diff options
Diffstat (limited to 'erts/emulator/drivers/common/inet_drv.c')
-rw-r--r-- | erts/emulator/drivers/common/inet_drv.c | 248 |
1 files changed, 188 insertions, 60 deletions
diff --git a/erts/emulator/drivers/common/inet_drv.c b/erts/emulator/drivers/common/inet_drv.c index ebd13e6f05..f6de343f72 100644 --- a/erts/emulator/drivers/common/inet_drv.c +++ b/erts/emulator/drivers/common/inet_drv.c @@ -1,7 +1,7 @@ /* * %CopyrightBegin% * - * Copyright Ericsson AB 1997-2016. All Rights Reserved. + * Copyright Ericsson AB 1997-2017. All Rights Reserved. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -591,7 +591,7 @@ static int my_strncasecmp(const char *s1, const char *s2, size_t n) (((unsigned char*) (s))[1] << 8) | \ (((unsigned char*) (s))[0])) -#ifdef HAVE_SYS_UN_H +#if defined(HAVE_SYS_UN_H) || defined(SO_BINDTODEVICE) /* strnlen doesn't exist everywhere */ static size_t my_strnlen(const char *s, size_t maxlen) @@ -602,14 +602,6 @@ static size_t my_strnlen(const char *s, size_t maxlen) return i; } -/* Check that some character in the buffer != '\0' */ -static int is_nonzero(const char *s, size_t n) -{ - size_t i; - for (i = 0; i < n; i++) if (s[i] != '\0') return !0; - return 0; -} - #endif #ifdef VALGRIND @@ -728,7 +720,7 @@ static int is_nonzero(const char *s, size_t n) #define TCP_ADDF_PENDING_SHUTDOWN \ (TCP_ADDF_PENDING_SHUT_WR | TCP_ADDF_PENDING_SHUT_RDWR) #define TCP_ADDF_SHOW_ECONNRESET 64 /* Tell user about incoming RST */ -#define TCP_ADDF_DELAYED_ECONNRESET 128 /* An ECONNRESET error occured on send or shutdown */ +#define TCP_ADDF_DELAYED_ECONNRESET 128 /* An ECONNRESET error occurred on send or shutdown */ #define TCP_ADDF_SHUTDOWN_WR_DONE 256 /* A shutdown(sock, SHUT_WR) or SHUT_RDWR was made */ #define TCP_ADDF_LINGER_ZERO 512 /* Discard driver queue on port close */ @@ -778,6 +770,7 @@ static int is_nonzero(const char *s, size_t n) #define INET_LOPT_TCP_SHOW_ECONNRESET 39 /* tell user about incoming RST */ #define INET_LOPT_LINE_DELIM 40 /* Line delimiting char */ #define INET_OPT_TCLASS 41 /* IPv6 transport class */ +#define INET_OPT_BIND_TO_DEVICE 42 /* get/set network device the socket is bound to */ /* SCTP options: a separate range, from 100: */ #define SCTP_OPT_RTOINFO 100 #define SCTP_OPT_ASSOCINFO 101 @@ -1025,6 +1018,7 @@ typedef struct { inet_async_op* oph; /* queue head or NULL */ inet_async_op* opt; /* queue tail or NULL */ inet_async_op op_queue[INET_MAX_ASYNC]; /* call queue */ + int op_ref; /* queue reference generator */ int active; /* 0 = passive, 1 = active, 2 = active once */ Sint16 active_count; /* counter for {active,N} */ @@ -1256,6 +1250,8 @@ static int tcp_shutdown_error(tcp_descriptor* desc, int err); static int tcp_inet_output(tcp_descriptor* desc, HANDLE event); static int tcp_inet_input(tcp_descriptor* desc, HANDLE event); +static void tcp_desc_close(tcp_descriptor*); + #ifdef HAVE_UDP typedef struct { inet_descriptor inet; /* common data structure (DON'T MOVE) */ @@ -1273,8 +1269,7 @@ static int packet_inet_output(udp_descriptor* udesc, HANDLE event); /* convert descriptor pointer to inet_descriptor pointer */ #define INETP(d) (&(d)->inet) -static int async_ref = 0; /* async reference id generator */ -#define NEW_ASYNC_ID() ((async_ref++) & 0xffff) +#define NEW_ASYNC_ID(desc) ((desc)->op_ref++ & 0xffff) /* check for transition from active to passive */ #define INET_CHECK_ACTIVE_TO_PASSIVE(inet) \ @@ -1334,6 +1329,7 @@ static ErlDrvTermData am_tos; static ErlDrvTermData am_tclass; static ErlDrvTermData am_ipv6_v6only; static