/* * Copyright 2000, International Business Machines Corporation and others. * All Rights Reserved. * * This software has been released under the terms of the IBM Public * License. For details, see the LICENSE file in the top-level source * directory or online at http://www.openafs.org/dl/license10.html */ /* * rx_kcommon.c - Common kernel RX code for all system types. */ #include #include #include "rx/rx_kcommon.h" #include "rx_atomic.h" #include "rx_packet.h" #include "rx_internal.h" #include "rx_stats.h" #include "rx_peer.h" #ifdef AFS_HPUX110_ENV #include "h/tihdr.h" #include #endif #include "afsint.h" #ifndef RXK_LISTENER_ENV int (*rxk_PacketArrivalProc) (struct rx_packet * ahandle, struct sockaddr_in * afrom, struct socket *arock, afs_int32 asize); /* set to packet allocation procedure */ int (*rxk_GetPacketProc) (struct rx_packet **ahandle, int asize); #endif osi_socket *rxk_NewSocketHost(afs_uint32 ahost, short aport); extern struct interfaceAddr afs_cb_interface; rxk_ports_t rxk_ports; rxk_portRocks_t rxk_portRocks; int rxk_initDone = 0; #if !defined(AFS_SUN5_ENV) && !defined(AFS_SGI62_ENV) #define ADDRSPERSITE 16 static afs_uint32 myNetAddrs[ADDRSPERSITE]; static int myNetMTUs[ADDRSPERSITE]; static int numMyNetAddrs = 0; #endif #if defined(AFS_DARWIN80_ENV) #define sobind sock_bind #define soclose sock_close #endif /* add a port to the monitored list, port # is in network order */ static int rxk_AddPort(u_short aport, char *arock) { int i; unsigned short *tsp, ts; int zslot; zslot = -1; /* look for an empty slot simultaneously */ for (i = 0, tsp = rxk_ports; i < MAXRXPORTS; i++, tsp++) { if (((ts = *tsp) == 0) && (zslot == -1)) zslot = i; if (ts == aport) { return 0; } } /* otherwise allocate a new port slot */ if (zslot < 0) return E2BIG; /* all full */ rxk_ports[zslot] = aport; rxk_portRocks[zslot] = arock; return 0; } /* remove as port from the monitored list, port # is in network order */ int rxk_DelPort(u_short aport) { int i; unsigned short *tsp; for (i = 0, tsp = rxk_ports; i < MAXRXPORTS; i++, tsp++) { if (*tsp == aport) { /* found it, adjust ref count and free the port reference if all gone */ *tsp = 0; return 0; } } /* otherwise port not found */ return ENOENT; } void rxk_shutdownPorts(void) { int i; for (i = 0; i < MAXRXPORTS; i++) { if (rxk_ports[i]) { rxk_ports[i] = 0; #if ! defined(AFS_SUN5_ENV) && ! defined(UKERNEL) && ! defined(RXK_LISTENER_ENV) soclose((struct socket *)rxk_portRocks[i]); #endif rxk_portRocks[i] = NULL; } } } osi_socket rxi_GetHostUDPSocket(u_int host, u_short port) { osi_socket *sockp; sockp = (osi_socket *)rxk_NewSocketHost(host, port); if (sockp == (osi_socket *)0) return OSI_NULLSOCKET; rxk_AddPort(port, (char *)sockp); return (osi_socket) sockp; } osi_socket rxi_GetUDPSocket(u_short port) { return rxi_GetHostUDPSocket(htonl(INADDR_ANY), port); } /* * osi_utoa() - write the NUL-terminated ASCII decimal form of the given * unsigned long value into the given buffer. Returns 0 on success, * and a value less than 0 on failure. The contents of the buffer is * defined only on success. */ int osi_utoa(char *buf, size_t len, unsigned long val) { long k; /* index of first byte of string value */ /* we definitely need room for at least one digit and NUL */ if (len < 2) { return -1; } /* compute the string form from the high end of the buffer */ buf[len - 1] = '\0'; for (k = len - 2; k >= 0; k--) { buf[k] = val % 10 + '0'; val /= 10; if (val == 0) break; } /* did we finish converting val to string form? */ if (val != 0) { return -2; } /* this should never happen */ if (k < 0) { return -3; } /* this should never happen */ if (k >= len) { return -4; } /* if necessary, relocate string to beginning of buf[] */ if (k > 0) { /* * We need to achieve the effect of calling * * memmove(buf, &buf[k], len - k); * * However, since memmove() is not available in all * kernels, we explicitly do an appropriate copy. */ char *dst = buf; char *src = buf + k; while ((*dst++ = *src++) != '\0') continue; } return 