net/ceph/osd_client.c

   1 // SPDX-License-Identifier: GPL-2.0
   2
   3 #include <linux/ceph/ceph_debug.h>
   4
   5 #include <linux/module.h>
   6 #include <linux/err.h>
   7 #include <linux/highmem.h>
   8 #include <linux/mm.h>
   9 #include <linux/pagemap.h>
  10 #include <linux/slab.h>
  11 #include <linux/uaccess.h>
  12 #ifdef CONFIG_BLOCK
  13 #include <linux/bio.h>
  14 #endif
  15
  16 #include <linux/ceph/ceph_features.h>
  17 #include <linux/ceph/libceph.h>
  18 #include <linux/ceph/osd_client.h>
  19 #include <linux/ceph/messenger.h>
  20 #include <linux/ceph/decode.h>
  21 #include <linux/ceph/auth.h>
  22 #include <linux/ceph/pagelist.h>
  23 #include <linux/ceph/striper.h>
  24
  25 #define OSD_OPREPLY_FRONT_LEN   512
  26
  27 static struct kmem_cache        *ceph_osd_request_cache;
  28
  29 static const struct ceph_connection_operations osd_con_ops;
  30
  31 /*
  32  * Implement client access to distributed object storage cluster.
  33  *
  34  * All data objects are stored within a cluster/cloud of OSDs, or
  35  * "object storage devices."  (Note that Ceph OSDs have _nothing_ to
  36  * do with the T10 OSD extensions to SCSI.)  Ceph OSDs are simply
  37  * remote daemons serving up and coordinating consistent and safe
  38  * access to storage.
  39  *
  40  * Cluster membership and the mapping of data objects onto storage devices
  41  * are described by the osd map.
  42  *
  43  * We keep track of pending OSD requests (read, write), resubmit
  44  * requests to different OSDs when the cluster topology/data layout
  45  * change, or retry the affected requests when the communications
  46  * channel with an OSD is reset.
  47  */
  48
  49 static void link_request(struct ceph_osd *osd, struct ceph_osd_request *req);
  50 static void unlink_request(struct ceph_osd *osd, struct ceph_osd_request *req);
  51 static void link_linger(struct ceph_osd *osd,
  52                         struct ceph_osd_linger_request *lreq);
  53 static void unlink_linger(struct ceph_osd *osd,
  54                           struct ceph_osd_linger_request *lreq);
  55 static void clear_backoffs(struct ceph_osd *osd);
  56
  57 #if 1
  58 static inline bool rwsem_is_wrlocked(struct rw_semaphore *sem)
  59 {
  60         bool wrlocked = true;
  61
  62         if (unlikely(down_read_trylock(sem))) {
  63                 wrlocked = false;
  64                 up_read(sem);
  65         }
  66
  67         return wrlocked;
  68 }
  69 static inline void verify_osdc_locked(struct ceph_osd_client *osdc)
  70 {
  71         WARN_ON(!rwsem_is_locked(&osdc->lock));
  72 }
  73 static inline void verify_osdc_wrlocked(struct ceph_osd_client *osdc)
  74 {
  75         WARN_ON(!rwsem_is_wrlocked(&osdc->lock));
  76 }
  77 static inline void verify_osd_locked(struct ceph_osd *osd)
  78 {
  79         struct ceph_osd_client *osdc = osd->o_osdc;
  80
  81         WARN_ON(!(mutex_is_locked(&osd->lock) &&
  82                   rwsem_is_locked(&osdc->lock)) &&
  83                 !rwsem_is_wrlocked(&osdc->lock));
  84 }
  85 static inline void verify_lreq_locked(struct ceph_osd_linger_request *lreq)
  86 {
  87         WARN_ON(!mutex_is_locked(&lreq->lock));
  88 }
  89 #else
  90 static inline void verify_osdc_locked(struct ceph_osd_client *osdc) { }
  91 static inline void verify_osdc_wrlocked(struct ceph_osd_client *osdc) { }
  92 static inline void verify_osd_locked(struct ceph_osd *osd) { }
  93 static inline void verify_lreq_locked(struct ceph_osd_linger_request *lreq) { }
  94 #endif
  95
  96 /*
  97  * calculate the mapping of a file extent onto an object, and fill out the
  98  * request accordingly.  shorten extent as necessary if it crosses an
  99  * object boundary.
 100  *
 101  * fill osd op in request message.
 102  */
 103 static int calc_layout(struct ceph_file_layout *layout, u64 off, u64 *plen,
 104                         u64 *objnum, u64 *objoff, u64 *objlen)
 105 {
 106         u64 orig_len = *plen;
 107         u32 xlen;
 108
 109         /* object extent? */
 110         ceph_calc_file_object_mapping(layout, off, orig_len, objnum,
 111                                           objoff, &xlen);
 112         *objlen = xlen;
 113         if (*objlen < orig_len) {
 114                 *plen = *objlen;
 115                 dout(" skipping last %llu, final file extent %llu~%llu\n",
 116                      orig_len - *plen, off, *plen);
 117         }
 118
 119         dout("calc_layout objnum=%llx %llu~%llu\n", *objnum, *objoff, *objlen);
 120         return 0;
 121 }
 122
 123 static void ceph_osd_data_init(struct ceph_osd_data *osd_data)
 124 {
 125         memset(osd_data, 0, sizeof (*osd_data));
 126         osd_data->type = CEPH_OSD_DATA_TYPE_NONE;
 127 }
 128
 129 /*
 130  * Consumes @pages if @own_pages is true.
 131  */
 132 static void ceph_osd_data_pages_init(struct ceph_osd_data *osd_data,
 133                         struct page **pages, u64 length, u32 alignment,
 134                         bool pages_from_pool, bool own_pages)
 135 {
 136         osd_data->type = CEPH_OSD_DATA_TYPE_PAGES;
 137         osd_data->pages = pages;
 138         osd_data->length = length;
 139         osd_data->alignment = alignment;
 140         osd_data->pages_from_pool = pages_from_pool;
 141         osd_data->own_pages = own_pages;
 142 }
 143
 144 /*
 145  * Consumes a ref on @pagelist.
 146  */
 147 static void ceph_osd_data_pagelist_init(struct ceph_osd_data *osd_data,
 148                         struct ceph_pagelist *pagelist)
 149 {
 150         osd_data->type = CEPH_OSD_DATA_TYPE_PAGELIST;
 151         osd_data->pagelist = pagelist;
 152 }
 153
 154 #ifdef CONFIG_BLOCK
 155 static void ceph_osd_data_bio_init(struct ceph_osd_data *osd_data,
 156                                    struct ceph_bio_iter *bio_pos,
 157                                    u32 bio_length)
 158 {
 159         osd_data->type = CEPH_OSD_DATA_TYPE_BIO;
 160         osd_data->bio_pos = *bio_pos;
 161         osd_data->bio_length = bio_length;
 162 }
 163 #endif /* CONFIG_BLOCK */
 164
 165 static void ceph_osd_data_bvecs_init(struct ceph_osd_data *osd_data,
 166                                      struct ceph_bvec_iter *bvec_pos,
 167                                      u32 num_bvecs)
 168 {
 169         osd_data->type = CEPH_OSD_DATA_TYPE_BVECS;
 170         osd_data->bvec_pos = *bvec_pos;
 171         osd_data->num_bvecs = num_bvecs;
 172 }
 173
 174 static struct ceph_osd_data *
 175 osd_req_op_raw_data_in(struct ceph_osd_request *osd_req, unsigned int which)
 176 {
 177         BUG_ON(which >= osd_req->r_num_ops);
 178
 179         return &osd_req->r_ops[which].raw_data_in;
 180 }
 181
 182 struct ceph_osd_data *
 183 osd_req_op_extent_osd_data(struct ceph_osd_request *osd_req,
 184                         unsigned int which)
 185 {
 186         return osd_req_op_data(osd_req, which, extent, osd_data);
 187 }
 188 EXPORT_SYMBOL(osd_req_op_extent_osd_data);
 189
 190 void osd_req_op_raw_data_in_pages(struct ceph_osd_request *osd_req,
 191                         unsigned int which, struct page **pages,
 192                         u64 length, u32 alignment,
 193                         bool pages_from_pool, bool own_pages)
 194 {
 195         struct ceph_osd_data *osd_data;
 196
 197         osd_data = osd_req_op_raw_data_in(osd_req, which);
 198         ceph_osd_data_pages_init(osd_data, pages, length, alignment,
 199                                 pages_from_pool, own_pages);
 200 }
 201 EXPORT_SYMBOL(osd_req_op_raw_data_in_pages);
 202
 203 void osd_req_op_extent_osd_data_pages(struct ceph_osd_request *osd_req,
 204                         unsigned int which, struct page **pages,
 205                         u64 length, u32 alignment,
 206                         bool pages_from_pool, bool own_pages)
 207 {
 208         struct ceph_osd_data *osd_data;
 209
 210         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 211         ceph_osd_data_pages_init(osd_data, pages, length, alignment,
 212                                 pages_from_pool, own_pages);
 213 }
 214 EXPORT_SYMBOL(osd_req_op_extent_osd_data_pages);
 215
 216 void osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request *osd_req,
 217                         unsigned int which, struct ceph_pagelist *pagelist)
 218 {
 219         struct ceph_osd_data *osd_data;
 220
 221         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 222         ceph_osd_data_pagelist_init(osd_data, pagelist);
 223 }
 224 EXPORT_SYMBOL(osd_req_op_extent_osd_data_pagelist);
 225
 226 #ifdef CONFIG_BLOCK
 227 void osd_req_op_extent_osd_data_bio(struct ceph_osd_request *osd_req,
 228                                     unsigned int which,
 229                                     struct ceph_bio_iter *bio_pos,
 230                                     u32 bio_length)
 231 {
 232         struct ceph_osd_data *osd_data;
 233
 234         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 235         ceph_osd_data_bio_init(osd_data, bio_pos, bio_length);
 236 }
 237 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bio);
 238 #endif /* CONFIG_BLOCK */
 239
 240 void osd_req_op_extent_osd_data_bvecs(struct ceph_osd_request *osd_req,
 241                                       unsigned int which,
 242                                       struct bio_vec *bvecs, u32 num_bvecs,
 243                                       u32 bytes)
 244 {
 245         struct ceph_osd_data *osd_data;
 246         struct ceph_bvec_iter it = {
 247                 .bvecs = bvecs,
 248                 .iter = { .bi_size = bytes },
 249         };
 250
 251         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 252         ceph_osd_data_bvecs_init(osd_data, &it, num_bvecs);
 253 }
 254 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvecs);
 255
 256 void osd_req_op_extent_osd_data_bvec_pos(struct ceph_osd_request *osd_req,
 257                                          unsigned int which,
 258                                          struct ceph_bvec_iter *bvec_pos)
 259 {
 260         struct ceph_osd_data *osd_data;
 261
 262         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 263         ceph_osd_data_bvecs_init(osd_data, bvec_pos, 0);
 264 }
 265 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvec_pos);
 266
 267 static void osd_req_op_cls_request_info_pagelist(
 268                         struct ceph_osd_request *osd_req,
 269                         unsigned int which, struct ceph_pagelist *pagelist)
 270 {
 271         struct ceph_osd_data *osd_data;
 272
 273         osd_data = osd_req_op_data(osd_req, which, cls, request_info);
 274         ceph_osd_data_pagelist_init(osd_data, pagelist);
 275 }
 276
 277 void osd_req_op_cls_request_data_pagelist(
 278                         struct ceph_osd_request *osd_req,
 279                         unsigned int which, struct ceph_pagelist *pagelist)
 280 {
 281         struct ceph_osd_data *osd_data;
 282
 283         osd_data = osd_req_op_data(osd_req, which, cls, request_data);
 284         ceph_osd_data_pagelist_init(osd_data, pagelist);
 285         osd_req->r_ops[which].cls.indata_len += pagelist->length;
 286         osd_req->r_ops[which].indata_len += pagelist->length;
 287 }
 288 EXPORT_SYMBOL(osd_req_op_cls_request_data_pagelist);
 289
 290 void osd_req_op_cls_request_data_pages(struct ceph_osd_request *osd_req,
 291                         unsigned int which, struct page **pages, u64 length,
 292                         u32 alignment, bool pages_from_pool, bool own_pages)
 293 {
 294         struct ceph_osd_data *osd_data;
 295
 296         osd_data = osd_req_op_data(osd_req, which, cls, request_data);
 297         ceph_osd_data_pages_init(osd_data, pages, length, alignment,
 298                                 pages_from_pool, own_pages);
 299         osd_req->r_ops[which].cls.indata_len += length;
 300         osd_req->r_ops[which].indata_len += length;
 301 }
 302 EXPORT_SYMBOL(osd_req_op_cls_request_data_pages);
 303
 304 void osd_req_op_cls_request_data_bvecs(struct ceph_osd_request *osd_req,
 305                                        unsigned int which,
 306                                        struct bio_vec *bvecs, u32 num_bvecs,
 307                                        u32 bytes)
 308 {
 309         struct ceph_osd_data *osd_data;
 310         struct ceph_bvec_iter it = {
 311                 .bvecs = bvecs,
 312                 .iter = { .bi_size = bytes },
 313         };
 314
 315         osd_data = osd_req_op_data(osd_req, which, cls, request_data);
 316         ceph_osd_data_bvecs_init(osd_data, &it, num_bvecs);
 317         osd_req->r_ops[which].cls.indata_len += bytes;
 318         osd_req->r_ops[which].indata_len += bytes;
 319 }
 320 EXPORT_SYMBOL(osd_req_op_cls_request_data_bvecs);
 321
 322 void osd_req_op_cls_response_data_pages(struct ceph_osd_request *osd_req,
 323                         unsigned int which, struct page **pages, u64 length,
 324                         u32 alignment, bool pages_from_pool, bool own_pages)
 325 {
 326         struct ceph_osd_data *osd_data;
 327
 328         osd_data = osd_req_op_data(osd_req, which, cls, response_data);
 329         ceph_osd_data_pages_init(osd_data, pages, length, alignment,
 330                                 pages_from_pool, own_pages);
 331 }
 332 EXPORT_SYMBOL(osd_req_op_cls_response_data_pages);
 333
 334 static u64 ceph_osd_data_length(struct ceph_osd_data *osd_data)
 335 {
 336         switch (osd_data->type) {
 337         case CEPH_OSD_DATA_TYPE_NONE:
 338                 return 0;
 339         case CEPH_OSD_DATA_TYPE_PAGES:
 340                 return osd_data->length;
 341         case CEPH_OSD_DATA_TYPE_PAGELIST:
 342                 return (u64)osd_data->pagelist->length;
 343 #ifdef CONFIG_BLOCK
 344         case CEPH_OSD_DATA_TYPE_BIO:
 345                 return (u64)osd_data->bio_length;
 346 #endif /* CONFIG_BLOCK */
 347         case CEPH_OSD_DATA_TYPE_BVECS:
 348                 return osd_data->bvec_pos.iter.bi_size;
 349         default:
 350                 WARN(true, "unrecognized data type %d\n", (int)osd_data->type);
 351                 return 0;
 352         }
 353 }
 354
 355 static void ceph_osd_data_release(struct ceph_osd_data *osd_data)
 356 {
 357         if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES && osd_data->own_pages) {
 358                 int num_pages;
 359
 360                 num_pages = calc_pages_for((u64)osd_data->alignment,
 361                                                 (u64)osd_data->length);
 362                 ceph_release_page_vector(osd_data->pages, num_pages);
 363         } else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
 364                 ceph_pagelist_release(osd_data->pagelist);
 365         }
 366         ceph_osd_data_init(osd_data);
 367 }
 368
 369 static void osd_req_op_data_release(struct ceph_osd_request *osd_req,
 370                         unsigned int which)
 371 {
 372         struct ceph_osd_req_op *op;
 373
 374         BUG_ON(which >= osd_req->r_num_ops);
 375         op = &osd_req->r_ops[which];
 376
 377         switch (op->op) {
 378         case CEPH_OSD_OP_READ:
 379         case CEPH_OSD_OP_WRITE:
 380         case CEPH_OSD_OP_WRITEFULL:
 381                 ceph_osd_data_release(&op->extent.osd_data);
 382                 break;
 383         case CEPH_OSD_OP_CALL:
 384                 ceph_osd_data_release(&op->cls.request_info);
 385                 ceph_osd_data_release(&op->cls.request_data);
 386                 ceph_osd_data_release(&op->cls.response_data);
 387                 break;
 388         case CEPH_OSD_OP_SETXATTR:
 389         case CEPH_OSD_OP_CMPXATTR:
 390                 ceph_osd_data_release(&op->xattr.osd_data);
 391                 break;
 392         case CEPH_OSD_OP_STAT:
 393                 ceph_osd_data_release(&op->raw_data_in);
 394                 break;
 395         case CEPH_OSD_OP_NOTIFY_ACK:
 396                 ceph_osd_data_release(&op->notify_ack.request_data);
 397                 break;
 398         case CEPH_OSD_OP_NOTIFY:
 399                 ceph_osd_data_release(&op->notify.request_data);
 400                 ceph_osd_data_release(&op->notify.response_data);
 401                 break;
 402         case CEPH_OSD_OP_LIST_WATCHERS:
 403                 ceph_osd_data_release(&op->list_watchers.response_data);
 404                 break;
 405         case CEPH_OSD_OP_COPY_FROM:
 406                 ceph_osd_data_release(&op->copy_from.osd_data);
 407                 break;
 408         default:
 409                 break;
 410         }
 411 }
 412
 413 /*
 414  * Assumes @t is zero-initialized.
 415  */
 416 static void target_init(struct ceph_osd_request_target *t)
 417 {
 418         ceph_oid_init(&t->base_oid);
 419         ceph_oloc_init(&t->base_oloc);
 420         ceph_oid_init(&t->target_oid);
 421         ceph_oloc_init(&t->target_oloc);
 422
 423         ceph_osds_init(&t->acting);
 424         ceph_osds_init(&t->up);
 425         t->size = -1;
 426         t->min_size = -1;
 427
 428         t->osd = CEPH_HOMELESS_OSD;
 429 }
 430
 431 static void target_copy(struct ceph_osd_request_target *dest,
 432                         const struct ceph_osd_request_target *src)
 433 {
 434         ceph_oid_copy(&dest->base_oid, &src->base_oid);
 435         ceph_oloc_copy(&dest->base_oloc, &src->base_oloc);
 436         ceph_oid_copy(&dest->target_oid, &src->target_oid);
 437         ceph_oloc_copy(&dest->target_oloc, &src->target_oloc);
 438
 439         dest->pgid = src->pgid; /* struct */
 440         dest->spgid = src->spgid; /* struct */
 441         dest->pg_num = src->pg_num;
 442         dest->pg_num_mask = src->pg_num_mask;
 443         ceph_osds_copy(&dest->acting, &src->acting);
 444         ceph_osds_copy(&dest->up, &src->up);
 445         dest->size = src->size;
 446         dest->min_size = src->min_size;
 447         dest->sort_bitwise = src->sort_bitwise;
 448
 449         dest->flags = src->flags;
 450         dest->paused = src->paused;
 451
 452         dest->epoch = src->epoch;
 453         dest->last_force_resend = src->last_force_resend;
 454
 455         dest->osd = src->osd;
 456 }
 457
 458 static void target_destroy(struct ceph_osd_request_target *t)
 459 {
 460         ceph_oid_destroy(&t->base_oid);
 461         ceph_oloc_destroy(&t->base_oloc);
 462         ceph_oid_destroy(&t->target_oid);
 463         ceph_oloc_destroy(&t->target_oloc);
 464 }
 465
 466 /*
 467  * requests
 468  */
 469 static void request_release_checks(struct ceph_osd_request *req)
 470 {
 471         WARN_ON(!RB_EMPTY_NODE(&req->r_node));
 472         WARN_ON(!RB_EMPTY_NODE(&req->r_mc_node));
 473         WARN_ON(!list_empty(&req->r_private_item));
 474         WARN_ON(req->r_osd);
 475 }
 476
 477 static void ceph_osdc_release_request(struct kref *kref)
 478 {
 479         struct ceph_osd_request *req = container_of(kref,
 480                                             struct ceph_osd_request, r_kref);
 481         unsigned int which;
 482
 483         dout("%s %p (r_request %p r_reply %p)\n", __func__, req,
 484              req->r_request, req->r_reply);
 485         request_release_checks(req);
 486
 487         if (req->r_request)
 488                 ceph_msg_put(req->r_request);
 489         if (req->r_reply)
 490                 ceph_msg_put(req->r_reply);
 491
 492         for (which = 0; which < req->r_num_ops; which++)
 493                 osd_req_op_data_release(req, which);
 494
 495         target_destroy(&req->r_t);
 496         ceph_put_snap_context(req->r_snapc);
 497
 498         if (req->r_mempool)
 499                 mempool_free(req, req->r_osdc->req_mempool);
 500         else if (req->r_num_ops <= CEPH_OSD_SLAB_OPS)
 501                 kmem_cache_free(ceph_osd_request_cache, req);
 502         else
 503                 kfree(req);
 504 }
 505
 506 void ceph_osdc_get_request(struct ceph_osd_request *req)
 507 {
 508         dout("%s %p (was %d)\n", __func__, req,
 509              kref_read(&req->r_kref));
 510         kref_get(&req->r_kref);
 511 }
 512 EXPORT_SYMBOL(ceph_osdc_get_request);
 513
 514 void ceph_osdc_put_request(struct ceph_osd_request *req)
 515 {
 516         if (req) {
 517                 dout("%s %p (was %d)\n", __func__, req,
 518                      kref_read(&req->r_kref));
 519                 kref_put(&req->r_kref, ceph_osdc_release_request);
 520         }
 521 }
 522 EXPORT_SYMBOL(ceph_osdc_put_request);
 523
 524 static void request_init(struct ceph_osd_request *req)
 525 {
 526         /* req only, each op is zeroed in _osd_req_op_init() */
 527         memset(req, 0, sizeof(*req));
 528
 529         kref_init(&req->r_kref);
 530         init_completion(&req->r_completion);
 531         RB_CLEAR_NODE(&req->r_node);
 532         RB_CLEAR_NODE(&req->r_mc_node);
 533         INIT_LIST_HEAD(&req->r_private_item);
 534
 535         target_init(&req->r_t);
 536 }
 537
 538 /*
 539  * This is ugly, but it allows us to reuse linger registration and ping
 540  * requests, keeping the structure of the code around send_linger{_ping}()
 541  * reasonable.  Setting up a min_nr=2 mempool for each linger request
 542  * and dealing with copying ops (this blasts req only, watch op remains
 543  * intact) isn't any better.
 544  */
 545 static void request_reinit(struct ceph_osd_request *req)
 546 {
 547         struct ceph_osd_client *osdc = req->r_osdc;
 548         bool mempool = req->r_mempool;
 549         unsigned int num_ops = req->r_num_ops;
 550         u64 snapid = req->r_snapid;
 551         struct ceph_snap_context *snapc = req->r_snapc;
 552         bool linger = req->r_linger;
 553         struct ceph_msg *request_msg = req->r_request;
 554         struct ceph_msg *reply_msg = req->r_reply;
 555
 556         dout("%s req %p\n", __func__, req);
 557         WARN_ON(kref_read(&req->r_kref) != 1);
 558         request_release_checks(req);
 559
 560         WARN_ON(kref_read(&request_msg->kref) != 1);
 561         WARN_ON(kref_read(&reply_msg->kref) != 1);
 562         target_destroy(&req->r_t);
 563
 564         request_init(req);
 565         req->r_osdc = osdc;
 566         req->r_mempool = mempool;
 567         req->r_num_ops = num_ops;
 568         req->r_snapid = snapid;
 569         req->r_snapc = snapc;
 570         req->r_linger = linger;
 571         req->r_request = request_msg;
 572         req->r_reply = reply_msg;
 573 }
 574
 575 struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
 576                                                struct ceph_snap_context *snapc,
 577                                                unsigned int num_ops,
 578                                                bool use_mempool,
 579                                                gfp_t gfp_flags)
 580 {
 581         struct ceph_osd_request *req;
 582
 583         if (use_mempool) {
 584                 BUG_ON(num_ops > CEPH_OSD_SLAB_OPS);
 585                 req = mempool_alloc(osdc->req_mempool, gfp_flags);
 586         } else if (num_ops <= CEPH_OSD_SLAB_OPS) {
 587                 req = kmem_cache_alloc(ceph_osd_request_cache, gfp_flags);
 588         } else {
 589                 BUG_ON(num_ops > CEPH_OSD_MAX_OPS);
 590                 req = kmalloc(struct_size(req, r_ops, num_ops), gfp_flags);
 591         }
 592         if (unlikely(!req))
 593                 return NULL;
 594
 595         request_init(req);
 596         req->r_osdc = osdc;
 597         req->r_mempool = use_mempool;
 598         req->r_num_ops = num_ops;
 599         req->r_snapid = CEPH_NOSNAP;
 600         req->r_snapc = ceph_get_snap_context(snapc);
 601
 602         dout("%s req %p\n", __func__, req);
 603         return req;
 604 }
 605 EXPORT_SYMBOL(ceph_osdc_alloc_request);
 606
 607 static int ceph_oloc_encoding_size(const struct ceph_object_locator *oloc)
 608 {
 609         return 8 + 4 + 4 + 4 + (oloc->pool_ns ? oloc->pool_ns->len : 0);
 610 }
 611
 612 static int __ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp,
 613                                       int num_request_data_items,
 614                                       int num_reply_data_items)
 615 {
 616         struct ceph_osd_client *osdc = req->r_osdc;
 617         struct ceph_msg *msg;
 618         int msg_size;
 619
 620         WARN_ON(req->r_request || req->r_reply);
 621         WARN_ON(ceph_oid_empty(&req->r_base_oid));
 622         WARN_ON(ceph_oloc_empty(&req->r_base_oloc));
 623
 624         /* create request message */
 625         msg_size = CEPH_ENCODING_START_BLK_LEN +
 626                         CEPH_PGID_ENCODING_LEN + 1; /* spgid */
 627         msg_size += 4 + 4 + 4; /* hash, osdmap_epoch, flags */
 628         msg_size += CEPH_ENCODING_START_BLK_LEN +
 629                         sizeof(struct ceph_osd_reqid); /* reqid */
 630         msg_size += sizeof(struct ceph_blkin_trace_info); /* trace */
 631         msg_size += 4 + sizeof(struct ceph_timespec); /* client_inc, mtime */
 632         msg_size += CEPH_ENCODING_START_BLK_LEN +
 633                         ceph_oloc_encoding_size(&req->r_base_oloc); /* oloc */
 634         msg_size += 4 + req->r_base_oid.name_len; /* oid */
 635         msg_size += 2 + req->r_num_ops * sizeof(struct ceph_osd_op);
 636         msg_size += 8; /* snapid */
 637         msg_size += 8; /* snap_seq */
 638         msg_size += 4 + 8 * (req->r_snapc ? req->r_snapc->num_snaps : 0);
 639         msg_size += 4 + 8; /* retry_attempt, features */
 640
 641         if (req->r_mempool)
 642                 msg = ceph_msgpool_get(&osdc->msgpool_op, msg_size,
 643                                        num_request_data_items);
 644         else
 645                 msg = ceph_msg_new2(CEPH_MSG_OSD_OP, msg_size,
 646                                     num_request_data_items, gfp, true);
 647         if (!msg)
 648                 return -ENOMEM;
 649
 650         memset(msg->front.iov_base, 0, msg->front.iov_len);
 651         req->r_request = msg;
 652
 653         /* create reply message */
 654         msg_size = OSD_OPREPLY_FRONT_LEN;
 655         msg_size += req->r_base_oid.name_len;
 656         msg_size += req->r_num_ops * sizeof(struct ceph_osd_op);
 657
 658         if (req->r_mempool)
 659                 msg = ceph_msgpool_get(&osdc->msgpool_op_reply, msg_size,
 660                                        num_reply_data_items);
 661         else
 662                 msg = ceph_msg_new2(CEPH_MSG_OSD_OPREPLY, msg_size,
 663                                     num_reply_data_items, gfp, true);
 664         if (!msg)
 665                 return -ENOMEM;
 666
 667         req->r_reply = msg;
 668
 669         return 0;
 670 }
 671
 672 static bool osd_req_opcode_valid(u16 opcode)
 673 {
 674         switch (opcode) {
 675 #define GENERATE_CASE(op, opcode, str)  case CEPH_OSD_OP_##op: return true;
 676 __CEPH_FORALL_OSD_OPS(GENERATE_CASE)
 677 #undef GENERATE_CASE
 678         default:
 679                 return false;
 680         }
 681 }
 682
 683 static void get_num_data_items(struct ceph_osd_request *req,
 684                                int *num_request_data_items,
 685                                int *num_reply_data_items)
 686 {
 687         struct ceph_osd_req_op *op;
 688
 689         *num_request_data_items = 0;
 690         *num_reply_data_items = 0;
 691
 692         for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) {
 693                 switch (op->op) {
 694                 /* request */
 695                 case CEPH_OSD_OP_WRITE:
 696                 case CEPH_OSD_OP_WRITEFULL:
 697                 case CEPH_OSD_OP_SETXATTR:
 698                 case CEPH_OSD_OP_CMPXATTR:
 699                 case CEPH_OSD_OP_NOTIFY_ACK:
 700                 case CEPH_OSD_OP_COPY_FROM:
 701                         *num_request_data_items += 1;
 702                         break;
 703
 704                 /* reply */
 705                 case CEPH_OSD_OP_STAT:
 706                 case CEPH_OSD_OP_READ:
 707                 case CEPH_OSD_OP_LIST_WATCHERS:
 708                         *num_reply_data_items += 1;
 709                         break;
 710
 711                 /* both */
 712                 case CEPH_OSD_OP_NOTIFY:
 713                         *num_request_data_items += 1;
 714                         *num_reply_data_items += 1;
 715                         break;
 716                 case CEPH_OSD_OP_CALL:
 717                         *num_request_data_items += 2;
 718                         *num_reply_data_items += 1;
 719                         break;
 720
 721                 default:
 722                         WARN_ON(!osd_req_opcode_valid(op->op));
 723                         break;
 724                 }
 725         }
 726 }
 727
 728 /*
 729  * oid, oloc and OSD op opcode(s) must be filled in before this function
 730  * is called.
