xref: /openbmc/qemu/target/riscv/vcrypto_helper.c (revision bad5cfcd)
1 /*
2  * RISC-V Vector Crypto Extension Helpers for QEMU.
3  *
4  * Copyright (C) 2023 SiFive, Inc.
5  * Written by Codethink Ltd and SiFive.
6  *
7  * This program is free software; you can redistribute it and/or modify it
8  * under the terms and conditions of the GNU General Public License,
9  * version 2 or later, as published by the Free Software Foundation.
10  *
11  * This program is distributed in the hope it will be useful, but WITHOUT
12  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
13  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
14  * more details.
15  *
16  * You should have received a copy of the GNU General Public License along with
17  * this program.  If not, see <http://www.gnu.org/licenses/>.
18  */
19 
20 #include "qemu/osdep.h"
21 #include "qemu/host-utils.h"
22 #include "qemu/bitops.h"
23 #include "qemu/bswap.h"
24 #include "cpu.h"
25 #include "crypto/aes.h"
26 #include "crypto/aes-round.h"
27 #include "crypto/sm4.h"
28 #include "exec/memop.h"
29 #include "exec/exec-all.h"
30 #include "exec/helper-proto.h"
31 #include "internals.h"
32 #include "vector_internals.h"
33 
34 static uint64_t clmul64(uint64_t y, uint64_t x)
35 {
36     uint64_t result = 0;
37     for (int j = 63; j >= 0; j--) {
38         if ((y >> j) & 1) {
39             result ^= (x << j);
40         }
41     }
42     return result;
43 }
44 
45 static uint64_t clmulh64(uint64_t y, uint64_t x)
46 {
47     uint64_t result = 0;
48     for (int j = 63; j >= 1; j--) {
49         if ((y >> j) & 1) {
50             result ^= (x >> (64 - j));
51         }
52     }
53     return result;
54 }
55 
56 RVVCALL(OPIVV2, vclmul_vv, OP_UUU_D, H8, H8, H8, clmul64)
57 GEN_VEXT_VV(vclmul_vv, 8)
58 RVVCALL(OPIVX2, vclmul_vx, OP_UUU_D, H8, H8, clmul64)
59 GEN_VEXT_VX(vclmul_vx, 8)
60 RVVCALL(OPIVV2, vclmulh_vv, OP_UUU_D, H8, H8, H8, clmulh64)
61 GEN_VEXT_VV(vclmulh_vv, 8)
62 RVVCALL(OPIVX2, vclmulh_vx, OP_UUU_D, H8, H8, clmulh64)
63 GEN_VEXT_VX(vclmulh_vx, 8)
64 
65 RVVCALL(OPIVV2, vror_vv_b, OP_UUU_B, H1, H1, H1, ror8)
66 RVVCALL(OPIVV2, vror_vv_h, OP_UUU_H, H2, H2, H2, ror16)
67 RVVCALL(OPIVV2, vror_vv_w, OP_UUU_W, H4, H4, H4, ror32)
68 RVVCALL(OPIVV2, vror_vv_d, OP_UUU_D, H8, H8, H8, ror64)
69 GEN_VEXT_VV(vror_vv_b, 1)
70 GEN_VEXT_VV(vror_vv_h, 2)
71 GEN_VEXT_VV(vror_vv_w, 4)
72 GEN_VEXT_VV(vror_vv_d, 8)
73 
74 RVVCALL(OPIVX2, vror_vx_b, OP_UUU_B, H1, H1, ror8)
75 RVVCALL(OPIVX2, vror_vx_h, OP_UUU_H, H2, H2, ror16)
76 RVVCALL(OPIVX2, vror_vx_w, OP_UUU_W, H4, H4, ror32)
77 RVVCALL(OPIVX2, vror_vx_d, OP_UUU_D, H8, H8, ror64)
78 GEN_VEXT_VX(vror_vx_b, 1)
79 GEN_VEXT_VX(vror_vx_h, 2)
80 GEN_VEXT_VX(vror_vx_w, 4)
81 GEN_VEXT_VX(vror_vx_d, 8)
82 
83 RVVCALL(OPIVV2, vrol_vv_b, OP_UUU_B, H1, H1, H1, rol8)
84 RVVCALL(OPIVV2, vrol_vv_h, OP_UUU_H, H2, H2, H2, rol16)
85 RVVCALL(OPIVV2, vrol_vv_w, OP_UUU_W, H4, H4, H4, rol32)
86 RVVCALL(OPIVV2, vrol_vv_d, OP_UUU_D, H8, H8, H8, rol64)
87 GEN_VEXT_VV(vrol_vv_b, 1)
88 GEN_VEXT_VV(vrol_vv_h, 2)
89 GEN_VEXT_VV(vrol_vv_w, 4)
90 GEN_VEXT_VV(vrol_vv_d, 8)
91 
92 RVVCALL(OPIVX2, vrol_vx_b, OP_UUU_B, H1, H1, rol8)
93 RVVCALL(OPIVX2, vrol_vx_h, OP_UUU_H, H2, H2, rol16)
94 RVVCALL(OPIVX2, vrol_vx_w, OP_UUU_W, H4, H4, rol32)
95 RVVCALL(OPIVX2, vrol_vx_d, OP_UUU_D, H8, H8, rol64)
96 GEN_VEXT_VX(vrol_vx_b, 1)
97 GEN_VEXT_VX(vrol_vx_h, 2)
98 GEN_VEXT_VX(vrol_vx_w, 4)
99 GEN_VEXT_VX(vrol_vx_d, 8)
100 
101 static uint64_t brev8(uint64_t val)
102 {
103     val = ((val & 0x5555555555555555ull) << 1) |
104           ((val & 0xAAAAAAAAAAAAAAAAull) >> 1);
105     val = ((val & 0x3333333333333333ull) << 2) |
106           ((val & 0xCCCCCCCCCCCCCCCCull) >> 2);
107     val = ((val & 0x0F0F0F0F0F0F0F0Full) << 4) |
108           ((val & 0xF0F0F0F0F0F0F0F0ull) >> 4);
109 
110     return val;
111 }
112 
113 RVVCALL(OPIVV1, vbrev8_v_b, OP_UU_B, H1, H1, brev8)
114 RVVCALL(OPIVV1, vbrev8_v_h, OP_UU_H, H2, H2, brev8)
115 RVVCALL(OPIVV1, vbrev8_v_w, OP_UU_W, H4, H4, brev8)
116 RVVCALL(OPIVV1, vbrev8_v_d, OP_UU_D, H8, H8, brev8)
117 GEN_VEXT_V(vbrev8_v_b, 1)
118 GEN_VEXT_V(vbrev8_v_h, 2)
119 GEN_VEXT_V(vbrev8_v_w, 4)
120 GEN_VEXT_V(vbrev8_v_d, 8)
121 
122 #define DO_IDENTITY(a) (a)
123 RVVCALL(OPIVV1, vrev8_v_b, OP_UU_B, H1, H1, DO_IDENTITY)
124 RVVCALL(OPIVV1, vrev8_v_h, OP_UU_H, H2, H2, bswap16)
125 RVVCALL(OPIVV1, vrev8_v_w, OP_UU_W, H4, H4, bswap32)
126 RVVCALL(OPIVV1, vrev8_v_d, OP_UU_D, H8, H8, bswap64)
127 GEN_VEXT_V(vrev8_v_b, 1)
128 GEN_VEXT_V(vrev8_v_h, 2)
129 GEN_VEXT_V(vrev8_v_w, 4)
