git.sesse.net Git - plocate/blob - plocate.cpp

   1 #include "db.h"
   2 #include "io_uring_engine.h"
   3
   4 #include <algorithm>
   5 #include <arpa/inet.h>
   6 #include <assert.h>
   7 #include <chrono>
   8 #include <endian.h>
   9 #include <fcntl.h>
  10 #include <functional>
  11 #include <getopt.h>
  12 #include <limits.h>
  13 #include <memory>
  14 #include <stdio.h>
  15 #include <string.h>
  16 #include <string>
  17 #include <unistd.h>
  18 #include <unordered_map>
  19 #include <vector>
  20 #include <zstd.h>
  21
  22 using namespace std;
  23 using namespace std::chrono;
  24
  25 #define dprintf(...)
  26 //#define dprintf(...) fprintf(stderr, __VA_ARGS__);
  27
  28 #include "turbopfor.h"
  29
  30 const char *dbpath = "/var/lib/mlocate/plocate.db";
  31 bool only_count = false;
  32 bool print_nul = false;
  33 int64_t limit_matches = -1;
  34
  35 class Serializer {
  36 public:
  37         bool ready_to_print(int seq) { return next_seq == seq; }
  38         void print_delayed(int seq, const vector<string> msg);
  39         void release_current();
  40
  41 private:
  42         int next_seq = 0;
  43         struct Element {
  44                 int seq;
  45                 vector<string> msg;
  46
  47                 bool operator<(const Element &other) const
  48                 {
  49                         return seq > other.seq;
  50                 }
  51         };
  52         priority_queue<Element> pending;
  53 };
  54
  55 void Serializer::print_delayed(int seq, const vector<string> msg)
  56 {
  57         pending.push(Element{ seq, move(msg) });
  58 }
  59
  60 void Serializer::release_current()
  61 {
  62         ++next_seq;
  63
  64         // See if any delayed prints can now be dealt with.
  65         while (!pending.empty() && pending.top().seq == next_seq) {
  66                 if (limit_matches-- <= 0) return;
  67                 for (const string &msg : pending.top().msg) {
  68                         if (print_nul) {
  69                                 printf("%s%c", msg.c_str(), 0);
  70                         } else {
  71                                 printf("%s\n", msg.c_str());
  72                         }
  73                 }
  74                 pending.pop();
  75                 ++next_seq;
  76         }
  77 }
  78
  79 static inline uint32_t read_unigram(const string &s, size_t idx)
  80 {
  81         if (idx < s.size()) {
  82                 return (unsigned char)s[idx];
  83         } else {
  84                 return 0;
  85         }
  86 }
  87
  88 static inline uint32_t read_trigram(const string &s, size_t start)
  89 {
  90         return read_unigram(s, start) | (read_unigram(s, start + 1) << 8) |
  91                 (read_unigram(s, start + 2) << 16);
  92 }
  93
  94 bool has_access(const char *filename,
  95                 unordered_map<string, bool> *access_rx_cache)
  96 {
  97         const char *end = strchr(filename + 1, '/');
  98         while (end != nullptr) {
  99                 string parent_path(filename, end);
 100                 auto it = access_rx_cache->find(parent_path);
 101                 bool ok;
 102                 if (it == access_rx_cache->end()) {
 103                         ok = access(parent_path.c_str(), R_OK | X_OK) == 0;
 104                         access_rx_cache->emplace(move(parent_path), ok);
 105                 } else {
 106                         ok = it->second;
 107                 }
 108                 if (!ok) {
 109                         return false;
 110                 }
 111                 end = strchr(end + 1, '/');
 112         }
 113
 114         return true;
 115 }
 116
 117 class Corpus {
 118 public:
 119         Corpus(int fd, IOUringEngine *engine);
 120         ~Corpus();
 121         void find_trigram(uint32_t trgm, function<void(const Trigram *trgmptr, size_t len)> cb);
 122         void get_compressed_filename_block(uint32_t docid, function<void(string)> cb) const;
 123         size_t get_num_filename_blocks() const;
 124         off_t offset_for_block(uint32_t docid) const
 125         {
 126                 return hdr.filename_index_offset_bytes + docid * sizeof(uint64_t);
 127         }
 128
 129 public:
 130         const int fd;
 131         IOUringEngine *const engine;
 132
 133         Header hdr;
 134 };
 135
 136 Corpus::Corpus(int fd, IOUringEngine *engine)
 137         : fd(fd), engine(engine)
 138 {
 139         // Enable to test cold-cache behavior (except for access()).
