[ARM] 2930/1: optimized sha1 implementation for ARM

Patch from Nicolas Pitre

Here's an ARM assembly SHA1 implementation to replace the default C
version. It is approximately 50% faster than the generic C version. On
an XScale processor running at 400MHz:
	generic C version:	9.8 MB/s
	my version:		14.5 MB/s
This code is useful to quite a few callers in the tree:
crypto/sha1.c:		sha_transform(sctx->state, sctx->buffer, temp);
crypto/sha1.c:			sha_transform(sctx->state, &data[i], temp);
drivers/char/random.c:		sha_transform(buf, (__u8 *)r->pool+i, buf + 5);
drivers/char/random.c:	sha_transform(buf, (__u8 *)data, buf + 5);
net/ipv4/syncookies.c:	sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5);
Signed-off-by: Nicolas Pitre <nico@cam.org>
Seems to work fine on big-endian as well.
Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

arch/arm/lib/Makefile

@@ -11,7 +11,7 @@ lib-y		:= backtrace.o changebit.o csumipv6.o csumpartial.o   \
 		   strnlen_user.o strchr.o strrchr.o testchangebit.o  \
 		   testclearbit.o testsetbit.o uaccess.o getuser.o    \
 		   putuser.o ashldi3.o ashrdi3.o lshrdi3.o muldi3.o   \
-		   ucmpdi2.o lib1funcs.o div64.o	              \
+		   ucmpdi2.o lib1funcs.o div64.o sha1.o               \
 		   io-readsb.o io-writesb.o io-readsl.o io-writesl.o
 
 ifeq ($(CONFIG_CPU_32v3),y)

