summaryrefslogtreecommitdiffstats
path: root/src/krbd.cc
blob: 5b379e6df14b1b3a49f56dc8ff46680d7ba51567 (plain)
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
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
/*
 * Ceph - scalable distributed file system
 *
 * Copyright (C) 2014 Inktank Storage, Inc.
 *
 * This is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License version 2.1, as published by the Free Software
 * Foundation.  See file COPYING.
 *
 */

#include <errno.h>
#include <fcntl.h>
#include <iostream>
#include <optional>
#include <poll.h>
#include <sstream>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <string>
#include <sys/stat.h>
#include <sys/sysmacros.h>
#include <sys/types.h>
#include <tuple>
#include <unistd.h>
#include <utility>

#include "auth/KeyRing.h"
#include "common/errno.h"
#include "common/Formatter.h"
#include "common/module.h"
#include "common/run_cmd.h"
#include "common/safe_io.h"
#include "common/secret.h"
#include "common/TextTable.h"
#include "common/Thread.h"
#include "include/ceph_assert.h"
#include "include/stringify.h"
#include "include/krbd.h"
#include "mon/MonMap.h"

#include <blkid/blkid.h>
#include <boost/algorithm/string/predicate.hpp>
#include <boost/tokenizer.hpp>
#include <libudev.h>

static const int UDEV_BUF_SIZE = 1 << 20;  /* doubled to 2M (SO_RCVBUFFORCE) */

struct krbd_ctx {
  CephContext *cct;
  struct udev *udev;
  uint32_t flags;  /* KRBD_CTX_F_* */
};

static const std::string SNAP_HEAD_NAME("-");

struct krbd_spec {
  std::string pool_name;
  std::string nspace_name;
  std::string image_name;
  std::string snap_name;

  krbd_spec(const char *pool_name, const char *nspace_name,
            const char *image_name, const char *snap_name)
      : pool_name(pool_name),
        nspace_name(nspace_name),
        image_name(image_name),
        snap_name(*snap_name ? snap_name : SNAP_HEAD_NAME) { }

  bool operator==(const krbd_spec& rhs) const {
    return pool_name == rhs.pool_name &&
           nspace_name == rhs.nspace_name &&
           image_name == rhs.image_name &&
           snap_name == rhs.snap_name;
  }
};

std::ostream& operator<<(std::ostream& os, const krbd_spec& spec) {
  os << spec.pool_name << "/";
  if (!spec.nspace_name.empty())
    os << spec.nspace_name << "/";
  os << spec.image_name;
  if (spec.snap_name != SNAP_HEAD_NAME)
    os << "@" << spec.snap_name;
  return os;
}

std::optional<krbd_spec> spec_from_dev(udev_device *dev) {
  const char *pool_name = udev_device_get_sysattr_value(dev, "pool");
  const char *nspace_name = udev_device_get_sysattr_value(dev, "pool_ns");
  const char *image_name = udev_device_get_sysattr_value(dev, "name");
  const char *snap_name = udev_device_get_sysattr_value(dev, "current_snap");

  if (!pool_name || !image_name || !snap_name)
    return std::nullopt;

  return std::make_optional<krbd_spec>(
      pool_name, nspace_name ?: "", image_name, snap_name);
}

static string get_kernel_rbd_name(const char *id)
{
  return string("/dev/rbd") + id;
}

static int sysfs_write_rbd(const char *which, const string& buf)
{
  const string s = string("/sys/bus/rbd/") + which;
  const string t = s + "_single_major";
  int fd;
  int r;

  /*
   * 'add' and 'add_single_major' interfaces are identical, but if rbd
   * kernel module is new enough and is configured to use single-major
   * scheme, 'add' is disabled in order to prevent old userspace from
   * doing weird things at unmap time.
   *
   * Same goes for 'remove' vs 'remove_single_major'.
   */
  fd = open(t.c_str(), O_WRONLY);
  if (fd < 0) {
    if (errno == ENOENT) {
      fd = open(s.c_str(), O_WRONLY);
      if (fd < 0)
        return -errno;
    } else {
      return -errno;
    }
  }

  r = safe_write(fd, buf.c_str(), buf.size());

  close(fd);
  return r;
}

static int sysfs_write_rbd_add(const string& buf)
{
  return sysfs_write_rbd("add", buf);
}

