| /* |
| * |
| * Copyright (c) 2021 Project CHIP Authors |
| * |
| * Licensed under the Apache License, Version 2.0 (the "License"); |
| * you may not use this file except in compliance with the License. |
| * You may obtain a copy of the License at |
| * |
| * http://www.apache.org/licenses/LICENSE-2.0 |
| * |
| * Unless required by applicable law or agreed to in writing, software |
| * distributed under the License is distributed on an "AS IS" BASIS, |
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| * See the License for the specific language governing permissions and |
| * limitations under the License. |
| */ |
| |
| #pragma once |
| |
| #include <cstddef> |
| #include <cstdint> |
| #include <optional> |
| |
| #include <lib/core/PeerId.h> |
| #include <lib/dnssd/Resolver.h> |
| #include <lib/dnssd/minimal_mdns/core/HeapQName.h> |
| #include <lib/support/Variant.h> |
| #include <system/SystemClock.h> |
| |
| namespace mdns { |
| namespace Minimal { |
| |
| /// Keeps track of active resolve attempts |
| /// |
| /// Maintains a list of 'pending mdns resolve queries' and provides operations |
| /// for: |
| /// - add/remove to the list |
| /// - figuring out a 'next query time' for items in the list |
| /// - iterating through the 'schedule now' items of the list |
| /// |
| class ActiveResolveAttempts |
| { |
| public: |
| static constexpr size_t kRetryQueueSize = 4; |
| static constexpr chip::System::Clock::Timeout kMaxRetryDelay = chip::System::Clock::Seconds16(16); |
| |
| struct ScheduledAttempt |
| { |
| struct Browse |
| { |
| Browse(const chip::Dnssd::DiscoveryFilter discoveryFilter, const chip::Dnssd::DiscoveryType discoveryType) : |
| filter(discoveryFilter), type(discoveryType) |
| {} |
| chip::Dnssd::DiscoveryFilter filter; |
| chip::Dnssd::DiscoveryType type; |
| }; |
| |
| struct Resolve |
| { |
| chip::PeerId peerId; |
| uint32_t consumerCount = 0; |
| |
| Resolve(chip::PeerId id) : peerId(id) {} |
| }; |
| |
| struct IpResolve |
| { |
| HeapQName hostName; |
| IpResolve(HeapQName && host) : hostName(std::move(host)) {} |
| }; |
| |
| ScheduledAttempt() |
| { |
| static_assert(sizeof(Resolve) <= sizeof(Browse) || sizeof(Resolve) <= sizeof(HeapQName), |
| "Figure out where to put the Resolve counter so that Resolve is not making ScheduledAttempt bigger than " |
| "it has to be anyway to handle the other attempt types."); |
| } |
| ScheduledAttempt(const chip::PeerId & peer, bool first) : |
| resolveData(chip::InPlaceTemplateType<Resolve>(), peer), firstSend(first) |
| {} |
| ScheduledAttempt(const chip::Dnssd::DiscoveryFilter discoveryFilter, const chip::Dnssd::DiscoveryType type, bool first) : |
| resolveData(chip::InPlaceTemplateType<Browse>(), discoveryFilter, type), firstSend(first) |
| {} |
| |
| ScheduledAttempt(IpResolve && ipResolve, bool first) : |
| resolveData(chip::InPlaceTemplateType<IpResolve>(), ipResolve), firstSend(first) |
| {} |
| |
| bool operator==(const ScheduledAttempt & other) const { return Matches(other) && other.firstSend == firstSend; } |
| bool Matches(const ScheduledAttempt & other) const |
| { |
| if (!resolveData.Valid()) |
| { |
| return !other.resolveData.Valid(); |
| } |
| |
| if (resolveData.Is<Browse>()) |
| { |
| if (!other.resolveData.Is<Browse>()) |
| { |
| return false; |
| } |
| |
| auto & a = resolveData.Get<Browse>(); |
| auto & b = other.resolveData.Get<Browse>(); |
| return (a.filter == b.filter && a.type == b.type); |
| } |
| |
| if (resolveData.Is<Resolve>()) |
| { |
| if (!other.resolveData.Is<Resolve>()) |
| { |
| return false; |
| } |
| auto & a = resolveData.Get<Resolve>(); |
| auto & b = other.resolveData.Get<Resolve>(); |
