generic-library/bionic
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

am aa66e882: Merge "Add __memcpy_chk assembly for 64 bit."

Browse Source 
* commit 'aa66e8828048d8d6fd7e4bb9c5e341ed89614266':
  Add __memcpy_chk assembly for 64 bit.
This commit is contained in:
Christopher Ferris
2014-09-26 21:04:56 +00:00
committed by Android Git Automerger
parent 0d7599e0b5 aa66e88280
commit 0fc3711b22
5 changed files with 484 additions and 370 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
libc/arch-arm64/arm64.mk
View File

@@ -5,7 +5,6 @@
# #
libc_bionic_src_files_arm64 += \ libc_bionic_src_files_arm64 += \
bionic/__memcpy_chk.cpp \
bionic/__memset_chk.cpp \ bionic/__memset_chk.cpp \
bionic/__strcpy_chk.cpp \ bionic/__strcpy_chk.cpp \
bionic/__strcat_chk.cpp \ bionic/__strcat_chk.cpp \

250
libc/arch-arm64/denver64/bionic/memcpy.S
View File

@@ -1,205 +1,63 @@
/* Copyright (c) 2012, Linaro Limited /*
All rights reserved. * Copyright (C) 2008 The Android Open Source Project
Copyright (c) 2014, NVIDIA Corporation. All rights reserved. * All rights reserved.
*
Redistribution and use in source and binary forms, with or without * Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met: * modification, are permitted provided that the following conditions
* Redistributions of source code must retain the above copyright * are met:
notice, this list of conditions and the following disclaimer. * * Redistributions of source code must retain the above copyright
* Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer.
notice, this list of conditions and the following disclaimer in the * * Redistributions in binary form must reproduce the above copyright
documentation and/or other materials provided with the distribution. * notice, this list of conditions and the following disclaimer in
* Neither the name of the Linaro nor the * the documentation and/or other materials provided with the
names of its contributors may be used to endorse or promote products * distribution.
derived from this software without specific prior written permission. *
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. * SUCH DAMAGE.
*/ */
/* Assumptions: // Prototype: void *memcpy (void *dst, const void *src, size_t count).
*
* denver, ARMv8-a, AArch64
* Unaligned accesses
*
*/
#include <private/bionic_asm.h> #include <private/bionic_asm.h>
#include <private/libc_events.h>
#define dstin x0 ENTRY(__memcpy_chk)
#define src x1 cmp x2, x3
#define count x2 b.hi __memcpy_chk_fail
#define tmp1 x3
#define tmp1w w3
#define tmp2 x4
#define tmp2w w4
#define tmp3 x5
#define tmp3w w5
#define dst x6
#define A_l x7 // Fall through to memcpy...
#define A_h x8 END(__memcpy_chk)
#define B_l x9
#define B_h x10
#define C_l x11
#define C_h x12
#define D_l x13
#define D_h x14
#define QA_l q0
#define QA_h q1
#define QB_l q2
#define QB_h q3
ENTRY(memcpy) ENTRY(memcpy)
#include "memcpy_base.S"
mov dst, dstin
cmp count, #64
b.ge .Lcpy_not_short
cmp count, #15
b.le .Ltail15tiny
/* Deal with small copies quickly by dropping straight into the
* exit block. */
.Ltail63:
/* Copy up to 48 bytes of data. At this point we only need the
* bottom 6 bits of count to be accurate. */
