fs/ceph/osd_client.c

   1 #include "ceph_debug.h"
   2
   3 #include <linux/err.h>
   4 #include <linux/highmem.h>
   5 #include <linux/mm.h>
   6 #include <linux/pagemap.h>
   7 #include <linux/slab.h>
   8 #include <linux/uaccess.h>
   9
  10 #include "super.h"
  11 #include "osd_client.h"
  12 #include "messenger.h"
  13 #include "decode.h"
  14 #include "auth.h"
  15
  16 #define OSD_REPLY_RESERVE_FRONT_LEN     512
  17
  18 const static struct ceph_connection_operations osd_con_ops;
  19
  20 static void kick_requests(struct ceph_osd_client *osdc, struct ceph_osd *osd);
  21
  22 /*
  23  * Implement client access to distributed object storage cluster.
  24  *
  25  * All data objects are stored within a cluster/cloud of OSDs, or
  26  * "object storage devices."  (Note that Ceph OSDs have _nothing_ to
  27  * do with the T10 OSD extensions to SCSI.)  Ceph OSDs are simply
  28  * remote daemons serving up and coordinating consistent and safe
  29  * access to storage.
  30  *
  31  * Cluster membership and the mapping of data objects onto storage devices
  32  * are described by the osd map.
  33  *
  34  * We keep track of pending OSD requests (read, write), resubmit
  35  * requests to different OSDs when the cluster topology/data layout
  36  * change, or retry the affected requests when the communications
  37  * channel with an OSD is reset.
  38  */
  39
  40 /*
  41  * calculate the mapping of a file extent onto an object, and fill out the
  42  * request accordingly.  shorten extent as necessary if it crosses an
  43  * object boundary.
  44  *
  45  * fill osd op in request message.
  46  */
  47 static void calc_layout(struct ceph_osd_client *osdc,
  48                         struct ceph_vino vino, struct ceph_file_layout *layout,
  49                         u64 off, u64 *plen,
  50                         struct ceph_osd_request *req)
  51 {
  52         struct ceph_osd_request_head *reqhead = req->r_request->front.iov_base;
  53         struct ceph_osd_op *op = (void *)(reqhead + 1);
  54         u64 orig_len = *plen;
  55         u64 objoff, objlen;    /* extent in object */
  56         u64 bno;
  57
  58         reqhead->snapid = cpu_to_le64(vino.snap);
  59
  60         /* object extent? */
  61         ceph_calc_file_object_mapping(layout, off, plen, &bno,
  62                                       &objoff, &objlen);
  63         if (*plen < orig_len)
  64                 dout(" skipping last %llu, final file extent %llu~%llu\n",
  65                      orig_len - *plen, off, *plen);
  66
  67         sprintf(req->r_oid, "%llx.%08llx", vino.ino, bno);
  68         req->r_oid_len = strlen(req->r_oid);
  69
  70         op->extent.offset = cpu_to_le64(objoff);
  71         op->extent.length = cpu_to_le64(objlen);
  72         req->r_num_pages = calc_pages_for(off, *plen);
  73
  74         dout("calc_layout %s (%d) %llu~%llu (%d pages)\n",
  75              req->r_oid, req->r_oid_len, objoff, objlen, req->r_num_pages);
  76 }
  77
  78 static void remove_replies(struct ceph_osd_request *req)
  79 {
  80         int i;
  81         int max = ARRAY_SIZE(req->replies);
  82
  83         for (i=0; i<max; i++) {
  84                 if (req->replies[i])
  85                         ceph_msg_put(req->replies[i]);
  86         }
  87 }
  88
  89 /*
  90  * requests
  91  */
  92 void ceph_osdc_release_request(struct kref *kref)
  93 {
  94         struct ceph_osd_request *req = container_of(kref,
  95                                                     struct ceph_osd_request,
  96                                                     r_kref);
  97
  98         if (req->r_request)
  99                 ceph_msg_put(req->r_request);
 100         if (req->r_reply)
 101                 ceph_msg_put(req->r_reply);
 102         remove_replies(req);
 103         if (req->r_con_filling_msg) {
 104                 dout("release_request revoking pages %p from con %p\n",
 105                      req->r_pages, req->r_con_filling_msg);
 106                 ceph_con_revoke_message(req->r_con_filling_msg,
 107                                       req->r_reply);
 108                 ceph_con_put(req->r_con_filling_msg);
 109         }
 110         if (req->r_own_pages)
 111                 ceph_release_page_vector(req->r_pages,
 112                                          req->r_num_pages);
 113         ceph_put_snap_context(req->r_snapc);
 114         if (req->r_mempool)
 115                 mempool_free(req, req->r_osdc->req_mempool);
 116         else
 117                 kfree(req);
 118 }
 119
 120 static int alloc_replies(struct ceph_osd_request *req, int num_reply)
 121 {
 122         int i;
 123         int max = ARRAY_SIZE(req->replies);
 124
 125         BUG_ON(num_reply > max);
 126
 127         for (i=0; i<num_reply; i++) {
 128                 req->replies[i] = ceph_msg_new(0, OSD_REPLY_RESERVE_FRONT_LEN, 0, 0, NULL);
 129                 if (IS_ERR(req->replies[i])) {
 130                         int j;
 131                         int err = PTR_ERR(req->replies[i]);
 132                         for (j = 0; j<=i; j++) {
 133                                 ceph_msg_put(req->replies[j]);
 134                         }
 135                         return err;
 136                 }
 137         }
 138
 139         for (; i<max; i++) {
 140                 req->replies[i] = NULL;
 141         }
 142
 143         req->cur_reply = 0;
 144
 145         return 0;
 146 }
 147
 148 static struct ceph_msg *__get_next_reply(struct ceph_connection *con,
 149                                        struct ceph_osd_request *req,
 150                                        int front_len)
 151 {
 152         struct ceph_msg *reply;
 153         if (req->r_con_filling_msg) {
 154                 dout("revoking reply msg %p from old con %p\n", req->r_reply,
 155                      req->r_con_filling_msg);
 156                 ceph_con_revoke_message(req->r_con_filling_msg, req->r_reply);
 157                 ceph_con_put(req->r_con_filling_msg);
 158                 req->cur_reply = 0;
 159         }
 160         reply = req->replies[req->cur_reply];
 161         if (!reply || front_len > OSD_REPLY_RESERVE_FRONT_LEN) {
 162                 /* maybe we can allocate it now? */
 163                 reply = ceph_msg_new(0, front_len, 0, 0, NULL);
 164                 if (!reply || IS_ERR(reply)) {
 165                         pr_err(" reply alloc failed, front_len=%d\n", front_len);
 166                         return ERR_PTR(-ENOMEM);
 167                 }
 168         }
 169         req->r_con_filling_msg = ceph_con_get(con);
 170         req->r_reply = ceph_msg_get(reply); /* for duration of read over socket */
 171         return ceph_msg_get(reply);
 172 }
 173
 174 /*
 175  * build new request AND message, calculate layout, and adjust file
 176  * extent as needed.
 177  *
 178  * if the file was recently truncated, we include information about its
 179  * old and new size so that the object can be updated appropriately.  (we
 180  * avoid synchronously deleting truncated objects because it's slow.)
 181  *
 182  * if @do_sync, include a 'startsync' command so that the osd will flush
 183  * data quickly.
