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 void ceph_osd_iter_init(struct ceph_osd_data *osd_data,
 175                                struct iov_iter *iter)
 176 {
 177         osd_data->type = CEPH_OSD_DATA_TYPE_ITER;
 178         osd_data->iter = *iter;
 179 }
 180
 181 static struct ceph_osd_data *
 182 osd_req_op_raw_data_in(struct ceph_osd_request *osd_req, unsigned int which)
 183 {
 184         BUG_ON(which >= osd_req->r_num_ops);
 185
 186         return &osd_req->r_ops[which].raw_data_in;
 187 }
 188
 189 struct ceph_osd_data *
 190 osd_req_op_extent_osd_data(struct ceph_osd_request *osd_req,
 191                         unsigned int which)
 192 {
 193         return osd_req_op_data(osd_req, which, extent, osd_data);
 194 }
 195 EXPORT_SYMBOL(osd_req_op_extent_osd_data);
 196
 197 void osd_req_op_raw_data_in_pages(struct ceph_osd_request *osd_req,
 198                         unsigned int which, struct page **pages,
 199                         u64 length, u32 alignment,
 200                         bool pages_from_pool, bool own_pages)
 201 {
 202         struct ceph_osd_data *osd_data;
 203
 204         osd_data = osd_req_op_raw_data_in(osd_req, which);
 205         ceph_osd_data_pages_init(osd_data, pages, length, alignment,
 206                                 pages_from_pool, own_pages);
 207 }
 208 EXPORT_SYMBOL(osd_req_op_raw_data_in_pages);
 209
 210 void osd_req_op_extent_osd_data_pages(struct ceph_osd_request *osd_req,
 211                         unsigned int which, struct page **pages,
 212                         u64 length, u32 alignment,
 213                         bool pages_from_pool, bool own_pages)
 214 {
 215         struct ceph_osd_data *osd_data;
 216
 217         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 218         ceph_osd_data_pages_init(osd_data, pages, length, alignment,
 219                                 pages_from_pool, own_pages);
 220 }
 221 EXPORT_SYMBOL(osd_req_op_extent_osd_data_pages);
 222
 223 void osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request *osd_req,
 224                         unsigned int which, struct ceph_pagelist *pagelist)
 225 {
 226         struct ceph_osd_data *osd_data;
 227
 228         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 229         ceph_osd_data_pagelist_init(osd_data, pagelist);
 230 }
 231 EXPORT_SYMBOL(osd_req_op_extent_osd_data_pagelist);
 232
 233 #ifdef CONFIG_BLOCK
 234 void osd_req_op_extent_osd_data_bio(struct ceph_osd_request *osd_req,
 235                                     unsigned int which,
 236                                     struct ceph_bio_iter *bio_pos,
 237                                     u32 bio_length)
 238 {
 239         struct ceph_osd_data *osd_data;
 240
 241         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 242         ceph_osd_data_bio_init(osd_data, bio_pos, bio_length);
 243 }
 244 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bio);
 245 #endif /* CONFIG_BLOCK */
 246
 247 void osd_req_op_extent_osd_data_bvecs(struct ceph_osd_request *osd_req,
 248                                       unsigned int which,
 249                                       struct bio_vec *bvecs, u32 num_bvecs,
 250                                       u32 bytes)
 251 {
 252         struct ceph_osd_data *osd_data;
 253         struct ceph_bvec_iter it = {
 254                 .bvecs = bvecs,
 255                 .iter = { .bi_size = bytes },
 256         };
 257
 258         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 259         ceph_osd_data_bvecs_init(osd_data, &it, num_bvecs);
 260 }
 261 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvecs);
 262
 263 void osd_req_op_extent_osd_data_bvec_pos(struct ceph_osd_request *osd_req,
 264                                          unsigned int which,
 265                                          struct ceph_bvec_iter *bvec_pos)
 266 {
 267         struct ceph_osd_data *osd_data;
 268
 269         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 270         ceph_osd_data_bvecs_init(osd_data, bvec_pos, 0);
 271 }
 272 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvec_pos);
 273
 274 /**
 275  * osd_req_op_extent_osd_iter - Set up an operation with an iterator buffer
 276  * @osd_req: The request to set up
 277  * @which: Index of the operation in which to set the iter
 278  * @iter: The buffer iterator
 279  */
 280 void osd_req_op_extent_osd_iter(struct ceph_osd_request *osd_req,
 281                                 unsigned int which, struct iov_iter *iter)
 282 {
 283         struct ceph_osd_data *osd_data;
 284
 285         osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
 286         ceph_osd_iter_init(osd_data, iter);
 287 }
 288 EXPORT_SYMBOL(osd_req_op_extent_osd_iter);
 289
 290 static void osd_req_op_cls_request_info_pagelist(
 291                         struct ceph_osd_request *osd_req,
 292                         unsigned int which, struct ceph_pagelist *pagelist)
 293 {
 294         struct ceph_osd_data *osd_data;
 295
 296         osd_data = osd_req_op_data(osd_req, which, cls, request_info);
 297         ceph_osd_data_pagelist_init(osd_data, pagelist);
 298 }
 299
 300 void osd_req_op_cls_request_data_pagelist(
 301                         struct ceph_osd_request *osd_req,
 302                         unsigned int which, struct ceph_pagelist *pagelist)
 303 {
 304         struct ceph_osd_data *osd_data;
 305
 306         osd_data = osd_req_op_data(osd_req, which, cls, request_data);
 307         ceph_osd_data_pagelist_init(osd_data, pagelist);
 308         osd_req->r_ops[which].cls.indata_len += pagelist->length;
 309         osd_req->r_ops[which].indata_len += pagelist->length;
 310 }
 311 EXPORT_SYMBOL(osd_req_op_cls_request_data_pagelist);
 312
 313 void osd_req_op_cls_request_data_pages(struct ceph_osd_request *osd_req,
 314                         unsigned int which, struct page **pages, u64 length,
 315                         u32 alignment, bool pages_from_pool, bool own_pages)
 316 {
 317         struct ceph_osd_data *osd_data;
 318
 319         osd_data = osd_req_op_data(osd_req, which, cls, request_data);
 320         ceph_osd_data_pages_init(osd_data, pages, length, alignment,
 321                                 pages_from_pool, own_pages);
 322         osd_req->r_ops[which].cls.indata_len += length;
 323         osd_req->r_ops[which].indata_len += length;
 324 }
 325 EXPORT_SYMBOL(osd_req_op_cls_request_data_pages);
 326
 327 void osd_req_op_cls_request_data_bvecs(struct ceph_osd_request *osd_req,
 328                                        unsigned int which,
 329                                        struct bio_vec *bvecs, u32 num_bvecs,
 330                                        u32 bytes)
 331 {
 332         struct ceph_osd_data *osd_data;
 333         struct ceph_bvec_iter it = {
 334                 .bvecs = bvecs,
 335                 .iter = { .bi_size = bytes },
 336         };
 337
 338         osd_data = osd_req_op_data(osd_req, which, cls, request_data);
 339         ceph_osd_data_bvecs_init(osd_data, &it, num_bvecs);
 340         osd_req->r_ops[which].cls.indata_len += bytes;
 341         osd_req->r_ops[which].indata_len += bytes;
 342 }
 343 EXPORT_SYMBOL(osd_req_op_cls_request_data_bvecs);
 344
 345 void osd_req_op_cls_response_data_pages(struct ceph_osd_request *osd_req,
 346                         unsigned int which, struct page **pages, u64 length,
 347                         u32 alignment, bool pages_from_pool, bool own_pages)
 348 {
 349         struct ceph_osd_data *osd_data;
 350
 351         osd_data = osd_req_op_data(osd_req, which, cls, response_data);
 352         ceph_osd_data_pages_init(osd_data, pages, length, alignment,
 353                                 pages_from_pool, own_pages);
 354 }
 355 EXPORT_SYMBOL(osd_req_op_cls_response_data_pages);
 356
 357 static u64 ceph_osd_data_length(struct ceph_osd_data *osd_data)
 358 {
 359         switch (osd_data->type) {
 360         case CEPH_OSD_DATA_TYPE_NONE:
 361                 return 0;
 362         case CEPH_OSD_DATA_TYPE_PAGES:
 363                 return osd_data->length;
 364         case CEPH_OSD_DATA_TYPE_PAGELIST:
 365                 return (u64)osd_data->pagelist->length;
 366 #ifdef CONFIG_BLOCK
 367         case CEPH_OSD_DATA_TYPE_BIO:
 368                 return (u64)osd_data->bio_length;
 369 #endif /* CONFIG_BLOCK */
 370         case CEPH_OSD_DATA_TYPE_BVECS:
 371                 return osd_data->bvec_pos.iter.bi_size;
 372         case CEPH_OSD_DATA_TYPE_ITER:
 373                 return iov_iter_count(&osd_data->iter);
 374         default:
 375                 WARN(true, "unrecognized data type %d\n", (int)osd_data->type);
 376                 return 0;
 377         }
 378 }
 379
 380 static void ceph_osd_data_release(struct ceph_osd_data *osd_data)
 381 {
 382         if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES && osd_data->own_pages) {
 383                 int num_pages;
 384
 385                 num_pages = calc_pages_for((u64)osd_data->alignment,
 386                                                 (u64)osd_data->length);
 387                 ceph_release_page_vector(osd_data->pages, num_pages);
 388         } else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
 389                 ceph_pagelist_release(osd_data->pagelist);
 390         }
 391         ceph_osd_data_init(osd_data);
 392 }
 393
 394 static void osd_req_op_data_release(struct ceph_osd_request *osd_req,
 395                         unsigned int which)
 396 {
 397         struct ceph_osd_req_op *op;
 398
 399         BUG_ON(which >= osd_req->r_num_ops);
 400         op = &osd_req->r_ops[which];
 401
 402         switch (op->op) {
 403         case CEPH_OSD_OP_READ:
 404         case CEPH_OSD_OP_SPARSE_READ:
 405         case CEPH_OSD_OP_WRITE:
 406         case CEPH_OSD_OP_WRITEFULL:
 407                 kfree(op->extent.sparse_ext);
 408                 ceph_osd_data_release(&op->extent.osd_data);
 409                 break;
 410         case CEPH_OSD_OP_CALL:
 411                 ceph_osd_data_release(&op->cls.request_info);
 412                 ceph_osd_data_release(&op->cls.request_data);
 413                 ceph_osd_data_release(&op->cls.response_data);
 414                 break;
 415         case CEPH_OSD_OP_SETXATTR:
 416         case CEPH_OSD_OP_CMPXATTR:
 417                 ceph_osd_data_release(&op->xattr.osd_data);
 418                 break;
 419         case CEPH_OSD_OP_STAT:
 420                 ceph_osd_data_release(&op->raw_data_in);
 421                 break;
 422         case CEPH_OSD_OP_NOTIFY_ACK:
 423                 ceph_osd_data_release(&op->notify_ack.request_data);
 424                 break;
 425         case CEPH_OSD_OP_NOTIFY:
 426                 ceph_osd_data_release(&op->notify.request_data);
 427                 ceph_osd_data_release(&op->notify.response_data);
 428                 break;
 429         case CEPH_OSD_OP_LIST_WATCHERS:
 430                 ceph_osd_data_release(&op->list_watchers.response_data);
 431                 break;
 432         case CEPH_OSD_OP_COPY_FROM2:
 433                 ceph_osd_data_release(&op->copy_from.osd_data);
 434                 break;
 435         default:
 436                 break;
 437         }
 438 }
 439
 440 /*
 441  * Assumes @t is zero-initialized.
 442  */
 443 static void target_init(struct ceph_osd_request_target *t)
 444 {
 445         ceph_oid_init(&t->base_oid);
 446         ceph_oloc_init(&t->base_oloc);
 447         ceph_oid_init(&t->target_oid);
 448         ceph_oloc_init(&t->target_oloc);
 449
 450         ceph_osds_init(&t->acting);
 451         ceph_osds_init(&t->up);
 452         t->size = -1;
 453         t->min_size = -1;
 454
 455         t->osd = CEPH_HOMELESS_OSD;
 456 }
 457
 458 static void target_copy(struct ceph_osd_request_target *dest,
 459                         const struct ceph_osd_request_target *src)
 460 {
 461         ceph_oid_copy(&dest->base_oid, &src->base_oid);
 462         ceph_oloc_copy(&dest->base_oloc, &src->base_oloc);
 463         ceph_oid_copy(&dest->target_oid, &src->target_oid);
 464         ceph_oloc_copy(&dest->target_oloc, &src->target_oloc);
 465
 466         dest->pgid = src->pgid; /* struct */
 467         dest->spgid = src->spgid; /* struct */
 468         dest->pg_num = src->pg_num;
 469         dest->pg_num_mask = src->pg_num_mask;
 470         ceph_osds_copy(&dest->acting, &src->acting);
 471         ceph_osds_copy(&dest->up, &src->up);
 472         dest->size = src->size;
 473         dest->min_size = src->min_size;
 474         dest->sort_bitwise = src->sort_bitwise;
 475         dest->recovery_deletes = src->recovery_deletes;
 476
 477         dest->flags = src->flags;
 478         dest->used_replica = src->used_replica;
 479         dest->paused = src->paused;
 480
 481         dest->epoch = src->epoch;
 482         dest->last_force_resend = src->last_force_resend;
 483
 484         dest->osd = src->osd;
 485 }
 486
 487 static void target_destroy(struct ceph_osd_request_target *t)
 488 {
 489         ceph_oid_destroy(&t->base_oid);
 490         ceph_oloc_destroy(&t->base_oloc);
 491         ceph_oid_destroy(&t->target_oid);
 492         ceph_oloc_destroy(&t->target_oloc);
 493 }
 494
 495 /*
 496  * requests
 497  */
 498 static void request_release_checks(struct ceph_osd_request *req)
 499 {
 500         WARN_ON(!RB_EMPTY_NODE(&req->r_node));
 501         WARN_ON(!RB_EMPTY_NODE(&req->r_mc_node));
 502         WARN_ON(!list_empty(&req->r_private_item));
 503         WARN_ON(req->r_osd);
 504 }
 505
 506 static void ceph_osdc_release_request(struct kref *kref)
 507 {
 508         struct ceph_osd_request *req = container_of(kref,
 509                                             struct ceph_osd_request, r_kref);
 510         unsigned int which;
 511
 512         dout("%s %p (r_request %p r_reply %p)\n", __func__, req,
 513              req->r_request, req->r_reply);
 514         request_release_checks(req);
 515
 516         if (req->r_request)
 517                 ceph_msg_put(req->r_request);
 518         if (req->r_reply)
 519                 ceph_msg_put(req->r_reply);
 520
 521         for (which = 0; which < req->r_num_ops; which++)
 522                 osd_req_op_data_release(req, which);
 523
 524         target_destroy(&req->r_t);
 525         ceph_put_snap_context(req->r_snapc);
 526
 527         if (req->r_mempool)
 528                 mempool_free(req, req->r_osdc->req_mempool);
 529         else if (req->r_num_ops <= CEPH_OSD_SLAB_OPS)
 530                 kmem_cache_free(ceph_osd_request_cache, req);
 531         else
 532                 kfree(req);
 533 }
 534
 535 void ceph_osdc_get_request(struct ceph_osd_request *req)
 536 {
 537         dout("%s %p (was %d)\n", __func__, req,
 538              kref_read(&req->r_kref));
 539         kref_get(&req->r_kref);
 540 }
 541 EXPORT_SYMBOL(ceph_osdc_get_request);
 542
 543 void ceph_osdc_put_request(struct ceph_osd_request *req)
 544 {
 545         if (req) {
 546                 dout("%s %p (was %d)\n", __func__, req,
 547                      kref_read(&req->r_kref));
 548                 kref_put(&req->r_kref, ceph_osdc_release_request);
 549         }
 550 }
 551 EXPORT_SYMBOL(ceph_osdc_put_request);
 552
 553 static void request_init(struct ceph_osd_request *req)
 554 {
 555         /* req only, each op is zeroed in osd_req_op_init() */
 556         memset(req, 0, sizeof(*req));
 557
 558         kref_init(&req->r_kref);
 559         init_completion(&req->r_completion);
 560         RB_CLEAR_NODE(&req->r_node);
 561         RB_CLEAR_NODE(&req->r_mc_node);
 562         INIT_LIST_HEAD(&req->r_private_item);
 563
 564         target_init(&req->r_t);
 565 }
 566
 567 struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
 568                                                struct ceph_snap_context *snapc,
 569                                                unsigned int num_ops,
 570                                                bool use_mempool,
 571                                                gfp_t gfp_flags)
 572 {
 573         struct ceph_osd_request *req;
 574
 575         if (use_mempool) {
 576                 BUG_ON(num_ops > CEPH_OSD_SLAB_OPS);
 577                 req = mempool_alloc(osdc->req_mempool, gfp_flags);
 578         } else if (num_ops <= CEPH_OSD_SLAB_OPS) {
 579                 req = kmem_cache_alloc(ceph_osd_request_cache, gfp_flags);
 580         } else {
 581                 BUG_ON(num_ops > CEPH_OSD_MAX_OPS);
 582                 req = kmalloc(struct_size(req, r_ops, num_ops), gfp_flags);
 583         }
 584         if (unlikely(!req))
 585                 return NULL;
 586
 587         request_init(req);
 588         req->r_osdc = osdc;
 589         req->r_mempool = use_mempool;
 590         req->r_num_ops = num_ops;
 591         req->r_snapid = CEPH_NOSNAP;
 592         req->r_snapc = ceph_get_snap_context(snapc);
 593
 594         dout("%s req %p\n", __func__, req);
 595         return req;
 596 }
 597 EXPORT_SYMBOL(ceph_osdc_alloc_request);
 598
 599 static int ceph_oloc_encoding_size(const struct ceph_object_locator *oloc)
 600 {
 601         return 8 + 4 + 4 + 4 + (oloc->pool_ns ? oloc->pool_ns->len : 0);
 602 }
 603
 604 static int __ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp,
 605                                       int num_request_data_items,
 606                                       int num_reply_data_items)
 607 {
 608         struct ceph_osd_client *osdc = req->r_osdc;
 609         struct ceph_msg *msg;
 610         int msg_size;
 611
 612         WARN_ON(req->r_request || req->r_reply);
 613         WARN_ON(ceph_oid_empty(&req->r_base_oid));
 614         WARN_ON(ceph_oloc_empty(&req->r_base_oloc));
 615
 616         /* create request message */
 617         msg_size = CEPH_ENCODING_START_BLK_LEN +
 618                         CEPH_PGID_ENCODING_LEN + 1; /* spgid */
 619         msg_size += 4 + 4 + 4; /* hash, osdmap_epoch, flags */
 620         msg_size += CEPH_ENCODING_START_BLK_LEN +
 621                         sizeof(struct ceph_osd_reqid); /* reqid */
 622         msg_size += sizeof(struct ceph_blkin_trace_info); /* trace */
 623         msg_size += 4 + sizeof(struct ceph_timespec); /* client_inc, mtime */
 624         msg_size += CEPH_ENCODING_START_BLK_LEN +
 625                         ceph_oloc_encoding_size(&req->r_base_oloc); /* oloc */
 626         msg_size += 4 + req->r_base_oid.name_len; /* oid */
 627         msg_size += 2 + req->r_num_ops * sizeof(struct ceph_osd_op);
 628         msg_size += 8; /* snapid */
 629         msg_size += 8; /* snap_seq */
 630         msg_size += 4 + 8 * (req->r_snapc ? req->r_snapc->num_snaps : 0);
 631         msg_size += 4 + 8; /* retry_attempt, features */
 632
 633         if (req->r_mempool)
 634                 msg = ceph_msgpool_get(&osdc->msgpool_op, msg_size,
 635                                        num_request_data_items);
 636         else
 637                 msg = ceph_msg_new2(CEPH_MSG_OSD_OP, msg_size,
 638                                     num_request_data_items, gfp, true);
 639         if (!msg)
 640                 return -ENOMEM;
 641
 642         memset(msg->front.iov_base, 0, msg->front.iov_len);
 643         req->r_request = msg;
 644
 645         /* create reply message */
 646         msg_size = OSD_OPREPLY_FRONT_LEN;
 647         msg_size += req->r_base_oid.name_len;
 648         msg_size += req->r_num_ops * sizeof(struct ceph_osd_op);
 649
 650         if (req->r_mempool)
 651                 msg = ceph_msgpool_get(&osdc->msgpool_op_reply, msg_size,
 652                                        num_reply_data_items);
 653         else
 654                 msg = ceph_msg_new2(CEPH_MSG_OSD_OPREPLY, msg_size,
 655                                     num_reply_data_items, gfp, true);
 656         if (!msg)
 657                 return -ENOMEM;
 658
 659         req->r_reply = msg;
 660
 661         return 0;
 662 }
 663
 664 static bool osd_req_opcode_valid(u16 opcode)
 665 {
 666         switch (opcode) {
 667 #define GENERATE_CASE(op, opcode, str)  case CEPH_OSD_OP_##op: return true;
 668 __CEPH_FORALL_OSD_OPS(GENERATE_CASE)
 669 #undef GENERATE_CASE
 670         default:
 671                 return false;
 672         }
 673 }
 674
 675 static void get_num_data_items(struct ceph_osd_request *req,
 676                                int *num_request_data_items,
 677                                int *num_reply_data_items)
 678 {
 679         struct ceph_osd_req_op *op;
 680
 681         *num_request_data_items = 0;
 682         *num_reply_data_items = 0;
 683
 684         for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) {
 685                 switch (op->op) {
 686                 /* request */
 687                 case CEPH_OSD_OP_WRITE:
 688                 case CEPH_OSD_OP_WRITEFULL:
 689                 case CEPH_OSD_OP_SETXATTR:
 690                 case CEPH_OSD_OP_CMPXATTR:
 691                 case CEPH_OSD_OP_NOTIFY_ACK:
 692                 case CEPH_OSD_OP_COPY_FROM2:
 693                         *num_request_data_items += 1;
 694                         break;
 695
 696                 /* reply */
 697                 case CEPH_OSD_OP_STAT:
 698                 case CEPH_OSD_OP_READ:
 699                 case CEPH_OSD_OP_SPARSE_READ:
 700                 case CEPH_OSD_OP_LIST_WATCHERS:
 701                         *num_reply_data_items += 1;
 702                         break;
 703
 704                 /* both */
 705                 case CEPH_OSD_OP_NOTIFY:
 706                         *num_request_data_items += 1;
 707                         *num_reply_data_items += 1;
 708                         break;
 709                 case CEPH_OSD_OP_CALL:
 710                         *num_request_data_items += 2;
 711                         *num_reply_data_items += 1;
 712                         break;
 713
 714                 default:
 715                         WARN_ON(!osd_req_opcode_valid(op->op));
 716                         break;
 717                 }
 718         }
 719 }
 720
 721 /*
 722  * oid, oloc and OSD op opcode(s) must be filled in before this function
 723  * is called.