ands tmp1, count, #0x30
b.eq .Ltail15
add dst, dst, tmp1
add src, src, tmp1
cmp tmp1w, #0x20
b.eq 1f
b.lt 2f
ldp A_l, A_h, [src, #-48]
stp A_l, A_h, [dst, #-48]
1:
ldp A_l, A_h, [src, #-32]
stp A_l, A_h, [dst, #-32]
2:
ldp A_l, A_h, [src, #-16]
stp A_l, A_h, [dst, #-16]
.Ltail15:
ands count, count, #15
beq 1f
add src, src, count
ldp A_l, A_h, [src, #-16]
add dst, dst, count
stp A_l, A_h, [dst, #-16]
1:
ret
.Ltail15tiny:
/* Copy up to 15 bytes of data. Does not assume additional data
being copied. */
tbz count, #3, 1f
ldr tmp1, [src], #8
str tmp1, [dst], #8
1:
tbz count, #2, 1f
ldr tmp1w, [src], #4
str tmp1w, [dst], #4
1:
tbz count, #1, 1f
ldrh tmp1w, [src], #2
strh tmp1w, [dst], #2
1:
tbz count, #0, 1f
ldrb tmp1w, [src]
strb tmp1w, [dst]
1:
ret
.Lcpy_not_short:
/* We don't much care about the alignment of DST, but we want SRC
* to be 128-bit (16 byte) aligned so that we don't cross cache line
* boundaries on both loads and stores. */
neg tmp2, src
ands tmp2, tmp2, #15 /* Bytes to reach alignment. */
b.eq 2f
sub count, count, tmp2
/* Copy more data than needed; it's faster than jumping
* around copying sub-Quadword quantities. We know that
* it can't overrun. */
ldp A_l, A_h, [src]
add src, src, tmp2
stp A_l, A_h, [dst]
add dst, dst, tmp2
/* There may be less than 63 bytes to go now. */
cmp count, #63
b.le .Ltail63
2:
subs count, count, #128
b.ge .Lcpy_body_large
/* Less than 128 bytes to copy, so handle 64 here and then jump
* to the tail. */
ldp QA_l, QA_h, [src]
ldp QB_l, QB_h, [src, #32]
stp QA_l, QA_h, [dst]
stp QB_l, QB_h, [dst, #32]
tst count, #0x3f
add src, src, #64
add dst, dst, #64
b.ne .Ltail63
ret
/* Critical loop. Start at a new cache line boundary. Assuming
* 64 bytes per line this ensures the entire loop is in one line. */
.p2align 6
.Lcpy_body_large:
cmp count, 65536
bhi .Lcpy_body_huge
/* There are at least 128 bytes to copy. */
ldp QA_l, QA_h, [src, #0]
sub dst, dst, #32 /* Pre-bias. */
ldp QB_l, QB_h, [src, #32]! /* src += 64 - Pre-bias. */
1:
stp QA_l, QA_h, [dst, #32]
ldp QA_l, QA_h, [src, #32]
stp QB_l, QB_h, [dst, #64]!
ldp QB_l, QB_h, [src, #64]!
subs count, count, #64
b.ge 1b
stp QA_l, QA_h, [dst, #32]
stp QB_l, QB_h, [dst, #64]
add src, src, #32
add dst, dst, #64 + 32
tst count, #0x3f
b.ne .Ltail63
ret
.Lcpy_body_huge:
/* There are at least 128 bytes to copy. */
ldp QA_l, QA_h, [src, #0]
sub dst, dst, #32 /* Pre-bias. */
ldp QB_l, QB_h, [src, #32]!
1:
stnp QA_l, QA_h, [dst, #32]
stnp QB_l, QB_h, [dst, #64]
ldp QA_l, QA_h, [src, #32]
ldp QB_l, QB_h, [src, #64]!
add dst, dst, #64
subs count, count, #64
b.ge 1b
stnp QA_l, QA_h, [dst, #32]
stnp QB_l, QB_h, [dst, #64]
add src, src, #32
add dst, dst, #64 + 32
tst count, #0x3f
b.ne .Ltail63
ret
END(memcpy) END(memcpy)
ENTRY_PRIVATE(__memcpy_chk_fail)
// Preserve for accurate backtrace.
stp x29, x30, [sp, -16]!
.cfi_def_cfa_offset 16
.cfi_rel_offset x29, 0
.cfi_rel_offset x30, 8
adrp x0, error_string
add x0, x0, :lo12:error_string
ldr x1, error_code
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_MEMCPY_BUFFER_OVERFLOW
END(__memcpy_chk_fail)
.data
.align 2
error_string:
.string "memcpy: prevented write past end of buffer"

