about summary refs log tree commit diff
path: root/src/nix-worker/nix-worker.cc
blob: a41fb2e154af1c78ab39b0b8ee02a332bb76bf4e (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
#include "shared.hh"
#include "local-store.hh"
#include "util.hh"
#include "serialise.hh"
#include "worker-protocol.hh"
#include "archive.hh"
#include "globals.hh"

#include <iostream>
#include <unistd.h>
#include <signal.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <sys/stat.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <fcntl.h>
#include <errno.h>

using namespace nix;


/* On platforms that have O_ASYNC, we can detect when a client
   disconnects and immediately kill any ongoing builds.  On platforms
   that lack it, we only notice the disconnection the next time we try
   to write to the client.  So if you have a builder that never
   generates output on stdout/stderr, the worker will never notice
   that the client has disconnected until the builder terminates. */
#ifdef O_ASYNC
#define HAVE_HUP_NOTIFICATION
#ifndef SIGPOLL
#define SIGPOLL SIGIO
#endif
#endif


static FdSource from(STDIN_FILENO);
static FdSink to(STDOUT_FILENO);

bool canSendStderr;
pid_t myPid;



/* This function is called anytime we want to write something to
   stderr.  If we're in a state where the protocol allows it (i.e.,
   when canSendStderr), send the message to the client over the
   socket. */
static void tunnelStderr(const unsigned char * buf, size_t count)
{
    /* Don't send the message to the client if we're a child of the
       process handling the connection.  Otherwise we could screw up
       the protocol.  It's up to the parent to redirect stderr and
       send it to the client somehow (e.g., as in build.cc). */
    if (canSendStderr && myPid == getpid()) {
        try {
            writeInt(STDERR_NEXT, to);
            writeString(string((char *) buf, count), to);
        } catch (...) {
            /* Write failed; that means that the other side is
               gone. */
            canSendStderr = false;
            throw;
        }
    } else
        writeFull(STDERR_FILENO, buf, count);
}


/* Return true if the remote side has closed its end of the
   connection, false otherwise.  Should not be called on any socket on
   which we expect input! */
static bool isFarSideClosed(int socket)
{
    struct timeval timeout;
    timeout.tv_sec = timeout.tv_usec = 0;

    fd_set fds;
    FD_ZERO(&fds);
    FD_SET(socket, &fds);

    while (select(socket + 1, &fds, 0, 0, &timeout) == -1)
        if (errno != EINTR) throw SysError("select()");

    if (!FD_ISSET(socket, &fds)) return false;

    /* Destructive read to determine whether the select() marked the
       socket as readable because there is actual input or because
       we've reached EOF (i.e., a read of size 0 is available). */
    char c;
    int rd;
    if ((rd = read(socket, &c, 1)) > 0)
        throw Error("EOF expected (protocol error?)");
    else if (rd == -1 && errno != ECONNRESET)
        throw SysError("expected connection reset or EOF");
    
    return true;
}


/* A SIGPOLL signal is received when data is available on the client
   communication socket, or when the client has closed its side of the
   socket.  This handler is enabled at precisely those moments in the
   protocol when we're doing work and the client is supposed to be
   quiet.  Thus, if we get a SIGPOLL signal, it means that the client
   has quit.  So we should quit as well.

   Too bad most operating systems don't support the POLL_HUP value for
   si_code in siginfo_t.  That would make most of the SIGPOLL
   complexity unnecessary, i.e., we could just enable SIGPOLL all the
   time and wouldn't have to worry about races. */
static void sigPollHandler(int sigNo)
{
    try {
        /* Check that the far side actually closed.  We're still
           getting spurious signals every once in a while.  I.e.,
           there is no input available, but we get a signal with
           POLL_IN set.  Maybe it's delayed or something. */
        if (isFarSideClosed(from.fd)) {
            if (!blockInt) {
                _isInterrupted = 1;
                blockInt = 1;
                canSendStderr = false;
                const char * s = "SIGPOLL\n";
                write(STDERR_FILENO, s, strlen(s));
            }
        } else {
            const char * s = "spurious SIGPOLL\n";
            write(STDERR_FILENO, s, strlen(s));
        }
    }
    catch (Error & e) {
        /* Shouldn't happen. */
        string s = "impossible: " + e.msg() + '\n';
        write(STDERR_FILENO, s.c_str(), s.size());
        throw;
    }
}


static void setSigPollAction(bool enable)
{
#ifdef HAVE_HUP_NOTIFICATION
    struct sigaction act, oact;
    act.sa_handler = enable ? sigPollHandler : SIG_IGN;
    sigfillset(&act.sa_mask);
    act.sa_flags = 0;
    if (sigaction(SIGPOLL, &act, &oact))
        throw SysError("setting handler for SIGPOLL");
#endif
}


