From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from localhost (localhost [127.0.0.1]) by turing.freelists.org (Avenir Technologies Mail Multiplex) with ESMTP id 6AF2221A92 for ; Tue, 10 Sep 2019 08:54:51 -0400 (EDT) Received: from turing.freelists.org ([127.0.0.1]) by localhost (turing.freelists.org [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id BOvb85pXd3rH for ; Tue, 10 Sep 2019 08:54:51 -0400 (EDT) Received: from smtpng1.m.smailru.net (smtpng1.m.smailru.net [94.100.181.251]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by turing.freelists.org (Avenir Technologies Mail Multiplex) with ESMTPS id 078F121A83 for ; Tue, 10 Sep 2019 08:54:49 -0400 (EDT) Content-Type: text/plain; charset=us-ascii Mime-Version: 1.0 (Mac OS X Mail 12.2 \(3445.102.3\)) Subject: [tarantool-patches] Re: [server-dev] Re: Re: [PATCH v2 1/2] lua: return getaddrinfo() errors From: Roman Khabibov In-Reply-To: <20190906134509.egjpa2vxfdolkeme@tkn_work_nb> Date: Tue, 10 Sep 2019 15:54:47 +0300 Content-Transfer-Encoding: quoted-printable Message-Id: References: <20190615165829.11888-1-roman.habibov@tarantool.org> <20190623203141.snnjgyqrntr6okp4@tkn_work_nb> <8C0820C4-7F5E-40A9-A634-E449E700D816@tarantool.org> <20190709080407.sv5quzrxpbnijwp4@tkn_work_nb> <20190723123954.dgtfsetqlmjhued3@tkn_work_nb> <704142F3-802F-4C1C-8FA2-6155BFF9A951@tarantool.org> <20190829004547.k24fktai33tlbl73@tkn_work_nb> <868EAF2C-A491-46C9-AD37-7512D6CAB213@tarantool.org> <20190906134509.egjpa2vxfdolkeme@tkn_work_nb> Sender: tarantool-patches-bounce@freelists.org Errors-to: tarantool-patches-bounce@freelists.org Reply-To: tarantool-patches@freelists.org List-Help: List-Unsubscribe: List-software: Ecartis version 1.0.0 List-Id: tarantool-patches List-Subscribe: List-Owner: List-post: List-Archive: To: tarantool-patches@freelists.org Cc: Alexander Turenko >>>> +-- gh-4138 Check getaddrinfo() error from socket:connect() only. >>>> +-- Error code and error message returned by getaddrinfo() depends >>>> +-- on system's gai_strerror(). So that there is no checking for >>>> +-- certain error message. >>>> +test_run:cmd("setopt delimiter ';'") >>>> +--- >>>> +- true >>>> +... >>>> +function check_err(err) >>>> + if err =3D=3D 'getaddrinfo: nodename nor servname provided, or = not known' or >>>> + err =3D=3D 'getaddrinfo: Servname not supported for = ai_socktype' or >>>> + err =3D=3D 'getaddrinfo: Name or service not known' then >>>> + return true >>>> + end >>>> + return false >>>> +end; >>>> +--- >>>> +... >>>> +test_run:cmd("setopt delimiter ''"); >>>> +--- >>>> +- true >>>> +... >>>> +s, err =3D socket:connect('non_exists_hostname:3301') >>>=20 >>> socket.connect() should be called, the colon here is the mistake. >=20 > Please, look again to the code. One call is socket.connect(...), = another > is socket:connect(). >=20 > Just in case, I suggest to add a patch to your patchset to keep = youself > from such mistakes: >=20 > | diff --git a/src/lua/socket.lua b/src/lua/socket.lua > | index e4815adbb..a1c13d4db 100644 > | --- a/src/lua/socket.lua > | +++ b/src/lua/socket.lua > | @@ -1554,7 +1554,8 @@ local function lsocket_tcp() > | end > | =20 > | local function lsocket_connect(host, port) > | - if host =3D=3D nil or port =3D=3D nil then > | + if type(host) ~=3D 'string' or (type(port) ~=3D 'string' and > | + type(port) ~=3D 'number') then > | error("Usage: luasocket.connect(host, port)") > | end > | local s, err =3D tcp_connect(host, port) > | @@ -1569,7 +1570,8 @@ local function lsocket_connect(host, port) > | end > | =20 > | local function lsocket_bind(host, port, backlog) > | - if