arch/arm/lib/sha1.S

+/*
+ *  linux/arch/arm/lib/sha1.S
+ *
+ *  SHA transform optimized for ARM
+ *
+ *  Copyright:	(C) 2005 by Nicolas Pitre <nico@cam.org>
+ *  Created:	September 17, 2005
+ *
+ *  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.
+ *
+ *  The reference implementation for this code is linux/lib/sha1.c
+ */
+
+#include <linux/linkage.h>
+
+	.text
+
+
+/*
+ * void sha_transform(__u32 *digest, const char *in, __u32 *W)
+ *
+ * Note: the "in" ptr may be unaligned.
+ */
+
+ENTRY(sha_transform)
+
+	stmfd	sp!, {r4 - r8, lr}
+
+	@ for (i = 0; i < 16; i++)
+	@         W[i] = be32_to_cpu(in[i]); */
+
+#ifdef __ARMEB__
+	mov	r4, r0
+	mov	r0, r2
+	mov	r2, #64
+	bl	memcpy
+	mov	r2, r0
+	mov	r0, r4
+#else
+	mov	r3, r2
+	mov	lr, #16
+1:	ldrb	r4, [r1], #1
+	ldrb	r5, [r1], #1
+	ldrb	r6, [r1], #1
+	ldrb	r7, [r1], #1
+	subs	lr, lr, #1
+	orr	r5, r5, r4, lsl #8
+	orr	r6, r6, r5, lsl #8
+	orr	r7, r7, r6, lsl #8
+	str	r7, [r3], #4
+	bne	1b
+#endif
+
+	@ for (i = 0; i < 64; i++)
+	@         W[i+16] = ror(W[i+13] ^ W[i+8] ^ W[i+2] ^ W[i], 31);
+
+	sub	r3, r2, #4
+	mov	lr, #64
+2:	ldr	r4, [r3, #4]!
+	subs	lr, lr, #1
+	ldr	r5, [r3, #8]
+	ldr	r6, [r3, #32]
+	ldr	r7, [r3, #52]
+	eor	r4, r4, r5
+	eor	r4, r4, r6
+	eor	r4, r4, r7
+	mov	r4, r4, ror #31
+	str	r4, [r3, #64]
+	bne	2b
+
+	/*
+	 * The SHA functions are:
+	 *
+	 * f1(B,C,D) = (D ^ (B & (C ^ D)))
+	 * f2(B,C,D) = (B ^ C ^ D)
+	 * f3(B,C,D) = ((B & C) | (D & (B | C)))
+	 *
+	 * Then the sub-blocks are processed as follows:
+	 *
+	 * A' = ror(A, 27) + f(B,C,D) + E + K + *W++
+	 * B' = A
+	 * C' = ror(B, 2)
+	 * D' = C
+	 * E' = D
+	 *
+	 * We therefore unroll each loop 5 times to avoid register shuffling.
+	 * Also the ror for C (and also D and E which are successivelyderived
+	 * from it) is applied in place to cut on an additional mov insn for
+	 * each round.
+	 */
+
+	.macro	sha_f1, A, B, C, D, E
+	ldr	r3, [r2], #4
+	eor	ip, \C, \D
+	add	\E, r1, \E, ror #2
+	and	ip, \B, ip, ror #2
+	add	\E, \E, \A, ror #27
+	eor	ip, ip, \D, ror #2
+	add	\E, \E, r3
+	add	\E, \E, ip
+	.endm
+
+	.macro	sha_f2, A, B, C, D, E
+	ldr	r3, [r2], #4
+	add	\E, r1, \E, ror #2
+	eor	ip, \B, \C, ror #2
+	add	\E, \E, \A, ror #27
+	eor	ip, ip, \D, ror #2
+	add	\E, \E, r3
+	add	\E, \E, ip
+	.endm
+
+	.macro	sha_f3, A, B, C, D, E
+	ldr	r3, [r2], #4
+	add	\E, r1, \E, ror #2
+	orr	ip, \B, \C, ror #2
+	add	\E, \E, \A, ror #27
+	and	ip, ip, \D, ror #2
+	add	\E, \E, r3
+	and	r3, \B, \C, ror #2
+	orr	ip, ip, r3
+	add	\E, \E, ip
+	.endm
+
+	ldmia	r0, {r4 - r8}
+
+	mov	lr, #4
+	ldr	r1, .L_sha_K + 0
+
+	/* adjust initial values */
+	mov	r6, r6, ror #30
+	mov	r7, r7, ror #30
+	mov	r8, r8, ror #30
+
+3:	subs	lr, lr, #1
+	sha_f1	r4, r5, r6, r7, r8
+	sha_f1	r8, r4, r5, r6, r7
+	sha_f1	r7, r8, r4, r5, r6
+	sha_f1	r6, r7, r8, r4, r5
+	sha_f1	r5, r6, r7, r8, r4
+	bne	3b
+
+	ldr	r1, .L_sha_K + 4
+	mov	lr, #4
+
+4:	subs	lr, lr, #1
+	sha_f2	r4, r5, r6, r7, r8
+	sha_f2	r8, r4, r5, r6, r7
+	sha_f2	r7, r8, r4, r5, r6
+	sha_f2	r6, r7, r8, r4, r5
+	sha_f2	r5, r6, r7, r8, r4
+	bne	4b
+
+	ldr	r1, .L_sha_K + 8
+	mov	lr, #4
+
+5:	subs	lr, lr, #1
+	sha_f3	r4, r5, r6, r7, r8
+	sha_f3	r8, r4, r5, r6, r7
+	sha_f3	r7, r8, r4, r5, r6
+	sha_f3	r6, r7, r8, r4, r5
+	sha_f3	r5, r6, r7, r8, r4
+	bne	5b
+
+	ldr	r1, .L_sha_K + 12
+	mov	lr, #4
+
+6:	subs	lr, lr, #1
+	sha_f2	r4, r5, r6, r7, r8
+	sha_f2	r8, r4, r5, r6, r7
+	sha_f2	r7, r8, r4, r5, r6
+	sha_f2	r6, r7, r8, r4, r5
+	sha_f2	r5, r6, r7, r8, r4
+	bne	6b
+
+	ldmia	r0, {r1, r2, r3, ip, lr}
+	add	r4, r1, r4
+	add	r5, r2, r5
+	add	r6, r3, r6, ror #2
+	add	r7, ip, r7, ror #2
+	add	r8, lr, r8, ror #2
+	stmia	r0, {r4 - r8}
+
+	ldmfd	sp!, {r4 - r8, pc}
+
+.L_sha_K:
+	.word	0x5a827999, 0x6ed9eba1, 0x8f1bbcdc, 0xca62c1d6
+
+
+/*
+ * void sha_init(__u32 *buf)
+ */
+
+.L_sha_initial_digest:
+	.word	0x67452301, 0xefcdab89, 0x98badcfe, 0x10325476, 0xc3d2e1f0
+
+ENTRY(sha_init)
+
+	str	lr, [sp, #-4]!
+	adr	r1, .L_sha_initial_digest
+	ldmia	r1, {r1, r2, r3, ip, lr}
+	stmia	r0, {r1, r2, r3, ip, lr}
+	ldr	pc, [sp], #4
+