ceph/src/osdc/Objecter.h

   1 // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
   2 // vim: ts=8 sw=2 smarttab
   3 /*
   4  * Ceph - scalable distributed file system
   5  *
   6  * Copyright (C) 2004-2006 Sage Weil <sage@newdream.net>
   7  *
   8  * This is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public
  10  * License version 2.1, as published by the Free Software
  11  * Foundation.  See file COPYING.
  12  *
  13  */
  14
  15 #ifndef CEPH_OBJECTER_H
  16 #define CEPH_OBJECTER_H
  17
  18 #include <condition_variable>
  19 #include <list>
  20 #include <map>
  21 #include <mutex>
  22 #include <memory>
  23 #include <sstream>
  24 #include <type_traits>
  25
  26 #include <boost/thread/shared_mutex.hpp>
  27
  28 #include "include/assert.h"
  29 #include "include/buffer.h"
  30 #include "include/types.h"
  31 #include "include/rados/rados_types.hpp"
  32
  33 #include "common/admin_socket.h"
  34 #include "common/ceph_time.h"
  35 #include "common/ceph_timer.h"
  36 #include "common/Finisher.h"
  37 #include "common/shunique_lock.h"
  38 #include "common/zipkin_trace.h"
  39
  40 #include "messages/MOSDOp.h"
  41 #include "osd/OSDMap.h"
  42
  43 using namespace std;
  44
  45 class Context;
  46 class Messenger;
  47 class OSDMap;
  48 class MonClient;
  49 class Message;
  50 class Finisher;
  51
  52 class MPoolOpReply;
  53
  54 class MGetPoolStatsReply;
  55 class MStatfsReply;
  56 class MCommandReply;
  57 class MWatchNotify;
  58
  59 class PerfCounters;
  60
  61 // -----------------------------------------
  62
  63 struct ObjectOperation {
  64   vector<OSDOp> ops;
  65   int flags;
  66   int priority;
  67
  68   vector<bufferlist*> out_bl;
  69   vector<Context*> out_handler;
  70   vector<int*> out_rval;
  71
  72   ObjectOperation() : flags(0), priority(0) {}
  73   ~ObjectOperation() {
  74     while (!out_handler.empty()) {
  75       delete out_handler.back();
  76       out_handler.pop_back();
  77     }
  78   }
  79
  80   size_t size() {
  81     return ops.size();
  82   }
  83
  84   void set_last_op_flags(int flags) {
  85     assert(!ops.empty());
  86     ops.rbegin()->op.flags = flags;
  87   }
  88
  89   class C_TwoContexts;
  90   /**
  91    * Add a callback to run when this operation completes,
  92    * after any other callbacks for it.
  93    */
  94   void add_handler(Context *extra);
  95
  96   OSDOp& add_op(int op) {
  97     int s = ops.size();
  98     ops.resize(s+1);
  99     ops[s].op.op = op;
 100     out_bl.resize(s+1);
 101     out_bl[s] = NULL;
 102     out_handler.resize(s+1);
 103     out_handler[s] = NULL;
 104     out_rval.resize(s+1);
 105     out_rval[s] = NULL;
 106     return ops[s];
 107   }
 108   void add_data(int op, uint64_t off, uint64_t len, bufferlist& bl) {
 109     OSDOp& osd_op = add_op(op);
 110     osd_op.op.extent.offset = off;
 111     osd_op.op.extent.length = len;
 112     osd_op.indata.claim_append(bl);
 113   }
 114   void add_writesame(int op, uint64_t off, uint64_t write_len,
 115                      bufferlist& bl) {
 116     OSDOp& osd_op = add_op(op);
 117     osd_op.op.writesame.offset = off;
 118     osd_op.op.writesame.length = write_len;
 119     osd_op.op.writesame.data_length = bl.length();
 120     osd_op.indata.claim_append(bl);
 121   }
 122   void add_xattr(int op, const char *name, const bufferlist& data) {
 123     OSDOp& osd_op = add_op(op);
 124     osd_op.op.xattr.name_len = (name ? strlen(name) : 0);
 125     osd_op.op.xattr.value_len = data.length();
 126     if (name)
 127       osd_op.indata.append(name);
 128     osd_op.indata.append(data);
 129   }
 130   void add_xattr_cmp(int op, const char *name, uint8_t cmp_op,
 131                      uint8_t cmp_mode, const bufferlist& data) {
 132     OSDOp& osd_op = add_op(op);
 133     osd_op.op.xattr.name_len = (name ? strlen(name) : 0);
 134     osd_op.op.xattr.value_len = data.length();
 135     osd_op.op.xattr.cmp_op = cmp_op;
 136     osd_op.op.xattr.cmp_mode = cmp_mode;
 137     if (name)
 138       osd_op.indata.append(name);
 139     osd_op.indata.append(data);
 140   }
 141   void add_call(int op, const char *cname, const char *method,
 142                 bufferlist &indata,
 143                 bufferlist *outbl, Context *ctx, int *prval) {
 144     OSDOp& osd_op = add_op(op);
 145
 146     unsigned p = ops.size() - 1;
 147     out_handler[p] = ctx;
 148     out_bl[p] = outbl;
 149     out_rval[p] = prval;
 150
 151     osd_op.op.cls.class_len = strlen(cname);
 152     osd_op.op.cls.method_len = strlen(method);
 153     osd_op.op.cls.indata_len = indata.length();
 154     osd_op.indata.append(cname, osd_op.op.cls.class_len);
 155     osd_op.indata.append(method, osd_op.op.cls.method_len);
 156     osd_op.indata.append(indata);
 157   }
 158   void add_pgls(int op, uint64_t count, collection_list_handle_t cookie,
 159                 epoch_t start_epoch) {
 160     OSDOp& osd_op = add_op(op);
 161     osd_op.op.pgls.count = count;
 162     osd_op.op.pgls.start_epoch = start_epoch;
 163     ::encode(cookie, osd_op.indata);
 164   }
 165   void add_pgls_filter(int op, uint64_t count, const bufferlist& filter,
 166                        collection_list_handle_t cookie, epoch_t start_epoch) {
 167     OSDOp& osd_op = add_op(op);
 168     osd_op.op.pgls.count = count;
 169     osd_op.op.pgls.start_epoch = start_epoch;
 170     string cname = "pg";
 171     string mname = "filter";
 172     ::encode(cname, osd_op.indata);
 173     ::encode(mname, osd_op.indata);
 174     osd_op.indata.append(filter);
 175     ::encode(cookie, osd_op.indata);
 176   }
 177   void add_alloc_hint(int op, uint64_t expected_object_size,
 178                       uint64_t expected_write_size,
 179                       uint32_t flags) {
 180     OSDOp& osd_op = add_op(op);
 181     osd_op.op.alloc_hint.expected_object_size = expected_object_size;
 182     osd_op.op.alloc_hint.expected_write_size = expected_write_size;
 183     osd_op.op.alloc_hint.flags = flags;
 184   }
 185
 186   // ------
 187
 188   // pg
 189   void pg_ls(uint64_t count, bufferlist& filter,
 190              collection_list_handle_t cookie, epoch_t start_epoch) {
 191     if (filter.length() == 0)
 192       add_pgls(CEPH_OSD_OP_PGLS, count, cookie, start_epoch);
 193     else
 194       add_pgls_filter(CEPH_OSD_OP_PGLS_FILTER, count, filter, cookie,
 195                       start_epoch);
 196     flags |= CEPH_OSD_FLAG_PGOP;
 197   }
 198
 199   void pg_nls(uint64_t count, const bufferlist& filter,
 200               collection_list_handle_t cookie, epoch_t start_epoch) {
 201     if (filter.length() == 0)
 202       add_pgls(CEPH_OSD_OP_PGNLS, count, cookie, start_epoch);
 203     else
 204       add_pgls_filter(CEPH_OSD_OP_PGNLS_FILTER, count, filter, cookie,
 205                       start_epoch);
 206     flags |= CEPH_OSD_FLAG_PGOP;
 207   }
 208
 209   void scrub_ls(const librados::object_id_t& start_after,
 210                 uint64_t max_to_get,
 211                 std::vector<librados::inconsistent_obj_t> *objects,
 212                 uint32_t *interval,
 213                 int *rval);
 214   void scrub_ls(const librados::object_id_t& start_after,
 215                 uint64_t max_to_get,
 216                 std::vector<librados::inconsistent_snapset_t> *objects,
 217                 uint32_t *interval,
 218                 int *rval);
 219
 220   void create(bool excl) {
 221     OSDOp& o = add_op(CEPH_OSD_OP_CREATE);
 222     o.op.flags = (excl ? CEPH_OSD_OP_FLAG_EXCL : 0);
 223   }
 224
 225   struct C_ObjectOperation_stat : public Context {
 226     bufferlist bl;
 227     uint64_t *psize;
 228     ceph::real_time *pmtime;
 229     time_t *ptime;
 230     struct timespec *pts;
 231     int *prval;
 232     C_ObjectOperation_stat(uint64_t *ps, ceph::real_time *pm, time_t *pt, struct timespec *_pts,
 233                            int *prval)
 234       : psize(ps), pmtime(pm), ptime(pt), pts(_pts), prval(prval) {}
 235     void finish(int r) override {
 236       if (r >= 0) {
 237         bufferlist::iterator p = bl.begin();
 238         try {
 239           uint64_t size;
 240           ceph::real_time mtime;
 241           ::decode(size, p);
 242           ::decode(mtime, p);
 243           if (psize)
 244             *psize = size;
 245           if (pmtime)
 246             *pmtime = mtime;
 247           if (ptime)
 248             *ptime = ceph::real_clock::to_time_t(mtime);
 249           if (pts)
 250             *pts = ceph::real_clock::to_timespec(mtime);
 251         } catch (buffer::error& e) {
 252           if (prval)
 253             *prval = -EIO;
 254         }
 255       }
 256     }
 257   };
 258   void stat(uint64_t *psize, ceph::real_time *pmtime, int *prval) {
 259     add_op(CEPH_OSD_OP_STAT);
 260     unsigned p = ops.size() - 1;
 261     C_ObjectOperation_stat *h = new C_ObjectOperation_stat(psize, pmtime, NULL, NULL,
 262                                                            prval);
 263     out_bl[p] = &h->bl;
 264     out_handler[p] = h;
 265     out_rval[p] = prval;
 266   }
 267   void stat(uint64_t *psize, time_t *ptime, int *prval) {
 268     add_op(CEPH_OSD_OP_STAT);
 269     unsigned p = ops.size() - 1;
 270     C_ObjectOperation_stat *h = new C_ObjectOperation_stat(psize, NULL, ptime, NULL,
 271                                                            prval);
 272     out_bl[p] = &h->bl;
 273     out_handler[p] = h;
 274     out_rval[p] = prval;
 275   }
 276   void stat(uint64_t *psize, struct timespec *pts, int *prval) {
 277     add_op(CEPH_OSD_OP_STAT);
 278     unsigned p = ops.size() - 1;
 279     C_ObjectOperation_stat *h = new C_ObjectOperation_stat(psize, NULL, NULL, pts,
 280                                                            prval);
 281     out_bl[p] = &h->bl;
 282     out_handler[p] = h;
 283     out_rval[p] = prval;
 284   }
 285   // object cmpext
 286   struct C_ObjectOperation_cmpext : public Context {
 287     int *prval;
 288     C_ObjectOperation_cmpext(int *prval)
 289       : prval(prval) {}
 290
 291     void finish(int r) {
 292       if (prval)
 293         *prval = r;
 294     }
 295   };
 296
 297   void cmpext(uint64_t off, bufferlist& cmp_bl, int *prval) {
 298     add_data(CEPH_OSD_OP_CMPEXT, off, cmp_bl.length(), cmp_bl);
 299     unsigned p = ops.size() - 1;
 300     C_ObjectOperation_cmpext *h = new C_ObjectOperation_cmpext(prval);
 301     out_handler[p] = h;
 302     out_rval[p] = prval;
 303   }
 304
 305   // Used by C API
 306   void cmpext(uint64_t off, uint64_t cmp_len, const char *cmp_buf, int *prval) {
 307     bufferlist cmp_bl;
 308     cmp_bl.append(cmp_buf, cmp_len);
 309     add_data(CEPH_OSD_OP_CMPEXT, off, cmp_len, cmp_bl);
 310     unsigned p = ops.size() - 1;
 311     C_ObjectOperation_cmpext *h = new C_ObjectOperation_cmpext(prval);
 312     out_handler[p] = h;
 313     out_rval[p] = prval;
 314   }
 315
 316   void read(uint64_t off, uint64_t len, bufferlist *pbl, int *prval,
 317             Context* ctx) {
 318     bufferlist bl;
 319     add_data(CEPH_OSD_OP_READ, off, len, bl);
 320     unsigned p = ops.size() - 1;
 321     out_bl[p] = pbl;
 322     out_rval[p] = prval;
 323     out_handler[p] = ctx;
 324   }
 325
 326   struct C_ObjectOperation_sparse_read : public Context {
 327     bufferlist bl;
 328     bufferlist *data_bl;
 329     std::map<uint64_t, uint64_t> *extents;
 330     int *prval;
 331     C_ObjectOperation_sparse_read(bufferlist *data_bl,
 332                                   std::map<uint64_t, uint64_t> *extents,
 333                                   int *prval)
 334       : data_bl(data_bl), extents(extents), prval(prval) {}
 335     void finish(int r) override {
 336       bufferlist::iterator iter = bl.begin();
 337       if (r >= 0) {
 338         try {
 339           ::decode(*extents, iter);
 340           ::decode(*data_bl, iter);
 341         } catch (buffer::error& e) {
 342           if (prval)
 343             *prval = -EIO;
 344         }
 345       }
 346     }
 347   };
 348   void sparse_read(uint64_t off, uint64_t len, std::map<uint64_t,uint64_t> *m,
 349                    bufferlist *data_bl, int *prval) {
 350     bufferlist bl;
 351     add_data(CEPH_OSD_OP_SPARSE_READ, off, len, bl);
 352     unsigned p = ops.size() - 1;
 353     C_ObjectOperation_sparse_read *h =
 354       new C_ObjectOperation_sparse_read(data_bl, m, prval);
 355     out_bl[p] = &h->bl;
 356     out_handler[p] = h;
 357     out_rval[p] = prval;
 358   }
 359   void write(uint64_t off, bufferlist& bl,
 360              uint64_t truncate_size,
 361              uint32_t truncate_seq) {
 362     add_data(CEPH_OSD_OP_WRITE, off, bl.length(), bl);
 363     OSDOp& o = *ops.rbegin();
 364     o.op.extent.truncate_size = truncate_size;
 365     o.op.extent.truncate_seq = truncate_seq;
 366   }
 367   void write(uint64_t off, bufferlist& bl) {
 368     write(off, bl, 0, 0);
 369   }
 370   void write_full(bufferlist& bl) {
 371     add_data(CEPH_OSD_OP_WRITEFULL, 0, bl.length(), bl);
 372   }
 373   void writesame(uint64_t off, uint64_t write_len, bufferlist& bl) {
 374     add_writesame(CEPH_OSD_OP_WRITESAME, off, write_len, bl);
 375   }
 376   void append(bufferlist& bl) {
 377     add_data(CEPH_OSD_OP_APPEND, 0, bl.length(), bl);
 378   }
 379   void zero(uint64_t off, uint64_t len) {
 380     bufferlist bl;
 381     add_data(CEPH_OSD_OP_ZERO, off, len, bl);
 382   }
 383   void truncate(uint64_t off) {
 384     bufferlist bl;
 385     add_data(CEPH_OSD_OP_TRUNCATE, off, 0, bl);
 386   }
 387   void remove() {
 388     bufferlist bl;
 389     add_data(CEPH_OSD_OP_DELETE, 0, 0, bl);
 390   }
 391   void mapext(uint64_t off, uint64_t len) {
 392     bufferlist bl;
 393     add_data(CEPH_OSD_OP_MAPEXT, off, len, bl);
 394   }
 395   void sparse_read(uint64_t off, uint64_t len) {
 396     bufferlist bl;