130 GEN_VEXT_V(vrev8_v_d, 8)
131 
132 #define DO_ANDN(a, b) ((a) & ~(b))
133 RVVCALL(OPIVV2, vandn_vv_b, OP_UUU_B, H1, H1, H1, DO_ANDN)
134 RVVCALL(OPIVV2, vandn_vv_h, OP_UUU_H, H2, H2, H2, DO_ANDN)
135 RVVCALL(OPIVV2, vandn_vv_w, OP_UUU_W, H4, H4, H4, DO_ANDN)
136 RVVCALL(OPIVV2, vandn_vv_d, OP_UUU_D, H8, H8, H8, DO_ANDN)
137 GEN_VEXT_VV(vandn_vv_b, 1)
138 GEN_VEXT_VV(vandn_vv_h, 2)
139 GEN_VEXT_VV(vandn_vv_w, 4)
140 GEN_VEXT_VV(vandn_vv_d, 8)
141 
142 RVVCALL(OPIVX2, vandn_vx_b, OP_UUU_B, H1, H1, DO_ANDN)
143 RVVCALL(OPIVX2, vandn_vx_h, OP_UUU_H, H2, H2, DO_ANDN)
144 RVVCALL(OPIVX2, vandn_vx_w, OP_UUU_W, H4, H4, DO_ANDN)
145 RVVCALL(OPIVX2, vandn_vx_d, OP_UUU_D, H8, H8, DO_ANDN)
146 GEN_VEXT_VX(vandn_vx_b, 1)
147 GEN_VEXT_VX(vandn_vx_h, 2)
148 GEN_VEXT_VX(vandn_vx_w, 4)
149 GEN_VEXT_VX(vandn_vx_d, 8)
150 
151 RVVCALL(OPIVV1, vbrev_v_b, OP_UU_B, H1, H1, revbit8)
152 RVVCALL(OPIVV1, vbrev_v_h, OP_UU_H, H2, H2, revbit16)
153 RVVCALL(OPIVV1, vbrev_v_w, OP_UU_W, H4, H4, revbit32)
154 RVVCALL(OPIVV1, vbrev_v_d, OP_UU_D, H8, H8, revbit64)
155 GEN_VEXT_V(vbrev_v_b, 1)
156 GEN_VEXT_V(vbrev_v_h, 2)
157 GEN_VEXT_V(vbrev_v_w, 4)
158 GEN_VEXT_V(vbrev_v_d, 8)
159 
160 RVVCALL(OPIVV1, vclz_v_b, OP_UU_B, H1, H1, clz8)
161 RVVCALL(OPIVV1, vclz_v_h, OP_UU_H, H2, H2, clz16)
162 RVVCALL(OPIVV1, vclz_v_w, OP_UU_W, H4, H4, clz32)
163 RVVCALL(OPIVV1, vclz_v_d, OP_UU_D, H8, H8, clz64)
164 GEN_VEXT_V(vclz_v_b, 1)
165 GEN_VEXT_V(vclz_v_h, 2)
166 GEN_VEXT_V(vclz_v_w, 4)
167 GEN_VEXT_V(vclz_v_d, 8)
168 
169 RVVCALL(OPIVV1, vctz_v_b, OP_UU_B, H1, H1, ctz8)
170 RVVCALL(OPIVV1, vctz_v_h, OP_UU_H, H2, H2, ctz16)
171 RVVCALL(OPIVV1, vctz_v_w, OP_UU_W, H4, H4, ctz32)
172 RVVCALL(OPIVV1, vctz_v_d, OP_UU_D, H8, H8, ctz64)
173 GEN_VEXT_V(vctz_v_b, 1)
174 GEN_VEXT_V(vctz_v_h, 2)
175 GEN_VEXT_V(vctz_v_w, 4)
176 GEN_VEXT_V(vctz_v_d, 8)
177 
178 RVVCALL(OPIVV1, vcpop_v_b, OP_UU_B, H1, H1, ctpop8)
179 RVVCALL(OPIVV1, vcpop_v_h, OP_UU_H, H2, H2, ctpop16)
180 RVVCALL(OPIVV1, vcpop_v_w, OP_UU_W, H4, H4, ctpop32)
181 RVVCALL(OPIVV1, vcpop_v_d, OP_UU_D, H8, H8, ctpop64)
182 GEN_VEXT_V(vcpop_v_b, 1)
183 GEN_VEXT_V(vcpop_v_h, 2)
184 GEN_VEXT_V(vcpop_v_w, 4)
185 GEN_VEXT_V(vcpop_v_d, 8)
186 
187 #define DO_SLL(N, M) (N << (M & (sizeof(N) * 8 - 1)))
188 RVVCALL(OPIVV2, vwsll_vv_b, WOP_UUU_B, H2, H1, H1, DO_SLL)
189 RVVCALL(OPIVV2, vwsll_vv_h, WOP_UUU_H, H4, H2, H2, DO_SLL)
190 RVVCALL(OPIVV2, vwsll_vv_w, WOP_UUU_W, H8, H4, H4, DO_SLL)
191 GEN_VEXT_VV(vwsll_vv_b, 2)
192 GEN_VEXT_VV(vwsll_vv_h, 4)
193 GEN_VEXT_VV(vwsll_vv_w, 8)
194 
195 RVVCALL(OPIVX2, vwsll_vx_b, WOP_UUU_B, H2, H1, DO_SLL)
196 RVVCALL(OPIVX2, vwsll_vx_h, WOP_UUU_H, H4, H2, DO_SLL)
197 RVVCALL(OPIVX2, vwsll_vx_w, WOP_UUU_W, H8, H4, DO_SLL)
198 GEN_VEXT_VX(vwsll_vx_b, 2)
199 GEN_VEXT_VX(vwsll_vx_h, 4)
200 GEN_VEXT_VX(vwsll_vx_w, 8)
201 
202 void HELPER(egs_check)(uint32_t egs, CPURISCVState *env)
203 {
204     uint32_t vl = env->vl;
205     uint32_t vstart = env->vstart;
206 
207     if (vl % egs != 0 || vstart % egs != 0) {
208         riscv_raise_exception(env, RISCV_EXCP_ILLEGAL_INST, GETPC());
209     }
210 }
211 
212 static inline void xor_round_key(AESState *round_state, AESState *round_key)
213 {
214     round_state->v = round_state->v ^ round_key->v;
215 }
216 
217 #define GEN_ZVKNED_HELPER_VV(NAME, ...)                                   \
218     void HELPER(NAME)(void *vd, void *vs2, CPURISCVState *env,            \
219                       uint32_t desc)                                      \
220     {                                                                     \
221         uint32_t vl = env->vl;                                            \
222         uint32_t total_elems = vext_get_total_elems(env, desc, 4);        \
223         uint32_t vta = vext_vta(desc);                                    \
224                                                                           \
225         for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {        \
226             AESState round_key;                                           \
227             round_key.d[0] = *((uint64_t *)vs2 + H8(i * 2 + 0));          \
228             round_key.d[1] = *((uint64_t *)vs2 + H8(i * 2 + 1));          \
229             AESState round_state;                                         \
230             round_state.d[0] = *((uint64_t *)vd + H8(i * 2 + 0));         \
231             round_state.d[1] = *((uint64_t *)vd + H8(i * 2 + 1));         \
232             __VA_ARGS__;                                                  \