 140         if (false) {
 141                 off_t len = lseek(fd, 0, SEEK_END);
 142                 if (len == -1) {
 143                         perror("lseek");
 144                         exit(1);
 145                 }
 146                 posix_fadvise(fd, 0, len, POSIX_FADV_DONTNEED);
 147         }
 148
 149         complete_pread(fd, &hdr, sizeof(hdr), /*offset=*/0);
 150         if (memcmp(hdr.magic, "\0plocate", 8) != 0) {
 151                 fprintf(stderr, "plocate.db is corrupt or an old version; please rebuild it.\n");
 152                 exit(1);
 153         }
 154         if (hdr.version != 0) {
 155                 fprintf(stderr, "plocate.db has version %u, expected 0; please rebuild it.\n", hdr.version);
 156                 exit(1);
 157         }
 158 }
 159
 160 Corpus::~Corpus()
 161 {
 162         close(fd);
 163 }
 164
 165 void Corpus::find_trigram(uint32_t trgm, function<void(const Trigram *trgmptr, size_t len)> cb)
 166 {
 167         uint32_t bucket = hash_trigram(trgm, hdr.hashtable_size);
 168         engine->submit_read(fd, sizeof(Trigram) * (hdr.extra_ht_slots + 2), hdr.hash_table_offset_bytes + sizeof(Trigram) * bucket, [this, trgm, cb{ move(cb) }](string s) {
 169                 const Trigram *trgmptr = reinterpret_cast<const Trigram *>(s.data());
 170                 for (unsigned i = 0; i < hdr.extra_ht_slots + 1; ++i) {
 171                         if (trgmptr[i].trgm == trgm) {
 172                                 cb(trgmptr + i, trgmptr[i + 1].offset - trgmptr[i].offset);
 173                                 return;
 174                         }
 175                 }
 176
 177                 // Not found.
 178                 cb(nullptr, 0);
 179         });
 180 }
 181
 182 void Corpus::get_compressed_filename_block(uint32_t docid, function<void(string)> cb) const
 183 {
 184         // Read the file offset from this docid and the next one.
 185         // This is always allowed, since we have a sentinel block at the end.
 186         engine->submit_read(fd, sizeof(uint64_t) * 2, offset_for_block(docid), [this, cb{ move(cb) }](string s) {
 187                 const uint64_t *ptr = reinterpret_cast<const uint64_t *>(s.data());
 188                 off_t offset = ptr[0];
 189                 size_t len = ptr[1] - ptr[0];
 190                 engine->submit_read(fd, len, offset, cb);
 191         });
 192 }
 193
 194 size_t Corpus::get_num_filename_blocks() const
 195 {
 196         return hdr.num_docids;
 197 }
 198
 199 uint64_t scan_file_block(const vector<string> &needles, string_view compressed,
 200                        unordered_map<string, bool> *access_rx_cache, int seq,
 201                        Serializer *serializer)
 202 {
 203         uint64_t matched = 0;
 204
 205         unsigned long long uncompressed_len = ZSTD_getFrameContentSize(compressed.data(), compressed.size());
 206         if (uncompressed_len == ZSTD_CONTENTSIZE_UNKNOWN || uncompressed_len == ZSTD_CONTENTSIZE_ERROR) {
 207                 fprintf(stderr, "ZSTD_getFrameContentSize() failed\n");
 208                 exit(1);
 209         }
 210
 211         string block;
 212         block.resize(uncompressed_len + 1);
 213
 214         size_t err = ZSTD_decompress(&block[0], block.size(), compressed.data(),
 215                                      compressed.size());
 216         if (ZSTD_isError(err)) {
 217                 fprintf(stderr, "ZSTD_decompress(): %s\n", ZSTD_getErrorName(err));
 218                 exit(1);
 219         }
 220         block[block.size() - 1] = '\0';
 221
 222         bool immediate_print = (serializer == nullptr || serializer->ready_to_print(seq));
 223         vector<string> delayed;
 224
 225         for (const char *filename = block.data();
 226              filename != block.data() + block.size();
 227              filename += strlen(filename) + 1) {
 228                 bool found = true;
 229                 for (const string &needle : needles) {
 230                         if (strstr(filename, needle.c_str()) == nullptr) {
 231                                 found = false;
 232                                 break;
 233                         }
