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 "repository.hpp"
17 
18 #include <sys/stat.h>
19 
20 #include <fstream>
21 #include <phosphor-logging/log.hpp>
22 #include <xyz/openbmc_project/Common/File/error.hpp>
23 
24 namespace openpower
25 {
26 namespace pels
27 {
28 
29 namespace fs = std::filesystem;
30 using namespace phosphor::logging;
31 namespace file_error = sdbusplus::xyz::openbmc_project::Common::File::Error;
32 
33 constexpr size_t warningPercentage = 95;
34 
35 /**
36  * @brief Returns the amount of space the file uses on disk.
37  *
38  * This is different than just the regular size of the file.
39  *
40  * @param[in] file - The file to get the size of
41  *
42  * @return size_t The disk space the file uses
43  */
44 size_t getFileDiskSize(const std::filesystem::path& file)
45 {
46     constexpr size_t statBlockSize = 512;
47     struct stat statData;
48     auto rc = stat(file.c_str(), &statData);
49     if (rc != 0)
50     {
51         auto e = errno;
52         std::string msg = "call to stat() failed on " + file.native() +
53                           " with errno " + std::to_string(e);
54         log<level::ERR>(msg.c_str());
55         abort();
56     }
57 
58     return statData.st_blocks * statBlockSize;
59 }
60 
61 Repository::Repository(const std::filesystem::path& basePath, size_t repoSize,
62                        size_t maxNumPELs) :
63     _logPath(basePath / "logs"),
64     _maxRepoSize(repoSize), _maxNumPELs(maxNumPELs),
65     _archivePath(basePath / "logs" / "archive")
66 {
67     if (!fs::exists(_logPath))
68     {
69         fs::create_directories(_logPath);
70     }
71 
72     if (!fs::exists(_archivePath))
73     {
74         fs::create_directories(_archivePath);
75     }
76 
77     restore();
78 }
79 
80 void Repository::restore()
81 {
82     for (auto& dirEntry : fs::directory_iterator(_logPath))
83     {
84         try
85         {
86             if (!fs::is_regular_file(dirEntry.path()))
87             {
88                 continue;
89             }
90 
91             std::ifstream file{dirEntry.path()};
92             std::vector<uint8_t> data{std::istreambuf_iterator<char>(file),
93                                       std::istreambuf_iterator<char>()};
94             file.close();
95 
96             PEL pel{data};
97             if (pel.valid())
98             {
99                 // If the host hasn't acked it, reset the host state so
100                 // it will get sent up again.
101                 if (pel.hostTransmissionState() == TransmissionState::sent)
102                 {
103                     pel.setHostTransmissionState(TransmissionState::newPEL);
104                     try
105                     {
106                         write(pel, dirEntry.path());
107                     }
108                     catch (const std::exception& e)
109                     {
110                         log<level::ERR>(
111                             "Failed to save PEL after updating host state",
112                             entry("PELID=0x%X", pel.id()));
113                     }
114                 }
115 
116                 PELAttributes attributes{dirEntry.path(),
117                                          getFileDiskSize(dirEntry.path()),
118                                          pel.privateHeader().creatorID(),
119                                          pel.userHeader().subsystem(),
120                                          pel.userHeader().severity(),
121                                          pel.userHeader().actionFlags(),
122                                          pel.hostTransmissionState(),
123                                          pel.hmcTransmissionState()};
124 
125                 using pelID = LogID::Pel;
126                 using obmcID = LogID::Obmc;
127                 _pelAttributes.emplace(
128                     LogID(pelID(pel.id()), obmcID(pel.obmcLogID())),
129                     attributes);
130 
131                 updateRepoStats(attributes, true);
132             }
133             else
134             {
135                 log<level::ERR>(
136                     "Found invalid PEL file while restoring.  Removing.",
137                     entry("FILENAME=%s", dirEntry.path().c_str()));
138                 fs::remove(dirEntry.path());
139             }
140         }
141         catch (const std::exception& e)
142         {
143             log<level::ERR>("Hit exception while restoring PEL File",
144                             entry("FILENAME=%s", dirEntry.path().c_str()),
145                             entry("ERROR=%s", e.what()));
146         }
147     }
148 
149     // Get size of archive folder
150     for (auto& dirEntry : fs::directory_iterator(_archivePath))
151     {
152         _archiveSize += getFileDiskSize(dirEntry);
153     }
154 }
155 
156 std::string Repository::getPELFilename(uint32_t pelID, const BCDTime& time)
157 {
158     char name[50];
159     sprintf(name, "%.2X%.2X%.2X%.2X%.2X%.2X%.2X%.2X_%.8X", time.yearMSB,