 184  */
 185 struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *osdc,
 186                                                struct ceph_file_layout *layout,
 187                                                struct ceph_vino vino,
 188                                                u64 off, u64 *plen,
 189                                                int opcode, int flags,
 190                                                struct ceph_snap_context *snapc,
 191                                                int do_sync,
 192                                                u32 truncate_seq,
 193                                                u64 truncate_size,
 194                                                struct timespec *mtime,
 195                                                bool use_mempool, int num_reply)
 196 {
 197         struct ceph_osd_request *req;
 198         struct ceph_msg *msg;
 199         struct ceph_osd_request_head *head;
 200         struct ceph_osd_op *op;
 201         void *p;
 202         int num_op = 1 + do_sync;
 203         size_t msg_size = sizeof(*head) + num_op*sizeof(*op);
 204         int err, i;
 205
 206         if (use_mempool) {
 207                 req = mempool_alloc(osdc->req_mempool, GFP_NOFS);
 208                 memset(req, 0, sizeof(*req));
 209         } else {
 210                 req = kzalloc(sizeof(*req), GFP_NOFS);
 211         }
 212         if (req == NULL)
 213                 return ERR_PTR(-ENOMEM);
 214
 215         err = alloc_replies(req, num_reply);
 216         if (err) {
 217                 ceph_osdc_put_request(req);
 218                 return ERR_PTR(-ENOMEM);
 219         }
 220         req->r_num_prealloc_reply = num_reply;
 221
 222         req->r_osdc = osdc;
 223         req->r_mempool = use_mempool;
 224         kref_init(&req->r_kref);
 225         init_completion(&req->r_completion);
 226         init_completion(&req->r_safe_completion);
 227         INIT_LIST_HEAD(&req->r_unsafe_item);
 228         req->r_flags = flags;
 229
 230         WARN_ON((flags & (CEPH_OSD_FLAG_READ|CEPH_OSD_FLAG_WRITE)) == 0);
 231
 232         /* create message; allow space for oid */
 233         msg_size += 40;
 234         if (snapc)
 235                 msg_size += sizeof(u64) * snapc->num_snaps;
 236         if (use_mempool)
 237                 msg = ceph_msgpool_get(&osdc->msgpool_op, 0);
 238         else
 239                 msg = ceph_msg_new(CEPH_MSG_OSD_OP, msg_size, 0, 0, NULL);
 240         if (IS_ERR(msg)) {
 241                 ceph_osdc_put_request(req);
 242                 return ERR_PTR(PTR_ERR(msg));
 243         }
 244         msg->hdr.type = cpu_to_le16(CEPH_MSG_OSD_OP);
 245         memset(msg->front.iov_base, 0, msg->front.iov_len);
 246         head = msg->front.iov_base;
 247         op = (void *)(head + 1);
 248         p = (void *)(op + num_op);
 249
 250         req->r_request = msg;
 251         req->r_snapc = ceph_get_snap_context(snapc);
 252
 253         head->client_inc = cpu_to_le32(1); /* always, for now. */
 254         head->flags = cpu_to_le32(flags);
 255         if (flags & CEPH_OSD_FLAG_WRITE)
 256                 ceph_encode_timespec(&head->mtime, mtime);
 257         head->num_ops = cpu_to_le16(num_op);
 258         op->op = cpu_to_le16(opcode);
 259
 260         /* calculate max write size */
 261         calc_layout(osdc, vino, layout, off, plen, req);
 262         req->r_file_layout = *layout;  /* keep a copy */
 263
 264         if (flags & CEPH_OSD_FLAG_WRITE) {
 265                 req->r_request->hdr.data_off = cpu_to_le16(off);
 266                 req->r_request->hdr.data_len = cpu_to_le32(*plen);
 267                 op->payload_len = cpu_to_le32(*plen);
 268         }
 269         op->extent.truncate_size = cpu_to_le64(truncate_size);
 270         op->extent.truncate_seq = cpu_to_le32(truncate_seq);
 271
 272         /* fill in oid */
 273         head->object_len = cpu_to_le32(req->r_oid_len);
 274         memcpy(p, req->r_oid, req->r_oid_len);
 275         p += req->r_oid_len;
 276
 277         if (do_sync) {
 278                 op++;
 279                 op->op = cpu_to_le16(CEPH_OSD_OP_STARTSYNC);
 280         }
 281         if (snapc) {
 282                 head->snap_seq = cpu_to_le64(snapc->seq);
 283                 head->num_snaps = cpu_to_le32(snapc->num_snaps);
 284                 for (i = 0; i < snapc->num_snaps; i++) {
 285                         put_unaligned_le64(snapc->snaps[i], p);
 286                         p += sizeof(u64);
 287                 }
 288         }
 289
 290         BUG_ON(p > msg->front.iov_base + msg->front.iov_len);
 291         return req;
 292 }
 293
 294 /*
 295  * We keep osd requests in an rbtree, sorted by ->r_tid.
 296  */
 297 static void __insert_request(struct ceph_osd_client *osdc,
 298                              struct ceph_osd_request *new)
 299 {
 300         struct rb_node **p = &osdc->requests.rb_node;
 301         struct rb_node *parent = NULL;
 302         struct ceph_osd_request *req = NULL;
 303
 304         while (*p) {
 305                 parent = *p;
 306                 req = rb_entry(parent, struct ceph_osd_request, r_node);
 307                 if (new->r_tid < req->r_tid)
 308                         p = &(*p)->rb_left;
 309                 else if (new->r_tid > req->r_tid)
 310                         p = &(*p)->rb_right;
 311                 else
 312                         BUG();
 313         }
 314
 315         rb_link_node(&new->r_node, parent, p);
 316         rb_insert_color(&new->r_node, &osdc->requests);
 317 }
 318
 319 static struct ceph_osd_request *__lookup_request(struct ceph_osd_client *osdc,
 320                                                  u64 tid)
 321 {
 322         struct ceph_osd_request *req;
 323         struct rb_node *n = osdc->requests.rb_node;
 324
 325         while (n) {
 326                 req = rb_entry(n, struct ceph_osd_request, r_node);
 327                 if (tid < req->r_tid)
 328                         n = n->rb_left;
 329                 else if (tid > req->r_tid)
 330                         n = n->rb_right;
 331                 else
 332                         return req;
 333         }
 334         return NULL;
 335 }
 336
 337 static struct ceph_osd_request *
 338 __lookup_request_ge(struct ceph_osd_client *osdc,
 339                     u64 tid)
 340 {
 341         struct ceph_osd_request *req;
 342         struct rb_node *n = osdc->requests.rb_node;
 343
 344         while (n) {
 345                 req = rb_entry(n, struct ceph_osd_request, r_node);
 346                 if (tid < req->r_tid) {
 347                         if (!n->rb_left)
 348                                 return req;
 349                         n = n->rb_left;
 350                 } else if (tid > req->r_tid) {
 351                         n = n->rb_right;
 352                 } else {
 353                         return req;
 354                 }
 355         }
 356         return NULL;
 357 }
 358
 359
 360 /*
 361  * If the osd connection drops, we need to resubmit all requests.
 362  */
 363 static void osd_reset(struct ceph_connection *con)
 364 {
 365         struct ceph_osd *osd = con->private;
 366         struct ceph_osd_client *osdc;
 367
 368         if (!osd)
 369                 return;
 370         dout("osd_reset osd%d\n", osd->o_osd);
 371         osdc = osd->o_osdc;
 372         osd->o_incarnation++;
 373         down_read(&osdc->map_sem);
 374         kick_requests(osdc, osd);
 375         up_read(&osdc->map_sem);
 376 }
 377
 378 /*
 379  * Track open sessions with osds.
