1/*
2 * Implement AES algorithm in Intel AES-NI instructions.
3 *
4 * The white paper of AES-NI instructions can be downloaded from:
5 *   http://softwarecommunity.intel.com/isn/downloads/intelavx/AES-Instructions-Set_WP.pdf
6 *
7 * Copyright (C) 2008, Intel Corp.
8 *    Author: Huang Ying <ying.huang@intel.com>
9 *            Vinodh Gopal <vinodh.gopal@intel.com>
10 *            Kahraman Akdemir
11 *
12 * Added RFC4106 AES-GCM support for 128-bit keys under the AEAD
13 * interface for 64-bit kernels.
14 *    Authors: Erdinc Ozturk (erdinc.ozturk@intel.com)
15 *             Aidan O'Mahony (aidan.o.mahony@intel.com)
16 *             Adrian Hoban <adrian.hoban@intel.com>
17 *             James Guilford (james.guilford@intel.com)
18 *             Gabriele Paoloni <gabriele.paoloni@intel.com>
19 *             Tadeusz Struk (tadeusz.struk@intel.com)
20 *             Wajdi Feghali (wajdi.k.feghali@intel.com)
21 *    Copyright (c) 2010, Intel Corporation.
22 *
23 * Ported x86_64 version to x86:
24 *    Author: Mathias Krause <minipli@googlemail.com>
25 *
26 * This program is free software; you can redistribute it and/or modify
27 * it under the terms of the GNU General Public License as published by
28 * the Free Software Foundation; either version 2 of the License, or
29 * (at your option) any later version.
30 */
31
32#include <linux/linkage.h>
33#include <asm/inst.h>
34#include <asm/frame.h>
35#include <asm/nospec-branch.h>
36
37/*
38 * The following macros are used to move an (un)aligned 16 byte value to/from
39 * an XMM register.  This can done for either FP or integer values, for FP use
40 * movaps (move aligned packed single) or integer use movdqa (move double quad
41 * aligned).  It doesn't make a performance difference which instruction is used
42 * since Nehalem (original Core i7) was released.  However, the movaps is a byte
43 * shorter, so that is the one we'll use for now. (same for unaligned).
44 */
45#define MOVADQ	movaps
46#define MOVUDQ	movups
47
48#ifdef __x86_64__
49
50# constants in mergeable sections, linker can reorder and merge
51.section	.rodata.cst16.gf128mul_x_ble_mask, "aM", @progbits, 16
52.align 16
53.Lgf128mul_x_ble_mask:
54	.octa 0x00000000000000010000000000000087
55.section	.rodata.cst16.POLY, "aM", @progbits, 16
56.align 16
57POLY:   .octa 0xC2000000000000000000000000000001
58.section	.rodata.cst16.TWOONE, "aM", @progbits, 16
59.align 16
60TWOONE: .octa 0x00000001000000000000000000000001
61
62.section	.rodata.cst16.SHUF_MASK, "aM", @progbits, 16
63.align 16
64SHUF_MASK:  .octa 0x000102030405060708090A0B0C0D0E0F
65.section	.rodata.cst16.MASK1, "aM", @progbits, 16
66.align 16
67MASK1:      .octa 0x0000000000000000ffffffffffffffff
68.section	.rodata.cst16.MASK2, "aM", @progbits, 16
69.align 16
70MASK2:      .octa 0xffffffffffffffff0000000000000000
71.section	.rodata.cst16.ONE, "aM", @progbits, 16
72.align 16
73ONE:        .octa 0x00000000000000000000000000000001
74.section	.rodata.cst16.F_MIN_MASK, "aM", @progbits, 16
75.align 16
76F_MIN_MASK: .octa 0xf1f2f3f4f5f6f7f8f9fafbfcfdfeff0
77.section	.rodata.cst16.dec, "aM", @progbits, 16
78.align 16
79dec:        .octa 0x1
80.section	.rodata.cst16.enc, "aM", @progbits, 16
81.align 16
82enc:        .octa 0x2
83
84# order of these constants should not change.
85# more specifically, ALL_F should follow SHIFT_MASK,
86# and zero should follow ALL_F
87.section	.rodata, "a", @progbits
88.align 16
89SHIFT_MASK: .octa 0x0f0e0d0c0b0a09080706050403020100
90ALL_F:      .octa 0xffffffffffffffffffffffffffffffff
91            .octa 0x00000000000000000000000000000000
92
93.text
94
95
96#define	STACK_OFFSET    8*3
97
98#define AadHash 16*0
99#define AadLen 16*1
100#define InLen (16*1)+8
101#define PBlockEncKey 16*2
102#define OrigIV 16*3
103#define CurCount 16*4
104#define PBlockLen 16*5
105#define	HashKey		16*6	// store HashKey <<1 mod poly here
106#define	HashKey_2	16*7	// store HashKey^2 <<1 mod poly here
107#define	HashKey_3	16*8	// store HashKey^3 <<1 mod poly here
108#define	HashKey_4	16*9	// store HashKey^4 <<1 mod poly here
109#define	HashKey_k	16*10	// store XOR of High 64 bits and Low 64
110				// bits of  HashKey <<1 mod poly here
111				//(for Karatsuba purposes)
112#define	HashKey_2_k	16*11	// store XOR of High 64 bits and Low 64
113				// bits of  HashKey^2 <<1 mod poly here
114				// (for Karatsuba purposes)
115#define	HashKey_3_k	16*12	// store XOR of High 64 bits and Low 64
116				// bits of  HashKey^3 <<1 mod poly here
117				// (for Karatsuba purposes)
118#define	HashKey_4_k	16*13	// store XOR of High 64 bits and Low 64
119				// bits of  HashKey^4 <<1 mod poly here
120				// (for Karatsuba purposes)
121
122#define arg1 rdi
123#define arg2 rsi
124#define arg3 rdx
125#define arg4 rcx
126#define arg5 r8
127#define arg6 r9
128#define arg7 STACK_OFFSET+8(%rsp)
129#define arg8 STACK_OFFSET+16(%rsp)
130#define arg9 STACK_OFFSET+24(%rsp)
131#define arg10 STACK_OFFSET+32(%rsp)
132#define arg11 STACK_OFFSET+40(%rsp)
133#define keysize 2*15*16(%arg1)
134#endif
135
136
137#define STATE1	%xmm0
138#define STATE2	%xmm4
139#define STATE3	%xmm5
140#define STATE4	%xmm6
141#define STATE	STATE1
142#define IN1	%xmm1
143#define IN2	%xmm7
144#define IN3	%xmm8
145#define IN4	%xmm9
146#define IN	IN1
147#define KEY	%xmm2
148#define IV	%xmm3
149
150#define BSWAP_MASK %xmm10
151#define CTR	%xmm11
152#define INC	%xmm12
153
154#define GF128MUL_MASK %xmm10
155
156#ifdef __x86_64__
157#define AREG	%rax
158#define KEYP	%rdi
159#define OUTP	%rsi
160#define UKEYP	OUTP
161#define INP	%rdx
162#define LEN	%rcx
163#define IVP	%r8
164#define KLEN	%r9d
165#define T1	%r10
166#define TKEYP	T1
167#define T2	%r11
168#define TCTR_LOW T2
169#else
170#define AREG	%eax
171#define KEYP	%edi
172#define OUTP	AREG
173#define UKEYP	OUTP
174#define INP	%edx
175#define LEN	%esi
176#define IVP	%ebp
177#define KLEN	%ebx
178#define T1	%ecx
179#define TKEYP	T1
180#endif
181
182.macro FUNC_SAVE
183	push	%r12
184	push	%r13
185	push	%r14
186#
187# states of %xmm registers %xmm6:%xmm15 not saved
188# all %xmm registers are clobbered
189#
190.endm
191
192
193.macro FUNC_RESTORE
194	pop	%r14
195	pop	%r13
196	pop	%r12
197.endm
198
199# Precompute hashkeys.
200# Input: Hash subkey.
201# Output: HashKeys stored in gcm_context_data.  Only needs to be called
202# once per key.
203# clobbers r12, and tmp xmm registers.
204.macro PRECOMPUTE SUBKEY TMP1 TMP2 TMP3 TMP4 TMP5 TMP6 TMP7
205	mov	\SUBKEY, %r12
206	movdqu	(%r12), \TMP3
207	movdqa	SHUF_MASK(%rip), \TMP2
208	PSHUFB_XMM \TMP2, \TMP3
209
210	# precompute HashKey<<1 mod poly from the HashKey (required for GHASH)
211
212	movdqa	\TMP3, \TMP2
213	psllq	$1, \TMP3
214	psrlq	$63, \TMP2
215	movdqa	\TMP2, \TMP1
216	pslldq	$8, \TMP2
217	psrldq	$8, \TMP1
218	por	\TMP2, \TMP3
219
220	# reduce HashKey<<1
221
222	pshufd	$0x24, \TMP1, \TMP2
223	pcmpeqd TWOONE(%rip), \TMP2
224	pand	POLY(%rip), \TMP2
225	pxor	\TMP2, \TMP3
226	movdqa	\TMP3, HashKey(%arg2)
227
228	movdqa	   \TMP3, \TMP5
229	pshufd	   $78, \TMP3, \TMP1
230	pxor	   \TMP3, \TMP1
231	movdqa	   \TMP1, HashKey_k(%arg2)
232
233	GHASH_MUL  \TMP5, \TMP3, \TMP1, \TMP2, \TMP4, \TMP6, \TMP7
234# TMP5 = HashKey^2<<1 (mod poly)
235	movdqa	   \TMP5, HashKey_2(%arg2)
236# HashKey_2 = HashKey^2<<1 (mod poly)
237	pshufd	   $78, \TMP5, \TMP1
238	pxor	   \TMP5, \TMP1
239	movdqa	   \TMP1, HashKey_2_k(%arg2)
240
241	GHASH_MUL  \TMP5, \TMP3, \TMP1, \TMP2, \TMP4, \TMP6, \TMP7
242# TMP5 = HashKey^3<<1 (mod poly)
243	movdqa	   \TMP5, HashKey_3(%arg2)
244	pshufd	   $78, \TMP5, \TMP1
245	pxor	   \TMP5, \TMP1
246	movdqa	   \TMP1, HashKey_3_k(%arg2)
247
248	GHASH_MUL  \TMP5, \TMP3, \TMP1, \TMP2, \TMP4, \TMP6, \TMP7
249# TMP5 = HashKey^3<<1 (mod poly)
250	movdqa	   \TMP5, HashKey_4(%arg2)
251	pshufd	   $78, \TMP5, \TMP1
252	pxor	   \TMP5, \TMP1
253	movdqa	   \TMP1, HashKey_4_k(%arg2)
254.endm
255
256# GCM_INIT initializes a gcm_context struct to prepare for encoding/decoding.
257# Clobbers rax, r10-r13 and xmm0-xmm6, %xmm13
258.macro GCM_INIT Iv SUBKEY AAD AADLEN
259	mov \AADLEN, %r11
260	mov %r11, AadLen(%arg2) # ctx_data.aad_length = aad_length
261	xor %r11d, %r11d
262	mov %r11, InLen(%arg2) # ctx_data.in_length = 0
263	mov %r11, PBlockLen(%arg2) # ctx_data.partial_block_length = 0
264	mov %r11, PBlockEncKey(%arg2) # ctx_data.partial_block_enc_key = 0
265	mov \Iv, %rax
266	movdqu (%rax), %xmm0
267	movdqu %xmm0, OrigIV(%arg2) # ctx_data.orig_IV = iv
268
269	movdqa  SHUF_MASK(%rip), %xmm2
270	PSHUFB_XMM %xmm2, %xmm0
271	movdqu %xmm0, CurCount(%arg2) # ctx_data.current_counter = iv
272
273	PRECOMPUTE \SUBKEY, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %xmm7,
274	movdqa HashKey(%arg2), %xmm13
275
276	CALC_AAD_HASH %xmm13, \AAD, \AADLEN, %xmm0, %xmm1, %xmm2, %xmm3, \
277	%xmm4, %xmm5, %xmm6
278.endm
279
280# GCM_ENC_DEC Encodes/Decodes given data. Assumes that the passed gcm_context
281# struct has been initialized by GCM_INIT.