 397     add_data(CEPH_OSD_OP_SPARSE_READ, off, len, bl);
 398   }
 399
 400   void checksum(uint8_t type, const bufferlist &init_value_bl,
 401                 uint64_t off, uint64_t len, size_t chunk_size,
 402                 bufferlist *pbl, int *prval, Context *ctx) {
 403     OSDOp& osd_op = add_op(CEPH_OSD_OP_CHECKSUM);
 404     osd_op.op.checksum.offset = off;
 405     osd_op.op.checksum.length = len;
 406     osd_op.op.checksum.type = type;
 407     osd_op.op.checksum.chunk_size = chunk_size;
 408     osd_op.indata.append(init_value_bl);
 409
 410     unsigned p = ops.size() - 1;
 411     out_bl[p] = pbl;
 412     out_rval[p] = prval;
 413     out_handler[p] = ctx;
 414   }
 415
 416   // object attrs
 417   void getxattr(const char *name, bufferlist *pbl, int *prval) {
 418     bufferlist bl;
 419     add_xattr(CEPH_OSD_OP_GETXATTR, name, bl);
 420     unsigned p = ops.size() - 1;
 421     out_bl[p] = pbl;
 422     out_rval[p] = prval;
 423   }
 424   struct C_ObjectOperation_decodevals : public Context {
 425     uint64_t max_entries;
 426     bufferlist bl;
 427     std::map<std::string,bufferlist> *pattrs;
 428     bool *ptruncated;
 429     int *prval;
 430     C_ObjectOperation_decodevals(uint64_t m, std::map<std::string,bufferlist> *pa,
 431                                  bool *pt, int *pr)
 432       : max_entries(m), pattrs(pa), ptruncated(pt), prval(pr) {
 433       if (ptruncated) {
 434         *ptruncated = false;
 435       }
 436     }
 437     void finish(int r) override {
 438       if (r >= 0) {
 439         bufferlist::iterator p = bl.begin();
 440         try {
 441           if (pattrs)
 442             ::decode(*pattrs, p);
 443           if (ptruncated) {
 444             std::map<std::string,bufferlist> ignore;
 445             if (!pattrs) {
 446               ::decode(ignore, p);
 447               pattrs = &ignore;
 448             }
 449             if (!p.end()) {
 450               ::decode(*ptruncated, p);
 451             } else {
 452               // the OSD did not provide this.  since old OSDs do not
 453               // enfoce omap result limits either, we can infer it from
 454               // the size of the result
 455               *ptruncated = (pattrs->size() == max_entries);
 456             }
 457           }
 458         }
 459         catch (buffer::error& e) {
 460           if (prval)
 461             *prval = -EIO;
 462         }
 463       }
 464     }
 465   };
 466   struct C_ObjectOperation_decodekeys : public Context {
 467     uint64_t max_entries;
 468     bufferlist bl;
 469     std::set<std::string> *pattrs;
 470     bool *ptruncated;
 471     int *prval;
 472     C_ObjectOperation_decodekeys(uint64_t m, std::set<std::string> *pa, bool *pt,
 473                                  int *pr)
 474       : max_entries(m), pattrs(pa), ptruncated(pt), prval(pr) {
 475       if (ptruncated) {
 476         *ptruncated = false;
 477       }
 478     }
 479     void finish(int r) override {
 480       if (r >= 0) {
 481         bufferlist::iterator p = bl.begin();
 482         try {
 483           if (pattrs)
 484             ::decode(*pattrs, p);
 485           if (ptruncated) {
 486             std::set<std::string> ignore;
 487             if (!pattrs) {
 488               ::decode(ignore, p);
 489               pattrs = &ignore;
 490             }
 491             if (!p.end()) {
 492               ::decode(*ptruncated, p);
 493             } else {
 494               // the OSD did not provide this.  since old OSDs do not
 495               // enfoce omap result limits either, we can infer it from
 496               // the size of the result
 497               *ptruncated = (pattrs->size() == max_entries);
 498             }
 499           }
 500         }
 501         catch (buffer::error& e) {
 502           if (prval)
 503             *prval = -EIO;
 504         }
 505       }
 506     }
 507   };
 508   struct C_ObjectOperation_decodewatchers : public Context {
 509     bufferlist bl;
 510     list<obj_watch_t> *pwatchers;
 511     int *prval;
 512     C_ObjectOperation_decodewatchers(list<obj_watch_t> *pw, int *pr)
 513       : pwatchers(pw), prval(pr) {}
 514     void finish(int r) override {
 515       if (r >= 0) {
 516         bufferlist::iterator p = bl.begin();
 517         try {
 518           obj_list_watch_response_t resp;
 519           ::decode(resp, p);
 520           if (pwatchers) {
 521             for (list<watch_item_t>::iterator i = resp.entries.begin() ;
 522                  i != resp.entries.end() ; ++i) {
 523               obj_watch_t ow;
 524               ostringstream sa;
 525               sa << i->addr;
 526               strncpy(ow.addr, sa.str().c_str(), 256);
 527               ow.watcher_id = i->name.num();
 528               ow.cookie = i->cookie;
 529               ow.timeout_seconds = i->timeout_seconds;
 530               pwatchers->push_back(ow);
 531             }
 532           }
 533         }
 534         catch (buffer::error& e) {
 535           if (prval)
 536             *prval = -EIO;
 537         }
 538       }
 539     }
 540   };
 541   struct C_ObjectOperation_decodesnaps : public Context {
 542     bufferlist bl;
 543     librados::snap_set_t *psnaps;
 544     int *prval;
 545     C_ObjectOperation_decodesnaps(librados::snap_set_t *ps, int *pr)
 546       : psnaps(ps), prval(pr) {}
 547     void finish(int r) override {
 548       if (r >= 0) {
 549         bufferlist::iterator p = bl.begin();
 550         try {
 551           obj_list_snap_response_t resp;
 552           ::decode(resp, p);
 553           if (psnaps) {
 554             psnaps->clones.clear();
 555             for (vector<clone_info>::iterator ci = resp.clones.begin();
 556                  ci != resp.clones.end();
 557                  ++ci) {
 558               librados::clone_info_t clone;
 559
 560               clone.cloneid = ci->cloneid;
 561               clone.snaps.reserve(ci->snaps.size());
 562               clone.snaps.insert(clone.snaps.end(), ci->snaps.begin(),
 563                                  ci->snaps.end());
 564               clone.overlap = ci->overlap;
 565               clone.size = ci->size;
 566
 567               psnaps->clones.push_back(clone);
 568             }
 569             psnaps->seq = resp.seq;
 570           }
 571         } catch (buffer::error& e) {
 572           if (prval)
 573             *prval = -EIO;
 574         }
 575       }
 576     }
 577   };
 578   void getxattrs(std::map<std::string,bufferlist> *pattrs, int *prval) {
 579     add_op(CEPH_OSD_OP_GETXATTRS);
 580     if (pattrs || prval) {
 581       unsigned p = ops.size() - 1;
 582       C_ObjectOperation_decodevals *h
 583         = new C_ObjectOperation_decodevals(0, pattrs, nullptr, prval);
 584       out_handler[p] = h;
 585       out_bl[p] = &h->bl;
 586       out_rval[p] = prval;
 587     }
 588   }
 589   void setxattr(const char *name, const bufferlist& bl) {
 590     add_xattr(CEPH_OSD_OP_SETXATTR, name, bl);
 591   }
 592   void setxattr(const char *name, const string& s) {
 593     bufferlist bl;
 594     bl.append(s);
 595     add_xattr(CEPH_OSD_OP_SETXATTR, name, bl);
 596   }
 597   void cmpxattr(const char *name, uint8_t cmp_op, uint8_t cmp_mode,
 598                 const bufferlist& bl) {
 599     add_xattr_cmp(CEPH_OSD_OP_CMPXATTR, name, cmp_op, cmp_mode, bl);
 600   }
 601   void rmxattr(const char *name) {
 602     bufferlist bl;
 603     add_xattr(CEPH_OSD_OP_RMXATTR, name, bl);
 604   }
 605   void setxattrs(map<string, bufferlist>& attrs) {
 606     bufferlist bl;
 607     ::encode(attrs, bl);
 608     add_xattr(CEPH_OSD_OP_RESETXATTRS, 0, bl.length());
 609   }
 610   void resetxattrs(const char *prefix, map<string, bufferlist>& attrs) {
 611     bufferlist bl;
 612     ::encode(attrs, bl);
 613     add_xattr(CEPH_OSD_OP_RESETXATTRS, prefix, bl);
 614   }
 615
 616   // trivialmap
 617   void tmap_update(bufferlist& bl) {
 618     add_data(CEPH_OSD_OP_TMAPUP, 0, 0, bl);
 619   }
 620   void tmap_put(bufferlist& bl) {
 621     add_data(CEPH_OSD_OP_TMAPPUT, 0, bl.length(), bl);
 622   }
 623   void tmap_get(bufferlist *pbl, int *prval) {
 624     add_op(CEPH_OSD_OP_TMAPGET);
 625     unsigned p = ops.size() - 1;
 626     out_bl[p] = pbl;
 627     out_rval[p] = prval;
 628   }
 629   void tmap_get() {
 630     add_op(CEPH_OSD_OP_TMAPGET);
 631   }
 632   void tmap_to_omap(bool nullok=false) {
 633      OSDOp& osd_op = add_op(CEPH_OSD_OP_TMAP2OMAP);
 634      if (nullok)
 635        osd_op.op.tmap2omap.flags = CEPH_OSD_TMAP2OMAP_NULLOK;
 636   }
 637
 638   // objectmap
 639   void omap_get_keys(const string &start_after,
 640                      uint64_t max_to_get,
 641                      std::set<std::string> *out_set,
 642                      bool *ptruncated,
 643                      int *prval) {
 644     OSDOp &op = add_op(CEPH_OSD_OP_OMAPGETKEYS);
 645     bufferlist bl;
 646     ::encode(start_after, bl);
 647     ::encode(max_to_get, bl);
 648     op.op.extent.offset = 0;
 649     op.op.extent.length = bl.length();
 650     op.indata.claim_append(bl);
 651     if (prval || ptruncated || out_set) {
 652       unsigned p = ops.size() - 1;
 653       C_ObjectOperation_decodekeys *h =
 654         new C_ObjectOperation_decodekeys(max_to_get, out_set, ptruncated, prval);
 655       out_handler[p] = h;
 656       out_bl[p] = &h->bl;
 657       out_rval[p] = prval;
 658     }
 659   }
 660
 661   void omap_get_vals(const string &start_after,
 662                      const string &filter_prefix,
 663                      uint64_t max_to_get,
 664                      std::map<std::string, bufferlist> *out_set,
 665                      bool *ptruncated,
 666                      int *prval) {
 667     OSDOp &op = add_op(CEPH_OSD_OP_OMAPGETVALS);
 668     bufferlist bl;
 669     ::encode(start_after, bl);
 670     ::encode(max_to_get, bl);
 671     ::encode(filter_prefix, bl);
 672     op.op.extent.offset = 0;
 673     op.op.extent.length = bl.length();
 674     op.indata.claim_append(bl);
 675     if (prval || out_set || ptruncated) {
 676       unsigned p = ops.size() - 1;
 677       C_ObjectOperation_decodevals *h =
 678         new C_ObjectOperation_decodevals(max_to_get, out_set, ptruncated, prval);
 679       out_handler[p] = h;
 680       out_bl[p] = &h->bl;
 681       out_rval[p] = prval;
 682     }
 683   }
 684
 685   void omap_get_vals_by_keys(const std::set<std::string> &to_get,
 686                             std::map<std::string, bufferlist> *out_set,
 687                             int *prval) {
 688     OSDOp &op = add_op(CEPH_OSD_OP_OMAPGETVALSBYKEYS);
 689     bufferlist bl;
 690     ::encode(to_get, bl);
 691     op.op.extent.offset = 0;
 692     op.op.extent.length = bl.length();
 693     op.indata.claim_append(bl);
 694     if (prval || out_set) {
 695       unsigned p = ops.size() - 1;
 696       C_ObjectOperation_decodevals *h =
 697         new C_ObjectOperation_decodevals(0, out_set, nullptr, prval);
 698       out_handler[p] = h;
 699       out_bl[p] = &h->bl;
 700       out_rval[p] = prval;
 701     }
 702   }
 703
 704   void omap_cmp(const std::map<std::string, pair<bufferlist,int> > &assertions,
 705                 int *prval) {
 706     OSDOp &op = add_op(CEPH_OSD_OP_OMAP_CMP);
 707     bufferlist bl;
 708     ::encode(assertions, bl);
 709     op.op.extent.offset = 0;
 710     op.op.extent.length = bl.length();
 711     op.indata.claim_append(bl);
 712     if (prval) {
 713       unsigned p = ops.size() - 1;
 714       out_rval[p] = prval;
 715     }
 716   }
 717
 718   struct C_ObjectOperation_copyget : public Context {
 719     bufferlist bl;
 720     object_copy_cursor_t *cursor;
 721     uint64_t *out_size;
 722     ceph::real_time *out_mtime;
 723     std::map<std::string,bufferlist> *out_attrs;
 724     bufferlist *out_data, *out_omap_header, *out_omap_data;
 725     vector<snapid_t> *out_snaps;
 726     snapid_t *out_snap_seq;
 727     uint32_t *out_flags;
 728     uint32_t *out_data_digest;
 729     uint32_t *out_omap_digest;
 730     mempool::osd_pglog::vector<pair<osd_reqid_t, version_t> > *out_reqids;
 731     uint64_t *out_truncate_seq;
 732     uint64_t *out_truncate_size;
 733     int *prval;
 734     C_ObjectOperation_copyget(object_copy_cursor_t *c,
 735                               uint64_t *s,
 736                               ceph::real_time *m,
 737                               std::map<std::string,bufferlist> *a,
 738                               bufferlist *d, bufferlist *oh,
 739                               bufferlist *o,
 740                               std::vector<snapid_t> *osnaps,
 741                               snapid_t *osnap_seq,
 742                               uint32_t *flags,
 743                               uint32_t *dd,
 744                               uint32_t *od,
 745                               mempool::osd_pglog::vector<pair<osd_reqid_t, version_t> > *oreqids,
 746                               uint64_t *otseq,
 747                               uint64_t *otsize,
 748                               int *r)
 749       : cursor(c),
 750         out_size(s), out_mtime(m),
 751         out_attrs(a), out_data(d), out_omap_header(oh),
 752         out_omap_data(o), out_snaps(osnaps), out_snap_seq(osnap_seq),
 753         out_flags(flags), out_data_digest(dd), out_omap_digest(od),
 754         out_reqids(oreqids),
 755         out_truncate_seq(otseq),
 756         out_truncate_size(otsize),
 757         prval(r) {}
 758     void finish(int r) override {
 759       // reqids are copied on ENOENT
 760       if (r < 0 && r != -ENOENT)
 761         return;
 762       try {
 763         bufferlist::iterator p = bl.begin();
 764         object_copy_data_t copy_reply;
 765         ::decode(copy_reply, p);
 766         if (r == -ENOENT) {
 767           if (out_reqids)
