md/raid6: implement recovery using ARM NEON intrinsics

Provide a NEON accelerated implementation of the recovery algorithm,
which supersedes the default byte-by-byte one.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
diff --git a/lib/raid6/Makefile b/lib/raid6/Makefile
index 3057011..a93adf6 100644
--- a/lib/raid6/Makefile
+++ b/lib/raid6/Makefile
@@ -5,7 +5,7 @@
 
 raid6_pq-$(CONFIG_X86) += recov_ssse3.o recov_avx2.o mmx.o sse1.o sse2.o avx2.o avx512.o recov_avx512.o
 raid6_pq-$(CONFIG_ALTIVEC) += altivec1.o altivec2.o altivec4.o altivec8.o
-raid6_pq-$(CONFIG_KERNEL_MODE_NEON) += neon.o neon1.o neon2.o neon4.o neon8.o
+raid6_pq-$(CONFIG_KERNEL_MODE_NEON) += neon.o neon1.o neon2.o neon4.o neon8.o recov_neon.o recov_neon_inner.o
 raid6_pq-$(CONFIG_TILEGX) += tilegx8.o
 raid6_pq-$(CONFIG_S390) += s390vx8.o recov_s390xc.o
 
@@ -26,7 +26,9 @@
 ifeq ($(ARCH),arm)
 NEON_FLAGS += -mfloat-abi=softfp -mfpu=neon
 endif
+CFLAGS_recov_neon_inner.o += $(NEON_FLAGS)
 ifeq ($(ARCH),arm64)
+CFLAGS_REMOVE_recov_neon_inner.o += -mgeneral-regs-only
 CFLAGS_REMOVE_neon1.o += -mgeneral-regs-only
 CFLAGS_REMOVE_neon2.o += -mgeneral-regs-only
 CFLAGS_REMOVE_neon4.o += -mgeneral-regs-only
diff --git a/lib/raid6/algos.c b/lib/raid6/algos.c
index 7857049..4769947 100644
--- a/lib/raid6/algos.c
+++ b/lib/raid6/algos.c
@@ -113,6 +113,9 @@
 #ifdef CONFIG_S390
 	&raid6_recov_s390xc,
 #endif
+#if defined(CONFIG_KERNEL_MODE_NEON)
+	&raid6_recov_neon,
+#endif
 	&raid6_recov_intx1,
 	NULL
 };
diff --git a/lib/raid6/recov_neon.c b/lib/raid6/recov_neon.c
new file mode 100644
index 0000000..eeb5c40
--- /dev/null
+++ b/lib/raid6/recov_neon.c
@@ -0,0 +1,110 @@
+/*
+ * Copyright (C) 2012 Intel Corporation
+ * Copyright (C) 2017 Linaro Ltd. <ard.biesheuvel@linaro.org>
+ *
+ * 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
+ * of the License.
+ */
+
+#include <linux/raid/pq.h>
+
+#ifdef __KERNEL__
+#include <asm/neon.h>
+#else
+#define kernel_neon_begin()
+#define kernel_neon_end()
+#define cpu_has_neon()		(1)
+#endif
+
+static int raid6_has_neon(void)
+{
+	return cpu_has_neon();
+}
+
+void __raid6_2data_recov_neon(int bytes, uint8_t *p, uint8_t *q, uint8_t *dp,
+			      uint8_t *dq, const uint8_t *pbmul,
+			      const uint8_t *qmul);
+
+void __raid6_datap_recov_neon(int bytes, uint8_t *p, uint8_t *q, uint8_t *dq,
+			      const uint8_t *qmul);
+
+static void raid6_2data_recov_neon(int disks, size_t bytes, int faila,
+		int failb, void **ptrs)
+{
+	u8 *p, *q, *dp, *dq;
+	const u8 *pbmul;	/* P multiplier table for B data */
+	const u8 *qmul;		/* Q multiplier table (for both) */
+
+	p = (u8 *)ptrs[disks - 2];
+	q = (u8 *)ptrs[disks - 1];
+
+	/*
+	 * Compute syndrome with zero for the missing data pages
+	 * Use the dead data pages as temporary storage for
+	 * delta p and delta q
+	 */
+	dp = (u8 *)ptrs[faila];
+	ptrs[faila] = (void *)raid6_empty_zero_page;
+	ptrs[disks - 2] = dp;
+	dq = (u8 *)ptrs[failb];
+	ptrs[failb] = (void *)raid6_empty_zero_page;
+	ptrs[disks - 1] = dq;
+
+	raid6_call.gen_syndrome(disks, bytes, ptrs);
+
+	/* Restore pointer table */
+	ptrs[faila]     = dp;
+	ptrs[failb]     = dq;
+	ptrs[disks - 2] = p;
+	ptrs[disks - 1] = q;
+
+	/* Now, pick the proper data tables */
+	pbmul = raid6_vgfmul[raid6_gfexi[failb-faila]];
+	qmul  = raid6_vgfmul[raid6_gfinv[raid6_gfexp[faila] ^
+					 raid6_gfexp[failb]]];
+
+	kernel_neon_begin();
+	__raid6_2data_recov_neon(bytes, p, q, dp, dq, pbmul, qmul);
+	kernel_neon_end();
+}
+
+static void raid6_datap_recov_neon(int disks, size_t bytes, int faila,
+		void **ptrs)
+{
+	u8 *p, *q, *dq;
+	const u8 *qmul;		/* Q multiplier table */
+
+	p = (u8 *)ptrs[disks - 2];
