diff options
Diffstat (limited to 'nsprpub/pr/src/io/prsocket.c')
-rw-r--r-- | nsprpub/pr/src/io/prsocket.c | 1794 |
1 files changed, 1794 insertions, 0 deletions
diff --git a/nsprpub/pr/src/io/prsocket.c b/nsprpub/pr/src/io/prsocket.c new file mode 100644 index 000000000..be9702408 --- /dev/null +++ b/nsprpub/pr/src/io/prsocket.c @@ -0,0 +1,1794 @@ +/* -*- Mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ +/* This Source Code Form is subject to the terms of the Mozilla Public + * License, v. 2.0. If a copy of the MPL was not distributed with this + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ + +#include "primpl.h" + +#include <string.h> + +/************************************************************************/ + +/* These two functions are only used in assertions. */ +#if defined(DEBUG) + +PRBool IsValidNetAddr(const PRNetAddr *addr) +{ + if ((addr != NULL) +#if defined(XP_UNIX) || defined(XP_OS2) + && (addr->raw.family != PR_AF_LOCAL) +#endif + && (addr->raw.family != PR_AF_INET6) + && (addr->raw.family != PR_AF_INET)) { + return PR_FALSE; + } + return PR_TRUE; +} + +static PRBool IsValidNetAddrLen(const PRNetAddr *addr, PRInt32 addr_len) +{ + /* + * The definition of the length of a Unix domain socket address + * is not uniform, so we don't check it. + */ + if ((addr != NULL) +#if defined(XP_UNIX) || defined(XP_OS2) + && (addr->raw.family != AF_UNIX) +#endif + && (PR_NETADDR_SIZE(addr) != addr_len)) { +#if defined(LINUX) && __GLIBC__ == 2 && __GLIBC_MINOR__ == 1 + /* + * In glibc 2.1, struct sockaddr_in6 is 24 bytes. In glibc 2.2 + * and in the 2.4 kernel, struct sockaddr_in6 has the scope_id + * field and is 28 bytes. It is possible for socket functions + * to return an addr_len greater than sizeof(struct sockaddr_in6). + * We need to allow that. (Bugzilla bug #77264) + */ + if ((PR_AF_INET6 == addr->raw.family) + && (sizeof(addr->ipv6) == addr_len)) { + return PR_TRUE; + } +#endif + /* + * The accept(), getsockname(), etc. calls on some platforms + * do not set the actual socket address length on return. + * In this case, we verifiy addr_len is still the value we + * passed in (i.e., sizeof(PRNetAddr)). + */ +#if defined(QNX) + if (sizeof(PRNetAddr) == addr_len) { + return PR_TRUE; + } +#endif + return PR_FALSE; + } + return PR_TRUE; +} + +#endif /* DEBUG */ + +static PRInt32 PR_CALLBACK SocketWritev(PRFileDesc *fd, const PRIOVec *iov, +PRInt32 iov_size, PRIntervalTime timeout) +{ + PRThread *me = _PR_MD_CURRENT_THREAD(); + int w = 0; + const PRIOVec *tmp_iov; +#define LOCAL_MAXIOV 8 + PRIOVec local_iov[LOCAL_MAXIOV]; + PRIOVec *iov_copy = NULL; + int tmp_out; + int index, iov_cnt; + int count=0, sz = 0; /* 'count' is the return value. */ + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return -1; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return -1; + } + + /* + * Assume the first writev will succeed. Copy iov's only on + * failure. + */ + tmp_iov = iov; + for (index = 0; index < iov_size; index++) + sz += iov[index].iov_len; + + iov_cnt = iov_size; + + while (sz > 0) { + + w = _PR_MD_WRITEV(fd, tmp_iov, iov_cnt, timeout); + if (w < 0) { + count = -1; + break; + } + count += w; + if (fd->secret->nonblocking) { + break; + } + sz -= w; + + if (sz > 0) { + /* find the next unwritten vector */ + for ( index = 0, tmp_out = count; + tmp_out >= iov[index].iov_len; + tmp_out -= iov[index].iov_len, index++){;} /* nothing to execute */ + + if (tmp_iov == iov) { + /* + * The first writev failed so we + * must copy iov's around. + * Avoid calloc/free if there + * are few enough iov's. + */ + if (iov_size - index <= LOCAL_MAXIOV) + iov_copy = local_iov; + else if ((iov_copy = (PRIOVec *) PR_CALLOC((iov_size - index) * + sizeof *iov_copy)) == NULL) { + PR_SetError(PR_OUT_OF_MEMORY_ERROR, 0); + return -1; + } + tmp_iov = iov_copy; + } + + PR_ASSERT(tmp_iov == iov_copy); + + /* fill in the first partial read */ + iov_copy[0].iov_base = &(((char *)iov[index].iov_base)[tmp_out]); + iov_copy[0].iov_len = iov[index].iov_len - tmp_out; + index++; + + /* copy the remaining vectors */ + for (iov_cnt=1; index<iov_size; iov_cnt++, index++) { + iov_copy[iov_cnt].iov_base = iov[index].iov_base; + iov_copy[iov_cnt].iov_len = iov[index].iov_len; + } + } + } + + if (iov_copy != local_iov) + PR_DELETE(iov_copy); + return count; +} + +/************************************************************************/ + +PR_IMPLEMENT(PRFileDesc *) PR_ImportTCPSocket(PROsfd osfd) +{ +PRFileDesc *fd; + + if (!_pr_initialized) _PR_ImplicitInitialization(); + fd = PR_AllocFileDesc(osfd, PR_GetTCPMethods()); + if (fd != NULL) { + _PR_MD_MAKE_NONBLOCK(fd); + _PR_MD_INIT_FD_INHERITABLE(fd, PR_TRUE); +#ifdef _PR_NEED_SECRET_AF + /* this means we can only import IPv4 sockets here. + * but this is what the function in ptio.c does. + * We need a way to import IPv6 sockets, too. + */ + fd->secret->af = AF_INET; +#endif + } else + _PR_MD_CLOSE_SOCKET(osfd); + return(fd); +} + +PR_IMPLEMENT(PRFileDesc *) PR_ImportUDPSocket(PROsfd osfd) +{ +PRFileDesc *fd; + + if (!_pr_initialized) _PR_ImplicitInitialization(); + fd = PR_AllocFileDesc(osfd, PR_GetUDPMethods()); + if (fd != NULL) { + _PR_MD_MAKE_NONBLOCK(fd); + _PR_MD_INIT_FD_INHERITABLE(fd, PR_TRUE); + } else + _PR_MD_CLOSE_SOCKET(osfd); + return(fd); +} + + +static const PRIOMethods* PR_GetSocketPollFdMethods(void); + +PR_IMPLEMENT(PRFileDesc*) PR_CreateSocketPollFd(PROsfd osfd) +{ + PRFileDesc *fd; + + if (!