 380  */
 381 static struct ceph_osd *create_osd(struct ceph_osd_client *osdc)
 382 {
 383         struct ceph_osd *osd;
 384
 385         osd = kzalloc(sizeof(*osd), GFP_NOFS);
 386         if (!osd)
 387                 return NULL;
 388
 389         atomic_set(&osd->o_ref, 1);
 390         osd->o_osdc = osdc;
 391         INIT_LIST_HEAD(&osd->o_requests);
 392         INIT_LIST_HEAD(&osd->o_osd_lru);
 393         osd->o_incarnation = 1;
 394
 395         ceph_con_init(osdc->client->msgr, &osd->o_con);
 396         osd->o_con.private = osd;
 397         osd->o_con.ops = &osd_con_ops;
 398         osd->o_con.peer_name.type = CEPH_ENTITY_TYPE_OSD;
 399
 400         return osd;
 401 }
 402
 403 static struct ceph_osd *get_osd(struct ceph_osd *osd)
 404 {
 405         if (atomic_inc_not_zero(&osd->o_ref)) {
 406                 dout("get_osd %p %d -> %d\n", osd, atomic_read(&osd->o_ref)-1,
 407                      atomic_read(&osd->o_ref));
 408                 return osd;
 409         } else {
 410                 dout("get_osd %p FAIL\n", osd);
 411                 return NULL;
 412         }
 413 }
 414
 415 static void put_osd(struct ceph_osd *osd)
 416 {
 417         dout("put_osd %p %d -> %d\n", osd, atomic_read(&osd->o_ref),
 418              atomic_read(&osd->o_ref) - 1);
 419         if (atomic_dec_and_test(&osd->o_ref))
 420                 kfree(osd);
 421 }
 422
 423 /*
 424  * remove an osd from our map
 425  */
 426 static void __remove_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd)
 427 {
 428         dout("__remove_osd %p\n", osd);
 429         BUG_ON(!list_empty(&osd->o_requests));
 430         rb_erase(&osd->o_node, &osdc->osds);
 431         list_del_init(&osd->o_osd_lru);
 432         ceph_con_close(&osd->o_con);
 433         put_osd(osd);
 434 }
 435
 436 static void __move_osd_to_lru(struct ceph_osd_client *osdc,
 437                               struct ceph_osd *osd)
 438 {
 439         dout("__move_osd_to_lru %p\n", osd);
 440         BUG_ON(!list_empty(&osd->o_osd_lru));
 441         list_add_tail(&osd->o_osd_lru, &osdc->osd_lru);
 442         osd->lru_ttl = jiffies + osdc->client->mount_args->osd_idle_ttl * HZ;
 443 }
 444
 445 static void __remove_osd_from_lru(struct ceph_osd *osd)
 446 {
 447         dout("__remove_osd_from_lru %p\n", osd);
 448         if (!list_empty(&osd->o_osd_lru))
 449                 list_del_init(&osd->o_osd_lru);
 450 }
 451
 452 static void remove_old_osds(struct ceph_osd_client *osdc, int remove_all)
 453 {
 454         struct ceph_osd *osd, *nosd;
 455
 456         dout("__remove_old_osds %p\n", osdc);
 457         mutex_lock(&osdc->request_mutex);
 458         list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) {
 459                 if (!remove_all && time_before(jiffies, osd->lru_ttl))
 460                         break;
 461                 __remove_osd(osdc, osd);
 462         }
 463         mutex_unlock(&osdc->request_mutex);
 464 }
 465
 466 /*
 467  * reset osd connect
 468  */
 469 static int __reset_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd)
 470 {
 471         int ret = 0;
 472
 473         dout("__reset_osd %p osd%d\n", osd, osd->o_osd);
 474         if (list_empty(&osd->o_requests)) {
 475                 __remove_osd(osdc, osd);
 476         } else {
 477                 ceph_con_close(&osd->o_con);
 478                 ceph_con_open(&osd->o_con, &osdc->osdmap->osd_addr[osd->o_osd]);
 479                 osd->o_incarnation++;
 480         }
 481         return ret;
 482 }
 483
 484 static void __insert_osd(struct ceph_osd_client *osdc, struct ceph_osd *new)
 485 {
 486         struct rb_node **p = &osdc->osds.rb_node;
 487         struct rb_node *parent = NULL;
 488         struct ceph_osd *osd = NULL;
 489
 490         while (*p) {
 491                 parent = *p;
 492                 osd = rb_entry(parent, struct ceph_osd, o_node);
 493                 if (new->o_osd < osd->o_osd)
 494                         p = &(*p)->rb_left;
 495                 else if (new->o_osd > osd->o_osd)
 496                         p = &(*p)->rb_right;
 497                 else
 498                         BUG();
 499         }
 500
 501         rb_link_node(&new->o_node, parent, p);
 502         rb_insert_color(&new->o_node, &osdc->osds);
 503 }
 504
 505 static struct ceph_osd *__lookup_osd(struct ceph_osd_client *osdc, int o)
 506 {
 507         struct ceph_osd *osd;
 508         struct rb_node *n = osdc->osds.rb_node;
 509
 510         while (n) {
 511                 osd = rb_entry(n, struct ceph_osd, o_node);
 512                 if (o < osd->o_osd)
 513                         n = n->rb_left;
 514                 else if (o > osd->o_osd)
 515                         n = n->rb_right;
 516                 else
 517                         return osd;
 518         }
 519         return NULL;
 520 }
 521
 522
 523 /*
 524  * Register request, assign tid.  If this is the first request, set up
 525  * the timeout event.
 526  */
 527 static void register_request(struct ceph_osd_client *osdc,
 528                              struct ceph_osd_request *req)
 529 {
 530         mutex_lock(&osdc->request_mutex);
 531         req->r_tid = ++osdc->last_tid;
 532         req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
 533
 534         dout("register_request %p tid %lld\n", req, req->r_tid);
 535         __insert_request(osdc, req);
 536         ceph_osdc_get_request(req);
 537         osdc->num_requests++;
 538
 539         req->r_timeout_stamp =
 540                 jiffies + osdc->client->mount_args->osd_timeout*HZ;
 541
 542         if (osdc->num_requests == 1) {
 543                 osdc->timeout_tid = req->r_tid;
 544                 dout("  timeout on tid %llu at %lu\n", req->r_tid,
 545                      req->r_timeout_stamp);
 546                 schedule_delayed_work(&osdc->timeout_work,
 547                       round_jiffies_relative(req->r_timeout_stamp - jiffies));
 548         }
 549         mutex_unlock(&osdc->request_mutex);
 550 }
 551
 552 /*
 553  * called under osdc->request_mutex
 554  */
 555 static void __unregister_request(struct ceph_osd_client *osdc,
 556                                  struct ceph_osd_request *req)
 557 {
 558         dout("__unregister_request %p tid %lld\n", req, req->r_tid);
 559         rb_erase(&req->r_node, &osdc->requests);
 560         osdc->num_requests--;
 561
 562         if (req->r_osd) {
 563                 /* make sure the original request isn't in flight. */
 564                 ceph_con_revoke(&req->r_osd->o_con, req->r_request);
 565
 566                 list_del_init(&req->r_osd_item);
 567                 if (list_empty(&req->r_osd->o_requests))
 568                         __move_osd_to_lru(osdc, req->r_osd);
 569                 req->r_osd = NULL;
 570         }
 571
 572         ceph_osdc_put_request(req);
 573
 574         if (req->r_tid == osdc->timeout_tid) {
 575                 if (osdc->num_requests == 0) {
 576                         dout("no requests, canceling timeout\n");
 577                         osdc->timeout_tid = 0;
 578                         cancel_delayed_work(&osdc->timeout_work);
 579                 } else {
 580                         req = rb_entry(rb_first(&osdc->requests),
 581                                        struct ceph_osd_request, r_node);
 582                         osdc->timeout_tid = req->r_tid;
 583                         dout("rescheduled timeout on tid %llu at %lu\n",
 584                              req->r_tid, req->r_timeout_stamp);
 585                         schedule_delayed_work(&osdc->timeout_work,
 586                               round_jiffies_relative(req->r_timeout_stamp -
 587                                                      jiffies));
 588                 }
 589         }
 590 }
 591
 592 /*
 593  * Cancel a previously queued request message
 594  */
 595 static void __cancel_request(struct ceph_osd_request *req)
 596 {
 597         if (req->r_sent) {
 598                 ceph_con_revoke(&req->r_osd->o_con, req->r_request);
 599                 req->r_sent = 0;
 600         }
 601 }
 602
 603 /*
 604  * Pick an osd (the first 'up' osd in the pg), allocate the osd struct
 605  * (as needed), and set the request r_osd appropriately.  If there is
 606  * no up osd, set r_osd to NULL.