 731  */
 732 int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp)
 733 {
 734         int num_request_data_items, num_reply_data_items;
 735
 736         get_num_data_items(req, &num_request_data_items, &num_reply_data_items);
 737         return __ceph_osdc_alloc_messages(req, gfp, num_request_data_items,
 738                                           num_reply_data_items);
 739 }
 740 EXPORT_SYMBOL(ceph_osdc_alloc_messages);
 741
 742 /*
 743  * This is an osd op init function for opcodes that have no data or
 744  * other information associated with them.  It also serves as a
 745  * common init routine for all the other init functions, below.
 746  */
 747 static struct ceph_osd_req_op *
 748 _osd_req_op_init(struct ceph_osd_request *osd_req, unsigned int which,
 749                  u16 opcode, u32 flags)
 750 {
 751         struct ceph_osd_req_op *op;
 752
 753         BUG_ON(which >= osd_req->r_num_ops);
 754         BUG_ON(!osd_req_opcode_valid(opcode));
 755
 756         op = &osd_req->r_ops[which];
 757         memset(op, 0, sizeof (*op));
 758         op->op = opcode;
 759         op->flags = flags;
 760
 761         return op;
 762 }
 763
 764 void osd_req_op_init(struct ceph_osd_request *osd_req,
 765                      unsigned int which, u16 opcode, u32 flags)
 766 {
 767         (void)_osd_req_op_init(osd_req, which, opcode, flags);
 768 }
 769 EXPORT_SYMBOL(osd_req_op_init);
 770
 771 void osd_req_op_extent_init(struct ceph_osd_request *osd_req,
 772                                 unsigned int which, u16 opcode,
 773                                 u64 offset, u64 length,
 774                                 u64 truncate_size, u32 truncate_seq)
 775 {
 776         struct ceph_osd_req_op *op = _osd_req_op_init(osd_req, which,
 777                                                       opcode, 0);
 778         size_t payload_len = 0;
 779
 780         BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
 781                opcode != CEPH_OSD_OP_WRITEFULL && opcode != CEPH_OSD_OP_ZERO &&
 782                opcode != CEPH_OSD_OP_TRUNCATE);
 783
 784         op->extent.offset = offset;
 785         op->extent.length = length;
 786         op->extent.truncate_size = truncate_size;
 787         op->extent.truncate_seq = truncate_seq;
 788         if (opcode == CEPH_OSD_OP_WRITE || opcode == CEPH_OSD_OP_WRITEFULL)
 789                 payload_len += length;
 790
 791         op->indata_len = payload_len;
 792 }
 793 EXPORT_SYMBOL(osd_req_op_extent_init);
 794
 795 void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
 796                                 unsigned int which, u64 length)
 797 {
 798         struct ceph_osd_req_op *op;
 799         u64 previous;
 800
 801         BUG_ON(which >= osd_req->r_num_ops);
 802         op = &osd_req->r_ops[which];
 803         previous = op->extent.length;
 804
 805         if (length == previous)
 806                 return;         /* Nothing to do */
 807         BUG_ON(length > previous);
 808
 809         op->extent.length = length;
 810         if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
 811                 op->indata_len -= previous - length;
 812 }
 813 EXPORT_SYMBOL(osd_req_op_extent_update);
 814
 815 void osd_req_op_extent_dup_last(struct ceph_osd_request *osd_req,
 816                                 unsigned int which, u64 offset_inc)
 817 {
 818         struct ceph_osd_req_op *op, *prev_op;
 819
 820         BUG_ON(which + 1 >= osd_req->r_num_ops);
 821
 822         prev_op = &osd_req->r_ops[which];
 823         op = _osd_req_op_init(osd_req, which + 1, prev_op->op, prev_op->flags);
 824         /* dup previous one */
 825         op->indata_len = prev_op->indata_len;
 826         op->outdata_len = prev_op->outdata_len;
 827         op->extent = prev_op->extent;
 828         /* adjust offset */
 829         op->extent.offset += offset_inc;
 830         op->extent.length -= offset_inc;
 831
 832         if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
 833                 op->indata_len -= offset_inc;
 834 }
 835 EXPORT_SYMBOL(osd_req_op_extent_dup_last);
 836
 837 int osd_req_op_cls_init(struct ceph_osd_request *osd_req, unsigned int which,
 838                         const char *class, const char *method)
 839 {
 840         struct ceph_osd_req_op *op;
 841         struct ceph_pagelist *pagelist;
 842         size_t payload_len = 0;
 843         size_t size;
 844
 845         op = _osd_req_op_init(osd_req, which, CEPH_OSD_OP_CALL, 0);
 846
 847         pagelist = ceph_pagelist_alloc(GFP_NOFS);
 848         if (!pagelist)
 849                 return -ENOMEM;
 850
 851         op->cls.class_name = class;
 852         size = strlen(class);
 853         BUG_ON(size > (size_t) U8_MAX);
 854         op->cls.class_len = size;
 855         ceph_pagelist_append(pagelist, class, size);
 856         payload_len += size;
 857
 858         op->cls.method_name = method;
 859         size = strlen(method);
 860         BUG_ON(size > (size_t) U8_MAX);
 861         op->cls.method_len = size;
 862         ceph_pagelist_append(pagelist, method, size);
 863         payload_len += size;
 864
 865         osd_req_op_cls_request_info_pagelist(osd_req, which, pagelist);
 866
 867         op->indata_len = payload_len;
 868         return 0;
 869 }
 870 EXPORT_SYMBOL(osd_req_op_cls_init);
 871
 872 int osd_req_op_xattr_init(struct ceph_osd_request *osd_req, unsigned int which,
 873                           u16 opcode, const char *name, const void *value,
 874                           size_t size, u8 cmp_op, u8 cmp_mode)
 875 {
 876         struct ceph_osd_req_op *op = _osd_req_op_init(osd_req, which,
 877                                                       opcode, 0);
 878         struct ceph_pagelist *pagelist;
 879         size_t payload_len;
 880
 881         BUG_ON(opcode != CEPH_OSD_OP_SETXATTR && opcode != CEPH_OSD_OP_CMPXATTR);
 882
 883         pagelist = ceph_pagelist_alloc(GFP_NOFS);
 884         if (!pagelist)
 885                 return -ENOMEM;
 886
 887         payload_len = strlen(name);
 888         op->xattr.name_len = payload_len;
 889         ceph_pagelist_append(pagelist, name, payload_len);
 890
 891         op->xattr.value_len = size;
 892         ceph_pagelist_append(pagelist, value, size);
 893         payload_len += size;
 894
 895         op->xattr.cmp_op = cmp_op;
 896         op->xattr.cmp_mode = cmp_mode;
 897
 898         ceph_osd_data_pagelist_init(&op->xattr.osd_data, pagelist);
 899         op->indata_len = payload_len;
 900         return 0;
 901 }
 902 EXPORT_SYMBOL(osd_req_op_xattr_init);
 903
 904 /*
 905  * @watch_opcode: CEPH_OSD_WATCH_OP_*
 906  */
 907 static void osd_req_op_watch_init(struct ceph_osd_request *req, int which,
 908                                   u64 cookie, u8 watch_opcode)
 909 {
 910         struct ceph_osd_req_op *op;
 911
 912         op = _osd_req_op_init(req, which, CEPH_OSD_OP_WATCH, 0);
 913         op->watch.cookie = cookie;
 914         op->watch.op = watch_opcode;
 915         op->watch.gen = 0;
 916 }
 917
 918 void osd_req_op_alloc_hint_init(struct ceph_osd_request *osd_req,
 919                                 unsigned int which,
 920                                 u64 expected_object_size,
 921                                 u64 expected_write_size)
 922 {
 923         struct ceph_osd_req_op *op = _osd_req_op_init(osd_req, which,
 924                                                       CEPH_OSD_OP_SETALLOCHINT,
 925                                                       0);
 926
 927         op->alloc_hint.expected_object_size = expected_object_size;
 928         op->alloc_hint.expected_write_size = expected_write_size;
 929
 930         /*
 931          * CEPH_OSD_OP_SETALLOCHINT op is advisory and therefore deemed
 932          * not worth a feature bit.  Set FAILOK per-op flag to make
 933          * sure older osds don't trip over an unsupported opcode.
 934          */
 935         op->flags |= CEPH_OSD_OP_FLAG_FAILOK;
 936 }
 937 EXPORT_SYMBOL(osd_req_op_alloc_hint_init);
 938
 939 static void ceph_osdc_msg_data_add(struct ceph_msg *msg,
 940                                 struct ceph_osd_data *osd_data)
 941 {
 942         u64 length = ceph_osd_data_length(osd_data);
 943
 944         if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES) {
 945                 BUG_ON(length > (u64) SIZE_MAX);
 946                 if (length)
 947                         ceph_msg_data_add_pages(msg, osd_data->pages,
 948                                         length, osd_data->alignment);
 949         } else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
 950                 BUG_ON(!length);
 951                 ceph_msg_data_add_pagelist(msg, osd_data->pagelist);
 952 #ifdef CONFIG_BLOCK
 953         } else if (osd_data->type == CEPH_OSD_DATA_TYPE_BIO) {
 954                 ceph_msg_data_add_bio(msg, &osd_data->bio_pos, length);
 955 #endif
 956         } else if (osd_data->type == CEPH_OSD_DATA_TYPE_BVECS) {
 957                 ceph_msg_data_add_bvecs(msg, &osd_data->bvec_pos);
 958         } else {
 959                 BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_NONE);
 960         }
 961 }
 962
 963 static u32 osd_req_encode_op(struct ceph_osd_op *dst,
 964                              const struct ceph_osd_req_op *src)
 965 {
 966         switch (src->op) {
 967         case CEPH_OSD_OP_STAT:
 968                 break;
 969         case CEPH_OSD_OP_READ:
 970         case CEPH_OSD_OP_WRITE:
 971         case CEPH_OSD_OP_WRITEFULL:
 972         case CEPH_OSD_OP_ZERO:
 973         case CEPH_OSD_OP_TRUNCATE:
 974                 dst->extent.offset = cpu_to_le64(src->extent.offset);
 975                 dst->extent.length = cpu_to_le64(src->extent.length);
 976                 dst->extent.truncate_size =
 977                         cpu_to_le64(src->extent.truncate_size);
 978                 dst->extent.truncate_seq =
 979                         cpu_to_le32(src->extent.truncate_seq);
 980                 break;
 981         case CEPH_OSD_OP_CALL:
 982                 dst->cls.class_len = src->cls.class_len;
 983                 dst->cls.method_len = src->cls.method_len;
 984                 dst->cls.indata_len = cpu_to_le32(src->cls.indata_len);
 985                 break;
 986         case CEPH_OSD_OP_WATCH:
 987                 dst->watch.cookie = cpu_to_le64(src->watch.cookie);
 988                 dst->watch.ver = cpu_to_le64(0);
 989                 dst->watch.op = src->watch.op;
 990                 dst->watch.gen = cpu_to_le32(src->watch.gen);
 991                 break;
 992         case CEPH_OSD_OP_NOTIFY_ACK:
 993                 break;
 994         case CEPH_OSD_OP_NOTIFY:
 995                 dst->notify.cookie = cpu_to_le64(src->notify.cookie);
 996                 break;
 997         case CEPH_OSD_OP_LIST_WATCHERS:
 998                 break;
 999         case CEPH_OSD_OP_SETALLOCHINT:
1000                 dst->alloc_hint.expected_object_size =
1001                     cpu_to_le64(src->alloc_hint.expected_object_size);
1002                 dst->alloc_hint.expected_write_size =
1003                     cpu_to_le64(src->alloc_hint.expected_write_size);
1004                 break;
1005         case CEPH_OSD_OP_SETXATTR:
1006         case CEPH_OSD_OP_CMPXATTR:
1007                 dst->xattr.name_len = cpu_to_le32(src->xattr.name_len);
1008                 dst->xattr.value_len = cpu_to_le32(src->xattr.value_len);
1009                 dst->xattr.cmp_op = src->xattr.cmp_op;
1010                 dst->xattr.cmp_mode = src->xattr.cmp_mode;
1011                 break;
1012         case CEPH_OSD_OP_CREATE:
1013         case CEPH_OSD_OP_DELETE:
1014                 break;
1015         case CEPH_OSD_OP_COPY_FROM:
1016                 dst->copy_from.snapid = cpu_to_le64(src->copy_from.snapid);
1017                 dst->copy_from.src_version =
1018                         cpu_to_le64(src->copy_from.src_version);
1019                 dst->copy_from.flags = src->copy_from.flags;
1020                 dst->copy_from.src_fadvise_flags =
1021                         cpu_to_le32(src->copy_from.src_fadvise_flags);
1022                 break;
1023         default:
1024                 pr_err("unsupported osd opcode %s\n",
1025                         ceph_osd_op_name(src->op));
1026                 WARN_ON(1);
1027
1028                 return 0;
1029         }
1030
1031         dst->op = cpu_to_le16(src->op);
1032         dst->flags = cpu_to_le32(src->flags);
1033         dst->payload_len = cpu_to_le32(src->indata_len);
1034
1035         return src->indata_len;
1036 }
1037
1038 /*
1039  * build new request AND message, calculate layout, and adjust file
1040  * extent as needed.
1041  *
1042  * if the file was recently truncated, we include information about its
1043  * old and new size so that the object can be updated appropriately.  (we
1044  * avoid synchronously deleting truncated objects because it's slow.)
1045  */
1046 struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *osdc,
1047                                                struct ceph_file_layout *layout,
1048                                                struct ceph_vino vino,
1049                                                u64 off, u64 *plen,
1050                                                unsigned int which, int num_ops,
1051                                                int opcode, int flags,
1052                                                struct ceph_snap_context *snapc,
1053                                                u32 truncate_seq,
1054                                                u64 truncate_size,
1055                                                bool use_mempool)
1056 {
1057         struct ceph_osd_request *req;
1058         u64 objnum = 0;
1059         u64 objoff = 0;
1060         u64 objlen = 0;
1061         int r;
1062
1063         BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
1064                opcode != CEPH_OSD_OP_ZERO && opcode != CEPH_OSD_OP_TRUNCATE &&
1065                opcode != CEPH_OSD_OP_CREATE && opcode != CEPH_OSD_OP_DELETE);
1066
1067         req = ceph_osdc_alloc_request(osdc, snapc, num_ops, use_mempool,
1068                                         GFP_NOFS);
1069         if (!req) {
1070                 r = -ENOMEM;
1071                 goto fail;
1072         }
1073
1074         /* calculate max write size */
1075         r = calc_layout(layout, off, plen, &objnum, &objoff, &objlen);
1076         if (r)
1077                 goto fail;
1078
1079         if (opcode == CEPH_OSD_OP_CREATE || opcode == CEPH_OSD_OP_DELETE) {
1080                 osd_req_op_init(req, which, opcode, 0);
1081         } else {
1082                 u32 object_size = layout->object_size;
1083                 u32 object_base = off - objoff;
1084                 if (!(truncate_seq == 1 && truncate_size == -1ULL)) {
1085                         if (truncate_size <= object_base) {
1086                                 truncate_size = 0;
1087                         } else {
1088                                 truncate_size -= object_base;
1089                                 if (truncate_size > object_size)
1090                                         truncate_size = object_size;
1091                         }
1092                 }
1093                 osd_req_op_extent_init(req, which, opcode, objoff, objlen,
1094                                        truncate_size, truncate_seq);
1095         }
1096
1097         req->r_flags = flags;
1098         req->r_base_oloc.pool = layout->pool_id;
1099         req->r_base_oloc.pool_ns = ceph_try_get_string(layout->pool_ns);
1100         ceph_oid_printf(&req->r_base_oid, "%llx.%08llx", vino.ino, objnum);
1101
1102         req->r_snapid = vino.snap;
1103         if (flags & CEPH_OSD_FLAG_WRITE)
1104                 req->r_data_offset = off;
1105
1106         if (num_ops > 1)
1107                 /*
1108                  * This is a special case for ceph_writepages_start(), but it
1109                  * also covers ceph_uninline_data().  If more multi-op request
1110                  * use cases emerge, we will need a separate helper.
1111                  */
1112                 r = __ceph_osdc_alloc_messages(req, GFP_NOFS, num_ops, 0);
1113         else
1114                 r = ceph_osdc_alloc_messages(req, GFP_NOFS);
1115         if (r)
1116                 goto fail;
1117
1118         return req;
1119
1120 fail:
1121         ceph_osdc_put_request(req);
1122         return ERR_PTR(r);
1123 }
1124 EXPORT_SYMBOL(ceph_osdc_new_request);
1125
1126 /*
1127  * We keep osd requests in an rbtree, sorted by ->r_tid.
1128  */
1129 DEFINE_RB_FUNCS(request, struct ceph_osd_request, r_tid, r_node)
1130 DEFINE_RB_FUNCS(request_mc, struct ceph_osd_request, r_tid, r_mc_node)
1131
1132 /*
1133  * Call @fn on each OSD request as long as @fn returns 0.
1134  */
1135 static void for_each_request(struct ceph_osd_client *osdc,
1136                         int (*fn)(struct ceph_osd_request *req, void *arg),
1137                         void *arg)
1138 {
1139         struct rb_node *n, *p;
1140
1141         for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
1142                 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
1143
1144                 for (p = rb_first(&osd->o_requests); p; ) {
1145                         struct ceph_osd_request *req =
1146                             rb_entry(p, struct ceph_osd_request, r_node);
1147
1148                         p = rb_next(p);
1149                         if (fn(req, arg))
1150                                 return;
1151                 }
1152         }
1153
1154         for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
1155                 struct ceph_osd_request *req =
1156                     rb_entry(p, struct ceph_osd_request, r_node);
1157
1158                 p = rb_next(p);
1159                 if (fn(req, arg))
1160                         return;
1161         }
1162 }
1163
1164 static bool osd_homeless(struct ceph_osd *osd)
1165 {
1166         return osd->o_osd == CEPH_HOMELESS_OSD;
1167 }
1168
1169 static bool osd_registered(struct ceph_osd *osd)
1170 {
1171         verify_osdc_locked(osd->o_osdc);
1172
1173         return !RB_EMPTY_NODE(&osd->o_node);
1174 }
1175
1176 /*
1177  * Assumes @osd is zero-initialized.
1178  */
1179 static void osd_init(struct ceph_osd *osd)
1180 {
1181         refcount_set(&osd->o_ref, 1);
1182         RB_CLEAR_NODE(&osd->o_node);
1183         osd->o_requests = RB_ROOT;
1184         osd->o_linger_requests = RB_ROOT;
1185         osd->o_backoff_mappings = RB_ROOT;
1186         osd->o_backoffs_by_id = RB_ROOT;
1187         INIT_LIST_HEAD(&osd->o_osd_lru);
1188         INIT_LIST_HEAD(&osd->o_keepalive_item);
1189         osd->o_incarnation = 1;
1190         mutex_init(&osd->lock);
1191 }
1192
1193 static void osd_cleanup(struct ceph_osd *osd)
1194 {
1195         WARN_ON(!RB_EMPTY_NODE(&osd->o_node));
1196         WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
1197         WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
1198         WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoff_mappings));
1199         WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoffs_by_id));
1200         WARN_ON(!list_empty(&osd->o_osd_lru));
1201         WARN_ON(!list_empty(&osd->o_keepalive_item));
1202
1203         if (osd->o_auth.authorizer) {
1204                 WARN_ON(osd_homeless(osd));
1205                 ceph_auth_destroy_authorizer(osd->o_auth.authorizer);
1206         }
1207 }
1208
1209 /*
1210  * Track open sessions with osds.
1211  */
1212 static struct ceph_osd *create_osd(struct ceph_osd_client *osdc, int onum)
1213 {
1214         struct ceph_osd *osd;
1215
1216         WARN_ON(onum == CEPH_HOMELESS_OSD);
1217
1218         osd = kzalloc(sizeof(*osd), GFP_NOIO | __GFP_NOFAIL);
1219         osd_init(osd);
1220         osd->o_osdc = osdc;
1221         osd->o_osd = onum;
1222
1223         ceph_con_init(&osd->o_con, osd, &osd_con_ops, &osdc->client->msgr);
1224
1225         return osd;
1226 }
1227
1228 static struct ceph_osd *get_osd(struct ceph_osd *osd)
1229 {
1230         if (refcount_inc_not_zero(&osd->o_ref)) {
1231                 dout("get_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref)-1,
1232                      refcount_read(&osd->o_ref));
1233                 return osd;
1234         } else {
1235                 dout("get_osd %p FAIL\n", osd);
1236                 return NULL;
1237         }
1238 }
1239
1240 static void put_osd(struct ceph_osd *osd)
1241 {
1242         dout("put_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref),
1243              refcount_read(&osd->o_ref) - 1);
1244         if (refcount_dec_and_test(&osd->o_ref)) {
1245                 osd_cleanup(osd);
1246                 kfree(osd);
1247         }
1248 }
1249
1250 DEFINE_RB_FUNCS(osd, struct ceph_osd, o_osd, o_node)
1251
1252 static void __move_osd_to_lru(struct ceph_osd *osd)
1253 {
1254         struct ceph_osd_client *osdc = osd->o_osdc;
1255
1256         dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1257         BUG_ON(!list_empty(&osd->o_osd_lru));
1258
1259         spin_lock(&osdc->osd_lru_lock);
1260         list_add_tail(&osd->o_osd_lru, &osdc->osd_lru);
1261         spin_unlock(&osdc->osd_lru_lock);
1262
1263         osd->lru_ttl = jiffies + osdc->client->options->osd_idle_ttl;
1264 }
1265
1266 static void maybe_move_osd_to_lru(struct ceph_osd *osd)
1267 {
1268         if (RB_EMPTY_ROOT(&osd->o_requests) &&
1269             RB_EMPTY_ROOT(&osd->o_linger_requests))
1270                 __move_osd_to_lru(osd);
1271 }
1272
1273 static void __remove_osd_from_lru(struct ceph_osd *osd)
1274 {
1275         struct ceph_osd_client *osdc = osd->o_osdc;
1276
1277         dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1278
1279         spin_lock(&osdc->osd_lru_lock);
1280         if (!list_empty(&osd->o_osd_lru))
1281                 list_del_init(&osd->o_osd_lru);
1282         spin_unlock(&osdc->osd_lru_lock);
1283 }
1284
1285 /*
1286  * Close the connection and assign any leftover requests to the
1287  * homeless session.
