src/fs/rock/RockSwapDir.cc

   1 /*
   2  * DEBUG: section 47    Store Directory Routines
   3  */
   4
   5 #include "squid.h"
   6 #include "cache_cf.h"
   7 #include "CollapsedForwarding.h"
   8 #include "ConfigOption.h"
   9 #include "DiskIO/DiskIOModule.h"
  10 #include "DiskIO/DiskIOStrategy.h"
  11 #include "DiskIO/ReadRequest.h"
  12 #include "DiskIO/WriteRequest.h"
  13 #include "fs/rock/RockSwapDir.h"
  14 #include "fs/rock/RockIoState.h"
  15 #include "fs/rock/RockIoRequests.h"
  16 #include "fs/rock/RockRebuild.h"
  17 #include "globals.h"
  18 #include "ipc/mem/Pages.h"
  19 #include "MemObject.h"
  20 #include "Parsing.h"
  21 #include "SquidConfig.h"
  22 #include "SquidMath.h"
  23 #include "tools.h"
  24
  25 #include <cstdlib>
  26 #include <iomanip>
  27
  28 #if HAVE_SYS_STAT_H
  29 #include <sys/stat.h>
  30 #endif
  31
  32 const int64_t Rock::SwapDir::HeaderSize = 16*1024;
  33
  34 Rock::SwapDir::SwapDir(): ::SwapDir("rock"),
  35     slotSize(HeaderSize), filePath(NULL), map(NULL), io(NULL),
  36     waitingForPage(NULL)
  37 {
  38 }
  39
  40 Rock::SwapDir::~SwapDir()
  41 {
  42     delete io;
  43     delete map;
  44     safe_free(filePath);
  45 }
  46
  47 StoreSearch *
  48 Rock::SwapDir::search(String const url, HttpRequest *)
  49 {
  50     assert(false);
  51     return NULL; // XXX: implement
  52 }
  53
  54 void
  55 Rock::SwapDir::get(String const key, STOREGETCLIENT cb, void *data)
  56 {
  57     ::SwapDir::get(key, cb, data);
  58 }
  59
  60 // called when Squid core needs a StoreEntry with a given key
  61 StoreEntry *
  62 Rock::SwapDir::get(const cache_key *key)
  63 {
  64     if (!map || !theFile || !theFile->canRead())
  65         return NULL;
  66
  67     sfileno filen;
  68     const Ipc::StoreMapAnchor *const slot = map->openForReading(key, filen);
  69     if (!slot)
  70         return NULL;
  71
  72     // create a brand new store entry and initialize it with stored basics
  73     StoreEntry *e = new StoreEntry();
  74     anchorEntry(*e, filen, *slot);
  75
  76     e->hashInsert(key);
  77     trackReferences(*e);
  78
  79     return e;
  80     // the disk entry remains open for reading, protected from modifications
  81 }
  82
  83 bool
  84 Rock::SwapDir::anchorCollapsed(StoreEntry &collapsed, bool &inSync)
  85 {
  86     if (!map || !theFile || !theFile->canRead())
  87         return false;
  88
  89     sfileno filen;
  90     const Ipc::StoreMapAnchor *const slot = map->openForReading(
  91         reinterpret_cast<cache_key*>(collapsed.key), filen);
  92     if (!slot)
  93         return false;
  94
  95     anchorEntry(collapsed, filen, *slot);
  96     inSync = updateCollapsedWith(collapsed, *slot);
  97     return false;
  98 }
  99
 100 bool
 101 Rock::SwapDir::updateCollapsed(StoreEntry &collapsed)
 102 {
 103     if (!map || !theFile || !theFile->canRead())
 104         return false;
 105
 106     if (collapsed.swap_filen < 0) // no longer using a disk cache
 107         return true;
 108     assert(collapsed.swap_dirn == index);
 109
 110     const Ipc::StoreMapAnchor &s = map->readableEntry(collapsed.swap_filen);
 111     return updateCollapsedWith(collapsed, s);
 112 }
 113
 114 bool
 115 Rock::SwapDir::updateCollapsedWith(StoreEntry &collapsed, const Ipc::StoreMapAnchor &anchor)
 116 {
 117     collapsed.swap_file_sz = anchor.basics.swap_file_sz; // XXX: make atomic
 118     return true;
 119 }
 120
 121 void
 122 Rock::SwapDir::anchorEntry(StoreEntry &e, const sfileno filen, const Ipc::StoreMapAnchor &anchor)
 123 {
 124     const Ipc::StoreMapAnchor::Basics &basics = anchor.basics;
 125
 126     e.swap_file_sz = basics.swap_file_sz;
 127     e.swap_dirn = index;
 128     e.swap_filen = filen;
 129     e.lastref = basics.lastref;
 130     e.timestamp = basics.timestamp;
 131     e.expires = basics.expires;
 132     e.lastmod = basics.lastmod;
 133     e.refcount = basics.refcount;
 134     e.flags = basics.flags;
 135
 136     e.store_status = STORE_OK;
 137     e.setMemStatus(NOT_IN_MEMORY);
 138     e.swap_status = SWAPOUT_DONE;
 139     e.ping_status = PING_NONE;
 140
 141     EBIT_CLR(e.flags, RELEASE_REQUEST);
 142     EBIT_CLR(e.flags, KEY_PRIVATE);
 143     EBIT_SET(e.flags, ENTRY_VALIDATED);
 144 }
 145
 146
 147 void Rock::SwapDir::disconnect(StoreEntry &e)
 148 {
 149     assert(e.swap_dirn == index);
 150     assert(e.swap_filen >= 0);
 151     // cannot have SWAPOUT_NONE entry with swap_filen >= 0
 152     assert(e.swap_status != SWAPOUT_NONE);
 153
 154     // do not rely on e.swap_status here because there is an async delay
 155     // before it switches from SWAPOUT_WRITING to SWAPOUT_DONE.
