OpenCloudOS-Kernel/arch/x86/crypto/sha256-ssse3-asm.S

514 lines
17 KiB
ArmAsm
Raw Normal View History

########################################################################
# Implement fast SHA-256 with SSSE3 instructions. (x86_64)
#
# Copyright (C) 2013 Intel Corporation.
#
# Authors:
# James Guilford <james.guilford@intel.com>
# Kirk Yap <kirk.s.yap@intel.com>
# Tim Chen <tim.c.chen@linux.intel.com>
#
# This software is available to you under a choice of one of two
# licenses. You may choose to be licensed under the terms of the GNU
# General Public License (GPL) Version 2, available from the file
# COPYING in the main directory of this source tree, or the
# OpenIB.org BSD license below:
#
# 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.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
# BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
# ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
#
########################################################################
#
# This code is described in an Intel White-Paper:
# "Fast SHA-256 Implementations on Intel Architecture Processors"
#
# To find it, surf to http://www.intel.com/p/en_US/embedded
# and search for that title.
#
########################################################################
#include <linux/linkage.h>
## assume buffers not aligned
#define MOVDQ movdqu
################################ Define Macros
# addm [mem], reg
# Add reg to mem using reg-mem add and store
.macro addm p1 p2
add \p1, \p2
mov \p2, \p1
.endm
################################
# COPY_XMM_AND_BSWAP xmm, [mem], byte_flip_mask
# Load xmm with mem and byte swap each dword
.macro COPY_XMM_AND_BSWAP p1 p2 p3
MOVDQ \p2, \p1
pshufb \p3, \p1
.endm
################################
X0 = %xmm4
X1 = %xmm5
X2 = %xmm6
X3 = %xmm7
XTMP0 = %xmm0
XTMP1 = %xmm1
XTMP2 = %xmm2
XTMP3 = %xmm3
XTMP4 = %xmm8
XFER = %xmm9
SHUF_00BA = %xmm10 # shuffle xBxA -> 00BA
SHUF_DC00 = %xmm11 # shuffle xDxC -> DC00
BYTE_FLIP_MASK = %xmm12
NUM_BLKS = %rdx # 3rd arg
INP = %rsi # 2nd arg
CTX = %rdi # 1st arg
SRND = %rsi # clobbers INP
c = %ecx
d = %r8d
e = %edx
TBL = %r12
a = %eax
b = %ebx
f = %r9d
g = %r10d
h = %r11d
y0 = %r13d
y1 = %r14d
y2 = %r15d
_INP_END_SIZE = 8
_INP_SIZE = 8
_XFER_SIZE = 16
_XMM_SAVE_SIZE = 0
_INP_END = 0
_INP = _INP_END + _INP_END_SIZE
_XFER = _INP + _INP_SIZE
_XMM_SAVE = _XFER + _XFER_SIZE
STACK_SIZE = _XMM_SAVE + _XMM_SAVE_SIZE
# rotate_Xs
# Rotate values of symbols X0...X3
.macro rotate_Xs
X_ = X0
X0 = X1
X1 = X2
X2 = X3
X3 = X_
.endm
# ROTATE_ARGS
# Rotate values of symbols a...h
.macro ROTATE_ARGS
TMP_ = h
h = g
g = f
f = e
e = d
d = c
c = b
b = a
a = TMP_
.endm
.macro FOUR_ROUNDS_AND_SCHED
## compute s0 four at a time and s1 two at a time
## compute W[-16] + W[-7] 4 at a time
movdqa X3, XTMP0
mov e, y0 # y0 = e
ror $(25-11), y0 # y0 = e >> (25-11)
mov a, y1 # y1 = a
palignr $4, X2, XTMP0 # XTMP0 = W[-7]
ror $(22-13), y1 # y1 = a >> (22-13)
xor e, y0 # y0 = e ^ (e >> (25-11))
mov f, y2 # y2 = f
ror $(11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
movdqa X1, XTMP1
xor a, y1 # y1 = a ^ (a >> (22-13)
xor g, y2 # y2 = f^g
paddd X0, XTMP0 # XTMP0 = W[-7] + W[-16]
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
and e, y2 # y2 = (f^g)&e
ror $(13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
## compute s0
palignr $4, X0, XTMP1 # XTMP1 = W[-15]
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
ror $6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
xor g, y2 # y2 = CH = ((f^g)&e)^g
movdqa XTMP1, XTMP2 # XTMP2 = W[-15]
ror $2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
add y0, y2 # y2 = S1 + CH
add _XFER(%rsp) , y2 # y2 = k + w + S1 + CH
movdqa XTMP1, XTMP3 # XTMP3 = W[-15]
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