| |
| return a.peerId == b.peerId; |
| } |
| |
| if (resolveData.Is<IpResolve>()) |
| { |
| if (!other.resolveData.Is<IpResolve>()) |
| { |
| return false; |
| } |
| auto & a = resolveData.Get<IpResolve>(); |
| auto & b = other.resolveData.Get<IpResolve>(); |
| |
| return a.hostName == b.hostName; |
| } |
| return false; |
| } |
| |
| bool MatchesIpResolve(SerializedQNameIterator hostName) const |
| { |
| return resolveData.Is<IpResolve>() && (hostName == resolveData.Get<IpResolve>().hostName.Content()); |
| } |
| bool Matches(const chip::PeerId & peer) const |
| { |
| return resolveData.Is<Resolve>() && (resolveData.Get<Resolve>().peerId == peer); |
| } |
| bool Matches(const chip::Dnssd::DiscoveredNodeData & data, chip::Dnssd::DiscoveryType type) const |
| { |
| if (!resolveData.Is<Browse>()) |
| { |
| return false; |
| } |
| |
| auto & browse = resolveData.Get<Browse>(); |
| |
| if (browse.type != type) |
| { |
| return false; |
| } |
| auto & nodeData = data.Get<chip::Dnssd::CommissionNodeData>(); |
| |
| switch (browse.filter.type) |
| { |
| case chip::Dnssd::DiscoveryFilterType::kNone: |
| return true; |
| case chip::Dnssd::DiscoveryFilterType::kShortDiscriminator: |
| return browse.filter.code == static_cast<uint64_t>((nodeData.longDiscriminator >> 8) & 0x0F); |
| case chip::Dnssd::DiscoveryFilterType::kLongDiscriminator: |
| return browse.filter.code == nodeData.longDiscriminator; |
| case chip::Dnssd::DiscoveryFilterType::kVendorId: |
| return browse.filter.code == nodeData.vendorId; |
| case chip::Dnssd::DiscoveryFilterType::kDeviceType: |
| return browse.filter.code == nodeData.deviceType; |
| case chip::Dnssd::DiscoveryFilterType::kCommissioningMode: |
| return browse.filter.code == nodeData.commissioningMode; |
| case chip::Dnssd::DiscoveryFilterType::kInstanceName: |
| return strncmp(browse.filter.instanceName, nodeData.instanceName, |
| chip::Dnssd::Commission::kInstanceNameMaxLength + 1) == 0; |
| case chip::Dnssd::DiscoveryFilterType::kCommissioner: |
| case chip::Dnssd::DiscoveryFilterType::kCompressedFabricId: |
| default: |
| // These are for other discovery types. |
| return false; |
| } |
| } |
| |
| bool IsEmpty() const { return !resolveData.Valid(); } |
| bool IsResolve() const { return resolveData.Is<Resolve>(); } |
| bool IsBrowse() const { return resolveData.Is<Browse>(); } |
| bool IsIpResolve() const { return resolveData.Is<IpResolve>(); } |
| void Clear() { resolveData = DataType(); } |
| |
| // Called when this scheduled attempt will replace an existing scheduled |
| // attempt, because either they match or we have run out of attempt |
| // slots. When this happens, we want to propagate the consumer count |
| // from the existing attempt to the new one, if they're matching resolve |
| // attempts. |
| void WillCoalesceWith(const ScheduledAttempt & existing) |
| { |
| if (!IsResolve()) |
| { |
| // Consumer count is only tracked for resolve requests |
| return; |
| } |
| |
| if (!existing.Matches(*this)) |
| { |
| // Out of attempt slots, so just dropping the existing attempt. |
| return; |
| } |
| |
| // Adding another consumer to the same query. Propagate along the |
| // consumer count to our new attempt, which will replace the |
| // existing one. |
| resolveData.Get<Resolve>().consumerCount = existing.resolveData.Get<Resolve>().consumerCount + 1; |
| } |
| |
| void ConsumerRemoved() |
| { |
| if (!IsResolve()) |
| { |
| return; |
| } |
| |
| auto & count = resolveData.Get<Resolve>().consumerCount; |
| if (count > 0) |
| { |
| --count; |
| } |
| |
| if (count == 0) |
| { |
| Clear(); |
| } |
| } |
| |
