OpenSSL is built with the generic linux settings for most targets,
including aarch64. These generic settings are designed for 32-bit CPU and
provide no assembler optmization: this is widely suboptimal for aarch64.
This patch simply switches to the aarch64 settings that are already
available in OpenSSL.
Here is the output of "openssl speed" before the optimization, with
"(...)" representing build flags that didn't change:
OpenSSL 1.0.2l 25 May 2017
options:bn(64,32) rc4(ptr,char) des(idx,cisc,2,int) aes(partial) blowfish(ptr)
compiler: aarch64-openwrt-linux-musl-gcc (...)
And after this patch, OpenSSL uses 64 bit mode and assembler optimizations:
OpenSSL 1.0.2l 25 May 2017
options:bn(64,64) rc4(ptr,char) des(idx,cisc,2,int) aes(partial) blowfish(ptr)
compiler: aarch64-openwrt-linux-musl-gcc (...) -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM
Here are some benchmarks on a pine64+ running latest LEDE master r5142-20d363aed3:
before# openssl speed sha aes blowfish
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
sha1 3918.89k 9982.43k 19148.03k 24933.03k 27325.78k
sha256 4604.51k 10240.64k 17472.51k 21355.18k 22801.07k
sha512 3662.19k 14539.41k 21443.16k 29544.11k 33177.60k
blowfish cbc 16266.63k 16940.86k 17176.92k 17237.33k 17252.35k
aes-128 cbc 19712.95k 21447.40k 22091.09k 22258.35k 22304.09k
aes-192 cbc 17680.12k 19064.47k 19572.14k 19703.13k 19737.26k
aes-256 cbc 15986.67k 17132.48k 17537.28k 17657.17k 17689.26k
after# openssl speed sha aes blowfish
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
sha1 6770.87k 26172.80k 86878.38k 205649.58k 345978.20k
sha256 20913.93k 74663.85k 184658.18k 290891.09k 351032.66k
sha512 7633.10k 30110.14k 50083.24k 71883.43k 82485.25k
blowfish cbc 16224.93k 16933.55k 17173.76k 17234.94k 17252.35k
aes-128 cbc 19425.74k 21193.31k 22065.74k 22304.77k 22380.54k
aes-192 cbc 17452.29k 18883.84k 19536.90k 19741.70k 19800.06k
aes-256 cbc 15815.89k 17003.01k 17530.03k 17695.40k 17746.60k
For some reason AES and blowfish do not benefit, but SHA performance
improves between 1.7x and 15x. SHA256 clearly benefits the most from the
optimization (4.5x on small blocks, 15x on large blocks!).
When using EVP (with "openssl speed -evp <algo>"):
# Before, EVP mode
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
sha1 3824.46k 10049.66k 19170.56k 24947.03k 27325.78k
sha256 3368.33k 8511.15k 16061.44k 20772.52k 22721.88k
sha512 2845.23k 11381.57k 19467.69k 28512.26k 33008.30k
bf-cbc 15146.74k 16623.83k 17092.01k 17211.39k 17249.62k
aes-128-cbc 17873.03k 20870.61k 21933.65k 22216.36k 22301.35k
aes-192-cbc 16184.18k 18607.15k 19447.13k 19670.02k 19737.26k
aes-256-cbc 14774.06k 16757.25k 17457.58k 17639.42k 17686.53k
# After, EVP mode
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
sha1 7056.97k 27142.10k 89515.86k 209155.41k 347419.99k
sha256 7745.70k 29750.06k 95341.48k 211001.69k 332376.75k
sha512 4550.47k 18086.06k 39997.10k 65880.75k 81431.21k
bf-cbc 15129.20k 16619.03k 17090.56k 17212.76k 17246.89k
aes-128-cbc 99619.74k 269032.34k 450214.23k 567353.00k 613933.06k
aes-192-cbc 93180.74k 231017.79k 361766.66k 433671.51k 461731.16k
aes-256-cbc 89343.23k 209858.58k 310160.04k 362234.88k 380878.85k
Blowfish does not seem to have assembler optimization at all, and SHA
still benefits (between 1.6x and 14.5x) but is generally slower than in
non-EVP mode.
