/* * Copyright (C) 1996-2024 The Squid Software Foundation and contributors * * Squid software is distributed under GPLv2+ license and includes * contributions from numerous individuals and organizations. * Please see the COPYING and CONTRIBUTORS files for details. */ /* DEBUG: section 44 Peer Selection Algorithm */ #include "squid.h" #include "acl/FilledChecklist.h" #include "base/AsyncCbdataCalls.h" #include "base/InstanceId.h" #include "base/TypeTraits.h" #include "CachePeer.h" #include "carp.h" #include "client_side.h" #include "dns/LookupDetails.h" #include "errorpage.h" #include "event.h" #include "FwdState.h" #include "globals.h" #include "hier_code.h" #include "htcp.h" #include "http/Stream.h" #include "HttpRequest.h" #include "icmp/net_db.h" #include "ICP.h" #include "ip/tools.h" #include "ipcache.h" #include "neighbors.h" #include "peer_sourcehash.h" #include "peer_userhash.h" #include "PeerSelectState.h" #include "SquidConfig.h" #include "Store.h" #include "time/gadgets.h" /** * A CachePeer which has been selected as a possible destination. * Listed as pointers here so as to prevent duplicates being added but will * be converted to a set of IP address path options before handing back out * to the caller. * * Certain connection flags and outgoing settings will also be looked up and * set based on the received request and CachePeer settings before handing back. */ class FwdServer { MEMPROXY_CLASS(FwdServer); public: FwdServer(CachePeer *p, hier_code c) : _peer(p), code(c), next(nullptr) {} CbcPointer _peer; /* NULL --> origin server */ hier_code code; FwdServer *next; }; static struct { int timeouts; } PeerStats; static const char *DirectStr[] = { "DIRECT_UNKNOWN", "DIRECT_NO", "DIRECT_MAYBE", "DIRECT_YES" }; /// a helper class to report a selected destination (for debugging) class PeerSelectionDumper { public: PeerSelectionDumper(const PeerSelector * const aSelector, const CachePeer * const aPeer, const hier_code aCode): selector(aSelector), peer(aPeer), code(aCode) {} const PeerSelector * const selector; ///< selection parameters const CachePeer * const peer; ///< successful selection info const hier_code code; ///< selection algorithm }; CBDATA_CLASS_INIT(PeerSelector); /// prints PeerSelectionDumper (for debugging) static std::ostream & operator <<(std::ostream &os, const PeerSelectionDumper &fsd) { os << hier_code_str[fsd.code]; if (fsd.peer) os << '/' << *fsd.peer; else if (fsd.selector) // useful for DIRECT and gone PINNED destinations os << '#' << fsd.selector->request->url.host(); return os; } /// An ICP ping timeout service. /// Protects event.cc (which is designed to handle a few unrelated timeouts) /// from exposure to thousands of ping-related timeouts on busy proxies. class PeerSelectorPingMonitor { public: /// registers the given selector to be notified about the IPC ping timeout void monitor(PeerSelector *); /// removes a PeerSelector from the waiting list void forget(PeerSelector *); /// \returns a (nil) registration of a non-waiting peer selector WaitingPeerSelectorPosition npos() { return selectors.end(); } private: static void NoteWaitOver(void *monitor); void startWaiting(); void abortWaiting(); void noteWaitOver(); WaitingPeerSelectors selectors; ///< \see WaitingPeerSelectors }; /// monitors all PeerSelector ICP ping timeouts static PeerSelectorPingMonitor & PingMonitor() { static const auto Instance = new PeerSelectorPingMonitor(); return *Instance; } /* PeerSelectorPingMonitor */ /// PeerSelectorPingMonitor::noteWaitOver() wrapper void PeerSelectorPingMonitor::NoteWaitOver(void *raw) { assert(raw); static_cast(raw)->noteWaitOver(); } /// schedules