X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=lib%2Fsocket-util.c;h=82a33c2077c9da0565e07fc4b1358f786af49d34;hb=be812f2d2b0c948e5f52017c4df79645809fcb90;hp=c436724cd501f44f701bb53e33046af20ffaddeb;hpb=13f2ef9709fbe24b56540ddc5db62c1098b4190f;p=openvswitch diff --git a/lib/socket-util.c b/lib/socket-util.c index c436724c..82a33c20 100644 --- a/lib/socket-util.c +++ b/lib/socket-util.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks. + * Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -17,6 +17,7 @@ #include #include "socket-util.h" #include +#include #include #include #include @@ -59,6 +60,9 @@ VLOG_DEFINE_THIS_MODULE(socket_util); #define O_DIRECTORY 0 #endif +static int getsockopt_int(int fd, int level, int option, const char *optname, + int *valuep); + /* Sets 'fd' to non-blocking mode. Returns 0 if successful, otherwise a * positive errno value. */ int @@ -78,6 +82,21 @@ set_nonblocking(int fd) } } +static int +set_dscp(int fd, uint8_t dscp) +{ + if (dscp > 63) { + return EINVAL; + } + + dscp = dscp << 2; + if (setsockopt(fd, IPPROTO_IP, IP_TOS, &dscp, sizeof dscp)) { + return errno; + } + + return 0; +} + static bool rlim_is_finite(rlim_t limit) { @@ -182,11 +201,9 @@ int get_socket_error(int fd) { int error; - socklen_t len = sizeof(error); - if (getsockopt(fd, SOL_SOCKET, SO_ERROR, &error, &len) < 0) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 10); + + if (getsockopt_int(fd, SOL_SOCKET, SO_ERROR, "SO_ERROR", &error)) { error = errno; - VLOG_ERR_RL(&rl, "getsockopt(SO_ERROR): %s", strerror(error)); } return error; } @@ -221,15 +238,13 @@ check_connection_completion(int fd) int drain_rcvbuf(int fd) { - socklen_t rcvbuf_len; - size_t rcvbuf; + int rcvbuf; - rcvbuf_len = sizeof rcvbuf; - if (getsockopt(fd, SOL_SOCKET, SO_RCVBUF, &rcvbuf, &rcvbuf_len) < 0) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 10); - VLOG_ERR_RL(&rl, "getsockopt(SO_RCVBUF) failed: %s", strerror(errno)); - return errno; + rcvbuf = get_socket_rcvbuf(fd); + if (rcvbuf < 0) { + return -rcvbuf; } + while (rcvbuf > 0) { /* In Linux, specifying MSG_TRUNC in the flags argument causes the * datagram length to be returned, even if that is longer than the @@ -250,6 +265,18 @@ drain_rcvbuf(int fd) return 0; } +/* Returns the size of socket 'sock''s receive buffer (SO_RCVBUF), or a + * negative errno value if an error occurs. */ +int +get_socket_rcvbuf(int sock) +{ + int rcvbuf; + int error; + + error = getsockopt_int(sock, SOL_SOCKET, SO_RCVBUF, "SO_RCVBUF", &rcvbuf); + return error ? -error : rcvbuf; +} + /* Reads and discards up to 'n' datagrams from 'fd', stopping as soon as no * more data can be immediately read. ('fd' should therefore be in * non-blocking mode.)*/ @@ -447,7 +474,7 @@ error: error = EPROTO; } if (bind_path) { - fatal_signal_remove_file_to_unlink(bind_path); + fatal_signal_unlink_file_now(bind_path); } close(fd); return -error; @@ -533,10 +560,14 @@ exit: * and stores -1 into '*fdp'. * * If 'sinp' is non-null, then on success the target address is stored into - * '*sinp'. */ + * '*sinp'. + * + * 'dscp' becomes the DSCP bits in the IP headers for the new connection. It + * should be in the range [0, 63] and will automatically be shifted to the + * appropriately place in the IP tos field. */ int inet_open_active(int style, const char *target, uint16_t default_port, - struct sockaddr_in *sinp, int *fdp) + struct sockaddr_in *sinp, int *fdp, uint8_t dscp) { struct sockaddr_in sin; int fd = -1; @@ -557,31 +588,33 @@ inet_open_active(int style, const char *target, uint16_t default_port, } error = set_nonblocking(fd); if (error) { - goto exit_close; + goto exit; + } + + /* The dscp bits must be configured before connect() to ensure that the TOS + * field is set during the connection establishment. If set after + * connect(), the handshake SYN frames will be sent with a TOS of 0. */ + error = set_dscp(fd, dscp); + if (error) { + VLOG_ERR("%s: socket: %s", target, strerror(error)); + goto exit; } /* Connect. */ error = connect(fd, (struct sockaddr *) &sin, sizeof sin) == 0 ? 