crypto: SHA1 multibuffer crypto hash infrastructure

index ce4012a..5d3c711 100644
@@ -157,6 +157,20 @@ config CRYPTO_CRYPTD
 	  converts an arbitrary synchronous software crypto algorithm
 	  into an asynchronous algorithm that executes in a kernel thread.
 
+config CRYPTO_MCRYPTD
+	tristate "Software async multi-buffer crypto daemon"
+	select CRYPTO_BLKCIPHER
+	select CRYPTO_HASH
+	select CRYPTO_MANAGER
+	select CRYPTO_WORKQUEUE
+	help
+	  This is a generic software asynchronous crypto daemon that
+	  provides the kernel thread to assist multi-buffer crypto
+	  algorithms for submitting jobs and flushing jobs in multi-buffer
+	  crypto algorithms.  Multi-buffer crypto algorithms are executed
+	  in the context of this kernel thread and drivers can post
+	  their crypto request asyncrhously and process by this daemon.
+
 config CRYPTO_AUTHENC
 	tristate "Authenc support"
 	select CRYPTO_AEAD

@@ -547,6 +561,22 @@ config CRYPTO_SHA1_PPC
 	  This is the powerpc hardware accelerated implementation of the
 	  SHA-1 secure hash standard (FIPS 180-1/DFIPS 180-2).
 
+config CRYPTO_SHA1_MB
+	tristate "SHA1 digest algorithm (x86_64 Multi-Buffer, Experimental)"
+	depends on X86 && 64BIT
+	select CRYPTO_SHA1
+	select CRYPTO_HASH
+	select CRYPTO_MCRYPTD
+	help
+	  SHA-1 secure hash standard (FIPS 180-1/DFIPS 180-2) implemented
+	  using multi-buffer technique.  This algorithm computes on
+	  multiple data lanes concurrently with SIMD instructions for
+	  better throughput.  It should not be enabled by default but
+	  used when there is significant amount of work to keep the keep
+	  the data lanes filled to get performance benefit.  If the data
+	  lanes remain unfilled, a flush operation will be initiated to
+	  process the crypto jobs, adding a slight latency.
+
 config CRYPTO_SHA256
 	tristate "SHA224 and SHA256 digest algorithm"
 	select CRYPTO_HASH

index 38e64231..27dbbad 100644
@@ -60,6 +60,7 @@ obj-$(CONFIG_CRYPTO_GCM) += gcm.o
 obj-$(CONFIG_CRYPTO_CCM) += ccm.o
 obj-$(CONFIG_CRYPTO_PCRYPT) += pcrypt.o
 obj-$(CONFIG_CRYPTO_CRYPTD) += cryptd.o
+obj-$(CONFIG_CRYPTO_MCRYPTD) += mcryptd.o
 obj-$(CONFIG_CRYPTO_DES) += des_generic.o
 obj-$(CONFIG_CRYPTO_FCRYPT) += fcrypt.o
 obj-$(CONFIG_CRYPTO_BLOWFISH) += blowfish_generic.o

new file mode 100644
index 0000000..86688c6
@@ -0,0 +1,287 @@
+/*
+ * Header file for multi buffer SHA1 algorithm data structure
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ *  Copyright(c) 2014 Intel Corporation.
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of version 2 of the GNU General Public License as
+ *  published by the Free Software Foundation.
+ *
+ *  This program is distributed in the hope that it will be useful, but
+ *  WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ *  General Public License for more details.
+ *
+ *  Contact Information:
+ *      James Guilford <[email protected]>
+ *	Tim Chen <[email protected]>
+ *
+ *  BSD LICENSE
+ *
+ *  Copyright(c) 2014 Intel Corporation.
+ *
+ *  Redistribution and use in source and binary forms, with or without
+ *  modification, are permitted provided that the following conditions
+ *  are met:
+ *
+ *    * Redistributions of source code must retain the above copyright
+ *      notice, this list of conditions and the following disclaimer.
+ *    * Redistributions in binary form must reproduce the above copyright
+ *      notice, this list of conditions and the following disclaimer in
+ *      the documentation and/or other materials provided with the
+ *      distribution.
+ *    * Neither the name of Intel Corporation nor the names of its
+ *      contributors may be used to endorse or promote products derived
+ *      from this software without specific prior written permission.
+ *
+ *  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *  OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *  SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *  LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *  DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *  THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+# Macros for defining data structures
+
+# Usage example
+
+#START_FIELDS	# JOB_AES
+###	name		size	align
+#FIELD	_plaintext,	8,	8	# pointer to plaintext
+#FIELD	_ciphertext,	8,	8	# pointer to ciphertext
+#FIELD	_IV,		16,	8	# IV
+#FIELD	_keys,		8,	8	# pointer to keys
+#FIELD	_len,		4,	4	# length in bytes
+#FIELD	_status,	4,	4	# status enumeration
+#FIELD	_user_data,	8,	8	# pointer to user data
+#UNION  _union,         size1,  align1, \
+#	                size2,  align2, \
+#	                size3,  align3, \
+#	                ...
+#END_FIELDS
+#%assign _JOB_AES_size	_FIELD_OFFSET
+#%assign _JOB_AES_align	_STRUCT_ALIGN
+
+#########################################################################
+
+# Alternate "struc-like" syntax:
+#	STRUCT job_aes2
+#	RES_Q	.plaintext,	1
+#	RES_Q	.ciphertext,	1
+#	RES_DQ	.IV,		1
+#	RES_B	.nested,	_JOB_AES_SIZE, _JOB_AES_ALIGN
+#	RES_U	.union,		size1, align1, \
+#				size2, align2, \
+#				...
+#	ENDSTRUCT
+#	# Following only needed if nesting
+#	%assign job_aes2_size	_FIELD_OFFSET
+#	%assign job_aes2_align	_STRUCT_ALIGN
+#
+# RES_* macros take a name, a count and an optional alignment.