233             *((uint64_t *)vd + H8(i * 2 + 0)) = round_state.d[0];         \
234             *((uint64_t *)vd + H8(i * 2 + 1)) = round_state.d[1];         \
235         }                                                                 \
236         env->vstart = 0;                                                  \
237         /* set tail elements to 1s */                                     \
238         vext_set_elems_1s(vd, vta, vl * 4, total_elems * 4);              \
239     }
240 
241 #define GEN_ZVKNED_HELPER_VS(NAME, ...)                                   \
242     void HELPER(NAME)(void *vd, void *vs2, CPURISCVState *env,            \
243                       uint32_t desc)                                      \
244     {                                                                     \
245         uint32_t vl = env->vl;                                            \
246         uint32_t total_elems = vext_get_total_elems(env, desc, 4);        \
247         uint32_t vta = vext_vta(desc);                                    \
248                                                                           \
249         for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {        \
250             AESState round_key;                                           \
251             round_key.d[0] = *((uint64_t *)vs2 + H8(0));                  \
252             round_key.d[1] = *((uint64_t *)vs2 + H8(1));                  \
253             AESState round_state;                                         \
254             round_state.d[0] = *((uint64_t *)vd + H8(i * 2 + 0));         \
255             round_state.d[1] = *((uint64_t *)vd + H8(i * 2 + 1));         \
256             __VA_ARGS__;                                                  \
257             *((uint64_t *)vd + H8(i * 2 + 0)) = round_state.d[0];         \
258             *((uint64_t *)vd + H8(i * 2 + 1)) = round_state.d[1];         \
259         }                                                                 \
260         env->vstart = 0;                                                  \
261         /* set tail elements to 1s */                                     \
262         vext_set_elems_1s(vd, vta, vl * 4, total_elems * 4);              \
263     }
264 
265 GEN_ZVKNED_HELPER_VV(vaesef_vv, aesenc_SB_SR_AK(&round_state,
266                                                 &round_state,
267                                                 &round_key,
268                                                 false);)
269 GEN_ZVKNED_HELPER_VS(vaesef_vs, aesenc_SB_SR_AK(&round_state,
270                                                 &round_state,
271                                                 &round_key,
272                                                 false);)
273 GEN_ZVKNED_HELPER_VV(vaesdf_vv, aesdec_ISB_ISR_AK(&round_state,
274                                                   &round_state,
275                                                   &round_key,
276                                                   false);)
277 GEN_ZVKNED_HELPER_VS(vaesdf_vs, aesdec_ISB_ISR_AK(&round_state,
278                                                   &round_state,
279                                                   &round_key,
280                                                   false);)
281 GEN_ZVKNED_HELPER_VV(vaesem_vv, aesenc_SB_SR_MC_AK(&round_state,
282                                                    &round_state,
283                                                    &round_key,
284                                                    false);)
285 GEN_ZVKNED_HELPER_VS(vaesem_vs, aesenc_SB_SR_MC_AK(&round_state,
286                                                    &round_state,
287                                                    &round_key,
288                                                    false);)
289 GEN_ZVKNED_HELPER_VV(vaesdm_vv, aesdec_ISB_ISR_AK_IMC(&round_state,
290                                                       &round_state,
291                                                       &round_key,
292                                                       false);)
293 GEN_ZVKNED_HELPER_VS(vaesdm_vs, aesdec_ISB_ISR_AK_IMC(&round_state,
294                                                       &round_state,
295                                                       &round_key,
296                                                       false);)
297 GEN_ZVKNED_HELPER_VS(vaesz_vs, xor_round_key(&round_state, &round_key);)
298 
299 void HELPER(vaeskf1_vi)(void *vd_vptr, void *vs2_vptr, uint32_t uimm,