0; } #ifndef AFS_LINUX26_ENV /* * osi_AssertFailK() -- used by the osi_Assert() macro. * * It essentially does * * osi_Panic("assertion failed: %s, file: %s, line: %d", expr, file, line); * * Since the kernel version of osi_Panic() only passes its first * argument to the native panic(), we construct a single string and hand * that to osi_Panic(). */ void osi_AssertFailK(const char *expr, const char *file, int line) { static const char msg0[] = "assertion failed: "; static const char msg1[] = ", file: "; static const char msg2[] = ", line: "; static const char msg3[] = "\n"; /* * These buffers add up to 1K, which is a pleasantly nice round * value, but probably not vital. */ char buf[1008]; char linebuf[16]; /* check line number conversion */ if (osi_utoa(linebuf, sizeof linebuf, line) < 0) { osi_Panic("osi_AssertFailK: error in osi_utoa()\n"); } /* okay, panic */ #define ADDBUF(BUF, STR) \ if (strlen(BUF) + strlen((char *)(STR)) + 1 <= sizeof BUF) { \ strcat(BUF, (char *)(STR)); \ } buf[0] = '\0'; ADDBUF(buf, msg0); ADDBUF(buf, expr); ADDBUF(buf, msg1); ADDBUF(buf, file); ADDBUF(buf, msg2); ADDBUF(buf, linebuf); ADDBUF(buf, msg3); #undef ADDBUF osi_Panic("%s", buf); } #endif #ifndef UKERNEL /* This is the server process request loop. Kernel server * processes never become listener threads */ void * rx_ServerProc(void *unused) { int threadID; rxi_MorePackets(rx_maxReceiveWindow + 2); /* alloc more packets */ MUTEX_ENTER(&rx_quota_mutex); rxi_dataQuota += rx_initSendWindow; /* Reserve some pkts for hard times */ /* threadID is used for making decisions in GetCall. Get it by bumping * number of threads handling incoming calls */ threadID = rxi_availProcs++; MUTEX_EXIT(&rx_quota_mutex); #ifdef RX_ENABLE_LOCKS AFS_GUNLOCK(); #endif /* RX_ENABLE_LOCKS */ rxi_ServerProc(threadID, NULL, NULL); #ifdef RX_ENABLE_LOCKS AFS_GLOCK(); #endif /* RX_ENABLE_LOCKS */ return NULL; } #endif /* !UKERNEL */ #ifndef RXK_LISTENER_ENV /* asize includes the Rx header */ static int MyPacketProc(struct rx_packet **ahandle, int asize) { struct rx_packet *tp; /* If this is larger than we expected, increase rx_maxReceiveDataSize */ /* If we can't scrounge enough cbufs, then we have to drop the packet, * but we should set a flag so we magic up some more at our leisure. */ if ((asize >= 0) && (asize <= RX_MAX_PACKET_SIZE)) { tp = rxi_AllocPacket(RX_PACKET_CLASS_RECEIVE); if (tp && (tp->length + RX_HEADER_SIZE) < asize) { if (0 < rxi_AllocDataBuf(tp, asize - (tp->length + RX_HEADER_SIZE), RX_PACKET_CLASS_RECV_CBUF)) { rxi_FreePacket(tp); tp = NULL; if (rx_stats_active) { rx_atomic_inc(&rx_stats.noPacketBuffersOnRead); } } } } else { /* * XXX if packet is too long for our buffer, * should do this at a higher layer and let other * end know we're losing. */ if (rx_stats_active) { rx_atomic_inc(&rx_stats.bogusPacketOnRead); } /* I DON"T LIKE THIS PRINTF -- PRINTFS MAKE THINGS VERY VERY SLOOWWW */ dpf(("rx: packet dropped: bad ulen=%d\n", asize)); tp = NULL; } if (!tp) return -1; /* otherwise we have a packet, set appropriate values */ *ahandle = tp; return 0; } static int MyArrivalProc(struct rx_packet *ahandle, struct sockaddr_in *afrom, struct socket *arock, afs_int32 asize) { /* handle basic rx packet */ ahandle->length = asize - RX_HEADER_SIZE; rxi_DecodePacketHeader(ahandle); ahandle = rxi_ReceivePacket(ahandle, arock, afrom->sin_addr.s_addr, afrom->sin_port, NULL, NULL); /* free the packet if it has been returned */ if (ahandle) rxi_FreePacket(ahandle); return 0; } #endif /* !RXK_LISTENER_ENV */ void rxi_StartListener(void) { #if !defined(RXK_LISTENER_ENV) && !defined(RXK_UPCALL_ENV) /* if kernel, give name of appropriate procedures */ rxk_GetPacketProc = MyPacketProc; rxk_PacketArrivalProc = MyArrivalProc; rxk_init(); #endif } /* Called from rxi_FindPeer, when initializing a clear rx_peer structure, to get interesting information. */ void rxi_InitPeerParams(struct rx_peer *pp) { u_short rxmtu; #ifndef AFS_SUN5_ENV # ifdef AFS_USERSPACE_IP_ADDR afs_int32 