 724  */
 725 int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp)
 726 {
 727         int num_request_data_items, num_reply_data_items;
 728
 729         get_num_data_items(req, &num_request_data_items, &num_reply_data_items);
 730         return __ceph_osdc_alloc_messages(req, gfp, num_request_data_items,
 731                                           num_reply_data_items);
 732 }
 733 EXPORT_SYMBOL(ceph_osdc_alloc_messages);
 734
 735 /*
 736  * This is an osd op init function for opcodes that have no data or
 737  * other information associated with them.  It also serves as a
 738  * common init routine for all the other init functions, below.
 739  */
 740 struct ceph_osd_req_op *
 741 osd_req_op_init(struct ceph_osd_request *osd_req, unsigned int which,
 742                  u16 opcode, u32 flags)
 743 {
 744         struct ceph_osd_req_op *op;
 745
 746         BUG_ON(which >= osd_req->r_num_ops);
 747         BUG_ON(!osd_req_opcode_valid(opcode));
 748
 749         op = &osd_req->r_ops[which];
 750         memset(op, 0, sizeof (*op));
 751         op->op = opcode;
 752         op->flags = flags;
 753
 754         return op;
 755 }
 756 EXPORT_SYMBOL(osd_req_op_init);
 757
 758 void osd_req_op_extent_init(struct ceph_osd_request *osd_req,
 759                                 unsigned int which, u16 opcode,
 760                                 u64 offset, u64 length,
 761                                 u64 truncate_size, u32 truncate_seq)
 762 {
 763         struct ceph_osd_req_op *op = osd_req_op_init(osd_req, which,
 764                                                      opcode, 0);
 765         size_t payload_len = 0;
 766
 767         BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
 768                opcode != CEPH_OSD_OP_WRITEFULL && opcode != CEPH_OSD_OP_ZERO &&
 769                opcode != CEPH_OSD_OP_TRUNCATE && opcode != CEPH_OSD_OP_SPARSE_READ);
 770
 771         op->extent.offset = offset;
 772         op->extent.length = length;
 773         op->extent.truncate_size = truncate_size;
 774         op->extent.truncate_seq = truncate_seq;
 775         if (opcode == CEPH_OSD_OP_WRITE || opcode == CEPH_OSD_OP_WRITEFULL)
 776                 payload_len += length;
 777
 778         op->indata_len = payload_len;
 779 }
 780 EXPORT_SYMBOL(osd_req_op_extent_init);
 781
 782 void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
 783                                 unsigned int which, u64 length)
 784 {
 785         struct ceph_osd_req_op *op;
 786         u64 previous;
 787
 788         BUG_ON(which >= osd_req->r_num_ops);
 789         op = &osd_req->r_ops[which];
 790         previous = op->extent.length;
 791
 792         if (length == previous)
 793                 return;         /* Nothing to do */
 794         BUG_ON(length > previous);
 795
 796         op->extent.length = length;
 797         if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
 798                 op->indata_len -= previous - length;
 799 }
 800 EXPORT_SYMBOL(osd_req_op_extent_update);
 801
 802 void osd_req_op_extent_dup_last(struct ceph_osd_request *osd_req,
 803                                 unsigned int which, u64 offset_inc)
 804 {
 805         struct ceph_osd_req_op *op, *prev_op;
 806
 807         BUG_ON(which + 1 >= osd_req->r_num_ops);
 808
 809         prev_op = &osd_req->r_ops[which];
 810         op = osd_req_op_init(osd_req, which + 1, prev_op->op, prev_op->flags);
 811         /* dup previous one */
 812         op->indata_len = prev_op->indata_len;
 813         op->outdata_len = prev_op->outdata_len;
 814         op->extent = prev_op->extent;
 815         /* adjust offset */
 816         op->extent.offset += offset_inc;
 817         op->extent.length -= offset_inc;
 818
 819         if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
 820                 op->indata_len -= offset_inc;
 821 }
 822 EXPORT_SYMBOL(osd_req_op_extent_dup_last);
 823
 824 int osd_req_op_cls_init(struct ceph_osd_request *osd_req, unsigned int which,
 825                         const char *class, const char *method)
 826 {
 827         struct ceph_osd_req_op *op;
 828         struct ceph_pagelist *pagelist;
 829         size_t payload_len = 0;
 830         size_t size;
 831         int ret;
 832
 833         op = osd_req_op_init(osd_req, which, CEPH_OSD_OP_CALL, 0);
 834
 835         pagelist = ceph_pagelist_alloc(GFP_NOFS);
 836         if (!pagelist)
 837                 return -ENOMEM;
 838
 839         op->cls.class_name = class;
 840         size = strlen(class);
 841         BUG_ON(size > (size_t) U8_MAX);
 842         op->cls.class_len = size;
 843         ret = ceph_pagelist_append(pagelist, class, size);
 844         if (ret)
 845                 goto err_pagelist_free;
 846         payload_len += size;
 847
 848         op->cls.method_name = method;
 849         size = strlen(method);
 850         BUG_ON(size > (size_t) U8_MAX);
 851         op->cls.method_len = size;
 852         ret = ceph_pagelist_append(pagelist, method, size);
 853         if (ret)
 854                 goto err_pagelist_free;
 855         payload_len += size;
 856
 857         osd_req_op_cls_request_info_pagelist(osd_req, which, pagelist);
 858         op->indata_len = payload_len;
 859         return 0;
 860
 861 err_pagelist_free:
 862         ceph_pagelist_release(pagelist);
 863         return ret;
 864 }
 865 EXPORT_SYMBOL(osd_req_op_cls_init);
 866
 867 int osd_req_op_xattr_init(struct ceph_osd_request *osd_req, unsigned int which,
 868                           u16 opcode, const char *name, const void *value,
 869                           size_t size, u8 cmp_op, u8 cmp_mode)
 870 {
 871         struct ceph_osd_req_op *op = osd_req_op_init(osd_req, which,
 872                                                      opcode, 0);
 873         struct ceph_pagelist *pagelist;
 874         size_t payload_len;
 875         int ret;
 876
 877         BUG_ON(opcode != CEPH_OSD_OP_SETXATTR && opcode != CEPH_OSD_OP_CMPXATTR);
 878
 879         pagelist = ceph_pagelist_alloc(GFP_NOFS);
 880         if (!pagelist)
 881                 return -ENOMEM;
 882
 883         payload_len = strlen(name);
 884         op->xattr.name_len = payload_len;
 885         ret = ceph_pagelist_append(pagelist, name, payload_len);
 886         if (ret)
 887                 goto err_pagelist_free;
 888
 889         op->xattr.value_len = size;
 890         ret = ceph_pagelist_append(pagelist, value, size);
 891         if (ret)
 892                 goto err_pagelist_free;
 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 err_pagelist_free:
 903         ceph_pagelist_release(pagelist);
 904         return ret;
 905 }
 906 EXPORT_SYMBOL(osd_req_op_xattr_init);
 907
 908 /*
 909  * @watch_opcode: CEPH_OSD_WATCH_OP_*
 910  */
 911 static void osd_req_op_watch_init(struct ceph_osd_request *req, int which,
 912                                   u8 watch_opcode, u64 cookie, u32 gen)
 913 {
 914         struct ceph_osd_req_op *op;
 915
 916         op = osd_req_op_init(req, which, CEPH_OSD_OP_WATCH, 0);
 917         op->watch.cookie = cookie;
 918         op->watch.op = watch_opcode;
 919         op->watch.gen = gen;
 920 }
 921
 922 /*
 923  * prot_ver, timeout and notify payload (may be empty) should already be
 924  * encoded in @request_pl
 925  */
 926 static void osd_req_op_notify_init(struct ceph_osd_request *req, int which,
 927                                    u64 cookie, struct ceph_pagelist *request_pl)
 928 {
 929         struct ceph_osd_req_op *op;
 930
 931         op = osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY, 0);
 932         op->notify.cookie = cookie;
 933
 934         ceph_osd_data_pagelist_init(&op->notify.request_data, request_pl);
 935         op->indata_len = request_pl->length;
 936 }
 937
 938 /*
 939  * @flags: CEPH_OSD_OP_ALLOC_HINT_FLAG_*
 940  */
 941 void osd_req_op_alloc_hint_init(struct ceph_osd_request *osd_req,
 942                                 unsigned int which,
 943                                 u64 expected_object_size,
 944                                 u64 expected_write_size,
 945                                 u32 flags)
 946 {
 947         struct ceph_osd_req_op *op;
 948
 949         op = osd_req_op_init(osd_req, which, CEPH_OSD_OP_SETALLOCHINT, 0);
 950         op->alloc_hint.expected_object_size = expected_object_size;
 951         op->alloc_hint.expected_write_size = expected_write_size;
 952         op->alloc_hint.flags = flags;
 953
 954         /*
 955          * CEPH_OSD_OP_SETALLOCHINT op is advisory and therefore deemed
 956          * not worth a feature bit.  Set FAILOK per-op flag to make
 957          * sure older osds don't trip over an unsupported opcode.
 958          */
 959         op->flags |= CEPH_OSD_OP_FLAG_FAILOK;
 960 }
 961 EXPORT_SYMBOL(osd_req_op_alloc_hint_init);
 962
 963 static void ceph_osdc_msg_data_add(struct ceph_msg *msg,
 964                                 struct ceph_osd_data *osd_data)
 965 {
 966         u64 length = ceph_osd_data_length(osd_data);
 967
 968         if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES) {
 969                 BUG_ON(length > (u64) SIZE_MAX);
 970                 if (length)
 971                         ceph_msg_data_add_pages(msg, osd_data->pages,
 972                                         length, osd_data->alignment, false);
 973         } else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
 974                 BUG_ON(!length);
 975                 ceph_msg_data_add_pagelist(msg, osd_data->pagelist);
 976 #ifdef CONFIG_BLOCK
 977         } else if (osd_data->type == CEPH_OSD_DATA_TYPE_BIO) {
 978                 ceph_msg_data_add_bio(msg, &osd_data->bio_pos, length);
 979 #endif
 980         } else if (osd_data->type == CEPH_OSD_DATA_TYPE_BVECS) {
 981                 ceph_msg_data_add_bvecs(msg, &osd_data->bvec_pos);
 982         } else if (osd_data->type == CEPH_OSD_DATA_TYPE_ITER) {
 983                 ceph_msg_data_add_iter(msg, &osd_data->iter);
 984         } else {
 985                 BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_NONE);
 986         }
 987 }
 988
 989 static u32 osd_req_encode_op(struct ceph_osd_op *dst,
 990                              const struct ceph_osd_req_op *src)
 991 {
 992         switch (src->op) {
 993         case CEPH_OSD_OP_STAT:
 994                 break;
 995         case CEPH_OSD_OP_READ:
 996         case CEPH_OSD_OP_SPARSE_READ:
 997         case CEPH_OSD_OP_WRITE:
 998         case CEPH_OSD_OP_WRITEFULL:
 999         case CEPH_OSD_OP_ZERO:
1000         case CEPH_OSD_OP_TRUNCATE:
1001                 dst->extent.offset = cpu_to_le64(src->extent.offset);
1002                 dst->extent.length = cpu_to_le64(src->extent.length);
1003                 dst->extent.truncate_size =
1004                         cpu_to_le64(src->extent.truncate_size);
1005                 dst->extent.truncate_seq =
1006                         cpu_to_le32(src->extent.truncate_seq);
1007                 break;
1008         case CEPH_OSD_OP_CALL:
1009                 dst->cls.class_len = src->cls.class_len;
1010                 dst->cls.method_len = src->cls.method_len;
1011                 dst->cls.indata_len = cpu_to_le32(src->cls.indata_len);
1012                 break;
1013         case CEPH_OSD_OP_WATCH:
1014                 dst->watch.cookie = cpu_to_le64(src->watch.cookie);
1015                 dst->watch.ver = cpu_to_le64(0);
1016                 dst->watch.op = src->watch.op;
1017                 dst->watch.gen = cpu_to_le32(src->watch.gen);
1018                 break;
1019         case CEPH_OSD_OP_NOTIFY_ACK:
1020                 break;
1021         case CEPH_OSD_OP_NOTIFY:
1022                 dst->notify.cookie = cpu_to_le64(src->notify.cookie);
1023                 break;
1024         case CEPH_OSD_OP_LIST_WATCHERS:
1025                 break;
1026         case CEPH_OSD_OP_SETALLOCHINT:
1027                 dst->alloc_hint.expected_object_size =
1028                     cpu_to_le64(src->alloc_hint.expected_object_size);
1029                 dst->alloc_hint.expected_write_size =
1030                     cpu_to_le64(src->alloc_hint.expected_write_size);
1031                 dst->alloc_hint.flags = cpu_to_le32(src->alloc_hint.flags);
1032                 break;
1033         case CEPH_OSD_OP_SETXATTR:
1034         case CEPH_OSD_OP_CMPXATTR:
1035                 dst->xattr.name_len = cpu_to_le32(src->xattr.name_len);
1036                 dst->xattr.value_len = cpu_to_le32(src->xattr.value_len);
1037                 dst->xattr.cmp_op = src->xattr.cmp_op;
1038                 dst->xattr.cmp_mode = src->xattr.cmp_mode;
1039                 break;
1040         case CEPH_OSD_OP_CREATE:
1041         case CEPH_OSD_OP_DELETE:
1042                 break;
1043         case CEPH_OSD_OP_COPY_FROM2:
1044                 dst->copy_from.snapid = cpu_to_le64(src->copy_from.snapid);
1045                 dst->copy_from.src_version =
1046                         cpu_to_le64(src->copy_from.src_version);
1047                 dst->copy_from.flags = src->copy_from.flags;
1048                 dst->copy_from.src_fadvise_flags =
1049                         cpu_to_le32(src->copy_from.src_fadvise_flags);
1050                 break;
1051         case CEPH_OSD_OP_ASSERT_VER:
1052                 dst->assert_ver.unused = cpu_to_le64(0);
1053                 dst->assert_ver.ver = cpu_to_le64(src->assert_ver.ver);
1054                 break;
1055         default:
1056                 pr_err("unsupported osd opcode %s\n",
1057                         ceph_osd_op_name(src->op));
1058                 WARN_ON(1);
1059
1060                 return 0;
1061         }
1062
1063         dst->op = cpu_to_le16(src->op);
1064         dst->flags = cpu_to_le32(src->flags);
1065         dst->payload_len = cpu_to_le32(src->indata_len);
1066
1067         return src->indata_len;
1068 }
1069
1070 /*
1071  * build new request AND message, calculate layout, and adjust file
1072  * extent as needed.
1073  *
1074  * if the file was recently truncated, we include information about its
1075  * old and new size so that the object can be updated appropriately.  (we
1076  * avoid synchronously deleting truncated objects because it's slow.)
1077  */
1078 struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *osdc,
1079                                                struct ceph_file_layout *layout,
1080                                                struct ceph_vino vino,
1081                                                u64 off, u64 *plen,
1082                                                unsigned int which, int num_ops,
1083                                                int opcode, int flags,
1084                                                struct ceph_snap_context *snapc,
1085                                                u32 truncate_seq,
1086                                                u64 truncate_size,
1087                                                bool use_mempool)
1088 {
1089         struct ceph_osd_request *req;
1090         u64 objnum = 0;
1091         u64 objoff = 0;
1092         u64 objlen = 0;
1093         int r;
1094
1095         BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
1096                opcode != CEPH_OSD_OP_ZERO && opcode != CEPH_OSD_OP_TRUNCATE &&
1097                opcode != CEPH_OSD_OP_CREATE && opcode != CEPH_OSD_OP_DELETE &&
1098                opcode != CEPH_OSD_OP_SPARSE_READ);
1099
1100         req = ceph_osdc_alloc_request(osdc, snapc, num_ops, use_mempool,
1101                                         GFP_NOFS);
1102         if (!req) {
1103                 r = -ENOMEM;
1104                 goto fail;
1105         }
1106
1107         /* calculate max write size */
1108         r = calc_layout(layout, off, plen, &objnum, &objoff, &objlen);
1109         if (r)
1110                 goto fail;
1111
1112         if (opcode == CEPH_OSD_OP_CREATE || opcode == CEPH_OSD_OP_DELETE) {
1113                 osd_req_op_init(req, which, opcode, 0);
1114         } else {
1115                 u32 object_size = layout->object_size;
1116                 u32 object_base = off - objoff;
1117                 if (!(truncate_seq == 1 && truncate_size == -1ULL)) {
1118                         if (truncate_size <= object_base) {
1119                                 truncate_size = 0;
1120                         } else {
1121                                 truncate_size -= object_base;
1122                                 if (truncate_size > object_size)
1123                                         truncate_size = object_size;
1124                         }
1125                 }
1126                 osd_req_op_extent_init(req, which, opcode, objoff, objlen,
1127                                        truncate_size, truncate_seq);
1128         }
1129
1130         req->r_base_oloc.pool = layout->pool_id;
1131         req->r_base_oloc.pool_ns = ceph_try_get_string(layout->pool_ns);
1132         ceph_oid_printf(&req->r_base_oid, "%llx.%08llx", vino.ino, objnum);
1133         req->r_flags = flags | osdc->client->options->read_from_replica;
1134
1135         req->r_snapid = vino.snap;
1136         if (flags & CEPH_OSD_FLAG_WRITE)
1137                 req->r_data_offset = off;
1138
1139         if (num_ops > 1) {
1140                 int num_req_ops, num_rep_ops;
1141
1142                 /*
1143                  * If this is a multi-op write request, assume that we'll need
1144                  * request ops. If it's a multi-op read then assume we'll need
1145                  * reply ops. Anything else and call it -EINVAL.
1146                  */
1147                 if (flags & CEPH_OSD_FLAG_WRITE) {
1148                         num_req_ops = num_ops;
1149                         num_rep_ops = 0;
1150                 } else if (flags & CEPH_OSD_FLAG_READ) {
1151                         num_req_ops = 0;
1152                         num_rep_ops = num_ops;
1153                 } else {
1154                         r = -EINVAL;
1155                         goto fail;
1156                 }
1157
1158                 r = __ceph_osdc_alloc_messages(req, GFP_NOFS, num_req_ops,
1159                                                num_rep_ops);
1160         } else {
1161                 r = ceph_osdc_alloc_messages(req, GFP_NOFS);
1162         }
1163         if (r)
1164                 goto fail;
1165
1166         return req;
1167
1168 fail:
1169         ceph_osdc_put_request(req);
1170         return ERR_PTR(r);
1171 }
1172 EXPORT_SYMBOL(ceph_osdc_new_request);
1173
1174 int __ceph_alloc_sparse_ext_map(struct ceph_osd_req_op *op, int cnt)
1175 {
1176         WARN_ON(op->op != CEPH_OSD_OP_SPARSE_READ);
1177
1178         op->extent.sparse_ext_cnt = cnt;
1179         op->extent.sparse_ext = kmalloc_array(cnt,
1180                                               sizeof(*op->extent.sparse_ext),
1181                                               GFP_NOFS);
1182         if (!op->extent.sparse_ext)
1183                 return -ENOMEM;
1184         return 0;
1185 }
1186 EXPORT_SYMBOL(__ceph_alloc_sparse_ext_map);
1187
1188 /*
1189  * We keep osd requests in an rbtree, sorted by ->r_tid.