282# Requires the input data be at least 1 byte long because of READ_PARTIAL_BLOCK
283# Clobbers rax, r10-r13, and xmm0-xmm15
284.macro GCM_ENC_DEC operation
285	movdqu AadHash(%arg2), %xmm8
286	movdqu HashKey(%arg2), %xmm13
287	add %arg5, InLen(%arg2)
288
289	xor %r11d, %r11d # initialise the data pointer offset as zero
290	PARTIAL_BLOCK %arg3 %arg4 %arg5 %r11 %xmm8 \operation
291
292	sub %r11, %arg5		# sub partial block data used
293	mov %arg5, %r13		# save the number of bytes
294
295	and $-16, %r13		# %r13 = %r13 - (%r13 mod 16)
296	mov %r13, %r12
297	# Encrypt/Decrypt first few blocks
298
299	and	$(3<<4), %r12
300	jz	_initial_num_blocks_is_0_\@
301	cmp	$(2<<4), %r12
302	jb	_initial_num_blocks_is_1_\@
303	je	_initial_num_blocks_is_2_\@
304_initial_num_blocks_is_3_\@:
305	INITIAL_BLOCKS_ENC_DEC	%xmm9, %xmm10, %xmm13, %xmm11, %xmm12, %xmm0, \
306%xmm1, %xmm2, %xmm3, %xmm4, %xmm8, %xmm5, %xmm6, 5, 678, \operation
307	sub	$48, %r13
308	jmp	_initial_blocks_\@
309_initial_num_blocks_is_2_\@:
310	INITIAL_BLOCKS_ENC_DEC	%xmm9, %xmm10, %xmm13, %xmm11, %xmm12, %xmm0, \
311%xmm1, %xmm2, %xmm3, %xmm4, %xmm8, %xmm5, %xmm6, 6, 78, \operation
312	sub	$32, %r13
313	jmp	_initial_blocks_\@
314_initial_num_blocks_is_1_\@:
315	INITIAL_BLOCKS_ENC_DEC	%xmm9, %xmm10, %xmm13, %xmm11, %xmm12, %xmm0, \
316%xmm1, %xmm2, %xmm3, %xmm4, %xmm8, %xmm5, %xmm6, 7, 8, \operation
317	sub	$16, %r13
318	jmp	_initial_blocks_\@
319_initial_num_blocks_is_0_\@:
320	INITIAL_BLOCKS_ENC_DEC	%xmm9, %xmm10, %xmm13, %xmm11, %xmm12, %xmm0, \
321%xmm1, %xmm2, %xmm3, %xmm4, %xmm8, %xmm5, %xmm6, 8, 0, \operation
322_initial_blocks_\@:
323
324	# Main loop - Encrypt/Decrypt remaining blocks
325
326	cmp	$0, %r13
327	je	_zero_cipher_left_\@
328	sub	$64, %r13
329	je	_four_cipher_left_\@
330_crypt_by_4_\@:
331	GHASH_4_ENCRYPT_4_PARALLEL_\operation	%xmm9, %xmm10, %xmm11, %xmm12, \
332	%xmm13, %xmm14, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, \
333	%xmm7, %xmm8, enc
334	add	$64, %r11
335	sub	$64, %r13
336	jne	_crypt_by_4_\@
337_four_cipher_left_\@:
338	GHASH_LAST_4	%xmm9, %xmm10, %xmm11, %xmm12, %xmm13, %xmm14, \
339%xmm15, %xmm1, %xmm2, %xmm3, %xmm4, %xmm8
340_zero_cipher_left_\@:
341	movdqu %xmm8, AadHash(%arg2)
342	movdqu %xmm0, CurCount(%arg2)
343
344	mov	%arg5, %r13
345	and	$15, %r13			# %r13 = arg5 (mod 16)
346	je	_multiple_of_16_bytes_\@
347
348	mov %r13, PBlockLen(%arg2)
349
350	# Handle the last <16 Byte block separately
351	paddd ONE(%rip), %xmm0                # INCR CNT to get Yn
352	movdqu %xmm0, CurCount(%arg2)
353	movdqa SHUF_MASK(%rip), %xmm10
354	PSHUFB_XMM %xmm10, %xmm0
355
356	ENCRYPT_SINGLE_BLOCK	%xmm0, %xmm1        # Encrypt(K, Yn)
357	movdqu %xmm0, PBlockEncKey(%arg2)
358
359	cmp	$16, %arg5
360	jge _large_enough_update_\@
361
362	lea (%arg4,%r11,1), %r10
363	mov %r13, %r12
364	READ_PARTIAL_BLOCK %r10 %r12 %xmm2 %xmm1
365	jmp _data_read_\@
366
367_large_enough_update_\@:
368	sub	$16, %r11
369	add	%r13, %r11
370
371	# receive the last <16 Byte block
372	movdqu	(%arg4, %r11, 1), %xmm1
373
374	sub	%r13, %r11
375	add	$16, %r11
376
377	lea	SHIFT_MASK+16(%rip), %r12
378	# adjust the shuffle mask pointer to be able to shift 16-r13 bytes
379	# (r13 is the number of bytes in plaintext mod 16)
380	sub	%r13, %r12
381	# get the appropriate shuffle mask
382	movdqu	(%r12), %xmm2
383	# shift right 16-r13 bytes
384	PSHUFB_XMM  %xmm2, %xmm1
385
386_data_read_\@:
387	lea ALL_F+16(%rip), %r12
388	sub %r13, %r12
389
390.ifc \operation, dec
391	movdqa  %xmm1, %xmm2
392.endif
393	pxor	%xmm1, %xmm0            # XOR Encrypt(K, Yn)
394	movdqu	(%r12), %xmm1
395	# get the appropriate mask to mask out top 16-r13 bytes of xmm0
396	pand	%xmm1, %xmm0            # mask out top 16-r13 bytes of xmm0
397.ifc \operation, dec
398	pand    %xmm1, %xmm2
399	movdqa SHUF_MASK(%rip), %xmm10
400	PSHUFB_XMM %xmm10 ,%xmm2
401
402	pxor %xmm2, %xmm8
403.else
404	movdqa SHUF_MASK(%rip), %xmm10
405	PSHUFB_XMM %xmm10,%xmm0
406
407	pxor	%xmm0, %xmm8
408.endif
409
410	movdqu %xmm8, AadHash(%arg2)
411.ifc \operation, enc
412	# GHASH computation for the last <16 byte block
413	movdqa SHUF_MASK(%rip), %xmm10
414	# shuffle xmm0 back to output as ciphertext
415	PSHUFB_XMM %xmm10, %xmm0
416.endif
417
418	# Output %r13 bytes
419	MOVQ_R64_XMM %xmm0, %rax
420	cmp $8, %r13
421	jle _less_than_8_bytes_left_\@
422	mov %rax, (%arg3 , %r11, 1)
423	add $8, %r11
424	psrldq $8, %xmm0
425	MOVQ_R64_XMM %xmm0, %rax
426	sub $8, %r13
427_less_than_8_bytes_left_\@:
428	mov %al,  (%arg3, %r11, 1)
429	add $1, %r11
430	shr $8, %rax
431	sub $1, %r13
432	jne _less_than_8_bytes_left_\@
433_multiple_of_16_bytes_\@:
434.endm
435
436# GCM_COMPLETE Finishes update of tag of last partial block
437# Output: Authorization Tag (AUTH_TAG)
438# Clobbers rax, r10-r12, and xmm0, xmm1, xmm5-xmm15
439.macro GCM_COMPLETE AUTHTAG AUTHTAGLEN
440	movdqu AadHash(%arg2), %xmm8
441	movdqu HashKey(%arg2), %xmm13
442
443	mov PBlockLen(%arg2), %r12
444
445	cmp $0, %r12
446	je _partial_done\@
447
448	GHASH_MUL %xmm8, %xmm13, %xmm9, %xmm10, %xmm11, %xmm5, %xmm6
449
450_partial_done\@:
451	mov AadLen(%arg2), %r12  # %r13 = aadLen (number of bytes)
452	shl	$3, %r12		  # convert into number of bits
453	movd	%r12d, %xmm15		  # len(A) in %xmm15
454	mov InLen(%arg2), %r12
455	shl     $3, %r12                  # len(C) in bits (*128)
456	MOVQ_R64_XMM    %r12, %xmm1
457
458	pslldq	$8, %xmm15		  # %xmm15 = len(A)||0x0000000000000000
459	pxor	%xmm1, %xmm15		  # %xmm15 = len(A)||len(C)
460	pxor	%xmm15, %xmm8
461	GHASH_MUL	%xmm8, %xmm13, %xmm9, %xmm10, %xmm11, %xmm5, %xmm6
462	# final GHASH computation
463	movdqa SHUF_MASK(%rip), %xmm10
464	PSHUFB_XMM %xmm10, %xmm8
465
466	movdqu OrigIV(%arg2), %xmm0       # %xmm0 = Y0
467	ENCRYPT_SINGLE_BLOCK	%xmm0,  %xmm1	  # E(K, Y0)
468	pxor	%xmm8, %xmm0
469_return_T_\@:
470	mov	\AUTHTAG, %r10                     # %r10 = authTag
471	mov	\AUTHTAGLEN, %r11                    # %r11 = auth_tag_len
472	cmp	$16, %r11
473	je	_T_16_\@
474	cmp	$8, %r11
475	jl	_T_4_\@
476_T_8_\@:
477	MOVQ_R64_XMM	%xmm0, %rax
478	mov	%rax, (%r10)
479	add	$8, %r10
480	sub	$8, %r11
481	psrldq	$8, %xmm0
482	cmp	$0, %r11
483	je	_return_T_done_\@
484_T_4_\@:
485	movd	%xmm0, %eax
486	mov	%eax, (%r10)
487	add	$4, %r10
488	sub	$4, %r11
489	psrldq	$4, %xmm0
490	cmp	$0, %r11
491	je	_return_T_done_\@
492_T_123_\@:
493	movd	%xmm0, %eax
494	cmp	$2, %r11
495	jl	_T_1_\@
496	mov	%ax, (%r10)
497	cmp	$2, %r11
498	je	_return_T_done_\@
499	add	$2, %r10
500	sar	$16, %eax
501_T_1_\@:
502	mov	%al, (%r10)
503	jmp	_return_T_done_\@
504_T_16_\@:
505	movdqu	%xmm0, (%r10)
506_return_T_done_\@:
507.endm
508
509#ifdef __x86_64__
510/* GHASH_MUL MACRO to implement: Data*HashKey mod (128,127,126,121,0)
511*
512*
513* Input: A and B (128-bits each, bit-reflected)
514* Output: C = A*B*x mod poly, (i.e. >>1 )
515* To compute GH = GH*HashKey mod poly, give HK = HashKey<<1 mod poly as input
516* GH = GH * HK * x mod poly which is equivalent to GH*HashKey mod poly.
517*
518*/
519.macro GHASH_MUL GH HK TMP1 TMP2 TMP3 TMP4 TMP5
520	movdqa	  \GH, \TMP1
521	pshufd	  $78, \GH, \TMP2
522	pshufd	  $78, \HK, \TMP3
523	pxor	  \GH, \TMP2            # TMP2 = a1+a0
524	pxor	  \HK, \TMP3            # TMP3 = b1+b0
525	PCLMULQDQ 0x11, \HK, \TMP1     # TMP1 = a1*b1
526	PCLMULQDQ 0x00, \HK, \GH       # GH = a0*b0
527	PCLMULQDQ 0x00, \TMP3, \TMP2   # TMP2 = (a0+a1)*(b1+b0)
528	pxor	  \GH, \TMP2
529	pxor	  \TMP1, \TMP2          # TMP2 = (a0*b0)+(a1*b0)
530	movdqa	  \TMP2, \TMP3
531	pslldq	  $8, \TMP3             # left shift TMP3 2 DWs
532	psrldq	  $8, \TMP2             # right shift TMP2 2 DWs
533	pxor	  \TMP3, \GH
534	pxor	  \TMP2, \TMP1          # TMP2:GH holds the result of GH*HK
535
536        # first phase of the reduction
537
538	movdqa    \GH, \TMP2
539	movdqa    \GH, \TMP3
540	movdqa    \GH, \TMP4            # copy GH into TMP2,TMP3 and TMP4
541					# in in order to perform
542					# independent shifts
543	pslld     $31, \TMP2            # packed right shift <<31
544	pslld     $30, \TMP3            # packed right shift <<30
545	pslld     $25, \TMP4            # packed right shift <<25
546	pxor      \TMP3, \TMP2          # xor the shifted versions
547	pxor      \TMP4, \TMP2
548	movdqa    \TMP2, \TMP5
549	psrldq    $4, \TMP5             # right shift TMP5 1 DW
550	pslldq    $12, \TMP2            # left shift TMP2 3 DWs
551	pxor      \TMP2, \GH
552
553        # second phase of the reduction
554
555	movdqa    \GH,\TMP2             # copy GH into TMP2,TMP3 and TMP4
556					# in in order to perform
557					# independent shifts
558	movdqa    \GH,\TMP3
559	movdqa    \GH,\TMP4
560	psrld     $1,\TMP2              # packed left shift >>1
561	psrld     $2,\TMP3              # packed left shift >>2
562	psrld     $7,\TMP4              # packed left shift >>7
563	pxor      \TMP3,\TMP2		# xor the shifted versions
564	pxor      \TMP4,\TMP2
565	pxor      \TMP5, \TMP2
566	pxor      \TMP2, \GH
567	pxor      \TMP1, \GH            # result is in TMP1
568.endm
569
570# Reads DLEN bytes starting at DPTR and stores in XMMDst
571# where 0 < DLEN < 16
572# Clobbers %rax, DLEN and XMM1
573.macro READ_PARTIAL_BLOCK DPTR DLEN XMM1 XMMDst
574        cmp $8, \DLEN
575        jl _read_lt8_\@
576        mov (\DPTR), %rax
577        MOVQ_R64_XMM %rax, \XMMDst
578        sub $8, \DLEN
579        jz _done_read_partial_block_\@
580	xor %eax, %eax
581_read_next_byte_\@:
582        shl $8, %rax
583        mov 7(\DPTR, \DLEN, 1), %al
584        dec \DLEN
585        jnz _read_next_byte_\@
586        MOVQ_R64_XMM %rax, \XMM1
587	pslldq $8, \XMM1
588        por \XMM1, \XMMDst
589	jmp _done_read_partial_block_\@
590_read_lt8_\@:
591	xor %eax, %eax
592_read_next_byte_lt8_\@:
593        shl $8, %rax
594        mov -1(\DPTR, \DLEN, 1), %al
595        dec \DLEN
596        jnz _read_next_byte_lt8_\@
597        MOVQ_R64_XMM %rax, \XMMDst
598_done_read_partial_block_\@:
599.endm
600
601# CALC_AAD_HASH: Calculates the hash of the data which will not be encrypted.
602# clobbers r10-11, xmm14
603.macro CALC_AAD_HASH HASHKEY AAD AADLEN TMP1 TMP2 TMP3 TMP4 TMP5 \
604	TMP6 TMP7
605	MOVADQ	   SHUF_MASK(%rip), %xmm14
606	mov	   \AAD, %r10		# %r10 = AAD
607	mov	   \AADLEN, %r11		# %r11 = aadLen
608	pxor	   \TMP7, \TMP7
609	pxor	   \TMP6, \TMP6
610
611	cmp	   $16, %r11
612	jl	   _get_AAD_rest\@
613_get_AAD_blocks\@:
614	movdqu	   (%r10), \TMP7
615	PSHUFB_XMM   %xmm14, \TMP7 # byte-reflect the AAD data
616	pxor	   \TMP7, \TMP6
617	GHASH_MUL  \TMP6, \HASHKEY, \TMP1, \TMP2, \TMP3, \TMP4, \TMP5
618	add	   $16, %r10
619	sub	   $16, %r11
620	cmp	   $16, %r11
621	jge	   _get_AAD_blocks\@
622
623	movdqu	   \TMP6, \TMP7
624
625	/* read the last <16B of AAD */
626_get_AAD_rest\@:
627	cmp	   $0, %r11
628	je	   _get_AAD_done\@
629
630	READ_PARTIAL_BLOCK %r10, %r11, \TMP1, \TMP7
631	PSHUFB_XMM   %xmm14, \TMP7 # byte-reflect the AAD data
632	pxor	   \TMP6, \TMP7
633	GHASH_MUL  \TMP7, \HASHKEY, \TMP1, \TMP2, \TMP3, \TMP4, \TMP5
634	movdqu \TMP7, \TMP6
635
636_get_AAD_done\@:
637	movdqu \TMP6, AadHash(%arg2)
638.endm
639
640# PARTIAL_BLOCK: Handles encryption/decryption and the tag partial blocks
641# between update calls.