host =3D=3D nil or port =3D=3D nil then > | + if type(host) ~=3D 'string' or (type(port) ~=3D 'string' and > | + type(port) ~=3D 'number') then > | error("Usage: luasocket.bind(host, port [, backlog])") > | end > | local function prepare(s) return backlog end +s, err =3D socket.getaddrinfo('non_exists_hostname', 3301) +check_err(err) +s, err =3D socket.connect('non_exists_hostname', 3301) +check_err(err) + Also, I added suggested patch. >>> I think we should test both native and Lua Socket API, because both = were >>> changed. >>>=20 >>> Should we ever change Lua Socket API? Please, verify it with a >>> documentation and an actual behaviour of the external module. We = emulate >>> it, so we should be compatible. >=20 > Are you look into this? http://w3.impa.br/~diego/software/luasocket/tcp.html#connect master:connect(address, port) Attempts to connect a master object to a remote host, transforming it = into a client object. Client objects support methods send, receive, = getsockname, getpeername, settimeout, and close. Address can be an IP address or a host name. Port must be an integer = number in the range [1..64K).=20 In case of error, the method returns nil followed by a string describing = the error. In case of success, the method returns 1. I looked to this doc. It seems like our behaviour corresponds this API. commit 9810ccd869183de068939cdce5a157ad35916d36 Author: Roman Khabibov Date: Tue Jul 30 15:49:13 2019 +0300 lua: return getaddrinfo() errors =20 Add getaddrinfo() errors into the several fuctions of socket. Now getaddrinfo() can return a pair of values (nil and error message) in case of error. =20 Closes #4138 =20 @TarantoolBot document Title: socket_object:connect =20 Connect to a remote host using tcp_connect() within. If it is no error, return a socket object. diff --git a/src/box/lua/net_box.lua b/src/box/lua/net_box.lua index 31a8c16b7..beab3b0a7 100644 --- a/src/box/lua/net_box.lua +++ b/src/box/lua/net_box.lua @@ -150,9 +150,12 @@ local function next_id(id) return band(id + 1, = 0x7FFFFFFF) end local function establish_connection(host, port, timeout) local timeout =3D timeout or DEFAULT_CONNECT_TIMEOUT local begin =3D fiber.clock() - local s =3D socket.tcp_connect(host, port, timeout) + local s, err =3D socket.tcp_connect(host, port, timeout) if not s then - return nil, errno.strerror(errno()) + if not err then + return nil, errno.strerror(errno()) + end + return nil, err end local msg =3D s:read({chunk =3D IPROTO_GREETING_SIZE}, timeout - (fiber.clock() - begin)) diff --git a/src/lua/socket.c b/src/lua/socket.c index 130378caf..dbade2d27 100644 --- a/src/lua/socket.c +++ b/src/lua/socket.c @@ -774,6 +774,18 @@ lbox_getaddrinfo_result_wrapper(struct lua_State = *L) return 1; } =20 +/** + * Wrap coio_getaddrinfo() and call it. Push returned values onto + * @a L Lua stack. + * + * @param L Lua stack. + * + * @retval 1 Number of returned values by Lua function if + * coio_getaddrinfo() success. + * @retval 2 Number of returned values by Lua function if + * coio_getaddrinfo() is failed (first is nil, second is error + * message). + */ static int lbox_socket_getaddrinfo(struct lua_State *L) { @@ -816,7 +828,9 @@ lbox_socket_getaddrinfo(struct lua_State *L) =20 if (dns_res !