i; afs_int32 mtu; i = rxi_Findcbi(pp->host); if (i == -1) { rx_rto_setPeerTimeoutSecs(pp, 3); pp->ifMTU = MIN(RX_REMOTE_PACKET_SIZE, rx_MyMaxSendSize); } else { rx_rto_setPeerTimeoutSecs(pp, 2); pp->ifMTU = MIN(RX_MAX_PACKET_SIZE, rx_MyMaxSendSize); mtu = ntohl(afs_cb_interface.mtu[i]); /* Diminish the packet size to one based on the MTU given by * the interface. */ if (mtu > (RX_IPUDP_SIZE + RX_HEADER_SIZE)) { rxmtu = mtu - RX_IPUDP_SIZE; if (rxmtu < pp->ifMTU) pp->ifMTU = rxmtu; } } # else /* AFS_USERSPACE_IP_ADDR */ rx_ifnet_t ifn; # if !defined(AFS_SGI62_ENV) if (numMyNetAddrs == 0) (void)rxi_GetIFInfo(); # endif ifn = rxi_FindIfnet(pp->host, NULL); if (ifn) { rx_rto_setPeerTimeoutSecs(pp, 2); pp->ifMTU = MIN(RX_MAX_PACKET_SIZE, rx_MyMaxSendSize); # ifdef IFF_POINTOPOINT if (rx_ifnet_flags(ifn) & IFF_POINTOPOINT) { /* wish we knew the bit rate and the chunk size, sigh. */ rx_rto_setPeerTimeoutSecs(pp, 4); pp->ifMTU = RX_PP_PACKET_SIZE; } # endif /* IFF_POINTOPOINT */ /* Diminish the packet size to one based on the MTU given by * the interface. */ if (rx_ifnet_mtu(ifn) > (RX_IPUDP_SIZE + RX_HEADER_SIZE)) { rxmtu = rx_ifnet_mtu(ifn) - RX_IPUDP_SIZE; if (rxmtu < pp->ifMTU) pp->ifMTU = rxmtu; } } else { /* couldn't find the interface, so assume the worst */ rx_rto_setPeerTimeoutSecs(pp, 3); pp->ifMTU = MIN(RX_REMOTE_PACKET_SIZE, rx_MyMaxSendSize); } # endif /* else AFS_USERSPACE_IP_ADDR */ #else /* AFS_SUN5_ENV */ afs_int32 mtu; mtu = rxi_FindIfMTU(pp->host); if (mtu <= 0) { rx_rto_setPeerTimeoutSecs(pp, 3); pp->ifMTU = MIN(RX_REMOTE_PACKET_SIZE, rx_MyMaxSendSize); } else { rx_rto_setPeerTimeoutSecs(pp, 2); pp->ifMTU = MIN(RX_MAX_PACKET_SIZE, rx_MyMaxSendSize); /* Diminish the packet size to one based on the MTU given by * the interface. */ if (mtu > (RX_IPUDP_SIZE + RX_HEADER_SIZE)) { rxmtu = mtu - RX_IPUDP_SIZE; if (rxmtu < pp->ifMTU) pp->ifMTU = rxmtu; } } #endif /* AFS_SUN5_ENV */ pp->ifMTU = rxi_AdjustIfMTU(pp->ifMTU); pp->maxMTU = OLD_MAX_PACKET_SIZE; /* for compatibility with old guys */ pp->natMTU = MIN(pp->ifMTU, OLD_MAX_PACKET_SIZE); pp->ifDgramPackets = MIN(rxi_nDgramPackets, rxi_AdjustDgramPackets(rxi_nSendFrags, pp->ifMTU)); pp->maxDgramPackets = 1; /* Initialize slow start parameters */ pp->MTU = MIN(pp->natMTU, pp->maxMTU); pp->cwind = 1; pp->nDgramPackets = 1; pp->congestSeq = 0; } /* The following code is common to several system types, but not all. The * separate ones are found in the system specific subdirectories. */ #if ! defined(AFS_AIX_ENV) && ! defined(AFS_SUN5_ENV) && ! defined(UKERNEL) && ! defined(AFS_LINUX20_ENV) && !defined (AFS_DARWIN_ENV) && !defined (AFS_XBSD_ENV) /* Routine called during the afsd "-shutdown" process to put things back to * the initial state. */ static struct protosw parent_proto; /* udp proto switch */ void shutdown_rxkernel(void) { struct protosw *tpro, *last; last = inetdomain.dom_protoswNPROTOSW; for (tpro = inetdomain.dom_protosw; tpro < last; tpro++) if (tpro->pr_protocol == IPPROTO_UDP) { /* restore original udp protocol switch */ memcpy((void *)tpro, (void *)&parent_proto, sizeof(parent_proto)); memset((void *)&parent_proto, 0, sizeof(parent_proto)); rxk_initDone = 0; rxk_shutdownPorts(); return; } dpf(("shutdown_rxkernel: no udp proto\n")); } #endif /* !AIX && !SUN && !NCR && !UKERNEL */ #if !defined(AFS_SUN5_ENV) && !defined(AFS_SGI62_ENV) /* Determine what the network interfaces are for this machine. */ #ifdef AFS_USERSPACE_IP_ADDR int rxi_GetcbiInfo(void) { int i, j, different = 0, num = ADDRSPERSITE; int rxmtu, maxmtu; afs_uint32 ifinaddr; afs_uint32 addrs[ADDRSPERSITE]; int mtus[ADDRSPERSITE]; memset((void *)addrs, 0, sizeof(addrs)); memset((void *)mtus, 0, sizeof(mtus)); if (afs_cb_interface.numberOfInterfaces < num) num = afs_cb_interface.numberOfInterfaces; for (i = 0; i < num; i++) { if (!afs_cb_interface.mtu[i]) afs_cb_interface.mtu[i] = htonl(1500); rxmtu = (ntohl(afs_cb_interface.mtu[i]) - RX_IPUDP_SIZE); ifinaddr = ntohl(afs_cb_interface.addr_in[i]); if (myNetAddrs[i] != ifinaddr) different++; mtus[i] = rxmtu; rxmtu = rxi_AdjustIfMTU(rxmtu); maxmtu = rxmtu * rxi_nRecvFrags + ((rxi_nRecvFrags - 1) * UDP_HDR_SIZE); maxmtu = rxi_AdjustMaxMTU(rxmtu, maxmtu); addrs[i++] = ifinaddr; if (!rx_IsLoopbackAddr(ifinaddr) && (maxmtu > rx_maxReceiveSize)) { rx_maxReceiveSize = MIN(RX_MAX_PACKET_SIZE, maxmtu); rx_maxReceiveSize = MIN(rx_maxReceiveSize, rx_maxReceiveSizeUser); } } rx_maxJumboRecvSize = RX_HEADER_SIZE + (rxi_nDgramPackets * RX_JUMBOBUFFERSIZE) + ((rxi_nDgramPackets - 1) * RX_JUMBOHEADERSIZE); rx_maxJumboRecvSize = MAX(rx_maxJumboRecvSize, rx_maxReceiveSize); if (different) { for (j = 0; j < i; j++) { myNetMTUs[j] = mtus[j]; myNetAddrs[j] = addrs[j]; } } return different; } /* Returns the afs_cb_interface inxex which best matches address. * If none is found, we return -1. */ afs_int32 rxi_Findcbi(afs_uint32 addr) { int j; afs_uint32 myAddr, thisAddr, netMask, subnetMask; afs_int32 rvalue = -1; int match_value = 0; if (numMyNetAddrs == 0) (void)rxi_GetcbiInfo(); myAddr = ntohl(addr); if (IN_CLASSA(myAddr)) netMask = IN_CLASSA_NET; else if (IN_CLASSB(myAddr)) netMask = IN_CLASSB_NET; else if (IN_CLASSC(myAddr)) netMask = IN_CLASSC_NET; else netMask = 0; for (j = 0; j < afs_cb_interface.numberOfInterfaces; j++) { thisAddr = ntohl(afs_cb_interface.addr_in[j]); subnetMask = ntohl(afs_cb_interface.subnetmask[j]); if ((myAddr & netMask) == (thisAddr & netMask)) { if ((myAddr & subnetMask) == (thisAddr & subnetMask)) { if (myAddr == thisAddr) { match_value = 4; rvalue = j; break; } if (match_value < 3) { match_value = 3; rvalue = j; } } else { if (match_value < 2) { match_value = 2; rvalue = j; } } } } return (rvalue); } #else /* AFS_USERSPACE_IP_ADDR */ #if !defined(AFS_AIX41_ENV) && !defined(AFS_DUX40_ENV) && !defined(AFS_DARWIN_ENV) && !defined(AFS_XBSD_ENV) #define IFADDR2SA(f) (&((f)->ifa_addr)) #else /* AFS_AIX41_ENV */ #define IFADDR2SA(f) ((f)->ifa_addr) #endif int rxi_GetIFInfo(void) { int i = 0; int different = 0; int rxmtu, maxmtu; afs_uint32 addrs[ADDRSPERSITE]; int mtus[ADDRSPERSITE]; afs_uint32 ifinaddr; #if defined(AFS_DARWIN80_ENV) errno_t t; unsigned int count; int cnt=0, m, j; rx_ifaddr_t *ifads; rx_ifnet_t *ifns; struct sockaddr sout; struct sockaddr_in *sin; struct in_addr pin; #else rx_ifaddr_t ifad; /* ifnet points to a if_addrlist of ifaddrs */ rx_ifnet_t ifn; #endif memset(addrs, 0, sizeof(addrs)); memset(mtus, 0, sizeof(mtus)); #if defined(AFS_DARWIN80_ENV) if (!ifnet_list_get(AF_INET, &ifns, &count)) { for (m = 0; m < count; m++) { if (!ifnet_get_address_list(ifns[m], &ifads)) { for (j = 0; ifads[j] != NULL && cnt < ADDRSPERSITE; j++) { if ((t = ifaddr_address(ifads[j], &sout, sizeof(struct sockaddr))) == 0) { sin = (struct sockaddr_in *)&sout; rxmtu = rx_ifnet_mtu(rx_ifaddr_ifnet(ifads[j])) - RX_IPUDP_SIZE; ifinaddr = ntohl(sin->sin_addr.s_addr); if (myNetAddrs[i] != ifinaddr) { different++; } mtus[i] = rxmtu; rxmtu = rxi_AdjustIfMTU(rxmtu); maxmtu = rxmtu * rxi_nRecvFrags + ((rxi_nRecvFrags - 1) * UDP_HDR_SIZE); maxmtu = rxi_AdjustMaxMTU(rxmtu, maxmtu); addrs[i++] = ifinaddr; if (!rx_IsLoopbackAddr(ifinaddr) && (maxmtu > rx_maxReceiveSize)) { rx_maxReceiveSize = MIN(RX_MAX_PACKET_SIZE, maxmtu); rx_maxReceiveSize = MIN(rx_maxReceiveSize, rx_maxReceiveSizeUser); } cnt++; } } ifnet_free_address_list(ifads); } } ifnet_list_free(ifns); } #else #if defined(AFS_DARWIN_ENV) || defined(AFS_FBSD_ENV) #if defined(AFS_FBSD80_ENV) TAILQ_FOREACH(ifn, &V_ifnet, if_link) { #else TAILQ_FOREACH(ifn, &ifnet, if_link) { #endif if (i >= ADDRSPERSITE) break; #elif defined(AFS_OBSD_ENV) || defined(AFS_NBSD_ENV) for (ifn = ifnet.tqh_first; i < ADDRSPERSITE && ifn != NULL; ifn = ifn->if_list.tqe_next) { #else for (ifn = ifnet; ifn != NULL && i < ADDRSPERSITE; ifn = ifn->if_next) { #endif rxmtu = (ifn->if_mtu - RX_IPUDP_SIZE); #if defined(AFS_DARWIN_ENV) || defined(AFS_FBSD_ENV) TAILQ_FOREACH(ifad, &ifn->if_addrhead, ifa_link) { if (i >= ADDRSPERSITE) break; #elif