However, AES performance is improved between 5.5x and 27.5x, which is
really impressive! For aes-128-cbc on large blocks, a core i7-6600U
@2.60GHz is only twice as fast...
Signed-off-by: Baptiste Jonglez <git@bitsofnetworks.org>
Changes in v1.27.0 :
build: Fixed accidental compiler flags concatenation for MSVC (Patch from LazyHamster) (GH-1029)
build: Reduce libxml2 version requirement to 2.6.26 (Patch from Mike Lothian) (GH-1020)
asio: Support for Windows / MinGW (Patch from Daniel Evers) (GH-1027)
h2load: Print out h2 header fields with --verbose option (GH-1015)
nghttpx: Send non-final response to HTTP/1.1 or HTTP/2 client only (GH-1016)
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
Due a compiler bug on ARM targets
( https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64516 )
unaligned access was disabled on all targets other than i386 and
x86_64 with commit 061319ec3d .
A fix has been added to lzo-2.09 so it is not necessary to disable
unaligned access within the Makefile anymore.
Signed-off-by: Stefan Oberhumer <stefan@obssys.com>
In order to build conntrack-tools from git, a newer version of
libnetfilter_conntrack is required. As 1.0.6 is currently the latest
release, switch to git.
b0a7cf7 include: expose a copy of nf_conntrack_common.h
f68f7b3 conntrack: fix missing break in setobjopt_undo_dnat()
79dac5a conntrack: revert getobjopt_is_nat() condition
b266523 libnetfilter_conntrack: bump version to 1.0.7
e870432 labels: don't crash on NULL labelmap
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
b39cac7 src: Correct typo in the location of internal.h in #include
58cb066 src: Declare the define visibility attribute together
e84b559 Revert "src: Declare the define visibility attribute together"
003c2b1 examples: set dummy connmark value to show use of NFQA_CT nested attribute
63973da doc: extend the doxygen section about NFQA_CFG_F_GSO
d7f74c7 build: bump version to 1.0.3
3f9eb57 build: bump library release version too
601abd1 doc: Add information about retrieving UID/GID/SECCTX fields
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
As git.netfilter.org seems to support HTTPS, use that instead of HTTP
which is insecure, or GIT which is blocked on many corporate networks.
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
The nghttp2 library is an implementation of the Hypertext Transfer
Protocol version 2 in C; it supports RFC7540 and RFC7541.
The package enables only the reusable C library; binary size is 130K (X86)
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
632688e utils: nuke bitfield functions and macros
f714be1 uloop: make SIGCHLD signal handling optional
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
Some symbols have been renamed.
Some are default enabled/disabled, so we need
to adjust semantics against that.
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
This seems to cause a false-positive warning/error
while building `libwebsockets-cyassl`.
```
make[6]: Leaving directory '/home/sandu/work/lede/build_dir/target-x86_64_musl/libwebsockets-cyassl/libwebsockets-2.2.1'
make[6]: Entering directory '/home/sandu/work/lede/build_dir/target-x86_64_musl/libwebsockets-cyassl/libwebsockets-2.2.1'
[ 2%] Building C object CMakeFiles/websockets.dir/lib/base64-decode.c.o
In file included from /home/sandu/work/lede/staging_dir/target-x86_64_musl/usr/include/wolfssl/ssl.h:31:0,
from /home/sandu/work/lede/staging_dir/target-x86_64_musl/usr/include/cyassl/ssl.h:33,
from /home/sandu/work/lede/staging_dir/target-x86_64_musl/usr/include/cyassl/openssl/ssl.h:30,
from /home/sandu/work/lede/build_dir/target-x86_64_musl/libwebsockets-cyassl/libwebsockets-2.2.1/lib/private-libwebsockets.h:256,
from /home/sandu/work/lede/build_dir/target-x86_64_musl/libwebsockets-cyassl/libwebsockets-2.2.1/lib/base64-decode.c:43:
/home/sandu/work/lede/staging_dir/target-x86_64_musl/usr/include/wolfssl/wolfcrypt/settings.h:1642:14: error: #warning "For timing resistance / side-channel attack prevention consider using harden options" [-Werror=cpp]
#warning "For timing resistance / side-channel attack prevention consider using harden options"
```
Hardening is enabled by default in libwolfssl at build-time.