=3D 0) { lua_pushnil(L); - return 1; + struct error *err =3D diag_get()->last; + lua_pushstring(L, err->errmsg); + return 2; } =20 /* no results */ diff --git a/src/lua/socket.lua b/src/lua/socket.lua index a334ad45b..e4815adbb 100644 --- a/src/lua/socket.lua +++ b/src/lua/socket.lua @@ -1041,9 +1041,12 @@ local function tcp_connect(host, port, timeout) end local timeout =3D timeout or TIMEOUT_INFINITY local stop =3D fiber.clock() + timeout - local dns =3D getaddrinfo(host, port, timeout, { type =3D = 'SOCK_STREAM', + local dns, err =3D getaddrinfo(host, port, timeout, { type =3D = 'SOCK_STREAM', protocol =3D 'tcp' }) - if dns =3D=3D nil or #dns =3D=3D 0 then + if dns =3D=3D nil then + return nil, err + end + if #dns =3D=3D 0 then boxerrno(boxerrno.EINVAL) return nil end @@ -1360,9 +1363,16 @@ local function lsocket_tcp_connect(self, host, = port) -- This function is broken by design local ga_opts =3D { family =3D 'AF_INET', type =3D 'SOCK_STREAM' } local timeout =3D deadline - fiber.clock() - local dns =3D getaddrinfo(host, port, timeout, ga_opts) + local dns, err =3D getaddrinfo(host, port, timeout, ga_opts) if dns =3D=3D nil or #dns =3D=3D 0 then - self._errno =3D boxerrno.EINVAL + return nil, err + end + if dns =3D=3D nil then + boxerrno(self._errno) + return nil, err + end + if #dns =3D=3D 0 then + boxerrno(self._errno) return nil, socket_error(self) end for _, remote in ipairs(dns) do @@ -1547,9 +1557,12 @@ local function lsocket_connect(host, port) if host =3D=3D nil or port =3D=3D nil then error("Usage: luasocket.connect(host, port)") end - local s =3D tcp_connect(host, port) + local s, err =3D tcp_connect(host, port) if not s then - return nil, boxerrno.strerror() + if not err then + return nil, boxerrno.strerror() + end + return nil, err end setmetatable(s, lsocket_tcp_client_mt) return s diff --git a/test/app/socket.result b/test/app/socket.result index fd299424c..87b1d0387 100644 --- a/test/app/socket.result +++ b/test/app/socket.result @@ -941,10 +941,20 @@ sc:close() - true ... -- tcp_connect --- test timeout -socket.tcp_connect('127.0.0.1', 80, 0.00000000001) +-- Test timeout. In this test, tcp_connect can return the second +-- output value from internal.getaddrinfo (usually on Mac OS, but +-- theoretically it can happen on Linux too). Sometimes +-- getaddrinfo() is timed out, sometimes connect. On Linux however +-- getaddrinfo is fast enough to never give timeout error in +-- the case. So, there are two sources of timeout errors that are +-- reported differently. This difference has appeared after +-- gh-4138 patch. +s, err =3D socket.tcp_connect('127.0.0.1', 80, 0.00000000001) --- -- null +... +s =3D=3D nil +--- +- true ... -- AF_INET s =3D socket('AF_INET', 'SOCK_STREAM', 'tcp') @@ -1606,7 +1616,7 @@ fio.stat(path) =3D=3D nil { socket.tcp_connect('abrakadabra#123') =3D=3D nil, errno.strerror() } --- - - true - - Invalid argument + - Input/output error ... -- wrong options for getaddrinfo socket.getaddrinfo('host', 'port', { type =3D 'WRONG' }) =3D=3D nil and = errno() =3D=3D errno.EINVAL @@ -2812,6 +2822,41 @@ server:close() --- - true ... +-- gh-4138 Check getaddrinfo() error from socket:connect() only. +-- Error code and error message returned by getaddrinfo() depends +-- on system's gai_strerror(). +test_run:cmd("setopt delimiter ';'") +--- +- true +... +function check_err(err) + if err =3D=3D 'getaddrinfo: nodename nor servname provided, or not = known' or + err =3D=3D 'getaddrinfo: Servname not supported for ai_socktype' = or + err =3D=3D 'getaddrinfo: Name or service not known' then + return true + end + return false +end; +--- +... +test_run:cmd("setopt delimiter ''"); +--- +- true +... +s, err =3D socket.getaddrinfo('non_exists_hostname', 3301) +--- +... +check_err(err) +--- +- true +... +s, err =3D socket.connect('non_exists_hostname', 3301) +--- +... +check_err(err) +--- +- true +... test_run:cmd("clear filter") --- - true diff --git