+# The count in in terms of the base size of the macro, and the
+# default alignment is the base size.
+# The macros are:
+# Macro    Base size
+# RES_B	    1
+# RES_W	    2
+# RES_D     4
+# RES_Q     8
+# RES_DQ   16
+# RES_Y    32
+# RES_Z    64
+#
+# RES_U defines a union. It's arguments are a name and two or more
+# pairs of "size, alignment"
+#
+# The two assigns are only needed if this structure is being nested
+# within another. Even if the assigns are not done, one can still use
+# STRUCT_NAME_size as the size of the structure.
+#
+# Note that for nesting, you still need to assign to STRUCT_NAME_size.
+#
+# The differences between this and using "struc" directly are that each
+# type is implicitly aligned to its natural length (although this can be
+# over-ridden with an explicit third parameter), and that the structure
+# is padded at the end to its overall alignment.
+#
+
+#########################################################################
+
+#ifndef _SHA1_MB_MGR_DATASTRUCT_ASM_
+#define _SHA1_MB_MGR_DATASTRUCT_ASM_
+
+## START_FIELDS
+.macro START_FIELDS
+ _FIELD_OFFSET = 0
+ _STRUCT_ALIGN = 0
+.endm
+
+## FIELD name size align
+.macro FIELD name size align
+ _FIELD_OFFSET = (_FIELD_OFFSET + (\align) - 1) & (~ ((\align)-1))
+ \name	= _FIELD_OFFSET
+ _FIELD_OFFSET = _FIELD_OFFSET + (\size)
+.if (\align > _STRUCT_ALIGN)
+ _STRUCT_ALIGN = \align
+.endif
+.endm
+
+## END_FIELDS
+.macro END_FIELDS
+ _FIELD_OFFSET = (_FIELD_OFFSET + _STRUCT_ALIGN-1) & (~ (_STRUCT_ALIGN-1))
+.endm
+
+########################################################################
+
+.macro STRUCT p1
+START_FIELDS
+.struc \p1
+.endm
+
+.macro ENDSTRUCT
+ tmp = _FIELD_OFFSET
+ END_FIELDS
+ tmp = (_FIELD_OFFSET - %%tmp)
+.if (tmp > 0)
+	.lcomm	tmp
+.endif
+.endstruc
+.endm
+
+## RES_int name size align
+.macro RES_int p1 p2 p3
+ name = \p1
+ size = \p2
+ align = .\p3
+
+ _FIELD_OFFSET = (_FIELD_OFFSET + (align) - 1) & (~ ((align)-1))
+.align align
+.lcomm name size
+ _FIELD_OFFSET = _FIELD_OFFSET + (size)
+.if (align > _STRUCT_ALIGN)
+ _STRUCT_ALIGN = align
+.endif
+.endm
+
+
+
+# macro RES_B name, size [, align]
+.macro RES_B _name, _size, _align=1
+RES_int _name _size _align
+.endm
+
+# macro RES_W name, size [, align]
+.macro RES_W _name, _size, _align=2
+RES_int _name 2*(_size) _align
+.endm
+
+# macro RES_D name, size [, align]
+.macro RES_D _name, _size, _align=4
+RES_int _name 4*(_size) _align
+.endm
+
+# macro RES_Q name, size [, align]
+.macro RES_Q _name, _size, _align=8
+RES_int _name 8*(_size) _align
+.endm
+
+# macro RES_DQ name, size [, align]
+.macro RES_DQ _name, _size, _align=16
+RES_int _name 16*(_size) _align
+.endm
+
+# macro RES_Y name, size [, align]
+.macro RES_Y _name, _size, _align=32
+RES_int _name 32*(_size) _align
+.endm
+
+# macro RES_Z name, size [, align]
+.macro RES_Z _name, _size, _align=64
+RES_int _name 64*(_size) _align
+.endm
+
+
+#endif
+
+########################################################################
+#### Define constants
+########################################################################
+
+########################################################################
+#### Define SHA1 Out Of Order Data Structures
+########################################################################
+
+START_FIELDS    # LANE_DATA
+###     name            size    align
+FIELD   _job_in_lane,   8,      8       # pointer to job object
+END_FIELDS
+
+_LANE_DATA_size = _FIELD_OFFSET
+_LANE_DATA_align = _STRUCT_ALIGN
+
+########################################################################
+
+START_FIELDS    # SHA1_ARGS_X8
+###     name            size    align
+FIELD   _digest,        4*5*8,  16      # transposed digest
+FIELD   _data_ptr,      8*8,    8       # array of pointers to data
+END_FIELDS
+
+_SHA1_ARGS_X4_size =     _FIELD_OFFSET
+_SHA1_ARGS_X4_align =    _STRUCT_ALIGN
+_SHA1_ARGS_X8_size =     _FIELD_OFFSET
+_SHA1_ARGS_X8_align =    _STRUCT_ALIGN
+
+########################################################################
+
+START_FIELDS    # MB_MGR
+###     name            size    align
+FIELD   _args,          _SHA1_ARGS_X4_size, _SHA1_ARGS_X4_align
+FIELD   _lens,          4*8,    8
+FIELD   _unused_lanes,  8,      8
+FIELD   _ldata,         _LANE_DATA_size*8, _LANE_DATA_align
+END_FIELDS
+
+_MB_MGR_size =   _FIELD_OFFSET
+_MB_MGR_align =  _STRUCT_ALIGN
+
+_args_digest    =     _args + _digest
+_args_data_ptr  =     _args + _data_ptr
+
+
+########################################################################
+#### Define constants
+########################################################################
+
+#define STS_UNKNOWN             0
+#define STS_BEING_PROCESSED     1
+#define STS_COMPLETED           2
+
+########################################################################
+#### Define JOB_SHA1 structure
+########################################################################
+
+START_FIELDS    # JOB_SHA1
+
+###     name                            size    align
+FIELD   _buffer,                        8,      8       # pointer to buffer
+FIELD   _len,                           4,      4       # length in bytes
+FIELD   _result_digest,                 5*4,    32      # Digest (output)
+FIELD   _status,                        4,      4
+FIELD   _user_data,                     8,      8
+END_FIELDS
+
+_JOB_SHA1_size =  _FIELD_OFFSET
+_JOB_SHA1_align = _STRUCT_ALIGN

new file mode 100644
index 0000000..85c4e1c
@@ -0,0 +1,327 @@
+/*
+ * Flush routine for SHA1 multibuffer
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ *  Copyright(c) 2014 Intel Corporation.