 768             *out_reqids = copy_reply.reqids;
 769           return;
 770         }
 771         if (out_size)
 772           *out_size = copy_reply.size;
 773         if (out_mtime)
 774           *out_mtime = ceph::real_clock::from_ceph_timespec(copy_reply.mtime);
 775         if (out_attrs)
 776           *out_attrs = copy_reply.attrs;
 777         if (out_data)
 778           out_data->claim_append(copy_reply.data);
 779         if (out_omap_header)
 780           out_omap_header->claim_append(copy_reply.omap_header);
 781         if (out_omap_data)
 782           *out_omap_data = copy_reply.omap_data;
 783         if (out_snaps)
 784           *out_snaps = copy_reply.snaps;
 785         if (out_snap_seq)
 786           *out_snap_seq = copy_reply.snap_seq;
 787         if (out_flags)
 788           *out_flags = copy_reply.flags;
 789         if (out_data_digest)
 790           *out_data_digest = copy_reply.data_digest;
 791         if (out_omap_digest)
 792           *out_omap_digest = copy_reply.omap_digest;
 793         if (out_reqids)
 794           *out_reqids = copy_reply.reqids;
 795         if (out_truncate_seq)
 796           *out_truncate_seq = copy_reply.truncate_seq;
 797         if (out_truncate_size)
 798           *out_truncate_size = copy_reply.truncate_size;
 799         *cursor = copy_reply.cursor;
 800       } catch (buffer::error& e) {
 801         if (prval)
 802           *prval = -EIO;
 803       }
 804     }
 805   };
 806
 807   void copy_get(object_copy_cursor_t *cursor,
 808                 uint64_t max,
 809                 uint64_t *out_size,
 810                 ceph::real_time *out_mtime,
 811                 std::map<std::string,bufferlist> *out_attrs,
 812                 bufferlist *out_data,
 813                 bufferlist *out_omap_header,
 814                 bufferlist *out_omap_data,
 815                 vector<snapid_t> *out_snaps,
 816                 snapid_t *out_snap_seq,
 817                 uint32_t *out_flags,
 818                 uint32_t *out_data_digest,
 819                 uint32_t *out_omap_digest,
 820                 mempool::osd_pglog::vector<pair<osd_reqid_t, version_t> > *out_reqids,
 821                 uint64_t *truncate_seq,
 822                 uint64_t *truncate_size,
 823                 int *prval) {
 824     OSDOp& osd_op = add_op(CEPH_OSD_OP_COPY_GET);
 825     osd_op.op.copy_get.max = max;
 826     ::encode(*cursor, osd_op.indata);
 827     ::encode(max, osd_op.indata);
 828     unsigned p = ops.size() - 1;
 829     out_rval[p] = prval;
 830     C_ObjectOperation_copyget *h =
 831       new C_ObjectOperation_copyget(cursor, out_size, out_mtime,
 832                                     out_attrs, out_data, out_omap_header,
 833                                     out_omap_data, out_snaps, out_snap_seq,
 834                                     out_flags, out_data_digest,
 835                                     out_omap_digest, out_reqids, truncate_seq,
 836                                     truncate_size, prval);
 837     out_bl[p] = &h->bl;
 838     out_handler[p] = h;
 839   }
 840
 841   void undirty() {
 842     add_op(CEPH_OSD_OP_UNDIRTY);
 843   }
 844
 845   struct C_ObjectOperation_isdirty : public Context {
 846     bufferlist bl;
 847     bool *pisdirty;
 848     int *prval;
 849     C_ObjectOperation_isdirty(bool *p, int *r)
 850       : pisdirty(p), prval(r) {}
 851     void finish(int r) override {
 852       if (r < 0)
 853         return;
 854       try {
 855         bufferlist::iterator p = bl.begin();
 856         bool isdirty;
 857         ::decode(isdirty, p);
 858         if (pisdirty)
 859           *pisdirty = isdirty;
 860       } catch (buffer::error& e) {
 861         if (prval)
 862           *prval = -EIO;
 863       }
 864     }
 865   };
 866
 867   void is_dirty(bool *pisdirty, int *prval) {
 868     add_op(CEPH_OSD_OP_ISDIRTY);
 869     unsigned p = ops.size() - 1;
 870     out_rval[p] = prval;
 871     C_ObjectOperation_isdirty *h =
 872       new C_ObjectOperation_isdirty(pisdirty, prval);
 873     out_bl[p] = &h->bl;
 874     out_handler[p] = h;
 875   }
 876
 877   struct C_ObjectOperation_hit_set_ls : public Context {
 878     bufferlist bl;
 879     std::list< std::pair<time_t, time_t> > *ptls;
 880     std::list< std::pair<ceph::real_time, ceph::real_time> > *putls;
 881     int *prval;
 882     C_ObjectOperation_hit_set_ls(std::list< std::pair<time_t, time_t> > *t,
 883                                  std::list< std::pair<ceph::real_time,
 884                                                       ceph::real_time> > *ut,
 885                                  int *r)
 886       : ptls(t), putls(ut), prval(r) {}
 887     void finish(int r) override {
 888       if (r < 0)
 889         return;
 890       try {
 891         bufferlist::iterator p = bl.begin();
 892         std::list< std::pair<ceph::real_time, ceph::real_time> > ls;
 893         ::decode(ls, p);
 894         if (ptls) {
 895           ptls->clear();
 896           for (auto p = ls.begin(); p != ls.end(); ++p)
 897             // round initial timestamp up to the next full second to
 898             // keep this a valid interval.
 899             ptls->push_back(
 900               make_pair(ceph::real_clock::to_time_t(
 901                           ceph::ceil(p->first,
 902                                      // Sadly, no time literals until C++14.
 903                                      std::chrono::seconds(1))),
 904                         ceph::real_clock::to_time_t(p->second)));
 905         }
 906         if (putls)
 907           putls->swap(ls);
 908       } catch (buffer::error& e) {
 909         r = -EIO;
 910       }
 911       if (prval)
 912         *prval = r;
 913     }
 914   };
 915
 916   /**
 917    * list available HitSets.
 918    *
 919    * We will get back a list of time intervals.  Note that the most
 920    * recent range may have an empty end timestamp if it is still
 921    * accumulating.
 922    *
 923    * @param pls [out] list of time intervals
 924    * @param prval [out] return value
 925    */
 926   void hit_set_ls(std::list< std::pair<time_t, time_t> > *pls, int *prval) {
 927     add_op(CEPH_OSD_OP_PG_HITSET_LS);
 928     unsigned p = ops.size() - 1;
 929     out_rval[p] = prval;
 930     C_ObjectOperation_hit_set_ls *h =
 931       new C_ObjectOperation_hit_set_ls(pls, NULL, prval);
 932     out_bl[p] = &h->bl;
 933     out_handler[p] = h;
 934   }
 935   void hit_set_ls(std::list<std::pair<ceph::real_time, ceph::real_time> > *pls,
 936                   int *prval) {
 937     add_op(CEPH_OSD_OP_PG_HITSET_LS);
 938     unsigned p = ops.size() - 1;
 939     out_rval[p] = prval;
 940     C_ObjectOperation_hit_set_ls *h =
 941       new C_ObjectOperation_hit_set_ls(NULL, pls, prval);
 942     out_bl[p] = &h->bl;
 943     out_handler[p] = h;
 944   }
 945
 946   /**
 947    * get HitSet
 948    *
 949    * Return an encoded HitSet that includes the provided time
 950    * interval.
 951    *
 952    * @param stamp [in] timestamp
 953    * @param pbl [out] target buffer for encoded HitSet
 954    * @param prval [out] return value
 955    */
 956   void hit_set_get(ceph::real_time stamp, bufferlist *pbl, int *prval) {
 957     OSDOp& op = add_op(CEPH_OSD_OP_PG_HITSET_GET);
 958     op.op.hit_set_get.stamp = ceph::real_clock::to_ceph_timespec(stamp);
 959     unsigned p = ops.size() - 1;
 960     out_rval[p] = prval;
 961     out_bl[p] = pbl;
 962   }
 963
 964   void omap_get_header(bufferlist *bl, int *prval) {
 965     add_op(CEPH_OSD_OP_OMAPGETHEADER);
 966     unsigned p = ops.size() - 1;
 967     out_bl[p] = bl;
 968     out_rval[p] = prval;
 969   }
 970
 971   void omap_set(const map<string, bufferlist> &map) {
 972     bufferlist bl;
 973     ::encode(map, bl);
 974     add_data(CEPH_OSD_OP_OMAPSETVALS, 0, bl.length(), bl);
 975   }
 976
 977   void omap_set_header(bufferlist &bl) {
 978     add_data(CEPH_OSD_OP_OMAPSETHEADER, 0, bl.length(), bl);
 979   }
 980
 981   void omap_clear() {
 982     add_op(CEPH_OSD_OP_OMAPCLEAR);
 983   }
 984
 985   void omap_rm_keys(const std::set<std::string> &to_remove) {
 986     bufferlist bl;
 987     ::encode(to_remove, bl);
 988     add_data(CEPH_OSD_OP_OMAPRMKEYS, 0, bl.length(), bl);
 989   }
 990
 991   // object classes
 992   void call(const char *cname, const char *method, bufferlist &indata) {
 993     add_call(CEPH_OSD_OP_CALL, cname, method, indata, NULL, NULL, NULL);
 994   }
 995
 996   void call(const char *cname, const char *method, bufferlist &indata,
 997             bufferlist *outdata, Context *ctx, int *prval) {
 998     add_call(CEPH_OSD_OP_CALL, cname, method, indata, outdata, ctx, prval);
 999   }
1000
1001   // watch/notify
1002   void watch(uint64_t cookie, __u8 op, uint32_t timeout = 0) {
1003     OSDOp& osd_op = add_op(CEPH_OSD_OP_WATCH);
1004     osd_op.op.watch.cookie = cookie;
1005     osd_op.op.watch.op = op;
1006     osd_op.op.watch.timeout = timeout;
1007   }
1008
1009   void notify(uint64_t cookie, uint32_t prot_ver, uint32_t timeout,
1010               bufferlist &bl, bufferlist *inbl) {
1011     OSDOp& osd_op = add_op(CEPH_OSD_OP_NOTIFY);
1012     osd_op.op.notify.cookie = cookie;
1013     ::encode(prot_ver, *inbl);
1014     ::encode(timeout, *inbl);
1015     ::encode(bl, *inbl);
1016     osd_op.indata.append(*inbl);
1017   }
1018
1019   void notify_ack(uint64_t notify_id, uint64_t cookie,
1020                   bufferlist& reply_bl) {
1021     OSDOp& osd_op = add_op(CEPH_OSD_OP_NOTIFY_ACK);
1022     bufferlist bl;
1023     ::encode(notify_id, bl);
1024     ::encode(cookie, bl);
1025     ::encode(reply_bl, bl);
1026     osd_op.indata.append(bl);
1027   }
1028
1029   void list_watchers(list<obj_watch_t> *out,
1030                      int *prval) {
1031     (void)add_op(CEPH_OSD_OP_LIST_WATCHERS);
1032     if (prval || out) {
1033       unsigned p = ops.size() - 1;
1034       C_ObjectOperation_decodewatchers *h =
1035         new C_ObjectOperation_decodewatchers(out, prval);
1036       out_handler[p] = h;
1037       out_bl[p] = &h->bl;
1038       out_rval[p] = prval;
1039     }
1040   }
1041
1042   void list_snaps(librados::snap_set_t *out, int *prval) {
1043     (void)add_op(CEPH_OSD_OP_LIST_SNAPS);
1044     if (prval || out) {
1045       unsigned p = ops.size() - 1;
1046       C_ObjectOperation_decodesnaps *h =
1047         new C_ObjectOperation_decodesnaps(out, prval);
1048       out_handler[p] = h;
1049       out_bl[p] = &h->bl;
1050       out_rval[p] = prval;
1051     }
1052   }
1053
1054   void assert_version(uint64_t ver) {
1055     OSDOp& osd_op = add_op(CEPH_OSD_OP_ASSERT_VER);
1056     osd_op.op.assert_ver.ver = ver;
1057   }
1058
1059   void cmpxattr(const char *name, const bufferlist& val,
1060                 int op, int mode) {
1061     add_xattr(CEPH_OSD_OP_CMPXATTR, name, val);
1062     OSDOp& o = *ops.rbegin();
1063     o.op.xattr.cmp_op = op;
1064     o.op.xattr.cmp_mode = mode;
1065   }
1066
1067   void rollback(uint64_t snapid) {
1068     OSDOp& osd_op = add_op(CEPH_OSD_OP_ROLLBACK);
1069     osd_op.op.snap.snapid = snapid;
1070   }
1071
1072   void copy_from(object_t src, snapid_t snapid, object_locator_t src_oloc,
1073                  version_t src_version, unsigned flags,
1074                  unsigned src_fadvise_flags) {
1075     OSDOp& osd_op = add_op(CEPH_OSD_OP_COPY_FROM);
1076     osd_op.op.copy_from.snapid = snapid;
1077     osd_op.op.copy_from.src_version = src_version;
1078     osd_op.op.copy_from.flags = flags;
1079     osd_op.op.copy_from.src_fadvise_flags = src_fadvise_flags;
1080     ::encode(src, osd_op.indata);
1081     ::encode(src_oloc, osd_op.indata);
1082   }
1083
1084   /**
1085    * writeback content to backing tier
1086    *
1087    * If object is marked dirty in the cache tier, write back content
1088    * to backing tier. If the object is clean this is a no-op.
1089    *
1090    * If writeback races with an update, the update will block.
1091    *
1092    * use with IGNORE_CACHE to avoid triggering promote.
1093    */
1094   void cache_flush() {
1095     add_op(CEPH_OSD_OP_CACHE_FLUSH);
1096   }
1097
1098   /**
1099    * writeback content to backing tier
1100    *
1101    * If object is marked dirty in the cache tier, write back content
1102    * to backing tier. If the object is clean this is a no-op.
1103    *
1104    * If writeback races with an update, return EAGAIN.  Requires that
1105    * the SKIPRWLOCKS flag be set.
1106    *
1107    * use with IGNORE_CACHE to avoid triggering promote.
1108    */
1109   void cache_try_flush() {
1110     add_op(CEPH_OSD_OP_CACHE_TRY_FLUSH);
1111   }
1112
1113   /**
1114    * evict object from cache tier
1115    *
1116    * If object is marked clean, remove the object from the cache tier.
1117    * Otherwise, return EBUSY.
1118    *
1119    * use with IGNORE_CACHE to avoid triggering promote.
1120    */
1121   void cache_evict() {
1122     add_op(CEPH_OSD_OP_CACHE_EVICT);
1123   }
1124
1125   /*
1126    * Extensible tier
1127    */
1128   void set_redirect(object_t tgt, snapid_t snapid, object_locator_t tgt_oloc,
1129                     version_t tgt_version) {
1130     OSDOp& osd_op = add_op(CEPH_OSD_OP_SET_REDIRECT);
1131     osd_op.op.copy_from.snapid = snapid;
1132     osd_op.op.copy_from.src_version = tgt_version;
1133     ::encode(tgt, osd_op.indata);
1134     ::encode(tgt_oloc, osd_op.indata);
1135   }
1136
1137   void set_alloc_hint(uint64_t expected_object_size,
1138                       uint64_t expected_write_size,
1139                       uint32_t flags) {
1140     add_alloc_hint(CEPH_OSD_OP_SETALLOCHINT, expected_object_size,
1141                    expected_write_size, flags);
1142
1143     // CEPH_OSD_OP_SETALLOCHINT op is advisory and therefore deemed
1144     // not worth a feature bit.  Set FAILOK per-op flag to make
1145     // sure older osds don't trip over an unsupported opcode.
