xref: /linux/arch/x86/crypto/camellia-aesni-avx2-asm_64.S (revision ca55b2fef3a9373fcfc30f82fd26bc7fccbda732) 
1/*
2 * x86_64/AVX2/AES-NI assembler implementation of Camellia
3 *
4 * Copyright © 2013 Jussi Kivilinna <jussi.kivilinna@iki.fi>
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 */
12
13#include <linux/linkage.h>
14
15#define CAMELLIA_TABLE_BYTE_LEN 272
16
17/* struct camellia_ctx: */
18#define key_table 0
19#define key_length CAMELLIA_TABLE_BYTE_LEN
20
21/* register macros */
22#define CTX %rdi
23#define RIO %r8
24
25/**********************************************************************
26  helper macros
27 **********************************************************************/
28#define filter_8bit(x, lo_t, hi_t, mask4bit, tmp0) \
29	vpand x, mask4bit, tmp0; \
30	vpandn x, mask4bit, x; \
31	vpsrld $4, x, x; \
32	\
33	vpshufb tmp0, lo_t, tmp0; \
34	vpshufb x, hi_t, x; \
35	vpxor tmp0, x, x;
36
37#define ymm0_x xmm0
38#define ymm1_x xmm1
39#define ymm2_x xmm2
40#define ymm3_x xmm3
41#define ymm4_x xmm4
42#define ymm5_x xmm5
43#define ymm6_x xmm6
44#define ymm7_x xmm7
45#define ymm8_x xmm8
46#define ymm9_x xmm9
47#define ymm10_x xmm10
48#define ymm11_x xmm11
49#define ymm12_x xmm12
50#define ymm13_x xmm13
51#define ymm14_x xmm14
52#define ymm15_x xmm15
53
54/**********************************************************************
55  32-way camellia
56 **********************************************************************/
57
58/*
59 * IN:
60 *   x0..x7: byte-sliced AB state
61 *   mem_cd: register pointer storing CD state
62 *   key: index for key material
63 * OUT:
64 *   x0..x7: new byte-sliced CD state
65 */
66#define roundsm32(x0, x1, x2, x3, x4, x5, x6, x7, t0, t1, t2, t3, t4, t5, t6, \
67		  t7, mem_cd, key) \
68	/* \
69	 * S-function with AES subbytes \
70	 */ \
71	vbroadcasti128 .Linv_shift_row, t4; \
72	vpbroadcastd .L0f0f0f0f, t7; \
73	vbroadcasti128 .Lpre_tf_lo_s1, t5; \
74	vbroadcasti128 .Lpre_tf_hi_s1, t6; \
75	vbroadcasti128 .Lpre_tf_lo_s4, t2; \
76	vbroadcasti128 .Lpre_tf_hi_s4, t3; \
77	\
78	/* AES inverse shift rows */ \
79	vpshufb t4, x0, x0; \
80	vpshufb t4, x7, x7; \
81	vpshufb t4, x3, x3; \
82	vpshufb t4, x6, x6; \
83	vpshufb t4, x2, x2; \
84	vpshufb t4, x5, x5; \
85	vpshufb t4, x1, x1; \
86	vpshufb t4, x4, x4; \
87	\
88	/* prefilter sboxes 1, 2 and 3 */ \
89	/* prefilter sbox 4 */ \
90	filter_8bit(x0, t5, t6, t7, t4); \
91	filter_8bit(x7, t5, t6, t7, t4); \
92	vextracti128 $1, x0, t0##_x; \
93	vextracti128 $1, x7, t1##_x; \
94	filter_8bit(x3, t2, t3, t7, t4); \
95	filter_8bit(x6, t2, t3, t7, t4); \
96	vextracti128 $1, x3, t3##_x; \
97	vextracti128 $1, x6, t2##_x; \
98	filter_8bit(x2, t5, t6, t7, t4); \
99	filter_8bit(x5, t5, t6, t7, t4); \
100	filter_8bit(x1, t5, t6, t7, t4); \
101	filter_8bit(x4, t5, t6, t7, t4); \
102	\
103	vpxor t4##_x, t4##_x, t4##_x; \
104	\
105	/* AES subbytes + AES shift rows */ \
106	vextracti128 $1, x2, t6##_x; \
107	vextracti128 $1, x5, t5##_x; \
108	vaesenclast t4##_x, x0##_x, x0##_x; \
109	vaesenclast t4##_x, t0##_x, t0##_x; \
110	vinserti128 $1, t0##_x, x0, x0; \
111	vaesenclast t4##_x, x7##_x, x7##_x; \
112	vaesenclast t4##_x, t1##_x, t1##_x; \
113	vinserti128 $1, t1##_x, x7, x7; \
114	vaesenclast t4##_x, x3##_x, x3##_x; \
115	vaesenclast t4##_x, t3##_x, t3##_x; \
116	vinserti128 $1, t3##_x, x3, x3; \
117	vaesenclast t4##_x, x6##_x, x6##_x; \
118	vaesenclast t4##_x, t2##_x, t2##_x; \
119	vinserti128 $1, t2##_x, x6, x6; \
120	vextracti128 $1, x1, t3##_x; \
121	vextracti128 $1, x4, t2##_x; \
122	vbroadcasti128 .Lpost_tf_lo_s1, t0; \
123	vbroadcasti128 .Lpost_tf_hi_s1, t1; \
124	vaesenclast t4##_x, x2##_x, x2##_x; \
125	vaesenclast t4##_x, t6##_x, t6##_x; \
126	vinserti128 $1, t6##_x, x2, x2; \
127	vaesenclast t4##_x, x5##_x, x5##_x; \
128	vaesenclast t4##_x, t5##_x, t5##_x; \
129	vinserti128 $1, t5##_x, x5, x5; \
130	vaesenclast t4##_x, x1##_x, x1##_x; \
131	vaesenclast t4##_x, t3##_x, t3##_x; \
132	vinserti128 $1, t3##_x, x1, x1; \
133	vaesenclast t4##_x, x4##_x, x4##_x; \
134	vaesenclast t4##_x, t2##_x, t2##_x; \
135	vinserti128 $1, t2##_x, x4, x4; \
136	\
137	/* postfilter sboxes 1 and 4 */ \
138	vbroadcasti128 .Lpost_tf_lo_s3, t2; \
139	vbroadcasti128 .Lpost_tf_hi_s3, t3; \
140	filter_8bit(x0, t0, t1, t7, t6); \
141	filter_8bit(x7, t0, t1, t7, t6); \
142	filter_8bit(x3, t0, t1, t7, t6); \
143	filter_8bit(x6, t0, t1, t7, t6); \
144	\
145	/* postfilter sbox 3 */ \
146	vbroadcasti128 .Lpost_tf_lo_s2, t4; \
147	vbroadcasti128 .Lpost_tf_hi_s2, t5; \
148	filter_8bit(x2, t2, t3, t7, t6); \
149	filter_8bit(x5, t2, t3, t7, t6); \
150	\
151	vpbroadcastq key, t0; /* higher 64-bit duplicate ignored */ \
152	\
153	/* postfilter sbox 2 */ \
154	filter_8bit(x1, t4, t5, t7, t2); \
155	filter_8bit(x4, t4, t5, t7, t2); \
156	vpxor t7, t7, t7; \
157	\
158	vpsrldq $1, t0, t1; \
159	vpsrldq $2, t0, t2; \
160	vpshufb t7, t1, t1; \
161	vpsrldq $3, t0, t3; \
162	\
163	/* P-function */ \
164	vpxor x5, x0, x0; \
165	vpxor x6, x1, x1; \
166	vpxor x7, x2, x2; \
167	vpxor x4, x3, x3; \
168	\
169	vpshufb t7, t2, t2; \
170	vpsrldq $4, t0, t4; \
171	vpshufb t7, t3, t3; \
172	vpsrldq $5, t0, t5; \
173	vpshufb t7, t4, t4; \
174	\
175	vpxor x2, x4, x4; \
176	vpxor x3, x5, x5; \
177	vpxor x0, x6, x6; \
178	vpxor x1, x7, x7; \
179	\
180	vpsrldq $6, t0, t6; \
181	vpshufb t7, t5, t5; \
182	vpshufb t7, t6, t6; \
183	\
184	vpxor x7, x0, x0; \
185	vpxor x4, x1, x1; \
186	vpxor x5, x2, x2; \
187	vpxor x6, x3, x3; \
188	\
189	vpxor x3, x4, x4; \
190	vpxor x0, x5, x5; \
191	vpxor x1, x6, x6; \
192	vpxor x2, x7, x7; /* note: high and low parts swapped */ \
193	\
194	/* Add key material and result to CD (x becomes new CD) */ \
195	\
196	vpxor t6, x1, x1; \
197	vpxor 5 * 32(mem_cd), x1, x1; \
198	\
199	vpsrldq $7, t0, t6; \
200	vpshufb t7, t0, t0; \
201	vpshufb t7, t6, t7; \
202	\
203	vpxor t7, x0, x0; \
204	vpxor 4 * 32(mem_cd), x0, x0; \
205	\
206	vpxor t5, x2, x2; \
207	vpxor 6 * 32(mem_cd), x2, x2; \
208	\
209	vpxor t4, x3, x3; \
210	vpxor 7 * 32(mem_cd), x3, x3; \
211	\
212	vpxor t3, x4, x4; \
213	vpxor 0 * 32(mem_cd), x4, x4; \
214	\
215	vpxor t2, x5, x5; \
216	vpxor 1 * 32(mem_cd), x5, x5; \
217	\
218	vpxor t1, x6, x6; \
219	vpxor 2 * 32(mem_cd), x6, x6; \
220	\
221	vpxor t0, x7, x7; \
222	vpxor 3 * 32(mem_cd), x7, x7;
223
224/*
225 * Size optimization... with inlined roundsm32 binary would be over 5 times
226 * larger and would only marginally faster.