 607  *
 608  * Return 0 if unchanged, 1 if changed, or negative on error.
 609  *
 610  * Caller should hold map_sem for read and request_mutex.
 611  */
 612 static int __map_osds(struct ceph_osd_client *osdc,
 613                       struct ceph_osd_request *req)
 614 {
 615         struct ceph_osd_request_head *reqhead = req->r_request->front.iov_base;
 616         struct ceph_pg pgid;
 617         int o = -1;
 618         int err;
 619         struct ceph_osd *newosd = NULL;
 620
 621         dout("map_osds %p tid %lld\n", req, req->r_tid);
 622         err = ceph_calc_object_layout(&reqhead->layout, req->r_oid,
 623                                       &req->r_file_layout, osdc->osdmap);
 624         if (err)
 625                 return err;
 626         pgid = reqhead->layout.ol_pgid;
 627         req->r_pgid = pgid;
 628
 629         o = ceph_calc_pg_primary(osdc->osdmap, pgid);
 630
 631         if ((req->r_osd && req->r_osd->o_osd == o &&
 632              req->r_sent >= req->r_osd->o_incarnation) ||
 633             (req->r_osd == NULL && o == -1))
 634                 return 0;  /* no change */
 635
 636         dout("map_osds tid %llu pgid %d.%x osd%d (was osd%d)\n",
 637              req->r_tid, le32_to_cpu(pgid.pool), le16_to_cpu(pgid.ps), o,
 638              req->r_osd ? req->r_osd->o_osd : -1);
 639
 640         if (req->r_osd) {
 641                 __cancel_request(req);
 642                 list_del_init(&req->r_osd_item);
 643                 if (list_empty(&req->r_osd->o_requests)) {
 644                         /* try to re-use r_osd if possible */
 645                         newosd = get_osd(req->r_osd);
 646                         __remove_osd(osdc, newosd);
 647                 }
 648                 req->r_osd = NULL;
 649         }
 650
 651         req->r_osd = __lookup_osd(osdc, o);
 652         if (!req->r_osd && o >= 0) {
 653                 if (newosd) {
 654                         req->r_osd = newosd;
 655                         newosd = NULL;
 656                 } else {
 657                         err = -ENOMEM;
 658                         req->r_osd = create_osd(osdc);
 659                         if (!req->r_osd)
 660                                 goto out;
 661                 }
 662
 663                 dout("map_osds osd %p is osd%d\n", req->r_osd, o);
 664                 req->r_osd->o_osd = o;
 665                 req->r_osd->o_con.peer_name.num = cpu_to_le64(o);
 666                 __insert_osd(osdc, req->r_osd);
 667
 668                 ceph_con_open(&req->r_osd->o_con, &osdc->osdmap->osd_addr[o]);
 669         }
 670
 671         if (req->r_osd) {
 672                 __remove_osd_from_lru(req->r_osd);
 673                 list_add(&req->r_osd_item, &req->r_osd->o_requests);
 674         }
 675         err = 1;   /* osd changed */
 676
 677 out:
 678         if (newosd)
 679                 put_osd(newosd);
 680         return err;
 681 }
 682
 683 /*
 684  * caller should hold map_sem (for read) and request_mutex
 685  */
 686 static int __send_request(struct ceph_osd_client *osdc,
 687                           struct ceph_osd_request *req)
 688 {
 689         struct ceph_osd_request_head *reqhead;
 690         int err;
 691
 692         err = __map_osds(osdc, req);
 693         if (err < 0)
 694                 return err;
 695         if (req->r_osd == NULL) {
 696                 dout("send_request %p no up osds in pg\n", req);
 697                 ceph_monc_request_next_osdmap(&osdc->client->monc);
 698                 return 0;
 699         }
 700
 701         dout("send_request %p tid %llu to osd%d flags %d\n",
 702              req, req->r_tid, req->r_osd->o_osd, req->r_flags);
 703
 704         reqhead = req->r_request->front.iov_base;
 705         reqhead->osdmap_epoch = cpu_to_le32(osdc->osdmap->epoch);
 706         reqhead->flags |= cpu_to_le32(req->r_flags);  /* e.g., RETRY */
 707         reqhead->reassert_version = req->r_reassert_version;
 708
 709         req->r_timeout_stamp = jiffies+osdc->client->mount_args->osd_timeout*HZ;
 710
 711         ceph_msg_get(req->r_request); /* send consumes a ref */
 712         ceph_con_send(&req->r_osd->o_con, req->r_request);
 713         req->r_sent = req->r_osd->o_incarnation;
 714         return 0;
 715 }
 716
 717 /*
 718  * Timeout callback, called every N seconds when 1 or more osd
 719  * requests has been active for more than N seconds.  When this
 720  * happens, we ping all OSDs with requests who have timed out to
 721  * ensure any communications channel reset is detected.  Reset the
 722  * request timeouts another N seconds in the future as we go.
 723  * Reschedule the timeout event another N seconds in future (unless
 724  * there are no open requests).