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of version 2 of the GNU General Public License as
+ *  published by the Free Software Foundation.
+ *
+ *  This program is distributed in the hope that it will be useful, but
+ *  WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ *  General Public License for more details.
+ *
+ *  Contact Information:
+ *      James Guilford <[email protected]>
+ *	Tim Chen <[email protected]>
+ *
+ *  BSD LICENSE
+ *
+ *  Copyright(c) 2014 Intel Corporation.
+ *
+ *  Redistribution and use in source and binary forms, with or without
+ *  modification, are permitted provided that the following conditions
+ *  are met:
+ *
+ *    * Redistributions of source code must retain the above copyright
+ *      notice, this list of conditions and the following disclaimer.
+ *    * Redistributions in binary form must reproduce the above copyright
+ *      notice, this list of conditions and the following disclaimer in
+ *      the documentation and/or other materials provided with the
+ *      distribution.
+ *    * Neither the name of Intel Corporation nor the names of its
+ *      contributors may be used to endorse or promote products derived
+ *      from this software without specific prior written permission.
+ *
+ *  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *  OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *  SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *  LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *  DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *  THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#include <linux/linkage.h>
+#include "sha1_mb_mgr_datastruct.S"
+
+
+.extern sha1_x8_avx2
+
+# LINUX register definitions
+#define arg1    %rdi
+#define arg2    %rsi
+
+# Common definitions
+#define state   arg1
+#define job     arg2
+#define len2    arg2
+
+# idx must be a register not clobbered by sha1_x8_avx2
+#define idx		%r8
+#define DWORD_idx	%r8d
+
+#define unused_lanes    %rbx
+#define lane_data       %rbx
+#define tmp2            %rbx
+#define tmp2_w		%ebx
+
+#define job_rax         %rax
+#define tmp1            %rax
+#define size_offset     %rax
+#define tmp             %rax
+#define start_offset    %rax
+
+#define tmp3            %arg1
+
+#define extra_blocks    %arg2
+#define p               %arg2
+
+
+# STACK_SPACE needs to be an odd multiple of 8
+_XMM_SAVE_SIZE  = 10*16
+_GPR_SAVE_SIZE  = 8*8
+_ALIGN_SIZE     = 8
+
+_XMM_SAVE       = 0
+_GPR_SAVE       = _XMM_SAVE + _XMM_SAVE_SIZE
+STACK_SPACE     = _GPR_SAVE + _GPR_SAVE_SIZE + _ALIGN_SIZE
+
+.macro LABEL prefix n
+\prefix\n\():
+.endm
+
+.macro JNE_SKIP i
+jne     skip_\i
+.endm
+
+.altmacro
+.macro SET_OFFSET _offset
+offset = \_offset
+.endm
+.noaltmacro
+
+# JOB* sha1_mb_mgr_flush_avx2(MB_MGR *state)
+# arg 1 : rcx : state
+ENTRY(sha1_mb_mgr_flush_avx2)
+	mov	%rsp, %r10
+	sub     $STACK_SPACE, %rsp
+	and     $~31, %rsp
+	mov     %rbx, _GPR_SAVE(%rsp)
+	mov     %r10, _GPR_SAVE+8*1(%rsp) #save rsp
+	mov	%rbp, _GPR_SAVE+8*3(%rsp)
+	mov	%r12, _GPR_SAVE+8*4(%rsp)
+	mov	%r13, _GPR_SAVE+8*5(%rsp)
+	mov	%r14, _GPR_SAVE+8*6(%rsp)
+	mov	%r15, _GPR_SAVE+8*7(%rsp)
+
+	# If bit (32+3) is set, then all lanes are empty
+	mov     _unused_lanes(state), unused_lanes
+	bt      $32+3, unused_lanes
+	jc      return_null
+
+	# find a lane with a non-null job
+	xor     idx, idx
+	offset = (_ldata + 1 * _LANE_DATA_size + _job_in_lane)
+	cmpq    $0, offset(state)
+	cmovne  one(%rip), idx
+	offset = (_ldata + 2 * _LANE_DATA_size + _job_in_lane)
+	cmpq    $0, offset(state)
+	cmovne  two(%rip), idx
+	offset = (_ldata + 3 * _LANE_DATA_size + _job_in_lane)
+	cmpq    $0, offset(state)
+	cmovne  three(%rip), idx
+	offset = (_ldata + 4 * _LANE_DATA_size + _job_in_lane)
+	cmpq    $0, offset(state)
+	cmovne  four(%rip), idx
+	offset = (_ldata + 5 * _LANE_DATA_size + _job_in_lane)
+	cmpq    $0, offset(state)
+	cmovne  five(%rip), idx
+	offset = (_ldata + 6 * _LANE_DATA_size + _job_in_lane)
+	cmpq    $0, offset(state)
+	cmovne  six(%rip), idx
+	offset = (_ldata + 7 * _LANE_DATA_size + _job_in_lane)
+	cmpq    $0, offset(state)
+	cmovne  seven(%rip), idx
+
+	# copy idx to empty lanes
+copy_lane_data:
+	offset =  (_args + _data_ptr)
+	mov     offset(state,idx,8), tmp
+
+	I = 0
+.rep 8
+	offset =  (_ldata + I * _LANE_DATA_size + _job_in_lane)
+	cmpq    $0, offset(state)
+.altmacro
