1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * This is a SIMD SHA-1 implementation. It requires the Intel(R) Supplemental
4 * SSE3 instruction set extensions introduced in Intel Core Microarchitecture
5 * processors. CPUs supporting Intel(R) AVX extensions will get an additional
6 * boost.
7 *
8 * This work was inspired by the vectorized implementation of Dean Gaudet.
9 * Additional information on it can be found at:
10 *    http://www.arctic.org/~dean/crypto/sha1.html
11 *
12 * It was improved upon with more efficient vectorization of the message
13 * scheduling. This implementation has also been optimized for all current and
14 * several future generations of Intel CPUs.
15 *
16 * See this article for more information about the implementation details:
17 *   http://software.intel.com/en-us/articles/improving-the-performance-of-the-secure-hash-algorithm-1/
18 *
19 * Copyright (C) 2010, Intel Corp.
20 *   Authors: Maxim Locktyukhin <maxim.locktyukhin@intel.com>
21 *            Ronen Zohar <ronen.zohar@intel.com>
22 *
23 * Converted to AT&T syntax and adapted for inclusion in the Linux kernel:
24 *   Author: Mathias Krause <minipli@googlemail.com>
25 */
26
27#include <linux/linkage.h>
28#include <linux/cfi_types.h>
29
30#define CTX	%rdi	// arg1
31#define BUF	%rsi	// arg2
32#define CNT	%rdx	// arg3
33
34#define REG_A	%ecx
35#define REG_B	%esi
36#define REG_C	%edi
37#define REG_D	%r12d
38#define REG_E	%edx
39
40#define REG_T1	%eax
41#define REG_T2	%ebx
42
43#define K_BASE		%r8
44#define HASH_PTR	%r9
45#define BUFFER_PTR	%r10
46#define BUFFER_END	%r11
47
48#define W_TMP1	%xmm0
49#define W_TMP2	%xmm9
50
51#define W0	%xmm1
52#define W4	%xmm2
53#define W8	%xmm3
54#define W12	%xmm4
55#define W16	%xmm5
56#define W20	%xmm6
57#define W24	%xmm7
58#define W28	%xmm8
59
60#define XMM_SHUFB_BSWAP	%xmm10
61
62/* we keep window of 64 w[i]+K pre-calculated values in a circular buffer */
63#define WK(t)	(((t) & 15) * 4)(%rsp)
64#define W_PRECALC_AHEAD	16
65
66/*
67 * This macro implements the SHA-1 function's body for single 64-byte block
68 * param: function's name
69 */
70.macro SHA1_VECTOR_ASM  name
71	SYM_TYPED_FUNC_START(\name)
72
73	push	%rbx
74	push	%r12
75	push	%rbp
76	mov	%rsp, %rbp
77
78	sub	$64, %rsp		# allocate workspace
79	and	$~15, %rsp		# align stack
80
81	mov	CTX, HASH_PTR
82	mov	BUF, BUFFER_PTR
83
84	shl	$6, CNT			# multiply by 64
85	add	BUF, CNT
86	mov	CNT, BUFFER_END
87
88	lea	K_XMM_AR(%rip), K_BASE
89	xmm_mov	BSWAP_SHUFB_CTL(%rip), XMM_SHUFB_BSWAP
90
91	SHA1_PIPELINED_MAIN_BODY
92
93	# cleanup workspace
94	mov	$8, %ecx
95	mov	%rsp, %rdi
96	xor	%eax, %eax
97	rep stosq
98
99	mov	%rbp, %rsp		# deallocate workspace
100	pop	%rbp
101	pop	%r12
102	pop	%rbx