 234                 }
 235                 if (found && has_access(filename, access_rx_cache)) {
 236                         if (limit_matches-- <= 0) break;
 237                         ++matched;
 238                         if (only_count) continue;
 239                         if (immediate_print) {
 240                                 if (print_nul) {
 241                                         printf("%s%c", filename, 0);
 242                                 } else {
 243                                         printf("%s\n", filename);
 244                                 }
 245                         } else {
 246                                 delayed.push_back(filename);
 247                         }
 248                 }
 249         }
 250         if (serializer != nullptr && !only_count) {
 251                 if (immediate_print) {
 252                         serializer->release_current();
 253                 } else {
 254                         serializer->print_delayed(seq, move(delayed));
 255                 }
 256         }
 257         return matched;
 258 }
 259
 260 size_t scan_docids(const vector<string> &needles, const vector<uint32_t> &docids, const Corpus &corpus, IOUringEngine *engine)
 261 {
 262         Serializer docids_in_order;
 263         unordered_map<string, bool> access_rx_cache;
 264         uint64_t matched = 0;
 265         for (size_t i = 0; i < docids.size(); ++i) {
 266                 uint32_t docid = docids[i];
 267                 corpus.get_compressed_filename_block(docid, [i, &matched, &needles, &access_rx_cache, &docids_in_order](string compressed) {
 268                         matched += scan_file_block(needles, compressed, &access_rx_cache, i, &docids_in_order);
 269                 });
 270         }
 271         engine->finish();
 272         return matched;
 273 }
 274
 275 // We do this sequentially, as it's faster than scattering
 276 // a lot of I/O through io_uring and hoping the kernel will
 277 // coalesce it plus readahead for us.
 278 uint64_t scan_all_docids(const vector<string> &needles, int fd, const Corpus &corpus, IOUringEngine *engine)
 279 {
 280         unordered_map<string, bool> access_rx_cache;
 281         uint32_t num_blocks = corpus.get_num_filename_blocks();
 282         unique_ptr<uint64_t[]> offsets(new uint64_t[num_blocks + 1]);
 283         complete_pread(fd, offsets.get(), (num_blocks + 1) * sizeof(uint64_t), corpus.offset_for_block(0));
 284         string compressed;
 285         uint64_t matched = 0;
 286         for (uint32_t io_docid = 0; io_docid < num_blocks; io_docid += 32) {
 287                 uint32_t last_docid = std::min(io_docid + 32, num_blocks);
 288                 size_t io_len = offsets[last_docid] - offsets[io_docid];
 289                 if (compressed.size() < io_len) {
 290                         compressed.resize(io_len);
 291                 }
 292                 complete_pread(fd, &compressed[0], io_len, offsets[io_docid]);
 293
 294                 for (uint32_t docid = io_docid; docid < last_docid; ++docid) {
 295                         size_t relative_offset = offsets[docid] - offsets[io_docid];
 296                         size_t len = offsets[docid + 1] - offsets[docid];
 297                         matched += scan_file_block(needles, { &compressed[relative_offset], len }, &access_rx_cache, 0, nullptr);
 298                         if (limit_matches <= 0) return matched;
 299                 }
 300         }
 301         return matched;
 302 }
 303
 304 void do_search_file(const vector<string> &needles, const char *filename)
 305 {
 306         int fd = open(filename, O_RDONLY);
 307         if (fd == -1) {
 308                 perror(filename);
 309                 exit(1);
 310         }
 311
 312         // Drop privileges.
 313         if (setgid(getgid()) != 0) {
 314                 perror("setgid");
 315                 exit(EXIT_FAILURE);
 316         }
 317
 318         steady_clock::time_point start __attribute__((unused)) = steady_clock::now();
 319         if (access("/", R_OK | X_OK)) {
 320                 // We can't find anything, no need to bother...