pslld $(32-7), XTMP1 #
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
psrld $7, XTMP2 #
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
por XTMP2, XTMP1 # XTMP1 = W[-15] ror 7
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
#
ROTATE_ARGS #
movdqa XTMP3, XTMP2 # XTMP2 = W[-15]
mov e, y0 # y0 = e
mov a, y1 # y1 = a
movdqa XTMP3, XTMP4 # XTMP4 = W[-15]
ror $(25-11), y0 # y0 = e >> (25-11)
xor e, y0 # y0 = e ^ (e >> (25-11))
mov f, y2 # y2 = f
ror $(22-13), y1 # y1 = a >> (22-13)
pslld $(32-18), XTMP3 #
xor a, y1 # y1 = a ^ (a >> (22-13)
ror $(11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
xor g, y2 # y2 = f^g
psrld $18, XTMP2 #
ror $(13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
and e, y2 # y2 = (f^g)&e
ror $6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
pxor XTMP3, XTMP1
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
xor g, y2 # y2 = CH = ((f^g)&e)^g
psrld $3, XTMP4 # XTMP4 = W[-15] >> 3
add y0, y2 # y2 = S1 + CH
add (1*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
ror $2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
pxor XTMP2, XTMP1 # XTMP1 = W[-15] ror 7 ^ W[-15] ror 18
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
pxor XTMP4, XTMP1 # XTMP1 = s0
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
## compute low s1
pshufd $0b11111010, X3, XTMP2 # XTMP2 = W[-2] {BBAA}
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
paddd XTMP1, XTMP0 # XTMP0 = W[-16] + W[-7] + s0
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
movdqa XTMP2, XTMP3 # XTMP3 = W[-2] {BBAA}
mov e, y0 # y0 = e
mov a, y1 # y1 = a
ror $(25-11), y0 # y0 = e >> (25-11)
movdqa XTMP2, XTMP4 # XTMP4 = W[-2] {BBAA}
xor e, y0 # y0 = e ^ (e >> (25-11))
ror $(22-13), y1 # y1 = a >> (22-13)
mov f, y2 # y2 = f
xor a, y1 # y1 = a ^ (a >> (22-13)
ror $(11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
psrlq $17, XTMP2 # XTMP2 = W[-2] ror 17 {xBxA}
xor g, y2 # y2 = f^g
psrlq $19, XTMP3 # XTMP3 = W[-2] ror 19 {xBxA}
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
and e, y2 # y2 = (f^g)&e
psrld $10, XTMP4 # XTMP4 = W[-2] >> 10 {BBAA}
ror $(13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
xor g, y2 # y2 = CH = ((f^g)&e)^g
ror $6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
pxor XTMP3, XTMP2
add y0, y2 # y2 = S1 + CH
ror $2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
add (2*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
pxor XTMP2, XTMP4 # XTMP4 = s1 {xBxA}
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
pshufb SHUF_00BA, XTMP4 # XTMP4 = s1 {00BA}
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
paddd XTMP4, XTMP0 # XTMP0 = {..., ..., W[1], W[0]}
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
## compute high s1
pshufd $0b01010000, XTMP0, XTMP2 # XTMP2 = W[-2] {BBAA}
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
#
ROTATE_ARGS #
movdqa XTMP2, XTMP3 # XTMP3 = W[-2] {DDCC}
mov e, y0 # y0 = e
ror $(25-11), y0 # y0 = e >> (25-11)
mov a, y1 # y1 = a
movdqa XTMP2, X0 # X0 = W[-2] {DDCC}
ror $(22-13), y1 # y1 = a >> (22-13)
xor e, y0 # y0 = e ^ (e >> (25-11))
mov f, y2 # y2 = f
ror $(11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
psrlq $17, XTMP2 # XTMP2 = W[-2] ror 17 {xDxC}
xor a, y1 # y1 = a ^ (a >> (22-13)
xor g, y2 # y2 = f^g
psrlq $19, XTMP3 # XTMP3 = W[-2] ror 19 {xDxC}
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25
and e, y2 # y2 = (f^g)&e
ror $(13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
psrld $10, X0 # X0 = W[-2] >> 10 {DDCC}
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22
ror $6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>2
xor g, y2 # y2 = CH = ((f^g)&e)^g
pxor XTMP3, XTMP2 #
ror $2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>2