 156
 157     // since e has swap_filen, its slot is locked for reading and/or writing
 158     // but it is difficult to know whether THIS worker is reading or writing e
 159     if (e.swap_status == SWAPOUT_WRITING ||
 160         (e.mem_obj && e.mem_obj->swapout.sio != NULL))
 161         map->abortWriting(e.swap_filen);
 162     else
 163         map->closeForReading(e.swap_filen);
 164     e.swap_dirn = -1;
 165     e.swap_filen = -1;
 166     e.swap_status = SWAPOUT_NONE;
 167 }
 168
 169 uint64_t
 170 Rock::SwapDir::currentSize() const
 171 {
 172     const uint64_t spaceSize = !freeSlots ?
 173         maxSize() : (slotSize * freeSlots->size());
 174     // everything that is not free is in use
 175     return maxSize() - spaceSize;
 176 }
 177
 178 uint64_t
 179 Rock::SwapDir::currentCount() const
 180 {
 181     return map ? map->entryCount() : 0;
 182 }
 183
 184 /// In SMP mode only the disker process reports stats to avoid
 185 /// counting the same stats by multiple processes.
 186 bool
 187 Rock::SwapDir::doReportStat() const
 188 {
 189     return ::SwapDir::doReportStat() && (!UsingSmp() || IamDiskProcess());
 190 }
 191
 192 void
 193 Rock::SwapDir::swappedOut(const StoreEntry &)
 194 {
 195     // stats are not stored but computed when needed
 196 }
 197
 198 int64_t
 199 Rock::SwapDir::entryLimitAllowed() const
 200 {
 201     const int64_t eLimitLo = map ? map->entryLimit() : 0; // dynamic shrinking unsupported
 202     const int64_t eWanted = (maxSize() - HeaderSize)/slotSize;
 203     return min(max(eLimitLo, eWanted), entryLimitHigh());
 204 }
 205
 206 // TODO: encapsulate as a tool; identical to CossSwapDir::create()
 207 void
 208 Rock::SwapDir::create()
 209 {
 210     assert(path);
 211     assert(filePath);
 212
 213     if (UsingSmp() && !IamDiskProcess()) {
 214         debugs (47,3, HERE << "disker will create in " << path);
 215         return;
 216     }
 217
 218     debugs (47,3, HERE << "creating in " << path);
 219
 220     struct stat dir_sb;
 221     if (::stat(path, &dir_sb) == 0) {
 222         struct stat file_sb;
 223         if (::stat(filePath, &file_sb) == 0) {
 224             debugs (47, DBG_IMPORTANT, "Skipping existing Rock db: " << filePath);
 225             return;
 226         }
 227         // else the db file is not there or is not accessible, and we will try
 228         // to create it later below, generating a detailed error on failures.
 229     } else { // path does not exist or is inaccessible
 230         // If path exists but is not accessible, mkdir() below will fail, and
 231         // the admin should see the error and act accordingly, so there is
 232         // no need to distinguish ENOENT from other possible stat() errors.
 233         debugs (47, DBG_IMPORTANT, "Creating Rock db directory: " << path);
 234         const int res = mkdir(path, 0700);
 235         if (res != 0)
 236             createError("mkdir");
 237     }
 238
 239     debugs (47, DBG_IMPORTANT, "Creating Rock db: " << filePath);
 240     const int swap = open(filePath, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY, 0600);
 241     if (swap < 0)
 242         createError("create");
 243
 244 #if SLOWLY_FILL_WITH_ZEROS
 245     char block[1024];
 246     Must(maxSize() % sizeof(block) == 0);
 247     memset(block, '\0', sizeof(block));
 248
 249     for (off_t offset = 0; offset < maxSize(); offset += sizeof(block)) {
 250         if (write(swap, block, sizeof(block)) != sizeof(block))
 251             createError("write");
 252     }
 253 #else
 254     if (ftruncate(swap, maxSize()) != 0)
 255         createError("truncate");
 256
 257     char header[HeaderSize];
 258     memset(header, '\0', sizeof(header));
 259     if (write(swap, header, sizeof(header)) != sizeof(header))
 260         createError("write");
 261 #endif
 262
 263     close(swap);
 264 }
 265
 266 // report Rock DB creation error and exit
 267 void
 268 Rock::SwapDir::createError(const char *const msg) {
 269     debugs(47, DBG_CRITICAL, "ERROR: Failed to initialize Rock Store db in " <<
 270            filePath << "; " << msg << " error: " << xstrerror());
 271     fatal("Rock Store db creation error");
 272 }
 273
 274 void
 275 Rock::SwapDir::init()
 276 {
 277     debugs(47,2, HERE);
 278
 279     // XXX: SwapDirs aren't refcounted. We make IORequestor calls, which
 280     // are refcounted. We up our count once to avoid implicit delete's.