 725  */
 726 static void handle_timeout(struct work_struct *work)
 727 {
 728         struct ceph_osd_client *osdc =
 729                 container_of(work, struct ceph_osd_client, timeout_work.work);
 730         struct ceph_osd_request *req;
 731         struct ceph_osd *osd;
 732         unsigned long timeout = osdc->client->mount_args->osd_timeout * HZ;
 733         unsigned long next_timeout = timeout + jiffies;
 734         struct rb_node *p;
 735
 736         dout("timeout\n");
 737         down_read(&osdc->map_sem);
 738
 739         ceph_monc_request_next_osdmap(&osdc->client->monc);
 740
 741         mutex_lock(&osdc->request_mutex);
 742         for (p = rb_first(&osdc->requests); p; p = rb_next(p)) {
 743                 req = rb_entry(p, struct ceph_osd_request, r_node);
 744
 745                 if (req->r_resend) {
 746                         int err;
 747
 748                         dout("osdc resending prev failed %lld\n", req->r_tid);
 749                         err = __send_request(osdc, req);
 750                         if (err)
 751                                 dout("osdc failed again on %lld\n", req->r_tid);
 752                         else
 753                                 req->r_resend = false;
 754                         continue;
 755                 }
 756         }
 757         for (p = rb_first(&osdc->osds); p; p = rb_next(p)) {
 758                 osd = rb_entry(p, struct ceph_osd, o_node);
 759                 if (list_empty(&osd->o_requests))
 760                         continue;
 761                 req = list_first_entry(&osd->o_requests,
 762                                        struct ceph_osd_request, r_osd_item);
 763                 if (time_before(jiffies, req->r_timeout_stamp))
 764                         continue;
 765
 766                 dout(" tid %llu (at least) timed out on osd%d\n",
 767                      req->r_tid, osd->o_osd);
 768                 req->r_timeout_stamp = next_timeout;
 769                 ceph_con_keepalive(&osd->o_con);
 770         }
 771
 772         if (osdc->timeout_tid)
 773                 schedule_delayed_work(&osdc->timeout_work,
 774                                       round_jiffies_relative(timeout));
 775
 776         mutex_unlock(&osdc->request_mutex);
 777
 778         up_read(&osdc->map_sem);
 779 }
 780
 781 static void handle_osds_timeout(struct work_struct *work)
 782 {
 783         struct ceph_osd_client *osdc =
 784                 container_of(work, struct ceph_osd_client,
 785                              osds_timeout_work.work);
 786         unsigned long delay =
 787                 osdc->client->mount_args->osd_idle_ttl * HZ >> 2;
 788
 789         dout("osds timeout\n");
 790         down_read(&osdc->map_sem);
 791         remove_old_osds(osdc, 0);
 792         up_read(&osdc->map_sem);
 793
 794         schedule_delayed_work(&osdc->osds_timeout_work,
 795                               round_jiffies_relative(delay));
 796 }
 797
 798 /*
 799  * handle osd op reply.  either call the callback if it is specified,
 800  * or do the completion to wake up the waiting thread.
 801  */
 802 static void handle_reply(struct ceph_osd_client *osdc, struct ceph_msg *msg,
 803                          struct ceph_connection *con)
 804 {
 805         struct ceph_osd_reply_head *rhead = msg->front.iov_base;
 806         struct ceph_osd_request *req;
 807         u64 tid;
 808         int numops, object_len, flags;
 809
 810         tid = le64_to_cpu(msg->hdr.tid);
 811         if (msg->front.iov_len < sizeof(*rhead))
 812                 goto bad;
 813         numops = le32_to_cpu(rhead->num_ops);
 814         object_len = le32_to_cpu(rhead->object_len);
 815         if (msg->front.iov_len != sizeof(*rhead) + object_len +
 816             numops * sizeof(struct ceph_osd_op))
 817                 goto bad;
 818         dout("handle_reply %p tid %llu\n", msg, tid);
 819
 820         /* lookup */
 821         mutex_lock(&osdc->request_mutex);
 822         req = __lookup_request(osdc, tid);
 823         if (req == NULL) {
 824                 dout("handle_reply tid %llu dne\n", tid);
 825                 mutex_unlock(&osdc->request_mutex);
 826                 return;
 827         }
 828         ceph_osdc_get_request(req);
 829         flags = le32_to_cpu(rhead->flags);
 830
 831         /*
 832          * if this connection filled our message, drop our reference now, to
 833          * avoid a (safe but slower) revoke later.
 834          */
 835         if (req->r_con_filling_msg == con && req->r_reply == msg) {
 836                 dout(" got pages, dropping con_filling_msg ref %p\n", con);
 837                 req->r_con_filling_msg = NULL;
 838                 ceph_con_put(con);
 839         }
 840
 841         if (req->r_reply) {
 842                 /*
 843                  * once we see the message has been received, we don't
 844                  * need a ref (which is only needed for revoking
 845                  * pages)
 846                  */
 847                 ceph_msg_put(req->r_reply);
 848                 req->r_reply = NULL;
 849         }
 850
 851         if (!req->r_got_reply) {
 852                 unsigned bytes;
 853
 854                 req->r_result = le32_to_cpu(rhead->result);
 855                 bytes = le32_to_cpu(msg->hdr.data_len);
 856                 dout("handle_reply result %d bytes %d\n", req->r_result,
 857                      bytes);
 858                 if (req->r_result == 0)
 859                         req->r_result = bytes;
 860
 861                 /* in case this is a write and we need to replay, */
 862                 req->r_reassert_version = rhead->reassert_version;
 863
 864                 req->r_got_reply = 1;
 865         } else if ((flags & CEPH_OSD_FLAG_ONDISK) == 0) {
 866                 dout("handle_reply tid %llu dup ack\n", tid);
 867                 mutex_unlock(&osdc->request_mutex);
 868                 goto done;
 869         }
 870
 871         dout("handle_reply tid %llu flags %d\n", tid, flags);
 872
 873         /* either this is a read, or we got the safe response */
 874         if ((flags & CEPH_OSD_FLAG_ONDISK) ||
 875             ((flags & CEPH_OSD_FLAG_WRITE) == 0))
 876                 __unregister_request(osdc, req);
 877
 878         mutex_unlock(&osdc->request_mutex);
 879
 880         if (req->r_callback)
 881                 req->r_callback(req, msg);
 882         else
 883                 complete(&req->r_completion);
 884
 885         if (flags & CEPH_OSD_FLAG_ONDISK) {
 886                 if (req->r_safe_callback)
 887                         req->r_safe_callback(req, msg);
 888                 complete(&req->r_safe_completion);  /* fsync waiter */
 889         }
 890
 891 done:
 892         ceph_osdc_put_request(req);
 893         return;
 894
 895 bad:
 896         pr_err("corrupt osd_op_reply got %d %d expected %d\n",
 897                (int)msg->front.iov_len, le32_to_cpu(msg->hdr.front_len),
 898                (int)sizeof(*rhead));
 899         ceph_msg_dump(msg);
 900 }
 901
 902
 903 /*
 904  * Resubmit osd requests whose osd or osd address has changed.  Request
 905  * a new osd map if osds are down, or we are otherwise unable to determine
 906  * how to direct a request.
 907  *
 908  * Close connections to down osds.
 909  *
 910  * If @who is specified, resubmit requests for that specific osd.
 911  *
 912  * Caller should hold map_sem for read and request_mutex.
 913  */
 914 static void kick_requests(struct ceph_osd_client *osdc,
 915                           struct ceph_osd *kickosd)
 916 {
 917         struct ceph_osd_request *req;
 918         struct rb_node *p, *n;
 919         int needmap = 0;
 920         int err;
 921
 922         dout("kick_requests osd%d\n", kickosd ? kickosd->o_osd : -1);
 923         mutex_lock(&osdc->request_mutex);
 924         if (!kickosd) {
 925                 for (p = rb_first(&osdc->osds); p; p = n) {
 926                         struct ceph_osd *osd =
 927                                 rb_entry(p, struct ceph_osd, o_node);
 928
 929                         n = rb_next(p);
 930                         if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) ||
 931                             memcmp(&osd->o_con.peer_addr,
 932                                    ceph_osd_addr(osdc->osdmap,
 933                                                  osd->o_osd),
 934                                    sizeof(struct ceph_entity_addr)) != 0)
 935                                 __reset_osd(osdc, osd);
 936                 }
 937         }
 938
 939         for (p = rb_first(&osdc->requests); p; p = rb_next(p)) {
 940                 req = rb_entry(p, struct ceph_osd_request, r_node);
 941
 942                 if (req->r_resend) {
 943                         dout(" r_resend set on tid %llu\n", req->r_tid);
 944                         __cancel_request(req);
 945                         goto kick;
 946                 }
 947                 if (req->r_osd && kickosd == req->r_osd) {
 948                         __cancel_request(req);
 949                         goto kick;
 950                 }
 951
 952                 err = __map_osds(osdc, req);
 953                 if (err == 0)
 954                         continue;  /* no change */
 955                 if (err < 0) {
 956                         /*
 957                          * FIXME: really, we should set the request
 958                          * error and fail if this isn't a 'nofail'
 959                          * request, but that's a fair bit more
 960                          * complicated to do.  So retry!