1146     set_last_op_flags(CEPH_OSD_OP_FLAG_FAILOK);
1147   }
1148
1149   void dup(vector<OSDOp>& sops) {
1150     ops = sops;
1151     out_bl.resize(sops.size());
1152     out_handler.resize(sops.size());
1153     out_rval.resize(sops.size());
1154     for (uint32_t i = 0; i < sops.size(); i++) {
1155       out_bl[i] = &sops[i].outdata;
1156       out_handler[i] = NULL;
1157       out_rval[i] = &sops[i].rval;
1158     }
1159   }
1160
1161   /**
1162    * Pin/unpin an object in cache tier
1163    */
1164   void cache_pin() {
1165     add_op(CEPH_OSD_OP_CACHE_PIN);
1166   }
1167
1168   void cache_unpin() {
1169     add_op(CEPH_OSD_OP_CACHE_UNPIN);
1170   }
1171 };
1172
1173
1174 // ----------------
1175
1176
1177 class Objecter : public md_config_obs_t, public Dispatcher {
1178 public:
1179   // config observer bits
1180   const char** get_tracked_conf_keys() const override;
1181   void handle_conf_change(const struct md_config_t *conf,
1182                           const std::set <std::string> &changed) override;
1183
1184 public:
1185   Messenger *messenger;
1186   MonClient *monc;
1187   Finisher *finisher;
1188   ZTracer::Endpoint trace_endpoint;
1189 private:
1190   OSDMap    *osdmap;
1191 public:
1192   using Dispatcher::cct;
1193   std::multimap<string,string> crush_location;
1194
1195   std::atomic<bool> initialized{false};
1196
1197 private:
1198   std::atomic<uint64_t> last_tid{0};
1199   std::atomic<unsigned> inflight_ops{0};
1200   std::atomic<int> client_inc{-1};
1201   uint64_t max_linger_id;
1202   std::atomic<unsigned> num_in_flight{0};
1203   std::atomic<int> global_op_flags{0}; // flags which are applied to each IO op
1204   bool keep_balanced_budget;
1205   bool honor_osdmap_full;
1206   bool osdmap_full_try;
1207
1208   // If this is true, accumulate a set of blacklisted entities
1209   // to be drained by consume_blacklist_events.
1210   bool blacklist_events_enabled;
1211   std::set<entity_addr_t> blacklist_events;
1212
1213 public:
1214   void maybe_request_map();
1215
1216   void enable_blacklist_events();
1217 private:
1218
1219   void _maybe_request_map();
1220
1221   version_t last_seen_osdmap_version;
1222   version_t last_seen_pgmap_version;
1223
1224   mutable boost::shared_mutex rwlock;
1225   using lock_guard = std::unique_lock<decltype(rwlock)>;
1226   using unique_lock = std::unique_lock<decltype(rwlock)>;
1227   using shared_lock = boost::shared_lock<decltype(rwlock)>;
1228   using shunique_lock = ceph::shunique_lock<decltype(rwlock)>;
1229   ceph::timer<ceph::mono_clock> timer;
1230
1231   PerfCounters *logger;
1232
1233   uint64_t tick_event;
1234
1235   void start_tick();
1236   void tick();
1237   void update_crush_location();
1238
1239   class RequestStateHook;
1240
1241   RequestStateHook *m_request_state_hook;
1242
1243 public:
1244   /*** track pending operations ***/
1245   // read
1246  public:
1247
1248   struct OSDSession;
1249
1250   struct op_target_t {
1251     int flags = 0;
1252
1253     epoch_t epoch = 0;  ///< latest epoch we calculated the mapping
1254
1255     object_t base_oid;
1256     object_locator_t base_oloc;
1257     object_t target_oid;
1258     object_locator_t target_oloc;
1259
1260     ///< true if we are directed at base_pgid, not base_oid
1261     bool precalc_pgid = false;
1262
1263     ///< true if we have ever mapped to a valid pool
1264     bool pool_ever_existed = false;
1265
1266     ///< explcit pg target, if any
1267     pg_t base_pgid;
1268
1269     pg_t pgid; ///< last (raw) pg we mapped to
1270     spg_t actual_pgid; ///< last (actual) spg_t we mapped to
1271     unsigned pg_num = 0; ///< last pg_num we mapped to
1272     unsigned pg_num_mask = 0; ///< last pg_num_mask we mapped to
1273     vector<int> up; ///< set of up osds for last pg we mapped to
1274     vector<int> acting; ///< set of acting osds for last pg we mapped to
1275     int up_primary = -1; ///< last up_primary we mapped to
1276     int acting_primary = -1;  ///< last acting_primary we mapped to
1277     int size = -1; ///< the size of the pool when were were last mapped
1278     int min_size = -1; ///< the min size of the pool when were were last mapped
1279     bool sort_bitwise = false; ///< whether the hobject_t sort order is bitwise
1280     bool recovery_deletes = false; ///< whether the deletes are performed during recovery instead of peering
1281
1282     bool used_replica = false;
1283     bool paused = false;
1284
1285     int osd = -1;      ///< the final target osd, or -1
1286
1287     epoch_t last_force_resend = 0;
1288
1289     op_target_t(object_t oid, object_locator_t oloc, int flags)
1290       : flags(flags),
1291         base_oid(oid),
1292         base_oloc(oloc)
1293       {}
1294
1295     op_target_t(pg_t pgid)
1296       : base_oloc(pgid.pool(), pgid.ps()),
1297         precalc_pgid(true),
1298         base_pgid(pgid)
1299       {}
1300
1301     op_target_t() = default;
1302
1303     hobject_t get_hobj() {
1304       return hobject_t(target_oid,
1305                        target_oloc.key,
1306                        CEPH_NOSNAP,
1307                        target_oloc.hash >= 0 ? target_oloc.hash : pgid.ps(),
1308                        target_oloc.pool,
1309                        target_oloc.nspace);
1310     }
1311
1312     bool contained_by(const hobject_t& begin, const hobject_t& end) {
1313       hobject_t h = get_hobj();
1314       int r = cmp(h, begin);
1315       return r == 0 || (r > 0 && h < end);
1316     }
1317
1318     void dump(Formatter *f) const;
1319   };
1320
1321   struct Op : public RefCountedObject {
1322     OSDSession *session;
1323     int incarnation;
1324
1325     op_target_t target;
1326
1327     ConnectionRef con;  // for rx buffer only
1328     uint64_t features;  // explicitly specified op features
1329
1330     vector<OSDOp> ops;
1331
1332     snapid_t snapid;
1333     SnapContext snapc;
1334     ceph::real_time mtime;
1335
1336     bufferlist *outbl;
1337     vector<bufferlist*> out_bl;
1338     vector<Context*> out_handler;
1339     vector<int*> out_rval;
1340
1341     int priority;
1342     Context *onfinish;
1343     uint64_t ontimeout;
1344
1345     ceph_tid_t tid;
1346     int attempts;
1347
1348     version_t *objver;
1349     epoch_t *reply_epoch;
1350
1351     ceph::mono_time stamp;
1352
1353     epoch_t map_dne_bound;
1354
1355     bool budgeted;
1356
1357     /// true if we should resend this message on failure
1358     bool should_resend;
1359
1360     /// true if the throttle budget is get/put on a series of OPs,
1361     /// instead of per OP basis, when this flag is set, the budget is
1362     /// acquired before sending the very first OP of the series and
1363     /// released upon receiving the last OP reply.
1364     bool ctx_budgeted;
1365
1366     int *data_offset;
1367
1368     osd_reqid_t reqid; // explicitly setting reqid
1369     ZTracer::Trace trace;
1370
1371     Op(const object_t& o, const object_locator_t& ol, vector<OSDOp>& op,
1372        int f, Context *fin, version_t *ov, int *offset = NULL,
1373        ZTracer::Trace *parent_trace = nullptr) :
1374       session(NULL), incarnation(0),
1375       target(o, ol, f),
1376       con(NULL),
1377       features(CEPH_FEATURES_SUPPORTED_DEFAULT),
1378       snapid(CEPH_NOSNAP),
1379       outbl(NULL),
1380       priority(0),
1381       onfinish(fin),
1382       ontimeout(0),
1383       tid(0),
1384       attempts(0),
1385       objver(ov),
1386       reply_epoch(NULL),
1387       map_dne_bound(0),
1388       budgeted(false),
1389       should_resend(true),
1390       ctx_budgeted(false),
1391       data_offset(offset) {
1392       ops.swap(op);
1393
1394       /* initialize out_* to match op vector */
1395       out_bl.resize(ops.size());
1396       out_rval.resize(ops.size());
1397       out_handler.resize(ops.size());
1398       for (unsigned i = 0; i < ops.size(); i++) {
1399         out_bl[i] = NULL;
1400         out_handler[i] = NULL;
1401         out_rval[i] = NULL;
1402       }
1403
1404       if (target.base_oloc.key == o)
1405         target.base_oloc.key.clear();
1406
1407       if (parent_trace && parent_trace->valid()) {
1408         trace.init("op", nullptr, parent_trace);
1409         trace.event("start");
1410       }
1411     }
1412
1413     bool operator<(const Op& other) const {
1414       return tid < other.tid;
1415     }
1416
1417     bool respects_full() const {
1418       return
1419         (target.flags & (CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_RWORDERED)) &&
1420         !(target.flags & (CEPH_OSD_FLAG_FULL_TRY | CEPH_OSD_FLAG_FULL_FORCE));
1421     }
1422
1423   private:
1424     ~Op() override {
1425       while (!out_handler.empty()) {
1426         delete out_handler.back();
1427         out_handler.pop_back();
1428       }
1429       trace.event("finish");
1430     }
1431   };
1432
1433   struct C_Op_Map_Latest : public Context {
1434     Objecter *objecter;
1435     ceph_tid_t tid;
1436     version_t latest;
1437     C_Op_Map_Latest(Objecter *o, ceph_tid_t t) : objecter(o), tid(t),
1438                                                  latest(0) {}
1439     void finish(int r) override;
1440   };
1441
1442   struct C_Command_Map_Latest : public Context {
1443     Objecter *objecter;
1444     uint64_t tid;
1445     version_t latest;
1446     C_Command_Map_Latest(Objecter *o, ceph_tid_t t) :  objecter(o), tid(t),
1447                                                        latest(0) {}
1448     void finish(int r) override;
1449   };
1450
1451   struct C_Stat : public Context {
1452     bufferlist bl;
1453     uint64_t *psize;
1454     ceph::real_time *pmtime;
1455     Context *fin;
1456     C_Stat(uint64_t *ps, ceph::real_time *pm, Context *c) :
1457       psize(ps), pmtime(pm), fin(c) {}
1458     void finish(int r) override {
1459       if (r >= 0) {
1460         bufferlist::iterator p = bl.begin();
1461         uint64_t s;
1462         ceph::real_time m;
1463         ::decode(s, p);
1464         ::decode(m, p);
1465         if (psize)
1466           *psize = s;
1467         if (pmtime)
1468           *pmtime = m;
1469       }
1470       fin->complete(r);
1471     }
1472   };
1473
1474   struct C_GetAttrs : public Context {
1475     bufferlist bl;
1476     map<string,bufferlist>& attrset;
1477     Context *fin;
1478     C_GetAttrs(map<string, bufferlist>& set, Context *c) : attrset(set),
1479                                                            fin(c) {}
1480     void finish(int r) override {
1481       if (r >= 0) {
1482         bufferlist::iterator p = bl.begin();
1483         ::decode(attrset, p);
1484       }
1485       fin->complete(r);
1486     }
1487   };
1488
1489
1490   // Pools and statistics
1491   struct NListContext {
1492     collection_list_handle_t pos;
1493
1494     // these are for !sortbitwise compat only
1495     int current_pg = 0;
1496     int starting_pg_num = 0;
1497     bool sort_bitwise = false;
1498
1499     bool at_end_of_pool = false; ///< publicly visible end flag
1500
1501     int64_t pool_id = -1;
1502     int pool_snap_seq = 0;
1503     uint64_t max_entries = 0;
1504     string nspace;
1505
1506     bufferlist bl;   // raw data read to here
1507     std::list<librados::ListObjectImpl> list;
1508
1509     bufferlist filter;
1510
1511     bufferlist extra_info;
1512
1513     // The budget associated with this context, once it is set (>= 0),
1514     // the budget is not get/released on OP basis, instead the budget
1515     // is acquired before sending the first OP and released upon receiving
1516     // the last op reply.
1517     int ctx_budget = -1;
1518
1519     bool at_end() const {
1520       return at_end_of_pool;
1521     }
1522
1523     uint32_t get_pg_hash_position() const {
1524       return pos.get_hash();
1525     }
1526   };
1527
1528   struct C_NList : public Context {
1529     NListContext *list_context;
1530     Context *final_finish;
1531     Objecter *objecter;
1532     epoch_t epoch;
1533     C_NList(NListContext *lc, Context * finish, Objecter *ob) :
1534       list_context(lc), final_finish(finish), objecter(ob), epoch(0) {}
1535     void finish(int r) override {
1536       if (r >= 0) {
1537         objecter->_nlist_reply(list_context, r, final_finish, epoch);
1538       } else {
1539         final_finish->complete(r);
1540       }
1541     }
1542   };
1543
1544   struct PoolStatOp {
1545     ceph_tid_t tid;
1546     list<string> pools;
1547
1548     map<string,pool_stat_t> *pool_stats;
1549     Context *onfinish;
1550     uint64_t ontimeout;
1551
1552     ceph::mono_time last_submit;
1553   };
1554
1555   struct StatfsOp {
1556     ceph_tid_t tid;
1557     struct ceph_statfs *stats;
1558     Context *onfinish;
1559     uint64_t ontimeout;
1560
1561     ceph::mono_time last_submit;
1562   };
1563
1564   struct PoolOp {
1565     ceph_tid_t tid;
1566     int64_t pool;
1567     string name;
1568     Context *onfinish;
1569     uint64_t ontimeout;
1570     int pool_op;
1571     uint64_t auid;
1572     int16_t crush_rule;
1573     snapid_t snapid;
1574     bufferlist *blp;
1575
1576     ceph::mono_time last_submit;
1577     PoolOp() : tid(0), pool(0), onfinish(NULL), ontimeout(0), pool_op(0),
1578                auid(0), crush_rule(0), snapid(0), blp(NULL) {}
1579   };
1580
1581   // -- osd commands --
1582   struct CommandOp : public RefCountedObject {
1583     OSDSession *session = nullptr;
1584     ceph_tid_t tid = 0;
1585     vector<string> cmd;
1586     bufferlist inbl;
1587     bufferlist *poutbl = nullptr;
1588     string *prs = nullptr;
1589
1590     // target_osd == -1 means target_pg is valid
1591     const int target_osd = -1;
1592     const pg_t target_pg;
1593
1594     op_target_t target;
1595
1596     epoch_t map_dne_bound = 0;
1597     int map_check_error = 0; // error to return if map check fails
1598     const char *map_check_error_str = nullptr;
1599
1600     Context *onfinish = nullptr;
1601     uint64_t ontimeout = 0;
1602     ceph::mono_time last_submit;
1603
1604     CommandOp(
1605       int target_osd,
1606       const vector<string> &cmd,
1607       bufferlist inbl,
1608       bufferlist *poutbl,
1609       string *prs,
1610       Context *onfinish)
1611       : cmd(cmd),
1612         inbl(inbl),
1613         poutbl(poutbl),
1614         prs(prs),
1615         target_osd(target_osd),
1616         onfinish(onfinish) {}
1617
1618     CommandOp(
1619       pg_t pgid,
1620       const vector<string> &cmd,
1621       bufferlist inbl,
1622       bufferlist *poutbl,
1623       string *prs,
1624       Context *onfinish)
1625       : cmd(cmd),
1626         inbl(inbl),
1627         poutbl(poutbl),
1628         prs(prs),
1629         target_pg(pgid),
1630         target(pgid),
1631         onfinish(onfinish) {}
1632
1633   };
1634
1635   void submit_command(CommandOp *c, ceph_tid_t *ptid);
1636   int _calc_command_target(CommandOp *c, shunique_lock &sul);
1637   void _assign_command_session(CommandOp *c, shunique_lock &sul);
1638   void _send_command(CommandOp *c);
1639   int command_op_cancel(OSDSession *s, ceph_tid_t tid, int r);
1640   void _finish_command(CommandOp *c, int r, string rs);
1641   void handle_command_reply(MCommandReply *m);
1642
1643
1644   // -- lingering ops --
1645
1646   struct WatchContext {
1647     // this simply mirrors librados WatchCtx2
1648     virtual void handle_notify(uint64_t notify_id,
1649                                uint64_t cookie,
1650                                uint64_t notifier_id,
1651                                bufferlist& bl) = 0;
1652     virtual void handle_error(uint64_t cookie, int err) = 0;
1653     virtual ~WatchContext() {}
1654   };
1655
1656   struct LingerOp : public RefCountedObject {
1657     uint64_t linger_id;
1658
1659     op_target_t target;
1660
1661     snapid_t snap;
1662     SnapContext snapc;
1663     ceph::real_time mtime;
1664
1665     vector<OSDOp> ops;
1666     bufferlist inbl;
1667     bufferlist *poutbl;
1668     version_t *pobjver;
1669
1670     bool is_watch;
1671     ceph::mono_time watch_valid_thru; ///< send time for last acked ping
1672     int last_error;  ///< error from last failed ping|reconnect, if any
1673     boost::shared_mutex watch_lock;
1674     using lock_guard = std::unique_lock<decltype(watch_lock)>;
1675     using unique_lock = std::unique_lock<decltype(watch_lock)>;
1676     using shared_lock = boost::shared_lock<decltype(watch_lock)>;
1677     using shunique_lock = ceph::shunique_lock<decltype(watch_lock)>;
1678
1679     // queue of pending async operations, with the timestamp of
1680     // when they were queued.