 281     lock();
 282
 283     freeSlots = shm_old(Ipc::Mem::PageStack)(freeSlotsPath());
 284
 285     Must(!map);
 286     map = new DirMap(inodeMapPath());
 287     map->cleaner = this;
 288
 289     const char *ioModule = needsDiskStrand() ? "IpcIo" : "Blocking";
 290     if (DiskIOModule *m = DiskIOModule::Find(ioModule)) {
 291         debugs(47,2, HERE << "Using DiskIO module: " << ioModule);
 292         io = m->createStrategy();
 293         io->init();
 294     } else {
 295         debugs(47, DBG_CRITICAL, "FATAL: Rock store is missing DiskIO module: " <<
 296                ioModule);
 297         fatal("Rock Store missing a required DiskIO module");
 298     }
 299
 300     theFile = io->newFile(filePath);
 301     theFile->configure(fileConfig);
 302     theFile->open(O_RDWR, 0644, this);
 303
 304     // Increment early. Otherwise, if one SwapDir finishes rebuild before
 305     // others start, storeRebuildComplete() will think the rebuild is over!
 306     // TODO: move store_dirs_rebuilding hack to store modules that need it.
 307     ++StoreController::store_dirs_rebuilding;
 308 }
 309
 310 bool
 311 Rock::SwapDir::needsDiskStrand() const
 312 {
 313     const bool wontEvenWorkWithoutDisker = Config.workers > 1;
 314     const bool wouldWorkBetterWithDisker = DiskIOModule::Find("IpcIo");
 315     return InDaemonMode() && (wontEvenWorkWithoutDisker ||
 316                               wouldWorkBetterWithDisker);
 317 }
 318
 319 void
 320 Rock::SwapDir::parse(int anIndex, char *aPath)
 321 {
 322     index = anIndex;
 323
 324     path = xstrdup(aPath);
 325
 326     // cache store is located at path/db
 327     String fname(path);
 328     fname.append("/rock");
 329     filePath = xstrdup(fname.termedBuf());
 330
 331     parseSize(false);
 332     parseOptions(0);
 333
 334     // Current openForWriting() code overwrites the old slot if needed
 335     // and possible, so proactively removing old slots is probably useless.
 336     assert(!repl); // repl = createRemovalPolicy(Config.replPolicy);
 337
 338     validateOptions();
 339 }
 340
 341 void
 342 Rock::SwapDir::reconfigure()
 343 {
 344     parseSize(true);
 345     parseOptions(1);
 346     // TODO: can we reconfigure the replacement policy (repl)?
 347     validateOptions();
 348 }
 349
 350 /// parse maximum db disk size
 351 void
 352 Rock::SwapDir::parseSize(const bool reconfig)
 353 {
 354     const int i = GetInteger();
 355     if (i < 0)
 356         fatal("negative Rock cache_dir size value");
 357     const uint64_t new_max_size =
 358         static_cast<uint64_t>(i) << 20; // MBytes to Bytes
 359     if (!reconfig)
 360         max_size = new_max_size;
 361     else if (new_max_size != max_size) {
 362         debugs(3, DBG_IMPORTANT, "WARNING: cache_dir '" << path << "' size "
 363                "cannot be changed dynamically, value left unchanged (" <<
 364                (max_size >> 20) << " MB)");
 365     }
 366 }
 367
 368 ConfigOption *
 369 Rock::SwapDir::getOptionTree() const
 370 {
 371     ConfigOptionVector *vector = dynamic_cast<ConfigOptionVector*>(::SwapDir::getOptionTree());
 372     assert(vector);
 373     vector->options.push_back(new ConfigOptionAdapter<SwapDir>(*const_cast<SwapDir *>(this), &SwapDir::parseSizeOption, &SwapDir::dumpSizeOption));
 374     vector->options.push_back(new ConfigOptionAdapter<SwapDir>(*const_cast<SwapDir *>(this), &SwapDir::parseTimeOption, &SwapDir::dumpTimeOption));
 375     vector->options.push_back(new ConfigOptionAdapter<SwapDir>(*const_cast<SwapDir *>(this), &SwapDir::parseRateOption, &SwapDir::dumpRateOption));
 376     return vector;
 377 }
 378
 379 bool
 380 Rock::SwapDir::allowOptionReconfigure(const char *const option) const
 381 {
 382     return strcmp(option, "slot-size") != 0 &&
 383            ::SwapDir::allowOptionReconfigure(option);
 384 }
 385
 386 /// parses time-specific options; mimics ::SwapDir::optionObjectSizeParse()
 387 bool
 388 Rock::SwapDir::parseTimeOption(char const *option, const char *value, int reconfig)
 389 {
 390     // TODO: ::SwapDir or, better, Config should provide time-parsing routines,
 391     // including time unit handling. Same for size and rate.