160             time.yearLSB, time.month, time.day, time.hour, time.minutes,
161             time.seconds, time.hundredths, pelID);
162     return std::string{name};
163 }
164 
165 void Repository::add(std::unique_ptr<PEL>& pel)
166 {
167     pel->setHostTransmissionState(TransmissionState::newPEL);
168     pel->setHMCTransmissionState(TransmissionState::newPEL);
169 
170     auto path = _logPath / getPELFilename(pel->id(), pel->commitTime());
171 
172     write(*(pel.get()), path);
173 
174     PELAttributes attributes{path,
175                              getFileDiskSize(path),
176                              pel->privateHeader().creatorID(),
177                              pel->userHeader().subsystem(),
178                              pel->userHeader().severity(),
179                              pel->userHeader().actionFlags(),
180                              pel->hostTransmissionState(),
181                              pel->hmcTransmissionState()};
182 
183     using pelID = LogID::Pel;
184     using obmcID = LogID::Obmc;
185     _pelAttributes.emplace(LogID(pelID(pel->id()), obmcID(pel->obmcLogID())),
186                            attributes);
187 
188     _lastPelID = pel->id();
189 
190     updateRepoStats(attributes, true);
191 
192     processAddCallbacks(*pel);
193 }
194 
195 void Repository::write(const PEL& pel, const fs::path& path)
196 {
197     std::ofstream file{path, std::ios::binary};
198 
199     if (!file.good())
200     {
201         // If this fails, the filesystem is probably full so it isn't like
202         // we could successfully create yet another error log here.
203         auto e = errno;
204         fs::remove(path);
205         log<level::ERR>("Unable to open PEL file for writing",
206                         entry("ERRNO=%d", e), entry("PATH=%s", path.c_str()));
207         throw file_error::Open();
208     }
209 
210     auto data = pel.data();
211     file.write(reinterpret_cast<const char*>(data.data()), data.size());
212 
213     if (file.fail())
214     {
215         // Same note as above about not being able to create an error log
216         // for this case even if we wanted.
217         auto e = errno;
218         file.close();
219         fs::remove(path);
220         log<level::ERR>("Unable to write PEL file", entry("ERRNO=%d", e),
221                         entry("PATH=%s", path.c_str()));
222         throw file_error::Write();
223     }
224 }
225 
226 std::optional<Repository::LogID> Repository::remove(const LogID& id)
227 {
228     auto pel = findPEL(id);
229     if (pel == _pelAttributes.end())
230     {
231         return std::nullopt;
232     }
233 
234     LogID actualID = pel->first;
235     updateRepoStats(pel->second, false);
236 
237     log<level::DEBUG>("Removing PEL from repository",
238                       entry("PEL_ID=0x%X", actualID.pelID.id),
239                       entry("OBMC_LOG_ID=%d", actualID.obmcID.id));
240 
241     if (fs::exists(pel->second.path))
242     {
243         // Check for existense of new archive folder
244         if (!fs::exists(_archivePath))
245         {
246             fs::create_directories(_archivePath);
247         }
248 
249         // Move log file to archive folder
250         auto fileName = _archivePath / pel->second.path.filename();
251         fs::rename(pel->second.path, fileName);
252 
253         // Update size of file
254         _archiveSize += getFileDiskSize(fileName);
255     }
256 
257     _pelAttributes.erase(pel);
258 
259     processDeleteCallbacks(actualID.pelID.id);
260 
261     return actualID;
262 }
263 
264 std::optional<std::vector<uint8_t>> Repository::getPELData(const LogID& id)
265 {
266     auto pel = findPEL(id);
267     if (pel != _pelAttributes.end())
268     {
269         std::ifstream file{pel->second.path.c_str()};
270         if (!file.good())
271         {
272             auto e = errno;
273             log<level::ERR>("Unable to open PEL file", entry("ERRNO=%d", e),
274                             entry("PATH=%s", pel->second.path.c_str()));
275             throw file_error::Open();
276         }
277 
278         std::vector<uint8_t> data{std::istreambuf_iterator<char>(file),
279                                   std::istreambuf_iterator<char>()};
280         return data;
281     }
282 
283     return std::nullopt;
284 }
285 
286 std::optional<sdbusplus::message::unix_fd> Repository::getPELFD(const LogID& id)
287 {
288     auto pel = findPEL(id);
289     if (pel != _pelAttributes.end())
290     {
291         FILE* fp = fopen(pel->second.path.c_str(), "rb");
292 
293         if (fp == nullptr)
294         {
295             auto e = errno;
296             log<level::ERR>("Unable to open PEL File", entry("ERRNO=%d", e),
297                             entry("PATH=%s", pel->second.path.c_str()));
298             throw file_error::Open();
299         }
300 
301         // Must leave the file open here.  It will be closed by sdbusplus
302         // when it sends it back over D-Bus.