1288  */
1289 static void close_osd(struct ceph_osd *osd)
1290 {
1291         struct ceph_osd_client *osdc = osd->o_osdc;
1292         struct rb_node *n;
1293
1294         verify_osdc_wrlocked(osdc);
1295         dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1296
1297         ceph_con_close(&osd->o_con);
1298
1299         for (n = rb_first(&osd->o_requests); n; ) {
1300                 struct ceph_osd_request *req =
1301                     rb_entry(n, struct ceph_osd_request, r_node);
1302
1303                 n = rb_next(n); /* unlink_request() */
1304
1305                 dout(" reassigning req %p tid %llu\n", req, req->r_tid);
1306                 unlink_request(osd, req);
1307                 link_request(&osdc->homeless_osd, req);
1308         }
1309         for (n = rb_first(&osd->o_linger_requests); n; ) {
1310                 struct ceph_osd_linger_request *lreq =
1311                     rb_entry(n, struct ceph_osd_linger_request, node);
1312
1313                 n = rb_next(n); /* unlink_linger() */
1314
1315                 dout(" reassigning lreq %p linger_id %llu\n", lreq,
1316                      lreq->linger_id);
1317                 unlink_linger(osd, lreq);
1318                 link_linger(&osdc->homeless_osd, lreq);
1319         }
1320         clear_backoffs(osd);
1321
1322         __remove_osd_from_lru(osd);
1323         erase_osd(&osdc->osds, osd);
1324         put_osd(osd);
1325 }
1326
1327 /*
1328  * reset osd connect
1329  */
1330 static int reopen_osd(struct ceph_osd *osd)
1331 {
1332         struct ceph_entity_addr *peer_addr;
1333
1334         dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1335
1336         if (RB_EMPTY_ROOT(&osd->o_requests) &&
1337             RB_EMPTY_ROOT(&osd->o_linger_requests)) {
1338                 close_osd(osd);
1339                 return -ENODEV;
1340         }
1341
1342         peer_addr = &osd->o_osdc->osdmap->osd_addr[osd->o_osd];
1343         if (!memcmp(peer_addr, &osd->o_con.peer_addr, sizeof (*peer_addr)) &&
1344                         !ceph_con_opened(&osd->o_con)) {
1345                 struct rb_node *n;
1346
1347                 dout("osd addr hasn't changed and connection never opened, "
1348                      "letting msgr retry\n");
1349                 /* touch each r_stamp for handle_timeout()'s benfit */
1350                 for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
1351                         struct ceph_osd_request *req =
1352                             rb_entry(n, struct ceph_osd_request, r_node);
1353                         req->r_stamp = jiffies;
1354                 }
1355
1356                 return -EAGAIN;
1357         }
1358
1359         ceph_con_close(&osd->o_con);
1360         ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd, peer_addr);
1361         osd->o_incarnation++;
1362
1363         return 0;
1364 }
1365
1366 static struct ceph_osd *lookup_create_osd(struct ceph_osd_client *osdc, int o,
1367                                           bool wrlocked)
1368 {
1369         struct ceph_osd *osd;
1370
1371         if (wrlocked)
1372                 verify_osdc_wrlocked(osdc);
1373         else
1374                 verify_osdc_locked(osdc);
1375
1376         if (o != CEPH_HOMELESS_OSD)
1377                 osd = lookup_osd(&osdc->osds, o);
1378         else
1379                 osd = &osdc->homeless_osd;
1380         if (!osd) {
1381                 if (!wrlocked)
1382                         return ERR_PTR(-EAGAIN);
1383
1384                 osd = create_osd(osdc, o);
1385                 insert_osd(&osdc->osds, osd);
1386                 ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd,
1387                               &osdc->osdmap->osd_addr[osd->o_osd]);
1388         }
1389
1390         dout("%s osdc %p osd%d -> osd %p\n", __func__, osdc, o, osd);
1391         return osd;
1392 }
1393
1394 /*
1395  * Create request <-> OSD session relation.
1396  *
1397  * @req has to be assigned a tid, @osd may be homeless.
1398  */
1399 static void link_request(struct ceph_osd *osd, struct ceph_osd_request *req)
1400 {
1401         verify_osd_locked(osd);
1402         WARN_ON(!req->r_tid || req->r_osd);
1403         dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1404              req, req->r_tid);
1405
1406         if (!osd_homeless(osd))
1407                 __remove_osd_from_lru(osd);
1408         else
1409                 atomic_inc(&osd->o_osdc->num_homeless);
1410
1411         get_osd(osd);
1412         insert_request(&osd->o_requests, req);
1413         req->r_osd = osd;
1414 }
1415
1416 static void unlink_request(struct ceph_osd *osd, struct ceph_osd_request *req)
1417 {
1418         verify_osd_locked(osd);
1419         WARN_ON(req->r_osd != osd);
1420         dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1421              req, req->r_tid);
1422
1423         req->r_osd = NULL;
1424         erase_request(&osd->o_requests, req);
1425         put_osd(osd);
1426
1427         if (!osd_homeless(osd))
1428                 maybe_move_osd_to_lru(osd);
1429         else
1430                 atomic_dec(&osd->o_osdc->num_homeless);
1431 }
1432
1433 static bool __pool_full(struct ceph_pg_pool_info *pi)
1434 {
1435         return pi->flags & CEPH_POOL_FLAG_FULL;
1436 }
1437
1438 static bool have_pool_full(struct ceph_osd_client *osdc)
1439 {
1440         struct rb_node *n;
1441
1442         for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
1443                 struct ceph_pg_pool_info *pi =
1444                     rb_entry(n, struct ceph_pg_pool_info, node);
1445
1446                 if (__pool_full(pi))
1447                         return true;
1448         }
1449
1450         return false;
1451 }
1452
1453 static bool pool_full(struct ceph_osd_client *osdc, s64 pool_id)
1454 {
1455         struct ceph_pg_pool_info *pi;
1456
1457         pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
1458         if (!pi)
1459                 return false;
1460
1461         return __pool_full(pi);
1462 }
1463
1464 /*
1465  * Returns whether a request should be blocked from being sent
1466  * based on the current osdmap and osd_client settings.
1467  */
1468 static bool target_should_be_paused(struct ceph_osd_client *osdc,
1469                                     const struct ceph_osd_request_target *t,
1470                                     struct ceph_pg_pool_info *pi)
1471 {
1472         bool pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
1473         bool pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
1474                        ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
1475                        __pool_full(pi);
1476
1477         WARN_ON(pi->id != t->target_oloc.pool);
1478         return ((t->flags & CEPH_OSD_FLAG_READ) && pauserd) ||
1479                ((t->flags & CEPH_OSD_FLAG_WRITE) && pausewr) ||
1480                (osdc->osdmap->epoch < osdc->epoch_barrier);
1481 }
1482
1483 enum calc_target_result {
1484         CALC_TARGET_NO_ACTION = 0,
1485         CALC_TARGET_NEED_RESEND,
1486         CALC_TARGET_POOL_DNE,
1487 };
1488
1489 static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
1490                                            struct ceph_osd_request_target *t,
1491                                            struct ceph_connection *con,
1492                                            bool any_change)
1493 {
1494         struct ceph_pg_pool_info *pi;
1495         struct ceph_pg pgid, last_pgid;
1496         struct ceph_osds up, acting;
1497         bool force_resend = false;
1498         bool unpaused = false;
1499         bool legacy_change = false;
1500         bool split = false;
1501         bool sort_bitwise = ceph_osdmap_flag(osdc, CEPH_OSDMAP_SORTBITWISE);
1502         bool recovery_deletes = ceph_osdmap_flag(osdc,
1503                                                  CEPH_OSDMAP_RECOVERY_DELETES);
1504         enum calc_target_result ct_res;
1505
1506         t->epoch = osdc->osdmap->epoch;
1507         pi = ceph_pg_pool_by_id(osdc->osdmap, t->base_oloc.pool);
1508         if (!pi) {
1509                 t->osd = CEPH_HOMELESS_OSD;
1510                 ct_res = CALC_TARGET_POOL_DNE;
1511                 goto out;
1512         }
1513
1514         if (osdc->osdmap->epoch == pi->last_force_request_resend) {
1515                 if (t->last_force_resend < pi->last_force_request_resend) {
1516                         t->last_force_resend = pi->last_force_request_resend;
1517                         force_resend = true;
1518                 } else if (t->last_force_resend == 0) {
1519                         force_resend = true;
1520                 }
1521         }
1522
1523         /* apply tiering */
1524         ceph_oid_copy(&t->target_oid, &t->base_oid);
1525         ceph_oloc_copy(&t->target_oloc, &t->base_oloc);
1526         if ((t->flags & CEPH_OSD_FLAG_IGNORE_OVERLAY) == 0) {
1527                 if (t->flags & CEPH_OSD_FLAG_READ && pi->read_tier >= 0)
1528                         t->target_oloc.pool = pi->read_tier;
1529                 if (t->flags & CEPH_OSD_FLAG_WRITE && pi->write_tier >= 0)
1530                         t->target_oloc.pool = pi->write_tier;
1531
1532                 pi = ceph_pg_pool_by_id(osdc->osdmap, t->target_oloc.pool);
1533                 if (!pi) {
1534                         t->osd = CEPH_HOMELESS_OSD;
1535                         ct_res = CALC_TARGET_POOL_DNE;
1536                         goto out;
1537                 }
1538         }
1539
1540         __ceph_object_locator_to_pg(pi, &t->target_oid, &t->target_oloc, &pgid);
1541         last_pgid.pool = pgid.pool;
1542         last_pgid.seed = ceph_stable_mod(pgid.seed, t->pg_num, t->pg_num_mask);
1543
1544         ceph_pg_to_up_acting_osds(osdc->osdmap, pi, &pgid, &up, &acting);
1545         if (any_change &&
1546             ceph_is_new_interval(&t->acting,
1547                                  &acting,
1548                                  &t->up,
1549                                  &up,
1550                                  t->size,
1551                                  pi->size,
1552                                  t->min_size,
1553                                  pi->min_size,
1554                                  t->pg_num,
1555                                  pi->pg_num,
1556                                  t->sort_bitwise,
1557                                  sort_bitwise,
1558                                  t->recovery_deletes,
1559                                  recovery_deletes,
1560                                  &last_pgid))
1561                 force_resend = true;
1562
1563         if (t->paused && !target_should_be_paused(osdc, t, pi)) {
1564                 t->paused = false;
1565                 unpaused = true;
1566         }
1567         legacy_change = ceph_pg_compare(&t->pgid, &pgid) ||
1568                         ceph_osds_changed(&t->acting, &acting, any_change);
1569         if (t->pg_num)
1570                 split = ceph_pg_is_split(&last_pgid, t->pg_num, pi->pg_num);
1571
1572         if (legacy_change || force_resend || split) {
1573                 t->pgid = pgid; /* struct */
1574                 ceph_pg_to_primary_shard(osdc->osdmap, pi, &pgid, &t->spgid);
1575                 ceph_osds_copy(&t->acting, &acting);
1576                 ceph_osds_copy(&t->up, &up);
1577                 t->size = pi->size;
1578                 t->min_size = pi->min_size;
1579                 t->pg_num = pi->pg_num;
1580                 t->pg_num_mask = pi->pg_num_mask;
1581                 t->sort_bitwise = sort_bitwise;
1582                 t->recovery_deletes = recovery_deletes;
1583
1584                 t->osd = acting.primary;
1585         }
1586
1587         if (unpaused || legacy_change || force_resend || split)
1588                 ct_res = CALC_TARGET_NEED_RESEND;
1589         else
1590                 ct_res = CALC_TARGET_NO_ACTION;
1591
1592 out:
1593         dout("%s t %p -> %d%d%d%d ct_res %d osd%d\n", __func__, t, unpaused,
1594              legacy_change, force_resend, split, ct_res, t->osd);
1595         return ct_res;
1596 }
1597
1598 static struct ceph_spg_mapping *alloc_spg_mapping(void)
1599 {
1600         struct ceph_spg_mapping *spg;
1601
1602         spg = kmalloc(sizeof(*spg), GFP_NOIO);
1603         if (!spg)
1604                 return NULL;
1605
1606         RB_CLEAR_NODE(&spg->node);
1607         spg->backoffs = RB_ROOT;
1608         return spg;
1609 }
1610
1611 static void free_spg_mapping(struct ceph_spg_mapping *spg)
1612 {
1613         WARN_ON(!RB_EMPTY_NODE(&spg->node));
1614         WARN_ON(!RB_EMPTY_ROOT(&spg->backoffs));
1615
1616         kfree(spg);
1617 }
1618
1619 /*
1620  * rbtree of ceph_spg_mapping for handling map<spg_t, ...>, similar to
1621  * ceph_pg_mapping.  Used to track OSD backoffs -- a backoff [range] is
1622  * defined only within a specific spgid; it does not pass anything to
1623  * children on split, or to another primary.
1624  */
1625 DEFINE_RB_FUNCS2(spg_mapping, struct ceph_spg_mapping, spgid, ceph_spg_compare,
1626                  RB_BYPTR, const struct ceph_spg *, node)
1627
1628 static u64 hoid_get_bitwise_key(const struct ceph_hobject_id *hoid)
1629 {
1630         return hoid->is_max ? 0x100000000ull : hoid->hash_reverse_bits;
1631 }
1632
1633 static void hoid_get_effective_key(const struct ceph_hobject_id *hoid,
1634                                    void **pkey, size_t *pkey_len)
1635 {
1636         if (hoid->key_len) {
1637                 *pkey = hoid->key;
1638                 *pkey_len = hoid->key_len;
1639         } else {
1640                 *pkey = hoid->oid;
1641                 *pkey_len = hoid->oid_len;
1642         }
1643 }
1644
1645 static int compare_names(const void *name1, size_t name1_len,
1646                          const void *name2, size_t name2_len)
1647 {
1648         int ret;
1649
1650         ret = memcmp(name1, name2, min(name1_len, name2_len));
1651         if (!ret) {
1652                 if (name1_len < name2_len)
1653                         ret = -1;
1654                 else if (name1_len > name2_len)
1655                         ret = 1;
1656         }
1657         return ret;
1658 }
1659
1660 static int hoid_compare(const struct ceph_hobject_id *lhs,
1661                         const struct ceph_hobject_id *rhs)
1662 {
1663         void *effective_key1, *effective_key2;
1664         size_t effective_key1_len, effective_key2_len;
1665         int ret;
1666
1667         if (lhs->is_max < rhs->is_max)
1668                 return -1;
1669         if (lhs->is_max > rhs->is_max)
1670                 return 1;
1671
1672         if (lhs->pool < rhs->pool)
1673                 return -1;
1674         if (lhs->pool > rhs->pool)
1675                 return 1;
1676
1677         if (hoid_get_bitwise_key(lhs) < hoid_get_bitwise_key(rhs))
1678                 return -1;
1679         if (hoid_get_bitwise_key(lhs) > hoid_get_bitwise_key(rhs))
1680                 return 1;
1681
1682         ret = compare_names(lhs->nspace, lhs->nspace_len,
1683                             rhs->nspace, rhs->nspace_len);
1684         if (ret)
1685                 return ret;
1686
1687         hoid_get_effective_key(lhs, &effective_key1, &effective_key1_len);
1688         hoid_get_effective_key(rhs, &effective_key2, &effective_key2_len);
1689         ret = compare_names(effective_key1, effective_key1_len,
1690                             effective_key2, effective_key2_len);
1691         if (ret)
1692                 return ret;
1693
1694         ret = compare_names(lhs->oid, lhs->oid_len, rhs->oid, rhs->oid_len);
1695         if (ret)
1696                 return ret;
1697
1698         if (lhs->snapid < rhs->snapid)
1699                 return -1;
1700         if (lhs->snapid > rhs->snapid)
1701                 return 1;
1702
1703         return 0;
1704 }
1705
1706 /*
1707  * For decoding ->begin and ->end of MOSDBackoff only -- no MIN/MAX
1708  * compat stuff here.
1709  *
1710  * Assumes @hoid is zero-initialized.
1711  */
1712 static int decode_hoid(void **p, void *end, struct ceph_hobject_id *hoid)
1713 {
1714         u8 struct_v;
1715         u32 struct_len;
1716         int ret;
1717
1718         ret = ceph_start_decoding(p, end, 4, "hobject_t", &struct_v,
1719                                   &struct_len);
1720         if (ret)
1721                 return ret;
1722
1723         if (struct_v < 4) {
1724                 pr_err("got struct_v %d < 4 of hobject_t\n", struct_v);
1725                 goto e_inval;
1726         }
1727
1728         hoid->key = ceph_extract_encoded_string(p, end, &hoid->key_len,
1729                                                 GFP_NOIO);
1730         if (IS_ERR(hoid->key)) {
1731                 ret = PTR_ERR(hoid->key);
1732                 hoid->key = NULL;
1733                 return ret;
1734         }
1735
1736         hoid->oid = ceph_extract_encoded_string(p, end, &hoid->oid_len,
1737                                                 GFP_NOIO);
1738         if (IS_ERR(hoid->oid)) {
1739                 ret = PTR_ERR(hoid->oid);
1740                 hoid->oid = NULL;
1741                 return ret;
1742         }
1743
1744         ceph_decode_64_safe(p, end, hoid->snapid, e_inval);
1745         ceph_decode_32_safe(p, end, hoid->hash, e_inval);
1746         ceph_decode_8_safe(p, end, hoid->is_max, e_inval);
1747
1748         hoid->nspace = ceph_extract_encoded_string(p, end, &hoid->nspace_len,
1749                                                    GFP_NOIO);
1750         if (IS_ERR(hoid->nspace)) {
1751                 ret = PTR_ERR(hoid->nspace);
1752                 hoid->nspace = NULL;
1753                 return ret;
1754         }
1755
1756         ceph_decode_64_safe(p, end, hoid->pool, e_inval);
1757
1758         ceph_hoid_build_hash_cache(hoid);
1759         return 0;
1760
1761 e_inval:
1762         return -EINVAL;
1763 }
1764
1765 static int hoid_encoding_size(const struct ceph_hobject_id *hoid)
1766 {
1767         return 8 + 4 + 1 + 8 + /* snapid, hash, is_max, pool */
1768                4 + hoid->key_len + 4 + hoid->oid_len + 4 + hoid->nspace_len;
1769 }
1770
1771 static void encode_hoid(void **p, void *end, const struct ceph_hobject_id *hoid)
1772 {
1773         ceph_start_encoding(p, 4, 3, hoid_encoding_size(hoid));
1774         ceph_encode_string(p, end, hoid->key, hoid->key_len);
1775         ceph_encode_string(p, end, hoid->oid, hoid->oid_len);
1776         ceph_encode_64(p, hoid->snapid);
1777         ceph_encode_32(p, hoid->hash);
1778         ceph_encode_8(p, hoid->is_max);
1779         ceph_encode_string(p, end, hoid->nspace, hoid->nspace_len);
1780         ceph_encode_64(p, hoid->pool);
1781 }
1782
1783 static void free_hoid(struct ceph_hobject_id *hoid)
1784 {
1785         if (hoid) {
1786                 kfree(hoid->key);
1787                 kfree(hoid->oid);
1788                 kfree(hoid->nspace);
1789                 kfree(hoid);
1790         }
1791 }
1792
1793 static struct ceph_osd_backoff *alloc_backoff(void)
1794 {
1795         struct ceph_osd_backoff *backoff;
1796
1797         backoff = kzalloc(sizeof(*backoff), GFP_NOIO);
1798         if (!backoff)
1799                 return NULL;
1800
1801         RB_CLEAR_NODE(&backoff->spg_node);
1802         RB_CLEAR_NODE(&backoff->id_node);
1803         return backoff;
1804 }
1805
1806 static void free_backoff(struct ceph_osd_backoff *backoff)
1807 {
1808         WARN_ON(!RB_EMPTY_NODE(&backoff->spg_node));
1809         WARN_ON(!RB_EMPTY_NODE(&backoff->id_node));
1810
1811         free_hoid(backoff->begin);
1812         free_hoid(backoff->end);
1813         kfree(backoff);
1814 }
1815
1816 /*
1817  * Within a specific spgid, backoffs are managed by ->begin hoid.
1818  */
1819 DEFINE_RB_INSDEL_FUNCS2(backoff, struct ceph_osd_backoff, begin, hoid_compare,
1820                         RB_BYVAL, spg_node);
1821
1822 static struct ceph_osd_backoff *lookup_containing_backoff(struct rb_root *root,
1823                                             const struct ceph_hobject_id *hoid)
1824 {
1825         struct rb_node *n = root->rb_node;
1826
1827         while (n) {
1828                 struct ceph_osd_backoff *cur =
1829                     rb_entry(n, struct ceph_osd_backoff, spg_node);
1830                 int cmp;
1831
1832                 cmp = hoid_compare(hoid, cur->begin);
1833                 if (cmp < 0) {
1834                         n = n->rb_left;
1835                 } else if (cmp > 0) {
1836                         if (hoid_compare(hoid, cur->end) < 0)
1837                                 return cur;
1838
1839                         n = n->rb_right;
1840                 } else {
1841                         return cur;
1842                 }
1843         }
1844
1845         return NULL;
1846 }
1847
1848 /*
1849  * Each backoff has a unique id within its OSD session.
1850  */
1851 DEFINE_RB_FUNCS(backoff_by_id, struct ceph_osd_backoff, id, id_node)
1852
1853 static void clear_backoffs(struct ceph_osd *osd)
1854 {
1855         while (!RB_EMPTY_ROOT(&osd->o_backoff_mappings)) {
1856                 struct ceph_spg_mapping *spg =
1857                     rb_entry(rb_first(&osd->o_backoff_mappings),
1858                              struct ceph_spg_mapping, node);
1859
1860                 while (!RB_EMPTY_ROOT(&spg->backoffs)) {
1861                         struct ceph_osd_backoff *backoff =
1862                             rb_entry(rb_first(&spg->backoffs),
1863                                      struct ceph_osd_backoff, spg_node);
1864
1865                         erase_backoff(&spg->backoffs, backoff);
1866                         erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
1867                         free_backoff(backoff);
1868                 }
1869                 erase_spg_mapping(&osd->o_backoff_mappings, spg);
1870                 free_spg_mapping(spg);
1871         }
1872 }
1873
1874 /*
1875  * Set up a temporary, non-owning view into @t.
1876  */
1877 static void hoid_fill_from_target(struct ceph_hobject_id *hoid,
1878                                   const struct ceph_osd_request_target *t)
1879 {
1880         hoid->key = NULL;
1881         hoid->key_len = 0;
1882         hoid->oid = t->target_oid.name;
1883         hoid->oid_len = t->target_oid.name_len;
1884         hoid->snapid = CEPH_NOSNAP;
1885         hoid->hash = t->pgid.seed;
1886         hoid->is_max = false;
1887         if (t->target_oloc.pool_ns) {
1888                 hoid->nspace = t->target_oloc.pool_ns->str;
1889                 hoid->nspace_len = t->target_oloc.pool_ns->len;
1890         } else {
1891                 hoid->nspace = NULL;
1892                 hoid->nspace_len = 0;
1893         }
1894         hoid->pool = t->target_oloc.pool;
1895         ceph_hoid_build_hash_cache(hoid);
1896 }
1897
1898 static bool should_plug_request(struct ceph_osd_request *req)
1899 {
1900         struct ceph_osd *osd = req->r_osd;
1901         struct ceph_spg_mapping *spg;
1902         struct ceph_osd_backoff *backoff;
1903         struct ceph_hobject_id hoid;
1904
1905         spg = lookup_spg_mapping(&osd->o_backoff_mappings, &req->r_t.spgid);
1906         if (!spg)
1907                 return false;
1908
1909         hoid_fill_from_target(&hoid, &req->r_t);
1910         backoff = lookup_containing_backoff(&spg->backoffs, &hoid);
1911         if (!backoff)
1912                 return false;
1913
1914         dout("%s req %p tid %llu backoff osd%d spgid %llu.%xs%d id %llu\n",
1915              __func__, req, req->r_tid, osd->o_osd, backoff->spgid.pgid.pool,
1916              backoff->spgid.pgid.seed, backoff->spgid.shard, backoff->id);
1917         return true;
1918 }
1919
1920 /*
1921  * Keep get_num_data_items() in sync with this function.
1922  */
1923 static void setup_request_data(struct ceph_osd_request *req)
1924 {
1925         struct ceph_msg *request_msg = req->r_request;
1926         struct ceph_msg *reply_msg = req->r_reply;
1927         struct ceph_osd_req_op *op;
1928
1929         if (req->r_request->num_data_items || req->r_reply->num_data_items)
1930                 return;
1931
1932         WARN_ON(request_msg->data_length || reply_msg->data_length);
1933         for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) {
1934                 switch (op->op) {
1935                 /* request */
1936                 case CEPH_OSD_OP_WRITE:
1937                 case CEPH_OSD_OP_WRITEFULL:
1938                         WARN_ON(op->indata_len != op->extent.length);
1939                         ceph_osdc_msg_data_add(request_msg,
1940                                                &op->extent.osd_data);
1941                         break;
1942                 case CEPH_OSD_OP_SETXATTR:
1943                 case CEPH_OSD_OP_CMPXATTR:
1944                         WARN_ON(op->indata_len != op->xattr.name_len +
1945                                                   op->xattr.value_len);
1946                         ceph_osdc_msg_data_add(request_msg,
1947                                                &op->xattr.osd_data);
1948                         break;
1949                 case CEPH_OSD_OP_NOTIFY_ACK:
1950                         ceph_osdc_msg_data_add(request_msg,
1951                                                &op->notify_ack.request_data);
1952                         break;
1953                 case CEPH_OSD_OP_COPY_FROM:
1954                         ceph_osdc_msg_data_add(request_msg,
1955                                                &op->copy_from.osd_data);
1956                         break;
1957
1958                 /* reply */
1959                 case CEPH_OSD_OP_STAT:
1960                         ceph_osdc_msg_data_add(reply_msg,
1961                                                &op->raw_data_in);
1962                         break;
1963                 case CEPH_OSD_OP_READ:
1964                         ceph_osdc_msg_data_add(reply_msg,
1965                                                &op->extent.osd_data);
1966                         break;
1967                 case CEPH_OSD_OP_LIST_WATCHERS:
1968                         ceph_osdc_msg_data_add(reply_msg,
1969                                                &op->list_watchers.response_data);
1970                         break;
1971
1972                 /* both */
1973                 case CEPH_OSD_OP_CALL:
1974                         WARN_ON(op->indata_len != op->cls.class_len +
1975                                                   op->cls.method_len +
1976                                                   op->cls.indata_len);
1977                         ceph_osdc_msg_data_add(request_msg,
1978                                                &op->cls.request_info);
1979                         /* optional, can be NONE */
1980                         ceph_osdc_msg_data_add(request_msg,
1981                                                &op->cls.request_data);
1982                         /* optional, can be NONE */
1983                         ceph_osdc_msg_data_add(reply_msg,
1984                                                &op->cls.response_data);
1985                         break;
1986                 case CEPH_OSD_OP_NOTIFY:
1987                         ceph_osdc_msg_data_add(request_msg,
1988                                                &op->notify.request_data);
1989                         ceph_osdc_msg_data_add(reply_msg,
1990                                                &op->notify.response_data);
1991                         break;
1992                 }
1993         }
1994 }
1995
1996 static void encode_pgid(void **p, const struct ceph_pg *pgid)
1997 {
1998         ceph_encode_8(p, 1);
1999         ceph_encode_64(p, pgid->pool);
2000         ceph_encode_32(p, pgid->seed);
2001         ceph_encode_32(p, -1); /* preferred */
2002 }
2003
2004 static void encode_spgid(void **p, const struct ceph_spg *spgid)
2005 {
2006         ceph_start_encoding(p, 1, 1, CEPH_PGID_ENCODING_LEN + 1);
2007         encode_pgid(p, &spgid->pgid);
2008         ceph_encode_8(p, spgid->shard);
2009 }
2010
2011 static void encode_oloc(void **p, void *end,
2012                         const struct ceph_object_locator *oloc)
2013 {
2014         ceph_start_encoding(p, 5, 4, ceph_oloc_encoding_size(oloc));
2015         ceph_encode_64(p, oloc->pool);
2016         ceph_encode_32(p, -1); /* preferred */
2017         ceph_encode_32(p, 0);  /* key len */
2018         if (oloc->pool_ns)
2019                 ceph_encode_string(p, end, oloc->pool_ns->str,
2020                                    oloc->pool_ns->len);
2021         else
2022                 ceph_encode_32(p, 0);
2023 }
2024
2025 static void encode_request_partial(struct ceph_osd_request *req,
2026                                    struct ceph_msg *msg)
2027 {
2028         void *p = msg->front.iov_base;
2029         void *const end = p + msg->front_alloc_len;
2030         u32 data_len = 0;
2031         int i;
2032
2033         if (req->r_flags & CEPH_OSD_FLAG_WRITE) {
2034                 /* snapshots aren't writeable */
2035                 WARN_ON(req->r_snapid != CEPH_NOSNAP);
2036         } else {
2037                 WARN_ON(req->r_mtime.tv_sec || req->r_mtime.tv_nsec ||
2038                         req->r_data_offset || req->r_snapc);
2039         }
2040
2041         setup_request_data(req);
2042
2043         encode_spgid(&p, &req->r_t.spgid); /* actual spg */
2044         ceph_encode_32(&p, req->r_t.pgid.seed); /* raw hash */
2045         ceph_encode_32(&p, req->r_osdc->osdmap->epoch);
2046         ceph_encode_32(&p, req->r_flags);
2047
2048         /* reqid */
2049         ceph_start_encoding(&p, 2, 2, sizeof(struct ceph_osd_reqid));
2050         memset(p, 0, sizeof(struct ceph_osd_reqid));
2051         p += sizeof(struct ceph_osd_reqid);
2052
2053         /* trace */
2054         memset(p, 0, sizeof(struct ceph_blkin_trace_info));
2055         p += sizeof(struct ceph_blkin_trace_info);
2056
2057         ceph_encode_32(&p, 0); /* client_inc, always 0 */
2058         ceph_encode_timespec64(p, &req->r_mtime);
2059         p += sizeof(struct ceph_timespec);
2060
2061         encode_oloc(&p, end, &req->r_t.target_oloc);
2062         ceph_encode_string(&p, end, req->r_t.target_oid.name,
2063                            req->r_t.target_oid.name_len);
2064
2065         /* ops, can imply data */
2066         ceph_encode_16(&p, req->r_num_ops);
2067         for (i = 0; i < req->r_num_ops; i++) {
2068                 data_len += osd_req_encode_op(p, &req->r_ops[i]);
2069                 p += sizeof(struct ceph_osd_op);
2070         }
2071
2072         ceph_encode_64(&p, req->r_snapid); /* snapid */
2073         if (req->r_snapc) {
2074                 ceph_encode_64(&p, req->r_snapc->seq);
2075                 ceph_encode_32(&p, req->r_snapc->num_snaps);
2076                 for (i = 0; i < req->r_snapc->num_snaps; i++)
2077                         ceph_encode_64(&p, req->r_snapc->snaps[i]);
2078         } else {
2079                 ceph_encode_64(&p, 0); /* snap_seq */
2080                 ceph_encode_32(&p, 0); /* snaps len */
2081         }
2082
2083         ceph_encode_32(&p, req->r_attempts); /* retry_attempt */
2084         BUG_ON(p > end - 8); /* space for features */
2085
2086         msg->hdr.version = cpu_to_le16(8); /* MOSDOp v8 */
2087         /* front_len is finalized in encode_request_finish() */
2088         msg->front.iov_len = p - msg->front.iov_base;
2089         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2090         msg->hdr.data_len = cpu_to_le32(data_len);
2091         /*
2092          * The header "data_off" is a hint to the receiver allowing it
2093          * to align received data into its buffers such that there's no
2094          * need to re-copy it before writing it to disk (direct I/O).