199
libc/arch-arm64/denver64/bionic/memcpy_base.S Normal file
View File

@@ -0,0 +1,199 @@
/* Copyright (c) 2012, Linaro Limited
All rights reserved.
Copyright (c) 2014, NVIDIA Corporation. All rights reserved.
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 the Linaro 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
HOLDER 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.
*/
/* Assumptions:
*
* denver, ARMv8-a, AArch64
* Unaligned accesses
*
*/
#define dstin x0
#define src x1
#define count x2
#define tmp1 x3
#define tmp1w w3
#define tmp2 x4
#define tmp2w w4
#define tmp3 x5
#define tmp3w w5
#define dst x6
#define A_l x7
#define A_h x8
#define B_l x9
#define B_h x10
#define C_l x11
#define C_h x12
#define D_l x13
#define D_h x14
#define QA_l q0
#define QA_h q1
#define QB_l q2
#define QB_h q3
mov dst, dstin
cmp count, #64
b.ge .Lcpy_not_short
cmp count, #15
b.le .Ltail15tiny
/* Deal with small copies quickly by dropping straight into the
* exit block. */
.Ltail63:
/* Copy up to 48 bytes of data. At this point we only need the
* bottom 6 bits of count to be accurate. */
ands tmp1, count, #0x30
b.eq .Ltail15
add dst, dst, tmp1
add src, src, tmp1
cmp tmp1w, #0x20
b.eq 1f
b.lt 2f
ldp A_l, A_h, [src, #-48]
stp A_l, A_h, [dst, #-48]
1:
ldp A_l, A_h, [src, #-32]
stp A_l, A_h, [dst, #-32]
2:
ldp A_l, A_h, [src, #-16]
stp A_l, A_h, [dst, #-16]
.Ltail15:
ands count, count, #15
beq 1f
add src, src, count
ldp A_l, A_h, [src, #-16]
add dst, dst, count
stp A_l, A_h, [dst, #-16]
1:
ret
.Ltail15tiny:
/* Copy up to 15 bytes of data. Does not assume additional data
being copied. */
tbz count, #3, 1f
ldr tmp1, [src], #8
str tmp1, [dst], #8
1:
tbz count, #2, 1f
ldr tmp1w, [src], #4
str tmp1w, [dst], #4
1:
tbz count, #1, 1f
ldrh tmp1w, [src], #2
strh tmp1w, [dst], #2
1:
tbz count, #0, 1f
ldrb tmp1w, [src]
strb tmp1w, [dst]
1:
ret
.Lcpy_not_short:
/* We don't much care about the alignment of DST, but we want SRC
* to be 128-bit (16 byte) aligned so that we don't cross cache line
* boundaries on both loads and stores. */
neg tmp2, src
ands tmp2, tmp2, #15 /* Bytes to reach alignment. */
b.eq 2f
sub count, count, tmp2
/* Copy more data than needed; it's faster than jumping
* around copying sub-Quadword quantities. We know that
* it can't overrun. */
ldp A_l, A_h, [src]
add src, src, tmp2
stp A_l, A_h, [dst]
add dst, dst, tmp2
/* There may be less than 63 bytes to go now. */
cmp count, #63
b.le .Ltail63
2:
subs count, count, #128
b.ge .Lcpy_body_large
/* Less than 128 bytes to copy, so handle 64 here and then jump
* to the tail. */
ldp QA_l, QA_h, [src]
ldp QB_l, QB_h, [src, #32]
stp QA_l, QA_h, [dst]
stp QB_l, QB_h, [dst, #32]
tst count, #0x3f
add src, src, #64
add dst, dst, #64
b.ne .Ltail63
ret
/* Critical loop. Start at a new cache line boundary. Assuming
* 64 bytes per line this ensures the entire loop is in one line. */
.p2align 6
.Lcpy_body_large:
cmp count, 65536
bhi .Lcpy_body_huge
/* There are at least 128 bytes to copy. */
ldp QA_l, QA_h, [src, #0]
sub dst, dst, #32 /* Pre-bias. */
ldp QB_l, QB_h, [src, #32]! /* src += 64 - Pre-bias. */
1:
stp QA_l, QA_h, [dst, #32]
ldp QA_l, QA_h, [src, #32]
stp QB_l, QB_h, [dst, #64]!
ldp QB_l, QB_h, [src, #64]!
subs count, count, #64
b.ge 1b
stp QA_l, QA_h, [dst, #32]
stp QB_l, QB_h, [dst, #64]
add src, src, #32
add dst, dst, #64 + 32
tst count, #0x3f
b.ne .Ltail63
ret
.Lcpy_body_huge:
/* There are at least 128 bytes to copy. */
ldp QA_l, QA_h, [src, #0]
sub dst, dst, #32 /* Pre-bias. */
ldp QB_l, QB_h, [src, #32]!
1:
stnp QA_l, QA_h, [dst, #32]
stnp QB_l, QB_h, [dst, #64]
ldp QA_l, QA_h, [src, #32]
ldp QB_l, QB_h, [src, #64]!
add dst, dst, #64
subs count, count, #64
b.ge 1b
stnp QA_l, QA_h, [dst, #32]
stnp QB_l, QB_h, [dst, #64]
add src, src, #32
add dst, dst, #64 + 32
tst count, #0x3f
b.ne .Ltail63
ret

