Matt Spinler | f60ac27 | 2019-12-11 13:47:50 -0600 | [diff] [blame] | 1 | /** |
| 2 | * Copyright © 2019 IBM Corporation |
| 3 | * |
| 4 | * Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | * you may not use this file except in compliance with the License. |
| 6 | * You may obtain a copy of the License at |
| 7 | * |
| 8 | * http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | * |
| 10 | * Unless required by applicable law or agreed to in writing, software |
| 11 | * distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | * See the License for the specific language governing permissions and |
| 14 | * limitations under the License. |
| 15 | */ |
| 16 | #include "host_notifier.hpp" |
| 17 | |
| 18 | #include <phosphor-logging/log.hpp> |
| 19 | |
| 20 | namespace openpower::pels |
| 21 | { |
| 22 | |
| 23 | const auto subscriptionName = "PELHostNotifier"; |
Matt Spinler | f77debb | 2019-12-12 10:04:33 -0600 | [diff] [blame] | 24 | const size_t maxRetryAttempts = 15; |
Matt Spinler | f60ac27 | 2019-12-11 13:47:50 -0600 | [diff] [blame] | 25 | |
| 26 | using namespace phosphor::logging; |
| 27 | |
| 28 | HostNotifier::HostNotifier(Repository& repo, DataInterfaceBase& dataIface, |
| 29 | std::unique_ptr<HostInterface> hostIface) : |
| 30 | _repo(repo), |
Matt Spinler | f869fcf | 2019-12-11 15:02:20 -0600 | [diff] [blame] | 31 | _dataIface(dataIface), _hostIface(std::move(hostIface)), |
| 32 | _retryTimer(_hostIface->getEvent(), |
| 33 | std::bind(std::mem_fn(&HostNotifier::retryTimerExpired), this)) |
Matt Spinler | f60ac27 | 2019-12-11 13:47:50 -0600 | [diff] [blame] | 34 | { |
| 35 | // Subscribe to be told about new PELs. |
| 36 | _repo.subscribeToAdds(subscriptionName, |
| 37 | std::bind(std::mem_fn(&HostNotifier::newLogCallback), |
| 38 | this, std::placeholders::_1)); |
| 39 | |
| 40 | // Add any existing PELs to the queue to send them if necessary. |
| 41 | _repo.for_each(std::bind(std::mem_fn(&HostNotifier::addPELToQueue), this, |
| 42 | std::placeholders::_1)); |
| 43 | |
| 44 | // Subscribe to be told about host state changes. |
| 45 | _dataIface.subscribeToHostStateChange( |
| 46 | subscriptionName, |
| 47 | std::bind(std::mem_fun(&HostNotifier::hostStateChange), this, |
| 48 | std::placeholders::_1)); |
| 49 | |
| 50 | // Set the function to call when the async reponse is received. |
| 51 | _hostIface->setResponseFunction( |
| 52 | std::bind(std::mem_fn(&HostNotifier::commandResponse), this, |
| 53 | std::placeholders::_1)); |
| 54 | |
| 55 | // Start sending logs if the host is running |
| 56 | if (!_pelQueue.empty() && _dataIface.isHostUp()) |
| 57 | { |
| 58 | doNewLogNotify(); |
| 59 | } |
| 60 | } |
| 61 | |
| 62 | HostNotifier::~HostNotifier() |
| 63 | { |
| 64 | _repo.unsubscribeFromAdds(subscriptionName); |
| 65 | _dataIface.unsubscribeFromHostStateChange(subscriptionName); |
| 66 | } |
| 67 | |
| 68 | bool HostNotifier::addPELToQueue(const PEL& pel) |
| 69 | { |
| 70 | if (enqueueRequired(pel.id())) |
| 71 | { |
| 72 | _pelQueue.push_back(pel.id()); |
| 73 | } |
| 74 | |
| 75 | // Return false so that Repo::for_each keeps going. |
| 76 | return false; |
| 77 | } |
| 78 | |
| 79 | bool HostNotifier::enqueueRequired(uint32_t id) const |
| 80 | { |
| 81 | bool required = true; |
Matt Spinler | a943b15 | 2019-12-11 14:44:50 -0600 | [diff] [blame] | 82 | Repository::LogID i{Repository::LogID::Pel{id}}; |
| 83 | |
| 84 | if (auto attributes = _repo.getPELAttributes(i); attributes) |
| 85 | { |
| 86 | auto a = attributes.value().get(); |
| 87 | |
| 88 | if ((a.hostState == TransmissionState::acked) || |