a single event to represent all waiting selectors void PeerSelectorPingMonitor::startWaiting() { assert(!selectors.empty()); const auto interval = tvSubDsec(current_time, selectors.begin()->first); eventAdd("PeerSelectorPingMonitor::NoteWaitOver", &PeerSelectorPingMonitor::NoteWaitOver, this, interval, 0, false); } /// undoes an earlier startWaiting() call void PeerSelectorPingMonitor::abortWaiting() { // our event may be already in the AsyncCallQueue but that is OK: // such queued calls cannot accumulate, and we ignore any stale ones eventDelete(&PeerSelectorPingMonitor::NoteWaitOver, nullptr); } /// calls back all ready PeerSelectors and continues to wait for others void PeerSelectorPingMonitor::noteWaitOver() { while (!selectors.empty() && current_time >= selectors.begin()->first) { const auto selector = selectors.begin()->second; CallBack(selector->al, [selector,this] { selector->ping.monitorRegistration = npos(); AsyncCall::Pointer callback = asyncCall(44, 4, "PeerSelector::HandlePingTimeout", cbdataDialer(PeerSelector::HandlePingTimeout, selector)); ScheduleCallHere(callback); }); selectors.erase(selectors.begin()); } if (!selectors.empty()) { // Since abortWaiting() is unreliable, we may have been awakened by a // stale event A after event B has been scheduled. Now we are going to // schedule event C. Prevent event accumulation by deleting B (if any). abortWaiting(); startWaiting(); } } void PeerSelectorPingMonitor::monitor(PeerSelector *selector) { assert(selector); const auto deadline = selector->ping.deadline(); const auto position = selectors.emplace(deadline, selector); selector->ping.monitorRegistration = position; if (position == selectors.begin()) { if (selectors.size() > 1) abortWaiting(); // remove the previously scheduled earlier event startWaiting(); } // else the already scheduled event is still the earliest one } void PeerSelectorPingMonitor::forget(PeerSelector *selector) { assert(selector); if (selector->ping.monitorRegistration == npos()) return; // already forgotten const auto wasFirst = selector->ping.monitorRegistration == selectors.begin(); selectors.erase(selector->ping.monitorRegistration); selector->ping.monitorRegistration = npos(); if (wasFirst) { // do not reschedule if there are still elements with the same deadline if (!selectors.empty() && selectors.begin()->first == selector->ping.deadline()) return; abortWaiting(); if (!selectors.empty()) startWaiting(); } // else do nothing since the old scheduled event is still the earliest one } /* PeerSelector */ PeerSelector::~PeerSelector() { while (servers) { FwdServer *next = servers->next; delete servers; servers = next; } cancelPingTimeoutMonitoring(); if (entry) { debugs(44, 3, entry->url()); entry->ping_status = PING_DONE; } if (acl_checklist) { debugs(44, DBG_IMPORTANT, "ERROR: Squid BUG: peer selector gone while waiting for a slow ACL"); delete acl_checklist; } HTTPMSGUNLOCK(request); if (entry) { assert(entry->ping_status != PING_WAITING); entry->unlock("peerSelect"); entry = nullptr; } delete lastError; } void PeerSelector::startPingWaiting() { assert(entry); assert(entry->ping_status != PING_WAITING); PingMonitor().monitor(this); entry->ping_status = PING_WAITING; } void PeerSelector::cancelPingTimeoutMonitoring() { PingMonitor().forget(this); } static int peerSelectIcpPing(PeerSelector *ps, int direct, StoreEntry * entry) { assert(ps); HttpRequest *request = ps->request; int n; assert(entry); assert(entry->ping_status == PING_NONE); assert(direct != DIRECT_YES); debugs(44, 3, entry->url()); if (!request->flags.hierarchical && direct != DIRECT_NO) return 0; if (EBIT_TEST(entry->flags, KEY_PRIVATE) && !neighbors_do_private_keys) if (direct != DIRECT_NO) return 0; n = neighborsCount(ps); debugs(44, 3, "counted " << n << " neighbors"); return n; } static void peerSelect(PeerSelectionInitiator *initiator, HttpRequest * request, AccessLogEntry::Pointer const &al, StoreEntry * entry) { if (entry) debugs(44, 3, *entry << ' ' << entry->url()); else debugs(44, 3, request->method); const auto selector = new PeerSelector(initiator); selector->request = request; HTTPMSGLOCK(selector->request); selector->al = al; selector->entry = entry; #if USE_CACHE_DIGESTS request->hier.peer_select_start = current_time; #endif if (selector->entry) selector->entry->lock("peerSelect"); selector->selectMore(); } void PeerSelectionInitiator::startSelectingDestinations(HttpRequest *request, const AccessLogEntry::Pointer &ale, StoreEntry *entry) { subscribed = true; peerSelect(this, request, ale, entry); // and wait for noteDestination() and/or noteDestinationsEnd() calls } void PeerSelector::checkNeverDirectDone(const Acl::Answer answer) { acl_checklist = nullptr; debugs(44, 3, answer); never_direct = answer; switch (answer) { case ACCESS_ALLOWED: /** if never_direct says YES, do that. */ direct = DIRECT_NO; debugs(44, 3, "direct = " << DirectStr[direct] << " (never_direct allow)"); break; case ACCESS_DENIED: // not relevant. case ACCESS_DUNNO: // not relevant. break; case ACCESS_AUTH_REQUIRED: debugs(44, DBG_IMPORTANT, "WARNING: never_direct resulted in " << answer << ". Username ACLs are not reliable here."); break; } selectMore(); } void PeerSelector::CheckNeverDirectDone(Acl::Answer answer, void *data) { static_cast(data)->checkNeverDirectDone(answer); } void PeerSelector::checkAlwaysDirectDone(const Acl::Answer answer) { acl_checklist = nullptr; debugs(44, 3, answer); always_direct = answer; switch (answer) { case ACCESS_ALLOWED: /** if always_direct says YES, do that. */ direct = DIRECT_YES; debugs(44, 3, "direct = " << DirectStr[direct] << " (always_direct allow)"); break; case ACCESS_DENIED: // not relevant. case ACCESS_DUNNO: // not relevant. break; case ACCESS_AUTH_REQUIRED: debugs(44, DBG_IMPORTANT, "WARNING: always_direct resulted in " << answer << ". Username ACLs are not reliable here."); break; } selectMore(); } void PeerSelector::CheckAlwaysDirectDone(Acl::Answer answer, void *data) { static_cast(data)->checkAlwaysDirectDone(answer); } /// \returns true (after destroying "this") if the peer initiator is gone /// \returns false (without side effects) otherwise bool PeerSelector::selectionAborted() { if (interestedInitiator()) return false; debugs(44, 3, "Aborting peer selection: Initiator gone or lost interest."); delete this; return true; } /// A single DNS resolution loop iteration: Converts selected FwdServer to IPs. void PeerSelector::resolveSelected() { if (selectionAborted()) return; FwdServer *fs = servers; // Bug 3243: CVE 2009-0801 // Bypass of browser same-origin access control in intercepted communication // To resolve this we must use only the original client destination when going DIRECT // on intercepted traffic which failed Host verification const HttpRequest *req = request; const bool isIntercepted = !req->flags.redirected && (req->flags.intercepted || req->flags.interceptTproxy); const bool useOriginalDst = Config.onoff.client_dst_passthru || !req->flags.hostVerified; const bool choseDirect = fs && fs->code == HIER_DIRECT; if (isIntercepted && useOriginalDst && choseDirect) { // check the client is still around before using any of its details if (req->clientConnectionManager.valid()) { // construct a "result" adding the ORIGINAL_DST to the set