add y0, y2 # y2 = S1 + CH
add (3*4 + _XFER)(%rsp), y2 # y2 = k + w + S1 + CH
pxor XTMP2, X0 # X0 = s1 {xDxC}
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
pshufb SHUF_DC00, X0 # X0 = s1 {DC00}
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
paddd XTMP0, X0 # X0 = {W[3], W[2], W[1], W[0]}
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
rotate_Xs
.endm
## input is [rsp + _XFER + %1 * 4]
.macro DO_ROUND round
mov e, y0 # y0 = e
ror $(25-11), y0 # y0 = e >> (25-11)
mov a, y1 # y1 = a
xor e, y0 # y0 = e ^ (e >> (25-11))
ror $(22-13), y1 # y1 = a >> (22-13)
mov f, y2 # y2 = f
xor a, y1 # y1 = a ^ (a >> (22-13)
ror $(11-6), y0 # y0 = (e >> (11-6)) ^ (e >> (25-6))
xor g, y2 # y2 = f^g
xor e, y0 # y0 = e ^ (e >> (11-6)) ^ (e >> (25-6))
ror $(13-2), y1 # y1 = (a >> (13-2)) ^ (a >> (22-2))
and e, y2 # y2 = (f^g)&e
xor a, y1 # y1 = a ^ (a >> (13-2)) ^ (a >> (22-2))
ror $6, y0 # y0 = S1 = (e>>6) & (e>>11) ^ (e>>25)
xor g, y2 # y2 = CH = ((f^g)&e)^g
add y0, y2 # y2 = S1 + CH
ror $2, y1 # y1 = S0 = (a>>2) ^ (a>>13) ^ (a>>22)
offset = \round * 4 + _XFER
add offset(%rsp), y2 # y2 = k + w + S1 + CH
mov a, y0 # y0 = a
add y2, h # h = h + S1 + CH + k + w
mov a, y2 # y2 = a
or c, y0 # y0 = a|c
add h, d # d = d + h + S1 + CH + k + w
and c, y2 # y2 = a&c
and b, y0 # y0 = (a|c)&b
add y1, h # h = h + S1 + CH + k + w + S0
or y2, y0 # y0 = MAJ = (a|c)&b)|(a&c)
add y0, h # h = h + S1 + CH + k + w + S0 + MAJ
ROTATE_ARGS
.endm
########################################################################
## void sha256_transform_ssse3(struct sha256_state *state, const u8 *data,
## int blocks);
## arg 1 : pointer to state
## (struct sha256_state is assumed to begin with u32 state[8])
## arg 2 : pointer to input data
## arg 3 : Num blocks
########################################################################
.text
x86/asm: Change all ENTRY+ENDPROC to SYM_FUNC_* These are all functions which are invoked from elsewhere, so annotate them as global using the new SYM_FUNC_START and their ENDPROC's by SYM_FUNC_END. Make sure ENTRY/ENDPROC is not defined on X86_64, given these were the last users. Signed-off-by: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Borislav Petkov <bp@suse.de> Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> [hibernate] Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> [xen bits] Acked-by: Herbert Xu <herbert@gondor.apana.org.au> [crypto] Cc: Allison Randal <allison@lohutok.net> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Andy Shevchenko <andy@infradead.org> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Armijn Hemel <armijn@tjaldur.nl> Cc: Cao jin <caoj.fnst@cn.fujitsu.com> Cc: Darren Hart <dvhart@infradead.org> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Enrico Weigelt <info@metux.net> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Herbert Xu <herbert@gondor.apana.org.au> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jim Mattson <jmattson@google.com> Cc: Joerg Roedel <joro@8bytes.org> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Kate Stewart <kstewart@linuxfoundation.org> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: kvm ML <kvm@vger.kernel.org> Cc: Len Brown <len.brown@intel.com> Cc: linux-arch@vger.kernel.org Cc: linux-crypto@vger.kernel.org Cc: linux-efi <linux-efi@vger.kernel.org> Cc: linux-efi@vger.kernel.org Cc: linux-pm@vger.kernel.org Cc: Mark Rutland <mark.rutland@arm.com> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: Peter Zijlstra <peterz@infradead.org> Cc: platform-driver-x86@vger.kernel.org Cc: "Radim Krčmář" <rkrcmar@redhat.com> Cc: Sean Christopherson <sean.j.christopherson@intel.com> Cc: Stefano Stabellini <sstabellini@kernel.org> Cc: "Steven Rostedt (VMware)" <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vitaly Kuznetsov <vkuznets@redhat.com> Cc: Wanpeng Li <wanpengli@tencent.com> Cc: Wei Huang <wei@redhat.com> Cc: x86-ml <x86@kernel.org> Cc: xen-devel@lists.xenproject.org Cc: Xiaoyao Li <xiaoyao.li@linux.intel.com> Link: https://lkml.kernel.org/r/20191011115108.12392-25-jslaby@suse.cz