| 89 | (a.hostState == TransmissionState::badPEL)) |
| 90 | { |
| 91 | required = false; |
| 92 | } |
| 93 | else if (a.actionFlags.test(hiddenFlagBit) && |
| 94 | (a.hmcState == TransmissionState::acked)) |
| 95 | { |
| 96 | required = false; |
| 97 | } |
| 98 | else if (a.actionFlags.test(dontReportToHostFlagBit)) |
| 99 | { |
| 100 | required = false; |
| 101 | } |
| 102 | } |
| 103 | else |
| 104 | { |
| 105 | using namespace phosphor::logging; |
| 106 | log<level::ERR>("Host Enqueue: Unable to find PEL ID in repository", |
| 107 | entry("PEL_ID=0x%X", id)); |
| 108 | required = false; |
| 109 | } |
Matt Spinler | f60ac27 | 2019-12-11 13:47:50 -0600 | [diff] [blame] | 110 | |
| 111 | return required; |
| 112 | } |
| 113 | |
Matt Spinler | f77debb | 2019-12-12 10:04:33 -0600 | [diff] [blame] | 114 | bool HostNotifier::notifyRequired(uint32_t id) const |
| 115 | { |
| 116 | bool notify = true; |
| 117 | Repository::LogID i{Repository::LogID::Pel{id}}; |
| 118 | |
| 119 | if (auto attributes = _repo.getPELAttributes(i); attributes) |
| 120 | { |
| 121 | // If already acked by the host, don't send again. |
| 122 | // (A safety check as it shouldn't get to this point.) |
| 123 | auto a = attributes.value().get(); |
| 124 | if (a.hostState == TransmissionState::acked) |
| 125 | { |
| 126 | notify = false; |
| 127 | } |
| 128 | else if (a.actionFlags.test(hiddenFlagBit)) |
| 129 | { |
| 130 | // If hidden and acked (or will be) acked by the HMC, |
| 131 | // also don't send it. (HMC management can come and |
| 132 | // go at any time) |
| 133 | if ((a.hmcState == TransmissionState::acked) || |
| 134 | _dataIface.isHMCManaged()) |
| 135 | { |
| 136 | notify = false; |
| 137 | } |
| 138 | } |
| 139 | } |
| 140 | else |
| 141 | { |
| 142 | // Must have been deleted since put on the queue. |
| 143 | notify = false; |
| 144 | } |
| 145 | |
| 146 | return notify; |
| 147 | } |
| 148 | |
Matt Spinler | f60ac27 | 2019-12-11 13:47:50 -0600 | [diff] [blame] | 149 | void HostNotifier::newLogCallback(const PEL& pel) |
| 150 | { |
| 151 | if (!enqueueRequired(pel.id())) |
| 152 | { |
| 153 | return; |
| 154 | } |
| 155 | |
| 156 | _pelQueue.push_back(pel.id()); |
| 157 | |
| 158 | // TODO: Check if a send is needed now |
| 159 | } |
| 160 | |
| 161 | void HostNotifier::doNewLogNotify() |
| 162 | { |
Matt Spinler | f77debb | 2019-12-12 10:04:33 -0600 | [diff] [blame] | 163 | if (!_dataIface.isHostUp() || _retryTimer.isEnabled()) |
| 164 | { |
| 165 | return; |
| 166 | } |
| 167 | |
| 168 | if (_retryCount >= maxRetryAttempts) |
| 169 | { |
| 170 | // Give up until a new log comes in. |
| 171 | if (_retryCount == maxRetryAttempts) |
| 172 | { |
| 173 | // If this were to really happen, the PLDM interface |
| 174 | // would be down and isolating that shouldn't left to |
| 175 | // a logging daemon, so just trace. Also, this will start |
| 176 | // trying again when the next new log comes in. |
| 177 | log<level::ERR>( |
| 178 | "PEL Host notifier hit max retry attempts. Giving up for now.", |
| 179 | entry("PEL_ID=0x%X", _pelQueue.front())); |
| 180 | } |
| 181 | return; |
| 182 | } |
| 183 | |
| 184 | bool doNotify = false; |
| 185 | uint32_t id = 0; |
| 186 | |
| 187 | // Find the PEL to send |
| 188 | while (!doNotify && !_pelQueue.empty()) |
| 189 | { |
| 190 | id = _pelQueue.front(); |
| 191 | _pelQueue.pop_front(); |
| 192 | |
| 193 | if (notifyRequired(id)) |
| 194 | { |
| 195 | doNotify = true; |
| 196 | } |
| 197 | } |
| 198 | |
| 199 | if (doNotify) |
| 200 | { |
| 201 | // Get the size using the repo attributes |
