1/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
2//
3// This file is dual-licensed, meaning that you can use it under your
4// choice of either of the following two licenses:
5//
6// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
7//
8// Licensed under the Apache License 2.0 (the "License"). You can obtain
9// a copy in the file LICENSE in the source distribution or at
10// https://www.openssl.org/source/license.html
11//
12// or
13//
14// Copyright (c) 2023, Christoph M��llner <christoph.muellner@vrull.eu>
15// Copyright (c) 2023, Phoebe Chen <phoebe.chen@sifive.com>
16// Copyright 2024 Google LLC
17// All rights reserved.
18//
19// Redistribution and use in source and binary forms, with or without
20// modification, are permitted provided that the following conditions
21// are met:
22// 1. Redistributions of source code must retain the above copyright
23//    notice, this list of conditions and the following disclaimer.
24// 2. Redistributions in binary form must reproduce the above copyright
25//    notice, this list of conditions and the following disclaimer in the
26//    documentation and/or other materials provided with the distribution.
27//
28// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
29// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
30// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
31// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
32// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
33// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
34// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
35// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
36// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
37// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
38// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
39
40// The generated code of this file depends on the following RISC-V extensions:
41// - RV64I
42// - RISC-V Vector ('V') with VLEN >= 128
43// - RISC-V Vector SHA-2 Secure Hash extension ('Zvknhb')
44// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb')
45
46#include <linux/cfi_types.h>
47
48.text
49.option arch, +zvknhb, +zvkb
50
51#define STATEP		a0
52#define DATA		a1
53#define NUM_BLOCKS	a2
54
55#define STATEP_C	a3
56#define K		a4
57
58#define MASK		v0
59#define INDICES		v1
60#define W0		v10	// LMUL=2
61#define W1		v12	// LMUL=2
62#define W2		v14	// LMUL=2
63#define W3		v16	// LMUL=2
64#define VTMP		v20	// LMUL=2
65#define FEBA		v22	// LMUL=2
66#define HGDC		v24	// LMUL=2
67#define PREV_FEBA	v26	// LMUL=2
68#define PREV_HGDC	v28	// LMUL=2
69
70// Do 4 rounds of SHA-512.  w0 contains the current 4 message schedule words.
71//
72// If not all the message schedule words have been computed yet, then this also
73// computes 4 more message schedule words.  w1-w3 contain the next 3 groups of 4
74// message schedule words; this macro computes the group after w3 and writes it
75// to w0.  This means that the next (w0, w1, w2, w3) is the current (w1, w2, w3,
76// w0), so the caller must cycle through the registers accordingly.
77.macro	sha512_4rounds	last, w0, w1, w2, w3
78	vle64.v		VTMP, (K)
79	addi		K, K, 32
80	vadd.vv		VTMP, VTMP, \w0
81	vsha2cl.vv	HGDC, FEBA, VTMP
82	vsha2ch.vv	FEBA, HGDC, VTMP
83.if !\last
84	vmerge.vvm	VTMP, \w2, \w1, MASK
85	vsha2ms.vv	\w0, VTMP, \w3
86.endif
87.endm
88
89.macro	sha512_16rounds	last
90	sha512_4rounds	\last, W0, W1, W2, W3
91	sha512_4rounds	\last, W1, W2, W3, W0
92	sha512_4rounds	\last, W2, W3, W0, W1
93	sha512_4rounds	\last, W3, W0, W1, W2
94.endm
95
96// void sha512_transform_zvknhb_zvkb(u64 state[8], const u8 *data,
97//				     int num_blocks);
98SYM_TYPED_FUNC_START(sha512_transform_zvknhb_zvkb)
99
100	// Setup mask for the vmerge to replace the first word (idx==0) in
101	// message scheduling.  There are 4 words, so an 8-bit mask suffices.
102	vsetivli	zero, 1, e8, m1, ta, ma
103	vmv.v.i		MASK, 0x01
104
105	// Load the state.  The state is stored as {a,b,c,d,e,f,g,h}, but we
106	// need {f,e,b,a},{h,g,d,c}.  The dst vtype is e64m2 and the index vtype
107	// is e8mf4.  We use index-load with the i8 indices {40, 32, 8, 0},
108	// loaded using the 32-bit little endian value 0x00082028.
109	li		t0, 0x00082028
110	vsetivli	zero, 1, e32, m1, ta, ma
111	vmv.v.x		INDICES, t0
112	addi		STATEP_C, STATEP, 16
113	vsetivli	zero, 4, e64, m2, ta, ma
114	vluxei8.v	FEBA, (STATEP), INDICES
115	vluxei8.v	HGDC, (STATEP_C), INDICES
116
117.Lnext_block:
118	la		K, K512
119	addi		NUM_BLOCKS, NUM_BLOCKS, -1
120
121	// Save the previous state, as it's needed later.