_pr_initialized) _PR_ImplicitInitialization(); + + fd = _PR_Getfd(); + + if (fd == NULL) PR_SetError(PR_OUT_OF_MEMORY_ERROR, 0); + else + { + fd->secret->md.osfd = osfd; + fd->secret->inheritable = _PR_TRI_FALSE; + fd->secret->state = _PR_FILEDESC_OPEN; + fd->methods = PR_GetSocketPollFdMethods(); + } + + return fd; +} /* PR_CreateSocketPollFD */ + +PR_IMPLEMENT(PRStatus) PR_DestroySocketPollFd(PRFileDesc *fd) +{ + if (NULL == fd) + { + PR_SetError(PR_BAD_DESCRIPTOR_ERROR, 0); + return PR_FAILURE; + } + fd->secret->state = _PR_FILEDESC_CLOSED; + _PR_Putfd(fd); + return PR_SUCCESS; +} /* PR_DestroySocketPollFd */ + +static PRStatus PR_CALLBACK SocketConnect( + PRFileDesc *fd, const PRNetAddr *addr, PRIntervalTime timeout) +{ + PRInt32 rv; /* Return value of _PR_MD_CONNECT */ + const PRNetAddr *addrp = addr; +#if defined(_PR_INET6) + PRNetAddr addrCopy; +#endif + PRThread *me = _PR_MD_CURRENT_THREAD(); + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return PR_FAILURE; + } +#if defined(_PR_INET6) + if (addr->raw.family == PR_AF_INET6) { + addrCopy = *addr; + addrCopy.raw.family = AF_INET6; + addrp = &addrCopy; + } +#endif + + rv = _PR_MD_CONNECT(fd, addrp, PR_NETADDR_SIZE(addr), timeout); + PR_LOG(_pr_io_lm, PR_LOG_MAX, ("connect -> %d", rv)); + if (rv == 0) + return PR_SUCCESS; + else + return PR_FAILURE; +} + +static PRStatus PR_CALLBACK SocketConnectContinue( + PRFileDesc *fd, PRInt16 out_flags) +{ + PROsfd osfd; + int err; + + if (out_flags & PR_POLL_NVAL) { + PR_SetError(PR_BAD_DESCRIPTOR_ERROR, 0); + return PR_FAILURE; + } + if ((out_flags & (PR_POLL_WRITE | PR_POLL_EXCEPT | PR_POLL_ERR)) == 0) { + PR_ASSERT(out_flags == 0); + PR_SetError(PR_IN_PROGRESS_ERROR, 0); + return PR_FAILURE; + } + + osfd = fd->secret->md.osfd; + +#if defined(XP_UNIX) + + err = _MD_unix_get_nonblocking_connect_error(osfd); + if (err != 0) { + _PR_MD_MAP_CONNECT_ERROR(err); + return PR_FAILURE; + } + return PR_SUCCESS; + +#elif defined(WIN32) || defined(WIN16) + + if (out_flags & PR_POLL_EXCEPT) { + int len = sizeof(err); + if (getsockopt(osfd, (int)SOL_SOCKET, SO_ERROR, (char *) &err, &len) + == SOCKET_ERROR) { + _PR_MD_MAP_GETSOCKOPT_ERROR(WSAGetLastError()); + return PR_FAILURE; + } + if (err != 0) { + _PR_MD_MAP_CONNECT_ERROR(err); + } else { + PR_SetError(PR_UNKNOWN_ERROR, 0); + } + return PR_FAILURE; + } + + PR_ASSERT(out_flags & PR_POLL_WRITE); + return PR_SUCCESS; + +#elif defined(XP_OS2) + + err = _MD_os2_get_nonblocking_connect_error(osfd); + if (err != 0) { + _PR_MD_MAP_CONNECT_ERROR(err); + return PR_FAILURE; + } + return PR_SUCCESS; + +#elif defined(XP_BEOS) + +#ifdef BONE_VERSION /* bug 122364 */ + /* temporary workaround until getsockopt(SO_ERROR) works in BONE */ + if (out_flags & PR_POLL_EXCEPT) { + PR_SetError(PR_CONNECT_REFUSED_ERROR, 0); + return PR_FAILURE; + } + PR_ASSERT(out_flags & PR_POLL_WRITE); + return PR_SUCCESS; +#else + err = _MD_beos_get_nonblocking_connect_error(fd); + if( err != 0 ) { + _PR_MD_MAP_CONNECT_ERROR(err); + return PR_FAILURE; + } + else + return PR_SUCCESS; +#endif /* BONE_VERSION */ + +#else + PR_SetError(PR_NOT_IMPLEMENTED_ERROR, 0); + return PR_FAILURE; +#endif +} + +PR_IMPLEMENT(PRStatus) PR_GetConnectStatus(const PRPollDesc *pd) +{ + /* Find the NSPR layer and invoke its connectcontinue method */ + PRFileDesc *bottom = PR_GetIdentitiesLayer(pd->fd, PR_NSPR_IO_LAYER); + + if (NULL == bottom) { + PR_SetError(PR_INVALID_ARGUMENT_ERROR, 0); + return PR_FAILURE; + } + return SocketConnectContinue(bottom, pd->out_flags); +} + +static PRFileDesc* PR_CALLBACK SocketAccept(PRFileDesc *fd, PRNetAddr *addr, +PRIntervalTime timeout) +{ + PROsfd osfd; + PRFileDesc *fd2; + PRUint32 al; + PRThread *me = _PR_MD_CURRENT_THREAD(); +#ifdef WINNT + PRNetAddr addrCopy; +#endif + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return 0; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return 0; + } + +#ifdef WINNT + if (addr == NULL) { + addr = &addrCopy; + } +#endif + al = sizeof(PRNetAddr); + osfd = _PR_MD_ACCEPT(fd, addr, &al, timeout); + if (osfd == -1) + return 0; + + fd2 = PR_AllocFileDesc(osfd, PR_GetTCPMethods()); + if (!fd2) { + _PR_MD_CLOSE_SOCKET(osfd); + return NULL; + } + + fd2->secret->nonblocking = fd->secret->nonblocking; + fd2->secret->inheritable = fd->secret->inheritable; +#ifdef WINNT + if (!fd2->secret->nonblocking && fd2->secret->inheritable != _PR_TRI_TRUE) { + /* + * The new socket has been associated with an I/O + * completion port. There is no going back. + */ + fd2->secret->md.io_model_committed = PR_TRUE; + } + PR_ASSERT(al == PR_NETADDR_SIZE(addr)); + fd2->secret->md.accepted_socket = PR_TRUE; + memcpy(&fd2->secret->md.peer_addr, addr, al); +#endif + + /* + * On some platforms, the new socket created by accept() + * inherits the nonblocking (or overlapped io) attribute + * of the listening socket. As an optimization, these + * platforms can skip the following _PR_MD_MAKE_NONBLOCK + * call. + */ +#if !defined(SOLARIS) && !defined(IRIX) && !defined(WINNT) + _PR_MD_MAKE_NONBLOCK(fd2); +#endif + +#ifdef _PR_INET6 + if (addr && (AF_INET6 == addr->raw.family)) + addr->raw.family = PR_AF_INET6; +#endif + PR_ASSERT(IsValidNetAddr(addr) == PR_TRUE); + PR_ASSERT(IsValidNetAddrLen(addr, al) == PR_TRUE); + + return fd2; +} + +#ifdef WINNT +PR_IMPLEMENT(PRFileDesc*) PR_NTFast_Accept(PRFileDesc *fd, PRNetAddr *addr, +PRIntervalTime timeout) +{ + PROsfd osfd; + PRFileDesc *fd2; + PRIntn al; + PRThread *me = _PR_MD_CURRENT_THREAD(); + PRNetAddr addrCopy; + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return 0; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return 0; + } + + if (addr == NULL) { + addr = &addrCopy; + } + al = PR_NETADDR_SIZE(addr); + osfd = _PR_MD_FAST_ACCEPT(fd, addr, &al, timeout, PR_TRUE, NULL, NULL); + if (osfd == -1) { + return 0; + } + + fd2 = PR_AllocFileDesc(osfd, PR_GetTCPMethods()); + if (!fd2) { + _PR_MD_CLOSE_SOCKET(osfd); + } else { + fd2->secret->nonblocking = fd->secret->nonblocking; + fd2->secret->md.io_model_committed = PR_TRUE; + PR_ASSERT(al == PR_NETADDR_SIZE(addr)); + fd2->secret->md.accepted_socket = PR_TRUE; + memcpy(&fd2->secret->md.peer_addr, addr, al); +#ifdef _PR_INET6 + if (AF_INET6 == addr->raw.family) + addr->raw.family = PR_AF_INET6; +#endif +#ifdef _PR_NEED_SECRET_AF + fd2->secret->af = fd->secret->af; +#endif + } + return fd2; +} +#endif /* WINNT */ + + +static PRStatus PR_CALLBACK SocketBind(PRFileDesc *fd, const PRNetAddr *addr) +{ + PRInt32 result; + const PRNetAddr *addrp = addr; +#if defined(_PR_INET6) + PRNetAddr addrCopy; +#endif + + PR_ASSERT(IsValidNetAddr(addr) == PR_TRUE); + +#ifdef XP_UNIX + if (addr->raw.family == AF_UNIX) { + /* Disallow relative pathnames */ + if (addr->local.path[0] != '/') { + PR_SetError(PR_INVALID_ARGUMENT_ERROR, 0); + return PR_FAILURE; + } + } +#endif /* XP_UNIX */ + +#if defined(_PR_INET6) + if (addr->raw.family == PR_AF_INET6) { + addrCopy = *addr; + addrCopy.raw.family = AF_INET6; + addrp = &addrCopy; + } +#endif + result = _PR_MD_BIND(fd, addrp, PR_NETADDR_SIZE(addr)); + if (result < 0) { + return PR_FAILURE; + } + return PR_SUCCESS; +} + +static PRStatus PR_CALLBACK SocketListen(PRFileDesc *fd, PRIntn backlog) +{ + PRInt32 result; + + result = _PR_MD_LISTEN(fd, backlog); + if (result < 0) { + return PR_FAILURE; + } + return PR_SUCCESS; +} + +static PRStatus PR_CALLBACK SocketShutdown(PRFileDesc *fd, PRIntn how) +{ + PRInt32 result; + + result = _PR_MD_SHUTDOWN(fd, how); + if (result < 0) { + return PR_FAILURE; + } + return PR_SUCCESS; +} + +static PRInt32 PR_CALLBACK SocketRecv(PRFileDesc *fd, void *buf, PRInt32 amount, PRIntn flags, +PRIntervalTime timeout) +{ + PRInt32 rv; + PRThread *me = _PR_MD_CURRENT_THREAD(); + + if ((flags != 0) && (flags != PR_MSG_PEEK)) { + PR_SetError(PR_INVALID_ARGUMENT_ERROR, 0); + return -1; + } + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return -1; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return -1; + } + + PR_LOG(_pr_io_lm, PR_LOG_MAX, + ("recv: fd=%p osfd=%" PR_PRIdOSFD " buf=%p amount=%d flags=%d", + fd, fd->secret->md.osfd, buf, amount, flags)); + +#ifdef _PR_HAVE_PEEK_BUFFER + if (fd->secret->peekBytes != 0) { + rv = (amount < fd->secret->peekBytes) ? + amount : fd->secret->peekBytes; + memcpy(buf, fd->secret->peekBuffer, rv); + if (flags == 0) { + /* consume the bytes in the peek buffer */ + fd->secret->peekBytes -= rv; + if (fd->secret->peekBytes != 0) { + memmove(fd->secret->peekBuffer, + fd->secret->peekBuffer + rv, + fd->secret->peekBytes); + } + } + return rv; + } + + /* allocate peek buffer, if necessary */ + if ((PR_MSG_PEEK == flags) && _PR_FD_NEED_EMULATE_MSG_PEEK(fd)) { + PR_ASSERT(0 == fd->secret->peekBytes); + /* impose a max size on the peek buffer */ + if (amount > _PR_PEEK_BUFFER_MAX) { + amount = _PR_PEEK_BUFFER_MAX; + } + if (fd->secret->peekBufSize < amount) { + if (fd->secret->peekBuffer) { + PR_Free(fd->secret->peekBuffer); + } + fd->secret->peekBufSize = amount; + fd->secret->peekBuffer = PR_Malloc(amount); + if (NULL == fd->secret->peekBuffer) { + fd->secret->peekBufSize = 0; + PR_SetError(PR_OUT_OF_MEMORY_ERROR, 0); + return -1; + } + } + } +#endif + + rv = _PR_MD_RECV(fd, buf, amount, flags, timeout); + PR_LOG(_pr_io_lm, PR_LOG_MAX, ("recv -> %d, error = %d, os error = %d", + rv, PR_GetError(), PR_GetOSError())); + +#ifdef _PR_HAVE_PEEK_BUFFER + if ((PR_MSG_PEEK == flags) && _PR_FD_NEED_EMULATE_MSG_PEEK(fd)) { + if (rv > 0) { + memcpy(fd->secret->peekBuffer, buf, rv); + fd->secret->peekBytes = rv; + } + } +#endif + + return rv; +} + +static PRInt32 PR_CALLBACK SocketRead(PRFileDesc *fd, void *buf, PRInt32 amount) +{ + return SocketRecv(fd, buf, amount, 0, PR_INTERVAL_NO_TIMEOUT); +} + +static PRInt32 PR_CALLBACK SocketSend(PRFileDesc *fd, const void *buf, PRInt32 amount, +PRIntn flags, PRIntervalTime timeout) +{ + PRInt32 temp, count; + PRThread *me = _PR_MD_CURRENT_THREAD(); + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return -1; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return -1; + } + + count = 0; + while (amount > 0) { + PR_LOG(_pr_io_lm, PR_LOG_MAX, + ("send: fd=%p osfd=%" PR_PRIdOSFD " buf=%p amount=%d", + fd, fd->secret->md.osfd, buf, amount)); + temp = _PR_MD_SEND(fd, buf, amount, flags, timeout); + if (temp < 0) { + count = -1; + break; + } + + count += temp; + if (fd->secret->nonblocking) { + break; + } + buf = (const void*) ((const char*)buf + temp); + + amount -= temp; + } + PR_LOG(_pr_io_lm, PR_LOG_MAX, ("send -> %d", count)); + return count; +} + +static PRInt32 PR_CALLBACK SocketWrite(PRFileDesc *fd, const void *buf, PRInt32 amount) +{ + return SocketSend(fd, buf, amount, 0, PR_INTERVAL_NO_TIMEOUT); +} + +static PRStatus PR_CALLBACK SocketClose(PRFileDesc *fd) +{ + if (!fd || !fd->secret + || (fd->secret->state != _PR_FILEDESC_OPEN + && fd->secret->state != _PR_FILEDESC_CLOSED)) { + PR_SetError(PR_BAD_DESCRIPTOR_ERROR, 