642# Requires the input data be at least 1 byte long due to READ_PARTIAL_BLOCK
643# Outputs encrypted bytes, and updates hash and partial info in gcm_data_context
644# Clobbers rax, r10, r12, r13, xmm0-6, xmm9-13
645.macro PARTIAL_BLOCK CYPH_PLAIN_OUT PLAIN_CYPH_IN PLAIN_CYPH_LEN DATA_OFFSET \
646	AAD_HASH operation
647	mov 	PBlockLen(%arg2), %r13
648	cmp	$0, %r13
649	je	_partial_block_done_\@	# Leave Macro if no partial blocks
650	# Read in input data without over reading
651	cmp	$16, \PLAIN_CYPH_LEN
652	jl	_fewer_than_16_bytes_\@
653	movups	(\PLAIN_CYPH_IN), %xmm1	# If more than 16 bytes, just fill xmm
654	jmp	_data_read_\@
655
656_fewer_than_16_bytes_\@:
657	lea	(\PLAIN_CYPH_IN, \DATA_OFFSET, 1), %r10
658	mov	\PLAIN_CYPH_LEN, %r12
659	READ_PARTIAL_BLOCK %r10 %r12 %xmm0 %xmm1
660
661	mov PBlockLen(%arg2), %r13
662
663_data_read_\@:				# Finished reading in data
664
665	movdqu	PBlockEncKey(%arg2), %xmm9
666	movdqu	HashKey(%arg2), %xmm13
667
668	lea	SHIFT_MASK(%rip), %r12
669
670	# adjust the shuffle mask pointer to be able to shift r13 bytes
671	# r16-r13 is the number of bytes in plaintext mod 16)
672	add	%r13, %r12
673	movdqu	(%r12), %xmm2		# get the appropriate shuffle mask
674	PSHUFB_XMM %xmm2, %xmm9		# shift right r13 bytes
675
676.ifc \operation, dec
677	movdqa	%xmm1, %xmm3
678	pxor	%xmm1, %xmm9		# Cyphertext XOR E(K, Yn)
679
680	mov	\PLAIN_CYPH_LEN, %r10
681	add	%r13, %r10
682	# Set r10 to be the amount of data left in CYPH_PLAIN_IN after filling
683	sub	$16, %r10
684	# Determine if if partial block is not being filled and
685	# shift mask accordingly
686	jge	_no_extra_mask_1_\@
687	sub	%r10, %r12
688_no_extra_mask_1_\@:
689
690	movdqu	ALL_F-SHIFT_MASK(%r12), %xmm1
691	# get the appropriate mask to mask out bottom r13 bytes of xmm9
692	pand	%xmm1, %xmm9		# mask out bottom r13 bytes of xmm9
693
694	pand	%xmm1, %xmm3
695	movdqa	SHUF_MASK(%rip), %xmm10
696	PSHUFB_XMM	%xmm10, %xmm3
697	PSHUFB_XMM	%xmm2, %xmm3
698	pxor	%xmm3, \AAD_HASH
699
700	cmp	$0, %r10
701	jl	_partial_incomplete_1_\@
702
703	# GHASH computation for the last <16 Byte block
704	GHASH_MUL \AAD_HASH, %xmm13, %xmm0, %xmm10, %xmm11, %xmm5, %xmm6
705	xor	%eax, %eax
706
707	mov	%rax, PBlockLen(%arg2)
708	jmp	_dec_done_\@
709_partial_incomplete_1_\@:
710	add	\PLAIN_CYPH_LEN, PBlockLen(%arg2)
711_dec_done_\@:
712	movdqu	\AAD_HASH, AadHash(%arg2)
713.else
714	pxor	%xmm1, %xmm9			# Plaintext XOR E(K, Yn)
715
716	mov	\PLAIN_CYPH_LEN, %r10
717	add	%r13, %r10
718	# Set r10 to be the amount of data left in CYPH_PLAIN_IN after filling
719	sub	$16, %r10
720	# Determine if if partial block is not being filled and
721	# shift mask accordingly
722	jge	_no_extra_mask_2_\@
723	sub	%r10, %r12
724_no_extra_mask_2_\@:
725
726	movdqu	ALL_F-SHIFT_MASK(%r12), %xmm1
727	# get the appropriate mask to mask out bottom r13 bytes of xmm9
728	pand	%xmm1, %xmm9
729
730	movdqa	SHUF_MASK(%rip), %xmm1
731	PSHUFB_XMM %xmm1, %xmm9
732	PSHUFB_XMM %xmm2, %xmm9
733	pxor	%xmm9, \AAD_HASH
734
735	cmp	$0, %r10
736	jl	_partial_incomplete_2_\@
737
738	# GHASH computation for the last <16 Byte block
739	GHASH_MUL \AAD_HASH, %xmm13, %xmm0, %xmm10, %xmm11, %xmm5, %xmm6
740	xor	%eax, %eax
741
742	mov	%rax, PBlockLen(%arg2)
743	jmp	_encode_done_\@
744_partial_incomplete_2_\@:
745	add	\PLAIN_CYPH_LEN, PBlockLen(%arg2)
746_encode_done_\@:
747	movdqu	\AAD_HASH, AadHash(%arg2)
748
749	movdqa	SHUF_MASK(%rip), %xmm10
750	# shuffle xmm9 back to output as ciphertext
751	PSHUFB_XMM	%xmm10, %xmm9
752	PSHUFB_XMM	%xmm2, %xmm9
753.endif
754	# output encrypted Bytes
755	cmp	$0, %r10
756	jl	_partial_fill_\@
757	mov	%r13, %r12
758	mov	$16, %r13
759	# Set r13 to be the number of bytes to write out
760	sub	%r12, %r13
761	jmp	_count_set_\@
762_partial_fill_\@:
763	mov	\PLAIN_CYPH_LEN, %r13
764_count_set_\@:
765	movdqa	%xmm9, %xmm0
766	MOVQ_R64_XMM	%xmm0, %rax
767	cmp	$8, %r13
768	jle	_less_than_8_bytes_left_\@
769
770	mov	%rax, (\CYPH_PLAIN_OUT, \DATA_OFFSET, 1)
771	add	$8, \DATA_OFFSET
772	psrldq	$8, %xmm0
773	MOVQ_R64_XMM	%xmm0, %rax
774	sub	$8, %r13
775_less_than_8_bytes_left_\@:
776	movb	%al, (\CYPH_PLAIN_OUT, \DATA_OFFSET, 1)
777	add	$1, \DATA_OFFSET
778	shr	$8, %rax
779	sub	$1, %r13
780	jne	_less_than_8_bytes_left_\@
781_partial_block_done_\@:
782.endm # PARTIAL_BLOCK
783
784/*
785* if a = number of total plaintext bytes
786* b = floor(a/16)
787* num_initial_blocks = b mod 4
788* encrypt the initial num_initial_blocks blocks and apply ghash on
789* the ciphertext
790* %r10, %r11, %r12, %rax, %xmm5, %xmm6, %xmm7, %xmm8, %xmm9 registers
791* are clobbered
792* arg1, %arg2, %arg3 are used as a pointer only, not modified
793*/
794
795
796.macro INITIAL_BLOCKS_ENC_DEC TMP1 TMP2 TMP3 TMP4 TMP5 XMM0 XMM1 \
797	XMM2 XMM3 XMM4 XMMDst TMP6 TMP7 i i_seq operation
798	MOVADQ		SHUF_MASK(%rip), %xmm14
799
800	movdqu AadHash(%arg2), %xmm\i		    # XMM0 = Y0
801
802	# start AES for num_initial_blocks blocks
803
804	movdqu CurCount(%arg2), \XMM0                # XMM0 = Y0
805
806.if (\i == 5) || (\i == 6) || (\i == 7)
807
808	MOVADQ		ONE(%RIP),\TMP1
809	MOVADQ		0(%arg1),\TMP2
810.irpc index, \i_seq
811	paddd		\TMP1, \XMM0                 # INCR Y0
812.ifc \operation, dec
813        movdqa     \XMM0, %xmm\index
814.else
815	MOVADQ		\XMM0, %xmm\index
816.endif
817	PSHUFB_XMM	%xmm14, %xmm\index      # perform a 16 byte swap
818	pxor		\TMP2, %xmm\index
819.endr
820	lea	0x10(%arg1),%r10
821	mov	keysize,%eax
822	shr	$2,%eax				# 128->4, 192->6, 256->8
823	add	$5,%eax			      # 128->9, 192->11, 256->13
824
825aes_loop_initial_\@:
826	MOVADQ	(%r10),\TMP1
827.irpc	index, \i_seq
828	AESENC	\TMP1, %xmm\index
829.endr
830	add	$16,%r10
831	sub	$1,%eax
832	jnz	aes_loop_initial_\@
833
834	MOVADQ	(%r10), \TMP1
835.irpc index, \i_seq
836	AESENCLAST \TMP1, %xmm\index         # Last Round
837.endr
838.irpc index, \i_seq
839	movdqu	   (%arg4 , %r11, 1), \TMP1
840	pxor	   \TMP1, %xmm\index
841	movdqu	   %xmm\index, (%arg3 , %r11, 1)
842	# write back plaintext/ciphertext for num_initial_blocks
843	add	   $16, %r11
844
845.ifc \operation, dec
846	movdqa     \TMP1, %xmm\index
847.endif
848	PSHUFB_XMM	   %xmm14, %xmm\index
849
850		# prepare plaintext/ciphertext for GHASH computation
851.endr
852.endif
853
854        # apply GHASH on num_initial_blocks blocks
855
856.if \i == 5
857        pxor       %xmm5, %xmm6
858	GHASH_MUL  %xmm6, \TMP3, \TMP1, \TMP2, \TMP4, \TMP5, \XMM1
859        pxor       %xmm6, %xmm7
860	GHASH_MUL  %xmm7, \TMP3, \TMP1, \TMP2, \TMP4, \TMP5, \XMM1
861        pxor       %xmm7, %xmm8
862	GHASH_MUL  %xmm8, \TMP3, \TMP1, \TMP2, \TMP4, \TMP5, \XMM1
863.elseif \i == 6
864        pxor       %xmm6, %xmm7
865	GHASH_MUL  %xmm7, \TMP3, \TMP1, \TMP2, \TMP4, \TMP5, \XMM1
866        pxor       %xmm7, %xmm8
867	GHASH_MUL  %xmm8, \TMP3, \TMP1, \TMP2, \TMP4, \TMP5, \XMM1
868.elseif \i == 7
869        pxor       %xmm7, %xmm8
870	GHASH_MUL  %xmm8, \TMP3, \TMP1, \TMP2, \TMP4, \TMP5, \XMM1
871.endif
872	cmp	   $64, %r13
873	jl	_initial_blocks_done\@
874	# no need for precomputed values
875/*
876*
877* Precomputations for HashKey parallel with encryption of first 4 blocks.
878* Haskey_i_k holds XORed values of the low and high parts of the Haskey_i
879*/
880	MOVADQ	   ONE(%RIP),\TMP1
881	paddd	   \TMP1, \XMM0              # INCR Y0
882	MOVADQ	   \XMM0, \XMM1
883	PSHUFB_XMM  %xmm14, \XMM1        # perform a 16 byte swap
884
885	paddd	   \TMP1, \XMM0              # INCR Y0
886	MOVADQ	   \XMM0, \XMM2
887	PSHUFB_XMM  %xmm14, \XMM2        # perform a 16 byte swap
888
889	paddd	   \TMP1, \XMM0              # INCR Y0
890	MOVADQ	   \XMM0, \XMM3
891	PSHUFB_XMM %xmm14, \XMM3        # perform a 16 byte swap
892
893	paddd	   \TMP1, \XMM0              # INCR Y0
894	MOVADQ	   \XMM0, \XMM4
895	PSHUFB_XMM %xmm14, \XMM4        # perform a 16 byte swap
896
897	MOVADQ	   0(%arg1),\TMP1
898	pxor	   \TMP1, \XMM1
899	pxor	   \TMP1, \XMM2
900	pxor	   \TMP1, \XMM3
901	pxor	   \TMP1, \XMM4
902.irpc index, 1234 # do 4 rounds
903	movaps 0x10*\index(%arg1), \TMP1
904	AESENC	   \TMP1, \XMM1
905	AESENC	   \TMP1, \XMM2
906	AESENC	   \TMP1, \XMM3
907	AESENC	   \TMP1, \XMM4
908.endr
909.irpc index, 56789 # do next 5 rounds
910	movaps 0x10*\index(%arg1), \TMP1
911	AESENC	   \TMP1, \XMM1
912	AESENC	   \TMP1, \XMM2
913	AESENC	   \TMP1, \XMM3
914	AESENC	   \TMP1, \XMM4
915.endr
916	lea	   0xa0(%arg1),%r10
917	mov	   keysize,%eax
918	shr	   $2,%eax			# 128->4, 192->6, 256->8
919	sub	   $4,%eax			# 128->0, 192->2, 256->4
920	jz	   aes_loop_pre_done\@
921
922aes_loop_pre_\@:
923	MOVADQ	   (%r10),\TMP2
924.irpc	index, 1234
925	AESENC	   \TMP2, %xmm\index
926.endr
927	add	   $16,%r10
928	sub	   $1,%eax
929	jnz	   aes_loop_pre_\@
930
931aes_loop_pre_done\@:
932	MOVADQ	   (%r10), \TMP2
933	AESENCLAST \TMP2, \XMM1
934	AESENCLAST \TMP2, \XMM2
935	AESENCLAST \TMP2, \XMM3
936	AESENCLAST \TMP2, \XMM4
937	movdqu	   16*0(%arg4 , %r11 , 1), \TMP1
938	pxor	   \TMP1, \XMM1
939.ifc \operation, dec
940	movdqu     \XMM1, 16*0(%arg3 , %r11 , 1)
941	movdqa     \TMP1, \XMM1
942.endif
943	movdqu	   16*1(%arg4 , %r11 , 1), \TMP1
944	pxor	   \TMP1, \XMM2
945.ifc \operation, dec
946	movdqu     \XMM2, 16*1(%arg3 , %r11 , 1)
947	movdqa     \TMP1, \XMM2
948.endif
949	movdqu	   16*2(%arg4 , %r11 , 1), \TMP1
950	pxor	   \TMP1, \XMM3
951.ifc \operation, dec
952	movdqu     \XMM3, 16*2(%arg3 , %r11 , 1)
953	movdqa     \TMP1, \XMM3
954.endif
955	movdqu	   16*3(%arg4 , %r11 , 1), \TMP1
956	pxor	   \TMP1, \XMM4
957.ifc \operation, dec
958	movdqu     \XMM4, 16*3(%arg3 , %r11 , 1)
959	movdqa     \TMP1, \XMM4
960.else
961	movdqu     \XMM1, 16*0(%arg3 , %r11 , 1)
962	movdqu     \XMM2, 16*1(%arg3 , %r11 , 1)
963	movdqu     \XMM3, 16*2(%arg3 , %r11 , 1)
964	movdqu     \XMM4, 16*3(%arg3 , %r11 , 1)
965.endif
966
967	add	   $64, %r11
968	PSHUFB_XMM %xmm14, \XMM1 # perform a 16 byte swap
969	pxor	   \XMMDst, \XMM1
970# combine GHASHed value with the corresponding ciphertext
971	PSHUFB_XMM %xmm14, \XMM2 # perform a 16 byte swap
972	PSHUFB_XMM %xmm14, \XMM3 # perform a 16 byte swap
973	PSHUFB_XMM %xmm14, \XMM4 # perform a 16 byte swap
974
975_initial_blocks_done\@:
976
977.endm
978
979/*
980* encrypt 4 blocks at a time
981* ghash the 4 previously encrypted ciphertext blocks
982* arg1, %arg3, %arg4 are used as pointers only, not modified
983* %r11 is the data offset value
984*/
985.macro GHASH_4_ENCRYPT_4_PARALLEL_ENC TMP1 TMP2 TMP3 TMP4 TMP5 \
986TMP6 XMM0 XMM1 XMM2 XMM3 XMM4 XMM5 XMM6 XMM7 XMM8 operation
987
988	movdqa	  \XMM1, \XMM5
989	movdqa	  \XMM2, \XMM6
990	movdqa	  \XMM3, \XMM7
991	movdqa	  \XMM4, \XMM8
992
993        movdqa    SHUF_MASK(%rip), %xmm15
994        # multiply TMP5 * HashKey using karatsuba
995
996	movdqa	  \XMM5, \TMP4
997	pshufd	  $78, \XMM5, \TMP6
998	pxor	  \XMM5, \TMP6
999	paddd     ONE(%rip), \XMM0		# INCR CNT
1000	movdqa	  HashKey_4(%arg2), \TMP5
1001	PCLMULQDQ 0x11, \TMP5, \TMP4           # TMP4 = a1*b1
1002	movdqa    \XMM0, \XMM1
1003	paddd     ONE(%rip), \XMM0		# INCR CNT
1004	movdqa    \XMM0, \XMM2
1005	paddd     ONE(%rip), \XMM0		# INCR CNT
1006	movdqa    \XMM0, \XMM3
1007	paddd     ONE(%rip), \XMM0		# INCR CNT
1008	movdqa    \XMM0, \XMM4
1009	PSHUFB_XMM %xmm15, \XMM1	# perform a 16 byte swap
1010	PCLMULQDQ 0x00, \TMP5, \XMM5           # XMM5 = a0*b0
1011	PSHUFB_XMM %xmm15, \XMM2	# perform a 16 byte swap
1012	PSHUFB_XMM %xmm15, \XMM3	# perform a 16 byte swap
1013	PSHUFB_XMM %xmm15, \XMM4	# perform a 16 byte swap
1014
1015	pxor	  (%arg1), \XMM1
1016	pxor	  (%arg1), \XMM2
1017	pxor	  (%arg1), \XMM3
1018	pxor	  (%arg1), \XMM4
1019	movdqa	  HashKey_4_k(%arg2), \TMP5
1020	PCLMULQDQ 0x00, \TMP5, \TMP6           # TMP6 = (a1+a0)*(b1+b0)
1021	movaps 0x10(%arg1), \TMP1
1022	AESENC	  \TMP1, \XMM1              # Round 1
1023	AESENC	  \TMP1, \XMM2
1024	AESENC	  \TMP1, \XMM3
