Compare commits

..

2 Commits

Author SHA1 Message Date
Baligh Uddin
8a6d66c7ef Merge "resolved conflicts for merge of 05292dd6 to jb-mr1-dev" into jb-mr1.1-dev 2013-09-25 02:31:23 +00:00
Elliott Hughes
61441240d9 resolved conflicts for merge of 05292dd6 to jb-mr1-dev
Change-Id: I051cad5dc3c985f11c52708f4487e64ea45ca617
2013-09-24 15:52:46 -07:00
1981 changed files with 47656 additions and 67620 deletions

View File

@@ -1,15 +0,0 @@
KNOWN ABI BUGS
--------------
time_t is 32-bit. http://b/5819737
off_t is 32-bit. There is off64_t, but no _FILE_OFFSET_BITS support.
sigset_t is too small on ARM and x86 (but correct on MIPS), so support
for real-time signals is broken. http://b/5828899
Too few TLS slots mean we can't allocate 128 pthread_key_t instances,
which POSIX says should be the minimum.
atexit(3) handlers registered by a shared library aren't called on
dlclose(3); this only affects ARM. http://b/4998315

File diff suppressed because it is too large Load Diff

View File

@@ -37,40 +37,6 @@ SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (C) 2004, 2005, 2008 Internet Systems Consortium, Inc. ("ISC")
Copyright (C) 1995-1999, 2001, 2003 Internet Software Consortium.
Permission to use, copy, modify, and/or distribute this software for any
purpose with or without fee is hereby granted, provided that the above
copyright notice and this permission notice appear in all copies.
THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES WITH
REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
AND FITNESS. IN NO EVENT SHALL ISC BE LIABLE FOR ANY SPECIAL, DIRECT,
INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM
LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
PERFORMANCE OF THIS SOFTWARE.
-------------------------------------------------------------------
Copyright (C) 2004, 2005, 2008 Internet Systems Consortium, Inc. ("ISC")
Copyright (C) 1997-2001 Internet Software Consortium.
Permission to use, copy, modify, and/or distribute this software for any
purpose with or without fee is hereby granted, provided that the above
copyright notice and this permission notice appear in all copies.
THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES WITH
REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
AND FITNESS. IN NO EVENT SHALL ISC BE LIABLE FOR ANY SPECIAL, DIRECT,
INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM
LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
PERFORMANCE OF THIS SOFTWARE.
-------------------------------------------------------------------
Copyright (C) 2006 The Android Open Source Project
All rights reserved.
@@ -99,6 +65,22 @@ SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (C) 2007 The Android Open Source Project
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-------------------------------------------------------------------
Copyright (C) 2008 The Android Open Source Project
All rights reserved.
@@ -346,50 +328,6 @@ SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (C) 2013 The Android Open Source Project
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-------------------------------------------------------------------
Copyright (C) 2013 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:
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.
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.
-------------------------------------------------------------------
Copyright (c) 1980, 1983, 1988, 1993
The Regents of the University of California. All rights reserved.
@@ -500,40 +438,6 @@ SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (c) 1982, 1986, 1993
The Regents of the University of California. All rights reserved.
(c) UNIX System Laboratories, Inc.
All or some portions of this file are derived from material licensed
to the University of California by American Telephone and Telegraph
Co. or Unix System Laboratories, Inc. and are reproduced herein with
the permission of UNIX System Laboratories, Inc.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. 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.
3. Neither the name of the University 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 REGENTS 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 REGENTS 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.
-------------------------------------------------------------------
Copyright (c) 1983, 1987, 1989
The Regents of the University of California. All rights reserved.
@@ -665,11 +569,7 @@ are met:
2. 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.
3. All advertising materials mentioning features or use of this software
must display the following acknowledgement:
This product includes software developed by the University of
California, Berkeley and its contributors.
4. Neither the name of the University nor the names of its contributors
3. Neither the name of the University nor the names of its contributors
may be used to endorse or promote products derived from this software
without specific prior written permission.
@@ -685,6 +585,24 @@ 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.
Portions Copyright (c) 1993 by Digital Equipment Corporation.
Permission to use, copy, modify, and distribute this software for any
purpose with or without fee is hereby granted, provided that the above
copyright notice and this permission notice appear in all copies, and that
the name of Digital Equipment Corporation not be used in advertising or
publicity pertaining to distribution of the document or software without
specific, written prior permission.
THE SOFTWARE IS PROVIDED "AS IS" AND DIGITAL EQUIPMENT CORP. DISCLAIMS ALL
WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES
OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL DIGITAL EQUIPMENT
CORPORATION BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
SOFTWARE.
-------------------------------------------------------------------
Copyright (c) 1983, 1993
@@ -1488,6 +1406,38 @@ SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (c) 1990 Regents of the University of California.
All rights reserved.
This code is derived from software contributed to Berkeley by
Chris Torek.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. 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.
3. Neither the name of the University 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 REGENTS 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 REGENTS 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.
-------------------------------------------------------------------
Copyright (c) 1990 The Regents of the University of California.
All rights reserved.
@@ -1641,6 +1591,35 @@ SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (c) 1990, 1993
The Regents of the University of California. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. 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.
4. Neither the name of the University 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 REGENTS 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 REGENTS 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.
-------------------------------------------------------------------
Copyright (c) 1990, 1993
The Regents of the University of California. All rights reserved.
@@ -1673,6 +1652,38 @@ SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (c) 1990, 1993
The Regents of the University of California. All rights reserved.
This code is derived from software contributed to Berkeley by
Chris Torek.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. 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.
4. Neither the name of the University 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 REGENTS 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 REGENTS 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.
-------------------------------------------------------------------
Copyright (c) 1990, 1993
The Regents of the University of California. All rights reserved.
@@ -2365,35 +2376,6 @@ to anyone/anything when using this software.
-------------------------------------------------------------------
Copyright (c) 1994 The NetBSD Foundation, Inc.
All rights reserved.
This code is derived from software contributed to The NetBSD Foundation
by Christos Zoulas.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. 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.
THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. 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 FOUNDATION 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.
-------------------------------------------------------------------
Copyright (c) 1995 - 2001 Kungliga Tekniska Högskolan
(Royal Institute of Technology, Stockholm, Sweden).
All rights reserved.
@@ -2481,6 +2463,32 @@ rights to redistribute these changes.
-------------------------------------------------------------------
Copyright (c) 1995, 1996 Erik Theisen. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. 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.
3. The name of the author may not be used to endorse or promote products
derived from this software without specific prior written permission
THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 AUTHOR 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.
-------------------------------------------------------------------
Copyright (c) 1996 by Internet Software Consortium.
Permission to use, copy, modify, and distribute this software for any
@@ -2820,7 +2828,7 @@ are met:
2. Redistributions in binary form must reproduce the above copyright
notices, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
3. Neither the name of the University nor the names of its contributors
4. Neither the name of the University nor the names of its contributors
may be used to endorse or promote products derived from this software
without specific prior written permission.
@@ -3060,6 +3068,13 @@ are met:
2. 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.
3. All advertising materials mentioning features or use of this software
must display the following acknowledgement:
This product includes software developed by the NetBSD
Foundation, Inc. and its contributors.
4. Neither the name of The NetBSD Foundation 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 NETBSD FOUNDATION, INC. AND CONTRIBUTORS
``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
@@ -3581,6 +3596,23 @@ OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
-------------------------------------------------------------------
Copyright (c) 2004 by Internet Systems Consortium, Inc. ("ISC")
Copyright (c) 1997-1999 by Internet Software Consortium.
Permission to use, copy, modify, and distribute this software for any
purpose with or without fee is hereby granted, provided that the above
copyright notice and this permission notice appear in all copies.
THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES
WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL ISC BE LIABLE FOR
ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT
OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
-------------------------------------------------------------------
Copyright (c) 2004 by Internet Systems Consortium, Inc. ("ISC")
Copyright (c) 1999 by Internet Software Consortium.
@@ -3641,23 +3673,6 @@ SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (c) 2005 by Internet Systems Consortium, Inc. ("ISC")
Copyright (c) 1995-1999 by Internet Software Consortium
Permission to use, copy, modify, and distribute this software for any
purpose with or without fee is hereby granted, provided that the above
copyright notice and this permission notice appear in all copies.
THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES
WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL ISC BE LIABLE FOR
ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT
OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
-------------------------------------------------------------------
Copyright (c) 2007-2008 Michael G Schwern
This software originally derived from Paul Sheer's pivotal_gmtime_r.c.
@@ -3818,33 +3833,6 @@ SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (c) 2009 The NetBSD Foundation, Inc.
This code is derived from software contributed to The NetBSD Foundation
by Roy Marples.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. 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.
THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``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 AUTHOR 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.
-------------------------------------------------------------------
Copyright (c) 2010 MIPS Technologies, Inc.
All rights reserved.
@@ -3877,36 +3865,6 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (c) 2010, 2011 Intel 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 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.
-------------------------------------------------------------------
Copyright (c) 2010, Intel Corporation
All rights reserved.
@@ -3937,36 +3895,6 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (c) 2011 Intel 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 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.
-------------------------------------------------------------------
Copyright (c) 2011 The Android Open Source Project
Copyright (c) 2008 ARM Ltd
All rights reserved.
@@ -3996,36 +3924,6 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (c) 2011, Intel 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 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.
-------------------------------------------------------------------
Copyright (c) 2011, VMware, Inc.
All rights reserved.
@@ -4081,66 +3979,6 @@ SUCH DAMAGE.
-------------------------------------------------------------------
Copyright (c) 2013 ARM Ltd
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. 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.
3. The name of the company may not be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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.
-------------------------------------------------------------------
Copyright (c) 2013, 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 Linaro Limited 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.
-------------------------------------------------------------------
Copyright (c)1999 Citrus Project,
All rights reserved.

54
libc/README Normal file
View File

@@ -0,0 +1,54 @@
Welcome to Bionic, Android's small and custom C library for the Android
platform.
Bionic is mainly a port of the BSD C library to our Linux kernel with the
following additions/changes:
- no support for locales
- no support for wide chars (i.e. multi-byte characters)
- its own smallish implementation of pthreads based on Linux futexes
- support for x86, ARM and ARM thumb CPU instruction sets and kernel interfaces
Bionic is released under the standard 3-clause BSD License
Bionic doesn't want to implement all features of a traditional C library, we only
add features to it as we need them, and we try to keep things as simple and small
as possible. Our goal is not to support scaling to thousands of concurrent threads
on multi-processors machines; we're running this on cell-phones, damnit !!
Note that Bionic doesn't provide a libthread_db or a libm implementation.
Adding new syscalls:
====================
Bionic provides the gensyscalls.py Python script to automatically generate syscall
stubs from the list defined in the file SYSCALLS.TXT. You can thus add a new syscall
by doing the following:
- edit SYSCALLS.TXT
- add a new line describing your syscall, it should look like:
return_type syscall_name(parameters) syscall_number
- in the event where you want to differentiate the syscall function from its entry name,
use the alternate:
return_type funcname:syscall_name(parameters) syscall_number
- additionally, if the syscall number is different between ARM and x86, use:
return_type funcname[:syscall_name](parameters) arm_number,x86_number
- a syscall number can be -1 to indicate that the syscall is not implemented on
a given platform, for example:
void __set_tls(void*) arm_number,-1
the comments in SYSCALLS.TXT contain more information about the line format
You can also use the 'checksyscalls.py' script to check that all the syscall
numbers you entered are correct. It does so by looking at the values defined in
your Linux kernel headers. The script indicates where the values are incorrect
and what is expected instead.

View File

@@ -1,11 +1,13 @@
# This file is used to automatically generate bionic's the system calls stubs.
# this file is used to list all the syscalls that will be supported by
# the Bionic C library. It is used to automatically generate the syscall
# stubs, the list of syscall constants (__NR_xxxx) and the content of <linux/_unistd.h>
#
# Each non comment line has the following format:
# each non comment line has the following format:
#
# return_type func_name[:syscall_name[:call_id]]([parameter_list]) (1|-1|"stub")
# return_type func_name[:syscall_name[:call_id]]([parameter_list]) (syscall_number|"stub")
#
# Note that:
# - syscall_name corresponds to the name of the syscall, which may differ from
# note that:
# - syscall_name correspond to the name of the syscall, which may differ from
# the exported function name (example: the exit syscall is implemented by the _exit()
# function, which is not the same as the standard C exit() function which calls it)
# The call_id parameter, given that func_name and syscall_name have
@@ -16,311 +18,298 @@
# - each parameter type is assumed to be stored on 32 bits, there is no plan to support
# 64-bit architectures at the moment
#
# - the final field can be "1", meaning: generate a stub for each architecture,
# taking the constants from the kernel header files.
# - it there is "stub" instead of a syscall number, the tool will not generate any
# assembler template for the syscall; it's up to the bionic implementation to provide
# a relevant C stub
#
# - the final field can be "stub" meaning: do not generate any stubs ---
# in this case, a hand-written custom stub must be provided.
# TODO: replace this with something like "custom" or "none", or remove
# it entirely.
# - additionally, if the syscall number is different amoung ARM, and x86, MIPS use:
# return_type funcname[:syscall_name](parameters) arm_number,x86_number,mips_number
#
# - the final field can be a three-element list of 1s and -1 meaning:
# this system call is only available on some of the architectures (1),
# and no stub should be generated for those architectures marked with -1.
# the order is arm,x86,mips.
# TODO: replace this with something more readable like "-arm,-mips" (meaning x86 only).
#
# This file is processed by a python script named gensyscalls.py.
# the file is processed by a python script named gensyscalls.py
#
# process management
void _exit:exit_group (int) 1
void _exit:exit_group (int) 248,252,246
void _exit_thread:exit (int) 1
pid_t __fork:fork (void) 1
pid_t _waitpid:waitpid (pid_t, int*, int, struct rusage*) -1,1,1
int __waitid:waitid(int, pid_t, struct siginfo_t*, int,void*) 1
pid_t wait4(pid_t pid, int *status, int options, struct rusage *rusage) 1
pid_t __fork:fork (void) 2
pid_t _waitpid:waitpid (pid_t, int*, int, struct rusage*) -1,7,7
int __waitid:waitid(int, pid_t, struct siginfo_t*, int,void*) 280,284,278
# NOTE: this system call is never called directly, but we list it there
# to have __NR_clone properly defined.
#
pid_t __sys_clone:clone (int, void*, int*, void*, int*) 1
pid_t __sys_clone:clone (int, void*, int*, void*, int*) 120
int execve (const char*, char* const*, char* const*) 1
int execve (const char*, char* const*, char* const*) 11
int __setuid:setuid32 (uid_t) 1,1,-1
int __setuid:setuid (uid_t) -1,-1,1
uid_t getuid:getuid32 () 1,1,-1
uid_t getuid:getuid () -1,-1,1
gid_t getgid:getgid32 () 1,1,-1
gid_t getgid:getgid () -1,-1,1
uid_t geteuid:geteuid32 () 1,1,-1
uid_t geteuid:geteuid () -1,-1,1
gid_t getegid:getegid32 () 1,1,-1
gid_t getegid:getegid () -1,-1,1
uid_t getresuid:getresuid32 (uid_t *ruid, uid_t *euid, uid_t *suid) 1,1,-1
uid_t getresuid:getresuid (uid_t *ruid, uid_t *euid, uid_t *suid) -1,-1,1
gid_t getresgid:getresgid32 (gid_t *rgid, gid_t *egid, gid_t *sgid) 1,1,-1
gid_t getresgid:getresgid (gid_t *rgid, gid_t *egid, gid_t *sgid) -1,-1,1
pid_t gettid() 1
ssize_t readahead(int, off64_t, size_t) 1
int getgroups:getgroups32(int, gid_t *) 1,1,-1
int getgroups:getgroups(int, gid_t *) -1,-1,1
pid_t getpgid(pid_t) 1
pid_t getppid() 1
pid_t getsid(pid_t) 1
pid_t setsid() 1
int setgid:setgid32(gid_t) 1,1,-1
int setgid:setgid(gid_t) -1,-1,1
int __setuid:setuid32 (uid_t) 213,213,-1
int __setuid:setuid (uid_t) -1,-1,23
uid_t getuid:getuid32 () 199,199,-1
uid_t getuid:getuid () -1,-1,24
gid_t getgid:getgid32 () 200,200,-1
gid_t getgid:getgid () -1,-1,47
uid_t geteuid:geteuid32 () 201,201,-1
uid_t geteuid:geteuid () -1,-1,49
gid_t getegid:getegid32 () 202,202,-1
gid_t getegid:getegid () -1,-1,50
uid_t getresuid:getresuid32 (uid_t *ruid, uid_t *euid, uid_t *suid) 209,209,-1
uid_t getresuid:getresuid (uid_t *ruid, uid_t *euid, uid_t *suid) -1,-1,186
gid_t getresgid:getresgid32 (gid_t *rgid, gid_t *egid, gid_t *sgid) 211,211,-1
gid_t getresgid:getresgid (gid_t *rgid, gid_t *egid, gid_t *sgid) -1,-1,191
pid_t gettid() 224,224,222
ssize_t readahead(int, off64_t, size_t) 225,225,223
int getgroups:getgroups32(int, gid_t *) 205,205,-1
int getgroups:getgroups(int, gid_t *) -1,-1,80
pid_t getpgid(pid_t) 132
pid_t getppid() 64
pid_t getsid(pid_t) 147
pid_t setsid() 66
int setgid:setgid32(gid_t) 214,214,-1
int setgid:setgid(gid_t) -1,-1,46
int seteuid:seteuid32(uid_t) stub
int __setreuid:setreuid32(uid_t, uid_t) 1,1,-1
int __setreuid:setreuid(uid_t, uid_t) -1,-1,1
int __setresuid:setresuid32(uid_t, uid_t, uid_t) 1,1,-1
int __setresuid:setresuid(uid_t, uid_t, uid_t) -1,-1,1
int setresgid:setresgid32(gid_t, gid_t, gid_t) 1,1,-1
int setresgid:setresgid(gid_t, gid_t, gid_t) -1,-1,1
void* __brk:brk(void*) 1
int __setreuid:setreuid32(uid_t, uid_t) 203,203,-1
int __setreuid:setreuid(uid_t, uid_t) -1,-1,70
int __setresuid:setresuid32(uid_t, uid_t, uid_t) 208,208,-1
int __setresuid:setresuid(uid_t, uid_t, uid_t) -1,-1,185
int setresgid:setresgid32(gid_t, gid_t, gid_t) 210,210,-1
int setresgid:setresgid(gid_t, gid_t, gid_t) -1,-1,190
void* __brk:brk(void*) 45
# see comments in arch-arm/bionic/kill.S to understand why we don't generate an ARM stub for kill/tkill
int kill(pid_t, int) -1,1,1
int tkill(pid_t tid, int sig) -1,1,1
int tgkill(pid_t tgid, pid_t tid, int sig) -1,1,1
int __ptrace:ptrace(int request, int pid, void* addr, void* data) 1
int __set_thread_area:set_thread_area(void* user_desc) -1,1,1
int __getpriority:getpriority(int, int) 1
int setpriority(int, int, int) 1
int setrlimit(int resource, const struct rlimit *rlp) 1
int getrlimit:ugetrlimit(int resource, struct rlimit *rlp) 1,1,-1
int getrlimit:getrlimit(int resource, struct rlimit *rlp) -1,-1,1
int getrusage(int who, struct rusage* r_usage) 1
int setgroups:setgroups32(int, const gid_t *) 1,1,-1
int setgroups:setgroups(int, const gid_t *) -1,-1,1
int kill(pid_t, int) -1,37,37
int tkill(pid_t tid, int sig) -1,238,236
int tgkill(pid_t tgid, pid_t tid, int sig) -1,270,266
int __ptrace:ptrace(int request, int pid, void* addr, void* data) 26
int __set_thread_area:set_thread_area(void* user_desc) -1,243,283
int __getpriority:getpriority(int, int) 96
int setpriority(int, int, int) 97
int setrlimit(int resource, const struct rlimit *rlp) 75
int getrlimit:ugetrlimit(int resource, struct rlimit *rlp) 191,191,-1
int getrlimit:getrlimit(int resource, struct rlimit *rlp) -1,-1,76
int getrusage(int who, struct rusage* r_usage) 77
int setgroups:setgroups32(int, const gid_t *) 206,206,-1
int setgroups:setgroups(int, const gid_t *) -1,-1,81
pid_t getpgrp(void) stub
int setpgid(pid_t, pid_t) 1
pid_t vfork(void) 1,-1,-1
int setregid:setregid32(gid_t, gid_t) 1,1,-1
int setregid:setregid(gid_t, gid_t) -1,-1,1
int chroot(const char *) 1
int setpgid(pid_t, pid_t) 57
pid_t vfork(void) 190,-1,-1
int setregid:setregid32(gid_t, gid_t) 204,204,-1
int setregid:setregid(gid_t, gid_t) -1,-1,71
int chroot(const char *) 61
# IMPORTANT: Even though <sys/prctl.h> declares prctl(int,...), the syscall stub must take 6 arguments
# to match the kernel implementation.
int prctl(int option, unsigned int arg2, unsigned int arg3, unsigned int arg4, unsigned int arg5) 1
int capget(cap_user_header_t header, cap_user_data_t data) 1
int capset(cap_user_header_t header, const cap_user_data_t data) 1
int sigaltstack(const stack_t*, stack_t*) 1
int acct(const char* filepath) 1
int prctl(int option, unsigned int arg2, unsigned int arg3, unsigned int arg4, unsigned int arg5) 172,172,192
int capget(cap_user_header_t header, cap_user_data_t data) 184,184,204
int capset(cap_user_header_t header, const cap_user_data_t data) 185,185,205
int sigaltstack(const stack_t*, stack_t*) 186,186,206
int acct(const char* filepath) 51
# file descriptors
ssize_t read (int, void*, size_t) 1
ssize_t write (int, const void*, size_t) 1
ssize_t pread64 (int, void *, size_t, off64_t) 1
ssize_t pwrite64 (int, void *, size_t, off64_t) 1
int __open:open (const char*, int, mode_t) 1
int __openat:openat (int, const char*, int, mode_t) 1
int close (int) 1
ssize_t read (int, void*, size_t) 3
ssize_t write (int, const void*, size_t) 4
ssize_t pread64 (int, void *, size_t, off64_t) 180,180,200
ssize_t pwrite64 (int, void *, size_t, off64_t) 181,181,201
int __open:open (const char*, int, mode_t) 5
int __openat:openat (int, const char*, int, mode_t) 322,295,288
int close (int) 6
int creat(const char*, mode_t) stub
off_t lseek(int, off_t, int) 1
int __llseek:_llseek (int, unsigned long, unsigned long, loff_t*, int) 1
pid_t getpid () 1
off_t lseek(int, off_t, int) 19
int __llseek:_llseek (int, unsigned long, unsigned long, loff_t*, int) 140
pid_t getpid () 20
void * mmap(void *, size_t, int, int, int, long) stub
void * __mmap2:mmap2(void*, size_t, int, int, int, long) 1
int munmap(void *, size_t) 1
void * mremap(void *, size_t, size_t, unsigned long) 1
int msync(const void *, size_t, int) 1
int mprotect(const void *, size_t, int) 1
int madvise(const void *, size_t, int) 1
int mlock(const void *addr, size_t len) 1
int munlock(const void *addr, size_t len) 1
int mlockall(int flags) 1
int munlockall() 1
int mincore(void* start, size_t length, unsigned char* vec) 1
int __ioctl:ioctl(int, int, void *) 1
int readv(int, const struct iovec *, int) 1
int writev(int, const struct iovec *, int) 1
int __fcntl:fcntl(int, int, void*) 1
int flock(int, int) 1
int fchmod(int, mode_t) 1
int dup(int) 1
int pipe(int *) 1,1,-1
int pipe2(int *, int) 1
int dup2(int, int) 1
int select:_newselect(int, struct fd_set *, struct fd_set *, struct fd_set *, struct timeval *) 1
int ftruncate(int, off_t) 1
int ftruncate64(int, off64_t) 1
int getdents:getdents64(unsigned int, struct dirent *, unsigned int) 1
int fsync(int) 1
int fdatasync(int) 1
int fchown:fchown32(int, uid_t, gid_t) 1,1,-1
int fchown:fchown(int, uid_t, gid_t) -1,-1,1
void sync(void) 1
int __fcntl64:fcntl64(int, int, void *) 1
int __fstatfs64:fstatfs64(int, size_t, struct statfs *) 1
ssize_t sendfile(int out_fd, int in_fd, off_t *offset, size_t count) 1
int fstatat:fstatat64(int dirfd, const char *path, struct stat *buf, int flags) 1
int mkdirat(int dirfd, const char *pathname, mode_t mode) 1
int fchownat(int dirfd, const char *path, uid_t owner, gid_t group, int flags) 1
int fchmodat(int dirfd, const char *path, mode_t mode, int flags) 1
int renameat(int olddirfd, const char *oldpath, int newdirfd, const char *newpath) 1
int fsetxattr(int, const char *, const void *, size_t, int) 1
ssize_t fgetxattr(int, const char *, void *, size_t) 1
ssize_t flistxattr(int, char *, size_t) 1
int fremovexattr(int, const char *) 1
void * __mmap2:mmap2(void*, size_t, int, int, int, long) 192,192,210
int munmap(void *, size_t) 91
void * mremap(void *, size_t, size_t, unsigned long) 163,163,167
int msync(const void *, size_t, int) 144
int mprotect(const void *, size_t, int) 125
int madvise(const void *, size_t, int) 220,219,218
int mlock(const void *addr, size_t len) 150,150,154
int munlock(const void *addr, size_t len) 151,151,155
int mlockall(int flags) 152,152,156
int munlockall() 153,153,157
int mincore(void* start, size_t length, unsigned char* vec) 219,218,217
int __ioctl:ioctl(int, int, void *) 54
int readv(int, const struct iovec *, int) 145
int writev(int, const struct iovec *, int) 146
int __fcntl:fcntl(int, int, void*) 55
int flock(int, int) 143
int fchmod(int, mode_t) 94
int dup(int) 41
int pipe(int *) 42,42,-1
int pipe2(int *, int) 359,331,328
int dup2(int, int) 63
int select:_newselect(int, struct fd_set *, struct fd_set *, struct fd_set *, struct timeval *) 142
int ftruncate(int, off_t) 93
int ftruncate64(int, off64_t) 194,194,212
int getdents:getdents64(unsigned int, struct dirent *, unsigned int) 217,220,219
int fsync(int) 118
int fdatasync(int) 148,148,152
int fchown:fchown32(int, uid_t, gid_t) 207,207,-1
int fchown:fchown(int, uid_t, gid_t) -1,-1,95
void sync(void) 36
int __fcntl64:fcntl64(int, int, void *) 221,221,220
int __fstatfs64:fstatfs64(int, size_t, struct statfs *) 267,269,256
ssize_t sendfile(int out_fd, int in_fd, off_t *offset, size_t count) 187,187,207
int fstatat:fstatat64(int dirfd, const char *path, struct stat *buf, int flags) 327,300,293
int mkdirat(int dirfd, const char *pathname, mode_t mode) 323,296,289
int fchownat(int dirfd, const char *path, uid_t owner, gid_t group, int flags) 325,298,291
int fchmodat(int dirfd, const char *path, mode_t mode, int flags) 333,306,299
int renameat(int olddirfd, const char *oldpath, int newdirfd, const char *newpath) 329,302,295
int fsetxattr(int, const char *, const void *, size_t, int) 228,228,226
ssize_t fgetxattr(int, const char *, void *, size_t) 231,231,229
ssize_t flistxattr(int, char *, size_t) 234,234,232
int fremovexattr(int, const char *) 237,237,235
# file system
int link (const char*, const char*) 1
int unlink (const char*) 1
int unlinkat (int, const char *, int) 1
int chdir (const char*) 1
int mknod (const char*, mode_t, dev_t) 1
int chmod (const char*,mode_t) 1
int chown:chown32(const char *, uid_t, gid_t) 1,1,-1
int chown:chown(const char *, uid_t, gid_t) -1,-1,1
int lchown:lchown32 (const char*, uid_t, gid_t) 1,1,-1
int lchown:lchown (const char*, uid_t, gid_t) -1,-1,1
int mount (const char*, const char*, const char*, unsigned long, const void*) 1
int link (const char*, const char*) 9
int unlink (const char*) 10
int unlinkat (int, const char *, int) 328,301,294
int chdir (const char*) 12
int mknod (const char*, mode_t, dev_t) 14
int chmod (const char*,mode_t) 15
int chown:chown32(const char *, uid_t, gid_t) 212,212,-1
int chown:chown(const char *, uid_t, gid_t) -1,-1,202
int lchown:lchown32 (const char*, uid_t, gid_t) 198,198,-1
int lchown:lchown (const char*, uid_t, gid_t) -1,-1,16
int mount (const char*, const char*, const char*, unsigned long, const void*) 21
int umount(const char*) stub
int umount2 (const char*, int) 1
int fstat:fstat64(int, struct stat*) 1
int stat:stat64(const char *, struct stat *) 1
int lstat:lstat64(const char *, struct stat *) 1
int mkdir(const char *, mode_t) 1
int readlink(const char *, char *, size_t) 1
int rmdir(const char *) 1
int rename(const char *, const char *) 1
int __getcwd:getcwd(char * buf, size_t size) 1
int access(const char *, int) 1
int faccessat(int, const char *, int, int) 1
int symlink(const char *, const char *) 1
int fchdir(int) 1
int truncate(const char*, off_t) 1
int setxattr(const char *, const char *, const void *, size_t, int) 1
int lsetxattr(const char *, const char *, const void *, size_t, int) 1
ssize_t getxattr(const char *, const char *, void *, size_t) 1
ssize_t lgetxattr(const char *, const char *, void *, size_t) 1
ssize_t listxattr(const char *, char *, size_t) 1
ssize_t llistxattr(const char *, char *, size_t) 1
int removexattr(const char *, const char *) 1
int lremovexattr(const char *, const char *) 1
int __statfs64:statfs64(const char *, size_t, struct statfs *) 1
long unshare(unsigned long) 1
int swapon(const char *, int) 1
int swapoff(const char *) 1
int umount2 (const char*, int) 52
int fstat:fstat64(int, struct stat*) 197,197,215
int stat:stat64(const char *, struct stat *) 195,195,213
int lstat:lstat64(const char *, struct stat *) 196,196,214
int mkdir(const char *, mode_t) 39
int readlink(const char *, char *, size_t) 85
int rmdir(const char *) 40
int rename(const char *, const char *) 38
int __getcwd:getcwd(char * buf, size_t size) 183,183,203
int access(const char *, int) 33
int faccessat(int, const char *, int, int) 334,307,300
int symlink(const char *, const char *) 83
int fchdir(int) 133
int truncate(const char*, off_t) 92
int setxattr(const char *, const char *, const void *, size_t, int) 226,226,224
int lsetxattr(const char *, const char *, const void *, size_t, int) 227,227,225
ssize_t getxattr(const char *, const char *, void *, size_t) 229,229,227
ssize_t lgetxattr(const char *, const char *, void *, size_t) 230,230,228
ssize_t listxattr(const char *, char *, size_t) 232,232,230
ssize_t llistxattr(const char *, char *, size_t) 233,233,231
int removexattr(const char *, const char *) 235,235,233
int lremovexattr(const char *, const char *) 236,236,234
int __statfs64:statfs64(const char *, size_t, struct statfs *) 266,268,255
long unshare(unsigned long) 337,310,303
# time
int pause () 1
int gettimeofday(struct timeval*, struct timezone*) 1
int settimeofday(const struct timeval*, const struct timezone*) 1
clock_t times(struct tms *) 1
int nanosleep(const struct timespec *, struct timespec *) 1
int clock_gettime(clockid_t clk_id, struct timespec *tp) 1
int clock_settime(clockid_t clk_id, const struct timespec *tp) 1
int clock_getres(clockid_t clk_id, struct timespec *res) 1
int clock_nanosleep(clockid_t clock_id, int flags, const struct timespec *req, struct timespec *rem) 1
int getitimer(int, const struct itimerval *) 1
int setitimer(int, const struct itimerval *, struct itimerval *) 1
int __timer_create:timer_create(clockid_t clockid, struct sigevent *evp, timer_t *timerid) 1
int __timer_settime:timer_settime(timer_t, int, const struct itimerspec*, struct itimerspec*) 1
int __timer_gettime:timer_gettime(timer_t, struct itimerspec*) 1
int __timer_getoverrun:timer_getoverrun(timer_t) 1
int __timer_delete:timer_delete(timer_t) 1
int utimes(const char*, const struct timeval tvp[2]) 1
int utimensat(int, const char *, const struct timespec times[2], int) 1
int timerfd_create(clockid_t, int) 1
int timerfd_settime(int, int, const struct itimerspec *, struct itimerspec *) 1
int timerfd_gettime(int, struct itimerspec *) 1
int pause () 29
int gettimeofday(struct timeval*, struct timezone*) 78
int settimeofday(const struct timeval*, const struct timezone*) 79
clock_t times(struct tms *) 43
int nanosleep(const struct timespec *, struct timespec *) 162,162,166
int clock_gettime(clockid_t clk_id, struct timespec *tp) 263,265,263
int clock_settime(clockid_t clk_id, const struct timespec *tp) 262,264,262
int clock_getres(clockid_t clk_id, struct timespec *res) 264,266,264
int clock_nanosleep(clockid_t clock_id, int flags, const struct timespec *req, struct timespec *rem) 265,267,265
int getitimer(int, const struct itimerval *) 105
int setitimer(int, const struct itimerval *, struct itimerval *) 104
int __timer_create:timer_create(clockid_t clockid, struct sigevent *evp, timer_t *timerid) 257,259,257
int __timer_settime:timer_settime(timer_t, int, const struct itimerspec*, struct itimerspec*) 258,260,258
int __timer_gettime:timer_gettime(timer_t, struct itimerspec*) 259,261,259
int __timer_getoverrun:timer_getoverrun(timer_t) 260,262,260
int __timer_delete:timer_delete(timer_t) 261,263,261
int utimes(const char*, const struct timeval tvp[2]) 269,271,267
int utimensat(int, const char *, const struct timespec times[2], int) 348,320,316
# signals
int sigaction(int, const struct sigaction *, struct sigaction *) 1
int sigprocmask(int, const sigset_t *, sigset_t *) 1
int __sigsuspend:sigsuspend(int unused1, int unused2, unsigned mask) 1,1,-1
int __sigsuspend:sigsuspend(const sigset_t *mask) -1,-1,1
int __rt_sigaction:rt_sigaction (int sig, const struct sigaction *act, struct sigaction *oact, size_t sigsetsize) 1
int __rt_sigprocmask:rt_sigprocmask (int how, const sigset_t *set, sigset_t *oset, size_t sigsetsize) 1
int __rt_sigtimedwait:rt_sigtimedwait(const sigset_t *set, struct siginfo_t *info, struct timespec_t *timeout, size_t sigset_size) 1
int sigpending(sigset_t *) 1
int signalfd4(int fd, const sigset_t *mask, size_t sizemask, int flags) 1
int sigaction(int, const struct sigaction *, struct sigaction *) 67
int sigprocmask(int, const sigset_t *, sigset_t *) 126
int __sigsuspend:sigsuspend(int unused1, int unused2, unsigned mask) 72,72,-1
int __sigsuspend:sigsuspend(const sigset_t *mask) -1,-1,72
int __rt_sigaction:rt_sigaction (int sig, const struct sigaction *act, struct sigaction *oact, size_t sigsetsize) 174,174,194
int __rt_sigprocmask:rt_sigprocmask (int how, const sigset_t *set, sigset_t *oset, size_t sigsetsize) 175,175,195
int __rt_sigtimedwait:rt_sigtimedwait(const sigset_t *set, struct siginfo_t *info, struct timespec_t *timeout, size_t sigset_size) 177,177,197
int sigpending(sigset_t *) 73
# sockets
int socket(int, int, int) 1,-1,1
int socketpair(int, int, int, int*) 1,-1,1
int bind(int, struct sockaddr *, int) 1,-1,1
int connect(int, struct sockaddr *, socklen_t) 1,-1,1
int listen(int, int) 1,-1,1
int accept(int, struct sockaddr *, socklen_t *) 1,-1,1
int getsockname(int, struct sockaddr *, socklen_t *) 1,-1,1
int getpeername(int, struct sockaddr *, socklen_t *) 1,-1,1
int sendto(int, const void *, size_t, int, const struct sockaddr *, socklen_t) 1,-1,1
int recvfrom(int, void *, size_t, unsigned int, struct sockaddr *, socklen_t *) 1,-1,1
int shutdown(int, int) 1,-1,1
int setsockopt(int, int, int, const void *, socklen_t) 1,-1,1
int getsockopt(int, int, int, void *, socklen_t *) 1,-1,1
int sendmsg(int, const struct msghdr *, unsigned int) 1,-1,1
int recvmsg(int, struct msghdr *, unsigned int) 1,-1,1
int socket(int, int, int) 281,-1,183
int socketpair(int, int, int, int*) 288,-1,184
int bind(int, struct sockaddr *, int) 282,-1,169
int connect(int, struct sockaddr *, socklen_t) 283,-1,170
int listen(int, int) 284,-1,174
int accept(int, struct sockaddr *, socklen_t *) 285,-1,168
int getsockname(int, struct sockaddr *, socklen_t *) 286,-1,172
int getpeername(int, struct sockaddr *, socklen_t *) 287,-1,171
int sendto(int, const void *, size_t, int, const struct sockaddr *, socklen_t) 290,-1,180
int recvfrom(int, void *, size_t, unsigned int, struct sockaddr *, socklen_t *) 292,-1,176
int shutdown(int, int) 293,-1,182
int setsockopt(int, int, int, const void *, socklen_t) 294,-1,181
int getsockopt(int, int, int, void *, socklen_t *) 295,-1,173
int sendmsg(int, const struct msghdr *, unsigned int) 296,-1,179
int recvmsg(int, struct msghdr *, unsigned int) 297,-1,177
# sockets for x86. These are done as an "indexed" call to socketcall syscall.
int socket:socketcall:1 (int, int, int) -1,1,-1
int bind:socketcall:2 (int, struct sockaddr *, int) -1,1,-1
int connect:socketcall:3(int, struct sockaddr *, socklen_t) -1,1,-1
int listen:socketcall:4(int, int) -1,1,-1
int accept:socketcall:5(int, struct sockaddr *, socklen_t *) -1,1,-1
int getsockname:socketcall:6(int, struct sockaddr *, socklen_t *) -1,1,-1
int getpeername:socketcall:7(int, struct sockaddr *, socklen_t *) -1,1,-1
int socketpair:socketcall:8(int, int, int, int*) -1,1,-1
int sendto:socketcall:11(int, const void *, size_t, int, const struct sockaddr *, socklen_t) -1,1,-1
int recvfrom:socketcall:12(int, void *, size_t, unsigned int, struct sockaddr *, socklen_t *) -1,1,-1
int shutdown:socketcall:13(int, int) -1,1,-1
int setsockopt:socketcall:14(int, int, int, const void *, socklen_t) -1,1,-1
int getsockopt:socketcall:15(int, int, int, void *, socklen_t *) -1,1,-1
int sendmsg:socketcall:16(int, const struct msghdr *, unsigned int) -1,1,-1
int recvmsg:socketcall:17(int, struct msghdr *, unsigned int) -1,1,-1
int socket:socketcall:1 (int, int, int) -1,102,-1
int bind:socketcall:2 (int, struct sockaddr *, int) -1,102,-1
int connect:socketcall:3(int, struct sockaddr *, socklen_t) -1,102,-1
int listen:socketcall:4(int, int) -1,102,-1
int accept:socketcall:5(int, struct sockaddr *, socklen_t *) -1,102,-1
int getsockname:socketcall:6(int, struct sockaddr *, socklen_t *) -1,102,-1
int getpeername:socketcall:7(int, struct sockaddr *, socklen_t *) -1,102,-1
int socketpair:socketcall:8(int, int, int, int*) -1,102,-1
int sendto:socketcall:11(int, const void *, size_t, int, const struct sockaddr *, socklen_t) -1,102,-1
int recvfrom:socketcall:12(int, void *, size_t, unsigned int, struct sockaddr *, socklen_t *) -1,102,-1
int shutdown:socketcall:13(int, int) -1,102,-1
int setsockopt:socketcall:14(int, int, int, const void *, socklen_t) -1,102,-1
int getsockopt:socketcall:15(int, int, int, void *, socklen_t *) -1,102,-1
int sendmsg:socketcall:16(int, const struct msghdr *, unsigned int) -1,102,-1
int recvmsg:socketcall:17(int, struct msghdr *, unsigned int) -1,102,-1
# scheduler & real-time
int sched_setscheduler(pid_t pid, int policy, const struct sched_param *param) 1
int sched_getscheduler(pid_t pid) 1
int sched_yield(void) 1
int sched_setparam(pid_t pid, const struct sched_param *param) 1
int sched_getparam(pid_t pid, struct sched_param *param) 1
int sched_get_priority_max(int policy) 1
int sched_get_priority_min(int policy) 1
int sched_rr_get_interval(pid_t pid, struct timespec *interval) 1
int sched_setaffinity(pid_t pid, size_t setsize, const cpu_set_t* set) 1
int __sched_getaffinity:sched_getaffinity(pid_t pid, size_t setsize, cpu_set_t* set) 1
int __getcpu:getcpu(unsigned *cpu, unsigned *node, void *unused) 1
int sched_setscheduler(pid_t pid, int policy, const struct sched_param *param) 156,156,160
int sched_getscheduler(pid_t pid) 157,157,161
int sched_yield(void) 158,158,162
int sched_setparam(pid_t pid, const struct sched_param *param) 154,154,158
int sched_getparam(pid_t pid, struct sched_param *param) 155,155,159
int sched_get_priority_max(int policy) 159,159,163
int sched_get_priority_min(int policy) 160,160,164
int sched_rr_get_interval(pid_t pid, struct timespec *interval) 161,161,165
int sched_setaffinity(pid_t pid, size_t setsize, const cpu_set_t* set) 241,241,239
int __sched_getaffinity:sched_getaffinity(pid_t pid, size_t setsize, cpu_set_t* set) 242,242,240
int __getcpu:getcpu(unsigned *cpu, unsigned *node, void *unused) 345,318,312
# io priorities
int ioprio_set(int which, int who, int ioprio) 1
int ioprio_get(int which, int who) 1
int ioprio_set(int which, int who, int ioprio) 314,289,314
int ioprio_get(int which, int who) 315,290,315
# other
int uname(struct utsname *) 1
mode_t umask(mode_t) 1
int __reboot:reboot(int, int, int, void *) 1
int __syslog:syslog(int, char *, int) 1
int init_module(void *, unsigned long, const char *) 1
int delete_module(const char*, unsigned int) 1
int klogctl:syslog(int, char *, int) 1
int sysinfo(struct sysinfo *) 1
int personality(unsigned long) 1
long perf_event_open(struct perf_event_attr *attr_uptr, pid_t pid, int cpu, int group_fd, unsigned long flags) 1
int uname(struct utsname *) 122
pid_t __wait4:wait4(pid_t pid, int *status, int options, struct rusage *rusage) 114
mode_t umask(mode_t) 60
int __reboot:reboot(int, int, int, void *) 88
int __syslog:syslog(int, char *, int) 103
int init_module(void *, unsigned long, const char *) 128
int delete_module(const char*, unsigned int) 129
int klogctl:syslog(int, char *, int) 103
int sysinfo(struct sysinfo *) 116
int personality(unsigned long) 136
long perf_event_open(struct perf_event_attr *attr_uptr, pid_t pid, int cpu, int group_fd, unsigned long flags) 364
# futex
int futex(void *, int, int, void *, void *, int) 1
int futex(void *, int, int, void *, void *, int) 240,240,238
# epoll
int epoll_create(int size) 1
int epoll_ctl(int epfd, int op, int fd, struct epoll_event *event) 1
int epoll_wait(int epfd, struct epoll_event *events, int max, int timeout) 1
int epoll_create(int size) 250,254,248
int epoll_ctl(int epfd, int op, int fd, struct epoll_event *event) 251,255,249
int epoll_wait(int epfd, struct epoll_event *events, int max, int timeout) 252,256,250
int inotify_init(void) 1
int inotify_add_watch(int, const char *, unsigned int) 1
int inotify_rm_watch(int, unsigned int) 1
int inotify_init(void) 316,291,284
int inotify_add_watch(int, const char *, unsigned int) 317,292,285
int inotify_rm_watch(int, unsigned int) 318,293,286
int poll(struct pollfd *, unsigned int, long) 1
int poll(struct pollfd *, unsigned int, long) 168,168,188
int eventfd:eventfd2(unsigned int, int) 1
int eventfd:eventfd2(unsigned int, int) 356,328,325
# ARM-specific ARM_NR_BASE == 0x0f0000 == 983040
int __set_tls:__ARM_NR_set_tls(void*) 1,-1,-1
int cacheflush:__ARM_NR_cacheflush(long start, long end, long flags) 1,-1,-1
int __set_tls:ARM_set_tls(void*) 983045,-1,-1
int cacheflush:ARM_cacheflush(long start, long end, long flags) 983042,-1,-1
# MIPS-specific
int _flush_cache:cacheflush(char *addr, const int nbytes, const int op) -1,-1,1
int syscall(int number,...) -1,-1,1
int _flush_cache:cacheflush(char *addr, const int nbytes, const int op) -1,-1,147
int syscall(int number,...) -1,-1,0