300                         CPURISCVState *env, uint32_t desc)
301 {
302     uint32_t *vd = vd_vptr;
303     uint32_t *vs2 = vs2_vptr;
304     uint32_t vl = env->vl;
305     uint32_t total_elems = vext_get_total_elems(env, desc, 4);
306     uint32_t vta = vext_vta(desc);
307 
308     uimm &= 0b1111;
309     if (uimm > 10 || uimm == 0) {
310         uimm ^= 0b1000;
311     }
312 
313     for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
314         uint32_t rk[8], tmp;
315         static const uint32_t rcon[] = {
316             0x00000001, 0x00000002, 0x00000004, 0x00000008, 0x00000010,
317             0x00000020, 0x00000040, 0x00000080, 0x0000001B, 0x00000036,
318         };
319 
320         rk[0] = vs2[i * 4 + H4(0)];
321         rk[1] = vs2[i * 4 + H4(1)];
322         rk[2] = vs2[i * 4 + H4(2)];
323         rk[3] = vs2[i * 4 + H4(3)];
324         tmp = ror32(rk[3], 8);
325 
326         rk[4] = rk[0] ^ (((uint32_t)AES_sbox[(tmp >> 24) & 0xff] << 24) |
327                          ((uint32_t)AES_sbox[(tmp >> 16) & 0xff] << 16) |
328                          ((uint32_t)AES_sbox[(tmp >> 8) & 0xff] << 8) |
329                          ((uint32_t)AES_sbox[(tmp >> 0) & 0xff] << 0))
330                       ^ rcon[uimm - 1];
331         rk[5] = rk[1] ^ rk[4];
332         rk[6] = rk[2] ^ rk[5];
333         rk[7] = rk[3] ^ rk[6];
334 
335         vd[i * 4 + H4(0)] = rk[4];
336         vd[i * 4 + H4(1)] = rk[5];
337         vd[i * 4 + H4(2)] = rk[6];
338         vd[i * 4 + H4(3)] = rk[7];
339     }
340     env->vstart = 0;
341     /* set tail elements to 1s */
342     vext_set_elems_1s(vd, vta, vl * 4, total_elems * 4);
343 }
344 
345 void HELPER(vaeskf2_vi)(void *vd_vptr, void *vs2_vptr, uint32_t uimm,
346                         CPURISCVState *env, uint32_t desc)
347 {
348     uint32_t *vd = vd_vptr;
349     uint32_t *vs2 = vs2_vptr;
350     uint32_t vl = env->vl;
351     uint32_t total_elems = vext_get_total_elems(env, desc, 4);
352     uint32_t vta = vext_vta(desc);
353 
354     uimm &= 0b1111;
355     if (uimm > 14 || uimm < 2) {
356         uimm ^= 0b1000;
357     }
358 
359     for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
360         uint32_t rk[12], tmp;
361         static const uint32_t rcon[] = {
362             0x00000001, 0x00000002, 0x00000004, 0x00000008, 0x00000010,
363             0x00000020, 0x00000040, 0x00000080, 0x0000001B, 0x00000036,
364         };
365 
366         rk[0] = vd[i * 4 + H4(0)];
367         rk[1] = vd[i * 4 + H4(1)];
368         rk[2] = vd[i * 4 + H4(2)];
369         rk[3] = vd[i * 4 + H4(3)];
370         rk[4] = vs2[i * 4 + H4(0)];
371         rk[5] = vs2[i * 4 + H4(1)];
372         rk[6] = vs2[i * 4 + H4(2)];
373         rk[7] = vs2[i * 4 + H4(3)];
374 
375         if (uimm % 2 == 0) {
376             tmp = ror32(rk[7], 8);
377             rk[8] = rk[0] ^ (((uint32_t)AES_sbox[(tmp >> 24) & 0xff] << 24) |
378                              ((uint32_t)AES_sbox[(tmp >> 16) & 0xff] << 16) |
379                              ((uint32_t)AES_sbox[(tmp >> 8) & 0xff] << 8) |
380                              ((uint32_t)AES_sbox[(tmp >> 0) & 0xff] << 0))
381                           ^ rcon[(uimm - 1) / 2];
382         } else {
383             rk[8] = rk[0] ^ (((uint32_t)AES_sbox[(rk[7] >> 24) & 0xff] << 24) |
384                              ((uint32_t)AES_sbox[(rk[7] >> 16) & 0xff] << 16) |
385                              ((uint32_t)AES_sbox[(rk[7] >> 8) & 0xff] << 8) |
386                              ((uint32_t)AES_sbox[(rk[7] >> 0) & 0xff] << 0));
387         }
388         rk[9] = rk[1] ^ rk[8];
389         rk[10] = rk[2] ^ rk[9];
390         rk[11] = rk[3] ^ rk[10];
391 
392         vd[i * 4 + H4(0)] = rk[8];
393         vd[i * 4 + H4(1)] = rk[9];
394         vd[i * 4 + H4(2)] = rk[10];
395         vd[i * 4 + H4(3)] = rk[11];
396     }
397     env->vstart = 0;
398     /* set tail elements to 1s */
399     vext_set_elems_1s(vd, vta, vl * 4, total_elems * 4);
400 }
401 
402 static inline uint32_t sig0_sha256(uint32_t x)
403 {
404     return ror32(x, 7) ^ ror32(x, 18) ^ (x >> 3);
405 }
406 
407 static inline uint32_t sig1_sha256(uint32_t x)
408 {
409     return ror32(x, 17) ^ ror32(x, 19) ^ (x >> 10);
410 }
411 
412 static inline uint64_t sig0_sha512(uint64_t x)
413 {
414     return ror64(x, 1) ^ ror64(x, 8) ^ (x >> 7);
415 }
416 
417 static inline uint64_t sig1_sha512(uint64_t x)
418 {
419     return ror64(x, 19) ^ ror64(x, 61) ^ (x >> 6);
420 }
421 
422 static inline void vsha2ms_e32(uint32_t *vd, uint32_t *vs1, uint32_t *vs2)