227 */
228.align 8
229roundsm32_x0_x1_x2_x3_x4_x5_x6_x7_y0_y1_y2_y3_y4_y5_y6_y7_cd:
230	roundsm32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
231		  %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14, %ymm15,
232		  %rcx, (%r9));
233	ret;
234ENDPROC(roundsm32_x0_x1_x2_x3_x4_x5_x6_x7_y0_y1_y2_y3_y4_y5_y6_y7_cd)
235
236.align 8
237roundsm32_x4_x5_x6_x7_x0_x1_x2_x3_y4_y5_y6_y7_y0_y1_y2_y3_ab:
238	roundsm32(%ymm4, %ymm5, %ymm6, %ymm7, %ymm0, %ymm1, %ymm2, %ymm3,
239		  %ymm12, %ymm13, %ymm14, %ymm15, %ymm8, %ymm9, %ymm10, %ymm11,
240		  %rax, (%r9));
241	ret;
242ENDPROC(roundsm32_x4_x5_x6_x7_x0_x1_x2_x3_y4_y5_y6_y7_y0_y1_y2_y3_ab)
243
244/*
245 * IN/OUT:
246 *  x0..x7: byte-sliced AB state preloaded
247 *  mem_ab: byte-sliced AB state in memory
248 *  mem_cb: byte-sliced CD state in memory
249 */
250#define two_roundsm32(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
251		      y6, y7, mem_ab, mem_cd, i, dir, store_ab) \
252	leaq (key_table + (i) * 8)(CTX), %r9; \
253	call roundsm32_x0_x1_x2_x3_x4_x5_x6_x7_y0_y1_y2_y3_y4_y5_y6_y7_cd; \
254	\
255	vmovdqu x0, 4 * 32(mem_cd); \
256	vmovdqu x1, 5 * 32(mem_cd); \
257	vmovdqu x2, 6 * 32(mem_cd); \
258	vmovdqu x3, 7 * 32(mem_cd); \
259	vmovdqu x4, 0 * 32(mem_cd); \
260	vmovdqu x5, 1 * 32(mem_cd); \
261	vmovdqu x6, 2 * 32(mem_cd); \
262	vmovdqu x7, 3 * 32(mem_cd); \
263	\
264	leaq (key_table + ((i) + (dir)) * 8)(CTX), %r9; \
265	call roundsm32_x4_x5_x6_x7_x0_x1_x2_x3_y4_y5_y6_y7_y0_y1_y2_y3_ab; \
266	\
267	store_ab(x0, x1, x2, x3, x4, x5, x6, x7, mem_ab);
268
269#define dummy_store(x0, x1, x2, x3, x4, x5, x6, x7, mem_ab) /* do nothing */
270
271#define store_ab_state(x0, x1, x2, x3, x4, x5, x6, x7, mem_ab) \
272	/* Store new AB state */ \
273	vmovdqu x4, 4 * 32(mem_ab); \
274	vmovdqu x5, 5 * 32(mem_ab); \
275	vmovdqu x6, 6 * 32(mem_ab); \
276	vmovdqu x7, 7 * 32(mem_ab); \
277	vmovdqu x0, 0 * 32(mem_ab); \
278	vmovdqu x1, 1 * 32(mem_ab); \
279	vmovdqu x2, 2 * 32(mem_ab); \
280	vmovdqu x3, 3 * 32(mem_ab);
281
282#define enc_rounds32(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
283		      y6, y7, mem_ab, mem_cd, i) \
284	two_roundsm32(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
285		      y6, y7, mem_ab, mem_cd, (i) + 2, 1, store_ab_state); \
286	two_roundsm32(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
287		      y6, y7, mem_ab, mem_cd, (i) + 4, 1, store_ab_state); \
288	two_roundsm32(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
289		      y6, y7, mem_ab, mem_cd, (i) + 6, 1, dummy_store);
290
291#define dec_rounds32(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
292		      y6, y7, mem_ab, mem_cd, i) \
293	two_roundsm32(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
294		      y6, y7, mem_ab, mem_cd, (i) + 7, -1, store_ab_state); \
295	two_roundsm32(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
296		      y6, y7, mem_ab, mem_cd, (i) + 5, -1, store_ab_state); \
297	two_roundsm32(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
298		      y6, y7, mem_ab, mem_cd, (i) + 3, -1, dummy_store);
299
300/*
301 * IN:
302 *  v0..3: byte-sliced 32-bit integers
303 * OUT:
304 *  v0..3: (IN <<< 1)
305 */
306#define rol32_1_32(v0, v1, v2, v3, t0, t1, t2, zero) \
307	vpcmpgtb v0, zero, t0; \
308	vpaddb v0, v0, v0; \
309	vpabsb t0, t0; \
310	\
311	vpcmpgtb v1, zero, t1; \
312	vpaddb v1, v1, v1; \
313	vpabsb t1, t1; \
314	\
315	vpcmpgtb v2, zero, t2; \
316	vpaddb v2, v2, v2; \
317	vpabsb t2, t2; \
318	\
319	vpor t0, v1, v1; \
320	\
321	vpcmpgtb v3, zero, t0; \
322	vpaddb v3, v3, v3; \
323	vpabsb t0, t0; \
324	\
325	vpor t1, v2, v2; \
326	vpor t2, v3, v3; \
327	vpor t0, v0, v0;
328
329/*
330 * IN:
331 *   r: byte-sliced AB state in memory
332 *   l: byte-sliced CD state in memory
333 * OUT:
334 *   x0..x7: new byte-sliced CD state
335 */
336#define fls32(l, l0, l1, l2, l3, l4, l5, l6, l7, r, t0, t1, t2, t3, tt0, \
337	      tt1, tt2, tt3, kll, klr, krl, krr) \
338	/* \
339	 * t0 = kll; \
340	 * t0 &= ll; \
341	 * lr ^= rol32(t0, 1); \
342	 */ \
343	vpbroadcastd kll, t0; /* only lowest 32-bit used */ \
344	vpxor tt0, tt0, tt0; \
345	vpshufb tt0, t0, t3; \
346	vpsrldq $1, t0, t0; \
347	vpshufb tt0, t0, t2; \
348	vpsrldq $1, t0, t0; \
349	vpshufb tt0, t0, t1; \
350	vpsrldq $1, t0, t0; \
351	vpshufb tt0, t0, t0; \
352	\
353	vpand l0, t0, t0; \
354	vpand l1, t1, t1; \
355	vpand l2, t2, t2; \
356	vpand l3, t3, t3; \
357	\
358	rol32_1_32(t3, t2, t1, t0, tt1, tt2, tt3, tt0); \
359	\
360	vpxor l4, t0, l4; \
361	vpbroadcastd krr, t0; /* only lowest 32-bit used */ \
362	vmovdqu l4, 4 * 32(l); \
363	vpxor l5, t1, l5; \
364	vmovdqu l5, 5 * 32(l); \
365	vpxor l6, t2, l6; \
366	vmovdqu l6, 6 * 32(l); \
367	vpxor l7, t3, l7; \
368	vmovdqu l7, 7 * 32(l); \
369	\
370	/* \
371	 * t2 = krr; \
372	 * t2 |= rr; \
373	 * rl ^= t2; \
374	 */ \
375	\
376	vpshufb tt0, t0, t3; \
377	vpsrldq $1, t0, t0; \
378	vpshufb tt0, t0, t2; \
379	vpsrldq $1, t0, t0; \
380	vpshufb tt0, t0, t1; \
381	vpsrldq $1, t0, t0; \
382	vpshufb tt0, t0, t0; \
383	\
384	vpor 4 * 32(r), t0, t0; \
385	vpor 5 * 32(r), t1, t1; \
386	vpor 6 * 32(r), t2, t2; \
387	vpor 7 * 32(r), t3, t3; \
388	\