1025	AESENC	  \TMP1, \XMM4
1026	movaps 0x20(%arg1), \TMP1
1027	AESENC	  \TMP1, \XMM1              # Round 2
1028	AESENC	  \TMP1, \XMM2
1029	AESENC	  \TMP1, \XMM3
1030	AESENC	  \TMP1, \XMM4
1031	movdqa	  \XMM6, \TMP1
1032	pshufd	  $78, \XMM6, \TMP2
1033	pxor	  \XMM6, \TMP2
1034	movdqa	  HashKey_3(%arg2), \TMP5
1035	PCLMULQDQ 0x11, \TMP5, \TMP1           # TMP1 = a1 * b1
1036	movaps 0x30(%arg1), \TMP3
1037	AESENC    \TMP3, \XMM1              # Round 3
1038	AESENC    \TMP3, \XMM2
1039	AESENC    \TMP3, \XMM3
1040	AESENC    \TMP3, \XMM4
1041	PCLMULQDQ 0x00, \TMP5, \XMM6           # XMM6 = a0*b0
1042	movaps 0x40(%arg1), \TMP3
1043	AESENC	  \TMP3, \XMM1              # Round 4
1044	AESENC	  \TMP3, \XMM2
1045	AESENC	  \TMP3, \XMM3
1046	AESENC	  \TMP3, \XMM4
1047	movdqa	  HashKey_3_k(%arg2), \TMP5
1048	PCLMULQDQ 0x00, \TMP5, \TMP2           # TMP2 = (a1+a0)*(b1+b0)
1049	movaps 0x50(%arg1), \TMP3
1050	AESENC	  \TMP3, \XMM1              # Round 5
1051	AESENC	  \TMP3, \XMM2
1052	AESENC	  \TMP3, \XMM3
1053	AESENC	  \TMP3, \XMM4
1054	pxor	  \TMP1, \TMP4
1055# accumulate the results in TMP4:XMM5, TMP6 holds the middle part
1056	pxor	  \XMM6, \XMM5
1057	pxor	  \TMP2, \TMP6
1058	movdqa	  \XMM7, \TMP1
1059	pshufd	  $78, \XMM7, \TMP2
1060	pxor	  \XMM7, \TMP2
1061	movdqa	  HashKey_2(%arg2), \TMP5
1062
1063        # Multiply TMP5 * HashKey using karatsuba
1064
1065	PCLMULQDQ 0x11, \TMP5, \TMP1           # TMP1 = a1*b1
1066	movaps 0x60(%arg1), \TMP3
1067	AESENC	  \TMP3, \XMM1              # Round 6
1068	AESENC	  \TMP3, \XMM2
1069	AESENC	  \TMP3, \XMM3
1070	AESENC	  \TMP3, \XMM4
1071	PCLMULQDQ 0x00, \TMP5, \XMM7           # XMM7 = a0*b0
1072	movaps 0x70(%arg1), \TMP3
1073	AESENC	  \TMP3, \XMM1             # Round 7
1074	AESENC	  \TMP3, \XMM2
1075	AESENC	  \TMP3, \XMM3
1076	AESENC	  \TMP3, \XMM4
1077	movdqa	  HashKey_2_k(%arg2), \TMP5
1078	PCLMULQDQ 0x00, \TMP5, \TMP2           # TMP2 = (a1+a0)*(b1+b0)
1079	movaps 0x80(%arg1), \TMP3
1080	AESENC	  \TMP3, \XMM1             # Round 8
1081	AESENC	  \TMP3, \XMM2
1082	AESENC	  \TMP3, \XMM3
1083	AESENC	  \TMP3, \XMM4
1084	pxor	  \TMP1, \TMP4
1085# accumulate the results in TMP4:XMM5, TMP6 holds the middle part
1086	pxor	  \XMM7, \XMM5
1087	pxor	  \TMP2, \TMP6
1088
1089        # Multiply XMM8 * HashKey
1090        # XMM8 and TMP5 hold the values for the two operands
1091
1092	movdqa	  \XMM8, \TMP1
1093	pshufd	  $78, \XMM8, \TMP2
1094	pxor	  \XMM8, \TMP2
1095	movdqa	  HashKey(%arg2), \TMP5
1096	PCLMULQDQ 0x11, \TMP5, \TMP1          # TMP1 = a1*b1
1097	movaps 0x90(%arg1), \TMP3
1098	AESENC	  \TMP3, \XMM1            # Round 9
1099	AESENC	  \TMP3, \XMM2
1100	AESENC	  \TMP3, \XMM3
1101	AESENC	  \TMP3, \XMM4
1102	PCLMULQDQ 0x00, \TMP5, \XMM8          # XMM8 = a0*b0
1103	lea	  0xa0(%arg1),%r10
1104	mov	  keysize,%eax
1105	shr	  $2,%eax			# 128->4, 192->6, 256->8
1106	sub	  $4,%eax			# 128->0, 192->2, 256->4
1107	jz	  aes_loop_par_enc_done\@
1108
1109aes_loop_par_enc\@:
1110	MOVADQ	  (%r10),\TMP3
1111.irpc	index, 1234
1112	AESENC	  \TMP3, %xmm\index
1113.endr
1114	add	  $16,%r10
1115	sub	  $1,%eax
1116	jnz	  aes_loop_par_enc\@
1117
1118aes_loop_par_enc_done\@:
1119	MOVADQ	  (%r10), \TMP3
1120	AESENCLAST \TMP3, \XMM1           # Round 10
1121	AESENCLAST \TMP3, \XMM2
1122	AESENCLAST \TMP3, \XMM3
1123	AESENCLAST \TMP3, \XMM4
1124	movdqa    HashKey_k(%arg2), \TMP5
1125	PCLMULQDQ 0x00, \TMP5, \TMP2          # TMP2 = (a1+a0)*(b1+b0)
1126	movdqu	  (%arg4,%r11,1), \TMP3
1127	pxor	  \TMP3, \XMM1                 # Ciphertext/Plaintext XOR EK
1128	movdqu	  16(%arg4,%r11,1), \TMP3
1129	pxor	  \TMP3, \XMM2                 # Ciphertext/Plaintext XOR EK
1130	movdqu	  32(%arg4,%r11,1), \TMP3
1131	pxor	  \TMP3, \XMM3                 # Ciphertext/Plaintext XOR EK
1132	movdqu	  48(%arg4,%r11,1), \TMP3
1133	pxor	  \TMP3, \XMM4                 # Ciphertext/Plaintext XOR EK
1134        movdqu    \XMM1, (%arg3,%r11,1)        # Write to the ciphertext buffer
1135        movdqu    \XMM2, 16(%arg3,%r11,1)      # Write to the ciphertext buffer
1136        movdqu    \XMM3, 32(%arg3,%r11,1)      # Write to the ciphertext buffer
1137        movdqu    \XMM4, 48(%arg3,%r11,1)      # Write to the ciphertext buffer
1138	PSHUFB_XMM %xmm15, \XMM1        # perform a 16 byte swap
1139	PSHUFB_XMM %xmm15, \XMM2	# perform a 16 byte swap
1140	PSHUFB_XMM %xmm15, \XMM3	# perform a 16 byte swap
1141	PSHUFB_XMM %xmm15, \XMM4	# perform a 16 byte swap
1142
1143	pxor	  \TMP4, \TMP1
1144	pxor	  \XMM8, \XMM5
1145	pxor	  \TMP6, \TMP2
1146	pxor	  \TMP1, \TMP2
1147	pxor	  \XMM5, \TMP2
1148	movdqa	  \TMP2, \TMP3
1149	pslldq	  $8, \TMP3                    # left shift TMP3 2 DWs
1150	psrldq	  $8, \TMP2                    # right shift TMP2 2 DWs
1151	pxor	  \TMP3, \XMM5
1152	pxor	  \TMP2, \TMP1	  # accumulate the results in TMP1:XMM5
1153
1154        # first phase of reduction
1155
1156	movdqa    \XMM5, \TMP2
1157	movdqa    \XMM5, \TMP3
1158	movdqa    \XMM5, \TMP4
1159# move XMM5 into TMP2, TMP3, TMP4 in order to perform shifts independently
1160	pslld     $31, \TMP2                   # packed right shift << 31
1161	pslld     $30, \TMP3                   # packed right shift << 30
1162	pslld     $25, \TMP4                   # packed right shift << 25
1163	pxor      \TMP3, \TMP2	               # xor the shifted versions
1164	pxor      \TMP4, \TMP2
1165	movdqa    \TMP2, \TMP5
1166	psrldq    $4, \TMP5                    # right shift T5 1 DW
1167	pslldq    $12, \TMP2                   # left shift T2 3 DWs
1168	pxor      \TMP2, \XMM5
1169
1170        # second phase of reduction
1171
1172	movdqa    \XMM5,\TMP2 # make 3 copies of XMM5 into TMP2, TMP3, TMP4
1173	movdqa    \XMM5,\TMP3
1174	movdqa    \XMM5,\TMP4
1175	psrld     $1, \TMP2                    # packed left shift >>1
1176	psrld     $2, \TMP3                    # packed left shift >>2
1177	psrld     $7, \TMP4                    # packed left shift >>7
1178	pxor      \TMP3,\TMP2		       # xor the shifted versions
1179	pxor      \TMP4,\TMP2
1180	pxor      \TMP5, \TMP2
1181	pxor      \TMP2, \XMM5
1182	pxor      \TMP1, \XMM5                 # result is in TMP1
1183
1184	pxor	  \XMM5, \XMM1
1185.endm
1186
1187/*
1188* decrypt 4 blocks at a time
1189* ghash the 4 previously decrypted ciphertext blocks
1190* arg1, %arg3, %arg4 are used as pointers only, not modified
1191* %r11 is the data offset value
1192*/
1193.macro GHASH_4_ENCRYPT_4_PARALLEL_DEC TMP1 TMP2 TMP3 TMP4 TMP5 \
1194TMP6 XMM0 XMM1 XMM2 XMM3 XMM4 XMM5 XMM6 XMM7 XMM8 operation
1195
1196	movdqa	  \XMM1, \XMM5
1197	movdqa	  \XMM2, \XMM6
1198	movdqa	  \XMM3, \XMM7
1199	movdqa	  \XMM4, \XMM8
1200
1201        movdqa    SHUF_MASK(%rip), %xmm15
1202        # multiply TMP5 * HashKey using karatsuba
1203
1204	movdqa	  \XMM5, \TMP4
1205	pshufd	  $78, \XMM5, \TMP6
1206	pxor	  \XMM5, \TMP6
1207	paddd     ONE(%rip), \XMM0		# INCR CNT
1208	movdqa	  HashKey_4(%arg2), \TMP5
1209	PCLMULQDQ 0x11, \TMP5, \TMP4           # TMP4 = a1*b1
1210	movdqa    \XMM0, \XMM1
1211	paddd     ONE(%rip), \XMM0		# INCR CNT
1212	movdqa    \XMM0, \XMM2
1213	paddd     ONE(%rip), \XMM0		# INCR CNT
1214	movdqa    \XMM0, \XMM3
1215	paddd     ONE(%rip), \XMM0		# INCR CNT
1216	movdqa    \XMM0, \XMM4
1217	PSHUFB_XMM %xmm15, \XMM1	# perform a 16 byte swap
1218	PCLMULQDQ 0x00, \TMP5, \XMM5           # XMM5 = a0*b0
1219	PSHUFB_XMM %xmm15, \XMM2	# perform a 16 byte swap
1220	PSHUFB_XMM %xmm15, \XMM3	# perform a 16 byte swap
1221	PSHUFB_XMM %xmm15, \XMM4	# perform a 16 byte swap
1222
1223	pxor	  (%arg1), \XMM1
1224	pxor	  (%arg1), \XMM2
1225	pxor	  (%arg1), \XMM3
1226	pxor	  (%arg1), \XMM4
1227	movdqa	  HashKey_4_k(%arg2), \TMP5
1228	PCLMULQDQ 0x00, \TMP5, \TMP6           # TMP6 = (a1+a0)*(b1+b0)
1229	movaps 0x10(%arg1), \TMP1
1230	AESENC	  \TMP1, \XMM1              # Round 1
1231	AESENC	  \TMP1, \XMM2
1232	AESENC	  \TMP1, \XMM3
1233	AESENC	  \TMP1, \XMM4
1234	movaps 0x20(%arg1), \TMP1
1235	AESENC	  \TMP1, \XMM1              # Round 2
1236	AESENC	  \TMP1, \XMM2
1237	AESENC	  \TMP1, \XMM3
1238	AESENC	  \TMP1, \XMM4
1239	movdqa	  \XMM6, \TMP1
1240	pshufd	  $78, \XMM6, \TMP2
1241	pxor	  \XMM6, \TMP2
1242	movdqa	  HashKey_3(%arg2), \TMP5
1243	PCLMULQDQ 0x11, \TMP5, \TMP1           # TMP1 = a1 * b1
1244	movaps 0x30(%arg1), \TMP3
1245	AESENC    \TMP3, \XMM1              # Round 3
1246	AESENC    \TMP3, \XMM2
1247	AESENC    \TMP3, \XMM3
1248	AESENC    \TMP3, \XMM4
1249	PCLMULQDQ 0x00, \TMP5, \XMM6           # XMM6 = a0*b0
1250	movaps 0x40(%arg1), \TMP3
1251	AESENC	  \TMP3, \XMM1              # Round 4
1252	AESENC	  \TMP3, \XMM2
1253	AESENC	  \TMP3, \XMM3
1254	AESENC	  \TMP3, \XMM4
1255	movdqa	  HashKey_3_k(%arg2), \TMP5
1256	PCLMULQDQ 0x00, \TMP5, \TMP2           # TMP2 = (a1+a0)*(b1+b0)
1257	movaps 0x50(%arg1), \TMP3
1258	AESENC	  \TMP3, \XMM1              # Round 5
1259	AESENC	  \TMP3, \XMM2
1260	AESENC	  \TMP3, \XMM3
1261	AESENC	  \TMP3, \XMM4
1262	pxor	  \TMP1, \TMP4
1263# accumulate the results in TMP4:XMM5, TMP6 holds the middle part
1264	pxor	  \XMM6, \XMM5
1265	pxor	  \TMP2, \TMP6
1266	movdqa	  \XMM7, \TMP1
1267	pshufd	  $78, \XMM7, \TMP2
1268	pxor	  \XMM7, \TMP2
1269	movdqa	  HashKey_2(%arg2), \TMP5
1270
1271        # Multiply TMP5 * HashKey using karatsuba
1272
1273	PCLMULQDQ 0x11, \TMP5, \TMP1           # TMP1 = a1*b1
1274	movaps 0x60(%arg1), \TMP3
1275	AESENC	  \TMP3, \XMM1              # Round 6
1276	AESENC	  \TMP3, \XMM2
1277	AESENC	  \TMP3, \XMM3
1278	AESENC	  \TMP3, \XMM4
1279	PCLMULQDQ 0x00, \TMP5, \XMM7           # XMM7 = a0*b0
1280	movaps 0x70(%arg1), \TMP3
1281	AESENC	  \TMP3, \XMM1             # Round 7
1282	AESENC	  \TMP3, \XMM2
1283	AESENC	  \TMP3, \XMM3
1284	AESENC	  \TMP3, \XMM4
1285	movdqa	  HashKey_2_k(%arg2), \TMP5
1286	PCLMULQDQ 0x00, \TMP5, \TMP2           # TMP2 = (a1+a0)*(b1+b0)
1287	movaps 0x80(%arg1), \TMP3
1288	AESENC	  \TMP3, \XMM1             # Round 8
1289	AESENC	  \TMP3, \XMM2
1290	AESENC	  \TMP3, \XMM3
1291	AESENC	  \TMP3, \XMM4
1292	pxor	  \TMP1, \TMP4
1293# accumulate the results in TMP4:XMM5, TMP6 holds the middle part
1294	pxor	  \XMM7, \XMM5
1295	pxor	  \TMP2, \TMP6
1296
1297        # Multiply XMM8 * HashKey
1298        # XMM8 and TMP5 hold the values for the two operands
1299
1300	movdqa	  \XMM8, \TMP1
1301	pshufd	  $78, \XMM8, \TMP2
1302	pxor	  \XMM8, \TMP2
1303	movdqa	  HashKey(%arg2), \TMP5
1304	PCLMULQDQ 0x11, \TMP5, \TMP1          # TMP1 = a1*b1
1305	movaps 0x90(%arg1), \TMP3
1306	AESENC	  \TMP3, \XMM1            # Round 9
1307	AESENC	  \TMP3, \XMM2
1308	AESENC	  \TMP3, \XMM3
1309	AESENC	  \TMP3, \XMM4
1310	PCLMULQDQ 0x00, \TMP5, \XMM8          # XMM8 = a0*b0
1311	lea	  0xa0(%arg1),%r10
1312	mov	  keysize,%eax
1313	shr	  $2,%eax		        # 128->4, 192->6, 256->8
1314	sub	  $4,%eax			# 128->0, 192->2, 256->4
1315	jz	  aes_loop_par_dec_done\@
1316
1317aes_loop_par_dec\@:
1318	MOVADQ	  (%r10),\TMP3
1319.irpc	index, 1234
1320	AESENC	  \TMP3, %xmm\index
1321.endr
1322	add	  $16,%r10
1323	sub	  $1,%eax
1324	jnz	  aes_loop_par_dec\@
1325
1326aes_loop_par_dec_done\@:
1327	MOVADQ	  (%r10), \TMP3
1328	AESENCLAST \TMP3, \XMM1           # last round
1329	AESENCLAST \TMP3, \XMM2
1330	AESENCLAST \TMP3, \XMM3
1331	AESENCLAST \TMP3, \XMM4
1332	movdqa    HashKey_k(%arg2), \TMP5
1333	PCLMULQDQ 0x00, \TMP5, \TMP2          # TMP2 = (a1+a0)*(b1+b0)
1334	movdqu	  (%arg4,%r11,1), \TMP3
1335	pxor	  \TMP3, \XMM1                 # Ciphertext/Plaintext XOR EK
1336	movdqu	  \XMM1, (%arg3,%r11,1)        # Write to plaintext buffer
1337	movdqa    \TMP3, \XMM1
1338	movdqu	  16(%arg4,%r11,1), \TMP3
1339	pxor	  \TMP3, \XMM2                 # Ciphertext/Plaintext XOR EK
1340	movdqu	  \XMM2, 16(%arg3,%r11,1)      # Write to plaintext buffer
1341	movdqa    \TMP3, \XMM2
1342	movdqu	  32(%arg4,%r11,1), \TMP3