2019-10-11 19:51:04 +08:00
SYM_FUNC_START(sha256_transform_ssse3)
.align 32
pushq %rbx
pushq %r12
pushq %r13
pushq %r14
pushq %r15
pushq %rbp
mov %rsp, %rbp
subq $STACK_SIZE, %rsp
and $~15, %rsp
shl $6, NUM_BLKS # convert to bytes
jz done_hash
add INP, NUM_BLKS
mov NUM_BLKS, _INP_END(%rsp) # pointer to end of data
## load initial digest
mov 4*0(CTX), a
mov 4*1(CTX), b
mov 4*2(CTX), c
mov 4*3(CTX), d
mov 4*4(CTX), e
mov 4*5(CTX), f
mov 4*6(CTX), g
mov 4*7(CTX), h
movdqa PSHUFFLE_BYTE_FLIP_MASK(%rip), BYTE_FLIP_MASK
movdqa _SHUF_00BA(%rip), SHUF_00BA
movdqa _SHUF_DC00(%rip), SHUF_DC00
loop0:
lea K256(%rip), TBL
## byte swap first 16 dwords
COPY_XMM_AND_BSWAP X0, 0*16(INP), BYTE_FLIP_MASK
COPY_XMM_AND_BSWAP X1, 1*16(INP), BYTE_FLIP_MASK
COPY_XMM_AND_BSWAP X2, 2*16(INP), BYTE_FLIP_MASK
COPY_XMM_AND_BSWAP X3, 3*16(INP), BYTE_FLIP_MASK
mov INP, _INP(%rsp)
## schedule 48 input dwords, by doing 3 rounds of 16 each
mov $3, SRND
.align 16
loop1:
movdqa (TBL), XFER
paddd X0, XFER
movdqa XFER, _XFER(%rsp)
FOUR_ROUNDS_AND_SCHED
movdqa 1*16(TBL), XFER
paddd X0, XFER
movdqa XFER, _XFER(%rsp)
FOUR_ROUNDS_AND_SCHED
movdqa 2*16(TBL), XFER
paddd X0, XFER
movdqa XFER, _XFER(%rsp)
FOUR_ROUNDS_AND_SCHED
movdqa 3*16(TBL), XFER
paddd X0, XFER
movdqa XFER, _XFER(%rsp)
add $4*16, TBL
FOUR_ROUNDS_AND_SCHED
sub $1, SRND
jne loop1
mov $2, SRND
loop2:
paddd (TBL), X0
movdqa X0, _XFER(%rsp)
DO_ROUND 0
DO_ROUND 1
DO_ROUND 2
DO_ROUND 3
paddd 1*16(TBL), X1
movdqa X1, _XFER(%rsp)
add $2*16, TBL
DO_ROUND 0
DO_ROUND 1
DO_ROUND 2
DO_ROUND 3
movdqa X2, X0
movdqa X3, X1
sub $1, SRND
jne loop2
addm (4*0)(CTX),a
addm (4*1)(CTX),b
addm (4*2)(CTX),c
addm (4*3)(CTX),d
addm (4*4)(CTX),e
addm (4*5)(CTX),f
addm (4*6)(CTX),g
addm (4*7)(CTX),h
mov _INP(%rsp), INP
add $64, INP
cmp _INP_END(%rsp), INP
jne loop0
done_hash:
mov %rbp, %rsp
popq %rbp
popq %r15
popq %r14
popq %r13
popq %r12
popq %rbx
ret
x86/asm: Change all ENTRY+ENDPROC to SYM_FUNC_* These are all functions which are invoked from elsewhere, so annotate them as global using the new SYM_FUNC_START and their ENDPROC's by SYM_FUNC_END. Make sure ENTRY/ENDPROC is not defined on X86_64, given these were the last users. Signed-off-by: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Borislav Petkov <bp@suse.de> Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> [hibernate] Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> [xen bits] Acked-by: Herbert Xu <herbert@gondor.apana.org.au> [crypto] Cc: Allison Randal <allison@lohutok.net> Cc: Andrey Ryabinin <aryabinin@virtuozzo.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Andy Shevchenko <andy@infradead.org> Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org> Cc: Armijn Hemel <armijn@tjaldur.nl> Cc: Cao jin <caoj.fnst@cn.fujitsu.com> Cc: Darren Hart <dvhart@infradead.org> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Enrico Weigelt <info@metux.net> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Herbert Xu <herbert@gondor.apana.org.au> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jim Mattson <jmattson@google.com> Cc: Joerg Roedel <joro@8bytes.org> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Juergen Gross <jgross@suse.com> Cc: Kate Stewart <kstewart@linuxfoundation.org> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: kvm ML <kvm@vger.kernel.org> Cc: Len Brown <len.brown@intel.com> Cc: linux-arch@vger.kernel.org Cc: linux-crypto@vger.kernel.org Cc: linux-efi <linux-efi@vger.kernel.org> Cc: linux-efi@vger.kernel.org Cc: linux-pm@vger.kernel.org Cc: Mark Rutland <mark.rutland@arm.com> Cc: Matt Fleming <matt@codeblueprint.co.uk> Cc: Paolo Bonzini <pbonzini@redhat.