389	vpxor 0 * 32(r), t0, t0; \
390	vpxor 1 * 32(r), t1, t1; \
391	vpxor 2 * 32(r), t2, t2; \
392	vpxor 3 * 32(r), t3, t3; \
393	vmovdqu t0, 0 * 32(r); \
394	vpbroadcastd krl, t0; /* only lowest 32-bit used */ \
395	vmovdqu t1, 1 * 32(r); \
396	vmovdqu t2, 2 * 32(r); \
397	vmovdqu t3, 3 * 32(r); \
398	\
399	/* \
400	 * t2 = krl; \
401	 * t2 &= rl; \
402	 * rr ^= rol32(t2, 1); \
403	 */ \
404	vpshufb tt0, t0, t3; \
405	vpsrldq $1, t0, t0; \
406	vpshufb tt0, t0, t2; \
407	vpsrldq $1, t0, t0; \
408	vpshufb tt0, t0, t1; \
409	vpsrldq $1, t0, t0; \
410	vpshufb tt0, t0, t0; \
411	\
412	vpand 0 * 32(r), t0, t0; \
413	vpand 1 * 32(r), t1, t1; \
414	vpand 2 * 32(r), t2, t2; \
415	vpand 3 * 32(r), t3, t3; \
416	\
417	rol32_1_32(t3, t2, t1, t0, tt1, tt2, tt3, tt0); \
418	\
419	vpxor 4 * 32(r), t0, t0; \
420	vpxor 5 * 32(r), t1, t1; \
421	vpxor 6 * 32(r), t2, t2; \
422	vpxor 7 * 32(r), t3, t3; \
423	vmovdqu t0, 4 * 32(r); \
424	vpbroadcastd klr, t0; /* only lowest 32-bit used */ \
425	vmovdqu t1, 5 * 32(r); \
426	vmovdqu t2, 6 * 32(r); \
427	vmovdqu t3, 7 * 32(r); \
428	\
429	/* \
430	 * t0 = klr; \
431	 * t0 |= lr; \
432	 * ll ^= t0; \
433	 */ \
434	\
435	vpshufb tt0, t0, t3; \
436	vpsrldq $1, t0, t0; \
437	vpshufb tt0, t0, t2; \
438	vpsrldq $1, t0, t0; \
439	vpshufb tt0, t0, t1; \
440	vpsrldq $1, t0, t0; \
441	vpshufb tt0, t0, t0; \
442	\
443	vpor l4, t0, t0; \
444	vpor l5, t1, t1; \
445	vpor l6, t2, t2; \
446	vpor l7, t3, t3; \
447	\
448	vpxor l0, t0, l0; \
449	vmovdqu l0, 0 * 32(l); \
450	vpxor l1, t1, l1; \
451	vmovdqu l1, 1 * 32(l); \
452	vpxor l2, t2, l2; \
453	vmovdqu l2, 2 * 32(l); \
454	vpxor l3, t3, l3; \
455	vmovdqu l3, 3 * 32(l);
456
457#define transpose_4x4(x0, x1, x2, x3, t1, t2) \
458	vpunpckhdq x1, x0, t2; \
459	vpunpckldq x1, x0, x0; \
460	\
461	vpunpckldq x3, x2, t1; \
462	vpunpckhdq x3, x2, x2; \
463	\
464	vpunpckhqdq t1, x0, x1; \
465	vpunpcklqdq t1, x0, x0; \
466	\
467	vpunpckhqdq x2, t2, x3; \
468	vpunpcklqdq x2, t2, x2;
469
470#define byteslice_16x16b_fast(a0, b0, c0, d0, a1, b1, c1, d1, a2, b2, c2, d2, \
471			      a3, b3, c3, d3, st0, st1) \
472	vmovdqu d2, st0; \
473	vmovdqu d3, st1; \
474	transpose_4x4(a0, a1, a2, a3, d2, d3); \
475	transpose_4x4(b0, b1, b2, b3, d2, d3); \
476	vmovdqu st0, d2; \
477	vmovdqu st1, d3; \
478	\
479	vmovdqu a0, st0; \
480	vmovdqu a1, st1; \
481	transpose_4x4(c0, c1, c2, c3, a0, a1); \
482	transpose_4x4(d0, d1, d2, d3, a0, a1); \
483	\
484	vbroadcasti128 .Lshufb_16x16b, a0; \
485	vmovdqu st1, a1; \
486	vpshufb a0, a2, a2; \
487	vpshufb a0, a3, a3; \
488	vpshufb a0, b0, b0; \
489	vpshufb a0, b1, b1; \
490	vpshufb a0, b2, b2; \
491	vpshufb a0, b3, b3; \
492	vpshufb a0, a1, a1; \
493	vpshufb a0, c0, c0; \
494	vpshufb a0, c1, c1; \
495	vpshufb a0, c2, c2; \
496	vpshufb a0, c3, c3; \
497	vpshufb a0, d0, d0; \
498	vpshufb a0, d1, d1; \
499	vpshufb a0, d2, d2; \
500	vpshufb a0, d3, d3; \
501	vmovdqu d3, st1; \
502	vmovdqu st0, d3; \
503	vpshufb a0, d3, a0; \
504	vmovdqu d2, st0; \
505	\
506	transpose_4x4(a0, b0, c0, d0, d2, d3); \
507	transpose_4x4(a1, b1, c1, d1, d2, d3); \
508	vmovdqu st0, d2; \
509	vmovdqu st1, d3; \
510	\
511	vmovdqu b0, st0; \
512	vmovdqu b1, st1; \
513	transpose_4x4(a2, b2, c2, d2, b0, b1); \
514	transpose_4x4(a3, b3, c3, d3, b0, b1); \
515	vmovdqu st0, b0; \
516	vmovdqu st1, b1; \
517	/* does not adjust output bytes inside vectors */
518
519/* load blocks to registers and apply pre-whitening */
520#define inpack32_pre(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
521		     y6, y7, rio, key) \
522	vpbroadcastq key, x0; \
523	vpshufb .Lpack_bswap, x0, x0; \
524	\
525	vpxor 0 * 32(rio), x0, y7; \
526	vpxor 1 * 32(rio), x0, y6; \
527	vpxor 2 * 32(rio), x0, y5; \
528	vpxor 3 * 32(rio), x0, y4; \
529	vpxor 4 * 32(rio), x0, y3; \
530	vpxor 5 * 32(rio), x0, y2; \
531	vpxor 6 * 32(rio), x0, y1; \
532	vpxor 7 * 32(rio), x0, y0; \
533	vpxor 8 * 32(rio), x0, x7; \
534	vpxor 9 * 32(rio), x0, x6; \
535	vpxor 10 * 32(rio), x0, x5; \
536	vpxor 11 * 32(rio), x0, x4; \
537	vpxor 12 * 32(rio), x0, x3; \
538	vpxor 13 * 32(rio), x0, x2; \
539	vpxor 14 * 32(rio), x0, x1; \
540	vpxor 15 * 32(rio), x0, x0;
541
542/* byteslice pre-whitened blocks and store to temporary memory */
543#define inpack32_post(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
544		      y6, y7, mem_ab, mem_cd) \
545	byteslice_16x16b_fast(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, \
546			      y4, y5, y6, y7, (mem_ab), (mem_cd)); \
547	\
548	vmovdqu x0, 0 * 32(mem_ab); \
549	vmovdqu x1, 1 * 32(mem_ab); \
550	vmovdqu x2, 2 * 32(mem_ab); \
551	vmovdqu x3, 3 * 32(mem_ab); \
552	vmovdqu x4, 4 * 32(mem_ab); \
553	vmovdqu x5, 5 * 32(mem_ab); \
554	vmovdqu x6, 6 * 32(mem_ab); \
555	vmovdqu x7, 7 * 32(mem_ab); \
556	vmovdqu y0, 0 * 32(mem_cd); \
557	vmovdqu y1, 1 * 32(mem_cd); \
558	vmovdqu y2, 2 * 32(mem_cd); \
559	vmovdqu y3, 3 * 32(mem_cd); \
560	vmovdqu y4, 4 * 32(mem_cd); \
561	vmovdqu y5, 5 * 32(mem_cd); \
562	vmovdqu y6, 6 * 32(mem_cd); \
563	vmovdqu y7, 7 * 32(mem_cd);
564
565/* de-byteslice, apply post-whitening and store blocks */
566#define outunpack32(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, \