2095          */
2096         msg->hdr.data_off = cpu_to_le16(req->r_data_offset);
2097
2098         dout("%s req %p msg %p oid %s oid_len %d\n", __func__, req, msg,
2099              req->r_t.target_oid.name, req->r_t.target_oid.name_len);
2100 }
2101
2102 static void encode_request_finish(struct ceph_msg *msg)
2103 {
2104         void *p = msg->front.iov_base;
2105         void *const partial_end = p + msg->front.iov_len;
2106         void *const end = p + msg->front_alloc_len;
2107
2108         if (CEPH_HAVE_FEATURE(msg->con->peer_features, RESEND_ON_SPLIT)) {
2109                 /* luminous OSD -- encode features and be done */
2110                 p = partial_end;
2111                 ceph_encode_64(&p, msg->con->peer_features);
2112         } else {
2113                 struct {
2114                         char spgid[CEPH_ENCODING_START_BLK_LEN +
2115                                    CEPH_PGID_ENCODING_LEN + 1];
2116                         __le32 hash;
2117                         __le32 epoch;
2118                         __le32 flags;
2119                         char reqid[CEPH_ENCODING_START_BLK_LEN +
2120                                    sizeof(struct ceph_osd_reqid)];
2121                         char trace[sizeof(struct ceph_blkin_trace_info)];
2122                         __le32 client_inc;
2123                         struct ceph_timespec mtime;
2124                 } __packed head;
2125                 struct ceph_pg pgid;
2126                 void *oloc, *oid, *tail;
2127                 int oloc_len, oid_len, tail_len;
2128                 int len;
2129
2130                 /*
2131                  * Pre-luminous OSD -- reencode v8 into v4 using @head
2132                  * as a temporary buffer.  Encode the raw PG; the rest
2133                  * is just a matter of moving oloc, oid and tail blobs
2134                  * around.
2135                  */
2136                 memcpy(&head, p, sizeof(head));
2137                 p += sizeof(head);
2138
2139                 oloc = p;
2140                 p += CEPH_ENCODING_START_BLK_LEN;
2141                 pgid.pool = ceph_decode_64(&p);
2142                 p += 4 + 4; /* preferred, key len */
2143                 len = ceph_decode_32(&p);
2144                 p += len;   /* nspace */
2145                 oloc_len = p - oloc;
2146
2147                 oid = p;
2148                 len = ceph_decode_32(&p);
2149                 p += len;
2150                 oid_len = p - oid;
2151
2152                 tail = p;
2153                 tail_len = partial_end - p;
2154
2155                 p = msg->front.iov_base;
2156                 ceph_encode_copy(&p, &head.client_inc, sizeof(head.client_inc));
2157                 ceph_encode_copy(&p, &head.epoch, sizeof(head.epoch));
2158                 ceph_encode_copy(&p, &head.flags, sizeof(head.flags));
2159                 ceph_encode_copy(&p, &head.mtime, sizeof(head.mtime));
2160
2161                 /* reassert_version */
2162                 memset(p, 0, sizeof(struct ceph_eversion));
2163                 p += sizeof(struct ceph_eversion);
2164
2165                 BUG_ON(p >= oloc);
2166                 memmove(p, oloc, oloc_len);
2167                 p += oloc_len;
2168
2169                 pgid.seed = le32_to_cpu(head.hash);
2170                 encode_pgid(&p, &pgid); /* raw pg */
2171
2172                 BUG_ON(p >= oid);
2173                 memmove(p, oid, oid_len);
2174                 p += oid_len;
2175
2176                 /* tail -- ops, snapid, snapc, retry_attempt */
2177                 BUG_ON(p >= tail);
2178                 memmove(p, tail, tail_len);
2179                 p += tail_len;
2180
2181                 msg->hdr.version = cpu_to_le16(4); /* MOSDOp v4 */
2182         }
2183
2184         BUG_ON(p > end);
2185         msg->front.iov_len = p - msg->front.iov_base;
2186         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2187
2188         dout("%s msg %p tid %llu %u+%u+%u v%d\n", __func__, msg,
2189              le64_to_cpu(msg->hdr.tid), le32_to_cpu(msg->hdr.front_len),
2190              le32_to_cpu(msg->hdr.middle_len), le32_to_cpu(msg->hdr.data_len),
2191              le16_to_cpu(msg->hdr.version));
2192 }
2193
2194 /*
2195  * @req has to be assigned a tid and registered.
2196  */
2197 static void send_request(struct ceph_osd_request *req)
2198 {
2199         struct ceph_osd *osd = req->r_osd;
2200
2201         verify_osd_locked(osd);
2202         WARN_ON(osd->o_osd != req->r_t.osd);
2203
2204         /* backoff? */
2205         if (should_plug_request(req))
2206                 return;
2207
2208         /*
2209          * We may have a previously queued request message hanging
2210          * around.  Cancel it to avoid corrupting the msgr.
2211          */
2212         if (req->r_sent)
2213                 ceph_msg_revoke(req->r_request);
2214
2215         req->r_flags |= CEPH_OSD_FLAG_KNOWN_REDIR;
2216         if (req->r_attempts)
2217                 req->r_flags |= CEPH_OSD_FLAG_RETRY;
2218         else
2219                 WARN_ON(req->r_flags & CEPH_OSD_FLAG_RETRY);
2220
2221         encode_request_partial(req, req->r_request);
2222
2223         dout("%s req %p tid %llu to pgid %llu.%x spgid %llu.%xs%d osd%d e%u flags 0x%x attempt %d\n",
2224              __func__, req, req->r_tid, req->r_t.pgid.pool, req->r_t.pgid.seed,
2225              req->r_t.spgid.pgid.pool, req->r_t.spgid.pgid.seed,
2226              req->r_t.spgid.shard, osd->o_osd, req->r_t.epoch, req->r_flags,
2227              req->r_attempts);
2228
2229         req->r_t.paused = false;
2230         req->r_stamp = jiffies;
2231         req->r_attempts++;
2232
2233         req->r_sent = osd->o_incarnation;
2234         req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
2235         ceph_con_send(&osd->o_con, ceph_msg_get(req->r_request));
2236 }
2237
2238 static void maybe_request_map(struct ceph_osd_client *osdc)
2239 {
2240         bool continuous = false;
2241
2242         verify_osdc_locked(osdc);
2243         WARN_ON(!osdc->osdmap->epoch);
2244
2245         if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2246             ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD) ||
2247             ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
2248                 dout("%s osdc %p continuous\n", __func__, osdc);
2249                 continuous = true;
2250         } else {
2251                 dout("%s osdc %p onetime\n", __func__, osdc);
2252         }
2253
2254         if (ceph_monc_want_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
2255                                osdc->osdmap->epoch + 1, continuous))
2256                 ceph_monc_renew_subs(&osdc->client->monc);
2257 }
2258
2259 static void complete_request(struct ceph_osd_request *req, int err);
2260 static void send_map_check(struct ceph_osd_request *req);
2261
2262 static void __submit_request(struct ceph_osd_request *req, bool wrlocked)
2263 {
2264         struct ceph_osd_client *osdc = req->r_osdc;
2265         struct ceph_osd *osd;
2266         enum calc_target_result ct_res;
2267         int err = 0;
2268         bool need_send = false;
2269         bool promoted = false;
2270
2271         WARN_ON(req->r_tid);
2272         dout("%s req %p wrlocked %d\n", __func__, req, wrlocked);
2273
2274 again:
2275         ct_res = calc_target(osdc, &req->r_t, NULL, false);
2276         if (ct_res == CALC_TARGET_POOL_DNE && !wrlocked)
2277                 goto promote;
2278
2279         osd = lookup_create_osd(osdc, req->r_t.osd, wrlocked);
2280         if (IS_ERR(osd)) {
2281                 WARN_ON(PTR_ERR(osd) != -EAGAIN || wrlocked);
2282                 goto promote;
2283         }
2284
2285         if (osdc->abort_err) {
2286                 dout("req %p abort_err %d\n", req, osdc->abort_err);
2287                 err = osdc->abort_err;
2288         } else if (osdc->osdmap->epoch < osdc->epoch_barrier) {
2289                 dout("req %p epoch %u barrier %u\n", req, osdc->osdmap->epoch,
2290                      osdc->epoch_barrier);
2291                 req->r_t.paused = true;
2292                 maybe_request_map(osdc);
2293         } else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2294                    ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
2295                 dout("req %p pausewr\n", req);
2296                 req->r_t.paused = true;
2297                 maybe_request_map(osdc);
2298         } else if ((req->r_flags & CEPH_OSD_FLAG_READ) &&
2299                    ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
2300                 dout("req %p pauserd\n", req);
2301                 req->r_t.paused = true;
2302                 maybe_request_map(osdc);
2303         } else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2304                    !(req->r_flags & (CEPH_OSD_FLAG_FULL_TRY |
2305                                      CEPH_OSD_FLAG_FULL_FORCE)) &&
2306                    (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2307                     pool_full(osdc, req->r_t.base_oloc.pool))) {
2308                 dout("req %p full/pool_full\n", req);
2309                 if (ceph_test_opt(osdc->client, ABORT_ON_FULL)) {
2310                         err = -ENOSPC;
2311                 } else {
2312                         pr_warn_ratelimited("FULL or reached pool quota\n");
2313                         req->r_t.paused = true;
2314                         maybe_request_map(osdc);
2315                 }
2316         } else if (!osd_homeless(osd)) {
2317                 need_send = true;
2318         } else {
2319                 maybe_request_map(osdc);
2320         }
2321
2322         mutex_lock(&osd->lock);
2323         /*
2324          * Assign the tid atomically with send_request() to protect
2325          * multiple writes to the same object from racing with each
2326          * other, resulting in out of order ops on the OSDs.
2327          */
2328         req->r_tid = atomic64_inc_return(&osdc->last_tid);
2329         link_request(osd, req);
2330         if (need_send)
2331                 send_request(req);
2332         else if (err)
2333                 complete_request(req, err);
2334         mutex_unlock(&osd->lock);
2335
2336         if (!err && ct_res == CALC_TARGET_POOL_DNE)
2337                 send_map_check(req);
2338
2339         if (promoted)
2340                 downgrade_write(&osdc->lock);
2341         return;
2342
2343 promote:
2344         up_read(&osdc->lock);
2345         down_write(&osdc->lock);
2346         wrlocked = true;
2347         promoted = true;
2348         goto again;
2349 }
2350
2351 static void account_request(struct ceph_osd_request *req)
2352 {
2353         WARN_ON(req->r_flags & (CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK));
2354         WARN_ON(!(req->r_flags & (CEPH_OSD_FLAG_READ | CEPH_OSD_FLAG_WRITE)));
2355
2356         req->r_flags |= CEPH_OSD_FLAG_ONDISK;
2357         atomic_inc(&req->r_osdc->num_requests);
2358
2359         req->r_start_stamp = jiffies;
2360 }
2361
2362 static void submit_request(struct ceph_osd_request *req, bool wrlocked)
2363 {
2364         ceph_osdc_get_request(req);
2365         account_request(req);
2366         __submit_request(req, wrlocked);
2367 }
2368
2369 static void finish_request(struct ceph_osd_request *req)
2370 {
2371         struct ceph_osd_client *osdc = req->r_osdc;
2372
2373         WARN_ON(lookup_request_mc(&osdc->map_checks, req->r_tid));
2374         dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2375
2376         if (req->r_osd)
2377                 unlink_request(req->r_osd, req);
2378         atomic_dec(&osdc->num_requests);
2379
2380         /*
2381          * If an OSD has failed or returned and a request has been sent
2382          * twice, it's possible to get a reply and end up here while the
2383          * request message is queued for delivery.  We will ignore the
2384          * reply, so not a big deal, but better to try and catch it.
2385          */
2386         ceph_msg_revoke(req->r_request);
2387         ceph_msg_revoke_incoming(req->r_reply);
2388 }
2389
2390 static void __complete_request(struct ceph_osd_request *req)
2391 {
2392         dout("%s req %p tid %llu cb %ps result %d\n", __func__, req,
2393              req->r_tid, req->r_callback, req->r_result);
2394
2395         if (req->r_callback)
2396                 req->r_callback(req);
2397         complete_all(&req->r_completion);
2398         ceph_osdc_put_request(req);
2399 }
2400
2401 static void complete_request_workfn(struct work_struct *work)
2402 {
2403         struct ceph_osd_request *req =
2404             container_of(work, struct ceph_osd_request, r_complete_work);
2405
2406         __complete_request(req);
2407 }
2408
2409 /*
2410  * This is open-coded in handle_reply().
2411  */
2412 static void complete_request(struct ceph_osd_request *req, int err)
2413 {
2414         dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2415
2416         req->r_result = err;
2417         finish_request(req);
2418
2419         INIT_WORK(&req->r_complete_work, complete_request_workfn);
2420         queue_work(req->r_osdc->completion_wq, &req->r_complete_work);
2421 }
2422
2423 static void cancel_map_check(struct ceph_osd_request *req)
2424 {
2425         struct ceph_osd_client *osdc = req->r_osdc;
2426         struct ceph_osd_request *lookup_req;
2427
2428         verify_osdc_wrlocked(osdc);
2429
2430         lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2431         if (!lookup_req)
2432                 return;
2433
2434         WARN_ON(lookup_req != req);
2435         erase_request_mc(&osdc->map_checks, req);
2436         ceph_osdc_put_request(req);
2437 }
2438
2439 static void cancel_request(struct ceph_osd_request *req)
2440 {
2441         dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2442
2443         cancel_map_check(req);
2444         finish_request(req);
2445         complete_all(&req->r_completion);
2446         ceph_osdc_put_request(req);
2447 }
2448
2449 static void abort_request(struct ceph_osd_request *req, int err)
2450 {
2451         dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2452
2453         cancel_map_check(req);
2454         complete_request(req, err);
2455 }
2456
2457 static int abort_fn(struct ceph_osd_request *req, void *arg)
2458 {
2459         int err = *(int *)arg;
2460
2461         abort_request(req, err);
2462         return 0; /* continue iteration */
2463 }
2464
2465 /*
2466  * Abort all in-flight requests with @err and arrange for all future
2467  * requests to be failed immediately.
2468  */
2469 void ceph_osdc_abort_requests(struct ceph_osd_client *osdc, int err)
2470 {
2471         dout("%s osdc %p err %d\n", __func__, osdc, err);
2472         down_write(&osdc->lock);
2473         for_each_request(osdc, abort_fn, &err);
2474         osdc->abort_err = err;
2475         up_write(&osdc->lock);
2476 }
2477 EXPORT_SYMBOL(ceph_osdc_abort_requests);
2478
2479 static void update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2480 {
2481         if (likely(eb > osdc->epoch_barrier)) {
2482                 dout("updating epoch_barrier from %u to %u\n",
2483                                 osdc->epoch_barrier, eb);
2484                 osdc->epoch_barrier = eb;
2485                 /* Request map if we're not to the barrier yet */
2486                 if (eb > osdc->osdmap->epoch)
2487                         maybe_request_map(osdc);
2488         }
2489 }
2490
2491 void ceph_osdc_update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2492 {
2493         down_read(&osdc->lock);
2494         if (unlikely(eb > osdc->epoch_barrier)) {
2495                 up_read(&osdc->lock);
2496                 down_write(&osdc->lock);
2497                 update_epoch_barrier(osdc, eb);
2498                 up_write(&osdc->lock);
2499         } else {
2500                 up_read(&osdc->lock);
2501         }
2502 }
2503 EXPORT_SYMBOL(ceph_osdc_update_epoch_barrier);
2504
2505 /*
2506  * We can end up releasing caps as a result of abort_request().
2507  * In that case, we probably want to ensure that the cap release message
2508  * has an updated epoch barrier in it, so set the epoch barrier prior to
2509  * aborting the first request.
2510  */
2511 static int abort_on_full_fn(struct ceph_osd_request *req, void *arg)
2512 {
2513         struct ceph_osd_client *osdc = req->r_osdc;
2514         bool *victims = arg;
2515
2516         if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2517             (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2518              pool_full(osdc, req->r_t.base_oloc.pool))) {
2519                 if (!*victims) {
2520                         update_epoch_barrier(osdc, osdc->osdmap->epoch);
2521                         *victims = true;
2522                 }
2523                 abort_request(req, -ENOSPC);
2524         }
2525
2526         return 0; /* continue iteration */
2527 }
2528
2529 /*
2530  * Drop all pending requests that are stalled waiting on a full condition to
2531  * clear, and complete them with ENOSPC as the return code. Set the
2532  * osdc->epoch_barrier to the latest map epoch that we've seen if any were
2533  * cancelled.
2534  */
2535 static void ceph_osdc_abort_on_full(struct ceph_osd_client *osdc)
2536 {
2537         bool victims = false;
2538
2539         if (ceph_test_opt(osdc->client, ABORT_ON_FULL) &&
2540             (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || have_pool_full(osdc)))
2541                 for_each_request(osdc, abort_on_full_fn, &victims);
2542 }
2543
2544 static void check_pool_dne(struct ceph_osd_request *req)
2545 {
2546         struct ceph_osd_client *osdc = req->r_osdc;
2547         struct ceph_osdmap *map = osdc->osdmap;
2548
2549         verify_osdc_wrlocked(osdc);
2550         WARN_ON(!map->epoch);
2551
2552         if (req->r_attempts) {
2553                 /*
2554                  * We sent a request earlier, which means that
2555                  * previously the pool existed, and now it does not
2556                  * (i.e., it was deleted).
2557                  */
2558                 req->r_map_dne_bound = map->epoch;
2559                 dout("%s req %p tid %llu pool disappeared\n", __func__, req,
2560                      req->r_tid);
2561         } else {
2562                 dout("%s req %p tid %llu map_dne_bound %u have %u\n", __func__,
2563                      req, req->r_tid, req->r_map_dne_bound, map->epoch);
2564         }
2565
2566         if (req->r_map_dne_bound) {
2567                 if (map->epoch >= req->r_map_dne_bound) {
2568                         /* we had a new enough map */
2569                         pr_info_ratelimited("tid %llu pool does not exist\n",
2570                                             req->r_tid);
2571                         complete_request(req, -ENOENT);
2572                 }
2573         } else {
2574                 send_map_check(req);
2575         }
2576 }
2577
2578 static void map_check_cb(struct ceph_mon_generic_request *greq)
2579 {
2580         struct ceph_osd_client *osdc = &greq->monc->client->osdc;
2581         struct ceph_osd_request *req;
2582         u64 tid = greq->private_data;
2583
2584         WARN_ON(greq->result || !greq->u.newest);
2585
2586         down_write(&osdc->lock);
2587         req = lookup_request_mc(&osdc->map_checks, tid);
2588         if (!req) {
2589                 dout("%s tid %llu dne\n", __func__, tid);
2590                 goto out_unlock;
2591         }
2592
2593         dout("%s req %p tid %llu map_dne_bound %u newest %llu\n", __func__,
2594              req, req->r_tid, req->r_map_dne_bound, greq->u.newest);
2595         if (!req->r_map_dne_bound)
2596                 req->r_map_dne_bound = greq->u.newest;
2597         erase_request_mc(&osdc->map_checks, req);
2598         check_pool_dne(req);
2599
2600         ceph_osdc_put_request(req);
2601 out_unlock:
2602         up_write(&osdc->lock);
2603 }
2604
2605 static void send_map_check(struct ceph_osd_request *req)
2606 {
2607         struct ceph_osd_client *osdc = req->r_osdc;
2608         struct ceph_osd_request *lookup_req;
2609         int ret;
2610
2611         verify_osdc_wrlocked(osdc);
2612
2613         lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2614         if (lookup_req) {
2615                 WARN_ON(lookup_req != req);
2616                 return;
2617         }
2618
2619         ceph_osdc_get_request(req);
2620         insert_request_mc(&osdc->map_checks, req);
2621         ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
2622                                           map_check_cb, req->r_tid);
2623         WARN_ON(ret);
2624 }
2625
2626 /*
2627  * lingering requests, watch/notify v2 infrastructure
2628  */
2629 static void linger_release(struct kref *kref)
2630 {
2631         struct ceph_osd_linger_request *lreq =
2632             container_of(kref, struct ceph_osd_linger_request, kref);
2633
2634         dout("%s lreq %p reg_req %p ping_req %p\n", __func__, lreq,
2635              lreq->reg_req, lreq->ping_req);
2636         WARN_ON(!RB_EMPTY_NODE(&lreq->node));
2637         WARN_ON(!RB_EMPTY_NODE(&lreq->osdc_node));
2638         WARN_ON(!RB_EMPTY_NODE(&lreq->mc_node));
2639         WARN_ON(!list_empty(&lreq->scan_item));
2640         WARN_ON(!list_empty(&lreq->pending_lworks));
2641         WARN_ON(lreq->osd);
2642
2643         if (lreq->reg_req)
2644                 ceph_osdc_put_request(lreq->reg_req);
2645         if (lreq->ping_req)
2646                 ceph_osdc_put_request(lreq->ping_req);
2647         target_destroy(&lreq->t);
2648         kfree(lreq);
2649 }
2650
2651 static void linger_put(struct ceph_osd_linger_request *lreq)
2652 {
2653         if (lreq)
2654                 kref_put(&lreq->kref, linger_release);
2655 }
2656
2657 static struct ceph_osd_linger_request *
2658 linger_get(struct ceph_osd_linger_request *lreq)
2659 {
2660         kref_get(&lreq->kref);
2661         return lreq;
2662 }
2663
2664 static struct ceph_osd_linger_request *
2665 linger_alloc(struct ceph_osd_client *osdc)
2666 {
2667         struct ceph_osd_linger_request *lreq;
2668
2669         lreq = kzalloc(sizeof(*lreq), GFP_NOIO);
2670         if (!lreq)
2671                 return NULL;
2672
2673         kref_init(&lreq->kref);
2674         mutex_init(&lreq->lock);
2675         RB_CLEAR_NODE(&lreq->node);
2676         RB_CLEAR_NODE(&lreq->osdc_node);
2677         RB_CLEAR_NODE(&lreq->mc_node);
2678         INIT_LIST_HEAD(&lreq->scan_item);
2679         INIT_LIST_HEAD(&lreq->pending_lworks);
2680         init_completion(&lreq->reg_commit_wait);
2681         init_completion(&lreq->notify_finish_wait);
2682
2683         lreq->osdc = osdc;
2684         target_init(&lreq->t);
2685
2686         dout("%s lreq %p\n", __func__, lreq);
2687         return lreq;
2688 }
2689
2690 DEFINE_RB_INSDEL_FUNCS(linger, struct ceph_osd_linger_request, linger_id, node)
2691 DEFINE_RB_FUNCS(linger_osdc, struct ceph_osd_linger_request, linger_id, osdc_node)
2692 DEFINE_RB_FUNCS(linger_mc, struct ceph_osd_linger_request, linger_id, mc_node)
2693
2694 /*
2695  * Create linger request <-> OSD session relation.
2696  *
2697  * @lreq has to be registered, @osd may be homeless.