1343	pxor	  \TMP3, \XMM3                 # Ciphertext/Plaintext XOR EK
1344	movdqu	  \XMM3, 32(%arg3,%r11,1)      # Write to plaintext buffer
1345	movdqa    \TMP3, \XMM3
1346	movdqu	  48(%arg4,%r11,1), \TMP3
1347	pxor	  \TMP3, \XMM4                 # Ciphertext/Plaintext XOR EK
1348	movdqu	  \XMM4, 48(%arg3,%r11,1)      # Write to plaintext buffer
1349	movdqa    \TMP3, \XMM4
1350	PSHUFB_XMM %xmm15, \XMM1        # perform a 16 byte swap
1351	PSHUFB_XMM %xmm15, \XMM2	# perform a 16 byte swap
1352	PSHUFB_XMM %xmm15, \XMM3	# perform a 16 byte swap
1353	PSHUFB_XMM %xmm15, \XMM4	# perform a 16 byte swap
1354
1355	pxor	  \TMP4, \TMP1
1356	pxor	  \XMM8, \XMM5
1357	pxor	  \TMP6, \TMP2
1358	pxor	  \TMP1, \TMP2
1359	pxor	  \XMM5, \TMP2
1360	movdqa	  \TMP2, \TMP3
1361	pslldq	  $8, \TMP3                    # left shift TMP3 2 DWs
1362	psrldq	  $8, \TMP2                    # right shift TMP2 2 DWs
1363	pxor	  \TMP3, \XMM5
1364	pxor	  \TMP2, \TMP1	  # accumulate the results in TMP1:XMM5
1365
1366        # first phase of reduction
1367
1368	movdqa    \XMM5, \TMP2
1369	movdqa    \XMM5, \TMP3
1370	movdqa    \XMM5, \TMP4
1371# move XMM5 into TMP2, TMP3, TMP4 in order to perform shifts independently
1372	pslld     $31, \TMP2                   # packed right shift << 31
1373	pslld     $30, \TMP3                   # packed right shift << 30
1374	pslld     $25, \TMP4                   # packed right shift << 25
1375	pxor      \TMP3, \TMP2	               # xor the shifted versions
1376	pxor      \TMP4, \TMP2
1377	movdqa    \TMP2, \TMP5
1378	psrldq    $4, \TMP5                    # right shift T5 1 DW
1379	pslldq    $12, \TMP2                   # left shift T2 3 DWs
1380	pxor      \TMP2, \XMM5
1381
1382        # second phase of reduction
1383
1384	movdqa    \XMM5,\TMP2 # make 3 copies of XMM5 into TMP2, TMP3, TMP4
1385	movdqa    \XMM5,\TMP3
1386	movdqa    \XMM5,\TMP4
1387	psrld     $1, \TMP2                    # packed left shift >>1
1388	psrld     $2, \TMP3                    # packed left shift >>2
1389	psrld     $7, \TMP4                    # packed left shift >>7
1390	pxor      \TMP3,\TMP2		       # xor the shifted versions
1391	pxor      \TMP4,\TMP2
1392	pxor      \TMP5, \TMP2
1393	pxor      \TMP2, \XMM5
1394	pxor      \TMP1, \XMM5                 # result is in TMP1
1395
1396	pxor	  \XMM5, \XMM1
1397.endm
1398
1399/* GHASH the last 4 ciphertext blocks. */
1400.macro	GHASH_LAST_4 TMP1 TMP2 TMP3 TMP4 TMP5 TMP6 \
1401TMP7 XMM1 XMM2 XMM3 XMM4 XMMDst
1402
1403        # Multiply TMP6 * HashKey (using Karatsuba)
1404
1405	movdqa	  \XMM1, \TMP6
1406	pshufd	  $78, \XMM1, \TMP2
1407	pxor	  \XMM1, \TMP2
1408	movdqa	  HashKey_4(%arg2), \TMP5
1409	PCLMULQDQ 0x11, \TMP5, \TMP6       # TMP6 = a1*b1
1410	PCLMULQDQ 0x00, \TMP5, \XMM1       # XMM1 = a0*b0
1411	movdqa	  HashKey_4_k(%arg2), \TMP4
1412	PCLMULQDQ 0x00, \TMP4, \TMP2       # TMP2 = (a1+a0)*(b1+b0)
1413	movdqa	  \XMM1, \XMMDst
1414	movdqa	  \TMP2, \XMM1              # result in TMP6, XMMDst, XMM1
1415
1416        # Multiply TMP1 * HashKey (using Karatsuba)
1417
1418	movdqa	  \XMM2, \TMP1
1419	pshufd	  $78, \XMM2, \TMP2
1420	pxor	  \XMM2, \TMP2
1421	movdqa	  HashKey_3(%arg2), \TMP5
1422	PCLMULQDQ 0x11, \TMP5, \TMP1       # TMP1 = a1*b1
1423	PCLMULQDQ 0x00, \TMP5, \XMM2       # XMM2 = a0*b0
1424	movdqa	  HashKey_3_k(%arg2), \TMP4
1425	PCLMULQDQ 0x00, \TMP4, \TMP2       # TMP2 = (a1+a0)*(b1+b0)
1426	pxor	  \TMP1, \TMP6
1427	pxor	  \XMM2, \XMMDst
1428	pxor	  \TMP2, \XMM1
1429# results accumulated in TMP6, XMMDst, XMM1
1430
1431        # Multiply TMP1 * HashKey (using Karatsuba)
1432
1433	movdqa	  \XMM3, \TMP1
1434	pshufd	  $78, \XMM3, \TMP2
1435	pxor	  \XMM3, \TMP2
1436	movdqa	  HashKey_2(%arg2), \TMP5
1437	PCLMULQDQ 0x11, \TMP5, \TMP1       # TMP1 = a1*b1
1438	PCLMULQDQ 0x00, \TMP5, \XMM3       # XMM3 = a0*b0
1439	movdqa	  HashKey_2_k(%arg2), \TMP4
1440	PCLMULQDQ 0x00, \TMP4, \TMP2       # TMP2 = (a1+a0)*(b1+b0)
1441	pxor	  \TMP1, \TMP6
1442	pxor	  \XMM3, \XMMDst
1443	pxor	  \TMP2, \XMM1   # results accumulated in TMP6, XMMDst, XMM1
1444
1445        # Multiply TMP1 * HashKey (using Karatsuba)
1446	movdqa	  \XMM4, \TMP1
1447	pshufd	  $78, \XMM4, \TMP2
1448	pxor	  \XMM4, \TMP2
1449	movdqa	  HashKey(%arg2), \TMP5
1450	PCLMULQDQ 0x11, \TMP5, \TMP1	    # TMP1 = a1*b1
1451	PCLMULQDQ 0x00, \TMP5, \XMM4       # XMM4 = a0*b0
1452	movdqa	  HashKey_k(%arg2), \TMP4
1453	PCLMULQDQ 0x00, \TMP4, \TMP2       # TMP2 = (a1+a0)*(b1+b0)
1454	pxor	  \TMP1, \TMP6
1455	pxor	  \XMM4, \XMMDst
1456	pxor	  \XMM1, \TMP2
1457	pxor	  \TMP6, \TMP2
1458	pxor	  \XMMDst, \TMP2
1459	# middle section of the temp results combined as in karatsuba algorithm
1460	movdqa	  \TMP2, \TMP4
1461	pslldq	  $8, \TMP4                 # left shift TMP4 2 DWs
1462	psrldq	  $8, \TMP2                 # right shift TMP2 2 DWs
1463	pxor	  \TMP4, \XMMDst
1464	pxor	  \TMP2, \TMP6
1465# TMP6:XMMDst holds the result of the accumulated carry-less multiplications
1466	# first phase of the reduction
1467	movdqa    \XMMDst, \TMP2
1468	movdqa    \XMMDst, \TMP3
1469	movdqa    \XMMDst, \TMP4
1470# move XMMDst into TMP2, TMP3, TMP4 in order to perform 3 shifts independently
1471	pslld     $31, \TMP2                # packed right shifting << 31
1472	pslld     $30, \TMP3                # packed right shifting << 30
1473	pslld     $25, \TMP4                # packed right shifting << 25
1474	pxor      \TMP3, \TMP2              # xor the shifted versions
1475	pxor      \TMP4, \TMP2
1476	movdqa    \TMP2, \TMP7
1477	psrldq    $4, \TMP7                 # right shift TMP7 1 DW
1478	pslldq    $12, \TMP2                # left shift TMP2 3 DWs
1479	pxor      \TMP2, \XMMDst
1480
1481        # second phase of the reduction
1482	movdqa    \XMMDst, \TMP2
1483	# make 3 copies of XMMDst for doing 3 shift operations
1484	movdqa    \XMMDst, \TMP3
1485	movdqa    \XMMDst, \TMP4
1486	psrld     $1, \TMP2                 # packed left shift >> 1
1487	psrld     $2, \TMP3                 # packed left shift >> 2
1488	psrld     $7, \TMP4                 # packed left shift >> 7
1489	pxor      \TMP3, \TMP2              # xor the shifted versions
1490	pxor      \TMP4, \TMP2
1491	pxor      \TMP7, \TMP2
1492	pxor      \TMP2, \XMMDst
1493	pxor      \TMP6, \XMMDst            # reduced result is in XMMDst
1494.endm
1495
1496
1497/* Encryption of a single block
1498* uses eax & r10
1499*/
1500
1501.macro ENCRYPT_SINGLE_BLOCK XMM0 TMP1
1502
1503	pxor		(%arg1), \XMM0
1504	mov		keysize,%eax
1505	shr		$2,%eax			# 128->4, 192->6, 256->8
1506	add		$5,%eax			# 128->9, 192->11, 256->13
1507	lea		16(%arg1), %r10	  # get first expanded key address
1508
1509_esb_loop_\@:
1510	MOVADQ		(%r10),\TMP1
1511	AESENC		\TMP1,\XMM0
1512	add		$16,%r10
1513	sub		$1,%eax
1514	jnz		_esb_loop_\@
1515
1516	MOVADQ		(%r10),\TMP1
1517	AESENCLAST	\TMP1,\XMM0
1518.endm
1519/*****************************************************************************
1520* void aesni_gcm_dec(void *aes_ctx,    // AES Key schedule. Starts on a 16 byte boundary.
1521*                   struct gcm_context_data *data
1522*                                      // Context data
1523*                   u8 *out,           // Plaintext output. Encrypt in-place is allowed.
1524*                   const u8 *in,      // Ciphertext input
1525*                   u64 plaintext_len, // Length of data in bytes for decryption.
1526*                   u8 *iv,            // Pre-counter block j0: 4 byte salt (from Security Association)
1527*                                      // concatenated with 8 byte Initialisation Vector (from IPSec ESP Payload)
1528*                                      // concatenated with 0x00000001. 16-byte aligned pointer.
1529*                   u8 *hash_subkey,   // H, the Hash sub key input. Data starts on a 16-byte boundary.
1530*                   const u8 *aad,     // Additional Authentication Data (AAD)
1531*                   u64 aad_len,       // Length of AAD in bytes. With RFC4106 this is going to be 8 or 12 bytes
1532*                   u8  *auth_tag,     // Authenticated Tag output. The driver will compare this to the
1533*                                      // given authentication tag and only return the plaintext if they match.
1534*                   u64 auth_tag_len); // Authenticated Tag Length in bytes. Valid values are 16
1535*                                      // (most likely), 12 or 8.
1536*
1537* Assumptions:
1538*
1539* keys:
1540*       keys are pre-expanded and aligned to 16 bytes. we are using the first
1541*       set of 11 keys in the data structure void *aes_ctx
1542*
1543* iv:
1544*       0                   1                   2                   3
1545*       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
1546*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1547*       |                             Salt  (From the SA)               |
1548*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1549*       |                     Initialization Vector                     |
1550*       |         (This is the sequence number from IPSec header)       |
1551*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1552*       |                              0x1                              |
1553*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1554*
1555*
1556*
1557* AAD:
1558*       AAD padded to 128 bits with 0
1559*       for example, assume AAD is a u32 vector
1560*
1561*       if AAD is 8 bytes:
1562*       AAD[3] = {A0, A1};
1563*       padded AAD in xmm register = {A1 A0 0 0}
1564*
1565*       0                   1                   2                   3
1566*       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
1567*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1568*       |                               SPI (A1)                        |
1569*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1570*       |                     32-bit Sequence Number (A0)               |
1571*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1572*       |                              0x0                              |
1573*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1574*
1575*                                       AAD Format with 32-bit Sequence Number
1576*
1577*       if AAD is 12 bytes:
1578*       AAD[3] = {A0, A1, A2};
1579*       padded AAD in xmm register = {A2 A1 A0 0}
1580*
1581*       0                   1                   2                   3
1582*       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
1583*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1584*       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
1585*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1586*       |                               SPI (A2)                        |
1587*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1588*       |                 64-bit Extended Sequence Number {A1,A0}       |
1589*       |                                                               |
1590*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1591*       |                              0x0                              |
1592*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1593*
1594*                        AAD Format with 64-bit Extended Sequence Number
1595*
1596* poly = x^128 + x^127 + x^126 + x^121 + 1
1597*
1598*****************************************************************************/
1599ENTRY(aesni_gcm_dec)
1600	FUNC_SAVE
1601
1602	GCM_INIT %arg6, arg7, arg8, arg9
1603	GCM_ENC_DEC dec
1604	GCM_COMPLETE arg10, arg11
1605	FUNC_RESTORE
1606	ret
1607ENDPROC(aesni_gcm_dec)
1608
1609
1610/*****************************************************************************
1611* void aesni_gcm_enc(void *aes_ctx,      // AES Key schedule. Starts on a 16 byte boundary.