567		    y5, y6, y7, key, stack_tmp0, stack_tmp1) \
568	byteslice_16x16b_fast(y0, y4, x0, x4, y1, y5, x1, x5, y2, y6, x2, x6, \
569			      y3, y7, x3, x7, stack_tmp0, stack_tmp1); \
570	\
571	vmovdqu x0, stack_tmp0; \
572	\
573	vpbroadcastq key, x0; \
574	vpshufb .Lpack_bswap, x0, x0; \
575	\
576	vpxor x0, y7, y7; \
577	vpxor x0, y6, y6; \
578	vpxor x0, y5, y5; \
579	vpxor x0, y4, y4; \
580	vpxor x0, y3, y3; \
581	vpxor x0, y2, y2; \
582	vpxor x0, y1, y1; \
583	vpxor x0, y0, y0; \
584	vpxor x0, x7, x7; \
585	vpxor x0, x6, x6; \
586	vpxor x0, x5, x5; \
587	vpxor x0, x4, x4; \
588	vpxor x0, x3, x3; \
589	vpxor x0, x2, x2; \
590	vpxor x0, x1, x1; \
591	vpxor stack_tmp0, x0, x0;
592
593#define write_output(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
594		     y6, y7, rio) \
595	vmovdqu x0, 0 * 32(rio); \
596	vmovdqu x1, 1 * 32(rio); \
597	vmovdqu x2, 2 * 32(rio); \
598	vmovdqu x3, 3 * 32(rio); \
599	vmovdqu x4, 4 * 32(rio); \
600	vmovdqu x5, 5 * 32(rio); \
601	vmovdqu x6, 6 * 32(rio); \
602	vmovdqu x7, 7 * 32(rio); \
603	vmovdqu y0, 8 * 32(rio); \
604	vmovdqu y1, 9 * 32(rio); \
605	vmovdqu y2, 10 * 32(rio); \
606	vmovdqu y3, 11 * 32(rio); \
607	vmovdqu y4, 12 * 32(rio); \
608	vmovdqu y5, 13 * 32(rio); \
609	vmovdqu y6, 14 * 32(rio); \
610	vmovdqu y7, 15 * 32(rio);
611
612.data
613.align 32
614
615#define SHUFB_BYTES(idx) \
616	0 + (idx), 4 + (idx), 8 + (idx), 12 + (idx)
617
618.Lshufb_16x16b:
619	.byte SHUFB_BYTES(0), SHUFB_BYTES(1), SHUFB_BYTES(2), SHUFB_BYTES(3)
620	.byte SHUFB_BYTES(0), SHUFB_BYTES(1), SHUFB_BYTES(2), SHUFB_BYTES(3)
621
622.Lpack_bswap:
623	.long 0x00010203, 0x04050607, 0x80808080, 0x80808080
624	.long 0x00010203, 0x04050607, 0x80808080, 0x80808080
625
626/* For CTR-mode IV byteswap */
627.Lbswap128_mask:
628	.byte 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0
629
630/* For XTS mode */
631.Lxts_gf128mul_and_shl1_mask_0:
632	.byte 0x87, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0
633.Lxts_gf128mul_and_shl1_mask_1:
634	.byte 0x0e, 1, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0
635
636/*
637 * pre-SubByte transform
638 *
639 * pre-lookup for sbox1, sbox2, sbox3:
640 *   swap_bitendianness(
641 *       isom_map_camellia_to_aes(
642 *           camellia_f(
643 *               swap_bitendianess(in)
644 *           )
645 *       )
646 *   )
647 *
648 * (note: '⊕ 0xc5' inside camellia_f())
649 */
650.Lpre_tf_lo_s1:
651	.byte 0x45, 0xe8, 0x40, 0xed, 0x2e, 0x83, 0x2b, 0x86
652	.byte 0x4b, 0xe6, 0x4e, 0xe3, 0x20, 0x8d, 0x25, 0x88
653.Lpre_tf_hi_s1:
654	.byte 0x00, 0x51, 0xf1, 0xa0, 0x8a, 0xdb, 0x7b, 0x2a
655	.byte 0x09, 0x58, 0xf8, 0xa9, 0x83, 0xd2, 0x72, 0x23
656
657/*
658 * pre-SubByte transform
659 *
660 * pre-lookup for sbox4:
661 *   swap_bitendianness(
662 *       isom_map_camellia_to_aes(
663 *           camellia_f(
664 *               swap_bitendianess(in <<< 1)
665 *           )
666 *       )
667 *   )
668 *
669 * (note: '⊕ 0xc5' inside camellia_f())
670 */
671.Lpre_tf_lo_s4:
672	.byte 0x45, 0x40, 0x2e, 0x2b, 0x4b, 0x4e, 0x20, 0x25
673	.byte 0x14, 0x11, 0x7f, 0x7a, 0x1a, 0x1f, 0x71, 0x74
674.Lpre_tf_hi_s4:
675	.byte 0x00, 0xf1, 0x8a, 0x7b, 0x09, 0xf8, 0x83, 0x72
676	.byte 0xad, 0x5c, 0x27, 0xd6, 0xa4, 0x55, 0x2e, 0xdf
677
678/*
679 * post-SubByte transform
680 *
681 * post-lookup for sbox1, sbox4:
682 *  swap_bitendianness(
683 *      camellia_h(
684 *          isom_map_aes_to_camellia(
685 *              swap_bitendianness(
686 *                  aes_inverse_affine_transform(in)
687 *              )
688 *          )
689 *      )
690 *  )
691 *
692 * (note: '⊕ 0x6e' inside camellia_h())
693 */
694.Lpost_tf_lo_s1:
695	.byte 0x3c, 0xcc, 0xcf, 0x3f, 0x32, 0xc2, 0xc1, 0x31
696	.byte 0xdc, 0x2c, 0x2f, 0xdf, 0xd2, 0x22, 0x21, 0xd1
697.Lpost_tf_hi_s1:
698	.byte 0x00, 0xf9, 0x86, 0x7f, 0xd7, 0x2e, 0x51, 0xa8
699	.byte 0xa4, 0x5d, 0x22, 0xdb, 0x73, 0x8a, 0xf5, 0x0c
700
701/*
702 * post-SubByte transform
703 *
704 * post-lookup for sbox2:
705 *  swap_bitendianness(
706 *      camellia_h(
707 *          isom_map_aes_to_camellia(
708 *              swap_bitendianness(
709 *                  aes_inverse_affine_transform(in)
710 *              )
711 *          )
712 *      )
713 *  ) <<< 1
714 *
715 * (note: '⊕ 0x6e' inside camellia_h())
716 */
717.Lpost_tf_lo_s2:
718	.byte 0x78, 0x99, 0x9f, 0x7e, 0x64, 0x85, 0x83, 0x62
719	.byte 0xb9, 0x58, 0x5e, 0xbf, 0xa5, 0x44, 0x42, 0xa3
720.Lpost_tf_hi_s2:
721	.byte 0x00, 0xf3, 0x0d, 0xfe, 0xaf, 0x5c, 0xa2, 0x51
722	.byte 0x49, 0xba, 0x44, 0xb7, 0xe6, 0x15, 0xeb, 0x18
723
724/*
725 * post-SubByte transform
726 *
727 * post-lookup for sbox3:
728 *  swap_bitendianness(
729 *      camellia_h(
730 *          isom_map_aes_to_camellia(
731 *              swap_bitendianness(
732 *                  aes_inverse_affine_transform(in)
733 *              )
734 *          )
735 *      )
736 *  ) >>> 1
737 *
738 * (note: '⊕ 0x6e' inside camellia_h())
739 */
740.Lpost_tf_lo_s3:
741	.byte 0x1e, 0x66, 0xe7, 0x9f, 0x19, 0x61, 0xe0, 0x98
742	.byte 0x6e, 0x16, 0x97, 0xef, 0x69, 0x11, 0x90, 0xe8
743.Lpost_tf_hi_s3:
744	.byte 0x00, 0xfc, 0x43, 0xbf, 0xeb, 0x17, 0xa8, 0x54
745	.byte 0x52, 0xae, 0x11, 0xed, 0xb9, 0x45, 0xfa, 0x06
