From 210665e8a252faecf28ccebade999e734ab92582 Mon Sep 17 00:00:00 2001 From: Igor Slepchin Date: Mon, 24 Sep 2018 22:44:03 -0400 Subject: Add nopush TCP socket option This translates to TCP_CORK on Linux and TCP_NOPUSH on BSD. In effect, this acts as super-Nagle: no partial TCP segments are sent out until this option is turned off. Once turned off, all accumulated unsent data is sent out immediately. The latter is *not* the case on OSX, hence the implementation ignores "nopush" on OSX to reduce confusion. --- erts/emulator/drivers/common/inet_drv.c | 30 ++++++++++++++++++++++++++++++ erts/preloaded/src/prim_inet.erl | 3 +++ lib/kernel/doc/src/inet.xml | 12 ++++++++++++ lib/kernel/src/inet_int.hrl | 1 + lib/kernel/test/inet_sockopt_SUITE.erl | 9 +++++++-- 5 files changed, 53 insertions(+), 2 deletions(-) diff --git a/erts/emulator/drivers/common/inet_drv.c b/erts/emulator/drivers/common/inet_drv.c index 7f20477363..c13ceb14ad 100644 --- a/erts/emulator/drivers/common/inet_drv.c +++ b/erts/emulator/drivers/common/inet_drv.c @@ -812,6 +812,7 @@ static size_t my_strnlen(const char *s, size_t maxlen) #define INET_OPT_PKTOPTIONS 45 /* IP(V6)_PKTOPTIONS get ancillary data */ #define INET_OPT_TTL 46 /* IP_TTL */ #define INET_OPT_RECVTTL 47 /* IP_RECVTTL ancillary data */ +#define TCP_OPT_NOPUSH 48 /* super-Nagle, aka TCP_CORK */ /* SCTP options: a separate range, from 100: */ #define SCTP_OPT_RTOINFO 100 #define SCTP_OPT_ASSOCINFO 101 @@ -955,6 +956,12 @@ static size_t my_strnlen(const char *s, size_t maxlen) #endif +#if defined(TCP_CORK) +#define INET_TCP_NOPUSH TCP_CORK +#elif defined(TCP_NOPUSH) && !defined(__DARWIN__) +#define INET_TCP_NOPUSH TCP_NOPUSH +#endif + #define BIN_REALLOC_MARGIN(x) ((x)/4) /* 25% */ /* The general purpose sockaddr */ @@ -6532,6 +6539,19 @@ static int inet_set_opts(inet_descriptor* desc, char* ptr, int len) (long)desc->port, desc->s, ival)); break; + case TCP_OPT_NOPUSH: +#if defined(INET_TCP_NOPUSH) + proto = IPPROTO_TCP; + type = INET_TCP_NOPUSH; + DEBUGF(("inet_set_opts(%ld): s=%d, t=%d TCP_NOPUSH=%d\r\n", + (long)desc->port, desc->s, type, ival)); + break; +#else + /* inet_fill_opts always returns a value for this option, + * so we need to ignore it if not implemented, just in case */ + continue; +#endif + #if defined(HAVE_MULTICAST_SUPPORT) && defined(IPPROTO_IP) case UDP_OPT_MULTICAST_TTL: @@ -7693,6 +7713,16 @@ static ErlDrvSSizeT inet_fill_opts(inet_descriptor* desc, proto = IPPROTO_TCP; type = TCP_NODELAY; break; + case TCP_OPT_NOPUSH: +#if defined(INET_TCP_NOPUSH) + proto = IPPROTO_TCP; + type = INET_TCP_NOPUSH; + break; +#else + *ptr++ = opt; + put_int32(0, ptr); + continue; +#endif #if defined(HAVE_MULTICAST_SUPPORT) && defined(IPPROTO_IP) case UDP_OPT_MULTICAST_TTL: diff --git a/erts/preloaded/src/prim_inet.erl b/erts/preloaded/src/prim_inet.erl index 8169943dde..c98c7937b7 100644 --- a/erts/preloaded/src/prim_inet.erl +++ b/erts/preloaded/src/prim_inet.erl @@ -1275,6 +1275,7 @@ enc_opt(pktoptions) -> ?INET_OPT_PKTOPTIONS; enc_opt(ttl) -> ?INET_OPT_TTL; enc_opt(recvttl) -> ?INET_OPT_RECVTTL; enc_opt(nodelay) -> ?TCP_OPT_NODELAY; +enc_opt(nopush) -> ?TCP_OPT_NOPUSH; enc_opt(multicast_if) -> ?UDP_OPT_MULTICAST_IF; enc_opt(multicast_ttl) -> ?UDP_OPT_MULTICAST_TTL; enc_opt(multicast_loop) -> ?UDP_OPT_MULTICAST_LOOP; @@ -1336,6 +1337,7 @@ dec_opt(?INET_OPT_PRIORITY) -> priority; dec_opt(?INET_OPT_TOS) -> tos; dec_opt(?INET_OPT_TCLASS) -> tclass; dec_opt(?TCP_OPT_NODELAY) -> nodelay; +dec_opt(?TCP_OPT_NOPUSH) -> nopush; dec_opt(?INET_OPT_RECVTOS) -> recvtos; dec_opt(?INET_OPT_RECVTCLASS) -> recvtclass; dec_opt(?INET_OPT_PKTOPTIONS) -> pktoptions; @@ -1422,6 +1424,7 @@ type_opt_1(pktoptions) -> opts; type_opt_1(ttl) -> int; type_opt_1(recvttl) -> bool; type_opt_1(nodelay) -> bool; +type_opt_1(nopush) -> bool; type_opt_1(ipv6_v6only) -> bool; %% multicast type_opt_1(multicast_ttl) -> int; diff --git a/lib/kernel/doc/src/inet.xml b/lib/kernel/doc/src/inet.xml index ed775d67eb..9fc361a234 100644 --- a/lib/kernel/doc/src/inet.xml +++ b/lib/kernel/doc/src/inet.xml @@ -1039,6 +1039,18 @@ setcap cap_sys_admin,cap_sys_ptrace,cap_dac_read_search+epi beam.smp is turned on for the socket, which means that also small amounts of data are sent immediately.