instead of DIRECT Comm::ConnectionPointer p = new Comm::Connection(); p->remote = req->clientConnectionManager->clientConnection->local; fs->code = ORIGINAL_DST; // fs->code is DIRECT. This fixes the display. handlePath(p, *fs); } // clear the used fs and continue servers = fs->next; delete fs; resolveSelected(); return; } if (fs && fs->code == PINNED) { // Nil path signals a PINNED destination selection. Our initiator should // borrow and use clientConnectionManager's pinned connection object // (regardless of that connection destination). handlePath(nullptr, *fs); servers = fs->next; delete fs; resolveSelected(); return; } // convert the list of FwdServer destinations into destinations IP addresses if (fs && wantsMoreDestinations()) { // send the next one off for DNS lookup. const char *host = fs->_peer.valid() ? fs->_peer->host : request->url.host(); debugs(44, 2, "Find IP destination for: " << url() << "' via " << host); Dns::nbgethostbyname(host, this); return; } // Bug 3605: clear any extra listed FwdServer destinations, when the options exceeds max_foward_tries. // due to the allocation method of fs, we must deallocate each manually. // TODO: use a std::list so we can get the size and abort adding whenever the selection loops reach Config.forward_max_tries if (fs) { assert(fs == servers); while (fs) { servers = fs->next; delete fs; fs = servers; } } // done with DNS lookups. pass back to caller debugs(44, 2, id << " found all " << foundPaths << " destinations for " << url()); debugs(44, 2, " always_direct = " << always_direct); debugs(44, 2, " never_direct = " << never_direct); debugs(44, 2, " timedout = " << ping.timedout); ping.stop = current_time; request->hier.ping = ping; // final result if (lastError && foundPaths) { // nobody cares about errors if we found destinations despite them debugs(44, 3, "forgetting the last error"); delete lastError; lastError = nullptr; } if (const auto initiator = interestedInitiator()) initiator->noteDestinationsEnd(lastError); lastError = nullptr; // initiator owns the ErrorState object now delete this; } void PeerSelector::noteLookup(const Dns::LookupDetails &details) { /* ignore lookup delays that occurred after the initiator moved on */ if (selectionAborted()) return; if (!wantsMoreDestinations()) return; request->recordLookup(details); } void PeerSelector::noteIp(const Ip::Address &ip) { if (selectionAborted()) return; if (!wantsMoreDestinations()) return; const auto peer = servers->_peer.valid(); // for TPROXY spoofing, we must skip unusable addresses if (request->flags.spoofClientIp && !(peer && peer->options.no_tproxy) ) { if (ip.isIPv4() != request->client_addr.isIPv4()) return; // cannot spoof the client address on this link } Comm::ConnectionPointer p = new Comm::Connection(); p->remote = ip; p->remote.port(peer ? peer->http_port : request->url.port()); handlePath(p, *servers); } void PeerSelector::noteIps(const Dns::CachedIps *ia, const Dns::LookupDetails &details) { if (selectionAborted()) return; FwdServer *fs = servers; if (!ia) { debugs(44, 3, "Unknown host: " << (fs->_peer.valid() ? fs->_peer->host : request->url.host())); // discard any previous error. delete lastError; lastError = nullptr; if (fs->code == HIER_DIRECT) { lastError = new ErrorState(ERR_DNS_FAIL, Http::scServiceUnavailable, request, al); lastError->dnsError = details.error; } } // else noteIp() calls have already processed all IPs in *ia servers = fs->next; delete fs; // continue resolving selected peers resolveSelected(); } int PeerSelector::checkNetdbDirect() { #if USE_ICMP CachePeer *p; int myrtt; int myhops; if (direct == DIRECT_NO) return 0; /* base