1681     list<ceph::mono_time> watch_pending_async;
1682
1683     uint32_t register_gen;
1684     bool registered;
1685     bool canceled;
1686     Context *on_reg_commit;
1687
1688     // we trigger these from an async finisher
1689     Context *on_notify_finish;
1690     bufferlist *notify_result_bl;
1691     uint64_t notify_id;
1692
1693     WatchContext *watch_context;
1694
1695     OSDSession *session;
1696
1697     ceph_tid_t register_tid;
1698     ceph_tid_t ping_tid;
1699     epoch_t map_dne_bound;
1700
1701     void _queued_async() {
1702       // watch_lock ust be locked unique
1703       watch_pending_async.push_back(ceph::mono_clock::now());
1704     }
1705     void finished_async() {
1706       unique_lock l(watch_lock);
1707       assert(!watch_pending_async.empty());
1708       watch_pending_async.pop_front();
1709     }
1710
1711     LingerOp() : linger_id(0),
1712                  target(object_t(), object_locator_t(), 0),
1713                  snap(CEPH_NOSNAP), poutbl(NULL), pobjver(NULL),
1714                  is_watch(false), last_error(0),
1715                  register_gen(0),
1716                  registered(false),
1717                  canceled(false),
1718                  on_reg_commit(NULL),
1719                  on_notify_finish(NULL),
1720                  notify_result_bl(NULL),
1721                  notify_id(0),
1722                  watch_context(NULL),
1723                  session(NULL),
1724                  register_tid(0),
1725                  ping_tid(0),
1726                  map_dne_bound(0) {}
1727
1728     // no copy!
1729     const LingerOp &operator=(const LingerOp& r);
1730     LingerOp(const LingerOp& o);
1731
1732     uint64_t get_cookie() {
1733       return reinterpret_cast<uint64_t>(this);
1734     }
1735
1736   private:
1737     ~LingerOp() override {
1738       delete watch_context;
1739     }
1740   };
1741
1742   struct C_Linger_Commit : public Context {
1743     Objecter *objecter;
1744     LingerOp *info;
1745     bufferlist outbl;  // used for notify only
1746     C_Linger_Commit(Objecter *o, LingerOp *l) : objecter(o), info(l) {
1747       info->get();
1748     }
1749     ~C_Linger_Commit() override {
1750       info->put();
1751     }
1752     void finish(int r) override {
1753       objecter->_linger_commit(info, r, outbl);
1754     }
1755   };
1756
1757   struct C_Linger_Reconnect : public Context {
1758     Objecter *objecter;
1759     LingerOp *info;
1760     C_Linger_Reconnect(Objecter *o, LingerOp *l) : objecter(o), info(l) {
1761       info->get();
1762     }
1763     ~C_Linger_Reconnect() override {
1764       info->put();
1765     }
1766     void finish(int r) override {
1767       objecter->_linger_reconnect(info, r);
1768     }
1769   };
1770
1771   struct C_Linger_Ping : public Context {
1772     Objecter *objecter;
1773     LingerOp *info;
1774     ceph::mono_time sent;
1775     uint32_t register_gen;
1776     C_Linger_Ping(Objecter *o, LingerOp *l)
1777       : objecter(o), info(l), register_gen(info->register_gen) {
1778       info->get();
1779     }
1780     ~C_Linger_Ping() override {
1781       info->put();
1782     }
1783     void finish(int r) override {
1784       objecter->_linger_ping(info, r, sent, register_gen);
1785     }
1786   };
1787
1788   struct C_Linger_Map_Latest : public Context {
1789     Objecter *objecter;
1790     uint64_t linger_id;
1791     version_t latest;
1792     C_Linger_Map_Latest(Objecter *o, uint64_t id) :
1793       objecter(o), linger_id(id), latest(0) {}
1794     void finish(int r) override;
1795   };
1796
1797   // -- osd sessions --
1798   struct OSDBackoff {
1799     spg_t pgid;
1800     uint64_t id;
1801     hobject_t begin, end;
1802   };
1803
1804   struct OSDSession : public RefCountedObject {
1805     boost::shared_mutex lock;
1806     using lock_guard = std::lock_guard<decltype(lock)>;
1807     using unique_lock = std::unique_lock<decltype(lock)>;
1808     using shared_lock = boost::shared_lock<decltype(lock)>;
1809     using shunique_lock = ceph::shunique_lock<decltype(lock)>;
1810
1811     // pending ops
1812     map<ceph_tid_t,Op*> ops;
1813     map<uint64_t, LingerOp*> linger_ops;
1814     map<ceph_tid_t,CommandOp*> command_ops;
1815
1816     // backoffs
1817     map<spg_t,map<hobject_t,OSDBackoff>> backoffs;
1818     map<uint64_t,OSDBackoff*> backoffs_by_id;
1819
1820     int osd;
1821     int incarnation;
1822     ConnectionRef con;
1823     int num_locks;
1824     std::unique_ptr<std::mutex[]> completion_locks;
1825     using unique_completion_lock = std::unique_lock<
1826       decltype(completion_locks)::element_type>;
1827
1828
1829     OSDSession(CephContext *cct, int o) :
1830       osd(o), incarnation(0), con(NULL),
1831       num_locks(cct->_conf->objecter_completion_locks_per_session),
1832       completion_locks(new std::mutex[num_locks]) {}
1833
1834     ~OSDSession() override;
1835
1836     bool is_homeless() { return (osd == -1); }
1837
1838     unique_completion_lock get_lock(object_t& oid);
1839   };
1840   map<int,OSDSession*> osd_sessions;
1841
1842   bool osdmap_full_flag() const;
1843   bool osdmap_pool_full(const int64_t pool_id) const;
1844
1845  private:
1846
1847   /**
1848    * Test pg_pool_t::FLAG_FULL on a pool
1849    *
1850    * @return true if the pool exists and has the flag set, or
1851    *         the global full flag is set, else false
1852    */
1853   bool _osdmap_pool_full(const int64_t pool_id) const;
1854   bool _osdmap_pool_full(const pg_pool_t &p) const;
1855   void update_pool_full_map(map<int64_t, bool>& pool_full_map);
1856
1857   map<uint64_t, LingerOp*> linger_ops;
1858   // we use this just to confirm a cookie is valid before dereferencing the ptr
1859   set<LingerOp*> linger_ops_set;
1860
1861   map<ceph_tid_t,PoolStatOp*> poolstat_ops;
1862   map<ceph_tid_t,StatfsOp*> statfs_ops;
1863   map<ceph_tid_t,PoolOp*> pool_ops;
1864   std::atomic<unsigned> num_homeless_ops{0};
1865
1866   OSDSession *homeless_session;
1867
1868   // ops waiting for an osdmap with a new pool or confirmation that
1869   // the pool does not exist (may be expanded to other uses later)
1870   map<uint64_t, LingerOp*> check_latest_map_lingers;
1871   map<ceph_tid_t, Op*> check_latest_map_ops;
1872   map<ceph_tid_t, CommandOp*> check_latest_map_commands;
1873
1874   map<epoch_t,list< pair<Context*, int> > > waiting_for_map;
1875
1876   ceph::timespan mon_timeout;
1877   ceph::timespan osd_timeout;
1878
1879   MOSDOp *_prepare_osd_op(Op *op);
1880   void _send_op(Op *op, MOSDOp *m = NULL);
1881   void _send_op_account(Op *op);
1882   void _cancel_linger_op(Op *op);
1883   void finish_op(OSDSession *session, ceph_tid_t tid);
1884   void _finish_op(Op *op, int r);
1885   static bool is_pg_changed(
1886     int oldprimary,
1887     const vector<int>& oldacting,
1888     int newprimary,
1889     const vector<int>& newacting,
1890     bool any_change=false);
1891   enum recalc_op_target_result {
1892     RECALC_OP_TARGET_NO_ACTION = 0,
1893     RECALC_OP_TARGET_NEED_RESEND,
1894     RECALC_OP_TARGET_POOL_DNE,
1895     RECALC_OP_TARGET_OSD_DNE,
1896     RECALC_OP_TARGET_OSD_DOWN,
1897   };
1898   bool _osdmap_full_flag() const;
1899   bool _osdmap_has_pool_full() const;
1900
1901   bool target_should_be_paused(op_target_t *op);
1902   int _calc_target(op_target_t *t, Connection *con,
1903                    bool any_change = false);
1904   int _map_session(op_target_t *op, OSDSession **s,
1905                    shunique_lock& lc);
1906
1907   void _session_op_assign(OSDSession *s, Op *op);
1908   void _session_op_remove(OSDSession *s, Op *op);
1909   void _session_linger_op_assign(OSDSession *to, LingerOp *op);
1910   void _session_linger_op_remove(OSDSession *from, LingerOp *op);
1911   void _session_command_op_assign(OSDSession *to, CommandOp *op);
1912   void _session_command_op_remove(OSDSession *from, CommandOp *op);
1913
1914   int _assign_op_target_session(Op *op, shunique_lock& lc,
1915                                 bool src_session_locked,
1916                                 bool dst_session_locked);
1917   int _recalc_linger_op_target(LingerOp *op, shunique_lock& lc);
1918
1919   void _linger_submit(LingerOp *info, shunique_lock& sul);
1920   void _send_linger(LingerOp *info, shunique_lock& sul);
1921   void _linger_commit(LingerOp *info, int r, bufferlist& outbl);
1922   void _linger_reconnect(LingerOp *info, int r);
1923   void _send_linger_ping(LingerOp *info);
1924   void _linger_ping(LingerOp *info, int r, ceph::mono_time sent,
1925                     uint32_t register_gen);
1926   int _normalize_watch_error(int r);
1927
1928   friend class C_DoWatchError;
1929 public:
1930   void linger_callback_flush(Context *ctx) {
1931     finisher->queue(ctx);
1932   }
1933
1934 private:
1935   void _check_op_pool_dne(Op *op, unique_lock *sl);
1936   void _send_op_map_check(Op *op);
1937   void _op_cancel_map_check(Op *op);
1938   void _check_linger_pool_dne(LingerOp *op, bool *need_unregister);
1939   void _send_linger_map_check(LingerOp *op);
1940   void _linger_cancel_map_check(LingerOp *op);
1941   void _check_command_map_dne(CommandOp *op);
1942   void _send_command_map_check(CommandOp *op);
1943   void _command_cancel_map_check(CommandOp *op);
1944
1945   void kick_requests(OSDSession *session);
1946   void _kick_requests(OSDSession *session, map<uint64_t, LingerOp *>& lresend);
1947   void _linger_ops_resend(map<uint64_t, LingerOp *>& lresend, unique_lock& ul);
1948
1949   int _get_session(int osd, OSDSession **session, shunique_lock& sul);
1950   void put_session(OSDSession *s);
1951   void get_session(OSDSession *s);
1952   void _reopen_session(OSDSession *session);
1953   void close_session(OSDSession *session);
1954
1955   void _nlist_reply(NListContext *list_context, int r, Context *final_finish,
1956                    epoch_t reply_epoch);
1957
1958   void resend_mon_ops();
1959
1960   /**
1961    * handle a budget for in-flight ops
1962    * budget is taken whenever an op goes into the ops map
1963    * and returned whenever an op is removed from the map
1964    * If throttle_op needs to throttle it will unlock client_lock.
1965    */
1966   int calc_op_budget(Op *op);
1967   void _throttle_op(Op *op, shunique_lock& sul, int op_size = 0);
1968   int _take_op_budget(Op *op, shunique_lock& sul) {
1969     assert(sul && sul.mutex() == &rwlock);
1970     int op_budget = calc_op_budget(op);
1971     if (keep_balanced_budget) {
1972       _throttle_op(op, sul, op_budget);
1973     } else {
1974       op_throttle_bytes.take(op_budget);
1975       op_throttle_ops.take(1);
1976     }
1977     op->budgeted = true;
1978     return op_budget;
1979   }
1980   void put_op_budget_bytes(int op_budget) {
1981     assert(op_budget >= 0);
1982     op_throttle_bytes.put(op_budget);
1983     op_throttle_ops.put(1);
1984   }
1985   void put_op_budget(Op *op) {
1986     assert(op->budgeted);
1987     int op_budget = calc_op_budget(op);
1988     put_op_budget_bytes(op_budget);
1989   }
1990   void put_nlist_context_budget(NListContext *list_context);
1991   Throttle op_throttle_bytes, op_throttle_ops;
1992
1993  public:
1994   Objecter(CephContext *cct_, Messenger *m, MonClient *mc,
1995            Finisher *fin,
1996            double mon_timeout,
1997            double osd_timeout) :
1998     Dispatcher(cct_), messenger(m), monc(mc), finisher(fin),
1999     trace_endpoint("0.0.0.0", 0, "Objecter"),
2000     osdmap(new OSDMap),
2001     max_linger_id(0),
2002     keep_balanced_budget(false), honor_osdmap_full(true), osdmap_full_try(false),
2003     blacklist_events_enabled(false),
2004     last_seen_osdmap_version(0), last_seen_pgmap_version(0),
2005     logger(NULL), tick_event(0), m_request_state_hook(NULL),
2006     homeless_session(new OSDSession(cct, -1)),
2007     mon_timeout(ceph::make_timespan(mon_timeout)),
2008     osd_timeout(ceph::make_timespan(osd_timeout)),
2009     op_throttle_bytes(cct, "objecter_bytes",
2010                       cct->_conf->objecter_inflight_op_bytes),
2011     op_throttle_ops(cct, "objecter_ops", cct->_conf->objecter_inflight_ops),
2012     epoch_barrier(0),
2013     retry_writes_after_first_reply(cct->_conf->objecter_retry_writes_after_first_reply)
2014   { }
2015   ~Objecter() override;
2016
2017   void init();
2018   void start(const OSDMap *o = nullptr);
2019   void shutdown();
2020
2021   // These two templates replace osdmap_(get)|(put)_read. Simply wrap
2022   // whatever functionality you want to use the OSDMap in a lambda like:
2023   //
2024   // with_osdmap([](const OSDMap& o) { o.do_stuff(); });
2025   //
2026   // or
2027   //
2028   // auto t = with_osdmap([&](const OSDMap& o) { return o.lookup_stuff(x); });
2029   //
2030   // Do not call into something that will try to lock the OSDMap from
2031   // here or you will have great woe and misery.