/* startWork() means that we're starting an operation for which we
   want to send out stderr to the client. */
static void startWork()
{
    canSendStderr = true;

    /* Handle client death asynchronously. */
    setSigPollAction(true);

    /* Of course, there is a race condition here: the socket could
       have closed between when we last read from / wrote to it, and
       between the time we set the handler for SIGPOLL.  In that case
       we won't get the signal.  So do a non-blocking select() to find
       out if any input is available on the socket.  If there is, it
       has to be the 0-byte read that indicates that the socket has
       closed. */
    if (isFarSideClosed(from.fd)) {
        _isInterrupted = 1;
        checkInterrupt();
    }
}


/* stopWork() means that we're done; stop sending stderr to the
   client. */
static void stopWork(bool success = true, const string & msg = "")
{
    /* Stop handling async client death; we're going to a state where
       we're either sending or receiving from the client, so we'll be
       notified of client death anyway. */
    setSigPollAction(false);
    
    canSendStderr = false;

    if (success)
        writeInt(STDERR_LAST, to);
    else {
        writeInt(STDERR_ERROR, to);
        writeString(msg, to);
    }
}


struct TunnelSink : Sink
{
    Sink & to;
    TunnelSink(Sink & to) : to(to)
    {
    }
    virtual void operator ()
        (const unsigned char * data, unsigned int len)
    {
        writeInt(STDERR_WRITE, to);
        writeString(string((const char *) data, len), to);
    }
};


struct TunnelSource : Source
{
    Source & from;
    TunnelSource(Source & from) : from(from)
    {
    }
    virtual void operator ()
        (unsigned char * data, unsigned int len)
    {
        /* Careful: we're going to receive data from the client now,
           so we have to disable the SIGPOLL handler. */
        setSigPollAction(false);
        canSendStderr = false;
        
        writeInt(STDERR_READ, to);
        writeInt(len, to);
        string s = readString(from);
        if (s.size() != len) throw Error("not enough data");
        memcpy(data, (const unsigned char *) s.c_str(), len);

        startWork();
    }
};


/* If the NAR archive contains a single file at top-level, then save
   the contents of the file to `s'.  Otherwise barf. */
struct RetrieveRegularNARSink : ParseSink
{
    string s;

    void createDirectory(const Path & path)
    {
        throw Error("regular file expected");
    }

    void receiveContents(unsigned char * data, unsigned int len)
    {
        s.append((const char *) data, len);
    }

    void createSymlink(const Path & path, const string & target)
    {
        throw Error("regular file expected");
    }
};


/* Adapter class of a Source that saves all data read to `s'. */
struct SavingSourceAdapter : Source
{
    Source & orig;
    string s;
    SavingSourceAdapter(Source & orig) : orig(orig) { }
    void operator () (unsigned char * data, unsigned int len)
    {
        orig(data, len);
        s.append((const char *) data, len);
    }
};


static void performOp(unsigned int clientVersion,
    Source & from, Sink & to, unsigned int op)
{
    switch (op) {