static int sysfs_write_rbd_remove(const string& buf)
{
  return sysfs_write_rbd("remove", buf);
}

static int have_minor_attr(void)
{
  /*
   * 'minor' attribute was added as part of single_major merge, which
   * exposed the 'single_major' parameter.  'minor' is always present,
   * regardless of whether single-major scheme is turned on or not.
   *
   * (Something like ver >= KERNEL_VERSION(3, 14, 0) is a no-go because
   * this has to work with rbd.ko backported to various kernels.)
   */
  return access("/sys/module/rbd/parameters/single_major", F_OK) == 0;
}

static int build_map_buf(CephContext *cct, const krbd_spec& spec,
                         const string& options, string *pbuf)
{
  bool msgr2 = false;
  ostringstream oss;
  int r;

  boost::char_separator<char> sep(",");
  boost::tokenizer<boost::char_separator<char>> tok(options, sep);
  for (const auto& t : tok) {
    if (boost::starts_with(t, "ms_mode=")) {
      /* msgr2 unless ms_mode=legacy */
      msgr2 = t.compare(8, t.npos, "legacy");
    }
  }

  MonMap monmap;
  r = monmap.build_initial(cct, false, cerr);
  if (r < 0)
    return r;

  /*
   * If msgr2, filter TYPE_MSGR2 addresses.  Otherwise, filter
   * TYPE_LEGACY addresses.
   */
  for (const auto& p : monmap.mon_info) {
    for (const auto& a : p.second.public_addrs.v) {
      if ((msgr2 && a.is_msgr2()) || (!msgr2 && a.is_legacy())) {
        if (oss.tellp() > 0) {
          oss << ",";
        }
        oss << a.get_sockaddr();
      }
    }
  }

  oss << " name=" << cct->_conf->name.get_id();

  KeyRing keyring;
  auto auth_client_required =
    cct->_conf.get_val<std::string>("auth_client_required");
  if (auth_client_required != "none") {
    r = keyring.from_ceph_context(cct);
    auto keyfile = cct->_conf.get_val<std::string>("keyfile");
    auto key = cct->_conf.get_val<std::string>("key");
    if (r == -ENOENT && keyfile.empty() && key.empty())
      r = 0;
    if (r < 0) {
      cerr << "rbd: failed to get secret" << std::endl;
      return r;
    }
  }

  CryptoKey secret;
  string key_name = string("client.") + cct->_conf->name.get_id();
  if (keyring.get_secret(cct->_conf->name, secret)) {
    string secret_str;
    secret.encode_base64(secret_str);

    r = set_kernel_secret(secret_str.c_str(), key_name.c_str());
    if (r >= 0) {
      if (r == 0)
        cerr << "rbd: warning: secret has length 0" << std::endl;
      oss << ",key=" << key_name;
    } else if (r == -ENODEV || r == -ENOSYS) {
      // running against older kernel; fall back to secret= in options
      oss << ",secret=" << secret_str;
    } else {
      cerr << "rbd: failed to add secret '" << key_name << "' to kernel"
           << std::endl;
      return r;
    }
  } else if (is_kernel_secret(key_name.c_str())) {
    oss << ",key=" << key_name;
  }

  if (!options.empty())
    oss << "," << options;
  if (!spec.nspace_name.empty())
    oss << ",_pool_ns=" << spec.nspace_name;

  oss << " " << spec.pool_name << " " << spec.image_name << " "
      << spec.snap_name;

  *pbuf = oss.str();
  return 0;
}

/*
 * Return:
 *   <kernel error, false> - didn't map
 *   <0 or udev error, true> - mapped
 */
template <typename F>
static std::pair<int, bool> wait_for_mapping(int sysfs_r_fd, udev_monitor *mon,
                                             F udev_device_handler)
{
  struct pollfd fds[2];
  int sysfs_r = INT_MAX, udev_r = INT_MAX;
  int r;

  fds[0].fd = sysfs_r_fd;
  fds[0].events = POLLIN;
  fds[1].fd = udev_monitor_get_fd(mon);
  fds[1].events = POLLIN;

  for (;;) {
    if (poll(fds, 2, -1) < 0) {
      ceph_abort_msgf("poll failed: %d", -errno);
    }