View File

@@ -1,44 +0,0 @@
_LIBC_ARCH_COMMON_SRC_FILES := \
arch-arm/bionic/abort_arm.S \
arch-arm/bionic/atomics_arm.c \
arch-arm/bionic/clone.S \
arch-arm/bionic/eabi.c \
arch-arm/bionic/_exit_with_stack_teardown.S \
arch-arm/bionic/ffs.S \
arch-arm/bionic/futex_arm.S \
arch-arm/bionic/__get_sp.S \
arch-arm/bionic/kill.S \
arch-arm/bionic/libgcc_compat.c \
arch-arm/bionic/memcmp16.S \
arch-arm/bionic/memcmp.S \
arch-arm/bionic/_setjmp.S \
arch-arm/bionic/setjmp.S \
arch-arm/bionic/sigsetjmp.S \
arch-arm/bionic/syscall.S \
arch-arm/bionic/tgkill.S \
arch-arm/bionic/tkill.S \
# These are used by the static and dynamic versions of the libc
# respectively.
_LIBC_ARCH_STATIC_SRC_FILES := \
arch-arm/bionic/exidx_static.c
_LIBC_ARCH_DYNAMIC_SRC_FILES := \
arch-arm/bionic/exidx_dynamic.c
# Remove the C++ fortify function implementations for which there is an
# arm assembler version.
_LIBC_FORTIFY_FILES_TO_REMOVE := \
bionic/__memcpy_chk.cpp \
bionic/__memset_chk.cpp \
bionic/__strcpy_chk.cpp \
bionic/__strcat_chk.cpp \
libc_common_src_files := \
$(filter-out $(_LIBC_FORTIFY_FILES_TO_REMOVE),$(libc_common_src_files))
ifeq ($(strip $(wildcard bionic/libc/arch-arm/$(TARGET_CPU_VARIANT)/$(TARGET_CPU_VARIANT).mk)),)
$(error "TARGET_CPU_VARIANT not set or set to an unknown value. Possible values are cortex-a7, cortex-a8, cortex-a9, cortex-a15, krait. Use generic for devices that do not have a CPU similar to any of the supported cpu variants.")
endif
include bionic/libc/arch-arm/$(TARGET_CPU_VARIANT)/$(TARGET_CPU_VARIANT).mk

View File

@@ -25,10 +25,10 @@
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
.global __get_sp
.type __get_sp, %function
#include <machine/asm.h>
__get_sp:
mov r0, sp
bx lr
ENTRY(__get_sp)
mov r0, sp
bx lr
END(__get_sp)

View File

@@ -25,20 +25,27 @@
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <machine/asm.h>
#include <asm/unistd.h>
#include <machine/asm.h>
// void _exit_with_stack_teardown(void* stackBase, int stackSize, int retCode)
@ void _exit_with_stack_teardown(void * stackBase, int stackSize, int retCode)
ENTRY(_exit_with_stack_teardown)
#if __ARM_EABI__
mov lr, r2
ldr r7, =__NR_munmap
swi #0 // the stack is destroyed by this call
swi #0 @ the stack is destroyed by this call
mov r0, lr
ldr r7, =__NR_exit
swi #0
#else
mov lr, r2
swi # __NR_munmap @ the stack is destroyed by this call
mov r0, lr
swi # __NR_exit
#endif
// exit() should never return, cause a crash if it does
mov r0, #0
ldr r0, [r0]
@ exit() should never return, cause a crash if it does
mov r0, #0
ldr r0, [r0]
END(_exit_with_stack_teardown)

View File

@@ -38,5 +38,5 @@
ENTRY(abort)
.save {r3, r14}
stmfd sp!, {r3, r14}
bl PIC_SYM(_C_LABEL(__libc_android_abort), PLT)
blx PIC_SYM(_C_LABEL(__libc_android_abort), PLT)
END(abort)

View File

@@ -27,7 +27,7 @@
*/
#include <sys/types.h>
#include <private/libc_logging.h>
#include <private/logd.h>
#include <stdio.h>
/*
@@ -50,9 +50,10 @@ atexit(void (*func)(void))
* calling library may have been dlclose()'d, causing the program to
* crash.
*/
static char const warning[] = "WARNING: generic atexit() called from legacy shared library\n";
static char const warning[] =
"WARNING: generic atexit() called from legacy shared library\n";
__libc_format_log(ANDROID_LOG_WARN, "libc", warning);
__libc_android_log_print(ANDROID_LOG_WARN, "libc", warning);
fprintf(stderr, warning);
return (__cxa_atexit((void (*)(void *))func, NULL, NULL));

View File

@@ -25,42 +25,47 @@
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <linux/err.h>
#include <sys/linux-syscalls.h>
#include <machine/asm.h>
#include <asm/unistd.h>
// int __pthread_clone(void* (*fn)(void*), void* child_stack, int flags, void* arg);
ENTRY(__pthread_clone)
# Push 'fn' and 'arg' onto 'child_stack'.
stmdb r1!, {r0, r3}
@ insert the args onto the new stack
stmdb r1!, {r0, r3}
@ do the system call
@ get flags
# The sys_clone system call only takes two arguments: 'flags' and 'child_stack'.
# 'child_stack' is already in r1, but we need to move 'flags' into position.
mov r0, r2
# System call.
mov ip, r7
@ new sp is already in r1
#if __ARM_EABI__
stmfd sp!, {r4, r7}
ldr r7, =__NR_clone
swi #0
#else
swi #__NR_clone
#endif
# Child?
movs r0, r0
beq 1f
#if __ARM_EABI__
ldmnefd sp!, {r4, r7}
#endif
blt __error
bxne lr
# Parent.
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
1: # Child.
# Pop 'fn' and 'arg' back off the stack and call __thread_entry.
@ pick the function arg and call address off the stack and jump
@ to the C __thread_entry function which does some setup and then
@ calls the thread's start function
pop {r0, r1}
# __thread_entry also needs our stack pointer.
mov r2, sp
mov r2, sp @ __thread_entry needs the TLS pointer
b __thread_entry
__error:
mov r0, #-1
bx lr
END(__pthread_clone)
@@ -71,8 +76,8 @@ END(__pthread_clone)
# pid_t *pid, void *tls, pid_t *ctid,
# int (*fn)(void *), void* arg );
#
# NOTE: This is not the same signature as the glibc
# __clone function. Placing 'fn' and 'arg'
# NOTE: This is not the same signature than the GLibc
# __clone function here !! Placing 'fn' and 'arg'
# at the end of the parameter list makes the
# implementation much simpler.
#
@@ -91,20 +96,20 @@ ENTRY(__bionic_clone)
str r5, [r1, #-4]
str r6, [r1, #-8]
# System call
# system call
ldr r7, =__NR_clone
swi #0
movs r0, r0
beq 1f
# In the parent, reload saved registers then either return or set errno.
# in parent, reload saved registers
# then either exit or error
#
ldmfd sp!, {r4, r5, r6, r7}
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
bxne lr
b __set_syscall_errno
1: # The child.
1: # in the child - pick arguments
ldr r0, [sp, #-4]
ldr r1, [sp, #-8]
b __bionic_clone_entry

View File

@@ -26,9 +26,21 @@
* SUCH DAMAGE.
*/
#include "../../bionic/libc_init_common.h"
#include <stddef.h>
#include <stdint.h>
typedef struct
{
void (**preinit_array)(void);
void (**init_array)(void);
void (**fini_array)(void);
} structors_array_t;
extern int main(int argc, char **argv, char **env);
extern void __libc_init(
unsigned int *elfdata,
void (*onexit)(void),
int (*slingshot)(int, char**, char**),
structors_array_t const * const structors
);
__attribute__ ((section (".preinit_array")))
void (*__PREINIT_ARRAY__)(void) = (void (*)(void)) -1;
@@ -39,14 +51,17 @@ void (*__INIT_ARRAY__)(void) = (void (*)(void)) -1;
__attribute__ ((section (".fini_array")))
void (*__FINI_ARRAY__)(void) = (void (*)(void)) -1;
__LIBC_HIDDEN__ void _start() {
__attribute__((visibility("hidden")))
void _start() {
structors_array_t array;
array.preinit_array = &__PREINIT_ARRAY__;
array.init_array = &__INIT_ARRAY__;
array.fini_array = &__FINI_ARRAY__;
void *elfdata;
void* raw_args = (void*) ((uintptr_t) __builtin_frame_address(0) + sizeof(void*));
__libc_init(raw_args, NULL, &main, &array);
array.preinit_array = &__PREINIT_ARRAY__;
array.init_array = &__INIT_ARRAY__;
array.fini_array = &__FINI_ARRAY__;
elfdata = __builtin_frame_address(0) + sizeof(void *);
__libc_init(elfdata, (void *) 0, &main, &array);
}
#include "__dso_handle.h"

View File

@@ -31,11 +31,55 @@
#include <machine/asm.h>
#include <machine/cpu-features.h>
/*
* ffs - find first set bit, this algorithm isolates the first set
* bit, then multiplies the number by 0x0450fbaf which leaves the top
* 6 bits as an index into the table. This algorithm should be a win
* over the checking each bit in turn as per the C compiled version.
*
* Some newer ARM architectures have an instruction named
* CLZ (count leading Zero's) that is used
*
* This is the ffs algorithm devised by d.seal and posted to comp.sys.arm on
* 16 Feb 1994.
*/
ENTRY(ffs)
/* Standard trick to isolate bottom bit in r0 or 0 if r0 = 0 on entry */
rsb r1, r0, #0
ands r0, r0, r1
#ifndef __ARM_HAVE_CLZ
/*
* now r0 has at most one set bit, call this X
* if X = 0, all further instructions are skipped
*/
adrne r2, .L_ffs_table
orrne r0, r0, r0, lsl #4 /* r0 = X * 0x11 */
orrne r0, r0, r0, lsl #6 /* r0 = X * 0x451 */
rsbne r0, r0, r0, lsl #16 /* r0 = X * 0x0450fbaf */
/* now lookup in table indexed on top 6 bits of r0 */
ldrneb r0, [ r2, r0, lsr #26 ]
bx lr
END(ffs)
.text;
.type .L_ffs_table, _ASM_TYPE_OBJECT;
.L_ffs_table:
/* 0 1 2 3 4 5 6 7 */
.byte 0, 1, 2, 13, 3, 7, 0, 14 /* 0- 7 */
.byte 4, 0, 8, 0, 0, 0, 0, 15 /* 8-15 */
.byte 11, 5, 0, 0, 9, 0, 0, 26 /* 16-23 */
.byte 0, 0, 0, 0, 0, 22, 28, 16 /* 24-31 */
.byte 32, 12, 6, 0, 0, 0, 0, 0 /* 32-39 */
.byte 10, 0, 0, 25, 0, 0, 21, 27 /* 40-47 */
.byte 31, 0, 0, 0, 0, 24, 0, 20 /* 48-55 */
.byte 30, 0, 23, 19, 29, 18, 17, 0 /* 56-63 */
#else /* !defined(__ARM_HAVE_CLZ) */
clzne r0, r0
rsbne r0, r0, #32
bx lr
END(ffs)
#endif /* !defined(__ARM_HAVE_CLZ) */

View File

@@ -25,46 +25,88 @@
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <asm/unistd.h>
#include <sys/linux-syscalls.h>
#include <machine/asm.h>
#include <machine/cpu-features.h>
#define FUTEX_WAIT 0
#define FUTEX_WAKE 1
// __futex_syscall3(*ftx, op, val)
/* __futex_wait(*ftx, val, *timespec) */
/* __futex_wake(*ftx, counter) */
/* __futex_syscall3(*ftx, op, val) */
/* __futex_syscall4(*ftx, op, val, *timespec) */
.global __futex_wait
.type __futex_wait, %function
.global __futex_wake
.type __futex_wake, %function
.global __futex_syscall3
.type __futex_syscall3, %function
.global __futex_syscall4
.type __futex_syscall4, %function
#if __ARM_EABI__
ENTRY(__futex_syscall3)
mov ip, r7
stmdb sp!, {r4, r7}
.save {r4, r7}
ldr r7, =__NR_futex
swi #0
mov r7, ip
ldmia sp!, {r4, r7}
bx lr
END(__futex_syscall3)
// __futex_syscall4(*ftx, op, val, *timespec)
ENTRY(__futex_syscall4)
b __futex_syscall3
END(__futex_syscall4)
// __futex_wait(*ftx, val, *timespec)
ENTRY(__futex_wait)
mov ip, r7
stmdb sp!, {r4, r7}
.save {r4, r7}
mov r3, r2
mov r2, r1
mov r1, #FUTEX_WAIT
ldr r7, =__NR_futex
swi #0
mov r7, ip
ldmia sp!, {r4, r7}
bx lr
END(__futex_wait)
// __futex_wake(*ftx, counter)
ENTRY(__futex_wake)
mov ip, r7
.save {r4, r7}
stmdb sp!, {r4, r7}
mov r2, r1
mov r1, #FUTEX_WAKE
ldr r7, =__NR_futex
swi #0
mov r7, ip
ldmia sp!, {r4, r7}
bx lr
END(__futex_wake)
#else
ENTRY(__futex_syscall3)
swi #__NR_futex
bx lr
END(__futex_syscall3)
ENTRY(__futex_wait)
mov r3, r2
mov r2, r1
mov r1, #FUTEX_WAIT
swi #__NR_futex
bx lr
END(__futex_wait)
ENTRY(__futex_wake)
mov r2, r1
mov r1, #FUTEX_WAKE
swi #__NR_futex
bx lr
END(__futex_wake)
#endif
ENTRY(__futex_syscall4)
b __futex_syscall3
END(__futex_syscall4)

View File

@@ -25,11 +25,6 @@
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <linux/err.h>
#include <asm/unistd.h>
#include <machine/asm.h>
/* unlike our auto-generated syscall stubs, this code saves lr
on the stack, as well as a few other registers. this makes
our stack unwinder happy, when we generate debug stack
@@ -37,14 +32,19 @@
abort due to a fatal runtime error (e.g. detection
of a corrupted malloc heap).
*/
#include <sys/linux-syscalls.h>
#include <machine/asm.h>
#ifndef __NR_kill
#define __NR_kill 37
#endif
ENTRY(kill)
stmfd sp!, {r4-r7, ip, lr}
ldr r7, =__NR_kill
swi #0
ldmfd sp!, {r4-r7, ip, lr}
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
movs r0, r0
bxpl lr
b __set_syscall_errno
END(kill)

View File

@@ -30,8 +30,8 @@
* dynamic linker to copy their definition into the final libc.so binary.
*
* They are required to ensure backwards binary compatibility with
* libc.so provided by the platform and binaries built with the NDK or
* different versions/configurations of toolchains.
* Android 1.5, 1.6 and even 3.0 system images. Some applications built
* using the NDK require them to be here.
*
* Now, for a more elaborate description of the issue:
*
@@ -48,9 +48,7 @@
* gcc <options> -o libfoo.so foo.a libgcc.a -lc -lm
*
* This ensures that any helper function needed by the code in foo.a is copied
* into the final libfoo.so. However, doing so will link a bunch of other __cxa
* functions from libgcc.a into each .so and executable, causing 4k+ increase
* in every binary. Therefore the Android platform build system has been
* into the final libfoo.so. Unfortunately, the Android build system has been
* using this instead:
*
* gcc <options> -o libfoo.so foo.a -lc -lm libgcc.a
@@ -60,10 +58,9 @@
* into libfoo.so. Instead, a symbol import definition will be added to it
* so libfoo.so can directly call the one in libc.so at runtime.
*
* When refreshing toolchains for new versions or using different architecture
* flags, the set of helper functions copied to libc.so may change, which
* resulted in some native shared libraries generated with the NDK or prebuilts
* from vendors to fail to load properly.
* When changing toolchains for 2.0, the set of helper functions copied to
* libc.so changed, which resulted in some native shared libraries generated
* with the NDK to fail to load properly.
*
* The NDK has been fixed after 1.6_r1 to use the correct link command, so
* any native shared library generated with it should now be safe from that
@@ -76,11 +73,6 @@
* but it is easier to add a single function here than asking several app
* developers to fix their build.
*
* The __aeabi_idiv function is added to the list since cortex-a15 supports
* HW idiv instructions so the system libc.so doesn't pull in the reference to
* __aeabi_idiv but legacy libraries built against cortex-a9 targets still need
* it.
*
* Final note: some of the functions below should really be in libm.so to
* completely reflect the state of 1.5/1.6 system images. However,
* since libm.so depends on libc.so, it's easier to put all of
@@ -118,10 +110,8 @@
XX(__aeabi_fsub) \
XX(__aeabi_i2d) \
XX(__aeabi_i2f) \
XX(__aeabi_idiv) \
XX(__aeabi_l2d) \
XX(__aeabi_l2f) \
XX(__aeabi_lasr) \
XX(__aeabi_lmul) \
XX(__aeabi_llsl) \
XX(__aeabi_llsr) \
@@ -132,7 +122,7 @@
XX(__cmpdf2) \
XX(__divdf3) \
XX(__divsf3) \
XX(__eqdf2) \
XX(__eqdf2) \
XX(__extendsfdf2) \
XX(__fixdfsi) \
XX(__fixsfsi) \

View File

@@ -29,93 +29,44 @@
#include <machine/cpu-features.h>
#include <machine/asm.h>
#ifdef HAVE_32_BYTE_CACHE_LINE
#define CACHE_LINE_SIZE 32
#else
#define CACHE_LINE_SIZE 64
#endif
/*
* Optimized memcmp() for Cortex-A9.
* Optimized memcmp() for ARM9.
* This would not be optimal on XScale or ARM11, where more prefetching
* and use of PLD will be needed.
* The 2 major optimzations here are
* (1) The main loop compares 16 bytes at a time
* (2) The loads are scheduled in a way they won't stall
*/
ENTRY(memcmp)
pld [r0, #(CACHE_LINE_SIZE * 0)]
pld [r0, #(CACHE_LINE_SIZE * 1)]
PLD (r0, #0)
PLD (r1, #0)
/* take of the case where length is 0 or the buffers are the same */
cmp r0, r1
cmpne r2, #0
moveq r0, #0
bxeq lr
pld [r1, #(CACHE_LINE_SIZE * 0)]
pld [r1, #(CACHE_LINE_SIZE * 1)]
/* make sure we have at least 8+4 bytes, this simplify things below
* and avoid some overhead for small blocks
*/
cmp r2, #(8+4)
bmi 10f
/*
* Neon optimization
* Comparing 32 bytes at a time
*/
#if defined(__ARM_NEON__) && defined(NEON_UNALIGNED_ACCESS)
subs r2, r2, #32
blo 3f
/* preload all the cache lines we need. */
pld [r0, #(CACHE_LINE_SIZE * 2)]
pld [r1, #(CACHE_LINE_SIZE * 2)]
1: /* The main loop compares 32 bytes at a time */
vld1.8 {d0 - d3}, [r0]!
pld [r0, #(CACHE_LINE_SIZE * 2)]
vld1.8 {d4 - d7}, [r1]!
pld [r1, #(CACHE_LINE_SIZE * 2)]
/* Start subtracting the values and merge results */
vsub.i8 q0, q2
vsub.i8 q1, q3
vorr q2, q0, q1
vorr d4, d5
vmov r3, ip, d4
/* Check if there are any differences among the 32 bytes */
orrs r3, ip
bne 2f
subs r2, r2, #32
bhs 1b
b 3f
2:
/* Check if the difference was in the first or last 16 bytes */
sub r0, #32
vorr d0, d1
sub r1, #32
vmov r3, ip, d0
orrs r3, ip
/* if the first 16 bytes are equal, we only have to rewind 16 bytes */
ittt eq
subeq r2, #16
addeq r0, #16
addeq r1, #16
3: /* fix-up the remaining count */
add r2, r2, #32
cmp r2, #(8+4)
bmi 10f
#endif
.save {r4, lr}
/* save registers */
stmfd sp!, {r4, lr}
PLD (r0, #32)
PLD (r1, #32)
/* since r0 hold the result, move the first source
* pointer somewhere else
*/
mov r4, r0
/* make sure we have at least 8+4 bytes, this simplify things below
* and avoid some overhead for small blocks
*/
cmp r2, #(8+4)
bmi 8f
/* align first pointer to word boundary
* offset = -src & 3
*/
@@ -151,9 +102,9 @@ ENTRY(memcmp)
ldr ip, [r1]
subs r2, r2, #(32 + 4)
bmi 1f
0: pld [r4, #(CACHE_LINE_SIZE * 2)]
pld [r1, #(CACHE_LINE_SIZE * 2)]
0: PLD (r4, #64)
PLD (r1, #64)
ldr r0, [r4], #4
ldr lr, [r1, #4]!
eors r0, r0, ip
@@ -178,14 +129,14 @@ ENTRY(memcmp)
ldreq r0, [r4], #4
ldreq ip, [r1, #4]!
eoreqs r0, r0, lr
bne 2f
bne 2f
subs r2, r2, #32
bhs 0b
/* do we have at least 4 bytes left? */
1: adds r2, r2, #(32 - 4 + 4)
bmi 4f
/* finish off 4 bytes at a time */
3: ldr r0, [r4], #4
ldr ip, [r1], #4
@@ -219,28 +170,19 @@ ENTRY(memcmp)
9: /* restore registers and return */
ldmfd sp!, {r4, lr}
bx lr
END(memcmp)
10: /* process less than 12 bytes */
cmp r2, #0
moveq r0, #0
bxeq lr
mov r3, r0
11:
ldrb r0, [r3], #1
ldrb ip, [r1], #1
subs r0, ip
bxne lr
subs r2, r2, #1
bne 11b
bx lr
5: /*************** non-congruent case ***************/
and r0, r1, #3
and r0, r1, #3
cmp r0, #2
bne 4f
/* here, offset is 2 (16-bits aligned, special cased) */
/* make sure we have at least 16 bytes to process */
subs r2, r2, #16
addmi r2, r2, #16
@@ -250,8 +192,8 @@ ENTRY(memcmp)
bic r1, r1, #3
ldr lr, [r1], #4
6: pld [r1, #(CACHE_LINE_SIZE * 2)]
pld [r4, #(CACHE_LINE_SIZE * 2)]
6: PLD (r1, #64)
PLD (r4, #64)
mov ip, lr, lsr #16
ldr lr, [r1], #4
ldr r0, [r4], #4
@@ -338,4 +280,3 @@ ENTRY(memcmp)
mov r2, #4
ldmfd sp!, {r5, r6, r7}
b 8b
END(memcmp)

View File

@@ -37,35 +37,28 @@
#ifdef HAVE_32_BYTE_CACHE_LINE
/* a prefetch distance of 2 cache-lines */
#define CACHE_LINE_SIZE 32
#define PREFETCH_DISTANCE (CACHE_LINE_SIZE*2)
#else
/* a prefetch distance of 4 cache-lines works best experimentally */
#define CACHE_LINE_SIZE 64
#define PREFETCH_DISTANCE (CACHE_LINE_SIZE*4)
#endif
ENTRY(memcpy)
.save {r0, lr}
/* start preloading as early as possible */
pld [r1, #(CACHE_LINE_SIZE * 0)]
pld [r1, #(CACHE_LINE_SIZE*0)]
stmfd sp!, {r0, lr}
pld [r1, #(CACHE_LINE_SIZE * 1)]
pld [r1, #(CACHE_LINE_SIZE*1)]
/* If Neon supports unaligned access then remove the align code,
* unless a size limit has been specified.
*/
#ifndef NEON_UNALIGNED_ACCESS
/* do we have at least 16-bytes to copy (needed for alignment below) */
cmp r2, #16
blo 5f
/* check if buffers are aligned. If so, run arm-only version */
eor r3, r0, r1
ands r3, r3, #0x3
beq 11f
/* align destination to cache-line for the write-buffer */
rsb r3, r0, #0
ands r3, r3, #0xF
beq 2f
beq 0f
/* copy up to 15-bytes (count in r3) */
sub r2, r2, r3
@@ -86,9 +79,10 @@ ENTRY(memcpy)
vld1.8 {d0}, [r1]!
vst1.8 {d0}, [r0, :64]!
2:
/* preload immediately the next cache line, which we may need */
pld [r1, #(CACHE_LINE_SIZE * 0)]
pld [r1, #(CACHE_LINE_SIZE * 1)]
0: /* preload immediately the next cache line, which we may need */
pld [r1, #(CACHE_LINE_SIZE*0)]
pld [r1, #(CACHE_LINE_SIZE*1)]
#ifdef HAVE_32_BYTE_CACHE_LINE
/* make sure we have at least 32 bytes to copy */
@@ -114,22 +108,23 @@ ENTRY(memcpy)
subs r2, r2, #64
blo 2f
/* preload all the cache lines we need. */
pld [r1, #(CACHE_LINE_SIZE * 2)]
pld [r1, #(CACHE_LINE_SIZE * 3)]
/* preload all the cache lines we need.
* NOTE: the number of pld below depends on PREFETCH_DISTANCE,
* ideally would would increase the distance in the main loop to
* avoid the goofy code below. In practice this doesn't seem to make
* a big difference.
*/
pld [r1, #(CACHE_LINE_SIZE*2)]
pld [r1, #(CACHE_LINE_SIZE*3)]
pld [r1, #(PREFETCH_DISTANCE)]
1: /* The main loop copies 64 bytes at a time */
vld1.8 {d0 - d3}, [r1]!
vld1.8 {d4 - d7}, [r1]!
#ifdef HAVE_32_BYTE_CACHE_LINE
pld [r1, #(CACHE_LINE_SIZE * 2)]
pld [r1, #(CACHE_LINE_SIZE * 3)]
#else
pld [r1, #(CACHE_LINE_SIZE * 3)]
#endif
vld1.8 {d0 - d3}, [r1]!
vld1.8 {d4 - d7}, [r1]!
pld [r1, #(PREFETCH_DISTANCE)]
subs r2, r2, #64
vst1.8 {d0 - d3}, [r0, :128]!
vst1.8 {d4 - d7}, [r0, :128]!
vst1.8 {d0 - d3}, [r0, :128]!
vst1.8 {d4 - d7}, [r0, :128]!
bhs 1b
2: /* fix-up the remaining count and make sure we have >= 32 bytes left */
@@ -138,9 +133,9 @@ ENTRY(memcpy)
blo 4f
3: /* 32 bytes at a time. These cache lines were already preloaded */
vld1.8 {d0 - d3}, [r1]!
vld1.8 {d0 - d3}, [r1]!
subs r2, r2, #32
vst1.8 {d0 - d3}, [r0, :128]!
vst1.8 {d0 - d3}, [r0, :128]!
bhs 3b
#endif
4: /* less than 32 left */
@@ -150,6 +145,7 @@ ENTRY(memcpy)
// copies 16 bytes, 128-bits aligned
vld1.8 {d0, d1}, [r1]!
vst1.8 {d0, d1}, [r0, :128]!
5: /* copy up to 15-bytes (count in r2) */
movs ip, r2, lsl #29
bcc 1f
@@ -168,164 +164,6 @@ ENTRY(memcpy)
ldmfd sp!, {r0, lr}
bx lr
#else /* NEON_UNALIGNED_ACCESS */
// Check so divider is at least 16 bytes, needed for alignment code.
cmp r2, #16
blo 5f
#ifdef NEON_MEMCPY_ALIGNMENT_DIVIDER
/* Check the upper size limit for Neon unaligned memory access in memcpy */
#if NEON_MEMCPY_ALIGNMENT_DIVIDER >= 16
cmp r2, #NEON_MEMCPY_ALIGNMENT_DIVIDER
blo 3f
#endif
/* check if buffers are aligned. If so, run arm-only version */
eor r3, r0, r1
ands r3, r3, #0x3
beq 11f
/* align destination to 16 bytes for the write-buffer */
rsb r3, r0, #0
ands r3, r3, #0xF
beq 3f
/* copy up to 15-bytes (count in r3) */
sub r2, r2, r3
movs ip, r3, lsl #31
ldrmib lr, [r1], #1
strmib lr, [r0], #1
ldrcsb ip, [r1], #1
ldrcsb lr, [r1], #1
strcsb ip, [r0], #1
strcsb lr, [r0], #1
movs ip, r3, lsl #29
bge 1f
// copies 4 bytes, destination 32-bits aligned
vld1.32 {d0[0]}, [r1]!
vst1.32 {d0[0]}, [r0, :32]!
1: bcc 2f
// copies 8 bytes, destination 64-bits aligned
vld1.8 {d0}, [r1]!
vst1.8 {d0}, [r0, :64]!
2:
/* preload immediately the next cache line, which we may need */
pld [r1, #(CACHE_LINE_SIZE * 0)]
pld [r1, #(CACHE_LINE_SIZE * 1)]
3:
#endif
/* make sure we have at least 64 bytes to copy */
subs r2, r2, #64
blo 2f
/* preload all the cache lines we need */
pld [r1, #(CACHE_LINE_SIZE * 2)]
pld [r1, #(CACHE_LINE_SIZE * 3)]
1: /* The main loop copies 64 bytes at a time */
vld1.8 {d0 - d3}, [r1]!
vld1.8 {d4 - d7}, [r1]!
#ifdef HAVE_32_BYTE_CACHE_LINE
pld [r1, #(CACHE_LINE_SIZE * 2)]
pld [r1, #(CACHE_LINE_SIZE * 3)]
#else
pld [r1, #(CACHE_LINE_SIZE * 3)]
#endif
subs r2, r2, #64
vst1.8 {d0 - d3}, [r0]!
vst1.8 {d4 - d7}, [r0]!
bhs 1b
2: /* fix-up the remaining count and make sure we have >= 32 bytes left */
add r2, r2, #64
subs r2, r2, #32
blo 4f
3: /* 32 bytes at a time. These cache lines were already preloaded */
vld1.8 {d0 - d3}, [r1]!
subs r2, r2, #32
vst1.8 {d0 - d3}, [r0]!
bhs 3b
4: /* less than 32 left */
add r2, r2, #32
tst r2, #0x10
beq 5f
// copies 16 bytes, 128-bits aligned
vld1.8 {d0, d1}, [r1]!
vst1.8 {d0, d1}, [r0]!
5: /* copy up to 15-bytes (count in r2) */
movs ip, r2, lsl #29
bcc 1f
vld1.8 {d0}, [r1]!
vst1.8 {d0}, [r0]!
1: bge 2f
vld1.32 {d0[0]}, [r1]!
vst1.32 {d0[0]}, [r0]!
2: movs ip, r2, lsl #31
ldrmib r3, [r1], #1
ldrcsb ip, [r1], #1
ldrcsb lr, [r1], #1
strmib r3, [r0], #1
strcsb ip, [r0], #1
strcsb lr, [r0], #1
ldmfd sp!, {r0, lr}
bx lr
#endif /* NEON_UNALIGNED_ACCESS */
11:
/* Simple arm-only copy loop to handle aligned copy operations */
stmfd sp!, {r4, r5, r6, r7, r8}
pld [r1, #(CACHE_LINE_SIZE * 2)]
/* Check alignment */
rsb r3, r1, #0
ands r3, #3
beq 2f
/* align source to 32 bits. We need to insert 2 instructions between
* a ldr[b|h] and str[b|h] because byte and half-word instructions
* stall 2 cycles.
*/
movs r12, r3, lsl #31
sub r2, r2, r3 /* we know that r3 <= r2 because r2 >= 4 */
ldrmib r3, [r1], #1
ldrcsb r4, [r1], #1
ldrcsb r5, [r1], #1
strmib r3, [r0], #1
strcsb r4, [r0], #1
strcsb r5, [r0], #1
2:
subs r2, #32
blt 5f
pld [r1, #(CACHE_LINE_SIZE * 3)]
3: /* Main copy loop, copying 32 bytes at a time */
pld [r1, #(CACHE_LINE_SIZE * 4)]
ldmia r1!, {r3, r4, r5, r6, r7, r8, r12, lr}
subs r2, r2, #32
stmia r0!, {r3, r4, r5, r6, r7, r8, r12, lr}
bge 3b
5: /* Handle any remaining bytes */
adds r2, #32
beq 6f
movs r12, r2, lsl #28
ldmcsia r1!, {r3, r4, r5, r6} /* 16 bytes */
ldmmiia r1!, {r7, r8} /* 8 bytes */
stmcsia r0!, {r3, r4, r5, r6}
stmmiia r0!, {r7, r8}
movs r12, r2, lsl #30
ldrcs r3, [r1], #4 /* 4 bytes */
ldrmih r4, [r1], #2 /* 2 bytes */
strcs r3, [r0], #4
strmih r4, [r0], #2
tst r2, #0x1
ldrneb r3, [r1] /* last byte */
strneb r3, [r0]
6:
ldmfd sp!, {r4, r5, r6, r7, r8}
ldmfd sp!, {r0, pc}
END(memcpy)