+	JNE_SKIP %I
+	offset =  (_args + _data_ptr + 8*I)
+	mov     tmp, offset(state)
+	offset =  (_lens + 4*I)
+	movl    $0xFFFFFFFF, offset(state)
+LABEL skip_ %I
+	I = (I+1)
+.noaltmacro
+.endr
+
+	# Find min length
+	vmovdqa _lens+0*16(state), %xmm0
+	vmovdqa _lens+1*16(state), %xmm1
+
+	vpminud %xmm1, %xmm0, %xmm2     # xmm2 has {D,C,B,A}
+	vpalignr $8, %xmm2, %xmm3, %xmm3   # xmm3 has {x,x,D,C}
+	vpminud %xmm3, %xmm2, %xmm2        # xmm2 has {x,x,E,F}
+	vpalignr $4, %xmm2, %xmm3, %xmm3    # xmm3 has {x,x,x,E}
+	vpminud %xmm3, %xmm2, %xmm2        # xmm2 has min value in low dword
+
+	vmovd   %xmm2, DWORD_idx
+	mov	idx, len2
+	and	$0xF, idx
+	shr	$4, len2
+	jz	len_is_0
+
+	vpand   clear_low_nibble(%rip), %xmm2, %xmm2
+	vpshufd $0, %xmm2, %xmm2
+
+	vpsubd  %xmm2, %xmm0, %xmm0
+	vpsubd  %xmm2, %xmm1, %xmm1
+
+	vmovdqa %xmm0, _lens+0*16(state)
+	vmovdqa %xmm1, _lens+1*16(state)
+
+	# "state" and "args" are the same address, arg1
+	# len is arg2
+	call	sha1_x8_avx2
+	# state and idx are intact
+
+
+len_is_0:
+	# process completed job "idx"
+	imul    $_LANE_DATA_size, idx, lane_data
+	lea     _ldata(state, lane_data), lane_data
+
+	mov     _job_in_lane(lane_data), job_rax
+	movq    $0, _job_in_lane(lane_data)
+	movl    $STS_COMPLETED, _status(job_rax)
+	mov     _unused_lanes(state), unused_lanes
+	shl     $4, unused_lanes
+	or      idx, unused_lanes
+	mov     unused_lanes, _unused_lanes(state)
+
+	movl	$0xFFFFFFFF, _lens(state, idx, 4)
+
+	vmovd    _args_digest(state , idx, 4) , %xmm0
+	vpinsrd  $1, _args_digest+1*32(state, idx, 4), %xmm0, %xmm0
+	vpinsrd  $2, _args_digest+2*32(state, idx, 4), %xmm0, %xmm0
+	vpinsrd  $3, _args_digest+3*32(state, idx, 4), %xmm0, %xmm0
+	movl    _args_digest+4*32(state, idx, 4), tmp2_w
+
+	vmovdqu  %xmm0, _result_digest(job_rax)
+	offset =  (_result_digest + 1*16)
+	mov     tmp2_w, offset(job_rax)
+
+return:
+
+	mov     _GPR_SAVE(%rsp), %rbx
+	mov     _GPR_SAVE+8*1(%rsp), %r10 #saved rsp
+	mov	_GPR_SAVE+8*3(%rsp), %rbp
+	mov	_GPR_SAVE+8*4(%rsp), %r12
+	mov	_GPR_SAVE+8*5(%rsp), %r13
+	mov	_GPR_SAVE+8*6(%rsp), %r14
+	mov	_GPR_SAVE+8*7(%rsp), %r15
+	mov     %r10, %rsp
+
+	ret
+
+return_null:
+	xor     job_rax, job_rax
+	jmp     return
+ENDPROC(sha1_mb_mgr_flush_avx2)
+
+
+#################################################################
+
+.align 16
+ENTRY(sha1_mb_mgr_get_comp_job_avx2)
+	push    %rbx
+
+	## if bit 32+3 is set, then all lanes are empty
+	mov     _unused_lanes(state), unused_lanes
+	bt      $(32+3), unused_lanes
+	jc      .return_null
+
+	# Find min length
+	vmovdqa _lens(state), %xmm0
+	vmovdqa _lens+1*16(state), %xmm1
+
+	vpminud %xmm1, %xmm0, %xmm2        # xmm2 has {D,C,B,A}
+	vpalignr $8, %xmm2, %xmm3, %xmm3   # xmm3 has {x,x,D,C}
+	vpminud %xmm3, %xmm2, %xmm2        # xmm2 has {x,x,E,F}
+	vpalignr $4, %xmm2, %xmm3, %xmm3    # xmm3 has {x,x,x,E}
+	vpminud %xmm3, %xmm2, %xmm2        # xmm2 has min value in low dword
+
+	vmovd   %xmm2, DWORD_idx
+	test    $~0xF, idx
+	jnz     .return_null
+
+	# process completed job "idx"
+	imul    $_LANE_DATA_size, idx, lane_data
+	lea     _ldata(state, lane_data), lane_data
+
+	mov     _job_in_lane(lane_data), job_rax
+	movq    $0,  _job_in_lane(lane_data)
+	movl    $STS_COMPLETED, _status(job_rax)
+	mov     _unused_lanes(state), unused_lanes
+	shl     $4, unused_lanes
+	or      idx, unused_lanes
+	mov     unused_lanes, _unused_lanes(state)
+
+	movl    $0xFFFFFFFF, _lens(state,  idx, 4)
+
+	vmovd   _args_digest(state, idx, 4), %xmm0
+	vpinsrd $1, _args_digest+1*32(state, idx, 4), %xmm0, %xmm0
+	vpinsrd $2, _args_digest+2*32(state, idx, 4), %xmm0, %xmm0
+	vpinsrd $3, _args_digest+3*32(state, idx, 4), %xmm0, %xmm0
+	movl    _args_digest+4*32(state, idx, 4), tmp2_w
+
+	vmovdqu %xmm0, _result_digest(job_rax)
+	movl    tmp2_w, _result_digest+1*16(job_rax)
+
+	pop     %rbx
+
+	ret
+
+.return_null:
+	xor     job_rax, job_rax
+	pop     %rbx
+	ret
+ENDPROC(sha1_mb_mgr_get_comp_job_avx2)
+
+.data
+
+.align 16
+clear_low_nibble:
+.octa	0x000000000000000000000000FFFFFFF0
+one:
+.quad  1
+two:
+.quad  2
+three:
+.quad  3
+four:
+.quad  4
+five:
+.quad  5
+six:
+.quad  6
+seven:
+.quad  7

228 lines not on this page · raw message

new file mode 100644
index 0000000..8e1b477
@@ -0,0 +1,472 @@
+/*
+ * Multi-buffer SHA1 algorithm hash compute routine
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ *  Copyright(c) 2014 Intel Corporation.