+ {nopush, Boolean}(TCP/IP sockets) + +

This translates to TCP_NOPUSH on BSD and + to TCP_CORK on Linux.

+

If Boolean == true, the corresponding option + is turned on for the socket, which means that small + amounts of data are accumulated until a full MSS-worth + of data is available or this option is turned off.

+

Note that while TCP_NOPUSH socket option is available on OSX, its semantics + is very different (e.g., unsetting it does not cause immediate send + of accumulated data). Hence, nopush option is intentionally ignored on OSX.

+ {packet, PacketType}(TCP/IP sockets)

Defines the type of packets to use for a socket. diff --git a/lib/kernel/src/inet_int.hrl b/lib/kernel/src/inet_int.hrl index c8e09d18ad..f6525d7261 100644 --- a/lib/kernel/src/inet_int.hrl +++ b/lib/kernel/src/inet_int.hrl @@ -162,6 +162,7 @@ -define(INET_OPT_PKTOPTIONS, 45). -define(INET_OPT_TTL, 46). -define(INET_OPT_RECVTTL, 47). +-define(TCP_OPT_NOPUSH, 48). % Specific SCTP options: separate range: -define(SCTP_OPT_RTOINFO, 100). -define(SCTP_OPT_ASSOCINFO, 101). diff --git a/lib/kernel/test/inet_sockopt_SUITE.erl b/lib/kernel/test/inet_sockopt_SUITE.erl index ada9c2689c..27ff74e309 100644 --- a/lib/kernel/test/inet_sockopt_SUITE.erl +++ b/lib/kernel/test/inet_sockopt_SUITE.erl @@ -110,9 +110,14 @@ simple(Config) when is_list(Config) -> {S1,S2} = create_socketpair(Opt, Opt), {ok,Opt} = inet:getopts(S1,OptTags), {ok,Opt} = inet:getopts(S2,OptTags), - COpt = [{X,case X of nodelay -> false;_ -> Y end} || {X,Y} <- Opt], + NoPushOpt = case os:type() of + {unix, Osname} when Osname =:= linux; Osname =:= freebsd -> {nopush, true}; + {_,_} -> {nopush, false} + end, + COpt = [{X,case X of nodelay -> false;_ -> Y end} || {X,Y} <- [NoPushOpt|Opt]], + COptTags = [X || {X,_} <- COpt], inet:setopts(S1,COpt), - {ok,COpt} = inet:getopts(S1,OptTags), + {ok,COpt} = inet:getopts(S1,COptTags), {ok,Opt} = inet:getopts(S2,OptTags), gen_tcp:close(S1), gen_tcp:close(S2), -- cgit v1.2.1 From 917eeea53273f00489715a94a90cc0c2bb129b74 Mon Sep 17 00:00:00 2001 From: Igor Slepchin Date: Tue, 25 Sep 2018 00:44:00 -0400 Subject: "cork" tcp socket around file:sendfile This fixes 200ms delay on the last TCP segment when using file:sendfile/2 on Linux (ERL-698). --- erts/preloaded/src/prim_inet.erl | 24 +++++++++++++++++++++++- 1 file changed, 23 insertions(+), 1 deletion(-) diff --git a/erts/preloaded/src/prim_inet.erl b/erts/preloaded/src/prim_inet.erl index c98c7937b7..abd322e5cf 100644 --- a/erts/preloaded/src/prim_inet.erl +++ b/erts/preloaded/src/prim_inet.erl @@ -520,13 +520,35 @@ sendfile(S, FileHandle, Offset, Length) sendfile(S, FileHandle, Offset, Length) -> case erlang:port_info(S, connected) of {connected, Pid} when Pid =:= self() -> - sendfile_1(S, FileHandle, Offset, Length); + Uncork = sendfile_maybe_cork(S), + Result = sendfile_1(S, FileHandle, Offset, Length), + sendfile_maybe_uncork(S, Uncork), + Result; {connected, Pid} when Pid =/= self() -> {error, not_owner}; _Other -> {error, einval} end. +sendfile_maybe_cork(S) -> + case getprotocol(S) of + tcp -> + case getopts(S, [nopush]) of + {ok, [{nopush,false}]} -> + _ = setopts(S, [{nopush,true}]), + true; + _ -> + false + end; + _ -> false + end. + +sendfile_maybe_uncork(S, true) -> + _ = setopts(S, [{nopush,false}]), + ok; +sendfile_maybe_uncork(_, false) -> + ok. + sendfile_1(S, FileHandle, Offset, 0) -> sendfile_1(S, FileHandle, Offset, (1 bsl 63) - 1); sendfile_1(_S, _FileHandle, Offset, Length) when -- cgit v1.2.1