    if (fds[0].revents) {
      r = safe_read_exact(sysfs_r_fd, &sysfs_r, sizeof(sysfs_r));
      if (r < 0) {
        ceph_abort_msgf("safe_read_exact failed: %d", r);
      }
      if (sysfs_r < 0) {
        return std::make_pair(sysfs_r, false);
      }
      if (udev_r != INT_MAX) {
        ceph_assert(!sysfs_r);
        return std::make_pair(udev_r, true);
      }
      fds[0].fd = -1;
    }

    if (fds[1].revents) {
      for (;;) {
        struct udev_device *dev;

        dev = udev_monitor_receive_device(mon);
        if (!dev) {
          if (errno != EINTR && errno != EAGAIN) {
            udev_r = -errno;
            if (sysfs_r != INT_MAX) {
              ceph_assert(!sysfs_r);
              return std::make_pair(udev_r, true);
            }
            fds[1].fd = -1;
          }
          break;
        }
        if (udev_device_handler(dev)) {
          udev_r = 0;
          if (sysfs_r != INT_MAX) {
            ceph_assert(!sysfs_r);
            return std::make_pair(udev_r, true);
          }
          fds[1].fd = -1;
          break;
        }
      }
    }
  }
}

class UdevMapHandler {
public:
  UdevMapHandler(const krbd_spec *spec, std::string *pdevnode) :
      m_spec(spec), m_pdevnode(pdevnode) {}

  /*
   * Catch /sys/devices/rbd/<id>/ and wait for the corresponding
   * block device to show up.  This is necessary because rbd devices
   * and block devices aren't linked together in our sysfs layout.
   */
  bool operator()(udev_device *dev) {
    if (strcmp(udev_device_get_action(dev), "add")) {
      goto next;
    }
    if (!strcmp(udev_device_get_subsystem(dev), "rbd")) {
      if (!m_bus_dev) {
        auto spec = spec_from_dev(dev);
        if (spec && *spec == *m_spec) {
          m_bus_dev = dev;
          goto check;
        }
      }
    } else if (!strcmp(udev_device_get_subsystem(dev), "block")) {
      m_block_devs.push_back(dev);
      goto check;
    }

next:
    udev_device_unref(dev);
    return false;

check:
    if (m_bus_dev && !m_block_devs.empty()) {
      const char *major = udev_device_get_sysattr_value(m_bus_dev, "major");
      const char *minor = udev_device_get_sysattr_value(m_bus_dev, "minor");
      ceph_assert(!minor ^ have_minor_attr());

      for (auto p : m_block_devs) {
        const char *this_major = udev_device_get_property_value(p, "MAJOR");
        const char *this_minor = udev_device_get_property_value(p, "MINOR");

        if (strcmp(this_major, major) == 0 &&
            (!minor || strcmp(this_minor, minor) == 0)) {
          string name = get_kernel_rbd_name(udev_device_get_sysname(m_bus_dev));

          ceph_assert(strcmp(udev_device_get_devnode(p), name.c_str()) == 0);
          *m_pdevnode = name;
          return true;
        }
      }
    }
    return false;
  }

  ~UdevMapHandler() {
    if (m_bus_dev) {
      udev_device_unref(m_bus_dev);
    }

    for (auto p : m_block_devs) {
      udev_device_unref(p);
    }
  }

private:
  udev_device *m_bus_dev = nullptr;
  std::vector<udev_device *> m_block_devs;
  const krbd_spec *m_spec;
  std::string *m_pdevnode;
};

static const char *get_event_source(const krbd_ctx *ctx)
{
  if (ctx->flags & KRBD_CTX_F_NOUDEV) {
    /*
     * For block devices (unlike network interfaces, they don't
     * carry any namespace tags), the kernel broadcasts uevents
     * into all network namespaces that are owned by the initial
     * user namespace.  This restriction is new in 4.18: starting
     * with 2.6.35 and through 4.17 the kernel broadcast uevents
     * into all network namespaces, period.
     *
     * However, when invoked from a non-initial user namespace,
     * udev_monitor_receive_device() has always ignored both kernel
     * and udev uevents by virtue of requiring SCM_CREDENTIALS and
     * checking that ucred->uid == 0.  When UIDs and GIDs are sent to
     * a process in a user namespace, they are translated according
     * to that process's UID and GID mappings and, unless root in the
     * user namespace is mapped to the global root, that check fails.
     * Normally they show up as 65534(nobody) because the global root
     * is not mapped.
     */
    return "kernel";
  }