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of version 2 of the GNU General Public License as
+ *  published by the Free Software Foundation.
+ *
+ *  This program is distributed in the hope that it will be useful, but
+ *  WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ *  General Public License for more details.
+ *
+ *  Contact Information:
+ *      James Guilford <[email protected]>
+ *	Tim Chen <[email protected]>
+ *
+ *  BSD LICENSE
+ *
+ *  Copyright(c) 2014 Intel Corporation.
+ *
+ *  Redistribution and use in source and binary forms, with or without
+ *  modification, are permitted provided that the following conditions
+ *  are met:
+ *
+ *    * Redistributions of source code must retain the above copyright
+ *      notice, this list of conditions and the following disclaimer.
+ *    * Redistributions in binary form must reproduce the above copyright
+ *      notice, this list of conditions and the following disclaimer in
+ *      the documentation and/or other materials provided with the
+ *      distribution.
+ *    * Neither the name of Intel Corporation nor the names of its
+ *      contributors may be used to endorse or promote products derived
+ *      from this software without specific prior written permission.
+ *
+ *  THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *  "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *  LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *  A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *  OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *  SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *  LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *  DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *  THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *  (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <linux/linkage.h>
+#include "sha1_mb_mgr_datastruct.S"
+
+## code to compute oct SHA1 using SSE-256
+## outer calling routine takes care of save and restore of XMM registers
+
+## Function clobbers: rax, rcx, rdx,   rbx, rsi, rdi, r9-r15# ymm0-15
+##
+## Linux clobbers:    rax rbx rcx rdx rsi            r9 r10 r11 r12 r13 r14 r15
+## Linux preserves:                       rdi rbp r8
+##
+## clobbers ymm0-15
+
+
+# TRANSPOSE8 r0, r1, r2, r3, r4, r5, r6, r7, t0, t1
+# "transpose" data in {r0...r7} using temps {t0...t1}
+# Input looks like: {r0 r1 r2 r3 r4 r5 r6 r7}
+# r0 = {a7 a6 a5 a4   a3 a2 a1 a0}
+# r1 = {b7 b6 b5 b4   b3 b2 b1 b0}
+# r2 = {c7 c6 c5 c4   c3 c2 c1 c0}
+# r3 = {d7 d6 d5 d4   d3 d2 d1 d0}
+# r4 = {e7 e6 e5 e4   e3 e2 e1 e0}
+# r5 = {f7 f6 f5 f4   f3 f2 f1 f0}
+# r6 = {g7 g6 g5 g4   g3 g2 g1 g0}
+# r7 = {h7 h6 h5 h4   h3 h2 h1 h0}
+#
+# Output looks like: {r0 r1 r2 r3 r4 r5 r6 r7}
+# r0 = {h0 g0 f0 e0   d0 c0 b0 a0}
+# r1 = {h1 g1 f1 e1   d1 c1 b1 a1}
+# r2 = {h2 g2 f2 e2   d2 c2 b2 a2}
+# r3 = {h3 g3 f3 e3   d3 c3 b3 a3}
+# r4 = {h4 g4 f4 e4   d4 c4 b4 a4}
+# r5 = {h5 g5 f5 e5   d5 c5 b5 a5}
+# r6 = {h6 g6 f6 e6   d6 c6 b6 a6}
+# r7 = {h7 g7 f7 e7   d7 c7 b7 a7}
+#
+
+.macro TRANSPOSE8 r0 r1 r2 r3 r4 r5 r6 r7 t0 t1
+	# process top half (r0..r3) {a...d}
+	vshufps  $0x44, \r1, \r0, \t0 # t0 = {b5 b4 a5 a4   b1 b0 a1 a0}
+	vshufps  $0xEE, \r1, \r0, \r0 # r0 = {b7 b6 a7 a6   b3 b2 a3 a2}
+	vshufps  $0x44, \r3, \r2, \t1 # t1 = {d5 d4 c5 c4   d1 d0 c1 c0}
+	vshufps  $0xEE, \r3, \r2, \r2 # r2 = {d7 d6 c7 c6   d3 d2 c3 c2}