423 {
424     uint32_t res[4];
425     res[0] = sig1_sha256(vs1[H4(2)]) + vs2[H4(1)] + sig0_sha256(vd[H4(1)]) +
426              vd[H4(0)];
427     res[1] = sig1_sha256(vs1[H4(3)]) + vs2[H4(2)] + sig0_sha256(vd[H4(2)]) +
428              vd[H4(1)];
429     res[2] =
430         sig1_sha256(res[0]) + vs2[H4(3)] + sig0_sha256(vd[H4(3)]) + vd[H4(2)];
431     res[3] =
432         sig1_sha256(res[1]) + vs1[H4(0)] + sig0_sha256(vs2[H4(0)]) + vd[H4(3)];
433     vd[H4(3)] = res[3];
434     vd[H4(2)] = res[2];
435     vd[H4(1)] = res[1];
436     vd[H4(0)] = res[0];
437 }
438 
439 static inline void vsha2ms_e64(uint64_t *vd, uint64_t *vs1, uint64_t *vs2)
440 {
441     uint64_t res[4];
442     res[0] = sig1_sha512(vs1[2]) + vs2[1] + sig0_sha512(vd[1]) + vd[0];
443     res[1] = sig1_sha512(vs1[3]) + vs2[2] + sig0_sha512(vd[2]) + vd[1];
444     res[2] = sig1_sha512(res[0]) + vs2[3] + sig0_sha512(vd[3]) + vd[2];
445     res[3] = sig1_sha512(res[1]) + vs1[0] + sig0_sha512(vs2[0]) + vd[3];
446     vd[3] = res[3];
447     vd[2] = res[2];
448     vd[1] = res[1];
449     vd[0] = res[0];
450 }
451 
452 void HELPER(vsha2ms_vv)(void *vd, void *vs1, void *vs2, CPURISCVState *env,
453                         uint32_t desc)
454 {
455     uint32_t sew = FIELD_EX64(env->vtype, VTYPE, VSEW);
456     uint32_t esz = sew == MO_32 ? 4 : 8;
457     uint32_t total_elems;
458     uint32_t vta = vext_vta(desc);
459 
460     for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
461         if (sew == MO_32) {
462             vsha2ms_e32(((uint32_t *)vd) + i * 4, ((uint32_t *)vs1) + i * 4,
463                         ((uint32_t *)vs2) + i * 4);
464         } else {
465             /* If not 32 then SEW should be 64 */
466             vsha2ms_e64(((uint64_t *)vd) + i * 4, ((uint64_t *)vs1) + i * 4,
467                         ((uint64_t *)vs2) + i * 4);
468         }
469     }
470     /* set tail elements to 1s */
471     total_elems = vext_get_total_elems(env, desc, esz);
472     vext_set_elems_1s(vd, vta, env->vl * esz, total_elems * esz);
473     env->vstart = 0;
474 }
475 
476 static inline uint64_t sum0_64(uint64_t x)
477 {
478     return ror64(x, 28) ^ ror64(x, 34) ^ ror64(x, 39);
479 }
480 
481 static inline uint32_t sum0_32(uint32_t x)
482 {
483     return ror32(x, 2) ^ ror32(x, 13) ^ ror32(x, 22);
484 }
485 
486 static inline uint64_t sum1_64(uint64_t x)
487 {
488     return ror64(x, 14) ^ ror64(x, 18) ^ ror64(x, 41);
489 }
490 
491 static inline uint32_t sum1_32(uint32_t x)
492 {
493     return ror32(x, 6) ^ ror32(x, 11) ^ ror32(x, 25);
494 }
495 
496 #define ch(x, y, z) ((x & y) ^ ((~x) & z))
497 
498 #define maj(x, y, z) ((x & y) ^ (x & z) ^ (y & z))
499 
500 static void vsha2c_64(uint64_t *vs2, uint64_t *vd, uint64_t *vs1)
501 {
502     uint64_t a = vs2[3], b = vs2[2], e = vs2[1], f = vs2[0];
503     uint64_t c = vd[3], d = vd[2], g = vd[1], h = vd[0];
504     uint64_t W0 = vs1[0], W1 = vs1[1];
505     uint64_t T1 = h + sum1_64(e) + ch(e, f, g) + W0;
506     uint64_t T2 = sum0_64(a) + maj(a, b, c);
507 
508     h = g;
509     g = f;
510     f = e;
511     e = d + T1;
512     d = c;
513     c = b;
514     b = a;
515     a = T1 + T2;
516 
517     T1 = h + sum1_64(e) + ch(e, f, g) + W1;
518     T2 = sum0_64(a) + maj(a, b, c);
519     h = g;
520     g = f;
521     f = e;
522     e = d + T1;
523     d = c;
524     c = b;
525     b = a;
526     a = T1 + T2;
527 
528     vd[0] = f;
529     vd[1] = e;
530     vd[2] = b;
531     vd[3] = a;
532 }
533 
534 static void vsha2c_32(uint32_t *vs2, uint32_t *vd, uint32_t *vs1)
535 {
536     uint32_t a = vs2[H4(3)], b = vs2[H4(2)], e = vs2[H4(1)], f = vs2[H4(0)];
537     uint32_t c = vd[H4(3)], d = vd[H4(2)], g = vd[H4(1)], h = vd[H4(0)];
538     uint32_t W0 = vs1[H4(0)], W1 = vs1[H4(1)];
539     uint32_t T1 = h + sum1_32(e) + ch(e, f, g) + W0;
540     uint32_t T2 = sum0_32(a) + maj(a, b, c);
541 
542     h = g;
543     g = f;
544     f = e;
545     e = d + T1;
546     d = c;
547     c = b;
548     b = a;
549     a = T1 + T2;
550 
551     T1 = h + sum1_32(e) + ch(e, f, g) + W1;
552     T2 = sum0_32(a) + maj(a, b, c);
553     h = g;
554     g = f;
555     f = e;
556     e = d + T1;
557     d = c;
558     c = b;
559     b = a;
560     a = T1 + T2;
561 
562     vd[H4(0)] = f;
563     vd[H4(1)] = e;
564     vd[H4(2)] = b;
565     vd[H4(3)] = a;
566 }
567 
568 void HELPER(vsha2ch32_vv)(void *vd, void *vs1, void *vs2, CPURISCVState *env,
569                           uint32_t desc)
570 {
571     const uint32_t esz = 4;
572     uint32_t total_elems;