229
libc/arch-arm64/generic/bionic/memcpy.S
View File

@@ -1,184 +1,63 @@
/* Copyright (c) 2012, Linaro Limited /*
All rights reserved. * Copyright (C) 2008 The Android Open Source Project
* All rights reserved.
Redistribution and use in source and binary forms, with or without *
modification, are permitted provided that the following conditions are met: * Redistribution and use in source and binary forms, with or without
* Redistributions of source code must retain the above copyright * modification, are permitted provided that the following conditions
notice, this list of conditions and the following disclaimer. * are met:
* Redistributions in binary form must reproduce the above copyright * * Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer in the * notice, this list of conditions and the following disclaimer.
documentation and/or other materials provided with the distribution. * * Redistributions in binary form must reproduce the above copyright
* Neither the name of the Linaro nor the * notice, this list of conditions and the following disclaimer in
names of its contributors may be used to endorse or promote products * the documentation and/or other materials provided with the
derived from this software without specific prior written permission. * distribution.
*
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/ */
/* Assumptions: // Prototype: void *memcpy (void *dst, const void *src, size_t count).
*
* ARMv8-a, AArch64
* Unaligned accesses
*
*/
#include <private/bionic_asm.h> #include <private/bionic_asm.h>
#include <private/libc_events.h>
#define dstin x0 ENTRY(__memcpy_chk)
#define src x1 cmp x2, x3
#define count x2 b.hi __memcpy_chk_fail
#define tmp1 x3
#define tmp1w w3
#define tmp2 x4
#define tmp2w w4
#define tmp3 x5
#define tmp3w w5
#define dst x6
#define A_l x7 // Fall through to memcpy...
#define A_h x8 END(__memcpy_chk)
#define B_l x9
#define B_h x10
#define C_l x11
#define C_h x12
#define D_l x13
#define D_h x14
ENTRY(memcpy) ENTRY(memcpy)
#include "memcpy_base.S"
mov dst, dstin
cmp count, #64
b.ge .Lcpy_not_short
cmp count, #15
b.le .Ltail15tiny
/* Deal with small copies quickly by dropping straight into the
* exit block. */
.Ltail63:
/* Copy up to 48 bytes of data. At this point we only need the
* bottom 6 bits of count to be accurate. */
ands tmp1, count, #0x30
b.eq .Ltail15
add dst, dst, tmp1
add src, src, tmp1
cmp tmp1w, #0x20
b.eq 1f
b.lt 2f
ldp A_l, A_h, [src, #-48]
stp A_l, A_h, [dst, #-48]
1:
ldp A_l, A_h, [src, #-32]
stp A_l, A_h, [dst, #-32]
2:
ldp A_l, A_h, [src, #-16]
stp A_l, A_h, [dst, #-16]
.Ltail15:
ands count, count, #15
beq 1f