defined(AFS_OBSD_ENV) || defined(AFS_NBSD_ENV) for (ifad = ifn->if_addrlist.tqh_first; ifad != NULL && i < ADDRSPERSITE; ifad = ifad->ifa_list.tqe_next) { #else for (ifad = ifn->if_addrlist; ifad != NULL && i < ADDRSPERSITE; ifad = ifad->ifa_next) { #endif if (IFADDR2SA(ifad)->sa_family == AF_INET) { ifinaddr = ntohl(((struct sockaddr_in *)IFADDR2SA(ifad))->sin_addr. s_addr); if (myNetAddrs[i] != ifinaddr) { different++; } mtus[i] = rxmtu; rxmtu = rxi_AdjustIfMTU(rxmtu); maxmtu = rxmtu * rxi_nRecvFrags + ((rxi_nRecvFrags - 1) * UDP_HDR_SIZE); maxmtu = rxi_AdjustMaxMTU(rxmtu, maxmtu); addrs[i++] = ifinaddr; if (!rx_IsLoopbackAddr(ifinaddr) && (maxmtu > rx_maxReceiveSize)) { rx_maxReceiveSize = MIN(RX_MAX_PACKET_SIZE, maxmtu); rx_maxReceiveSize = MIN(rx_maxReceiveSize, rx_maxReceiveSizeUser); } } } } #endif rx_maxJumboRecvSize = RX_HEADER_SIZE + rxi_nDgramPackets * RX_JUMBOBUFFERSIZE + (rxi_nDgramPackets - 1) * RX_JUMBOHEADERSIZE; rx_maxJumboRecvSize = MAX(rx_maxJumboRecvSize, rx_maxReceiveSize); if (different) { int l; for (l = 0; l < i; l++) { myNetMTUs[l] = mtus[l]; myNetAddrs[l] = addrs[l]; } } return different; } #if defined(AFS_DARWIN_ENV) || defined(AFS_XBSD_ENV) /* Returns ifnet which best matches address */ rx_ifnet_t rxi_FindIfnet(afs_uint32 addr, afs_uint32 * maskp) { struct sockaddr_in s, sr; rx_ifaddr_t ifad; s.sin_family = AF_INET; s.sin_addr.s_addr = addr; ifad = rx_ifaddr_withnet((struct sockaddr *)&s); if (ifad && maskp) { rx_ifaddr_netmask(ifad, (struct sockaddr *)&sr, sizeof(sr)); *maskp = sr.sin_addr.s_addr; } return (ifad ? rx_ifaddr_ifnet(ifad) : NULL); } #else /* DARWIN || XBSD */ /* Returns ifnet which best matches address */ rx_ifnet_t rxi_FindIfnet(afs_uint32 addr, afs_uint32 * maskp) { int match_value = 0; extern struct in_ifaddr *in_ifaddr; struct in_ifaddr *ifa, *ifad = NULL; addr = ntohl(addr); for (ifa = in_ifaddr; ifa; ifa = ifa->ia_next) { if ((addr & ifa->ia_netmask) == ifa->ia_net) { if ((addr & ifa->ia_subnetmask) == ifa->ia_subnet) { if (IA_SIN(ifa)->sin_addr.s_addr == addr) { /* ie, ME!!! */ match_value = 4; ifad = ifa; goto done; } if (match_value < 3) { ifad = ifa; match_value = 3; } } else { if (match_value < 2) { ifad = ifa; match_value = 2; } } } /* if net matches */ } /* for all in_ifaddrs */ done: if (ifad && maskp) *maskp = ifad->ia_subnetmask; return (ifad ? ifad->ia_ifp : NULL); } #endif /* else DARWIN || XBSD */ #endif /* else AFS_USERSPACE_IP_ADDR */ #endif /* !SUN5 && !SGI62 */ /* rxk_NewSocket, rxk_FreeSocket and osi_NetSend are from the now defunct * afs_osinet.c. One could argue that rxi_NewSocket could go into the * system specific subdirectories for all systems. But for the moment, * most of it is simple to follow common code. */ #if !defined(UKERNEL) #if !defined(AFS_SUN5_ENV) && !defined(AFS_LINUX20_ENV) /* rxk_NewSocket creates a new socket on the specified port. The port is * in network byte order. */ osi_socket * rxk_NewSocketHost(afs_uint32 ahost, short aport) { afs_int32 code; #ifdef AFS_DARWIN80_ENV socket_t newSocket; #else struct socket *newSocket; #endif #if (!defined(AFS_HPUX1122_ENV) && !defined(AFS_FBSD_ENV)) struct mbuf *nam; #endif struct sockaddr_in myaddr; #ifdef AFS_HPUX110_ENV /* prototype copied from kernel source file streams/str_proto.h */ extern MBLKP allocb_wait(int, int); MBLKP bindnam; int addrsize = sizeof(struct sockaddr_in); struct file *fp; extern struct fileops socketops; #endif #ifdef AFS_SGI65_ENV bhv_desc_t bhv; #endif AFS_STATCNT(osi_NewSocket); #if (defined(AFS_DARWIN_ENV) || defined(AFS_XBSD_ENV)) && defined(KERNEL_FUNNEL) thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); #endif AFS_ASSERT_GLOCK(); AFS_GUNLOCK(); #if defined(AFS_HPUX102_ENV) #if defined(AFS_HPUX110_ENV) /* we need a file associated with the socket so sosend in NetSend * will not fail */ /* blocking socket */ code = socreate(AF_INET, &newSocket, SOCK_DGRAM, 0, 0); fp = falloc(); if (!fp) goto bad; fp->f_flag = FREAD | FWRITE; fp->f_type = DTYPE_SOCKET; fp->f_ops = &socketops; fp->f_data = (void *)newSocket; newSocket->so_fp = (void *)fp; #else /* AFS_HPUX110_ENV */ code = socreate(AF_INET, &newSocket, SOCK_DGRAM, 0, SS_NOWAIT); #endif /* else AFS_HPUX110_ENV */ #elif defined(AFS_SGI65_ENV) || defined(AFS_OBSD_ENV) code = socreate(AF_INET, &newSocket, SOCK_DGRAM, IPPROTO_UDP); #elif defined(AFS_FBSD_ENV) code = socreate(AF_INET, &newSocket, SOCK_DGRAM, IPPROTO_UDP, afs_osi_credp, curthread); #elif defined(AFS_DARWIN80_ENV) #ifdef RXK_LISTENER_ENV code = sock_socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP, NULL, NULL, &newSocket); #else code = sock_socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP, rx_upcall, NULL, &newSocket); #endif #elif defined(AFS_NBSD50_ENV) code = socreate(AF_INET, &newSocket, SOCK_DGRAM, 0, osi_curproc(), NULL); #elif defined(AFS_NBSD40_ENV) code = socreate(AF_INET, &newSocket, SOCK_DGRAM, 0, osi_curproc()); #else code = socreate(AF_INET, &newSocket, SOCK_DGRAM, 0); #endif /* AFS_HPUX102_ENV */ if (code) goto bad; memset(&myaddr, 0, sizeof myaddr); myaddr.sin_family = AF_INET; myaddr.sin_port = aport; myaddr.sin_addr.s_addr = ahost; #ifdef STRUCT_SOCKADDR_HAS_SA_LEN myaddr.sin_len = sizeof(myaddr); #endif #ifdef AFS_HPUX110_ENV bindnam = allocb_wait((addrsize + SO_MSGOFFSET + 1), BPRI_MED); if (!bindnam) { setuerror(ENOBUFS); goto bad; } memcpy((caddr_t) bindnam->b_rptr + SO_MSGOFFSET, (caddr_t) & myaddr, addrsize); bindnam->b_wptr = bindnam->b_rptr + (addrsize + SO_MSGOFFSET + 1); #if defined(AFS_NBSD40_ENV) code = sobind(newSocket, bindnam, addrsize, osi_curproc()); #else code = sobind(newSocket, bindnam, addrsize); #endif if (code) { soclose(newSocket); #if !defined(AFS_HPUX1122_ENV) m_freem(nam); #endif goto bad; } freeb(bindnam); #else /* AFS_HPUX110_ENV */ #if defined(AFS_DARWIN80_ENV) { int buflen = 50000; int i,code2; for (i=0;i<2;i++) { code = sock_setsockopt(newSocket, SOL_SOCKET, SO_SNDBUF, &buflen, sizeof(buflen)); code2 = sock_setsockopt(newSocket, SOL_SOCKET, SO_RCVBUF, &buflen, sizeof(buflen)); if (!code && !code2) break; if (i == 2) osi_Panic("osi_NewSocket: last attempt to reserve 32K failed!\n"); buflen = 32766; } } #else #if defined(AFS_NBSD_ENV) solock(newSocket); #endif code = soreserve(newSocket, 50000, 50000); if (code) { code = soreserve(newSocket, 32766, 32766); if (code) osi_Panic("osi_NewSocket: last attempt to reserve 32K failed!\n"); } #if defined(AFS_NBSD_ENV) sounlock(newSocket); #endif #endif #if defined(AFS_DARWIN_ENV) || defined(AFS_FBSD_ENV) #if defined(AFS_FBSD_ENV) code = sobind(newSocket, (struct sockaddr *)&myaddr, curthread); #else code = sobind(newSocket, (struct sockaddr *)&myaddr); #endif if (code) { dpf(("sobind fails (%d)\n", (int)code)); soclose(newSocket); goto bad; } #else /* defined(AFS_DARWIN_ENV) || defined(AFS_FBSD_ENV) */ #ifdef AFS_OSF_ENV nam = m_getclr(M_WAIT, MT_SONAME); #else /* AFS_OSF_ENV */ nam = m_get(M_WAIT, MT_SONAME); #endif if (nam == NULL) { #if defined(KERNEL_HAVE_UERROR) setuerror(ENOBUFS); #endif goto bad; } nam->m_len = sizeof(myaddr); memcpy(mtod(nam, caddr_t), &myaddr, sizeof(myaddr)); #if defined(AFS_SGI65_ENV) BHV_PDATA(&bhv) = (void *)newSocket; code = sobind(&bhv, nam); m_freem(nam); #elif defined(AFS_OBSD44_ENV) || defined(AFS_NBSD40_ENV) code = sobind(newSocket, nam, osi_curproc()); #else code = sobind(newSocket, nam); #endif if (code) { dpf(("sobind fails (%d)\n", (int)code)); soclose(newSocket); #ifndef AFS_SGI65_ENV m_freem(nam); #endif goto bad; } #endif /* else AFS_DARWIN_ENV */ #endif /* else AFS_HPUX110_ENV */ AFS_GLOCK(); #if defined(AFS_DARWIN_ENV) && defined(KERNEL_FUNNEL) thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); #endif return (osi_socket *)newSocket; bad: AFS_GLOCK(); #if defined(AFS_DARWIN_ENV) && defined(KERNEL_FUNNEL) thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); #endif return (osi_socket *)0; } osi_socket * rxk_NewSocket(short aport) { return rxk_NewSocketHost(0, aport); } /* free socket allocated by rxk_NewSocket */ int