However, the `settings.h` header is exported (along with other headers)
for build (via Build/InstallDev).
This looks like a small bug/issue with wolfssl.
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
This is to eliminate any ambiguity about the cyassl/wolfssl lib.
The rename happened some time ago (~3+ years).
As time goes by, people will start to forget cyassl and
start to get confused about the wolfSSL vs cyassl thing.
It's a good idea to keep up with the times (moving forward).
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
Until other packages from feeds decide to rename the
dependency of `+libcyassl` to `+libwolfssl`, this allows
for a bit of backwards compatibility with those packages.
Signed-off-by: Alexandru Ardelean <ardeleanalex@gmail.com>
Fixed an authentication bypass issue in SSL/TLS. When the TLS
authentication mode was set to 'optional',
mbedtls_ssl_get_verify_result() would incorrectly return 0 when the
peer's X.509 certificate chain had more than
MBEDTLS_X509_MAX_INTERMEDIATE_CA intermediates (default: 8), even when
it was not trusted. This could be triggered remotely on both the client
and server side. (Note, with the authentication mode set by
mbedtls_ssl_conf_authmode()to be 'required' (the default), the handshake
was correctly aborted).
Signed-off-by: Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk>
Tested-by: Magnus Kroken <mkroken@gmail.com>
Fixes some security issues (no remote exploits), and introduces
some changes. See release notes for details:
https://tls.mbed.org/tech-updates/releases/mbedtls-2.5.1-2.1.8-and-1.3.20-released
* Fixes an unlimited overread of heap-based buffers in mbedtls_ssl_read()
* Adds exponent blinding to RSA private operations
* Wipes stack buffers in RSA private key operations (rsa_rsaes_pkcs1_v15_decrypt(), rsa_rsaes_oaep_decrypt())
* Removes SHA-1 and RIPEMD-160 from the default hash algorithms for certificate verification.
* Fixes offset in FALLBACK_SCSV parsing that caused TLS server to fail to detect it sometimes.
* Tighten parsing of RSA PKCS#1 v1.5 signatures, to avoid a potential Bleichenbacher/BERserk-style attack.
Signed-off-by: Magnus Kroken <mkroken@gmail.com>
elfutils turns on -Werror by default, and patch 100-musl-compat.patch
changes how strerror_r is used and we no longer use the function's
return value. This causes the following build error/warning to occur
with glibc-based toolchains:
dwfl_error.c: In function 'dwfl_errmsg':
dwfl_error.c:158:18: error: ignoring return value of 'strerror_r',
declared with attribute warn_unused_result [-Werror=unused-result]
strerror_r (error & 0xffff, s, sizeof(s));
^
cc1: all warnings being treated as errors
Fixing this would be tricky as there are two possible signatures for
strerror_r (XSI and GNU), just turn off unused-result warnings instead.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Addresses CVE-2015-3239: Off-by-one error in the dwarf_to_unw_regnum
function in include/dwarf_i.h in libunwind 1.1 allows local users to
have unspecified impact via invalid dwarf opcodes.
Upstream stable-v1.2 fixed the missing unwind_i.h issue but no new
tarball is released yet
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Some external toolchains may be configured to enable OpenMP. Provide a
package for these libraries which can be used by other packages.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Add mirror and use main site as last resort.