#if 0        
    case wopQuit: {
        /* Close the database. */
        store.reset((StoreAPI *) 0);
        writeInt(1, to);
        break;
    }
#endif

    case wopIsValidPath: {
        Path path = readStorePath(from);
        startWork();
        bool result = store->isValidPath(path);
        stopWork();
        writeInt(result, to);
        break;
    }

    case wopHasSubstitutes: {
        Path path = readStorePath(from);
        startWork();
        bool result = store->hasSubstitutes(path);
        stopWork();
        writeInt(result, to);
        break;
    }

    case wopQueryPathHash: {
        Path path = readStorePath(from);
        startWork();
        Hash hash = store->queryPathHash(path);
        stopWork();
        writeString(printHash(hash), to);
        break;
    }

    case wopQueryReferences:
    case wopQueryReferrers:
    case wopQueryDerivationOutputs: {
        Path path = readStorePath(from);
        startWork();
        PathSet paths;
        if (op == wopQueryReferences)
            store->queryReferences(path, paths);
        else if (op == wopQueryReferrers)
            store->queryReferrers(path, paths);
        else paths = store->queryDerivationOutputs(path);
        stopWork();
        writeStringSet(paths, to);
        break;
    }

    case wopQueryDeriver: {
        Path path = readStorePath(from);
        startWork();
        Path deriver = store->queryDeriver(path);
        stopWork();
        writeString(deriver, to);
        break;
    }

    case wopAddToStore: {
        string baseName = readString(from);
        bool fixed = readInt(from) == 1; /* obsolete */
        bool recursive = readInt(from) == 1;
        string s = readString(from);
        /* Compatibility hack. */
        if (!fixed) {
            s = "sha256";
            recursive = true;
        }
        HashType hashAlgo = parseHashType(s);

        SavingSourceAdapter savedNAR(from);
        RetrieveRegularNARSink savedRegular;
        
        if (recursive) {
            /* Get the entire NAR dump from the client and save it to
               a string so that we can pass it to
               addToStoreFromDump(). */
            ParseSink sink; /* null sink; just parse the NAR */
            parseDump(sink, savedNAR);
        } else {
            parseDump(savedRegular, from);
        }
            
        startWork();
        Path path = dynamic_cast<LocalStore *>(store.get())
            ->addToStoreFromDump(recursive ? savedNAR.s : savedRegular.s, baseName, recursive, hashAlgo);
        stopWork();
        
        writeString(path, to);
        break;
    }

    case wopAddTextToStore: {
        string suffix = readString(from);
        string s = readString(from);
        PathSet refs = readStorePaths(from);
        startWork();
        Path path = store->addTextToStore(suffix, s, refs);
        stopWork();
        writeString(path, to);
        break;
    }

    case wopExportPath: {
        Path path = readStorePath(from);
        bool sign = readInt(from) == 1;
        startWork();
        TunnelSink sink(to);
        store->exportPath(path, sign, sink);
        stopWork();
        writeInt(1, to);
        break;
    }

    case wopImportPath: {
        startWork();
        TunnelSource source(from);
        Path path = store->importPath(true, source);
        stopWork();
        writeString(path, to);
        break;
    }

    case wopBuildDerivations: {
        PathSet drvs = readStorePaths(from);
        startWork();
        store->buildDerivations(drvs);
        stopWork();
        writeInt(1, to);
        break;
    }

    case wopEnsurePath: {
        Path path = readStorePath(from);
        startWork();
        store->ensurePath(path);
        stopWork();
        writeInt(1, to);
        break;
    }

    case wopAddTempRoot: {
        Path path = readStorePath(from);
        startWork();
        store->addTempRoot(path);
        stopWork();
        writeInt(1, to);
        break;
    }

    case wopAddIndirectRoot: {
        Path path = absPath(readString(from));
        startWork();
        store->addIndirectRoot(path);
        stopWork();
        writeInt(1, to);
        break;
    }

    case wopSyncWithGC: {
        startWork();
        store->syncWithGC();
        stopWork();
        writeInt(1, to);
        break;
    }

    case wopFindRoots: {
        startWork();
        Roots roots = store->findRoots();
        stopWork();
        writeInt(roots.size(), to);
        for (Roots::iterator i = roots.begin(); i != roots.end(); ++i) {
            writeString(i->first, to);
            writeString(i->second, to);
        }
        break;
    }

    case wopCollectGarbage: {
        GCOptions options;
        options.action = (GCOptions::GCAction) readInt(from);
        options.pathsToDelete = readStorePaths(from);
        options.ignoreLiveness = readInt(from);
        options.maxFreed = readLongLong(from);
        options.maxLinks = readInt(from);
        if (GET_PROTOCOL_MINOR(clientVersion) >= 5) {
            /* removed options */
            readInt(from);
            readInt(from);
        }