746
747/* For isolating SubBytes from AESENCLAST, inverse shift row */
748.Linv_shift_row:
749	.byte 0x00, 0x0d, 0x0a, 0x07, 0x04, 0x01, 0x0e, 0x0b
750	.byte 0x08, 0x05, 0x02, 0x0f, 0x0c, 0x09, 0x06, 0x03
751
752.align 4
753/* 4-bit mask */
754.L0f0f0f0f:
755	.long 0x0f0f0f0f
756
757.text
758
759.align 8
760__camellia_enc_blk32:
761	/* input:
762	 *	%rdi: ctx, CTX
763	 *	%rax: temporary storage, 512 bytes
764	 *	%ymm0..%ymm15: 32 plaintext blocks
765	 * output:
766	 *	%ymm0..%ymm15: 32 encrypted blocks, order swapped:
767	 *       7, 8, 6, 5, 4, 3, 2, 1, 0, 15, 14, 13, 12, 11, 10, 9, 8
768	 */
769
770	leaq 8 * 32(%rax), %rcx;
771
772	inpack32_post(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
773		      %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
774		      %ymm15, %rax, %rcx);
775
776	enc_rounds32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
777		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
778		     %ymm15, %rax, %rcx, 0);
779
780	fls32(%rax, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
781	      %rcx, %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
782	      %ymm15,
783	      ((key_table + (8) * 8) + 0)(CTX),
784	      ((key_table + (8) * 8) + 4)(CTX),
785	      ((key_table + (8) * 8) + 8)(CTX),
786	      ((key_table + (8) * 8) + 12)(CTX));
787
788	enc_rounds32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
789		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
790		     %ymm15, %rax, %rcx, 8);
791
792	fls32(%rax, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
793	      %rcx, %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
794	      %ymm15,
795	      ((key_table + (16) * 8) + 0)(CTX),
796	      ((key_table + (16) * 8) + 4)(CTX),
797	      ((key_table + (16) * 8) + 8)(CTX),
798	      ((key_table + (16) * 8) + 12)(CTX));
799
800	enc_rounds32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
801		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
802		     %ymm15, %rax, %rcx, 16);
803
804	movl $24, %r8d;
805	cmpl $16, key_length(CTX);
806	jne .Lenc_max32;
807
808.Lenc_done:
809	/* load CD for output */
810	vmovdqu 0 * 32(%rcx), %ymm8;
811	vmovdqu 1 * 32(%rcx), %ymm9;
812	vmovdqu 2 * 32(%rcx), %ymm10;
813	vmovdqu 3 * 32(%rcx), %ymm11;
814	vmovdqu 4 * 32(%rcx), %ymm12;
815	vmovdqu 5 * 32(%rcx), %ymm13;
816	vmovdqu 6 * 32(%rcx), %ymm14;
817	vmovdqu 7 * 32(%rcx), %ymm15;
818
819	outunpack32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
820		    %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
821		    %ymm15, (key_table)(CTX, %r8, 8), (%rax), 1 * 32(%rax));
822
823	ret;
824
825.align 8
826.Lenc_max32:
827	movl $32, %r8d;
828
829	fls32(%rax, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
830	      %rcx, %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
831	      %ymm15,
832	      ((key_table + (24) * 8) + 0)(CTX),
833	      ((key_table + (24) * 8) + 4)(CTX),
834	      ((key_table + (24) * 8) + 8)(CTX),
835	      ((key_table + (24) * 8) + 12)(CTX));
836
837	enc_rounds32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
838		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
839		     %ymm15, %rax, %rcx, 24);
840
841	jmp .Lenc_done;
842ENDPROC(__camellia_enc_blk32)
843
844.align 8
845__camellia_dec_blk32:
846	/* input:
847	 *	%rdi: ctx, CTX
848	 *	%rax: temporary storage, 512 bytes
849	 *	%r8d: 24 for 16 byte key, 32 for larger
850	 *	%ymm0..%ymm15: 16 encrypted blocks
851	 * output:
852	 *	%ymm0..%ymm15: 16 plaintext blocks, order swapped:
853	 *       7, 8, 6, 5, 4, 3, 2, 1, 0, 15, 14, 13, 12, 11, 10, 9, 8
854	 */
855
856	leaq 8 * 32(%rax), %rcx;
857
858	inpack32_post(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
859		      %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
860		      %ymm15, %rax, %rcx);
861
862	cmpl $32, %r8d;
863	je .Ldec_max32;
864
865.Ldec_max24:
866	dec_rounds32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
867		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
868		     %ymm15, %rax, %rcx, 16);
869
870	fls32(%rax, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
871	      %rcx, %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
872	      %ymm15,
873	      ((key_table + (16) * 8) + 8)(CTX),
874	      ((key_table + (16) * 8) + 12)(CTX),
875	      ((key_table + (16) * 8) + 0)(CTX),
876	      ((key_table + (16) * 8) + 4)(CTX));
877
878	dec_rounds32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
879		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
880		     %ymm15, %rax, %rcx, 8);
881
882	fls32(%rax, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
883	      %rcx, %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
884	      %ymm15,
885	      ((key_table + (8) * 8) + 8)(CTX),
886	      ((key_table + (8) * 8) + 12)(CTX),
887	      ((key_table + (8) * 8) + 0)(CTX),
888	      ((key_table + (8) * 8) + 4)(CTX));
889
890	dec_rounds32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
891		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
892		     %ymm15, %rax, %rcx, 0);
893
894	/* load CD for output */
895	vmovdqu 0 * 32(%rcx), %ymm8;
896	vmovdqu 1 * 32(%rcx), %ymm9;