ErlDrvTermData am_netns; +static ErlDrvTermData am_bind_to_device; #endif static char str_eafnosupport[] = "eafnosupport"; @@ -1925,7 +1921,7 @@ static void enq_multi_op(tcp_descriptor *desc, char *buf, int req, ErlDrvTermData caller, MultiTimerData *timeout, ErlDrvMonitor *monitorp) { - int id = NEW_ASYNC_ID(); + int id = NEW_ASYNC_ID(INETP(desc)); enq_old_multi_op(desc,id,req,caller,timeout,monitorp); if (buf != NULL) put_int16(id, buf); @@ -1994,7 +1990,7 @@ static int remove_multi_op(tcp_descriptor *desc, int *id_p, int *req_p, static int enq_async_w_tmo(inet_descriptor* desc, char* buf, int req, unsigned timeout, ErlDrvMonitor *monitorp) { - int id = NEW_ASYNC_ID(); + int id = NEW_ASYNC_ID(desc); inet_async_op* opp; if ((opp = desc->oph) == NULL) /* queue empty */ @@ -2203,13 +2199,16 @@ static int inet_reply_ok(inet_descriptor* desc) ErlDrvTermData caller = desc->caller; int i = 0; + desc->caller = 0; + if (is_not_internal_pid(caller)) + return 0; + i = LOAD_ATOM(spec, i, am_inet_reply); i = LOAD_PORT(spec, i, desc->dport); i = LOAD_ATOM(spec, i, am_ok); i = LOAD_TUPLE(spec, i, 3); ASSERT(i == sizeof(spec)/sizeof(*spec)); - desc->caller = 0; return erl_drv_send_term(desc->dport, caller, spec, i); } @@ -3725,6 +3724,7 @@ static void inet_init_sctp(void) { INIT_ATOM(tclass); INIT_ATOM(ipv6_v6only); INIT_ATOM(netns); + INIT_ATOM(bind_to_device); /* Option names */ INIT_ATOM(sctp_rtoinfo); @@ -4018,13 +4018,30 @@ static char* inet_set_address(int family, inet_address* dst, int n; if (*len == 0) return str_einval; n = *((unsigned char*)(*src)); /* Length field */ - if ((*len < 1+n) || (sizeof(dst->sal.sun_path) < n+1)) { + if (*len < 1+n) return str_einval; + if (n + +#ifdef __linux__ + /* Make sure the address gets zero terminated + * except when the first byte is \0 because then it is + * sort of zero terminated although the zero termination + * comes before the address... + * This fix handles Linux's nonportable + * abstract socket address extension. + */ + ((*len) > 1 && (*src)[1] == '\0' ? 0 : 1) +#else + 1 +#endif + > sizeof(dst->sal.sun_path)) { return str_einval; } sys_memzero((char*)dst, sizeof(struct sockaddr_un)); dst->sal.sun_family = family; sys_memcpy(dst->sal.sun_path, (*src)+1, n); *len = offsetof(struct sockaddr_un, sun_path) + n; +#ifndef NO_SA_LEN + dst->sal.sun_len = *len; +#endif *src += 1 + n; return NULL; } @@ -4132,8 +4149,8 @@ static char *inet_set_faddress(int family, inet_address* dst, /* Get a inaddr structure ** src = inaddr structure -** *len is the lenght of structure ** dst is filled with [F,P1,P0,X1,....] +** *len is the length of structure ** where F is the family code (coded) ** and *len is the length of dst on return ** (suitable to deliver to erlang) @@ -4169,15 +4186,16 @@ static int inet_get_address(char* dst, inet_address* src, unsigned int* len) if (*len < offsetof(struct sockaddr_un, sun_path)) return -1; n = *len - offsetof(struct sockaddr_un, sun_path); if (255 < n) return -1; - /* Portability fix: Assume that the address is a zero terminated - * string, except when the first byte is \0 i.e the - * string length is 0. Then use the reported length instead. - * This fix handles Linux's abstract socket address - * nonportable extension. - */ m = my_strnlen(src->sal.sun_path, n); - if ((m == 0) && is_nonzero(src->sal.sun_path, n)) - m = n; +#ifdef __linux__ + /* Assume that the address is a zero terminated string, + * except when the first byte is \0 i.e the string length is 0, + * then use the reported length instead. + * This fix handles Linux's