  /*
   * Like most netlink messages, udev uevents don't cross network
   * namespace boundaries and are therefore confined to the initial
   * network namespace.
   */
  return "udev";
}

static int do_map(krbd_ctx *ctx, const krbd_spec& spec, const string& buf,
                  string *pname)
{
  struct udev_monitor *mon;
  std::thread mapper;
  bool mapped;
  int fds[2];
  int r;

  mon = udev_monitor_new_from_netlink(ctx->udev, get_event_source(ctx));
  if (!mon)
    return -ENOMEM;

  r = udev_monitor_filter_add_match_subsystem_devtype(mon, "rbd", nullptr);
  if (r < 0)
    goto out_mon;

  r = udev_monitor_filter_add_match_subsystem_devtype(mon, "block", "disk");
  if (r < 0)
    goto out_mon;

  r = udev_monitor_set_receive_buffer_size(mon, UDEV_BUF_SIZE);
  if (r < 0) {
    std::cerr << "rbd: failed to set udev buffer size: " << cpp_strerror(r)
              << std::endl;
    /* not fatal */
  }

  r = udev_monitor_enable_receiving(mon);
  if (r < 0)
    goto out_mon;

  if (pipe2(fds, O_NONBLOCK) < 0) {
    r = -errno;
    goto out_mon;
  }

  mapper = make_named_thread("mapper", [&buf, sysfs_r_fd = fds[1]]() {
    int sysfs_r = sysfs_write_rbd_add(buf);
    int r = safe_write(sysfs_r_fd, &sysfs_r, sizeof(sysfs_r));
    if (r < 0) {
      ceph_abort_msgf("safe_write failed: %d", r);
    }
  });

  std::tie(r, mapped) = wait_for_mapping(fds[0], mon,
                                         UdevMapHandler(&spec, pname));
  if (r < 0) {
    if (!mapped) {
      std::cerr << "rbd: sysfs write failed" << std::endl;
    } else {
      std::cerr << "rbd: udev wait failed" << std::endl;
      /* TODO: fall back to enumeration */
    }
  }

  mapper.join();
  close(fds[0]);
  close(fds[1]);

out_mon:
  udev_monitor_unref(mon);
  return r;
}

static int map_image(struct krbd_ctx *ctx, const krbd_spec& spec,
                     const char *options, string *pname)
{
  string buf;
  int r;

  r = build_map_buf(ctx->cct, spec, options, &buf);
  if (r < 0)
    return r;

  /*
   * Modprobe rbd kernel module.  If it supports single-major device
   * number allocation scheme, make sure it's turned on.
   */
  if (access("/sys/bus/rbd", F_OK) != 0) {
    const char *module_options = NULL;
    if (module_has_param("rbd", "single_major"))
      module_options = "single_major=Y";

    r = module_load("rbd", module_options);
    if (r) {
      cerr << "rbd: failed to load rbd kernel module (" << r << ")"
           << std::endl;
      /*
       * Ignore the error: modprobe failing doesn't necessarily prevent
       * from working.
       */
    }
  }

  return do_map(ctx, spec, buf, pname);
}

static int devno_to_krbd_id(struct udev *udev, dev_t devno, string *pid)
{
  struct udev_enumerate *enm;
  struct udev_list_entry *l;
  struct udev_device *dev;
  int r;

retry:
  enm = udev_enumerate_new(udev);
  if (!enm)
    return -ENOMEM;

  r = udev_enumerate_add_match_subsystem(enm, "rbd");
  if (r < 0)
    goto out_enm;

  r = udev_enumerate_add_match_sysattr(enm, "major",
                                       stringify(major(devno)).c_str());
  if (r < 0)
    goto out_enm;

  if (have_minor_attr()) {
    r = udev_enumerate_add_match_sysattr(enm, "minor",
                                         stringify(minor(devno)).c_str());
    if (r < 0)
      goto out_enm;
  }

  r = udev_enumerate_scan_devices(enm);
  if (r < 0) {
    if (r == -ENOENT || r == -ENODEV) {
      std::cerr << "rbd: udev enumerate failed, retrying" << std::endl;
      udev_enumerate_unref(enm);
      goto retry;
    }
    goto out_enm;
  }

  l = udev_enumerate_get_list_entry(enm);
  if (!l) {
    r = -ENOENT;
    goto out_enm;
  }