897	vmovdqu 2 * 32(%rcx), %ymm10;
898	vmovdqu 3 * 32(%rcx), %ymm11;
899	vmovdqu 4 * 32(%rcx), %ymm12;
900	vmovdqu 5 * 32(%rcx), %ymm13;
901	vmovdqu 6 * 32(%rcx), %ymm14;
902	vmovdqu 7 * 32(%rcx), %ymm15;
903
904	outunpack32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
905		    %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
906		    %ymm15, (key_table)(CTX), (%rax), 1 * 32(%rax));
907
908	ret;
909
910.align 8
911.Ldec_max32:
912	dec_rounds32(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
913		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
914		     %ymm15, %rax, %rcx, 24);
915
916	fls32(%rax, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
917	      %rcx, %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
918	      %ymm15,
919	      ((key_table + (24) * 8) + 8)(CTX),
920	      ((key_table + (24) * 8) + 12)(CTX),
921	      ((key_table + (24) * 8) + 0)(CTX),
922	      ((key_table + (24) * 8) + 4)(CTX));
923
924	jmp .Ldec_max24;
925ENDPROC(__camellia_dec_blk32)
926
927ENTRY(camellia_ecb_enc_32way)
928	/* input:
929	 *	%rdi: ctx, CTX
930	 *	%rsi: dst (32 blocks)
931	 *	%rdx: src (32 blocks)
932	 */
933
934	vzeroupper;
935
936	inpack32_pre(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
937		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
938		     %ymm15, %rdx, (key_table)(CTX));
939
940	/* now dst can be used as temporary buffer (even in src == dst case) */
941	movq	%rsi, %rax;
942
943	call __camellia_enc_blk32;
944
945	write_output(%ymm7, %ymm6, %ymm5, %ymm4, %ymm3, %ymm2, %ymm1, %ymm0,
946		     %ymm15, %ymm14, %ymm13, %ymm12, %ymm11, %ymm10, %ymm9,
947		     %ymm8, %rsi);
948
949	vzeroupper;
950
951	ret;
952ENDPROC(camellia_ecb_enc_32way)
953
954ENTRY(camellia_ecb_dec_32way)
955	/* input:
956	 *	%rdi: ctx, CTX
957	 *	%rsi: dst (32 blocks)
958	 *	%rdx: src (32 blocks)
959	 */
960
961	vzeroupper;
962
963	cmpl $16, key_length(CTX);
964	movl $32, %r8d;
965	movl $24, %eax;
966	cmovel %eax, %r8d; /* max */
967
968	inpack32_pre(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
969		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
970		     %ymm15, %rdx, (key_table)(CTX, %r8, 8));
971
972	/* now dst can be used as temporary buffer (even in src == dst case) */
973	movq	%rsi, %rax;
974
975	call __camellia_dec_blk32;
976
977	write_output(%ymm7, %ymm6, %ymm5, %ymm4, %ymm3, %ymm2, %ymm1, %ymm0,
978		     %ymm15, %ymm14, %ymm13, %ymm12, %ymm11, %ymm10, %ymm9,
979		     %ymm8, %rsi);
980
981	vzeroupper;
982
983	ret;
984ENDPROC(camellia_ecb_dec_32way)
985
986ENTRY(camellia_cbc_dec_32way)
987	/* input:
988	 *	%rdi: ctx, CTX
989	 *	%rsi: dst (32 blocks)
990	 *	%rdx: src (32 blocks)
991	 */
992
993	vzeroupper;
994
995	cmpl $16, key_length(CTX);
996	movl $32, %r8d;
997	movl $24, %eax;
998	cmovel %eax, %r8d; /* max */
999
1000	inpack32_pre(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %ymm7,
1001		     %ymm8, %ymm9, %ymm10, %ymm11, %ymm12, %ymm13, %ymm14,
1002		     %ymm15, %rdx, (key_table)(CTX, %r8, 8));
1003
1004	movq %rsp, %r10;
1005	cmpq %rsi, %rdx;
1006	je .Lcbc_dec_use_stack;
1007
1008	/* dst can be used as temporary storage, src is not overwritten. */
1009	movq %rsi, %rax;
1010	jmp .Lcbc_dec_continue;
1011
1012.Lcbc_dec_use_stack:
1013	/*
1014	 * dst still in-use (because dst == src), so use stack for temporary
1015	 * storage.
1016	 */
1017	subq $(16 * 32), %rsp;
1018	movq %rsp, %rax;
1019
1020.Lcbc_dec_continue:
1021	call __camellia_dec_blk32;
1022
1023	vmovdqu %ymm7, (%rax);
1024	vpxor %ymm7, %ymm7, %ymm7;
1025	vinserti128 $1, (%rdx), %ymm7, %ymm7;
1026	vpxor (%rax), %ymm7, %ymm7;
1027	movq %r10, %rsp;
1028	vpxor (0 * 32 + 16)(%rdx), %ymm6, %ymm6;
1029	vpxor (1 * 32 + 16)(%rdx), %ymm5, %ymm5;
1030	vpxor (2 * 32 + 16)(%rdx), %ymm4, %ymm4;
1031	vpxor (3 * 32 + 16)(%rdx), %ymm3, %ymm3;
1032	vpxor (4 * 32 + 16)(%rdx), %ymm2, %ymm2;
1033	vpxor (5 * 32 + 16)(%rdx), %ymm1, %ymm1;
1034	vpxor (6 * 32 + 16)(%rdx), %ymm0, %ymm0;
1035	vpxor (7 * 32 + 16)(%rdx), %ymm15, %ymm15;
1036	vpxor (8 * 32 + 16)(%rdx), %ymm14, %ymm14;
1037	vpxor (9 * 32 + 16)(%rdx), %ymm13, %ymm13;
1038	vpxor (10 * 32 + 16)(%rdx), %ymm12, %ymm12;
1039	vpxor (11 * 32 + 16)(%rdx), %ymm11, %ymm11;
1040	vpxor (12 * 32 + 16)(%rdx), %ymm10, %ymm10;
1041	vpxor (13 * 32 + 16)(%rdx), %ymm9, %ymm9;
1042	vpxor (14 * 32 + 16)(%rdx), %ymm8, %ymm8;
1043	write_output(%ymm7, %ymm6, %ymm5, %ymm4, %ymm3, %ymm2, %ymm1, %ymm0,
1044		     %ymm15, %ymm14, %ymm13, %ymm12, %ymm11, %ymm10, %ymm9,
1045		     %ymm8, %rsi);
1046
1047	vzeroupper;
1048
1049	ret;
1050ENDPROC(camellia_cbc_dec_32way)
1051
1052#define inc_le128(x, minus_one, tmp) \
1053	vpcmpeqq minus_one, x, tmp; \
1054	vpsubq minus_one, x, x; \
1055	vpslldq $8, tmp, tmp; \
1056	vpsubq tmp, x, x;
1057
1058#define add2_le128(x, minus_one, minus_two, tmp1, tmp2) \
1059	vpcmpeqq minus_one, x, tmp1; \
1060	vpcmpeqq minus_two, x, tmp2; \
1061	vpsubq minus_two, x, x; \
1062	vpor tmp2, tmp1, tmp1; \
1063	vpslldq $8, tmp1, tmp1; \
1064	vpsubq tmp1, x, x;
1065
1066ENTRY(camellia_ctr_32way)
1067	/* input:
1068	 *	%rdi: ctx, CTX
1069	 *	%rsi: dst (32 blocks)
1070	 *	%rdx: src (32 blocks)
1071	 *	%rcx: iv (little endian, 128bit)