lookup on RTT and Hops if ICMP NetDB is enabled. */ myrtt = netdbHostRtt(request->url.host()); debugs(44, 3, "MY RTT = " << myrtt << " msec"); debugs(44, 3, "minimum_direct_rtt = " << Config.minDirectRtt << " msec"); if (myrtt && myrtt <= Config.minDirectRtt) return 1; myhops = netdbHostHops(request->url.host()); debugs(44, 3, "MY hops = " << myhops); debugs(44, 3, "minimum_direct_hops = " << Config.minDirectHops); if (myhops && myhops <= Config.minDirectHops) return 1; p = whichPeer(closest_parent_miss); if (p == nullptr) return 0; debugs(44, 3, "closest_parent_miss RTT = " << ping.p_rtt << " msec"); if (myrtt && myrtt <= ping.p_rtt) return 1; #endif /* USE_ICMP */ return 0; } void PeerSelector::selectMore() { if (selectionAborted()) return; debugs(44, 3, request->method << ' ' << request->url.host()); /** If we don't know whether DIRECT is permitted ... */ if (direct == DIRECT_UNKNOWN) { if (always_direct == ACCESS_DUNNO) { debugs(44, 3, "direct = " << DirectStr[direct] << " (always_direct to be checked)"); /** check always_direct; */ ACLFilledChecklist *ch = new ACLFilledChecklist(Config.accessList.AlwaysDirect, request, nullptr); ch->al = al; acl_checklist = ch; acl_checklist->syncAle(request, nullptr); acl_checklist->nonBlockingCheck(CheckAlwaysDirectDone, this); return; } else if (never_direct == ACCESS_DUNNO) { debugs(44, 3, "direct = " << DirectStr[direct] << " (never_direct to be checked)"); /** check never_direct; */ ACLFilledChecklist *ch = new ACLFilledChecklist(Config.accessList.NeverDirect, request, nullptr); ch->al = al; acl_checklist = ch; acl_checklist->syncAle(request, nullptr); acl_checklist->nonBlockingCheck(CheckNeverDirectDone, this); return; } else if (request->flags.noDirect) { /** if we are accelerating, direct is not an option. */ direct = DIRECT_NO; debugs(44, 3, "direct = " << DirectStr[direct] << " (forced non-direct)"); } else if (request->flags.loopDetected) { /** if we are in a forwarding-loop, direct is not an option. */ direct = DIRECT_YES; debugs(44, 3, "direct = " << DirectStr[direct] << " (forwarding loop detected)"); } else if (checkNetdbDirect()) { direct = DIRECT_YES; debugs(44, 3, "direct = " << DirectStr[direct] << " (checkNetdbDirect)"); } else { direct = DIRECT_MAYBE; debugs(44, 3, "direct = " << DirectStr[direct] << " (default)"); } debugs(44, 3, "direct = " << DirectStr[direct]); } if (!entry || entry->ping_status == PING_NONE) selectPinned(); if (entry == nullptr) { (void) 0; } else if (entry->ping_status == PING_NONE) { selectSomeNeighbor(); if (entry->ping_status == PING_WAITING) return; } else if (entry->ping_status == PING_WAITING) { selectSomeNeighborReplies(); cancelPingTimeoutMonitoring(); entry->ping_status = PING_DONE; } switch (direct) { case DIRECT_YES: selectSomeDirect(); break; case DIRECT_NO: selectSomeParent(); selectAllParents(); break; default: if (Config.onoff.prefer_direct) selectSomeDirect(); if (request->flags.hierarchical || !Config.onoff.nonhierarchical_direct) { selectSomeParent(); selectAllParents(); } if (!Config.onoff.prefer_direct) selectSomeDirect(); break; } // end peer selection; start resolving selected peers resolveSelected(); } bool peerAllowedToUse(const CachePeer *, PeerSelector*); /// Selects a pinned connection if it exists, is valid, and is allowed. void PeerSelector::selectPinned() { // TODO: Avoid all repeated calls. Relying on PING_DONE is not enough. if (!request->pinnedConnection()) return; const auto peer = request->pinnedConnection()->pinnedPeer(); const auto usePinned = peer ? peerAllowedToUse(peer, this) : (direct != DIRECT_NO); // If the pinned connection is prohibited (for