573     uint32_t vta = vext_vta(desc);
574 
575     for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
576         vsha2c_32(((uint32_t *)vs2) + 4 * i, ((uint32_t *)vd) + 4 * i,
577                   ((uint32_t *)vs1) + 4 * i + 2);
578     }
579 
580     /* set tail elements to 1s */
581     total_elems = vext_get_total_elems(env, desc, esz);
582     vext_set_elems_1s(vd, vta, env->vl * esz, total_elems * esz);
583     env->vstart = 0;
584 }
585 
586 void HELPER(vsha2ch64_vv)(void *vd, void *vs1, void *vs2, CPURISCVState *env,
587                           uint32_t desc)
588 {
589     const uint32_t esz = 8;
590     uint32_t total_elems;
591     uint32_t vta = vext_vta(desc);
592 
593     for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
594         vsha2c_64(((uint64_t *)vs2) + 4 * i, ((uint64_t *)vd) + 4 * i,
595                   ((uint64_t *)vs1) + 4 * i + 2);
596     }
597 
598     /* set tail elements to 1s */
599     total_elems = vext_get_total_elems(env, desc, esz);
600     vext_set_elems_1s(vd, vta, env->vl * esz, total_elems * esz);
601     env->vstart = 0;
602 }
603 
604 void HELPER(vsha2cl32_vv)(void *vd, void *vs1, void *vs2, CPURISCVState *env,
605                           uint32_t desc)
606 {
607     const uint32_t esz = 4;
608     uint32_t total_elems;
609     uint32_t vta = vext_vta(desc);
610 
611     for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
612         vsha2c_32(((uint32_t *)vs2) + 4 * i, ((uint32_t *)vd) + 4 * i,
613                   (((uint32_t *)vs1) + 4 * i));
614     }
615 
616     /* set tail elements to 1s */
617     total_elems = vext_get_total_elems(env, desc, esz);
618     vext_set_elems_1s(vd, vta, env->vl * esz, total_elems * esz);
619     env->vstart = 0;
620 }
621 
622 void HELPER(vsha2cl64_vv)(void *vd, void *vs1, void *vs2, CPURISCVState *env,
623                           uint32_t desc)
624 {
625     uint32_t esz = 8;
626     uint32_t total_elems;
627     uint32_t vta = vext_vta(desc);
628 
629     for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
630         vsha2c_64(((uint64_t *)vs2) + 4 * i, ((uint64_t *)vd) + 4 * i,
631                   (((uint64_t *)vs1) + 4 * i));
632     }
633 
634     /* set tail elements to 1s */
635     total_elems = vext_get_total_elems(env, desc, esz);
636     vext_set_elems_1s(vd, vta, env->vl * esz, total_elems * esz);
637     env->vstart = 0;
638 }
639 
640 static inline uint32_t p1(uint32_t x)
641 {
642     return x ^ rol32(x, 15) ^ rol32(x, 23);
643 }
644 
645 static inline uint32_t zvksh_w(uint32_t m16, uint32_t m9, uint32_t m3,
646                                uint32_t m13, uint32_t m6)
647 {
648     return p1(m16 ^ m9 ^ rol32(m3, 15)) ^ rol32(m13, 7) ^ m6;
649 }
650 
651 void HELPER(vsm3me_vv)(void *vd_vptr, void *vs1_vptr, void *vs2_vptr,
652                        CPURISCVState *env, uint32_t desc)
653 {
654     uint32_t esz = memop_size(FIELD_EX64(env->vtype, VTYPE, VSEW));
655     uint32_t total_elems = vext_get_total_elems(env, desc, esz);
656     uint32_t vta = vext_vta(desc);
657     uint32_t *vd = vd_vptr;
658     uint32_t *vs1 = vs1_vptr;
659     uint32_t *vs2 = vs2_vptr;
660 
661     for (int i = env->vstart / 8; i < env->vl / 8; i++) {
662         uint32_t w[24];
663         for (int j = 0; j < 8; j++) {
664             w[j] = bswap32(vs1[H4((i * 8) + j)]);
665             w[j + 8] = bswap32(vs2[H4((i * 8) + j)]);
666         }
667         for (int j = 0; j < 8; j++) {
668             w[j + 16] =
669                 zvksh_w(w[j], w[j + 7], w[j + 13], w[j + 3], w[j + 10]);
670         }
671         for (int j = 0; j < 8; j++) {
672             vd[(i * 8) + j] = bswap32(w[H4(j + 16)]);
673         }
674     }
675     vext_set_elems_1s(vd_vptr, vta, env->vl * esz, total_elems * esz);
676     env->vstart = 0;
677 }
678 
679 static inline uint32_t ff1(uint32_t x, uint32_t y, uint32_t z)
680 {
681     return x ^ y ^ z;
682 }
683 
684 static inline uint32_t ff2(uint32_t x, uint32_t y, uint32_t z)
685 {
686     return (x & y) | (x & z) | (y & z);
687 }
688 
689 static inline uint32_t ff_j(uint32_t x, uint32_t y, uint32_t z, uint32_t j)
690 {
691     return (j <= 15) ? ff1(x, y, z) : ff2(x, y, z);
692 }
693 
694 static inline uint32_t gg1(uint32_t x, uint32_t y, uint32_t z)
695 {
696     return x ^ y ^ z;
697 }
698 
699 static inline uint32_t gg2(uint32_t x, uint32_t y, uint32_t z)
700 {
701     return (x & y) | (~x & z);
702 }
703 
704 static inline uint32_t gg_j(uint32_t x, uint32_t y, uint32_t z, uint32_t j)
705 {