1190  */
1191 DEFINE_RB_FUNCS(request, struct ceph_osd_request, r_tid, r_node)
1192 DEFINE_RB_FUNCS(request_mc, struct ceph_osd_request, r_tid, r_mc_node)
1193
1194 /*
1195  * Call @fn on each OSD request as long as @fn returns 0.
1196  */
1197 static void for_each_request(struct ceph_osd_client *osdc,
1198                         int (*fn)(struct ceph_osd_request *req, void *arg),
1199                         void *arg)
1200 {
1201         struct rb_node *n, *p;
1202
1203         for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
1204                 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
1205
1206                 for (p = rb_first(&osd->o_requests); p; ) {
1207                         struct ceph_osd_request *req =
1208                             rb_entry(p, struct ceph_osd_request, r_node);
1209
1210                         p = rb_next(p);
1211                         if (fn(req, arg))
1212                                 return;
1213                 }
1214         }
1215
1216         for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
1217                 struct ceph_osd_request *req =
1218                     rb_entry(p, struct ceph_osd_request, r_node);
1219
1220                 p = rb_next(p);
1221                 if (fn(req, arg))
1222                         return;
1223         }
1224 }
1225
1226 static bool osd_homeless(struct ceph_osd *osd)
1227 {
1228         return osd->o_osd == CEPH_HOMELESS_OSD;
1229 }
1230
1231 static bool osd_registered(struct ceph_osd *osd)
1232 {
1233         verify_osdc_locked(osd->o_osdc);
1234
1235         return !RB_EMPTY_NODE(&osd->o_node);
1236 }
1237
1238 /*
1239  * Assumes @osd is zero-initialized.
1240  */
1241 static void osd_init(struct ceph_osd *osd)
1242 {
1243         refcount_set(&osd->o_ref, 1);
1244         RB_CLEAR_NODE(&osd->o_node);
1245         spin_lock_init(&osd->o_requests_lock);
1246         osd->o_requests = RB_ROOT;
1247         osd->o_linger_requests = RB_ROOT;
1248         osd->o_backoff_mappings = RB_ROOT;
1249         osd->o_backoffs_by_id = RB_ROOT;
1250         INIT_LIST_HEAD(&osd->o_osd_lru);
1251         INIT_LIST_HEAD(&osd->o_keepalive_item);
1252         osd->o_incarnation = 1;
1253         mutex_init(&osd->lock);
1254 }
1255
1256 static void ceph_init_sparse_read(struct ceph_sparse_read *sr)
1257 {
1258         kfree(sr->sr_extent);
1259         memset(sr, '\0', sizeof(*sr));
1260         sr->sr_state = CEPH_SPARSE_READ_HDR;
1261 }
1262
1263 static void osd_cleanup(struct ceph_osd *osd)
1264 {
1265         WARN_ON(!RB_EMPTY_NODE(&osd->o_node));
1266         WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
1267         WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
1268         WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoff_mappings));
1269         WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoffs_by_id));
1270         WARN_ON(!list_empty(&osd->o_osd_lru));
1271         WARN_ON(!list_empty(&osd->o_keepalive_item));
1272
1273         ceph_init_sparse_read(&osd->o_sparse_read);
1274
1275         if (osd->o_auth.authorizer) {
1276                 WARN_ON(osd_homeless(osd));
1277                 ceph_auth_destroy_authorizer(osd->o_auth.authorizer);
1278         }
1279 }
1280
1281 /*
1282  * Track open sessions with osds.
1283  */
1284 static struct ceph_osd *create_osd(struct ceph_osd_client *osdc, int onum)
1285 {
1286         struct ceph_osd *osd;
1287
1288         WARN_ON(onum == CEPH_HOMELESS_OSD);
1289
1290         osd = kzalloc(sizeof(*osd), GFP_NOIO | __GFP_NOFAIL);
1291         osd_init(osd);
1292         osd->o_osdc = osdc;
1293         osd->o_osd = onum;
1294         osd->o_sparse_op_idx = -1;
1295
1296         ceph_init_sparse_read(&osd->o_sparse_read);
1297
1298         ceph_con_init(&osd->o_con, osd, &osd_con_ops, &osdc->client->msgr);
1299
1300         return osd;
1301 }
1302
1303 static struct ceph_osd *get_osd(struct ceph_osd *osd)
1304 {
1305         if (refcount_inc_not_zero(&osd->o_ref)) {
1306                 dout("get_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref)-1,
1307                      refcount_read(&osd->o_ref));
1308                 return osd;
1309         } else {
1310                 dout("get_osd %p FAIL\n", osd);
1311                 return NULL;
1312         }
1313 }
1314
1315 static void put_osd(struct ceph_osd *osd)
1316 {
1317         dout("put_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref),
1318              refcount_read(&osd->o_ref) - 1);
1319         if (refcount_dec_and_test(&osd->o_ref)) {
1320                 osd_cleanup(osd);
1321                 kfree(osd);
1322         }
1323 }
1324
1325 DEFINE_RB_FUNCS(osd, struct ceph_osd, o_osd, o_node)
1326
1327 static void __move_osd_to_lru(struct ceph_osd *osd)
1328 {
1329         struct ceph_osd_client *osdc = osd->o_osdc;
1330
1331         dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1332         BUG_ON(!list_empty(&osd->o_osd_lru));
1333
1334         spin_lock(&osdc->osd_lru_lock);
1335         list_add_tail(&osd->o_osd_lru, &osdc->osd_lru);
1336         spin_unlock(&osdc->osd_lru_lock);
1337
1338         osd->lru_ttl = jiffies + osdc->client->options->osd_idle_ttl;
1339 }
1340
1341 static void maybe_move_osd_to_lru(struct ceph_osd *osd)
1342 {
1343         if (RB_EMPTY_ROOT(&osd->o_requests) &&
1344             RB_EMPTY_ROOT(&osd->o_linger_requests))
1345                 __move_osd_to_lru(osd);
1346 }
1347
1348 static void __remove_osd_from_lru(struct ceph_osd *osd)
1349 {
1350         struct ceph_osd_client *osdc = osd->o_osdc;
1351
1352         dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1353
1354         spin_lock(&osdc->osd_lru_lock);
1355         if (!list_empty(&osd->o_osd_lru))
1356                 list_del_init(&osd->o_osd_lru);
1357         spin_unlock(&osdc->osd_lru_lock);
1358 }
1359
1360 /*
1361  * Close the connection and assign any leftover requests to the
1362  * homeless session.
1363  */
1364 static void close_osd(struct ceph_osd *osd)
1365 {
1366         struct ceph_osd_client *osdc = osd->o_osdc;
1367         struct rb_node *n;
1368
1369         verify_osdc_wrlocked(osdc);
1370         dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1371
1372         ceph_con_close(&osd->o_con);
1373
1374         for (n = rb_first(&osd->o_requests); n; ) {
1375                 struct ceph_osd_request *req =
1376                     rb_entry(n, struct ceph_osd_request, r_node);
1377
1378                 n = rb_next(n); /* unlink_request() */
1379
1380                 dout(" reassigning req %p tid %llu\n", req, req->r_tid);
1381                 unlink_request(osd, req);
1382                 link_request(&osdc->homeless_osd, req);
1383         }
1384         for (n = rb_first(&osd->o_linger_requests); n; ) {
1385                 struct ceph_osd_linger_request *lreq =
1386                     rb_entry(n, struct ceph_osd_linger_request, node);
1387
1388                 n = rb_next(n); /* unlink_linger() */
1389
1390                 dout(" reassigning lreq %p linger_id %llu\n", lreq,
1391                      lreq->linger_id);
1392                 unlink_linger(osd, lreq);
1393                 link_linger(&osdc->homeless_osd, lreq);
1394         }
1395         clear_backoffs(osd);
1396
1397         __remove_osd_from_lru(osd);
1398         erase_osd(&osdc->osds, osd);
1399         put_osd(osd);
1400 }
1401
1402 /*
1403  * reset osd connect
1404  */
1405 static int reopen_osd(struct ceph_osd *osd)
1406 {
1407         struct ceph_entity_addr *peer_addr;
1408
1409         dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1410
1411         if (RB_EMPTY_ROOT(&osd->o_requests) &&
1412             RB_EMPTY_ROOT(&osd->o_linger_requests)) {
1413                 close_osd(osd);
1414                 return -ENODEV;
1415         }
1416
1417         peer_addr = &osd->o_osdc->osdmap->osd_addr[osd->o_osd];
1418         if (!memcmp(peer_addr, &osd->o_con.peer_addr, sizeof (*peer_addr)) &&
1419                         !ceph_con_opened(&osd->o_con)) {
1420                 struct rb_node *n;
1421
1422                 dout("osd addr hasn't changed and connection never opened, "
1423                      "letting msgr retry\n");
1424                 /* touch each r_stamp for handle_timeout()'s benfit */
1425                 for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
1426                         struct ceph_osd_request *req =
1427                             rb_entry(n, struct ceph_osd_request, r_node);
1428                         req->r_stamp = jiffies;
1429                 }
1430
1431                 return -EAGAIN;
1432         }
1433
1434         ceph_con_close(&osd->o_con);
1435         ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd, peer_addr);
1436         osd->o_incarnation++;
1437
1438         return 0;
1439 }
1440
1441 static struct ceph_osd *lookup_create_osd(struct ceph_osd_client *osdc, int o,
1442                                           bool wrlocked)
1443 {
1444         struct ceph_osd *osd;
1445
1446         if (wrlocked)
1447                 verify_osdc_wrlocked(osdc);
1448         else
1449                 verify_osdc_locked(osdc);
1450
1451         if (o != CEPH_HOMELESS_OSD)
1452                 osd = lookup_osd(&osdc->osds, o);
1453         else
1454                 osd = &osdc->homeless_osd;
1455         if (!osd) {
1456                 if (!wrlocked)
1457                         return ERR_PTR(-EAGAIN);
1458
1459                 osd = create_osd(osdc, o);
1460                 insert_osd(&osdc->osds, osd);
1461                 ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd,
1462                               &osdc->osdmap->osd_addr[osd->o_osd]);
1463         }
1464
1465         dout("%s osdc %p osd%d -> osd %p\n", __func__, osdc, o, osd);
1466         return osd;
1467 }
1468
1469 /*
1470  * Create request <-> OSD session relation.
1471  *
1472  * @req has to be assigned a tid, @osd may be homeless.
1473  */
1474 static void link_request(struct ceph_osd *osd, struct ceph_osd_request *req)
1475 {
1476         verify_osd_locked(osd);
1477         WARN_ON(!req->r_tid || req->r_osd);
1478         dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1479              req, req->r_tid);
1480
1481         if (!osd_homeless(osd))
1482                 __remove_osd_from_lru(osd);
1483         else
1484                 atomic_inc(&osd->o_osdc->num_homeless);
1485
1486         get_osd(osd);
1487         spin_lock(&osd->o_requests_lock);
1488         insert_request(&osd->o_requests, req);
1489         spin_unlock(&osd->o_requests_lock);
1490         req->r_osd = osd;
1491 }
1492
1493 static void unlink_request(struct ceph_osd *osd, struct ceph_osd_request *req)
1494 {
1495         verify_osd_locked(osd);
1496         WARN_ON(req->r_osd != osd);
1497         dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1498              req, req->r_tid);
1499
1500         req->r_osd = NULL;
1501         spin_lock(&osd->o_requests_lock);
1502         erase_request(&osd->o_requests, req);
1503         spin_unlock(&osd->o_requests_lock);
1504         put_osd(osd);
1505
1506         if (!osd_homeless(osd))
1507                 maybe_move_osd_to_lru(osd);
1508         else
1509                 atomic_dec(&osd->o_osdc->num_homeless);
1510 }
1511
1512 static bool __pool_full(struct ceph_pg_pool_info *pi)
1513 {
1514         return pi->flags & CEPH_POOL_FLAG_FULL;
1515 }
1516
1517 static bool have_pool_full(struct ceph_osd_client *osdc)
1518 {
1519         struct rb_node *n;
1520
1521         for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
1522                 struct ceph_pg_pool_info *pi =
1523                     rb_entry(n, struct ceph_pg_pool_info, node);
1524
1525                 if (__pool_full(pi))
1526                         return true;
1527         }
1528
1529         return false;
1530 }
1531
1532 static bool pool_full(struct ceph_osd_client *osdc, s64 pool_id)
1533 {
1534         struct ceph_pg_pool_info *pi;
1535
1536         pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
1537         if (!pi)
1538                 return false;
1539
1540         return __pool_full(pi);
1541 }
1542
1543 /*
1544  * Returns whether a request should be blocked from being sent
1545  * based on the current osdmap and osd_client settings.
1546  */
1547 static bool target_should_be_paused(struct ceph_osd_client *osdc,
1548                                     const struct ceph_osd_request_target *t,
1549                                     struct ceph_pg_pool_info *pi)
1550 {
1551         bool pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
1552         bool pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
1553                        ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
1554                        __pool_full(pi);
1555
1556         WARN_ON(pi->id != t->target_oloc.pool);
1557         return ((t->flags & CEPH_OSD_FLAG_READ) && pauserd) ||
1558                ((t->flags & CEPH_OSD_FLAG_WRITE) && pausewr) ||
1559                (osdc->osdmap->epoch < osdc->epoch_barrier);
1560 }
1561
1562 static int pick_random_replica(const struct ceph_osds *acting)
1563 {
1564         int i = get_random_u32_below(acting->size);
1565
1566         dout("%s picked osd%d, primary osd%d\n", __func__,
1567              acting->osds[i], acting->primary);
1568         return i;
1569 }
1570
1571 /*
1572  * Picks the closest replica based on client's location given by
1573  * crush_location option.  Prefers the primary if the locality is
1574  * the same.
1575  */
1576 static int pick_closest_replica(struct ceph_osd_client *osdc,
1577                                 const struct ceph_osds *acting)
1578 {
1579         struct ceph_options *opt = osdc->client->options;
1580         int best_i, best_locality;
1581         int i = 0, locality;
1582
1583         do {
1584                 locality = ceph_get_crush_locality(osdc->osdmap,
1585                                                    acting->osds[i],
1586                                                    &opt->crush_locs);
1587                 if (i == 0 ||
1588                     (locality >= 0 && best_locality < 0) ||
1589                     (locality >= 0 && best_locality >= 0 &&
1590                      locality < best_locality)) {
1591                         best_i = i;
1592                         best_locality = locality;
1593                 }
1594         } while (++i < acting->size);
1595
1596         dout("%s picked osd%d with locality %d, primary osd%d\n", __func__,
1597              acting->osds[best_i], best_locality, acting->primary);
1598         return best_i;
1599 }
1600
1601 enum calc_target_result {
1602         CALC_TARGET_NO_ACTION = 0,
1603         CALC_TARGET_NEED_RESEND,
1604         CALC_TARGET_POOL_DNE,
1605 };
1606
1607 static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
1608                                            struct ceph_osd_request_target *t,
1609                                            bool any_change)
1610 {
1611         struct ceph_pg_pool_info *pi;
1612         struct ceph_pg pgid, last_pgid;
1613         struct ceph_osds up, acting;
1614         bool is_read = t->flags & CEPH_OSD_FLAG_READ;
1615         bool is_write = t->flags & CEPH_OSD_FLAG_WRITE;
1616         bool force_resend = false;
1617         bool unpaused = false;
1618         bool legacy_change = false;
1619         bool split = false;
1620         bool sort_bitwise = ceph_osdmap_flag(osdc, CEPH_OSDMAP_SORTBITWISE);
1621         bool recovery_deletes = ceph_osdmap_flag(osdc,
1622                                                  CEPH_OSDMAP_RECOVERY_DELETES);
1623         enum calc_target_result ct_res;
1624
1625         t->epoch = osdc->osdmap->epoch;
1626         pi = ceph_pg_pool_by_id(osdc->osdmap, t->base_oloc.pool);
1627         if (!pi) {
1628                 t->osd = CEPH_HOMELESS_OSD;
1629                 ct_res = CALC_TARGET_POOL_DNE;
1630                 goto out;
1631         }
1632
1633         if (osdc->osdmap->epoch == pi->last_force_request_resend) {
1634                 if (t->last_force_resend < pi->last_force_request_resend) {
1635                         t->last_force_resend = pi->last_force_request_resend;
1636                         force_resend = true;
1637                 } else if (t->last_force_resend == 0) {
1638                         force_resend = true;
1639                 }
1640         }
1641
1642         /* apply tiering */
1643         ceph_oid_copy(&t->target_oid, &t->base_oid);
1644         ceph_oloc_copy(&t->target_oloc, &t->base_oloc);
1645         if ((t->flags & CEPH_OSD_FLAG_IGNORE_OVERLAY) == 0) {
1646                 if (is_read && pi->read_tier >= 0)
1647                         t->target_oloc.pool = pi->read_tier;
1648                 if (is_write && pi->write_tier >= 0)
1649                         t->target_oloc.pool = pi->write_tier;
1650
1651                 pi = ceph_pg_pool_by_id(osdc->osdmap, t->target_oloc.pool);
1652                 if (!pi) {
1653                         t->osd = CEPH_HOMELESS_OSD;
1654                         ct_res = CALC_TARGET_POOL_DNE;
1655                         goto out;
1656                 }
1657         }
1658
1659         __ceph_object_locator_to_pg(pi, &t->target_oid, &t->target_oloc, &pgid);
1660         last_pgid.pool = pgid.pool;
1661         last_pgid.seed = ceph_stable_mod(pgid.seed, t->pg_num, t->pg_num_mask);
1662
1663         ceph_pg_to_up_acting_osds(osdc->osdmap, pi, &pgid, &up, &acting);
1664         if (any_change &&
1665             ceph_is_new_interval(&t->acting,
1666                                  &acting,
1667                                  &t->up,
1668                                  &up,
1669                                  t->size,
1670                                  pi->size,
1671                                  t->min_size,
1672                                  pi->min_size,
1673                                  t->pg_num,
1674                                  pi->pg_num,
1675                                  t->sort_bitwise,
1676                                  sort_bitwise,
1677                                  t->recovery_deletes,
1678                                  recovery_deletes,
1679                                  &last_pgid))
1680                 force_resend = true;
1681
1682         if (t->paused && !target_should_be_paused(osdc, t, pi)) {
1683                 t->paused = false;
1684                 unpaused = true;
1685         }
1686         legacy_change = ceph_pg_compare(&t->pgid, &pgid) ||
1687                         ceph_osds_changed(&t->acting, &acting,
1688                                           t->used_replica || any_change);
1689         if (t->pg_num)
1690                 split = ceph_pg_is_split(&last_pgid, t->pg_num, pi->pg_num);
1691
1692         if (legacy_change || force_resend || split) {
1693                 t->pgid = pgid; /* struct */
1694                 ceph_pg_to_primary_shard(osdc->osdmap, pi, &pgid, &t->spgid);
1695                 ceph_osds_copy(&t->acting, &acting);
1696                 ceph_osds_copy(&t->up, &up);
1697                 t->size = pi->size;
1698                 t->min_size = pi->min_size;
1699                 t->pg_num = pi->pg_num;
1700                 t->pg_num_mask = pi->pg_num_mask;
1701                 t->sort_bitwise = sort_bitwise;
1702                 t->recovery_deletes = recovery_deletes;
1703
1704                 if ((t->flags & (CEPH_OSD_FLAG_BALANCE_READS |
1705                                  CEPH_OSD_FLAG_LOCALIZE_READS)) &&
1706                     !is_write && pi->type == CEPH_POOL_TYPE_REP &&
1707                     acting.size > 1) {
1708                         int pos;
1709
1710                         WARN_ON(!is_read || acting.osds[0] != acting.primary);
1711                         if (t->flags & CEPH_OSD_FLAG_BALANCE_READS) {
1712                                 pos = pick_random_replica(&acting);
1713                         } else {
1714                                 pos = pick_closest_replica(osdc, &acting);
1715                         }
1716                         t->osd = acting.osds[pos];
1717                         t->used_replica = pos > 0;
1718                 } else {
1719                         t->osd = acting.primary;
1720                         t->used_replica = false;
1721                 }
1722         }
1723
1724         if (unpaused || legacy_change || force_resend || split)
1725                 ct_res = CALC_TARGET_NEED_RESEND;
1726         else
1727                 ct_res = CALC_TARGET_NO_ACTION;
1728
1729 out:
1730         dout("%s t %p -> %d%d%d%d ct_res %d osd%d\n", __func__, t, unpaused,
1731              legacy_change, force_resend, split, ct_res, t->osd);
1732         return ct_res;
1733 }
1734
1735 static struct ceph_spg_mapping *alloc_spg_mapping(void)
1736 {
1737         struct ceph_spg_mapping *spg;
1738
1739         spg = kmalloc(sizeof(*spg), GFP_NOIO);
1740         if (!spg)
1741                 return NULL;
1742
1743         RB_CLEAR_NODE(&spg->node);
1744         spg->backoffs = RB_ROOT;
1745         return spg;
1746 }
1747
1748 static void free_spg_mapping(struct ceph_spg_mapping *spg)
1749 {
1750         WARN_ON(!RB_EMPTY_NODE(&spg->node));
1751         WARN_ON(!RB_EMPTY_ROOT(&spg->backoffs));
1752
1753         kfree(spg);
1754 }
1755
1756 /*
1757  * rbtree of ceph_spg_mapping for handling map<spg_t, ...>, similar to
1758  * ceph_pg_mapping.  Used to track OSD backoffs -- a backoff [range] is
1759  * defined only within a specific spgid; it does not pass anything to
1760  * children on split, or to another primary.