103	RET
104
105	SYM_FUNC_END(\name)
106.endm
107
108/*
109 * This macro implements 80 rounds of SHA-1 for one 64-byte block
110 */
111.macro SHA1_PIPELINED_MAIN_BODY
112	INIT_REGALLOC
113
114	mov	  (HASH_PTR), A
115	mov	 4(HASH_PTR), B
116	mov	 8(HASH_PTR), C
117	mov	12(HASH_PTR), D
118	mov	16(HASH_PTR), E
119
120  .set i, 0
121  .rept W_PRECALC_AHEAD
122	W_PRECALC i
123    .set i, (i+1)
124  .endr
125
126.align 4
1271:
128	RR F1,A,B,C,D,E,0
129	RR F1,D,E,A,B,C,2
130	RR F1,B,C,D,E,A,4
131	RR F1,E,A,B,C,D,6
132	RR F1,C,D,E,A,B,8
133
134	RR F1,A,B,C,D,E,10
135	RR F1,D,E,A,B,C,12
136	RR F1,B,C,D,E,A,14
137	RR F1,E,A,B,C,D,16
138	RR F1,C,D,E,A,B,18
139
140	RR F2,A,B,C,D,E,20
141	RR F2,D,E,A,B,C,22
142	RR F2,B,C,D,E,A,24
143	RR F2,E,A,B,C,D,26
144	RR F2,C,D,E,A,B,28
145
146	RR F2,A,B,C,D,E,30
147	RR F2,D,E,A,B,C,32
148	RR F2,B,C,D,E,A,34
149	RR F2,E,A,B,C,D,36
150	RR F2,C,D,E,A,B,38
151
152	RR F3,A,B,C,D,E,40
153	RR F3,D,E,A,B,C,42
154	RR F3,B,C,D,E,A,44
155	RR F3,E,A,B,C,D,46
156	RR F3,C,D,E,A,B,48
157
158	RR F3,A,B,C,D,E,50
159	RR F3,D,E,A,B,C,52
160	RR F3,B,C,D,E,A,54
161	RR F3,E,A,B,C,D,56
162	RR F3,C,D,E,A,B,58
163
164	add	$64, BUFFER_PTR		# move to the next 64-byte block
165	cmp	BUFFER_END, BUFFER_PTR	# if the current is the last one use
166	cmovae	K_BASE, BUFFER_PTR	# dummy source to avoid buffer overrun
167
168	RR F4,A,B,C,D,E,60
169	RR F4,D,E,A,B,C,62
170	RR F4,B,C,D,E,A,64
171	RR F4,E,A,B,C,D,66
172	RR F4,C,D,E,A,B,68
173
174	RR F4,A,B,C,D,E,70
175	RR F4,D,E,A,B,C,72
176	RR F4,B,C,D,E,A,74
177	RR F4,E,A,B,C,D,76
178	RR F4,C,D,E,A,B,78
179
180	UPDATE_HASH   (HASH_PTR), A
181	UPDATE_HASH  4(HASH_PTR), B
182	UPDATE_HASH  8(HASH_PTR), C
183	UPDATE_HASH 12(HASH_PTR), D
184	UPDATE_HASH 16(HASH_PTR), E
185
186	RESTORE_RENAMED_REGS
187	cmp	K_BASE, BUFFER_PTR	# K_BASE means, we reached the end
188	jne	1b
189.endm
190
191.macro INIT_REGALLOC
192  .set A, REG_A
193  .set B, REG_B
194  .set C, REG_C
195  .set D, REG_D
196  .set E, REG_E
197  .set T1, REG_T1
198  .set T2, REG_T2
199.endm
200
201.macro RESTORE_RENAMED_REGS
202	# order is important (REG_C is where it should be)
203	mov	B, REG_B
204	mov	D, REG_D
205	mov	A, REG_A
206	mov	E, REG_E
207.endm
208
209.macro SWAP_REG_NAMES  a, b
210  .set _T, \a
211  .set \a, \b
212  .set \b, _T
213.endm
214
215.macro F1  b, c, d
216	mov	\c, T1
217	SWAP_REG_NAMES \c, T1
218	xor	\d, T1
219	and	\b, T1
220	xor	\d, T1
221.endm
222
223.macro F2  b, c, d
224	mov	\d, T1
225	SWAP_REG_NAMES \d, T1
226	xor	\c, T1
227	xor	\b, T1
228.endm
229
230.macro F3  b, c ,d
231	mov	\c, T1
232	SWAP_REG_NAMES \c, T1
233	mov	\b, T2
234	or	\b, T1
235	and	\c, T2
236	and	\d, T1
237	or	T2, T1
238.endm
239
240.macro F4  b, c, d
241	F2 \b, \c, \d
242.endm
243