0); + return PR_FAILURE; + } + + if (fd->secret->state == _PR_FILEDESC_OPEN) { + if (_PR_MD_CLOSE_SOCKET(fd->secret->md.osfd) < 0) { + return PR_FAILURE; + } + fd->secret->state = _PR_FILEDESC_CLOSED; + } + +#ifdef _PR_HAVE_PEEK_BUFFER + if (fd->secret->peekBuffer) { + PR_ASSERT(fd->secret->peekBufSize > 0); + PR_DELETE(fd->secret->peekBuffer); + fd->secret->peekBufSize = 0; + fd->secret->peekBytes = 0; + } +#endif + + PR_FreeFileDesc(fd); + return PR_SUCCESS; +} + +static PRInt32 PR_CALLBACK SocketAvailable(PRFileDesc *fd) +{ + PRInt32 rv; +#ifdef _PR_HAVE_PEEK_BUFFER + if (fd->secret->peekBytes != 0) { + return fd->secret->peekBytes; + } +#endif + rv = _PR_MD_SOCKETAVAILABLE(fd); + return rv; +} + +static PRInt64 PR_CALLBACK SocketAvailable64(PRFileDesc *fd) +{ + PRInt64 rv; +#ifdef _PR_HAVE_PEEK_BUFFER + if (fd->secret->peekBytes != 0) { + LL_I2L(rv, fd->secret->peekBytes); + return rv; + } +#endif + LL_I2L(rv, _PR_MD_SOCKETAVAILABLE(fd)); + return rv; +} + +static PRStatus PR_CALLBACK SocketSync(PRFileDesc *fd) +{ + return PR_SUCCESS; +} + +static PRInt32 PR_CALLBACK SocketSendTo( + PRFileDesc *fd, const void *buf, PRInt32 amount, + PRIntn flags, const PRNetAddr *addr, PRIntervalTime timeout) +{ + PRInt32 temp, count; + const PRNetAddr *addrp = addr; +#if defined(_PR_INET6) + PRNetAddr addrCopy; +#endif + PRThread *me = _PR_MD_CURRENT_THREAD(); + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return -1; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return -1; + } + + PR_ASSERT(IsValidNetAddr(addr) == PR_TRUE); +#if defined(_PR_INET6) + if (addr->raw.family == PR_AF_INET6) { + addrCopy = *addr; + addrCopy.raw.family = AF_INET6; + addrp = &addrCopy; + } +#endif + + count = 0; + while (amount > 0) { + temp = _PR_MD_SENDTO(fd, buf, amount, flags, + addrp, PR_NETADDR_SIZE(addr), timeout); + if (temp < 0) { + count = -1; + break; + } + count += temp; + if (fd->secret->nonblocking) { + break; + } + buf = (const void*) ((const char*)buf + temp); + amount -= temp; + } + return count; +} + +static PRInt32 PR_CALLBACK SocketRecvFrom(PRFileDesc *fd, void *buf, PRInt32 amount, +PRIntn flags, PRNetAddr *addr, PRIntervalTime timeout) +{ + PRInt32 rv; + PRUint32 al; + PRThread *me = _PR_MD_CURRENT_THREAD(); + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return -1; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return -1; + } + + al = sizeof(PRNetAddr); + rv = _PR_MD_RECVFROM(fd, buf, amount, flags, addr, &al, timeout); +#ifdef _PR_INET6 + if (addr && (AF_INET6 == addr->raw.family)) + addr->raw.family = PR_AF_INET6; +#endif + return rv; +} + +static PRInt32 PR_CALLBACK SocketAcceptRead(PRFileDesc *sd, PRFileDesc **nd, +PRNetAddr **raddr, void *buf, PRInt32 amount, +PRIntervalTime timeout) +{ + PRInt32 rv; + PRThread *me = _PR_MD_CURRENT_THREAD(); + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return -1; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return -1; + } + /* The socket must be in blocking mode. */ + if (sd->secret->nonblocking) { + PR_SetError(PR_INVALID_ARGUMENT_ERROR, 0); + return -1; + } + *nd = NULL; + +#if defined(WINNT) + { + PROsfd newSock; + PRNetAddr *raddrCopy; + + if (raddr == NULL) { + raddr = &raddrCopy; + } + rv = _PR_MD_ACCEPT_READ(sd, &newSock, raddr, buf, amount, timeout); + if (rv < 0) { + rv = -1; + } else { + /* Successfully accepted and read; create the new PRFileDesc */ + *nd = PR_AllocFileDesc(newSock, PR_GetTCPMethods()); + if (*nd == 0) { + _PR_MD_CLOSE_SOCKET(newSock); + /* PR_AllocFileDesc() has invoked PR_SetError(). */ + rv = -1; + } else { + (*nd)->secret->md.io_model_committed = PR_TRUE; + (*nd)->secret->md.accepted_socket = PR_TRUE; + memcpy(&(*nd)->secret->md.peer_addr, *raddr, + PR_NETADDR_SIZE(*raddr)); +#ifdef _PR_INET6 + if (AF_INET6 == *raddr->raw.family) + *raddr->raw.family = PR_AF_INET6; +#endif + } + } + } +#else + rv = PR_EmulateAcceptRead(sd, nd, raddr, buf, amount, timeout); +#endif + return rv; +} + +#ifdef WINNT +PR_IMPLEMENT(PRInt32) PR_NTFast_AcceptRead(PRFileDesc *sd, PRFileDesc **nd, +PRNetAddr **raddr, void *buf, PRInt32 amount, +PRIntervalTime timeout) +{ + PRInt32 rv; + PROsfd newSock; + PRThread *me = _PR_MD_CURRENT_THREAD(); + PRNetAddr *raddrCopy; + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return -1; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return -1; + } + *nd = NULL; + + if (raddr == NULL) { + raddr = &raddrCopy; + } + rv = _PR_MD_FAST_ACCEPT_READ(sd, &newSock, raddr, buf, amount, + timeout, PR_TRUE, NULL, NULL); + if (rv < 0) { + rv = -1; + } else { + /* Successfully accepted and read; create the new PRFileDesc */ + *nd = PR_AllocFileDesc(newSock, PR_GetTCPMethods()); + if (*nd == 0) { + _PR_MD_CLOSE_SOCKET(newSock); + /* PR_AllocFileDesc() has invoked PR_SetError(). */ + rv = -1; + } else { + (*nd)->secret->md.io_model_committed = PR_TRUE; + (*nd)->secret->md.accepted_socket = PR_TRUE; + memcpy(&(*nd)->secret->md.peer_addr, *raddr, + PR_NETADDR_SIZE(*raddr)); +#ifdef _PR_INET6 + if (AF_INET6 == *raddr->raw.family) + *raddr->raw.family = PR_AF_INET6; +#endif +#ifdef _PR_NEED_SECRET_AF + (*nd)->secret->af = sd->secret->af; +#endif + } + } + return rv; +} + +PR_IMPLEMENT(PRInt32) PR_NTFast_AcceptRead_WithTimeoutCallback( +PRFileDesc *sd, PRFileDesc **nd, +PRNetAddr **raddr, void *buf, PRInt32 amount, +PRIntervalTime timeout, +_PR_AcceptTimeoutCallback callback, +void *callbackArg) +{ + PRInt32 rv; + PROsfd newSock; + PRThread *me = _PR_MD_CURRENT_THREAD(); + PRNetAddr *raddrCopy; + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return -1; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return -1; + } + *nd = NULL; + + if (raddr == NULL) { + raddr = &raddrCopy; + } + rv = _PR_MD_FAST_ACCEPT_READ(sd, &newSock, raddr, buf, amount, + timeout, PR_TRUE, callback, callbackArg); + if (rv < 0) { + rv = -1; + } else { + /* Successfully accepted and read; create the new PRFileDesc */ + *nd = PR_AllocFileDesc(newSock, PR_GetTCPMethods()); + if (*nd == 0) { + _PR_MD_CLOSE_SOCKET(newSock); + /* PR_AllocFileDesc() has invoked PR_SetError(). */ + rv = -1; + } else { + (*nd)->secret->md.io_model_committed = PR_TRUE; + (*nd)->secret->md.accepted_socket = PR_TRUE; + memcpy(&(*nd)->secret->md.peer_addr, *raddr, + PR_NETADDR_SIZE(*raddr)); +#ifdef _PR_INET6 + if (AF_INET6 == *raddr->raw.family) + *raddr->raw.family = PR_AF_INET6; +#endif +#ifdef _PR_NEED_SECRET_AF + (*nd)->secret->af = sd->secret->af; +#endif + } + } + return rv; +} +#endif /* WINNT */ + +#ifdef WINNT +PR_IMPLEMENT(void) +PR_NTFast_UpdateAcceptContext(PRFileDesc *socket, PRFileDesc *acceptSocket) +{ + _PR_MD_UPDATE_ACCEPT_CONTEXT( + socket->secret->md.osfd, acceptSocket->secret->md.osfd); +} +#endif /* WINNT */ + +static PRInt32 PR_CALLBACK SocketSendFile( + PRFileDesc *sd, PRSendFileData *sfd, + PRTransmitFileFlags flags, PRIntervalTime timeout) +{ + PRInt32 rv; + PRThread *me = _PR_MD_CURRENT_THREAD(); + + if (_PR_PENDING_INTERRUPT(me)) { + me->flags &= ~_PR_INTERRUPT; + PR_SetError(PR_PENDING_INTERRUPT_ERROR, 0); + return -1; + } + if (_PR_IO_PENDING(me)) { + PR_SetError(PR_IO_PENDING_ERROR, 0); + return -1; + } + /* The socket must be in blocking mode. */ + if (sd->secret->nonblocking) { + PR_SetError(PR_INVALID_ARGUMENT_ERROR, 0); + return -1; + } +#if defined(WINNT) + rv = _PR_MD_SENDFILE(sd, sfd, flags, timeout); + if ((rv >= 0) && (flags == PR_TRANSMITFILE_CLOSE_SOCKET)) { + /* + * This should be kept the same as SocketClose, except + * that _PR_MD_CLOSE_SOCKET(sd->secret->md.osfd) should + * not be called because the socket will be recycled. + */ + PR_FreeFileDesc(sd); + } +#else + rv = PR_EmulateSendFile(sd, sfd, flags, timeout); +#endif /* WINNT */ + + return rv; +} + +static PRInt32 PR_CALLBACK SocketTransmitFile(PRFileDesc *sd, PRFileDesc *fd, +const void *headers, PRInt32 hlen, PRTransmitFileFlags flags, +PRIntervalTime timeout) +{ + PRSendFileData sfd; + + sfd.fd = fd; + sfd.file_offset = 0; + sfd.file_nbytes = 0; + sfd.header = headers; + sfd.hlen = hlen; + sfd.trailer = NULL; + sfd.tlen = 0; + + return(SocketSendFile(sd, &sfd, flags, timeout)); +} + +static PRStatus PR_CALLBACK SocketGetName(PRFileDesc *fd, PRNetAddr *addr) +{ + PRInt32 result; + PRUint32 addrlen; + + addrlen = sizeof(PRNetAddr); + result = _PR_MD_GETSOCKNAME(fd, addr, &addrlen); + if (result < 0) { + return PR_FAILURE; + } +#ifdef _PR_INET6 + if (AF_INET6 == addr->raw.family) + addr->raw.family = PR_AF_INET6; +#endif + PR_ASSERT(IsValidNetAddr(addr) == PR_TRUE); + PR_ASSERT(IsValidNetAddrLen(addr, addrlen) == PR_TRUE); + return PR_SUCCESS; +} + +static PRStatus PR_CALLBACK SocketGetPeerName(PRFileDesc *fd, PRNetAddr *addr) +{ + PRInt32 result; + PRUint32 addrlen; + + addrlen = sizeof(PRNetAddr); + result = _PR_MD_GETPEERNAME(fd, addr, &addrlen); + if (result < 0) { + return PR_FAILURE; + } +#ifdef _PR_INET6 + if (AF_INET6 == addr->raw.family) + addr->raw.family = PR_AF_INET6; +#endif + PR_ASSERT(IsValidNetAddr(addr) == PR_TRUE); + PR_ASSERT(IsValidNetAddrLen(addr, addrlen) == PR_TRUE); + return PR_SUCCESS; +} + +static PRInt16 PR_CALLBACK SocketPoll( + PRFileDesc *fd, PRInt16 in_flags, PRInt16 *out_flags) +{ + *out_flags = 0; + return in_flags; +} /* SocketPoll */ + +static PRIOMethods tcpMethods = { + PR_DESC_SOCKET_TCP, + SocketClose, + SocketRead, + SocketWrite, + SocketAvailable, + SocketAvailable64, + SocketSync, + (PRSeekFN)_PR_InvalidInt, + (PRSeek64FN)_PR_InvalidInt64, + (PRFileInfoFN)_PR_InvalidStatus, + (PRFileInfo64FN)_PR_InvalidStatus, + SocketWritev, + SocketConnect, + SocketAccept, + SocketBind, + SocketListen, + SocketShutdown, + SocketRecv, + SocketSend, + (PRRecvfromFN)_PR_InvalidInt, + (PRSendtoFN)_PR_InvalidInt, + SocketPoll, + SocketAcceptRead, + SocketTransmitFile, + SocketGetName, + SocketGetPeerName, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt, + _PR_SocketGetSocketOption, + _PR_SocketSetSocketOption, + SocketSendFile, + SocketConnectContinue, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt +}; + +static PRIOMethods udpMethods = { + PR_DESC_SOCKET_UDP, + SocketClose, + SocketRead, + SocketWrite, + SocketAvailable, + SocketAvailable64, + SocketSync, + (PRSeekFN)_PR_InvalidInt, + (PRSeek64FN)_PR_InvalidInt64, + (PRFileInfoFN)_PR_InvalidStatus, + (PRFileInfo64FN)_PR_InvalidStatus, + SocketWritev, + SocketConnect, + (PRAcceptFN)_PR_InvalidDesc, + SocketBind, + SocketListen, + SocketShutdown, + SocketRecv, + SocketSend, + SocketRecvFrom, + SocketSendTo, + SocketPoll, + (PRAcceptreadFN)_PR_InvalidInt, + (PRTransmitfileFN)_PR_InvalidInt, + SocketGetName, + SocketGetPeerName, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt, + _PR_SocketGetSocketOption, + _PR_SocketSetSocketOption, + (PRSendfileFN)_PR_InvalidInt, + (PRConnectcontinueFN)_PR_InvalidStatus, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt +}; + + +static PRIOMethods socketpollfdMethods = { + (PRDescType) 