 961                          */
 962                         dout(" setting r_resend on %llu\n", req->r_tid);
 963                         req->r_resend = true;
 964                         continue;
 965                 }
 966                 if (req->r_osd == NULL) {
 967                         dout("tid %llu maps to no valid osd\n", req->r_tid);
 968                         needmap++;  /* request a newer map */
 969                         continue;
 970                 }
 971
 972 kick:
 973                 dout("kicking %p tid %llu osd%d\n", req, req->r_tid,
 974                      req->r_osd->o_osd);
 975                 req->r_flags |= CEPH_OSD_FLAG_RETRY;
 976                 err = __send_request(osdc, req);
 977                 if (err) {
 978                         dout(" setting r_resend on %llu\n", req->r_tid);
 979                         req->r_resend = true;
 980                 }
 981         }
 982         mutex_unlock(&osdc->request_mutex);
 983
 984         if (needmap) {
 985                 dout("%d requests for down osds, need new map\n", needmap);
 986                 ceph_monc_request_next_osdmap(&osdc->client->monc);
 987         }
 988 }
 989
 990 /*
 991  * Process updated osd map.
 992  *
 993  * The message contains any number of incremental and full maps, normally
 994  * indicating some sort of topology change in the cluster.  Kick requests
 995  * off to different OSDs as needed.
 996  */
 997 void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
 998 {
 999         void *p, *end, *next;
1000         u32 nr_maps, maplen;
1001         u32 epoch;
1002         struct ceph_osdmap *newmap = NULL, *oldmap;
1003         int err;
1004         struct ceph_fsid fsid;
1005
1006         dout("handle_map have %u\n", osdc->osdmap ? osdc->osdmap->epoch : 0);
1007         p = msg->front.iov_base;
1008         end = p + msg->front.iov_len;
1009
1010         /* verify fsid */
1011         ceph_decode_need(&p, end, sizeof(fsid), bad);
1012         ceph_decode_copy(&p, &fsid, sizeof(fsid));
1013         if (ceph_check_fsid(osdc->client, &fsid) < 0)
1014                 return;
1015
1016         down_write(&osdc->map_sem);
1017
1018         /* incremental maps */
1019         ceph_decode_32_safe(&p, end, nr_maps, bad);
1020         dout(" %d inc maps\n", nr_maps);
1021         while (nr_maps > 0) {
1022                 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
1023                 epoch = ceph_decode_32(&p);
1024                 maplen = ceph_decode_32(&p);
1025                 ceph_decode_need(&p, end, maplen, bad);
1026                 next = p + maplen;
1027                 if (osdc->osdmap && osdc->osdmap->epoch+1 == epoch) {
1028                         dout("applying incremental map %u len %d\n",
1029                              epoch, maplen);
1030                         newmap = osdmap_apply_incremental(&p, next,
1031                                                           osdc->osdmap,
1032                                                           osdc->client->msgr);
1033                         if (IS_ERR(newmap)) {
1034                                 err = PTR_ERR(newmap);
1035                                 goto bad;
1036                         }
1037                         BUG_ON(!newmap);
1038                         if (newmap != osdc->osdmap) {
1039                                 ceph_osdmap_destroy(osdc->osdmap);
1040                                 osdc->osdmap = newmap;
1041                         }
1042                 } else {
1043                         dout("ignoring incremental map %u len %d\n",
1044                              epoch, maplen);
1045                 }
1046                 p = next;
1047                 nr_maps--;
1048         }
1049         if (newmap)
1050                 goto done;
1051
1052         /* full maps */
1053         ceph_decode_32_safe(&p, end, nr_maps, bad);
1054         dout(" %d full maps\n", nr_maps);
1055         while (nr_maps) {
1056                 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
1057                 epoch = ceph_decode_32(&p);
1058                 maplen = ceph_decode_32(&p);
1059                 ceph_decode_need(&p, end, maplen, bad);
1060                 if (nr_maps > 1) {
1061                         dout("skipping non-latest full map %u len %d\n",
1062                              epoch, maplen);
1063                 } else if (osdc->osdmap && osdc->osdmap->epoch >= epoch) {
1064                         dout("skipping full map %u len %d, "
1065                              "older than our %u\n", epoch, maplen,
1066                              osdc->osdmap->epoch);
1067                 } else {
1068                         dout("taking full map %u len %d\n", epoch, maplen);
1069                         newmap = osdmap_decode(&p, p+maplen);
1070                         if (IS_ERR(newmap)) {
1071                                 err = PTR_ERR(newmap);
1072                                 goto bad;
1073                         }
1074                         BUG_ON(!newmap);
1075                         oldmap = osdc->osdmap;
1076                         osdc->osdmap = newmap;
1077                         if (oldmap)
1078                                 ceph_osdmap_destroy(oldmap);
1079                 }
1080                 p += maplen;
1081                 nr_maps--;
1082         }
1083
1084 done:
1085         downgrade_write(&osdc->map_sem);
1086         ceph_monc_got_osdmap(&osdc->client->monc, osdc->osdmap->epoch);
1087         if (newmap)
1088                 kick_requests(osdc, NULL);
1089         up_read(&osdc->map_sem);
1090         return;
1091
1092 bad:
1093         pr_err("osdc handle_map corrupt msg\n");
1094         ceph_msg_dump(msg);
1095         up_write(&osdc->map_sem);
1096         return;
1097 }
1098
1099
1100 /*
1101  * A read request prepares specific pages that data is to be read into.
1102  * When a message is being read off the wire, we call prepare_pages to
1103  * find those pages.
1104  *  0 = success, -1 failure.
1105  */
1106 static int __prepare_pages(struct ceph_connection *con,
1107                          struct ceph_msg_header *hdr,
1108                          struct ceph_osd_request *req,
1109                          u64 tid,
1110                          struct ceph_msg *m)
1111 {
1112         struct ceph_osd *osd = con->private;
1113         struct ceph_osd_client *osdc;
1114         int ret = -1;
1115         int data_len = le32_to_cpu(hdr->data_len);
1116         unsigned data_off = le16_to_cpu(hdr->data_off);
1117
1118         int want = calc_pages_for(data_off & ~PAGE_MASK, data_len);
1119
1120         if (!osd)
1121                 return -1;
1122
1123         osdc = osd->o_osdc;
1124
1125         dout("__prepare_pages on msg %p tid %llu, has %d pages, want %d\n", m,
1126              tid, req->r_num_pages, want);
1127         if (unlikely(req->r_num_pages < want))
1128                 goto out;
1129         m->pages = req->r_pages;
1130         m->nr_pages = req->r_num_pages;
1131         ret = 0; /* success */
1132 out:
1133         BUG_ON(ret < 0 || m->nr_pages < want);
1134
1135         return ret;
1136 }
1137
1138 /*
1139  * Register request, send initial attempt.