1072	 */
1073
1074	vzeroupper;
1075
1076	movq %rsp, %r10;
1077	cmpq %rsi, %rdx;
1078	je .Lctr_use_stack;
1079
1080	/* dst can be used as temporary storage, src is not overwritten. */
1081	movq %rsi, %rax;
1082	jmp .Lctr_continue;
1083
1084.Lctr_use_stack:
1085	subq $(16 * 32), %rsp;
1086	movq %rsp, %rax;
1087
1088.Lctr_continue:
1089	vpcmpeqd %ymm15, %ymm15, %ymm15;
1090	vpsrldq $8, %ymm15, %ymm15; /* ab: -1:0 ; cd: -1:0 */
1091	vpaddq %ymm15, %ymm15, %ymm12; /* ab: -2:0 ; cd: -2:0 */
1092
1093	/* load IV and byteswap */
1094	vmovdqu (%rcx), %xmm0;
1095	vmovdqa %xmm0, %xmm1;
1096	inc_le128(%xmm0, %xmm15, %xmm14);
1097	vbroadcasti128 .Lbswap128_mask, %ymm14;
1098	vinserti128 $1, %xmm0, %ymm1, %ymm0;
1099	vpshufb %ymm14, %ymm0, %ymm13;
1100	vmovdqu %ymm13, 15 * 32(%rax);
1101
1102	/* construct IVs */
1103	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13); /* ab:le2 ; cd:le3 */
1104	vpshufb %ymm14, %ymm0, %ymm13;
1105	vmovdqu %ymm13, 14 * 32(%rax);
1106	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1107	vpshufb %ymm14, %ymm0, %ymm13;
1108	vmovdqu %ymm13, 13 * 32(%rax);
1109	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1110	vpshufb %ymm14, %ymm0, %ymm13;
1111	vmovdqu %ymm13, 12 * 32(%rax);
1112	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1113	vpshufb %ymm14, %ymm0, %ymm13;
1114	vmovdqu %ymm13, 11 * 32(%rax);
1115	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1116	vpshufb %ymm14, %ymm0, %ymm10;
1117	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1118	vpshufb %ymm14, %ymm0, %ymm9;
1119	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1120	vpshufb %ymm14, %ymm0, %ymm8;
1121	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1122	vpshufb %ymm14, %ymm0, %ymm7;
1123	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1124	vpshufb %ymm14, %ymm0, %ymm6;
1125	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1126	vpshufb %ymm14, %ymm0, %ymm5;
1127	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1128	vpshufb %ymm14, %ymm0, %ymm4;
1129	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1130	vpshufb %ymm14, %ymm0, %ymm3;
1131	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1132	vpshufb %ymm14, %ymm0, %ymm2;
1133	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1134	vpshufb %ymm14, %ymm0, %ymm1;
1135	add2_le128(%ymm0, %ymm15, %ymm12, %ymm11, %ymm13);
1136	vextracti128 $1, %ymm0, %xmm13;
1137	vpshufb %ymm14, %ymm0, %ymm0;
1138	inc_le128(%xmm13, %xmm15, %xmm14);
1139	vmovdqu %xmm13, (%rcx);
1140
1141	/* inpack32_pre: */
1142	vpbroadcastq (key_table)(CTX), %ymm15;
1143	vpshufb .Lpack_bswap, %ymm15, %ymm15;
1144	vpxor %ymm0, %ymm15, %ymm0;
1145	vpxor %ymm1, %ymm15, %ymm1;
1146	vpxor %ymm2, %ymm15, %ymm2;
1147	vpxor %ymm3, %ymm15, %ymm3;
1148	vpxor %ymm4, %ymm15, %ymm4;
1149	vpxor %ymm5, %ymm15, %ymm5;
1150	vpxor %ymm6, %ymm15, %ymm6;
1151	vpxor %ymm7, %ymm15, %ymm7;
1152	vpxor %ymm8, %ymm15, %ymm8;
1153	vpxor %ymm9, %ymm15, %ymm9;
1154	vpxor %ymm10, %ymm15, %ymm10;
1155	vpxor 11 * 32(%rax), %ymm15, %ymm11;
1156	vpxor 12 * 32(%rax), %ymm15, %ymm12;
1157	vpxor 13 * 32(%rax), %ymm15, %ymm13;
1158	vpxor 14 * 32(%rax), %ymm15, %ymm14;
1159	vpxor 15 * 32(%rax), %ymm15, %ymm15;
1160
1161	call __camellia_enc_blk32;
1162
1163	movq %r10, %rsp;
1164
1165	vpxor 0 * 32(%rdx), %ymm7, %ymm7;
1166	vpxor 1 * 32(%rdx), %ymm6, %ymm6;
1167	vpxor 2 * 32(%rdx), %ymm5, %ymm5;
1168	vpxor 3 * 32(%rdx), %ymm4, %ymm4;
1169	vpxor 4 * 32(%rdx), %ymm3, %ymm3;
1170	vpxor 5 * 32(%rdx), %ymm2, %ymm2;
1171	vpxor 6 * 32(%rdx), %ymm1, %ymm1;
1172	vpxor 7 * 32(%rdx), %ymm0, %ymm0;
1173	vpxor 8 * 32(%rdx), %ymm15, %ymm15;
1174	vpxor 9 * 32(%rdx), %ymm14, %ymm14;
1175	vpxor 10 * 32(%rdx), %ymm13, %ymm13;
1176	vpxor 11 * 32(%rdx), %ymm12, %ymm12;
1177	vpxor 12 * 32(%rdx), %ymm11, %ymm11;
1178	vpxor 13 * 32(%rdx), %ymm10, %ymm10;
1179	vpxor 14 * 32(%rdx), %ymm9, %ymm9;
1180	vpxor 15 * 32(%rdx), %ymm8, %ymm8;
1181	write_output(%ymm7, %ymm6, %ymm5, %ymm4, %ymm3, %ymm2, %ymm1, %ymm0,
1182		     %ymm15, %ymm14, %ymm13, %ymm12, %ymm11, %ymm10, %ymm9,
1183		     %ymm8, %rsi);
1184
1185	vzeroupper;
1186
1187	ret;
1188ENDPROC(camellia_ctr_32way)
1189
1190#define gf128mul_x_ble(iv, mask, tmp) \
1191	vpsrad $31, iv, tmp; \
1192	vpaddq iv, iv, iv; \
1193	vpshufd $0x13, tmp, tmp; \
1194	vpand mask, tmp, tmp; \
1195	vpxor tmp, iv, iv;
1196
1197#define gf128mul_x2_ble(iv, mask1, mask2, tmp0, tmp1) \
1198	vpsrad $31, iv, tmp0; \
1199	vpaddq iv, iv, tmp1; \
1200	vpsllq $2, iv, iv; \
1201	vpshufd $0x13, tmp0, tmp0; \
1202	vpsrad $31, tmp1, tmp1; \
1203	vpand mask2, tmp0, tmp0; \
1204	vpshufd $0x13, tmp1, tmp1; \
1205	vpxor tmp0, iv, iv; \
1206	vpand mask1, tmp1, tmp1; \
1207	vpxor tmp1, iv, iv;
1208
1209.align 8
1210camellia_xts_crypt_32way:
1211	/* input:
1212	 *	%rdi: ctx, CTX
1213	 *	%rsi: dst (32 blocks)
1214	 *	%rdx: src (32 blocks)
1215	 *	%rcx: iv (t ⊕ αⁿ ∈ GF(2¹²⁸))
1216	 *	%r8: index for input whitening key
1217	 *	%r9: pointer to  __camellia_enc_blk32 or __camellia_dec_blk32
1218	 */
1219
1220	vzeroupper;
1221
1222	subq $(16 * 32), %rsp;
1223	movq %rsp, %rax;
1224
1225	vbroadcasti128 .Lxts_gf128mul_and_shl1_mask_0, %ymm12;
1226