add src, src, count
ldp A_l, A_h, [src, #-16]
add dst, dst, count
stp A_l, A_h, [dst, #-16]
1:
ret
.Ltail15tiny:
/* Copy up to 15 bytes of data. Does not assume additional data
being copied. */
tbz count, #3, 1f
ldr tmp1, [src], #8
str tmp1, [dst], #8
1:
tbz count, #2, 1f
ldr tmp1w, [src], #4
str tmp1w, [dst], #4
1:
tbz count, #1, 1f
ldrh tmp1w, [src], #2
strh tmp1w, [dst], #2
1:
tbz count, #0, 1f
ldrb tmp1w, [src]
strb tmp1w, [dst]
1:
ret
.Lcpy_not_short:
/* We don't much care about the alignment of DST, but we want SRC
* to be 128-bit (16 byte) aligned so that we don't cross cache line
* boundaries on both loads and stores. */
neg tmp2, src
ands tmp2, tmp2, #15 /* Bytes to reach alignment. */
b.eq 2f
sub count, count, tmp2
/* Copy more data than needed; it's faster than jumping
* around copying sub-Quadword quantities. We know that
* it can't overrun. */
ldp A_l, A_h, [src]
add src, src, tmp2
stp A_l, A_h, [dst]
add dst, dst, tmp2
/* There may be less than 63 bytes to go now. */
cmp count, #63
b.le .Ltail63
2:
subs count, count, #128
b.ge .Lcpy_body_large
/* Less than 128 bytes to copy, so handle 64 here and then jump
* to the tail. */
ldp A_l, A_h, [src]
ldp B_l, B_h, [src, #16]
ldp C_l, C_h, [src, #32]
ldp D_l, D_h, [src, #48]
stp A_l, A_h, [dst]
stp B_l, B_h, [dst, #16]
stp C_l, C_h, [dst, #32]
stp D_l, D_h, [dst, #48]
tst count, #0x3f
add src, src, #64
add dst, dst, #64
b.ne .Ltail63
ret
/* Critical loop. Start at a new cache line boundary. Assuming
* 64 bytes per line this ensures the entire loop is in one line. */
.p2align 6
.Lcpy_body_large:
/* There are at least 128 bytes to copy. */
ldp A_l, A_h, [src, #0]
sub dst, dst, #16 /* Pre-bias. */
ldp B_l, B_h, [src, #16]
ldp C_l, C_h, [src, #32]
ldp D_l, D_h, [src, #48]! /* src += 64 - Pre-bias. */
1:
stp A_l, A_h, [dst, #16]
ldp A_l, A_h, [src, #16]
stp B_l, B_h, [dst, #32]
ldp B_l, B_h, [src, #32]
stp C_l, C_h, [dst, #48]
ldp C_l, C_h, [src, #48]
stp D_l, D_h, [dst, #64]!
ldp D_l, D_h, [src, #64]!
subs count, count, #64
b.ge 1b
stp A_l, A_h, [dst, #16]
stp B_l, B_h, [dst, #32]
stp C_l, C_h, [dst, #48]
stp D_l, D_h, [dst, #64]
add src, src, #16
add dst, dst, #64 + 16
tst count, #0x3f
b.ne .Ltail63
ret
END(memcpy) END(memcpy)
ENTRY_PRIVATE(__memcpy_chk_fail)
// Preserve for accurate backtrace.
stp x29, x30, [sp, -16]!
.cfi_def_cfa_offset 16
.cfi_rel_offset x29, 0
.cfi_rel_offset x30, 8
adrp x0, error_string
add x0, x0, :lo12:error_string
ldr x1, error_code
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_MEMCPY_BUFFER_OVERFLOW
END(__memcpy_chk_fail)
.data
.align 2
error_string:
.string "memcpy: prevented write past end of buffer"