a/test/app/socket.test.lua b/test/app/socket.test.lua index c72d41763..08cb7fc9e 100644 --- a/test/app/socket.test.lua +++ b/test/app/socket.test.lua @@ -300,8 +300,16 @@ sc:close() =20 -- tcp_connect =20 --- test timeout -socket.tcp_connect('127.0.0.1', 80, 0.00000000001) +-- Test timeout. In this test, tcp_connect can return the second +-- output value from internal.getaddrinfo (usually on Mac OS, but +-- theoretically it can happen on Linux too). Sometimes +-- getaddrinfo() is timed out, sometimes connect. On Linux however +-- getaddrinfo is fast enough to never give timeout error in +-- the case. So, there are two sources of timeout errors that are +-- reported differently. This difference has appeared after +-- gh-4138 patch. +s, err =3D socket.tcp_connect('127.0.0.1', 80, 0.00000000001) +s =3D=3D nil =20 -- AF_INET s =3D socket('AF_INET', 'SOCK_STREAM', 'tcp') @@ -958,6 +966,25 @@ fiber.cancel(echo_fiber) client:read(1, 5) =3D=3D '' server:close() =20 +-- gh-4138 Check getaddrinfo() error from socket:connect() only. +-- Error code and error message returned by getaddrinfo() depends +-- on system's gai_strerror(). +test_run:cmd("setopt delimiter ';'") +function check_err(err) + if err =3D=3D 'getaddrinfo: nodename nor servname provided, or not = known' or + err =3D=3D 'getaddrinfo: Servname not supported for ai_socktype' = or + err =3D=3D 'getaddrinfo: Name or service not known' then + return true + end + return false +end; +test_run:cmd("setopt delimiter ''"); + +s, err =3D socket.getaddrinfo('non_exists_hostname', 3301) +check_err(err) +s, err =3D socket.connect('non_exists_hostname', 3301) +check_err(err) + test_run:cmd("clear filter") =20 -- case: sicket receive inconsistent behavior diff --git a/test/box/net.box.result b/test/box/net.box.result index e3dabf7d9..0635ca9e7 100644 --- a/test/box/net.box.result +++ b/test/box/net.box.result @@ -3927,6 +3927,34 @@ test_run:grep_log('default', '00000040:.*') --- - null ... +-- gh-4138 Check getaddrinfo() error from connect() only. Error +-- code and error message returned by getaddrinfo() depends on +-- system's gai_strerror(). +test_run:cmd("setopt delimiter ';'") +--- +- true +... +function check_err(err) + if err =3D=3D 'getaddrinfo: nodename nor servname provided, or not = known' or + err =3D=3D 'getaddrinfo: Servname not supported for ai_socktype' = or + err =3D=3D 'getaddrinfo: Name or service not known' then + return true + end + return false +end; +--- +... +test_run:cmd("setopt delimiter ''"); +--- +- true +... +s =3D remote.connect('non_exists_hostname:3301') +--- +... +check_err(s['error']) +--- +- true +... box.cfg{log_level=3Dlog_level} --- ... diff --git a/test/box/net.box.test.lua b/test/box/net.box.test.lua index 8e65ff470..9c9aee1ad 100644 --- a/test/box/net.box.test.lua +++ b/test/box/net.box.test.lua @@ -1582,4 +1582,21 @@ test_run:wait_log('default', '00000030:.*', nil, = 10) -- we expect nothing below, so don't wait test_run:grep_log('default', '00000040:.*') =20 +-- gh-4138 Check getaddrinfo() error from connect() only. Error +-- code and error message returned by getaddrinfo() depends on +-- system's gai_strerror(). +test_run:cmd("setopt delimiter ';'") +function check_err(err) + if err =3D=3D 'getaddrinfo: nodename nor servname provided, or not = known' or + err =3D=3D 'getaddrinfo: Servname not supported for ai_socktype' = or + err =3D=3D 'getaddrinfo: Name or service not known' then + return true + end + return false +end; +test_run:cmd("setopt delimiter ''"); + +s =3D remote.connect('non_exists_hostname:3301') +check_err(s['error']) + box.cfg{log_level=3Dlog_level}