View File

@@ -1,406 +0,0 @@
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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.
*/
/* Prototype: void *memcpy (void *dst, const void *src, size_t count). */
/* Use the version of memcpy implemented using LDRD and STRD.
This version is tuned for Cortex-A15.
This might not be the best for other ARMv7-A CPUs,
but there is no predefine to distinguish between
different CPUs in the same architecture,
and this version is better than the plain memcpy provided in newlib.
Therefore, we use this version for all ARMv7-A CPUS. */
/* To make the same code compile for both ARM and Thumb instruction
sets, switch to unified syntax at the beginning of this function.
However, by using the same code, we may be missing optimization
opportunities. For instance, in LDRD/STRD instructions, the first
destination register must be even and the second consecutive in
ARM state, but not in Thumb state. */
#include <machine/cpu-features.h>
#include <machine/asm.h>
.syntax unified
ENTRY(memcpy)
/* Assumes that n >= 0, and dst, src are valid pointers.
If there is at least 8 bytes to copy, use LDRD/STRD.
If src and dst are misaligned with different offsets,
first copy byte by byte until dst is aligned,
and then copy using LDRD/STRD and shift if needed.
When less than 8 left, copy a word and then byte by byte. */
/* Save registers (r0 holds the return value):
optimized push {r0, r4, r5, r6, r7, lr}.
To try and improve performance, stack layout changed,
i.e., not keeping the stack looking like users expect
(highest numbered register at highest address). */
.save {r0, lr}
push {r0, lr}
.save {r4, r5}
strd r4, r5, [sp, #-8]!
.save {r6, r7}
strd r6, r7, [sp, #-8]!
/* TODO: Add debug frame directives.
We don't need exception unwind directives, because the code below
does not throw any exceptions and does not call any other functions.
Generally, newlib functions like this lack debug information for
assembler source. */
/* Get copying of tiny blocks out of the way first. */
/* Is there at least 4 bytes to copy? */
subs r2, r2, #4
blt copy_less_than_4 /* If n < 4. */
/* Check word alignment. */
ands ip, r0, #3 /* ip = last 2 bits of dst. */
bne dst_not_word_aligned /* If dst is not word-aligned. */
/* Get here if dst is word-aligned. */
ands ip, r1, #3 /* ip = last 2 bits of src. */
bne src_not_word_aligned /* If src is not word-aligned. */
word_aligned:
/* Get here if source and dst both are word-aligned.
The number of bytes remaining to copy is r2+4. */
/* Is there is at least 64 bytes to copy? */
subs r2, r2, #60
blt copy_less_than_64 /* If r2 + 4 < 64. */
/* First, align the destination buffer to 8-bytes,
to make sure double loads and stores don't cross cache line boundary,
as they are then more expensive even if the data is in the cache
(require two load/store issue cycles instead of one).
If only one of the buffers is not 8-bytes aligned,
then it's more important to align dst than src,
because there is more penalty for stores
than loads that cross cacheline boundary.
This check and realignment are only worth doing
if there is a lot to copy. */
/* Get here if dst is word aligned,
i.e., the 2 least significant bits are 0.
If dst is not 2w aligned (i.e., the 3rd bit is not set in dst),
then copy 1 word (4 bytes). */
ands r3, r0, #4
beq 11f /* If dst already two-word aligned. */
ldr r3, [r1], #4
str r3, [r0], #4
subs r2, r2, #4
blt copy_less_than_64
11:
/* TODO: Align to cacheline (useful for PLD optimization). */
/* Every loop iteration copies 64 bytes. */
1:
.irp offset, #0, #8, #16, #24, #32, #40, #48, #56
ldrd r4, r5, [r1, \offset]
strd r4, r5, [r0, \offset]
.endr
add r0, r0, #64
add r1, r1, #64
subs r2, r2, #64
bge 1b /* If there is more to copy. */
copy_less_than_64:
/* Get here if less than 64 bytes to copy, -64 <= r2 < 0.
Restore the count if there is more than 7 bytes to copy. */
adds r2, r2, #56
blt copy_less_than_8
/* Copy 8 bytes at a time. */
2:
ldrd r4, r5, [r1], #8
strd r4, r5, [r0], #8
subs r2, r2, #8
bge 2b /* If there is more to copy. */
copy_less_than_8:
/* Get here if less than 8 bytes to copy, -8 <= r2 < 0.
Check if there is more to copy. */
cmn r2, #8
beq return /* If r2 + 8 == 0. */
/* Restore the count if there is more than 3 bytes to copy. */
adds r2, r2, #4
blt copy_less_than_4
/* Copy 4 bytes. */
ldr r3, [r1], #4
str r3, [r0], #4
copy_less_than_4:
/* Get here if less than 4 bytes to copy, -4 <= r2 < 0. */
/* Restore the count, check if there is more to copy. */
adds r2, r2, #4
beq return /* If r2 == 0. */
/* Get here with r2 is in {1,2,3}={01,10,11}. */
/* Logical shift left r2, insert 0s, update flags. */
lsls r2, r2, #31
/* Copy byte by byte.
Condition ne means the last bit of r2 is 0.
Condition cs means the second to last bit of r2 is set,
i.e., r2 is 1 or 3. */
itt ne
ldrbne r3, [r1], #1
strbne r3, [r0], #1
itttt cs
ldrbcs r4, [r1], #1
ldrbcs r5, [r1]
strbcs r4, [r0], #1
strbcs r5, [r0]
return:
/* Restore registers: optimized pop {r0, r4, r5, r6, r7, pc} */
/* This is the only return point of memcpy. */
ldrd r6, r7, [sp], #8
ldrd r4, r5, [sp], #8
pop {r0, pc}
#ifndef __ARM_FEATURE_UNALIGNED
/* The following assembly macro implements misaligned copy in software.
Assumes that dst is word aligned, src is at offset "pull" bits from
word, push = 32 - pull, and the number of bytes that remain to copy
is r2 + 4, r2 >= 0. */
/* In the code below, r2 is the number of bytes that remain to be
written. The number of bytes read is always larger, because we have
partial words in the shift queue. */
.macro miscopy pull push shiftleft shiftright
/* Align src to the previous word boundary. */
bic r1, r1, #3
/* Initialize the shift queue. */
ldr r5, [r1], #4 /* Load a word from source. */
subs r2, r2, #4
blt 6f /* Go to misaligned copy of less than 8 bytes. */
/* Get here if there is more than 8 bytes to copy.
The number of bytes to copy is r2+8, r2 >= 0. */
subs r2, r2, #56
blt 4f /* Go to misaligned copy of less than 64 bytes. */
3:
/* Get here if there is more than 64 bytes to copy.
The number of bytes to copy is r2+64, r2 >= 0. */
/* Copy 64 bytes in every iteration.
Use a partial word from the shift queue. */
.irp offset, #0, #8, #16, #24, #32, #40, #48, #56
mov r6, r5, \shiftleft #\pull
ldrd r4, r5, [r1, \offset]
orr r6, r6, r4, \shiftright #\push
mov r7, r4, \shiftleft #\pull
orr r7, r7, r5, \shiftright #\push
strd r6, r7, [r0, \offset]
.endr
add r1, r1, #64
add r0, r0, #64
subs r2, r2, #64
bge 3b
4:
/* Get here if there is less than 64 bytes to copy (-64 <= r2 < 0)
and they are misaligned. */
/* Restore the count if there is more than 7 bytes to copy. */
adds r2, r2, #56
blt 6f /* Go to misaligned copy of less than 8 bytes. */
5:
/* Copy 8 bytes at a time.
Use a partial word from the shift queue. */
mov r6, r5, \shiftleft #\pull
ldrd r4, r5, [r1], #8
orr r6, r6, r4, \shiftright #\push
mov r7, r4, \shiftleft #\pull
orr r7, r7, r5, \shiftright #\push
strd r6, r7, [r0], #8
subs r2, r2, #8
bge 5b /* If there is more to copy. */
6:
/* Get here if there less than 8 bytes to copy (-8 <= r2 < 0)
and they are misaligned. */
/* Check if there is more to copy. */
cmn r2, #8
beq return
/* Check if there is less than 4 bytes to copy. */
cmn r2, #4
itt lt
/* Restore src offset from word-align. */
sublt r1, r1, #(\push / 8)
blt copy_less_than_4
/* Use a partial word from the shift queue. */
mov r3, r5, \shiftleft #\pull
/* Load a word from src, but without writeback
(this word is not fully written to dst). */
ldr r5, [r1]
/* Restore src offset from word-align. */
add r1, r1, #(\pull / 8)
/* Shift bytes to create one dst word and store it. */
orr r3, r3, r5, \shiftright #\push
str r3, [r0], #4
/* Use single byte copying of the remaining bytes. */
b copy_less_than_4
.endm
#endif /* not __ARM_FEATURE_UNALIGNED */
dst_not_word_aligned:
/* Get here when dst is not aligned and ip has the last 2 bits of dst,
i.e., ip is the offset of dst from word.
The number of bytes that remains to copy is r2 + 4,
i.e., there are at least 4 bytes to copy.
Write a partial word (0 to 3 bytes), such that dst becomes
word-aligned. */
/* If dst is at ip bytes offset from a word (with 0 < ip < 4),
then there are (4 - ip) bytes to fill up to align dst to the next
word. */
rsb ip, ip, #4 /* ip = #4 - ip. */
cmp ip, #2
/* Copy byte by byte with conditionals. */
itt gt
ldrbgt r3, [r1], #1
strbgt r3, [r0], #1
itt ge
ldrbge r4, [r1], #1
strbge r4, [r0], #1
ldrb lr, [r1], #1
strb lr, [r0], #1
/* Update the count.
ip holds the number of bytes we have just copied. */
subs r2, r2, ip /* r2 = r2 - ip. */
blt copy_less_than_4 /* If r2 < ip. */
/* Get here if there are more than 4 bytes to copy.
Check if src is aligned. If beforehand src and dst were not word
aligned but congruent (same offset), then now they are both
word-aligned, and we can copy the rest efficiently (without
shifting). */
ands ip, r1, #3 /* ip = last 2 bits of src. */
beq word_aligned /* If r1 is word-aligned. */
src_not_word_aligned:
/* Get here when src is not word-aligned, but dst is word-aligned.
The number of bytes that remains to copy is r2+4. */
#ifdef __ARM_FEATURE_UNALIGNED
/* Copy word by word using LDR when alignment can be done in hardware,
i.e., SCTLR.A is set, supporting unaligned access in LDR and STR. */
subs r2, r2, #60
blt 8f
7:
/* Copy 64 bytes in every loop iteration. */
.irp offset, #0, #4, #8, #12, #16, #20, #24, #28, #32, #36, #40, #44, #48, #52, #56, #60
ldr r3, [r1, \offset]
str r3, [r0, \offset]
.endr
add r0, r0, #64
add r1, r1, #64
subs r2, r2, #64
bge 7b
8:
/* Get here if less than 64 bytes to copy, -64 <= r2 < 0.
Check if there is more than 3 bytes to copy. */
adds r2, r2, #60
blt copy_less_than_4
9:
/* Get here if there is less than 64 but at least 4 bytes to copy,
where the number of bytes to copy is r2+4. */
ldr r3, [r1], #4
str r3, [r0], #4
subs r2, r2, #4
bge 9b
b copy_less_than_4
#else /* not __ARM_FEATURE_UNALIGNED */
/* ip has last 2 bits of src,
i.e., ip is the offset of src from word, and ip > 0.
Compute shifts needed to copy from src to dst. */
cmp ip, #2
beq miscopy_16_16 /* If ip == 2. */
bge miscopy_24_8 /* If ip == 3. */
/* Get here if ip == 1. */
/* Endian independent macros for shifting bytes within registers. */
#ifndef __ARMEB__
miscopy_8_24: miscopy pull=8 push=24 shiftleft=lsr shiftright=lsl
miscopy_16_16: miscopy pull=16 push=16 shiftleft=lsr shiftright=lsl
miscopy_24_8: miscopy pull=24 push=8 shiftleft=lsr shiftright=lsl
#else /* not __ARMEB__ */
miscopy_8_24: miscopy pull=8 push=24 shiftleft=lsl shiftright=lsr
miscopy_16_16: miscopy pull=16 push=16 shiftleft=lsl shiftright=lsr
miscopy_24_8: miscopy pull=24 push=8 shiftleft=lsl shiftright=lsr
#endif /* not __ARMEB__ */
#endif /* not __ARM_FEATURE_UNALIGNED */
END(memcpy)

View File

@@ -1,614 +0,0 @@
/* Copyright (c) 2013, 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 Linaro Limited 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.
*/
/*
This memcpy routine is optimised for Cortex-A15 cores and takes advantage
of VFP or NEON when built with the appropriate flags.
Assumptions:
ARMv6 (ARMv7-a if using Neon)
ARM state
Unaligned accesses
LDRD/STRD support unaligned word accesses
*/
#include <machine/cpu-features.h>
#include <machine/asm.h>
.syntax unified
/* This implementation requires ARM state. */
.arm
#ifdef __ARM_NEON__
.fpu neon
.arch armv7-a
# define FRAME_SIZE 4
# define USE_VFP
# define USE_NEON
#elif !defined (__SOFTFP__)
.arch armv6
.fpu vfpv2
# define FRAME_SIZE 32
# define USE_VFP
#else
.arch armv6
# define FRAME_SIZE 32
#endif
/* Old versions of GAS incorrectly implement the NEON align semantics. */
#ifdef BROKEN_ASM_NEON_ALIGN
#define ALIGN(addr, align) addr,:align
#else
#define ALIGN(addr, align) addr:align
#endif
#define PC_OFFSET 8 /* PC pipeline compensation. */
#define INSN_SIZE 4
/* Call parameters. */
#define dstin r0
#define src r1
#define count r2
/* Locals. */
#define tmp1 r3
#define dst ip
#define tmp2 r10
#ifndef USE_NEON
/* For bulk copies using GP registers. */
#define A_l r2 /* Call-clobbered. */
#define A_h r3 /* Call-clobbered. */
#define B_l r4
#define B_h r5
#define C_l r6
#define C_h r7
#define D_l r8
#define D_h r9
#endif
/* Number of lines ahead to pre-fetch data. If you change this the code
below will need adjustment to compensate. */
#define prefetch_lines 5
#ifdef USE_VFP
.macro cpy_line_vfp vreg, base
vstr \vreg, [dst, #\base]
vldr \vreg, [src, #\base]
vstr d0, [dst, #\base + 8]
vldr d0, [src, #\base + 8]
vstr d1, [dst, #\base + 16]
vldr d1, [src, #\base + 16]
vstr d2, [dst, #\base + 24]
vldr d2, [src, #\base + 24]
vstr \vreg, [dst, #\base + 32]
vldr \vreg, [src, #\base + prefetch_lines * 64 - 32]
vstr d0, [dst, #\base + 40]
vldr d0, [src, #\base + 40]
vstr d1, [dst, #\base + 48]
vldr d1, [src, #\base + 48]
vstr d2, [dst, #\base + 56]
vldr d2, [src, #\base + 56]
.endm
.macro cpy_tail_vfp vreg, base
vstr \vreg, [dst, #\base]
vldr \vreg, [src, #\base]
vstr d0, [dst, #\base + 8]
vldr d0, [src, #\base + 8]
vstr d1, [dst, #\base + 16]
vldr d1, [src, #\base + 16]
vstr d2, [dst, #\base + 24]
vldr d2, [src, #\base + 24]
vstr \vreg, [dst, #\base + 32]
vstr d0, [dst, #\base + 40]
vldr d0, [src, #\base + 40]
vstr d1, [dst, #\base + 48]
vldr d1, [src, #\base + 48]
vstr d2, [dst, #\base + 56]
vldr d2, [src, #\base + 56]
.endm
#endif
.p2align 6
ENTRY(memcpy)
mov dst, dstin /* Preserve dstin, we need to return it. */
cmp count, #64
bge .Lcpy_not_short
/* Deal with small copies quickly by dropping straight into the
exit block. */
.Ltail63unaligned:
#ifdef USE_NEON
and tmp1, count, #0x38
rsb tmp1, tmp1, #(56 - PC_OFFSET + INSN_SIZE)
add pc, pc, tmp1
vld1.8 {d0}, [src]! /* 14 words to go. */
vst1.8 {d0}, [dst]!
vld1.8 {d0}, [src]! /* 12 words to go. */
vst1.8 {d0}, [dst]!
vld1.8 {d0}, [src]! /* 10 words to go. */
vst1.8 {d0}, [dst]!
vld1.8 {d0}, [src]! /* 8 words to go. */
vst1.8 {d0}, [dst]!
vld1.8 {d0}, [src]! /* 6 words to go. */
vst1.8 {d0}, [dst]!
vld1.8 {d0}, [src]! /* 4 words to go. */
vst1.8 {d0}, [dst]!
vld1.8 {d0}, [src]! /* 2 words to go. */
vst1.8 {d0}, [dst]!
tst count, #4
ldrne tmp1, [src], #4
strne tmp1, [dst], #4
#else
/* Copy up to 15 full words of data. May not be aligned. */
/* Cannot use VFP for unaligned data. */
and tmp1, count, #0x3c
add dst, dst, tmp1
add src, src, tmp1
rsb tmp1, tmp1, #(60 - PC_OFFSET/2 + INSN_SIZE/2)
/* Jump directly into the sequence below at the correct offset. */
add pc, pc, tmp1, lsl #1
ldr tmp1, [src, #-60] /* 15 words to go. */
str tmp1, [dst, #-60]
ldr tmp1, [src, #-56] /* 14 words to go. */
str tmp1, [dst, #-56]
ldr tmp1, [src, #-52]
str tmp1, [dst, #-52]
ldr tmp1, [src, #-48] /* 12 words to go. */
str tmp1, [dst, #-48]
ldr tmp1, [src, #-44]
str tmp1, [dst, #-44]
ldr tmp1, [src, #-40] /* 10 words to go. */
str tmp1, [dst, #-40]
ldr tmp1, [src, #-36]
str tmp1, [dst, #-36]
ldr tmp1, [src, #-32] /* 8 words to go. */
str tmp1, [dst, #-32]
ldr tmp1, [src, #-28]
str tmp1, [dst, #-28]
ldr tmp1, [src, #-24] /* 6 words to go. */
str tmp1, [dst, #-24]
ldr tmp1, [src, #-20]
str tmp1, [dst, #-20]
ldr tmp1, [src, #-16] /* 4 words to go. */
str tmp1, [dst, #-16]
ldr tmp1, [src, #-12]
str tmp1, [dst, #-12]
ldr tmp1, [src, #-8] /* 2 words to go. */
str tmp1, [dst, #-8]
ldr tmp1, [src, #-4]
str tmp1, [dst, #-4]
#endif
lsls count, count, #31
ldrhcs tmp1, [src], #2
ldrbne src, [src] /* Src is dead, use as a scratch. */
strhcs tmp1, [dst], #2
strbne src, [dst]
bx lr
.Lcpy_not_short:
/* At least 64 bytes to copy, but don't know the alignment yet. */
str tmp2, [sp, #-FRAME_SIZE]!
and tmp2, src, #7
and tmp1, dst, #7
cmp tmp1, tmp2
bne .Lcpy_notaligned
#ifdef USE_VFP
/* Magic dust alert! Force VFP on Cortex-A9. Experiments show
that the FP pipeline is much better at streaming loads and
stores. This is outside the critical loop. */
vmov.f32 s0, s0
#endif
/* SRC and DST have the same mutual 32-bit alignment, but we may
still need to pre-copy some bytes to get to natural alignment.
We bring DST into full 64-bit alignment. */
lsls tmp2, dst, #29
beq 1f
rsbs tmp2, tmp2, #0
sub count, count, tmp2, lsr #29
ldrmi tmp1, [src], #4
strmi tmp1, [dst], #4
lsls tmp2, tmp2, #2
ldrhcs tmp1, [src], #2
ldrbne tmp2, [src], #1
strhcs tmp1, [dst], #2
strbne tmp2, [dst], #1
1:
subs tmp2, count, #64 /* Use tmp2 for count. */
blt .Ltail63aligned
cmp tmp2, #512
bge .Lcpy_body_long
.Lcpy_body_medium: /* Count in tmp2. */
#ifdef USE_VFP
1:
vldr d0, [src, #0]
subs tmp2, tmp2, #64
vldr d1, [src, #8]
vstr d0, [dst, #0]
vldr d0, [src, #16]
vstr d1, [dst, #8]
vldr d1, [src, #24]
vstr d0, [dst, #16]
vldr d0, [src, #32]
vstr d1, [dst, #24]
vldr d1, [src, #40]
vstr d0, [dst, #32]
vldr d0, [src, #48]
vstr d1, [dst, #40]
vldr d1, [src, #56]
vstr d0, [dst, #48]
add src, src, #64
vstr d1, [dst, #56]
add dst, dst, #64
bge 1b
tst tmp2, #0x3f
beq .Ldone
.Ltail63aligned: /* Count in tmp2. */
and tmp1, tmp2, #0x38
add dst, dst, tmp1
add src, src, tmp1
rsb tmp1, tmp1, #(56 - PC_OFFSET + INSN_SIZE)
add pc, pc, tmp1
vldr d0, [src, #-56] /* 14 words to go. */
vstr d0, [dst, #-56]
vldr d0, [src, #-48] /* 12 words to go. */
vstr d0, [dst, #-48]
vldr d0, [src, #-40] /* 10 words to go. */
vstr d0, [dst, #-40]
vldr d0, [src, #-32] /* 8 words to go. */
vstr d0, [dst, #-32]
vldr d0, [src, #-24] /* 6 words to go. */
vstr d0, [dst, #-24]
vldr d0, [src, #-16] /* 4 words to go. */
vstr d0, [dst, #-16]
vldr d0, [src, #-8] /* 2 words to go. */
vstr d0, [dst, #-8]
#else
sub src, src, #8
sub dst, dst, #8
1:
ldrd A_l, A_h, [src, #8]
strd A_l, A_h, [dst, #8]
ldrd A_l, A_h, [src, #16]
strd A_l, A_h, [dst, #16]
ldrd A_l, A_h, [src, #24]
strd A_l, A_h, [dst, #24]
ldrd A_l, A_h, [src, #32]
strd A_l, A_h, [dst, #32]
ldrd A_l, A_h, [src, #40]
strd A_l, A_h, [dst, #40]
ldrd A_l, A_h, [src, #48]
strd A_l, A_h, [dst, #48]
ldrd A_l, A_h, [src, #56]
strd A_l, A_h, [dst, #56]
ldrd A_l, A_h, [src, #64]!
strd A_l, A_h, [dst, #64]!
subs tmp2, tmp2, #64
bge 1b
tst tmp2, #0x3f
bne 1f
ldr tmp2,[sp], #FRAME_SIZE
bx lr
1:
add src, src, #8
add dst, dst, #8
.Ltail63aligned: /* Count in tmp2. */
/* Copy up to 7 d-words of data. Similar to Ltail63unaligned, but
we know that the src and dest are 32-bit aligned so we can use
LDRD/STRD to improve efficiency. */
/* TMP2 is now negative, but we don't care about that. The bottom
six bits still tell us how many bytes are left to copy. */
and tmp1, tmp2, #0x38
add dst, dst, tmp1
add src, src, tmp1
rsb tmp1, tmp1, #(56 - PC_OFFSET + INSN_SIZE)
add pc, pc, tmp1
ldrd A_l, A_h, [src, #-56] /* 14 words to go. */
strd A_l, A_h, [dst, #-56]
ldrd A_l, A_h, [src, #-48] /* 12 words to go. */
strd A_l, A_h, [dst, #-48]
ldrd A_l, A_h, [src, #-40] /* 10 words to go. */
strd A_l, A_h, [dst, #-40]
ldrd A_l, A_h, [src, #-32] /* 8 words to go. */
strd A_l, A_h, [dst, #-32]
ldrd A_l, A_h, [src, #-24] /* 6 words to go. */
strd A_l, A_h, [dst, #-24]
ldrd A_l, A_h, [src, #-16] /* 4 words to go. */
strd A_l, A_h, [dst, #-16]
ldrd A_l, A_h, [src, #-8] /* 2 words to go. */
strd A_l, A_h, [dst, #-8]
#endif
tst tmp2, #4
ldrne tmp1, [src], #4
strne tmp1, [dst], #4
lsls tmp2, tmp2, #31 /* Count (tmp2) now dead. */
ldrhcs tmp1, [src], #2
ldrbne tmp2, [src]
strhcs tmp1, [dst], #2
strbne tmp2, [dst]
.Ldone:
ldr tmp2, [sp], #FRAME_SIZE
bx lr
.Lcpy_body_long: /* Count in tmp2. */
/* Long copy. We know that there's at least (prefetch_lines * 64)
bytes to go. */
#ifdef USE_VFP
/* Don't use PLD. Instead, read some data in advance of the current
copy position into a register. This should act like a PLD
operation but we won't have to repeat the transfer. */
vldr d3, [src, #0]
vldr d4, [src, #64]
vldr d5, [src, #128]
vldr d6, [src, #192]
vldr d7, [src, #256]
vldr d0, [src, #8]
vldr d1, [src, #16]
vldr d2, [src, #24]
add src, src, #32
subs tmp2, tmp2, #prefetch_lines * 64 * 2
blt 2f
1:
cpy_line_vfp d3, 0
cpy_line_vfp d4, 64
cpy_line_vfp d5, 128
add dst, dst, #3 * 64
add src, src, #3 * 64
cpy_line_vfp d6, 0
cpy_line_vfp d7, 64
add dst, dst, #2 * 64
add src, src, #2 * 64
subs tmp2, tmp2, #prefetch_lines * 64
bge 1b
2:
cpy_tail_vfp d3, 0
cpy_tail_vfp d4, 64
cpy_tail_vfp d5, 128
add src, src, #3 * 64
add dst, dst, #3 * 64
cpy_tail_vfp d6, 0
vstr d7, [dst, #64]
vldr d7, [src, #64]
vstr d0, [dst, #64 + 8]
vldr d0, [src, #64 + 8]
vstr d1, [dst, #64 + 16]
vldr d1, [src, #64 + 16]
vstr d2, [dst, #64 + 24]
vldr d2, [src, #64 + 24]
vstr d7, [dst, #64 + 32]
add src, src, #96
vstr d0, [dst, #64 + 40]
vstr d1, [dst, #64 + 48]
vstr d2, [dst, #64 + 56]
add dst, dst, #128
add tmp2, tmp2, #prefetch_lines * 64
b .Lcpy_body_medium
#else
/* Long copy. Use an SMS style loop to maximize the I/O
bandwidth of the core. We don't have enough spare registers
to synthesise prefetching, so use PLD operations. */
/* Pre-bias src and dst. */
sub src, src, #8
sub dst, dst, #8
pld [src, #8]
pld [src, #72]
subs tmp2, tmp2, #64
pld [src, #136]
ldrd A_l, A_h, [src, #8]
strd B_l, B_h, [sp, #8]
ldrd B_l, B_h, [src, #16]
strd C_l, C_h, [sp, #16]
ldrd C_l, C_h, [src, #24]
strd D_l, D_h, [sp, #24]
pld [src, #200]
ldrd D_l, D_h, [src, #32]!
b 1f
.p2align 6
2:
pld [src, #232]
strd A_l, A_h, [dst, #40]
ldrd A_l, A_h, [src, #40]
strd B_l, B_h, [dst, #48]
ldrd B_l, B_h, [src, #48]
strd C_l, C_h, [dst, #56]
ldrd C_l, C_h, [src, #56]
strd D_l, D_h, [dst, #64]!
ldrd D_l, D_h, [src, #64]!
subs tmp2, tmp2, #64
1:
strd A_l, A_h, [dst, #8]
ldrd A_l, A_h, [src, #8]
strd B_l, B_h, [dst, #16]
ldrd B_l, B_h, [src, #16]
strd C_l, C_h, [dst, #24]
ldrd C_l, C_h, [src, #24]
strd D_l, D_h, [dst, #32]
ldrd D_l, D_h, [src, #32]
bcs 2b
/* Save the remaining bytes and restore the callee-saved regs. */
strd A_l, A_h, [dst, #40]
add src, src, #40
strd B_l, B_h, [dst, #48]
ldrd B_l, B_h, [sp, #8]
strd C_l, C_h, [dst, #56]
ldrd C_l, C_h, [sp, #16]
strd D_l, D_h, [dst, #64]
ldrd D_l, D_h, [sp, #24]
add dst, dst, #72
tst tmp2, #0x3f
bne .Ltail63aligned
ldr tmp2, [sp], #FRAME_SIZE
bx lr
#endif
.Lcpy_notaligned:
pld [src]
pld [src, #64]
/* There's at least 64 bytes to copy, but there is no mutual
alignment. */
/* Bring DST to 64-bit alignment. */
lsls tmp2, dst, #29
pld [src, #(2 * 64)]
beq 1f
rsbs tmp2, tmp2, #0
sub count, count, tmp2, lsr #29
ldrmi tmp1, [src], #4
strmi tmp1, [dst], #4
lsls tmp2, tmp2, #2
ldrbne tmp1, [src], #1
ldrhcs tmp2, [src], #2
strbne tmp1, [dst], #1
strhcs tmp2, [dst], #2
1:
pld [src, #(3 * 64)]
subs count, count, #64
ldrmi tmp2, [sp], #FRAME_SIZE
bmi .Ltail63unaligned
pld [src, #(4 * 64)]
#ifdef USE_NEON
vld1.8 {d0-d3}, [src]!
vld1.8 {d4-d7}, [src]!
subs count, count, #64
bmi 2f
1:
pld [src, #(4 * 64)]
vst1.8 {d0-d3}, [ALIGN (dst, 64)]!
vld1.8 {d0-d3}, [src]!
vst1.8 {d4-d7}, [ALIGN (dst, 64)]!
vld1.8 {d4-d7}, [src]!
subs count, count, #64
bpl 1b
2:
vst1.8 {d0-d3}, [ALIGN (dst, 64)]!
vst1.8 {d4-d7}, [ALIGN (dst, 64)]!
ands count, count, #0x3f
#else
/* Use an SMS style loop to maximize the I/O bandwidth. */
sub src, src, #4
sub dst, dst, #8
subs tmp2, count, #64 /* Use tmp2 for count. */
ldr A_l, [src, #4]
ldr A_h, [src, #8]
strd B_l, B_h, [sp, #8]
ldr B_l, [src, #12]
ldr B_h, [src, #16]
strd C_l, C_h, [sp, #16]
ldr C_l, [src, #20]
ldr C_h, [src, #24]
strd D_l, D_h, [sp, #24]
ldr D_l, [src, #28]
ldr D_h, [src, #32]!
b 1f
.p2align 6
2:
pld [src, #(5 * 64) - (32 - 4)]
strd A_l, A_h, [dst, #40]
ldr A_l, [src, #36]
ldr A_h, [src, #40]
strd B_l, B_h, [dst, #48]
ldr B_l, [src, #44]
ldr B_h, [src, #48]
strd C_l, C_h, [dst, #56]
ldr C_l, [src, #52]
ldr C_h, [src, #56]
strd D_l, D_h, [dst, #64]!
ldr D_l, [src, #60]
ldr D_h, [src, #64]!
subs tmp2, tmp2, #64
1:
strd A_l, A_h, [dst, #8]
ldr A_l, [src, #4]
ldr A_h, [src, #8]
strd B_l, B_h, [dst, #16]
ldr B_l, [src, #12]
ldr B_h, [src, #16]
strd C_l, C_h, [dst, #24]
ldr C_l, [src, #20]
ldr C_h, [src, #24]
strd D_l, D_h, [dst, #32]
ldr D_l, [src, #28]
ldr D_h, [src, #32]
bcs 2b
/* Save the remaining bytes and restore the callee-saved regs. */
strd A_l, A_h, [dst, #40]
add src, src, #36
strd B_l, B_h, [dst, #48]
ldrd B_l, B_h, [sp, #8]
strd C_l, C_h, [dst, #56]
ldrd C_l, C_h, [sp, #16]
strd D_l, D_h, [dst, #64]
ldrd D_l, D_h, [sp, #24]
add dst, dst, #72
ands count, tmp2, #0x3f
#endif
ldr tmp2, [sp], #FRAME_SIZE
bne .Ltail63unaligned
bx lr
END(memcpy)

View File

@@ -26,114 +26,23 @@
* SUCH DAMAGE.
*/
#include <machine/cpu-features.h>
#include <machine/asm.h>
/*
* Optimized memset() for ARM.
*
* memset() returns its first argument.
*/
#if defined(__ARM_NEON__)
.fpu neon
#endif
ENTRY(bzero)
mov r2, r1
mov r1, #0
// Fall through to memset...
END(bzero)
ENTRY(memset)
#if defined(__ARM_NEON__)
#ifdef NEON_MEMSET_DIVIDER
cmp r2, #NEON_MEMSET_DIVIDER
bhi 11f
#endif
.save {r0}
stmfd sp!, {r0}
vdup.8 q0, r1
#ifndef NEON_UNALIGNED_ACCESS
/* do we have at least 16-bytes to write (needed for alignment below) */
cmp r2, #16
blo 3f
/* align destination to 16 bytes for the write-buffer */
rsb r3, r0, #0
ands r3, r3, #0xF
beq 2f
/* write up to 15-bytes (count in r3) */
sub r2, r2, r3
movs ip, r3, lsl #31
strmib r1, [r0], #1
strcsb r1, [r0], #1
strcsb r1, [r0], #1
movs ip, r3, lsl #29
bge 1f
// writes 4 bytes, 32-bits aligned
vst1.32 {d0[0]}, [r0, :32]!
1: bcc 2f
// writes 8 bytes, 64-bits aligned
vst1.8 {d0}, [r0, :64]!
2:
#endif
/* make sure we have at least 32 bytes to write */
subs r2, r2, #32
blo 2f
vmov q1, q0
1: /* The main loop writes 32 bytes at a time */
subs r2, r2, #32
#ifndef NEON_UNALIGNED_ACCESS
vst1.8 {d0 - d3}, [r0, :128]!
#else
vst1.8 {d0 - d3}, [r0]!
#endif
bhs 1b
2: /* less than 32 left */
add r2, r2, #32
tst r2, #0x10
beq 3f
// writes 16 bytes, 128-bits aligned
#ifndef NEON_UNALIGNED_ACCESS
vst1.8 {d0, d1}, [r0, :128]!
#else
vst1.8 {d0, d1}, [r0]!
#endif
3: /* write up to 15-bytes (count in r2) */
movs ip, r2, lsl #29
bcc 1f
vst1.8 {d0}, [r0]!
1: bge 2f
vst1.32 {d0[0]}, [r0]!
2: movs ip, r2, lsl #31
strmib r1, [r0], #1
strcsb r1, [r0], #1
strcsb r1, [r0], #1
ldmfd sp!, {r0}
bx lr
11:
#endif
/*
* Optimized memset() for ARM.
*
* memset() returns its first argument.
*/
/* compute the offset to align the destination
* offset = (4-(src&3))&3 = -src & 3
*/
.save {r0, r4-r7, lr}
stmfd sp!, {r0, r4-r7, lr}
rsb r3, r0, #0
@@ -161,7 +70,7 @@ ENTRY(memset)
mov r5, r1
mov r6, r1
mov r7, r1
rsb r3, r0, #0
ands r3, r3, #0x1C
beq 3f
@@ -169,7 +78,7 @@ ENTRY(memset)
andhi r3, r2, #0x1C
sub r2, r2, r3
/* conditionally writes 0 to 7 words (length in r3) */
/* conditionnaly writes 0 to 7 words (length in r3) */
movs r3, r3, lsl #28
stmcsia r0!, {r1, lr}
stmcsia r0!, {r1, lr}
@@ -186,7 +95,7 @@ ENTRY(memset)
bhs 1b
2: add r2, r2, #32
/* conditionally stores 0 to 31 bytes */
/* conditionnaly stores 0 to 31 bytes */
movs r2, r2, lsl #28
stmcsia r0!, {r1,r3,r12,lr}
stmmiia r0!, {r1, lr}