1140  */
1141 int ceph_osdc_start_request(struct ceph_osd_client *osdc,
1142                             struct ceph_osd_request *req,
1143                             bool nofail)
1144 {
1145         int rc = 0;
1146
1147         req->r_request->pages = req->r_pages;
1148         req->r_request->nr_pages = req->r_num_pages;
1149
1150         register_request(osdc, req);
1151
1152         down_read(&osdc->map_sem);
1153         mutex_lock(&osdc->request_mutex);
1154         /*
1155          * a racing kick_requests() may have sent the message for us
1156          * while we dropped request_mutex above, so only send now if
1157          * the request still han't been touched yet.
1158          */
1159         if (req->r_sent == 0) {
1160                 rc = __send_request(osdc, req);
1161                 if (rc) {
1162                         if (nofail) {
1163                                 dout("osdc_start_request failed send, "
1164                                      " marking %lld\n", req->r_tid);
1165                                 req->r_resend = true;
1166                                 rc = 0;
1167                         } else {
1168                                 __unregister_request(osdc, req);
1169                         }
1170                 }
1171         }
1172         mutex_unlock(&osdc->request_mutex);
1173         up_read(&osdc->map_sem);
1174         return rc;
1175 }
1176
1177 /*
1178  * wait for a request to complete
1179  */
1180 int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
1181                            struct ceph_osd_request *req)
1182 {
1183         int rc;
1184
1185         rc = wait_for_completion_interruptible(&req->r_completion);
1186         if (rc < 0) {
1187                 mutex_lock(&osdc->request_mutex);
1188                 __cancel_request(req);
1189                 __unregister_request(osdc, req);
1190                 mutex_unlock(&osdc->request_mutex);
1191                 dout("wait_request tid %llu canceled/timed out\n", req->r_tid);
1192                 return rc;
1193         }
1194
1195         dout("wait_request tid %llu result %d\n", req->r_tid, req->r_result);
1196         return req->r_result;
1197 }
1198
1199 /*
1200  * sync - wait for all in-flight requests to flush.  avoid starvation.
1201  */
1202 void ceph_osdc_sync(struct ceph_osd_client *osdc)
1203 {
1204         struct ceph_osd_request *req;
1205         u64 last_tid, next_tid = 0;
1206
1207         mutex_lock(&osdc->request_mutex);
1208         last_tid = osdc->last_tid;
1209         while (1) {
1210                 req = __lookup_request_ge(osdc, next_tid);
1211                 if (!req)
1212                         break;
1213                 if (req->r_tid > last_tid)
1214                         break;
1215
1216                 next_tid = req->r_tid + 1;
1217                 if ((req->r_flags & CEPH_OSD_FLAG_WRITE) == 0)
1218                         continue;
1219
1220                 ceph_osdc_get_request(req);
1221                 mutex_unlock(&osdc->request_mutex);
1222                 dout("sync waiting on tid %llu (last is %llu)\n",
1223                      req->r_tid, last_tid);
1224                 wait_for_completion(&req->r_safe_completion);
1225                 mutex_lock(&osdc->request_mutex);
1226                 ceph_osdc_put_request(req);
1227         }
1228         mutex_unlock(&osdc->request_mutex);
1229         dout("sync done (thru tid %llu)\n", last_tid);
1230 }
1231
1232 /*
1233  * init, shutdown
1234  */
1235 int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client)
1236 {
1237         int err;
1238
1239         dout("init\n");
1240         osdc->client = client;
1241         osdc->osdmap = NULL;
1242         init_rwsem(&osdc->map_sem);
1243         init_completion(&osdc->map_waiters);
1244         osdc->last_requested_map = 0;
1245         mutex_init(&osdc->request_mutex);
1246         osdc->timeout_tid = 0;
1247         osdc->last_tid = 0;
1248         osdc->osds = RB_ROOT;
1249         INIT_LIST_HEAD(&osdc->osd_lru);
1250         osdc->requests = RB_ROOT;
1251         osdc->num_requests = 0;
1252         INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout);
1253         INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout);
1254
1255         schedule_delayed_work(&osdc->osds_timeout_work,
1256            round_jiffies_relative(osdc->client->mount_args->osd_idle_ttl * HZ));
1257
1258         err = -ENOMEM;
1259         osdc->req_mempool = mempool_create_kmalloc_pool(10,
1260                                         sizeof(struct ceph_osd_request));
1261         if (!osdc->req_mempool)
1262                 goto out;
1263
1264         err = ceph_msgpool_init(&osdc->msgpool_op, 4096, 10, true);
1265         if (err < 0)
1266                 goto out_mempool;
1267         return 0;
1268
1269 out_mempool:
1270         mempool_destroy(osdc->req_mempool);
1271 out:
1272         return err;
1273 }
1274
1275 void ceph_osdc_stop(struct ceph_osd_client *osdc)
1276 {
1277         cancel_delayed_work_sync(&osdc->timeout_work);
1278         cancel_delayed_work_sync(&osdc->osds_timeout_work);
1279         if (osdc->osdmap) {
1280                 ceph_osdmap_destroy(osdc->osdmap);
1281                 osdc->osdmap = NULL;
1282         }
1283         remove_old_osds(osdc, 1);
1284         mempool_destroy(osdc->req_mempool);
1285         ceph_msgpool_destroy(&osdc->msgpool_op);
1286 }
1287
1288 /*
1289  * Read some contiguous pages.  If we cross a stripe boundary, shorten
1290  * *plen.  Return number of bytes read, or error.
