copy_load(&buf, T2, off + 2 * copy_word_size);
+			build_copy_load(&buf, T3, off + 3 * copy_word_size);
+			build_copy_store(&buf, T0, off);
+			build_copy_store(&buf, T1, off + copy_word_size);
+			build_copy_store(&buf, T2, off + 2 * copy_word_size);
+			if (off == -(4 * copy_word_size))
+				uasm_il_bne(&buf, &r, A2, A0,
+					    label_copy_nopref);
+			build_copy_store(&buf, T3, off + 3 * copy_word_size);
+			off += 4 * copy_word_size;
+		} while (off < 0);
+	}
+
+	uasm_i_jr(&buf, RA);
+	uasm_i_nop(&buf);
+
+	BUG_ON(buf > copy_page_array + ARRAY_SIZE(copy_page_array));
+
+	uasm_resolve_relocs(relocs, labels);
+
+	pr_debug("Synthesized copy page handler (%u instructions).\n",
+		 (u32)(buf - copy_page_array));
+
+	pr_debug("\t.set push\n");
+	pr_debug("\t.set noreorder\n");
+	for (i = 0; i < (buf - copy_page_array); i++)
+		pr_debug("\t.word 0x%08x\n", copy_page_array[i]);
+	pr_debug("\t.set pop\n");
+}
+
+#ifdef CONFIG_SIBYTE_DMA_PAGEOPS
+
+/*
+ * Pad descriptors to cacheline, since each is exclusively owned by a
+ * particular CPU.
+ */
+struct dmadscr {
+	u64 dscr_a;
+	u64 dscr_b;
+	u64 pad_a;
+	u64 pad_b;
+} ____cacheline_aligned_in_smp page_descr[DM_NUM_CHANNELS];
+
+void sb1_dma_init(void)
+{
+	int i;
+
+	for (i = 0; i < DM_NUM_CHANNELS; i++) {
+		const u64 base_val = CPHYSADDR((unsigned long)&page_descr[i]) |
+				     V_DM_DSCR_BASE_RINGSZ(1);
+		void *base_reg = IOADDR(A_DM_REGISTER(i, R_DM_DSCR_BASE));
+
+		__raw_writeq(base_val, base_reg);
+		__raw_writeq(base_val | M_DM_DSCR_BASE_RESET, base_reg);
+		__raw_writeq(base_val | M_DM_DSCR_BASE_ENABL, base_reg);
+	}
+}
+
+void clear_page(void *page)
+{
+	u64 to_phys = CPHYSADDR((unsigned long)page);
+	unsigned int cpu = smp_processor_id();
+
+	/* if the page is not in KSEG0, use old way */
+	if ((long)KSEGX((unsigned long)page) != (long)CKSEG0)
+		return clear_page_cpu(page);
+
+	page_descr[cpu].dscr_a = to_phys | M_DM_DSCRA_ZERO_MEM |
+				 M_DM_DSCRA_L2C_DEST | M_DM_DSCRA_INTERRUPT;
+	page_descr[cpu].dscr_b = V_DM_DSCRB_SRC_LENGTH(PAGE_SIZE);
+	__raw_writeq(1, IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_COUNT)));
+
+	/*
+	 * Don't really want to do it this way, but there's no
+	 * reliable way to delay completion detection.
+	 */
+	while (!(__raw_readq(IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_BASE_DEBUG)))
+		 & M_DM_DSCR_BASE_INTERRUPT))
+		;
+	__raw_readq(IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_BASE)));
+}
+
+void copy_page(void *to, void *from)
+{
+	u64 from_phys = CPHYSADDR((unsigned long)from);
+	u64 to_phys = CPHYSADDR((unsigned long)to);
+	unsigned int cpu = smp_processor_id();
+
+	/* if any page is not in KSEG0, use old way */
+	if ((long)KSEGX((unsigned long)to) != (long)CKSEG0
+	    || (long)KSEGX((unsigned long)from) != (long)CKSEG0)
+		return copy_page_cpu(to, from);
+
+	page_descr[cpu].dscr_a = to_phys | M_DM_DSCRA_L2C_DEST |
+				 M_DM_DSCRA_INTERRUPT;
+	page_descr[cpu].dscr_b = from_phys | V_DM_DSCRB_SRC_LENGTH(PAGE_SIZE);
+	__raw_writeq(1, IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_COUNT)));
+
+	/*
+	 * Don't really want to do it this way, but there's no
+	 * reliable way to delay completion detection.
+	 */
+	while (!(__raw_readq(IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_BASE_DEBUG)))
+		 & M_DM_DSCR_BASE_INTERRUPT))
+		;
+	__raw_readq(IOADDR(A_DM_REGISTER(cpu, R_DM_DSCR_BASE)));
+}
+
+#endif /* CONFIG_SIBYTE_DMA_PAGEOPS */