1612*                    struct gcm_context_data *data
1613*                                        // Context data
1614*                    u8 *out,            // Ciphertext output. Encrypt in-place is allowed.
1615*                    const u8 *in,       // Plaintext input
1616*                    u64 plaintext_len,  // Length of data in bytes for encryption.
1617*                    u8 *iv,             // Pre-counter block j0: 4 byte salt (from Security Association)
1618*                                        // concatenated with 8 byte Initialisation Vector (from IPSec ESP Payload)
1619*                                        // concatenated with 0x00000001. 16-byte aligned pointer.
1620*                    u8 *hash_subkey,    // H, the Hash sub key input. Data starts on a 16-byte boundary.
1621*                    const u8 *aad,      // Additional Authentication Data (AAD)
1622*                    u64 aad_len,        // Length of AAD in bytes. With RFC4106 this is going to be 8 or 12 bytes
1623*                    u8 *auth_tag,       // Authenticated Tag output.
1624*                    u64 auth_tag_len);  // Authenticated Tag Length in bytes. Valid values are 16 (most likely),
1625*                                        // 12 or 8.
1626*
1627* Assumptions:
1628*
1629* keys:
1630*       keys are pre-expanded and aligned to 16 bytes. we are using the
1631*       first set of 11 keys in the data structure void *aes_ctx
1632*
1633*
1634* iv:
1635*       0                   1                   2                   3
1636*       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
1637*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1638*       |                             Salt  (From the SA)               |
1639*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1640*       |                     Initialization Vector                     |
1641*       |         (This is the sequence number from IPSec header)       |
1642*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1643*       |                              0x1                              |
1644*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1645*
1646*
1647*
1648* AAD:
1649*       AAD padded to 128 bits with 0
1650*       for example, assume AAD is a u32 vector
1651*
1652*       if AAD is 8 bytes:
1653*       AAD[3] = {A0, A1};
1654*       padded AAD in xmm register = {A1 A0 0 0}
1655*
1656*       0                   1                   2                   3
1657*       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
1658*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1659*       |                               SPI (A1)                        |
1660*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1661*       |                     32-bit Sequence Number (A0)               |
1662*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1663*       |                              0x0                              |
1664*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1665*
1666*                                 AAD Format with 32-bit Sequence Number
1667*
1668*       if AAD is 12 bytes:
1669*       AAD[3] = {A0, A1, A2};
1670*       padded AAD in xmm register = {A2 A1 A0 0}
1671*
1672*       0                   1                   2                   3
1673*       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
1674*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1675*       |                               SPI (A2)                        |
1676*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1677*       |                 64-bit Extended Sequence Number {A1,A0}       |
1678*       |                                                               |
1679*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1680*       |                              0x0                              |
1681*       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1682*
1683*                         AAD Format with 64-bit Extended Sequence Number
1684*
1685* poly = x^128 + x^127 + x^126 + x^121 + 1
1686***************************************************************************/
1687ENTRY(aesni_gcm_enc)
1688	FUNC_SAVE
1689
1690	GCM_INIT %arg6, arg7, arg8, arg9
1691	GCM_ENC_DEC enc
1692
1693	GCM_COMPLETE arg10, arg11
1694	FUNC_RESTORE
1695	ret
1696ENDPROC(aesni_gcm_enc)
1697
1698/*****************************************************************************
1699* void aesni_gcm_init(void *aes_ctx,      // AES Key schedule. Starts on a 16 byte boundary.
1700*                     struct gcm_context_data *data,
1701*                                         // context data
1702*                     u8 *iv,             // Pre-counter block j0: 4 byte salt (from Security Association)
1703*                                         // concatenated with 8 byte Initialisation Vector (from IPSec ESP Payload)
1704*                                         // concatenated with 0x00000001. 16-byte aligned pointer.
1705*                     u8 *hash_subkey,    // H, the Hash sub key input. Data starts on a 16-byte boundary.
1706*                     const u8 *aad,      // Additional Authentication Data (AAD)
1707*                     u64 aad_len)        // Length of AAD in bytes.
1708*/
1709ENTRY(aesni_gcm_init)
1710	FUNC_SAVE
1711	GCM_INIT %arg3, %arg4,%arg5, %arg6
1712	FUNC_RESTORE
1713	ret
1714ENDPROC(aesni_gcm_init)
1715
1716/*****************************************************************************
1717* void aesni_gcm_enc_update(void *aes_ctx,      // AES Key schedule. Starts on a 16 byte boundary.
1718*                    struct gcm_context_data *data,
1719*                                        // context data
1720*                    u8 *out,            // Ciphertext output. Encrypt in-place is allowed.
1721*                    const u8 *in,       // Plaintext input
1722*                    u64 plaintext_len,  // Length of data in bytes for encryption.
1723*/
1724ENTRY(aesni_gcm_enc_update)
1725	FUNC_SAVE
1726	GCM_ENC_DEC enc
1727	FUNC_RESTORE
1728	ret
1729ENDPROC(aesni_gcm_enc_update)
1730
1731/*****************************************************************************
1732* void aesni_gcm_dec_update(void *aes_ctx,      // AES Key schedule. Starts on a 16 byte boundary.
1733*                    struct gcm_context_data *data,
1734*                                        // context data
1735*                    u8 *out,            // Ciphertext output. Encrypt in-place is allowed.
1736*                    const u8 *in,       // Plaintext input
1737*                    u64 plaintext_len,  // Length of data in bytes for encryption.
1738*/
1739ENTRY(aesni_gcm_dec_update)
1740	FUNC_SAVE
1741	GCM_ENC_DEC dec
1742	FUNC_RESTORE
1743	ret
1744ENDPROC(aesni_gcm_dec_update)
1745
1746/*****************************************************************************
1747* void aesni_gcm_finalize(void *aes_ctx,      // AES Key schedule. Starts on a 16 byte boundary.
1748*                    struct gcm_context_data *data,
1749*                                        // context data
1750*                    u8 *auth_tag,       // Authenticated Tag output.
1751*                    u64 auth_tag_len);  // Authenticated Tag Length in bytes. Valid values are 16 (most likely),
1752*                                        // 12 or 8.
1753*/
1754ENTRY(aesni_gcm_finalize)
1755	FUNC_SAVE
1756	GCM_COMPLETE %arg3 %arg4
1757	FUNC_RESTORE
1758	ret
1759ENDPROC(aesni_gcm_finalize)
1760
1761#endif
1762
1763
1764.align 4
1765_key_expansion_128:
1766_key_expansion_256a:
1767	pshufd $0b11111111, %xmm1, %xmm1
1768	shufps $0b00010000, %xmm0, %xmm4
1769	pxor %xmm4, %xmm0
1770	shufps $0b10001100, %xmm0, %xmm4
1771	pxor %xmm4, %xmm0
1772	pxor %xmm1, %xmm0
1773	movaps %xmm0, (TKEYP)
1774	add $0x10, TKEYP
1775	ret
1776ENDPROC(_key_expansion_128)
1777ENDPROC(_key_expansion_256a)
1778
1779.align 4
1780_key_expansion_192a:
1781	pshufd $0b01010101, %xmm1, %xmm1
1782	shufps $0b00010000, %xmm0, %xmm4
1783	pxor %xmm4, %xmm0
1784	shufps $0b10001100, %xmm0, %xmm4
1785	pxor %xmm4, %xmm0
1786	pxor %xmm1, %xmm0
1787
1788	movaps %xmm2, %xmm5
1789	movaps %xmm2, %xmm6
1790	pslldq $4, %xmm5
1791	pshufd $0b11111111, %xmm0, %xmm3
1792	pxor %xmm3, %xmm2
1793	pxor %xmm5, %xmm2
1794
1795	movaps %xmm0, %xmm1
1796	shufps $0b01000100, %xmm0, %xmm6
1797	movaps %xmm6, (TKEYP)
1798	shufps $0b01001110, %xmm2, %xmm1
1799	movaps %xmm1, 0x10(TKEYP)
1800	add $0x20, TKEYP
1801	ret
1802ENDPROC(_key_expansion_192a)
1803
1804.align 4
1805_key_expansion_192b:
1806	pshufd $0b01010101, %xmm1, %xmm1
1807	shufps $0b00010000, %xmm0, %xmm4
1808	pxor %xmm4, %xmm0
1809	shufps $0b10001100, %xmm0, %xmm4
1810	pxor %xmm4, %xmm0
1811	pxor %xmm1, %xmm0
1812
1813	movaps %xmm2, %xmm5
1814	pslldq $4, %xmm5
1815	pshufd $0b11111111, %xmm0, %xmm3
1816	pxor %xmm3, %xmm2
1817	pxor %xmm5, %xmm2
1818
1819	movaps %xmm0, (TKEYP)
1820	add $0x10, TKEYP
1821	ret
1822ENDPROC(_key_expansion_192b)
1823
1824.align 4
1825_key_expansion_256b:
1826	pshufd $0b10101010, %xmm1, %xmm1
1827	shufps $0b00010000, %xmm2, %xmm4
1828	pxor %xmm4, %xmm2
1829	shufps $0b10001100, %xmm2, %xmm4
1830	pxor %xmm4, %xmm2
1831	pxor %xmm1, %xmm2
1832	movaps %xmm2, (TKEYP)
1833	add $0x10, TKEYP
1834	ret
1835ENDPROC(_key_expansion_256b)
1836
1837/*
1838 * int aesni_set_key(struct crypto_aes_ctx *ctx, const u8 *in_key,
1839 *                   unsigned int key_len)
1840 */
1841ENTRY(aesni_set_key)
1842	FRAME_BEGIN
1843#ifndef __x86_64__
1844	pushl KEYP
1845	movl (FRAME_OFFSET+8)(%esp), KEYP	# ctx
1846	movl (FRAME_OFFSET+12)(%esp), UKEYP	# in_key
1847	movl (FRAME_OFFSET+16)(%esp), %edx	# key_len
1848#endif
1849	movups (UKEYP), %xmm0		# user key (first 16 bytes)
1850	movaps %xmm0, (KEYP)
1851	lea 0x10(KEYP), TKEYP		# key addr
1852	movl %edx, 480(KEYP)
1853	pxor %xmm4, %xmm4		# xmm4 is assumed 0 in _key_expansion_x
1854	cmp $24, %dl
1855	jb .Lenc_key128
1856	je .Lenc_key192
1857	movups 0x10(UKEYP), %xmm2	# other user key
1858	movaps %xmm2, (TKEYP)
1859	add $0x10, TKEYP
1860	AESKEYGENASSIST 0x1 %xmm2 %xmm1		# round 1
1861	call _key_expansion_256a
1862	AESKEYGENASSIST 0x1 %xmm0 %xmm1
1863	call _key_expansion_256b
1864	AESKEYGENASSIST 0x2 %xmm2 %xmm1		# round 2
1865	call _key_expansion_256a
1866	AESKEYGENASSIST 0x2 %xmm0 %xmm1
1867	call _key_expansion_256b
1868	AESKEYGENASSIST 0x4 %xmm2 %xmm1		# round 3
1869	call _key_expansion_256a
1870	AESKEYGENASSIST 0x4 %xmm0 %xmm1
1871	call _key_expansion_256b
1872	AESKEYGENASSIST 0x8 %xmm2 %xmm1		# round 4
1873	call _key_expansion_256a
1874	AESKEYGENASSIST 0x8 %xmm0 %xmm1
1875	call _key_expansion_256b
1876	AESKEYGENASSIST 0x10 %xmm2 %xmm1	# round 5
1877	call _key_expansion_256a
1878	AESKEYGENASSIST 0x10 %xmm0 %xmm1
1879	call _key_expansion_256b
1880	AESKEYGENASSIST 0x20 %xmm2 %xmm1	# round 6
1881	call _key_expansion_256a
1882	AESKEYGENASSIST 0x20 %xmm0 %xmm1
1883	call _key_expansion_256b
1884	AESKEYGENASSIST 0x40 %xmm2 %xmm1	# round 7
1885	call _key_expansion_256a
1886	jmp .Ldec_key
1887.Lenc_key192:
1888	movq 0x10(UKEYP), %xmm2		# other user key
1889	AESKEYGENASSIST 0x1 %xmm2 %xmm1		# round 1