rxk_FreeSocket(struct socket *asocket) { AFS_STATCNT(osi_FreeSocket); #if defined(AFS_DARWIN_ENV) && defined(KERNEL_FUNNEL) thread_funnel_switch(KERNEL_FUNNEL, NETWORK_FUNNEL); #endif #ifdef AFS_HPUX110_ENV if (asocket->so_fp) { struct file *fp = asocket->so_fp; #if !defined(AFS_HPUX1123_ENV) /* 11.23 still has falloc, but not FPENTRYFREE ! * so for now if we shutdown, we will waist a file * structure */ FPENTRYFREE(fp); asocket->so_fp = NULL; #endif } #endif /* AFS_HPUX110_ENV */ soclose(asocket); #if defined(AFS_DARWIN_ENV) && defined(KERNEL_FUNNEL) thread_funnel_switch(NETWORK_FUNNEL, KERNEL_FUNNEL); #endif return 0; } #endif /* !SUN5 && !LINUX20 */ #if defined(RXK_LISTENER_ENV) || defined(AFS_SUN5_ENV) || defined(RXK_UPCALL_ENV) #ifdef RXK_TIMEDSLEEP_ENV /* Shutting down should wake us up, as should an earlier event. */ void rxi_ReScheduleEvents(void) { /* needed to allow startup */ int glock = ISAFS_GLOCK(); if (!glock) AFS_GLOCK(); osi_rxWakeup(&afs_termState); if (!glock) AFS_GUNLOCK(); } #endif /* * Run RX event daemon every second (5 times faster than rest of systems) */ void afs_rxevent_daemon(void) { struct clock temp; SPLVAR; while (1) { #ifdef RX_ENABLE_LOCKS AFS_GUNLOCK(); #endif /* RX_ENABLE_LOCKS */ NETPRI; rxevent_RaiseEvents(&temp); USERPRI; #ifdef RX_ENABLE_LOCKS AFS_GLOCK(); #endif /* RX_ENABLE_LOCKS */ #ifdef RX_KERNEL_TRACE afs_Trace1(afs_iclSetp, CM_TRACE_TIMESTAMP, ICL_TYPE_STRING, "before afs_osi_Wait()"); #endif #ifdef RXK_TIMEDSLEEP_ENV afs_osi_TimedSleep(&afs_termState, MAX(500, ((temp.sec * 1000) + (temp.usec / 1000))), 0); #else afs_osi_Wait(500, NULL, 0); #endif #ifdef RX_KERNEL_TRACE afs_Trace1(afs_iclSetp, CM_TRACE_TIMESTAMP, ICL_TYPE_STRING, "after afs_osi_Wait()"); #endif if (afs_termState == AFSOP_STOP_RXEVENT) { #ifdef RXK_LISTENER_ENV afs_termState = AFSOP_STOP_RXK_LISTENER; #elif defined(AFS_SUN510_ENV) || defined(RXK_UPCALL_ENV) afs_termState = AFSOP_STOP_NETIF; #else afs_termState = AFSOP_STOP_COMPLETE; #endif osi_rxWakeup(&afs_termState); return; } } } #endif #ifdef RXK_LISTENER_ENV /* rxk_ReadPacket returns 1 if valid packet, 0 on error. */ int rxk_ReadPacket(osi_socket so, struct rx_packet *p, int *host, int *port) { int code; struct sockaddr_in from; int nbytes; afs_int32 rlen; afs_int32 tlen; afs_int32 savelen; /* was using rlen but had aliasing problems */ rx_computelen(p, tlen); rx_SetDataSize(p, tlen); /* this is the size of the user data area */ tlen += RX_HEADER_SIZE; /* now this is the size of the entire packet */ rlen = rx_maxJumboRecvSize; /* this is what I am advertising. Only check * it once in order to avoid races. */ tlen = rlen - tlen; if (tlen > 0) { tlen = rxi_AllocDataBuf(p, tlen, RX_PACKET_CLASS_RECV_CBUF); if (tlen > 0) { tlen = rlen - tlen; } else tlen = rlen; } else tlen = rlen; /* add some padding to the last iovec, it's just to make sure that the * read doesn't return more data than we expect, and is done to get around * our problems caused by the lack of a length field in the rx header. */ savelen = p->wirevec[p->niovecs - 1].iov_len; p->wirevec[p->niovecs - 1].iov_len = savelen + RX_EXTRABUFFERSIZE; nbytes = tlen + sizeof(afs_int32); #ifdef RX_KERNEL_TRACE if (ICL_SETACTIVE(afs_iclSetp)) { AFS_GLOCK(); afs_Trace1(afs_iclSetp, CM_TRACE_TIMESTAMP, ICL_TYPE_STRING, "before osi_NetRecive()"); AFS_GUNLOCK(); } #endif code = osi_NetReceive(rx_socket, &from, p->wirevec, p->niovecs, &nbytes); #ifdef RX_KERNEL_TRACE if (ICL_SETACTIVE(afs_iclSetp)) { AFS_GLOCK(); afs_Trace1(afs_iclSetp, CM_TRACE_TIMESTAMP, ICL_TYPE_STRING, "after osi_NetRecive()"); AFS_GUNLOCK(); } #endif /* restore the vec to its correct state */ p->wirevec[p->niovecs - 1].iov_len = savelen; if (!code) { p->length = nbytes - RX_HEADER_SIZE;; if ((nbytes > tlen) || (p->length & 0x8000)) { /* Bogus packet */ if (nbytes <= 0) { if (rx_stats_active) { MUTEX_ENTER(&rx_stats_mutex); rx_atomic_inc(&rx_stats.bogusPacketOnRead); rx_stats.bogusHost = from.sin_addr.s_addr; MUTEX_EXIT(&rx_stats_mutex); } dpf(("B: bogus