303 
304         return fileno(fp);
305     }
306     return std::nullopt;
307 }
308 
309 void Repository::for_each(ForEachFunc func) const
310 {
311     for (const auto& [id, attributes] : _pelAttributes)
312     {
313         std::ifstream file{attributes.path};
314 
315         if (!file.good())
316         {
317             auto e = errno;
318             log<level::ERR>("Repository::for_each: Unable to open PEL file",
319                             entry("ERRNO=%d", e),
320                             entry("PATH=%s", attributes.path.c_str()));
321             continue;
322         }
323 
324         std::vector<uint8_t> data{std::istreambuf_iterator<char>(file),
325                                   std::istreambuf_iterator<char>()};
326         file.close();
327 
328         PEL pel{data};
329 
330         try
331         {
332             if (func(pel))
333             {
334                 break;
335             }
336         }
337         catch (const std::exception& e)
338         {
339             log<level::ERR>("Repository::for_each function exception",
340                             entry("ERROR=%s", e.what()));
341         }
342     }
343 }
344 
345 void Repository::processAddCallbacks(const PEL& pel) const
346 {
347     for (auto& [name, func] : _addSubscriptions)
348     {
349         try
350         {
351             func(pel);
352         }
353         catch (const std::exception& e)
354         {
355             log<level::ERR>("PEL Repository add callback exception",
356                             entry("NAME=%s", name.c_str()),
357                             entry("ERROR=%s", e.what()));
358         }
359     }
360 }
361 
362 void Repository::processDeleteCallbacks(uint32_t id) const
363 {
364     for (auto& [name, func] : _deleteSubscriptions)
365     {
366         try
367         {
368             func(id);
369         }
370         catch (const std::exception& e)
371         {
372             log<level::ERR>("PEL Repository delete callback exception",
373                             entry("NAME=%s", name.c_str()),
374                             entry("ERROR=%s", e.what()));
375         }
376     }
377 }
378 
379 std::optional<std::reference_wrapper<const Repository::PELAttributes>>
380     Repository::getPELAttributes(const LogID& id) const
381 {
382     auto pel = findPEL(id);
383     if (pel != _pelAttributes.end())
384     {
385         return pel->second;
386     }
387 
388     return std::nullopt;
389 }
390 
391 void Repository::setPELHostTransState(uint32_t pelID, TransmissionState state)
392 {
393     LogID id{LogID::Pel{pelID}};
394     auto attr = std::find_if(_pelAttributes.begin(), _pelAttributes.end(),
395                              [&id](const auto& a) { return a.first == id; });
396 
397     if ((attr != _pelAttributes.end()) && (attr->second.hostState != state))
398     {
399         PELUpdateFunc func = [state](PEL& pel) {
400             pel.setHostTransmissionState(state);
401         };
402 
403         try
404         {
405             updatePEL(attr->second.path, func);
406 
407             attr->second.hostState = state;
408         }
409         catch (const std::exception& e)
410         {
411             log<level::ERR>("Unable to update PEL host transmission state",
412                             entry("PATH=%s", attr->second.path.c_str()),
413                             entry("ERROR=%s", e.what()));
414         }
415     }
416 }
417 
418 void Repository::setPELHMCTransState(uint32_t pelID, TransmissionState state)
419 {
420     LogID id{LogID::Pel{pelID}};
421     auto attr = std::find_if(_pelAttributes.begin(), _pelAttributes.end(),
422                              [&id](const auto& a) { return a.first == id; });
423 
424     if ((attr != _pelAttributes.end()) && (attr->second.hmcState != state))
425     {
426         PELUpdateFunc func = [state](PEL& pel) {
427             pel.setHMCTransmissionState(state);
428         };
429 
430         try
431         {