+	vshufps  $0xDD, \t1, \t0, \r3 # r3 = {d5 c5 b5 a5   d1 c1 b1 a1}
+	vshufps  $0x88, \r2, \r0, \r1 # r1 = {d6 c6 b6 a6   d2 c2 b2 a2}
+	vshufps  $0xDD, \r2, \r0, \r0 # r0 = {d7 c7 b7 a7   d3 c3 b3 a3}
+	vshufps  $0x88, \t1, \t0, \t0 # t0 = {d4 c4 b4 a4   d0 c0 b0 a0}
+
+	# use r2 in place of t0
+	# process bottom half (r4..r7) {e...h}
+	vshufps  $0x44, \r5, \r4, \r2 # r2 = {f5 f4 e5 e4   f1 f0 e1 e0}
+	vshufps  $0xEE, \r5, \r4, \r4 # r4 = {f7 f6 e7 e6   f3 f2 e3 e2}
+	vshufps  $0x44, \r7, \r6, \t1 # t1 = {h5 h4 g5 g4   h1 h0 g1 g0}
+	vshufps  $0xEE, \r7, \r6, \r6 # r6 = {h7 h6 g7 g6   h3 h2 g3 g2}
+	vshufps  $0xDD, \t1, \r2, \r7 # r7 = {h5 g5 f5 e5   h1 g1 f1 e1}
+	vshufps  $0x88, \r6, \r4, \r5 # r5 = {h6 g6 f6 e6   h2 g2 f2 e2}
+	vshufps  $0xDD, \r6, \r4, \r4 # r4 = {h7 g7 f7 e7   h3 g3 f3 e3}
+	vshufps  $0x88, \t1, \r2, \t1 # t1 = {h4 g4 f4 e4   h0 g0 f0 e0}
+
+	vperm2f128      $0x13, \r1, \r5, \r6  # h6...a6
+	vperm2f128      $0x02, \r1, \r5, \r2  # h2...a2
+	vperm2f128      $0x13, \r3, \r7, \r5  # h5...a5
+	vperm2f128      $0x02, \r3, \r7, \r1  # h1...a1
+	vperm2f128      $0x13, \r0, \r4, \r7  # h7...a7
+	vperm2f128      $0x02, \r0, \r4, \r3  # h3...a3
+	vperm2f128      $0x13, \t0, \t1, \r4  # h4...a4
+	vperm2f128      $0x02, \t0, \t1, \r0  # h0...a0
+
+.endm
+##
+## Magic functions defined in FIPS 180-1
+##
+# macro MAGIC_F0 F,B,C,D,T   ## F = (D ^ (B & (C ^ D)))
+.macro MAGIC_F0 regF regB regC regD regT
+    vpxor \regD, \regC, \regF
+    vpand \regB, \regF, \regF
+    vpxor \regD, \regF, \regF
+.endm
+
+# macro MAGIC_F1 F,B,C,D,T   ## F = (B ^ C ^ D)
+.macro MAGIC_F1 regF regB regC regD regT
+    vpxor  \regC, \regD, \regF
+    vpxor  \regB, \regF, \regF
+.endm
+
+# macro MAGIC_F2 F,B,C,D,T   ## F = ((B & C) | (B & D) | (C & D))
+.macro MAGIC_F2 regF regB regC regD regT
+    vpor  \regC, \regB, \regF
+    vpand \regC, \regB, \regT
+    vpand \regD, \regF, \regF
+    vpor  \regT, \regF, \regF
+.endm
+
+# macro MAGIC_F3 F,B,C,D,T   ## F = (B ^ C ^ D)
+.macro MAGIC_F3 regF regB regC regD regT
+    MAGIC_F1 \regF,\regB,\regC,\regD,\regT
+.endm
+
+# PROLD reg, imm, tmp
+.macro PROLD reg imm tmp
+	vpsrld  $(32-\imm), \reg, \tmp
+	vpslld  $\imm, \reg, \reg
+	vpor    \tmp, \reg, \reg
+.endm
+
+.macro PROLD_nd reg imm tmp src
+	vpsrld  $(32-\imm), \src, \tmp
+	vpslld  $\imm, \src, \reg
+	vpor	\tmp, \reg, \reg
+.endm
+
+.macro SHA1_STEP_00_15 regA regB regC regD regE regT regF memW immCNT MAGIC
+	vpaddd	\immCNT, \regE, \regE
+	vpaddd	\memW*32(%rsp), \regE, \regE
+	PROLD_nd \regT, 5, \regF, \regA
+	vpaddd	\regT, \regE, \regE
+	\MAGIC  \regF, \regB, \regC, \regD, \regT
+        PROLD   \regB, 30, \regT
+        vpaddd  \regF, \regE, \regE
+.endm
+
+.macro SHA1_STEP_16_79 regA regB regC regD regE regT regF memW immCNT MAGIC
+	vpaddd	\immCNT, \regE, \regE
+	offset = ((\memW - 14) & 15) * 32
+	vmovdqu offset(%rsp), W14
+	vpxor	W14, W16, W16
+	offset = ((\memW -  8) & 15) * 32
+	vpxor	offset(%rsp), W16, W16
+	offset = ((\memW -  3) & 15) * 32
+	vpxor	offset(%rsp), W16, W16
+	vpsrld	$(32-1), W16, \regF
+	vpslld	$1, W16, W16
+	vpor	W16, \regF, \regF
+
+	ROTATE_W
+
+	offset = ((\memW - 0) & 15) * 32
+	vmovdqu	\regF, offset(%rsp)
+	vpaddd	\regF, \regE, \regE
+	PROLD_nd \regT, 5, \regF, \regA
+	vpaddd	\regT, \regE, \regE
+	\MAGIC \regF,\regB,\regC,\regD,\regT      ## FUN  = MAGIC_Fi(B,C,D)
+	PROLD   \regB,30, \regT
+	vpaddd  \regF, \regE, \regE
+.endm
+