244.macro UPDATE_HASH  hash, val
245	add	\hash, \val
246	mov	\val, \hash
247.endm
248
249/*
250 * RR does two rounds of SHA-1 back to back with W[] pre-calc
251 *   t1 = F(b, c, d);   e += w(i)
252 *   e += t1;           b <<= 30;   d  += w(i+1);
253 *   t1 = F(a, b, c);
254 *   d += t1;           a <<= 5;
255 *   e += a;
256 *   t1 = e;            a >>= 7;
257 *   t1 <<= 5;
258 *   d += t1;
259 */
260.macro RR  F, a, b, c, d, e, round
261	add	WK(\round), \e
262	\F   \b, \c, \d		# t1 = F(b, c, d);
263	W_PRECALC (\round + W_PRECALC_AHEAD)
264	rol	$30, \b
265	add	T1, \e
266	add	WK(\round + 1), \d
267
268	\F   \a, \b, \c
269	W_PRECALC (\round + W_PRECALC_AHEAD + 1)
270	rol	$5, \a
271	add	\a, \e
272	add	T1, \d
273	ror	$7, \a		# (a <<r 5) >>r 7) => a <<r 30)
274
275	mov	\e, T1
276	SWAP_REG_NAMES \e, T1
277
278	rol	$5, T1
279	add	T1, \d
280
281	# write:  \a, \b
282	# rotate: \a<=\d, \b<=\e, \c<=\a, \d<=\b, \e<=\c
283.endm
284
285.macro W_PRECALC  r
286  .set i, \r
287
288  .if (i < 20)
289    .set K_XMM, 0
290  .elseif (i < 40)
291    .set K_XMM, 16
292  .elseif (i < 60)
293    .set K_XMM, 32
294  .elseif (i < 80)
295    .set K_XMM, 48
296  .endif
297
298  .if ((i < 16) || ((i >= 80) && (i < (80 + W_PRECALC_AHEAD))))
299    .set i, ((\r) % 80)	    # pre-compute for the next iteration
300    .if (i == 0)
301	W_PRECALC_RESET
302    .endif
303	W_PRECALC_00_15
304  .elseif (i<32)
305	W_PRECALC_16_31
306  .elseif (i < 80)   // rounds 32-79
307	W_PRECALC_32_79
308  .endif
309.endm
310
311.macro W_PRECALC_RESET
312  .set W,          W0
313  .set W_minus_04, W4
314  .set W_minus_08, W8
315  .set W_minus_12, W12
316  .set W_minus_16, W16
317  .set W_minus_20, W20
318  .set W_minus_24, W24
319  .set W_minus_28, W28
320  .set W_minus_32, W
321.endm
322
323.macro W_PRECALC_ROTATE
324  .set W_minus_32, W_minus_28
325  .set W_minus_28, W_minus_24
326  .set W_minus_24, W_minus_20
327  .set W_minus_20, W_minus_16
328  .set W_minus_16, W_minus_12
329  .set W_minus_12, W_minus_08
330  .set W_minus_08, W_minus_04
331  .set W_minus_04, W
332  .set W,          W_minus_32
333.endm
334
335.macro W_PRECALC_SSSE3
336
337.macro W_PRECALC_00_15
338	W_PRECALC_00_15_SSSE3
339.endm
340.macro W_PRECALC_16_31
341	W_PRECALC_16_31_SSSE3
342.endm
343.macro W_PRECALC_32_79
344	W_PRECALC_32_79_SSSE3
345.endm
346
347/* message scheduling pre-compute for rounds 0-15 */
348.macro W_PRECALC_00_15_SSSE3
349  .if ((i & 3) == 0)
350	movdqu	(i*4)(BUFFER_PTR), W_TMP1
351  .elseif ((i & 3) == 1)
352	pshufb	XMM_SHUFB_BSWAP, W_TMP1
353	movdqa	W_TMP1, W
354  .elseif ((i & 3) == 2)
355	paddd	(K_BASE), W_TMP1
356  .elseif ((i & 3) == 3)
357	movdqa  W_TMP1, WK(i&~3)
358	W_PRECALC_ROTATE
359  .endif
360.endm
361