0, + (PRCloseFN)_PR_InvalidStatus, + (PRReadFN)_PR_InvalidInt, + (PRWriteFN)_PR_InvalidInt, + (PRAvailableFN)_PR_InvalidInt, + (PRAvailable64FN)_PR_InvalidInt64, + (PRFsyncFN)_PR_InvalidStatus, + (PRSeekFN)_PR_InvalidInt, + (PRSeek64FN)_PR_InvalidInt64, + (PRFileInfoFN)_PR_InvalidStatus, + (PRFileInfo64FN)_PR_InvalidStatus, + (PRWritevFN)_PR_InvalidInt, + (PRConnectFN)_PR_InvalidStatus, + (PRAcceptFN)_PR_InvalidDesc, + (PRBindFN)_PR_InvalidStatus, + (PRListenFN)_PR_InvalidStatus, + (PRShutdownFN)_PR_InvalidStatus, + (PRRecvFN)_PR_InvalidInt, + (PRSendFN)_PR_InvalidInt, + (PRRecvfromFN)_PR_InvalidInt, + (PRSendtoFN)_PR_InvalidInt, + SocketPoll, + (PRAcceptreadFN)_PR_InvalidInt, + (PRTransmitfileFN)_PR_InvalidInt, + (PRGetsocknameFN)_PR_InvalidStatus, + (PRGetpeernameFN)_PR_InvalidStatus, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt, + (PRGetsocketoptionFN)_PR_InvalidStatus, + (PRSetsocketoptionFN)_PR_InvalidStatus, + (PRSendfileFN)_PR_InvalidInt, + (PRConnectcontinueFN)_PR_InvalidStatus, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt, + (PRReservedFN)_PR_InvalidInt +}; + +PR_IMPLEMENT(const PRIOMethods*) PR_GetTCPMethods() +{ + return &tcpMethods; +} + +PR_IMPLEMENT(const PRIOMethods*) PR_GetUDPMethods() +{ + return &udpMethods; +} + +static const PRIOMethods* PR_GetSocketPollFdMethods() +{ + return &socketpollfdMethods; +} /* PR_GetSocketPollFdMethods */ + +#if !defined(_PR_INET6) || defined(_PR_INET6_PROBE) +PR_EXTERN(PRStatus) _pr_push_ipv6toipv4_layer(PRFileDesc *fd); + +#if defined(_PR_INET6_PROBE) + +extern PRBool _pr_ipv6_is_present(void); + +PR_IMPLEMENT(PRBool) _pr_test_ipv6_socket() +{ + PROsfd osfd; + + osfd = _PR_MD_SOCKET(AF_INET6, SOCK_STREAM, 0); + if (osfd != -1) { + _PR_MD_CLOSE_SOCKET(osfd); + return PR_TRUE; + } + return PR_FALSE; +} +#endif /* _PR_INET6_PROBE */ + +#endif + +PR_IMPLEMENT(PRFileDesc*) PR_Socket(PRInt32 domain, PRInt32 type, PRInt32 proto) +{ + PROsfd osfd; + PRFileDesc *fd; + PRInt32 tmp_domain = domain; + + if (!_pr_initialized) _PR_ImplicitInitialization(); + if (PR_AF_INET != domain + && PR_AF_INET6 != domain +#if defined(XP_UNIX) || defined(XP_OS2) + && PR_AF_LOCAL != domain +#endif + ) { + PR_SetError(PR_ADDRESS_NOT_SUPPORTED_ERROR, 0); + return NULL; + } + +#if defined(_PR_INET6_PROBE) + if (PR_AF_INET6 == domain) + domain = _pr_ipv6_is_present() ? AF_INET6 : AF_INET; +#elif defined(_PR_INET6) + if (PR_AF_INET6 == domain) + domain = AF_INET6; +#else + if (PR_AF_INET6 == domain) + domain = AF_INET; +#endif /* _PR_INET6 */ + osfd = _PR_MD_SOCKET(domain, type, proto); + if (osfd == -1) { + return 0; + } + if (type == SOCK_STREAM) + fd = PR_AllocFileDesc(osfd, PR_GetTCPMethods()); + else + fd = PR_AllocFileDesc(osfd, PR_GetUDPMethods()); + /* + * Make the sockets non-blocking + */ + if (fd != NULL) { + _PR_MD_MAKE_NONBLOCK(fd); + _PR_MD_INIT_FD_INHERITABLE(fd, PR_FALSE); +#ifdef _PR_NEED_SECRET_AF + fd->secret->af = domain; +#endif +#if defined(_PR_INET6_PROBE) || !defined(_PR_INET6) + /* + * For platforms with no support for IPv6 + * create layered socket for IPv4-mapped IPv6 addresses + */ + if (PR_AF_INET6 == tmp_domain && PR_AF_INET == domain) { + if (PR_FAILURE == _pr_push_ipv6toipv4_layer(fd)) { + PR_Close(fd); + fd = NULL; + } + } +#endif + } else + _PR_MD_CLOSE_SOCKET(osfd); + + return fd; +} + +PR_IMPLEMENT(PRFileDesc *) PR_NewTCPSocket(void) +{ + PRInt32 domain = AF_INET; + + return PR_Socket(domain, SOCK_STREAM, 0); +} + +PR_IMPLEMENT(PRFileDesc*) PR_NewUDPSocket(void) +{ + PRInt32 domain = AF_INET; + + return PR_Socket(domain, SOCK_DGRAM, 0); +} + +PR_IMPLEMENT(PRFileDesc *) PR_OpenTCPSocket(PRIntn af) +{ + return PR_Socket(af, SOCK_STREAM, 0); +} + +PR_IMPLEMENT(PRFileDesc*) PR_OpenUDPSocket(PRIntn af) +{ + return PR_Socket(af, SOCK_DGRAM, 0); +} + +PR_IMPLEMENT(PRStatus) PR_NewTCPSocketPair(PRFileDesc *f[]) +{ +#ifdef XP_UNIX + PRInt32 rv, osfd[2]; + + if (!_pr_initialized) _PR_ImplicitInitialization(); + + rv = _PR_MD_SOCKETPAIR(AF_UNIX, SOCK_STREAM, 0, osfd); + if (rv == -1) { + return PR_FAILURE; + } + + f[0] = PR_AllocFileDesc(osfd[0], PR_GetTCPMethods()); + if (!f[0]) { + _PR_MD_CLOSE_SOCKET(osfd[0]); + _PR_MD_CLOSE_SOCKET(osfd[1]); + /* PR_AllocFileDesc() has invoked PR_SetError(). */ + return PR_FAILURE; + } + f[1] = PR_AllocFileDesc(osfd[1], PR_GetTCPMethods()); + if (!f[1]) { + PR_Close(f[0]); + _PR_MD_CLOSE_SOCKET(osfd[1]); + /* PR_AllocFileDesc() has invoked PR_SetError(). */ + return PR_FAILURE; + } + _PR_MD_MAKE_NONBLOCK(f[0]); + _PR_MD_INIT_FD_INHERITABLE(f[0], PR_FALSE); + _PR_MD_MAKE_NONBLOCK(f[1]); + _PR_MD_INIT_FD_INHERITABLE(f[1], PR_FALSE); + return PR_SUCCESS; +#elif defined(WINNT) + /* + * A socket pair is often used for interprocess communication, + * so we need to make sure neither socket is associated with + * the I/O completion port; otherwise it can't be used by a + * child process. + * + * The default implementation below cannot be used for NT + * because PR_Accept would have associated the I/O completion + * port with the listening and accepted sockets. + */ + SOCKET listenSock; + SOCKET osfd[2]; + struct sockaddr_in selfAddr, peerAddr; + int addrLen; + + if (!