1761  */
1762 DEFINE_RB_FUNCS2(spg_mapping, struct ceph_spg_mapping, spgid, ceph_spg_compare,
1763                  RB_BYPTR, const struct ceph_spg *, node)
1764
1765 static u64 hoid_get_bitwise_key(const struct ceph_hobject_id *hoid)
1766 {
1767         return hoid->is_max ? 0x100000000ull : hoid->hash_reverse_bits;
1768 }
1769
1770 static void hoid_get_effective_key(const struct ceph_hobject_id *hoid,
1771                                    void **pkey, size_t *pkey_len)
1772 {
1773         if (hoid->key_len) {
1774                 *pkey = hoid->key;
1775                 *pkey_len = hoid->key_len;
1776         } else {
1777                 *pkey = hoid->oid;
1778                 *pkey_len = hoid->oid_len;
1779         }
1780 }
1781
1782 static int compare_names(const void *name1, size_t name1_len,
1783                          const void *name2, size_t name2_len)
1784 {
1785         int ret;
1786
1787         ret = memcmp(name1, name2, min(name1_len, name2_len));
1788         if (!ret) {
1789                 if (name1_len < name2_len)
1790                         ret = -1;
1791                 else if (name1_len > name2_len)
1792                         ret = 1;
1793         }
1794         return ret;
1795 }
1796
1797 static int hoid_compare(const struct ceph_hobject_id *lhs,
1798                         const struct ceph_hobject_id *rhs)
1799 {
1800         void *effective_key1, *effective_key2;
1801         size_t effective_key1_len, effective_key2_len;
1802         int ret;
1803
1804         if (lhs->is_max < rhs->is_max)
1805                 return -1;
1806         if (lhs->is_max > rhs->is_max)
1807                 return 1;
1808
1809         if (lhs->pool < rhs->pool)
1810                 return -1;
1811         if (lhs->pool > rhs->pool)
1812                 return 1;
1813
1814         if (hoid_get_bitwise_key(lhs) < hoid_get_bitwise_key(rhs))
1815                 return -1;
1816         if (hoid_get_bitwise_key(lhs) > hoid_get_bitwise_key(rhs))
1817                 return 1;
1818
1819         ret = compare_names(lhs->nspace, lhs->nspace_len,
1820                             rhs->nspace, rhs->nspace_len);
1821         if (ret)
1822                 return ret;
1823
1824         hoid_get_effective_key(lhs, &effective_key1, &effective_key1_len);
1825         hoid_get_effective_key(rhs, &effective_key2, &effective_key2_len);
1826         ret = compare_names(effective_key1, effective_key1_len,
1827                             effective_key2, effective_key2_len);
1828         if (ret)
1829                 return ret;
1830
1831         ret = compare_names(lhs->oid, lhs->oid_len, rhs->oid, rhs->oid_len);
1832         if (ret)
1833                 return ret;
1834
1835         if (lhs->snapid < rhs->snapid)
1836                 return -1;
1837         if (lhs->snapid > rhs->snapid)
1838                 return 1;
1839
1840         return 0;
1841 }
1842
1843 /*
1844  * For decoding ->begin and ->end of MOSDBackoff only -- no MIN/MAX
1845  * compat stuff here.
1846  *
1847  * Assumes @hoid is zero-initialized.
1848  */
1849 static int decode_hoid(void **p, void *end, struct ceph_hobject_id *hoid)
1850 {
1851         u8 struct_v;
1852         u32 struct_len;
1853         int ret;
1854
1855         ret = ceph_start_decoding(p, end, 4, "hobject_t", &struct_v,
1856                                   &struct_len);
1857         if (ret)
1858                 return ret;
1859
1860         if (struct_v < 4) {
1861                 pr_err("got struct_v %d < 4 of hobject_t\n", struct_v);
1862                 goto e_inval;
1863         }
1864
1865         hoid->key = ceph_extract_encoded_string(p, end, &hoid->key_len,
1866                                                 GFP_NOIO);
1867         if (IS_ERR(hoid->key)) {
1868                 ret = PTR_ERR(hoid->key);
1869                 hoid->key = NULL;
1870                 return ret;
1871         }
1872
1873         hoid->oid = ceph_extract_encoded_string(p, end, &hoid->oid_len,
1874                                                 GFP_NOIO);
1875         if (IS_ERR(hoid->oid)) {
1876                 ret = PTR_ERR(hoid->oid);
1877                 hoid->oid = NULL;
1878                 return ret;
1879         }
1880
1881         ceph_decode_64_safe(p, end, hoid->snapid, e_inval);
1882         ceph_decode_32_safe(p, end, hoid->hash, e_inval);
1883         ceph_decode_8_safe(p, end, hoid->is_max, e_inval);
1884
1885         hoid->nspace = ceph_extract_encoded_string(p, end, &hoid->nspace_len,
1886                                                    GFP_NOIO);
1887         if (IS_ERR(hoid->nspace)) {
1888                 ret = PTR_ERR(hoid->nspace);
1889                 hoid->nspace = NULL;
1890                 return ret;
1891         }
1892
1893         ceph_decode_64_safe(p, end, hoid->pool, e_inval);
1894
1895         ceph_hoid_build_hash_cache(hoid);
1896         return 0;
1897
1898 e_inval:
1899         return -EINVAL;
1900 }
1901
1902 static int hoid_encoding_size(const struct ceph_hobject_id *hoid)
1903 {
1904         return 8 + 4 + 1 + 8 + /* snapid, hash, is_max, pool */
1905                4 + hoid->key_len + 4 + hoid->oid_len + 4 + hoid->nspace_len;
1906 }
1907
1908 static void encode_hoid(void **p, void *end, const struct ceph_hobject_id *hoid)
1909 {
1910         ceph_start_encoding(p, 4, 3, hoid_encoding_size(hoid));
1911         ceph_encode_string(p, end, hoid->key, hoid->key_len);
1912         ceph_encode_string(p, end, hoid->oid, hoid->oid_len);
1913         ceph_encode_64(p, hoid->snapid);
1914         ceph_encode_32(p, hoid->hash);
1915         ceph_encode_8(p, hoid->is_max);
1916         ceph_encode_string(p, end, hoid->nspace, hoid->nspace_len);
1917         ceph_encode_64(p, hoid->pool);
1918 }
1919
1920 static void free_hoid(struct ceph_hobject_id *hoid)
1921 {
1922         if (hoid) {
1923                 kfree(hoid->key);
1924                 kfree(hoid->oid);
1925                 kfree(hoid->nspace);
1926                 kfree(hoid);
1927         }
1928 }
1929
1930 static struct ceph_osd_backoff *alloc_backoff(void)
1931 {
1932         struct ceph_osd_backoff *backoff;
1933
1934         backoff = kzalloc(sizeof(*backoff), GFP_NOIO);
1935         if (!backoff)
1936                 return NULL;
1937
1938         RB_CLEAR_NODE(&backoff->spg_node);
1939         RB_CLEAR_NODE(&backoff->id_node);
1940         return backoff;
1941 }
1942
1943 static void free_backoff(struct ceph_osd_backoff *backoff)
1944 {
1945         WARN_ON(!RB_EMPTY_NODE(&backoff->spg_node));
1946         WARN_ON(!RB_EMPTY_NODE(&backoff->id_node));
1947
1948         free_hoid(backoff->begin);
1949         free_hoid(backoff->end);
1950         kfree(backoff);
1951 }
1952
1953 /*
1954  * Within a specific spgid, backoffs are managed by ->begin hoid.
1955  */
1956 DEFINE_RB_INSDEL_FUNCS2(backoff, struct ceph_osd_backoff, begin, hoid_compare,
1957                         RB_BYVAL, spg_node);
1958
1959 static struct ceph_osd_backoff *lookup_containing_backoff(struct rb_root *root,
1960                                             const struct ceph_hobject_id *hoid)
1961 {
1962         struct rb_node *n = root->rb_node;
1963
1964         while (n) {
1965                 struct ceph_osd_backoff *cur =
1966                     rb_entry(n, struct ceph_osd_backoff, spg_node);
1967                 int cmp;
1968
1969                 cmp = hoid_compare(hoid, cur->begin);
1970                 if (cmp < 0) {
1971                         n = n->rb_left;
1972                 } else if (cmp > 0) {
1973                         if (hoid_compare(hoid, cur->end) < 0)
1974                                 return cur;
1975
1976                         n = n->rb_right;
1977                 } else {
1978                         return cur;
1979                 }
1980         }
1981
1982         return NULL;
1983 }
1984
1985 /*
1986  * Each backoff has a unique id within its OSD session.
1987  */
1988 DEFINE_RB_FUNCS(backoff_by_id, struct ceph_osd_backoff, id, id_node)
1989
1990 static void clear_backoffs(struct ceph_osd *osd)
1991 {
1992         while (!RB_EMPTY_ROOT(&osd->o_backoff_mappings)) {
1993                 struct ceph_spg_mapping *spg =
1994                     rb_entry(rb_first(&osd->o_backoff_mappings),
1995                              struct ceph_spg_mapping, node);
1996
1997                 while (!RB_EMPTY_ROOT(&spg->backoffs)) {
1998                         struct ceph_osd_backoff *backoff =
1999                             rb_entry(rb_first(&spg->backoffs),
2000                                      struct ceph_osd_backoff, spg_node);
2001
2002                         erase_backoff(&spg->backoffs, backoff);
2003                         erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
2004                         free_backoff(backoff);
2005                 }
2006                 erase_spg_mapping(&osd->o_backoff_mappings, spg);
2007                 free_spg_mapping(spg);
2008         }
2009 }
2010
2011 /*
2012  * Set up a temporary, non-owning view into @t.
2013  */
2014 static void hoid_fill_from_target(struct ceph_hobject_id *hoid,
2015                                   const struct ceph_osd_request_target *t)
2016 {
2017         hoid->key = NULL;
2018         hoid->key_len = 0;
2019         hoid->oid = t->target_oid.name;
2020         hoid->oid_len = t->target_oid.name_len;
2021         hoid->snapid = CEPH_NOSNAP;
2022         hoid->hash = t->pgid.seed;
2023         hoid->is_max = false;
2024         if (t->target_oloc.pool_ns) {
2025                 hoid->nspace = t->target_oloc.pool_ns->str;
2026                 hoid->nspace_len = t->target_oloc.pool_ns->len;
2027         } else {
2028                 hoid->nspace = NULL;
2029                 hoid->nspace_len = 0;
2030         }
2031         hoid->pool = t->target_oloc.pool;
2032         ceph_hoid_build_hash_cache(hoid);
2033 }
2034
2035 static bool should_plug_request(struct ceph_osd_request *req)
2036 {
2037         struct ceph_osd *osd = req->r_osd;
2038         struct ceph_spg_mapping *spg;
2039         struct ceph_osd_backoff *backoff;
2040         struct ceph_hobject_id hoid;
2041
2042         spg = lookup_spg_mapping(&osd->o_backoff_mappings, &req->r_t.spgid);
2043         if (!spg)
2044                 return false;
2045
2046         hoid_fill_from_target(&hoid, &req->r_t);
2047         backoff = lookup_containing_backoff(&spg->backoffs, &hoid);
2048         if (!backoff)
2049                 return false;
2050
2051         dout("%s req %p tid %llu backoff osd%d spgid %llu.%xs%d id %llu\n",
2052              __func__, req, req->r_tid, osd->o_osd, backoff->spgid.pgid.pool,
2053              backoff->spgid.pgid.seed, backoff->spgid.shard, backoff->id);
2054         return true;
2055 }
2056
2057 /*
2058  * Keep get_num_data_items() in sync with this function.
2059  */
2060 static void setup_request_data(struct ceph_osd_request *req)
2061 {
2062         struct ceph_msg *request_msg = req->r_request;
2063         struct ceph_msg *reply_msg = req->r_reply;
2064         struct ceph_osd_req_op *op;
2065
2066         if (req->r_request->num_data_items || req->r_reply->num_data_items)
2067                 return;
2068
2069         WARN_ON(request_msg->data_length || reply_msg->data_length);
2070         for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) {
2071                 switch (op->op) {
2072                 /* request */
2073                 case CEPH_OSD_OP_WRITE:
2074                 case CEPH_OSD_OP_WRITEFULL:
2075                         WARN_ON(op->indata_len != op->extent.length);
2076                         ceph_osdc_msg_data_add(request_msg,
2077                                                &op->extent.osd_data);
2078                         break;
2079                 case CEPH_OSD_OP_SETXATTR:
2080                 case CEPH_OSD_OP_CMPXATTR:
2081                         WARN_ON(op->indata_len != op->xattr.name_len +
2082                                                   op->xattr.value_len);
2083                         ceph_osdc_msg_data_add(request_msg,
2084                                                &op->xattr.osd_data);
2085                         break;
2086                 case CEPH_OSD_OP_NOTIFY_ACK:
2087                         ceph_osdc_msg_data_add(request_msg,
2088                                                &op->notify_ack.request_data);
2089                         break;
2090                 case CEPH_OSD_OP_COPY_FROM2:
2091                         ceph_osdc_msg_data_add(request_msg,
2092                                                &op->copy_from.osd_data);
2093                         break;
2094
2095                 /* reply */
2096                 case CEPH_OSD_OP_STAT:
2097                         ceph_osdc_msg_data_add(reply_msg,
2098                                                &op->raw_data_in);
2099                         break;
2100                 case CEPH_OSD_OP_READ:
2101                 case CEPH_OSD_OP_SPARSE_READ:
2102                         ceph_osdc_msg_data_add(reply_msg,
2103                                                &op->extent.osd_data);
2104                         break;
2105                 case CEPH_OSD_OP_LIST_WATCHERS:
2106                         ceph_osdc_msg_data_add(reply_msg,
2107                                                &op->list_watchers.response_data);
2108                         break;
2109
2110                 /* both */
2111                 case CEPH_OSD_OP_CALL:
2112                         WARN_ON(op->indata_len != op->cls.class_len +
2113                                                   op->cls.method_len +
2114                                                   op->cls.indata_len);
2115                         ceph_osdc_msg_data_add(request_msg,
2116                                                &op->cls.request_info);
2117                         /* optional, can be NONE */
2118                         ceph_osdc_msg_data_add(request_msg,
2119                                                &op->cls.request_data);
2120                         /* optional, can be NONE */
2121                         ceph_osdc_msg_data_add(reply_msg,
2122                                                &op->cls.response_data);
2123                         break;
2124                 case CEPH_OSD_OP_NOTIFY:
2125                         ceph_osdc_msg_data_add(request_msg,
2126                                                &op->notify.request_data);
2127                         ceph_osdc_msg_data_add(reply_msg,
2128                                                &op->notify.response_data);
2129                         break;
2130                 }
2131         }
2132 }
2133
2134 static void encode_pgid(void **p, const struct ceph_pg *pgid)
2135 {
2136         ceph_encode_8(p, 1);
2137         ceph_encode_64(p, pgid->pool);
2138         ceph_encode_32(p, pgid->seed);
2139         ceph_encode_32(p, -1); /* preferred */
2140 }
2141
2142 static void encode_spgid(void **p, const struct ceph_spg *spgid)
2143 {
2144         ceph_start_encoding(p, 1, 1, CEPH_PGID_ENCODING_LEN + 1);
2145         encode_pgid(p, &spgid->pgid);
2146         ceph_encode_8(p, spgid->shard);
2147 }
2148
2149 static void encode_oloc(void **p, void *end,
2150                         const struct ceph_object_locator *oloc)
2151 {
2152         ceph_start_encoding(p, 5, 4, ceph_oloc_encoding_size(oloc));
2153         ceph_encode_64(p, oloc->pool);
2154         ceph_encode_32(p, -1); /* preferred */
2155         ceph_encode_32(p, 0);  /* key len */
2156         if (oloc->pool_ns)
2157                 ceph_encode_string(p, end, oloc->pool_ns->str,
2158                                    oloc->pool_ns->len);
2159         else
2160                 ceph_encode_32(p, 0);
2161 }
2162
2163 static void encode_request_partial(struct ceph_osd_request *req,
2164                                    struct ceph_msg *msg)
2165 {
2166         void *p = msg->front.iov_base;
2167         void *const end = p + msg->front_alloc_len;
2168         u32 data_len = 0;
2169         int i;
2170
2171         if (req->r_flags & CEPH_OSD_FLAG_WRITE) {
2172                 /* snapshots aren't writeable */
2173                 WARN_ON(req->r_snapid != CEPH_NOSNAP);
2174         } else {
2175                 WARN_ON(req->r_mtime.tv_sec || req->r_mtime.tv_nsec ||
2176                         req->r_data_offset || req->r_snapc);
2177         }
2178
2179         setup_request_data(req);
2180
2181         encode_spgid(&p, &req->r_t.spgid); /* actual spg */
2182         ceph_encode_32(&p, req->r_t.pgid.seed); /* raw hash */
2183         ceph_encode_32(&p, req->r_osdc->osdmap->epoch);
2184         ceph_encode_32(&p, req->r_flags);
2185
2186         /* reqid */
2187         ceph_start_encoding(&p, 2, 2, sizeof(struct ceph_osd_reqid));
2188         memset(p, 0, sizeof(struct ceph_osd_reqid));
2189         p += sizeof(struct ceph_osd_reqid);
2190
2191         /* trace */
2192         memset(p, 0, sizeof(struct ceph_blkin_trace_info));
2193         p += sizeof(struct ceph_blkin_trace_info);
2194
2195         ceph_encode_32(&p, 0); /* client_inc, always 0 */
2196         ceph_encode_timespec64(p, &req->r_mtime);
2197         p += sizeof(struct ceph_timespec);
2198
2199         encode_oloc(&p, end, &req->r_t.target_oloc);
2200         ceph_encode_string(&p, end, req->r_t.target_oid.name,
2201                            req->r_t.target_oid.name_len);
2202
2203         /* ops, can imply data */
2204         ceph_encode_16(&p, req->r_num_ops);
2205         for (i = 0; i < req->r_num_ops; i++) {
2206                 data_len += osd_req_encode_op(p, &req->r_ops[i]);
2207                 p += sizeof(struct ceph_osd_op);
2208         }
2209
2210         ceph_encode_64(&p, req->r_snapid); /* snapid */
2211         if (req->r_snapc) {
2212                 ceph_encode_64(&p, req->r_snapc->seq);
2213                 ceph_encode_32(&p, req->r_snapc->num_snaps);
2214                 for (i = 0; i < req->r_snapc->num_snaps; i++)
2215                         ceph_encode_64(&p, req->r_snapc->snaps[i]);
2216         } else {
2217                 ceph_encode_64(&p, 0); /* snap_seq */
2218                 ceph_encode_32(&p, 0); /* snaps len */
2219         }
2220
2221         ceph_encode_32(&p, req->r_attempts); /* retry_attempt */
2222         BUG_ON(p > end - 8); /* space for features */
2223
2224         msg->hdr.version = cpu_to_le16(8); /* MOSDOp v8 */
2225         /* front_len is finalized in encode_request_finish() */
2226         msg->front.iov_len = p - msg->front.iov_base;
2227         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2228         msg->hdr.data_len = cpu_to_le32(data_len);
2229         /*
2230          * The header "data_off" is a hint to the receiver allowing it
2231          * to align received data into its buffers such that there's no
2232          * need to re-copy it before writing it to disk (direct I/O).
2233          */
2234         msg->hdr.data_off = cpu_to_le16(req->r_data_offset);
2235
2236         dout("%s req %p msg %p oid %s oid_len %d\n", __func__, req, msg,
2237              req->r_t.target_oid.name, req->r_t.target_oid.name_len);
2238 }
2239
2240 static void encode_request_finish(struct ceph_msg *msg)
2241 {
2242         void *p = msg->front.iov_base;
2243         void *const partial_end = p + msg->front.iov_len;
2244         void *const end = p + msg->front_alloc_len;
2245
2246         if (CEPH_HAVE_FEATURE(msg->con->peer_features, RESEND_ON_SPLIT)) {
2247                 /* luminous OSD -- encode features and be done */
2248                 p = partial_end;
2249                 ceph_encode_64(&p, msg->con->peer_features);
2250         } else {
2251                 struct {
2252                         char spgid[CEPH_ENCODING_START_BLK_LEN +
2253                                    CEPH_PGID_ENCODING_LEN + 1];
2254                         __le32 hash;
2255                         __le32 epoch;
2256                         __le32 flags;
2257                         char reqid[CEPH_ENCODING_START_BLK_LEN +
2258                                    sizeof(struct ceph_osd_reqid)];
2259                         char trace[sizeof(struct ceph_blkin_trace_info)];
2260                         __le32 client_inc;
2261                         struct ceph_timespec mtime;
2262                 } __packed head;
2263                 struct ceph_pg pgid;
2264                 void *oloc, *oid, *tail;
2265                 int oloc_len, oid_len, tail_len;
2266                 int len;
2267
2268                 /*
2269                  * Pre-luminous OSD -- reencode v8 into v4 using @head
2270                  * as a temporary buffer.  Encode the raw PG; the rest
2271                  * is just a matter of moving oloc, oid and tail blobs
2272                  * around.