1890	call _key_expansion_192a
1891	AESKEYGENASSIST 0x2 %xmm2 %xmm1		# round 2
1892	call _key_expansion_192b
1893	AESKEYGENASSIST 0x4 %xmm2 %xmm1		# round 3
1894	call _key_expansion_192a
1895	AESKEYGENASSIST 0x8 %xmm2 %xmm1		# round 4
1896	call _key_expansion_192b
1897	AESKEYGENASSIST 0x10 %xmm2 %xmm1	# round 5
1898	call _key_expansion_192a
1899	AESKEYGENASSIST 0x20 %xmm2 %xmm1	# round 6
1900	call _key_expansion_192b
1901	AESKEYGENASSIST 0x40 %xmm2 %xmm1	# round 7
1902	call _key_expansion_192a
1903	AESKEYGENASSIST 0x80 %xmm2 %xmm1	# round 8
1904	call _key_expansion_192b
1905	jmp .Ldec_key
1906.Lenc_key128:
1907	AESKEYGENASSIST 0x1 %xmm0 %xmm1		# round 1
1908	call _key_expansion_128
1909	AESKEYGENASSIST 0x2 %xmm0 %xmm1		# round 2
1910	call _key_expansion_128
1911	AESKEYGENASSIST 0x4 %xmm0 %xmm1		# round 3
1912	call _key_expansion_128
1913	AESKEYGENASSIST 0x8 %xmm0 %xmm1		# round 4
1914	call _key_expansion_128
1915	AESKEYGENASSIST 0x10 %xmm0 %xmm1	# round 5
1916	call _key_expansion_128
1917	AESKEYGENASSIST 0x20 %xmm0 %xmm1	# round 6
1918	call _key_expansion_128
1919	AESKEYGENASSIST 0x40 %xmm0 %xmm1	# round 7
1920	call _key_expansion_128
1921	AESKEYGENASSIST 0x80 %xmm0 %xmm1	# round 8
1922	call _key_expansion_128
1923	AESKEYGENASSIST 0x1b %xmm0 %xmm1	# round 9
1924	call _key_expansion_128
1925	AESKEYGENASSIST 0x36 %xmm0 %xmm1	# round 10
1926	call _key_expansion_128
1927.Ldec_key:
1928	sub $0x10, TKEYP
1929	movaps (KEYP), %xmm0
1930	movaps (TKEYP), %xmm1
1931	movaps %xmm0, 240(TKEYP)
1932	movaps %xmm1, 240(KEYP)
1933	add $0x10, KEYP
1934	lea 240-16(TKEYP), UKEYP
1935.align 4
1936.Ldec_key_loop:
1937	movaps (KEYP), %xmm0
1938	AESIMC %xmm0 %xmm1
1939	movaps %xmm1, (UKEYP)
1940	add $0x10, KEYP
1941	sub $0x10, UKEYP
1942	cmp TKEYP, KEYP
1943	jb .Ldec_key_loop
1944	xor AREG, AREG
1945#ifndef __x86_64__
1946	popl KEYP
1947#endif
1948	FRAME_END
1949	ret
1950ENDPROC(aesni_set_key)
1951
1952/*
1953 * void aesni_enc(struct crypto_aes_ctx *ctx, u8 *dst, const u8 *src)
1954 */
1955ENTRY(aesni_enc)
1956	FRAME_BEGIN
1957#ifndef __x86_64__
1958	pushl KEYP
1959	pushl KLEN
1960	movl (FRAME_OFFSET+12)(%esp), KEYP	# ctx
1961	movl (FRAME_OFFSET+16)(%esp), OUTP	# dst
1962	movl (FRAME_OFFSET+20)(%esp), INP	# src
1963#endif
1964	movl 480(KEYP), KLEN		# key length
1965	movups (INP), STATE		# input
1966	call _aesni_enc1
1967	movups STATE, (OUTP)		# output
1968#ifndef __x86_64__
1969	popl KLEN
1970	popl KEYP
1971#endif
1972	FRAME_END
1973	ret
1974ENDPROC(aesni_enc)
1975
1976/*
1977 * _aesni_enc1:		internal ABI
1978 * input:
1979 *	KEYP:		key struct pointer
1980 *	KLEN:		round count
1981 *	STATE:		initial state (input)
1982 * output:
1983 *	STATE:		finial state (output)
1984 * changed:
1985 *	KEY
1986 *	TKEYP (T1)
1987 */
1988.align 4
1989_aesni_enc1:
1990	movaps (KEYP), KEY		# key
1991	mov KEYP, TKEYP
1992	pxor KEY, STATE		# round 0
1993	add $0x30, TKEYP
1994	cmp $24, KLEN
1995	jb .Lenc128
1996	lea 0x20(TKEYP), TKEYP
1997	je .Lenc192
1998	add $0x20, TKEYP
1999	movaps -0x60(TKEYP), KEY
2000	AESENC KEY STATE
2001	movaps -0x50(TKEYP), KEY
2002	AESENC KEY STATE
2003.align 4
2004.Lenc192:
2005	movaps -0x40(TKEYP), KEY
2006	AESENC KEY STATE
2007	movaps -0x30(TKEYP), KEY
2008	AESENC KEY STATE
2009.align 4
2010.Lenc128:
2011	movaps -0x20(TKEYP), KEY
2012	AESENC KEY STATE
2013	movaps -0x10(TKEYP), KEY
2014	AESENC KEY STATE
2015	movaps (TKEYP), KEY
2016	AESENC KEY STATE
2017	movaps 0x10(TKEYP), KEY
2018	AESENC KEY STATE
2019	movaps 0x20(TKEYP), KEY
2020	AESENC KEY STATE
2021	movaps 0x30(TKEYP), KEY
2022	AESENC KEY STATE
2023	movaps 0x40(TKEYP), KEY
2024	AESENC KEY STATE
2025	movaps 0x50(TKEYP), KEY
2026	AESENC KEY STATE
2027	movaps 0x60(TKEYP), KEY
2028	AESENC KEY STATE
2029	movaps 0x70(TKEYP), KEY
2030	AESENCLAST KEY STATE
2031	ret
2032ENDPROC(_aesni_enc1)
2033
2034/*
2035 * _aesni_enc4:	internal ABI
2036 * input:
2037 *	KEYP:		key struct pointer
2038 *	KLEN:		round count
2039 *	STATE1:		initial state (input)
2040 *	STATE2
2041 *	STATE3
2042 *	STATE4
2043 * output:
2044 *	STATE1:		finial state (output)
2045 *	STATE2
2046 *	STATE3
2047 *	STATE4
2048 * changed:
2049 *	KEY
2050 *	TKEYP (T1)
2051 */
2052.align 4
2053_aesni_enc4:
2054	movaps (KEYP), KEY		# key
2055	mov KEYP, TKEYP
2056	pxor KEY, STATE1		# round 0
2057	pxor KEY, STATE2
2058	pxor KEY, STATE3
2059	pxor KEY, STATE4
2060	add $0x30, TKEYP
2061	cmp $24, KLEN
2062	jb .L4enc128
2063	lea 0x20(TKEYP), TKEYP
2064	je .L4enc192
2065	add $0x20, TKEYP
2066	movaps -0x60(TKEYP), KEY
2067	AESENC KEY STATE1
2068	AESENC KEY STATE2
2069	AESENC KEY STATE3
2070	AESENC KEY STATE4
2071	movaps -0x50(TKEYP), KEY
2072	AESENC KEY STATE1
2073	AESENC KEY STATE2
2074	AESENC KEY STATE3
2075	AESENC KEY STATE4
2076#.align 4
2077.L4enc192:
2078	movaps -0x40(TKEYP), KEY
2079	AESENC KEY STATE1
2080	AESENC KEY STATE2
2081	AESENC KEY STATE3
2082	AESENC KEY STATE4
2083	movaps -0x30(TKEYP), KEY
2084	AESENC KEY STATE1
2085	AESENC KEY STATE2
2086	AESENC KEY STATE3
2087	AESENC KEY STATE4
2088#.align 4
2089.L4enc128:
2090	movaps -0x20(TKEYP), KEY
2091	AESENC KEY STATE1
2092	AESENC KEY STATE2
2093	AESENC KEY STATE3
2094	AESENC KEY STATE4
2095	movaps -0x10(TKEYP), KEY
2096	AESENC KEY STATE1
2097	AESENC KEY STATE2
2098	AESENC KEY STATE3
2099	AESENC KEY STATE4
2100	movaps (TKEYP), KEY
2101	AESENC KEY STATE1
2102	AESENC KEY STATE2
2103	AESENC KEY STATE3
2104	AESENC KEY STATE4
2105	movaps 0x10(TKEYP), KEY
2106	AESENC KEY STATE1
2107	AESENC KEY STATE2
2108	AESENC KEY STATE3
2109	AESENC KEY STATE4
2110	movaps 0x20(TKEYP), KEY
2111	AESENC KEY STATE1
2112	AESENC KEY STATE2
2113	AESENC KEY STATE3
2114	AESENC KEY STATE4
2115	movaps 0x30(TKEYP), KEY
2116	AESENC KEY STATE1
2117	AESENC KEY STATE2
2118	AESENC KEY STATE3
2119	AESENC KEY STATE4
2120	movaps 0x40(TKEYP), KEY
2121	AESENC KEY STATE1
2122	AESENC KEY STATE2
2123	AESENC KEY STATE3
2124	AESENC KEY STATE4
2125	movaps 0x50(TKEYP), KEY
2126	AESENC KEY STATE1
2127	AESENC KEY STATE2
2128	AESENC KEY STATE3
2129	AESENC KEY STATE4
2130	movaps 0x60(TKEYP), KEY
2131	AESENC KEY STATE1
2132	AESENC KEY STATE2
2133	AESENC KEY STATE3
2134	AESENC KEY STATE4
2135	movaps 0x70(TKEYP), KEY
2136	AESENCLAST KEY STATE1		# last round
2137	AESENCLAST KEY STATE2
2138	AESENCLAST KEY STATE3
2139	AESENCLAST KEY STATE4
2140	ret
2141ENDPROC(_aesni_enc4)
2142
2143/*
2144 * void aesni_dec (struct crypto_aes_ctx *ctx, u8 *dst, const u8 *src)
2145 */
2146ENTRY(aesni_dec)
2147	FRAME_BEGIN
2148#ifndef __x86_64__
2149	pushl KEYP
2150	pushl KLEN
2151	movl (FRAME_OFFSET+12)(%esp), KEYP	# ctx
2152	movl (FRAME_OFFSET+16)(%esp), OUTP	# dst
2153	movl (FRAME_OFFSET+20)(%esp), INP	# src
2154#endif
2155	mov 480(KEYP), KLEN		# key length
2156	add $240, KEYP
2157	movups (INP), STATE		# input
2158	call _aesni_dec1
2159	movups STATE, (OUTP)		#output
2160#ifndef __x86_64__
2161	popl KLEN
2162	popl KEYP
2163#endif
2164	FRAME_END
2165	ret
2166ENDPROC(aesni_dec)
2167
2168/*
2169 * _aesni_dec1:		internal ABI
2170 * input:
2171 *	KEYP:		key struct pointer
2172 *	KLEN:		key length
2173 *	STATE:		initial state (input)
2174 * output:
2175 *	STATE:		finial state (output)
2176 * changed:
2177 *	KEY
2178 *	TKEYP (T1)
2179 */
2180.align 4
2181_aesni_dec1:
2182	movaps (KEYP), KEY		# key
2183	mov KEYP, TKEYP
2184	pxor KEY, STATE		# round 0
2185	add $0x30, TKEYP
2186	cmp $24, KLEN
2187	jb .Ldec128
2188	lea 0x20(TKEYP), TKEYP
2189	je .Ldec192
2190	add $0x20, TKEYP
2191	movaps -0x60(TKEYP), KEY
2192	AESDEC KEY STATE
2193	movaps -0x50(TKEYP), KEY
2194	AESDEC KEY STATE
2195.align 4
2196.Ldec192:
2197	movaps -0x40(TKEYP), KEY
2198	AESDEC KEY STATE
2199	movaps -0x30(TKEYP), KEY
2200	AESDEC KEY STATE
2201.align 4
2202.Ldec128:
2203	movaps -0x20(TKEYP), KEY
2204	AESDEC KEY STATE
2205	movaps -0x10(TKEYP), KEY
2206	AESDEC KEY STATE
2207	movaps (TKEYP), KEY
2208	AESDEC KEY STATE
2209	movaps 0x10(TKEYP), KEY
2210	AESDEC KEY STATE
2211	movaps 0x20(TKEYP), KEY
2212	AESDEC KEY STATE
2213	movaps 0x30(TKEYP), KEY
2214	AESDEC KEY STATE
2215	movaps 0x40(TKEYP), KEY
2216	AESDEC KEY STATE
2217	movaps 0x50(TKEYP), KEY
2218	AESDEC KEY STATE
2219	movaps 0x60(TKEYP), KEY
2220	AESDEC KEY STATE
2221	movaps 0x70(TKEYP), KEY
2222	AESDECLAST KEY STATE
2223	ret
2224ENDPROC(_aesni_dec1)
2225
2226/*
2227 * _aesni_dec4:	internal ABI
2228 * input:
2229 *	KEYP:		key struct pointer
2230 *	KLEN:		key length
2231 *	STATE1:		initial state (input)
2232 *	STATE2
2233 *	STATE3
2234 *	STATE4
2235 * output:
2236 *	STATE1:		finial state (output)
2237 *	STATE2
2238 *	STATE3
2239 *	STATE4
2240 * changed:
2241 *	KEY
2242 *	TKEYP (T1)
2243 */
2244.align 4
2245_aesni_dec4:
2246	movaps (KEYP), KEY		# key
2247	mov KEYP, TKEYP
2248	pxor KEY, STATE1		# round 0
2249	pxor KEY, STATE2
2250	pxor KEY, STATE3
2251	pxor KEY, STATE4
2252	add $0x30, TKEYP
2253	cmp $24, KLEN
2254	jb .L4dec128
2255	lea 0x20(TKEYP), TKEYP
2256	je .L4dec192
2257	add $0x20, TKEYP
2258	movaps -0x60(TKEYP), KEY
2259	AESDEC KEY STATE1
2260	AESDEC KEY STATE2
2261	AESDEC KEY STATE3
2262	AESDEC KEY STATE4
2263	movaps -0x50(TKEYP), KEY
2264	AESDEC KEY STATE1
2265	AESDEC KEY STATE2
2266	AESDEC KEY STATE3
2267	AESDEC KEY STATE4
2268.align 4
2269.L4dec192:
2270	movaps -0x40(TKEYP), KEY
2271	AESDEC KEY STATE1
2272	AESDEC KEY STATE2
2273	AESDEC KEY STATE3
2274	AESDEC KEY STATE4
2275	movaps -0x30(TKEYP), KEY
2276	AESDEC KEY STATE1
2277	AESDEC KEY STATE2
2278	AESDEC KEY STATE3
2279	AESDEC KEY STATE4
2280.align 4
2281.L4dec128:
2282	movaps -0x20(TKEYP), KEY
2283	AESDEC KEY STATE1
2284	AESDEC KEY STATE2
2285	AESDEC KEY STATE3
2286	AESDEC KEY STATE4
2287	movaps -0x10(TKEYP), KEY
2288	AESDEC KEY STATE1
2289	AESDEC KEY STATE2
2290	AESDEC KEY STATE3
2291	AESDEC KEY STATE4
2292	movaps (TKEYP), KEY
2293	AESDEC KEY STATE1
2294	AESDEC KEY STATE2
2295	AESDEC KEY STATE3
2296	AESDEC KEY STATE4
2297	movaps 0x10(TKEYP), KEY
2298	AESDEC KEY STATE1
2299	AESDEC KEY STATE2
2300	AESDEC KEY STATE3
2301	AESDEC KEY STATE4
2302	movaps 0x20(TKEYP), KEY
2303	AESDEC KEY STATE1
2304	AESDEC KEY STATE2
2305	AESDEC KEY STATE3
2306	AESDEC KEY STATE4
2307	movaps 0x30(TKEYP), KEY
2308	AESDEC KEY STATE1
2309	AESDEC KEY STATE2
2310	AESDEC KEY STATE3
2311	AESDEC KEY STATE4
2312	movaps 0x40(TKEYP), KEY
2313	AESDEC KEY STATE1
2314	AESDEC KEY STATE2
2315	AESDEC KEY STATE3
2316	AESDEC KEY STATE4
2317	movaps 0x50(TKEYP), KEY
2318	AESDEC KEY STATE1
2319	AESDEC KEY STATE2
2320	AESDEC KEY STATE3
2321	AESDEC KEY STATE4
2322	movaps 0x60(TKEYP), KEY
2323	AESDEC KEY STATE1
2324	AESDEC KEY STATE2
2325	AESDEC KEY STATE3
2326	AESDEC KEY STATE4
2327	movaps 0x70(TKEYP), KEY
2328	AESDECLAST KEY STATE1		# last round
2329	AESDECLAST KEY STATE2
2330	AESDECLAST KEY STATE3
2331	AESDECLAST KEY STATE4
2332	ret
2333ENDPROC(_aesni_dec4)
2334
2335/*
2336 * void aesni_ecb_enc(struct crypto_aes_ctx *ctx, const u8 *dst, u8 *src,
2337 *		      size_t len)
2338 */
2339ENTRY(aesni_ecb_enc)
2340	FRAME_BEGIN
2341#ifndef __x86_64__
2342	pushl LEN
2343	pushl KEYP
2344	pushl KLEN
2345	movl (FRAME_OFFSET+16)(%esp), KEYP	# ctx
2346	movl (FRAME_OFFSET+20)(%esp), OUTP	# dst
2347	movl (FRAME_OFFSET+24)(%esp), INP	# src