 392
 393     time_msec_t *storedTime;
 394     if (strcmp(option, "swap-timeout") == 0)
 395         storedTime = &fileConfig.ioTimeout;
 396     else
 397         return false;
 398
 399     if (!value)
 400         self_destruct();
 401
 402     // TODO: handle time units and detect parsing errors better
 403     const int64_t parsedValue = strtoll(value, NULL, 10);
 404     if (parsedValue < 0) {
 405         debugs(3, DBG_CRITICAL, "FATAL: cache_dir " << path << ' ' << option << " must not be negative but is: " << parsedValue);
 406         self_destruct();
 407     }
 408
 409     const time_msec_t newTime = static_cast<time_msec_t>(parsedValue);
 410
 411     if (!reconfig)
 412         *storedTime = newTime;
 413     else if (*storedTime != newTime) {
 414         debugs(3, DBG_IMPORTANT, "WARNING: cache_dir " << path << ' ' << option
 415                << " cannot be changed dynamically, value left unchanged: " <<
 416                *storedTime);
 417     }
 418
 419     return true;
 420 }
 421
 422 /// reports time-specific options; mimics ::SwapDir::optionObjectSizeDump()
 423 void
 424 Rock::SwapDir::dumpTimeOption(StoreEntry * e) const
 425 {
 426     if (fileConfig.ioTimeout)
 427         storeAppendPrintf(e, " swap-timeout=%" PRId64,
 428                           static_cast<int64_t>(fileConfig.ioTimeout));
 429 }
 430
 431 /// parses rate-specific options; mimics ::SwapDir::optionObjectSizeParse()
 432 bool
 433 Rock::SwapDir::parseRateOption(char const *option, const char *value, int isaReconfig)
 434 {
 435     int *storedRate;
 436     if (strcmp(option, "max-swap-rate") == 0)
 437         storedRate = &fileConfig.ioRate;
 438     else
 439         return false;
 440
 441     if (!value)
 442         self_destruct();
 443
 444     // TODO: handle time units and detect parsing errors better
 445     const int64_t parsedValue = strtoll(value, NULL, 10);
 446     if (parsedValue < 0) {
 447         debugs(3, DBG_CRITICAL, "FATAL: cache_dir " << path << ' ' << option << " must not be negative but is: " << parsedValue);
 448         self_destruct();
 449     }
 450
 451     const int newRate = static_cast<int>(parsedValue);
 452
 453     if (newRate < 0) {
 454         debugs(3, DBG_CRITICAL, "FATAL: cache_dir " << path << ' ' << option << " must not be negative but is: " << newRate);
 455         self_destruct();
 456     }
 457
 458     if (!isaReconfig)
 459         *storedRate = newRate;
 460     else if (*storedRate != newRate) {
 461         debugs(3, DBG_IMPORTANT, "WARNING: cache_dir " << path << ' ' << option
 462                << " cannot be changed dynamically, value left unchanged: " <<
 463                *storedRate);
 464     }
 465
 466     return true;
 467 }
 468
 469 /// reports rate-specific options; mimics ::SwapDir::optionObjectSizeDump()
 470 void
 471 Rock::SwapDir::dumpRateOption(StoreEntry * e) const
 472 {
 473     if (fileConfig.ioRate >= 0)
 474         storeAppendPrintf(e, " max-swap-rate=%d", fileConfig.ioRate);
 475 }
 476
 477 /// parses size-specific options; mimics ::SwapDir::optionObjectSizeParse()
 478 bool
 479 Rock::SwapDir::parseSizeOption(char const *option, const char *value, int reconfig)
 480 {
 481     uint64_t *storedSize;
 482     if (strcmp(option, "slot-size") == 0)
 483         storedSize = &slotSize;
 484     else
 485         return false;
 486
 487     if (!value)
 488         self_destruct();
 489
 490     // TODO: handle size units and detect parsing errors better
 491     const uint64_t newSize = strtoll(value, NULL, 10);
 492     if (newSize <= 0) {
 493         debugs(3, DBG_CRITICAL, "FATAL: cache_dir " << path << ' ' << option << " must be positive; got: " << newSize);
 494         self_destruct();
 495     }
 496
 497     if (newSize <= sizeof(DbCellHeader)) {
 498         debugs(3, DBG_CRITICAL, "FATAL: cache_dir " << path << ' ' << option << " must exceed " << sizeof(DbCellHeader) << "; got: " << newSize);