| const Browse & BrowseData() const { return resolveData.Get<Browse>(); } |
| const Resolve & ResolveData() const { return resolveData.Get<Resolve>(); } |
| const IpResolve & IpResolveData() const { return resolveData.Get<IpResolve>(); } |
| |
| using DataType = chip::Variant<Browse, Resolve, IpResolve>; |
| |
| DataType resolveData; |
| |
| // First packet send is marked separately: minMDNS logic can choose |
| // to first send a unicast query followed by a multicast one. |
| bool firstSend = false; |
| }; |
| |
| ActiveResolveAttempts(chip::System::Clock::ClockBase * clock) : mClock(clock) { Reset(); } |
| |
| /// Clear out the internal queue |
| void Reset(); |
| |
| /// Mark a resolution as a success, removing it from the internal list |
| void Complete(const chip::PeerId & peerId); |
| void CompleteIpResolution(SerializedQNameIterator targetHostName); |
| |
| /// Mark all browse-type scheduled attemptes as a success, removing them |
| /// from the internal list. |
| CHIP_ERROR CompleteAllBrowses(); |
| |
| /// Note that resolve attempts for the given peer id now have one fewer |
| /// consumer. |
| void NodeIdResolutionNoLongerNeeded(const chip::PeerId & peerId); |
| |
| /// Mark that a resolution is pending, adding it to the internal list |
| /// |
| /// Once this complete, this peer id will be returned immediately |
| /// by NextScheduled (potentially with others as well) |
| void MarkPending(const chip::PeerId & peerId); |
| void MarkPending(const chip::Dnssd::DiscoveryFilter & filter, const chip::Dnssd::DiscoveryType type); |
| void MarkPending(ScheduledAttempt::IpResolve && resolve); |
| |
| // Get minimum time until the next pending reply is required. |
| // |
| // Returns missing if no actively tracked elements exist. |
| std::optional<chip::System::Clock::Timeout> GetTimeUntilNextExpectedResponse() const; |
| |
| // Get the peer Id that needs scheduling for a query |
| // |
| // Assumes that the resolution is being sent and will apply internal |
| // query logic. This means: |
| // - internal tracking of 'next due time' will updated as 'request sent |
| // now' |
| // - there is NO sorting implied by this call. Returned value will be |
| // any peer that needs a new request sent |
| std::optional<ScheduledAttempt> NextScheduled(); |
| |
| /// Check if any of the pending queries are for the given host name for |
| /// IP resolution. |
| bool IsWaitingForIpResolutionFor(SerializedQNameIterator hostName) const; |
| |
| /// Determines if address resolution for the given peer ID is required |
| /// |
| /// IP Addresses are required for active operational discovery of specific peers |
| /// or if an active browse is being performed. |
| bool ShouldResolveIpAddress(chip::PeerId peerId) const; |
| |
| /// Check if a browse operation is active for the given discovery type |
| bool HasBrowseFor(chip::Dnssd::DiscoveryType type) const; |
| |
| private: |
| struct RetryEntry |
| { |
| ScheduledAttempt attempt; |
| // When a reply is expected for this item |
| chip::System::Clock::Timestamp queryDueTime; |
| |
| // Next expected delay for sending if reply is not reached by |
| // 'queryDueTimeMs' |
| // |
| // Based on RFC 6762 expectations are: |
| // - the interval between the first two queries MUST be at least |
| // one second |
| // - the intervals between successive queries MUST increase by at |
| // least a factor of two |
| chip::System::Clock::Timeout nextRetryDelay = chip::System::Clock::Seconds16(1); |
| }; |
| void MarkPending(ScheduledAttempt && attempt); |
| chip::System::Clock::ClockBase * mClock; |
| RetryEntry mRetryQueue[kRetryQueueSize]; |
| }; |
| |
| } // namespace Minimal |
| } // namespace mdns |