2698  */
2699 static void link_linger(struct ceph_osd *osd,
2700                         struct ceph_osd_linger_request *lreq)
2701 {
2702         verify_osd_locked(osd);
2703         WARN_ON(!lreq->linger_id || lreq->osd);
2704         dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2705              osd->o_osd, lreq, lreq->linger_id);
2706
2707         if (!osd_homeless(osd))
2708                 __remove_osd_from_lru(osd);
2709         else
2710                 atomic_inc(&osd->o_osdc->num_homeless);
2711
2712         get_osd(osd);
2713         insert_linger(&osd->o_linger_requests, lreq);
2714         lreq->osd = osd;
2715 }
2716
2717 static void unlink_linger(struct ceph_osd *osd,
2718                           struct ceph_osd_linger_request *lreq)
2719 {
2720         verify_osd_locked(osd);
2721         WARN_ON(lreq->osd != osd);
2722         dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2723              osd->o_osd, lreq, lreq->linger_id);
2724
2725         lreq->osd = NULL;
2726         erase_linger(&osd->o_linger_requests, lreq);
2727         put_osd(osd);
2728
2729         if (!osd_homeless(osd))
2730                 maybe_move_osd_to_lru(osd);
2731         else
2732                 atomic_dec(&osd->o_osdc->num_homeless);
2733 }
2734
2735 static bool __linger_registered(struct ceph_osd_linger_request *lreq)
2736 {
2737         verify_osdc_locked(lreq->osdc);
2738
2739         return !RB_EMPTY_NODE(&lreq->osdc_node);
2740 }
2741
2742 static bool linger_registered(struct ceph_osd_linger_request *lreq)
2743 {
2744         struct ceph_osd_client *osdc = lreq->osdc;
2745         bool registered;
2746
2747         down_read(&osdc->lock);
2748         registered = __linger_registered(lreq);
2749         up_read(&osdc->lock);
2750
2751         return registered;
2752 }
2753
2754 static void linger_register(struct ceph_osd_linger_request *lreq)
2755 {
2756         struct ceph_osd_client *osdc = lreq->osdc;
2757
2758         verify_osdc_wrlocked(osdc);
2759         WARN_ON(lreq->linger_id);
2760
2761         linger_get(lreq);
2762         lreq->linger_id = ++osdc->last_linger_id;
2763         insert_linger_osdc(&osdc->linger_requests, lreq);
2764 }
2765
2766 static void linger_unregister(struct ceph_osd_linger_request *lreq)
2767 {
2768         struct ceph_osd_client *osdc = lreq->osdc;
2769
2770         verify_osdc_wrlocked(osdc);
2771
2772         erase_linger_osdc(&osdc->linger_requests, lreq);
2773         linger_put(lreq);
2774 }
2775
2776 static void cancel_linger_request(struct ceph_osd_request *req)
2777 {
2778         struct ceph_osd_linger_request *lreq = req->r_priv;
2779
2780         WARN_ON(!req->r_linger);
2781         cancel_request(req);
2782         linger_put(lreq);
2783 }
2784
2785 struct linger_work {
2786         struct work_struct work;
2787         struct ceph_osd_linger_request *lreq;
2788         struct list_head pending_item;
2789         unsigned long queued_stamp;
2790
2791         union {
2792                 struct {
2793                         u64 notify_id;
2794                         u64 notifier_id;
2795                         void *payload; /* points into @msg front */
2796                         size_t payload_len;
2797
2798                         struct ceph_msg *msg; /* for ceph_msg_put() */
2799                 } notify;
2800                 struct {
2801                         int err;
2802                 } error;
2803         };
2804 };
2805
2806 static struct linger_work *lwork_alloc(struct ceph_osd_linger_request *lreq,
2807                                        work_func_t workfn)
2808 {
2809         struct linger_work *lwork;
2810
2811         lwork = kzalloc(sizeof(*lwork), GFP_NOIO);
2812         if (!lwork)
2813                 return NULL;
2814
2815         INIT_WORK(&lwork->work, workfn);
2816         INIT_LIST_HEAD(&lwork->pending_item);
2817         lwork->lreq = linger_get(lreq);
2818
2819         return lwork;
2820 }
2821
2822 static void lwork_free(struct linger_work *lwork)
2823 {
2824         struct ceph_osd_linger_request *lreq = lwork->lreq;
2825
2826         mutex_lock(&lreq->lock);
2827         list_del(&lwork->pending_item);
2828         mutex_unlock(&lreq->lock);
2829
2830         linger_put(lreq);
2831         kfree(lwork);
2832 }
2833
2834 static void lwork_queue(struct linger_work *lwork)
2835 {
2836         struct ceph_osd_linger_request *lreq = lwork->lreq;
2837         struct ceph_osd_client *osdc = lreq->osdc;
2838
2839         verify_lreq_locked(lreq);
2840         WARN_ON(!list_empty(&lwork->pending_item));
2841
2842         lwork->queued_stamp = jiffies;
2843         list_add_tail(&lwork->pending_item, &lreq->pending_lworks);
2844         queue_work(osdc->notify_wq, &lwork->work);
2845 }
2846
2847 static void do_watch_notify(struct work_struct *w)
2848 {
2849         struct linger_work *lwork = container_of(w, struct linger_work, work);
2850         struct ceph_osd_linger_request *lreq = lwork->lreq;
2851
2852         if (!linger_registered(lreq)) {
2853                 dout("%s lreq %p not registered\n", __func__, lreq);
2854                 goto out;
2855         }
2856
2857         WARN_ON(!lreq->is_watch);
2858         dout("%s lreq %p notify_id %llu notifier_id %llu payload_len %zu\n",
2859              __func__, lreq, lwork->notify.notify_id, lwork->notify.notifier_id,
2860              lwork->notify.payload_len);
2861         lreq->wcb(lreq->data, lwork->notify.notify_id, lreq->linger_id,
2862                   lwork->notify.notifier_id, lwork->notify.payload,
2863                   lwork->notify.payload_len);
2864
2865 out:
2866         ceph_msg_put(lwork->notify.msg);
2867         lwork_free(lwork);
2868 }
2869
2870 static void do_watch_error(struct work_struct *w)
2871 {
2872         struct linger_work *lwork = container_of(w, struct linger_work, work);
2873         struct ceph_osd_linger_request *lreq = lwork->lreq;
2874
2875         if (!linger_registered(lreq)) {
2876                 dout("%s lreq %p not registered\n", __func__, lreq);
2877                 goto out;
2878         }
2879
2880         dout("%s lreq %p err %d\n", __func__, lreq, lwork->error.err);
2881         lreq->errcb(lreq->data, lreq->linger_id, lwork->error.err);
2882
2883 out:
2884         lwork_free(lwork);
2885 }
2886
2887 static void queue_watch_error(struct ceph_osd_linger_request *lreq)
2888 {
2889         struct linger_work *lwork;
2890
2891         lwork = lwork_alloc(lreq, do_watch_error);
2892         if (!lwork) {
2893                 pr_err("failed to allocate error-lwork\n");
2894                 return;
2895         }
2896
2897         lwork->error.err = lreq->last_error;
2898         lwork_queue(lwork);
2899 }
2900
2901 static void linger_reg_commit_complete(struct ceph_osd_linger_request *lreq,
2902                                        int result)
2903 {
2904         if (!completion_done(&lreq->reg_commit_wait)) {
2905                 lreq->reg_commit_error = (result <= 0 ? result : 0);
2906                 complete_all(&lreq->reg_commit_wait);
2907         }
2908 }
2909
2910 static void linger_commit_cb(struct ceph_osd_request *req)
2911 {
2912         struct ceph_osd_linger_request *lreq = req->r_priv;
2913
2914         mutex_lock(&lreq->lock);
2915         dout("%s lreq %p linger_id %llu result %d\n", __func__, lreq,
2916              lreq->linger_id, req->r_result);
2917         linger_reg_commit_complete(lreq, req->r_result);
2918         lreq->committed = true;
2919
2920         if (!lreq->is_watch) {
2921                 struct ceph_osd_data *osd_data =
2922                     osd_req_op_data(req, 0, notify, response_data);
2923                 void *p = page_address(osd_data->pages[0]);
2924
2925                 WARN_ON(req->r_ops[0].op != CEPH_OSD_OP_NOTIFY ||
2926                         osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
2927
2928                 /* make note of the notify_id */
2929                 if (req->r_ops[0].outdata_len >= sizeof(u64)) {
2930                         lreq->notify_id = ceph_decode_64(&p);
2931                         dout("lreq %p notify_id %llu\n", lreq,
2932                              lreq->notify_id);
2933                 } else {
2934                         dout("lreq %p no notify_id\n", lreq);
2935                 }
2936         }
2937
2938         mutex_unlock(&lreq->lock);
2939         linger_put(lreq);
2940 }
2941
2942 static int normalize_watch_error(int err)
2943 {
2944         /*
2945          * Translate ENOENT -> ENOTCONN so that a delete->disconnection
2946          * notification and a failure to reconnect because we raced with
2947          * the delete appear the same to the user.
2948          */
2949         if (err == -ENOENT)
2950                 err = -ENOTCONN;
2951
2952         return err;
2953 }
2954
2955 static void linger_reconnect_cb(struct ceph_osd_request *req)
2956 {
2957         struct ceph_osd_linger_request *lreq = req->r_priv;
2958
2959         mutex_lock(&lreq->lock);
2960         dout("%s lreq %p linger_id %llu result %d last_error %d\n", __func__,
2961              lreq, lreq->linger_id, req->r_result, lreq->last_error);
2962         if (req->r_result < 0) {
2963                 if (!lreq->last_error) {
2964                         lreq->last_error = normalize_watch_error(req->r_result);
2965                         queue_watch_error(lreq);
2966                 }
2967         }
2968
2969         mutex_unlock(&lreq->lock);
2970         linger_put(lreq);
2971 }
2972
2973 static void send_linger(struct ceph_osd_linger_request *lreq)
2974 {
2975         struct ceph_osd_request *req = lreq->reg_req;
2976         struct ceph_osd_req_op *op = &req->r_ops[0];
2977
2978         verify_osdc_wrlocked(req->r_osdc);
2979         dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
2980
2981         if (req->r_osd)
2982                 cancel_linger_request(req);
2983
2984         request_reinit(req);
2985         ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
2986         ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
2987         req->r_flags = lreq->t.flags;
2988         req->r_mtime = lreq->mtime;
2989
2990         mutex_lock(&lreq->lock);
2991         if (lreq->is_watch && lreq->committed) {
2992                 WARN_ON(op->op != CEPH_OSD_OP_WATCH ||
2993                         op->watch.cookie != lreq->linger_id);
2994                 op->watch.op = CEPH_OSD_WATCH_OP_RECONNECT;
2995                 op->watch.gen = ++lreq->register_gen;
2996                 dout("lreq %p reconnect register_gen %u\n", lreq,
2997                      op->watch.gen);
2998                 req->r_callback = linger_reconnect_cb;
2999         } else {
3000                 if (!lreq->is_watch)
3001                         lreq->notify_id = 0;
3002                 else
3003                         WARN_ON(op->watch.op != CEPH_OSD_WATCH_OP_WATCH);
3004                 dout("lreq %p register\n", lreq);
3005                 req->r_callback = linger_commit_cb;
3006         }
3007         mutex_unlock(&lreq->lock);
3008
3009         req->r_priv = linger_get(lreq);
3010         req->r_linger = true;
3011
3012         submit_request(req, true);
3013 }
3014
3015 static void linger_ping_cb(struct ceph_osd_request *req)
3016 {
3017         struct ceph_osd_linger_request *lreq = req->r_priv;
3018
3019         mutex_lock(&lreq->lock);
3020         dout("%s lreq %p linger_id %llu result %d ping_sent %lu last_error %d\n",
3021              __func__, lreq, lreq->linger_id, req->r_result, lreq->ping_sent,
3022              lreq->last_error);
3023         if (lreq->register_gen == req->r_ops[0].watch.gen) {
3024                 if (!req->r_result) {
3025                         lreq->watch_valid_thru = lreq->ping_sent;
3026                 } else if (!lreq->last_error) {
3027                         lreq->last_error = normalize_watch_error(req->r_result);
3028                         queue_watch_error(lreq);
3029                 }
3030         } else {
3031                 dout("lreq %p register_gen %u ignoring old pong %u\n", lreq,
3032                      lreq->register_gen, req->r_ops[0].watch.gen);
3033         }
3034
3035         mutex_unlock(&lreq->lock);
3036         linger_put(lreq);
3037 }
3038
3039 static void send_linger_ping(struct ceph_osd_linger_request *lreq)
3040 {
3041         struct ceph_osd_client *osdc = lreq->osdc;
3042         struct ceph_osd_request *req = lreq->ping_req;
3043         struct ceph_osd_req_op *op = &req->r_ops[0];
3044
3045         if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
3046                 dout("%s PAUSERD\n", __func__);
3047                 return;
3048         }
3049
3050         lreq->ping_sent = jiffies;
3051         dout("%s lreq %p linger_id %llu ping_sent %lu register_gen %u\n",
3052              __func__, lreq, lreq->linger_id, lreq->ping_sent,
3053              lreq->register_gen);
3054
3055         if (req->r_osd)
3056                 cancel_linger_request(req);
3057
3058         request_reinit(req);
3059         target_copy(&req->r_t, &lreq->t);
3060
3061         WARN_ON(op->op != CEPH_OSD_OP_WATCH ||
3062                 op->watch.cookie != lreq->linger_id ||
3063                 op->watch.op != CEPH_OSD_WATCH_OP_PING);
3064         op->watch.gen = lreq->register_gen;
3065         req->r_callback = linger_ping_cb;
3066         req->r_priv = linger_get(lreq);
3067         req->r_linger = true;
3068
3069         ceph_osdc_get_request(req);
3070         account_request(req);
3071         req->r_tid = atomic64_inc_return(&osdc->last_tid);
3072         link_request(lreq->osd, req);
3073         send_request(req);
3074 }
3075
3076 static void linger_submit(struct ceph_osd_linger_request *lreq)
3077 {
3078         struct ceph_osd_client *osdc = lreq->osdc;
3079         struct ceph_osd *osd;
3080
3081         down_write(&osdc->lock);
3082         linger_register(lreq);
3083         if (lreq->is_watch) {
3084                 lreq->reg_req->r_ops[0].watch.cookie = lreq->linger_id;
3085                 lreq->ping_req->r_ops[0].watch.cookie = lreq->linger_id;
3086         } else {
3087                 lreq->reg_req->r_ops[0].notify.cookie = lreq->linger_id;
3088         }
3089
3090         calc_target(osdc, &lreq->t, NULL, false);
3091         osd = lookup_create_osd(osdc, lreq->t.osd, true);
3092         link_linger(osd, lreq);
3093
3094         send_linger(lreq);
3095         up_write(&osdc->lock);
3096 }
3097
3098 static void cancel_linger_map_check(struct ceph_osd_linger_request *lreq)
3099 {
3100         struct ceph_osd_client *osdc = lreq->osdc;
3101         struct ceph_osd_linger_request *lookup_lreq;
3102
3103         verify_osdc_wrlocked(osdc);
3104
3105         lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3106                                        lreq->linger_id);
3107         if (!lookup_lreq)
3108                 return;
3109
3110         WARN_ON(lookup_lreq != lreq);
3111         erase_linger_mc(&osdc->linger_map_checks, lreq);
3112         linger_put(lreq);
3113 }
3114
3115 /*
3116  * @lreq has to be both registered and linked.
3117  */
3118 static void __linger_cancel(struct ceph_osd_linger_request *lreq)
3119 {
3120         if (lreq->is_watch && lreq->ping_req->r_osd)
3121                 cancel_linger_request(lreq->ping_req);
3122         if (lreq->reg_req->r_osd)
3123                 cancel_linger_request(lreq->reg_req);
3124         cancel_linger_map_check(lreq);
3125         unlink_linger(lreq->osd, lreq);
3126         linger_unregister(lreq);
3127 }
3128
3129 static void linger_cancel(struct ceph_osd_linger_request *lreq)
3130 {
3131         struct ceph_osd_client *osdc = lreq->osdc;
3132
3133         down_write(&osdc->lock);
3134         if (__linger_registered(lreq))
3135                 __linger_cancel(lreq);
3136         up_write(&osdc->lock);
3137 }
3138
3139 static void send_linger_map_check(struct ceph_osd_linger_request *lreq);
3140
3141 static void check_linger_pool_dne(struct ceph_osd_linger_request *lreq)
3142 {
3143         struct ceph_osd_client *osdc = lreq->osdc;
3144         struct ceph_osdmap *map = osdc->osdmap;
3145
3146         verify_osdc_wrlocked(osdc);
3147         WARN_ON(!map->epoch);
3148
3149         if (lreq->register_gen) {
3150                 lreq->map_dne_bound = map->epoch;
3151                 dout("%s lreq %p linger_id %llu pool disappeared\n", __func__,
3152                      lreq, lreq->linger_id);
3153         } else {
3154                 dout("%s lreq %p linger_id %llu map_dne_bound %u have %u\n",
3155                      __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3156                      map->epoch);
3157         }
3158
3159         if (lreq->map_dne_bound) {
3160                 if (map->epoch >= lreq->map_dne_bound) {
3161                         /* we had a new enough map */
3162                         pr_info("linger_id %llu pool does not exist\n",
3163                                 lreq->linger_id);
3164                         linger_reg_commit_complete(lreq, -ENOENT);
3165                         __linger_cancel(lreq);
3166                 }
3167         } else {
3168                 send_linger_map_check(lreq);
3169         }
3170 }
3171
3172 static void linger_map_check_cb(struct ceph_mon_generic_request *greq)
3173 {
3174         struct ceph_osd_client *osdc = &greq->monc->client->osdc;
3175         struct ceph_osd_linger_request *lreq;
3176         u64 linger_id = greq->private_data;
3177
3178         WARN_ON(greq->result || !greq->u.newest);
3179
3180         down_write(&osdc->lock);
3181         lreq = lookup_linger_mc(&osdc->linger_map_checks, linger_id);
3182         if (!lreq) {
3183                 dout("%s linger_id %llu dne\n", __func__, linger_id);
3184                 goto out_unlock;
3185         }
3186
3187         dout("%s lreq %p linger_id %llu map_dne_bound %u newest %llu\n",
3188              __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3189              greq->u.newest);
3190         if (!lreq->map_dne_bound)
3191                 lreq->map_dne_bound = greq->u.newest;
3192         erase_linger_mc(&osdc->linger_map_checks, lreq);
3193         check_linger_pool_dne(lreq);
3194
3195         linger_put(lreq);
3196 out_unlock:
3197         up_write(&osdc->lock);
3198 }
3199
3200 static void send_linger_map_check(struct ceph_osd_linger_request *lreq)
3201 {
3202         struct ceph_osd_client *osdc = lreq->osdc;
3203         struct ceph_osd_linger_request *lookup_lreq;
3204         int ret;
3205
3206         verify_osdc_wrlocked(osdc);
3207
3208         lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3209                                        lreq->linger_id);
3210         if (lookup_lreq) {
3211                 WARN_ON(lookup_lreq != lreq);
3212                 return;
3213         }
3214
3215         linger_get(lreq);
3216         insert_linger_mc(&osdc->linger_map_checks, lreq);
3217         ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
3218                                           linger_map_check_cb, lreq->linger_id);
3219         WARN_ON(ret);
3220 }
3221
3222 static int linger_reg_commit_wait(struct ceph_osd_linger_request *lreq)
3223 {
3224         int ret;
3225
3226         dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3227         ret = wait_for_completion_interruptible(&lreq->reg_commit_wait);
3228         return ret ?: lreq->reg_commit_error;
3229 }
3230
3231 static int linger_notify_finish_wait(struct ceph_osd_linger_request *lreq)
3232 {
3233         int ret;
3234
3235         dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3236         ret = wait_for_completion_interruptible(&lreq->notify_finish_wait);
3237         return ret ?: lreq->notify_finish_error;
3238 }
3239
3240 /*
3241  * Timeout callback, called every N seconds.  When 1 or more OSD
3242  * requests has been active for more than N seconds, we send a keepalive
3243  * (tag + timestamp) to its OSD to ensure any communications channel
3244  * reset is detected.
3245  */
3246 static void handle_timeout(struct work_struct *work)
3247 {
3248         struct ceph_osd_client *osdc =
3249                 container_of(work, struct ceph_osd_client, timeout_work.work);
3250         struct ceph_options *opts = osdc->client->options;
3251         unsigned long cutoff = jiffies - opts->osd_keepalive_timeout;
3252         unsigned long expiry_cutoff = jiffies - opts->osd_request_timeout;
3253         LIST_HEAD(slow_osds);
3254         struct rb_node *n, *p;
3255
3256         dout("%s osdc %p\n", __func__, osdc);
3257         down_write(&osdc->lock);
3258
3259         /*
3260          * ping osds that are a bit slow.  this ensures that if there
3261          * is a break in the TCP connection we will notice, and reopen
3262          * a connection with that osd (from the fault callback).
3263          */
3264         for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
3265                 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
3266                 bool found = false;
3267
3268                 for (p = rb_first(&osd->o_requests); p; ) {
3269                         struct ceph_osd_request *req =
3270                             rb_entry(p, struct ceph_osd_request, r_node);
3271
3272                         p = rb_next(p); /* abort_request() */
3273
3274                         if (time_before(req->r_stamp, cutoff)) {
3275                                 dout(" req %p tid %llu on osd%d is laggy\n",
3276                                      req, req->r_tid, osd->o_osd);
3277                                 found = true;
3278                         }
3279                         if (opts->osd_request_timeout &&
3280                             time_before(req->r_start_stamp, expiry_cutoff)) {
3281                                 pr_err_ratelimited("tid %llu on osd%d timeout\n",
3282                                        req->r_tid, osd->o_osd);
3283                                 abort_request(req, -ETIMEDOUT);
3284                         }
3285                 }
3286                 for (p = rb_first(&osd->o_linger_requests); p; p = rb_next(p)) {
3287                         struct ceph_osd_linger_request *lreq =
3288                             rb_entry(p, struct ceph_osd_linger_request, node);
3289
3290                         dout(" lreq %p linger_id %llu is served by osd%d\n",
3291                              lreq, lreq->linger_id, osd->o_osd);
3292                         found = true;
3293
3294                         mutex_lock(&lreq->lock);
3295                         if (lreq->is_watch && lreq->committed && !lreq->last_error)
3296                                 send_linger_ping(lreq);
3297                         mutex_unlock(&lreq->lock);
3298                 }
3299
3300                 if (found)
3301                         list_move_tail(&osd->o_keepalive_item, &slow_osds);
3302         }
3303
3304         if (opts->osd_request_timeout) {
3305                 for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
3306                         struct ceph_osd_request *req =
3307                             rb_entry(p, struct ceph_osd_request, r_node);
3308
3309                         p = rb_next(p); /* abort_request() */
3310
3311                         if (time_before(req->r_start_stamp, expiry_cutoff)) {
3312                                 pr_err_ratelimited("tid %llu on osd%d timeout\n",
3313                                        req->r_tid, osdc->homeless_osd.o_osd);
3314                                 abort_request(req, -ETIMEDOUT);
3315                         }
3316                 }
3317         }
3318
3319         if (atomic_read(&osdc->num_homeless) || !list_empty(&slow_osds))
3320                 maybe_request_map(osdc);
3321
3322         while (!list_empty(&slow_osds)) {
3323                 struct ceph_osd *osd = list_first_entry(&slow_osds,
3324                                                         struct ceph_osd,
3325                                                         o_keepalive_item);
3326                 list_del_init(&osd->o_keepalive_item);
3327                 ceph_con_keepalive(&osd->o_con);
3328         }
3329
3330         up_write(&osdc->lock);
3331         schedule_delayed_work(&osdc->timeout_work,
3332                               osdc->client->options->osd_keepalive_timeout);
3333 }
3334
3335 static void handle_osds_timeout(struct work_struct *work)
3336 {
3337         struct ceph_osd_client *osdc =
3338                 container_of(work, struct ceph_osd_client,
3339                              osds_timeout_work.work);
3340         unsigned long delay = osdc->client->options->osd_idle_ttl / 4;
3341         struct ceph_osd *osd, *nosd;
3342
3343         dout("%s osdc %p\n", __func__, osdc);
3344         down_write(&osdc->lock);
3345         list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) {
3346                 if (time_before(jiffies, osd->lru_ttl))
3347                         break;
3348
3349                 WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
3350                 WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
3351                 close_osd(osd);
3352         }
3353
3354         up_write(&osdc->lock);
3355         schedule_delayed_work(&osdc->osds_timeout_work,
3356                               round_jiffies_relative(delay));
3357 }
3358
3359 static int ceph_oloc_decode(void **p, void *end,
3360                             struct ceph_object_locator *oloc)
3361 {
3362         u8 struct_v, struct_cv;
3363         u32 len;
3364         void *struct_end;
3365         int ret = 0;
3366
3367         ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3368         struct_v = ceph_decode_8(p);
3369         struct_cv = ceph_decode_8(p);
3370         if (struct_v < 3) {
3371                 pr_warn("got v %d < 3 cv %d of ceph_object_locator\n",
3372                         struct_v, struct_cv);
3373                 goto e_inval;
3374         }
3375         if (struct_cv > 6) {
3376                 pr_warn("got v %d cv %d > 6 of ceph_object_locator\n",
3377                         struct_v, struct_cv);
3378                 goto e_inval;
3379         }
3380         len = ceph_decode_32(p);
3381         ceph_decode_need(p, end, len, e_inval);
3382         struct_end = *p + len;
3383
3384         oloc->pool = ceph_decode_64(p);
3385         *p += 4; /* skip preferred */
3386
3387         len = ceph_decode_32(p);
3388         if (len > 0) {
3389                 pr_warn("ceph_object_locator::key is set\n");
3390                 goto e_inval;
3391         }
3392
3393         if (struct_v >= 5) {
3394                 bool changed = false;
3395
3396                 len = ceph_decode_32(p);
3397                 if (len > 0) {
3398                         ceph_decode_need(p, end, len, e_inval);
3399                         if (!oloc->pool_ns ||
3400                             ceph_compare_string(oloc->pool_ns, *p, len))
3401                                 changed = true;
3402                         *p += len;
3403                 } else {
3404                         if (oloc->pool_ns)
3405                                 changed = true;
3406                 }
3407                 if (changed) {
3408                         /* redirect changes namespace */
3409                         pr_warn("ceph_object_locator::nspace is changed\n");
3410                         goto e_inval;
3411                 }
3412         }
3413
3414         if (struct_v >= 6) {
3415                 s64 hash = ceph_decode_64(p);
3416                 if (hash != -1) {
3417                         pr_warn("ceph_object_locator::hash is set\n");
3418                         goto e_inval;
3419                 }
3420         }
3421
3422         /* skip the rest */
3423         *p = struct_end;
3424 out:
3425         return ret;
3426
3427 e_inval:
3428         ret = -EINVAL;
3429         goto out;
3430 }
3431
3432 static int ceph_redirect_decode(void **p, void *end,
3433                                 struct ceph_request_redirect *redir)
3434 {
3435         u8 struct_v, struct_cv;
3436         u32 len;
3437         void *struct_end;
3438         int ret;
3439
3440         ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3441         struct_v = ceph_decode_8(p);
3442         struct_cv = ceph_decode_8(p);
3443         if (struct_cv > 1) {
3444                 pr_warn("got v %d cv %d > 1 of ceph_request_redirect\n",
3445                         struct_v, struct_cv);
3446                 goto e_inval;
3447         }
3448         len = ceph_decode_32(p);
3449         ceph_decode_need(p, end, len, e_inval);
3450         struct_end = *p + len;
3451
3452         ret = ceph_oloc_decode(p, end, &redir->oloc);
3453         if (ret)
3454                 goto out;
3455
3456         len = ceph_decode_32(p);
3457         if (len > 0) {
3458                 pr_warn("ceph_request_redirect::object_name is set\n");
3459                 goto e_inval;
3460         }
3461
3462         len = ceph_decode_32(p);
3463         *p += len; /* skip osd_instructions */
3464
3465         /* skip the rest */
3466         *p = struct_end;
3467 out:
3468         return ret;
3469
3470 e_inval:
3471         ret = -EINVAL;
3472         goto out;
3473 }
3474
3475 struct MOSDOpReply {
3476         struct ceph_pg pgid;
3477         u64 flags;
3478         int result;
3479         u32 epoch;
3480         int num_ops;
3481         u32 outdata_len[CEPH_OSD_MAX_OPS];
3482         s32 rval[CEPH_OSD_MAX_OPS];
3483         int retry_attempt;
3484         struct ceph_eversion replay_version;
3485         u64 user_version;
3486         struct ceph_request_redirect redirect;
3487 };
3488
3489 static int decode_MOSDOpReply(const struct ceph_msg *msg, struct MOSDOpReply *m)
3490 {
3491         void *p = msg->front.iov_base;
3492         void *const end = p + msg->front.iov_len;
3493         u16 version = le16_to_cpu(msg->hdr.version);
3494         struct ceph_eversion bad_replay_version;
3495         u8 decode_redir;
3496         u32 len;
3497         int ret;
3498         int i;
3499
3500         ceph_decode_32_safe(&p, end, len, e_inval);
3501         ceph_decode_need(&p, end, len, e_inval);
3502         p += len; /* skip oid */
3503
3504         ret = ceph_decode_pgid(&p, end, &m->pgid);
3505         if (ret)
3506                 return ret;
3507
3508         ceph_decode_64_safe(&p, end, m->flags, e_inval);
3509         ceph_decode_32_safe(&p, end, m->result, e_inval);
3510         ceph_decode_need(&p, end, sizeof(bad_replay_version), e_inval);
3511         memcpy(&bad_replay_version, p, sizeof(bad_replay_version));
3512         p += sizeof(bad_replay_version);
3513         ceph_decode_32_safe(&p, end, m->epoch, e_inval);
3514
3515         ceph_decode_32_safe(&p, end, m->num_ops, e_inval);
3516         if (m->num_ops > ARRAY_SIZE(m->outdata_len))
3517                 goto e_inval;
3518
3519         ceph_decode_need(&p, end, m->num_ops * sizeof(struct ceph_osd_op),
3520                          e_inval);
3521         for (i = 0; i < m->num_ops; i++) {
3522                 struct ceph_osd_op *op = p;
3523
3524                 m->outdata_len[i] = le32_to_cpu(op->payload_len);
3525                 p += sizeof(*op);
3526         }
3527
3528         ceph_decode_32_safe(&p, end, m->retry_attempt, e_inval);
3529         for (i = 0; i < m->num_ops; i++)
3530                 ceph_decode_32_safe(&p, end, m->rval[i], e_inval);
3531
3532         if (version >= 5) {
3533                 ceph_decode_need(&p, end, sizeof(m->replay_version), e_inval);
3534                 memcpy(&m->replay_version, p, sizeof(m->replay_version));
3535                 p += sizeof(m->replay_version);
3536                 ceph_decode_64_safe(&p, end, m->user_version, e_inval);
3537         } else {
3538                 m->replay_version = bad_replay_version; /* struct */
3539                 m->user_version = le64_to_cpu(m->replay_version.version);
3540         }
3541
3542         if (version >= 6) {
3543                 if (version >= 7)
3544                         ceph_decode_8_safe(&p, end, decode_redir, e_inval);
3545                 else
3546                         decode_redir = 1;
3547         } else {
3548                 decode_redir = 0;
3549         }
3550
3551         if (decode_redir) {
3552                 ret = ceph_redirect_decode(&p, end, &m->redirect);
3553                 if (ret)
3554                         return ret;
3555         } else {
3556                 ceph_oloc_init(&m->redirect.oloc);
3557         }
3558
3559         return 0;
3560
3561 e_inval:
3562         return -EINVAL;
3563 }
3564
3565 /*
3566  * Handle MOSDOpReply.  Set ->r_result and call the callback if it is
3567  * specified.