| 202 | Repository::LogID i{Repository::LogID::Pel{id}}; |
| 203 | if (auto attributes = _repo.getPELAttributes(i); attributes) |
| 204 | { |
| 205 | auto size = static_cast<size_t>( |
| 206 | std::filesystem::file_size((*attributes).get().path)); |
| 207 | auto rc = _hostIface->sendNewLogCmd(id, size); |
| 208 | |
| 209 | if (rc == CmdStatus::success) |
| 210 | { |
| 211 | _inProgressPEL = id; |
| 212 | } |
| 213 | else |
| 214 | { |
| 215 | // It failed. Retry |
| 216 | log<level::ERR>("PLDM send failed", entry("PEL_ID=0x%X", id)); |
| 217 | _pelQueue.push_front(id); |
| 218 | _inProgressPEL = 0; |
| 219 | _retryTimer.restartOnce(_hostIface->getSendRetryDelay()); |
| 220 | } |
| 221 | } |
| 222 | else |
| 223 | { |
| 224 | log<level::ERR>("PEL ID not in repository. Cannot notify host", |
| 225 | entry("PEL_ID=0x%X", id)); |
| 226 | } |
| 227 | } |
Matt Spinler | f60ac27 | 2019-12-11 13:47:50 -0600 | [diff] [blame] | 228 | } |
| 229 | |
| 230 | void HostNotifier::hostStateChange(bool hostUp) |
| 231 | { |
Matt Spinler | 3019c6f | 2019-12-11 15:24:45 -0600 | [diff] [blame] | 232 | _retryCount = 0; |
| 233 | |
| 234 | if (hostUp && !_pelQueue.empty()) |
| 235 | { |
| 236 | doNewLogNotify(); |
| 237 | } |
| 238 | else if (!hostUp) |
| 239 | { |
| 240 | stopCommand(); |
| 241 | |
| 242 | // Reset the state on any PELs that were sent but not acked back |
| 243 | // to new so they'll get sent again. |
| 244 | for (auto id : _sentPELs) |
| 245 | { |
| 246 | _pelQueue.push_back(id); |
| 247 | _repo.setPELHostTransState(id, TransmissionState::newPEL); |
| 248 | } |
| 249 | |
| 250 | _sentPELs.clear(); |
| 251 | } |
Matt Spinler | f60ac27 | 2019-12-11 13:47:50 -0600 | [diff] [blame] | 252 | } |
| 253 | |
| 254 | void HostNotifier::commandResponse(ResponseStatus status) |
| 255 | { |
Matt Spinler | f869fcf | 2019-12-11 15:02:20 -0600 | [diff] [blame] | 256 | auto id = _inProgressPEL; |
| 257 | _inProgressPEL = 0; |
| 258 | |
| 259 | if (status == ResponseStatus::success) |
| 260 | { |
| 261 | _retryCount = 0; |
| 262 | |
| 263 | _sentPELs.push_back(id); |
| 264 | |
| 265 | _repo.setPELHostTransState(id, TransmissionState::sent); |
| 266 | |
| 267 | if (!_pelQueue.empty()) |
| 268 | { |
| 269 | doNewLogNotify(); |
| 270 | } |
| 271 | } |
| 272 | else |
| 273 | { |
| 274 | log<level::ERR>("PLDM command response failure", |
| 275 | entry("PEL_ID=0x%X", id)); |
| 276 | // Retry |
| 277 | _pelQueue.push_front(id); |
| 278 | _retryTimer.restartOnce(_hostIface->getReceiveRetryDelay()); |
| 279 | } |
| 280 | } |
| 281 | |
| 282 | void HostNotifier::retryTimerExpired() |
| 283 | { |
| 284 | if (_dataIface.isHostUp()) |
| 285 | { |
| 286 | log<level::INFO>("Attempting command retry", |
| 287 | entry("PEL_ID=0x%X", _pelQueue.front())); |
| 288 | _retryCount++; |
| 289 | doNewLogNotify(); |
| 290 | } |
Matt Spinler | f60ac27 | 2019-12-11 13:47:50 -0600 | [diff] [blame] | 291 | } |
| 292 | |
Matt Spinler | 3019c6f | 2019-12-11 15:24:45 -0600 | [diff] [blame] | 293 | void HostNotifier::stopCommand() |
| 294 | { |
| 295 | _retryCount = 0; |
| 296 | |
| 297 | if (_inProgressPEL != 0) |
| 298 | { |
| 299 | _pelQueue.push_front(_inProgressPEL); |
| 300 | _inProgressPEL = 0; |
| 301 | } |
| 302 | |
| 303 | if (_retryTimer.isEnabled()) |
| 304 | { |
| 305 | _retryTimer.setEnabled(false); |
| 306 | } |
| 307 | |
| 308 | if (_hostIface->cmdInProgress()) |
| 309 | { |
| 310 | _hostIface->cancelCmd(); |
| 311 | } |
| 312 | } |
| 313 | |
Matt Spinler | f60ac27 | 2019-12-11 13:47:50 -0600 | [diff] [blame] | 314 | } // namespace openpower::pels |