  /* make sure there is only one match */
  ceph_assert(!udev_list_entry_get_next(l));

  dev = udev_device_new_from_syspath(udev, udev_list_entry_get_name(l));
  if (!dev) {
    r = -ENOMEM;
    goto out_enm;
  }

  *pid = udev_device_get_sysname(dev);

  udev_device_unref(dev);
out_enm:
  udev_enumerate_unref(enm);
  return r;
}

static int __enumerate_devices(struct udev *udev, const krbd_spec& spec,
                               bool match_nspace, struct udev_enumerate **penm)
{
  struct udev_enumerate *enm;
  int r;

retry:
  enm = udev_enumerate_new(udev);
  if (!enm)
    return -ENOMEM;

  r = udev_enumerate_add_match_subsystem(enm, "rbd");
  if (r < 0)
    goto out_enm;

  r = udev_enumerate_add_match_sysattr(enm, "pool", spec.pool_name.c_str());
  if (r < 0)
    goto out_enm;

  if (match_nspace) {
    r = udev_enumerate_add_match_sysattr(enm, "pool_ns",
                                         spec.nspace_name.c_str());
  } else {
    /*
     * Match _only_ devices that don't have pool_ns attribute.
     * If the kernel supports namespaces, the result will be empty.
     */
    r = udev_enumerate_add_nomatch_sysattr(enm, "pool_ns", nullptr);
  }
  if (r < 0)
    goto out_enm;

  r = udev_enumerate_add_match_sysattr(enm, "name", spec.image_name.c_str());
  if (r < 0)
    goto out_enm;

  r = udev_enumerate_add_match_sysattr(enm, "current_snap",
                                       spec.snap_name.c_str());
  if (r < 0)
    goto out_enm;

  r = udev_enumerate_scan_devices(enm);
  if (r < 0) {
    if (r == -ENOENT || r == -ENODEV) {
      std::cerr << "rbd: udev enumerate failed, retrying" << std::endl;
      udev_enumerate_unref(enm);
      goto retry;
    }
    goto out_enm;
  }

  *penm = enm;
  return 0;

out_enm:
  udev_enumerate_unref(enm);
  return r;
}

static int enumerate_devices(struct udev *udev, const krbd_spec& spec,
                             struct udev_enumerate **penm)
{
  struct udev_enumerate *enm;
  int r;

  r = __enumerate_devices(udev, spec, true, &enm);
  if (r < 0)
    return r;

  /*
   * If no namespace is set, try again with match_nspace=false to
   * handle older kernels.  On a newer kernel the result will remain
   * the same (i.e. empty).
   */
  if (!udev_enumerate_get_list_entry(enm) && spec.nspace_name.empty()) {
    udev_enumerate_unref(enm);
    r = __enumerate_devices(udev, spec, false, &enm);
    if (r < 0)
      return r;
  }

  *penm = enm;
  return 0;
}

static int spec_to_devno_and_krbd_id(struct udev *udev, const krbd_spec& spec,
                                     dev_t *pdevno, string *pid)
{
  struct udev_enumerate *enm;
  struct udev_list_entry *l;
  struct udev_device *dev;
  unsigned int maj, min = 0;
  string err;
  int r;

  r = enumerate_devices(udev, spec, &enm);
  if (r < 0)
    return r;

  l = udev_enumerate_get_list_entry(enm);
  if (!l) {
    r = -ENOENT;
    goto out_enm;
  }

  dev = udev_device_new_from_syspath(udev, udev_list_entry_get_name(l));
  if (!dev) {
    r = -ENOMEM;
    goto out_enm;
  }

  maj = strict_strtoll(udev_device_get_sysattr_value(dev, "major"), 10, &err);
  if (!err.empty()) {
    cerr << "rbd: couldn't parse major: " << err << std::endl;
    r = -EINVAL;
    goto out_dev;
  }
  if (have_minor_attr()) {
    min = strict_strtoll(udev_device_get_sysattr_value(dev, "minor"), 10, &err);
    if (!err.empty()) {
      cerr << "rbd: couldn't parse minor: " << err << std::endl;
      r = -EINVAL;
      goto out_dev;
    }
  }