 321                 return;
 322         }
 323
 324         IOUringEngine engine;
 325         Corpus corpus(fd, &engine);
 326         dprintf("Corpus init done after %.1f ms.\n", 1e3 * duration<float>(steady_clock::now() - start).count());
 327
 328         vector<pair<Trigram, size_t>> trigrams;
 329         uint64_t shortest_so_far = numeric_limits<uint32_t>::max();
 330         for (const string &needle : needles) {
 331                 if (needle.size() < 3)
 332                         continue;
 333                 for (size_t i = 0; i < needle.size() - 2; ++i) {
 334                         uint32_t trgm = read_trigram(needle, i);
 335                         corpus.find_trigram(trgm, [trgm, &trigrams, &shortest_so_far](const Trigram *trgmptr, size_t len) {
 336                                 if (trgmptr == nullptr) {
 337                                         dprintf("trigram '%c%c%c' isn't found, we abort the search\n",
 338                                                 trgm & 0xff, (trgm >> 8) & 0xff, (trgm >> 16) & 0xff);
 339                                         if (only_count) {
 340                                                 printf("0\n");
 341                                         }
 342                                         exit(0);
 343                                 }
 344                                 if (trgmptr->num_docids > shortest_so_far * 100) {
 345                                         dprintf("not loading trigram '%c%c%c' with %u docids, it would be ignored later anyway\n",
 346                                                 trgm & 0xff, (trgm >> 8) & 0xff, (trgm >> 16) & 0xff,
 347                                                 trgmptr->num_docids);
 348                                 } else {
 349                                         trigrams.emplace_back(*trgmptr, len);
 350                                         shortest_so_far = std::min<uint64_t>(shortest_so_far, trgmptr->num_docids);
 351                                 }
 352                         });
 353                 }
 354         }
 355         engine.finish();
 356         dprintf("Hashtable lookups done after %.1f ms.\n", 1e3 * duration<float>(steady_clock::now() - start).count());
 357
 358         if (trigrams.empty()) {
 359                 // Too short for trigram matching. Apply brute force.
 360                 // (We could have searched through all trigrams that matched
 361                 // the pattern and done a union of them, but that's a lot of
 362                 // work for fairly unclear gain.)
 363                 uint64_t matched = scan_all_docids(needles, fd, corpus, &engine);
 364                 printf("%zu\n", matched);
 365                 return;
 366         }
 367         sort(trigrams.begin(), trigrams.end());
 368         {
 369                 auto last = unique(trigrams.begin(), trigrams.end());
 370                 trigrams.erase(last, trigrams.end());
 371         }
 372         sort(trigrams.begin(), trigrams.end(),
 373              [&](const pair<Trigram, size_t> &a, const pair<Trigram, size_t> &b) {
 374                      return a.first.num_docids < b.first.num_docids;
 375              });
 376
 377         vector<uint32_t> in1, in2, out;
 378         bool done = false;
 379         for (auto [trgmptr, len] : trigrams) {
 380                 if (!in1.empty() && trgmptr.num_docids > in1.size() * 100) {
 381                         uint32_t trgm __attribute__((unused)) = trgmptr.trgm;
 382                         dprintf("trigram '%c%c%c' (%zu bytes) has %u entries, ignoring the rest (will "
 383                                 "weed out false positives later)\n",
 384                                 trgm & 0xff, (trgm >> 8) & 0xff, (trgm >> 16) & 0xff,
 385                                 len, trgmptr.num_docids);
 386                         break;
 387                 }
 388
 389                 // Only stay a certain amount ahead, so that we don't spend I/O
 390                 // on reading the latter, large posting lists. We are unlikely
 391                 // to need them anyway, even if they should come in first.