nonportable + * abstract socket address extension. + */ + if (m == 0) m = n; +#endif dst[0] = INET_AF_LOCAL; dst[1] = (char) ((unsigned char) m); sys_memcpy(dst+2, src->sal.sun_path, m); @@ -4234,15 +4252,16 @@ inet_address_to_erlang(char *dst, inet_address **src, SOCKLEN_T sz) { if (sz < offsetof(struct sockaddr_un, sun_path)) return -1; n = sz - offsetof(struct sockaddr_un, sun_path); if (255 < n) return -1; - /* Portability fix: Assume that the address is a zero terminated - * string, except when the first byte is \0 i.e the - * string length is 0. Then use the reported length instead. - * This fix handles Linux's abstract socket address - * nonportable extension. - */ m = my_strnlen((*src)->sal.sun_path, n); - if ((m == 0) && is_nonzero((*src)->sal.sun_path, n)) - m = n; +#ifdef __linux__ + /* Assume that the address is a zero terminated string, + * except when the first byte is \0 i.e the string length is 0, + * Then use the reported length instead. + * This fix handles Linux's nonportable + * abstract socket address extension. + */ + if (m == 0) m = n; +#endif if (dst) { dst[0] = INET_AF_LOCAL; dst[1] = (char) ((unsigned char) m); @@ -4320,6 +4339,12 @@ static void desc_close(inet_descriptor* desc) desc->event = INVALID_EVENT; /* closed by stop_select callback */ desc->s = INVALID_SOCKET; desc->event_mask = 0; + + /* mark as disconnected in case when socket is left lingering due to + * {exit_on_close, false} option in gen_tcp socket creation. Next + * write to socket should produce {error, enotconn} and send a + * message {tcp_error,#Port<>,econnreset} */ + desc->state &= ~INET_STATE_CONNECTED; } } @@ -5946,6 +5971,9 @@ static int inet_set_opts(inet_descriptor* desc, char* ptr, int len) int ival; char* arg_ptr; int arg_sz; +#ifdef SO_BINDTODEVICE + char ifname[IFNAMSIZ]; +#endif enum PacketParseType old_htype = desc->htype; int old_active = desc->active; int propagate; /* Set to 1 if failure to set this option @@ -6331,6 +6359,29 @@ static int inet_set_opts(inet_descriptor* desc, char* ptr, int len) len -= arg_sz; break; +#ifdef SO_BINDTODEVICE + case INET_OPT_BIND_TO_DEVICE: + if (ival < 0) return -1; + if (len < ival) return -1; + if (ival > sizeof(ifname)) { + return -1; + } + memcpy(ifname, ptr, ival); + ifname[ival] = '\0'; + ptr += ival; + len -= ival; + + proto = SOL_SOCKET; + type = SO_BINDTODEVICE; + arg_ptr = (char*)&ifname; + arg_sz = sizeof(ifname); + propagate = 1; /* We do want to know if this fails */ + + DEBUGF(("inet_set_opts(%ld): s=%d, SO_BINDTODEVICE=%s\r\n", + (long)desc->port, desc->s, ifname)); + break; +#endif + default: return -1; } @@ -6463,6 +6514,9 @@ static int sctp_set_opts(inet_descriptor* desc, char* ptr, int len) # ifdef SCTP_DELAYED_ACK_TIME struct sctp_assoc_value av; /* Not in SOLARIS10 */ # endif +# ifdef SO_BINDTODEVICE + char ifname[IFNAMSIZ]; +# endif } arg; @@ -6702,6 +6756,23 @@ static int sctp_set_opts(inet_descriptor* desc, char* ptr, int len) continue; /* Option not supported -- ignore it */ # endif +#ifdef SO_BINDTODEVICE + case INET_OPT_BIND_TO_DEVICE: + arg_sz = get_int32(curr); curr += 4; + CHKLEN(curr, arg_sz); + if (arg_sz >= sizeof(arg.ifname)) + return -1; + memcpy(arg.ifname, curr, arg_sz); + arg.ifname[arg_sz] = '\0'; + curr += arg_sz; + + proto = SOL_SOCKET; + type = SO_BINDTODEVICE; + arg_ptr = (char*) (&arg.ifname); + arg_sz = sizeof ( arg.ifname); + break; +#endif + case SCTP_OPT_AUTOCLOSE: { arg.ival= get_int32 (curr); curr += 4; @@ -6967,6 +7038,9 @@ static ErlDrvSSizeT inet_fill_opts(inet_descriptor* desc, ErlDrvSizeT