packet from [%x,%d] nb=%d\n", from.sin_addr.s_addr, from.sin_port, nbytes)); } return -1; } else { /* Extract packet header. */ rxi_DecodePacketHeader(p); *host = from.sin_addr.s_addr; *port = from.sin_port; if (p->header.type > 0 && p->header.type < RX_N_PACKET_TYPES) { if (rx_stats_active) { rx_atomic_inc(&rx_stats.packetsRead[p->header.type - 1]); } } #ifdef RX_TRIMDATABUFS /* Free any empty packet buffers at the end of this packet */ rxi_TrimDataBufs(p, 1); #endif return 0; } } else return code; } /* rxk_Listener() * * Listen for packets on socket. This thread is typically started after * rx_Init has called rxi_StartListener(), but nevertheless, ensures that * the start state is set before proceeding. * * Note that this thread is outside the AFS global lock for much of * it's existence. * * In many OS's, the socket receive code sleeps interruptibly. That's not what * we want here. So we need to either block all signals (including SIGKILL * and SIGSTOP) or reset the thread's signal state to unsignalled when the * OS's socket receive routine returns as a result of a signal. */ int rxk_ListenerPid; /* Used to signal process to wakeup at shutdown */ #ifdef AFS_LINUX20_ENV struct task_struct *rxk_ListenerTask; #endif void rxk_Listener(void) { struct rx_packet *rxp = NULL; int code; int host, port; #ifdef AFS_LINUX20_ENV rxk_ListenerPid = current->pid; rxk_ListenerTask = current; #endif #ifdef AFS_SUN5_ENV rxk_ListenerPid = 1; /* No PID, just a flag that we're alive */ #endif /* AFS_SUN5_ENV */ #ifdef AFS_XBSD_ENV rxk_ListenerPid = curproc->p_pid; #endif /* AFS_FBSD_ENV */ #ifdef AFS_DARWIN80_ENV rxk_ListenerPid = proc_selfpid(); #elif defined(AFS_DARWIN_ENV) rxk_ListenerPid = current_proc()->p_pid; #endif #ifdef RX_ENABLE_LOCKS AFS_GUNLOCK(); #endif /* RX_ENABLE_LOCKS */ while (afs_termState != AFSOP_STOP_RXK_LISTENER) { /* See if a check for additional packets was issued */ rx_CheckPackets(); if (rxp) { rxi_RestoreDataBufs(rxp); } else { rxp = rxi_AllocPacket(RX_PACKET_CLASS_RECEIVE); if (!rxp) osi_Panic("rxk_Listener: No more Rx buffers!\n"); } if (!(code = rxk_ReadPacket(rx_socket, rxp, &host, &port))) { rxp = rxi_ReceivePacket(rxp, rx_socket, host, port, 0, 0); } } #ifdef RX_ENABLE_LOCKS AFS_GLOCK(); #endif /* RX_ENABLE_LOCKS */ if (afs_termState == AFSOP_STOP_RXK_LISTENER) { #ifdef AFS_SUN510_ENV afs_termState = AFSOP_STOP_NETIF; #else afs_termState = AFSOP_STOP_COMPLETE; #endif osi_rxWakeup(&afs_termState); } rxk_ListenerPid = 0; #ifdef AFS_LINUX20_ENV rxk_ListenerTask = 0; osi_rxWakeup(&rxk_ListenerTask); #endif #if defined(AFS_SUN5_ENV) || defined(AFS_FBSD_ENV) osi_rxWakeup(&rxk_ListenerPid); #endif } #if !defined(AFS_LINUX20_ENV) && !defined(AFS_SUN5_ENV) && !defined(AFS_DARWIN_ENV) && !defined(AFS_XBSD_ENV) /* The manner of stopping the rx listener thread may vary. Most unix's should * be able to call soclose. */ void osi_StopListener(void) { soclose(rx_socket); } #endif #endif /* RXK_LISTENER_ENV */ #endif /* !NCR && !UKERNEL */ #if !defined(AFS_LINUX26_ENV) void #if defined(AFS_AIX_ENV) osi_Panic(char *msg, void *a1, void *a2, void *a3) #else osi_Panic(char *msg, ...) #endif { #ifdef AFS_AIX_ENV if (!msg) msg = "Unknown AFS panic"; /* * we should probably use the errsave facility here. it is not * varargs-aware */ printf(msg, a1, a2, a3); panic(msg); #elif defined(AFS_SGI_ENV) va_list ap; /* Solaris has vcmn_err, Sol10 01/06 may have issues. Beware. */ if (!msg) { cmn_err(CE_PANIC, "Unknown AFS panic"); } else { va_start(ap, msg); icmn_err(CE_PANIC, msg, ap); va_end(ap); } #elif defined(AFS_DARWIN80_ENV) || (defined(AFS_LINUX22_ENV) && !defined(AFS_LINUX_26_ENV)) char buf[256]; va_list ap; if (!msg) msg = "Unknown AFS panic"; va_start(ap, msg); vsnprintf(buf, sizeof(buf), msg, ap); va_end(ap); printf("%s", buf); panic(buf); #else va_list ap; if (!msg) msg = "Unknown AFS panic"; va_start(ap, msg); vprintf(msg, ap); va_end(ap); # ifdef AFS_LINUX20_ENV * ((char *) 0) = 0; # else panic(msg); # endif #endif } #endif