432             updatePEL(attr->second.path, func);
433 
434             attr->second.hmcState = state;
435         }
436         catch (const std::exception& e)
437         {
438             log<level::ERR>("Unable to update PEL HMC transmission state",
439                             entry("PATH=%s", attr->second.path.c_str()),
440                             entry("ERROR=%s", e.what()));
441         }
442     }
443 }
444 
445 void Repository::updatePEL(const fs::path& path, PELUpdateFunc updateFunc)
446 {
447     std::ifstream file{path};
448     std::vector<uint8_t> data{std::istreambuf_iterator<char>(file),
449                               std::istreambuf_iterator<char>()};
450     file.close();
451 
452     PEL pel{data};
453 
454     if (pel.valid())
455     {
456         updateFunc(pel);
457 
458         write(pel, path);
459     }
460     else
461     {
462         throw std::runtime_error(
463             "Unable to read a valid PEL when trying to update it");
464     }
465 }
466 
467 bool Repository::isServiceableSev(const PELAttributes& pel)
468 {
469     auto sevType = static_cast<SeverityType>(pel.severity & 0xF0);
470     auto sevPVEntry =
471         pel_values::findByValue(pel.severity, pel_values::severityValues);
472     std::string sevName = std::get<pel_values::registryNamePos>(*sevPVEntry);
473 
474     bool check1 = (sevType == SeverityType::predictive) ||
475                   (sevType == SeverityType::unrecoverable) ||
476                   (sevType == SeverityType::critical);
477 
478     bool check2 = ((sevType == SeverityType::recovered) ||
479                    (sevName == "symptom_recovered")) &&
480                   !pel.actionFlags.test(hiddenFlagBit);
481 
482     bool check3 = (sevName == "symptom_predictive") ||
483                   (sevName == "symptom_unrecoverable") ||
484                   (sevName == "symptom_critical");
485 
486     return check1 || check2 || check3;
487 }
488 
489 void Repository::updateRepoStats(const PELAttributes& pel, bool pelAdded)
490 {
491     auto isServiceable = Repository::isServiceableSev(pel);
492     auto bmcPEL = CreatorID::openBMC == static_cast<CreatorID>(pel.creator);
493 
494     auto adjustSize = [pelAdded, &pel](auto& runningSize) {
495         if (pelAdded)
496         {
497             runningSize += pel.sizeOnDisk;
498         }
499         else
500         {
501             runningSize = std::max(static_cast<int64_t>(runningSize) -
502                                        static_cast<int64_t>(pel.sizeOnDisk),
503                                    static_cast<int64_t>(0));
504         }
505     };
506 
507     adjustSize(_sizes.total);
508 
509     if (bmcPEL)
510     {
511         adjustSize(_sizes.bmc);
512         if (isServiceable)
513         {
514             adjustSize(_sizes.bmcServiceable);
515         }
516         else
517         {
518             adjustSize(_sizes.bmcInfo);
519         }
520     }
521     else
522     {
523         adjustSize(_sizes.nonBMC);
524         if (isServiceable)
525         {
526             adjustSize(_sizes.nonBMCServiceable);
527         }
528         else
529         {
530             adjustSize(_sizes.nonBMCInfo);
531         }
532     }
533 }
534 
535 bool Repository::sizeWarning()
536 {
537     std::error_code ec;
538 
539     if ((_archiveSize > 0) && ((_sizes.total + _archiveSize) >
540                                ((_maxRepoSize * warningPercentage) / 100)))
541     {
542         log<level::INFO>(
543             "Repository::sizeWarning function:Deleting the files in archive");
544 
545         for (const auto& dirEntry : fs::directory_iterator(_archivePath))
546         {
547             fs::remove(dirEntry.path(), ec);
548             if (ec)
549             {
550                 log<level::INFO>(
551                     "Repository::sizeWarning function:Could not delete "
552                     "a file in PEL archive",
553                     entry("FILENAME=%s", dirEntry.path().c_str()));