dest_used = 0; ErlDrvSizeT dest_allocated = destlen; char *orig_dest = *dest; +#ifdef SO_BINDTODEVICE + char ifname[IFNAMSIZ]; +#endif /* Ptr is a name parameter */ #define RETURN_ERROR() \ @@ -7302,6 +7376,26 @@ static ErlDrvSSizeT inet_fill_opts(inet_descriptor* desc, put_int32(arg_sz,ptr); continue; } + +#ifdef SO_BINDTODEVICE + case INET_OPT_BIND_TO_DEVICE: + arg_sz = sizeof(ifname); + TRUNCATE_TO(0,ptr); + PLACE_FOR(5 + arg_sz,ptr); + arg_ptr = ptr + 5; + if (IS_SOCKET_ERROR(sock_getopt(desc->s,SOL_SOCKET,SO_BINDTODEVICE, + arg_ptr,&arg_sz))) { + TRUNCATE_TO(0,ptr); + continue; + } + arg_sz = my_strnlen(arg_ptr, arg_sz); + TRUNCATE_TO(arg_sz + 5,ptr); + *ptr++ = opt; + put_int32(arg_sz,ptr); + ptr += arg_sz; + continue; +#endif + default: RETURN_ERROR(); } @@ -7583,6 +7677,25 @@ static ErlDrvSSizeT sctp_fill_opts(inet_descriptor* desc, i = LOAD_TUPLE (spec, i, 2); break; } + +#ifdef SO_BINDTODEVICE + /* The following option returns a binary: */ + case INET_OPT_BIND_TO_DEVICE: { + char ifname[IFNAMSIZ]; + unsigned int sz = sizeof(ifname); + + if (sock_getopt(desc->s, SOL_SOCKET, SO_BINDTODEVICE, + &ifname, &sz) < 0) continue; + /* Fill in the response: */ + PLACE_FOR(spec, i, + LOAD_ATOM_CNT + LOAD_BUF2BINARY_CNT + LOAD_TUPLE_CNT); + i = LOAD_ATOM (spec, i, am_bind_to_device); + i = LOAD_BUF2BINARY(spec, i, ifname, my_strnlen(ifname, sz)); + i = LOAD_TUPLE (spec, i, 2); + break; + } +#endif + /* The following options just return an integer value: */ case INET_OPT_RCVBUF : case INET_OPT_SNDBUF : @@ -8310,6 +8423,7 @@ static ErlDrvData inet_start(ErlDrvPort port, int size, int protocol) desc->delimiter = '\n'; /* line delimiting char */ desc->oph = NULL; desc->opt = NULL; + desc->op_ref = 0; desc->peer_ptr = NULL; desc->name_ptr = NULL; @@ -8345,10 +8459,10 @@ static ErlDrvData inet_start(ErlDrvPort port, int size, int protocol) return (ErlDrvData)desc; } - -#ifndef MAXHOSTNAMELEN -#define MAXHOSTNAMELEN 256 -#endif +/* MAXHOSTNAMELEN could be 64 or 255 depending +on the platform. Instead, use INET_MAXHOSTNAMELEN +which is always 255 across all platforms */ +#define INET_MAXHOSTNAMELEN 255 /* ** common TCP/UDP/SCTP control command @@ -8525,13 +8639,14 @@ static ErlDrvSSizeT inet_ctl(inet_descriptor* desc, int cmd, char* buf, } case INET_REQ_GETHOSTNAME: { /* get host name */ - char tbuf[MAXHOSTNAMELEN]; + char tbuf[INET_MAXHOSTNAMELEN + 1]; DEBUGF(("inet_ctl(%ld): GETHOSTNAME\r\n", (long)desc->port)); if (len != 0) return ctl_error(EINVAL, rbuf, rsize); - if (IS_SOCKET_ERROR(sock_hostname(tbuf, MAXHOSTNAMELEN))) + /* gethostname requires len to be max(hostname) + 1 */ + if (IS_SOCKET_ERROR(sock_hostname(tbuf, INET_MAXHOSTNAMELEN + 1))) return ctl_error(sock_errno(), rbuf, rsize); return ctl_reply(INET_REP_OK, tbuf, strlen(tbuf), rbuf, rsize); } @@ -8584,6 +8699,7 @@ static ErlDrvSSizeT inet_ctl(inet_descriptor* desc, int cmd, char* buf, else { ptr = &peer; sz = sizeof(peer); + sys_memzero((char *) &peer, sz); if (IS_SOCKET_ERROR (sock_peer (desc->s, (struct sockaddr*)ptr, &sz))) @@ -9138,16 +9254,31 @@ static void tcp_inet_stop(ErlDrvData e) tcp_descriptor* desc = (tcp_descriptor*)e; DEBUGF(("tcp_inet_stop(%ld) {s=%d\r\n", (long)desc->inet.port, desc->inet.s)); + tcp_close_check(desc); - /* free input buffer & output buffer */ - if (desc->i_buf != NULL) - release_buffer(desc->i_buf); - desc->i_buf = NULL; /* net_mess2 may call this function recursively when - faulty messages arrive on dist ports*/ + tcp_clear_input(desc); + DEBUGF(("tcp_inet_stop(%ld) }\r\n", (long)desc->inet.port)); inet_stop(INETP(desc)); } +/* Closes