362/* message scheduling pre-compute for rounds 16-31
363 *
364 * - calculating last 32 w[i] values in 8 XMM registers
365 * - pre-calculate K+w[i] values and store to mem, for later load by ALU add
366 *   instruction
367 *
368 * some "heavy-lifting" vectorization for rounds 16-31 due to w[i]->w[i-3]
369 * dependency, but improves for 32-79
370 */
371.macro W_PRECALC_16_31_SSSE3
372  # blended scheduling of vector and scalar instruction streams, one 4-wide
373  # vector iteration / 4 scalar rounds
374  .if ((i & 3) == 0)
375	movdqa	W_minus_12, W
376	palignr	$8, W_minus_16, W	# w[i-14]
377	movdqa	W_minus_04, W_TMP1
378	psrldq	$4, W_TMP1		# w[i-3]
379	pxor	W_minus_08, W
380  .elseif ((i & 3) == 1)
381	pxor	W_minus_16, W_TMP1
382	pxor	W_TMP1, W
383	movdqa	W, W_TMP2
384	movdqa	W, W_TMP1
385	pslldq	$12, W_TMP2
386  .elseif ((i & 3) == 2)
387	psrld	$31, W
388	pslld	$1, W_TMP1
389	por	W, W_TMP1
390	movdqa	W_TMP2, W
391	psrld	$30, W_TMP2
392	pslld	$2, W
393  .elseif ((i & 3) == 3)
394	pxor	W, W_TMP1
395	pxor	W_TMP2, W_TMP1
396	movdqa	W_TMP1, W
397	paddd	K_XMM(K_BASE), W_TMP1
398	movdqa	W_TMP1, WK(i&~3)
399	W_PRECALC_ROTATE
400  .endif
401.endm
402
403/* message scheduling pre-compute for rounds 32-79
404 *
405 * in SHA-1 specification: w[i] = (w[i-3] ^ w[i-8]  ^ w[i-14] ^ w[i-16]) rol 1
406 * instead we do equal:    w[i] = (w[i-6] ^ w[i-16] ^ w[i-28] ^ w[i-32]) rol 2
407 * allows more efficient vectorization since w[i]=>w[i-3] dependency is broken
408 */
409.macro W_PRECALC_32_79_SSSE3
410  .if ((i & 3) == 0)
411	movdqa	W_minus_04, W_TMP1
412	pxor	W_minus_28, W		# W is W_minus_32 before xor
413	palignr	$8, W_minus_08, W_TMP1
414  .elseif ((i & 3) == 1)
415	pxor	W_minus_16, W
416	pxor	W_TMP1, W
417	movdqa	W, W_TMP1
418  .elseif ((i & 3) == 2)
419	psrld	$30, W
420	pslld	$2, W_TMP1
421	por	W, W_TMP1
422  .elseif ((i & 3) == 3)
423	movdqa	W_TMP1, W
424	paddd	K_XMM(K_BASE), W_TMP1
425	movdqa	W_TMP1, WK(i&~3)
426	W_PRECALC_ROTATE
427  .endif
428.endm
429
430.endm		// W_PRECALC_SSSE3
431
432
433#define K1	0x5a827999
434#define K2	0x6ed9eba1
435#define K3	0x8f1bbcdc
436#define K4	0xca62c1d6
437
438.section .rodata
439.align 16
440
441K_XMM_AR:
442	.long K1, K1, K1, K1
443	.long K2, K2, K2, K2
444	.long K3, K3, K3, K3
445	.long K4, K4, K4, K4
446
447BSWAP_SHUFB_CTL:
448	.long 0x00010203
449	.long 0x04050607
450	.long 0x08090a0b
451	.long 0x0c0d0e0f
452
453
454.section .text
455
456W_PRECALC_SSSE3
457.macro xmm_mov a, b
458	movdqu	\a,\b
459.endm
460
461/*
462 * SSSE3 optimized implementation:
463 *
464 * extern "C" void sha1_transform_ssse3(struct sha1_state *state,
465 *					const u8 *data, int blocks);
466 *
467 * Note that struct sha1_state is assumed to begin with u32 state[5].