  /*
   * If an image is mapped more than once don't bother trying to unmap
   * all devices - let users run unmap the same number of times they
   * ran map.
   */
  if (udev_list_entry_get_next(l))
    cerr << "rbd: " << spec << ": mapped more than once, unmapping "
         << get_kernel_rbd_name(udev_device_get_sysname(dev))
         << " only" << std::endl;

  *pdevno = makedev(maj, min);
  *pid = udev_device_get_sysname(dev);

out_dev:
  udev_device_unref(dev);
out_enm:
  udev_enumerate_unref(enm);
  return r;
}

static string build_unmap_buf(const string& id, const char *options)
{
  string buf(id);
  if (strcmp(options, "") != 0) {
    buf += " ";
    buf += options;
  }
  return buf;
}

class UdevUnmapHandler {
public:
  UdevUnmapHandler(dev_t devno) : m_devno(devno) {}

  bool operator()(udev_device *dev) {
    bool match = false;

    if (!strcmp(udev_device_get_action(dev), "remove") &&
        udev_device_get_devnum(dev) == m_devno) {
      match = true;
    }
    udev_device_unref(dev);
    return match;
  }

private:
  dev_t m_devno;
};

static int do_unmap(krbd_ctx *ctx, dev_t devno, const string& buf)
{
  struct udev_monitor *mon;
  std::thread unmapper;
  bool unmapped;
  int fds[2];
  int r;

  mon = udev_monitor_new_from_netlink(ctx->udev, get_event_source(ctx));
  if (!mon)
    return -ENOMEM;

  r = udev_monitor_filter_add_match_subsystem_devtype(mon, "block", "disk");
  if (r < 0)
    goto out_mon;

  r = udev_monitor_set_receive_buffer_size(mon, UDEV_BUF_SIZE);
  if (r < 0) {
    std::cerr << "rbd: failed to set udev buffer size: " << cpp_strerror(r)
              << std::endl;
    /* not fatal */
  }

  r = udev_monitor_enable_receiving(mon);
  if (r < 0)
    goto out_mon;

  if (pipe2(fds, O_NONBLOCK) < 0) {
    r = -errno;
    goto out_mon;
  }

  unmapper = make_named_thread(
      "unmapper", [&buf, sysfs_r_fd = fds[1], flags = ctx->flags]() {
    /*
     * On final device close(), kernel sends a block change event, in
     * response to which udev apparently runs blkid on the device.  This
     * makes unmap fail with EBUSY, if issued right after final close().
     * Try to circumvent this with a retry before turning to udev.
     */
    for (int tries = 0; ; tries++) {
      int sysfs_r = sysfs_write_rbd_remove(buf);
      if (sysfs_r == -EBUSY && tries < 2) {
        if (!tries) {
          usleep(250 * 1000);
        } else if (!(flags & KRBD_CTX_F_NOUDEV)) {
          /*
           * libudev does not provide the "wait until the queue is empty"
           * API or the sufficient amount of primitives to build it from.
           */
          std::string err = run_cmd("udevadm", "settle", "--timeout", "10",
                                    (char *)NULL);
          if (!err.empty())
            std::cerr << "rbd: " << err << std::endl;
        }
      } else {
        int r = safe_write(sysfs_r_fd, &sysfs_r, sizeof(sysfs_r));
        if (r < 0) {
          ceph_abort_msgf("safe_write failed: %d", r);
        }
        break;
      }
    }
  });

  std::tie(r, unmapped) = wait_for_mapping(fds[0], mon,
                                           UdevUnmapHandler(devno));
  if (r < 0) {
    if (!unmapped) {
      std::cerr << "rbd: sysfs write failed" << std::endl;
    } else {
      std::cerr << "rbd: udev wait failed: " << cpp_strerror(r) << std::endl;
      r = 0;
    }
  }

  unmapper.join();
  close(fds[0]);
  close(fds[1]);