3568  */
3569 static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg)
3570 {
3571         struct ceph_osd_client *osdc = osd->o_osdc;
3572         struct ceph_osd_request *req;
3573         struct MOSDOpReply m;
3574         u64 tid = le64_to_cpu(msg->hdr.tid);
3575         u32 data_len = 0;
3576         int ret;
3577         int i;
3578
3579         dout("%s msg %p tid %llu\n", __func__, msg, tid);
3580
3581         down_read(&osdc->lock);
3582         if (!osd_registered(osd)) {
3583                 dout("%s osd%d unknown\n", __func__, osd->o_osd);
3584                 goto out_unlock_osdc;
3585         }
3586         WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
3587
3588         mutex_lock(&osd->lock);
3589         req = lookup_request(&osd->o_requests, tid);
3590         if (!req) {
3591                 dout("%s osd%d tid %llu unknown\n", __func__, osd->o_osd, tid);
3592                 goto out_unlock_session;
3593         }
3594
3595         m.redirect.oloc.pool_ns = req->r_t.target_oloc.pool_ns;
3596         ret = decode_MOSDOpReply(msg, &m);
3597         m.redirect.oloc.pool_ns = NULL;
3598         if (ret) {
3599                 pr_err("failed to decode MOSDOpReply for tid %llu: %d\n",
3600                        req->r_tid, ret);
3601                 ceph_msg_dump(msg);
3602                 goto fail_request;
3603         }
3604         dout("%s req %p tid %llu flags 0x%llx pgid %llu.%x epoch %u attempt %d v %u'%llu uv %llu\n",
3605              __func__, req, req->r_tid, m.flags, m.pgid.pool, m.pgid.seed,
3606              m.epoch, m.retry_attempt, le32_to_cpu(m.replay_version.epoch),
3607              le64_to_cpu(m.replay_version.version), m.user_version);
3608
3609         if (m.retry_attempt >= 0) {
3610                 if (m.retry_attempt != req->r_attempts - 1) {
3611                         dout("req %p tid %llu retry_attempt %d != %d, ignoring\n",
3612                              req, req->r_tid, m.retry_attempt,
3613                              req->r_attempts - 1);
3614                         goto out_unlock_session;
3615                 }
3616         } else {
3617                 WARN_ON(1); /* MOSDOpReply v4 is assumed */
3618         }
3619
3620         if (!ceph_oloc_empty(&m.redirect.oloc)) {
3621                 dout("req %p tid %llu redirect pool %lld\n", req, req->r_tid,
3622                      m.redirect.oloc.pool);
3623                 unlink_request(osd, req);
3624                 mutex_unlock(&osd->lock);
3625
3626                 /*
3627                  * Not ceph_oloc_copy() - changing pool_ns is not
3628                  * supported.
3629                  */
3630                 req->r_t.target_oloc.pool = m.redirect.oloc.pool;
3631                 req->r_flags |= CEPH_OSD_FLAG_REDIRECTED;
3632                 req->r_tid = 0;
3633                 __submit_request(req, false);
3634                 goto out_unlock_osdc;
3635         }
3636
3637         if (m.num_ops != req->r_num_ops) {
3638                 pr_err("num_ops %d != %d for tid %llu\n", m.num_ops,
3639                        req->r_num_ops, req->r_tid);
3640                 goto fail_request;
3641         }
3642         for (i = 0; i < req->r_num_ops; i++) {
3643                 dout(" req %p tid %llu op %d rval %d len %u\n", req,
3644                      req->r_tid, i, m.rval[i], m.outdata_len[i]);
3645                 req->r_ops[i].rval = m.rval[i];
3646                 req->r_ops[i].outdata_len = m.outdata_len[i];
3647                 data_len += m.outdata_len[i];
3648         }
3649         if (data_len != le32_to_cpu(msg->hdr.data_len)) {
3650                 pr_err("sum of lens %u != %u for tid %llu\n", data_len,
3651                        le32_to_cpu(msg->hdr.data_len), req->r_tid);
3652                 goto fail_request;
3653         }
3654         dout("%s req %p tid %llu result %d data_len %u\n", __func__,
3655              req, req->r_tid, m.result, data_len);
3656
3657         /*
3658          * Since we only ever request ONDISK, we should only ever get
3659          * one (type of) reply back.
3660          */
3661         WARN_ON(!(m.flags & CEPH_OSD_FLAG_ONDISK));
3662         req->r_result = m.result ?: data_len;
3663         finish_request(req);
3664         mutex_unlock(&osd->lock);
3665         up_read(&osdc->lock);
3666
3667         __complete_request(req);
3668         return;
3669
3670 fail_request:
3671         complete_request(req, -EIO);
3672 out_unlock_session:
3673         mutex_unlock(&osd->lock);
3674 out_unlock_osdc:
3675         up_read(&osdc->lock);
3676 }
3677
3678 static void set_pool_was_full(struct ceph_osd_client *osdc)
3679 {
3680         struct rb_node *n;
3681
3682         for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
3683                 struct ceph_pg_pool_info *pi =
3684                     rb_entry(n, struct ceph_pg_pool_info, node);
3685
3686                 pi->was_full = __pool_full(pi);
3687         }
3688 }
3689
3690 static bool pool_cleared_full(struct ceph_osd_client *osdc, s64 pool_id)
3691 {
3692         struct ceph_pg_pool_info *pi;
3693
3694         pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
3695         if (!pi)
3696                 return false;
3697
3698         return pi->was_full && !__pool_full(pi);
3699 }
3700
3701 static enum calc_target_result
3702 recalc_linger_target(struct ceph_osd_linger_request *lreq)
3703 {
3704         struct ceph_osd_client *osdc = lreq->osdc;
3705         enum calc_target_result ct_res;
3706
3707         ct_res = calc_target(osdc, &lreq->t, NULL, true);
3708         if (ct_res == CALC_TARGET_NEED_RESEND) {
3709                 struct ceph_osd *osd;
3710
3711                 osd = lookup_create_osd(osdc, lreq->t.osd, true);
3712                 if (osd != lreq->osd) {
3713                         unlink_linger(lreq->osd, lreq);
3714                         link_linger(osd, lreq);
3715                 }
3716         }
3717
3718         return ct_res;
3719 }
3720
3721 /*
3722  * Requeue requests whose mapping to an OSD has changed.
3723  */
3724 static void scan_requests(struct ceph_osd *osd,
3725                           bool force_resend,
3726                           bool cleared_full,
3727                           bool check_pool_cleared_full,
3728                           struct rb_root *need_resend,
3729                           struct list_head *need_resend_linger)
3730 {
3731         struct ceph_osd_client *osdc = osd->o_osdc;
3732         struct rb_node *n;
3733         bool force_resend_writes;
3734
3735         for (n = rb_first(&osd->o_linger_requests); n; ) {
3736                 struct ceph_osd_linger_request *lreq =
3737                     rb_entry(n, struct ceph_osd_linger_request, node);
3738                 enum calc_target_result ct_res;
3739
3740                 n = rb_next(n); /* recalc_linger_target() */
3741
3742                 dout("%s lreq %p linger_id %llu\n", __func__, lreq,
3743                      lreq->linger_id);
3744                 ct_res = recalc_linger_target(lreq);
3745                 switch (ct_res) {
3746                 case CALC_TARGET_NO_ACTION:
3747                         force_resend_writes = cleared_full ||
3748                             (check_pool_cleared_full &&
3749                              pool_cleared_full(osdc, lreq->t.base_oloc.pool));
3750                         if (!force_resend && !force_resend_writes)
3751                                 break;
3752
3753                         /* fall through */
3754                 case CALC_TARGET_NEED_RESEND:
3755                         cancel_linger_map_check(lreq);
3756                         /*
3757                          * scan_requests() for the previous epoch(s)
3758                          * may have already added it to the list, since
3759                          * it's not unlinked here.
3760                          */
3761                         if (list_empty(&lreq->scan_item))
3762                                 list_add_tail(&lreq->scan_item, need_resend_linger);
3763                         break;
3764                 case CALC_TARGET_POOL_DNE:
3765                         list_del_init(&lreq->scan_item);
3766                         check_linger_pool_dne(lreq);
3767                         break;
3768                 }
3769         }
3770
3771         for (n = rb_first(&osd->o_requests); n; ) {
3772                 struct ceph_osd_request *req =
3773                     rb_entry(n, struct ceph_osd_request, r_node);
3774                 enum calc_target_result ct_res;
3775
3776                 n = rb_next(n); /* unlink_request(), check_pool_dne() */
3777
3778                 dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
3779                 ct_res = calc_target(osdc, &req->r_t, &req->r_osd->o_con,
3780                                      false);
3781                 switch (ct_res) {
3782                 case CALC_TARGET_NO_ACTION:
3783                         force_resend_writes = cleared_full ||
3784                             (check_pool_cleared_full &&
3785                              pool_cleared_full(osdc, req->r_t.base_oloc.pool));
3786                         if (!force_resend &&
3787                             (!(req->r_flags & CEPH_OSD_FLAG_WRITE) ||
3788                              !force_resend_writes))
3789                                 break;
3790
3791                         /* fall through */
3792                 case CALC_TARGET_NEED_RESEND:
3793                         cancel_map_check(req);
3794                         unlink_request(osd, req);
3795                         insert_request(need_resend, req);
3796                         break;
3797                 case CALC_TARGET_POOL_DNE:
3798                         check_pool_dne(req);
3799                         break;
3800                 }
3801         }
3802 }
3803
3804 static int handle_one_map(struct ceph_osd_client *osdc,
3805                           void *p, void *end, bool incremental,
3806                           struct rb_root *need_resend,
3807                           struct list_head *need_resend_linger)
3808 {
3809         struct ceph_osdmap *newmap;
3810         struct rb_node *n;
3811         bool skipped_map = false;
3812         bool was_full;
3813
3814         was_full = ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
3815         set_pool_was_full(osdc);
3816
3817         if (incremental)
3818                 newmap = osdmap_apply_incremental(&p, end, osdc->osdmap);
3819         else
3820                 newmap = ceph_osdmap_decode(&p, end);
3821         if (IS_ERR(newmap))
3822                 return PTR_ERR(newmap);
3823
3824         if (newmap != osdc->osdmap) {
3825                 /*
3826                  * Preserve ->was_full before destroying the old map.
3827                  * For pools that weren't in the old map, ->was_full
3828                  * should be false.
3829                  */
3830                 for (n = rb_first(&newmap->pg_pools); n; n = rb_next(n)) {
3831                         struct ceph_pg_pool_info *pi =
3832                             rb_entry(n, struct ceph_pg_pool_info, node);
3833                         struct ceph_pg_pool_info *old_pi;
3834
3835                         old_pi = ceph_pg_pool_by_id(osdc->osdmap, pi->id);
3836                         if (old_pi)
3837                                 pi->was_full = old_pi->was_full;
3838                         else
3839                                 WARN_ON(pi->was_full);
3840                 }
3841
3842                 if (osdc->osdmap->epoch &&
3843                     osdc->osdmap->epoch + 1 < newmap->epoch) {
3844                         WARN_ON(incremental);
3845                         skipped_map = true;
3846                 }
3847
3848                 ceph_osdmap_destroy(osdc->osdmap);
3849                 osdc->osdmap = newmap;
3850         }
3851
3852         was_full &= !ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
3853         scan_requests(&osdc->homeless_osd, skipped_map, was_full, true,
3854                       need_resend, need_resend_linger);
3855
3856         for (n = rb_first(&osdc->osds); n; ) {
3857                 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
3858
3859                 n = rb_next(n); /* close_osd() */
3860
3861                 scan_requests(osd, skipped_map, was_full, true, need_resend,
3862                               need_resend_linger);
3863                 if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) ||
3864                     memcmp(&osd->o_con.peer_addr,
3865                            ceph_osd_addr(osdc->osdmap, osd->o_osd),
3866                            sizeof(struct ceph_entity_addr)))
3867                         close_osd(osd);
3868         }
3869
3870         return 0;
3871 }
3872
3873 static void kick_requests(struct ceph_osd_client *osdc,
3874                           struct rb_root *need_resend,
3875                           struct list_head *need_resend_linger)
3876 {
3877         struct ceph_osd_linger_request *lreq, *nlreq;
3878         enum calc_target_result ct_res;
3879         struct rb_node *n;
3880
3881         /* make sure need_resend targets reflect latest map */
3882         for (n = rb_first(need_resend); n; ) {
3883                 struct ceph_osd_request *req =
3884                     rb_entry(n, struct ceph_osd_request, r_node);
3885
3886                 n = rb_next(n);
3887
3888                 if (req->r_t.epoch < osdc->osdmap->epoch) {
3889                         ct_res = calc_target(osdc, &req->r_t, NULL, false);
3890                         if (ct_res == CALC_TARGET_POOL_DNE) {
3891                                 erase_request(need_resend, req);
3892                                 check_pool_dne(req);
3893                         }
3894                 }
3895         }
3896
3897         for (n = rb_first(need_resend); n; ) {
3898                 struct ceph_osd_request *req =
3899                     rb_entry(n, struct ceph_osd_request, r_node);
3900                 struct ceph_osd *osd;
3901
3902                 n = rb_next(n);
3903                 erase_request(need_resend, req); /* before link_request() */
3904
3905                 osd = lookup_create_osd(osdc, req->r_t.osd, true);
3906                 link_request(osd, req);
3907                 if (!req->r_linger) {
3908                         if (!osd_homeless(osd) && !req->r_t.paused)
3909                                 send_request(req);
3910                 } else {
3911                         cancel_linger_request(req);
3912                 }
3913         }
3914
3915         list_for_each_entry_safe(lreq, nlreq, need_resend_linger, scan_item) {
3916                 if (!osd_homeless(lreq->osd))
3917                         send_linger(lreq);
3918
3919                 list_del_init(&lreq->scan_item);
3920         }
3921 }
3922
3923 /*
3924  * Process updated osd map.
3925  *
3926  * The message contains any number of incremental and full maps, normally
3927  * indicating some sort of topology change in the cluster.  Kick requests
3928  * off to different OSDs as needed.
3929  */
3930 void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
3931 {
3932         void *p = msg->front.iov_base;
3933         void *const end = p + msg->front.iov_len;
3934         u32 nr_maps, maplen;
3935         u32 epoch;
3936         struct ceph_fsid fsid;
3937         struct rb_root need_resend = RB_ROOT;
3938         LIST_HEAD(need_resend_linger);
3939         bool handled_incremental = false;
3940         bool was_pauserd, was_pausewr;
3941         bool pauserd, pausewr;
3942         int err;
3943
3944         dout("%s have %u\n", __func__, osdc->osdmap->epoch);
3945         down_write(&osdc->lock);
3946
3947         /* verify fsid */
3948         ceph_decode_need(&p, end, sizeof(fsid), bad);
3949         ceph_decode_copy(&p, &fsid, sizeof(fsid));
3950         if (ceph_check_fsid(osdc->client, &fsid) < 0)
3951                 goto bad;
3952
3953         was_pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
3954         was_pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
3955                       ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
3956                       have_pool_full(osdc);
3957
3958         /* incremental maps */
3959         ceph_decode_32_safe(&p, end, nr_maps, bad);
3960         dout(" %d inc maps\n", nr_maps);
3961         while (nr_maps > 0) {
3962                 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
3963                 epoch = ceph_decode_32(&p);
3964                 maplen = ceph_decode_32(&p);
3965                 ceph_decode_need(&p, end, maplen, bad);
3966                 if (osdc->osdmap->epoch &&
3967                     osdc->osdmap->epoch + 1 == epoch) {
3968                         dout("applying incremental map %u len %d\n",
3969                              epoch, maplen);
3970                         err = handle_one_map(osdc, p, p + maplen, true,
3971                                              &need_resend, &need_resend_linger);
3972                         if (err)
3973                                 goto bad;
3974                         handled_incremental = true;
3975                 } else {
3976                         dout("ignoring incremental map %u len %d\n",
3977                              epoch, maplen);
3978                 }
3979                 p += maplen;
3980                 nr_maps--;
3981         }
3982         if (handled_incremental)
3983                 goto done;
3984
3985         /* full maps */
3986         ceph_decode_32_safe(&p, end, nr_maps, bad);
3987         dout(" %d full maps\n", nr_maps);
3988         while (nr_maps) {
3989                 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
3990                 epoch = ceph_decode_32(&p);
3991                 maplen = ceph_decode_32(&p);
3992                 ceph_decode_need(&p, end, maplen, bad);
3993                 if (nr_maps > 1) {
3994                         dout("skipping non-latest full map %u len %d\n",
3995                              epoch, maplen);
3996                 } else if (osdc->osdmap->epoch >= epoch) {
3997                         dout("skipping full map %u len %d, "
3998                              "older than our %u\n", epoch, maplen,
3999                              osdc->osdmap->epoch);
4000                 } else {
4001                         dout("taking full map %u len %d\n", epoch, maplen);
4002                         err = handle_one_map(osdc, p, p + maplen, false,
4003                                              &need_resend, &need_resend_linger);
4004                         if (err)
4005                                 goto bad;
4006                 }
4007                 p += maplen;
4008                 nr_maps--;
4009         }
4010
4011 done:
4012         /*
4013          * subscribe to subsequent osdmap updates if full to ensure
4014          * we find out when we are no longer full and stop returning
4015          * ENOSPC.
4016          */
4017         pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
4018         pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
4019                   ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
4020                   have_pool_full(osdc);
4021         if (was_pauserd || was_pausewr || pauserd || pausewr ||
4022             osdc->osdmap->epoch < osdc->epoch_barrier)
4023                 maybe_request_map(osdc);
4024
4025         kick_requests(osdc, &need_resend, &need_resend_linger);
4026
4027         ceph_osdc_abort_on_full(osdc);
4028         ceph_monc_got_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
4029                           osdc->osdmap->epoch);
4030         up_write(&osdc->lock);
4031         wake_up_all(&osdc->client->auth_wq);
4032         return;
4033
4034 bad:
4035         pr_err("osdc handle_map corrupt msg\n");
4036         ceph_msg_dump(msg);
4037         up_write(&osdc->lock);
4038 }
4039
4040 /*
4041  * Resubmit requests pending on the given osd.
4042  */
4043 static void kick_osd_requests(struct ceph_osd *osd)
4044 {
4045         struct rb_node *n;
4046
4047         clear_backoffs(osd);
4048
4049         for (n = rb_first(&osd->o_requests); n; ) {
4050                 struct ceph_osd_request *req =
4051                     rb_entry(n, struct ceph_osd_request, r_node);
4052
4053                 n = rb_next(n); /* cancel_linger_request() */
4054
4055                 if (!req->r_linger) {
4056                         if (!req->r_t.paused)
4057                                 send_request(req);
4058                 } else {
4059                         cancel_linger_request(req);
4060                 }
4061         }
4062         for (n = rb_first(&osd->o_linger_requests); n; n = rb_next(n)) {
4063                 struct ceph_osd_linger_request *lreq =
4064                     rb_entry(n, struct ceph_osd_linger_request, node);
4065
4066                 send_linger(lreq);
4067         }
4068 }
4069
4070 /*
4071  * If the osd connection drops, we need to resubmit all requests.
4072  */
4073 static void osd_fault(struct ceph_connection *con)
4074 {
4075         struct ceph_osd *osd = con->private;
4076         struct ceph_osd_client *osdc = osd->o_osdc;
4077
4078         dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
4079
4080         down_write(&osdc->lock);
4081         if (!osd_registered(osd)) {
4082                 dout("%s osd%d unknown\n", __func__, osd->o_osd);
4083                 goto out_unlock;
4084         }
4085
4086         if (!reopen_osd(osd))
4087                 kick_osd_requests(osd);
4088         maybe_request_map(osdc);
4089
4090 out_unlock:
4091         up_write(&osdc->lock);
4092 }
4093
4094 struct MOSDBackoff {
4095         struct ceph_spg spgid;
4096         u32 map_epoch;
4097         u8 op;
4098         u64 id;
4099         struct ceph_hobject_id *begin;
4100         struct ceph_hobject_id *end;
4101 };
4102
4103 static int decode_MOSDBackoff(const struct ceph_msg *msg, struct MOSDBackoff *m)
4104 {
4105         void *p = msg->front.iov_base;
4106         void *const end = p + msg->front.iov_len;
4107         u8 struct_v;
4108         u32 struct_len;
4109         int ret;
4110
4111         ret = ceph_start_decoding(&p, end, 1, "spg_t", &struct_v, &struct_len);
4112         if (ret)
4113                 return ret;
4114
4115         ret = ceph_decode_pgid(&p, end, &m->spgid.pgid);
4116         if (ret)
4117                 return ret;
4118
4119         ceph_decode_8_safe(&p, end, m->spgid.shard, e_inval);
4120         ceph_decode_32_safe(&p, end, m->map_epoch, e_inval);
4121         ceph_decode_8_safe(&p, end, m->op, e_inval);
4122         ceph_decode_64_safe(&p, end, m->id, e_inval);
4123
4124         m->begin = kzalloc(sizeof(*m->begin), GFP_NOIO);
4125         if (!m->begin)
4126                 return -ENOMEM;
4127
4128         ret = decode_hoid(&p, end, m->begin);
4129         if (ret) {
4130                 free_hoid(m->begin);
4131                 return ret;
4132         }
4133
4134         m->end = kzalloc(sizeof(*m->end), GFP_NOIO);
4135         if (!m->end) {
4136                 free_hoid(m->begin);
4137                 return -ENOMEM;
4138         }
4139
4140         ret = decode_hoid(&p, end, m->end);
4141         if (ret) {
4142                 free_hoid(m->begin);
4143                 free_hoid(m->end);
4144                 return ret;
4145         }
4146
4147         return 0;
4148
4149 e_inval:
4150         return -EINVAL;
4151 }
4152
4153 static struct ceph_msg *create_backoff_message(
4154                                 const struct ceph_osd_backoff *backoff,
4155                                 u32 map_epoch)
4156 {
4157         struct ceph_msg *msg;
4158         void *p, *end;
4159         int msg_size;
4160
4161         msg_size = CEPH_ENCODING_START_BLK_LEN +
4162                         CEPH_PGID_ENCODING_LEN + 1; /* spgid */
4163         msg_size += 4 + 1 + 8; /* map_epoch, op, id */
4164         msg_size += CEPH_ENCODING_START_BLK_LEN +
4165                         hoid_encoding_size(backoff->begin);
4166         msg_size += CEPH_ENCODING_START_BLK_LEN +
4167                         hoid_encoding_size(backoff->end);
4168
4169         msg = ceph_msg_new(CEPH_MSG_OSD_BACKOFF, msg_size, GFP_NOIO, true);
4170         if (!msg)
4171                 return NULL;
4172
4173         p = msg->front.iov_base;
4174         end = p + msg->front_alloc_len;
4175
4176         encode_spgid(&p, &backoff->spgid);
4177         ceph_encode_32(&p, map_epoch);
4178         ceph_encode_8(&p, CEPH_OSD_BACKOFF_OP_ACK_BLOCK);
4179         ceph_encode_64(&p, backoff->id);
4180         encode_hoid(&p, end, backoff->begin);
4181         encode_hoid(&p, end, backoff->end);
4182         BUG_ON(p != end);
4183
4184         msg->front.iov_len = p - msg->front.iov_base;
4185         msg->hdr.version = cpu_to_le16(1); /* MOSDBackoff v1 */
4186         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
4187
4188         return msg;
4189 }
4190
4191 static void handle_backoff_block(struct ceph_osd *osd, struct MOSDBackoff *m)
4192 {
4193         struct ceph_spg_mapping *spg;
4194         struct ceph_osd_backoff *backoff;
4195         struct ceph_msg *msg;
4196
4197         dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4198              m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4199
4200         spg = lookup_spg_mapping(&osd->o_backoff_mappings, &m->spgid);
4201         if (!spg) {
4202                 spg = alloc_spg_mapping();
4203                 if (!spg) {
4204                         pr_err("%s failed to allocate spg\n", __func__);
4205                         return;
4206                 }
4207                 spg->spgid = m->spgid; /* struct */
4208                 insert_spg_mapping(&osd->o_backoff_mappings, spg);
4209         }
4210
4211         backoff = alloc_backoff();
4212         if (!backoff) {
4213                 pr_err("%s failed to allocate backoff\n", __func__);
4214                 return;
4215         }
4216         backoff->spgid = m->spgid; /* struct */
4217         backoff->id = m->id;
4218         backoff->begin = m->begin;
4219         m->begin = NULL; /* backoff now owns this */
4220         backoff->end = m->end;
4221         m->end = NULL;   /* ditto */
4222
4223         insert_backoff(&spg->backoffs, backoff);
4224         insert_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4225
4226         /*
4227          * Ack with original backoff's epoch so that the OSD can
4228          * discard this if there was a PG split.