2273                  */
2274                 memcpy(&head, p, sizeof(head));
2275                 p += sizeof(head);
2276
2277                 oloc = p;
2278                 p += CEPH_ENCODING_START_BLK_LEN;
2279                 pgid.pool = ceph_decode_64(&p);
2280                 p += 4 + 4; /* preferred, key len */
2281                 len = ceph_decode_32(&p);
2282                 p += len;   /* nspace */
2283                 oloc_len = p - oloc;
2284
2285                 oid = p;
2286                 len = ceph_decode_32(&p);
2287                 p += len;
2288                 oid_len = p - oid;
2289
2290                 tail = p;
2291                 tail_len = partial_end - p;
2292
2293                 p = msg->front.iov_base;
2294                 ceph_encode_copy(&p, &head.client_inc, sizeof(head.client_inc));
2295                 ceph_encode_copy(&p, &head.epoch, sizeof(head.epoch));
2296                 ceph_encode_copy(&p, &head.flags, sizeof(head.flags));
2297                 ceph_encode_copy(&p, &head.mtime, sizeof(head.mtime));
2298
2299                 /* reassert_version */
2300                 memset(p, 0, sizeof(struct ceph_eversion));
2301                 p += sizeof(struct ceph_eversion);
2302
2303                 BUG_ON(p >= oloc);
2304                 memmove(p, oloc, oloc_len);
2305                 p += oloc_len;
2306
2307                 pgid.seed = le32_to_cpu(head.hash);
2308                 encode_pgid(&p, &pgid); /* raw pg */
2309
2310                 BUG_ON(p >= oid);
2311                 memmove(p, oid, oid_len);
2312                 p += oid_len;
2313
2314                 /* tail -- ops, snapid, snapc, retry_attempt */
2315                 BUG_ON(p >= tail);
2316                 memmove(p, tail, tail_len);
2317                 p += tail_len;
2318
2319                 msg->hdr.version = cpu_to_le16(4); /* MOSDOp v4 */
2320         }
2321
2322         BUG_ON(p > end);
2323         msg->front.iov_len = p - msg->front.iov_base;
2324         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2325
2326         dout("%s msg %p tid %llu %u+%u+%u v%d\n", __func__, msg,
2327              le64_to_cpu(msg->hdr.tid), le32_to_cpu(msg->hdr.front_len),
2328              le32_to_cpu(msg->hdr.middle_len), le32_to_cpu(msg->hdr.data_len),
2329              le16_to_cpu(msg->hdr.version));
2330 }
2331
2332 /*
2333  * @req has to be assigned a tid and registered.
2334  */
2335 static void send_request(struct ceph_osd_request *req)
2336 {
2337         struct ceph_osd *osd = req->r_osd;
2338
2339         verify_osd_locked(osd);
2340         WARN_ON(osd->o_osd != req->r_t.osd);
2341
2342         /* backoff? */
2343         if (should_plug_request(req))
2344                 return;
2345
2346         /*
2347          * We may have a previously queued request message hanging
2348          * around.  Cancel it to avoid corrupting the msgr.
2349          */
2350         if (req->r_sent)
2351                 ceph_msg_revoke(req->r_request);
2352
2353         req->r_flags |= CEPH_OSD_FLAG_KNOWN_REDIR;
2354         if (req->r_attempts)
2355                 req->r_flags |= CEPH_OSD_FLAG_RETRY;
2356         else
2357                 WARN_ON(req->r_flags & CEPH_OSD_FLAG_RETRY);
2358
2359         encode_request_partial(req, req->r_request);
2360
2361         dout("%s req %p tid %llu to pgid %llu.%x spgid %llu.%xs%d osd%d e%u flags 0x%x attempt %d\n",
2362              __func__, req, req->r_tid, req->r_t.pgid.pool, req->r_t.pgid.seed,
2363              req->r_t.spgid.pgid.pool, req->r_t.spgid.pgid.seed,
2364              req->r_t.spgid.shard, osd->o_osd, req->r_t.epoch, req->r_flags,
2365              req->r_attempts);
2366
2367         req->r_t.paused = false;
2368         req->r_stamp = jiffies;
2369         req->r_attempts++;
2370
2371         req->r_sent = osd->o_incarnation;
2372         req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
2373         ceph_con_send(&osd->o_con, ceph_msg_get(req->r_request));
2374 }
2375
2376 static void maybe_request_map(struct ceph_osd_client *osdc)
2377 {
2378         bool continuous = false;
2379
2380         verify_osdc_locked(osdc);
2381         WARN_ON(!osdc->osdmap->epoch);
2382
2383         if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2384             ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD) ||
2385             ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
2386                 dout("%s osdc %p continuous\n", __func__, osdc);
2387                 continuous = true;
2388         } else {
2389                 dout("%s osdc %p onetime\n", __func__, osdc);
2390         }
2391
2392         if (ceph_monc_want_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
2393                                osdc->osdmap->epoch + 1, continuous))
2394                 ceph_monc_renew_subs(&osdc->client->monc);
2395 }
2396
2397 static void complete_request(struct ceph_osd_request *req, int err);
2398 static void send_map_check(struct ceph_osd_request *req);
2399
2400 static void __submit_request(struct ceph_osd_request *req, bool wrlocked)
2401 {
2402         struct ceph_osd_client *osdc = req->r_osdc;
2403         struct ceph_osd *osd;
2404         enum calc_target_result ct_res;
2405         int err = 0;
2406         bool need_send = false;
2407         bool promoted = false;
2408
2409         WARN_ON(req->r_tid);
2410         dout("%s req %p wrlocked %d\n", __func__, req, wrlocked);
2411
2412 again:
2413         ct_res = calc_target(osdc, &req->r_t, false);
2414         if (ct_res == CALC_TARGET_POOL_DNE && !wrlocked)
2415                 goto promote;
2416
2417         osd = lookup_create_osd(osdc, req->r_t.osd, wrlocked);
2418         if (IS_ERR(osd)) {
2419                 WARN_ON(PTR_ERR(osd) != -EAGAIN || wrlocked);
2420                 goto promote;
2421         }
2422
2423         if (osdc->abort_err) {
2424                 dout("req %p abort_err %d\n", req, osdc->abort_err);
2425                 err = osdc->abort_err;
2426         } else if (osdc->osdmap->epoch < osdc->epoch_barrier) {
2427                 dout("req %p epoch %u barrier %u\n", req, osdc->osdmap->epoch,
2428                      osdc->epoch_barrier);
2429                 req->r_t.paused = true;
2430                 maybe_request_map(osdc);
2431         } else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2432                    ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
2433                 dout("req %p pausewr\n", req);
2434                 req->r_t.paused = true;
2435                 maybe_request_map(osdc);
2436         } else if ((req->r_flags & CEPH_OSD_FLAG_READ) &&
2437                    ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
2438                 dout("req %p pauserd\n", req);
2439                 req->r_t.paused = true;
2440                 maybe_request_map(osdc);
2441         } else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2442                    !(req->r_flags & (CEPH_OSD_FLAG_FULL_TRY |
2443                                      CEPH_OSD_FLAG_FULL_FORCE)) &&
2444                    (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2445                     pool_full(osdc, req->r_t.base_oloc.pool))) {
2446                 dout("req %p full/pool_full\n", req);
2447                 if (ceph_test_opt(osdc->client, ABORT_ON_FULL)) {
2448                         err = -ENOSPC;
2449                 } else {
2450                         if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL))
2451                                 pr_warn_ratelimited("cluster is full (osdmap FULL)\n");
2452                         else
2453                                 pr_warn_ratelimited("pool %lld is full or reached quota\n",
2454                                                     req->r_t.base_oloc.pool);
2455                         req->r_t.paused = true;
2456                         maybe_request_map(osdc);
2457                 }
2458         } else if (!osd_homeless(osd)) {
2459                 need_send = true;
2460         } else {
2461                 maybe_request_map(osdc);
2462         }
2463
2464         mutex_lock(&osd->lock);
2465         /*
2466          * Assign the tid atomically with send_request() to protect
2467          * multiple writes to the same object from racing with each
2468          * other, resulting in out of order ops on the OSDs.
2469          */
2470         req->r_tid = atomic64_inc_return(&osdc->last_tid);
2471         link_request(osd, req);
2472         if (need_send)
2473                 send_request(req);
2474         else if (err)
2475                 complete_request(req, err);
2476         mutex_unlock(&osd->lock);
2477
2478         if (!err && ct_res == CALC_TARGET_POOL_DNE)
2479                 send_map_check(req);
2480
2481         if (promoted)
2482                 downgrade_write(&osdc->lock);
2483         return;
2484
2485 promote:
2486         up_read(&osdc->lock);
2487         down_write(&osdc->lock);
2488         wrlocked = true;
2489         promoted = true;
2490         goto again;
2491 }
2492
2493 static void account_request(struct ceph_osd_request *req)
2494 {
2495         WARN_ON(req->r_flags & (CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK));
2496         WARN_ON(!(req->r_flags & (CEPH_OSD_FLAG_READ | CEPH_OSD_FLAG_WRITE)));
2497
2498         req->r_flags |= CEPH_OSD_FLAG_ONDISK;
2499         atomic_inc(&req->r_osdc->num_requests);
2500
2501         req->r_start_stamp = jiffies;
2502         req->r_start_latency = ktime_get();
2503 }
2504
2505 static void submit_request(struct ceph_osd_request *req, bool wrlocked)
2506 {
2507         ceph_osdc_get_request(req);
2508         account_request(req);
2509         __submit_request(req, wrlocked);
2510 }
2511
2512 static void finish_request(struct ceph_osd_request *req)
2513 {
2514         struct ceph_osd_client *osdc = req->r_osdc;
2515
2516         WARN_ON(lookup_request_mc(&osdc->map_checks, req->r_tid));
2517         dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2518
2519         req->r_end_latency = ktime_get();
2520
2521         if (req->r_osd) {
2522                 ceph_init_sparse_read(&req->r_osd->o_sparse_read);
2523                 unlink_request(req->r_osd, req);
2524         }
2525         atomic_dec(&osdc->num_requests);
2526
2527         /*
2528          * If an OSD has failed or returned and a request has been sent
2529          * twice, it's possible to get a reply and end up here while the
2530          * request message is queued for delivery.  We will ignore the
2531          * reply, so not a big deal, but better to try and catch it.
2532          */
2533         ceph_msg_revoke(req->r_request);
2534         ceph_msg_revoke_incoming(req->r_reply);
2535 }
2536
2537 static void __complete_request(struct ceph_osd_request *req)
2538 {
2539         dout("%s req %p tid %llu cb %ps result %d\n", __func__, req,
2540              req->r_tid, req->r_callback, req->r_result);
2541
2542         if (req->r_callback)
2543                 req->r_callback(req);
2544         complete_all(&req->r_completion);
2545         ceph_osdc_put_request(req);
2546 }
2547
2548 static void complete_request_workfn(struct work_struct *work)
2549 {
2550         struct ceph_osd_request *req =
2551             container_of(work, struct ceph_osd_request, r_complete_work);
2552
2553         __complete_request(req);
2554 }
2555
2556 /*
2557  * This is open-coded in handle_reply().
2558  */
2559 static void complete_request(struct ceph_osd_request *req, int err)
2560 {
2561         dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2562
2563         req->r_result = err;
2564         finish_request(req);
2565
2566         INIT_WORK(&req->r_complete_work, complete_request_workfn);
2567         queue_work(req->r_osdc->completion_wq, &req->r_complete_work);
2568 }
2569
2570 static void cancel_map_check(struct ceph_osd_request *req)
2571 {
2572         struct ceph_osd_client *osdc = req->r_osdc;
2573         struct ceph_osd_request *lookup_req;
2574
2575         verify_osdc_wrlocked(osdc);
2576
2577         lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2578         if (!lookup_req)
2579                 return;
2580
2581         WARN_ON(lookup_req != req);
2582         erase_request_mc(&osdc->map_checks, req);
2583         ceph_osdc_put_request(req);
2584 }
2585
2586 static void cancel_request(struct ceph_osd_request *req)
2587 {
2588         dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2589
2590         cancel_map_check(req);
2591         finish_request(req);
2592         complete_all(&req->r_completion);
2593         ceph_osdc_put_request(req);
2594 }
2595
2596 static void abort_request(struct ceph_osd_request *req, int err)
2597 {
2598         dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2599
2600         cancel_map_check(req);
2601         complete_request(req, err);
2602 }
2603
2604 static int abort_fn(struct ceph_osd_request *req, void *arg)
2605 {
2606         int err = *(int *)arg;
2607
2608         abort_request(req, err);
2609         return 0; /* continue iteration */
2610 }
2611
2612 /*
2613  * Abort all in-flight requests with @err and arrange for all future
2614  * requests to be failed immediately.
2615  */
2616 void ceph_osdc_abort_requests(struct ceph_osd_client *osdc, int err)
2617 {
2618         dout("%s osdc %p err %d\n", __func__, osdc, err);
2619         down_write(&osdc->lock);
2620         for_each_request(osdc, abort_fn, &err);
2621         osdc->abort_err = err;
2622         up_write(&osdc->lock);
2623 }
2624 EXPORT_SYMBOL(ceph_osdc_abort_requests);
2625
2626 void ceph_osdc_clear_abort_err(struct ceph_osd_client *osdc)
2627 {
2628         down_write(&osdc->lock);
2629         osdc->abort_err = 0;
2630         up_write(&osdc->lock);
2631 }
2632 EXPORT_SYMBOL(ceph_osdc_clear_abort_err);
2633
2634 static void update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2635 {
2636         if (likely(eb > osdc->epoch_barrier)) {
2637                 dout("updating epoch_barrier from %u to %u\n",
2638                                 osdc->epoch_barrier, eb);
2639                 osdc->epoch_barrier = eb;
2640                 /* Request map if we're not to the barrier yet */
2641                 if (eb > osdc->osdmap->epoch)
2642                         maybe_request_map(osdc);
2643         }
2644 }
2645
2646 void ceph_osdc_update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2647 {
2648         down_read(&osdc->lock);
2649         if (unlikely(eb > osdc->epoch_barrier)) {
2650                 up_read(&osdc->lock);
2651                 down_write(&osdc->lock);
2652                 update_epoch_barrier(osdc, eb);
2653                 up_write(&osdc->lock);
2654         } else {
2655                 up_read(&osdc->lock);
2656         }
2657 }
2658 EXPORT_SYMBOL(ceph_osdc_update_epoch_barrier);
2659
2660 /*
2661  * We can end up releasing caps as a result of abort_request().
2662  * In that case, we probably want to ensure that the cap release message
2663  * has an updated epoch barrier in it, so set the epoch barrier prior to
2664  * aborting the first request.
2665  */
2666 static int abort_on_full_fn(struct ceph_osd_request *req, void *arg)
2667 {
2668         struct ceph_osd_client *osdc = req->r_osdc;
2669         bool *victims = arg;
2670
2671         if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2672             (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2673              pool_full(osdc, req->r_t.base_oloc.pool))) {
2674                 if (!*victims) {
2675                         update_epoch_barrier(osdc, osdc->osdmap->epoch);
2676                         *victims = true;
2677                 }
2678                 abort_request(req, -ENOSPC);
2679         }
2680
2681         return 0; /* continue iteration */
2682 }
2683
2684 /*
2685  * Drop all pending requests that are stalled waiting on a full condition to
2686  * clear, and complete them with ENOSPC as the return code. Set the
2687  * osdc->epoch_barrier to the latest map epoch that we've seen if any were
2688  * cancelled.
2689  */
2690 static void ceph_osdc_abort_on_full(struct ceph_osd_client *osdc)
2691 {
2692         bool victims = false;
2693
2694         if (ceph_test_opt(osdc->client, ABORT_ON_FULL) &&
2695             (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || have_pool_full(osdc)))
2696                 for_each_request(osdc, abort_on_full_fn, &victims);
2697 }
2698
2699 static void check_pool_dne(struct ceph_osd_request *req)
2700 {
2701         struct ceph_osd_client *osdc = req->r_osdc;
2702         struct ceph_osdmap *map = osdc->osdmap;
2703
2704         verify_osdc_wrlocked(osdc);
2705         WARN_ON(!map->epoch);
2706
2707         if (req->r_attempts) {
2708                 /*
2709                  * We sent a request earlier, which means that
2710                  * previously the pool existed, and now it does not
2711                  * (i.e., it was deleted).
2712                  */
2713                 req->r_map_dne_bound = map->epoch;
2714                 dout("%s req %p tid %llu pool disappeared\n", __func__, req,
2715                      req->r_tid);
2716         } else {
2717                 dout("%s req %p tid %llu map_dne_bound %u have %u\n", __func__,
2718                      req, req->r_tid, req->r_map_dne_bound, map->epoch);
2719         }
2720
2721         if (req->r_map_dne_bound) {
2722                 if (map->epoch >= req->r_map_dne_bound) {
2723                         /* we had a new enough map */
2724                         pr_info_ratelimited("tid %llu pool does not exist\n",
2725                                             req->r_tid);
2726                         complete_request(req, -ENOENT);
2727                 }
2728         } else {
2729                 send_map_check(req);
2730         }
2731 }
2732
2733 static void map_check_cb(struct ceph_mon_generic_request *greq)
2734 {
2735         struct ceph_osd_client *osdc = &greq->monc->client->osdc;
2736         struct ceph_osd_request *req;
2737         u64 tid = greq->private_data;
2738
2739         WARN_ON(greq->result || !greq->u.newest);
2740
2741         down_write(&osdc->lock);
2742         req = lookup_request_mc(&osdc->map_checks, tid);
2743         if (!req) {
2744                 dout("%s tid %llu dne\n", __func__, tid);
2745                 goto out_unlock;
2746         }
2747
2748         dout("%s req %p tid %llu map_dne_bound %u newest %llu\n", __func__,
2749              req, req->r_tid, req->r_map_dne_bound, greq->u.newest);
2750         if (!req->r_map_dne_bound)
2751                 req->r_map_dne_bound = greq->u.newest;
2752         erase_request_mc(&osdc->map_checks, req);
2753         check_pool_dne(req);
2754
2755         ceph_osdc_put_request(req);
2756 out_unlock:
2757         up_write(&osdc->lock);
2758 }
2759
2760 static void send_map_check(struct ceph_osd_request *req)
2761 {
2762         struct ceph_osd_client *osdc = req->r_osdc;
2763         struct ceph_osd_request *lookup_req;
2764         int ret;
2765
2766         verify_osdc_wrlocked(osdc);
2767
2768         lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2769         if (lookup_req) {
2770                 WARN_ON(lookup_req != req);
2771                 return;
2772         }
2773
2774         ceph_osdc_get_request(req);
2775         insert_request_mc(&osdc->map_checks, req);
2776         ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
2777                                           map_check_cb, req->r_tid);
2778         WARN_ON(ret);
2779 }
2780
2781 /*
2782  * lingering requests, watch/notify v2 infrastructure
2783  */
2784 static void linger_release(struct kref *kref)
2785 {
2786         struct ceph_osd_linger_request *lreq =
2787             container_of(kref, struct ceph_osd_linger_request, kref);
2788
2789         dout("%s lreq %p reg_req %p ping_req %p\n", __func__, lreq,
2790              lreq->reg_req, lreq->ping_req);
2791         WARN_ON(!RB_EMPTY_NODE(&lreq->node));
2792         WARN_ON(!RB_EMPTY_NODE(&lreq->osdc_node));
2793         WARN_ON(!RB_EMPTY_NODE(&lreq->mc_node));
2794         WARN_ON(!list_empty(&lreq->scan_item));
2795         WARN_ON(!list_empty(&lreq->pending_lworks));
2796         WARN_ON(lreq->osd);
2797
2798         if (lreq->request_pl)
2799                 ceph_pagelist_release(lreq->request_pl);
2800         if (lreq->notify_id_pages)
2801                 ceph_release_page_vector(lreq->notify_id_pages, 1);
2802
2803         ceph_osdc_put_request(lreq->reg_req);
2804         ceph_osdc_put_request(lreq->ping_req);
2805         target_destroy(&lreq->t);
2806         kfree(lreq);
2807 }
2808
2809 static void linger_put(struct ceph_osd_linger_request *lreq)
2810 {
2811         if (lreq)
2812                 kref_put(&lreq->kref, linger_release);
2813 }
2814
2815 static struct ceph_osd_linger_request *
2816 linger_get(struct ceph_osd_linger_request *lreq)
2817 {
2818         kref_get(&lreq->kref);
2819         return lreq;
2820 }
2821
2822 static struct ceph_osd_linger_request *
2823 linger_alloc(struct ceph_osd_client *osdc)
2824 {
2825         struct ceph_osd_linger_request *lreq;
2826
2827         lreq = kzalloc(sizeof(*lreq), GFP_NOIO);
2828         if (!lreq)
2829                 return NULL;
2830
2831         kref_init(&lreq->kref);
2832         mutex_init(&lreq->lock);
2833         RB_CLEAR_NODE(&lreq->node);
2834         RB_CLEAR_NODE(&lreq->osdc_node);
2835         RB_CLEAR_NODE(&lreq->mc_node);
2836         INIT_LIST_HEAD(&lreq->scan_item);
2837         INIT_LIST_HEAD(&lreq->pending_lworks);
2838         init_completion(&lreq->reg_commit_wait);
2839         init_completion(&lreq->notify_finish_wait);
2840
2841         lreq->osdc = osdc;
2842         target_init(&lreq->t);
2843
2844         dout("%s lreq %p\n", __func__, lreq);
2845         return lreq;
2846 }
2847
2848 DEFINE_RB_INSDEL_FUNCS(linger, struct ceph_osd_linger_request, linger_id, node)
2849 DEFINE_RB_FUNCS(linger_osdc, struct ceph_osd_linger_request, linger_id, osdc_node)
2850 DEFINE_RB_FUNCS(linger_mc, struct ceph_osd_linger_request, linger_id, mc_node)
2851
2852 /*
2853  * Create linger request <-> OSD session relation.