        GCResults results;
        
        startWork();
        if (options.ignoreLiveness)
            throw Error("you are not allowed to ignore liveness");
        store->collectGarbage(options, results);
        stopWork();
        
        writeStringSet(results.paths, to);
        writeLongLong(results.bytesFreed, to);
        writeLongLong(results.blocksFreed, to);
        
        break;
    }

    case wopSetOptions: {
        keepFailed = readInt(from) != 0;
        keepGoing = readInt(from) != 0;
        tryFallback = readInt(from) != 0;
        verbosity = (Verbosity) readInt(from);
        maxBuildJobs = readInt(from);
        maxSilentTime = readInt(from);
        if (GET_PROTOCOL_MINOR(clientVersion) >= 2)
            useBuildHook = readInt(from) != 0;
        if (GET_PROTOCOL_MINOR(clientVersion) >= 4) {
            buildVerbosity = (Verbosity) readInt(from);
            logType = (LogType) readInt(from);
            printBuildTrace = readInt(from) != 0;
        }
        startWork();
        stopWork();
        break;
    }

    case wopQuerySubstitutablePathInfo: {
        Path path = absPath(readString(from));
        startWork();
        SubstitutablePathInfo info;
        bool res = store->querySubstitutablePathInfo(path, info);
        stopWork();
        writeInt(res ? 1 : 0, to);
        if (res) {
            writeString(info.deriver, to);
            writeStringSet(info.references, to);
            writeLongLong(info.downloadSize, to);
        }
        break;
    }
            
    case wopQueryValidPaths: {
        startWork();
        PathSet paths = store->queryValidPaths();
        stopWork();
        writeStringSet(paths, to);
        break;
    }

    default:
        throw Error(format("invalid operation %1%") % op);
    }
}


static void processConnection()
{
    RemoveTempRoots removeTempRoots __attribute__((unused));

    canSendStderr = false;
    myPid = getpid();    
    writeToStderr = tunnelStderr;

#ifdef HAVE_HUP_NOTIFICATION
    /* Allow us to receive SIGPOLL for events on the client socket. */
    setSigPollAction(false);
    if (fcntl(from.fd, F_SETOWN, getpid()) == -1)
        throw SysError("F_SETOWN");
    if (fcntl(from.fd, F_SETFL, fcntl(from.fd, F_GETFL, 0) | O_ASYNC) == -1)
        throw SysError("F_SETFL");
#endif

    /* Exchange the greeting. */
    unsigned int magic = readInt(from);
    if (magic != WORKER_MAGIC_1) throw Error("protocol mismatch");
    writeInt(WORKER_MAGIC_2, to);

    writeInt(PROTOCOL_VERSION, to);
    unsigned int clientVersion = readInt(from);

    /* Send startup error messages to the client. */
    startWork();

    try {

        /* If we can't accept clientVersion, then throw an error
           *here* (not above). */

#if 0
        /* Prevent users from doing something very dangerous. */
        if (geteuid() == 0 &&
            querySetting("build-users-group", "") == "")
            throw Error("if you run `nix-worker' as root, then you MUST set `build-users-group'!");
#endif

        /* Open the store. */
        store = boost::shared_ptr<StoreAPI>(new LocalStore());

        stopWork();
        
    } catch (Error & e) {
        stopWork(false, e.msg());
        return;
    }

    /* Process client requests. */
    unsigned int opCount = 0;
    
    while (true) {
        WorkerOp op;
        try {
            op = (WorkerOp) readInt(from);
        } catch (EndOfFile & e) {
            break;
        }

        opCount++;

        try {
            performOp(clientVersion, from, to, op);
        } catch (Error & e) {
            stopWork(false, e.msg());
        }

        assert(!canSendStderr);
    };

    printMsg(lvlError, format("%1% worker operations") % opCount);
}


static void sigChldHandler(int sigNo)
{
    /* Reap all dead children. */
    while (waitpid(-1, 0, WNOHANG) > 0) ;
}


static void setSigChldAction(bool autoReap)
{
    struct sigaction act, oact;
    act.sa_handler = autoReap ? sigChldHandler : SIG_DFL;
    sigfillset(&act.sa_mask);
    act.sa_flags = 0;
    if (sigaction(SIGCHLD, &act, &oact))
        throw SysError("setting SIGCHLD handler");
}


static void daemonLoop()
{
    /* Get rid of children automatically; don't let them become
       zombies. */
    setSigChldAction(true);
    