View File

@@ -1,787 +0,0 @@
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 "arm_asm.h"
#ifdef __ARMEB__
#define S2LOMEM lsl
#define S2LOMEMEQ lsleq
#define S2HIMEM lsr
#define MSB 0x000000ff
#define LSB 0xff000000
#define BYTE0_OFFSET 24
#define BYTE1_OFFSET 16
#define BYTE2_OFFSET 8
#define BYTE3_OFFSET 0
#else /* not __ARMEB__ */
#define S2LOMEM lsr
#define S2LOMEMEQ lsreq
#define S2HIMEM lsl
#define BYTE0_OFFSET 0
#define BYTE1_OFFSET 8
#define BYTE2_OFFSET 16
#define BYTE3_OFFSET 24
#define MSB 0xff000000
#define LSB 0x000000ff
#endif /* not __ARMEB__ */
.syntax unified
#if defined (__thumb__)
.thumb
.thumb_func
#endif
.global strcmp
.type strcmp, %function
strcmp:
#if (defined (__thumb__) && !defined (__thumb2__))
1:
ldrb r2, [r0]
ldrb r3, [r1]
adds r0, r0, #1
adds r1, r1, #1
cmp r2, #0
beq 2f
cmp r2, r3
beq 1b
2:
subs r0, r2, r3
bx lr
#elif (defined (__OPTIMIZE_SIZE__) || defined (PREFER_SIZE_OVER_SPEED))
1:
ldrb r2, [r0], #1
ldrb r3, [r1], #1
cmp r2, #1
it cs
cmpcs r2, r3
beq 1b
subs r0, r2, r3
RETURN
#elif (defined (_ISA_THUMB_2) || defined (_ISA_ARM_6))
/* Use LDRD whenever possible. */
/* The main thing to look out for when comparing large blocks is that
the loads do not cross a page boundary when loading past the index
of the byte with the first difference or the first string-terminator.
For example, if the strings are identical and the string-terminator
is at index k, byte by byte comparison will not load beyond address
s1+k and s2+k; word by word comparison may load up to 3 bytes beyond
k; double word - up to 7 bytes. If the load of these bytes crosses
a page boundary, it might cause a memory fault (if the page is not mapped)
that would not have happened in byte by byte comparison.
If an address is (double) word aligned, then a load of a (double) word
from that address will not cross a page boundary.
Therefore, the algorithm below considers word and double-word alignment
of strings separately. */
/* High-level description of the algorithm.
* The fast path: if both strings are double-word aligned,
use LDRD to load two words from each string in every loop iteration.
* If the strings have the same offset from a word boundary,
use LDRB to load and compare byte by byte until
the first string is aligned to a word boundary (at most 3 bytes).
This is optimized for quick return on short unaligned strings.
* If the strings have the same offset from a double-word boundary,
use LDRD to load two words from each string in every loop iteration, as in the fast path.
* If the strings do not have the same offset from a double-word boundary,
load a word from the second string before the loop to initialize the queue.
Use LDRD to load two words from every string in every loop iteration.
Inside the loop, load the second word from the second string only after comparing
the first word, using the queued value, to guarantee safety across page boundaries.
* If the strings do not have the same offset from a word boundary,
use LDR and a shift queue. Order of loads and comparisons matters,
similarly to the previous case.
* Use UADD8 and SEL to compare words, and use REV and CLZ to compute the return value.
* The only difference between ARM and Thumb modes is the use of CBZ instruction.
* The only difference between big and little endian is the use of REV in little endian
to compute the return value, instead of MOV.
* No preload. [TODO.]
*/
.macro m_cbz reg label
#ifdef __thumb2__
cbz \reg, \label
#else /* not defined __thumb2__ */
cmp \reg, #0
beq \label
#endif /* not defined __thumb2__ */
.endm /* m_cbz */
.macro m_cbnz reg label
#ifdef __thumb2__
cbnz \reg, \label
#else /* not defined __thumb2__ */
cmp \reg, #0
bne \label
#endif /* not defined __thumb2__ */
.endm /* m_cbnz */
.macro init
/* Macro to save temporary registers and prepare magic values. */
subs sp, sp, #16
strd r4, r5, [sp, #8]
strd r6, r7, [sp]
mvn r6, #0 /* all F */
mov r7, #0 /* all 0 */
.endm /* init */
.macro magic_compare_and_branch w1 w2 label
/* Macro to compare registers w1 and w2 and conditionally branch to label. */
cmp \w1, \w2 /* Are w1 and w2 the same? */
magic_find_zero_bytes \w1
it eq
cmpeq ip, #0 /* Is there a zero byte in w1? */
bne \label
.endm /* magic_compare_and_branch */
.macro magic_find_zero_bytes w1
/* Macro to find all-zero bytes in w1, result is in ip. */
#if (defined (__ARM_FEATURE_DSP))
uadd8 ip, \w1, r6
sel ip, r7, r6
#else /* not defined (__ARM_FEATURE_DSP) */
/* __ARM_FEATURE_DSP is not defined for some Cortex-M processors.
Coincidently, these processors only have Thumb-2 mode, where we can use the
the (large) magic constant available directly as an immediate in instructions.
Note that we cannot use the magic constant in ARM mode, where we need
to create the constant in a register. */
sub ip, \w1, #0x01010101
bic ip, ip, \w1
and ip, ip, #0x80808080
#endif /* not defined (__ARM_FEATURE_DSP) */
.endm /* magic_find_zero_bytes */
.macro setup_return w1 w2
#ifdef __ARMEB__
mov r1, \w1
mov r2, \w2
#else /* not __ARMEB__ */
rev r1, \w1
rev r2, \w2
#endif /* not __ARMEB__ */
.endm /* setup_return */
/*
optpld r0, #0
optpld r1, #0
*/
/* Are both strings double-word aligned? */
orr ip, r0, r1
tst ip, #7
bne do_align
/* Fast path. */
init
doubleword_aligned:
/* Get here when the strings to compare are double-word aligned. */
/* Compare two words in every iteration. */
.p2align 2
2:
/*
optpld r0, #16
optpld r1, #16
*/
/* Load the next double-word from each string. */
ldrd r2, r3, [r0], #8
ldrd r4, r5, [r1], #8
magic_compare_and_branch w1=r2, w2=r4, label=return_24
magic_compare_and_branch w1=r3, w2=r5, label=return_35
b 2b
do_align:
/* Is the first string word-aligned? */
ands ip, r0, #3
beq word_aligned_r0
/* Fast compare byte by byte until the first string is word-aligned. */
/* The offset of r0 from a word boundary is in ip. Thus, the number of bytes
to read until the next word boudnary is 4-ip. */
bic r0, r0, #3
ldr r2, [r0], #4
lsls ip, ip, #31
beq byte2
bcs byte3
byte1:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE1_OFFSET
subs ip, r3, ip
bne fast_return
m_cbz reg=r3, label=fast_return
byte2:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE2_OFFSET
subs ip, r3, ip
bne fast_return
m_cbz reg=r3, label=fast_return
byte3:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE3_OFFSET
subs ip, r3, ip
bne fast_return
m_cbnz reg=r3, label=word_aligned_r0
fast_return:
mov r0, ip
bx lr
word_aligned_r0:
init
/* The first string is word-aligned. */
/* Is the second string word-aligned? */
ands ip, r1, #3
bne strcmp_unaligned
word_aligned:
/* The strings are word-aligned. */
/* Is the first string double-word aligned? */
tst r0, #4
beq doubleword_aligned_r0
/* If r0 is not double-word aligned yet, align it by loading
and comparing the next word from each string. */
ldr r2, [r0], #4
ldr r4, [r1], #4
magic_compare_and_branch w1=r2 w2=r4 label=return_24
doubleword_aligned_r0:
/* Get here when r0 is double-word aligned. */
/* Is r1 doubleword_aligned? */
tst r1, #4
beq doubleword_aligned
/* Get here when the strings to compare are word-aligned,
r0 is double-word aligned, but r1 is not double-word aligned. */
/* Initialize the queue. */
ldr r5, [r1], #4
/* Compare two words in every iteration. */
.p2align 2
3:
/*
optpld r0, #16
optpld r1, #16
*/
/* Load the next double-word from each string and compare. */
ldrd r2, r3, [r0], #8
magic_compare_and_branch w1=r2 w2=r5 label=return_25
ldrd r4, r5, [r1], #8
magic_compare_and_branch w1=r3 w2=r4 label=return_34
b 3b
.macro miscmp_word offsetlo offsethi
/* Macro to compare misaligned strings. */
/* r0, r1 are word-aligned, and at least one of the strings
is not double-word aligned. */
/* Compare one word in every loop iteration. */
/* OFFSETLO is the original bit-offset of r1 from a word-boundary,
OFFSETHI is 32 - OFFSETLO (i.e., offset from the next word). */
/* Initialize the shift queue. */
ldr r5, [r1], #4
/* Compare one word from each string in every loop iteration. */
.p2align 2
7:
ldr r3, [r0], #4
S2LOMEM r5, r5, #\offsetlo
magic_find_zero_bytes w1=r3
cmp r7, ip, S2HIMEM #\offsetlo
and r2, r3, r6, S2LOMEM #\offsetlo
it eq
cmpeq r2, r5
bne return_25
ldr r5, [r1], #4
cmp ip, #0
eor r3, r2, r3
S2HIMEM r2, r5, #\offsethi
it eq
cmpeq r3, r2
bne return_32
b 7b
.endm /* miscmp_word */
strcmp_unaligned:
/* r0 is word-aligned, r1 is at offset ip from a word. */
/* Align r1 to the (previous) word-boundary. */
bic r1, r1, #3
/* Unaligned comparison word by word using LDRs. */
cmp ip, #2
beq miscmp_word_16 /* If ip == 2. */
bge miscmp_word_24 /* If ip == 3. */
miscmp_word offsetlo=8 offsethi=24 /* If ip == 1. */
miscmp_word_16: miscmp_word offsetlo=16 offsethi=16
miscmp_word_24: miscmp_word offsetlo=24 offsethi=8
return_32:
setup_return w1=r3, w2=r2
b do_return
return_34:
setup_return w1=r3, w2=r4
b do_return
return_25:
setup_return w1=r2, w2=r5
b do_return
return_35:
setup_return w1=r3, w2=r5
b do_return
return_24:
setup_return w1=r2, w2=r4
do_return:
#ifdef __ARMEB__
mov r0, ip
#else /* not __ARMEB__ */
rev r0, ip
#endif /* not __ARMEB__ */
/* Restore temporaries early, before computing the return value. */
ldrd r6, r7, [sp]
ldrd r4, r5, [sp, #8]
adds sp, sp, #16
/* There is a zero or a different byte between r1 and r2. */
/* r0 contains a mask of all-zero bytes in r1. */
/* Using r0 and not ip here because cbz requires low register. */
m_cbz reg=r0, label=compute_return_value
clz r0, r0
/* r0 contains the number of bits on the left of the first all-zero byte in r1. */
rsb r0, r0, #24
/* Here, r0 contains the number of bits on the right of the first all-zero byte in r1. */
lsr r1, r1, r0
lsr r2, r2, r0
compute_return_value:
movs r0, #1
cmp r1, r2
/* The return value is computed as follows.
If r1>r2 then (C==1 and Z==0) and LS doesn't hold and r0 is #1 at return.
If r1<r2 then (C==0 and Z==0) and we execute SBC with carry_in=0,
which means r0:=r0-r0-1 and r0 is #-1 at return.
If r1=r2 then (C==1 and Z==1) and we execute SBC with carry_in=1,
which means r0:=r0-r0 and r0 is #0 at return.
(C==0 and Z==1) cannot happen because the carry bit is "not borrow". */
it ls
sbcls r0, r0, r0
bx lr
#else /* !(defined (_ISA_THUMB_2) || defined (_ISA_ARM_6)
defined (__OPTIMIZE_SIZE__) || defined (PREFER_SIZE_OVER_SPEED) ||
(defined (__thumb__) && !defined (__thumb2__))) */
/* Use LDR whenever possible. */
#ifdef __thumb2__
#define magic1(REG) 0x01010101
#define magic2(REG) 0x80808080
#else
#define magic1(REG) REG
#define magic2(REG) REG, lsl #7
#endif
optpld r0
optpld r1
eor r2, r0, r1
tst r2, #3
/* Strings not at same byte offset from a word boundary. */
bne strcmp_unaligned
ands r2, r0, #3
bic r0, r0, #3
bic r1, r1, #3
ldr ip, [r0], #4
it eq
ldreq r3, [r1], #4
beq 1f
/* Although s1 and s2 have identical initial alignment, they are
not currently word aligned. Rather than comparing bytes,
make sure that any bytes fetched from before the addressed
bytes are forced to 0xff. Then they will always compare
equal. */
eor r2, r2, #3
lsl r2, r2, #3
mvn r3, MSB
S2LOMEM r2, r3, r2
ldr r3, [r1], #4
orr ip, ip, r2
orr r3, r3, r2
1:
#ifndef __thumb2__
/* Load the 'magic' constant 0x01010101. */
str r4, [sp, #-4]!
mov r4, #1
orr r4, r4, r4, lsl #8
orr r4, r4, r4, lsl #16
#endif
.p2align 2
4:
optpld r0, #8
optpld r1, #8
sub r2, ip, magic1(r4)
cmp ip, r3
itttt eq
/* check for any zero bytes in first word */
biceq r2, r2, ip
tsteq r2, magic2(r4)
ldreq ip, [r0], #4
ldreq r3, [r1], #4
beq 4b
2:
/* There's a zero or a different byte in the word */
S2HIMEM r0, ip, #24
S2LOMEM ip, ip, #8
cmp r0, #1
it cs
cmpcs r0, r3, S2HIMEM #24
it eq
S2LOMEMEQ r3, r3, #8
beq 2b
/* On a big-endian machine, r0 contains the desired byte in bits
0-7; on a little-endian machine they are in bits 24-31. In
both cases the other bits in r0 are all zero. For r3 the
interesting byte is at the other end of the word, but the
other bits are not necessarily zero. We need a signed result
representing the differnece in the unsigned bytes, so for the
little-endian case we can't just shift the interesting bits
up. */
#ifdef __ARMEB__
sub r0, r0, r3, lsr #24
#else
and r3, r3, #255
#ifdef __thumb2__
/* No RSB instruction in Thumb2 */
lsr r0, r0, #24
sub r0, r0, r3
#else
rsb r0, r3, r0, lsr #24
#endif
#endif
#ifndef __thumb2__
ldr r4, [sp], #4
#endif
RETURN
strcmp_unaligned:
#if 0
/* The assembly code below is based on the following alogrithm. */
#ifdef __ARMEB__
#define RSHIFT <<
#define LSHIFT >>
#else
#define RSHIFT >>
#define LSHIFT <<
#endif
#define body(shift) \
mask = 0xffffffffU RSHIFT shift; \
w1 = *wp1++; \
w2 = *wp2++; \
do \
{ \
t1 = w1 & mask; \
if (__builtin_expect(t1 != w2 RSHIFT shift, 0)) \
{ \
w2 RSHIFT= shift; \
break; \
} \
if (__builtin_expect(((w1 - b1) & ~w1) & (b1 << 7), 0)) \
{ \
/* See comment in assembler below re syndrome on big-endian */\
if ((((w1 - b1) & ~w1) & (b1 << 7)) & mask) \
w2 RSHIFT= shift; \
else \
{ \
w2 = *wp2; \
t1 = w1 RSHIFT (32 - shift); \
w2 = (w2 LSHIFT (32 - shift)) RSHIFT (32 - shift); \
} \
break; \
} \
w2 = *wp2++; \
t1 ^= w1; \
if (__builtin_expect(t1 != w2 LSHIFT (32 - shift), 0)) \
{ \
t1 = w1 >> (32 - shift); \
w2 = (w2 << (32 - shift)) RSHIFT (32 - shift); \
break; \
} \
w1 = *wp1++; \
} while (1)
const unsigned* wp1;
const unsigned* wp2;
unsigned w1, w2;
unsigned mask;
unsigned shift;
unsigned b1 = 0x01010101;
char c1, c2;
unsigned t1;
while (((unsigned) s1) & 3)
{
c1 = *s1++;
c2 = *s2++;
if (c1 == 0 || c1 != c2)
return c1 - (int)c2;
}
wp1 = (unsigned*) (((unsigned)s1) & ~3);
wp2 = (unsigned*) (((unsigned)s2) & ~3);
t1 = ((unsigned) s2) & 3;
if (t1 == 1)
{
body(8);
}
else if (t1 == 2)
{
body(16);
}
else
{
body (24);
}
do
{
#ifdef __ARMEB__
c1 = (char) t1 >> 24;
c2 = (char) w2 >> 24;
#else /* not __ARMEB__ */
c1 = (char) t1;
c2 = (char) w2;
#endif /* not __ARMEB__ */
t1 RSHIFT= 8;
w2 RSHIFT= 8;
} while (c1 != 0 && c1 == c2);
return c1 - c2;
#endif /* 0 */
wp1 .req r0
wp2 .req r1
b1 .req r2
w1 .req r4
w2 .req r5
t1 .req ip
@ r3 is scratch
/* First of all, compare bytes until wp1(sp1) is word-aligned. */
1:
tst wp1, #3
beq 2f
ldrb r2, [wp1], #1
ldrb r3, [wp2], #1
cmp r2, #1
it cs
cmpcs r2, r3
beq 1b
sub r0, r2, r3
RETURN
2:
str r5, [sp, #-4]!
str r4, [sp, #-4]!
//stmfd sp!, {r4, r5}
mov b1, #1
orr b1, b1, b1, lsl #8
orr b1, b1, b1, lsl #16
and t1, wp2, #3
bic wp2, wp2, #3
ldr w1, [wp1], #4
ldr w2, [wp2], #4
cmp t1, #2
beq 2f
bhi 3f
/* Critical inner Loop: Block with 3 bytes initial overlap */
.p2align 2
1:
bic t1, w1, MSB
cmp t1, w2, S2LOMEM #8
sub r3, w1, b1
bic r3, r3, w1
bne 4f
ands r3, r3, b1, lsl #7
it eq
ldreq w2, [wp2], #4
bne 5f
eor t1, t1, w1
cmp t1, w2, S2HIMEM #24
bne 6f
ldr w1, [wp1], #4
b 1b
4:
S2LOMEM w2, w2, #8
b 8f
5:
#ifdef __ARMEB__
/* The syndrome value may contain false ones if the string ends
with the bytes 0x01 0x00 */
tst w1, #0xff000000
itt ne
tstne w1, #0x00ff0000
tstne w1, #0x0000ff00
beq 7f
#else
bics r3, r3, #0xff000000
bne 7f
#endif
ldrb w2, [wp2]
S2LOMEM t1, w1, #24
#ifdef __ARMEB__
lsl w2, w2, #24
#endif
b 8f
6:
S2LOMEM t1, w1, #24
and w2, w2, LSB
b 8f
/* Critical inner Loop: Block with 2 bytes initial overlap */
.p2align 2
2:
S2HIMEM t1, w1, #16
sub r3, w1, b1
S2LOMEM t1, t1, #16
bic r3, r3, w1
cmp t1, w2, S2LOMEM #16
bne 4f
ands r3, r3, b1, lsl #7
it eq
ldreq w2, [wp2], #4
bne 5f
eor t1, t1, w1
cmp t1, w2, S2HIMEM #16
bne 6f
ldr w1, [wp1], #4
b 2b
5:
#ifdef __ARMEB__
/* The syndrome value may contain false ones if the string ends
with the bytes 0x01 0x00 */
tst w1, #0xff000000
it ne
tstne w1, #0x00ff0000
beq 7f
#else
lsls r3, r3, #16
bne 7f
#endif
ldrh w2, [wp2]
S2LOMEM t1, w1, #16
#ifdef __ARMEB__
lsl w2, w2, #16
#endif
b 8f
6:
S2HIMEM w2, w2, #16
S2LOMEM t1, w1, #16
4:
S2LOMEM w2, w2, #16
b 8f
/* Critical inner Loop: Block with 1 byte initial overlap */
.p2align 2
3:
and t1, w1, LSB
cmp t1, w2, S2LOMEM #24
sub r3, w1, b1
bic r3, r3, w1
bne 4f
ands r3, r3, b1, lsl #7
it eq
ldreq w2, [wp2], #4
bne 5f
eor t1, t1, w1
cmp t1, w2, S2HIMEM #8
bne 6f
ldr w1, [wp1], #4
b 3b
4:
S2LOMEM w2, w2, #24
b 8f
5:
/* The syndrome value may contain false ones if the string ends
with the bytes 0x01 0x00 */
tst w1, LSB
beq 7f
ldr w2, [wp2], #4
6:
S2LOMEM t1, w1, #8
bic w2, w2, MSB
b 8f
7:
mov r0, #0
//ldmfd sp!, {r4, r5}
ldr r4, [sp], #4
ldr r5, [sp], #4
RETURN
8:
and r2, t1, LSB
and r0, w2, LSB
cmp r0, #1
it cs
cmpcs r0, r2
itt eq
S2LOMEMEQ t1, t1, #8
S2LOMEMEQ w2, w2, #8
beq 8b
sub r0, r2, r0
//ldmfd sp!, {r4, r5}
ldr r4, [sp], #4
ldr r5, [sp], #4
RETURN
#endif /* !(defined (_ISA_THUMB_2) || defined (_ISA_ARM_6)
defined (__OPTIMIZE_SIZE__) || defined (PREFER_SIZE_OVER_SPEED) ||
(defined (__thumb__) && !defined (__thumb2__))) */

View File

@@ -33,16 +33,16 @@ size_t strlen(const char *s)
{
__builtin_prefetch(s);
__builtin_prefetch(s+32);
union {
const char *b;
const uint32_t *w;
uintptr_t i;
} u;
// these are some scratch variables for the asm code below
uint32_t v, t;
// initialize the string length to zero
size_t l = 0;
@@ -60,60 +60,52 @@ size_t strlen(const char *s)
// We need to process 32 bytes per loop to schedule PLD properly
// and achieve the maximum bus speed.
asm(
"ldr %[v], [%[s]], #4 \n"
"ldr %[v], [ %[s] ], #4 \n"
"sub %[l], %[l], %[s] \n"
"0: \n"
#if __ARM_HAVE_PLD
"pld [%[s], #64] \n"
"pld [ %[s], #64 ] \n"
#endif
"sub %[t], %[v], %[mask], lsr #7\n"
"and %[t], %[t], %[mask] \n"
"bics %[t], %[t], %[v] \n"
"it eq \n"
"ldreq %[v], [%[s]], #4 \n"
"ldreq %[v], [ %[s] ], #4 \n"
#if !defined(__OPTIMIZE_SIZE__)
"bne 1f \n"
"sub %[t], %[v], %[mask], lsr #7\n"
"and %[t], %[t], %[mask] \n"
"bics %[t], %[t], %[v] \n"
"it eq \n"
"ldreq %[v], [%[s]], #4 \n"
"ldreq %[v], [ %[s] ], #4 \n"
"bne 1f \n"
"sub %[t], %[v], %[mask], lsr #7\n"
"and %[t], %[t], %[mask] \n"
"bics %[t], %[t], %[v] \n"
"it eq \n"
"ldreq %[v], [%[s]], #4 \n"
"ldreq %[v], [ %[s] ], #4 \n"
"bne 1f \n"
"sub %[t], %[v], %[mask], lsr #7\n"
"and %[t], %[t], %[mask] \n"
"bics %[t], %[t], %[v] \n"
"it eq \n"
"ldreq %[v], [%[s]], #4 \n"
"ldreq %[v], [ %[s] ], #4 \n"
"bne 1f \n"
"sub %[t], %[v], %[mask], lsr #7\n"
"and %[t], %[t], %[mask] \n"
"bics %[t], %[t], %[v] \n"
"it eq \n"
"ldreq %[v], [%[s]], #4 \n"
"ldreq %[v], [ %[s] ], #4 \n"
"bne 1f \n"
"sub %[t], %[v], %[mask], lsr #7\n"
"and %[t], %[t], %[mask] \n"
"bics %[t], %[t], %[v] \n"
"it eq \n"
"ldreq %[v], [%[s]], #4 \n"
"ldreq %[v], [ %[s] ], #4 \n"
"bne 1f \n"
"sub %[t], %[v], %[mask], lsr #7\n"
"and %[t], %[t], %[mask] \n"
"bics %[t], %[t], %[v] \n"
"it eq \n"
"ldreq %[v], [%[s]], #4 \n"
"ldreq %[v], [ %[s] ], #4 \n"
"bne 1f \n"
"sub %[t], %[v], %[mask], lsr #7\n"
"and %[t], %[t], %[mask] \n"
"bics %[t], %[t], %[v] \n"
"it eq \n"
"ldreq %[v], [%[s]], #4 \n"
"ldreq %[v], [ %[s] ], #4 \n"
#endif
"beq 0b \n"
"1: \n"
@@ -125,14 +117,13 @@ size_t strlen(const char *s)
"beq 2f \n"
"add %[l], %[l], #1 \n"
"tst %[v], #0xFF0000 \n"
"it ne \n"
"addne %[l], %[l], #1 \n"
"2: \n"
: [l]"=&r"(l), [v]"=&r"(v), [t]"=&r"(t), [s]"=&r"(u.b)
: "%[l]"(l), "%[s]"(u.b), [mask]"r"(0x80808080UL)
: "cc"
);
done:
return l;
}

View File

@@ -25,22 +25,47 @@
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <sys/linux-syscalls.h>
#include <linux/err.h>
#include <machine/asm.h>
ENTRY(syscall)
.text
.align 4
.type syscall,#function
.globl syscall
.text
.align
#if __ARM_EABI__
syscall:
mov ip, sp
stmfd sp!, {r4, r5, r6, r7}
stmfd sp!, {r4, r5, r6, r7}
mov r7, r0
mov r0, r1
mov r1, r2
mov r2, r3
ldmfd ip, {r3, r4, r5, r6}
ldmfd ip, {r3, r4, r5, r6}
swi #0
ldmfd sp!, {r4, r5, r6, r7}
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
END(syscall)
movs r0, r0
bxpl lr
b __set_syscall_errno
#else
#ifndef __NR_syscall
#define __NR_syscall 113
#endif
syscall:
stmfd sp!, {r4, r5, lr}
ldr r4, [sp, #12]
ldr r5, [sp, #16]
swi __NR_syscall
ldmfd sp!, {r4, r5, lr}
movs r0, r0
bxpl lr
b __set_syscall_errno
#endif

View File

@@ -25,11 +25,6 @@
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <linux/err.h>
#include <asm/unistd.h>
#include <machine/asm.h>
/* unlike our auto-generated syscall stubs, this code saves lr
on the stack, as well as a few other registers. this makes
our stack unwinder happy, when we generate debug stack
@@ -38,14 +33,19 @@
of a corrupted malloc heap).
*/
#include <sys/linux-syscalls.h>
#include <machine/asm.h>
#ifndef __NR_tgkill
#define __NR_tgkill 268
#endif
ENTRY(tgkill)
.save {r4-r7, ip, lr}
stmfd sp!, {r4-r7, ip, lr}
ldr r7, =__NR_tgkill
swi #0
ldmfd sp!, {r4-r7, ip, lr}
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
movs r0, r0
bxpl lr
b __set_syscall_errno
END(tgkill)

View File

@@ -25,11 +25,6 @@
* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <linux/err.h>
#include <asm/unistd.h>
#include <machine/asm.h>
/* unlike our auto-generated syscall stubs, this code saves lr
on the stack, as well as a few other registers. this makes
our stack unwinder happy, when we generate debug stack
@@ -38,13 +33,19 @@
of a corrupted malloc heap).
*/
#include <sys/linux-syscalls.h>
#include <machine/asm.h>
#ifndef __NR_tkill
#define __NR_tkill 238
#endif
ENTRY(tkill)
stmfd sp!, {r4-r7, ip, lr}
ldr r7, =__NR_tkill
swi #0
ldmfd sp!, {r4-r7, ip, lr}
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
movs r0, r0
bxpl lr
b __set_syscall_errno
END(tkill)

View File

@@ -1,227 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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 <machine/asm.h>
#include "libc_events.h"
.syntax unified
.thumb
.thumb_func
// Get the length of src string, then get the source of the dst string.
// Check that the two lengths together don't exceed the threshold, then
// do a memcpy of the data.
ENTRY(__strcat_chk)
.cfi_startproc
pld [r0, #0]
push {r0, lr}
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
push {r4, r5}
.save {r4, r5}
.cfi_adjust_cfa_offset 8
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
mov lr, r2
// Save the dst register to r5
mov r5, r0
// Zero out r4
eor r4, r4, r4
// r1 contains the address of the string to count.
.L_strlen_start:
mov r0, r1
ands r3, r1, #7
beq .L_mainloop
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq .L_align_to_32
ldrb r2, [r1], #1
cbz r2, .L_update_count_and_finish
.L_align_to_32:
bcc .L_align_to_64
ands ip, r3, #2
beq .L_align_to_64
ldrb r2, [r1], #1
cbz r2, .L_update_count_and_finish
ldrb r2, [r1], #1
cbz r2, .L_update_count_and_finish
.L_align_to_64:
tst r3, #4
beq .L_mainloop
ldr r3, [r1], #4
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
.p2align 2
.L_mainloop:
ldrd r2, r3, [r1], #8
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne .L_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
b .L_mainloop
.L_update_count_and_finish:
sub r3, r1, r0
sub r3, r3, #1
b .L_finish
.L_zero_in_first_register:
sub r3, r1, r0
lsls r2, ip, #17
bne .L_sub8_and_finish
bcs .L_sub7_and_finish
lsls ip, ip, #1
bne .L_sub6_and_finish
sub r3, r3, #5
b .L_finish
.L_sub8_and_finish:
sub r3, r3, #8
b .L_finish
.L_sub7_and_finish:
sub r3, r3, #7
b .L_finish
.L_sub6_and_finish:
sub r3, r3, #6
b .L_finish
.L_zero_in_second_register:
sub r3, r1, r0
lsls r2, ip, #17
bne .L_sub4_and_finish
bcs .L_sub3_and_finish
lsls ip, ip, #1
bne .L_sub2_and_finish
sub r3, r3, #1
b .L_finish
.L_sub4_and_finish:
sub r3, r3, #4
b .L_finish
.L_sub3_and_finish:
sub r3, r3, #3
b .L_finish
.L_sub2_and_finish:
sub r3, r3, #2
.L_finish:
cmp r4, #0
bne .L_strlen_done
// Time to get the dst string length.
mov r1, r5
// Save the original source address to r5.
mov r5, r0
// Save the current length (adding 1 for the terminator).
add r4, r3, #1
b .L_strlen_start
// r0 holds the pointer to the dst string.
// r3 holds the dst string length.
// r4 holds the src string length + 1.
.L_strlen_done:
add r2, r3, r4
cmp r2, lr
bhi __strcat_chk_failed
// Set up the registers for the memcpy code.
mov r1, r5
pld [r1, #64]
mov r2, r4
add r0, r0, r3
pop {r4, r5}
.cfi_endproc
END(__strcat_chk)
#define MEMCPY_BASE __strcat_chk_memcpy_base
#define MEMCPY_BASE_ALIGNED __strcat_chk_memcpy_base_aligned
#include "memcpy_base.S"
ENTRY(__strcat_chk_failed)
.cfi_startproc
.save {r0, lr}
.save {r4, r5}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
.cfi_adjust_cfa_offset 8
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_STRCAT_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+4)
.cfi_endproc
END(__strcat_chk_failed)
.data
error_string:
.string "strcat buffer overflow"

View File

@@ -1,188 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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 <machine/asm.h>
#include "libc_events.h"
.syntax unified
.thumb
.thumb_func
// Get the length of the source string first, then do a memcpy of the data
// instead of a strcpy.
ENTRY(__strcpy_chk)
.cfi_startproc
pld [r0, #0]
push {r0, lr}
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
mov lr, r2
mov r0, r1
ands r3, r1, #7
beq .L_mainloop
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq .L_align_to_32
ldrb r2, [r0], #1
cbz r2, .L_update_count_and_finish
.L_align_to_32:
bcc .L_align_to_64
ands ip, r3, #2
beq .L_align_to_64
ldrb r2, [r0], #1
cbz r2, .L_update_count_and_finish
ldrb r2, [r0], #1
cbz r2, .L_update_count_and_finish
.L_align_to_64:
tst r3, #4
beq .L_mainloop
ldr r3, [r0], #4
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
.p2align 2
.L_mainloop:
ldrd r2, r3, [r0], #8
pld [r0, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne .L_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
b .L_mainloop
.L_update_count_and_finish:
sub r3, r0, r1
sub r3, r3, #1
b .L_check_size
.L_zero_in_first_register:
sub r3, r0, r1
lsls r2, ip, #17
bne .L_sub8_and_finish
bcs .L_sub7_and_finish
lsls ip, ip, #1
bne .L_sub6_and_finish
sub r3, r3, #5
b .L_check_size
.L_sub8_and_finish:
sub r3, r3, #8
b .L_check_size
.L_sub7_and_finish:
sub r3, r3, #7
b .L_check_size
.L_sub6_and_finish:
sub r3, r3, #6
b .L_check_size
.L_zero_in_second_register:
sub r3, r0, r1
lsls r2, ip, #17
bne .L_sub4_and_finish
bcs .L_sub3_and_finish
lsls ip, ip, #1
bne .L_sub2_and_finish
sub r3, r3, #1
b .L_check_size
.L_sub4_and_finish:
sub r3, r3, #4
b .L_check_size
.L_sub3_and_finish:
sub r3, r3, #3
b .L_check_size
.L_sub2_and_finish:
sub r3, r3, #2
.L_check_size:
pld [r1, #0]
pld [r1, #64]
ldr r0, [sp]
cmp r3, lr
bhs __strcpy_chk_failed
// Add 1 for copy length to get the string terminator.
add r2, r3, #1
.cfi_endproc
END(__strcpy_chk)
#define MEMCPY_BASE __strcpy_chk_memcpy_base
#define MEMCPY_BASE_ALIGNED __strcpy_chk_memcpy_base_aligned
#include "memcpy_base.S"
ENTRY(__strcpy_chk_failed)
.cfi_startproc
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_STRCPY_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+4)
.cfi_endproc
END(__strcpy_chk_failed)
.data
error_string:
.string "strcpy buffer overflow"

View File

@@ -1,112 +0,0 @@
/*
* 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:
* * 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.
*
* 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.
*/
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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.
*/
// Prototype: void *memcpy (void *dst, const void *src, size_t count).
#include <machine/asm.h>
#include "libc_events.h"
.text
.syntax unified
.fpu neon
ENTRY(__memcpy_chk)
.cfi_startproc
cmp r2, r3
bhi __memcpy_chk_fail
// Fall through to memcpy...
.cfi_endproc
END(__memcpy_chk)
ENTRY(memcpy)
.cfi_startproc
pld [r1, #64]
push {r0, lr}
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
.cfi_endproc
END(memcpy)
#define MEMCPY_BASE __memcpy_base
#define MEMCPY_BASE_ALIGNED __memcpy_base_aligned
#include "memcpy_base.S"
ENTRY(__memcpy_chk_fail)
.cfi_startproc
// Preserve lr for backtrace.
push {lr}
.save {lr}
.cfi_def_cfa_offset 4
.cfi_rel_offset lr, 0
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_MEMCPY_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+8)
.cfi_endproc
END(__memcpy_chk_fail)
.data
error_string:
.string "memcpy buffer overflow"

View File

@@ -1,329 +0,0 @@
/*
* 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:
* * 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.
*
* 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.
*/
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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.
*/
ENTRY(MEMCPY_BASE)
.cfi_startproc
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
// Assumes that n >= 0, and dst, src are valid pointers.
// For any sizes less than 832 use the neon code that doesn't
// care about the src alignment. This avoids any checks
// for src alignment, and offers the best improvement since
// smaller sized copies are dominated by the overhead of
// the pre and post main loop.
// For larger copies, if src and dst cannot both be aligned to
// word boundaries, use the neon code.
// For all other copies, align dst to a double word boundary
// and copy using LDRD/STRD instructions.
cmp r2, #16
blo .L_copy_less_than_16_unknown_align
// TODO: The aligned copy code is extremely slow copying some large
// buffers so always go through the unaligned path for now.
//cmp r2, #832
//bge .L_check_alignment
.L_copy_unknown_alignment:
// Unknown alignment of src and dst.
// Assumes that the first few bytes have already been prefetched.
// Align destination to 128 bits. The mainloop store instructions
// require this alignment or they will throw an exception.
rsb r3, r0, #0
ands r3, r3, #0xF
beq 2f
// Copy up to 15 bytes (count in r3).
sub r2, r2, r3
movs ip, r3, lsl #31
itt mi
ldrbmi lr, [r1], #1
strbmi lr, [r0], #1
itttt cs
ldrbcs ip, [r1], #1
ldrbcs lr, [r1], #1
strbcs ip, [r0], #1
strbcs lr, [r0], #1
movs ip, r3, lsl #29
bge 1f
// Copies 4 bytes, dst 32 bits aligned before, at least 64 bits after.
vld4.8 {d0[0], d1[0], d2[0], d3[0]}, [r1]!
vst4.8 {d0[0], d1[0], d2[0], d3[0]}, [r0, :32]!
1: bcc 2f
// Copies 8 bytes, dst 64 bits aligned before, at least 128 bits after.
vld1.8 {d0}, [r1]!
vst1.8 {d0}, [r0, :64]!
2: // Make sure we have at least 64 bytes to copy.
subs r2, r2, #64
blo 2f
1: // The main loop copies 64 bytes at a time.
vld1.8 {d0 - d3}, [r1]!
vld1.8 {d4 - d7}, [r1]!
pld [r1, #(64*4)]
subs r2, r2, #64
vst1.8 {d0 - d3}, [r0, :128]!
vst1.8 {d4 - d7}, [r0, :128]!
bhs 1b
2: // Fix-up the remaining count and make sure we have >= 32 bytes left.
adds r2, r2, #32
blo 3f
// 32 bytes. These cache lines were already preloaded.
vld1.8 {d0 - d3}, [r1]!
sub r2, r2, #32
vst1.8 {d0 - d3}, [r0, :128]!
3: // Less than 32 left.
add r2, r2, #32
tst r2, #0x10
beq .L_copy_less_than_16_unknown_align
// Copies 16 bytes, destination 128 bits aligned.
vld1.8 {d0, d1}, [r1]!
vst1.8 {d0, d1}, [r0, :128]!
.L_copy_less_than_16_unknown_align:
// Copy up to 15 bytes (count in r2).
movs ip, r2, lsl #29
bcc 1f
vld1.8 {d0}, [r1]!
vst1.8 {d0}, [r0]!
1: bge 2f
vld4.8 {d0[0], d1[0], d2[0], d3[0]}, [r1]!
vst4.8 {d0[0], d1[0], d2[0], d3[0]}, [r0]!
2: // Copy 0 to 4 bytes.
lsls r2, r2, #31
itt ne
ldrbne lr, [r1], #1
strbne lr, [r0], #1
itttt cs
ldrbcs ip, [r1], #1
ldrbcs lr, [r1]
strbcs ip, [r0], #1
strbcs lr, [r0]
pop {r0, pc}
.L_check_alignment:
// If src and dst cannot both be aligned to a word boundary,
// use the unaligned copy version.
eor r3, r0, r1
ands r3, r3, #0x3
bne .L_copy_unknown_alignment
.cfi_endproc
END(MEMCPY_BASE)
ENTRY(MEMCPY_BASE_ALIGNED)
.cfi_startproc
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
// To try and improve performance, stack layout changed,
// i.e., not keeping the stack looking like users expect
// (highest numbered register at highest address).
strd r4, r5, [sp, #-8]!
.save {r4, r5}
.cfi_adjust_cfa_offset 8
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
strd r6, r7, [sp, #-8]!
.save {r6, r7}
.cfi_adjust_cfa_offset 8
.cfi_rel_offset r6, 0
.cfi_rel_offset r7, 0
strd r8, r9, [sp, #-8]!
.save {r8, r9}
.cfi_adjust_cfa_offset 8
.cfi_rel_offset r8, 0
.cfi_rel_offset r9, 4
// Optimized for already aligned dst code.
ands ip, r0, #3
bne .L_dst_not_word_aligned
.L_word_aligned:
// Align the destination buffer to 8 bytes, to make sure double
// loads and stores don't cross a cache line boundary,
// as they are then more expensive even if the data is in the cache
// (require two load/store issue cycles instead of one).
// If only one of the buffers is not 8 bytes aligned,
// then it's more important to align dst than src,
// because there is more penalty for stores
// than loads that cross a cacheline boundary.
// This check and realignment are only done if there is >= 832
// bytes to copy.
// Dst is word aligned, but check if it is already double word aligned.
ands r3, r0, #4
beq 1f
ldr r3, [r1], #4
str r3, [r0], #4
sub r2, #4
1: // Can only get here if > 64 bytes to copy, so don't do check r2.
sub r2, #64
2: // Every loop iteration copies 64 bytes.
.irp offset, #0, #8, #16, #24, #32
ldrd r4, r5, [r1, \offset]
strd r4, r5, [r0, \offset]
.endr
ldrd r4, r5, [r1, #40]
ldrd r6, r7, [r1, #48]
ldrd r8, r9, [r1, #56]
// Keep the pld as far from the next load as possible.
// The amount to prefetch was determined experimentally using
// large sizes, and verifying the prefetch size does not affect
// the smaller copies too much.
// WARNING: If the ldrd and strd instructions get too far away
// from each other, performance suffers. Three loads
// in a row is the best tradeoff.
pld [r1, #(64*16)]
strd r4, r5, [r0, #40]
strd r6, r7, [r0, #48]
strd r8, r9, [r0, #56]
add r0, r0, #64
add r1, r1, #64
subs r2, r2, #64
bge 2b
// Fix-up the remaining count and make sure we have >= 32 bytes left.
adds r2, r2, #32
blo 4f
// Copy 32 bytes. These cache lines were already preloaded.
.irp offset, #0, #8, #16, #24
ldrd r4, r5, [r1, \offset]
strd r4, r5, [r0, \offset]
.endr
add r1, r1, #32
add r0, r0, #32
sub r2, r2, #32
4: // Less than 32 left.
add r2, r2, #32
tst r2, #0x10
beq 5f
// Copy 16 bytes.
.irp offset, #0, #8
ldrd r4, r5, [r1, \offset]
strd r4, r5, [r0, \offset]
.endr
add r1, r1, #16
add r0, r0, #16
5: // Copy up to 15 bytes (count in r2).
movs ip, r2, lsl #29
bcc 1f
// Copy 8 bytes.
ldrd r4, r5, [r1], #8
strd r4, r5, [r0], #8
1: bge 2f
// Copy 4 bytes.
ldr r4, [r1], #4
str r4, [r0], #4
2: // Copy 0 to 4 bytes.
lsls r2, r2, #31
itt ne
ldrbne lr, [r1], #1
strbne lr, [r0], #1
itttt cs
ldrbcs ip, [r1], #1
ldrbcs lr, [r1]
strbcs ip, [r0], #1
strbcs lr, [r0]
// Restore registers: optimized pop {r0, pc}
ldrd r8, r9, [sp], #8
ldrd r6, r7, [sp], #8
ldrd r4, r5, [sp], #8
pop {r0, pc}
.L_dst_not_word_aligned:
// Align dst to word.
rsb ip, ip, #4
cmp ip, #2
itt gt
ldrbgt lr, [r1], #1
strbgt lr, [r0], #1
itt ge
ldrbge lr, [r1], #1
strbge lr, [r0], #1
ldrb lr, [r1], #1
strb lr, [r0], #1
sub r2, r2, ip
// Src is guaranteed to be at least word aligned by this point.
b .L_word_aligned
.cfi_endproc
END(MEMCPY_BASE_ALIGNED)