+	q = (u8 *)ptrs[disks - 1];
+
+	/*
+	 * Compute syndrome with zero for the missing data page
+	 * Use the dead data page as temporary storage for delta q
+	 */
+	dq = (u8 *)ptrs[faila];
+	ptrs[faila] = (void *)raid6_empty_zero_page;
+	ptrs[disks - 1] = dq;
+
+	raid6_call.gen_syndrome(disks, bytes, ptrs);
+
+	/* Restore pointer table */
+	ptrs[faila]     = dq;
+	ptrs[disks - 1] = q;
+
+	/* Now, pick the proper data tables */
+	qmul = raid6_vgfmul[raid6_gfinv[raid6_gfexp[faila]]];
+
+	kernel_neon_begin();
+	__raid6_datap_recov_neon(bytes, p, q, dq, qmul);
+	kernel_neon_end();
+}
+
+const struct raid6_recov_calls raid6_recov_neon = {
+	.data2		= raid6_2data_recov_neon,
+	.datap		= raid6_datap_recov_neon,
+	.valid		= raid6_has_neon,
+	.name		= "neon",
+	.priority	= 10,
+};
diff --git a/lib/raid6/recov_neon_inner.c b/lib/raid6/recov_neon_inner.c
new file mode 100644
index 0000000..8cd20c9
--- /dev/null
+++ b/lib/raid6/recov_neon_inner.c
@@ -0,0 +1,117 @@
+/*
+ * Copyright (C) 2012 Intel Corporation
+ * Copyright (C) 2017 Linaro Ltd. <ard.biesheuvel@linaro.org>
+ *
+ * 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
+ * of the License.
+ */
+
+#include <arm_neon.h>
+
+static const uint8x16_t x0f = {
+	0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f,
+	0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f, 0x0f,
+};
+
+#ifdef CONFIG_ARM
+/*
+ * AArch32 does not provide this intrinsic natively because it does not
+ * implement the underlying instruction. AArch32 only provides a 64-bit
+ * wide vtbl.8 instruction, so use that instead.
+ */
+static uint8x16_t vqtbl1q_u8(uint8x16_t a, uint8x16_t b)
+{
+	union {
+		uint8x16_t	val;
+		uint8x8x2_t	pair;
+	} __a = { a };
+
+	return vcombine_u8(vtbl2_u8(__a.pair, vget_low_u8(b)),
+			   vtbl2_u8(__a.pair, vget_high_u8(b)));
+}
+#endif
+
+void __raid6_2data_recov_neon(int bytes, uint8_t *p, uint8_t *q, uint8_t *dp,
+			      uint8_t *dq, const uint8_t *pbmul,
+			      const uint8_t *qmul)
+{
+	uint8x16_t pm0 = vld1q_u8(pbmul);
+	uint8x16_t pm1 = vld1q_u8(pbmul + 16);
+	uint8x16_t qm0 = vld1q_u8(qmul);
+	uint8x16_t qm1 = vld1q_u8(qmul + 16);
+
+	/*
+	 * while ( bytes-- ) {
+	 *	uint8_t px, qx, db;
+	 *
+	 *	px    = *p ^ *dp;
+	 *	qx    = qmul[*q ^ *dq];
+	 *	*dq++ = db = pbmul[px] ^ qx;
+	 *	*dp++ = db ^ px;
+	 *	p++; q++;
+	 * }
+	 */
+
+	while (bytes) {
+		uint8x16_t vx, vy, px, qx, db;
+
+		px = veorq_u8(vld1q_u8(p), vld1q_u8(dp));
+		vx = veorq_u8(vld1q_u8(q), vld1q_u8(dq));
+
+		vy = (uint8x16_t)vshrq_n_s16((int16x8_t)vx, 4);
+		vx = vqtbl1q_u8(qm0, vandq_u8(vx, x0f));
+		vy = vqtbl1q_u8(qm1, vandq_u8(vy, x0f));
+		qx = veorq_u8(vx, vy);
+
+		vy = (uint8x16_t)vshrq_n_s16((int16x8_t)px, 4);
+		vx = vqtbl1q_u8(pm0, vandq_u8(px, x0f));
+		vy = vqtbl1q_u8(pm1, vandq_u8(vy, x0f));
+		vx = veorq_u8(vx, vy);
+		db = veorq_u8(vx, qx);
+
+		vst1q_u8(dq, db);
+		vst1q_u8(dp, veorq_u8(db, px));
+
+		bytes -= 16;
+		p += 16;
+		q += 16;
+		dp += 16;
+		dq += 16;
+	}
+}
+
+void __raid6_datap_recov_neon(int bytes, uint8_t *p, uint8_t *q, uint8_t *dq,
+			      const uint8_t *qmul)
+{
+	uint8x16_t qm0 = vld1q_u8(qmul);
+	uint8x16_t qm1 = vld1q_u8(qmul + 16);
+
+	/*
+	 * while (bytes--) {
+	 *	*p++ ^= *dq = qmul[*q ^ *dq];
+	 *	q++; dq++;
+	 * }
+	 */
+
+	while (bytes) {
+		uint8x16_t vx, vy;
+
+		vx = veorq_u8(vld1q_u8(q), vld1q_u8(dq));
+
+		vy = (uint8x16_t)vshrq_n_s16((int16x8_t)vx, 4);
+		vx = vqtbl1q_u8(qm0, vandq_u8(vx, x0f));
+		vy = vqtbl1q_u8(qm1, vandq_u8(vy, x0f));
+		vx = veorq_u8(vx, vy);
+		vy = veorq_u8(vx, vld1q_u8(p));
+
+		vst1q_u8(dq, vx);
+		vst1q_u8(p, vy);
+
+		bytes -= 16;
+		p += 16;
+		q += 16;
+		dq += 16;
+	}
+}