 499         self_destruct();
 500     }
 501
 502     if (!reconfig)
 503         *storedSize = newSize;
 504     else if (*storedSize != newSize) {
 505         debugs(3, DBG_IMPORTANT, "WARNING: cache_dir " << path << ' ' << option
 506                << " cannot be changed dynamically, value left unchanged: " <<
 507                *storedSize);
 508     }
 509
 510     return true;
 511 }
 512
 513 /// reports size-specific options; mimics ::SwapDir::optionObjectSizeDump()
 514 void
 515 Rock::SwapDir::dumpSizeOption(StoreEntry * e) const
 516 {
 517     storeAppendPrintf(e, " slot-size=%" PRId64, slotSize);
 518 }
 519
 520 /// check the results of the configuration; only level-0 debugging works here
 521 void
 522 Rock::SwapDir::validateOptions()
 523 {
 524     if (slotSize <= 0)
 525         fatal("Rock store requires a positive slot-size");
 526
 527     const int64_t maxSizeRoundingWaste = 1024 * 1024; // size is configured in MB
 528     const int64_t slotSizeRoundingWaste = slotSize;
 529     const int64_t maxRoundingWaste =
 530         max(maxSizeRoundingWaste, slotSizeRoundingWaste);
 531     const int64_t usableDiskSize = diskOffset(entryLimitAllowed());
 532     const int64_t diskWasteSize = maxSize() - usableDiskSize;
 533     Must(diskWasteSize >= 0);
 534
 535     // warn if maximum db size is not reachable due to sfileno limit
 536     if (entryLimitAllowed() == entryLimitHigh() &&
 537             diskWasteSize >= maxRoundingWaste) {
 538         debugs(47, DBG_CRITICAL, "Rock store cache_dir[" << index << "] '" << path << "':");
 539         debugs(47, DBG_CRITICAL, "\tmaximum number of entries: " << entryLimitAllowed());
 540         debugs(47, DBG_CRITICAL, "\tdb slot size: " << slotSize << " Bytes");
 541         debugs(47, DBG_CRITICAL, "\tmaximum db size: " << maxSize() << " Bytes");
 542         debugs(47, DBG_CRITICAL, "\tusable db size:  " << usableDiskSize << " Bytes");
 543         debugs(47, DBG_CRITICAL, "\tdisk space waste: " << diskWasteSize << " Bytes");
 544         debugs(47, DBG_CRITICAL, "WARNING: Rock store config wastes space.");
 545     }
 546 }
 547
 548 void
 549 Rock::SwapDir::rebuild()
 550 {
 551     //++StoreController::store_dirs_rebuilding; // see Rock::SwapDir::init()
 552     AsyncJob::Start(new Rebuild(this));
 553 }
 554
 555 bool
 556 Rock::SwapDir::canStore(const StoreEntry &e, int64_t diskSpaceNeeded, int &load) const
 557 {
 558     if (!::SwapDir::canStore(e, sizeof(DbCellHeader)+diskSpaceNeeded, load))
 559         return false;
 560
 561     if (!theFile || !theFile->canWrite())
 562         return false;
 563
 564     if (!map)
 565         return false;
 566
 567     // Do not start I/O transaction if there are less than 10% free pages left.
 568     // TODO: reserve page instead
 569     if (needsDiskStrand() &&
 570             Ipc::Mem::PageLevel(Ipc::Mem::PageId::ioPage) >= 0.9 * Ipc::Mem::PageLimit(Ipc::Mem::PageId::ioPage)) {
 571         debugs(47, 5, HERE << "too few shared pages for IPC I/O left");
 572         return false;
 573     }
 574
 575     if (io->shedLoad())
 576         return false;
 577
 578     load = io->load();
 579     return true;
 580 }
 581
 582 StoreIOState::Pointer
 583 Rock::SwapDir::createStoreIO(StoreEntry &e, StoreIOState::STFNCB *cbFile, StoreIOState::STIOCB *cbIo, void *data)
 584 {
 585     if (!theFile || theFile->error()) {
 586         debugs(47,4, HERE << theFile);
 587         return NULL;
 588     }
 589
 590     sfileno filen;
 591     Ipc::StoreMapAnchor *const slot =
 592         map->openForWriting(reinterpret_cast<const cache_key *>(e.key), filen);
 593     if (!slot) {
 594         debugs(47, 5, HERE << "map->add failed");
 595         return NULL;
 596     }
 597
 598     assert(filen >= 0);
 599     slot->set(e);
 600
 601     // XXX: We rely on our caller, storeSwapOutStart(), to set e.fileno.
 602     // If that does not happen, the entry will not decrement the read level!