View File

@@ -1,200 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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 <machine/cpu-features.h>
#include <machine/asm.h>
#include "libc_events.h"
/*
* Optimized memset() for ARM.
*
* memset() returns its first argument.
*/
.fpu neon
.syntax unified
ENTRY(__memset_chk)
.cfi_startproc
cmp r2, r3
bls .L_done
// Preserve lr for backtrace.
.save {lr}
push {lr}
.cfi_def_cfa_offset 4
.cfi_rel_offset lr, 0
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_MEMSET_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+8)
.cfi_endproc
END(__memset_chk)
ENTRY(bzero)
.cfi_startproc
mov r2, r1
mov r1, #0
.L_done:
// Fall through to memset...
.cfi_endproc
END(bzero)
ENTRY(memset)
.cfi_startproc
.save {r0}
stmfd sp!, {r0}
.cfi_def_cfa_offset 4
.cfi_rel_offset r0, 0
// The new algorithm is slower for copies < 16 so use the old
// neon code in that case.
cmp r2, #16
blo .L_set_less_than_16_unknown_align
// Use strd which requires an even and odd register so move the
// values so that:
// r0 and r1 contain the memset value
// r2 is the number of bytes to set
// r3 is the destination pointer
mov r3, r0
// Copy the byte value in every byte of r1.
mov r1, r1, lsl #24
orr r1, r1, r1, lsr #8
orr r1, r1, r1, lsr #16
.L_check_alignment:
// Align destination to a double word to avoid the strd crossing
// a cache line boundary.
ands ip, r3, #7
bne .L_do_double_word_align
.L_double_word_aligned:
mov r0, r1
subs r2, #64
blo .L_set_less_than_64
1: // Main loop sets 64 bytes at a time.
.irp offset, #0, #8, #16, #24, #32, #40, #48, #56
strd r0, r1, [r3, \offset]
.endr
add r3, #64
subs r2, #64
bge 1b
.L_set_less_than_64:
// Restore r2 to the count of bytes left to set.
add r2, #64
lsls ip, r2, #27
bcc .L_set_less_than_32
// Set 32 bytes.
.irp offset, #0, #8, #16, #24
strd r0, r1, [r3, \offset]
.endr
add r3, #32
.L_set_less_than_32:
bpl .L_set_less_than_16
// Set 16 bytes.
.irp offset, #0, #8
strd r0, r1, [r3, \offset]
.endr
add r3, #16
.L_set_less_than_16:
// Less than 16 bytes to set.
lsls ip, r2, #29
bcc .L_set_less_than_8
// Set 8 bytes.
strd r0, r1, [r3], #8
.L_set_less_than_8:
bpl .L_set_less_than_4
// Set 4 bytes
str r1, [r3], #4
.L_set_less_than_4:
lsls ip, r2, #31
it ne
strbne r1, [r3], #1
itt cs
strbcs r1, [r3], #1
strbcs r1, [r3]
ldmfd sp!, {r0}
bx lr
.L_do_double_word_align:
rsb ip, ip, #8
sub r2, r2, ip
movs r0, ip, lsl #31
it mi
strbmi r1, [r3], #1
itt cs
strbcs r1, [r3], #1
strbcs r1, [r3], #1
// Dst is at least word aligned by this point.
cmp ip, #4
blo .L_double_word_aligned
str r1, [r3], #4
b .L_double_word_aligned
.L_set_less_than_16_unknown_align:
// Set up to 15 bytes.
vdup.8 d0, r1
movs ip, r2, lsl #29
bcc 1f
vst1.8 {d0}, [r0]!
1: bge 2f
vst1.32 {d0[0]}, [r0]!
2: movs ip, r2, lsl #31
it mi
strbmi r1, [r0], #1
itt cs
strbcs r1, [r0], #1
strbcs r1, [r0], #1
ldmfd sp!, {r0}
bx lr
.cfi_endproc
END(memset)
.data
error_string:
.string "memset buffer overflow"

View File

@@ -1,568 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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.
*/
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 <machine/asm.h>
.syntax unified
.thumb
.thumb_func
.macro m_push
push {r0, r4, r5, lr}
.endm // m_push
.macro m_pop
pop {r0, r4, r5, pc}
.endm // m_pop
.macro m_scan_byte
ldrb r3, [r0]
cbz r3, strcat_r0_scan_done
add r0, #1
.endm // m_scan_byte
.macro m_copy_byte reg, cmd, label
ldrb \reg, [r1], #1
strb \reg, [r0], #1
\cmd \reg, \label
.endm // m_copy_byte
ENTRY(strcat)
// Quick check to see if src is empty.
ldrb r2, [r1]
pld [r1, #0]
cbnz r2, strcat_continue
bx lr
strcat_continue:
// To speed up really small dst strings, unroll checking the first 4 bytes.
m_push
m_scan_byte
m_scan_byte
m_scan_byte
m_scan_byte
ands r3, r0, #7
beq strcat_mainloop
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq strcat_align_to_32
ldrb r5, [r0]
cbz r5, strcat_r0_scan_done
add r0, r0, #1
strcat_align_to_32:
bcc strcat_align_to_64
ldrb r2, [r0]
cbz r2, strcat_r0_scan_done
add r0, r0, #1
ldrb r4, [r0]
cbz r4, strcat_r0_scan_done
add r0, r0, #1
strcat_align_to_64:
tst r3, #4
beq strcat_mainloop
ldr r3, [r0], #4
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcat_zero_in_second_register
b strcat_mainloop
strcat_r0_scan_done:
// For short copies, hard-code checking the first 8 bytes since this
// new code doesn't win until after about 8 bytes.
m_copy_byte reg=r2, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r3, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r4, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r5, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r2, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r3, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r4, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r5, cmd=cbnz, label=strcpy_continue
strcpy_finish:
m_pop
strcpy_continue:
ands r3, r0, #7
beq strcpy_check_src_align
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq strcpy_align_to_32
ldrb r2, [r1], #1
strb r2, [r0], #1
cbz r2, strcpy_complete
strcpy_align_to_32:
bcc strcpy_align_to_64
ldrb r2, [r1], #1
strb r2, [r0], #1
cbz r2, strcpy_complete
ldrb r2, [r1], #1
strb r2, [r0], #1
cbz r2, strcpy_complete
strcpy_align_to_64:
tst r3, #4
beq strcpy_check_src_align
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
str r2, [r0], #4
strcpy_check_src_align:
// At this point dst is aligned to a double word, check if src
// is also aligned to a double word.
ands r3, r1, #7
bne strcpy_unaligned_copy
.p2align 2
strcpy_mainloop:
ldrd r2, r3, [r1], #8
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_mainloop
strcpy_complete:
m_pop
strcpy_zero_in_first_register:
lsls lr, ip, #17
bne strcpy_copy1byte
bcs strcpy_copy2bytes
lsls ip, ip, #1
bne strcpy_copy3bytes
strcpy_copy4bytes:
// Copy 4 bytes to the destiniation.
str r2, [r0]
m_pop
strcpy_copy1byte:
strb r2, [r0]
m_pop
strcpy_copy2bytes:
strh r2, [r0]
m_pop
strcpy_copy3bytes:
strh r2, [r0], #2
lsr r2, #16
strb r2, [r0]
m_pop
strcpy_zero_in_second_register:
lsls lr, ip, #17
bne strcpy_copy5bytes
bcs strcpy_copy6bytes
lsls ip, ip, #1
bne strcpy_copy7bytes
// Copy 8 bytes to the destination.
strd r2, r3, [r0]
m_pop
strcpy_copy5bytes:
str r2, [r0], #4
strb r3, [r0]
m_pop
strcpy_copy6bytes:
str r2, [r0], #4
strh r3, [r0]
m_pop
strcpy_copy7bytes:
str r2, [r0], #4
strh r3, [r0], #2
lsr r3, #16
strb r3, [r0]
m_pop
strcpy_unaligned_copy:
// Dst is aligned to a double word, while src is at an unknown alignment.
// There are 7 different versions of the unaligned copy code
// to prevent overreading the src. The mainloop of every single version
// will store 64 bits per loop. The difference is how much of src can
// be read without potentially crossing a page boundary.
tbb [pc, r3]
strcpy_unaligned_branchtable:
.byte 0
.byte ((strcpy_unalign7 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign6 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign5 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign4 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign3 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign2 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign1 - strcpy_unaligned_branchtable)/2)
.p2align 2
// Can read 7 bytes before possibly crossing a page.
strcpy_unalign7:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r3, [r1]
cbz r3, strcpy_unalign7_copy5bytes
ldrb r4, [r1, #1]
cbz r4, strcpy_unalign7_copy6bytes
ldrb r5, [r1, #2]
cbz r5, strcpy_unalign7_copy7bytes
ldr r3, [r1], #4
pld [r1, #64]
lsrs ip, r3, #24
strd r2, r3, [r0], #8
beq strcpy_unalign_return
b strcpy_unalign7
strcpy_unalign7_copy5bytes:
str r2, [r0], #4
strb r3, [r0]
strcpy_unalign_return:
m_pop
strcpy_unalign7_copy6bytes:
str r2, [r0], #4
strb r3, [r0], #1
strb r4, [r0], #1
m_pop
strcpy_unalign7_copy7bytes:
str r2, [r0], #4
strb r3, [r0], #1
strb r4, [r0], #1
strb r5, [r0], #1
m_pop
.p2align 2
// Can read 6 bytes before possibly crossing a page.
strcpy_unalign6:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r4, [r1]
cbz r4, strcpy_unalign_copy5bytes
ldrb r5, [r1, #1]
cbz r5, strcpy_unalign_copy6bytes
ldr r3, [r1], #4
pld [r1, #64]
tst r3, #0xff0000
beq strcpy_copy7bytes
lsrs ip, r3, #24
strd r2, r3, [r0], #8
beq strcpy_unalign_return
b strcpy_unalign6
.p2align 2
// Can read 5 bytes before possibly crossing a page.
strcpy_unalign5:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r4, [r1]
cbz r4, strcpy_unalign_copy5bytes
ldr r3, [r1], #4
pld [r1, #64]
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_unalign5
strcpy_unalign_copy5bytes:
str r2, [r0], #4
strb r4, [r0]
m_pop
strcpy_unalign_copy6bytes:
str r2, [r0], #4
strb r4, [r0], #1
strb r5, [r0]
m_pop
.p2align 2
// Can read 4 bytes before possibly crossing a page.
strcpy_unalign4:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldr r3, [r1], #4
pld [r1, #64]
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_unalign4
.p2align 2
// Can read 3 bytes before possibly crossing a page.
strcpy_unalign3:
ldrb r2, [r1]
cbz r2, strcpy_unalign3_copy1byte
ldrb r3, [r1, #1]
cbz r3, strcpy_unalign3_copy2bytes
ldrb r4, [r1, #2]
cbz r4, strcpy_unalign3_copy3bytes
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
lsrs lr, r2, #24
beq strcpy_copy4bytes
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_unalign3
strcpy_unalign3_copy1byte:
strb r2, [r0]
m_pop
strcpy_unalign3_copy2bytes:
strb r2, [r0], #1
strb r3, [r0]
m_pop
strcpy_unalign3_copy3bytes:
strb r2, [r0], #1
strb r3, [r0], #1
strb r4, [r0]
m_pop
.p2align 2
// Can read 2 bytes before possibly crossing a page.
strcpy_unalign2:
ldrb r2, [r1]
cbz r2, strcpy_unalign_copy1byte
ldrb r4, [r1, #1]
cbz r4, strcpy_unalign_copy2bytes
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
tst r2, #0xff0000
beq strcpy_copy3bytes
lsrs ip, r2, #24
beq strcpy_copy4bytes
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_unalign2
.p2align 2
// Can read 1 byte before possibly crossing a page.
strcpy_unalign1:
ldrb r2, [r1]
cbz r2, strcpy_unalign_copy1byte
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_unalign1
strcpy_unalign_copy1byte:
strb r2, [r0]
m_pop
strcpy_unalign_copy2bytes:
strb r2, [r0], #1
strb r4, [r0]
m_pop
.p2align 2
strcat_mainloop:
ldrd r2, r3, [r0], #8
pld [r0, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcat_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcat_zero_in_second_register
b strcat_mainloop
strcat_zero_in_first_register:
// Prefetch the src now, it's going to be used soon.
pld [r1, #0]
lsls lr, ip, #17
bne strcat_sub8
bcs strcat_sub7
lsls ip, ip, #1
bne strcat_sub6
sub r0, r0, #5
b strcat_r0_scan_done
strcat_sub8:
sub r0, r0, #8
b strcat_r0_scan_done
strcat_sub7:
sub r0, r0, #7
b strcat_r0_scan_done
strcat_sub6:
sub r0, r0, #6
b strcat_r0_scan_done
strcat_zero_in_second_register:
// Prefetch the src now, it's going to be used soon.
pld [r1, #0]
lsls lr, ip, #17
bne strcat_sub4
bcs strcat_sub3
lsls ip, ip, #1
bne strcat_sub2
sub r0, r0, #1
b strcat_r0_scan_done
strcat_sub4:
sub r0, r0, #4
b strcat_r0_scan_done
strcat_sub3:
sub r0, r0, #3
b strcat_r0_scan_done
strcat_sub2:
sub r0, r0, #2
b strcat_r0_scan_done
END(strcat)

View File

@@ -1,390 +0,0 @@
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 <machine/cpu-features.h>
#include <machine/asm.h>
#ifdef __ARMEB__
#define S2LOMEM lsl
#define S2LOMEMEQ lsleq
#define S2HIMEM lsr
#define MSB 0x000000ff
#define LSB 0xff000000
#define BYTE0_OFFSET 24
#define BYTE1_OFFSET 16
#define BYTE2_OFFSET 8
#define BYTE3_OFFSET 0
#else /* not __ARMEB__ */
#define S2LOMEM lsr
#define S2LOMEMEQ lsreq
#define S2HIMEM lsl
#define BYTE0_OFFSET 0
#define BYTE1_OFFSET 8
#define BYTE2_OFFSET 16
#define BYTE3_OFFSET 24
#define MSB 0xff000000
#define LSB 0x000000ff
#endif /* not __ARMEB__ */
.syntax unified
#if defined (__thumb__)
.thumb
.thumb_func
#endif
ENTRY(strcmp)
/* Use LDRD whenever possible. */
/* The main thing to look out for when comparing large blocks is that
the loads do not cross a page boundary when loading past the index
of the byte with the first difference or the first string-terminator.
For example, if the strings are identical and the string-terminator
is at index k, byte by byte comparison will not load beyond address
s1+k and s2+k; word by word comparison may load up to 3 bytes beyond
k; double word - up to 7 bytes. If the load of these bytes crosses
a page boundary, it might cause a memory fault (if the page is not mapped)
that would not have happened in byte by byte comparison.
If an address is (double) word aligned, then a load of a (double) word
from that address will not cross a page boundary.
Therefore, the algorithm below considers word and double-word alignment
of strings separately. */
/* High-level description of the algorithm.
* The fast path: if both strings are double-word aligned,
use LDRD to load two words from each string in every loop iteration.
* If the strings have the same offset from a word boundary,
use LDRB to load and compare byte by byte until
the first string is aligned to a word boundary (at most 3 bytes).
This is optimized for quick return on short unaligned strings.
* If the strings have the same offset from a double-word boundary,
use LDRD to load two words from each string in every loop iteration, as in the fast path.
* If the strings do not have the same offset from a double-word boundary,
load a word from the second string before the loop to initialize the queue.
Use LDRD to load two words from every string in every loop iteration.
Inside the loop, load the second word from the second string only after comparing
the first word, using the queued value, to guarantee safety across page boundaries.
* If the strings do not have the same offset from a word boundary,
use LDR and a shift queue. Order of loads and comparisons matters,
similarly to the previous case.
* Use UADD8 and SEL to compare words, and use REV and CLZ to compute the return value.
* The only difference between ARM and Thumb modes is the use of CBZ instruction.
* The only difference between big and little endian is the use of REV in little endian
to compute the return value, instead of MOV.
*/
.macro m_cbz reg label
#ifdef __thumb2__
cbz \reg, \label
#else /* not defined __thumb2__ */
cmp \reg, #0
beq \label
#endif /* not defined __thumb2__ */
.endm /* m_cbz */
.macro m_cbnz reg label
#ifdef __thumb2__
cbnz \reg, \label
#else /* not defined __thumb2__ */
cmp \reg, #0
bne \label
#endif /* not defined __thumb2__ */
.endm /* m_cbnz */
.macro init
/* Macro to save temporary registers and prepare magic values. */
subs sp, sp, #16
.cfi_def_cfa_offset 16
strd r4, r5, [sp, #8]
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
strd r6, r7, [sp]
.cfi_rel_offset r6, 8
.cfi_rel_offset r7, 12
mvn r6, #0 /* all F */
mov r7, #0 /* all 0 */
.endm /* init */
.macro magic_compare_and_branch w1 w2 label
/* Macro to compare registers w1 and w2 and conditionally branch to label. */
cmp \w1, \w2 /* Are w1 and w2 the same? */
magic_find_zero_bytes \w1
it eq
cmpeq ip, #0 /* Is there a zero byte in w1? */
bne \label
.endm /* magic_compare_and_branch */
.macro magic_find_zero_bytes w1
/* Macro to find all-zero bytes in w1, result is in ip. */
#if (defined (__ARM_FEATURE_DSP))
uadd8 ip, \w1, r6
sel ip, r7, r6
#else /* not defined (__ARM_FEATURE_DSP) */
/* __ARM_FEATURE_DSP is not defined for some Cortex-M processors.
Coincidently, these processors only have Thumb-2 mode, where we can use the
the (large) magic constant available directly as an immediate in instructions.
Note that we cannot use the magic constant in ARM mode, where we need
to create the constant in a register. */
sub ip, \w1, #0x01010101
bic ip, ip, \w1
and ip, ip, #0x80808080
#endif /* not defined (__ARM_FEATURE_DSP) */
.endm /* magic_find_zero_bytes */
.macro setup_return w1 w2
#ifdef __ARMEB__
mov r1, \w1
mov r2, \w2
#else /* not __ARMEB__ */
rev r1, \w1
rev r2, \w2
#endif /* not __ARMEB__ */
.endm /* setup_return */
.cfi_startproc
pld [r0, #0]
pld [r1, #0]
/* Are both strings double-word aligned? */
orr ip, r0, r1
tst ip, #7
bne .L_do_align
/* Fast path. */
.save {r4-r7}
init
.L_doubleword_aligned:
/* Get here when the strings to compare are double-word aligned. */
/* Compare two words in every iteration. */
.p2align 2
2:
pld [r0, #16]
pld [r1, #16]
/* Load the next double-word from each string. */
ldrd r2, r3, [r0], #8
ldrd r4, r5, [r1], #8
magic_compare_and_branch w1=r2, w2=r4, label=.L_return_24
magic_compare_and_branch w1=r3, w2=r5, label=.L_return_35
b 2b
.L_do_align:
/* Is the first string word-aligned? */
ands ip, r0, #3
beq .L_word_aligned_r0
/* Fast compare byte by byte until the first string is word-aligned. */
/* The offset of r0 from a word boundary is in ip. Thus, the number of bytes
to read until the next word boundary is 4-ip. */
bic r0, r0, #3
ldr r2, [r0], #4
lsls ip, ip, #31
beq .L_byte2
bcs .L_byte3
.L_byte1:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE1_OFFSET
subs ip, r3, ip
bne .L_fast_return
m_cbz reg=r3, label=.L_fast_return
.L_byte2:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE2_OFFSET
subs ip, r3, ip
bne .L_fast_return
m_cbz reg=r3, label=.L_fast_return
.L_byte3:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE3_OFFSET
subs ip, r3, ip
bne .L_fast_return
m_cbnz reg=r3, label=.L_word_aligned_r0
.L_fast_return:
mov r0, ip
bx lr
.L_word_aligned_r0:
init
/* The first string is word-aligned. */
/* Is the second string word-aligned? */
ands ip, r1, #3
bne .L_strcmp_unaligned
.L_word_aligned:
/* The strings are word-aligned. */
/* Is the first string double-word aligned? */
tst r0, #4
beq .L_doubleword_aligned_r0
/* If r0 is not double-word aligned yet, align it by loading
and comparing the next word from each string. */
ldr r2, [r0], #4
ldr r4, [r1], #4
magic_compare_and_branch w1=r2 w2=r4 label=.L_return_24
.L_doubleword_aligned_r0:
/* Get here when r0 is double-word aligned. */
/* Is r1 doubleword_aligned? */
tst r1, #4
beq .L_doubleword_aligned
/* Get here when the strings to compare are word-aligned,
r0 is double-word aligned, but r1 is not double-word aligned. */
/* Initialize the queue. */
ldr r5, [r1], #4
/* Compare two words in every iteration. */
.p2align 2
3:
pld [r0, #16]
pld [r1, #16]
/* Load the next double-word from each string and compare. */
ldrd r2, r3, [r0], #8
magic_compare_and_branch w1=r2 w2=r5 label=.L_return_25
ldrd r4, r5, [r1], #8
magic_compare_and_branch w1=r3 w2=r4 label=.L_return_34
b 3b
.macro miscmp_word offsetlo offsethi
/* Macro to compare misaligned strings. */
/* r0, r1 are word-aligned, and at least one of the strings
is not double-word aligned. */
/* Compare one word in every loop iteration. */
/* OFFSETLO is the original bit-offset of r1 from a word-boundary,
OFFSETHI is 32 - OFFSETLO (i.e., offset from the next word). */
/* Initialize the shift queue. */
ldr r5, [r1], #4
/* Compare one word from each string in every loop iteration. */
.p2align 2
7:
ldr r3, [r0], #4
S2LOMEM r5, r5, #\offsetlo
magic_find_zero_bytes w1=r3
cmp r7, ip, S2HIMEM #\offsetlo
and r2, r3, r6, S2LOMEM #\offsetlo
it eq
cmpeq r2, r5
bne .L_return_25
ldr r5, [r1], #4
cmp ip, #0
eor r3, r2, r3
S2HIMEM r2, r5, #\offsethi
it eq
cmpeq r3, r2
bne .L_return_32
b 7b
.endm /* miscmp_word */
.L_strcmp_unaligned:
/* r0 is word-aligned, r1 is at offset ip from a word. */
/* Align r1 to the (previous) word-boundary. */
bic r1, r1, #3
/* Unaligned comparison word by word using LDRs. */
cmp ip, #2
beq .L_miscmp_word_16 /* If ip == 2. */
bge .L_miscmp_word_24 /* If ip == 3. */
miscmp_word offsetlo=8 offsethi=24 /* If ip == 1. */
.L_miscmp_word_16: miscmp_word offsetlo=16 offsethi=16
.L_miscmp_word_24: miscmp_word offsetlo=24 offsethi=8
.L_return_32:
setup_return w1=r3, w2=r2
b .L_do_return
.L_return_34:
setup_return w1=r3, w2=r4
b .L_do_return
.L_return_25:
setup_return w1=r2, w2=r5
b .L_do_return
.L_return_35:
setup_return w1=r3, w2=r5
b .L_do_return
.L_return_24:
setup_return w1=r2, w2=r4
.L_do_return:
#ifdef __ARMEB__
mov r0, ip
#else /* not __ARMEB__ */
rev r0, ip
#endif /* not __ARMEB__ */
/* Restore temporaries early, before computing the return value. */
ldrd r6, r7, [sp]
ldrd r4, r5, [sp, #8]
adds sp, sp, #16
.cfi_def_cfa_offset 0
.cfi_restore r4
.cfi_restore r5
.cfi_restore r6
.cfi_restore r7
/* There is a zero or a different byte between r1 and r2. */
/* r0 contains a mask of all-zero bytes in r1. */
/* Using r0 and not ip here because cbz requires low register. */
m_cbz reg=r0, label=.L_compute_return_value
clz r0, r0
/* r0 contains the number of bits on the left of the first all-zero byte in r1. */
rsb r0, r0, #24
/* Here, r0 contains the number of bits on the right of the first all-zero byte in r1. */
lsr r1, r1, r0
lsr r2, r2, r0
.L_compute_return_value:
movs r0, #1
cmp r1, r2
/* The return value is computed as follows.
If r1>r2 then (C==1 and Z==0) and LS doesn't hold and r0 is #1 at return.
If r1<r2 then (C==0 and Z==0) and we execute SBC with carry_in=0,
which means r0:=r0-r0-1 and r0 is #-1 at return.
If r1=r2 then (C==1 and Z==1) and we execute SBC with carry_in=1,
which means r0:=r0-r0 and r0 is #0 at return.
(C==0 and Z==1) cannot happen because the carry bit is "not borrow". */
it ls
sbcls r0, r0, r0
bx lr
.cfi_endproc
END(strcmp)

View File

@@ -1,451 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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.
*/
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 <machine/asm.h>
.syntax unified
.thumb
.thumb_func
.macro m_push
push {r0, r4, r5, lr}
.endm // m_push
.macro m_pop
pop {r0, r4, r5, pc}
.endm // m_pop
.macro m_copy_byte reg, cmd, label
ldrb \reg, [r1], #1
strb \reg, [r0], #1
\cmd \reg, \label
.endm // m_copy_byte
ENTRY(strcpy)
// For short copies, hard-code checking the first 8 bytes since this
// new code doesn't win until after about 8 bytes.
m_push
m_copy_byte reg=r2, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r3, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r4, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r5, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r2, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r3, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r4, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r5, cmd=cbnz, label=strcpy_continue
strcpy_finish:
m_pop
strcpy_continue:
pld [r1, #0]
ands r3, r0, #7
beq strcpy_check_src_align
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq strcpy_align_to_32
ldrb r2, [r1], #1
strb r2, [r0], #1
cbz r2, strcpy_complete
strcpy_align_to_32:
bcc strcpy_align_to_64
ldrb r2, [r1], #1
strb r2, [r0], #1
cbz r2, strcpy_complete
ldrb r2, [r1], #1
strb r2, [r0], #1
cbz r2, strcpy_complete
strcpy_align_to_64:
tst r3, #4
beq strcpy_check_src_align
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
str r2, [r0], #4
strcpy_check_src_align:
// At this point dst is aligned to a double word, check if src
// is also aligned to a double word.
ands r3, r1, #7
bne strcpy_unaligned_copy
.p2align 2
strcpy_mainloop:
ldrd r2, r3, [r1], #8
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_mainloop
strcpy_complete:
m_pop
strcpy_zero_in_first_register:
lsls lr, ip, #17
bne strcpy_copy1byte
bcs strcpy_copy2bytes
lsls ip, ip, #1
bne strcpy_copy3bytes
strcpy_copy4bytes:
// Copy 4 bytes to the destiniation.
str r2, [r0]
m_pop
strcpy_copy1byte:
strb r2, [r0]
m_pop
strcpy_copy2bytes:
strh r2, [r0]
m_pop
strcpy_copy3bytes:
strh r2, [r0], #2
lsr r2, #16
strb r2, [r0]
m_pop
strcpy_zero_in_second_register:
lsls lr, ip, #17
bne strcpy_copy5bytes
bcs strcpy_copy6bytes
lsls ip, ip, #1
bne strcpy_copy7bytes
// Copy 8 bytes to the destination.
strd r2, r3, [r0]
m_pop
strcpy_copy5bytes:
str r2, [r0], #4
strb r3, [r0]
m_pop
strcpy_copy6bytes:
str r2, [r0], #4
strh r3, [r0]
m_pop
strcpy_copy7bytes:
str r2, [r0], #4
strh r3, [r0], #2
lsr r3, #16
strb r3, [r0]
m_pop
strcpy_unaligned_copy:
// Dst is aligned to a double word, while src is at an unknown alignment.
// There are 7 different versions of the unaligned copy code
// to prevent overreading the src. The mainloop of every single version
// will store 64 bits per loop. The difference is how much of src can
// be read without potentially crossing a page boundary.
tbb [pc, r3]
strcpy_unaligned_branchtable:
.byte 0
.byte ((strcpy_unalign7 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign6 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign5 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign4 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign3 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign2 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign1 - strcpy_unaligned_branchtable)/2)
.p2align 2
// Can read 7 bytes before possibly crossing a page.
strcpy_unalign7:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r3, [r1]
cbz r3, strcpy_unalign7_copy5bytes
ldrb r4, [r1, #1]
cbz r4, strcpy_unalign7_copy6bytes
ldrb r5, [r1, #2]
cbz r5, strcpy_unalign7_copy7bytes
ldr r3, [r1], #4
pld [r1, #64]
lsrs ip, r3, #24
strd r2, r3, [r0], #8
beq strcpy_unalign_return
b strcpy_unalign7
strcpy_unalign7_copy5bytes:
str r2, [r0], #4
strb r3, [r0]
strcpy_unalign_return:
m_pop
strcpy_unalign7_copy6bytes:
str r2, [r0], #4
strb r3, [r0], #1
strb r4, [r0], #1
m_pop
strcpy_unalign7_copy7bytes:
str r2, [r0], #4
strb r3, [r0], #1
strb r4, [r0], #1
strb r5, [r0], #1
m_pop
.p2align 2
// Can read 6 bytes before possibly crossing a page.
strcpy_unalign6:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r4, [r1]
cbz r4, strcpy_unalign_copy5bytes
ldrb r5, [r1, #1]
cbz r5, strcpy_unalign_copy6bytes
ldr r3, [r1], #4
pld [r1, #64]
tst r3, #0xff0000
beq strcpy_copy7bytes
lsrs ip, r3, #24
strd r2, r3, [r0], #8
beq strcpy_unalign_return
b strcpy_unalign6
.p2align 2
// Can read 5 bytes before possibly crossing a page.
strcpy_unalign5:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r4, [r1]
cbz r4, strcpy_unalign_copy5bytes
ldr r3, [r1], #4
pld [r1, #64]
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_unalign5
strcpy_unalign_copy5bytes:
str r2, [r0], #4
strb r4, [r0]
m_pop
strcpy_unalign_copy6bytes:
str r2, [r0], #4
strb r4, [r0], #1
strb r5, [r0]
m_pop
.p2align 2
// Can read 4 bytes before possibly crossing a page.
strcpy_unalign4:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldr r3, [r1], #4
pld [r1, #64]
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_unalign4
.p2align 2
// Can read 3 bytes before possibly crossing a page.
strcpy_unalign3:
ldrb r2, [r1]
cbz r2, strcpy_unalign3_copy1byte
ldrb r3, [r1, #1]
cbz r3, strcpy_unalign3_copy2bytes
ldrb r4, [r1, #2]
cbz r4, strcpy_unalign3_copy3bytes
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
lsrs lr, r2, #24
beq strcpy_copy4bytes
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_unalign3
strcpy_unalign3_copy1byte:
strb r2, [r0]
m_pop
strcpy_unalign3_copy2bytes:
strb r2, [r0], #1
strb r3, [r0]
m_pop
strcpy_unalign3_copy3bytes:
strb r2, [r0], #1
strb r3, [r0], #1
strb r4, [r0]
m_pop
.p2align 2
// Can read 2 bytes before possibly crossing a page.
strcpy_unalign2:
ldrb r2, [r1]
cbz r2, strcpy_unalign_copy1byte
ldrb r4, [r1, #1]
cbz r4, strcpy_unalign_copy2bytes
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
tst r2, #0xff0000
beq strcpy_copy3bytes
lsrs ip, r2, #24
beq strcpy_copy4bytes
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_unalign2
.p2align 2
// Can read 1 byte before possibly crossing a page.
strcpy_unalign1:
ldrb r2, [r1]
cbz r2, strcpy_unalign_copy1byte
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
strd r2, r3, [r0], #8
b strcpy_unalign1
strcpy_unalign_copy1byte:
strb r2, [r0]
m_pop
strcpy_unalign_copy2bytes:
strb r2, [r0], #1
strb r4, [r0]
m_pop
END(strcpy)

View File

@@ -1,165 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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.
*/
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 <machine/asm.h>
.syntax unified
.thumb
.thumb_func
ENTRY(strlen)
pld [r0, #0]
mov r1, r0
ands r3, r0, #7
beq mainloop
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq align_to_32
ldrb r2, [r1], #1
cbz r2, update_count_and_return
align_to_32:
bcc align_to_64
ands ip, r3, #2
beq align_to_64
ldrb r2, [r1], #1
cbz r2, update_count_and_return
ldrb r2, [r1], #1
cbz r2, update_count_and_return
align_to_64:
tst r3, #4
beq mainloop
ldr r3, [r1], #4
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne zero_in_second_register
.p2align 2
mainloop:
ldrd r2, r3, [r1], #8
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne zero_in_second_register
b mainloop
update_count_and_return:
sub r0, r1, r0
sub r0, r0, #1
bx lr
zero_in_first_register:
sub r0, r1, r0
lsls r3, ip, #17
bne sub8_and_return
bcs sub7_and_return
lsls ip, ip, #1
bne sub6_and_return
sub r0, r0, #5
bx lr
sub8_and_return:
sub r0, r0, #8
bx lr
sub7_and_return:
sub r0, r0, #7
bx lr
sub6_and_return:
sub r0, r0, #6
bx lr
zero_in_second_register:
sub r0, r1, r0
lsls r3, ip, #17
bne sub4_and_return
bcs sub3_and_return
lsls ip, ip, #1
bne sub2_and_return
sub r0, r0, #1
bx lr
sub4_and_return:
sub r0, r0, #4
bx lr
sub3_and_return:
sub r0, r0, #3
bx lr
sub2_and_return:
sub r0, r0, #2
bx lr
END(strlen)

View File

@@ -1,10 +0,0 @@
$(call libc-add-cpu-variant-src,MEMCPY,arch-arm/cortex-a15/bionic/memcpy.S)
$(call libc-add-cpu-variant-src,MEMSET,arch-arm/cortex-a15/bionic/memset.S)
$(call libc-add-cpu-variant-src,STRCAT,arch-arm/cortex-a15/bionic/strcat.S)
$(call libc-add-cpu-variant-src,STRCMP,arch-arm/cortex-a15/bionic/strcmp.S)
$(call libc-add-cpu-variant-src,STRCPY,arch-arm/cortex-a15/bionic/strcpy.S)
$(call libc-add-cpu-variant-src,STRLEN,arch-arm/cortex-a15/bionic/strlen.S)
$(call libc-add-cpu-variant-src,__STRCAT_CHK,arch-arm/cortex-a15/bionic/__strcat_chk.S)
$(call libc-add-cpu-variant-src,__STRCPY_CHK,arch-arm/cortex-a15/bionic/__strcpy_chk.S)
include bionic/libc/arch-arm/generic/generic.mk