2032
2033   template<typename Callback, typename...Args>
2034   auto with_osdmap(Callback&& cb, Args&&... args) const ->
2035     decltype(cb(*osdmap, std::forward<Args>(args)...)) {
2036     shared_lock l(rwlock);
2037     return std::forward<Callback>(cb)(*osdmap, std::forward<Args>(args)...);
2038   }
2039
2040
2041   /**
2042    * Tell the objecter to throttle outgoing ops according to its
2043    * budget (in _conf). If you do this, ops can block, in
2044    * which case it will unlock client_lock and sleep until
2045    * incoming messages reduce the used budget low enough for
2046    * the ops to continue going; then it will lock client_lock again.
2047    */
2048   void set_balanced_budget() { keep_balanced_budget = true; }
2049   void unset_balanced_budget() { keep_balanced_budget = false; }
2050
2051   void set_honor_osdmap_full() { honor_osdmap_full = true; }
2052   void unset_honor_osdmap_full() { honor_osdmap_full = false; }
2053
2054   void set_osdmap_full_try() { osdmap_full_try = true; }
2055   void unset_osdmap_full_try() { osdmap_full_try = false; }
2056
2057   void _scan_requests(OSDSession *s,
2058                       bool force_resend,
2059                       bool cluster_full,
2060                       map<int64_t, bool> *pool_full_map,
2061                       map<ceph_tid_t, Op*>& need_resend,
2062                       list<LingerOp*>& need_resend_linger,
2063                       map<ceph_tid_t, CommandOp*>& need_resend_command,
2064                       shunique_lock& sul);
2065
2066   int64_t get_object_hash_position(int64_t pool, const string& key,
2067                                    const string& ns);
2068   int64_t get_object_pg_hash_position(int64_t pool, const string& key,
2069                                       const string& ns);
2070
2071   // messages
2072  public:
2073   bool ms_dispatch(Message *m) override;
2074   bool ms_can_fast_dispatch_any() const override {
2075     return true;
2076   }
2077   bool ms_can_fast_dispatch(const Message *m) const override {
2078     switch (m->get_type()) {
2079     case CEPH_MSG_OSD_OPREPLY:
2080     case CEPH_MSG_WATCH_NOTIFY:
2081       return true;
2082     default:
2083       return false;
2084     }
2085   }
2086   void ms_fast_dispatch(Message *m) override {
2087     if (!ms_dispatch(m)) {
2088       m->put();
2089     }
2090   }
2091
2092   void handle_osd_op_reply(class MOSDOpReply *m);
2093   void handle_osd_backoff(class MOSDBackoff *m);
2094   void handle_watch_notify(class MWatchNotify *m);
2095   void handle_osd_map(class MOSDMap *m);
2096   void wait_for_osd_map();
2097
2098   /**
2099    * Get list of entities blacklisted since this was last called,
2100    * and reset the list.
2101    *
2102    * Uses a std::set because typical use case is to compare some
2103    * other list of clients to see which overlap with the blacklisted
2104    * addrs.
2105    *
2106    */
2107   void consume_blacklist_events(std::set<entity_addr_t> *events);
2108
2109   int pool_snap_by_name(int64_t poolid,
2110                         const char *snap_name,
2111                         snapid_t *snap) const;
2112   int pool_snap_get_info(int64_t poolid, snapid_t snap,
2113                          pool_snap_info_t *info) const;
2114   int pool_snap_list(int64_t poolid, vector<uint64_t> *snaps);
2115 private:
2116
2117   void emit_blacklist_events(const OSDMap::Incremental &inc);
2118   void emit_blacklist_events(const OSDMap &old_osd_map,
2119                              const OSDMap &new_osd_map);
2120
2121   // low-level
2122   void _op_submit(Op *op, shunique_lock& lc, ceph_tid_t *ptid);
2123   void _op_submit_with_budget(Op *op, shunique_lock& lc,
2124                               ceph_tid_t *ptid,
2125                               int *ctx_budget = NULL);
2126   inline void unregister_op(Op *op);
2127
2128   // public interface
2129 public:
2130   void op_submit(Op *op, ceph_tid_t *ptid = NULL, int *ctx_budget = NULL);
2131   bool is_active() {
2132     shared_lock l(rwlock);
2133     return !((!inflight_ops) && linger_ops.empty() &&
2134              poolstat_ops.empty() && statfs_ops.empty());
2135   }
2136
2137   /**
2138    * Output in-flight requests
2139    */
2140   void _dump_active(OSDSession *s);
2141   void _dump_active();
2142   void dump_active();
2143   void dump_requests(Formatter *fmt);
2144   void _dump_ops(const OSDSession *s, Formatter *fmt);
2145   void dump_ops(Formatter *fmt);
2146   void _dump_linger_ops(const OSDSession *s, Formatter *fmt);
2147   void dump_linger_ops(Formatter *fmt);
2148   void _dump_command_ops(const OSDSession *s, Formatter *fmt);
2149   void dump_command_ops(Formatter *fmt);
2150   void dump_pool_ops(Formatter *fmt) const;
2151   void dump_pool_stat_ops(Formatter *fmt) const;
2152   void dump_statfs_ops(Formatter *fmt) const;
2153
2154   int get_client_incarnation() const { return client_inc; }
2155   void set_client_incarnation(int inc) { client_inc = inc; }
2156
2157   bool have_map(epoch_t epoch);
2158   /// wait for epoch; true if we already have it
2159   bool wait_for_map(epoch_t epoch, Context *c, int err=0);
2160   void _wait_for_new_map(Context *c, epoch_t epoch, int err=0);
2161   void wait_for_latest_osdmap(Context *fin);
2162   void get_latest_version(epoch_t oldest, epoch_t neweset, Context *fin);
2163   void _get_latest_version(epoch_t oldest, epoch_t neweset, Context *fin);
2164
2165   /** Get the current set of global op flags */
2166   int get_global_op_flags() const { return global_op_flags; }
2167   /** Add a flag to the global op flags, not really atomic operation */
2168   void add_global_op_flags(int flag) {
2169     global_op_flags.fetch_or(flag);
2170   }
2171   /** Clear the passed flags from the global op flag set */
2172   void clear_global_op_flag(int flags) {
2173     global_op_flags.fetch_and(~flags);
2174   }
2175
2176   /// cancel an in-progress request with the given return code
2177 private:
2178   int op_cancel(OSDSession *s, ceph_tid_t tid, int r);
2179   int _op_cancel(ceph_tid_t tid, int r);
2180 public:
2181   int op_cancel(ceph_tid_t tid, int r);
2182
2183   /**
2184    * Any write op which is in progress at the start of this call shall no
2185    * longer be in progress when this call ends.  Operations started after the
2186    * start of this call may still be in progress when this call ends.
2187    *
2188    * @return the latest possible epoch in which a cancelled op could have
2189    *         existed, or -1 if nothing was cancelled.
2190    */
2191   epoch_t op_cancel_writes(int r, int64_t pool=-1);
2192
2193   // commands
2194   void osd_command(int osd, const std::vector<string>& cmd,
2195                   const bufferlist& inbl, ceph_tid_t *ptid,
2196                   bufferlist *poutbl, string *prs, Context *onfinish) {
2197     assert(osd >= 0);
2198     CommandOp *c = new CommandOp(
2199       osd,
2200       cmd,
2201       inbl,
2202       poutbl,
2203       prs,
2204       onfinish);
2205     submit_command(c, ptid);
2206   }
2207   void pg_command(pg_t pgid, const vector<string>& cmd,
2208                  const bufferlist& inbl, ceph_tid_t *ptid,
2209                  bufferlist *poutbl, string *prs, Context *onfinish) {
2210     CommandOp *c = new CommandOp(
2211       pgid,
2212       cmd,
2213       inbl,
2214       poutbl,
2215       prs,
2216       onfinish);
2217     submit_command(c, ptid);
2218   }
2219
2220   // mid-level helpers
2221   Op *prepare_mutate_op(
2222     const object_t& oid, const object_locator_t& oloc,
2223     ObjectOperation& op, const SnapContext& snapc,
2224     ceph::real_time mtime, int flags,
2225     Context *oncommit, version_t *objver = NULL,
2226     osd_reqid_t reqid = osd_reqid_t(),
2227     ZTracer::Trace *parent_trace = nullptr) {
2228     Op *o = new Op(oid, oloc, op.ops, flags | global_op_flags |
2229                    CEPH_OSD_FLAG_WRITE, oncommit, objver, nullptr, parent_trace);
2230     o->priority = op.priority;
2231     o->mtime = mtime;
2232     o->snapc = snapc;
2233     o->out_rval.swap(op.out_rval);
2234     o->reqid = reqid;
2235     return o;
2236   }
2237   ceph_tid_t mutate(
2238     const object_t& oid, const object_locator_t& oloc,
2239     ObjectOperation& op, const SnapContext& snapc,
2240     ceph::real_time mtime, int flags,
2241     Context *oncommit, version_t *objver = NULL,
2242     osd_reqid_t reqid = osd_reqid_t()) {
2243     Op *o = prepare_mutate_op(oid, oloc, op, snapc, mtime, flags,
2244                               oncommit, objver, reqid);
2245     ceph_tid_t tid;
2246     op_submit(o, &tid);
2247     return tid;
2248   }
2249   Op *prepare_read_op(
2250     const object_t& oid, const object_locator_t& oloc,
2251     ObjectOperation& op,
2252     snapid_t snapid, bufferlist *pbl, int flags,
2253     Context *onack, version_t *objver = NULL,
2254     int *data_offset = NULL,
2255     uint64_t features = 0,
2256     ZTracer::Trace *parent_trace = nullptr) {
2257     Op *o = new Op(oid, oloc, op.ops, flags | global_op_flags |
2258                    CEPH_OSD_FLAG_READ, onack, objver, data_offset, parent_trace);
2259     o->priority = op.priority;
2260     o->snapid = snapid;
2261     o->outbl = pbl;
2262     if (!o->outbl && op.size() == 1 && op.out_bl[0]->length())
2263         o->outbl = op.out_bl[0];
2264     o->out_bl.swap(op.out_bl);
2265     o->out_handler.swap(op.out_handler);
2266     o->out_rval.swap(op.out_rval);
2267     return o;
2268   }
2269   ceph_tid_t read(
2270     const object_t& oid, const object_locator_t& oloc,
2271     ObjectOperation& op,
2272     snapid_t snapid, bufferlist *pbl, int flags,
2273     Context *onack, version_t *objver = NULL,
2274     int *data_offset = NULL,
2275     uint64_t features = 0) {
2276     Op *o = prepare_read_op(oid, oloc, op, snapid, pbl, flags, onack, objver,
2277                             data_offset);
2278     if (features)
2279       o->features = features;
2280     ceph_tid_t tid;
2281     op_submit(o, &tid);
2282     return tid;
2283   }
2284   Op *prepare_pg_read_op(
2285     uint32_t hash, object_locator_t oloc,
2286     ObjectOperation& op, bufferlist *pbl, int flags,
2287     Context *onack, epoch_t *reply_epoch,
2288     int *ctx_budget) {
2289     Op *o = new Op(object_t(), oloc,
2290                    op.ops,
2291                    flags | global_op_flags | CEPH_OSD_FLAG_READ |
2292                    CEPH_OSD_FLAG_IGNORE_OVERLAY,
2293                    onack, NULL);
2294     o->target.precalc_pgid = true;
2295     o->target.base_pgid = pg_t(hash, oloc.pool);
2296     o->priority = op.priority;
2297     o->snapid = CEPH_NOSNAP;
2298     o->outbl = pbl;
2299     o->out_bl.swap(op.out_bl);
2300     o->out_handler.swap(op.out_handler);
2301     o->out_rval.swap(op.out_rval);
2302     o->reply_epoch = reply_epoch;
2303     if (ctx_budget) {
2304       // budget is tracked by listing context
2305       o->ctx_budgeted = true;
2306     }
2307     return o;
2308   }
2309   ceph_tid_t pg_read(
2310     uint32_t hash, object_locator_t oloc,
2311     ObjectOperation& op, bufferlist *pbl, int flags,
2312     Context *onack, epoch_t *reply_epoch,
2313     int *ctx_budget) {
2314     Op *o = prepare_pg_read_op(hash, oloc, op, pbl, flags,
2315                                onack, reply_epoch, ctx_budget);
2316     ceph_tid_t tid;
2317     op_submit(o, &tid, ctx_budget);
2318     return tid;
2319   }
2320
2321   // caller owns a ref
2322   LingerOp *linger_register(const object_t& oid, const object_locator_t& oloc,
2323                             int flags);
2324   ceph_tid_t linger_watch(LingerOp *info,
2325                           ObjectOperation& op,
2326                           const SnapContext& snapc, ceph::real_time mtime,
2327                           bufferlist& inbl,
2328                           Context *onfinish,
2329                           version_t *objver);
2330   ceph_tid_t linger_notify(LingerOp *info,
2331                            ObjectOperation& op,
2332                            snapid_t snap, bufferlist& inbl,
2333                            bufferlist *poutbl,
2334                            Context *onack,
2335                            version_t *objver);
2336   int linger_check(LingerOp *info);
2337   void linger_cancel(LingerOp *info);  // releases a reference
2338   void _linger_cancel(LingerOp *info);
2339
2340   void _do_watch_notify(LingerOp *info, MWatchNotify *m);
2341
2342   /**
2343    * set up initial ops in the op vector, and allocate a final op slot.
2344    *
2345    * The caller is responsible for filling in the final ops_count ops.