122	vmv.v.v		PREV_FEBA, FEBA
123	vmv.v.v		PREV_HGDC, HGDC
124
125	// Load the next 1024-bit message block and endian-swap each 64-bit word
126	vle64.v		W0, (DATA)
127	vrev8.v		W0, W0
128	addi		DATA, DATA, 32
129	vle64.v		W1, (DATA)
130	vrev8.v		W1, W1
131	addi		DATA, DATA, 32
132	vle64.v		W2, (DATA)
133	vrev8.v		W2, W2
134	addi		DATA, DATA, 32
135	vle64.v		W3, (DATA)
136	vrev8.v		W3, W3
137	addi		DATA, DATA, 32
138
139	// Do the 80 rounds of SHA-512.
140	sha512_16rounds 0
141	sha512_16rounds 0
142	sha512_16rounds 0
143	sha512_16rounds 0
144	sha512_16rounds 1
145
146	// Add the previous state.
147	vadd.vv		FEBA, FEBA, PREV_FEBA
148	vadd.vv		HGDC, HGDC, PREV_HGDC
149
150	// Repeat if more blocks remain.
151	bnez		NUM_BLOCKS, .Lnext_block
152
153	// Store the new state and return.
154	vsuxei8.v	FEBA, (STATEP), INDICES
155	vsuxei8.v	HGDC, (STATEP_C), INDICES
156	ret
157SYM_FUNC_END(sha512_transform_zvknhb_zvkb)
158
159.section ".rodata"
160.p2align 3
161.type K512, @object
162K512:
163	.dword		0x428a2f98d728ae22, 0x7137449123ef65cd
164	.dword		0xb5c0fbcfec4d3b2f, 0xe9b5dba58189dbbc
165	.dword		0x3956c25bf348b538, 0x59f111f1b605d019
166	.dword		0x923f82a4af194f9b, 0xab1c5ed5da6d8118
167	.dword		0xd807aa98a3030242, 0x12835b0145706fbe
168	.dword		0x243185be4ee4b28c, 0x550c7dc3d5ffb4e2
169	.dword		0x72be5d74f27b896f, 0x80deb1fe3b1696b1
170	.dword		0x9bdc06a725c71235, 0xc19bf174cf692694
171	.dword		0xe49b69c19ef14ad2, 0xefbe4786384f25e3
172	.dword		0x0fc19dc68b8cd5b5, 0x240ca1cc77ac9c65
173	.dword		0x2de92c6f592b0275, 0x4a7484aa6ea6e483
174	.dword		0x5cb0a9dcbd41fbd4, 0x76f988da831153b5
175	.dword		0x983e5152ee66dfab, 0xa831c66d2db43210
176	.dword		0xb00327c898fb213f, 0xbf597fc7beef0ee4
177	.dword		0xc6e00bf33da88fc2, 0xd5a79147930aa725
178	.dword		0x06ca6351e003826f, 0x142929670a0e6e70
179	.dword		0x27b70a8546d22ffc, 0x2e1b21385c26c926
180	.dword		0x4d2c6dfc5ac42aed, 0x53380d139d95b3df
181	.dword		0x650a73548baf63de, 0x766a0abb3c77b2a8
182	.dword		0x81c2c92e47edaee6, 0x92722c851482353b
183	.dword		0xa2bfe8a14cf10364, 0xa81a664bbc423001
184	.dword		0xc24b8b70d0f89791, 0xc76c51a30654be30
185	.dword		0xd192e819d6ef5218, 0xd69906245565a910
186	.dword		0xf40e35855771202a, 0x106aa07032bbd1b8
187	.dword		0x19a4c116b8d2d0c8, 0x1e376c085141ab53
188	.dword		0x2748774cdf8eeb99, 0x34b0bcb5e19b48a8
189	.dword		0x391c0cb3c5c95a63, 0x4ed8aa4ae3418acb
190	.dword		0x5b9cca4f7763e373, 0x682e6ff3d6b2b8a3
191	.dword		0x748f82ee5defb2fc, 0x78a5636f43172f60
192	.dword		0x84c87814a1f0ab72, 0x8cc702081a6439ec
193	.dword		0x90befffa23631e28, 0xa4506cebde82bde9
194	.dword		0xbef9a3f7b2c67915, 0xc67178f2e372532b
195	.dword		0xca273eceea26619c, 0xd186b8c721c0c207
196	.dword		0xeada7dd6cde0eb1e, 0xf57d4f7fee6ed178
197	.dword		0x06f067aa72176fba, 0x0a637dc5a2c898a6
198	.dword		0x113f9804bef90dae, 0x1b710b35131c471b
199	.dword		0x28db77f523047d84, 0x32caab7b40c72493
200	.dword		0x3c9ebe0a15c9bebc, 0x431d67c49c100d4c
201	.dword		0x4cc5d4becb3e42b6, 0x597f299cfc657e2a
202	.dword		0x5fcb6fab3ad6faec, 0x6c44198c4a475817
203.size K512, . - K512
204