View File

@@ -1 +0,0 @@
include bionic/libc/arch-arm/cortex-a15/cortex-a15.mk

View File

@@ -1 +0,0 @@
include bionic/libc/arch-arm/cortex-a15/cortex-a15.mk

View File

@@ -1,230 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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 <machine/asm.h>
#include "libc_events.h"
.syntax unified
.fpu neon
.thumb
.thumb_func
// Get the length of src string, then get the source of the dst string.
// Check that the two lengths together don't exceed the threshold, then
// do a memcpy of the data.
ENTRY(__strcat_chk)
.cfi_startproc
pld [r0, #0]
push {r0, lr}
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
push {r4, r5}
.save {r4, r5}
.cfi_adjust_cfa_offset 8
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
mov lr, r2
// Save the dst register to r5
mov r5, r0
// Zero out r4
eor r4, r4, r4
// r1 contains the address of the string to count.
.L_strlen_start:
mov r0, r1
ands r3, r0, #7
bne .L_align_src
.p2align 2
.L_mainloop:
ldmia r1!, {r2, r3}
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne .L_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
b .L_mainloop
.L_zero_in_first_register:
sub r3, r1, r0
// Check for zero in byte 0.
lsls r2, ip, #17
beq .L_check_byte1_reg1
sub r3, r3, #8
b .L_finish
.L_check_byte1_reg1:
bcc .L_check_byte2_reg1
sub r3, r3, #7
b .L_finish
.L_check_byte2_reg1:
// Check for zero in byte 2.
tst ip, #0x800000
it ne
subne r3, r3, #6
bne .L_finish
sub r3, r3, #5
b .L_finish
.L_zero_in_second_register:
sub r3, r1, r0
// Check for zero in byte 0.
lsls r2, ip, #17
beq .L_check_byte1_reg2
sub r3, r3, #4
b .L_finish
.L_check_byte1_reg2:
bcc .L_check_byte2_reg2
sub r3, r3, #3
b .L_finish
.L_check_byte2_reg2:
// Check for zero in byte 2.
tst ip, #0x800000
it ne
subne r3, r3, #2
bne .L_finish
sub r3, r3, #1
b .L_finish
.L_align_src:
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq .L_align_to_32
ldrb r2, [r1], #1
cbz r2, .L_done
.L_align_to_32:
bcc .L_align_to_64
ldrb r2, [r1], #1
cbz r2, .L_done
ldrb r2, [r1], #1
cbz r2, .L_done
.L_align_to_64:
tst r3, #4
beq .L_mainloop
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
b .L_mainloop
.L_done:
sub r3, r1, r0
sub r3, r3, #1
.L_finish:
cmp r4, #0
bne .L_strlen_done
// Time to get the dst string length.
mov r1, r5
// Save the original source address to r5.
mov r5, r0
// Save the current length (adding 1 for the terminator).
add r4, r3, #1
b .L_strlen_start
// r0 holds the pointer to the dst string.
// r3 holds the dst string length.
// r4 holds the src string length + 1.
.L_strlen_done:
add r2, r3, r4
cmp r2, lr
bhi __strcat_chk_fail
// Set up the registers for the memcpy code.
mov r1, r5
pld [r1, #64]
mov r2, r4
add r0, r0, r3
pop {r4, r5}
// Fall through into the memcpy_base function.
.cfi_endproc
END(__strcat_chk)
#define MEMCPY_BASE __strcat_chk_memcpy_base
#define MEMCPY_BASE_ALIGNED __strcat_chk_memcpy_base_aligned
#include "memcpy_base.S"
ENTRY(__strcat_chk_fail)
.cfi_startproc
.save {r0, lr}
.save {r4, r5}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
.cfi_adjust_cfa_offset 8
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_STRCAT_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+4)
.cfi_endproc
END(__strcat_chk_fail)
.data
error_string:
.string "strcat buffer overflow"

View File

@@ -1,194 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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 <machine/asm.h>
#include "libc_events.h"
.syntax unified
.fpu neon
.thumb
.thumb_func
// Get the length of the source string first, then do a memcpy of the data
// instead of a strcpy.
ENTRY(__strcpy_chk)
.cfi_startproc
pld [r0, #0]
push {r0, lr}
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
mov lr, r2
mov r0, r1
ands r3, r0, #7
bne .L_align_src
.p2align 2
.L_mainloop:
ldmia r0!, {r2, r3}
pld [r0, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne .L_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
b .L_mainloop
.L_zero_in_first_register:
sub r3, r0, r1
// Check for zero in byte 0.
lsls r2, ip, #17
beq .L_check_byte1_reg1
sub r3, r3, #8
b .L_check_size
.L_check_byte1_reg1:
bcc .L_check_byte2_reg1
sub r3, r3, #7
b .L_check_size
.L_check_byte2_reg1:
// Check for zero in byte 2.
tst ip, #0x800000
it ne
subne r3, r3, #6
bne .L_check_size
sub r3, r3, #5
b .L_check_size
.L_zero_in_second_register:
sub r3, r0, r1
// Check for zero in byte 0.
lsls r2, ip, #17
beq .L_check_byte1_reg2
sub r3, r3, #4
b .L_check_size
.L_check_byte1_reg2:
bcc .L_check_byte2_reg2
sub r3, r3, #3
b .L_check_size
.L_check_byte2_reg2:
// Check for zero in byte 2.
tst ip, #0x800000
it ne
subne r3, r3, #2
bne .L_check_size
sub r3, r3, #1
b .L_check_size
.L_align_src:
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq .L_align_to_32
ldrb r2, [r0], #1
cbz r2, .L_done
.L_align_to_32:
bcc .L_align_to_64
ldrb r2, [r0], #1
cbz r2, .L_done
ldrb r2, [r0], #1
cbz r2, .L_done
.L_align_to_64:
tst r3, #4
beq .L_mainloop
ldr r2, [r0], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
b .L_mainloop
.L_done:
sub r3, r0, r1
sub r3, r3, #1
.L_check_size:
pld [r1, #0]
pld [r1, #64]
ldr r0, [sp]
cmp r3, lr
bhs __strcpy_chk_fail
// Add 1 for copy length to get the string terminator.
add r2, r3, #1
.cfi_endproc
// Fall through into the memcpy_base function.
END(__strcpy_chk)
#define MEMCPY_BASE __strcpy_chk_memcpy_base
#define MEMCPY_BASE_ALIGNED __strcpy_chk_memcpy_base_aligned
#include "memcpy_base.S"
ENTRY(__strcpy_chk_fail)
.cfi_startproc
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_STRCPY_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+4)
.cfi_endproc
END(__strcpy_chk_fail)
.data
error_string:
.string "strcpy buffer overflow"

View File

@@ -1,92 +0,0 @@
/*
* 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:
* * 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.
*
* 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 <machine/asm.h>
#include "libc_events.h"
/*
* This code assumes it is running on a processor that supports all arm v7
* instructions, that supports neon instructions, and that has a 32 byte
* cache line.
*/
.syntax unified
.fpu neon
.thumb
.thumb_func
ENTRY(__memcpy_chk)
.cfi_startproc
cmp r2, r3
bhi __memcpy_chk_fail
// Fall through to memcpy...
.cfi_endproc
END(__memcpy_chk)
ENTRY(memcpy)
.cfi_startproc
pld [r1, #0]
stmfd sp!, {r0, lr}
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
pld [r1, #64]
.cfi_endproc
END(memcpy)
#define MEMCPY_BASE __memcpy_base
#define MEMCPY_BASE_ALIGNED __memcpy_base_aligned
#include "memcpy_base.S"
ENTRY(__memcpy_chk_fail)
.cfi_startproc
// Preserve lr for backtrace.
push {lr}
.save {lr}
.cfi_def_cfa_offset 4
.cfi_rel_offset lr, 0
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_MEMCPY_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+4)
.cfi_endproc
END(__memcpy_chk_fail)
.data
error_string:
.string "memcpy buffer overflow"

View File

@@ -1,233 +0,0 @@
/*
* 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:
* * 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.
*
* 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.
*/
/*
* This code assumes it is running on a processor that supports all arm v7
* instructions, that supports neon instructions, and that has a 32 byte
* cache line.
*/
ENTRY(MEMCPY_BASE)
.cfi_startproc
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
// Check so divider is at least 16 bytes, needed for alignment code.
cmp r2, #16
blo 5f
/* check if buffers are aligned. If so, run arm-only version */
eor r3, r0, r1
ands r3, r3, #0x3
beq __memcpy_base_aligned
/* Check the upper size limit for Neon unaligned memory access in memcpy */
cmp r2, #224
blo 3f
/* align destination to 16 bytes for the write-buffer */
rsb r3, r0, #0
ands r3, r3, #0xF
beq 3f
/* copy up to 15-bytes (count in r3) */
sub r2, r2, r3
movs ip, r3, lsl #31
itt mi
ldrbmi lr, [r1], #1
strbmi lr, [r0], #1
itttt cs
ldrbcs ip, [r1], #1
ldrbcs lr, [r1], #1
strbcs ip, [r0], #1
strbcs lr, [r0], #1
movs ip, r3, lsl #29
bge 1f
// copies 4 bytes, destination 32-bits aligned
vld1.32 {d0[0]}, [r1]!
vst1.32 {d0[0]}, [r0, :32]!
1: bcc 2f
// copies 8 bytes, destination 64-bits aligned
vld1.8 {d0}, [r1]!
vst1.8 {d0}, [r0, :64]!
2:
/* preload immediately the next cache line, which we may need */
pld [r1, #0]
pld [r1, #(32 * 2)]
3:
/* make sure we have at least 64 bytes to copy */
subs r2, r2, #64
blo 2f
/* preload all the cache lines we need */
pld [r1, #(32 * 4)]
pld [r1, #(32 * 6)]
1: /* The main loop copies 64 bytes at a time */
vld1.8 {d0 - d3}, [r1]!
vld1.8 {d4 - d7}, [r1]!
pld [r1, #(32 * 6)]
subs r2, r2, #64
vst1.8 {d0 - d3}, [r0]!
vst1.8 {d4 - d7}, [r0]!
bhs 1b
2: /* fix-up the remaining count and make sure we have >= 32 bytes left */
add r2, r2, #64
subs r2, r2, #32
blo 4f
3: /* 32 bytes at a time. These cache lines were already preloaded */
vld1.8 {d0 - d3}, [r1]!
subs r2, r2, #32
vst1.8 {d0 - d3}, [r0]!
bhs 3b
4: /* less than 32 left */
add r2, r2, #32
tst r2, #0x10
beq 5f
// copies 16 bytes, 128-bits aligned
vld1.8 {d0, d1}, [r1]!
vst1.8 {d0, d1}, [r0]!
5: /* copy up to 15-bytes (count in r2) */
movs ip, r2, lsl #29
bcc 1f
vld1.8 {d0}, [r1]!
vst1.8 {d0}, [r0]!
1: bge 2f
vld1.32 {d0[0]}, [r1]!
vst1.32 {d0[0]}, [r0]!
2: movs ip, r2, lsl #31
itt mi
ldrbmi r3, [r1], #1
strbmi r3, [r0], #1
itttt cs
ldrbcs ip, [r1], #1
ldrbcs lr, [r1], #1
strbcs ip, [r0], #1
strbcs lr, [r0], #1
ldmfd sp!, {r0, lr}
bx lr
.cfi_endproc
END(MEMCPY_BASE)
ENTRY(MEMCPY_BASE_ALIGNED)
.cfi_startproc
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
/* Simple arm-only copy loop to handle aligned copy operations */
stmfd sp!, {r4-r8}
.save {r4-r8}
.cfi_adjust_cfa_offset 20
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
.cfi_rel_offset r6, 8
.cfi_rel_offset r7, 12
.cfi_rel_offset r8, 16
pld [r1, #(32 * 4)]
/* Check alignment */
rsb r3, r1, #0
ands r3, #3
beq 2f
/* align source to 32 bits. We need to insert 2 instructions between
* a ldr[b|h] and str[b|h] because byte and half-word instructions
* stall 2 cycles.
*/
movs r12, r3, lsl #31
sub r2, r2, r3 /* we know that r3 <= r2 because r2 >= 4 */
itt mi
ldrbmi r3, [r1], #1
strbmi r3, [r0], #1
itttt cs
ldrbcs r4, [r1], #1
ldrbcs r5, [r1], #1
strbcs r4, [r0], #1
strbcs r5, [r0], #1
2:
subs r2, r2, #64
blt 4f
3: /* Main copy loop, copying 64 bytes at a time */
pld [r1, #(32 * 8)]
ldmia r1!, {r3, r4, r5, r6, r7, r8, r12, lr}
stmia r0!, {r3, r4, r5, r6, r7, r8, r12, lr}
ldmia r1!, {r3, r4, r5, r6, r7, r8, r12, lr}
stmia r0!, {r3, r4, r5, r6, r7, r8, r12, lr}
subs r2, r2, #64
bge 3b
4: /* Check if there are > 32 bytes left */
adds r2, r2, #64
subs r2, r2, #32
blt 5f
/* Copy 32 bytes */
ldmia r1!, {r3, r4, r5, r6, r7, r8, r12, lr}
stmia r0!, {r3, r4, r5, r6, r7, r8, r12, lr}
subs r2, #32
5: /* Handle any remaining bytes */
adds r2, #32
beq 6f
movs r12, r2, lsl #28
itt cs
ldmiacs r1!, {r3, r4, r5, r6} /* 16 bytes */
stmiacs r0!, {r3, r4, r5, r6}
itt mi
ldmiami r1!, {r7, r8} /* 8 bytes */
stmiami r0!, {r7, r8}
movs r12, r2, lsl #30
itt cs
ldrcs r3, [r1], #4 /* 4 bytes */
strcs r3, [r0], #4
itt mi
ldrhmi r4, [r1], #2 /* 2 bytes */
strhmi r4, [r0], #2
tst r2, #0x1
itt ne
ldrbne r3, [r1] /* last byte */
strbne r3, [r0]
6:
ldmfd sp!, {r4-r8}
ldmfd sp!, {r0, pc}
.cfi_endproc
END(MEMCPY_BASE_ALIGNED)

View File

@@ -1,204 +0,0 @@
/*
* 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:
* * 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.
*
* 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 <machine/cpu-features.h>
#include <machine/asm.h>
#include "libc_events.h"
/*
* This code assumes it is running on a processor that supports all arm v7
* instructions and that supports neon instructions.
*/
.fpu neon
ENTRY(__memset_chk)
.cfi_startproc
cmp r2, r3
bls .L_done
// Preserve lr for backtrace.
push {lr}
.save {lr}
.cfi_def_cfa_offset 4
.cfi_rel_offset lr, 0
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_MEMSET_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+8)
.cfi_endproc
END(__memset_chk)
ENTRY(bzero)
.cfi_startproc
mov r2, r1
mov r1, #0
.L_done:
// Fall through to memset...
.cfi_endproc
END(bzero)
/* memset() returns its first argument. */
ENTRY(memset)
.cfi_startproc
# The neon memset only wins for less than 132.
cmp r2, #132
bhi __memset_large_copy
stmfd sp!, {r0}
.save {r0}
.cfi_def_cfa_offset 4
.cfi_rel_offset r0, 0
vdup.8 q0, r1
/* make sure we have at least 32 bytes to write */
subs r2, r2, #32
blo 2f
vmov q1, q0
1: /* The main loop writes 32 bytes at a time */
subs r2, r2, #32
vst1.8 {d0 - d3}, [r0]!
bhs 1b
2: /* less than 32 left */
add r2, r2, #32
tst r2, #0x10
beq 3f
// writes 16 bytes, 128-bits aligned
vst1.8 {d0, d1}, [r0]!
3: /* write up to 15-bytes (count in r2) */
movs ip, r2, lsl #29
bcc 1f
vst1.8 {d0}, [r0]!
1: bge 2f
vst1.32 {d0[0]}, [r0]!
2: movs ip, r2, lsl #31
strmib r1, [r0], #1
strcsb r1, [r0], #1
strcsb r1, [r0], #1
ldmfd sp!, {r0}
bx lr
.cfi_endproc
END(memset)
ENTRY(__memset_large_copy)
.cfi_startproc
/* compute the offset to align the destination
* offset = (4-(src&3))&3 = -src & 3
*/
stmfd sp!, {r0, r4-r7, lr}
.save {r0, r4-r7, lr}
.cfi_def_cfa_offset 24
.cfi_rel_offset r0, 0
.cfi_rel_offset r4, 4
.cfi_rel_offset r5, 8
.cfi_rel_offset r6, 12
.cfi_rel_offset r7, 16
.cfi_rel_offset lr, 20
rsb r3, r0, #0
ands r3, r3, #3
cmp r3, r2
movhi r3, r2
/* splat r1 */
mov r1, r1, lsl #24
orr r1, r1, r1, lsr #8
orr r1, r1, r1, lsr #16
movs r12, r3, lsl #31
strcsb r1, [r0], #1 /* can't use strh (alignment unknown) */
strcsb r1, [r0], #1
strmib r1, [r0], #1
subs r2, r2, r3
ldmlsfd sp!, {r0, r4-r7, lr} /* return */
bxls lr
/* align the destination to a cache-line */
mov r12, r1
mov lr, r1
mov r4, r1
mov r5, r1
mov r6, r1
mov r7, r1
rsb r3, r0, #0
ands r3, r3, #0x1C
beq 3f
cmp r3, r2
andhi r3, r2, #0x1C
sub r2, r2, r3
/* conditionally writes 0 to 7 words (length in r3) */
movs r3, r3, lsl #28
stmcsia r0!, {r1, lr}
stmcsia r0!, {r1, lr}
stmmiia r0!, {r1, lr}
movs r3, r3, lsl #2
strcs r1, [r0], #4
3:
subs r2, r2, #32
mov r3, r1
bmi 2f
1: subs r2, r2, #32
stmia r0!, {r1,r3,r4,r5,r6,r7,r12,lr}
bhs 1b
2: add r2, r2, #32
/* conditionally stores 0 to 31 bytes */
movs r2, r2, lsl #28
stmcsia r0!, {r1,r3,r12,lr}
stmmiia r0!, {r1, lr}
movs r2, r2, lsl #2
strcs r1, [r0], #4
strmih r1, [r0], #2
movs r2, r2, lsl #2
strcsb r1, [r0]
ldmfd sp!, {r0, r4-r7, lr}
bx lr
.cfi_endproc
END(__memset_large_copy)
.data
error_string:
.string "memset buffer overflow"

View File

@@ -1,548 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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.
*/
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 <machine/asm.h>
.syntax unified
.thumb
.thumb_func
.macro m_push
push {r0, r4, r5, lr}
.endm // m_push
.macro m_ret inst
\inst {r0, r4, r5, pc}
.endm // m_ret
.macro m_scan_byte
ldrb r3, [r0]
cbz r3, strcat_r0_scan_done
add r0, #1
.endm // m_scan_byte
.macro m_copy_byte reg, cmd, label
ldrb \reg, [r1], #1
strb \reg, [r0], #1
\cmd \reg, \label
.endm // m_copy_byte
ENTRY(strcat)
// Quick check to see if src is empty.
ldrb r2, [r1]
pld [r1, #0]
cbnz r2, strcat_continue
bx lr
strcat_continue:
// To speed up really small dst strings, unroll checking the first 4 bytes.
m_push
m_scan_byte
m_scan_byte
m_scan_byte
m_scan_byte
ands r3, r0, #7
bne strcat_align_src
.p2align 2
strcat_mainloop:
ldmia r0!, {r2, r3}
pld [r0, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcat_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcat_zero_in_second_register
b strcat_mainloop
strcat_zero_in_first_register:
sub r0, r0, #4
strcat_zero_in_second_register:
// Check for zero in byte 0.
tst ip, #0x80
it ne
subne r0, r0, #4
bne strcat_r0_scan_done
// Check for zero in byte 1.
tst ip, #0x8000
it ne
subne r0, r0, #3
bne strcat_r0_scan_done
// Check for zero in byte 2.
tst ip, #0x800000
it ne
subne r0, r0, #2
it eq
// Zero is in byte 3.
subeq r0, r0, #1
strcat_r0_scan_done:
// Unroll the first 8 bytes that will be copied.
m_copy_byte reg=r2, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r3, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r4, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r5, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r2, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r3, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r4, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r5, cmd=cbnz, label=strcpy_continue
strcpy_finish:
m_ret inst=pop
strcpy_continue:
pld [r1, #0]
ands r3, r0, #7
bne strcpy_align_dst
strcpy_check_src_align:
// At this point dst is aligned to a double word, check if src
// is also aligned to a double word.
ands r3, r1, #7
bne strcpy_unaligned_copy
.p2align 2
strcpy_mainloop:
ldmia r1!, {r2, r3}
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_mainloop
strcpy_zero_in_first_register:
lsls lr, ip, #17
itt ne
strbne r2, [r0]
m_ret inst=popne
itt cs
strhcs r2, [r0]
m_ret inst=popcs
lsls ip, ip, #1
itt eq
streq r2, [r0]
m_ret inst=popeq
strh r2, [r0], #2
lsr r3, r2, #16
strb r3, [r0]
m_ret inst=pop
strcpy_zero_in_second_register:
lsls lr, ip, #17
ittt ne
stmiane r0!, {r2}
strbne r3, [r0]
m_ret inst=popne
ittt cs
strcs r2, [r0], #4
strhcs r3, [r0]
m_ret inst=popcs
lsls ip, ip, #1
itt eq
stmiaeq r0, {r2, r3}
m_ret inst=popeq
stmia r0!, {r2}
strh r3, [r0], #2
lsr r4, r3, #16
strb r4, [r0]
m_ret inst=pop
strcpy_align_dst:
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq strcpy_align_to_32
ldrb r2, [r1], #1
strb r2, [r0], #1
cbz r2, strcpy_complete
strcpy_align_to_32:
bcc strcpy_align_to_64
ldrb r4, [r1], #1
strb r4, [r0], #1
cmp r4, #0
it eq
m_ret inst=popeq
ldrb r5, [r1], #1
strb r5, [r0], #1
cmp r5, #0
it eq
m_ret inst=popeq
strcpy_align_to_64:
tst r3, #4
beq strcpy_check_src_align
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
stmia r0!, {r2}
b strcpy_check_src_align
strcpy_complete:
m_ret inst=pop
strcpy_unaligned_copy:
// Dst is aligned to a double word, while src is at an unknown alignment.
// There are 7 different versions of the unaligned copy code
// to prevent overreading the src. The mainloop of every single version
// will store 64 bits per loop. The difference is how much of src can
// be read without potentially crossing a page boundary.
tbb [pc, r3]
strcpy_unaligned_branchtable:
.byte 0
.byte ((strcpy_unalign7 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign6 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign5 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign4 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign3 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign2 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign1 - strcpy_unaligned_branchtable)/2)
.p2align 2
// Can read 7 bytes before possibly crossing a page.
strcpy_unalign7:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r3, [r1]
cbz r3, strcpy_unalign7_copy5bytes
ldrb r4, [r1, #1]
cbz r4, strcpy_unalign7_copy6bytes
ldrb r5, [r1, #2]
cbz r5, strcpy_unalign7_copy7bytes
ldr r3, [r1], #4
pld [r1, #64]
lsrs ip, r3, #24
stmia r0!, {r2, r3}
beq strcpy_unalign_return
b strcpy_unalign7
strcpy_unalign7_copy5bytes:
stmia r0!, {r2}
strb r3, [r0]
strcpy_unalign_return:
m_ret inst=pop
strcpy_unalign7_copy6bytes:
stmia r0!, {r2}
strb r3, [r0], #1
strb r4, [r0], #1
m_ret inst=pop
strcpy_unalign7_copy7bytes:
stmia r0!, {r2}
strb r3, [r0], #1
strb r4, [r0], #1
strb r5, [r0], #1
m_ret inst=pop
.p2align 2
// Can read 6 bytes before possibly crossing a page.
strcpy_unalign6:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r4, [r1]
cbz r4, strcpy_unalign_copy5bytes
ldrb r5, [r1, #1]
cbz r5, strcpy_unalign_copy6bytes
ldr r3, [r1], #4
pld [r1, #64]
tst r3, #0xff0000
beq strcpy_unalign6_copy7bytes
lsrs ip, r3, #24
stmia r0!, {r2, r3}
beq strcpy_unalign_return
b strcpy_unalign6
strcpy_unalign6_copy7bytes:
stmia r0!, {r2}
strh r3, [r0], #2
lsr r3, #16
strb r3, [r0]
m_ret inst=pop
.p2align 2
// Can read 5 bytes before possibly crossing a page.
strcpy_unalign5:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r4, [r1]
cbz r4, strcpy_unalign_copy5bytes
ldr r3, [r1], #4
pld [r1, #64]
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_unalign5
strcpy_unalign_copy5bytes:
stmia r0!, {r2}
strb r4, [r0]
m_ret inst=pop
strcpy_unalign_copy6bytes:
stmia r0!, {r2}
strb r4, [r0], #1
strb r5, [r0]
m_ret inst=pop
.p2align 2
// Can read 4 bytes before possibly crossing a page.
strcpy_unalign4:
ldmia r1!, {r2}
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldmia r1!, {r3}
pld [r1, #64]
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_unalign4
.p2align 2
// Can read 3 bytes before possibly crossing a page.
strcpy_unalign3:
ldrb r2, [r1]
cbz r2, strcpy_unalign3_copy1byte
ldrb r3, [r1, #1]
cbz r3, strcpy_unalign3_copy2bytes
ldrb r4, [r1, #2]
cbz r4, strcpy_unalign3_copy3bytes
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
lsrs lr, r2, #24
beq strcpy_unalign_copy4bytes
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_unalign3
strcpy_unalign3_copy1byte:
strb r2, [r0]
m_ret inst=pop
strcpy_unalign3_copy2bytes:
strb r2, [r0], #1
strb r3, [r0]
m_ret inst=pop
strcpy_unalign3_copy3bytes:
strb r2, [r0], #1
strb r3, [r0], #1
strb r4, [r0]
m_ret inst=pop
.p2align 2
// Can read 2 bytes before possibly crossing a page.
strcpy_unalign2:
ldrb r2, [r1]
cbz r2, strcpy_unalign_copy1byte
ldrb r3, [r1, #1]
cbz r3, strcpy_unalign_copy2bytes
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
tst r2, #0xff0000
beq strcpy_unalign_copy3bytes
lsrs ip, r2, #24
beq strcpy_unalign_copy4bytes
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_unalign2
.p2align 2
// Can read 1 byte before possibly crossing a page.
strcpy_unalign1:
ldrb r2, [r1]
cbz r2, strcpy_unalign_copy1byte
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_unalign1
strcpy_unalign_copy1byte:
strb r2, [r0]
m_ret inst=pop
strcpy_unalign_copy2bytes:
strb r2, [r0], #1
strb r3, [r0]
m_ret inst=pop
strcpy_unalign_copy3bytes:
strh r2, [r0], #2
lsr r2, #16
strb r2, [r0]
m_ret inst=pop
strcpy_unalign_copy4bytes:
stmia r0, {r2}
m_ret inst=pop
strcat_align_src:
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq strcat_align_to_32
ldrb r2, [r0], #1
cbz r2, strcat_r0_update
strcat_align_to_32:
bcc strcat_align_to_64
ldrb r2, [r0], #1
cbz r2, strcat_r0_update
ldrb r2, [r0], #1
cbz r2, strcat_r0_update
strcat_align_to_64:
tst r3, #4
beq strcat_mainloop
ldr r3, [r0], #4
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcat_zero_in_second_register
b strcat_mainloop
strcat_r0_update:
sub r0, r0, #1
b strcat_r0_scan_done
END(strcat)

View File

@@ -1,562 +0,0 @@
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 <machine/cpu-features.h>
#include <machine/asm.h>
#ifdef __ARMEB__
#define S2LOMEM lsl
#define S2LOMEMEQ lsleq
#define S2HIMEM lsr
#define MSB 0x000000ff
#define LSB 0xff000000
#define BYTE0_OFFSET 24
#define BYTE1_OFFSET 16
#define BYTE2_OFFSET 8
#define BYTE3_OFFSET 0
#else /* not __ARMEB__ */
#define S2LOMEM lsr
#define S2LOMEMEQ lsreq
#define S2HIMEM lsl
#define BYTE0_OFFSET 0
#define BYTE1_OFFSET 8
#define BYTE2_OFFSET 16
#define BYTE3_OFFSET 24
#define MSB 0xff000000
#define LSB 0x000000ff
#endif /* not __ARMEB__ */
.syntax unified
#if defined (__thumb__)
.thumb
.thumb_func
#endif
ENTRY(strcmp)
/* Use LDRD whenever possible. */
/* The main thing to look out for when comparing large blocks is that
the loads do not cross a page boundary when loading past the index
of the byte with the first difference or the first string-terminator.
For example, if the strings are identical and the string-terminator
is at index k, byte by byte comparison will not load beyond address
s1+k and s2+k; word by word comparison may load up to 3 bytes beyond
k; double word - up to 7 bytes. If the load of these bytes crosses
a page boundary, it might cause a memory fault (if the page is not mapped)
that would not have happened in byte by byte comparison.
If an address is (double) word aligned, then a load of a (double) word
from that address will not cross a page boundary.
Therefore, the algorithm below considers word and double-word alignment
of strings separately. */
/* High-level description of the algorithm.
* The fast path: if both strings are double-word aligned,
use LDRD to load two words from each string in every loop iteration.
* If the strings have the same offset from a word boundary,
use LDRB to load and compare byte by byte until
the first string is aligned to a word boundary (at most 3 bytes).
This is optimized for quick return on short unaligned strings.
* If the strings have the same offset from a double-word boundary,
use LDRD to load two words from each string in every loop iteration, as in the fast path.
* If the strings do not have the same offset from a double-word boundary,
load a word from the second string before the loop to initialize the queue.
Use LDRD to load two words from every string in every loop iteration.
Inside the loop, load the second word from the second string only after comparing
the first word, using the queued value, to guarantee safety across page boundaries.
* If the strings do not have the same offset from a word boundary,
use LDR and a shift queue. Order of loads and comparisons matters,
similarly to the previous case.
* Use UADD8 and SEL to compare words, and use REV and CLZ to compute the return value.
* The only difference between ARM and Thumb modes is the use of CBZ instruction.
* The only difference between big and little endian is the use of REV in little endian
to compute the return value, instead of MOV.
*/
.macro m_cbz reg label
#ifdef __thumb2__
cbz \reg, \label
#else /* not defined __thumb2__ */
cmp \reg, #0
beq \label
#endif /* not defined __thumb2__ */
.endm /* m_cbz */
.macro m_cbnz reg label
#ifdef __thumb2__
cbnz \reg, \label
#else /* not defined __thumb2__ */
cmp \reg, #0
bne \label
#endif /* not defined __thumb2__ */
.endm /* m_cbnz */
.macro init
/* Macro to save temporary registers and prepare magic values. */
subs sp, sp, #16
.cfi_def_cfa_offset 16
strd r4, r5, [sp, #8]
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
strd r6, r7, [sp]
.cfi_rel_offset r6, 8
.cfi_rel_offset r7, 12
mvn r6, #0 /* all F */
mov r7, #0 /* all 0 */
.endm /* init */
.macro magic_compare_and_branch w1 w2 label
/* Macro to compare registers w1 and w2 and conditionally branch to label. */
cmp \w1, \w2 /* Are w1 and w2 the same? */
magic_find_zero_bytes \w1
it eq
cmpeq ip, #0 /* Is there a zero byte in w1? */
bne \label
.endm /* magic_compare_and_branch */
.macro magic_find_zero_bytes w1
/* Macro to find all-zero bytes in w1, result is in ip. */
#if (defined (__ARM_FEATURE_DSP))
uadd8 ip, \w1, r6
sel ip, r7, r6
#else /* not defined (__ARM_FEATURE_DSP) */
/* __ARM_FEATURE_DSP is not defined for some Cortex-M processors.
Coincidently, these processors only have Thumb-2 mode, where we can use the
the (large) magic constant available directly as an immediate in instructions.
Note that we cannot use the magic constant in ARM mode, where we need
to create the constant in a register. */
sub ip, \w1, #0x01010101
bic ip, ip, \w1
and ip, ip, #0x80808080
#endif /* not defined (__ARM_FEATURE_DSP) */
.endm /* magic_find_zero_bytes */
.macro setup_return w1 w2
#ifdef __ARMEB__
mov r1, \w1
mov r2, \w2
#else /* not __ARMEB__ */
rev r1, \w1
rev r2, \w2
#endif /* not __ARMEB__ */
.endm /* setup_return */
.cfi_startproc
pld [r0, #0]
pld [r1, #0]
/* Are both strings double-word aligned? */
orr ip, r0, r1
tst ip, #7
bne .L_do_align
/* Fast path. */
.save {r4-r7}
init
.L_doubleword_aligned:
/* Get here when the strings to compare are double-word aligned. */
/* Compare two words in every iteration. */
.p2align 2
2:
pld [r0, #16]
pld [r1, #16]
/* Load the next double-word from each string. */
ldrd r2, r3, [r0], #8
ldrd r4, r5, [r1], #8
magic_compare_and_branch w1=r2, w2=r4, label=.L_return_24
magic_compare_and_branch w1=r3, w2=r5, label=.L_return_35
b 2b
.L_do_align:
/* Is the first string word-aligned? */
ands ip, r0, #3
beq .L_word_aligned_r0
/* Fast compare byte by byte until the first string is word-aligned. */
/* The offset of r0 from a word boundary is in ip. Thus, the number of bytes
to read until the next word boundary is 4-ip. */
bic r0, r0, #3
ldr r2, [r0], #4
lsls ip, ip, #31
beq .L_byte2
bcs .L_byte3
.L_byte1:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE1_OFFSET
subs ip, r3, ip
bne .L_fast_return
m_cbz reg=r3, label=.L_fast_return
.L_byte2:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE2_OFFSET
subs ip, r3, ip
bne .L_fast_return
m_cbz reg=r3, label=.L_fast_return
.L_byte3:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE3_OFFSET
subs ip, r3, ip
bne .L_fast_return
m_cbnz reg=r3, label=.L_word_aligned_r0
.L_fast_return:
mov r0, ip
bx lr
.L_word_aligned_r0:
init
/* The first string is word-aligned. */
/* Is the second string word-aligned? */
ands ip, r1, #3
bne .L_strcmp_unaligned
.L_word_aligned:
/* The strings are word-aligned. */
/* Is the first string double-word aligned? */
tst r0, #4
beq .L_doubleword_aligned_r0
/* If r0 is not double-word aligned yet, align it by loading
and comparing the next word from each string. */
ldr r2, [r0], #4
ldr r4, [r1], #4
magic_compare_and_branch w1=r2 w2=r4 label=.L_return_24
.L_doubleword_aligned_r0:
/* Get here when r0 is double-word aligned. */
/* Is r1 doubleword_aligned? */
tst r1, #4
beq .L_doubleword_aligned
/* Get here when the strings to compare are word-aligned,
r0 is double-word aligned, but r1 is not double-word aligned. */
/* Initialize the queue. */
ldr r5, [r1], #4
/* Compare two words in every iteration. */
.p2align 2
3:
pld [r0, #16]
pld [r1, #16]
/* Load the next double-word from each string and compare. */
ldrd r2, r3, [r0], #8
magic_compare_and_branch w1=r2 w2=r5 label=.L_return_25
ldrd r4, r5, [r1], #8
magic_compare_and_branch w1=r3 w2=r4 label=.L_return_34
b 3b
.macro miscmp_word offsetlo offsethi
/* Macro to compare misaligned strings. */
/* r0, r1 are word-aligned, and at least one of the strings
is not double-word aligned. */
/* Compare one word in every loop iteration. */
/* OFFSETLO is the original bit-offset of r1 from a word-boundary,
OFFSETHI is 32 - OFFSETLO (i.e., offset from the next word). */
/* Initialize the shift queue. */
ldr r5, [r1], #4
/* Compare one word from each string in every loop iteration. */
.p2align 2
7:
ldr r3, [r0], #4
S2LOMEM r5, r5, #\offsetlo
magic_find_zero_bytes w1=r3
cmp r7, ip, S2HIMEM #\offsetlo
and r2, r3, r6, S2LOMEM #\offsetlo
it eq
cmpeq r2, r5
bne .L_return_25
ldr r5, [r1], #4
cmp ip, #0
eor r3, r2, r3
S2HIMEM r2, r5, #\offsethi
it eq
cmpeq r3, r2
bne .L_return_32
b 7b
.endm /* miscmp_word */
.L_return_32:
setup_return w1=r3, w2=r2
b .L_do_return
.L_return_34:
setup_return w1=r3, w2=r4
b .L_do_return
.L_return_25:
setup_return w1=r2, w2=r5
b .L_do_return
.L_return_35:
setup_return w1=r3, w2=r5
b .L_do_return
.L_return_24:
setup_return w1=r2, w2=r4
.L_do_return:
#ifdef __ARMEB__
mov r0, ip
#else /* not __ARMEB__ */
rev r0, ip
#endif /* not __ARMEB__ */
/* Restore temporaries early, before computing the return value. */
ldrd r6, r7, [sp]
ldrd r4, r5, [sp, #8]
adds sp, sp, #16
.cfi_def_cfa_offset 0
.cfi_restore r4
.cfi_restore r5
.cfi_restore r6
.cfi_restore r7
/* There is a zero or a different byte between r1 and r2. */
/* r0 contains a mask of all-zero bytes in r1. */
/* Using r0 and not ip here because cbz requires low register. */
m_cbz reg=r0, label=.L_compute_return_value
clz r0, r0
/* r0 contains the number of bits on the left of the first all-zero byte in r1. */
rsb r0, r0, #24
/* Here, r0 contains the number of bits on the right of the first all-zero byte in r1. */
lsr r1, r1, r0
lsr r2, r2, r0
.L_compute_return_value:
movs r0, #1
cmp r1, r2
/* The return value is computed as follows.
If r1>r2 then (C==1 and Z==0) and LS doesn't hold and r0 is #1 at return.
If r1<r2 then (C==0 and Z==0) and we execute SBC with carry_in=0,
which means r0:=r0-r0-1 and r0 is #-1 at return.
If r1=r2 then (C==1 and Z==1) and we execute SBC with carry_in=1,
which means r0:=r0-r0 and r0 is #0 at return.
(C==0 and Z==1) cannot happen because the carry bit is "not borrow". */
it ls
sbcls r0, r0, r0
bx lr
/* The code from the previous version of strcmp.S handles all of the
* cases where the first string and seconds string cannot both be
* aligned to a word boundary faster than the new algorithm. See
* bionic/libc/arch-arm/cortex-a15/bionic/strcmp.S for the unedited
* version of the code.
*/
.L_strcmp_unaligned:
wp1 .req r0
wp2 .req r1
b1 .req r2
w1 .req r4
w2 .req r5
t1 .req ip
@ r3 is scratch
2:
mov b1, #1
orr b1, b1, b1, lsl #8
orr b1, b1, b1, lsl #16
and t1, wp2, #3
bic wp2, wp2, #3
ldr w1, [wp1], #4
ldr w2, [wp2], #4
cmp t1, #2
beq 2f
bhi 3f
/* Critical inner Loop: Block with 3 bytes initial overlap */
.p2align 2
1:
bic t1, w1, #MSB
cmp t1, w2, S2LOMEM #8
sub r3, w1, b1
bic r3, r3, w1
bne 4f
ands r3, r3, b1, lsl #7
it eq
ldreq w2, [wp2], #4
bne 5f
eor t1, t1, w1
cmp t1, w2, S2HIMEM #24
bne 6f
ldr w1, [wp1], #4
b 1b
4:
S2LOMEM w2, w2, #8
b 8f
5:
#ifdef __ARMEB__
/* The syndrome value may contain false ones if the string ends
* with the bytes 0x01 0x00
*/
tst w1, #0xff000000
itt ne
tstne w1, #0x00ff0000
tstne w1, #0x0000ff00
beq 7f
#else
bics r3, r3, #0xff000000
bne 7f
#endif
ldrb w2, [wp2]
S2LOMEM t1, w1, #24
#ifdef __ARMEB__
lsl w2, w2, #24
#endif
b 8f
6:
S2LOMEM t1, w1, #24
and w2, w2, #LSB
b 8f
/* Critical inner Loop: Block with 2 bytes initial overlap */
.p2align 2
2:
S2HIMEM t1, w1, #16
sub r3, w1, b1
S2LOMEM t1, t1, #16
bic r3, r3, w1
cmp t1, w2, S2LOMEM #16
bne 4f
ands r3, r3, b1, lsl #7
it eq
ldreq w2, [wp2], #4
bne 5f
eor t1, t1, w1
cmp t1, w2, S2HIMEM #16
bne 6f
ldr w1, [wp1], #4
b 2b
5:
#ifdef __ARMEB__
/* The syndrome value may contain false ones if the string ends
* with the bytes 0x01 0x00
*/
tst w1, #0xff000000
it ne
tstne w1, #0x00ff0000
beq 7f
#else
lsls r3, r3, #16
bne 7f
#endif
ldrh w2, [wp2]
S2LOMEM t1, w1, #16
#ifdef __ARMEB__
lsl w2, w2, #16
#endif
b 8f
6:
S2HIMEM w2, w2, #16
S2LOMEM t1, w1, #16
4:
S2LOMEM w2, w2, #16
b 8f
/* Critical inner Loop: Block with 1 byte initial overlap */
.p2align 2
3:
and t1, w1, #LSB
cmp t1, w2, S2LOMEM #24
sub r3, w1, b1
bic r3, r3, w1
bne 4f
ands r3, r3, b1, lsl #7
it eq
ldreq w2, [wp2], #4
bne 5f
eor t1, t1, w1
cmp t1, w2, S2HIMEM #8
bne 6f
ldr w1, [wp1], #4
b 3b
4:
S2LOMEM w2, w2, #24
b 8f
5:
/* The syndrome value may contain false ones if the string ends
* with the bytes 0x01 0x00
*/
tst w1, #LSB
beq 7f
ldr w2, [wp2], #4
6:
S2LOMEM t1, w1, #8
bic w2, w2, #MSB
b 8f
7:
mov r0, #0
/* Restore registers and stack. */
ldrd r6, r7, [sp]
ldrd r4, r5, [sp, #8]
adds sp, sp, #16
.cfi_def_cfa_offset 0
.cfi_restore r4
.cfi_restore r5
.cfi_restore r6
.cfi_restore r7
bx lr
8:
and r2, t1, #LSB
and r0, w2, #LSB
cmp r0, #1
it cs
cmpcs r0, r2
itt eq
S2LOMEMEQ t1, t1, #8
S2LOMEMEQ w2, w2, #8
beq 8b
sub r0, r2, r0
/* Restore registers and stack. */
ldrd r6, r7, [sp]
ldrd r4, r5, [sp, #8]
adds sp, sp, #16
bx lr
.cfi_endproc
END(strcmp)