2346    *
2347    * @param ops op vector
2348    * @param ops_count number of final ops the caller will fill in
2349    * @param extra_ops pointer to [array of] initial op[s]
2350    * @return index of final op (for caller to fill in)
2351    */
2352   int init_ops(vector<OSDOp>& ops, int ops_count, ObjectOperation *extra_ops) {
2353     int i;
2354     int extra = 0;
2355
2356     if (extra_ops)
2357       extra = extra_ops->ops.size();
2358
2359     ops.resize(ops_count + extra);
2360
2361     for (i=0; i<extra; i++) {
2362       ops[i] = extra_ops->ops[i];
2363     }
2364
2365     return i;
2366   }
2367
2368
2369   // high-level helpers
2370   Op *prepare_stat_op(
2371     const object_t& oid, const object_locator_t& oloc,
2372     snapid_t snap, uint64_t *psize, ceph::real_time *pmtime,
2373     int flags, Context *onfinish, version_t *objver = NULL,
2374     ObjectOperation *extra_ops = NULL) {
2375     vector<OSDOp> ops;
2376     int i = init_ops(ops, 1, extra_ops);
2377     ops[i].op.op = CEPH_OSD_OP_STAT;
2378     C_Stat *fin = new C_Stat(psize, pmtime, onfinish);
2379     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2380                    CEPH_OSD_FLAG_READ, fin, objver);
2381     o->snapid = snap;
2382     o->outbl = &fin->bl;
2383     return o;
2384   }
2385   ceph_tid_t stat(
2386     const object_t& oid, const object_locator_t& oloc,
2387     snapid_t snap, uint64_t *psize, ceph::real_time *pmtime,
2388     int flags, Context *onfinish, version_t *objver = NULL,
2389     ObjectOperation *extra_ops = NULL) {
2390     Op *o = prepare_stat_op(oid, oloc, snap, psize, pmtime, flags,
2391                             onfinish, objver, extra_ops);
2392     ceph_tid_t tid;
2393     op_submit(o, &tid);
2394     return tid;
2395   }
2396
2397   Op *prepare_read_op(
2398     const object_t& oid, const object_locator_t& oloc,
2399     uint64_t off, uint64_t len, snapid_t snap, bufferlist *pbl,
2400     int flags, Context *onfinish, version_t *objver = NULL,
2401     ObjectOperation *extra_ops = NULL, int op_flags = 0,
2402     ZTracer::Trace *parent_trace = nullptr) {
2403     vector<OSDOp> ops;
2404     int i = init_ops(ops, 1, extra_ops);
2405     ops[i].op.op = CEPH_OSD_OP_READ;
2406     ops[i].op.extent.offset = off;
2407     ops[i].op.extent.length = len;
2408     ops[i].op.extent.truncate_size = 0;
2409     ops[i].op.extent.truncate_seq = 0;
2410     ops[i].op.flags = op_flags;
2411     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2412                    CEPH_OSD_FLAG_READ, onfinish, objver, nullptr, parent_trace);
2413     o->snapid = snap;
2414     o->outbl = pbl;
2415     return o;
2416   }
2417   ceph_tid_t read(
2418     const object_t& oid, const object_locator_t& oloc,
2419     uint64_t off, uint64_t len, snapid_t snap, bufferlist *pbl,
2420     int flags, Context *onfinish, version_t *objver = NULL,
2421     ObjectOperation *extra_ops = NULL, int op_flags = 0) {
2422     Op *o = prepare_read_op(oid, oloc, off, len, snap, pbl, flags,
2423                             onfinish, objver, extra_ops, op_flags);
2424     ceph_tid_t tid;
2425     op_submit(o, &tid);
2426     return tid;
2427   }
2428
2429   Op *prepare_cmpext_op(
2430     const object_t& oid, const object_locator_t& oloc,
2431     uint64_t off, bufferlist &cmp_bl,
2432     snapid_t snap, int flags, Context *onfinish, version_t *objver = NULL,
2433     ObjectOperation *extra_ops = NULL, int op_flags = 0) {
2434     vector<OSDOp> ops;
2435     int i = init_ops(ops, 1, extra_ops);
2436     ops[i].op.op = CEPH_OSD_OP_CMPEXT;
2437     ops[i].op.extent.offset = off;
2438     ops[i].op.extent.length = cmp_bl.length();
2439     ops[i].op.extent.truncate_size = 0;
2440     ops[i].op.extent.truncate_seq = 0;
2441     ops[i].indata = cmp_bl;
2442     ops[i].op.flags = op_flags;
2443     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2444                    CEPH_OSD_FLAG_READ, onfinish, objver);
2445     o->snapid = snap;
2446     return o;
2447   }
2448
2449   ceph_tid_t cmpext(
2450     const object_t& oid, const object_locator_t& oloc,
2451     uint64_t off, bufferlist &cmp_bl,
2452     snapid_t snap, int flags, Context *onfinish, version_t *objver = NULL,
2453     ObjectOperation *extra_ops = NULL, int op_flags = 0) {
2454     Op *o = prepare_cmpext_op(oid, oloc, off, cmp_bl, snap,
2455                               flags, onfinish, objver, extra_ops, op_flags);
2456     ceph_tid_t tid;
2457     op_submit(o, &tid);
2458     return tid;
2459   }
2460
2461   ceph_tid_t read_trunc(const object_t& oid, const object_locator_t& oloc,
2462                         uint64_t off, uint64_t len, snapid_t snap,
2463                         bufferlist *pbl, int flags, uint64_t trunc_size,
2464                         __u32 trunc_seq, Context *onfinish,
2465                         version_t *objver = NULL,
2466                         ObjectOperation *extra_ops = NULL, int op_flags = 0) {
2467     vector<OSDOp> ops;
2468     int i = init_ops(ops, 1, extra_ops);
2469     ops[i].op.op = CEPH_OSD_OP_READ;
2470     ops[i].op.extent.offset = off;
2471     ops[i].op.extent.length = len;
2472     ops[i].op.extent.truncate_size = trunc_size;
2473     ops[i].op.extent.truncate_seq = trunc_seq;
2474     ops[i].op.flags = op_flags;
2475     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2476                    CEPH_OSD_FLAG_READ, onfinish, objver);
2477     o->snapid = snap;
2478     o->outbl = pbl;
2479     ceph_tid_t tid;
2480     op_submit(o, &tid);
2481     return tid;
2482   }
2483   ceph_tid_t mapext(const object_t& oid, const object_locator_t& oloc,
2484                     uint64_t off, uint64_t len, snapid_t snap, bufferlist *pbl,
2485                     int flags, Context *onfinish, version_t *objver = NULL,
2486                     ObjectOperation *extra_ops = NULL) {
2487     vector<OSDOp> ops;
2488     int i = init_ops(ops, 1, extra_ops);
2489     ops[i].op.op = CEPH_OSD_OP_MAPEXT;
2490     ops[i].op.extent.offset = off;
2491     ops[i].op.extent.length = len;
2492     ops[i].op.extent.truncate_size = 0;
2493     ops[i].op.extent.truncate_seq = 0;
2494     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2495                    CEPH_OSD_FLAG_READ, onfinish, objver);
2496     o->snapid = snap;
2497     o->outbl = pbl;
2498     ceph_tid_t tid;
2499     op_submit(o, &tid);
2500     return tid;
2501   }
2502   ceph_tid_t getxattr(const object_t& oid, const object_locator_t& oloc,
2503              const char *name, snapid_t snap, bufferlist *pbl, int flags,
2504              Context *onfinish,
2505              version_t *objver = NULL, ObjectOperation *extra_ops = NULL) {
2506     vector<OSDOp> ops;
2507     int i = init_ops(ops, 1, extra_ops);
2508     ops[i].op.op = CEPH_OSD_OP_GETXATTR;
2509     ops[i].op.xattr.name_len = (name ? strlen(name) : 0);
2510     ops[i].op.xattr.value_len = 0;
2511     if (name)
2512       ops[i].indata.append(name);
2513     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2514                    CEPH_OSD_FLAG_READ, onfinish, objver);
2515     o->snapid = snap;
2516     o->outbl = pbl;
2517     ceph_tid_t tid;
2518     op_submit(o, &tid);
2519     return tid;
2520   }
2521
2522   ceph_tid_t getxattrs(const object_t& oid, const object_locator_t& oloc,
2523                        snapid_t snap, map<string,bufferlist>& attrset,
2524                        int flags, Context *onfinish, version_t *objver = NULL,
2525                        ObjectOperation *extra_ops = NULL) {
2526     vector<OSDOp> ops;
2527     int i = init_ops(ops, 1, extra_ops);
2528     ops[i].op.op = CEPH_OSD_OP_GETXATTRS;
2529     C_GetAttrs *fin = new C_GetAttrs(attrset, onfinish);
2530     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2531                    CEPH_OSD_FLAG_READ, fin, objver);
2532     o->snapid = snap;
2533     o->outbl = &fin->bl;
2534     ceph_tid_t tid;
2535     op_submit(o, &tid);
2536     return tid;
2537   }
2538
2539   ceph_tid_t read_full(const object_t& oid, const object_locator_t& oloc,
2540                        snapid_t snap, bufferlist *pbl, int flags,
2541                        Context *onfinish, version_t *objver = NULL,
2542                        ObjectOperation *extra_ops = NULL) {
2543     return read(oid, oloc, 0, 0, snap, pbl, flags | global_op_flags |
2544                 CEPH_OSD_FLAG_READ, onfinish, objver, extra_ops);
2545   }
2546
2547
2548   // writes
2549   ceph_tid_t _modify(const object_t& oid, const object_locator_t& oloc,
2550                      vector<OSDOp>& ops, ceph::real_time mtime,
2551                      const SnapContext& snapc, int flags,
2552                      Context *oncommit,
2553                      version_t *objver = NULL) {
2554     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2555                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2556     o->mtime = mtime;
2557     o->snapc = snapc;
2558     ceph_tid_t tid;
2559     op_submit(o, &tid);
2560     return tid;
2561   }
2562   Op *prepare_write_op(
2563     const object_t& oid, const object_locator_t& oloc,
2564     uint64_t off, uint64_t len, const SnapContext& snapc,
2565     const bufferlist &bl, ceph::real_time mtime, int flags,
2566     Context *oncommit, version_t *objver = NULL,
2567     ObjectOperation *extra_ops = NULL, int op_flags = 0,
2568     ZTracer::Trace *parent_trace = nullptr) {
2569     vector<OSDOp> ops;
2570     int i = init_ops(ops, 1, extra_ops);
2571     ops[i].op.op = CEPH_OSD_OP_WRITE;
2572     ops[i].op.extent.offset = off;
2573     ops[i].op.extent.length = len;
2574     ops[i].op.extent.truncate_size = 0;
2575     ops[i].op.extent.truncate_seq = 0;
2576     ops[i].indata = bl;
2577     ops[i].op.flags = op_flags;
2578     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2579                    CEPH_OSD_FLAG_WRITE, oncommit, objver,
2580                    nullptr, parent_trace);
2581     o->mtime = mtime;
2582     o->snapc = snapc;
2583     return o;
2584   }
2585   ceph_tid_t write(
2586     const object_t& oid, const object_locator_t& oloc,
2587     uint64_t off, uint64_t len, const SnapContext& snapc,
2588     const bufferlist &bl, ceph::real_time mtime, int flags,
2589     Context *oncommit, version_t *objver = NULL,
2590     ObjectOperation *extra_ops = NULL, int op_flags = 0) {
2591     Op *o = prepare_write_op(oid, oloc, off, len, snapc, bl, mtime, flags,
2592                              oncommit, objver, extra_ops, op_flags);
2593     ceph_tid_t tid;
2594     op_submit(o, &tid);
2595     return tid;
2596   }
2597   Op *prepare_append_op(
2598     const object_t& oid, const object_locator_t& oloc,
2599     uint64_t len, const SnapContext& snapc,
2600     const bufferlist &bl, ceph::real_time mtime, int flags,
2601     Context *oncommit,
2602     version_t *objver = NULL,
2603     ObjectOperation *extra_ops = NULL) {
2604     vector<OSDOp> ops;
2605     int i = init_ops(ops, 1, extra_ops);
2606     ops[i].op.op = CEPH_OSD_OP_APPEND;
2607     ops[i].op.extent.offset = 0;
2608     ops[i].op.extent.length = len;
2609     ops[i].op.extent.truncate_size = 0;
2610     ops[i].op.extent.truncate_seq = 0;
2611     ops[i].indata = bl;
2612     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2613                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2614     o->mtime = mtime;
2615     o->snapc = snapc;
2616     return o;
2617   }
2618   ceph_tid_t append(
2619     const object_t& oid, const object_locator_t& oloc,
2620     uint64_t len, const SnapContext& snapc,
2621     const bufferlist &bl, ceph::real_time mtime, int flags,
2622     Context *oncommit,
2623     version_t *objver = NULL,
2624     ObjectOperation *extra_ops = NULL) {
2625     Op *o = prepare_append_op(oid, oloc, len, snapc, bl, mtime, flags,
2626                               oncommit, objver, extra_ops);
2627     ceph_tid_t tid;
2628     op_submit(o, &tid);
2629     return tid;
2630   }
2631   ceph_tid_t write_trunc(const object_t& oid, const object_locator_t& oloc,
2632                          uint64_t off, uint64_t len, const SnapContext& snapc,
2633                          const bufferlist &bl, ceph::real_time mtime, int flags,
2634                          uint64_t trunc_size, __u32 trunc_seq,
2635                          Context *oncommit,
2636                          version_t *objver = NULL,
2637                          ObjectOperation *extra_ops = NULL, int op_flags = 0) {
2638     vector<OSDOp> ops;
2639     int i = init_ops(ops, 1, extra_ops);
2640     ops[i].op.op = CEPH_OSD_OP_WRITE;
2641     ops[i].op.extent.offset = off;
2642     ops[i].op.extent.length = len;
2643     ops[i].op.extent.truncate_size = trunc_size;
2644     ops[i].op.extent.truncate_seq = trunc_seq;
2645     ops[i].indata = bl;
2646     ops[i].op.flags = op_flags;
2647     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2648                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2649     o->mtime = mtime;
2650     o->snapc = snapc;
2651     ceph_tid_t tid;
2652     op_submit(o, &tid);
2653     return tid;
2654   }
2655   Op *prepare_write_full_op(
2656     const object_t& oid, const object_locator_t& oloc,
2657     const SnapContext& snapc, const bufferlist &bl,
2658     ceph::real_time mtime, int flags,
2659     Context *oncommit, version_t *objver = NULL,
2660     ObjectOperation *extra_ops = NULL, int op_flags = 0) {
2661     vector<OSDOp> ops;
2662     int i = init_ops(ops, 1, extra_ops);
2663     ops[i].op.op = CEPH_OSD_OP_WRITEFULL;
2664     ops[i].op.extent.offset = 0;
2665     ops[i].op.extent.length = bl.length();
2666     ops[i].indata = bl;
2667     ops[i].op.flags = op_flags;
2668     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2669                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2670     o->mtime = mtime;
2671     o->snapc = snapc;
2672     return o;
2673   }
2674   ceph_tid_t write_full(
2675     const object_t& oid, const object_locator_t& oloc,
2676     const SnapContext& snapc, const bufferlist &bl,
2677     ceph::real_time mtime, int flags,
2678     Context *oncommit, version_t *objver = NULL,
2679     ObjectOperation *extra_ops = NULL, int op_flags = 0) {
2680     Op *o = prepare_write_full_op(oid, oloc, snapc, bl, mtime, flags,
2681                                   oncommit, objver, extra_ops, op_flags);
2682     ceph_tid_t tid;
2683     op_submit(o, &tid);
2684     return tid;
2685   }
2686   Op *prepare_writesame_op(
2687     const object_t& oid, const object_locator_t& oloc,
2688     uint64_t write_len, uint64_t off,
2689     const SnapContext& snapc, const bufferlist &bl,
2690     ceph::real_time mtime, int flags,
2691     Context *oncommit, version_t *objver = NULL,
2692     ObjectOperation *extra_ops = NULL, int op_flags = 0) {
2693
2694     vector<OSDOp> ops;