    /* Create and bind to a Unix domain socket. */
    AutoCloseFD fdSocket = socket(PF_UNIX, SOCK_STREAM, 0);
    if (fdSocket == -1)
        throw SysError("cannot create Unix domain socket");

    string socketPath = nixStateDir + DEFAULT_SOCKET_PATH;

    createDirs(dirOf(socketPath));

    /* Urgh, sockaddr_un allows path names of only 108 characters.  So
       chdir to the socket directory so that we can pass a relative
       path name. */
    chdir(dirOf(socketPath).c_str());
    Path socketPathRel = "./" + baseNameOf(socketPath);
    
    struct sockaddr_un addr;
    addr.sun_family = AF_UNIX;
    if (socketPathRel.size() >= sizeof(addr.sun_path))
        throw Error(format("socket path `%1%' is too long") % socketPathRel);
    strcpy(addr.sun_path, socketPathRel.c_str());

    unlink(socketPath.c_str());

    /* Make sure that the socket is created with 0666 permission
       (everybody can connect --- provided they have access to the
       directory containing the socket). */
    mode_t oldMode = umask(0111);
    int res = bind(fdSocket, (struct sockaddr *) &addr, sizeof(addr));
    umask(oldMode);
    if (res == -1)
        throw SysError(format("cannot bind to socket `%1%'") % socketPath);

    chdir("/"); /* back to the root */

    if (listen(fdSocket, 5) == -1)
        throw SysError(format("cannot listen on socket `%1%'") % socketPath);

    /* Loop accepting connections. */
    while (1) {

        try {
            /* Important: the server process *cannot* open the
               Berkeley DB environment, because it doesn't like forks
               very much. */
            assert(!store);
            
            /* Accept a connection. */
            struct sockaddr_un remoteAddr;
            socklen_t remoteAddrLen = sizeof(remoteAddr);

            AutoCloseFD remote = accept(fdSocket,
                (struct sockaddr *) &remoteAddr, &remoteAddrLen);
            checkInterrupt();
            if (remote == -1) {
		if (errno == EINTR)
		    continue;
		else
		    throw SysError("accepting connection");
            }

            printMsg(lvlInfo, format("accepted connection %1%") % remote);

            /* Fork a child to handle the connection. */
            pid_t child;
            child = fork();
    
            switch (child) {
        
            case -1:
                throw SysError("unable to fork");

            case 0:
                try { /* child */

                    /* Background the worker. */
                    if (setsid() == -1)
                        throw SysError(format("creating a new session"));

                    /* Restore normal handling of SIGCHLD. */
                    setSigChldAction(false);

                    /* Since the daemon can be long-running, the
                       settings may have changed.  So force a reload. */
                    reloadSettings();
                    
                    /* Handle the connection. */
                    from.fd = remote;
                    to.fd = remote;
                    processConnection();
                    
                } catch (std::exception & e) {
                    std::cerr << format("child error: %1%\n") % e.what();
                }
                exit(0);
            }

        } catch (Interrupted & e) {
            throw;
        } catch (Error & e) {
            printMsg(lvlError, format("error processing connection: %1%") % e.msg());
        }
    }
}


void run(Strings args)
{
    bool slave = false;
    bool daemon = false;
    
    for (Strings::iterator i = args.begin(); i != args.end(); ) {
        string arg = *i++;
        if (arg == "--slave") slave = true;
        if (arg == "--daemon") daemon = true;
    }

    if (slave) {
        /* This prevents us from receiving signals from the terminal
           when we're running in setuid mode. */
        if (setsid() == -1)
            throw SysError(format("creating a new session"));

        processConnection();
    }

    else if (daemon) {
        if (setuidMode)
            throw Error("daemon cannot be started in setuid mode");
        chdir("/");
        daemonLoop();
    }

    else
        throw Error("must be run in either --slave or --daemon mode");
}


#include "help.txt.hh"

void printHelp()
{
    std::cout << string((char *) helpText);
}


string programId = "nix-worker";