554             }
555         }
556 
557         _archiveSize = 0;
558     }
559 
560     return (_sizes.total > (_maxRepoSize * warningPercentage / 100)) ||
561            (_pelAttributes.size() > _maxNumPELs);
562 }
563 
564 std::vector<Repository::AttributesReference>
565     Repository::getAllPELAttributes(SortOrder order) const
566 {
567     std::vector<Repository::AttributesReference> attributes;
568 
569     std::for_each(
570         _pelAttributes.begin(), _pelAttributes.end(),
571         [&attributes](auto& pelEntry) { attributes.push_back(pelEntry); });
572 
573     std::sort(attributes.begin(), attributes.end(),
574               [order](const auto& left, const auto& right) {
575                   if (order == SortOrder::ascending)
576                   {
577                       return left.get().second.path < right.get().second.path;
578                   }
579                   return left.get().second.path > right.get().second.path;
580               });
581 
582     return attributes;
583 }
584 
585 std::vector<uint32_t>
586     Repository::prune(const std::vector<uint32_t>& idsWithHwIsoEntry)
587 {
588     std::vector<uint32_t> obmcLogIDs;
589     std::string msg = "Pruning PEL repository that takes up " +
590                       std::to_string(_sizes.total) + " bytes and has " +
591                       std::to_string(_pelAttributes.size()) + " PELs";
592     log<level::INFO>(msg.c_str());
593 
594     // Set up the 5 functions to check if the PEL category
595     // is still over its limits.
596 
597     // BMC informational PELs should only take up 15%
598     IsOverLimitFunc overBMCInfoLimit = [this]() {
599         return _sizes.bmcInfo > _maxRepoSize * 15 / 100;
600     };
601 
602     // BMC non informational PELs should only take up 30%
603     IsOverLimitFunc overBMCNonInfoLimit = [this]() {
604         return _sizes.bmcServiceable > _maxRepoSize * 30 / 100;
605     };
606 
607     // Non BMC informational PELs should only take up 15%
608     IsOverLimitFunc overNonBMCInfoLimit = [this]() {
609         return _sizes.nonBMCInfo > _maxRepoSize * 15 / 100;
610     };
611 
612     // Non BMC non informational PELs should only take up 15%
613     IsOverLimitFunc overNonBMCNonInfoLimit = [this]() {
614         return _sizes.nonBMCServiceable > _maxRepoSize * 30 / 100;
615     };
616 
617     // Bring the total number of PELs down to 80% of the max
618     IsOverLimitFunc tooManyPELsLimit = [this]() {
619         return _pelAttributes.size() > _maxNumPELs * 80 / 100;
620     };
621 
622     // Set up the functions to determine which category a PEL is in.
623     // TODO: Return false in these functions if a PEL caused a guard record.
624 
625     // A BMC informational PEL
626     IsPELTypeFunc isBMCInfo = [](const PELAttributes& pel) {
627         return (CreatorID::openBMC == static_cast<CreatorID>(pel.creator)) &&
628                !Repository::isServiceableSev(pel);
629     };
630 
631     // A BMC non informational PEL
632     IsPELTypeFunc isBMCNonInfo = [](const PELAttributes& pel) {
633         return (CreatorID::openBMC == static_cast<CreatorID>(pel.creator)) &&
634                Repository::isServiceableSev(pel);
635     };
636 
637     // A non BMC informational PEL
638     IsPELTypeFunc isNonBMCInfo = [](const PELAttributes& pel) {
639         return (CreatorID::openBMC != static_cast<CreatorID>(pel.creator)) &&
640                !Repository::isServiceableSev(pel);
641     };
642 
643     // A non BMC non informational PEL
644     IsPELTypeFunc isNonBMCNonInfo = [](const PELAttributes& pel) {
645         return (CreatorID::openBMC != static_cast<CreatorID>(pel.creator)) &&
646                Repository::isServiceableSev(pel);
647     };
648 
649     // When counting PELs, count every PEL
650     IsPELTypeFunc isAnyPEL = [](const PELAttributes& /*pel*/) { return true; };
651 
652     // Check all 4 categories, which will result in at most 90%
653     // usage (15 + 30 + 15 + 30).