706     return (j <= 15) ? gg1(x, y, z) : gg2(x, y, z);
707 }
708 
709 static inline uint32_t t_j(uint32_t j)
710 {
711     return (j <= 15) ? 0x79cc4519 : 0x7a879d8a;
712 }
713 
714 static inline uint32_t p_0(uint32_t x)
715 {
716     return x ^ rol32(x, 9) ^ rol32(x, 17);
717 }
718 
719 static void sm3c(uint32_t *vd, uint32_t *vs1, uint32_t *vs2, uint32_t uimm)
720 {
721     uint32_t x0, x1;
722     uint32_t j;
723     uint32_t ss1, ss2, tt1, tt2;
724     x0 = vs2[0] ^ vs2[4];
725     x1 = vs2[1] ^ vs2[5];
726     j = 2 * uimm;
727     ss1 = rol32(rol32(vs1[0], 12) + vs1[4] + rol32(t_j(j), j % 32), 7);
728     ss2 = ss1 ^ rol32(vs1[0], 12);
729     tt1 = ff_j(vs1[0], vs1[1], vs1[2], j) + vs1[3] + ss2 + x0;
730     tt2 = gg_j(vs1[4], vs1[5], vs1[6], j) + vs1[7] + ss1 + vs2[0];
731     vs1[3] = vs1[2];
732     vd[3] = rol32(vs1[1], 9);
733     vs1[1] = vs1[0];
734     vd[1] = tt1;
735     vs1[7] = vs1[6];
736     vd[7] = rol32(vs1[5], 19);
737     vs1[5] = vs1[4];
738     vd[5] = p_0(tt2);
739     j = 2 * uimm + 1;
740     ss1 = rol32(rol32(vd[1], 12) + vd[5] + rol32(t_j(j), j % 32), 7);
741     ss2 = ss1 ^ rol32(vd[1], 12);
742     tt1 = ff_j(vd[1], vs1[1], vd[3], j) + vs1[3] + ss2 + x1;
743     tt2 = gg_j(vd[5], vs1[5], vd[7], j) + vs1[7] + ss1 + vs2[1];
744     vd[2] = rol32(vs1[1], 9);
745     vd[0] = tt1;
746     vd[6] = rol32(vs1[5], 19);
747     vd[4] = p_0(tt2);
748 }
749 
750 void HELPER(vsm3c_vi)(void *vd_vptr, void *vs2_vptr, uint32_t uimm,
751                       CPURISCVState *env, uint32_t desc)
752 {
753     uint32_t esz = memop_size(FIELD_EX64(env->vtype, VTYPE, VSEW));
754     uint32_t total_elems = vext_get_total_elems(env, desc, esz);
755     uint32_t vta = vext_vta(desc);
756     uint32_t *vd = vd_vptr;
757     uint32_t *vs2 = vs2_vptr;
758     uint32_t v1[8], v2[8], v3[8];
759 
760     for (int i = env->vstart / 8; i < env->vl / 8; i++) {
761         for (int k = 0; k < 8; k++) {
762             v2[k] = bswap32(vd[H4(i * 8 + k)]);
763             v3[k] = bswap32(vs2[H4(i * 8 + k)]);
764         }
765         sm3c(v1, v2, v3, uimm);
766         for (int k = 0; k < 8; k++) {
767             vd[i * 8 + k] = bswap32(v1[H4(k)]);
768         }
769     }
770     vext_set_elems_1s(vd_vptr, vta, env->vl * esz, total_elems * esz);
771     env->vstart = 0;
772 }
773 
774 void HELPER(vghsh_vv)(void *vd_vptr, void *vs1_vptr, void *vs2_vptr,
775                       CPURISCVState *env, uint32_t desc)
776 {
777     uint64_t *vd = vd_vptr;
778     uint64_t *vs1 = vs1_vptr;
779     uint64_t *vs2 = vs2_vptr;
780     uint32_t vta = vext_vta(desc);
781     uint32_t total_elems = vext_get_total_elems(env, desc, 4);
782 
783     for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
784         uint64_t Y[2] = {vd[i * 2 + 0], vd[i * 2 + 1]};
785         uint64_t H[2] = {brev8(vs2[i * 2 + 0]), brev8(vs2[i * 2 + 1])};
786         uint64_t X[2] = {vs1[i * 2 + 0], vs1[i * 2 + 1]};
787         uint64_t Z[2] = {0, 0};
788 
789         uint64_t S[2] = {brev8(Y[0] ^ X[0]), brev8(Y[1] ^ X[1])};
790 
791         for (int j = 0; j < 128; j++) {
792             if ((S[j / 64] >> (j % 64)) & 1) {
793                 Z[0] ^= H[0];
794                 Z[1] ^= H[1];
795             }
796             bool reduce = ((H[1] >> 63) & 1);
797             H[1] = H[1] << 1 | H[0] >> 63;
798             H[0] = H[0] << 1;
799             if (reduce) {
800                 H[0] ^= 0x87;
801             }
802         }
803 
804         vd[i * 2 + 0] = brev8(Z[0]);
805         vd[i * 2 + 1] = brev8(Z[1]);
806     }
807     /* set tail elements to 1s */
808     vext_set_elems_1s(vd, vta, env->vl * 4, total_elems * 4);
809     env->vstart = 0;
810 }
811 
812 void HELPER(vgmul_vv)(void *vd_vptr, void *vs2_vptr, CPURISCVState *env,
813                       uint32_t desc)
814 {
815     uint64_t *vd = vd_vptr;
816     uint64_t *vs2 = vs2_vptr;
817     uint32_t vta = vext_vta(desc);
818     uint32_t total_elems = vext_get_total_elems(env, desc, 4);
819 
820     for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
821         uint64_t Y[2] = {brev8(vd[i * 2 + 0]), brev8(vd[i * 2 + 1])};
822         uint64_t H[2] = {brev8(vs2[i * 2 + 0]), brev8(vs2[i * 2 + 1])};
823         uint64_t Z[2] = {0, 0};
824 
825         for (int j = 0; j < 128; j++) {
826             if ((Y[j / 64] >> (j % 64)) & 1) {
827                 Z[0] ^= H[0];
828                 Z[1] ^= H[1];
829             }
830             bool reduce = ((H[1] >> 63) & 1);
831             H[1] = H[1] << 1 | H[0] >> 63;
832             H[0] = H[0] << 1;
