Notable changes since 3.1.0
26ae365: fix possible socks5 exchange corruption caused by bad
state transition when parsing responses
f19a96e: fix segfault when presented with config {"mode": null}
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Notable changes since 3.0.8
ede744a: depends on libcares now instead of libudns
1c64829: new cmdline option --no-delay for not turning off TCP_NODELAY
9201619: ss-local: check if client supports socks5 protocol and no-auth-required method
f8283fc: Fix potential buffer overflow when parsing json config
380fddb: redir: fix conversion from DSCP to ToS
The two patches are now in the offical repo
ipset command line utility supports ranges of address: IP-IP, but the
dash character is also valid character in host names. If we have a
remote server ss-00.example.com, ipset may complain that
ipset v6.32: Syntax error: cannot parse ss: resolving to IPv4 address failed
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
ubox 'list' type is for validating multiple elements separated by
tabs/whitespaces in a single value. E.g. The following should not be
accepted
list src_ip_bypass '1.2.3.4 4.3.2.1'
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
- New UCI options ifnames, dst_default
- UCI options src_ips_xxx now accept cidr as their values
- Export ipset names as part of the interface so that it can be
depended on and used by other programs
- Bypass only remote servers used ss-redir instances, so that it's
possible to let other servers to go through existing re-redir
instances
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
This is the default value taken by ss-server and ss-redir. After this
change ss_rules section can still use those ss-redir instances who do
not have mode explicitly specified.
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
- It's a common practice that assert should be ignored in release build
- Whether to enable ssp should be decided by the config of build system
This was taken from Makefile in shadowsocks/openwrt-shadowsocks.
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Notable changes since 3.0.6
f308dde ss-server: fix a use-after-free bug
0b2dce5 ss-redir: fix a mem leak
b7bdb16 ss-local: SOCKS5 UDP associate terminates the connection prematurely
3f0d39a ss-local: use getsockname udp_fd where it applies (fixes local_port==0)
eb30a3d fix possible data loss with salsa20 cipher
0559d8c fix partial nonce data being overwritten
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
The change is mainly for keeping compatibility with old validate_data
before ubox commit ac481cdd999ee84d3f115c33a56397237e95ec64 in Sat Jul
16 14:52:36 2016 +0200. The behaviour change comes with that commit can
be seen with the following command line session
root@LEDE:/usr/bin# validate_data network interface wan 'disabled:bool:false'
network.wan.disabled is unset and defaults to bool false
disabled=0; root@LEDE:/usr/bin#
root@OpenWrt:/# validate_data network interface lan 'disabled:bool:false'
disabled='false'; root@OpenWrt:/#
This will cause shadowsocks-libev in current master branch fail on OpenWrt
15.01 though they actually should only use packages from the 15.01 branch...
Fixesopenwrt/packages#4614
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
- Selecting only a single or subset of all components of shadowsocks-libev is
now possible (this is the main motivation behind the rewrite)
- Configuring multiple instances of the same component is now also possible
- Same option names as with the json config
- Unified configuration generation method for each component
- Add support for ss-local, ss-tunnel, ss-server
- Most data validation is now done with validate_data
- USE_PROCD=1
- Update ss-rules with the one from shadowsocks/luci-app-shadowsocks
- Add README.md
- Set myself as the maintainer
Addresses #4435
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
- use exec directly to eliminate a level in the process tree
- use "$@" instead of "$*" to pass arguments to openconnect
According to openconnect(8), openconnect will call vpnc-script to
cleanup before quit when it received SIGINT(2) and will quit immediately
when it received SIGTERM (the default signal by kill command)
Before and after the change, openconnect process will be killed first
with SIGINT sent from netifd. This was decided by the
'proto_kill_command "$config" 2' notify call in the proto script.
SIGKILL is the only other signal that can be sent from netifd when the
process did not quit on SIGINT on time. There should be no need to trap
on signal 1 3 6 9 (HUP QUIT ABRT KILL)
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
According to openconnect --help output:
-m, --mtu=MTU Request MTU from server
--base-mtu=MTU Indicate path MTU to/from server
Fixes#2099 by allowing setting tunnel mtu
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
It was introduced with 41f8d5465 ("openconnect: fix a couple of minor
things and add an interface option") and not needed since 4083de9d7
("openconnect: use proto_add_host_dependency")
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Below are changes in the packaging method
- Use tarball provided by the upstream project maintainer instead of
cloning the whole git-repo
- Drop openssl variant as it was deprecated then removed by the
upstream project
- Add dependency on libev, libsodium, libudns as they are not bundled
with the source code anymore
Addresses issue #4191
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
At the moment, only build these softmmu emulators for x86_64 and sunxi
target. The decision was made for the following reasons
- It seems that interests of virtualization with qemu are mostly from
x86, and ARM recently.
- x86, sunxi boards/boxes capable of running qemu with accel=kvm are more
widely available
- Not all host, target combinations of qemu works, or even compiles
- Extra maintenance work and server resources
Test results are as the following
- Nested vmx works: lede-qemu-x86_64-kvm on lede-qemu-x86_64-kvm
- KVM on Cubieboard2 works
- tcg with malta works: lede-qemu-malta-tcg on lede-qemu-malta-tcg.
But it's too slow to be useful thus not included in this version
- mips64 host does not compile
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
This is caused by a toolchain bug and has been reported to GCC a
while a ago without receiving any responses yet [1]. Fix it at the
moment by not compiling with mips16 enabled.
Should fix#1750 and #1904.
[1] Bug 71519 - "Out of range operand" bteqz inst generated by
"casesi_internal_mips16_<mode>",
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71519
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
This has two effects
- Saves time and computing resources. Non-relevant tools like
qemu-img, qemu-io will not be built
- Fixes packaging failure on octeon target caused by failure of
building internal pixman library
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
$(CP) was defined as `cp -fpR' since the very begining of OpenWrt build
system (2006-06-22). The -R option should be enough and base packages
use only $(CP) for the same purposes just fine and BSD manual of cp also
discourages the use of `-r' option. So let's just tidy up the usage now.
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
Yousong Zhou