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: Peter Zijlstra <peterz@infradead.org> Cc: platform-driver-x86@vger.kernel.org Cc: "Radim Krčmář" <rkrcmar@redhat.com> Cc: Sean Christopherson <sean.j.christopherson@intel.com> Cc: Stefano Stabellini <sstabellini@kernel.org> Cc: "Steven Rostedt (VMware)" <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vitaly Kuznetsov <vkuznets@redhat.com> Cc: Wanpeng Li <wanpengli@tencent.com> Cc: Wei Huang <wei@redhat.com> Cc: x86-ml <x86@kernel.org> Cc: xen-devel@lists.xenproject.org Cc: Xiaoyao Li <xiaoyao.li@linux.intel.com> Link: https://lkml.kernel.org/r/20191011115108.12392-25-jslaby@suse.cz
2019-10-11 19:51:04 +08:00
SYM_FUNC_END(sha256_transform_ssse3)
crypto: x86 - make constants readonly, allow linker to merge them A lot of asm-optimized routines in arch/x86/crypto/ keep its constants in .data. This is wrong, they should be on .rodata. Mnay of these constants are the same in different modules. For example, 128-bit shuffle mask 0x000102030405060708090A0B0C0D0E0F exists in at least half a dozen places. There is a way to let linker merge them and use just one copy. The rules are as follows: mergeable objects of different sizes should not share sections. You can't put them all in one .rodata section, they will lose "mergeability". GCC puts its mergeable constants in ".rodata.cstSIZE" sections, or ".rodata.cstSIZE.<object_name>" if -fdata-sections is used. This patch does the same: .section .rodata.cst16.SHUF_MASK, "aM", @progbits, 16 It is important that all data in such section consists of 16-byte elements, not larger ones, and there are no implicit use of one element from another. When this is not the case, use non-mergeable section: .section .rodata[.VAR_NAME], "a", @progbits This reduces .data by ~15 kbytes: text data bss dec hex filename 11097415 2705840 2630712 16433967 fac32f vmlinux-prev.o 11112095 2690672 2630712 16433479 fac147 vmlinux.o Merged objects are visible in System.map: ffffffff81a28810 r POLY ffffffff81a28810 r POLY ffffffff81a28820 r TWOONE ffffffff81a28820 r TWOONE ffffffff81a28830 r PSHUFFLE_BYTE_FLIP_MASK <- merged regardless of ffffffff81a28830 r SHUF_MASK <------------- the name difference ffffffff81a28830 r SHUF_MASK ffffffff81a28830 r SHUF_MASK .. ffffffff81a28d00 r K512 <- merged three identical 640-byte tables ffffffff81a28d00 r K512 ffffffff81a28d00 r K512 Use of object names in section name suffixes is not strictly necessary, but might help if someday link stage will use garbage collection to eliminate unused sections (ld --gc-sections). Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com> CC: Herbert Xu <herbert@gondor.apana.org.au> CC: Josh Poimboeuf <jpoimboe@redhat.com> CC: Xiaodong Liu <xiaodong.liu@intel.com> CC: Megha Dey <megha.dey@intel.com> CC: linux-crypto@vger.kernel.org CC: x86@kernel.org CC: linux-kernel@vger.kernel.org Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2017-01-20 05:33:04 +08:00
.section .rodata.cst256.K256, "aM", @progbits, 256
.align 64
K256:
.long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
.long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5
.long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3
.long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174
.long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc
.long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da
.long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7
.long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967
.long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13
.long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85