+########################################################################
+########################################################################
+########################################################################
+
+## FRAMESZ plus pushes must be an odd multiple of 8
+YMM_SAVE = (15-15)*32
+FRAMESZ = 32*16 + YMM_SAVE
+_YMM  =   FRAMESZ - YMM_SAVE
+
+#define VMOVPS   vmovups
+
+IDX  = %rax
+inp0 = %r9
+inp1 = %r10
+inp2 = %r11
+inp3 = %r12
+inp4 = %r13
+inp5 = %r14
+inp6 = %r15
+inp7 = %rcx
+arg1 = %rdi
+arg2 = %rsi
+RSP_SAVE = %rdx
+
+# ymm0 A
+# ymm1 B
+# ymm2 C
+# ymm3 D
+# ymm4 E
+# ymm5         F       AA
+# ymm6         T0      BB
+# ymm7         T1      CC
+# ymm8         T2      DD
+# ymm9         T3      EE
+# ymm10                T4      TMP
+# ymm11                T5      FUN
+# ymm12                T6      K
+# ymm13                T7      W14
+# ymm14                T8      W15
+# ymm15                T9      W16
+
+
+A  =     %ymm0
+B  =     %ymm1
+C  =     %ymm2
+D  =     %ymm3
+E  =     %ymm4
+F  =     %ymm5
+T0 =	 %ymm6
+T1 =     %ymm7
+T2 =     %ymm8
+T3 =     %ymm9
+T4 =     %ymm10
+T5 =     %ymm11
+T6 =     %ymm12
+T7 =     %ymm13
+T8  =     %ymm14
+T9  =     %ymm15
+
+AA  =     %ymm5
+BB  =     %ymm6
+CC  =     %ymm7
+DD  =     %ymm8
+EE  =     %ymm9
+TMP =     %ymm10
+FUN =     %ymm11
+K   =     %ymm12
+W14 =     %ymm13
+W15 =     %ymm14
+W16 =     %ymm15
+
+.macro ROTATE_ARGS
+ TMP_ = E
+ E = D
+ D = C
+ C = B
+ B = A
+ A = TMP_
+.endm
+
+.macro ROTATE_W
+TMP_  = W16
+W16  = W15
+W15  = W14
+W14  = TMP_
+.endm
+
+# 8 streams x 5 32bit words per digest x 4 bytes per word
+#define DIGEST_SIZE (8*5*4)
+
+.align 32
+
+# void sha1_x8_avx2(void **input_data, UINT128 *digest, UINT32 size)
+# arg 1 : pointer to array[4] of pointer to input data
+# arg 2 : size (in blocks) ;; assumed to be >= 1
+#
+ENTRY(sha1_x8_avx2)
+
+	push	RSP_SAVE
+
+	#save rsp
+	mov	%rsp, RSP_SAVE
+	sub     $FRAMESZ, %rsp
+
+	#align rsp to 32 Bytes
+	and	$~0x1F, %rsp
+
+	## Initialize digests
+	vmovdqu  0*32(arg1), A
+	vmovdqu  1*32(arg1), B
+	vmovdqu  2*32(arg1), C
+	vmovdqu  3*32(arg1), D
+	vmovdqu  4*32(arg1), E
+
+	## transpose input onto stack
+	mov     _data_ptr+0*8(arg1),inp0
+	mov     _data_ptr+1*8(arg1),inp1
+	mov     _data_ptr+2*8(arg1),inp2
+	mov     _data_ptr+3*8(arg1),inp3
+	mov     _data_ptr+4*8(arg1),inp4
+	mov     _data_ptr+5*8(arg1),inp5
+	mov     _data_ptr+6*8(arg1),inp6
+	mov     _data_ptr+7*8(arg1),inp7
+
+	xor     IDX, IDX
+lloop:
+	vmovdqu  PSHUFFLE_BYTE_FLIP_MASK(%rip), F
+	I=0
+.rep 2
+	VMOVPS   (inp0, IDX), T0
+	VMOVPS   (inp1, IDX), T1
+	VMOVPS   (inp2, IDX), T2
+	VMOVPS   (inp3, IDX), T3
+	VMOVPS   (inp4, IDX), T4
+	VMOVPS   (inp5, IDX), T5
+	VMOVPS   (inp6, IDX), T6
+	VMOVPS   (inp7, IDX), T7
+
+	TRANSPOSE8       T0, T1, T2, T3, T4, T5, T6, T7, T8, T9
+	vpshufb  F, T0, T0
+	vmovdqu  T0, (I*8)*32(%rsp)
+	vpshufb  F, T1, T1
+	vmovdqu  T1, (I*8+1)*32(%rsp)
+	vpshufb  F, T2, T2
+	vmovdqu  T2, (I*8+2)*32(%rsp)
+	vpshufb  F, T3, T3
+	vmovdqu  T3, (I*8+3)*32(%rsp)
+	vpshufb  F, T4, T4
+	vmovdqu  T4, (I*8+4)*32(%rsp)
+	vpshufb  F, T5, T5
+	vmovdqu  T5, (I*8+5)*32(%rsp)
+	vpshufb  F, T6, T6
+	vmovdqu  T6, (I*8+6)*32(%rsp)
+	vpshufb  F, T7, T7
+	vmovdqu  T7, (I*8+7)*32(%rsp)
+	add     $32, IDX
+	I = (I+1)
+.endr
+	# save old digests
+	vmovdqu  A,AA
+	vmovdqu  B,BB
+	vmovdqu  C,CC
+	vmovdqu  D,DD
+	vmovdqu  E,EE
+
+##