833             if (reduce) {
834                 H[0] ^= 0x87;
835             }
836         }
837 
838         vd[i * 2 + 0] = brev8(Z[0]);
839         vd[i * 2 + 1] = brev8(Z[1]);
840     }
841     /* set tail elements to 1s */
842     vext_set_elems_1s(vd, vta, env->vl * 4, total_elems * 4);
843     env->vstart = 0;
844 }
845 
846 void HELPER(vsm4k_vi)(void *vd, void *vs2, uint32_t uimm5, CPURISCVState *env,
847                       uint32_t desc)
848 {
849     const uint32_t egs = 4;
850     uint32_t rnd = uimm5 & 0x7;
851     uint32_t group_start = env->vstart / egs;
852     uint32_t group_end = env->vl / egs;
853     uint32_t esz = sizeof(uint32_t);
854     uint32_t total_elems = vext_get_total_elems(env, desc, esz);
855 
856     for (uint32_t i = group_start; i < group_end; ++i) {
857         uint32_t vstart = i * egs;
858         uint32_t vend = (i + 1) * egs;
859         uint32_t rk[4] = {0};
860         uint32_t tmp[8] = {0};
861 
862         for (uint32_t j = vstart; j < vend; ++j) {
863             rk[j - vstart] = *((uint32_t *)vs2 + H4(j));
864         }
865 
866         for (uint32_t j = 0; j < egs; ++j) {
867             tmp[j] = rk[j];
868         }
869 
870         for (uint32_t j = 0; j < egs; ++j) {
871             uint32_t b, s;
872             b = tmp[j + 1] ^ tmp[j + 2] ^ tmp[j + 3] ^ sm4_ck[rnd * 4 + j];
873 
874             s = sm4_subword(b);
875 
876             tmp[j + 4] = tmp[j] ^ (s ^ rol32(s, 13) ^ rol32(s, 23));
877         }
878 
879         for (uint32_t j = vstart; j < vend; ++j) {
880             *((uint32_t *)vd + H4(j)) = tmp[egs + (j - vstart)];
881         }
882     }
883 
884     env->vstart = 0;
885     /* set tail elements to 1s */
886     vext_set_elems_1s(vd, vext_vta(desc), env->vl * esz, total_elems * esz);
887 }
888 
889 static void do_sm4_round(uint32_t *rk, uint32_t *buf)
890 {
891     const uint32_t egs = 4;
892     uint32_t s, b;
893 
894     for (uint32_t j = egs; j < egs * 2; ++j) {
895         b = buf[j - 3] ^ buf[j - 2] ^ buf[j - 1] ^ rk[j - 4];
896 
897         s = sm4_subword(b);
898 
899         buf[j] = buf[j - 4] ^ (s ^ rol32(s, 2) ^ rol32(s, 10) ^ rol32(s, 18) ^
900                                rol32(s, 24));
901     }
902 }
903 
904 void HELPER(vsm4r_vv)(void *vd, void *vs2, CPURISCVState *env, uint32_t desc)
905 {
906     const uint32_t egs = 4;
907     uint32_t group_start = env->vstart / egs;
908     uint32_t group_end = env->vl / egs;
909     uint32_t esz = sizeof(uint32_t);
910     uint32_t total_elems = vext_get_total_elems(env, desc, esz);
911 
912     for (uint32_t i = group_start; i < group_end; ++i) {
913         uint32_t vstart = i * egs;
914         uint32_t vend = (i + 1) * egs;
915         uint32_t rk[4] = {0};
916         uint32_t tmp[8] = {0};
917 
918         for (uint32_t j = vstart; j < vend; ++j) {
919             rk[j - vstart] = *((uint32_t *)vs2 + H4(j));
920         }
921 
922         for (uint32_t j = vstart; j < vend; ++j) {
923             tmp[j - vstart] = *((uint32_t *)vd + H4(j));
924         }
925 
926         do_sm4_round(rk, tmp);
927 
928         for (uint32_t j = vstart; j < vend; ++j) {
929             *((uint32_t *)vd + H4(j)) = tmp[egs + (j - vstart)];
930         }
931     }
932 
933     env->vstart = 0;
934     /* set tail elements to 1s */
935     vext_set_elems_1s(vd, vext_vta(desc), env->vl * esz, total_elems * esz);
936 }
937 
938 void HELPER(vsm4r_vs)(void *vd, void *vs2, CPURISCVState *env, uint32_t desc)
939 {
940     const uint32_t egs = 4;
941     uint32_t group_start = env->vstart / egs;
942     uint32_t group_end = env->vl / egs;
943     uint32_t esz = sizeof(uint32_t);
944     uint32_t total_elems = vext_get_total_elems(env, desc, esz);
945 
946     for (uint32_t i = group_start; i < group_end; ++i) {
947         uint32_t vstart = i * egs;
948         uint32_t vend = (i + 1) * egs;
949         uint32_t rk[4] = {0};
950         uint32_t tmp[8] = {0};
951 
952         for (uint32_t j = 0; j < egs; ++j) {
953             rk[j] = *((uint32_t *)vs2 + H4(j));
954         }
955 
956         for (uint32_t j = vstart; j < vend; ++j) {
957             tmp[j - vstart] = *((uint32_t *)vd + H4(j));
958         }
959 
960         do_sm4_round(rk, tmp);
961 
962         for (uint32_t j = vstart; j < vend; ++j) {
963             *((uint32_t *)vd + H4(j)) = tmp[egs + (j - vstart)];
964         }
965     }
966 
967     env->vstart = 0;
968     /* set tail elements to 1s */
969     vext_set_elems_1s(vd, vext_vta(desc), env->vl * esz, total_elems * esz);
970 }
971