2854  *
2855  * @lreq has to be registered, @osd may be homeless.
2856  */
2857 static void link_linger(struct ceph_osd *osd,
2858                         struct ceph_osd_linger_request *lreq)
2859 {
2860         verify_osd_locked(osd);
2861         WARN_ON(!lreq->linger_id || lreq->osd);
2862         dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2863              osd->o_osd, lreq, lreq->linger_id);
2864
2865         if (!osd_homeless(osd))
2866                 __remove_osd_from_lru(osd);
2867         else
2868                 atomic_inc(&osd->o_osdc->num_homeless);
2869
2870         get_osd(osd);
2871         insert_linger(&osd->o_linger_requests, lreq);
2872         lreq->osd = osd;
2873 }
2874
2875 static void unlink_linger(struct ceph_osd *osd,
2876                           struct ceph_osd_linger_request *lreq)
2877 {
2878         verify_osd_locked(osd);
2879         WARN_ON(lreq->osd != osd);
2880         dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2881              osd->o_osd, lreq, lreq->linger_id);
2882
2883         lreq->osd = NULL;
2884         erase_linger(&osd->o_linger_requests, lreq);
2885         put_osd(osd);
2886
2887         if (!osd_homeless(osd))
2888                 maybe_move_osd_to_lru(osd);
2889         else
2890                 atomic_dec(&osd->o_osdc->num_homeless);
2891 }
2892
2893 static bool __linger_registered(struct ceph_osd_linger_request *lreq)
2894 {
2895         verify_osdc_locked(lreq->osdc);
2896
2897         return !RB_EMPTY_NODE(&lreq->osdc_node);
2898 }
2899
2900 static bool linger_registered(struct ceph_osd_linger_request *lreq)
2901 {
2902         struct ceph_osd_client *osdc = lreq->osdc;
2903         bool registered;
2904
2905         down_read(&osdc->lock);
2906         registered = __linger_registered(lreq);
2907         up_read(&osdc->lock);
2908
2909         return registered;
2910 }
2911
2912 static void linger_register(struct ceph_osd_linger_request *lreq)
2913 {
2914         struct ceph_osd_client *osdc = lreq->osdc;
2915
2916         verify_osdc_wrlocked(osdc);
2917         WARN_ON(lreq->linger_id);
2918
2919         linger_get(lreq);
2920         lreq->linger_id = ++osdc->last_linger_id;
2921         insert_linger_osdc(&osdc->linger_requests, lreq);
2922 }
2923
2924 static void linger_unregister(struct ceph_osd_linger_request *lreq)
2925 {
2926         struct ceph_osd_client *osdc = lreq->osdc;
2927
2928         verify_osdc_wrlocked(osdc);
2929
2930         erase_linger_osdc(&osdc->linger_requests, lreq);
2931         linger_put(lreq);
2932 }
2933
2934 static void cancel_linger_request(struct ceph_osd_request *req)
2935 {
2936         struct ceph_osd_linger_request *lreq = req->r_priv;
2937
2938         WARN_ON(!req->r_linger);
2939         cancel_request(req);
2940         linger_put(lreq);
2941 }
2942
2943 struct linger_work {
2944         struct work_struct work;
2945         struct ceph_osd_linger_request *lreq;
2946         struct list_head pending_item;
2947         unsigned long queued_stamp;
2948
2949         union {
2950                 struct {
2951                         u64 notify_id;
2952                         u64 notifier_id;
2953                         void *payload; /* points into @msg front */
2954                         size_t payload_len;
2955
2956                         struct ceph_msg *msg; /* for ceph_msg_put() */
2957                 } notify;
2958                 struct {
2959                         int err;
2960                 } error;
2961         };
2962 };
2963
2964 static struct linger_work *lwork_alloc(struct ceph_osd_linger_request *lreq,
2965                                        work_func_t workfn)
2966 {
2967         struct linger_work *lwork;
2968
2969         lwork = kzalloc(sizeof(*lwork), GFP_NOIO);
2970         if (!lwork)
2971                 return NULL;
2972
2973         INIT_WORK(&lwork->work, workfn);
2974         INIT_LIST_HEAD(&lwork->pending_item);
2975         lwork->lreq = linger_get(lreq);
2976
2977         return lwork;
2978 }
2979
2980 static void lwork_free(struct linger_work *lwork)
2981 {
2982         struct ceph_osd_linger_request *lreq = lwork->lreq;
2983
2984         mutex_lock(&lreq->lock);
2985         list_del(&lwork->pending_item);
2986         mutex_unlock(&lreq->lock);
2987
2988         linger_put(lreq);
2989         kfree(lwork);
2990 }
2991
2992 static void lwork_queue(struct linger_work *lwork)
2993 {
2994         struct ceph_osd_linger_request *lreq = lwork->lreq;
2995         struct ceph_osd_client *osdc = lreq->osdc;
2996
2997         verify_lreq_locked(lreq);
2998         WARN_ON(!list_empty(&lwork->pending_item));
2999
3000         lwork->queued_stamp = jiffies;
3001         list_add_tail(&lwork->pending_item, &lreq->pending_lworks);
3002         queue_work(osdc->notify_wq, &lwork->work);
3003 }
3004
3005 static void do_watch_notify(struct work_struct *w)
3006 {
3007         struct linger_work *lwork = container_of(w, struct linger_work, work);
3008         struct ceph_osd_linger_request *lreq = lwork->lreq;
3009
3010         if (!linger_registered(lreq)) {
3011                 dout("%s lreq %p not registered\n", __func__, lreq);
3012                 goto out;
3013         }
3014
3015         WARN_ON(!lreq->is_watch);
3016         dout("%s lreq %p notify_id %llu notifier_id %llu payload_len %zu\n",
3017              __func__, lreq, lwork->notify.notify_id, lwork->notify.notifier_id,
3018              lwork->notify.payload_len);
3019         lreq->wcb(lreq->data, lwork->notify.notify_id, lreq->linger_id,
3020                   lwork->notify.notifier_id, lwork->notify.payload,
3021                   lwork->notify.payload_len);
3022
3023 out:
3024         ceph_msg_put(lwork->notify.msg);
3025         lwork_free(lwork);
3026 }
3027
3028 static void do_watch_error(struct work_struct *w)
3029 {
3030         struct linger_work *lwork = container_of(w, struct linger_work, work);
3031         struct ceph_osd_linger_request *lreq = lwork->lreq;
3032
3033         if (!linger_registered(lreq)) {
3034                 dout("%s lreq %p not registered\n", __func__, lreq);
3035                 goto out;
3036         }
3037
3038         dout("%s lreq %p err %d\n", __func__, lreq, lwork->error.err);
3039         lreq->errcb(lreq->data, lreq->linger_id, lwork->error.err);
3040
3041 out:
3042         lwork_free(lwork);
3043 }
3044
3045 static void queue_watch_error(struct ceph_osd_linger_request *lreq)
3046 {
3047         struct linger_work *lwork;
3048
3049         lwork = lwork_alloc(lreq, do_watch_error);
3050         if (!lwork) {
3051                 pr_err("failed to allocate error-lwork\n");
3052                 return;
3053         }
3054
3055         lwork->error.err = lreq->last_error;
3056         lwork_queue(lwork);
3057 }
3058
3059 static void linger_reg_commit_complete(struct ceph_osd_linger_request *lreq,
3060                                        int result)
3061 {
3062         if (!completion_done(&lreq->reg_commit_wait)) {
3063                 lreq->reg_commit_error = (result <= 0 ? result : 0);
3064                 complete_all(&lreq->reg_commit_wait);
3065         }
3066 }
3067
3068 static void linger_commit_cb(struct ceph_osd_request *req)
3069 {
3070         struct ceph_osd_linger_request *lreq = req->r_priv;
3071
3072         mutex_lock(&lreq->lock);
3073         if (req != lreq->reg_req) {
3074                 dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n",
3075                      __func__, lreq, lreq->linger_id, req, lreq->reg_req);
3076                 goto out;
3077         }
3078
3079         dout("%s lreq %p linger_id %llu result %d\n", __func__, lreq,
3080              lreq->linger_id, req->r_result);
3081         linger_reg_commit_complete(lreq, req->r_result);
3082         lreq->committed = true;
3083
3084         if (!lreq->is_watch) {
3085                 struct ceph_osd_data *osd_data =
3086                     osd_req_op_data(req, 0, notify, response_data);
3087                 void *p = page_address(osd_data->pages[0]);
3088
3089                 WARN_ON(req->r_ops[0].op != CEPH_OSD_OP_NOTIFY ||
3090                         osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
3091
3092                 /* make note of the notify_id */
3093                 if (req->r_ops[0].outdata_len >= sizeof(u64)) {
3094                         lreq->notify_id = ceph_decode_64(&p);
3095                         dout("lreq %p notify_id %llu\n", lreq,
3096                              lreq->notify_id);
3097                 } else {
3098                         dout("lreq %p no notify_id\n", lreq);
3099                 }
3100         }
3101
3102 out:
3103         mutex_unlock(&lreq->lock);
3104         linger_put(lreq);
3105 }
3106
3107 static int normalize_watch_error(int err)
3108 {
3109         /*
3110          * Translate ENOENT -> ENOTCONN so that a delete->disconnection
3111          * notification and a failure to reconnect because we raced with
3112          * the delete appear the same to the user.
3113          */
3114         if (err == -ENOENT)
3115                 err = -ENOTCONN;
3116
3117         return err;
3118 }
3119
3120 static void linger_reconnect_cb(struct ceph_osd_request *req)
3121 {
3122         struct ceph_osd_linger_request *lreq = req->r_priv;
3123
3124         mutex_lock(&lreq->lock);
3125         if (req != lreq->reg_req) {
3126                 dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n",
3127                      __func__, lreq, lreq->linger_id, req, lreq->reg_req);
3128                 goto out;
3129         }
3130
3131         dout("%s lreq %p linger_id %llu result %d last_error %d\n", __func__,
3132              lreq, lreq->linger_id, req->r_result, lreq->last_error);
3133         if (req->r_result < 0) {
3134                 if (!lreq->last_error) {
3135                         lreq->last_error = normalize_watch_error(req->r_result);
3136                         queue_watch_error(lreq);
3137                 }
3138         }
3139
3140 out:
3141         mutex_unlock(&lreq->lock);
3142         linger_put(lreq);
3143 }
3144
3145 static void send_linger(struct ceph_osd_linger_request *lreq)
3146 {
3147         struct ceph_osd_client *osdc = lreq->osdc;
3148         struct ceph_osd_request *req;
3149         int ret;
3150
3151         verify_osdc_wrlocked(osdc);
3152         mutex_lock(&lreq->lock);
3153         dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3154
3155         if (lreq->reg_req) {
3156                 if (lreq->reg_req->r_osd)
3157                         cancel_linger_request(lreq->reg_req);
3158                 ceph_osdc_put_request(lreq->reg_req);
3159         }
3160
3161         req = ceph_osdc_alloc_request(osdc, NULL, 1, true, GFP_NOIO);
3162         BUG_ON(!req);
3163
3164         target_copy(&req->r_t, &lreq->t);
3165         req->r_mtime = lreq->mtime;
3166
3167         if (lreq->is_watch && lreq->committed) {
3168                 osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_RECONNECT,
3169                                       lreq->linger_id, ++lreq->register_gen);
3170                 dout("lreq %p reconnect register_gen %u\n", lreq,
3171                      req->r_ops[0].watch.gen);
3172                 req->r_callback = linger_reconnect_cb;
3173         } else {
3174                 if (lreq->is_watch) {
3175                         osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_WATCH,
3176                                               lreq->linger_id, 0);
3177                 } else {
3178                         lreq->notify_id = 0;
3179
3180                         refcount_inc(&lreq->request_pl->refcnt);
3181                         osd_req_op_notify_init(req, 0, lreq->linger_id,
3182                                                lreq->request_pl);
3183                         ceph_osd_data_pages_init(
3184                             osd_req_op_data(req, 0, notify, response_data),
3185                             lreq->notify_id_pages, PAGE_SIZE, 0, false, false);
3186                 }
3187                 dout("lreq %p register\n", lreq);
3188                 req->r_callback = linger_commit_cb;
3189         }
3190
3191         ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
3192         BUG_ON(ret);
3193
3194         req->r_priv = linger_get(lreq);
3195         req->r_linger = true;
3196         lreq->reg_req = req;
3197         mutex_unlock(&lreq->lock);
3198
3199         submit_request(req, true);
3200 }
3201
3202 static void linger_ping_cb(struct ceph_osd_request *req)
3203 {
3204         struct ceph_osd_linger_request *lreq = req->r_priv;
3205
3206         mutex_lock(&lreq->lock);
3207         if (req != lreq->ping_req) {
3208                 dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n",
3209                      __func__, lreq, lreq->linger_id, req, lreq->ping_req);
3210                 goto out;
3211         }
3212
3213         dout("%s lreq %p linger_id %llu result %d ping_sent %lu last_error %d\n",
3214              __func__, lreq, lreq->linger_id, req->r_result, lreq->ping_sent,
3215              lreq->last_error);
3216         if (lreq->register_gen == req->r_ops[0].watch.gen) {
3217                 if (!req->r_result) {
3218                         lreq->watch_valid_thru = lreq->ping_sent;
3219                 } else if (!lreq->last_error) {
3220                         lreq->last_error = normalize_watch_error(req->r_result);
3221                         queue_watch_error(lreq);
3222                 }
3223         } else {
3224                 dout("lreq %p register_gen %u ignoring old pong %u\n", lreq,
3225                      lreq->register_gen, req->r_ops[0].watch.gen);
3226         }
3227
3228 out:
3229         mutex_unlock(&lreq->lock);
3230         linger_put(lreq);
3231 }
3232
3233 static void send_linger_ping(struct ceph_osd_linger_request *lreq)
3234 {
3235         struct ceph_osd_client *osdc = lreq->osdc;
3236         struct ceph_osd_request *req;
3237         int ret;
3238
3239         if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
3240                 dout("%s PAUSERD\n", __func__);
3241                 return;
3242         }
3243
3244         lreq->ping_sent = jiffies;
3245         dout("%s lreq %p linger_id %llu ping_sent %lu register_gen %u\n",
3246              __func__, lreq, lreq->linger_id, lreq->ping_sent,
3247              lreq->register_gen);
3248
3249         if (lreq->ping_req) {
3250                 if (lreq->ping_req->r_osd)
3251                         cancel_linger_request(lreq->ping_req);
3252                 ceph_osdc_put_request(lreq->ping_req);
3253         }
3254
3255         req = ceph_osdc_alloc_request(osdc, NULL, 1, true, GFP_NOIO);
3256         BUG_ON(!req);
3257
3258         target_copy(&req->r_t, &lreq->t);
3259         osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_PING, lreq->linger_id,
3260                               lreq->register_gen);
3261         req->r_callback = linger_ping_cb;
3262
3263         ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
3264         BUG_ON(ret);
3265
3266         req->r_priv = linger_get(lreq);
3267         req->r_linger = true;
3268         lreq->ping_req = req;
3269
3270         ceph_osdc_get_request(req);
3271         account_request(req);
3272         req->r_tid = atomic64_inc_return(&osdc->last_tid);
3273         link_request(lreq->osd, req);
3274         send_request(req);
3275 }
3276
3277 static void linger_submit(struct ceph_osd_linger_request *lreq)
3278 {
3279         struct ceph_osd_client *osdc = lreq->osdc;
3280         struct ceph_osd *osd;
3281
3282         down_write(&osdc->lock);
3283         linger_register(lreq);
3284
3285         calc_target(osdc, &lreq->t, false);
3286         osd = lookup_create_osd(osdc, lreq->t.osd, true);
3287         link_linger(osd, lreq);
3288
3289         send_linger(lreq);
3290         up_write(&osdc->lock);
3291 }
3292
3293 static void cancel_linger_map_check(struct ceph_osd_linger_request *lreq)
3294 {
3295         struct ceph_osd_client *osdc = lreq->osdc;
3296         struct ceph_osd_linger_request *lookup_lreq;
3297
3298         verify_osdc_wrlocked(osdc);
3299
3300         lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3301                                        lreq->linger_id);
3302         if (!lookup_lreq)
3303                 return;
3304
3305         WARN_ON(lookup_lreq != lreq);
3306         erase_linger_mc(&osdc->linger_map_checks, lreq);
3307         linger_put(lreq);
3308 }
3309
3310 /*
3311  * @lreq has to be both registered and linked.
3312  */
3313 static void __linger_cancel(struct ceph_osd_linger_request *lreq)
3314 {
3315         if (lreq->ping_req && lreq->ping_req->r_osd)
3316                 cancel_linger_request(lreq->ping_req);
3317         if (lreq->reg_req && lreq->reg_req->r_osd)
3318                 cancel_linger_request(lreq->reg_req);
3319         cancel_linger_map_check(lreq);
3320         unlink_linger(lreq->osd, lreq);
3321         linger_unregister(lreq);
3322 }
3323
3324 static void linger_cancel(struct ceph_osd_linger_request *lreq)
3325 {
3326         struct ceph_osd_client *osdc = lreq->osdc;
3327
3328         down_write(&osdc->lock);
3329         if (__linger_registered(lreq))
3330                 __linger_cancel(lreq);
3331         up_write(&osdc->lock);
3332 }
3333
3334 static void send_linger_map_check(struct ceph_osd_linger_request *lreq);
3335
3336 static void check_linger_pool_dne(struct ceph_osd_linger_request *lreq)
3337 {
3338         struct ceph_osd_client *osdc = lreq->osdc;
3339         struct ceph_osdmap *map = osdc->osdmap;
3340
3341         verify_osdc_wrlocked(osdc);
3342         WARN_ON(!map->epoch);
3343
3344         if (lreq->register_gen) {
3345                 lreq->map_dne_bound = map->epoch;
3346                 dout("%s lreq %p linger_id %llu pool disappeared\n", __func__,
3347                      lreq, lreq->linger_id);
3348         } else {
3349                 dout("%s lreq %p linger_id %llu map_dne_bound %u have %u\n",
3350                      __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3351                      map->epoch);
3352         }
3353
3354         if (lreq->map_dne_bound) {
3355                 if (map->epoch >= lreq->map_dne_bound) {
3356                         /* we had a new enough map */
3357                         pr_info("linger_id %llu pool does not exist\n",
3358                                 lreq->linger_id);
3359                         linger_reg_commit_complete(lreq, -ENOENT);
3360                         __linger_cancel(lreq);
3361                 }
3362         } else {
3363                 send_linger_map_check(lreq);
3364         }
3365 }
3366
3367 static void linger_map_check_cb(struct ceph_mon_generic_request *greq)
3368 {
3369         struct ceph_osd_client *osdc = &greq->monc->client->osdc;
3370         struct ceph_osd_linger_request *lreq;
3371         u64 linger_id = greq->private_data;
3372
3373         WARN_ON(greq->result || !greq->u.newest);
3374
3375         down_write(&osdc->lock);
3376         lreq = lookup_linger_mc(&osdc->linger_map_checks, linger_id);
3377         if (!lreq) {
3378                 dout("%s linger_id %llu dne\n", __func__, linger_id);
3379                 goto out_unlock;
3380         }
3381
3382         dout("%s lreq %p linger_id %llu map_dne_bound %u newest %llu\n",
3383              __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3384              greq->u.newest);
3385         if (!lreq->map_dne_bound)
3386                 lreq->map_dne_bound = greq->u.newest;
3387         erase_linger_mc(&osdc->linger_map_checks, lreq);
3388         check_linger_pool_dne(lreq);
3389
3390         linger_put(lreq);
3391 out_unlock:
3392         up_write(&osdc->lock);
3393 }
3394
3395 static void send_linger_map_check(struct ceph_osd_linger_request *lreq)
3396 {
3397         struct ceph_osd_client *osdc = lreq->osdc;
3398         struct ceph_osd_linger_request *lookup_lreq;
3399         int ret;
3400
3401         verify_osdc_wrlocked(osdc);
3402
3403         lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3404                                        lreq->linger_id);
3405         if (lookup_lreq) {
3406                 WARN_ON(lookup_lreq != lreq);
3407                 return;
3408         }
3409
3410         linger_get(lreq);
3411         insert_linger_mc(&osdc->linger_map_checks, lreq);
3412         ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
3413                                           linger_map_check_cb, lreq->linger_id);
3414         WARN_ON(ret);
3415 }
3416
3417 static int linger_reg_commit_wait(struct ceph_osd_linger_request *lreq)
3418 {
3419         int ret;
3420
3421         dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3422         ret = wait_for_completion_killable(&lreq->reg_commit_wait);
3423         return ret ?: lreq->reg_commit_error;
3424 }
3425
3426 static int linger_notify_finish_wait(struct ceph_osd_linger_request *lreq,
3427                                      unsigned long timeout)
3428 {
3429         long left;
3430
3431         dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3432         left = wait_for_completion_killable_timeout(&lreq->notify_finish_wait,
3433                                                 ceph_timeout_jiffies(timeout));
3434         if (left <= 0)
3435                 left = left ?: -ETIMEDOUT;
3436         else
3437                 left = lreq->notify_finish_error; /* completed */
3438
3439         return left;
3440 }
3441
3442 /*
3443  * Timeout callback, called every N seconds.  When 1 or more OSD
3444  * requests has been active for more than N seconds, we send a keepalive
3445  * (tag + timestamp) to its OSD to ensure any communications channel
3446  * reset is detected.