out_mon:
  udev_monitor_unref(mon);
  return r;
}

static int unmap_image(struct krbd_ctx *ctx, const char *devnode,
                       const char *options)
{
  struct stat sb;
  dev_t wholedevno = 0;
  string id;
  int r;

  if (stat(devnode, &sb) < 0 || !S_ISBLK(sb.st_mode)) {
    cerr << "rbd: '" << devnode << "' is not a block device" << std::endl;
    return -EINVAL;
  }

  r = blkid_devno_to_wholedisk(sb.st_rdev, NULL, 0, &wholedevno);
  if (r < 0) {
    cerr << "rbd: couldn't compute wholedevno: " << cpp_strerror(r)
         << std::endl;
    /*
     * Ignore the error: we are given whole disks most of the time, and
     * if it turns out this is a partition we will fail later anyway.
     */
    wholedevno = sb.st_rdev;
  }

  for (int tries = 0; ; tries++) {
    r = devno_to_krbd_id(ctx->udev, wholedevno, &id);
    if (r == -ENOENT && tries < 2) {
      usleep(250 * 1000);
    } else {
      if (r < 0) {
        if (r == -ENOENT) {
          std::cerr << "rbd: '" << devnode << "' is not an rbd device"
                    << std::endl;
          r = -EINVAL;
        }
        return r;
      }
      if (tries) {
        std::cerr << "rbd: udev enumerate missed a device, tries = " << tries
                  << std::endl;
      }
      break;
    }
  }

  return do_unmap(ctx, wholedevno, build_unmap_buf(id, options));
}

static int unmap_image(struct krbd_ctx *ctx, const krbd_spec& spec,
                       const char *options)
{
  dev_t devno = 0;
  string id;
  int r;

  for (int tries = 0; ; tries++) {
    r = spec_to_devno_and_krbd_id(ctx->udev, spec, &devno, &id);
    if (r == -ENOENT && tries < 2) {
      usleep(250 * 1000);
    } else {
      if (r < 0) {
        if (r == -ENOENT) {
          std::cerr << "rbd: " << spec << ": not a mapped image or snapshot"
                    << std::endl;
          r = -EINVAL;
        }
        return r;
      }
      if (tries) {
        std::cerr << "rbd: udev enumerate missed a device, tries = " << tries
                  << std::endl;
      }
      break;
    }
  }

  return do_unmap(ctx, devno, build_unmap_buf(id, options));
}

static bool dump_one_image(Formatter *f, TextTable *tbl,
                           struct udev_device *dev)
{
  const char *id = udev_device_get_sysname(dev);
  auto spec = spec_from_dev(dev);
  string kname = get_kernel_rbd_name(id);

  if (!spec)
    return false;

  if (f) {
    f->open_object_section("device");
    f->dump_string("id", id);
    f->dump_string("pool", spec->pool_name);
    f->dump_string("namespace", spec->nspace_name);
    f->dump_string("name", spec->image_name);
    f->dump_string("snap", spec->snap_name);
    f->dump_string("device", kname);
    f->close_section();
  } else {
    *tbl << id << spec->pool_name << spec->nspace_name << spec->image_name
         << spec->snap_name << kname << TextTable::endrow;
  }

  return true;
}

static int do_dump(struct udev *udev, Formatter *f, TextTable *tbl)
{
  struct udev_enumerate *enm;
  struct udev_list_entry *l = NULL;
  bool have_output = false;
  int r;

retry:
  enm = udev_enumerate_new(udev);
  if (!enm)
    return -ENOMEM;

  r = udev_enumerate_add_match_subsystem(enm, "rbd");
  if (r < 0)
    goto out_enm;

  r = udev_enumerate_scan_devices(enm);
  if (r < 0) {
    if (r == -ENOENT || r == -ENODEV) {
      std::cerr << "rbd: udev enumerate failed, retrying" << std::endl;
      udev_enumerate_unref(enm);
      goto retry;
    }
    goto out_enm;
  }

  udev_list_entry_foreach(l, udev_enumerate_get_list_entry(enm)) {
    struct udev_device *dev;

    dev = udev_device_new_from_syspath(udev, udev_list_entry_get_name(l));
    if (dev) {
      have_output |= dump_one_image(f, tbl, dev);
      udev_device_unref(dev);
    }
  }