0 : errno; if (error == EINPROGRESS) { error = EAGAIN; - } else if (error && error != EAGAIN) { - goto exit_close; } - /* Success: error is 0 or EAGAIN. */ - goto exit; - -exit_close: - close(fd); exit: if (!error || error == EAGAIN) { if (sinp) { *sinp = sin; } - *fdp = fd; - } else { - *fdp = -1; + } else if (fd >= 0) { + close(fd); } + *fdp = fd; return error; } @@ -598,7 +631,7 @@ exit: * If successful, stores the address into '*sinp' and returns true; otherwise * zeros '*sinp' and returns false. */ bool -inet_parse_passive(const char *target_, uint16_t default_port, +inet_parse_passive(const char *target_, int default_port, struct sockaddr_in *sinp) { char *target = xstrdup(target_); @@ -652,17 +685,21 @@ exit: * negative errno value. * * If 'sinp' is non-null, then on success the bound address is stored into - * '*sinp'. */ + * '*sinp'. + * + * 'dscp' becomes the DSCP bits in the IP headers for the new connection. It + * should be in the range [0, 63] and will automatically be shifted to the + * appropriately place in the IP tos field. */ int inet_open_passive(int style, const char *target, int default_port, - struct sockaddr_in *sinp) + struct sockaddr_in *sinp, uint8_t dscp) { struct sockaddr_in sin; int fd = 0, error; unsigned int yes = 1; if (!inet_parse_passive(target, default_port, &sin)) { - return EAFNOSUPPORT; + return -EAFNOSUPPORT; } /* Create non-blocking socket, set SO_REUSEADDR. */ @@ -670,7 +707,7 @@ inet_open_passive(int style, const char *target, int default_port, if (fd < 0) { error = errno; VLOG_ERR("%s: socket: %s", target, strerror(error)); - return error; + return -error; } error = set_nonblocking(fd); if (error) { @@ -690,8 +727,17 @@ inet_open_passive(int style, const char *target, int default_port, goto error; } + /* The dscp bits must be configured before connect() to ensure that the TOS + * field is set during the connection establishment. If set after + * connect(), the handshake SYN frames will be sent with a TOS of 0. */ + error = set_dscp(fd, dscp); + if (error) { + VLOG_ERR("%s: socket: %s", target, strerror(error)); + goto error; + } + /* Listen. */ - if (listen(fd, 10) < 0) { + if (style == SOCK_STREAM && listen(fd, 10) < 0) { error = errno; VLOG_ERR("%s: listen: %s", target, strerror(error)); goto error; @@ -705,6 +751,7 @@ inet_open_passive(int style, const char *target, int default_port, goto error; } if (sin.sin_family != AF_INET || sin_len != sizeof sin) { + error = EAFNOSUPPORT; VLOG_ERR("%s: getsockname: invalid socket name", target); goto error; } @@ -715,7 +762,7 @@ inet_open_passive(int style, const char *target, int default_port, error: close(fd); - return error; + return -error; } /* Returns a readable and writable fd for /dev/null, if successful, otherwise @@ -846,13 +893,27 @@ xpipe(int fds[2]) } static int -getsockopt_int(int fd, int level, int optname, int *valuep) +getsockopt_int(int fd, int level, int option, const char *optname, int *valuep) { - socklen_t len = sizeof *valuep; + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 10); + socklen_t len; + int value; + int error; - return (getsockopt(fd, level, optname, valuep, &len) ? errno - : len == sizeof *valuep ? 0 - : EINVAL); + len = sizeof value; + if (getsockopt(fd, level, option, &value, &len)) { + error = errno; + VLOG_ERR_RL(&rl, "getsockopt(%s): %s", optname, strerror(error)); + } else if (len != sizeof value) { + error = EINVAL; + VLOG_ERR_RL(&rl, "getsockopt(%s): value is %u bytes (expected %zu)", + optname, (unsigned int) len, sizeof value); + } else { + error = 0; + } + + *valuep = error ? 0 : value; + return error; } static void @@ -890,7 +951,8 @@ describe_sockaddr(struct ds *string, int fd, #define SO_PROTOCOL 38 #endif - if (!getsockopt_int(fd, SOL_SOCKET, SO_PROTOCOL, &protocol)) { + if (!getsockopt_int(fd, SOL_SOCKET, SO_PROTOCOL, "SO_PROTOCOL", + &protocol)) { switch (protocol) { case NETLINK_ROUTE: ds_put_cstr(string, "NETLINK_ROUTE");