 603
 604     Rock::SwapDir::Pointer self(this);
 605     IoState *sio = new IoState(self, &e, cbFile, cbIo, data);
 606
 607     sio->swap_dirn = index;
 608     sio->swap_filen = filen;
 609     sio->writeableAnchor_ = slot;
 610
 611     debugs(47,5, HERE << "dir " << index << " created new filen " <<
 612            std::setfill('0') << std::hex << std::uppercase << std::setw(8) <<
 613            sio->swap_filen << std::dec << " starting at " <<
 614            diskOffset(sio->swap_filen));
 615
 616     sio->file(theFile);
 617
 618     trackReferences(e);
 619     return sio;
 620 }
 621
 622 int64_t
 623 Rock::SwapDir::diskOffset(int filen) const
 624 {
 625     assert(filen >= 0);
 626     return HeaderSize + slotSize*filen;
 627 }
 628
 629 int64_t
 630 Rock::SwapDir::diskOffset(Ipc::Mem::PageId &pageId) const
 631 {
 632     assert(pageId);
 633     return diskOffset(pageId.number - 1);
 634 }
 635
 636 int64_t
 637 Rock::SwapDir::diskOffsetLimit() const
 638 {
 639     assert(map);
 640     return diskOffset(map->entryLimit());
 641 }
 642
 643 int
 644 Rock::SwapDir::entryMaxPayloadSize() const
 645 {
 646     return slotSize - sizeof(DbCellHeader);
 647 }
 648
 649 int
 650 Rock::SwapDir::entriesNeeded(const int64_t objSize) const
 651 {
 652     return (objSize + entryMaxPayloadSize() - 1) / entryMaxPayloadSize();
 653 }
 654
 655 bool
 656 Rock::SwapDir::useFreeSlot(Ipc::Mem::PageId &pageId)
 657 {
 658     if (freeSlots->pop(pageId)) {
 659         debugs(47, 5, "got a previously free slot: " << pageId);
 660         return true;
 661     }
 662
 663     // catch free slots delivered to noteFreeMapSlice()
 664     assert(!waitingForPage);
 665     waitingForPage = &pageId;
 666     if (map->purgeOne()) {
 667         assert(!waitingForPage); // noteFreeMapSlice() should have cleared it
 668         assert(pageId.set());
 669         debugs(47, 5, "got a previously busy slot: " << pageId);
 670         return true;
 671     }
 672     assert(waitingForPage == &pageId);
 673     waitingForPage = NULL;
 674
 675     debugs(47, 3, "cannot get a slot; entries: " << map->entryCount());
 676     return false;
 677 }
 678
 679 bool
 680 Rock::SwapDir::validSlotId(const SlotId slotId) const
 681 {
 682     return 0 <= slotId && slotId < entryLimitAllowed();
 683 }
 684
 685 void
 686 Rock::SwapDir::noteFreeMapSlice(const sfileno sliceId)
 687 {
 688     Ipc::Mem::PageId pageId;
 689     pageId.pool = index+1;
 690     pageId.number = sliceId+1;
 691     if (waitingForPage) {
 692         *waitingForPage = pageId;
 693         waitingForPage = NULL;
 694     } else {
 695         freeSlots->push(pageId);
 696     }
 697 }
 698
 699 // tries to open an old entry with swap_filen for reading
 700 StoreIOState::Pointer
 701 Rock::SwapDir::openStoreIO(StoreEntry &e, StoreIOState::STFNCB *cbFile, StoreIOState::STIOCB *cbIo, void *data)
 702 {
 703     if (!theFile || theFile->error()) {
 704         debugs(47,4, HERE << theFile);
 705         return NULL;
 706     }
 707
 708     if (e.swap_filen < 0) {
 709         debugs(47,4, HERE << e);
 710         return NULL;
 711     }
 712
 713     // Do not start I/O transaction if there are less than 10% free pages left.
 714     // TODO: reserve page instead
 715     if (needsDiskStrand() &&
 716             Ipc::Mem::PageLevel(Ipc::Mem::PageId::ioPage) >= 0.9 * Ipc::Mem::PageLimit(Ipc::Mem::PageId::ioPage)) {
 717         debugs(47, 5, HERE << "too few shared pages for IPC I/O left");
 718         return NULL;
 719     }
 720
 721     // The are two ways an entry can get swap_filen: our get() locked it for
 722     // reading or our storeSwapOutStart() locked it for writing. Peeking at our
 723     // locked entry is safe, but no support for reading a filling entry.