654     removePELs(overBMCInfoLimit, isBMCInfo, idsWithHwIsoEntry, obmcLogIDs);
655     removePELs(overBMCNonInfoLimit, isBMCNonInfo, idsWithHwIsoEntry,
656                obmcLogIDs);
657     removePELs(overNonBMCInfoLimit, isNonBMCInfo, idsWithHwIsoEntry,
658                obmcLogIDs);
659     removePELs(overNonBMCNonInfoLimit, isNonBMCNonInfo, idsWithHwIsoEntry,
660                obmcLogIDs);
661 
662     // After the above pruning check if there are still too many PELs,
663     // which can happen depending on PEL sizes.
664     if (_pelAttributes.size() > _maxNumPELs)
665     {
666         removePELs(tooManyPELsLimit, isAnyPEL, idsWithHwIsoEntry, obmcLogIDs);
667     }
668 
669     if (!obmcLogIDs.empty())
670     {
671         std::string m = "Number of PELs removed to save space: " +
672                         std::to_string(obmcLogIDs.size());
673         log<level::INFO>(m.c_str());
674     }
675 
676     return obmcLogIDs;
677 }
678 
679 void Repository::removePELs(const IsOverLimitFunc& isOverLimit,
680                             const IsPELTypeFunc& isPELType,
681                             const std::vector<uint32_t>& idsWithHwIsoEntry,
682                             std::vector<uint32_t>& removedBMCLogIDs)
683 {
684     if (!isOverLimit())
685     {
686         return;
687     }
688 
689     auto attributes = getAllPELAttributes(SortOrder::ascending);
690 
691     // Make 4 passes on the PELs, stopping as soon as isOverLimit
692     // returns false.
693     //   Pass 1: only delete HMC acked PELs
694     //   Pass 2: only delete OS acked PELs
695     //   Pass 3: only delete PHYP sent PELs
696     //   Pass 4: delete all PELs
697     static const std::vector<std::function<bool(const PELAttributes& pel)>>
698         stateChecks{[](const auto& pel) {
699                         return pel.hmcState == TransmissionState::acked;
700                     },
701 
702                     [](const auto& pel) {
703                         return pel.hostState == TransmissionState::acked;
704                     },
705 
706                     [](const auto& pel) {
707                         return pel.hostState == TransmissionState::sent;
708                     },
709 
710                     [](const auto& /*pel*/) { return true; }};
711 
712     for (const auto& stateCheck : stateChecks)
713     {
714         for (auto it = attributes.begin(); it != attributes.end();)
715         {
716             const auto& pel = it->get();
717             if (isPELType(pel.second) && stateCheck(pel.second))
718             {
719                 auto removedID = pel.first.obmcID.id;
720 
721                 auto idFound = std::find(idsWithHwIsoEntry.begin(),
722                                          idsWithHwIsoEntry.end(), removedID);
723                 if (idFound != idsWithHwIsoEntry.end())
724                 {
725                     ++it;
726                     continue;
727                 }
728 
729                 remove(pel.first);
730 
731                 removedBMCLogIDs.push_back(removedID);
732 
733                 attributes.erase(it);
734 
735                 if (!isOverLimit())
736                 {
737                     break;
738                 }
739             }
740             else
741             {
742                 ++it;
743             }
744         }
745 
746         if (!isOverLimit())
747         {
748             break;
749         }
750     }
751 }
752 
753 void Repository::archivePEL(const PEL& pel)
754 {
755     if (pel.valid())
756     {
757         auto path = _archivePath / getPELFilename(pel.id(), pel.commitTime());
758 
759         write(pel, path);
760 
761         _archiveSize += getFileDiskSize(path);
762     }
763 }
764 
765 } // namespace pels
766 } // namespace openpower
767