a tcp descriptor without leaving things hanging; the VM keeps trying + * to flush IO queues as long as it contains anything even after the port has + * been closed from the erlang side, which is desired behavior (Think escripts + * writing to files) but pretty hopeless if the underlying fd has been set to + * INVALID_SOCKET through desc_close. + * + * This function should be used in place of desc_close/erl_inet_close in all + * TCP-related operations. Note that this only closes the desc cleanly; it + * will be freed through tcp_inet_stop later on. */ +static void tcp_desc_close(tcp_descriptor* desc) +{ + tcp_clear_input(desc); + tcp_clear_output(desc); + + erl_inet_close(INETP(desc)); +} + /* TCP requests from Erlang */ static ErlDrvSSizeT tcp_inet_ctl(ErlDrvData e, unsigned int cmd, char* buf, ErlDrvSizeT len, @@ -9392,7 +9523,7 @@ static ErlDrvSSizeT tcp_inet_ctl(ErlDrvData e, unsigned int cmd, case INET_REQ_CLOSE: DEBUGF(("tcp_inet_ctl(%ld): CLOSE\r\n", (long)desc->inet.port)); tcp_close_check(desc); - erl_inet_close(INETP(desc)); + tcp_desc_close(desc); return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize); @@ -9516,7 +9647,7 @@ static void tcp_inet_timeout(ErlDrvData e) set_busy_port(desc->inet.port, 0); inet_reply_error_am(INETP(desc), am_timeout); if (desc->send_timeout_close) { - erl_inet_close(INETP(desc)); + tcp_desc_close(desc); } } else { @@ -9530,7 +9661,7 @@ static void tcp_inet_timeout(ErlDrvData e) else if ((state & INET_STATE_CONNECTING) == INET_STATE_CONNECTING) { /* assume connect timeout */ /* close the socket since it's not usable (see man pages) */ - erl_inet_close(INETP(desc)); + tcp_desc_close(desc); async_error_am(INETP(desc), am_timeout); } else if ((state & INET_STATE_ACCEPTING) == INET_STATE_ACCEPTING) { @@ -9693,8 +9824,7 @@ static int tcp_recv_closed(tcp_descriptor* desc) /* passive mode do not terminate port ! */ tcp_clear_input(desc); if (desc->inet.exitf) { - tcp_clear_output(desc); - desc_close(INETP(desc)); + tcp_desc_close(desc); } else { desc_close_read(INETP(desc)); } @@ -9737,7 +9867,7 @@ static int tcp_recv_error(tcp_descriptor* desc, int err) driver_cancel_timer(desc->inet.port); tcp_clear_input(desc); if (desc->inet.exitf) { - desc_close(INETP(desc)); + tcp_desc_close(desc); } else { desc_close_read(INETP(desc)); } @@ -10386,9 +10516,6 @@ static int tcp_send_or_shutdown_error(tcp_descriptor* desc, int err) set_busy_port(desc->inet.port, 0); } - tcp_clear_output(desc); - tcp_clear_input(desc); - /* * We used to handle "expected errors" differently from unexpected ones. * Now we handle all errors in the same way (unless the show_econnreset @@ -10409,10 +10536,10 @@ static int tcp_send_or_shutdown_error(tcp_descriptor* desc, int err) if (desc->inet.exitf) driver_exit(desc->inet.port, 0); else - desc_close(INETP(desc)); + tcp_desc_close(desc); } else { tcp_close_check(desc); - erl_inet_close(INETP(desc)); + tcp_desc_close(desc); if (desc->inet.caller) { if (show_econnreset) @@ -10735,10 +10862,11 @@ static int tcp_inet_output(tcp_descriptor* desc, HANDLE event) #ifndef SO_ERROR { - int sz = sizeof(desc->inet.remote); - int code = sock_peer(desc->inet.s, - (struct sockaddr*) &desc->inet.remote, &sz); - + int sz, code; + sz = sizeof(desc->inet.remote); + sys_memzero((char *) &desc->inet.remote, sz); + code = sock_peer(desc->inet.s, + (struct sockaddr*) &desc->inet.remote, &sz); if (IS_SOCKET_ERROR(code)) { desc->inet.state = INET_STATE_OPEN; /* restore state */ ret = async_error(INETP(desc), sock_errno()); |