This _partially_ reverts commit
d2ba8ea54a4089959afdeecdd963e3c4ff391748, but for UDP sockets only.
With TCP sockets (and other things which use `rsock_init_inetsock`), the
order of operations is to call `getaddrinfo(3)` with AF_UNSPEC, look at
the returned addresses, pick one, and then call `socket(2)` with the
family for that address (i.e. AF_INET or AF_INET6).
With UDP sockets, however, this is reversed; `UDPSocket.new` takes an
address family as an argument, and then calls `socket(2)` with that
family. A subsequent call to UDPSocket#connect will then call
`getaddrinfo(3)` with that family.
The problem here is that...
* If you are in a networking situation that _only_ has loopback addrs,
* And you want to look up a name like "localhost" (or NULL)
* And you pass AF_INET or AF_INET6 as the ai_family argument to
getaddrinfo(3),
* And you pass AI_ADDRCONFIG to the hints argument as well,
then glibc on Linux will not return an address. This is because
AI_ADDRCONFIG is supposed to return addresses for families we actually
have an address for and could conceivably connect to, but also is
documented to explicitly ignore localhost in that situation.
It honestly doesn't make a ton of sense to pass AI_ADDRCONFIG if you're
explicitly passing the address family anyway, because you're not looking
for "an address for this name we can connect to"; you're looking for "an
IPv(4|6) address for this name". And the original glibc bug that
d2ba8ea5 was supposed to work around was related to parallel issuance of
A and AAAA queries, which of course won't happen if an address family is
explicitly specified.
So, we fix this by not passing AI_ADDRCONFIG for calls to
`rsock_addrinfo` that we also pass an explicit family to (i.e. for
UDPsocket).
[Bug #20048]
When making an outgoing TCP or UDP connection, set AI_ADDRCONFIG in the
hints we send to getaddrinfo(3) (if supported). This will prompt the
resolver to _NOT_ issue A or AAAA queries if the system does not
actually have an IPv4 or IPv6 address (respectively).
This makes outgoing connections marginally more efficient on
non-dual-stack systems, since we don't have to try connecting to an
address which can't possibly work.
More importantly, however, this works around a race condition present
in some older versions of glibc on aarch64 where it could accidently
send the two outgoing DNS queries with the same DNS txnid, and get
confused when receiving the responses. This manifests as outgoing
connections sometimes taking 5 seconds (the DNS timeout before retry) to
be made.
Fixes#19144
After 5e86b005c0f2ef30df2f9906c7e2f3abefe286a2, I now think ANYARGS is
dangerous and should be extinct. This commit deletes ANYARGS from
rb_ensure, which also revealed many arity / type mismatches.
* ext/socket/init.c (rsock_s_recvfrom_nonblock):
avoid arg parsing with C API
[ruby-core:71439] [Feature #11339]
* ext/socket/basicsocket.c (bsock_recv_nonblock):
adjust for above change, make private
* ext/socket/socket.c (sock_recvfrom_nonblock): ditto
* ext/socket/udpsocket.c (udp_recvfrom_nonblock): ditto
* ext/socket/lib/socket.rb (BasicSocket#recv_nonblock):
new wrapper for private method, move RDoc
(Socket#recvfrom_nonblock): ditto
(UDPSocket#recvfrom_nonblock): ditto
Note, not adding bm_recv_nonblock.rb to benchmark/ directory
since it is non-portable. It is only in this commit message.
Benchmark results + code
target 0: a (ruby 2.3.0dev (2015-11-12 trunk 52540) [x86_64-linux])
target 1: b (ruby 2.3.0dev (2015-11-12 avoid-kwarg-capi 52540) [x86_64-linux]
-----------------------------------------------------------
recv_nonblock
require 'socket'
nr = 1000000
msg = 'hello world'
buf = ''
size = msg.bytesize
UNIXSocket.pair(:SEQPACKET) do |a, b|
nr.times do
a.sendmsg(msg)
b.recv_nonblock(size, 0, buf, exception: false)
end
end
-----------------------------------------------------------
raw data:
[["recv_nonblock",
[[1.83511221408844,
1.8703329525887966,
1.8448856547474861,
1.859263762831688,
1.8331583738327026],
[1.5637447573244572,
1.4062932096421719,
1.4247371144592762,
1.4108827747404575,
1.4802536629140377]]]]
Elapsed time: 16.530452496 (sec)
-----------------------------------------------------------
benchmark results:
minimum results in each 5 measurements.
Execution time (sec)
name a b
recv_nonblock 1.833 1.406
Speedup ratio: compare with the result of `a' (greater is better)
name b
recv_nonblock 1.304
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@52598 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* ext/socket/udpsocket.c (udp_connect, udp_bind): check if the
socket is opened once before retreiving address infos.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@52111 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
As documented before, exceptions are expensive and IO::Wait*able are too
common in socket applications to be the exceptional case. Datagram
sockets deserve the same API which stream sockets are allowed with
read_nonblock and write_nonblock.
Note: this does not offer a performance advantage under optimal
conditions when both ends are equally matched in speed, but it it
does make debug output cleaner by avoiding exceptions whenever
the receiver slows down.
* ext/socket/ancdata.c (bsock_sendmsg_internal, bsock_recvmsg_internal):
support "exception: false" kwarg
* ext/socket/init.c (rsock_s_recvfrom_nonblock):
ditto
* ext/socket/init.c (rsock_s_recvfrom_nonblock): use rsock_opt_false_p
* ext/socket/socket.c (sock_connect_nonblock): ditto
* ext/socket/rubysocket.h (rsock_opt_false_p): new function
* ext/socket/basicsocket.c (bsock_recv_nonblock): update rdoc
* ext/socket/udpsocket.c (udp_recvfrom_nonblock): ditto
* test/socket/test_nonblock.rb: new tests
[ruby-core:69542] [Feature #11229]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@50910 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
r36944. it breaks mswin/mingw ruby and brought into many many
crashes.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@36946 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
blocking region with GIL released, for fd.
* thread.c (rb_thread_fd_close): implement. [ruby-core:35203]
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@30852 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
from void-returning function to VALUE-returning one. It causes SEGV
on RubySpec with mingw32.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@26768 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
(rsock_addrinfo): renamed from sock_addrinfo.
(rsock_getaddrinfo): renamed from sock_getaddrinfo.
(rsock_socket): renamed from ruby_socket.
(rsock_sock_s_socketpair): renamed from sock_s_socketpair.
(rsock_connect): renamed from ruby_connect.
* ext/socket/socket.c (sock_listen): make it static.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@22686 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
* ext/socket/rubysocket.h: common header.
* ext/socket/basicsocket.c: new file for BasicSocket.
* ext/socket/ipsocket.c: new file for IPSocket.
* ext/socket/tcpsocket.c: new file for TCPSocket.
* ext/socket/tcpserver.c: new file for TCPServer.
* ext/socket/sockssocket.c: new file for SOCKSSocket.
* ext/socket/udpsocket.c: new file for UDPSocket.
* ext/socket/unixsocket.c: new file for UNIXSocket.
* ext/socket/unixserver.c: new file for UNIXServer.
* ext/socket/socket.c: now for Socket.
* ext/socket/raddrinfo.c: new file for AddrInfo and name resolution.
* ext/socket/constants.c: new file for constants.
* ext/socket/init.c: new file for utilities.
* ext/socket/mkconstants.rb: export *_to_int.
* ext/socket/extconf.rb: add new object files.
* ext/socket/depend: add dependencies for new files.
* ext/.document: add new files.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@21619 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