View File

@@ -1,456 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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.
*/
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 <machine/asm.h>
.syntax unified
.thumb
.thumb_func
.macro m_push
push {r0, r4, r5, lr}
.endm // m_push
.macro m_ret inst
\inst {r0, r4, r5, pc}
.endm // m_ret
.macro m_copy_byte reg, cmd, label
ldrb \reg, [r1], #1
strb \reg, [r0], #1
\cmd \reg, \label
.endm // m_copy_byte
ENTRY(strcpy)
// Unroll the first 8 bytes that will be copied.
m_push
m_copy_byte reg=r2, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r3, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r4, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r5, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r2, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r3, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r4, cmd=cbz, label=strcpy_finish
m_copy_byte reg=r5, cmd=cbnz, label=strcpy_continue
strcpy_finish:
m_ret inst=pop
strcpy_continue:
pld [r1, #0]
ands r3, r0, #7
bne strcpy_align_dst
strcpy_check_src_align:
// At this point dst is aligned to a double word, check if src
// is also aligned to a double word.
ands r3, r1, #7
bne strcpy_unaligned_copy
.p2align 2
strcpy_mainloop:
ldmia r1!, {r2, r3}
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_mainloop
strcpy_zero_in_first_register:
lsls lr, ip, #17
itt ne
strbne r2, [r0]
m_ret inst=popne
itt cs
strhcs r2, [r0]
m_ret inst=popcs
lsls ip, ip, #1
itt eq
streq r2, [r0]
m_ret inst=popeq
strh r2, [r0], #2
lsr r3, r2, #16
strb r3, [r0]
m_ret inst=pop
strcpy_zero_in_second_register:
lsls lr, ip, #17
ittt ne
stmiane r0!, {r2}
strbne r3, [r0]
m_ret inst=popne
ittt cs
strcs r2, [r0], #4
strhcs r3, [r0]
m_ret inst=popcs
lsls ip, ip, #1
itt eq
stmiaeq r0, {r2, r3}
m_ret inst=popeq
stmia r0!, {r2}
strh r3, [r0], #2
lsr r4, r3, #16
strb r4, [r0]
m_ret inst=pop
strcpy_align_dst:
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq strcpy_align_to_32
ldrb r2, [r1], #1
strb r2, [r0], #1
cbz r2, strcpy_complete
strcpy_align_to_32:
bcc strcpy_align_to_64
ldrb r4, [r1], #1
strb r4, [r0], #1
cmp r4, #0
it eq
m_ret inst=popeq
ldrb r5, [r1], #1
strb r5, [r0], #1
cmp r5, #0
it eq
m_ret inst=popeq
strcpy_align_to_64:
tst r3, #4
beq strcpy_check_src_align
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
stmia r0!, {r2}
b strcpy_check_src_align
strcpy_complete:
m_ret inst=pop
strcpy_unaligned_copy:
// Dst is aligned to a double word, while src is at an unknown alignment.
// There are 7 different versions of the unaligned copy code
// to prevent overreading the src. The mainloop of every single version
// will store 64 bits per loop. The difference is how much of src can
// be read without potentially crossing a page boundary.
tbb [pc, r3]
strcpy_unaligned_branchtable:
.byte 0
.byte ((strcpy_unalign7 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign6 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign5 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign4 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign3 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign2 - strcpy_unaligned_branchtable)/2)
.byte ((strcpy_unalign1 - strcpy_unaligned_branchtable)/2)
.p2align 2
// Can read 7 bytes before possibly crossing a page.
strcpy_unalign7:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r3, [r1]
cbz r3, strcpy_unalign7_copy5bytes
ldrb r4, [r1, #1]
cbz r4, strcpy_unalign7_copy6bytes
ldrb r5, [r1, #2]
cbz r5, strcpy_unalign7_copy7bytes
ldr r3, [r1], #4
pld [r1, #64]
lsrs ip, r3, #24
stmia r0!, {r2, r3}
beq strcpy_unalign_return
b strcpy_unalign7
strcpy_unalign7_copy5bytes:
stmia r0!, {r2}
strb r3, [r0]
strcpy_unalign_return:
m_ret inst=pop
strcpy_unalign7_copy6bytes:
stmia r0!, {r2}
strb r3, [r0], #1
strb r4, [r0], #1
m_ret inst=pop
strcpy_unalign7_copy7bytes:
stmia r0!, {r2}
strb r3, [r0], #1
strb r4, [r0], #1
strb r5, [r0], #1
m_ret inst=pop
.p2align 2
// Can read 6 bytes before possibly crossing a page.
strcpy_unalign6:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r4, [r1]
cbz r4, strcpy_unalign_copy5bytes
ldrb r5, [r1, #1]
cbz r5, strcpy_unalign_copy6bytes
ldr r3, [r1], #4
pld [r1, #64]
tst r3, #0xff0000
beq strcpy_unalign6_copy7bytes
lsrs ip, r3, #24
stmia r0!, {r2, r3}
beq strcpy_unalign_return
b strcpy_unalign6
strcpy_unalign6_copy7bytes:
stmia r0!, {r2}
strh r3, [r0], #2
lsr r3, #16
strb r3, [r0]
m_ret inst=pop
.p2align 2
// Can read 5 bytes before possibly crossing a page.
strcpy_unalign5:
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldrb r4, [r1]
cbz r4, strcpy_unalign_copy5bytes
ldr r3, [r1], #4
pld [r1, #64]
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_unalign5
strcpy_unalign_copy5bytes:
stmia r0!, {r2}
strb r4, [r0]
m_ret inst=pop
strcpy_unalign_copy6bytes:
stmia r0!, {r2}
strb r4, [r0], #1
strb r5, [r0]
m_ret inst=pop
.p2align 2
// Can read 4 bytes before possibly crossing a page.
strcpy_unalign4:
ldmia r1!, {r2}
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
ldmia r1!, {r3}
pld [r1, #64]
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_unalign4
.p2align 2
// Can read 3 bytes before possibly crossing a page.
strcpy_unalign3:
ldrb r2, [r1]
cbz r2, strcpy_unalign3_copy1byte
ldrb r3, [r1, #1]
cbz r3, strcpy_unalign3_copy2bytes
ldrb r4, [r1, #2]
cbz r4, strcpy_unalign3_copy3bytes
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
lsrs lr, r2, #24
beq strcpy_unalign_copy4bytes
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_unalign3
strcpy_unalign3_copy1byte:
strb r2, [r0]
m_ret inst=pop
strcpy_unalign3_copy2bytes:
strb r2, [r0], #1
strb r3, [r0]
m_ret inst=pop
strcpy_unalign3_copy3bytes:
strb r2, [r0], #1
strb r3, [r0], #1
strb r4, [r0]
m_ret inst=pop
.p2align 2
// Can read 2 bytes before possibly crossing a page.
strcpy_unalign2:
ldrb r2, [r1]
cbz r2, strcpy_unalign_copy1byte
ldrb r3, [r1, #1]
cbz r3, strcpy_unalign_copy2bytes
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
tst r2, #0xff0000
beq strcpy_unalign_copy3bytes
lsrs ip, r2, #24
beq strcpy_unalign_copy4bytes
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_unalign2
.p2align 2
// Can read 1 byte before possibly crossing a page.
strcpy_unalign1:
ldrb r2, [r1]
cbz r2, strcpy_unalign_copy1byte
ldr r2, [r1], #4
ldr r3, [r1], #4
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne strcpy_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne strcpy_zero_in_second_register
stmia r0!, {r2, r3}
b strcpy_unalign1
strcpy_unalign_copy1byte:
strb r2, [r0]
m_ret inst=pop
strcpy_unalign_copy2bytes:
strb r2, [r0], #1
strb r3, [r0]
m_ret inst=pop
strcpy_unalign_copy3bytes:
strh r2, [r0], #2
lsr r2, #16
strb r2, [r0]
m_ret inst=pop
strcpy_unalign_copy4bytes:
stmia r0, {r2}
m_ret inst=pop
END(strcpy)

View File

@@ -1,167 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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.
*/
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 <machine/asm.h>
.syntax unified
.thumb
.thumb_func
ENTRY(strlen)
pld [r0, #0]
mov r1, r0
ands r3, r0, #7
bne align_src
.p2align 2
mainloop:
ldmia r1!, {r2, r3}
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne zero_in_second_register
b mainloop
zero_in_first_register:
sub r0, r1, r0
// Check for zero in byte 0.
lsls r2, ip, #17
beq check_byte1_reg1
sub r0, r0, #8
bx lr
check_byte1_reg1:
bcc check_byte2_reg1
sub r0, r0, #7
bx lr
check_byte2_reg1:
// Check for zero in byte 2.
tst ip, #0x800000
itt ne
subne r0, r0, #6
bxne lr
sub r0, r0, #5
bx lr
zero_in_second_register:
sub r0, r1, r0
// Check for zero in byte 0.
lsls r2, ip, #17
beq check_byte1_reg2
sub r0, r0, #4
bx lr
check_byte1_reg2:
bcc check_byte2_reg2
sub r0, r0, #3
bx lr
check_byte2_reg2:
// Check for zero in byte 2.
tst ip, #0x800000
itt ne
subne r0, r0, #2
bxne lr
sub r0, r0, #1
bx lr
align_src:
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq align_to_32
ldrb r2, [r1], #1
cbz r2, done
align_to_32:
bcc align_to_64
ldrb r2, [r1], #1
cbz r2, done
ldrb r2, [r1], #1
cbz r2, done
align_to_64:
tst r3, #4
beq mainloop
ldr r2, [r1], #4
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne zero_in_second_register
b mainloop
done:
sub r0, r1, r0
sub r0, r0, #1
bx lr
END(strlen)

View File

@@ -1,10 +0,0 @@
$(call libc-add-cpu-variant-src,MEMCPY,arch-arm/cortex-a9/bionic/memcpy.S)
$(call libc-add-cpu-variant-src,MEMSET,arch-arm/cortex-a9/bionic/memset.S)
$(call libc-add-cpu-variant-src,STRCAT,arch-arm/cortex-a9/bionic/strcat.S)
$(call libc-add-cpu-variant-src,STRCMP,arch-arm/cortex-a9/bionic/strcmp.S)
$(call libc-add-cpu-variant-src,STRCPY,arch-arm/cortex-a9/bionic/strcpy.S)
$(call libc-add-cpu-variant-src,STRLEN,arch-arm/cortex-a9/bionic/strlen.S)
$(call libc-add-cpu-variant-src,__STRCAT_CHK,arch-arm/cortex-a9/bionic/__strcat_chk.S)
$(call libc-add-cpu-variant-src,__STRCPY_CHK,arch-arm/cortex-a9/bionic/__strcpy_chk.S)
include bionic/libc/arch-arm/generic/generic.mk

View File

@@ -1,404 +0,0 @@
/*
* 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:
* * 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.
*
* 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 <machine/cpu-features.h>
#include <machine/asm.h>
#include "libc_events.h"
/*
* Optimized memcpy() for ARM.
*
* note that memcpy() always returns the destination pointer,
* so we have to preserve R0.
*/
ENTRY(__memcpy_chk)
cmp r2, r3
bgt fortify_check_failed
// Fall through to memcpy...
END(__memcpy_chk)
ENTRY(memcpy)
/* The stack must always be 64-bits aligned to be compliant with the
* ARM ABI. Since we have to save R0, we might as well save R4
* which we can use for better pipelining of the reads below
*/
.save {r0, r4, lr}
stmfd sp!, {r0, r4, lr}
/* Making room for r5-r11 which will be spilled later */
.pad #28
sub sp, sp, #28
// preload the destination because we'll align it to a cache line
// with small writes. Also start the source "pump".
PLD (r0, #0)
PLD (r1, #0)
PLD (r1, #32)
/* it simplifies things to take care of len<4 early */
cmp r2, #4
blo copy_last_3_and_return
/* compute the offset to align the source
* offset = (4-(src&3))&3 = -src & 3
*/
rsb r3, r1, #0
ands r3, r3, #3
beq src_aligned
/* align source to 32 bits. We need to insert 2 instructions between
* a ldr[b|h] and str[b|h] because byte and half-word instructions
* stall 2 cycles.
*/
movs r12, r3, lsl #31
sub r2, r2, r3 /* we know that r3 <= r2 because r2 >= 4 */
ldrmib r3, [r1], #1
ldrcsb r4, [r1], #1
ldrcsb r12,[r1], #1
strmib r3, [r0], #1
strcsb r4, [r0], #1
strcsb r12,[r0], #1
src_aligned:
/* see if src and dst are aligned together (congruent) */
eor r12, r0, r1
tst r12, #3
bne non_congruent
/* Use post-incriment mode for stm to spill r5-r11 to reserved stack
* frame. Don't update sp.
*/
stmea sp, {r5-r11}
/* align the destination to a cache-line */
rsb r3, r0, #0
ands r3, r3, #0x1C
beq congruent_aligned32
cmp r3, r2
andhi r3, r2, #0x1C
/* conditionally copies 0 to 7 words (length in r3) */
movs r12, r3, lsl #28
ldmcsia r1!, {r4, r5, r6, r7} /* 16 bytes */
ldmmiia r1!, {r8, r9} /* 8 bytes */
stmcsia r0!, {r4, r5, r6, r7}
stmmiia r0!, {r8, r9}
tst r3, #0x4
ldrne r10,[r1], #4 /* 4 bytes */
strne r10,[r0], #4
sub r2, r2, r3
congruent_aligned32:
/*
* here source is aligned to 32 bytes.
*/
cached_aligned32:
subs r2, r2, #32
blo less_than_32_left
/*
* We preload a cache-line up to 64 bytes ahead. On the 926, this will
* stall only until the requested world is fetched, but the linefill
* continues in the the background.
* While the linefill is going, we write our previous cache-line
* into the write-buffer (which should have some free space).
* When the linefill is done, the writebuffer will
* start dumping its content into memory
*
* While all this is going, we then load a full cache line into
* 8 registers, this cache line should be in the cache by now
* (or partly in the cache).
*
* This code should work well regardless of the source/dest alignment.
*
*/
// Align the preload register to a cache-line because the cpu does
// "critical word first" (the first word requested is loaded first).
bic r12, r1, #0x1F
add r12, r12, #64
1: ldmia r1!, { r4-r11 }
PLD (r12, #64)
subs r2, r2, #32
// NOTE: if r12 is more than 64 ahead of r1, the following ldrhi
// for ARM9 preload will not be safely guarded by the preceding subs.
// When it is safely guarded the only possibility to have SIGSEGV here
// is because the caller overstates the length.
ldrhi r3, [r12], #32 /* cheap ARM9 preload */
stmia r0!, { r4-r11 }
bhs 1b
add r2, r2, #32
less_than_32_left:
/*
* less than 32 bytes left at this point (length in r2)
*/
/* skip all this if there is nothing to do, which should
* be a common case (if not executed the code below takes
* about 16 cycles)
*/
tst r2, #0x1F
beq 1f
/* conditionnaly copies 0 to 31 bytes */
movs r12, r2, lsl #28
ldmcsia r1!, {r4, r5, r6, r7} /* 16 bytes */
ldmmiia r1!, {r8, r9} /* 8 bytes */
stmcsia r0!, {r4, r5, r6, r7}
stmmiia r0!, {r8, r9}
movs r12, r2, lsl #30
ldrcs r3, [r1], #4 /* 4 bytes */
ldrmih r4, [r1], #2 /* 2 bytes */
strcs r3, [r0], #4
strmih r4, [r0], #2
tst r2, #0x1
ldrneb r3, [r1] /* last byte */
strneb r3, [r0]
/* we're done! restore everything and return */
1: ldmfd sp!, {r5-r11}
ldmfd sp!, {r0, r4, lr}
bx lr
/********************************************************************/
non_congruent:
/*
* here source is aligned to 4 bytes
* but destination is not.
*
* in the code below r2 is the number of bytes read
* (the number of bytes written is always smaller, because we have
* partial words in the shift queue)
*/
cmp r2, #4
blo copy_last_3_and_return
/* Use post-incriment mode for stm to spill r5-r11 to reserved stack
* frame. Don't update sp.
*/
stmea sp, {r5-r11}
/* compute shifts needed to align src to dest */
rsb r5, r0, #0
and r5, r5, #3 /* r5 = # bytes in partial words */
mov r12, r5, lsl #3 /* r12 = right */
rsb lr, r12, #32 /* lr = left */
/* read the first word */
ldr r3, [r1], #4
sub r2, r2, #4
/* write a partial word (0 to 3 bytes), such that destination
* becomes aligned to 32 bits (r5 = nb of words to copy for alignment)
*/
movs r5, r5, lsl #31
strmib r3, [r0], #1
movmi r3, r3, lsr #8
strcsb r3, [r0], #1
movcs r3, r3, lsr #8
strcsb r3, [r0], #1
movcs r3, r3, lsr #8
cmp r2, #4
blo partial_word_tail
/* Align destination to 32 bytes (cache line boundary) */
1: tst r0, #0x1c
beq 2f
ldr r5, [r1], #4
sub r2, r2, #4
orr r4, r3, r5, lsl lr
mov r3, r5, lsr r12
str r4, [r0], #4
cmp r2, #4
bhs 1b
blo partial_word_tail
/* copy 32 bytes at a time */
2: subs r2, r2, #32
blo less_than_thirtytwo
/* Use immediate mode for the shifts, because there is an extra cycle
* for register shifts, which could account for up to 50% of
* performance hit.
*/
cmp r12, #24
beq loop24
cmp r12, #8
beq loop8
loop16:
ldr r12, [r1], #4
1: mov r4, r12
ldmia r1!, { r5,r6,r7, r8,r9,r10,r11}
PLD (r1, #64)
subs r2, r2, #32
ldrhs r12, [r1], #4
orr r3, r3, r4, lsl #16
mov r4, r4, lsr #16
orr r4, r4, r5, lsl #16
mov r5, r5, lsr #16
orr r5, r5, r6, lsl #16
mov r6, r6, lsr #16
orr r6, r6, r7, lsl #16
mov r7, r7, lsr #16
orr r7, r7, r8, lsl #16
mov r8, r8, lsr #16
orr r8, r8, r9, lsl #16
mov r9, r9, lsr #16
orr r9, r9, r10, lsl #16
mov r10, r10, lsr #16
orr r10, r10, r11, lsl #16
stmia r0!, {r3,r4,r5,r6, r7,r8,r9,r10}
mov r3, r11, lsr #16
bhs 1b
b less_than_thirtytwo
loop8:
ldr r12, [r1], #4
1: mov r4, r12
ldmia r1!, { r5,r6,r7, r8,r9,r10,r11}
PLD (r1, #64)
subs r2, r2, #32
ldrhs r12, [r1], #4
orr r3, r3, r4, lsl #24
mov r4, r4, lsr #8
orr r4, r4, r5, lsl #24
mov r5, r5, lsr #8
orr r5, r5, r6, lsl #24
mov r6, r6, lsr #8
orr r6, r6, r7, lsl #24
mov r7, r7, lsr #8
orr r7, r7, r8, lsl #24
mov r8, r8, lsr #8
orr r8, r8, r9, lsl #24
mov r9, r9, lsr #8
orr r9, r9, r10, lsl #24
mov r10, r10, lsr #8
orr r10, r10, r11, lsl #24
stmia r0!, {r3,r4,r5,r6, r7,r8,r9,r10}
mov r3, r11, lsr #8
bhs 1b
b less_than_thirtytwo
loop24:
ldr r12, [r1], #4
1: mov r4, r12
ldmia r1!, { r5,r6,r7, r8,r9,r10,r11}
PLD (r1, #64)
subs r2, r2, #32
ldrhs r12, [r1], #4
orr r3, r3, r4, lsl #8
mov r4, r4, lsr #24
orr r4, r4, r5, lsl #8
mov r5, r5, lsr #24
orr r5, r5, r6, lsl #8
mov r6, r6, lsr #24
orr r6, r6, r7, lsl #8
mov r7, r7, lsr #24
orr r7, r7, r8, lsl #8
mov r8, r8, lsr #24
orr r8, r8, r9, lsl #8
mov r9, r9, lsr #24
orr r9, r9, r10, lsl #8
mov r10, r10, lsr #24
orr r10, r10, r11, lsl #8
stmia r0!, {r3,r4,r5,r6, r7,r8,r9,r10}
mov r3, r11, lsr #24
bhs 1b
less_than_thirtytwo:
/* copy the last 0 to 31 bytes of the source */
rsb r12, lr, #32 /* we corrupted r12, recompute it */
add r2, r2, #32
cmp r2, #4
blo partial_word_tail
1: ldr r5, [r1], #4
sub r2, r2, #4
orr r4, r3, r5, lsl lr
mov r3, r5, lsr r12
str r4, [r0], #4
cmp r2, #4
bhs 1b
partial_word_tail:
/* we have a partial word in the input buffer */
movs r5, lr, lsl #(31-3)
strmib r3, [r0], #1
movmi r3, r3, lsr #8
strcsb r3, [r0], #1
movcs r3, r3, lsr #8
strcsb r3, [r0], #1
/* Refill spilled registers from the stack. Don't update sp. */
ldmfd sp, {r5-r11}
copy_last_3_and_return:
movs r2, r2, lsl #31 /* copy remaining 0, 1, 2 or 3 bytes */
ldrmib r2, [r1], #1
ldrcsb r3, [r1], #1
ldrcsb r12,[r1]
strmib r2, [r0], #1
strcsb r3, [r0], #1
strcsb r12,[r0]
/* we're done! restore sp and spilled registers and return */
add sp, sp, #28
ldmfd sp!, {r0, r4, lr}
bx lr
// Only reached when the __memcpy_chk check fails.
fortify_check_failed:
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_MEMCPY_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+8)
END(memcpy)
.data
error_string:
.string "memcpy buffer overflow"

View File

@@ -1,133 +0,0 @@
/*
* 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:
* * 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.
*
* 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 <machine/asm.h>
#include "libc_events.h"
/*
* Optimized memset() for ARM.
*
* memset() returns its first argument.
*/
ENTRY(__memset_chk)
cmp r2, r3
bls done
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_MEMSET_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+8)
END(__memset_chk)
ENTRY(bzero)
mov r2, r1
mov r1, #0
done:
// Fall through to memset...
END(bzero)
ENTRY(memset)
/* compute the offset to align the destination
* offset = (4-(src&3))&3 = -src & 3
*/
.save {r0, r4-r7, lr}
stmfd sp!, {r0, r4-r7, lr}
rsb r3, r0, #0
ands r3, r3, #3
cmp r3, r2
movhi r3, r2
/* splat r1 */
mov r1, r1, lsl #24
orr r1, r1, r1, lsr #8
orr r1, r1, r1, lsr #16
movs r12, r3, lsl #31
strcsb r1, [r0], #1 /* can't use strh (alignment unknown) */
strcsb r1, [r0], #1
strmib r1, [r0], #1
subs r2, r2, r3
ldmlsfd sp!, {r0, r4-r7, lr} /* return */
bxls lr
/* align the destination to a cache-line */
mov r12, r1
mov lr, r1
mov r4, r1
mov r5, r1
mov r6, r1
mov r7, r1
rsb r3, r0, #0
ands r3, r3, #0x1C
beq 3f
cmp r3, r2
andhi r3, r2, #0x1C
sub r2, r2, r3
/* conditionally writes 0 to 7 words (length in r3) */
movs r3, r3, lsl #28
stmcsia r0!, {r1, lr}
stmcsia r0!, {r1, lr}
stmmiia r0!, {r1, lr}
movs r3, r3, lsl #2
strcs r1, [r0], #4
3:
subs r2, r2, #32
mov r3, r1
bmi 2f
1: subs r2, r2, #32
stmia r0!, {r1,r3,r4,r5,r6,r7,r12,lr}
bhs 1b
2: add r2, r2, #32
/* conditionally stores 0 to 31 bytes */
movs r2, r2, lsl #28
stmcsia r0!, {r1,r3,r12,lr}
stmmiia r0!, {r1, lr}
movs r2, r2, lsl #2
strcs r1, [r0], #4
strmih r1, [r0], #2
movs r2, r2, lsl #2
strcsb r1, [r0]
ldmfd sp!, {r0, r4-r7, lr}
bx lr
END(memset)
.data
error_string:
.string "memset buffer overflow"

View File

@@ -1,317 +0,0 @@
/*
* Copyright (c) 2011 The Android Open Source Project
* Copyright (c) 2008 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 <machine/cpu-features.h>
#include <machine/asm.h>
.text
#ifdef __ARMEB__
#define SHFT2LSB lsl
#define SHFT2LSBEQ lsleq
#define SHFT2MSB lsr
#define SHFT2MSBEQ lsreq
#define MSB 0x000000ff
#define LSB 0xff000000
#else
#define SHFT2LSB lsr
#define SHFT2LSBEQ lsreq
#define SHFT2MSB lsl
#define SHFT2MSBEQ lsleq
#define MSB 0xff000000
#define LSB 0x000000ff
#endif
#define magic1(REG) REG
#define magic2(REG) REG, lsl #7
ENTRY(strcmp)
PLD(r0, #0)
PLD(r1, #0)
eor r2, r0, r1
tst r2, #3
/* Strings not at same byte offset from a word boundary. */
bne .Lstrcmp_unaligned
ands r2, r0, #3
bic r0, r0, #3
bic r1, r1, #3
ldr ip, [r0], #4
it eq
ldreq r3, [r1], #4
beq 1f
/* Although s1 and s2 have identical initial alignment, they are
* not currently word aligned. Rather than comparing bytes,
* make sure that any bytes fetched from before the addressed
* bytes are forced to 0xff. Then they will always compare
* equal.
*/
eor r2, r2, #3
lsl r2, r2, #3
mvn r3, #MSB
SHFT2LSB r2, r3, r2
ldr r3, [r1], #4
orr ip, ip, r2
orr r3, r3, r2
1:
/* Load the 'magic' constant 0x01010101. */
str r4, [sp, #-4]!
mov r4, #1
orr r4, r4, r4, lsl #8
orr r4, r4, r4, lsl #16
.p2align 2
4:
PLD(r0, #8)
PLD(r1, #8)
sub r2, ip, magic1(r4)
cmp ip, r3
itttt eq
/* check for any zero bytes in first word */
biceq r2, r2, ip
tsteq r2, magic2(r4)
ldreq ip, [r0], #4
ldreq r3, [r1], #4
beq 4b
2:
/* There's a zero or a different byte in the word */
SHFT2MSB r0, ip, #24
SHFT2LSB ip, ip, #8
cmp r0, #1
it cs
cmpcs r0, r3, SHFT2MSB #24
it eq
SHFT2LSBEQ r3, r3, #8
beq 2b
/* On a big-endian machine, r0 contains the desired byte in bits
* 0-7; on a little-endian machine they are in bits 24-31. In
* both cases the other bits in r0 are all zero. For r3 the
* interesting byte is at the other end of the word, but the
* other bits are not necessarily zero. We need a signed result
* representing the differnece in the unsigned bytes, so for the
* little-endian case we can't just shift the interesting bits up.
*/
#ifdef __ARMEB__
sub r0, r0, r3, lsr #24
#else
and r3, r3, #255
/* No RSB instruction in Thumb2 */
#ifdef __thumb2__
lsr r0, r0, #24
sub r0, r0, r3
#else
rsb r0, r3, r0, lsr #24
#endif
#endif
ldr r4, [sp], #4
bx lr
.Lstrcmp_unaligned:
wp1 .req r0
wp2 .req r1
b1 .req r2
w1 .req r4
w2 .req r5
t1 .req ip
@ r3 is scratch
/* First of all, compare bytes until wp1(sp1) is word-aligned. */
1:
tst wp1, #3
beq 2f
ldrb r2, [wp1], #1
ldrb r3, [wp2], #1
cmp r2, #1
it cs
cmpcs r2, r3
beq 1b
sub r0, r2, r3
bx lr
2:
str r5, [sp, #-4]!
str r4, [sp, #-4]!
mov b1, #1
orr b1, b1, b1, lsl #8
orr b1, b1, b1, lsl #16
and t1, wp2, #3
bic wp2, wp2, #3
ldr w1, [wp1], #4
ldr w2, [wp2], #4
cmp t1, #2
beq 2f
bhi 3f
/* Critical inner Loop: Block with 3 bytes initial overlap */
.p2align 2
1:
bic t1, w1, #MSB
cmp t1, w2, SHFT2LSB #8
sub r3, w1, b1
bic r3, r3, w1
bne 4f
ands r3, r3, b1, lsl #7
it eq
ldreq w2, [wp2], #4
bne 5f
eor t1, t1, w1
cmp t1, w2, SHFT2MSB #24
bne 6f
ldr w1, [wp1], #4
b 1b
4:
SHFT2LSB w2, w2, #8
b 8f
5:
#ifdef __ARMEB__
/* The syndrome value may contain false ones if the string ends
* with the bytes 0x01 0x00
*/
tst w1, #0xff000000
itt ne
tstne w1, #0x00ff0000
tstne w1, #0x0000ff00
beq 7f
#else
bics r3, r3, #0xff000000
bne 7f
#endif
ldrb w2, [wp2]
SHFT2LSB t1, w1, #24
#ifdef __ARMEB__
lsl w2, w2, #24
#endif
b 8f
6:
SHFT2LSB t1, w1, #24
and w2, w2, #LSB
b 8f
/* Critical inner Loop: Block with 2 bytes initial overlap */
.p2align 2
2:
SHFT2MSB t1, w1, #16
sub r3, w1, b1
SHFT2LSB t1, t1, #16
bic r3, r3, w1
cmp t1, w2, SHFT2LSB #16
bne 4f
ands r3, r3, b1, lsl #7
it eq
ldreq w2, [wp2], #4
bne 5f
eor t1, t1, w1
cmp t1, w2, SHFT2MSB #16
bne 6f
ldr w1, [wp1], #4
b 2b
5:
#ifdef __ARMEB__
/* The syndrome value may contain false ones if the string ends
* with the bytes 0x01 0x00
*/
tst w1, #0xff000000
it ne
tstne w1, #0x00ff0000
beq 7f
#else
lsls r3, r3, #16
bne 7f
#endif
ldrh w2, [wp2]
SHFT2LSB t1, w1, #16
#ifdef __ARMEB__
lsl w2, w2, #16
#endif
b 8f
6:
SHFT2MSB w2, w2, #16
SHFT2LSB t1, w1, #16
4:
SHFT2LSB w2, w2, #16
b 8f
/* Critical inner Loop: Block with 1 byte initial overlap */
.p2align 2
3:
and t1, w1, #LSB
cmp t1, w2, SHFT2LSB #24
sub r3, w1, b1
bic r3, r3, w1
bne 4f
ands r3, r3, b1, lsl #7
it eq
ldreq w2, [wp2], #4
bne 5f
eor t1, t1, w1
cmp t1, w2, SHFT2MSB #8
bne 6f
ldr w1, [wp1], #4
b 3b
4:
SHFT2LSB w2, w2, #24
b 8f
5:
/* The syndrome value may contain false ones if the string ends
* with the bytes 0x01 0x00
*/
tst w1, #LSB
beq 7f
ldr w2, [wp2], #4
6:
SHFT2LSB t1, w1, #8
bic w2, w2, #MSB
b 8f
7:
mov r0, #0
ldr r4, [sp], #4
ldr r5, [sp], #4
bx lr
8:
and r2, t1, #LSB
and r0, w2, #LSB
cmp r0, #1
it cs
cmpcs r0, r2
itt eq
SHFT2LSBEQ t1, t1, #8
SHFT2LSBEQ w2, w2, #8
beq 8b
sub r0, r2, r0
ldr r4, [sp], #4
ldr r5, [sp], #4
bx lr
END(strcmp)

View File

@@ -1,8 +0,0 @@
$(call libc-add-cpu-variant-src,MEMCPY,arch-arm/generic/bionic/memcpy.S)
$(call libc-add-cpu-variant-src,MEMSET,arch-arm/generic/bionic/memset.S)
$(call libc-add-cpu-variant-src,STRCAT,string/strcat.c)
$(call libc-add-cpu-variant-src,STRCMP,arch-arm/generic/bionic/strcmp.S)
$(call libc-add-cpu-variant-src,STRCPY,arch-arm/generic/bionic/strcpy.S)
$(call libc-add-cpu-variant-src,STRLEN,arch-arm/generic/bionic/strlen.c)
$(call libc-add-cpu-variant-src,__STRCAT_CHK,bionic/__strcat_chk.cpp)
$(call libc-add-cpu-variant-src,__STRCPY_CHK,bionic/__strcpy_chk.cpp)

View File

@@ -77,7 +77,7 @@
#endif /* __ARM_ARCH__ */
#endif /* __GNUC__ */
#if defined(__ARMEB__)
#ifdef __ARMEB__
#define _BYTE_ORDER _BIG_ENDIAN
#else
#define _BYTE_ORDER _LITTLE_ENDIAN

View File

@@ -35,6 +35,17 @@
#ifndef _ARM__TYPES_H_
#define _ARM__TYPES_H_
#if !defined(__ARM_EABI__)
/* the kernel defines size_t as unsigned int, but g++ wants it to be unsigned long */
#define _SIZE_T
#define _SSIZE_T
#define _PTRDIFF_T
typedef unsigned long size_t;
typedef long ssize_t;
typedef long ptrdiff_t;
#endif
/* 7.18.1.1 Exact-width integer types */
typedef __signed char __int8_t;
typedef unsigned char __uint8_t;
@@ -87,8 +98,6 @@ typedef unsigned long __psize_t;
/* Standard system types */
typedef int __clock_t;
typedef int __clockid_t;
typedef double __double_t;
typedef float __float_t;
typedef long __ptrdiff_t;
typedef int __time_t;
typedef int __timer_t;
@@ -107,4 +116,10 @@ typedef int __rune_t;
typedef void * __wctrans_t;
typedef void * __wctype_t;
#ifdef __ARMEB__
#define _BYTE_ORDER _BIG_ENDIAN
#else
#define _BYTE_ORDER _LITTLE_ENDIAN
#endif
#endif /* _ARM__TYPES_H_ */