2348	movl (FRAME_OFFSET+28)(%esp), LEN	# len
2349#endif
2350	test LEN, LEN		# check length
2351	jz .Lecb_enc_ret
2352	mov 480(KEYP), KLEN
2353	cmp $16, LEN
2354	jb .Lecb_enc_ret
2355	cmp $64, LEN
2356	jb .Lecb_enc_loop1
2357.align 4
2358.Lecb_enc_loop4:
2359	movups (INP), STATE1
2360	movups 0x10(INP), STATE2
2361	movups 0x20(INP), STATE3
2362	movups 0x30(INP), STATE4
2363	call _aesni_enc4
2364	movups STATE1, (OUTP)
2365	movups STATE2, 0x10(OUTP)
2366	movups STATE3, 0x20(OUTP)
2367	movups STATE4, 0x30(OUTP)
2368	sub $64, LEN
2369	add $64, INP
2370	add $64, OUTP
2371	cmp $64, LEN
2372	jge .Lecb_enc_loop4
2373	cmp $16, LEN
2374	jb .Lecb_enc_ret
2375.align 4
2376.Lecb_enc_loop1:
2377	movups (INP), STATE1
2378	call _aesni_enc1
2379	movups STATE1, (OUTP)
2380	sub $16, LEN
2381	add $16, INP
2382	add $16, OUTP
2383	cmp $16, LEN
2384	jge .Lecb_enc_loop1
2385.Lecb_enc_ret:
2386#ifndef __x86_64__
2387	popl KLEN
2388	popl KEYP
2389	popl LEN
2390#endif
2391	FRAME_END
2392	ret
2393ENDPROC(aesni_ecb_enc)
2394
2395/*
2396 * void aesni_ecb_dec(struct crypto_aes_ctx *ctx, const u8 *dst, u8 *src,
2397 *		      size_t len);
2398 */
2399ENTRY(aesni_ecb_dec)
2400	FRAME_BEGIN
2401#ifndef __x86_64__
2402	pushl LEN
2403	pushl KEYP
2404	pushl KLEN
2405	movl (FRAME_OFFSET+16)(%esp), KEYP	# ctx
2406	movl (FRAME_OFFSET+20)(%esp), OUTP	# dst
2407	movl (FRAME_OFFSET+24)(%esp), INP	# src
2408	movl (FRAME_OFFSET+28)(%esp), LEN	# len
2409#endif
2410	test LEN, LEN
2411	jz .Lecb_dec_ret
2412	mov 480(KEYP), KLEN
2413	add $240, KEYP
2414	cmp $16, LEN
2415	jb .Lecb_dec_ret
2416	cmp $64, LEN
2417	jb .Lecb_dec_loop1
2418.align 4
2419.Lecb_dec_loop4:
2420	movups (INP), STATE1
2421	movups 0x10(INP), STATE2
2422	movups 0x20(INP), STATE3
2423	movups 0x30(INP), STATE4
2424	call _aesni_dec4
2425	movups STATE1, (OUTP)
2426	movups STATE2, 0x10(OUTP)
2427	movups STATE3, 0x20(OUTP)
2428	movups STATE4, 0x30(OUTP)
2429	sub $64, LEN
2430	add $64, INP
2431	add $64, OUTP
2432	cmp $64, LEN
2433	jge .Lecb_dec_loop4
2434	cmp $16, LEN
2435	jb .Lecb_dec_ret
2436.align 4
2437.Lecb_dec_loop1:
2438	movups (INP), STATE1
2439	call _aesni_dec1
2440	movups STATE1, (OUTP)
2441	sub $16, LEN
2442	add $16, INP
2443	add $16, OUTP
2444	cmp $16, LEN
2445	jge .Lecb_dec_loop1
2446.Lecb_dec_ret:
2447#ifndef __x86_64__
2448	popl KLEN
2449	popl KEYP
2450	popl LEN
2451#endif
2452	FRAME_END
2453	ret
2454ENDPROC(aesni_ecb_dec)
2455
2456/*
2457 * void aesni_cbc_enc(struct crypto_aes_ctx *ctx, const u8 *dst, u8 *src,
2458 *		      size_t len, u8 *iv)
2459 */
2460ENTRY(aesni_cbc_enc)
2461	FRAME_BEGIN
2462#ifndef __x86_64__
2463	pushl IVP
2464	pushl LEN
2465	pushl KEYP
2466	pushl KLEN
2467	movl (FRAME_OFFSET+20)(%esp), KEYP	# ctx
2468	movl (FRAME_OFFSET+24)(%esp), OUTP	# dst
2469	movl (FRAME_OFFSET+28)(%esp), INP	# src
2470	movl (FRAME_OFFSET+32)(%esp), LEN	# len
2471	movl (FRAME_OFFSET+36)(%esp), IVP	# iv
2472#endif
2473	cmp $16, LEN
2474	jb .Lcbc_enc_ret
2475	mov 480(KEYP), KLEN
2476	movups (IVP), STATE	# load iv as initial state
2477.align 4
2478.Lcbc_enc_loop:
2479	movups (INP), IN	# load input
2480	pxor IN, STATE
2481	call _aesni_enc1
2482	movups STATE, (OUTP)	# store output
2483	sub $16, LEN
2484	add $16, INP
2485	add $16, OUTP
2486	cmp $16, LEN
2487	jge .Lcbc_enc_loop
2488	movups STATE, (IVP)
2489.Lcbc_enc_ret:
2490#ifndef __x86_64__
2491	popl KLEN
2492	popl KEYP
2493	popl LEN
2494	popl IVP
2495#endif
2496	FRAME_END
2497	ret
2498ENDPROC(aesni_cbc_enc)
2499
2500/*
2501 * void aesni_cbc_dec(struct crypto_aes_ctx *ctx, const u8 *dst, u8 *src,
2502 *		      size_t len, u8 *iv)
2503 */
2504ENTRY(aesni_cbc_dec)
2505	FRAME_BEGIN
2506#ifndef __x86_64__
2507	pushl IVP
2508	pushl LEN
2509	pushl KEYP
2510	pushl KLEN
2511	movl (FRAME_OFFSET+20)(%esp), KEYP	# ctx
2512	movl (FRAME_OFFSET+24)(%esp), OUTP	# dst
2513	movl (FRAME_OFFSET+28)(%esp), INP	# src
2514	movl (FRAME_OFFSET+32)(%esp), LEN	# len
2515	movl (FRAME_OFFSET+36)(%esp), IVP	# iv
2516#endif
2517	cmp $16, LEN
2518	jb .Lcbc_dec_just_ret
2519	mov 480(KEYP), KLEN
2520	add $240, KEYP
2521	movups (IVP), IV
2522	cmp $64, LEN
2523	jb .Lcbc_dec_loop1
2524.align 4
2525.Lcbc_dec_loop4:
2526	movups (INP), IN1
2527	movaps IN1, STATE1
2528	movups 0x10(INP), IN2
2529	movaps IN2, STATE2
2530#ifdef __x86_64__
2531	movups 0x20(INP), IN3
2532	movaps IN3, STATE3
2533	movups 0x30(INP), IN4
2534	movaps IN4, STATE4
2535#else
2536	movups 0x20(INP), IN1
2537	movaps IN1, STATE3
2538	movups 0x30(INP), IN2
2539	movaps IN2, STATE4
2540#endif
2541	call _aesni_dec4
2542	pxor IV, STATE1
2543#ifdef __x86_64__
2544	pxor IN1, STATE2
2545	pxor IN2, STATE3
2546	pxor IN3, STATE4
2547	movaps IN4, IV
2548#else
2549	pxor IN1, STATE4
2550	movaps IN2, IV
2551	movups (INP), IN1
2552	pxor IN1, STATE2
2553	movups 0x10(INP), IN2
2554	pxor IN2, STATE3
2555#endif
2556	movups STATE1, (OUTP)
2557	movups STATE2, 0x10(OUTP)
2558	movups STATE3, 0x20(OUTP)
2559	movups STATE4, 0x30(OUTP)
2560	sub $64, LEN
2561	add $64, INP
2562	add $64, OUTP
2563	cmp $64, LEN
2564	jge .Lcbc_dec_loop4
2565	cmp $16, LEN
2566	jb .Lcbc_dec_ret
2567.align 4
2568.Lcbc_dec_loop1:
2569	movups (INP), IN
2570	movaps IN, STATE
2571	call _aesni_dec1
2572	pxor IV, STATE
2573	movups STATE, (OUTP)
2574	movaps IN, IV
2575	sub $16, LEN
2576	add $16, INP
2577	add $16, OUTP
2578	cmp $16, LEN
2579	jge .Lcbc_dec_loop1
2580.Lcbc_dec_ret:
2581	movups IV, (IVP)
2582.Lcbc_dec_just_ret:
2583#ifndef __x86_64__
2584	popl KLEN
2585	popl KEYP
2586	popl LEN
2587	popl IVP
2588#endif
2589	FRAME_END
2590	ret
2591ENDPROC(aesni_cbc_dec)
2592
2593#ifdef __x86_64__
2594.pushsection .rodata
2595.align 16
2596.Lbswap_mask:
2597	.byte 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0
2598.popsection
2599
2600/*
2601 * _aesni_inc_init:	internal ABI
2602 *	setup registers used by _aesni_inc
2603 * input:
2604 *	IV
2605 * output:
2606 *	CTR:	== IV, in little endian
2607 *	TCTR_LOW: == lower qword of CTR
2608 *	INC:	== 1, in little endian
2609 *	BSWAP_MASK == endian swapping mask
2610 */
2611.align 4
2612_aesni_inc_init:
2613	movaps .Lbswap_mask, BSWAP_MASK
2614	movaps IV, CTR
2615	PSHUFB_XMM BSWAP_MASK CTR
2616	mov $1, TCTR_LOW
2617	MOVQ_R64_XMM TCTR_LOW INC
2618	MOVQ_R64_XMM CTR TCTR_LOW
2619	ret
2620ENDPROC(_aesni_inc_init)
2621
2622/*
2623 * _aesni_inc:		internal ABI
2624 *	Increase IV by 1, IV is in big endian
2625 * input:
2626 *	IV
2627 *	CTR:	== IV, in little endian
2628 *	TCTR_LOW: == lower qword of CTR
2629 *	INC:	== 1, in little endian
2630 *	BSWAP_MASK == endian swapping mask
2631 * output:
2632 *	IV:	Increase by 1
2633 * changed:
2634 *	CTR:	== output IV, in little endian
2635 *	TCTR_LOW: == lower qword of CTR
2636 */
2637.align 4
2638_aesni_inc:
2639	paddq INC, CTR
2640	add $1, TCTR_LOW
2641	jnc .Linc_low
2642	pslldq $8, INC
2643	paddq INC, CTR
2644	psrldq $8, INC
2645.Linc_low:
2646	movaps CTR, IV
2647	PSHUFB_XMM BSWAP_MASK IV
2648	ret
2649ENDPROC(_aesni_inc)
2650
2651/*
2652 * void aesni_ctr_enc(struct crypto_aes_ctx *ctx, const u8 *dst, u8 *src,
2653 *		      size_t len, u8 *iv)
2654 */
2655ENTRY(aesni_ctr_enc)
2656	FRAME_BEGIN
2657	cmp $16, LEN
2658	jb .Lctr_enc_just_ret
2659	mov 480(KEYP), KLEN
2660	movups (IVP), IV
2661	call _aesni_inc_init
2662	cmp $64, LEN
2663	jb .Lctr_enc_loop1
2664.align 4
2665.Lctr_enc_loop4:
2666	movaps IV, STATE1
2667	call _aesni_inc
2668	movups (INP), IN1
2669	movaps IV, STATE2
2670	call _aesni_inc
2671	movups 0x10(INP), IN2
2672	movaps IV, STATE3
2673	call _aesni_inc
2674	movups 0x20(INP), IN3
2675	movaps IV, STATE4
2676	call _aesni_inc
2677	movups 0x30(INP), IN4
2678	call _aesni_enc4
2679	pxor IN1, STATE1
2680	movups STATE1, (OUTP)
2681	pxor IN2, STATE2
2682	movups STATE2, 0x10(OUTP)
2683	pxor IN3, STATE3
2684	movups STATE3, 0x20(OUTP)
2685	pxor IN4, STATE4
2686	movups STATE4, 0x30(OUTP)
2687	sub $64, LEN
2688	add $64, INP
2689	add $64, OUTP
2690	cmp $64, LEN
2691	jge .Lctr_enc_loop4
2692	cmp $16, LEN
2693	jb .Lctr_enc_ret
2694.align 4
2695.Lctr_enc_loop1:
2696	movaps IV, STATE
2697	call _aesni_inc
2698	movups (INP), IN
2699	call _aesni_enc1
2700	pxor IN, STATE
2701	movups STATE, (OUTP)
2702	sub $16, LEN
2703	add $16, INP
2704	add $16, OUTP
2705	cmp $16, LEN
2706	jge .Lctr_enc_loop1
2707.Lctr_enc_ret:
2708	movups IV, (IVP)
2709.Lctr_enc_just_ret:
2710	FRAME_END
2711	ret
2712ENDPROC(aesni_ctr_enc)
2713
2714/*
2715 * _aesni_gf128mul_x_ble:		internal ABI
2716 *	Multiply in GF(2^128) for XTS IVs
2717 * input:
2718 *	IV:	current IV
2719 *	GF128MUL_MASK == mask with 0x87 and 0x01
2720 * output:
2721 *	IV:	next IV
2722 * changed:
2723 *	CTR:	== temporary value
2724 */
2725#define _aesni_gf128mul_x_ble() \
2726	pshufd $0x13, IV, CTR; \
2727	paddq IV, IV; \
2728	psrad $31, CTR; \
2729	pand GF128MUL_MASK, CTR; \
2730	pxor CTR, IV;
2731
2732/*
2733 * void aesni_xts_crypt8(struct crypto_aes_ctx *ctx, const u8 *dst, u8 *src,
2734 *			 bool enc, u8 *iv)
2735 */
2736ENTRY(aesni_xts_crypt8)
2737	FRAME_BEGIN
2738	cmpb $0, %cl
2739	movl $0, %ecx
2740	movl $240, %r10d
2741	leaq _aesni_enc4, %r11
2742	leaq _aesni_dec4, %rax
2743	cmovel %r10d, %ecx
2744	cmoveq %rax, %r11
2745
2746	movdqa .Lgf128mul_x_ble_mask, GF128MUL_MASK
2747	movups (IVP), IV
2748
2749	mov 480(KEYP), KLEN
2750	addq %rcx, KEYP
2751
2752	movdqa IV, STATE1
2753	movdqu 0x00(INP), INC
2754	pxor INC, STATE1
2755	movdqu IV, 0x00(OUTP)
2756
2757	_aesni_gf128mul_x_ble()
2758	movdqa IV, STATE2
2759	movdqu 0x10(INP), INC
2760	pxor INC, STATE2
2761	movdqu IV, 0x10(OUTP)
2762
2763	_aesni_gf128mul_x_ble()
2764	movdqa IV, STATE3
2765	movdqu 0x20(INP), INC
2766	pxor INC, STATE3
2767	movdqu IV, 0x20(OUTP)
2768
2769	_aesni_gf128mul_x_ble()
2770	movdqa IV, STATE4
2771	movdqu 0x30(INP), INC
2772	pxor INC, STATE4
2773	movdqu IV, 0x30(OUTP)
2774
2775	CALL_NOSPEC %r11
2776
2777	movdqu 0x00(OUTP), INC
2778	pxor INC, STATE1
2779	movdqu STATE1, 0x00(OUTP)
2780
2781	_aesni_gf128mul_x_ble()
2782	movdqa IV, STATE1
2783	movdqu 0x40(INP), INC
2784	pxor INC, STATE1
2785	movdqu IV, 0x40(OUTP)
2786
2787	movdqu 0x10(OUTP), INC
2788	pxor INC, STATE2
2789	movdqu STATE2, 0x10(OUTP)
2790
2791	_aesni_gf128mul_x_ble()
2792	movdqa IV, STATE2
2793	movdqu 0x50(INP), INC
2794	pxor INC, STATE2
2795	movdqu IV, 0x50(OUTP)
2796
2797	movdqu 0x20(OUTP), INC
2798	pxor INC, STATE3
2799	movdqu STATE3, 0x20(OUTP)
2800
2801	_aesni_gf128mul_x_ble()
2802	movdqa IV, STATE3
2803	movdqu 0x60(INP), INC
2804	pxor INC, STATE3
2805	movdqu IV, 0x60(OUTP)
2806
2807	movdqu 0x30(OUTP), INC
2808	pxor INC, STATE4
2809	movdqu STATE4, 0x30(OUTP)
2810
2811	_aesni_gf128mul_x_ble()
2812	movdqa IV, STATE4
2813	movdqu 0x70(INP), INC
2814	pxor INC, STATE4
2815	movdqu IV, 0x70(OUTP)
2816
2817	_aesni_gf128mul_x_ble()
2818	movups IV, (IVP)
2819
2820	CALL_NOSPEC %r11
2821
2822	movdqu 0x40(OUTP), INC
2823	pxor INC, STATE1
2824	movdqu STATE1, 0x40(OUTP)
2825
2826	movdqu 0x50(OUTP), INC
2827	pxor INC, STATE2
2828	movdqu STATE2, 0x50(OUTP)
2829
2830	movdqu 0x60(OUTP), INC
2831	pxor INC, STATE3
2832	movdqu STATE3, 0x60(OUTP)
2833
2834	movdqu 0x70(OUTP), INC
2835	pxor INC, STATE4
2836	movdqu STATE4, 0x70(OUTP)
2837
2838	FRAME_END
2839	ret
2840ENDPROC(aesni_xts_crypt8)
2841
2842#endif
2843