 392                 if (engine.get_waiting_reads() >= 5) {
 393                         engine.finish();
 394                         if (done)
 395                                 break;
 396                 }
 397                 engine.submit_read(fd, len, trgmptr.offset, [trgmptr{trgmptr}, len{len}, &done, &in1, &in2, &out](string s) {
 398                         if (done)
 399                                 return;
 400                         uint32_t trgm __attribute__((unused)) = trgmptr.trgm;
 401                         size_t num = trgmptr.num_docids;
 402                         unsigned char *pldata = reinterpret_cast<unsigned char *>(s.data());
 403                         if (in1.empty()) {
 404                                 in1.resize(num + 128);
 405                                 decode_pfor_delta1<128>(pldata, num, /*interleaved=*/true, &in1[0]);
 406                                 in1.resize(num);
 407                                 dprintf("trigram '%c%c%c' (%zu bytes) decoded to %zu entries\n", trgm & 0xff,
 408                                         (trgm >> 8) & 0xff, (trgm >> 16) & 0xff, len, num);
 409                         } else {
 410                                 if (in2.size() < num + 128) {
 411                                         in2.resize(num + 128);
 412                                 }
 413                                 decode_pfor_delta1<128>(pldata, num, /*interleaved=*/true, &in2[0]);
 414
 415                                 out.clear();
 416                                 set_intersection(in1.begin(), in1.end(), in2.begin(), in2.begin() + num,
 417                                                  back_inserter(out));
 418                                 swap(in1, out);
 419                                 dprintf("trigram '%c%c%c' (%zu bytes) decoded to %zu entries, %zu left\n",
 420                                         trgm & 0xff, (trgm >> 8) & 0xff, (trgm >> 16) & 0xff,
 421                                         len, num, in1.size());
 422                                 if (in1.empty()) {
 423                                         dprintf("no matches (intersection list is empty)\n");
 424                                         done = true;
 425                                 }
 426                         }
 427                 });
 428         }
 429         engine.finish();
 430         if (done) {
 431                 return;
 432         }
 433         dprintf("Intersection done after %.1f ms. Doing final verification and printing:\n",
 434                 1e3 * duration<float>(steady_clock::now() - start).count());
 435
 436         uint64_t matched = scan_docids(needles, in1, corpus, &engine);
 437         dprintf("Done in %.1f ms, found %zu matches.\n",
 438                 1e3 * duration<float>(steady_clock::now() - start).count(), matched);
 439
 440         if (only_count) {
 441                 printf("%zu\n", matched);
 442         }
 443 }
 444
 445 void usage()
 446 {
 447         // The help text comes from mlocate.
 448         printf("Usage: plocate [OPTION]... PATTERN...\n");
 449         printf("\n");
 450         printf("  -c, --count            only print number of found entries\n");
 451         printf("  -d, --database DBPATH  use DBPATH instead of default database (which is\n");
 452         printf("                         %s)\n", dbpath);
 453         printf("  -h, --help             print this help\n");
 454         printf("  -l, --limit, -n LIMIT  limit output (or counting) to LIMIT entries\n");
 455         printf("  -0, --null             separate entries with NUL on output\n");
 456 }
 457
 458 int main(int argc, char **argv)
 459 {
 460         static const struct option long_options[] = {
 461                 { "help", no_argument, 0, 'h' },
 462                 { "count", no_argument, 0, 'c' },
 463                 { "database", required_argument, 0, 'd' },
 464                 { "limit", required_argument, 0, 'l' },
 465                 { nullptr, required_argument, 0, 'n' },
 466                 { "null", no_argument, 0, '0' },
 467                 { 0, 0, 0, 0 }
 468         };
 469
 470         for (;;) {
 471                 int option_index = 0;
 472                 int c = getopt_long(argc, argv, "cd:hl:n:0", long_options, &option_index);
 473                 if (c == -1) {
 474                         break;
 475                 }
 476                 switch (c) {
 477                 case 'c':
 478                         only_count = true;
 479                         break;
 480                 case 'd':
 481                         dbpath = strdup(optarg);
 482                         break;
 483                 case 'h':
 484                         usage();
 485                         exit(0);
 486                 case 'l':
 487                 case 'n':
 488                         limit_matches = atoll(optarg);
 489                         break;
 490                 case '0':
 491                         print_nul = true;
 492                         break;
 493                 default:
 494                         exit(1);
 495                 }
 496         }
 497
 498         vector<string> needles;
 499         for (int i = optind; i < argc; ++i) {
 500                 needles.push_back(argv[i]);
 501         }
 502         if (needles.empty()) {
 503                 fprintf(stderr, "plocate: no pattern to search for specified\n");
 504                 exit(0);
 505         }
 506         do_search_file(needles, dbpath);
 507 }