View File

@@ -1,138 +0,0 @@
/* $NetBSD: elf_machdep.h,v 1.10 2012/08/05 04:12:46 matt Exp $ */
#ifndef _ARM_ELF_MACHDEP_H_
#define _ARM_ELF_MACHDEP_H_
#if defined(__ARMEB__)
#define ELF32_MACHDEP_ENDIANNESS ELFDATA2MSB
#else
#define ELF32_MACHDEP_ENDIANNESS ELFDATA2LSB
#endif
#define ELF64_MACHDEP_ENDIANNESS XXX /* break compilation */
#define ELF64_MACHDEP_ID_CASES \
/* no 64-bit ELF machine types supported */
/* Processor specific flags for the ELF header e_flags field. */
#define EF_ARM_RELEXEC 0x00000001
#define EF_ARM_HASENTRY 0x00000002
#define EF_ARM_INTERWORK 0x00000004 /* GNU binutils 000413 */
#define EF_ARM_SYMSARESORTED 0x00000004 /* ARM ELF A08 */
#define EF_ARM_APCS_26 0x00000008 /* GNU binutils 000413 */
#define EF_ARM_DYNSYMSUSESEGIDX 0x00000008 /* ARM ELF B01 */
#define EF_ARM_APCS_FLOAT 0x00000010 /* GNU binutils 000413 */
#define EF_ARM_MAPSYMSFIRST 0x00000010 /* ARM ELF B01 */
#define EF_ARM_PIC 0x00000020
#define EF_ARM_ALIGN8 0x00000040 /* 8-bit structure alignment. */
#define EF_ARM_NEW_ABI 0x00000080
#define EF_ARM_OLD_ABI 0x00000100
#define EF_ARM_SOFT_FLOAT 0x00000200
#define EF_ARM_EABIMASK 0xff000000
#define EF_ARM_EABI_VER1 0x01000000
#define EF_ARM_EABI_VER2 0x02000000
#define EF_ARM_EABI_VER3 0x03000000
#define EF_ARM_EABI_VER4 0x04000000
#define EF_ARM_EABI_VER5 0x05000000
#define ELF32_MACHDEP_ID_CASES \
case EM_ARM: \
break;
#define ELF32_MACHDEP_ID EM_ARM
#define ARCH_ELFSIZE 32 /* MD native binary size */
/* Processor specific relocation types */
#define R_ARM_NONE 0
#define R_ARM_PC24 1
#define R_ARM_ABS32 2
#define R_ARM_REL32 3
#define R_ARM_PC13 4
#define R_ARM_ABS16 5
#define R_ARM_ABS12 6
#define R_ARM_THM_ABS5 7
#define R_ARM_ABS8 8
#define R_ARM_SBREL32 9
#define R_ARM_THM_PC22 10
#define R_ARM_THM_PC8 11
#define R_ARM_AMP_VCALL9 12
#define R_ARM_SWI24 13
#define R_ARM_THM_SWI8 14
#define R_ARM_XPC25 15
#define R_ARM_THM_XPC22 16
/* TLS relocations */
#define R_ARM_TLS_DTPMOD32 17 /* ID of module containing symbol */
#define R_ARM_TLS_DTPOFF32 18 /* Offset in TLS block */
#define R_ARM_TLS_TPOFF32 19 /* Offset in static TLS block */
/* 20-31 are reserved for ARM Linux. */
#define R_ARM_COPY 20
#define R_ARM_GLOB_DAT 21
#define R_ARM_JUMP_SLOT 22
#define R_ARM_RELATIVE 23
#define R_ARM_GOTOFF 24
#define R_ARM_GOTPC 25
#define R_ARM_GOT32 26
#define R_ARM_PLT32 27
#define R_ARM_ALU_PCREL_7_0 32
#define R_ARM_ALU_PCREL_15_8 33
#define R_ARM_ALU_PCREL_23_15 34
#define R_ARM_ALU_SBREL_11_0 35
#define R_ARM_ALU_SBREL_19_12 36
#define R_ARM_ALU_SBREL_27_20 37
/* 96-111 are reserved to G++. */
#define R_ARM_GNU_VTENTRY 100
#define R_ARM_GNU_VTINHERIT 101
#define R_ARM_THM_PC11 102
#define R_ARM_THM_PC9 103
/* More TLS relocations */
#define R_ARM_TLS_GD32 104 /* PC-rel 32 bit for global dynamic */
#define R_ARM_TLS_LDM32 105 /* PC-rel 32 bit for local dynamic */
#define R_ARM_TLS_LDO32 106 /* 32 bit offset relative to TLS */
#define R_ARM_TLS_IE32 107 /* PC-rel 32 bit for GOT entry of */
#define R_ARM_TLS_LE32 108
#define R_ARM_TLS_LDO12 109
#define R_ARM_TLS_LE12 110
#define R_ARM_TLS_IE12GP 111
/* 112-127 are reserved for private experiments. */
#define R_ARM_RXPC25 249
#define R_ARM_RSBREL32 250
#define R_ARM_THM_RPC22 251
#define R_ARM_RREL32 252
#define R_ARM_RABS32 253
#define R_ARM_RPC24 254
#define R_ARM_RBASE 255
#define R_TYPE(name) __CONCAT(R_ARM_,name)
/* Processor specific program header flags */
#define PF_ARM_SB 0x10000000
#define PF_ARM_PI 0x20000000
#define PF_ARM_ENTRY 0x80000000
/* Processor specific section header flags */
#define SHF_ENTRYSECT 0x10000000
#define SHF_COMDEF 0x80000000
/* Processor specific symbol types */
#define STT_ARM_TFUNC STT_LOPROC
#ifdef _KERNEL
#ifdef ELFSIZE
#define ELF_MD_PROBE_FUNC ELFNAME2(arm_netbsd,probe)
#endif
struct exec_package;
int arm_netbsd_elf32_probe(struct lwp *, struct exec_package *, void *, char *,
vaddr_t *);
#endif
#endif /* _ARM_ELF_MACHDEP_H_ */

View File

@@ -1,225 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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 <machine/asm.h>
#include "libc_events.h"
.syntax unified
.thumb
.thumb_func
// Get the length of src string, then get the source of the dst string.
// Check that the two lengths together don't exceed the threshold, then
// do a memcpy of the data.
ENTRY(__strcat_chk)
.cfi_startproc
pld [r0, #0]
push {r0, lr}
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
push {r4, r5}
.save {r4, r5}
.cfi_adjust_cfa_offset 8
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
mov lr, r2
// Save the dst register to r5
mov r5, r0
// Zero out r4
eor r4, r4, r4
// r1 contains the address of the string to count.
.L_strlen_start:
mov r0, r1
ands r3, r1, #7
beq .L_mainloop
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq .L_align_to_32
ldrb r2, [r1], #1
cbz r2, .L_update_count_and_finish
.L_align_to_32:
bcc .L_align_to_64
ands ip, r3, #2
beq .L_align_to_64
ldrb r2, [r1], #1
cbz r2, .L_update_count_and_finish
ldrb r2, [r1], #1
cbz r2, .L_update_count_and_finish
.L_align_to_64:
tst r3, #4
beq .L_mainloop
ldr r3, [r1], #4
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
.p2align 2
.L_mainloop:
ldrd r2, r3, [r1], #8
pld [r1, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne .L_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
b .L_mainloop
.L_update_count_and_finish:
sub r3, r1, r0
sub r3, r3, #1
b .L_finish
.L_zero_in_first_register:
sub r3, r1, r0
lsls r2, ip, #17
bne .L_sub8_and_finish
bcs .L_sub7_and_finish
lsls ip, ip, #1
bne .L_sub6_and_finish
sub r3, r3, #5
b .L_finish
.L_sub8_and_finish:
sub r3, r3, #8
b .L_finish
.L_sub7_and_finish:
sub r3, r3, #7
b .L_finish
.L_sub6_and_finish:
sub r3, r3, #6
b .L_finish
.L_zero_in_second_register:
sub r3, r1, r0
lsls r2, ip, #17
bne .L_sub4_and_finish
bcs .L_sub3_and_finish
lsls ip, ip, #1
bne .L_sub2_and_finish
sub r3, r3, #1
b .L_finish
.L_sub4_and_finish:
sub r3, r3, #4
b .L_finish
.L_sub3_and_finish:
sub r3, r3, #3
b .L_finish
.L_sub2_and_finish:
sub r3, r3, #2
.L_finish:
cmp r4, #0
bne .L_strlen_done
// Time to get the dst string length.
mov r1, r5
// Save the original source address to r5.
mov r5, r0
// Save the current length (adding 1 for the terminator).
add r4, r3, #1
b .L_strlen_start
// r0 holds the pointer to the dst string.
// r3 holds the dst string length.
// r4 holds the src string length + 1.
.L_strlen_done:
add r2, r3, r4
cmp r2, lr
bhi __strcat_chk_failed
// Set up the registers for the memcpy code.
mov r1, r5
pld [r1, #64]
mov r2, r4
add r0, r0, r3
pop {r4, r5}
.cfi_endproc
END(__strcat_chk)
#define MEMCPY_BASE __strcat_chk_memcpy_base
#define MEMCPY_BASE_ALIGNED __strcat_chk_memcpy_base_aligned
#include "memcpy_base.S"
ENTRY(__strcat_chk_failed)
.cfi_startproc
.save {r0, lr}
.save {r4, r5}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
.cfi_adjust_cfa_offset 8
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_STRCAT_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+4)
.cfi_endproc
END(__strcat_chk_failed)
.data
error_string:
.string "strcat buffer overflow"

View File

@@ -1,187 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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 <machine/asm.h>
#include "libc_events.h"
.syntax unified
.thumb
.thumb_func
// Get the length of the source string first, then do a memcpy of the data
// instead of a strcpy.
ENTRY(__strcpy_chk)
.cfi_startproc
pld [r0, #0]
push {r0, lr}
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
mov lr, r2
mov r0, r1
ands r3, r1, #7
beq .L_mainloop
// Align to a double word (64 bits).
rsb r3, r3, #8
lsls ip, r3, #31
beq .L_align_to_32
ldrb r2, [r0], #1
cbz r2, .L_update_count_and_finish
.L_align_to_32:
bcc .L_align_to_64
ands ip, r3, #2
beq .L_align_to_64
ldrb r2, [r0], #1
cbz r2, .L_update_count_and_finish
ldrb r2, [r0], #1
cbz r2, .L_update_count_and_finish
.L_align_to_64:
tst r3, #4
beq .L_mainloop
ldr r3, [r0], #4
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
.p2align 2
.L_mainloop:
ldrd r2, r3, [r0], #8
pld [r0, #64]
sub ip, r2, #0x01010101
bic ip, ip, r2
ands ip, ip, #0x80808080
bne .L_zero_in_first_register
sub ip, r3, #0x01010101
bic ip, ip, r3
ands ip, ip, #0x80808080
bne .L_zero_in_second_register
b .L_mainloop
.L_update_count_and_finish:
sub r3, r0, r1
sub r3, r3, #1
b .L_check_size
.L_zero_in_first_register:
sub r3, r0, r1
lsls r2, ip, #17
bne .L_sub8_and_finish
bcs .L_sub7_and_finish
lsls ip, ip, #1
bne .L_sub6_and_finish
sub r3, r3, #5
b .L_check_size
.L_sub8_and_finish:
sub r3, r3, #8
b .L_check_size
.L_sub7_and_finish:
sub r3, r3, #7
b .L_check_size
.L_sub6_and_finish:
sub r3, r3, #6
b .L_check_size
.L_zero_in_second_register:
sub r3, r0, r1
lsls r2, ip, #17
bne .L_sub4_and_finish
bcs .L_sub3_and_finish
lsls ip, ip, #1
bne .L_sub2_and_finish
sub r3, r3, #1
b .L_check_size
.L_sub4_and_finish:
sub r3, r3, #4
b .L_check_size
.L_sub3_and_finish:
sub r3, r3, #3
b .L_check_size
.L_sub2_and_finish:
sub r3, r3, #2
.L_check_size:
pld [r1, #0]
pld [r1, #64]
ldr r0, [sp]
cmp r3, lr
bhs __strcpy_chk_failed
// Add 1 for copy length to get the string terminator.
add r2, r3, #1
.cfi_endproc
END(__strcpy_chk)
#define MEMCPY_BASE __strcpy_chk_memcpy_base
#define MEMCPY_BASE_ALIGNED __strcpy_chk_memcpy_base_aligned
#include "memcpy_base.S"
ENTRY(__strcpy_chk_failed)
.cfi_startproc
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_STRCPY_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+4)
.cfi_endproc
END(__strcpy_chk_failed)
.data
error_string:
.string "strcpy buffer overflow"

View File

@@ -1,92 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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.
*/
/* Assumes neon instructions and a cache line size of 32 bytes. */
#include <machine/asm.h>
#include "libc_events.h"
/*
* This code assumes it is running on a processor that supports all arm v7
* instructions, that supports neon instructions, and that has a 32 byte
* cache line.
*/
.text
.syntax unified
.fpu neon
.thumb
.thumb_func
ENTRY(__memcpy_chk)
.cfi_startproc
cmp r2, r3
bhi __memcpy_chk_fail
// Fall through to memcpy...
.cfi_endproc
END(__memcpy_chk)
ENTRY(memcpy)
.cfi_startproc
pld [r1, #64]
stmfd sp!, {r0, lr}
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
.cfi_endproc
END(memcpy)
#define MEMCPY_BASE __memcpy_base
#define MEMCPY_BASE_ALIGNED __memcpy_base_aligned
#include "memcpy_base.S"
ENTRY(__memcpy_chk_fail)
.cfi_startproc
// Preserve lr for backtrace.
push {lr}
.save {lr}
.cfi_def_cfa_offset 4
.cfi_rel_offset lr, 0
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_MEMCPY_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+4)
.cfi_endproc
END(__memcpy_chk_fail)
.data
error_string:
.string "memcpy buffer overflow"

View File

@@ -1,127 +0,0 @@
/*
* Copyright (C) 2013 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:
* * 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.
*
* 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.
*/
/*
* This code assumes it is running on a processor that supports all arm v7
* instructions, that supports neon instructions, and that has a 32 byte
* cache line.
*/
// Assumes neon instructions and a cache line size of 32 bytes.
ENTRY(MEMCPY_BASE)
.cfi_startproc
.save {r0, lr}
.cfi_def_cfa_offset 8
.cfi_rel_offset r0, 0
.cfi_rel_offset lr, 4
/* do we have at least 16-bytes to copy (needed for alignment below) */
cmp r2, #16
blo 5f
/* align destination to cache-line for the write-buffer */
rsb r3, r0, #0
ands r3, r3, #0xF
beq 2f
/* copy up to 15-bytes (count in r3) */
sub r2, r2, r3
movs ip, r3, lsl #31
itt mi
ldrbmi lr, [r1], #1
strbmi lr, [r0], #1
itttt cs
ldrbcs ip, [r1], #1
ldrbcs lr, [r1], #1
strbcs ip, [r0], #1
strbcs lr, [r0], #1
movs ip, r3, lsl #29
bge 1f
// copies 4 bytes, destination 32-bits aligned
vld4.8 {d0[0], d1[0], d2[0], d3[0]}, [r1]!
vst4.8 {d0[0], d1[0], d2[0], d3[0]}, [r0, :32]!
1: bcc 2f
// copies 8 bytes, destination 64-bits aligned
vld1.8 {d0}, [r1]!
vst1.8 {d0}, [r0, :64]!
2: /* make sure we have at least 64 bytes to copy */
subs r2, r2, #64
blo 2f
1: /* The main loop copies 64 bytes at a time */
vld1.8 {d0 - d3}, [r1]!
vld1.8 {d4 - d7}, [r1]!
pld [r1, #(32*8)]
subs r2, r2, #64
vst1.8 {d0 - d3}, [r0, :128]!
vst1.8 {d4 - d7}, [r0, :128]!
bhs 1b
2: /* fix-up the remaining count and make sure we have >= 32 bytes left */
adds r2, r2, #32
blo 4f
/* Copy 32 bytes. These cache lines were already preloaded */
vld1.8 {d0 - d3}, [r1]!
sub r2, r2, #32
vst1.8 {d0 - d3}, [r0, :128]!
4: /* less than 32 left */
add r2, r2, #32
tst r2, #0x10
beq 5f
// copies 16 bytes, 128-bits aligned
vld1.8 {d0, d1}, [r1]!
vst1.8 {d0, d1}, [r0, :128]!
5: /* copy up to 15-bytes (count in r2) */
movs ip, r2, lsl #29
bcc 1f
vld1.8 {d0}, [r1]!
vst1.8 {d0}, [r0]!
1: bge 2f
vld4.8 {d0[0], d1[0], d2[0], d3[0]}, [r1]!
vst4.8 {d0[0], d1[0], d2[0], d3[0]}, [r0]!
2: movs ip, r2, lsl #31
itt mi
ldrbmi r3, [r1], #1
strbmi r3, [r0], #1
itttt cs
ldrbcs ip, [r1], #1
ldrbcs lr, [r1], #1
strbcs ip, [r0], #1
strbcs lr, [r0], #1
ldmfd sp!, {r0, lr}
bx lr
.cfi_endproc
END(MEMCPY_BASE)

View File

@@ -1,119 +0,0 @@
/*
* 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:
* * 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.
*
* 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 <machine/cpu-features.h>
#include <machine/asm.h>
#include "libc_events.h"
/*
* This code assumes it is running on a processor that supports all arm v7
* instructions, that supports neon instructions, and that supports
* unaligned neon instruction accesses to memory.
*/
.fpu neon
ENTRY(__memset_chk)
.cfi_startproc
cmp r2, r3
bls .L_done
// Preserve lr for backtrace.
.save {lr}
push {lr}
.cfi_def_cfa_offset 4
.cfi_rel_offset lr, 0
ldr r0, error_message
ldr r1, error_code
1:
add r0, pc
bl __fortify_chk_fail
error_code:
.word BIONIC_EVENT_MEMSET_BUFFER_OVERFLOW
error_message:
.word error_string-(1b+8)
.cfi_endproc
END(__memset_chk)
ENTRY(bzero)
.cfi_startproc
mov r2, r1
mov r1, #0
.L_done:
// Fall through to memset...
.cfi_endproc
END(bzero)
/* memset() returns its first argument. */
ENTRY(memset)
.cfi_startproc
.save {r0}
stmfd sp!, {r0}
.cfi_def_cfa_offset 4
.cfi_rel_offset r0, 0
vdup.8 q0, r1
/* make sure we have at least 32 bytes to write */
subs r2, r2, #32
blo 2f
vmov q1, q0
1: /* The main loop writes 32 bytes at a time */
subs r2, r2, #32
vst1.8 {d0 - d3}, [r0]!
bhs 1b
2: /* less than 32 left */
add r2, r2, #32
tst r2, #0x10
beq 3f
// writes 16 bytes, 128-bits aligned
vst1.8 {d0, d1}, [r0]!
3: /* write up to 15-bytes (count in r2) */
movs ip, r2, lsl #29
bcc 1f
vst1.8 {d0}, [r0]!
1: bge 2f
vst1.32 {d0[0]}, [r0]!
2: movs ip, r2, lsl #31
strmib r1, [r0], #1
strcsb r1, [r0], #1
strcsb r1, [r0], #1
ldmfd sp!, {r0}
bx lr
.cfi_endproc
END(memset)
.data
error_string:
.string "memset buffer overflow"

View File

@@ -1,500 +0,0 @@
/*
* Copyright (c) 2013 ARM Ltd
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 3. The name of the company may not be used to endorse or promote
* products derived from this software without specific prior written
* permission.
*
* THIS SOFTWARE IS PROVIDED BY ARM LTD ``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 ARM LTD 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 <machine/cpu-features.h>
#include <machine/asm.h>
#ifdef __ARMEB__
#define S2LOMEM lsl
#define S2LOMEMEQ lsleq
#define S2HIMEM lsr
#define MSB 0x000000ff
#define LSB 0xff000000
#define BYTE0_OFFSET 24
#define BYTE1_OFFSET 16
#define BYTE2_OFFSET 8
#define BYTE3_OFFSET 0
#else /* not __ARMEB__ */
#define S2LOMEM lsr
#define S2LOMEMEQ lsreq
#define S2HIMEM lsl
#define BYTE0_OFFSET 0
#define BYTE1_OFFSET 8
#define BYTE2_OFFSET 16
#define BYTE3_OFFSET 24
#define MSB 0xff000000
#define LSB 0x000000ff
#endif /* not __ARMEB__ */
.syntax unified
#if defined (__thumb__)
.thumb
.thumb_func
#endif
ENTRY(strcmp)
/* Use LDRD whenever possible. */
/* The main thing to look out for when comparing large blocks is that
the loads do not cross a page boundary when loading past the index
of the byte with the first difference or the first string-terminator.
For example, if the strings are identical and the string-terminator
is at index k, byte by byte comparison will not load beyond address
s1+k and s2+k; word by word comparison may load up to 3 bytes beyond
k; double word - up to 7 bytes. If the load of these bytes crosses
a page boundary, it might cause a memory fault (if the page is not mapped)
that would not have happened in byte by byte comparison.
If an address is (double) word aligned, then a load of a (double) word
from that address will not cross a page boundary.
Therefore, the algorithm below considers word and double-word alignment
of strings separately. */
/* High-level description of the algorithm.
* The fast path: if both strings are double-word aligned,
use LDRD to load two words from each string in every loop iteration.
* If the strings have the same offset from a word boundary,
use LDRB to load and compare byte by byte until
the first string is aligned to a word boundary (at most 3 bytes).
This is optimized for quick return on short unaligned strings.
* If the strings have the same offset from a double-word boundary,
use LDRD to load two words from each string in every loop iteration, as in the fast path.
* If the strings do not have the same offset from a double-word boundary,
load a word from the second string before the loop to initialize the queue.
Use LDRD to load two words from every string in every loop iteration.
Inside the loop, load the second word from the second string only after comparing
the first word, using the queued value, to guarantee safety across page boundaries.
* If the strings do not have the same offset from a word boundary,
use LDR and a shift queue. Order of loads and comparisons matters,
similarly to the previous case.
* Use UADD8 and SEL to compare words, and use REV and CLZ to compute the return value.
* The only difference between ARM and Thumb modes is the use of CBZ instruction.
* The only difference between big and little endian is the use of REV in little endian
to compute the return value, instead of MOV.
*/
.macro m_cbz reg label
#ifdef __thumb2__
cbz \reg, \label
#else /* not defined __thumb2__ */
cmp \reg, #0
beq \label
#endif /* not defined __thumb2__ */
.endm /* m_cbz */
.macro m_cbnz reg label
#ifdef __thumb2__
cbnz \reg, \label
#else /* not defined __thumb2__ */
cmp \reg, #0
bne \label
#endif /* not defined __thumb2__ */
.endm /* m_cbnz */
.macro init
/* Macro to save temporary registers and prepare magic values. */
subs sp, sp, #16
.cfi_def_cfa_offset 16
strd r4, r5, [sp, #8]
.cfi_rel_offset r4, 0
.cfi_rel_offset r5, 4
strd r6, r7, [sp]
.cfi_rel_offset r6, 8
.cfi_rel_offset r7, 12
mvn r6, #0 /* all F */
mov r7, #0 /* all 0 */
.endm /* init */
.macro magic_compare_and_branch w1 w2 label
/* Macro to compare registers w1 and w2 and conditionally branch to label. */
cmp \w1, \w2 /* Are w1 and w2 the same? */
magic_find_zero_bytes \w1
it eq
cmpeq ip, #0 /* Is there a zero byte in w1? */
bne \label
.endm /* magic_compare_and_branch */
.macro magic_find_zero_bytes w1
/* Macro to find all-zero bytes in w1, result is in ip. */
#if (defined (__ARM_FEATURE_DSP))
uadd8 ip, \w1, r6
sel ip, r7, r6
#else /* not defined (__ARM_FEATURE_DSP) */
/* __ARM_FEATURE_DSP is not defined for some Cortex-M processors.
Coincidently, these processors only have Thumb-2 mode, where we can use the
the (large) magic constant available directly as an immediate in instructions.
Note that we cannot use the magic constant in ARM mode, where we need
to create the constant in a register. */
sub ip, \w1, #0x01010101
bic ip, ip, \w1
and ip, ip, #0x80808080
#endif /* not defined (__ARM_FEATURE_DSP) */
.endm /* magic_find_zero_bytes */
.macro setup_return w1 w2
#ifdef __ARMEB__
mov r1, \w1
mov r2, \w2
#else /* not __ARMEB__ */
rev r1, \w1
rev r2, \w2
#endif /* not __ARMEB__ */
.endm /* setup_return */
.cfi_startproc
pld [r0, #0]
pld [r1, #0]
/* Are both strings double-word aligned? */
orr ip, r0, r1
tst ip, #7
bne .L_do_align
/* Fast path. */
.save {r4-r7}
init
.L_doubleword_aligned:
/* Get here when the strings to compare are double-word aligned. */
/* Compare two words in every iteration. */
.p2align 2
2:
pld [r0, #16]
pld [r1, #16]
/* Load the next double-word from each string. */
ldrd r2, r3, [r0], #8
ldrd r4, r5, [r1], #8
magic_compare_and_branch w1=r2, w2=r4, label=.L_return_24
magic_compare_and_branch w1=r3, w2=r5, label=.L_return_35
b 2b
.L_do_align:
/* Is the first string word-aligned? */
ands ip, r0, #3
beq .L_word_aligned_r0
/* Fast compare byte by byte until the first string is word-aligned. */
/* The offset of r0 from a word boundary is in ip. Thus, the number of bytes
to read until the next word boundary is 4-ip. */
bic r0, r0, #3
ldr r2, [r0], #4
lsls ip, ip, #31
beq .L_byte2
bcs .L_byte3
.L_byte1:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE1_OFFSET
subs ip, r3, ip
bne .L_fast_return
m_cbz reg=r3, label=.L_fast_return
.L_byte2:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE2_OFFSET
subs ip, r3, ip
bne .L_fast_return
m_cbz reg=r3, label=.L_fast_return
.L_byte3:
ldrb ip, [r1], #1
uxtb r3, r2, ror #BYTE3_OFFSET
subs ip, r3, ip
bne .L_fast_return
m_cbnz reg=r3, label=.L_word_aligned_r0
.L_fast_return:
mov r0, ip
bx lr
.L_word_aligned_r0:
init
/* The first string is word-aligned. */
/* Is the second string word-aligned? */
ands ip, r1, #3
bne .L_strcmp_unaligned
.L_word_aligned:
/* The strings are word-aligned. */
/* Is the first string double-word aligned? */
tst r0, #4
beq .L_doubleword_aligned_r0
/* If r0 is not double-word aligned yet, align it by loading
and comparing the next word from each string. */
ldr r2, [r0], #4
ldr r4, [r1], #4
magic_compare_and_branch w1=r2 w2=r4 label=.L_return_24
.L_doubleword_aligned_r0:
/* Get here when r0 is double-word aligned. */
/* Is r1 doubleword_aligned? */
tst r1, #4
beq .L_doubleword_aligned
/* Get here when the strings to compare are word-aligned,
r0 is double-word aligned, but r1 is not double-word aligned. */
/* Initialize the queue. */
ldr r5, [r1], #4
/* Compare two words in every iteration. */
.p2align 2
3:
pld [r0, #16]
pld [r1, #16]
/* Load the next double-word from each string and compare. */
ldrd r2, r3, [r0], #8
magic_compare_and_branch w1=r2 w2=r5 label=.L_return_25
ldrd r4, r5, [r1], #8
magic_compare_and_branch w1=r3 w2=r4 label=.L_return_34
b 3b
.macro miscmp_word offsetlo offsethi
/* Macro to compare misaligned strings. */
/* r0, r1 are word-aligned, and at least one of the strings
is not double-word aligned. */
/* Compare one word in every loop iteration. */
/* OFFSETLO is the original bit-offset of r1 from a word-boundary,
OFFSETHI is 32 - OFFSETLO (i.e., offset from the next word). */
/* Initialize the shift queue. */
ldr r5, [r1], #4
/* Compare one word from each string in every loop iteration. */
.p2align 2
7:
ldr r3, [r0], #4
S2LOMEM r5, r5, #\offsetlo
magic_find_zero_bytes w1=r3
cmp r7, ip, S2HIMEM #\offsetlo
and r2, r3, r6, S2LOMEM #\offsetlo
it eq
cmpeq r2, r5
bne .L_return_25
ldr r5, [r1], #4
cmp ip, #0
eor r3, r2, r3
S2HIMEM r2, r5, #\offsethi
it eq
cmpeq r3, r2
bne .L_return_32
b 7b
.endm /* miscmp_word */
.L_strcmp_unaligned:
/* r0 is word-aligned, r1 is at offset ip from a word. */
/* Align r1 to the (previous) word-boundary. */
bic r1, r1, #3
/* Unaligned comparison word by word using LDRs. */
cmp ip, #2
beq .L_miscmp_word_16 /* If ip == 2. */
bge .L_miscmp_word_24 /* If ip == 3. */
miscmp_word offsetlo=8 offsethi=24 /* If ip == 1. */
.L_miscmp_word_24: miscmp_word offsetlo=24 offsethi=8
.L_return_32:
setup_return w1=r3, w2=r2
b .L_do_return
.L_return_34:
setup_return w1=r3, w2=r4
b .L_do_return
.L_return_25:
setup_return w1=r2, w2=r5
b .L_do_return
.L_return_35:
setup_return w1=r3, w2=r5
b .L_do_return
.L_return_24:
setup_return w1=r2, w2=r4
.L_do_return:
#ifdef __ARMEB__
mov r0, ip
#else /* not __ARMEB__ */
rev r0, ip
#endif /* not __ARMEB__ */
/* Restore temporaries early, before computing the return value. */
ldrd r6, r7, [sp]
ldrd r4, r5, [sp, #8]
adds sp, sp, #16
.cfi_def_cfa_offset 0
.cfi_restore r4
.cfi_restore r5
.cfi_restore r6
.cfi_restore r7
/* There is a zero or a different byte between r1 and r2. */
/* r0 contains a mask of all-zero bytes in r1. */
/* Using r0 and not ip here because cbz requires low register. */
m_cbz reg=r0, label=.L_compute_return_value
clz r0, r0
/* r0 contains the number of bits on the left of the first all-zero byte in r1. */
rsb r0, r0, #24
/* Here, r0 contains the number of bits on the right of the first all-zero byte in r1. */
lsr r1, r1, r0
lsr r2, r2, r0
.L_compute_return_value:
movs r0, #1
cmp r1, r2
/* The return value is computed as follows.
If r1>r2 then (C==1 and Z==0) and LS doesn't hold and r0 is #1 at return.
If r1<r2 then (C==0 and Z==0) and we execute SBC with carry_in=0,
which means r0:=r0-r0-1 and r0 is #-1 at return.
If r1=r2 then (C==1 and Z==1) and we execute SBC with carry_in=1,
which means r0:=r0-r0 and r0 is #0 at return.
(C==0 and Z==1) cannot happen because the carry bit is "not borrow". */
it ls
sbcls r0, r0, r0
bx lr
/* The code from the previous version of strcmp.S handles this
* particular case (the second string is 2 bytes off a word alignment)
* faster than any current version. In this very specific case, use the
* previous version. See bionic/libc/arch-arm/cortex-a15/bionic/strcmp.S
* for the unedited version of this code.
*/
.L_miscmp_word_16:
wp1 .req r0
wp2 .req r1
b1 .req r2
w1 .req r4
w2 .req r5
t1 .req ip
@ r3 is scratch
/* At this point, wp1 (r0) has already been word-aligned. */
2:
mov b1, #1
orr b1, b1, b1, lsl #8
orr b1, b1, b1, lsl #16
and t1, wp2, #3
bic wp2, wp2, #3
ldr w1, [wp1], #4
ldr w2, [wp2], #4
/* Critical inner Loop: Block with 2 bytes initial overlap */
.p2align 2
2:
S2HIMEM t1, w1, #16
sub r3, w1, b1
S2LOMEM t1, t1, #16
bic r3, r3, w1
cmp t1, w2, S2LOMEM #16
bne 4f
ands r3, r3, b1, lsl #7
it eq
ldreq w2, [wp2], #4
bne 5f
eor t1, t1, w1
cmp t1, w2, S2HIMEM #16
bne 6f
ldr w1, [wp1], #4
b 2b
5:
#ifdef __ARMEB__
/* The syndrome value may contain false ones if the string ends
* with the bytes 0x01 0x00
*/
tst w1, #0xff000000
it ne
tstne w1, #0x00ff0000
beq 7f
#else
lsls r3, r3, #16
bne 7f
#endif
ldrh w2, [wp2]
S2LOMEM t1, w1, #16
#ifdef __ARMEB__
lsl w2, w2, #16
#endif
b 8f
6:
S2HIMEM w2, w2, #16
S2LOMEM t1, w1, #16
4:
S2LOMEM w2, w2, #16
b 8f
7:
mov r0, #0
/* Restore registers and stack. */
ldrd r6, r7, [sp]
ldrd r4, r5, [sp, #8]
adds sp, sp, #16
.cfi_def_cfa_offset 0
.cfi_restore r4
.cfi_restore r5
.cfi_restore r6
.cfi_restore r7
bx lr
8:
and r2, t1, #LSB
and r0, w2, #LSB
cmp r0, #1
it cs
cmpcs r0, r2
itt eq
S2LOMEMEQ t1, t1, #8
S2LOMEMEQ w2, w2, #8
beq 8b
sub r0, r2, r0
/* Restore registers and stack. */
ldrd r6, r7, [sp]
ldrd r4, r5, [sp, #8]
adds sp, sp, #16
.cfi_def_cfa_offset 0
.cfi_restore r4
.cfi_restore r5
.cfi_restore r6
.cfi_restore r7
bx lr
.cfi_endproc
END(strcmp)

View File

@@ -1,11 +0,0 @@
$(call libc-add-cpu-variant-src,MEMCPY,arch-arm/krait/bionic/memcpy.S)
$(call libc-add-cpu-variant-src,MEMSET,arch-arm/krait/bionic/memset.S)
$(call libc-add-cpu-variant-src,STRCMP,arch-arm/krait/bionic/strcmp.S)
$(call libc-add-cpu-variant-src,__STRCAT_CHK,arch-arm/krait/bionic/__strcat_chk.S)
$(call libc-add-cpu-variant-src,__STRCPY_CHK,arch-arm/krait/bionic/__strcpy_chk.S)
# Use cortex-a15 versions of strcat/strcpy/strlen.
$(call libc-add-cpu-variant-src,STRCAT,arch-arm/cortex-a15/bionic/strcat.S)
$(call libc-add-cpu-variant-src,STRCPY,arch-arm/cortex-a15/bionic/strcpy.S)
$(call libc-add-cpu-variant-src,STRLEN,arch-arm/cortex-a15/bionic/strlen.S)
include bionic/libc/arch-arm/generic/generic.mk

View File

@@ -4,7 +4,6 @@ syscall_src += arch-arm/syscalls/_exit.S
syscall_src += arch-arm/syscalls/_exit_thread.S
syscall_src += arch-arm/syscalls/__fork.S
syscall_src += arch-arm/syscalls/__waitid.S
syscall_src += arch-arm/syscalls/wait4.S
syscall_src += arch-arm/syscalls/__sys_clone.S
syscall_src += arch-arm/syscalls/execve.S
syscall_src += arch-arm/syscalls/__setuid.S
@@ -126,8 +125,6 @@ syscall_src += arch-arm/syscalls/removexattr.S
syscall_src += arch-arm/syscalls/lremovexattr.S
syscall_src += arch-arm/syscalls/__statfs64.S
syscall_src += arch-arm/syscalls/unshare.S
syscall_src += arch-arm/syscalls/swapon.S
syscall_src += arch-arm/syscalls/swapoff.S
syscall_src += arch-arm/syscalls/pause.S
syscall_src += arch-arm/syscalls/gettimeofday.S
syscall_src += arch-arm/syscalls/settimeofday.S
@@ -146,9 +143,6 @@ syscall_src += arch-arm/syscalls/__timer_getoverrun.S
syscall_src += arch-arm/syscalls/__timer_delete.S
syscall_src += arch-arm/syscalls/utimes.S
syscall_src += arch-arm/syscalls/utimensat.S
syscall_src += arch-arm/syscalls/timerfd_create.S
syscall_src += arch-arm/syscalls/timerfd_settime.S
syscall_src += arch-arm/syscalls/timerfd_gettime.S
syscall_src += arch-arm/syscalls/sigaction.S
syscall_src += arch-arm/syscalls/sigprocmask.S
syscall_src += arch-arm/syscalls/__sigsuspend.S
@@ -156,7 +150,6 @@ syscall_src += arch-arm/syscalls/__rt_sigaction.S
syscall_src += arch-arm/syscalls/__rt_sigprocmask.S
syscall_src += arch-arm/syscalls/__rt_sigtimedwait.S
syscall_src += arch-arm/syscalls/sigpending.S
syscall_src += arch-arm/syscalls/signalfd4.S
syscall_src += arch-arm/syscalls/socket.S
syscall_src += arch-arm/syscalls/socketpair.S
syscall_src += arch-arm/syscalls/bind.S
@@ -186,6 +179,7 @@ syscall_src += arch-arm/syscalls/__getcpu.S
syscall_src += arch-arm/syscalls/ioprio_set.S
syscall_src += arch-arm/syscalls/ioprio_get.S
syscall_src += arch-arm/syscalls/uname.S
syscall_src += arch-arm/syscalls/__wait4.S
syscall_src += arch-arm/syscalls/umask.S
syscall_src += arch-arm/syscalls/__reboot.S
syscall_src += arch-arm/syscalls/__syslog.S

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__brk)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_brk
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__brk)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__fcntl)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_fcntl
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__fcntl)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__fcntl64)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_fcntl64
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__fcntl64)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__fork)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_fork
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__fork)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__fstatfs64)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_fstatfs64
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__fstatfs64)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__getcpu)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_getcpu
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__getcpu)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__getcwd)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_getcwd
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__getcwd)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__getpriority)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_getpriority
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__getpriority)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__ioctl)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_ioctl
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__ioctl)

View File

@@ -1,7 +1,6 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__llseek)
mov ip, sp
@@ -11,8 +10,7 @@ ENTRY(__llseek)
ldr r7, =__NR__llseek
swi #0
ldmfd sp!, {r4, r5, r6, r7}
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__llseek)

View File

@@ -1,7 +1,6 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__mmap2)
mov ip, sp
@@ -11,8 +10,7 @@ ENTRY(__mmap2)
ldr r7, =__NR_mmap2
swi #0
ldmfd sp!, {r4, r5, r6, r7}
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__mmap2)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__open)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_open
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__open)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__openat)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_openat
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__openat)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__ptrace)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_ptrace
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__ptrace)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__reboot)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_reboot
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__reboot)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__rt_sigaction)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_rt_sigaction
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__rt_sigaction)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__rt_sigprocmask)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_rt_sigprocmask
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__rt_sigprocmask)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__rt_sigtimedwait)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_rt_sigtimedwait
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__rt_sigtimedwait)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__sched_getaffinity)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_sched_getaffinity
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__sched_getaffinity)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__set_tls)
mov ip, r7
ldr r7, =__ARM_NR_set_tls
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_ARM_set_tls
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__set_tls)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__setresuid)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_setresuid32
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__setresuid)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__setreuid)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_setreuid32
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__setreuid)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__setuid)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_setuid32
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__setuid)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__sigsuspend)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_sigsuspend
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__sigsuspend)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__statfs64)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_statfs64
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__statfs64)

View File

@@ -1,7 +1,6 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__sys_clone)
mov ip, sp
@@ -11,8 +10,7 @@ ENTRY(__sys_clone)
ldr r7, =__NR_clone
swi #0
ldmfd sp!, {r4, r5, r6, r7}
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__sys_clone)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__syslog)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_syslog
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__syslog)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__timer_create)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_timer_create
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__timer_create)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__timer_delete)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_timer_delete
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__timer_delete)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__timer_getoverrun)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_timer_getoverrun
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__timer_getoverrun)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__timer_gettime)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_timer_gettime
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__timer_gettime)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__timer_settime)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_timer_settime
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__timer_settime)

View File

@@ -0,0 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__wait4)
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_wait4
swi #0
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__wait4)

View File

@@ -1,7 +1,6 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(__waitid)
mov ip, sp
@@ -11,8 +10,7 @@ ENTRY(__waitid)
ldr r7, =__NR_waitid
swi #0
ldmfd sp!, {r4, r5, r6, r7}
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
movs r0, r0
bxpl lr
b __set_syscall_errno
END(__waitid)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(_exit)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_exit_group
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(_exit)

View File

@@ -1,15 +1,14 @@
/* autogenerated by gensyscalls.py */
#include <asm/unistd.h>
#include <linux/err.h>
#include <machine/asm.h>
#include <sys/linux-syscalls.h>
ENTRY(_exit_thread)
mov ip, r7
.save {r4, r7}
stmfd sp!, {r4, r7}
ldr r7, =__NR_exit
swi #0
mov r7, ip
cmn r0, #(MAX_ERRNO + 1)
bxls lr
neg r0, r0
b __set_errno
ldmfd sp!, {r4, r7}
movs r0, r0
bxpl lr
b __set_syscall_errno
END(_exit_thread)

Some files were not shown because too many files have changed in this diff Show More