4229          */
4230         msg = create_backoff_message(backoff, m->map_epoch);
4231         if (!msg) {
4232                 pr_err("%s failed to allocate msg\n", __func__);
4233                 return;
4234         }
4235         ceph_con_send(&osd->o_con, msg);
4236 }
4237
4238 static bool target_contained_by(const struct ceph_osd_request_target *t,
4239                                 const struct ceph_hobject_id *begin,
4240                                 const struct ceph_hobject_id *end)
4241 {
4242         struct ceph_hobject_id hoid;
4243         int cmp;
4244
4245         hoid_fill_from_target(&hoid, t);
4246         cmp = hoid_compare(&hoid, begin);
4247         return !cmp || (cmp > 0 && hoid_compare(&hoid, end) < 0);
4248 }
4249
4250 static void handle_backoff_unblock(struct ceph_osd *osd,
4251                                    const struct MOSDBackoff *m)
4252 {
4253         struct ceph_spg_mapping *spg;
4254         struct ceph_osd_backoff *backoff;
4255         struct rb_node *n;
4256
4257         dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4258              m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4259
4260         backoff = lookup_backoff_by_id(&osd->o_backoffs_by_id, m->id);
4261         if (!backoff) {
4262                 pr_err("%s osd%d spgid %llu.%xs%d id %llu backoff dne\n",
4263                        __func__, osd->o_osd, m->spgid.pgid.pool,
4264                        m->spgid.pgid.seed, m->spgid.shard, m->id);
4265                 return;
4266         }
4267
4268         if (hoid_compare(backoff->begin, m->begin) &&
4269             hoid_compare(backoff->end, m->end)) {
4270                 pr_err("%s osd%d spgid %llu.%xs%d id %llu bad range?\n",
4271                        __func__, osd->o_osd, m->spgid.pgid.pool,
4272                        m->spgid.pgid.seed, m->spgid.shard, m->id);
4273                 /* unblock it anyway... */
4274         }
4275
4276         spg = lookup_spg_mapping(&osd->o_backoff_mappings, &backoff->spgid);
4277         BUG_ON(!spg);
4278
4279         erase_backoff(&spg->backoffs, backoff);
4280         erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4281         free_backoff(backoff);
4282
4283         if (RB_EMPTY_ROOT(&spg->backoffs)) {
4284                 erase_spg_mapping(&osd->o_backoff_mappings, spg);
4285                 free_spg_mapping(spg);
4286         }
4287
4288         for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
4289                 struct ceph_osd_request *req =
4290                     rb_entry(n, struct ceph_osd_request, r_node);
4291
4292                 if (!ceph_spg_compare(&req->r_t.spgid, &m->spgid)) {
4293                         /*
4294                          * Match against @m, not @backoff -- the PG may
4295                          * have split on the OSD.
4296                          */
4297                         if (target_contained_by(&req->r_t, m->begin, m->end)) {
4298                                 /*
4299                                  * If no other installed backoff applies,
4300                                  * resend.
4301                                  */
4302                                 send_request(req);
4303                         }
4304                 }
4305         }
4306 }
4307
4308 static void handle_backoff(struct ceph_osd *osd, struct ceph_msg *msg)
4309 {
4310         struct ceph_osd_client *osdc = osd->o_osdc;
4311         struct MOSDBackoff m;
4312         int ret;
4313
4314         down_read(&osdc->lock);
4315         if (!osd_registered(osd)) {
4316                 dout("%s osd%d unknown\n", __func__, osd->o_osd);
4317                 up_read(&osdc->lock);
4318                 return;
4319         }
4320         WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
4321
4322         mutex_lock(&osd->lock);
4323         ret = decode_MOSDBackoff(msg, &m);
4324         if (ret) {
4325                 pr_err("failed to decode MOSDBackoff: %d\n", ret);
4326                 ceph_msg_dump(msg);
4327                 goto out_unlock;
4328         }
4329
4330         switch (m.op) {
4331         case CEPH_OSD_BACKOFF_OP_BLOCK:
4332                 handle_backoff_block(osd, &m);
4333                 break;
4334         case CEPH_OSD_BACKOFF_OP_UNBLOCK:
4335                 handle_backoff_unblock(osd, &m);
4336                 break;
4337         default:
4338                 pr_err("%s osd%d unknown op %d\n", __func__, osd->o_osd, m.op);
4339         }
4340
4341         free_hoid(m.begin);
4342         free_hoid(m.end);
4343
4344 out_unlock:
4345         mutex_unlock(&osd->lock);
4346         up_read(&osdc->lock);
4347 }
4348
4349 /*
4350  * Process osd watch notifications
4351  */
4352 static void handle_watch_notify(struct ceph_osd_client *osdc,
4353                                 struct ceph_msg *msg)
4354 {
4355         void *p = msg->front.iov_base;
4356         void *const end = p + msg->front.iov_len;
4357         struct ceph_osd_linger_request *lreq;
4358         struct linger_work *lwork;
4359         u8 proto_ver, opcode;
4360         u64 cookie, notify_id;
4361         u64 notifier_id = 0;
4362         s32 return_code = 0;
4363         void *payload = NULL;
4364         u32 payload_len = 0;
4365
4366         ceph_decode_8_safe(&p, end, proto_ver, bad);
4367         ceph_decode_8_safe(&p, end, opcode, bad);
4368         ceph_decode_64_safe(&p, end, cookie, bad);
4369         p += 8; /* skip ver */
4370         ceph_decode_64_safe(&p, end, notify_id, bad);
4371
4372         if (proto_ver >= 1) {
4373                 ceph_decode_32_safe(&p, end, payload_len, bad);
4374                 ceph_decode_need(&p, end, payload_len, bad);
4375                 payload = p;
4376                 p += payload_len;
4377         }
4378
4379         if (le16_to_cpu(msg->hdr.version) >= 2)
4380                 ceph_decode_32_safe(&p, end, return_code, bad);
4381
4382         if (le16_to_cpu(msg->hdr.version) >= 3)
4383                 ceph_decode_64_safe(&p, end, notifier_id, bad);
4384
4385         down_read(&osdc->lock);
4386         lreq = lookup_linger_osdc(&osdc->linger_requests, cookie);
4387         if (!lreq) {
4388                 dout("%s opcode %d cookie %llu dne\n", __func__, opcode,
4389                      cookie);
4390                 goto out_unlock_osdc;
4391         }
4392
4393         mutex_lock(&lreq->lock);
4394         dout("%s opcode %d cookie %llu lreq %p is_watch %d\n", __func__,
4395              opcode, cookie, lreq, lreq->is_watch);
4396         if (opcode == CEPH_WATCH_EVENT_DISCONNECT) {
4397                 if (!lreq->last_error) {
4398                         lreq->last_error = -ENOTCONN;
4399                         queue_watch_error(lreq);
4400                 }
4401         } else if (!lreq->is_watch) {
4402                 /* CEPH_WATCH_EVENT_NOTIFY_COMPLETE */
4403                 if (lreq->notify_id && lreq->notify_id != notify_id) {
4404                         dout("lreq %p notify_id %llu != %llu, ignoring\n", lreq,
4405                              lreq->notify_id, notify_id);
4406                 } else if (!completion_done(&lreq->notify_finish_wait)) {
4407                         struct ceph_msg_data *data =
4408                             msg->num_data_items ? &msg->data[0] : NULL;
4409
4410                         if (data) {
4411                                 if (lreq->preply_pages) {
4412                                         WARN_ON(data->type !=
4413                                                         CEPH_MSG_DATA_PAGES);
4414                                         *lreq->preply_pages = data->pages;
4415                                         *lreq->preply_len = data->length;
4416                                 } else {
4417                                         ceph_release_page_vector(data->pages,
4418                                                calc_pages_for(0, data->length));
4419                                 }
4420                         }
4421                         lreq->notify_finish_error = return_code;
4422                         complete_all(&lreq->notify_finish_wait);
4423                 }
4424         } else {
4425                 /* CEPH_WATCH_EVENT_NOTIFY */
4426                 lwork = lwork_alloc(lreq, do_watch_notify);
4427                 if (!lwork) {
4428                         pr_err("failed to allocate notify-lwork\n");
4429                         goto out_unlock_lreq;
4430                 }
4431
4432                 lwork->notify.notify_id = notify_id;
4433                 lwork->notify.notifier_id = notifier_id;
4434                 lwork->notify.payload = payload;
4435                 lwork->notify.payload_len = payload_len;
4436                 lwork->notify.msg = ceph_msg_get(msg);
4437                 lwork_queue(lwork);
4438         }
4439
4440 out_unlock_lreq:
4441         mutex_unlock(&lreq->lock);
4442 out_unlock_osdc:
4443         up_read(&osdc->lock);
4444         return;
4445
4446 bad:
4447         pr_err("osdc handle_watch_notify corrupt msg\n");
4448 }
4449
4450 /*
4451  * Register request, send initial attempt.
4452  */
4453 int ceph_osdc_start_request(struct ceph_osd_client *osdc,
4454                             struct ceph_osd_request *req,
4455                             bool nofail)
4456 {
4457         down_read(&osdc->lock);
4458         submit_request(req, false);
4459         up_read(&osdc->lock);
4460
4461         return 0;
4462 }
4463 EXPORT_SYMBOL(ceph_osdc_start_request);
4464
4465 /*
4466  * Unregister a registered request.  The request is not completed:
4467  * ->r_result isn't set and __complete_request() isn't called.
4468  */
4469 void ceph_osdc_cancel_request(struct ceph_osd_request *req)
4470 {
4471         struct ceph_osd_client *osdc = req->r_osdc;
4472
4473         down_write(&osdc->lock);
4474         if (req->r_osd)
4475                 cancel_request(req);
4476         up_write(&osdc->lock);
4477 }
4478 EXPORT_SYMBOL(ceph_osdc_cancel_request);
4479
4480 /*
4481  * @timeout: in jiffies, 0 means "wait forever"
4482  */
4483 static int wait_request_timeout(struct ceph_osd_request *req,
4484                                 unsigned long timeout)
4485 {
4486         long left;
4487
4488         dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
4489         left = wait_for_completion_killable_timeout(&req->r_completion,
4490                                                 ceph_timeout_jiffies(timeout));
4491         if (left <= 0) {
4492                 left = left ?: -ETIMEDOUT;
4493                 ceph_osdc_cancel_request(req);
4494         } else {
4495                 left = req->r_result; /* completed */
4496         }
4497
4498         return left;
4499 }
4500
4501 /*
4502  * wait for a request to complete
4503  */
4504 int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
4505                            struct ceph_osd_request *req)
4506 {
4507         return wait_request_timeout(req, 0);
4508 }
4509 EXPORT_SYMBOL(ceph_osdc_wait_request);
4510
4511 /*
4512  * sync - wait for all in-flight requests to flush.  avoid starvation.
4513  */
4514 void ceph_osdc_sync(struct ceph_osd_client *osdc)
4515 {
4516         struct rb_node *n, *p;
4517         u64 last_tid = atomic64_read(&osdc->last_tid);
4518
4519 again:
4520         down_read(&osdc->lock);
4521         for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
4522                 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
4523
4524                 mutex_lock(&osd->lock);
4525                 for (p = rb_first(&osd->o_requests); p; p = rb_next(p)) {
4526                         struct ceph_osd_request *req =
4527                             rb_entry(p, struct ceph_osd_request, r_node);
4528
4529                         if (req->r_tid > last_tid)
4530                                 break;
4531
4532                         if (!(req->r_flags & CEPH_OSD_FLAG_WRITE))
4533                                 continue;
4534
4535                         ceph_osdc_get_request(req);
4536                         mutex_unlock(&osd->lock);
4537                         up_read(&osdc->lock);
4538                         dout("%s waiting on req %p tid %llu last_tid %llu\n",
4539                              __func__, req, req->r_tid, last_tid);
4540                         wait_for_completion(&req->r_completion);
4541                         ceph_osdc_put_request(req);
4542                         goto again;
4543                 }
4544
4545                 mutex_unlock(&osd->lock);
4546         }
4547
4548         up_read(&osdc->lock);
4549         dout("%s done last_tid %llu\n", __func__, last_tid);
4550 }
4551 EXPORT_SYMBOL(ceph_osdc_sync);
4552
4553 static struct ceph_osd_request *
4554 alloc_linger_request(struct ceph_osd_linger_request *lreq)
4555 {
4556         struct ceph_osd_request *req;
4557
4558         req = ceph_osdc_alloc_request(lreq->osdc, NULL, 1, false, GFP_NOIO);
4559         if (!req)
4560                 return NULL;
4561
4562         ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
4563         ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
4564         return req;
4565 }
4566
4567 static struct ceph_osd_request *
4568 alloc_watch_request(struct ceph_osd_linger_request *lreq, u8 watch_opcode)
4569 {
4570         struct ceph_osd_request *req;
4571
4572         req = alloc_linger_request(lreq);
4573         if (!req)
4574                 return NULL;
4575
4576         /*
4577          * Pass 0 for cookie because we don't know it yet, it will be
4578          * filled in by linger_submit().
4579          */
4580         osd_req_op_watch_init(req, 0, 0, watch_opcode);
4581
4582         if (ceph_osdc_alloc_messages(req, GFP_NOIO)) {
4583                 ceph_osdc_put_request(req);
4584                 return NULL;
4585         }
4586
4587         return req;
4588 }
4589
4590 /*
4591  * Returns a handle, caller owns a ref.
4592  */
4593 struct ceph_osd_linger_request *
4594 ceph_osdc_watch(struct ceph_osd_client *osdc,
4595                 struct ceph_object_id *oid,
4596                 struct ceph_object_locator *oloc,
4597                 rados_watchcb2_t wcb,
4598                 rados_watcherrcb_t errcb,
4599                 void *data)
4600 {
4601         struct ceph_osd_linger_request *lreq;
4602         int ret;
4603
4604         lreq = linger_alloc(osdc);
4605         if (!lreq)
4606                 return ERR_PTR(-ENOMEM);
4607
4608         lreq->is_watch = true;
4609         lreq->wcb = wcb;
4610         lreq->errcb = errcb;
4611         lreq->data = data;
4612         lreq->watch_valid_thru = jiffies;
4613
4614         ceph_oid_copy(&lreq->t.base_oid, oid);
4615         ceph_oloc_copy(&lreq->t.base_oloc, oloc);
4616         lreq->t.flags = CEPH_OSD_FLAG_WRITE;
4617         ktime_get_real_ts64(&lreq->mtime);
4618
4619         lreq->reg_req = alloc_watch_request(lreq, CEPH_OSD_WATCH_OP_WATCH);
4620         if (!lreq->reg_req) {
4621                 ret = -ENOMEM;
4622                 goto err_put_lreq;
4623         }
4624
4625         lreq->ping_req = alloc_watch_request(lreq, CEPH_OSD_WATCH_OP_PING);
4626         if (!lreq->ping_req) {
4627                 ret = -ENOMEM;
4628                 goto err_put_lreq;
4629         }
4630
4631         linger_submit(lreq);
4632         ret = linger_reg_commit_wait(lreq);
4633         if (ret) {
4634                 linger_cancel(lreq);
4635                 goto err_put_lreq;
4636         }
4637
4638         return lreq;
4639
4640 err_put_lreq:
4641         linger_put(lreq);
4642         return ERR_PTR(ret);
4643 }
4644 EXPORT_SYMBOL(ceph_osdc_watch);
4645
4646 /*
4647  * Releases a ref.
4648  *
4649  * Times out after mount_timeout to preserve rbd unmap behaviour
4650  * introduced in 2894e1d76974 ("rbd: timeout watch teardown on unmap
4651  * with mount_timeout").
4652  */
4653 int ceph_osdc_unwatch(struct ceph_osd_client *osdc,
4654                       struct ceph_osd_linger_request *lreq)
4655 {
4656         struct ceph_options *opts = osdc->client->options;
4657         struct ceph_osd_request *req;
4658         int ret;
4659
4660         req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4661         if (!req)
4662                 return -ENOMEM;
4663
4664         ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
4665         ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
4666         req->r_flags = CEPH_OSD_FLAG_WRITE;
4667         ktime_get_real_ts64(&req->r_mtime);
4668         osd_req_op_watch_init(req, 0, lreq->linger_id,
4669                               CEPH_OSD_WATCH_OP_UNWATCH);
4670
4671         ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4672         if (ret)
4673                 goto out_put_req;
4674
4675         ceph_osdc_start_request(osdc, req, false);
4676         linger_cancel(lreq);
4677         linger_put(lreq);
4678         ret = wait_request_timeout(req, opts->mount_timeout);
4679
4680 out_put_req:
4681         ceph_osdc_put_request(req);
4682         return ret;
4683 }
4684 EXPORT_SYMBOL(ceph_osdc_unwatch);
4685
4686 static int osd_req_op_notify_ack_init(struct ceph_osd_request *req, int which,
4687                                       u64 notify_id, u64 cookie, void *payload,
4688                                       u32 payload_len)
4689 {
4690         struct ceph_osd_req_op *op;
4691         struct ceph_pagelist *pl;
4692         int ret;
4693
4694         op = _osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY_ACK, 0);
4695
4696         pl = ceph_pagelist_alloc(GFP_NOIO);
4697         if (!pl)
4698                 return -ENOMEM;
4699
4700         ret = ceph_pagelist_encode_64(pl, notify_id);
4701         ret |= ceph_pagelist_encode_64(pl, cookie);
4702         if (payload) {
4703                 ret |= ceph_pagelist_encode_32(pl, payload_len);
4704                 ret |= ceph_pagelist_append(pl, payload, payload_len);
4705         } else {
4706                 ret |= ceph_pagelist_encode_32(pl, 0);
4707         }
4708         if (ret) {
4709                 ceph_pagelist_release(pl);
4710                 return -ENOMEM;
4711         }
4712
4713         ceph_osd_data_pagelist_init(&op->notify_ack.request_data, pl);
4714         op->indata_len = pl->length;
4715         return 0;
4716 }
4717
4718 int ceph_osdc_notify_ack(struct ceph_osd_client *osdc,
4719                          struct ceph_object_id *oid,
4720                          struct ceph_object_locator *oloc,
4721                          u64 notify_id,
4722                          u64 cookie,
4723                          void *payload,
4724                          u32 payload_len)
4725 {
4726         struct ceph_osd_request *req;
4727         int ret;
4728
4729         req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4730         if (!req)
4731                 return -ENOMEM;
4732
4733         ceph_oid_copy(&req->r_base_oid, oid);
4734         ceph_oloc_copy(&req->r_base_oloc, oloc);
4735         req->r_flags = CEPH_OSD_FLAG_READ;
4736
4737         ret = osd_req_op_notify_ack_init(req, 0, notify_id, cookie, payload,
4738                                          payload_len);
4739         if (ret)
4740                 goto out_put_req;
4741
4742         ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4743         if (ret)
4744                 goto out_put_req;
4745
4746         ceph_osdc_start_request(osdc, req, false);
4747         ret = ceph_osdc_wait_request(osdc, req);
4748
4749 out_put_req:
4750         ceph_osdc_put_request(req);
4751         return ret;
4752 }
4753 EXPORT_SYMBOL(ceph_osdc_notify_ack);
4754
4755 static int osd_req_op_notify_init(struct ceph_osd_request *req, int which,
4756                                   u64 cookie, u32 prot_ver, u32 timeout,
4757                                   void *payload, u32 payload_len)
4758 {
4759         struct ceph_osd_req_op *op;
4760         struct ceph_pagelist *pl;
4761         int ret;
4762
4763         op = _osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY, 0);
4764         op->notify.cookie = cookie;
4765
4766         pl = ceph_pagelist_alloc(GFP_NOIO);
4767         if (!pl)
4768                 return -ENOMEM;
4769
4770         ret = ceph_pagelist_encode_32(pl, 1); /* prot_ver */
4771         ret |= ceph_pagelist_encode_32(pl, timeout);
4772         ret |= ceph_pagelist_encode_32(pl, payload_len);
4773         ret |= ceph_pagelist_append(pl, payload, payload_len);
4774         if (ret) {
4775                 ceph_pagelist_release(pl);
4776                 return -ENOMEM;
4777         }
4778
4779         ceph_osd_data_pagelist_init(&op->notify.request_data, pl);
4780         op->indata_len = pl->length;
4781         return 0;
4782 }
4783
4784 /*
4785  * @timeout: in seconds
4786  *
4787  * @preply_{pages,len} are initialized both on success and error.
4788  * The caller is responsible for:
4789  *
4790  *     ceph_release_page_vector(reply_pages, calc_pages_for(0, reply_len))
4791  */
4792 int ceph_osdc_notify(struct ceph_osd_client *osdc,
4793                      struct ceph_object_id *oid,
4794                      struct ceph_object_locator *oloc,
4795                      void *payload,
4796                      u32 payload_len,
4797                      u32 timeout,
4798                      struct page ***preply_pages,
4799                      size_t *preply_len)
4800 {
4801         struct ceph_osd_linger_request *lreq;
4802         struct page **pages;
4803         int ret;
4804
4805         WARN_ON(!timeout);
4806         if (preply_pages) {
4807                 *preply_pages = NULL;
4808                 *preply_len = 0;
4809         }
4810
4811         lreq = linger_alloc(osdc);
4812         if (!lreq)
4813                 return -ENOMEM;
4814
4815         lreq->preply_pages = preply_pages;
4816         lreq->preply_len = preply_len;
4817
4818         ceph_oid_copy(&lreq->t.base_oid, oid);
4819         ceph_oloc_copy(&lreq->t.base_oloc, oloc);
4820         lreq->t.flags = CEPH_OSD_FLAG_READ;
4821
4822         lreq->reg_req = alloc_linger_request(lreq);
4823         if (!lreq->reg_req) {
4824                 ret = -ENOMEM;
4825                 goto out_put_lreq;
4826         }
4827
4828         /*
4829          * Pass 0 for cookie because we don't know it yet, it will be
4830          * filled in by linger_submit().
4831          */
4832         ret = osd_req_op_notify_init(lreq->reg_req, 0, 0, 1, timeout,
4833                                      payload, payload_len);
4834         if (ret)
4835                 goto out_put_lreq;
4836
4837         /* for notify_id */
4838         pages = ceph_alloc_page_vector(1, GFP_NOIO);
4839         if (IS_ERR(pages)) {
4840                 ret = PTR_ERR(pages);
4841                 goto out_put_lreq;
4842         }
4843         ceph_osd_data_pages_init(osd_req_op_data(lreq->reg_req, 0, notify,
4844                                                  response_data),
4845                                  pages, PAGE_SIZE, 0, false, true);
4846
4847         ret = ceph_osdc_alloc_messages(lreq->reg_req, GFP_NOIO);
4848         if (ret)
4849                 goto out_put_lreq;
4850
4851         linger_submit(lreq);
4852         ret = linger_reg_commit_wait(lreq);
4853         if (!ret)
4854                 ret = linger_notify_finish_wait(lreq);
4855         else
4856                 dout("lreq %p failed to initiate notify %d\n", lreq, ret);
4857
4858         linger_cancel(lreq);
4859 out_put_lreq:
4860         linger_put(lreq);
4861         return ret;
4862 }
4863 EXPORT_SYMBOL(ceph_osdc_notify);
4864
4865 /*
4866  * Return the number of milliseconds since the watch was last
4867  * confirmed, or an error.  If there is an error, the watch is no
4868  * longer valid, and should be destroyed with ceph_osdc_unwatch().