_pr_initialized) _PR_ImplicitInitialization(); + + osfd[0] = osfd[1] = INVALID_SOCKET; + listenSock = socket(AF_INET, SOCK_STREAM, 0); + if (listenSock == INVALID_SOCKET) { + goto failed; + } + selfAddr.sin_family = AF_INET; + selfAddr.sin_port = 0; + selfAddr.sin_addr.s_addr = htonl(INADDR_LOOPBACK); /* BugZilla: 35408 */ + addrLen = sizeof(selfAddr); + if (bind(listenSock, (struct sockaddr *) &selfAddr, + addrLen) == SOCKET_ERROR) { + goto failed; + } + if (getsockname(listenSock, (struct sockaddr *) &selfAddr, + &addrLen) == SOCKET_ERROR) { + goto failed; + } + if (listen(listenSock, 5) == SOCKET_ERROR) { + goto failed; + } + osfd[0] = socket(AF_INET, SOCK_STREAM, 0); + if (osfd[0] == INVALID_SOCKET) { + goto failed; + } + selfAddr.sin_addr.s_addr = htonl(INADDR_LOOPBACK); + + /* + * Only a thread is used to do the connect and accept. + * I am relying on the fact that connect returns + * successfully as soon as the connect request is put + * into the listen queue (but before accept is called). + * This is the behavior of the BSD socket code. If + * connect does not return until accept is called, we + * will need to create another thread to call connect. + */ + if (connect(osfd[0], (struct sockaddr *) &selfAddr, + addrLen) == SOCKET_ERROR) { + goto failed; + } + /* + * A malicious local process may connect to the listening + * socket, so we need to verify that the accepted connection + * is made from our own socket osfd[0]. + */ + if (getsockname(osfd[0], (struct sockaddr *) &selfAddr, + &addrLen) == SOCKET_ERROR) { + goto failed; + } + osfd[1] = accept(listenSock, (struct sockaddr *) &peerAddr, &addrLen); + if (osfd[1] == INVALID_SOCKET) { + goto failed; + } + if (peerAddr.sin_port != selfAddr.sin_port) { + /* the connection we accepted is not from osfd[0] */ + PR_SetError(PR_INSUFFICIENT_RESOURCES_ERROR, 0); + goto failed; + } + closesocket(listenSock); + + f[0] = PR_AllocFileDesc(osfd[0], PR_GetTCPMethods()); + if (!f[0]) { + closesocket(osfd[0]); + closesocket(osfd[1]); + /* PR_AllocFileDesc() has invoked PR_SetError(). */ + return PR_FAILURE; + } + f[1] = PR_AllocFileDesc(osfd[1], PR_GetTCPMethods()); + if (!f[1]) { + PR_Close(f[0]); + closesocket(osfd[1]); + /* PR_AllocFileDesc() has invoked PR_SetError(). */ + return PR_FAILURE; + } + _PR_MD_INIT_FD_INHERITABLE(f[0], PR_FALSE); + _PR_MD_INIT_FD_INHERITABLE(f[1], PR_FALSE); + return PR_SUCCESS; + +failed: + if (listenSock != INVALID_SOCKET) { + closesocket(listenSock); + } + if (osfd[0] != INVALID_SOCKET) { + closesocket(osfd[0]); + } + if (osfd[1] != INVALID_SOCKET) { + closesocket(osfd[1]); + } + return PR_FAILURE; +#else /* not Unix or NT */ + /* + * default implementation + */ + PRFileDesc *listenSock; + PRNetAddr selfAddr, peerAddr; + PRUint16 port; + + f[0] = f[1] = NULL; + listenSock = PR_NewTCPSocket(); + if (listenSock == NULL) { + goto failed; + } + PR_InitializeNetAddr(PR_IpAddrLoopback, 0, &selfAddr); /* BugZilla: 35408 */ + if (PR_Bind(listenSock, &selfAddr) == PR_FAILURE) { + goto failed; + } + if (PR_GetSockName(listenSock, &selfAddr) == PR_FAILURE) { + goto failed; + } + port = ntohs(selfAddr.inet.port); + if (PR_Listen(listenSock, 5) == PR_FAILURE) { + goto failed; + } + f[0] = PR_NewTCPSocket(); + if (f[0] == NULL) { + goto failed; + } +#ifdef _PR_CONNECT_DOES_NOT_BIND + /* + * If connect does not implicitly bind the socket (e.g., on + * BeOS), we have to bind the socket so that we can get its + * port with getsockname later. + */ + PR_InitializeNetAddr(PR_IpAddrLoopback, 0, &selfAddr); + if (PR_Bind(f[0], &selfAddr) == PR_FAILURE) { + goto failed; + } +#endif + PR_InitializeNetAddr(PR_IpAddrLoopback, port, &selfAddr); + + /* + * Only a thread is used to do the connect and accept. + * I am relying on the fact that PR_Connect returns + * successfully as soon as the connect request is put + * into the listen queue (but before PR_Accept is called). + * This is the behavior of the BSD socket code. If + * connect does not return until accept is called, we + * will need to create another thread to call connect. + */ + if (PR_Connect(f[0], &selfAddr, PR_INTERVAL_NO_TIMEOUT) + == PR_FAILURE) { + goto failed; + } + /* + * A malicious local process may connect to the listening + * socket, so we need to verify that the accepted connection + * is made from our own socket f[0]. + */ + if (PR_GetSockName(f[0], &selfAddr) == PR_FAILURE) { + goto failed; + } + f[1] = PR_Accept(listenSock, &peerAddr, PR_INTERVAL_NO_TIMEOUT); + if (f[1] == NULL) { + goto failed; + } + if (peerAddr.inet.port != selfAddr.inet.port) { + /* the connection we accepted is not from f[0] */ + PR_SetError(PR_INSUFFICIENT_RESOURCES_ERROR, 0); + goto failed; + } + PR_Close(listenSock); + return PR_SUCCESS; + +failed: + if (listenSock) { + PR_Close(listenSock); + } + if (f[0]) { + PR_Close(f[0]); + } + if (f[1]) { + PR_Close(f[1]); + } + return PR_FAILURE; +#endif +} + +PR_IMPLEMENT(PROsfd) +PR_FileDesc2NativeHandle(PRFileDesc *fd) +{ + if (fd) { + fd = PR_GetIdentitiesLayer(fd, PR_NSPR_IO_LAYER); + } + if (!fd) { + PR_SetError(PR_INVALID_ARGUMENT_ERROR, 0); + return -1; + } + return fd->secret->md.osfd; +} + +PR_IMPLEMENT(void) +PR_ChangeFileDescNativeHandle(PRFileDesc *fd, PROsfd handle) +{ + if (fd) + fd->secret->md.osfd = handle; +} + +/* +** Select compatibility +** +*/ + +PR_IMPLEMENT(void) PR_FD_ZERO(PR_fd_set *set) +{ + memset(set, 0, sizeof(PR_fd_set)); +} + +PR_IMPLEMENT(void) PR_FD_SET(PRFileDesc *fh, PR_fd_set *set) +{ + PR_ASSERT( set->hsize < PR_MAX_SELECT_DESC ); + + set->harray[set->hsize++] = fh; +} + +PR_IMPLEMENT(void) PR_FD_CLR(PRFileDesc *fh, PR_fd_set *set) +{ + PRUint32 index, index2; + + for (index = 0; index<set->hsize; index++) + if (set->harray[index] == fh) { + for (index2=index; index2 < (set->hsize-1); index2++) { + set->harray[index2] = set->harray[index2+1]; + } + set->hsize--; + break; + } +} + +PR_IMPLEMENT(PRInt32) PR_FD_ISSET(PRFileDesc *fh, PR_fd_set *set) +{ + PRUint32 index; + for (index = 0; index<set->hsize; index++) + if (set->harray[index] == fh) { + return 1; + } + return 0; +} + +PR_IMPLEMENT(void) PR_FD_NSET(PROsfd fd, PR_fd_set *set) +{ + PR_ASSERT( set->nsize < PR_MAX_SELECT_DESC ); + + set->narray[set->nsize++] = fd; +} + +PR_IMPLEMENT(void) PR_FD_NCLR(PROsfd fd, PR_fd_set *set) +{ + PRUint32 index, index2; + + for (index = 0; index<set->nsize; index++) + if (set->narray[index] == fd) { + for (index2=index; index2 < (set->nsize-1); index2++) { + set->narray[index2] = set->narray[index2+1]; + } + set->nsize--; + break; + } +} + +PR_IMPLEMENT(PRInt32) PR_FD_NISSET(PROsfd fd, PR_fd_set *set) +{ + PRUint32 index; + for (index = 0; index<set->nsize; index++) + if (set->narray[index] == fd) { + return 1; + } + return 0; +} + + +#if !defined(NEED_SELECT) +#include "obsolete/probslet.h" + +#define PD_INCR 20 + +static PRPollDesc *_pr_setfd( + PR_fd_set *set, PRInt16 flags, PRPollDesc *polldesc) +{ + PRUintn fsidx, pdidx; + PRPollDesc *poll = polldesc; + + if (NULL == set) return poll; + + /* First set the pr file handle osfds */ + for (fsidx = 0; fsidx < set->hsize; fsidx++) + { + for (pdidx = 0; 1; pdidx++) + { + if ((PRFileDesc*)-1 == poll[pdidx].fd) + { + /* our vector is full - extend and condition it */ + poll = (PRPollDesc*)PR_Realloc( + poll, (pdidx + 1 + PD_INCR) * sizeof(PRPollDesc)); + if (NULL == poll) goto out_of_memory; + memset( + poll + pdidx * sizeof(PRPollDesc), + 0, PD_INCR * sizeof(PRPollDesc)); + poll[pdidx + PD_INCR].fd = (PRFileDesc*)-1; + } + if ((NULL == poll[pdidx].fd) + || (poll[pdidx].fd == set->harray[fsidx])) + { + /* PR_ASSERT(0 == (poll[pdidx].in_flags & flags)); */ + /* either empty or prevously defined */ + poll[pdidx].fd = set->harray[fsidx]; /* possibly redundant */ + poll[pdidx].in_flags |= flags; /* possibly redundant */ + break; + } + } + } + +#if 0 + /* Second set the native osfds */ + for (fsidx = 0; fsidx < set->nsize; fsidx++) + { + for (pdidx = 0; ((PRFileDesc*)-1 != poll[pdidx].fd); pdidx++) + { + if ((PRFileDesc*)-1 == poll[pdidx].fd) + { + /* our vector is full - extend and condition it */ + poll = PR_Realloc( + poll, (pdidx + PD_INCR) * sizeof(PRPollDesc)); + if (NULL == poll) goto out_of_memory; + memset( + poll + pdidx * sizeof(PRPollDesc), + 0, PD_INCR * sizeof(PRPollDesc)); + poll[(pdidx + PD_INCR)].fd = (PRFileDesc*)-1; + } + if ((NULL == poll[pdidx].fd) + || (poll[pdidx].fd == set->narray[fsidx])) + { + /* either empty or prevously defined */ + poll[pdidx].fd = set->narray[fsidx]; + PR_ASSERT(0 == (poll[pdidx].in_flags & flags)); + poll[pdidx].in_flags |= flags; + break; + } + } + } +#endif /* 0 */ + + return poll; + +out_of_memory: + if (NULL != polldesc) PR_DELETE(polldesc); + return NULL; +} /* _pr_setfd */ + +#endif /* !defined(NEED_SELECT) */ + +PR_IMPLEMENT(PRInt32) PR_Select( + PRInt32 unused, PR_fd_set *pr_rd, PR_fd_set *pr_wr, + PR_fd_set *pr_ex, PRIntervalTime timeout) +{ + +#if !defined(NEED_SELECT) + PRInt32 npds = 0; + /* + ** Find out how many fds are represented in the three lists. + ** Then allocate a polling descriptor for the logical union + ** (there can't be any overlapping) and call PR_Poll(). + */ + + PRPollDesc *copy, *poll; + + static PRBool warning = PR_TRUE; + if (warning) warning = _PR_Obsolete( "PR_Select()", "PR_Poll()"); + + /* try to get an initial guesss at how much space we need */ + npds = 0; + if ((NULL != pr_rd) && ((pr_rd->hsize + pr_rd->nsize - npds) > 0)) + npds = pr_rd->hsize + pr_rd->nsize; + if ((NULL != pr_wr) && ((pr_wr->hsize + pr_wr->nsize - npds) > 0)) + npds = pr_wr->hsize + pr_wr->nsize; + if ((NULL != pr_ex) && ((pr_ex->hsize + pr_ex->nsize - npds) > 0)) + npds = pr_ex->hsize + pr_ex->nsize; + + if (0 == npds) + { + PR_Sleep(timeout); + return 0; + } + + copy = poll = (PRPollDesc*)PR_Calloc(npds + PD_INCR, sizeof(PRPollDesc)); + if (NULL == poll) goto out_of_memory; + poll[npds + PD_INCR - 1].fd = (PRFileDesc*)-1; + + poll = _pr_setfd(pr_rd, PR_POLL_READ, poll); + if (NULL == poll) goto out_of_memory; + poll = _pr_setfd(pr_wr, PR_POLL_WRITE, poll); + if (NULL == poll) goto out_of_memory; + poll = _pr_setfd(pr_ex, PR_POLL_EXCEPT, poll); + if (NULL == poll) goto out_of_memory; + unused = 0; + while (NULL != poll[unused].fd && (PRFileDesc*)-1 != poll[unused].fd) + { + ++unused; + } + + PR_ASSERT(unused > 0); + npds = PR_Poll(poll, unused, timeout); + + if (npds > 0) + { + /* Copy the results back into the fd sets */ + if (NULL != pr_rd) pr_rd->nsize = pr_rd->hsize = 0; + if (NULL != pr_wr) pr_wr->nsize = pr_wr->hsize = 0; + if (NULL != pr_ex) pr_ex->nsize = pr_ex->hsize = 0; + for (copy = &poll[unused - 1]; copy >= poll; --copy) + { + if (copy->out_flags & PR_POLL_NVAL) + { + PR_SetError(PR_BAD_DESCRIPTOR_ERROR, 0); + npds = -1; + break; + } + if (copy->out_flags & PR_POLL_READ) + if (NULL != pr_rd) pr_rd->harray[pr_rd->hsize++] = copy->fd; + if (copy->out_flags & PR_POLL_WRITE) + if (NULL != pr_wr) pr_wr->harray[pr_wr->hsize++] = copy->fd; + if (copy->out_flags & PR_POLL_EXCEPT) + if (NULL != pr_ex) pr_ex->harray[pr_ex->hsize++] = copy->fd; + } + } + PR_DELETE(poll); + + return npds; +out_of_memory: + PR_SetError(PR_OUT_OF_MEMORY_ERROR, 0); + return -1; + +#endif /* !defined(NEED_SELECT) */ + +} |