.long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3
.long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070
.long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5
.long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3
.long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
.long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
crypto: x86 - make constants readonly, allow linker to merge them A lot of asm-optimized routines in arch/x86/crypto/ keep its constants in .data. This is wrong, they should be on .rodata. Mnay of these constants are the same in different modules. For example, 128-bit shuffle mask 0x000102030405060708090A0B0C0D0E0F exists in at least half a dozen places. There is a way to let linker merge them and use just one copy. The rules are as follows: mergeable objects of different sizes should not share sections. You can't put them all in one .rodata section, they will lose "mergeability". GCC puts its mergeable constants in ".rodata.cstSIZE" sections, or ".rodata.cstSIZE.<object_name>" if -fdata-sections is used. This patch does the same: .section .rodata.cst16.SHUF_MASK, "aM", @progbits, 16 It is important that all data in such section consists of 16-byte elements, not larger ones, and there are no implicit use of one element from another. When this is not the case, use non-mergeable section: .section .rodata[.VAR_NAME], "a", @progbits This reduces .data by ~15 kbytes: text data bss dec hex filename 11097415 2705840 2630712 16433967 fac32f vmlinux-prev.o 11112095 2690672 2630712 16433479 fac147 vmlinux.o Merged objects are visible in System.map: ffffffff81a28810 r POLY ffffffff81a28810 r POLY ffffffff81a28820 r TWOONE ffffffff81a28820 r TWOONE ffffffff81a28830 r PSHUFFLE_BYTE_FLIP_MASK <- merged regardless of ffffffff81a28830 r SHUF_MASK <------------- the name difference ffffffff81a28830 r SHUF_MASK ffffffff81a28830 r SHUF_MASK .. ffffffff81a28d00 r K512 <- merged three identical 640-byte tables ffffffff81a28d00 r K512 ffffffff81a28d00 r K512 Use of object names in section name suffixes is not strictly necessary, but might help if someday link stage will use garbage collection to eliminate unused sections (ld --gc-sections). Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com> CC: Herbert Xu <herbert@gondor.apana.org.au> CC: Josh Poimboeuf <jpoimboe@redhat.com> CC: Xiaodong Liu <xiaodong.liu@intel.com> CC: Megha Dey <megha.dey@intel.com> CC: linux-crypto@vger.kernel.org CC: x86@kernel.org CC: linux-kernel@vger.kernel.org Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2017-01-20 05:33:04 +08:00
.section .rodata.cst16.PSHUFFLE_BYTE_FLIP_MASK, "aM", @progbits, 16
.align 16
PSHUFFLE_BYTE_FLIP_MASK:
.octa 0x0c0d0e0f08090a0b0405060700010203
crypto: x86 - make constants readonly, allow linker to merge them A lot of asm-optimized routines in arch/x86/crypto/ keep its constants in .data. This is wrong, they should be on .rodata. Mnay of these constants are the same in different modules. For example, 128-bit shuffle mask 0x000102030405060708090A0B0C0D0E0F exists in at least half a dozen places. There is a way to let linker merge them and use just one copy. The rules are as follows: mergeable objects of different sizes should not share sections. You can't put them all in one .rodata section, they will lose "mergeability". GCC puts its mergeable constants in ".rodata.cstSIZE" sections, or ".rodata.cstSIZE.