  r = have_output;
out_enm:
  udev_enumerate_unref(enm);
  return r;
}

int dump_images(struct krbd_ctx *ctx, Formatter *f)
{
  TextTable tbl;
  int r;

  if (f) {
    f->open_array_section("devices");
  } else {
    tbl.define_column("id", TextTable::LEFT, TextTable::LEFT);
    tbl.define_column("pool", TextTable::LEFT, TextTable::LEFT);
    tbl.define_column("namespace", TextTable::LEFT, TextTable::LEFT);
    tbl.define_column("image", TextTable::LEFT, TextTable::LEFT);
    tbl.define_column("snap", TextTable::LEFT, TextTable::LEFT);
    tbl.define_column("device", TextTable::LEFT, TextTable::LEFT);
  }

  r = do_dump(ctx->udev, f, &tbl);

  if (f) {
    f->close_section();
    f->flush(cout);
  } else {
    if (r > 0)
      cout << tbl;
  }

  return r;
}

static int is_mapped_image(struct udev *udev, const krbd_spec& spec,
                           string *pname)
{
  struct udev_enumerate *enm;
  struct udev_list_entry *l;
  int r;

  r = enumerate_devices(udev, spec, &enm);
  if (r < 0)
    return r;

  l = udev_enumerate_get_list_entry(enm);
  if (l) {
    struct udev_device *dev;

    dev = udev_device_new_from_syspath(udev, udev_list_entry_get_name(l));
    if (!dev) {
      r = -ENOMEM;
      goto out_enm;
    }

    r = 1;
    *pname = get_kernel_rbd_name(udev_device_get_sysname(dev));
    udev_device_unref(dev);
  } else {
    r = 0;  /* not mapped */
  }

out_enm:
  udev_enumerate_unref(enm);
  return r;
}

extern "C" int krbd_create_from_context(rados_config_t cct, uint32_t flags,
                                        struct krbd_ctx **pctx)
{
  struct krbd_ctx *ctx = new struct krbd_ctx();

  ctx->cct = reinterpret_cast<CephContext *>(cct);
  ctx->udev = udev_new();
  if (!ctx->udev) {
    delete ctx;
    return -ENOMEM;
  }
  ctx->flags = flags;

  *pctx = ctx;
  return 0;
}

extern "C" void krbd_destroy(struct krbd_ctx *ctx)
{
  if (!ctx)
    return;

  udev_unref(ctx->udev);

  delete ctx;
}

extern "C" int krbd_map(struct krbd_ctx *ctx,
                        const char *pool_name,
                        const char *nspace_name,
                        const char *image_name,
                        const char *snap_name,
                        const char *options,
                        char **pdevnode)
{
  krbd_spec spec(pool_name, nspace_name, image_name, snap_name);
  string name;
  char *devnode;
  int r;

  r = map_image(ctx, spec, options, &name);
  if (r < 0)
    return r;

  devnode = strdup(name.c_str());
  if (!devnode)
    return -ENOMEM;

  *pdevnode = devnode;
  return r;
}

extern "C" int krbd_unmap(struct krbd_ctx *ctx, const char *devnode,
                          const char *options)
{
  return unmap_image(ctx, devnode, options);
}

extern "C" int krbd_unmap_by_spec(struct krbd_ctx *ctx,
                                  const char *pool_name,
                                  const char *nspace_name,
                                  const char *image_name,
                                  const char *snap_name,
                                  const char *options)
{
  krbd_spec spec(pool_name, nspace_name, image_name, snap_name);
  return unmap_image(ctx, spec, options);
}

int krbd_showmapped(struct krbd_ctx *ctx, Formatter *f)
{
  return dump_images(ctx, f);
}

extern "C" int krbd_is_mapped(struct krbd_ctx *ctx,
                              const char *pool_name,
                              const char *nspace_name,
                              const char *image_name,
                              const char *snap_name,
                              char **pdevnode)
{
  krbd_spec spec(pool_name, nspace_name, image_name, snap_name);
  string name;
  char *devnode;
  int r;

  r = is_mapped_image(ctx->udev, spec, &name);
  if (r <= 0)  /* error or not mapped */
    return r;

  devnode = strdup(name.c_str());
  if (!devnode)
    return -ENOMEM;

  *pdevnode = devnode;
  return r;
}