1291  */
1292 int ceph_osdc_readpages(struct ceph_osd_client *osdc,
1293                         struct ceph_vino vino, struct ceph_file_layout *layout,
1294                         u64 off, u64 *plen,
1295                         u32 truncate_seq, u64 truncate_size,
1296                         struct page **pages, int num_pages)
1297 {
1298         struct ceph_osd_request *req;
1299         int rc = 0;
1300
1301         dout("readpages on ino %llx.%llx on %llu~%llu\n", vino.ino,
1302              vino.snap, off, *plen);
1303         req = ceph_osdc_new_request(osdc, layout, vino, off, plen,
1304                                     CEPH_OSD_OP_READ, CEPH_OSD_FLAG_READ,
1305                                     NULL, 0, truncate_seq, truncate_size, NULL,
1306                                     false, 1);
1307         if (IS_ERR(req))
1308                 return PTR_ERR(req);
1309
1310         /* it may be a short read due to an object boundary */
1311         req->r_pages = pages;
1312         num_pages = calc_pages_for(off, *plen);
1313         req->r_num_pages = num_pages;
1314
1315         dout("readpages  final extent is %llu~%llu (%d pages)\n",
1316              off, *plen, req->r_num_pages);
1317
1318         rc = ceph_osdc_start_request(osdc, req, false);
1319         if (!rc)
1320                 rc = ceph_osdc_wait_request(osdc, req);
1321
1322         ceph_osdc_put_request(req);
1323         dout("readpages result %d\n", rc);
1324         return rc;
1325 }
1326
1327 /*
1328  * do a synchronous write on N pages
1329  */
1330 int ceph_osdc_writepages(struct ceph_osd_client *osdc, struct ceph_vino vino,
1331                          struct ceph_file_layout *layout,
1332                          struct ceph_snap_context *snapc,
1333                          u64 off, u64 len,
1334                          u32 truncate_seq, u64 truncate_size,
1335                          struct timespec *mtime,
1336                          struct page **pages, int num_pages,
1337                          int flags, int do_sync, bool nofail)
1338 {
1339         struct ceph_osd_request *req;
1340         int rc = 0;
1341
1342         BUG_ON(vino.snap != CEPH_NOSNAP);
1343         req = ceph_osdc_new_request(osdc, layout, vino, off, &len,
1344                                     CEPH_OSD_OP_WRITE,
1345                                     flags | CEPH_OSD_FLAG_ONDISK |
1346                                             CEPH_OSD_FLAG_WRITE,
1347                                     snapc, do_sync,
1348                                     truncate_seq, truncate_size, mtime,
1349                                     nofail, 1);
1350         if (IS_ERR(req))
1351                 return PTR_ERR(req);
1352
1353         /* it may be a short write due to an object boundary */
1354         req->r_pages = pages;
1355         req->r_num_pages = calc_pages_for(off, len);
1356         dout("writepages %llu~%llu (%d pages)\n", off, len,
1357              req->r_num_pages);
1358
1359         rc = ceph_osdc_start_request(osdc, req, nofail);
1360         if (!rc)
1361                 rc = ceph_osdc_wait_request(osdc, req);
1362
1363         ceph_osdc_put_request(req);
1364         if (rc == 0)
1365                 rc = len;
1366         dout("writepages result %d\n", rc);
1367         return rc;
1368 }
1369
1370 /*
1371  * handle incoming message
1372  */
1373 static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
1374 {
1375         struct ceph_osd *osd = con->private;
1376         struct ceph_osd_client *osdc;
1377         int type = le16_to_cpu(msg->hdr.type);
1378
1379         if (!osd)
1380                 return;
1381         osdc = osd->o_osdc;
1382
1383         switch (type) {
1384         case CEPH_MSG_OSD_MAP:
1385                 ceph_osdc_handle_map(osdc, msg);
1386                 break;
1387         case CEPH_MSG_OSD_OPREPLY:
1388                 handle_reply(osdc, msg, con);
1389                 break;
1390
1391         default:
1392                 pr_err("received unknown message type %d %s\n", type,
1393                        ceph_msg_type_name(type));
1394         }
1395         ceph_msg_put(msg);
1396 }
1397
1398 static struct ceph_msg *alloc_msg(struct ceph_connection *con,
1399                                   struct ceph_msg_header *hdr,
1400                                   int *skip)
1401 {
1402         struct ceph_osd *osd = con->private;
1403         struct ceph_osd_client *osdc = osd->o_osdc;
1404         int type = le16_to_cpu(hdr->type);
1405         int front = le32_to_cpu(hdr->front_len);
1406         int data_len = le32_to_cpu(hdr->data_len);
1407         struct ceph_msg *m;
1408         struct ceph_osd_request *req;
1409         u64 tid;
1410         int err;
1411
1412         *skip = 0;
1413         if (type != CEPH_MSG_OSD_OPREPLY)
1414                 return NULL;
1415
1416         tid = le64_to_cpu(hdr->tid);
1417         mutex_lock(&osdc->request_mutex);
1418         req = __lookup_request(osdc, tid);
1419         if (!req) {
1420                 *skip = 1;
1421                 m = NULL;
1422                 dout("alloc_msg unknown tid %llu\n", tid);
1423                 goto out;
1424         }
1425         m = __get_next_reply(con, req, front);
1426         if (!m || IS_ERR(m)) {
1427                 *skip = 1;
1428                 goto out;
1429         }
1430
1431         if (data_len > 0) {
1432                 err = __prepare_pages(con, hdr, req, tid, m);
1433                 if (err < 0) {
1434                         *skip = 1;
1435                         ceph_msg_put(m);
1436                         m = ERR_PTR(err);
1437                 }
1438         }
1439
1440 out:
1441         mutex_unlock(&osdc->request_mutex);
1442
1443         return m;
1444 }
1445
1446 /*
1447  * Wrappers to refcount containing ceph_osd struct
1448  */
1449 static struct ceph_connection *get_osd_con(struct ceph_connection *con)
1450 {
1451         struct ceph_osd *osd = con->private;
1452         if (get_osd(osd))
1453                 return con;
1454         return NULL;
1455 }
1456
1457 static void put_osd_con(struct ceph_connection *con)
1458 {
1459         struct ceph_osd *osd = con->private;
1460         put_osd(osd);
1461 }
1462
1463 /*
1464  * authentication
1465  */
1466 static int get_authorizer(struct ceph_connection *con,
1467                           void **buf, int *len, int *proto,
1468                           void **reply_buf, int *reply_len, int force_new)
1469 {
1470         struct ceph_osd *o = con->private;
1471         struct ceph_osd_client *osdc = o->o_osdc;
1472         struct ceph_auth_client *ac = osdc->client->monc.auth;
1473         int ret = 0;
1474
1475         if (force_new && o->o_authorizer) {
1476                 ac->ops->destroy_authorizer(ac, o->o_authorizer);
1477                 o->o_authorizer = NULL;
1478         }
1479         if (o->o_authorizer == NULL) {
1480                 ret = ac->ops->create_authorizer(
1481                         ac, CEPH_ENTITY_TYPE_OSD,
1482                         &o->o_authorizer,
1483                         &o->o_authorizer_buf,
1484                         &o->o_authorizer_buf_len,
1485                         &o->o_authorizer_reply_buf,
1486                         &o->o_authorizer_reply_buf_len);
1487                 if (ret)
1488                 return ret;
1489         }
1490
1491         *proto = ac->protocol;
1492         *buf = o->o_authorizer_buf;
1493         *len = o->o_authorizer_buf_len;
1494         *reply_buf = o->o_authorizer_reply_buf;
1495         *reply_len = o->o_authorizer_reply_buf_len;
1496         return 0;
1497 }
1498
1499
1500 static int verify_authorizer_reply(struct ceph_connection *con, int len)
1501 {
1502         struct ceph_osd *o = con->private;
1503         struct ceph_osd_client *osdc = o->o_osdc;
1504         struct ceph_auth_client *ac = osdc->client->monc.auth;
1505
1506         return ac->ops->verify_authorizer_reply(ac, o->o_authorizer, len);
1507 }
1508
1509 static int invalidate_authorizer(struct ceph_connection *con)
1510 {
1511         struct ceph_osd *o = con->private;
1512         struct ceph_osd_client *osdc = o->o_osdc;
1513         struct ceph_auth_client *ac = osdc->client->monc.auth;
1514
1515         if (ac->ops->invalidate_authorizer)
1516                 ac->ops->invalidate_authorizer(ac, CEPH_ENTITY_TYPE_OSD);
1517
1518         return ceph_monc_validate_auth(&osdc->client->monc);
1519 }
1520
1521 const static struct ceph_connection_operations osd_con_ops = {
1522         .get = get_osd_con,
1523         .put = put_osd_con,
1524         .dispatch = dispatch,
1525         .get_authorizer = get_authorizer,
1526         .verify_authorizer_reply = verify_authorizer_reply,
1527         .invalidate_authorizer = invalidate_authorizer,
1528         .alloc_msg = alloc_msg,
1529         .fault = osd_reset,
1530 };