4869  */
4870 int ceph_osdc_watch_check(struct ceph_osd_client *osdc,
4871                           struct ceph_osd_linger_request *lreq)
4872 {
4873         unsigned long stamp, age;
4874         int ret;
4875
4876         down_read(&osdc->lock);
4877         mutex_lock(&lreq->lock);
4878         stamp = lreq->watch_valid_thru;
4879         if (!list_empty(&lreq->pending_lworks)) {
4880                 struct linger_work *lwork =
4881                     list_first_entry(&lreq->pending_lworks,
4882                                      struct linger_work,
4883                                      pending_item);
4884
4885                 if (time_before(lwork->queued_stamp, stamp))
4886                         stamp = lwork->queued_stamp;
4887         }
4888         age = jiffies - stamp;
4889         dout("%s lreq %p linger_id %llu age %lu last_error %d\n", __func__,
4890              lreq, lreq->linger_id, age, lreq->last_error);
4891         /* we are truncating to msecs, so return a safe upper bound */
4892         ret = lreq->last_error ?: 1 + jiffies_to_msecs(age);
4893
4894         mutex_unlock(&lreq->lock);
4895         up_read(&osdc->lock);
4896         return ret;
4897 }
4898
4899 static int decode_watcher(void **p, void *end, struct ceph_watch_item *item)
4900 {
4901         u8 struct_v;
4902         u32 struct_len;
4903         int ret;
4904
4905         ret = ceph_start_decoding(p, end, 2, "watch_item_t",
4906                                   &struct_v, &struct_len);
4907         if (ret)
4908                 goto bad;
4909
4910         ret = -EINVAL;
4911         ceph_decode_copy_safe(p, end, &item->name, sizeof(item->name), bad);
4912         ceph_decode_64_safe(p, end, item->cookie, bad);
4913         ceph_decode_skip_32(p, end, bad); /* skip timeout seconds */
4914
4915         if (struct_v >= 2) {
4916                 ret = ceph_decode_entity_addr(p, end, &item->addr);
4917                 if (ret)
4918                         goto bad;
4919         } else {
4920                 ret = 0;
4921         }
4922
4923         dout("%s %s%llu cookie %llu addr %s\n", __func__,
4924              ENTITY_NAME(item->name), item->cookie,
4925              ceph_pr_addr(&item->addr));
4926 bad:
4927         return ret;
4928 }
4929
4930 static int decode_watchers(void **p, void *end,
4931                            struct ceph_watch_item **watchers,
4932                            u32 *num_watchers)
4933 {
4934         u8 struct_v;
4935         u32 struct_len;
4936         int i;
4937         int ret;
4938
4939         ret = ceph_start_decoding(p, end, 1, "obj_list_watch_response_t",
4940                                   &struct_v, &struct_len);
4941         if (ret)
4942                 return ret;
4943
4944         *num_watchers = ceph_decode_32(p);
4945         *watchers = kcalloc(*num_watchers, sizeof(**watchers), GFP_NOIO);
4946         if (!*watchers)
4947                 return -ENOMEM;
4948
4949         for (i = 0; i < *num_watchers; i++) {
4950                 ret = decode_watcher(p, end, *watchers + i);
4951                 if (ret) {
4952                         kfree(*watchers);
4953                         return ret;
4954                 }
4955         }
4956
4957         return 0;
4958 }
4959
4960 /*
4961  * On success, the caller is responsible for:
4962  *
4963  *     kfree(watchers);
4964  */
4965 int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
4966                             struct ceph_object_id *oid,
4967                             struct ceph_object_locator *oloc,
4968                             struct ceph_watch_item **watchers,
4969                             u32 *num_watchers)
4970 {
4971         struct ceph_osd_request *req;
4972         struct page **pages;
4973         int ret;
4974
4975         req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4976         if (!req)
4977                 return -ENOMEM;
4978
4979         ceph_oid_copy(&req->r_base_oid, oid);
4980         ceph_oloc_copy(&req->r_base_oloc, oloc);
4981         req->r_flags = CEPH_OSD_FLAG_READ;
4982
4983         pages = ceph_alloc_page_vector(1, GFP_NOIO);
4984         if (IS_ERR(pages)) {
4985                 ret = PTR_ERR(pages);
4986                 goto out_put_req;
4987         }
4988
4989         osd_req_op_init(req, 0, CEPH_OSD_OP_LIST_WATCHERS, 0);
4990         ceph_osd_data_pages_init(osd_req_op_data(req, 0, list_watchers,
4991                                                  response_data),
4992                                  pages, PAGE_SIZE, 0, false, true);
4993
4994         ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4995         if (ret)
4996                 goto out_put_req;
4997
4998         ceph_osdc_start_request(osdc, req, false);
4999         ret = ceph_osdc_wait_request(osdc, req);
5000         if (ret >= 0) {
5001                 void *p = page_address(pages[0]);
5002                 void *const end = p + req->r_ops[0].outdata_len;
5003
5004                 ret = decode_watchers(&p, end, watchers, num_watchers);
5005         }
5006
5007 out_put_req:
5008         ceph_osdc_put_request(req);
5009         return ret;
5010 }
5011 EXPORT_SYMBOL(ceph_osdc_list_watchers);
5012
5013 /*
5014  * Call all pending notify callbacks - for use after a watch is
5015  * unregistered, to make sure no more callbacks for it will be invoked
5016  */
5017 void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc)
5018 {
5019         dout("%s osdc %p\n", __func__, osdc);
5020         flush_workqueue(osdc->notify_wq);
5021 }
5022 EXPORT_SYMBOL(ceph_osdc_flush_notifies);
5023
5024 void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc)
5025 {
5026         down_read(&osdc->lock);
5027         maybe_request_map(osdc);
5028         up_read(&osdc->lock);
5029 }
5030 EXPORT_SYMBOL(ceph_osdc_maybe_request_map);
5031
5032 /*
5033  * Execute an OSD class method on an object.
5034  *
5035  * @flags: CEPH_OSD_FLAG_*
5036  * @resp_len: in/out param for reply length
5037  */
5038 int ceph_osdc_call(struct ceph_osd_client *osdc,
5039                    struct ceph_object_id *oid,
5040                    struct ceph_object_locator *oloc,
5041                    const char *class, const char *method,
5042                    unsigned int flags,
5043                    struct page *req_page, size_t req_len,
5044                    struct page **resp_pages, size_t *resp_len)
5045 {
5046         struct ceph_osd_request *req;
5047         int ret;
5048
5049         if (req_len > PAGE_SIZE)
5050                 return -E2BIG;
5051
5052         req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
5053         if (!req)
5054                 return -ENOMEM;
5055
5056         ceph_oid_copy(&req->r_base_oid, oid);
5057         ceph_oloc_copy(&req->r_base_oloc, oloc);
5058         req->r_flags = flags;
5059
5060         ret = osd_req_op_cls_init(req, 0, class, method);
5061         if (ret)
5062                 goto out_put_req;
5063
5064         if (req_page)
5065                 osd_req_op_cls_request_data_pages(req, 0, &req_page, req_len,
5066                                                   0, false, false);
5067         if (resp_pages)
5068                 osd_req_op_cls_response_data_pages(req, 0, resp_pages,
5069                                                    *resp_len, 0, false, false);
5070
5071         ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
5072         if (ret)
5073                 goto out_put_req;
5074
5075         ceph_osdc_start_request(osdc, req, false);
5076         ret = ceph_osdc_wait_request(osdc, req);
5077         if (ret >= 0) {
5078                 ret = req->r_ops[0].rval;
5079                 if (resp_pages)
5080                         *resp_len = req->r_ops[0].outdata_len;
5081         }
5082
5083 out_put_req:
5084         ceph_osdc_put_request(req);
5085         return ret;
5086 }
5087 EXPORT_SYMBOL(ceph_osdc_call);
5088
5089 /*
5090  * init, shutdown
5091  */
5092 int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client)
5093 {
5094         int err;
5095
5096         dout("init\n");
5097         osdc->client = client;
5098         init_rwsem(&osdc->lock);
5099         osdc->osds = RB_ROOT;
5100         INIT_LIST_HEAD(&osdc->osd_lru);
5101         spin_lock_init(&osdc->osd_lru_lock);
5102         osd_init(&osdc->homeless_osd);
5103         osdc->homeless_osd.o_osdc = osdc;
5104         osdc->homeless_osd.o_osd = CEPH_HOMELESS_OSD;
5105         osdc->last_linger_id = CEPH_LINGER_ID_START;
5106         osdc->linger_requests = RB_ROOT;
5107         osdc->map_checks = RB_ROOT;
5108         osdc->linger_map_checks = RB_ROOT;
5109         INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout);
5110         INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout);
5111
5112         err = -ENOMEM;
5113         osdc->osdmap = ceph_osdmap_alloc();
5114         if (!osdc->osdmap)
5115                 goto out;
5116
5117         osdc->req_mempool = mempool_create_slab_pool(10,
5118                                                      ceph_osd_request_cache);
5119         if (!osdc->req_mempool)
5120                 goto out_map;
5121
5122         err = ceph_msgpool_init(&osdc->msgpool_op, CEPH_MSG_OSD_OP,
5123                                 PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10, "osd_op");
5124         if (err < 0)
5125                 goto out_mempool;
5126         err = ceph_msgpool_init(&osdc->msgpool_op_reply, CEPH_MSG_OSD_OPREPLY,
5127                                 PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10,
5128                                 "osd_op_reply");
5129         if (err < 0)
5130                 goto out_msgpool;
5131
5132         err = -ENOMEM;
5133         osdc->notify_wq = create_singlethread_workqueue("ceph-watch-notify");
5134         if (!osdc->notify_wq)
5135                 goto out_msgpool_reply;
5136
5137         osdc->completion_wq = create_singlethread_workqueue("ceph-completion");
5138         if (!osdc->completion_wq)
5139                 goto out_notify_wq;
5140
5141         schedule_delayed_work(&osdc->timeout_work,
5142                               osdc->client->options->osd_keepalive_timeout);
5143         schedule_delayed_work(&osdc->osds_timeout_work,
5144             round_jiffies_relative(osdc->client->options->osd_idle_ttl));
5145
5146         return 0;
5147
5148 out_notify_wq:
5149         destroy_workqueue(osdc->notify_wq);
5150 out_msgpool_reply:
5151         ceph_msgpool_destroy(&osdc->msgpool_op_reply);
5152 out_msgpool:
5153         ceph_msgpool_destroy(&osdc->msgpool_op);
5154 out_mempool:
5155         mempool_destroy(osdc->req_mempool);
5156 out_map:
5157         ceph_osdmap_destroy(osdc->osdmap);
5158 out:
5159         return err;
5160 }
5161
5162 void ceph_osdc_stop(struct ceph_osd_client *osdc)
5163 {
5164         destroy_workqueue(osdc->completion_wq);
5165         destroy_workqueue(osdc->notify_wq);
5166         cancel_delayed_work_sync(&osdc->timeout_work);
5167         cancel_delayed_work_sync(&osdc->osds_timeout_work);
5168
5169         down_write(&osdc->lock);
5170         while (!RB_EMPTY_ROOT(&osdc->osds)) {
5171                 struct ceph_osd *osd = rb_entry(rb_first(&osdc->osds),
5172                                                 struct ceph_osd, o_node);
5173                 close_osd(osd);
5174         }
5175         up_write(&osdc->lock);
5176         WARN_ON(refcount_read(&osdc->homeless_osd.o_ref) != 1);
5177         osd_cleanup(&osdc->homeless_osd);
5178
5179         WARN_ON(!list_empty(&osdc->osd_lru));
5180         WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_requests));
5181         WARN_ON(!RB_EMPTY_ROOT(&osdc->map_checks));
5182         WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_map_checks));
5183         WARN_ON(atomic_read(&osdc->num_requests));
5184         WARN_ON(atomic_read(&osdc->num_homeless));
5185
5186         ceph_osdmap_destroy(osdc->osdmap);
5187         mempool_destroy(osdc->req_mempool);
5188         ceph_msgpool_destroy(&osdc->msgpool_op);
5189         ceph_msgpool_destroy(&osdc->msgpool_op_reply);
5190 }
5191
5192 /*
5193  * Read some contiguous pages.  If we cross a stripe boundary, shorten
5194  * *plen.  Return number of bytes read, or error.
5195  */
5196 int ceph_osdc_readpages(struct ceph_osd_client *osdc,
5197                         struct ceph_vino vino, struct ceph_file_layout *layout,
5198                         u64 off, u64 *plen,
5199                         u32 truncate_seq, u64 truncate_size,
5200                         struct page **pages, int num_pages, int page_align)
5201 {
5202         struct ceph_osd_request *req;
5203         int rc = 0;
5204
5205         dout("readpages on ino %llx.%llx on %llu~%llu\n", vino.ino,
5206              vino.snap, off, *plen);
5207         req = ceph_osdc_new_request(osdc, layout, vino, off, plen, 0, 1,
5208                                     CEPH_OSD_OP_READ, CEPH_OSD_FLAG_READ,
5209                                     NULL, truncate_seq, truncate_size,
5210                                     false);
5211         if (IS_ERR(req))
5212                 return PTR_ERR(req);
5213
5214         /* it may be a short read due to an object boundary */
5215         osd_req_op_extent_osd_data_pages(req, 0,
5216                                 pages, *plen, page_align, false, false);
5217
5218         dout("readpages  final extent is %llu~%llu (%llu bytes align %d)\n",
5219              off, *plen, *plen, page_align);
5220
5221         rc = ceph_osdc_start_request(osdc, req, false);
5222         if (!rc)
5223                 rc = ceph_osdc_wait_request(osdc, req);
5224
5225         ceph_osdc_put_request(req);
5226         dout("readpages result %d\n", rc);
5227         return rc;
5228 }
5229 EXPORT_SYMBOL(ceph_osdc_readpages);
5230
5231 /*
5232  * do a synchronous write on N pages
5233  */
5234 int ceph_osdc_writepages(struct ceph_osd_client *osdc, struct ceph_vino vino,
5235                          struct ceph_file_layout *layout,
5236                          struct ceph_snap_context *snapc,
5237                          u64 off, u64 len,
5238                          u32 truncate_seq, u64 truncate_size,
5239                          struct timespec64 *mtime,
5240                          struct page **pages, int num_pages)
5241 {
5242         struct ceph_osd_request *req;
5243         int rc = 0;
5244         int page_align = off & ~PAGE_MASK;
5245
5246         req = ceph_osdc_new_request(osdc, layout, vino, off, &len, 0, 1,
5247                                     CEPH_OSD_OP_WRITE, CEPH_OSD_FLAG_WRITE,
5248                                     snapc, truncate_seq, truncate_size,
5249                                     true);
5250         if (IS_ERR(req))
5251                 return PTR_ERR(req);
5252
5253         /* it may be a short write due to an object boundary */
5254         osd_req_op_extent_osd_data_pages(req, 0, pages, len, page_align,
5255                                 false, false);
5256         dout("writepages %llu~%llu (%llu bytes)\n", off, len, len);
5257
5258         req->r_mtime = *mtime;
5259         rc = ceph_osdc_start_request(osdc, req, true);
5260         if (!rc)
5261                 rc = ceph_osdc_wait_request(osdc, req);
5262
5263         ceph_osdc_put_request(req);
5264         if (rc == 0)
5265                 rc = len;
5266         dout("writepages result %d\n", rc);
5267         return rc;
5268 }
5269 EXPORT_SYMBOL(ceph_osdc_writepages);
5270
5271 static int osd_req_op_copy_from_init(struct ceph_osd_request *req,
5272                                      u64 src_snapid, u64 src_version,
5273                                      struct ceph_object_id *src_oid,
5274                                      struct ceph_object_locator *src_oloc,
5275                                      u32 src_fadvise_flags,
5276                                      u32 dst_fadvise_flags,
5277                                      u8 copy_from_flags)
5278 {
5279         struct ceph_osd_req_op *op;
5280         struct page **pages;
5281         void *p, *end;
5282
5283         pages = ceph_alloc_page_vector(1, GFP_KERNEL);
5284         if (IS_ERR(pages))
5285                 return PTR_ERR(pages);
5286
5287         op = _osd_req_op_init(req, 0, CEPH_OSD_OP_COPY_FROM, dst_fadvise_flags);
5288         op->copy_from.snapid = src_snapid;
5289         op->copy_from.src_version = src_version;
5290         op->copy_from.flags = copy_from_flags;
5291         op->copy_from.src_fadvise_flags = src_fadvise_flags;
5292
5293         p = page_address(pages[0]);
5294         end = p + PAGE_SIZE;
5295         ceph_encode_string(&p, end, src_oid->name, src_oid->name_len);
5296         encode_oloc(&p, end, src_oloc);
5297         op->indata_len = PAGE_SIZE - (end - p);
5298
5299         ceph_osd_data_pages_init(&op->copy_from.osd_data, pages,
5300                                  op->indata_len, 0, false, true);
5301         return 0;
5302 }
5303
5304 int ceph_osdc_copy_from(struct ceph_osd_client *osdc,
5305                         u64 src_snapid, u64 src_version,
5306                         struct ceph_object_id *src_oid,
5307                         struct ceph_object_locator *src_oloc,
5308                         u32 src_fadvise_flags,
5309                         struct ceph_object_id *dst_oid,
5310                         struct ceph_object_locator *dst_oloc,
5311                         u32 dst_fadvise_flags,
5312                         u8 copy_from_flags)
5313 {
5314         struct ceph_osd_request *req;
5315         int ret;
5316
5317         req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_KERNEL);
5318         if (!req)
5319                 return -ENOMEM;
5320
5321         req->r_flags = CEPH_OSD_FLAG_WRITE;
5322
5323         ceph_oloc_copy(&req->r_t.base_oloc, dst_oloc);
5324         ceph_oid_copy(&req->r_t.base_oid, dst_oid);
5325
5326         ret = osd_req_op_copy_from_init(req, src_snapid, src_version, src_oid,
5327                                         src_oloc, src_fadvise_flags,
5328                                         dst_fadvise_flags, copy_from_flags);
5329         if (ret)
5330                 goto out;
5331
5332         ret = ceph_osdc_alloc_messages(req, GFP_KERNEL);
5333         if (ret)
5334                 goto out;
5335
5336         ceph_osdc_start_request(osdc, req, false);
5337         ret = ceph_osdc_wait_request(osdc, req);
5338
5339 out:
5340         ceph_osdc_put_request(req);
5341         return ret;
5342 }
5343 EXPORT_SYMBOL(ceph_osdc_copy_from);
5344
5345 int __init ceph_osdc_setup(void)
5346 {
5347         size_t size = sizeof(struct ceph_osd_request) +
5348             CEPH_OSD_SLAB_OPS * sizeof(struct ceph_osd_req_op);
5349
5350         BUG_ON(ceph_osd_request_cache);
5351         ceph_osd_request_cache = kmem_cache_create("ceph_osd_request", size,
5352                                                    0, 0, NULL);
5353
5354         return ceph_osd_request_cache ? 0 : -ENOMEM;
5355 }
5356
5357 void ceph_osdc_cleanup(void)
5358 {
5359         BUG_ON(!ceph_osd_request_cache);
5360         kmem_cache_destroy(ceph_osd_request_cache);
5361         ceph_osd_request_cache = NULL;
5362 }
5363
5364 /*
5365  * handle incoming message
5366  */
5367 static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
5368 {
5369         struct ceph_osd *osd = con->private;
5370         struct ceph_osd_client *osdc = osd->o_osdc;
5371         int type = le16_to_cpu(msg->hdr.type);
5372
5373         switch (type) {
5374         case CEPH_MSG_OSD_MAP:
5375                 ceph_osdc_handle_map(osdc, msg);
5376                 break;
5377         case CEPH_MSG_OSD_OPREPLY:
5378                 handle_reply(osd, msg);
5379                 break;
5380         case CEPH_MSG_OSD_BACKOFF:
5381                 handle_backoff(osd, msg);
5382                 break;
5383         case CEPH_MSG_WATCH_NOTIFY:
5384                 handle_watch_notify(osdc, msg);
5385                 break;
5386
5387         default:
5388                 pr_err("received unknown message type %d %s\n", type,
5389                        ceph_msg_type_name(type));
5390         }
5391
5392         ceph_msg_put(msg);
5393 }
5394
5395 /*
5396  * Lookup and return message for incoming reply.  Don't try to do
5397  * anything about a larger than preallocated data portion of the
5398  * message at the moment - for now, just skip the message.
5399  */
5400 static struct ceph_msg *get_reply(struct ceph_connection *con,
5401                                   struct ceph_msg_header *hdr,
5402                                   int *skip)
5403 {
5404         struct ceph_osd *osd = con->private;
5405         struct ceph_osd_client *osdc = osd->o_osdc;
5406         struct ceph_msg *m = NULL;
5407         struct ceph_osd_request *req;
5408         int front_len = le32_to_cpu(hdr->front_len);
5409         int data_len = le32_to_cpu(hdr->data_len);
5410         u64 tid = le64_to_cpu(hdr->tid);
5411
5412         down_read(&osdc->lock);
5413         if (!osd_registered(osd)) {
5414                 dout("%s osd%d unknown, skipping\n", __func__, osd->o_osd);
5415                 *skip = 1;
5416                 goto out_unlock_osdc;
5417         }
5418         WARN_ON(osd->o_osd != le64_to_cpu(hdr->src.num));
5419
5420         mutex_lock(&osd->lock);
5421         req = lookup_request(&osd->o_requests, tid);
5422         if (!req) {
5423                 dout("%s osd%d tid %llu unknown, skipping\n", __func__,
5424                      osd->o_osd, tid);
5425                 *skip = 1;
5426                 goto out_unlock_session;
5427         }
5428
5429         ceph_msg_revoke_incoming(req->r_reply);
5430
5431         if (front_len > req->r_reply->front_alloc_len) {
5432                 pr_warn("%s osd%d tid %llu front %d > preallocated %d\n",
5433                         __func__, osd->o_osd, req->r_tid, front_len,
5434                         req->r_reply->front_alloc_len);
5435                 m = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, front_len, GFP_NOFS,
5436                                  false);
5437                 if (!m)
5438                         goto out_unlock_session;
5439                 ceph_msg_put(req->r_reply);
5440                 req->r_reply = m;
5441         }
5442
5443         if (data_len > req->r_reply->data_length) {
5444                 pr_warn("%s osd%d tid %llu data %d > preallocated %zu, skipping\n",
5445                         __func__, osd->o_osd, req->r_tid, data_len,
5446                         req->r_reply->data_length);
5447                 m = NULL;
5448                 *skip = 1;
5449                 goto out_unlock_session;
5450         }
5451
5452         m = ceph_msg_get(req->r_reply);
5453         dout("get_reply tid %lld %p\n", tid, m);
5454
5455 out_unlock_session:
5456         mutex_unlock(&osd->lock);
5457 out_unlock_osdc:
5458         up_read(&osdc->lock);
5459         return m;
5460 }
5461
5462 /*
5463  * TODO: switch to a msg-owned pagelist
5464  */
5465 static struct ceph_msg *alloc_msg_with_page_vector(struct ceph_msg_header *hdr)
5466 {
5467         struct ceph_msg *m;
5468         int type = le16_to_cpu(hdr->type);
5469         u32 front_len = le32_to_cpu(hdr->front_len);
5470         u32 data_len = le32_to_cpu(hdr->data_len);
5471
5472         m = ceph_msg_new2(type, front_len, 1, GFP_NOIO, false);
5473         if (!m)
5474                 return NULL;
5475
5476         if (data_len) {
5477                 struct page **pages;
5478                 struct ceph_osd_data osd_data;
5479
5480                 pages = ceph_alloc_page_vector(calc_pages_for(0, data_len),
5481                                                GFP_NOIO);
5482                 if (IS_ERR(pages)) {
5483                         ceph_msg_put(m);
5484                         return NULL;
5485                 }
5486
5487                 ceph_osd_data_pages_init(&osd_data, pages, data_len, 0, false,
5488                                          false);
5489                 ceph_osdc_msg_data_add(m, &osd_data);
5490         }
5491
5492         return m;
5493 }
5494
5495 static struct ceph_msg *alloc_msg(struct ceph_connection *con,
5496                                   struct ceph_msg_header *hdr,
5497                                   int *skip)
5498 {
5499         struct ceph_osd *osd = con->private;
5500         int type = le16_to_cpu(hdr->type);
5501
5502         *skip = 0;
5503         switch (type) {
5504         case CEPH_MSG_OSD_MAP:
5505         case CEPH_MSG_OSD_BACKOFF:
5506         case CEPH_MSG_WATCH_NOTIFY:
5507                 return alloc_msg_with_page_vector(hdr);
5508         case CEPH_MSG_OSD_OPREPLY:
5509                 return get_reply(con, hdr, skip);
5510         default:
5511                 pr_warn("%s osd%d unknown msg type %d, skipping\n", __func__,
5512                         osd->o_osd, type);
5513                 *skip = 1;
5514                 return NULL;
5515         }
5516 }
5517
5518 /*
5519  * Wrappers to refcount containing ceph_osd struct
5520  */
5521 static struct ceph_connection *get_osd_con(struct ceph_connection *con)
5522 {
5523         struct ceph_osd *osd = con->private;
5524         if (get_osd(osd))
5525                 return con;
5526         return NULL;
5527 }
5528
5529 static void put_osd_con(struct ceph_connection *con)
5530 {
5531         struct ceph_osd *osd = con->private;
5532         put_osd(osd);
5533 }
5534
5535 /*
5536  * authentication
5537  */
5538 /*
5539  * Note: returned pointer is the address of a structure that's
5540  * managed separately.  Caller must *not* attempt to free it.
5541  */
5542 static struct ceph_auth_handshake *get_authorizer(struct ceph_connection *con,
5543                                         int *proto, int force_new)
5544 {
5545         struct ceph_osd *o = con->private;
5546         struct ceph_osd_client *osdc = o->o_osdc;
5547         struct ceph_auth_client *ac = osdc->client->monc.auth;
5548         struct ceph_auth_handshake *auth = &o->o_auth;
5549
5550         if (force_new && auth->authorizer) {
5551                 ceph_auth_destroy_authorizer(auth->authorizer);
5552                 auth->authorizer = NULL;
5553         }
5554         if (!auth->authorizer) {
5555                 int ret = ceph_auth_create_authorizer(ac, CEPH_ENTITY_TYPE_OSD,
5556                                                       auth);
5557                 if (ret)
5558                         return ERR_PTR(ret);
5559         } else {
5560                 int ret = ceph_auth_update_authorizer(ac, CEPH_ENTITY_TYPE_OSD,
5561                                                      auth);
5562                 if (ret)
5563                         return ERR_PTR(ret);
5564         }
5565         *proto = ac->protocol;
5566
5567         return auth;
5568 }
5569
5570 static int add_authorizer_challenge(struct ceph_connection *con,
5571                                     void *challenge_buf, int challenge_buf_len)
5572 {
5573         struct ceph_osd *o = con->private;
5574         struct ceph_osd_client *osdc = o->o_osdc;
5575         struct ceph_auth_client *ac = osdc->client->monc.auth;
5576
5577         return ceph_auth_add_authorizer_challenge(ac, o->o_auth.authorizer,
5578                                             challenge_buf, challenge_buf_len);
5579 }
5580
5581 static int verify_authorizer_reply(struct ceph_connection *con)
5582 {
5583         struct ceph_osd *o = con->private;
5584         struct ceph_osd_client *osdc = o->o_osdc;
5585         struct ceph_auth_client *ac = osdc->client->monc.auth;
5586
5587         return ceph_auth_verify_authorizer_reply(ac, o->o_auth.authorizer);
5588 }
5589
5590 static int invalidate_authorizer(struct ceph_connection *con)
5591 {
5592         struct ceph_osd *o = con->private;
5593         struct ceph_osd_client *osdc = o->o_osdc;
5594         struct ceph_auth_client *ac = osdc->client->monc.auth;
5595
5596         ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_OSD);
5597         return ceph_monc_validate_auth(&osdc->client->monc);
5598 }
5599
5600 static void osd_reencode_message(struct ceph_msg *msg)
5601 {
5602         int type = le16_to_cpu(msg->hdr.type);
5603
5604         if (type == CEPH_MSG_OSD_OP)
5605                 encode_request_finish(msg);
5606 }
5607
5608 static int osd_sign_message(struct ceph_msg *msg)
5609 {
5610         struct ceph_osd *o = msg->con->private;
5611         struct ceph_auth_handshake *auth = &o->o_auth;
5612
5613         return ceph_auth_sign_message(auth, msg);
5614 }
5615
5616 static int osd_check_message_signature(struct ceph_msg *msg)
5617 {
5618         struct ceph_osd *o = msg->con->private;
5619         struct ceph_auth_handshake *auth = &o->o_auth;
5620
5621         return ceph_auth_check_message_signature(auth, msg);
5622 }
5623
5624 static const struct ceph_connection_operations osd_con_ops = {
5625         .get = get_osd_con,
5626         .put = put_osd_con,
5627         .dispatch = dispatch,
5628         .get_authorizer = get_authorizer,
5629         .add_authorizer_challenge = add_authorizer_challenge,
5630         .verify_authorizer_reply = verify_authorizer_reply,
5631         .invalidate_authorizer = invalidate_authorizer,
5632         .alloc_msg = alloc_msg,
5633         .reencode_message = osd_reencode_message,
5634         .sign_message = osd_sign_message,
5635         .check_message_signature = osd_check_message_signature,
5636         .fault = osd_fault,
5637 };