this request) then // the initiator must decide whether it is OK to open a new one instead. request->pinnedConnection()->pinning.peerAccessDenied = !usePinned; addSelection(peer, PINNED); if (entry) entry->ping_status = PING_DONE; // skip ICP } /** * Selects a neighbor (parent or sibling) based on one of the * following methods: * Cache Digests * CARP * ICMP Netdb RTT estimates * ICP/HTCP queries */ void PeerSelector::selectSomeNeighbor() { CachePeer *p; hier_code code = HIER_NONE; assert(entry->ping_status == PING_NONE); if (direct == DIRECT_YES) { entry->ping_status = PING_DONE; return; } #if USE_CACHE_DIGESTS if ((p = neighborsDigestSelect(this))) { if (neighborType(p, request->url) == PEER_PARENT) code = CD_PARENT_HIT; else code = CD_SIBLING_HIT; } else #endif if ((p = netdbClosestParent(this))) { code = CLOSEST_PARENT; } else if (peerSelectIcpPing(this, direct, entry)) { debugs(44, 3, "Doing ICP pings"); ping.start = current_time; ping.n_sent = neighborsUdpPing(request, entry, HandlePingReply, this, &ping.n_replies_expected, &ping.timeout); // TODO: Refactor neighborsUdpPing() to guarantee positive timeouts. if (ping.timeout < 0) ping.timeout = 0; if (ping.n_sent == 0) debugs(44, DBG_CRITICAL, "WARNING: neighborsUdpPing returned 0"); debugs(44, 3, ping.n_replies_expected << " ICP replies expected, RTT " << ping.timeout << " msec"); if (ping.n_replies_expected > 0) { startPingWaiting(); return; } } if (code != HIER_NONE) { assert(p); addSelection(p, code); } entry->ping_status = PING_DONE; } /// Selects a neighbor (parent or sibling) based on ICP/HTCP replies. void PeerSelector::selectSomeNeighborReplies() { CachePeer *p = nullptr; hier_code code = HIER_NONE; assert(entry->ping_status == PING_WAITING); assert(direct != DIRECT_YES); if (checkNetdbDirect()) { code = CLOSEST_DIRECT; addSelection(nullptr, code); return; } if ((p = hit)) { code = hit_type == PEER_PARENT ? PARENT_HIT : SIBLING_HIT; } else { if (!closest_parent_miss.isAnyAddr()) { p = whichPeer(closest_parent_miss); code = CLOSEST_PARENT_MISS; } else if (!first_parent_miss.isAnyAddr()) { p = whichPeer(first_parent_miss); code = FIRST_PARENT_MISS; } } if (p && code != HIER_NONE) { addSelection(p, code); } } /// Adds a "direct" entry if the request can be forwarded to the origin server. void PeerSelector::selectSomeDirect() { if (direct == DIRECT_NO) return; /* WAIS is not implemented natively */ if (request->url.getScheme() == AnyP::PROTO_WAIS) return; addSelection(nullptr, HIER_DIRECT); } void PeerSelector::selectSomeParent() { CachePeer *p; hier_code code = HIER_NONE; debugs(44, 3, request->method << ' ' << request->url.host()); if (direct == DIRECT_YES) return; if ((p = peerSourceHashSelectParent(this))) { code = SOURCEHASH_PARENT; #if USE_AUTH } else if ((p = peerUserHashSelectParent(this))) { code = USERHASH_PARENT; #endif } else if ((p = carpSelectParent(this))) { code = CARP; } else if ((p = getRoundRobinParent(this))) { code = ROUNDROBIN_PARENT; } else if ((p = getWeightedRoundRobinParent(this))) { code = ROUNDROBIN_PARENT; } else if ((p = getFirstUpParent(this))) { code = FIRSTUP_PARENT; } else if ((p = getDefaultParent(this))) { code = DEFAULT_PARENT; } if (code != HIER_NONE) { addSelection(p, code); } } /// Adds alive parents. Used as a last resort for never_direct. void PeerSelector::selectAllParents() { CachePeer *p; /* Add all alive parents */ for (p = Config.peers; p; p = p->next) { /* XXX: neighbors.c lacks a public interface for enumerating * parents to a request so we have to dig some here.. */ if (neighborType(p, request->url) != PEER_PARENT) continue; if (!peerHTTPOkay(p, this)) continue; addSelection(p, ANY_OLD_PARENT); } /* XXX: should add dead parents here, but it is currently * not possible to find out which parents are dead or which * simply are not configured to handle the request. */ /* Add default parent as a last resort */ if ((p = getDefaultParent(this))) { addSelection(p, DEFAULT_PARENT); } } void PeerSelector::handlePingTimeout() { debugs(44, 3, url()); // do nothing if ping reply came while handlePingTimeout() was queued if (!entry || entry->ping_status != PING_WAITING) return; entry->ping_status = PING_DONE; if (selectionAborted()) return; ++PeerStats.timeouts; ping.timedout = 1; selectMore(); } void PeerSelector::HandlePingTimeout(PeerSelector *selector) { selector->handlePingTimeout(); } void peerSelectInit(void) { memset(&PeerStats, '\0', sizeof(PeerStats)); } void PeerSelector::handleIcpParentMiss(CachePeer *p, icp_common_t *header) { int rtt; #if USE_ICMP if (Config.onoff.query_icmp) { if (header->flags & ICP_FLAG_SRC_RTT) { rtt = header->pad & 0xFFFF; int hops = (header->pad >> 16) & 0xFFFF; if (rtt > 0 && rtt < 0xFFFF) netdbUpdatePeer(request->url, p, rtt, hops); if (rtt && (ping.p_rtt == 0 || rtt < ping.p_rtt)) { closest_parent_miss = p->in_addr; ping.p_rtt = rtt; } } } #else (void)header; #endif /* USE_ICMP */ /* if closest-only is set, then don't allow FIRST_PARENT_MISS */ if (p->options.closest_only) return; /* set FIRST_MISS if there is no CLOSEST parent */ if (!closest_parent_miss.isAnyAddr()) return; rtt = (tvSubMsec(ping.start, current_time) - p->basetime) / p->weight; if (rtt < 1) rtt = 1; if (first_parent_miss.isAnyAddr() || rtt < ping.w_rtt) { first_parent_miss = p->in_addr; ping.w_rtt = rtt; } } void PeerSelector::handleIcpReply(CachePeer *p, const peer_t type, icp_common_t *header) { icp_opcode op = header->getOpCode(); debugs(44, 3, icp_opcode_str[op] << ' ' << url()); #if USE_CACHE_DIGESTS && 0 /* do cd lookup to count false misses */ if (p && request) peerNoteDigestLookup(request, p, peerDigestLookup(p, this)); #endif ++ping.n_recv; if (op == ICP_MISS || op == ICP_DECHO) { if (type == PEER_PARENT) handleIcpParentMiss(p, header); } else if (op == ICP_HIT) { hit = p; hit_type = type; selectMore(); return; } if (ping.n_recv < ping.n_replies_expected) return; selectMore(); } #if USE_HTCP void PeerSelector::handleHtcpReply(CachePeer *p, const peer_t type, HtcpReplyData *htcp) { debugs(44, 3, (htcp->hit ? "HIT" : "MISS") << ' ' << url()); ++ping.n_recv; if (htcp->hit) { hit = p; hit_type = type; selectMore(); return; } if (type == PEER_PARENT) handleHtcpParentMiss(p, htcp); if (ping.n_recv < ping.n_replies_expected) return; selectMore(); } void PeerSelector::handleHtcpParentMiss(CachePeer *p, HtcpReplyData *htcp) { int rtt; #if USE_ICMP if (Config.onoff.query_icmp) { if (htcp->cto.rtt > 0) { rtt = (int) htcp->cto.rtt * 1000; int hops = (int) htcp->cto.hops * 1000; netdbUpdatePeer(request->url, p, rtt, hops); if (rtt && (ping.p_rtt == 0 || rtt < ping.p_rtt)) { closest_parent_miss = p->in_addr; ping.p_rtt = rtt; } } } #else (void)htcp; #endif /* USE_ICMP */ /* if closest-only is set, then don't allow FIRST_PARENT_MISS */ if (p->options.closest_only) return; /* set FIRST_MISS if there is no CLOSEST parent */ if (!closest_parent_miss.isAnyAddr()) return; rtt = (tvSubMsec(ping.start, current_time) - p->basetime) / p->weight; if (rtt < 1) rtt = 1; if (first_parent_miss.isAnyAddr() || rtt < ping.w_rtt) { first_parent_miss = p->in_addr; ping.w_rtt = rtt; } } #endif void PeerSelector::HandlePingReply(CachePeer * p, peer_t type, AnyP::ProtocolType proto, void *pingdata, void *data) { if (proto == AnyP::PROTO_ICP) static_cast(data)->handleIcpReply(p, type, static_cast(pingdata)); #if USE_HTCP else if (proto == AnyP::PROTO_HTCP) static_cast(data)->handleHtcpReply(p, type, static_cast(pingdata)); #endif else debugs(44, DBG_IMPORTANT, "ERROR: ignoring an ICP reply with unknown protocol " << proto); } void PeerSelector::addSelection(CachePeer *peer, const hier_code code) { // Find the end of the servers list. Bail on a duplicate destination. auto **serversTail = &servers; while (const auto server = *serversTail) { // There can be at most one PINNED destination. // Non-PINNED destinations are uniquely identified by their CachePeer // (even though a DIRECT destination might match a cache_peer address). // TODO: We may still add duplicates because the same peer could have // been removed from `servers` already (and given to the requestor). const bool duplicate = (server->code == PINNED) ? (code == PINNED) : (server->_peer == peer); if (duplicate) { debugs(44, 3, "skipping " << PeerSelectionDumper(this, peer, code) << "; have " << PeerSelectionDumper(this, server->_peer.get(), server->code)); return; } serversTail = &server->next; } debugs(44, 3, "adding " << PeerSelectionDumper(this, peer, code)); *serversTail = new FwdServer(peer, code); } PeerSelector::PeerSelector(PeerSelectionInitiator *initiator): request(nullptr), entry (nullptr), always_direct(Config.accessList.AlwaysDirect?ACCESS_DUNNO:ACCESS_DENIED), never_direct(Config.accessList.NeverDirect?ACCESS_DUNNO:ACCESS_DENIED), direct(DIRECT_UNKNOWN), lastError(nullptr), servers (nullptr), first_parent_miss(), closest_parent_miss(), hit(nullptr), hit_type(PEER_NONE), acl_checklist (nullptr), initiator_(initiator) { ; // no local defaults. } const SBuf PeerSelector::url() const { if (entry) return SBuf(entry->url()); if (request) return request->effectiveRequestUri(); static const SBuf noUrl("[no URL]"); return noUrl; } /// \returns valid/interested peer initiator or nil PeerSelectionInitiator * PeerSelector::interestedInitiator() { const auto initiator = initiator_.valid(); if (!initiator) { debugs(44, 3, id << " initiator gone"); return nullptr; } if (!initiator->subscribed) { debugs(44, 3, id << " initiator lost interest"); return nullptr; } debugs(44, 7, id); return initiator; } bool PeerSelector::wantsMoreDestinations() const { const auto maxCount = Config.forward_max_tries; return maxCount >= 0 && foundPaths < static_cast(maxCount); } void PeerSelector::handlePath(const Comm::ConnectionPointer &path, FwdServer &fs) { ++foundPaths; if (path) { path->peerType = fs.code; path->setPeer(fs._peer.get()); // check for a configured outgoing address for this destination... getOutgoingAddress(request, path); debugs(44, 2, id << " found " << path << ", destination #" << foundPaths << " for " << url()); } else debugs(44, 2, id << " found pinned, destination #" << foundPaths << " for " << url()); request->hier.ping = ping; // may be updated later debugs(44, 2, " always_direct = " << always_direct); debugs(44, 2, " never_direct = " << never_direct); debugs(44, 2, " timedout = " << ping.timedout); if (const auto initiator = interestedInitiator()) initiator->noteDestination(path); } InstanceIdDefinitions(PeerSelector, "PeerSelector"); ping_data::ping_data() : n_sent(0), n_recv(0), n_replies_expected(0), timeout(0), timedout(0), w_rtt(0), p_rtt(0), monitorRegistration(PingMonitor().npos()) { start.tv_sec = 0; start.tv_usec = 0; stop.tv_sec = 0; stop.tv_usec = 0; } timeval ping_data::deadline() const { timeval timeInterval; timeInterval.tv_sec = timeout / 1000; timeInterval.tv_usec = (timeout % 1000) * 1000; timeval result; tvAdd(result, start, timeInterval); return result; }