+## perform 0-79 steps
+##
+	vmovdqu  K00_19(%rip), K
+## do rounds 0...15
+	I = 0
+.rep 16
+	SHA1_STEP_00_15 A,B,C,D,E, TMP,FUN, I, K, MAGIC_F0
+	ROTATE_ARGS
+	I = (I+1)
+.endr
+
+## do rounds 16...19
+	vmovdqu  ((16 - 16) & 15) * 32 (%rsp), W16
+	vmovdqu  ((16 - 15) & 15) * 32 (%rsp), W15
+.rep 4
+	SHA1_STEP_16_79 A,B,C,D,E, TMP,FUN, I, K, MAGIC_F0
+	ROTATE_ARGS
+	I = (I+1)
+.endr
+
+## do rounds 20...39
+	vmovdqu  K20_39(%rip), K
+.rep 20
+	SHA1_STEP_16_79 A,B,C,D,E, TMP,FUN, I, K, MAGIC_F1
+	ROTATE_ARGS
+	I = (I+1)
+.endr
+
+## do rounds 40...59
+	vmovdqu  K40_59(%rip), K
+.rep 20
+	SHA1_STEP_16_79 A,B,C,D,E, TMP,FUN, I, K, MAGIC_F2
+	ROTATE_ARGS
+	I = (I+1)
+.endr
+
+## do rounds 60...79
+	vmovdqu  K60_79(%rip), K
+.rep 20
+	SHA1_STEP_16_79 A,B,C,D,E, TMP,FUN, I, K, MAGIC_F3
+	ROTATE_ARGS
+	I = (I+1)
+.endr
+
+	vpaddd   AA,A,A
+	vpaddd   BB,B,B
+	vpaddd   CC,C,C
+	vpaddd   DD,D,D
+	vpaddd   EE,E,E
+
+	sub     $1, arg2
+	jne     lloop
+
+	# write out digests
+	vmovdqu  A, 0*32(arg1)
+	vmovdqu  B, 1*32(arg1)
+	vmovdqu  C, 2*32(arg1)
+	vmovdqu  D, 3*32(arg1)
+	vmovdqu  E, 4*32(arg1)
+
+	# update input pointers
+	add     IDX, inp0
+	add     IDX, inp1
+	add     IDX, inp2
+	add     IDX, inp3
+	add     IDX, inp4
+	add     IDX, inp5
+	add     IDX, inp6
+	add     IDX, inp7
+	mov     inp0, _data_ptr (arg1)
+	mov     inp1, _data_ptr + 1*8(arg1)
+	mov     inp2, _data_ptr + 2*8(arg1)
+	mov     inp3, _data_ptr + 3*8(arg1)
+	mov     inp4, _data_ptr + 4*8(arg1)
+	mov     inp5, _data_ptr + 5*8(arg1)
+	mov     inp6, _data_ptr + 6*8(arg1)
+	mov     inp7, _data_ptr + 7*8(arg1)
+
+	################
+	## Postamble
+
+	mov     RSP_SAVE, %rsp
+	pop	RSP_SAVE
+
+	ret
+ENDPROC(sha1_x8_avx2)
+
+
+.data
+
+.align 32
+K00_19:
+.octa 0x5A8279995A8279995A8279995A827999
+.octa 0x5A8279995A8279995A8279995A827999
+K20_39:
+.octa 0x6ED9EBA16ED9EBA16ED9EBA16ED9EBA1
+.octa 0x6ED9EBA16ED9EBA16ED9EBA16ED9EBA1
+K40_59:
+.octa 0x8F1BBCDC8F1BBCDC8F1BBCDC8F1BBCDC
+.octa 0x8F1BBCDC8F1BBCDC8F1BBCDC8F1BBCDC
+K60_79:
+.octa 0xCA62C1D6CA62C1D6CA62C1D6CA62C1D6
+.octa 0xCA62C1D6CA62C1D6CA62C1D6CA62C1D6
+PSHUFFLE_BYTE_FLIP_MASK:
+.octa 0x0c0d0e0f08090a0b0405060700010203
+.octa 0x0c0d0e0f08090a0b0405060700010203

index 61d6e28..1189267 100644
@@ -25,6 +25,7 @@ obj-$(CONFIG_CRYPTO_GHASH_CLMUL_NI_INTEL) += ghash-clmulni-intel.o
 
 obj-$(CONFIG_CRYPTO_CRC32C_INTEL) += crc32c-intel.o
 obj-$(CONFIG_CRYPTO_SHA1_SSSE3) += sha1-ssse3.o
+obj-$(CONFIG_CRYPTO_SHA1_MB) += sha-mb/
 obj-$(CONFIG_CRYPTO_CRC32_PCLMUL) += crc32-pclmul.o
 obj-$(CONFIG_CRYPTO_SHA256_SSSE3) += sha256-ssse3.o
 obj-$(CONFIG_CRYPTO_SHA512_SSSE3) += sha512-ssse3.o

@@ -88,3 +89,4 @@ crc32-pclmul-y := crc32-pclmul_asm.o crc32-pclmul_glue.o
 sha256-ssse3-y := sha256-ssse3-asm.o sha256-avx-asm.o sha256-avx2-asm.o sha256_ssse3_glue.o
 sha512-ssse3-y := sha512-ssse3-asm.o sha512-avx-asm.o sha512-avx2-asm.o sha512_ssse3_glue.o
 crct10dif-pclmul-y := crct10dif-pcl-asm_64.o crct10dif-pclmul_glue.o
+

new file mode 100644
index 0000000..2f87563
@@ -0,0 +1,11 @@
+#
+# Arch-specific CryptoAPI modules.
+#
+
+avx2_supported := $(call as-instr,vpgatherdd %ymm0$(comma)(%eax$(comma)%ymm1\
+                                $(comma)4)$(comma)%ymm2,yes,no)
+ifeq ($(avx2_supported),yes)
+	obj-$(CONFIG_CRYPTO_SHA1_MB) += sha1-mb.o
+	sha1-mb-y := sha1_mb.o sha1_mb_mgr_flush_avx2.o \
+	     sha1_mb_mgr_init_avx2.o sha1_mb_mgr_submit_avx2.o sha1_x8_avx2.o
+endif