2695     int i = init_ops(ops, 1, extra_ops);
2696     ops[i].op.op = CEPH_OSD_OP_WRITESAME;
2697     ops[i].op.writesame.offset = off;
2698     ops[i].op.writesame.length = write_len;
2699     ops[i].op.writesame.data_length = bl.length();
2700     ops[i].indata = bl;
2701     ops[i].op.flags = op_flags;
2702     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2703                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2704     o->mtime = mtime;
2705     o->snapc = snapc;
2706     return o;
2707   }
2708   ceph_tid_t writesame(
2709     const object_t& oid, const object_locator_t& oloc,
2710     uint64_t write_len, uint64_t off,
2711     const SnapContext& snapc, const bufferlist &bl,
2712     ceph::real_time mtime, int flags,
2713     Context *oncommit, version_t *objver = NULL,
2714     ObjectOperation *extra_ops = NULL, int op_flags = 0) {
2715
2716     Op *o = prepare_writesame_op(oid, oloc, write_len, off, snapc, bl,
2717                                  mtime, flags, oncommit, objver,
2718                                  extra_ops, op_flags);
2719
2720     ceph_tid_t tid;
2721     op_submit(o, &tid);
2722     return tid;
2723   }
2724   ceph_tid_t trunc(const object_t& oid, const object_locator_t& oloc,
2725                    const SnapContext& snapc, ceph::real_time mtime, int flags,
2726                    uint64_t trunc_size, __u32 trunc_seq,
2727                    Context *oncommit, version_t *objver = NULL,
2728                    ObjectOperation *extra_ops = NULL) {
2729     vector<OSDOp> ops;
2730     int i = init_ops(ops, 1, extra_ops);
2731     ops[i].op.op = CEPH_OSD_OP_TRUNCATE;
2732     ops[i].op.extent.offset = trunc_size;
2733     ops[i].op.extent.truncate_size = trunc_size;
2734     ops[i].op.extent.truncate_seq = trunc_seq;
2735     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2736                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2737     o->mtime = mtime;
2738     o->snapc = snapc;
2739     ceph_tid_t tid;
2740     op_submit(o, &tid);
2741     return tid;
2742   }
2743   ceph_tid_t zero(const object_t& oid, const object_locator_t& oloc,
2744                   uint64_t off, uint64_t len, const SnapContext& snapc,
2745                   ceph::real_time mtime, int flags, Context *oncommit,
2746              version_t *objver = NULL, ObjectOperation *extra_ops = NULL) {
2747     vector<OSDOp> ops;
2748     int i = init_ops(ops, 1, extra_ops);
2749     ops[i].op.op = CEPH_OSD_OP_ZERO;
2750     ops[i].op.extent.offset = off;
2751     ops[i].op.extent.length = len;
2752     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2753                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2754     o->mtime = mtime;
2755     o->snapc = snapc;
2756     ceph_tid_t tid;
2757     op_submit(o, &tid);
2758     return tid;
2759   }
2760   ceph_tid_t rollback_object(const object_t& oid, const object_locator_t& oloc,
2761                              const SnapContext& snapc, snapid_t snapid,
2762                              ceph::real_time mtime, Context *oncommit,
2763                              version_t *objver = NULL,
2764                              ObjectOperation *extra_ops = NULL) {
2765     vector<OSDOp> ops;
2766     int i = init_ops(ops, 1, extra_ops);
2767     ops[i].op.op = CEPH_OSD_OP_ROLLBACK;
2768     ops[i].op.snap.snapid = snapid;
2769     Op *o = new Op(oid, oloc, ops, CEPH_OSD_FLAG_WRITE, oncommit, objver);
2770     o->mtime = mtime;
2771     o->snapc = snapc;
2772     ceph_tid_t tid;
2773     op_submit(o, &tid);
2774     return tid;
2775   }
2776   ceph_tid_t create(const object_t& oid, const object_locator_t& oloc,
2777                     const SnapContext& snapc, ceph::real_time mtime, int global_flags,
2778                     int create_flags, Context *oncommit,
2779                     version_t *objver = NULL,
2780                     ObjectOperation *extra_ops = NULL) {
2781     vector<OSDOp> ops;
2782     int i = init_ops(ops, 1, extra_ops);
2783     ops[i].op.op = CEPH_OSD_OP_CREATE;
2784     ops[i].op.flags = create_flags;
2785     Op *o = new Op(oid, oloc, ops, global_flags | global_op_flags |
2786                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2787     o->mtime = mtime;
2788     o->snapc = snapc;
2789     ceph_tid_t tid;
2790     op_submit(o, &tid);
2791     return tid;
2792   }
2793   Op *prepare_remove_op(
2794     const object_t& oid, const object_locator_t& oloc,
2795     const SnapContext& snapc, ceph::real_time mtime, int flags,
2796     Context *oncommit,
2797     version_t *objver = NULL, ObjectOperation *extra_ops = NULL) {
2798     vector<OSDOp> ops;
2799     int i = init_ops(ops, 1, extra_ops);
2800     ops[i].op.op = CEPH_OSD_OP_DELETE;
2801     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2802                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2803     o->mtime = mtime;
2804     o->snapc = snapc;
2805     return o;
2806   }
2807   ceph_tid_t remove(
2808     const object_t& oid, const object_locator_t& oloc,
2809     const SnapContext& snapc, ceph::real_time mtime, int flags,
2810     Context *oncommit,
2811     version_t *objver = NULL, ObjectOperation *extra_ops = NULL) {
2812     Op *o = prepare_remove_op(oid, oloc, snapc, mtime, flags,
2813                               oncommit, objver, extra_ops);
2814     ceph_tid_t tid;
2815     op_submit(o, &tid);
2816     return tid;
2817   }
2818
2819   ceph_tid_t setxattr(const object_t& oid, const object_locator_t& oloc,
2820               const char *name, const SnapContext& snapc, const bufferlist &bl,
2821               ceph::real_time mtime, int flags,
2822               Context *oncommit,
2823               version_t *objver = NULL, ObjectOperation *extra_ops = NULL) {
2824     vector<OSDOp> ops;
2825     int i = init_ops(ops, 1, extra_ops);
2826     ops[i].op.op = CEPH_OSD_OP_SETXATTR;
2827     ops[i].op.xattr.name_len = (name ? strlen(name) : 0);
2828     ops[i].op.xattr.value_len = bl.length();
2829     if (name)
2830       ops[i].indata.append(name);
2831     ops[i].indata.append(bl);
2832     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2833                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2834     o->mtime = mtime;
2835     o->snapc = snapc;
2836     ceph_tid_t tid;
2837     op_submit(o, &tid);
2838     return tid;
2839   }
2840   ceph_tid_t removexattr(const object_t& oid, const object_locator_t& oloc,
2841               const char *name, const SnapContext& snapc,
2842               ceph::real_time mtime, int flags,
2843               Context *oncommit,
2844               version_t *objver = NULL, ObjectOperation *extra_ops = NULL) {
2845     vector<OSDOp> ops;
2846     int i = init_ops(ops, 1, extra_ops);
2847     ops[i].op.op = CEPH_OSD_OP_RMXATTR;
2848     ops[i].op.xattr.name_len = (name ? strlen(name) : 0);
2849     ops[i].op.xattr.value_len = 0;
2850     if (name)
2851       ops[i].indata.append(name);
2852     Op *o = new Op(oid, oloc, ops, flags | global_op_flags |
2853                    CEPH_OSD_FLAG_WRITE, oncommit, objver);
2854     o->mtime = mtime;
2855     o->snapc = snapc;
2856     ceph_tid_t tid;
2857     op_submit(o, &tid);
2858     return tid;
2859   }
2860
2861   void list_nobjects(NListContext *p, Context *onfinish);
2862   uint32_t list_nobjects_seek(NListContext *p, uint32_t pos);
2863   uint32_t list_nobjects_seek(NListContext *list_context, const hobject_t& c);
2864   void list_nobjects_get_cursor(NListContext *list_context, hobject_t *c);
2865
2866   hobject_t enumerate_objects_begin();
2867   hobject_t enumerate_objects_end();
2868   //hobject_t enumerate_objects_begin(int n, int m);
2869   void enumerate_objects(
2870     int64_t pool_id,
2871     const std::string &ns,
2872     const hobject_t &start,
2873     const hobject_t &end,
2874     const uint32_t max,
2875     const bufferlist &filter_bl,
2876     std::list<librados::ListObjectImpl> *result,
2877     hobject_t *next,
2878     Context *on_finish);
2879
2880   void _enumerate_reply(
2881       bufferlist &bl,
2882       int r,
2883       const hobject_t &end,
2884       const int64_t pool_id,
2885       int budget,
2886       epoch_t reply_epoch,
2887       std::list<librados::ListObjectImpl> *result,
2888       hobject_t *next,
2889       Context *on_finish);
2890   friend class C_EnumerateReply;
2891
2892   // -------------------------
2893   // pool ops
2894 private:
2895   void pool_op_submit(PoolOp *op);
2896   void _pool_op_submit(PoolOp *op);
2897   void _finish_pool_op(PoolOp *op, int r);
2898   void _do_delete_pool(int64_t pool, Context *onfinish);
2899 public:
2900   int create_pool_snap(int64_t pool, string& snapName, Context *onfinish);
2901   int allocate_selfmanaged_snap(int64_t pool, snapid_t *psnapid,
2902                                 Context *onfinish);
2903   int delete_pool_snap(int64_t pool, string& snapName, Context *onfinish);
2904   int delete_selfmanaged_snap(int64_t pool, snapid_t snap, Context *onfinish);
2905
2906   int create_pool(string& name, Context *onfinish, uint64_t auid=0,
2907                   int crush_rule=-1);
2908   int delete_pool(int64_t pool, Context *onfinish);
2909   int delete_pool(const string& name, Context *onfinish);
2910   int change_pool_auid(int64_t pool, Context *onfinish, uint64_t auid);
2911
2912   void handle_pool_op_reply(MPoolOpReply *m);
2913   int pool_op_cancel(ceph_tid_t tid, int r);
2914
2915   // --------------------------
2916   // pool stats
2917 private:
2918   void _poolstat_submit(PoolStatOp *op);
2919 public:
2920   void handle_get_pool_stats_reply(MGetPoolStatsReply *m);
2921   void get_pool_stats(list<string>& pools, map<string,pool_stat_t> *result,
2922                       Context *onfinish);
2923   int pool_stat_op_cancel(ceph_tid_t tid, int r);
2924   void _finish_pool_stat_op(PoolStatOp *op, int r);
2925
2926   // ---------------------------
2927   // df stats
2928 private:
2929   void _fs_stats_submit(StatfsOp *op);
2930 public:
2931   void handle_fs_stats_reply(MStatfsReply *m);
2932   void get_fs_stats(struct ceph_statfs& result, Context *onfinish);
2933   int statfs_op_cancel(ceph_tid_t tid, int r);
2934   void _finish_statfs_op(StatfsOp *op, int r);
2935
2936   // ---------------------------
2937   // some scatter/gather hackery
2938
2939   void _sg_read_finish(vector<ObjectExtent>& extents,
2940                        vector<bufferlist>& resultbl,
2941                        bufferlist *bl, Context *onfinish);
2942
2943   struct C_SGRead : public Context {
2944     Objecter *objecter;
2945     vector<ObjectExtent> extents;
2946     vector<bufferlist> resultbl;
2947     bufferlist *bl;
2948     Context *onfinish;
2949     C_SGRead(Objecter *ob,
2950              vector<ObjectExtent>& e, vector<bufferlist>& r, bufferlist *b,
2951              Context *c) :
2952       objecter(ob), bl(b), onfinish(c) {
2953       extents.swap(e);
2954       resultbl.swap(r);
2955     }
2956     void finish(int r) override {
2957       objecter->_sg_read_finish(extents, resultbl, bl, onfinish);
2958     }
2959   };
2960
2961   void sg_read_trunc(vector<ObjectExtent>& extents, snapid_t snap,
2962                      bufferlist *bl, int flags, uint64_t trunc_size,
2963                      __u32 trunc_seq, Context *onfinish, int op_flags = 0) {
2964     if (extents.size() == 1) {
2965       read_trunc(extents[0].oid, extents[0].oloc, extents[0].offset,
2966                  extents[0].length, snap, bl, flags, extents[0].truncate_size,
2967                  trunc_seq, onfinish, 0, 0, op_flags);
2968     } else {
2969       C_GatherBuilder gather(cct);
2970       vector<bufferlist> resultbl(extents.size());
2971       int i=0;
2972       for (vector<ObjectExtent>::iterator p = extents.begin();
2973            p != extents.end();
2974            ++p) {
2975         read_trunc(p->oid, p->oloc, p->offset, p->length, snap, &resultbl[i++],
2976                    flags, p->truncate_size, trunc_seq, gather.new_sub(),
2977                    0, 0, op_flags);
2978       }
2979       gather.set_finisher(new C_SGRead(this, extents, resultbl, bl, onfinish));
2980       gather.activate();
2981     }
2982   }
2983
2984   void sg_read(vector<ObjectExtent>& extents, snapid_t snap, bufferlist *bl,
2985                int flags, Context *onfinish, int op_flags = 0) {
2986     sg_read_trunc(extents, snap, bl, flags, 0, 0, onfinish, op_flags);
2987   }
2988
2989   void sg_write_trunc(vector<ObjectExtent>& extents, const SnapContext& snapc,
2990                       const bufferlist& bl, ceph::real_time mtime, int flags,
2991                       uint64_t trunc_size, __u32 trunc_seq,
2992                       Context *oncommit, int op_flags = 0) {
2993     if (extents.size() == 1) {
2994       write_trunc(extents[0].oid, extents[0].oloc, extents[0].offset,
2995                   extents[0].length, snapc, bl, mtime, flags,
2996                   extents[0].truncate_size, trunc_seq, oncommit,
2997                   0, 0, op_flags);
2998     } else {
2999       C_GatherBuilder gcom(cct, oncommit);
3000       for (vector<ObjectExtent>::iterator p = extents.begin();
3001            p != extents.end();
3002            ++p) {
3003         bufferlist cur;
3004         for (vector<pair<uint64_t,uint64_t> >::iterator bit
3005                = p->buffer_extents.begin();
3006              bit != p->buffer_extents.end();
3007              ++bit)
3008           bl.copy(bit->first, bit->second, cur);
3009         assert(cur.length() == p->length);
3010         write_trunc(p->oid, p->oloc, p->offset, p->length,
3011               snapc, cur, mtime, flags, p->truncate_size, trunc_seq,
3012               oncommit ? gcom.new_sub():0,
3013               0, 0, op_flags);
3014       }
3015       gcom.activate();
3016     }
3017   }
3018
3019   void sg_write(vector<ObjectExtent>& extents, const SnapContext& snapc,
3020                 const bufferlist& bl, ceph::real_time mtime, int flags,
3021                 Context *oncommit, int op_flags = 0) {
3022     sg_write_trunc(extents, snapc, bl, mtime, flags, 0, 0, oncommit,
3023                    op_flags);
3024   }
3025
3026   void ms_handle_connect(Connection *con) override;
3027   bool ms_handle_reset(Connection *con) override;
3028   void ms_handle_remote_reset(Connection *con) override;
3029   bool ms_handle_refused(Connection *con) override;
3030   bool ms_get_authorizer(int dest_type,
3031                          AuthAuthorizer **authorizer,
3032                          bool force_new) override;
3033
3034   void blacklist_self(bool set);
3035
3036 private:
3037   epoch_t epoch_barrier;
3038   bool retry_writes_after_first_reply;
3039 public:
3040   void set_epoch_barrier(epoch_t epoch);
3041
3042   PerfCounters *get_logger() {
3043     return logger;
3044   }
3045 };
3046
3047 #endif