 724     const Ipc::StoreMapAnchor *slot = map->peekAtReader(e.swap_filen);
 725     if (!slot)
 726         return NULL; // we were writing afterall
 727
 728     Rock::SwapDir::Pointer self(this);
 729     IoState *sio = new IoState(self, &e, cbFile, cbIo, data);
 730
 731     sio->swap_dirn = index;
 732     sio->swap_filen = e.swap_filen;
 733     sio->readableAnchor_ = slot;
 734     sio->file(theFile);
 735
 736     debugs(47,5, HERE << "dir " << index << " has old filen: " <<
 737            std::setfill('0') << std::hex << std::uppercase << std::setw(8) <<
 738            sio->swap_filen);
 739
 740     assert(slot->sameKey(static_cast<const cache_key*>(e.key)));
 741     assert(slot->basics.swap_file_sz > 0);
 742     assert(slot->basics.swap_file_sz == e.swap_file_sz);
 743
 744     return sio;
 745 }
 746
 747 void
 748 Rock::SwapDir::ioCompletedNotification()
 749 {
 750     if (!theFile)
 751         fatalf("Rock cache_dir failed to initialize db file: %s", filePath);
 752
 753     if (theFile->error())
 754         fatalf("Rock cache_dir at %s failed to open db file: %s", filePath,
 755                xstrerror());
 756
 757     debugs(47, 2, "Rock cache_dir[" << index << "] limits: " <<
 758            std::setw(12) << maxSize() << " disk bytes and " <<
 759            std::setw(7) << map->entryLimit() << " entries");
 760
 761     rebuild();
 762 }
 763
 764 void
 765 Rock::SwapDir::closeCompleted()
 766 {
 767     theFile = NULL;
 768 }
 769
 770 void
 771 Rock::SwapDir::readCompleted(const char *buf, int rlen, int errflag, RefCount< ::ReadRequest> r)
 772 {
 773     ReadRequest *request = dynamic_cast<Rock::ReadRequest*>(r.getRaw());
 774     assert(request);
 775     IoState::Pointer sio = request->sio;
 776
 777     if (errflag == DISK_OK && rlen > 0)
 778         sio->offset_ += rlen;
 779
 780     StoreIOState::STRCB *callb = sio->read.callback;
 781     assert(callb);
 782     sio->read.callback = NULL;
 783     void *cbdata;
 784     if (cbdataReferenceValidDone(sio->read.callback_data, &cbdata))
 785         callb(cbdata, r->buf, rlen, sio.getRaw());
 786 }
 787
 788 void
 789 Rock::SwapDir::writeCompleted(int errflag, size_t rlen, RefCount< ::WriteRequest> r)
 790 {
 791     Rock::WriteRequest *request = dynamic_cast<Rock::WriteRequest*>(r.getRaw());
 792     assert(request);
 793     assert(request->sio !=  NULL);
 794     IoState &sio = *request->sio;
 795
 796     // quit if somebody called IoState::close() while we were waiting
 797     if (!sio.stillWaiting()) {
 798         debugs(79, 3, "ignoring closed entry " << sio.swap_filen);
 799         return;
 800     }
 801
 802     if (errflag == DISK_OK) {
 803         // do not increment sio.offset_ because we do it in sio->write()
 804
 805         // finalize the shared slice info after writing slice contents to disk
 806         Ipc::StoreMap::Slice &slice =
 807             map->writeableSlice(sio.swap_filen, request->sidCurrent);
 808         slice.size = request->len - sizeof(DbCellHeader);
 809         slice.next = request->sidNext;
 810
 811         if (request->sidNext < 0) {
 812             // close, the entry gets the read lock
 813             map->closeForWriting(sio.swap_filen, true);
 814             sio.finishedWriting(errflag);
 815         }
 816     } else {
 817         writeError(*sio.e);
 818         sio.finishedWriting(errflag);
 819         // and hope that Core will call disconnect() to close the map entry
 820     }
 821
 822     CollapsedForwarding::Broadcast(*sio.e);
 823 }
 824
 825 void
 826 Rock::SwapDir::writeError(StoreEntry &e)
 827 {
 828     // Do not abortWriting here. The entry should keep the write lock
 829     // instead of losing association with the store and confusing core.
 830     map->freeEntry(e.swap_filen); // will mark as unusable, just in case
 831
 832     Store::Root().transientsAbandon(e);
 833
 834     // All callers must also call IoState callback, to propagate the error.
 835 }
 836
 837 bool
 838 Rock::SwapDir::full() const
 839 {
 840     return freeSlots != NULL && !freeSlots->size();
 841 }
 842
 843 // storeSwapOutFileClosed calls this nethod on DISK_NO_SPACE_LEFT,
 844 // but it should not happen for us
 845 void
 846 Rock::SwapDir::diskFull()
 847 {
 848     debugs(20, DBG_IMPORTANT, "BUG: No space left with rock cache_dir: " <<
 849            filePath);
 850 }
 851
 852 /// purge while full(); it should be sufficient to purge just one
 853 void
 854 Rock::SwapDir::maintain()
 855 {
 856     // The Store calls this to free some db space, but there is nothing wrong
 857     // with a full() db, except when db has to shrink after reconfigure, and
 858     // we do not support shrinking yet (it would have to purge specific slots).
 859     // TODO: Disable maintain() requests when they are pointless.