3447  */
3448 static void handle_timeout(struct work_struct *work)
3449 {
3450         struct ceph_osd_client *osdc =
3451                 container_of(work, struct ceph_osd_client, timeout_work.work);
3452         struct ceph_options *opts = osdc->client->options;
3453         unsigned long cutoff = jiffies - opts->osd_keepalive_timeout;
3454         unsigned long expiry_cutoff = jiffies - opts->osd_request_timeout;
3455         LIST_HEAD(slow_osds);
3456         struct rb_node *n, *p;
3457
3458         dout("%s osdc %p\n", __func__, osdc);
3459         down_write(&osdc->lock);
3460
3461         /*
3462          * ping osds that are a bit slow.  this ensures that if there
3463          * is a break in the TCP connection we will notice, and reopen
3464          * a connection with that osd (from the fault callback).
3465          */
3466         for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
3467                 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
3468                 bool found = false;
3469
3470                 for (p = rb_first(&osd->o_requests); p; ) {
3471                         struct ceph_osd_request *req =
3472                             rb_entry(p, struct ceph_osd_request, r_node);
3473
3474                         p = rb_next(p); /* abort_request() */
3475
3476                         if (time_before(req->r_stamp, cutoff)) {
3477                                 dout(" req %p tid %llu on osd%d is laggy\n",
3478                                      req, req->r_tid, osd->o_osd);
3479                                 found = true;
3480                         }
3481                         if (opts->osd_request_timeout &&
3482                             time_before(req->r_start_stamp, expiry_cutoff)) {
3483                                 pr_err_ratelimited("tid %llu on osd%d timeout\n",
3484                                        req->r_tid, osd->o_osd);
3485                                 abort_request(req, -ETIMEDOUT);
3486                         }
3487                 }
3488                 for (p = rb_first(&osd->o_linger_requests); p; p = rb_next(p)) {
3489                         struct ceph_osd_linger_request *lreq =
3490                             rb_entry(p, struct ceph_osd_linger_request, node);
3491
3492                         dout(" lreq %p linger_id %llu is served by osd%d\n",
3493                              lreq, lreq->linger_id, osd->o_osd);
3494                         found = true;
3495
3496                         mutex_lock(&lreq->lock);
3497                         if (lreq->is_watch && lreq->committed && !lreq->last_error)
3498                                 send_linger_ping(lreq);
3499                         mutex_unlock(&lreq->lock);
3500                 }
3501
3502                 if (found)
3503                         list_move_tail(&osd->o_keepalive_item, &slow_osds);
3504         }
3505
3506         if (opts->osd_request_timeout) {
3507                 for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
3508                         struct ceph_osd_request *req =
3509                             rb_entry(p, struct ceph_osd_request, r_node);
3510
3511                         p = rb_next(p); /* abort_request() */
3512
3513                         if (time_before(req->r_start_stamp, expiry_cutoff)) {
3514                                 pr_err_ratelimited("tid %llu on osd%d timeout\n",
3515                                        req->r_tid, osdc->homeless_osd.o_osd);
3516                                 abort_request(req, -ETIMEDOUT);
3517                         }
3518                 }
3519         }
3520
3521         if (atomic_read(&osdc->num_homeless) || !list_empty(&slow_osds))
3522                 maybe_request_map(osdc);
3523
3524         while (!list_empty(&slow_osds)) {
3525                 struct ceph_osd *osd = list_first_entry(&slow_osds,
3526                                                         struct ceph_osd,
3527                                                         o_keepalive_item);
3528                 list_del_init(&osd->o_keepalive_item);
3529                 ceph_con_keepalive(&osd->o_con);
3530         }
3531
3532         up_write(&osdc->lock);
3533         schedule_delayed_work(&osdc->timeout_work,
3534                               osdc->client->options->osd_keepalive_timeout);
3535 }
3536
3537 static void handle_osds_timeout(struct work_struct *work)
3538 {
3539         struct ceph_osd_client *osdc =
3540                 container_of(work, struct ceph_osd_client,
3541                              osds_timeout_work.work);
3542         unsigned long delay = osdc->client->options->osd_idle_ttl / 4;
3543         struct ceph_osd *osd, *nosd;
3544
3545         dout("%s osdc %p\n", __func__, osdc);
3546         down_write(&osdc->lock);
3547         list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) {
3548                 if (time_before(jiffies, osd->lru_ttl))
3549                         break;
3550
3551                 WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
3552                 WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
3553                 close_osd(osd);
3554         }
3555
3556         up_write(&osdc->lock);
3557         schedule_delayed_work(&osdc->osds_timeout_work,
3558                               round_jiffies_relative(delay));
3559 }
3560
3561 static int ceph_oloc_decode(void **p, void *end,
3562                             struct ceph_object_locator *oloc)
3563 {
3564         u8 struct_v, struct_cv;
3565         u32 len;
3566         void *struct_end;
3567         int ret = 0;
3568
3569         ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3570         struct_v = ceph_decode_8(p);
3571         struct_cv = ceph_decode_8(p);
3572         if (struct_v < 3) {
3573                 pr_warn("got v %d < 3 cv %d of ceph_object_locator\n",
3574                         struct_v, struct_cv);
3575                 goto e_inval;
3576         }
3577         if (struct_cv > 6) {
3578                 pr_warn("got v %d cv %d > 6 of ceph_object_locator\n",
3579                         struct_v, struct_cv);
3580                 goto e_inval;
3581         }
3582         len = ceph_decode_32(p);
3583         ceph_decode_need(p, end, len, e_inval);
3584         struct_end = *p + len;
3585
3586         oloc->pool = ceph_decode_64(p);
3587         *p += 4; /* skip preferred */
3588
3589         len = ceph_decode_32(p);
3590         if (len > 0) {
3591                 pr_warn("ceph_object_locator::key is set\n");
3592                 goto e_inval;
3593         }
3594
3595         if (struct_v >= 5) {
3596                 bool changed = false;
3597
3598                 len = ceph_decode_32(p);
3599                 if (len > 0) {
3600                         ceph_decode_need(p, end, len, e_inval);
3601                         if (!oloc->pool_ns ||
3602                             ceph_compare_string(oloc->pool_ns, *p, len))
3603                                 changed = true;
3604                         *p += len;
3605                 } else {
3606                         if (oloc->pool_ns)
3607                                 changed = true;
3608                 }
3609                 if (changed) {
3610                         /* redirect changes namespace */
3611                         pr_warn("ceph_object_locator::nspace is changed\n");
3612                         goto e_inval;
3613                 }
3614         }
3615
3616         if (struct_v >= 6) {
3617                 s64 hash = ceph_decode_64(p);
3618                 if (hash != -1) {
3619                         pr_warn("ceph_object_locator::hash is set\n");
3620                         goto e_inval;
3621                 }
3622         }
3623
3624         /* skip the rest */
3625         *p = struct_end;
3626 out:
3627         return ret;
3628
3629 e_inval:
3630         ret = -EINVAL;
3631         goto out;
3632 }
3633
3634 static int ceph_redirect_decode(void **p, void *end,
3635                                 struct ceph_request_redirect *redir)
3636 {
3637         u8 struct_v, struct_cv;
3638         u32 len;
3639         void *struct_end;
3640         int ret;
3641
3642         ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3643         struct_v = ceph_decode_8(p);
3644         struct_cv = ceph_decode_8(p);
3645         if (struct_cv > 1) {
3646                 pr_warn("got v %d cv %d > 1 of ceph_request_redirect\n",
3647                         struct_v, struct_cv);
3648                 goto e_inval;
3649         }
3650         len = ceph_decode_32(p);
3651         ceph_decode_need(p, end, len, e_inval);
3652         struct_end = *p + len;
3653
3654         ret = ceph_oloc_decode(p, end, &redir->oloc);
3655         if (ret)
3656                 goto out;
3657
3658         len = ceph_decode_32(p);
3659         if (len > 0) {
3660                 pr_warn("ceph_request_redirect::object_name is set\n");
3661                 goto e_inval;
3662         }
3663
3664         /* skip the rest */
3665         *p = struct_end;
3666 out:
3667         return ret;
3668
3669 e_inval:
3670         ret = -EINVAL;
3671         goto out;
3672 }
3673
3674 struct MOSDOpReply {
3675         struct ceph_pg pgid;
3676         u64 flags;
3677         int result;
3678         u32 epoch;
3679         int num_ops;
3680         u32 outdata_len[CEPH_OSD_MAX_OPS];
3681         s32 rval[CEPH_OSD_MAX_OPS];
3682         int retry_attempt;
3683         struct ceph_eversion replay_version;
3684         u64 user_version;
3685         struct ceph_request_redirect redirect;
3686 };
3687
3688 static int decode_MOSDOpReply(const struct ceph_msg *msg, struct MOSDOpReply *m)
3689 {
3690         void *p = msg->front.iov_base;
3691         void *const end = p + msg->front.iov_len;
3692         u16 version = le16_to_cpu(msg->hdr.version);
3693         struct ceph_eversion bad_replay_version;
3694         u8 decode_redir;
3695         u32 len;
3696         int ret;
3697         int i;
3698
3699         ceph_decode_32_safe(&p, end, len, e_inval);
3700         ceph_decode_need(&p, end, len, e_inval);
3701         p += len; /* skip oid */
3702
3703         ret = ceph_decode_pgid(&p, end, &m->pgid);
3704         if (ret)
3705                 return ret;
3706
3707         ceph_decode_64_safe(&p, end, m->flags, e_inval);
3708         ceph_decode_32_safe(&p, end, m->result, e_inval);
3709         ceph_decode_need(&p, end, sizeof(bad_replay_version), e_inval);
3710         memcpy(&bad_replay_version, p, sizeof(bad_replay_version));
3711         p += sizeof(bad_replay_version);
3712         ceph_decode_32_safe(&p, end, m->epoch, e_inval);
3713
3714         ceph_decode_32_safe(&p, end, m->num_ops, e_inval);
3715         if (m->num_ops > ARRAY_SIZE(m->outdata_len))
3716                 goto e_inval;
3717
3718         ceph_decode_need(&p, end, m->num_ops * sizeof(struct ceph_osd_op),
3719                          e_inval);
3720         for (i = 0; i < m->num_ops; i++) {
3721                 struct ceph_osd_op *op = p;
3722
3723                 m->outdata_len[i] = le32_to_cpu(op->payload_len);
3724                 p += sizeof(*op);
3725         }
3726
3727         ceph_decode_32_safe(&p, end, m->retry_attempt, e_inval);
3728         for (i = 0; i < m->num_ops; i++)
3729                 ceph_decode_32_safe(&p, end, m->rval[i], e_inval);
3730
3731         if (version >= 5) {
3732                 ceph_decode_need(&p, end, sizeof(m->replay_version), e_inval);
3733                 memcpy(&m->replay_version, p, sizeof(m->replay_version));
3734                 p += sizeof(m->replay_version);
3735                 ceph_decode_64_safe(&p, end, m->user_version, e_inval);
3736         } else {
3737                 m->replay_version = bad_replay_version; /* struct */
3738                 m->user_version = le64_to_cpu(m->replay_version.version);
3739         }
3740
3741         if (version >= 6) {
3742                 if (version >= 7)
3743                         ceph_decode_8_safe(&p, end, decode_redir, e_inval);
3744                 else
3745                         decode_redir = 1;
3746         } else {
3747                 decode_redir = 0;
3748         }
3749
3750         if (decode_redir) {
3751                 ret = ceph_redirect_decode(&p, end, &m->redirect);
3752                 if (ret)
3753                         return ret;
3754         } else {
3755                 ceph_oloc_init(&m->redirect.oloc);
3756         }
3757
3758         return 0;
3759
3760 e_inval:
3761         return -EINVAL;
3762 }
3763
3764 /*
3765  * Handle MOSDOpReply.  Set ->r_result and call the callback if it is
3766  * specified.
3767  */
3768 static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg)
3769 {
3770         struct ceph_osd_client *osdc = osd->o_osdc;
3771         struct ceph_osd_request *req;
3772         struct MOSDOpReply m;
3773         u64 tid = le64_to_cpu(msg->hdr.tid);
3774         u32 data_len = 0;
3775         int ret;
3776         int i;
3777
3778         dout("%s msg %p tid %llu\n", __func__, msg, tid);
3779
3780         down_read(&osdc->lock);
3781         if (!osd_registered(osd)) {
3782                 dout("%s osd%d unknown\n", __func__, osd->o_osd);
3783                 goto out_unlock_osdc;
3784         }
3785         WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
3786
3787         mutex_lock(&osd->lock);
3788         req = lookup_request(&osd->o_requests, tid);
3789         if (!req) {
3790                 dout("%s osd%d tid %llu unknown\n", __func__, osd->o_osd, tid);
3791                 goto out_unlock_session;
3792         }
3793
3794         m.redirect.oloc.pool_ns = req->r_t.target_oloc.pool_ns;
3795         ret = decode_MOSDOpReply(msg, &m);
3796         m.redirect.oloc.pool_ns = NULL;
3797         if (ret) {
3798                 pr_err("failed to decode MOSDOpReply for tid %llu: %d\n",
3799                        req->r_tid, ret);
3800                 ceph_msg_dump(msg);
3801                 goto fail_request;
3802         }
3803         dout("%s req %p tid %llu flags 0x%llx pgid %llu.%x epoch %u attempt %d v %u'%llu uv %llu\n",
3804              __func__, req, req->r_tid, m.flags, m.pgid.pool, m.pgid.seed,
3805              m.epoch, m.retry_attempt, le32_to_cpu(m.replay_version.epoch),
3806              le64_to_cpu(m.replay_version.version), m.user_version);
3807
3808         if (m.retry_attempt >= 0) {
3809                 if (m.retry_attempt != req->r_attempts - 1) {
3810                         dout("req %p tid %llu retry_attempt %d != %d, ignoring\n",
3811                              req, req->r_tid, m.retry_attempt,
3812                              req->r_attempts - 1);
3813                         goto out_unlock_session;
3814                 }
3815         } else {
3816                 WARN_ON(1); /* MOSDOpReply v4 is assumed */
3817         }
3818
3819         if (!ceph_oloc_empty(&m.redirect.oloc)) {
3820                 dout("req %p tid %llu redirect pool %lld\n", req, req->r_tid,
3821                      m.redirect.oloc.pool);
3822                 unlink_request(osd, req);
3823                 mutex_unlock(&osd->lock);
3824
3825                 /*
3826                  * Not ceph_oloc_copy() - changing pool_ns is not
3827                  * supported.
3828                  */
3829                 req->r_t.target_oloc.pool = m.redirect.oloc.pool;
3830                 req->r_flags |= CEPH_OSD_FLAG_REDIRECTED |
3831                                 CEPH_OSD_FLAG_IGNORE_OVERLAY |
3832                                 CEPH_OSD_FLAG_IGNORE_CACHE;
3833                 req->r_tid = 0;
3834                 __submit_request(req, false);
3835                 goto out_unlock_osdc;
3836         }
3837
3838         if (m.result == -EAGAIN) {
3839                 dout("req %p tid %llu EAGAIN\n", req, req->r_tid);
3840                 unlink_request(osd, req);
3841                 mutex_unlock(&osd->lock);
3842
3843                 /*
3844                  * The object is missing on the replica or not (yet)
3845                  * readable.  Clear pgid to force a resend to the primary
3846                  * via legacy_change.
3847                  */
3848                 req->r_t.pgid.pool = 0;
3849                 req->r_t.pgid.seed = 0;
3850                 WARN_ON(!req->r_t.used_replica);
3851                 req->r_flags &= ~(CEPH_OSD_FLAG_BALANCE_READS |
3852                                   CEPH_OSD_FLAG_LOCALIZE_READS);
3853                 req->r_tid = 0;
3854                 __submit_request(req, false);
3855                 goto out_unlock_osdc;
3856         }
3857
3858         if (m.num_ops != req->r_num_ops) {
3859                 pr_err("num_ops %d != %d for tid %llu\n", m.num_ops,
3860                        req->r_num_ops, req->r_tid);
3861                 goto fail_request;
3862         }
3863         for (i = 0; i < req->r_num_ops; i++) {
3864                 dout(" req %p tid %llu op %d rval %d len %u\n", req,
3865                      req->r_tid, i, m.rval[i], m.outdata_len[i]);
3866                 req->r_ops[i].rval = m.rval[i];
3867                 req->r_ops[i].outdata_len = m.outdata_len[i];
3868                 data_len += m.outdata_len[i];
3869         }
3870         if (data_len != le32_to_cpu(msg->hdr.data_len)) {
3871                 pr_err("sum of lens %u != %u for tid %llu\n", data_len,
3872                        le32_to_cpu(msg->hdr.data_len), req->r_tid);
3873                 goto fail_request;
3874         }
3875         dout("%s req %p tid %llu result %d data_len %u\n", __func__,
3876              req, req->r_tid, m.result, data_len);
3877
3878         /*
3879          * Since we only ever request ONDISK, we should only ever get
3880          * one (type of) reply back.
3881          */
3882         WARN_ON(!(m.flags & CEPH_OSD_FLAG_ONDISK));
3883         req->r_version = m.user_version;
3884         req->r_result = m.result ?: data_len;
3885         finish_request(req);
3886         mutex_unlock(&osd->lock);
3887         up_read(&osdc->lock);
3888
3889         __complete_request(req);
3890         return;
3891
3892 fail_request:
3893         complete_request(req, -EIO);
3894 out_unlock_session:
3895         mutex_unlock(&osd->lock);
3896 out_unlock_osdc:
3897         up_read(&osdc->lock);
3898 }
3899
3900 static void set_pool_was_full(struct ceph_osd_client *osdc)
3901 {
3902         struct rb_node *n;
3903
3904         for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
3905                 struct ceph_pg_pool_info *pi =
3906                     rb_entry(n, struct ceph_pg_pool_info, node);
3907
3908                 pi->was_full = __pool_full(pi);
3909         }
3910 }
3911
3912 static bool pool_cleared_full(struct ceph_osd_client *osdc, s64 pool_id)
3913 {
3914         struct ceph_pg_pool_info *pi;
3915
3916         pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
3917         if (!pi)
3918                 return false;
3919
3920         return pi->was_full && !__pool_full(pi);
3921 }
3922
3923 static enum calc_target_result
3924 recalc_linger_target(struct ceph_osd_linger_request *lreq)
3925 {
3926         struct ceph_osd_client *osdc = lreq->osdc;
3927         enum calc_target_result ct_res;
3928
3929         ct_res = calc_target(osdc, &lreq->t, true);
3930         if (ct_res == CALC_TARGET_NEED_RESEND) {
3931                 struct ceph_osd *osd;
3932
3933                 osd = lookup_create_osd(osdc, lreq->t.osd, true);
3934                 if (osd != lreq->osd) {
3935                         unlink_linger(lreq->osd, lreq);
3936                         link_linger(osd, lreq);
3937                 }
3938         }
3939
3940         return ct_res;
3941 }
3942
3943 /*
3944  * Requeue requests whose mapping to an OSD has changed.