<object_name>" if -fdata-sections is used. This patch does the same: .section .rodata.cst16.SHUF_MASK, "aM", @progbits, 16 It is important that all data in such section consists of 16-byte elements, not larger ones, and there are no implicit use of one element from another. When this is not the case, use non-mergeable section: .section .rodata[.VAR_NAME], "a", @progbits This reduces .data by ~15 kbytes: text data bss dec hex filename 11097415 2705840 2630712 16433967 fac32f vmlinux-prev.o 11112095 2690672 2630712 16433479 fac147 vmlinux.o Merged objects are visible in System.map: ffffffff81a28810 r POLY ffffffff81a28810 r POLY ffffffff81a28820 r TWOONE ffffffff81a28820 r TWOONE ffffffff81a28830 r PSHUFFLE_BYTE_FLIP_MASK <- merged regardless of ffffffff81a28830 r SHUF_MASK <------------- the name difference ffffffff81a28830 r SHUF_MASK ffffffff81a28830 r SHUF_MASK .. ffffffff81a28d00 r K512 <- merged three identical 640-byte tables ffffffff81a28d00 r K512 ffffffff81a28d00 r K512 Use of object names in section name suffixes is not strictly necessary, but might help if someday link stage will use garbage collection to eliminate unused sections (ld --gc-sections). Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com> CC: Herbert Xu <herbert@gondor.apana.org.au> CC: Josh Poimboeuf <jpoimboe@redhat.com> CC: Xiaodong Liu <xiaodong.liu@intel.com> CC: Megha Dey <megha.dey@intel.com> CC: linux-crypto@vger.kernel.org CC: x86@kernel.org CC: linux-kernel@vger.kernel.org Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2017-01-20 05:33:04 +08:00
.section .rodata.cst16._SHUF_00BA, "aM", @progbits, 16
.align 16
# shuffle xBxA -> 00BA
_SHUF_00BA:
.octa 0xFFFFFFFFFFFFFFFF0b0a090803020100
crypto: x86 - make constants readonly, allow linker to merge them A lot of asm-optimized routines in arch/x86/crypto/ keep its constants in .data. This is wrong, they should be on .rodata. Mnay of these constants are the same in different modules. For example, 128-bit shuffle mask 0x000102030405060708090A0B0C0D0E0F exists in at least half a dozen places. There is a way to let linker merge them and use just one copy. The rules are as follows: mergeable objects of different sizes should not share sections. You can't put them all in one .rodata section, they will lose "mergeability". GCC puts its mergeable constants in ".rodata.cstSIZE" sections, or ".rodata.cstSIZE.<object_name>" if -fdata-sections is used. This patch does the same: .section .rodata.cst16.SHUF_MASK, "aM", @progbits, 16 It is important that all data in such section consists of 16-byte elements, not larger ones, and there are no implicit use of one element from another. When this is not the case, use non-mergeable section: .section .rodata[.VAR_NAME], "a", @progbits This reduces .data by ~15 kbytes: text data bss dec hex filename 11097415 2705840 2630712 16433967 fac32f vmlinux-prev.o 11112095 2690672 2630712 16433479 fac147 vmlinux.o Merged objects are visible in System.map: ffffffff81a28810 r POLY ffffffff81a28810 r POLY ffffffff81a28820 r TWOONE ffffffff81a28820 r TWOONE ffffffff81a28830 r PSHUFFLE_BYTE_FLIP_MASK <- merged regardless of ffffffff81a28830 r SHUF_MASK <------------- the name difference ffffffff81a28830 r SHUF_MASK ffffffff81a28830 r SHUF_MASK .. ffffffff81a28d00 r K512 <- merged three identical 640-byte tables ffffffff81a28d00 r K512 ffffffff81a28d00 r K512 Use of object names in section name suffixes is not strictly necessary, but might help if someday link stage will use garbage collection to eliminate unused sections (ld --gc-sections). Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com> CC: Herbert Xu <herbert@gondor.apana.org.au> CC: Josh Poimboeuf <jpoimboe@redhat.com> CC: Xiaodong Liu <xiaodong.liu@intel.com> CC: Megha Dey <megha.dey@intel.com> CC: linux-crypto@vger.kernel.org CC: x86@kernel.org CC: linux-kernel@vger.kernel.org Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2017-01-20 05:33:04 +08:00
.section .rodata.cst16._SHUF_DC00, "aM", @progbits, 16
.align 16
# shuffle xDxC -> DC00
_SHUF_DC00:
.octa 0x0b0a090803020100FFFFFFFFFFFFFFFF