Source: http://www.tcpdump.org/mirrors.html
Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
Refresh mirror list, some doesn't offer OpenSSL and add main site as last resort.
Source: https://www.openssl.org/source/mirror.html
Signed-off-by: Daniel Engberg <daniel.engberg.lists@pyret.net>
uClibc doesn't implement strerror_l() and thus libnl starting from
3.2.29 couldn't be compiled with it any longer, see
6c2d111177
To work-around that problem we'll just do a check on strerror_l()
availability during configuration and if it's not there just fall back
to locale-less strerror().
Patch for libnl is alreadfy merged upstream, see
e15966ac7f
and once the next libnl release happens this one must be removed from
Lede/OpenWrt.
Signed-off-by: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
Cc: Felix Fietkau <nbd@nbd.name>
Cc: John Crispin <john@phrozen.org>
Cc: Daniel Engberg <daniel.engberg.lists@pyret.net>
musl provides a /lib/libc.so file which should be integrated into the libc
package when the external toolchain with musl is used.
Signed-off-by: Hauke Mehrtens <hauke.mehrtens@intel.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Jo-Philipp Wich <jo@mein.io>
This fixes the following security problems:
* CVE-2017-2784: Freeing of memory allocated on stack when validating a public key with a secp224k1 curve
* SLOTH vulnerability
* Denial of Service through Certificate Revocation List
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
We need to let external toolchains be able to specify the path and
specification file to the libthread-db POSIX thread debugging shared
libraries.
This fixes GDB not being able to be installed because it is depending on
libthread-db:
Collected errors:
* satisfy_dependencies_for: Cannot satisfy the following dependencies
* for gdb:
* libthread-db *
* opkg_install_cmd: Cannot install package gdb.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
The current way of creating a STAMP_CONFIGURED filename for OpenSSL can
lead to an extremely long filename that makes touch unable to create it,
and fail the build.
Use mkhash to produce a hash against OPENSSL_OPTIONS which creates a
shortert stamp file,
Fixes#572
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Important change was made in 1.0.18: all sub-libs were merged
in one and only libc similarly to musl.
See [1] for more details.
To support that we had to remove refences to those sub-libs like
libpthread, libcrypt, libdl, libm, libutil etc.
[1] http://cgit.uclibc-ng.org/cgi/cgit/uclibc-ng.git/commit/?id=29ff9055c80efe77a7130767a9fcb3ab8c67e8ce
Signed-off-by: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
When calling erase() on a containers derived from __base_associative
(e.g. multimap) and providing a pair of iterators a segfault will
occur.
Example code to reproduce:
typedef std::multimap<int, int> testmap;
testmap t;
t.insert(std::pair<int, int>(1, 1));
t.insert(std::pair<int, int>(2, 1));
t.insert(std::pair<int, int>(3, 1));
t.erase(t.begin(), t.end());
Signed-off-by: Ben Kelly <ben@benjii.net>
Adds the following changes:
de3f14b uloop: add uloop_cancelling function
3b6181b utils: fix build on Mac OS X 10.12
7f671b1 blobmsg: add support for double
0fe1374 utils: add helper functions useful for allocating a ring buffer
8fc1c30 libubox: replace strtok with _r version.
4a9f74f libubox: allow reading out the pid of uloop process in lua
372e1e6 uloop: remove useless epoll data assignment
f9db1cb libubox: allow reading out the remaining time of a uloop timer in Lua
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Some toolchains will produce executables with an interpreter that is e.g:
ld.so.1 (typically a symbolic link). Due to our current LIBC_SPEC_FILE value,
we would not be able to copy this symbolic link/file over to the rootfs and
executables would fail to load. Extend the search pattern to include all
ld*.so* files that could be needed.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
This fixes the following security problems:
CVE-2017-3731: Truncated packet could crash via OOB read
CVE-2017-3732: BN_mod_exp may produce incorrect results on x86_64
CVE-2016-7055: Montgomery multiplication may produce incorrect results
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