179
libc/arch-arm64/generic/bionic/memcpy_base.S Normal file
View File

@@ -0,0 +1,179 @@
/* Copyright (c) 2012, Linaro Limited
All rights reserved.
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 the Linaro 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
HOLDER 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.
*/
/* Assumptions:
*
* ARMv8-a, AArch64
* Unaligned accesses
*
*/
#define dstin x0
#define src x1
#define count x2
#define tmp1 x3
#define tmp1w w3
#define tmp2 x4
#define tmp2w w4
#define tmp3 x5
#define tmp3w w5
#define dst x6
#define A_l x7
#define A_h x8
#define B_l x9
#define B_h x10
#define C_l x11
#define C_h x12
#define D_l x13
#define D_h x14
mov dst, dstin
cmp count, #64
b.ge .Lcpy_not_short
cmp count, #15
b.le .Ltail15tiny
/* Deal with small copies quickly by dropping straight into the
* exit block. */
.Ltail63:
/* Copy up to 48 bytes of data. At this point we only need the
* bottom 6 bits of count to be accurate. */
ands tmp1, count, #0x30
b.eq .Ltail15
add dst, dst, tmp1
add src, src, tmp1
cmp tmp1w, #0x20
b.eq 1f
b.lt 2f
ldp A_l, A_h, [src, #-48]
stp A_l, A_h, [dst, #-48]
1:
ldp A_l, A_h, [src, #-32]
stp A_l, A_h, [dst, #-32]
2:
ldp A_l, A_h, [src, #-16]
stp A_l, A_h, [dst, #-16]
.Ltail15:
ands count, count, #15
beq 1f
add src, src, count
ldp A_l, A_h, [src, #-16]
add dst, dst, count
stp A_l, A_h, [dst, #-16]
1:
ret
.Ltail15tiny:
/* Copy up to 15 bytes of data. Does not assume additional data
being copied. */
tbz count, #3, 1f
ldr tmp1, [src], #8
str tmp1, [dst], #8
1:
tbz count, #2, 1f
ldr tmp1w, [src], #4
str tmp1w, [dst], #4
1:
tbz count, #1, 1f
ldrh tmp1w, [src], #2
strh tmp1w, [dst], #2
1:
tbz count, #0, 1f
ldrb tmp1w, [src]
strb tmp1w, [dst]
1:
ret
.Lcpy_not_short:
/* We don't much care about the alignment of DST, but we want SRC
* to be 128-bit (16 byte) aligned so that we don't cross cache line
* boundaries on both loads and stores. */
neg tmp2, src
ands tmp2, tmp2, #15 /* Bytes to reach alignment. */
b.eq 2f
sub count, count, tmp2
/* Copy more data than needed; it's faster than jumping
* around copying sub-Quadword quantities. We know that
* it can't overrun. */
ldp A_l, A_h, [src]
add src, src, tmp2
stp A_l, A_h, [dst]
add dst, dst, tmp2
/* There may be less than 63 bytes to go now. */
cmp count, #63
b.le .Ltail63
2:
subs count, count, #128
b.ge .Lcpy_body_large
/* Less than 128 bytes to copy, so handle 64 here and then jump
* to the tail. */
ldp A_l, A_h, [src]
ldp B_l, B_h, [src, #16]
ldp C_l, C_h, [src, #32]
ldp D_l, D_h, [src, #48]
stp A_l, A_h, [dst]
stp B_l, B_h, [dst, #16]
stp C_l, C_h, [dst, #32]
stp D_l, D_h, [dst, #48]
tst count, #0x3f
add src, src, #64
add dst, dst, #64
b.ne .Ltail63
ret
/* Critical loop. Start at a new cache line boundary. Assuming
* 64 bytes per line this ensures the entire loop is in one line. */
.p2align 6
.Lcpy_body_large:
/* There are at least 128 bytes to copy. */
ldp A_l, A_h, [src, #0]
sub dst, dst, #16 /* Pre-bias. */
ldp B_l, B_h, [src, #16]
ldp C_l, C_h, [src, #32]
ldp D_l, D_h, [src, #48]! /* src += 64 - Pre-bias. */
1:
stp A_l, A_h, [dst, #16]
ldp A_l, A_h, [src, #16]
stp B_l, B_h, [dst, #32]
ldp B_l, B_h, [src, #32]
stp C_l, C_h, [dst, #48]
ldp C_l, C_h, [src, #48]
stp D_l, D_h, [dst, #64]!
ldp D_l, D_h, [src, #64]!
subs count, count, #64
b.ge 1b
stp A_l, A_h, [dst, #16]
stp B_l, B_h, [dst, #32]
stp C_l, C_h, [dst, #48]
stp D_l, D_h, [dst, #64]
add src, src, #16
add dst, dst, #64 + 16
tst count, #0x3f
b.ne .Ltail63
ret