1227	/* load IV and construct second IV */
1228	vmovdqu (%rcx), %xmm0;
1229	vmovdqa %xmm0, %xmm15;
1230	gf128mul_x_ble(%xmm0, %xmm12, %xmm13);
1231	vbroadcasti128 .Lxts_gf128mul_and_shl1_mask_1, %ymm13;
1232	vinserti128 $1, %xmm0, %ymm15, %ymm0;
1233	vpxor 0 * 32(%rdx), %ymm0, %ymm15;
1234	vmovdqu %ymm15, 15 * 32(%rax);
1235	vmovdqu %ymm0, 0 * 32(%rsi);
1236
1237	/* construct IVs */
1238	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1239	vpxor 1 * 32(%rdx), %ymm0, %ymm15;
1240	vmovdqu %ymm15, 14 * 32(%rax);
1241	vmovdqu %ymm0, 1 * 32(%rsi);
1242
1243	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1244	vpxor 2 * 32(%rdx), %ymm0, %ymm15;
1245	vmovdqu %ymm15, 13 * 32(%rax);
1246	vmovdqu %ymm0, 2 * 32(%rsi);
1247
1248	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1249	vpxor 3 * 32(%rdx), %ymm0, %ymm15;
1250	vmovdqu %ymm15, 12 * 32(%rax);
1251	vmovdqu %ymm0, 3 * 32(%rsi);
1252
1253	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1254	vpxor 4 * 32(%rdx), %ymm0, %ymm11;
1255	vmovdqu %ymm0, 4 * 32(%rsi);
1256
1257	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1258	vpxor 5 * 32(%rdx), %ymm0, %ymm10;
1259	vmovdqu %ymm0, 5 * 32(%rsi);
1260
1261	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1262	vpxor 6 * 32(%rdx), %ymm0, %ymm9;
1263	vmovdqu %ymm0, 6 * 32(%rsi);
1264
1265	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1266	vpxor 7 * 32(%rdx), %ymm0, %ymm8;
1267	vmovdqu %ymm0, 7 * 32(%rsi);
1268
1269	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1270	vpxor 8 * 32(%rdx), %ymm0, %ymm7;
1271	vmovdqu %ymm0, 8 * 32(%rsi);
1272
1273	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1274	vpxor 9 * 32(%rdx), %ymm0, %ymm6;
1275	vmovdqu %ymm0, 9 * 32(%rsi);
1276
1277	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1278	vpxor 10 * 32(%rdx), %ymm0, %ymm5;
1279	vmovdqu %ymm0, 10 * 32(%rsi);
1280
1281	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1282	vpxor 11 * 32(%rdx), %ymm0, %ymm4;
1283	vmovdqu %ymm0, 11 * 32(%rsi);
1284
1285	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1286	vpxor 12 * 32(%rdx), %ymm0, %ymm3;
1287	vmovdqu %ymm0, 12 * 32(%rsi);
1288
1289	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1290	vpxor 13 * 32(%rdx), %ymm0, %ymm2;
1291	vmovdqu %ymm0, 13 * 32(%rsi);
1292
1293	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1294	vpxor 14 * 32(%rdx), %ymm0, %ymm1;
1295	vmovdqu %ymm0, 14 * 32(%rsi);
1296
1297	gf128mul_x2_ble(%ymm0, %ymm12, %ymm13, %ymm14, %ymm15);
1298	vpxor 15 * 32(%rdx), %ymm0, %ymm15;
1299	vmovdqu %ymm15, 0 * 32(%rax);
1300	vmovdqu %ymm0, 15 * 32(%rsi);
1301
1302	vextracti128 $1, %ymm0, %xmm0;
1303	gf128mul_x_ble(%xmm0, %xmm12, %xmm15);
1304	vmovdqu %xmm0, (%rcx);
1305
1306	/* inpack32_pre: */
1307	vpbroadcastq (key_table)(CTX, %r8, 8), %ymm15;
1308	vpshufb .Lpack_bswap, %ymm15, %ymm15;
1309	vpxor 0 * 32(%rax), %ymm15, %ymm0;
1310	vpxor %ymm1, %ymm15, %ymm1;
1311	vpxor %ymm2, %ymm15, %ymm2;
1312	vpxor %ymm3, %ymm15, %ymm3;
1313	vpxor %ymm4, %ymm15, %ymm4;
1314	vpxor %ymm5, %ymm15, %ymm5;
1315	vpxor %ymm6, %ymm15, %ymm6;
1316	vpxor %ymm7, %ymm15, %ymm7;
1317	vpxor %ymm8, %ymm15, %ymm8;
1318	vpxor %ymm9, %ymm15, %ymm9;
1319	vpxor %ymm10, %ymm15, %ymm10;
1320	vpxor %ymm11, %ymm15, %ymm11;
1321	vpxor 12 * 32(%rax), %ymm15, %ymm12;
1322	vpxor 13 * 32(%rax), %ymm15, %ymm13;
1323	vpxor 14 * 32(%rax), %ymm15, %ymm14;
1324	vpxor 15 * 32(%rax), %ymm15, %ymm15;
1325
1326	call *%r9;
1327
1328	addq $(16 * 32), %rsp;
1329
1330	vpxor 0 * 32(%rsi), %ymm7, %ymm7;
1331	vpxor 1 * 32(%rsi), %ymm6, %ymm6;
1332	vpxor 2 * 32(%rsi), %ymm5, %ymm5;
1333	vpxor 3 * 32(%rsi), %ymm4, %ymm4;
1334	vpxor 4 * 32(%rsi), %ymm3, %ymm3;
1335	vpxor 5 * 32(%rsi), %ymm2, %ymm2;
1336	vpxor 6 * 32(%rsi), %ymm1, %ymm1;
1337	vpxor 7 * 32(%rsi), %ymm0, %ymm0;
1338	vpxor 8 * 32(%rsi), %ymm15, %ymm15;
1339	vpxor 9 * 32(%rsi), %ymm14, %ymm14;
1340	vpxor 10 * 32(%rsi), %ymm13, %ymm13;
1341	vpxor 11 * 32(%rsi), %ymm12, %ymm12;
1342	vpxor 12 * 32(%rsi), %ymm11, %ymm11;
1343	vpxor 13 * 32(%rsi), %ymm10, %ymm10;
1344	vpxor 14 * 32(%rsi), %ymm9, %ymm9;
1345	vpxor 15 * 32(%rsi), %ymm8, %ymm8;
1346	write_output(%ymm7, %ymm6, %ymm5, %ymm4, %ymm3, %ymm2, %ymm1, %ymm0,
1347		     %ymm15, %ymm14, %ymm13, %ymm12, %ymm11, %ymm10, %ymm9,
1348		     %ymm8, %rsi);
1349
1350	vzeroupper;
1351
1352	ret;
1353ENDPROC(camellia_xts_crypt_32way)
1354
1355ENTRY(camellia_xts_enc_32way)
1356	/* input:
1357	 *	%rdi: ctx, CTX
1358	 *	%rsi: dst (32 blocks)
1359	 *	%rdx: src (32 blocks)
1360	 *	%rcx: iv (t ⊕ αⁿ ∈ GF(2¹²⁸))
1361	 */
1362
1363	xorl %r8d, %r8d; /* input whitening key, 0 for enc */
1364
1365	leaq __camellia_enc_blk32, %r9;
1366
1367	jmp camellia_xts_crypt_32way;
1368ENDPROC(camellia_xts_enc_32way)
1369
1370ENTRY(camellia_xts_dec_32way)
1371	/* input:
1372	 *	%rdi: ctx, CTX
1373	 *	%rsi: dst (32 blocks)
1374	 *	%rdx: src (32 blocks)
1375	 *	%rcx: iv (t ⊕ αⁿ ∈ GF(2¹²⁸))
1376	 */
1377
1378	cmpl $16, key_length(CTX);
1379	movl $32, %r8d;
1380	movl $24, %eax;
1381	cmovel %eax, %r8d;  /* input whitening key, last for dec */
1382
1383	leaq __camellia_dec_blk32, %r9;
1384
1385	jmp camellia_xts_crypt_32way;
1386ENDPROC(camellia_xts_dec_32way)
1387