 860 }
 861
 862 void
 863 Rock::SwapDir::reference(StoreEntry &e)
 864 {
 865     debugs(47, 5, HERE << &e << ' ' << e.swap_dirn << ' ' << e.swap_filen);
 866     if (repl && repl->Referenced)
 867         repl->Referenced(repl, &e, &e.repl);
 868 }
 869
 870 bool
 871 Rock::SwapDir::dereference(StoreEntry &e, bool)
 872 {
 873     debugs(47, 5, HERE << &e << ' ' << e.swap_dirn << ' ' << e.swap_filen);
 874     if (repl && repl->Dereferenced)
 875         repl->Dereferenced(repl, &e, &e.repl);
 876
 877     // no need to keep e in the global store_table for us; we have our own map
 878     return false;
 879 }
 880
 881 bool
 882 Rock::SwapDir::unlinkdUseful() const
 883 {
 884     // no entry-specific files to unlink
 885     return false;
 886 }
 887
 888 void
 889 Rock::SwapDir::unlink(StoreEntry &e)
 890 {
 891     debugs(47, 5, HERE << e);
 892     ignoreReferences(e);
 893     map->freeEntry(e.swap_filen);
 894     disconnect(e);
 895 }
 896
 897 void
 898 Rock::SwapDir::markForUnlink(StoreEntry &e)
 899 {
 900     debugs(47, 5, e);
 901     map->freeEntry(e.swap_filen);
 902 }
 903
 904 void
 905 Rock::SwapDir::trackReferences(StoreEntry &e)
 906 {
 907     debugs(47, 5, HERE << e);
 908     if (repl)
 909         repl->Add(repl, &e, &e.repl);
 910 }
 911
 912 void
 913 Rock::SwapDir::ignoreReferences(StoreEntry &e)
 914 {
 915     debugs(47, 5, HERE << e);
 916     if (repl)
 917         repl->Remove(repl, &e, &e.repl);
 918 }
 919
 920 void
 921 Rock::SwapDir::statfs(StoreEntry &e) const
 922 {
 923     storeAppendPrintf(&e, "\n");
 924     storeAppendPrintf(&e, "Maximum Size: %" PRIu64 " KB\n", maxSize() >> 10);
 925     storeAppendPrintf(&e, "Current Size: %.2f KB %.2f%%\n",
 926                       currentSize() / 1024.0,
 927                       Math::doublePercent(currentSize(), maxSize()));
 928
 929     if (map) {
 930         const int limit = map->entryLimit();
 931         storeAppendPrintf(&e, "Maximum entries: %9d\n", limit);
 932         if (limit > 0) {
 933             const int entryCount = map->entryCount();
 934             storeAppendPrintf(&e, "Current entries: %9d %.2f%%\n",
 935                               entryCount, (100.0 * entryCount / limit));
 936
 937             const unsigned int slotsFree = !freeSlots ? 0 : freeSlots->size();
 938             if (slotsFree <= static_cast<const unsigned int>(limit)) {
 939                 const int usedSlots = limit - static_cast<const int>(slotsFree);
 940                 storeAppendPrintf(&e, "Used slots:      %9d %.2f%%\n",
 941                                   usedSlots, (100.0 * usedSlots / limit));
 942             }
 943             if (limit < 100) { // XXX: otherwise too expensive to count
 944                 Ipc::ReadWriteLockStats stats;
 945                 map->updateStats(stats);
 946                 stats.dump(e);
 947             }
 948         }
 949     }
 950
 951     storeAppendPrintf(&e, "Pending operations: %d out of %d\n",
 952                       store_open_disk_fd, Config.max_open_disk_fds);
 953
 954     storeAppendPrintf(&e, "Flags:");
 955
 956     if (flags.selected)
 957         storeAppendPrintf(&e, " SELECTED");
 958
 959     if (flags.read_only)
 960         storeAppendPrintf(&e, " READ-ONLY");
 961
 962     storeAppendPrintf(&e, "\n");
 963
 964 }
 965
 966 const char *
 967 Rock::SwapDir::inodeMapPath() const {
 968     static String inodesPath;
 969     inodesPath = path;
 970     inodesPath.append("_inodes");
 971     return inodesPath.termedBuf();
 972 }
 973
 974 const char *
 975 Rock::SwapDir::freeSlotsPath() const {
 976     static String spacesPath;
 977     spacesPath = path;
 978     spacesPath.append("_spaces");
 979     return spacesPath.termedBuf();
 980 }
 981
 982 namespace Rock
 983 {
 984 RunnerRegistrationEntry(rrAfterConfig, SwapDirRr);
 985 }
 986
 987 void Rock::SwapDirRr::create(const RunnerRegistry &)
 988 {
 989     Must(mapOwners.empty() && freeSlotsOwners.empty());
 990     for (int i = 0; i < Config.cacheSwap.n_configured; ++i) {
 991         if (const Rock::SwapDir *const sd = dynamic_cast<Rock::SwapDir *>(INDEXSD(i))) {
 992             const int64_t capacity = sd->entryLimitAllowed();
 993
 994             SwapDir::DirMap::Owner *const mapOwner =
 995                 SwapDir::DirMap::Init(sd->inodeMapPath(), capacity);
 996             mapOwners.push_back(mapOwner);
 997
 998             // XXX: remove pool id and counters from PageStack
 999             Ipc::Mem::Owner<Ipc::Mem::PageStack> *const freeSlotsOwner =
1000                 shm_new(Ipc::Mem::PageStack)(sd->freeSlotsPath(),
1001                                              i+1, capacity,
1002                                              sizeof(DbCellHeader));
1003             freeSlotsOwners.push_back(freeSlotsOwner);
1004
1005             // XXX: add method to initialize PageStack with no free pages
1006             while (true) {
1007                 Ipc::Mem::PageId pageId;
1008                 if (!freeSlotsOwner->object()->pop(pageId))
1009                     break;
1010             }
1011         }
1012     }
1013 }
1014
1015 Rock::SwapDirRr::~SwapDirRr()
1016 {
1017     for (size_t i = 0; i < mapOwners.size(); ++i) {
1018         delete mapOwners[i];
1019         delete freeSlotsOwners[i];
1020     }
1021 }