468 */
469SHA1_VECTOR_ASM     sha1_transform_ssse3
470
471.macro W_PRECALC_AVX
472
473.purgem W_PRECALC_00_15
474.macro  W_PRECALC_00_15
475    W_PRECALC_00_15_AVX
476.endm
477.purgem W_PRECALC_16_31
478.macro  W_PRECALC_16_31
479    W_PRECALC_16_31_AVX
480.endm
481.purgem W_PRECALC_32_79
482.macro  W_PRECALC_32_79
483    W_PRECALC_32_79_AVX
484.endm
485
486.macro W_PRECALC_00_15_AVX
487  .if ((i & 3) == 0)
488	vmovdqu	(i*4)(BUFFER_PTR), W_TMP1
489  .elseif ((i & 3) == 1)
490	vpshufb	XMM_SHUFB_BSWAP, W_TMP1, W
491  .elseif ((i & 3) == 2)
492	vpaddd	(K_BASE), W, W_TMP1
493  .elseif ((i & 3) == 3)
494	vmovdqa	W_TMP1, WK(i&~3)
495	W_PRECALC_ROTATE
496  .endif
497.endm
498
499.macro W_PRECALC_16_31_AVX
500  .if ((i & 3) == 0)
501	vpalignr $8, W_minus_16, W_minus_12, W	# w[i-14]
502	vpsrldq	$4, W_minus_04, W_TMP1		# w[i-3]
503	vpxor	W_minus_08, W, W
504	vpxor	W_minus_16, W_TMP1, W_TMP1
505  .elseif ((i & 3) == 1)
506	vpxor	W_TMP1, W, W
507	vpslldq	$12, W, W_TMP2
508	vpslld	$1, W, W_TMP1
509  .elseif ((i & 3) == 2)
510	vpsrld	$31, W, W
511	vpor	W, W_TMP1, W_TMP1
512	vpslld	$2, W_TMP2, W
513	vpsrld	$30, W_TMP2, W_TMP2
514  .elseif ((i & 3) == 3)
515	vpxor	W, W_TMP1, W_TMP1
516	vpxor	W_TMP2, W_TMP1, W
517	vpaddd	K_XMM(K_BASE), W, W_TMP1
518	vmovdqu	W_TMP1, WK(i&~3)
519	W_PRECALC_ROTATE
520  .endif
521.endm
522
523.macro W_PRECALC_32_79_AVX
524  .if ((i & 3) == 0)
525	vpalignr $8, W_minus_08, W_minus_04, W_TMP1
526	vpxor	W_minus_28, W, W		# W is W_minus_32 before xor
527  .elseif ((i & 3) == 1)
528	vpxor	W_minus_16, W_TMP1, W_TMP1
529	vpxor	W_TMP1, W, W
530  .elseif ((i & 3) == 2)
531	vpslld	$2, W, W_TMP1
532	vpsrld	$30, W, W
533	vpor	W, W_TMP1, W
534  .elseif ((i & 3) == 3)
535	vpaddd	K_XMM(K_BASE), W, W_TMP1
536	vmovdqu	W_TMP1, WK(i&~3)
537	W_PRECALC_ROTATE
538  .endif
539.endm
540
541.endm    // W_PRECALC_AVX
542
543W_PRECALC_AVX
544.purgem xmm_mov
545.macro xmm_mov a, b
546	vmovdqu	\a,\b
547.endm
548
549
550/* AVX optimized implementation:
551 *  extern "C" void sha1_transform_avx(struct sha1_state *state,
552 *				       const u8 *data, int blocks);
553 */
554SHA1_VECTOR_ASM     sha1_transform_avx
555