3945  */
3946 static void scan_requests(struct ceph_osd *osd,
3947                           bool force_resend,
3948                           bool cleared_full,
3949                           bool check_pool_cleared_full,
3950                           struct rb_root *need_resend,
3951                           struct list_head *need_resend_linger)
3952 {
3953         struct ceph_osd_client *osdc = osd->o_osdc;
3954         struct rb_node *n;
3955         bool force_resend_writes;
3956
3957         for (n = rb_first(&osd->o_linger_requests); n; ) {
3958                 struct ceph_osd_linger_request *lreq =
3959                     rb_entry(n, struct ceph_osd_linger_request, node);
3960                 enum calc_target_result ct_res;
3961
3962                 n = rb_next(n); /* recalc_linger_target() */
3963
3964                 dout("%s lreq %p linger_id %llu\n", __func__, lreq,
3965                      lreq->linger_id);
3966                 ct_res = recalc_linger_target(lreq);
3967                 switch (ct_res) {
3968                 case CALC_TARGET_NO_ACTION:
3969                         force_resend_writes = cleared_full ||
3970                             (check_pool_cleared_full &&
3971                              pool_cleared_full(osdc, lreq->t.base_oloc.pool));
3972                         if (!force_resend && !force_resend_writes)
3973                                 break;
3974
3975                         fallthrough;
3976                 case CALC_TARGET_NEED_RESEND:
3977                         cancel_linger_map_check(lreq);
3978                         /*
3979                          * scan_requests() for the previous epoch(s)
3980                          * may have already added it to the list, since
3981                          * it's not unlinked here.
3982                          */
3983                         if (list_empty(&lreq->scan_item))
3984                                 list_add_tail(&lreq->scan_item, need_resend_linger);
3985                         break;
3986                 case CALC_TARGET_POOL_DNE:
3987                         list_del_init(&lreq->scan_item);
3988                         check_linger_pool_dne(lreq);
3989                         break;
3990                 }
3991         }
3992
3993         for (n = rb_first(&osd->o_requests); n; ) {
3994                 struct ceph_osd_request *req =
3995                     rb_entry(n, struct ceph_osd_request, r_node);
3996                 enum calc_target_result ct_res;
3997
3998                 n = rb_next(n); /* unlink_request(), check_pool_dne() */
3999
4000                 dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
4001                 ct_res = calc_target(osdc, &req->r_t, false);
4002                 switch (ct_res) {
4003                 case CALC_TARGET_NO_ACTION:
4004                         force_resend_writes = cleared_full ||
4005                             (check_pool_cleared_full &&
4006                              pool_cleared_full(osdc, req->r_t.base_oloc.pool));
4007                         if (!force_resend &&
4008                             (!(req->r_flags & CEPH_OSD_FLAG_WRITE) ||
4009                              !force_resend_writes))
4010                                 break;
4011
4012                         fallthrough;
4013                 case CALC_TARGET_NEED_RESEND:
4014                         cancel_map_check(req);
4015                         unlink_request(osd, req);
4016                         insert_request(need_resend, req);
4017                         break;
4018                 case CALC_TARGET_POOL_DNE:
4019                         check_pool_dne(req);
4020                         break;
4021                 }
4022         }
4023 }
4024
4025 static int handle_one_map(struct ceph_osd_client *osdc,
4026                           void *p, void *end, bool incremental,
4027                           struct rb_root *need_resend,
4028                           struct list_head *need_resend_linger)
4029 {
4030         struct ceph_osdmap *newmap;
4031         struct rb_node *n;
4032         bool skipped_map = false;
4033         bool was_full;
4034
4035         was_full = ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
4036         set_pool_was_full(osdc);
4037
4038         if (incremental)
4039                 newmap = osdmap_apply_incremental(&p, end,
4040                                                   ceph_msgr2(osdc->client),
4041                                                   osdc->osdmap);
4042         else
4043                 newmap = ceph_osdmap_decode(&p, end, ceph_msgr2(osdc->client));
4044         if (IS_ERR(newmap))
4045                 return PTR_ERR(newmap);
4046
4047         if (newmap != osdc->osdmap) {
4048                 /*
4049                  * Preserve ->was_full before destroying the old map.
4050                  * For pools that weren't in the old map, ->was_full
4051                  * should be false.
4052                  */
4053                 for (n = rb_first(&newmap->pg_pools); n; n = rb_next(n)) {
4054                         struct ceph_pg_pool_info *pi =
4055                             rb_entry(n, struct ceph_pg_pool_info, node);
4056                         struct ceph_pg_pool_info *old_pi;
4057
4058                         old_pi = ceph_pg_pool_by_id(osdc->osdmap, pi->id);
4059                         if (old_pi)
4060                                 pi->was_full = old_pi->was_full;
4061                         else
4062                                 WARN_ON(pi->was_full);
4063                 }
4064
4065                 if (osdc->osdmap->epoch &&
4066                     osdc->osdmap->epoch + 1 < newmap->epoch) {
4067                         WARN_ON(incremental);
4068                         skipped_map = true;
4069                 }
4070
4071                 ceph_osdmap_destroy(osdc->osdmap);
4072                 osdc->osdmap = newmap;
4073         }
4074
4075         was_full &= !ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
4076         scan_requests(&osdc->homeless_osd, skipped_map, was_full, true,
4077                       need_resend, need_resend_linger);
4078
4079         for (n = rb_first(&osdc->osds); n; ) {
4080                 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
4081
4082                 n = rb_next(n); /* close_osd() */
4083
4084                 scan_requests(osd, skipped_map, was_full, true, need_resend,
4085                               need_resend_linger);
4086                 if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) ||
4087                     memcmp(&osd->o_con.peer_addr,
4088                            ceph_osd_addr(osdc->osdmap, osd->o_osd),
4089                            sizeof(struct ceph_entity_addr)))
4090                         close_osd(osd);
4091         }
4092
4093         return 0;
4094 }
4095
4096 static void kick_requests(struct ceph_osd_client *osdc,
4097                           struct rb_root *need_resend,
4098                           struct list_head *need_resend_linger)
4099 {
4100         struct ceph_osd_linger_request *lreq, *nlreq;
4101         enum calc_target_result ct_res;
4102         struct rb_node *n;
4103
4104         /* make sure need_resend targets reflect latest map */
4105         for (n = rb_first(need_resend); n; ) {
4106                 struct ceph_osd_request *req =
4107                     rb_entry(n, struct ceph_osd_request, r_node);
4108
4109                 n = rb_next(n);
4110
4111                 if (req->r_t.epoch < osdc->osdmap->epoch) {
4112                         ct_res = calc_target(osdc, &req->r_t, false);
4113                         if (ct_res == CALC_TARGET_POOL_DNE) {
4114                                 erase_request(need_resend, req);
4115                                 check_pool_dne(req);
4116                         }
4117                 }
4118         }
4119
4120         for (n = rb_first(need_resend); n; ) {
4121                 struct ceph_osd_request *req =
4122                     rb_entry(n, struct ceph_osd_request, r_node);
4123                 struct ceph_osd *osd;
4124
4125                 n = rb_next(n);
4126                 erase_request(need_resend, req); /* before link_request() */
4127
4128                 osd = lookup_create_osd(osdc, req->r_t.osd, true);
4129                 link_request(osd, req);
4130                 if (!req->r_linger) {
4131                         if (!osd_homeless(osd) && !req->r_t.paused)
4132                                 send_request(req);
4133                 } else {
4134                         cancel_linger_request(req);
4135                 }
4136         }
4137
4138         list_for_each_entry_safe(lreq, nlreq, need_resend_linger, scan_item) {
4139                 if (!osd_homeless(lreq->osd))
4140                         send_linger(lreq);
4141
4142                 list_del_init(&lreq->scan_item);
4143         }
4144 }
4145
4146 /*
4147  * Process updated osd map.
4148  *
4149  * The message contains any number of incremental and full maps, normally
4150  * indicating some sort of topology change in the cluster.  Kick requests
4151  * off to different OSDs as needed.
4152  */
4153 void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
4154 {
4155         void *p = msg->front.iov_base;
4156         void *const end = p + msg->front.iov_len;
4157         u32 nr_maps, maplen;
4158         u32 epoch;
4159         struct ceph_fsid fsid;
4160         struct rb_root need_resend = RB_ROOT;
4161         LIST_HEAD(need_resend_linger);
4162         bool handled_incremental = false;
4163         bool was_pauserd, was_pausewr;
4164         bool pauserd, pausewr;
4165         int err;
4166
4167         dout("%s have %u\n", __func__, osdc->osdmap->epoch);
4168         down_write(&osdc->lock);
4169
4170         /* verify fsid */
4171         ceph_decode_need(&p, end, sizeof(fsid), bad);
4172         ceph_decode_copy(&p, &fsid, sizeof(fsid));
4173         if (ceph_check_fsid(osdc->client, &fsid) < 0)
4174                 goto bad;
4175
4176         was_pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
4177         was_pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
4178                       ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
4179                       have_pool_full(osdc);
4180
4181         /* incremental maps */
4182         ceph_decode_32_safe(&p, end, nr_maps, bad);
4183         dout(" %d inc maps\n", nr_maps);
4184         while (nr_maps > 0) {
4185                 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
4186                 epoch = ceph_decode_32(&p);
4187                 maplen = ceph_decode_32(&p);
4188                 ceph_decode_need(&p, end, maplen, bad);
4189                 if (osdc->osdmap->epoch &&
4190                     osdc->osdmap->epoch + 1 == epoch) {
4191                         dout("applying incremental map %u len %d\n",
4192                              epoch, maplen);
4193                         err = handle_one_map(osdc, p, p + maplen, true,
4194                                              &need_resend, &need_resend_linger);
4195                         if (err)
4196                                 goto bad;
4197                         handled_incremental = true;
4198                 } else {
4199                         dout("ignoring incremental map %u len %d\n",
4200                              epoch, maplen);
4201                 }
4202                 p += maplen;
4203                 nr_maps--;
4204         }
4205         if (handled_incremental)
4206                 goto done;
4207
4208         /* full maps */
4209         ceph_decode_32_safe(&p, end, nr_maps, bad);
4210         dout(" %d full maps\n", nr_maps);
4211         while (nr_maps) {
4212                 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
4213                 epoch = ceph_decode_32(&p);
4214                 maplen = ceph_decode_32(&p);
4215                 ceph_decode_need(&p, end, maplen, bad);
4216                 if (nr_maps > 1) {
4217                         dout("skipping non-latest full map %u len %d\n",
4218                              epoch, maplen);
4219                 } else if (osdc->osdmap->epoch >= epoch) {
4220                         dout("skipping full map %u len %d, "
4221                              "older than our %u\n", epoch, maplen,
4222                              osdc->osdmap->epoch);
4223                 } else {
4224                         dout("taking full map %u len %d\n", epoch, maplen);
4225                         err = handle_one_map(osdc, p, p + maplen, false,
4226                                              &need_resend, &need_resend_linger);
4227                         if (err)
4228                                 goto bad;
4229                 }
4230                 p += maplen;
4231                 nr_maps--;
4232         }
4233
4234 done:
4235         /*
4236          * subscribe to subsequent osdmap updates if full to ensure
4237          * we find out when we are no longer full and stop returning
4238          * ENOSPC.
4239          */
4240         pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
4241         pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
4242                   ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
4243                   have_pool_full(osdc);
4244         if (was_pauserd || was_pausewr || pauserd || pausewr ||
4245             osdc->osdmap->epoch < osdc->epoch_barrier)
4246                 maybe_request_map(osdc);
4247
4248         kick_requests(osdc, &need_resend, &need_resend_linger);
4249
4250         ceph_osdc_abort_on_full(osdc);
4251         ceph_monc_got_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
4252                           osdc->osdmap->epoch);
4253         up_write(&osdc->lock);
4254         wake_up_all(&osdc->client->auth_wq);
4255         return;
4256
4257 bad:
4258         pr_err("osdc handle_map corrupt msg\n");
4259         ceph_msg_dump(msg);
4260         up_write(&osdc->lock);
4261 }
4262
4263 /*
4264  * Resubmit requests pending on the given osd.
4265  */
4266 static void kick_osd_requests(struct ceph_osd *osd)
4267 {
4268         struct rb_node *n;
4269
4270         clear_backoffs(osd);
4271
4272         for (n = rb_first(&osd->o_requests); n; ) {
4273                 struct ceph_osd_request *req =
4274                     rb_entry(n, struct ceph_osd_request, r_node);
4275
4276                 n = rb_next(n); /* cancel_linger_request() */
4277
4278                 if (!req->r_linger) {
4279                         if (!req->r_t.paused)
4280                                 send_request(req);
4281                 } else {
4282                         cancel_linger_request(req);
4283                 }
4284         }
4285         for (n = rb_first(&osd->o_linger_requests); n; n = rb_next(n)) {
4286                 struct ceph_osd_linger_request *lreq =
4287                     rb_entry(n, struct ceph_osd_linger_request, node);
4288
4289                 send_linger(lreq);
4290         }
4291 }
4292
4293 /*
4294  * If the osd connection drops, we need to resubmit all requests.
4295  */
4296 static void osd_fault(struct ceph_connection *con)
4297 {
4298         struct ceph_osd *osd = con->private;
4299         struct ceph_osd_client *osdc = osd->o_osdc;
4300
4301         dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
4302
4303         down_write(&osdc->lock);
4304         if (!osd_registered(osd)) {
4305                 dout("%s osd%d unknown\n", __func__, osd->o_osd);
4306                 goto out_unlock;
4307         }
4308
4309         if (!reopen_osd(osd))
4310                 kick_osd_requests(osd);
4311         maybe_request_map(osdc);
4312
4313 out_unlock:
4314         up_write(&osdc->lock);
4315 }
4316
4317 struct MOSDBackoff {
4318         struct ceph_spg spgid;
4319         u32 map_epoch;
4320         u8 op;
4321         u64 id;
4322         struct ceph_hobject_id *begin;
4323         struct ceph_hobject_id *end;
4324 };
4325
4326 static int decode_MOSDBackoff(const struct ceph_msg *msg, struct MOSDBackoff *m)
4327 {
4328         void *p = msg->front.iov_base;
4329         void *const end = p + msg->front.iov_len;
4330         u8 struct_v;
4331         u32 struct_len;
4332         int ret;
4333
4334         ret = ceph_start_decoding(&p, end, 1, "spg_t", &struct_v, &struct_len);
4335         if (ret)
4336                 return ret;
4337
4338         ret = ceph_decode_pgid(&p, end, &m->spgid.pgid);
4339         if (ret)
4340                 return ret;
4341
4342         ceph_decode_8_safe(&p, end, m->spgid.shard, e_inval);
4343         ceph_decode_32_safe(&p, end, m->map_epoch, e_inval);
4344         ceph_decode_8_safe(&p, end, m->op, e_inval);
4345         ceph_decode_64_safe(&p, end, m->id, e_inval);
4346
4347         m->begin = kzalloc(sizeof(*m->begin), GFP_NOIO);
4348         if (!m->begin)
4349                 return -ENOMEM;
4350
4351         ret = decode_hoid(&p, end, m->begin);
4352         if (ret) {
4353                 free_hoid(m->begin);
4354                 return ret;
4355         }
4356
4357         m->end = kzalloc(sizeof(*m->end), GFP_NOIO);
4358         if (!m->end) {
4359                 free_hoid(m->begin);
4360                 return -ENOMEM;
4361         }
4362
4363         ret = decode_hoid(&p, end, m->end);
4364         if (ret) {
4365                 free_hoid(m->begin);
4366                 free_hoid(m->end);
4367                 return ret;
4368         }
4369
4370         return 0;
4371
4372 e_inval:
4373         return -EINVAL;
4374 }
4375
4376 static struct ceph_msg *create_backoff_message(
4377                                 const struct ceph_osd_backoff *backoff,
4378                                 u32 map_epoch)
4379 {
4380         struct ceph_msg *msg;
4381         void *p, *end;
4382         int msg_size;
4383
4384         msg_size = CEPH_ENCODING_START_BLK_LEN +
4385                         CEPH_PGID_ENCODING_LEN + 1; /* spgid */
4386         msg_size += 4 + 1 + 8; /* map_epoch, op, id */
4387         msg_size += CEPH_ENCODING_START_BLK_LEN +
4388                         hoid_encoding_size(backoff->begin);
4389         msg_size += CEPH_ENCODING_START_BLK_LEN +
4390                         hoid_encoding_size(backoff->end);
4391
4392         msg = ceph_msg_new(CEPH_MSG_OSD_BACKOFF, msg_size, GFP_NOIO, true);
4393         if (!msg)
4394                 return NULL;
4395
4396         p = msg->front.iov_base;
4397         end = p + msg->front_alloc_len;
4398
4399         encode_spgid(&p, &backoff->spgid);
4400         ceph_encode_32(&p, map_epoch);
4401         ceph_encode_8(&p, CEPH_OSD_BACKOFF_OP_ACK_BLOCK);
4402         ceph_encode_64(&p, backoff->id);
4403         encode_hoid(&p, end, backoff->begin);
4404         encode_hoid(&p, end, backoff->end);
4405         BUG_ON(p != end);
4406
4407         msg->front.iov_len = p - msg->front.iov_base;
4408         msg->hdr.version = cpu_to_le16(1); /* MOSDBackoff v1 */
4409         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
4410
4411         return msg;
4412 }
4413
4414 static void handle_backoff_block(struct ceph_osd *osd, struct MOSDBackoff *m)
4415 {
4416         struct ceph_spg_mapping *spg;
4417         struct ceph_osd_backoff *backoff;
4418         struct ceph_msg *msg;
4419
4420         dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4421              m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4422
4423         spg = lookup_spg_mapping(&osd->o_backoff_mappings, &m->spgid);
4424         if (!spg) {
4425                 spg = alloc_spg_mapping();
4426                 if (!spg) {
4427                         pr_err("%s failed to allocate spg\n", __func__);
4428                         return;
4429                 }
4430                 spg->spgid = m->spgid; /* struct */
4431                 insert_spg_mapping(&osd->o_backoff_mappings, spg);
4432         }
4433
4434         backoff = alloc_backoff();
4435         if (!backoff) {
4436                 pr_err("%s failed to allocate backoff\n", __func__);
4437                 return;
4438         }
4439         backoff->spgid = m->spgid; /* struct */
4440         backoff->id = m->id;
4441         backoff->begin = m->begin;
4442         m->begin = NULL; /* backoff now owns this */
4443         backoff->end = m->end;
4444         m->end = NULL;   /* ditto */
4445
4446         insert_backoff(&spg->backoffs, backoff);
4447         insert_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4448
4449         /*
4450          * Ack with original backoff's epoch so that the OSD can
4451          * discard this if there was a PG split.
4452          */
4453         msg = create_backoff_message(backoff, m->map_epoch);
4454         if (!msg) {
4455                 pr_err("%s failed to allocate msg\n", __func__);
4456                 return;
4457         }
4458         ceph_con_send(&osd->o_con, msg);
4459 }
4460
4461 static bool target_contained_by(const struct ceph_osd_request_target *t,
4462                                 const struct ceph_hobject_id *begin,
4463                                 const struct ceph_hobject_id *end)
4464 {
4465         struct ceph_hobject_id hoid;
4466         int cmp;
4467
4468         hoid_fill_from_target(&hoid, t);
4469         cmp = hoid_compare(&hoid, begin);
4470         return !cmp || (cmp > 0 && hoid_compare(&hoid, end) < 0);
4471 }
4472
4473 static void handle_backoff_unblock(struct ceph_osd *osd,
4474                                    const struct MOSDBackoff *m)
4475 {
4476         struct ceph_spg_mapping *spg;
4477         struct ceph_osd_backoff *backoff;
4478         struct rb_node *n;
4479
4480         dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4481              m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4482
4483         backoff = lookup_backoff_by_id(&osd->o_backoffs_by_id, m->id);
4484         if (!backoff) {
4485                 pr_err("%s osd%d spgid %llu.%xs%d id %llu backoff dne\n",
4486                        __func__, osd->o_osd, m->spgid.pgid.pool,
4487                        m->spgid.pgid.seed, m->spgid.shard, m->id);
4488                 return;
4489         }
4490
4491         if (hoid_compare(backoff->begin, m->begin) &&
4492             hoid_compare(backoff->end, m->end)) {
4493                 pr_err("%s osd%d spgid %llu.%xs%d id %llu bad range?\n",
4494                        __func__, osd->o_osd, m->spgid.pgid.pool,
4495                        m->spgid.pgid.seed, m->spgid.shard, m->id);
4496                 /* unblock it anyway... */
4497         }
4498
4499         spg = lookup_spg_mapping(&osd->o_backoff_mappings, &backoff->spgid);
4500         BUG_ON(!spg);
4501
4502         erase_backoff(&spg->backoffs, backoff);
4503         erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4504         free_backoff(backoff);
4505
4506         if (RB_EMPTY_ROOT(&spg->backoffs)) {
4507                 erase_spg_mapping(&osd->o_backoff_mappings, spg);
4508                 free_spg_mapping(spg);
4509         }
4510
4511         for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
4512                 struct ceph_osd_request *req =
4513                     rb_entry(n, struct ceph_osd_request, r_node);
4514
4515                 if (!ceph_spg_compare(&req->r_t.spgid, &m->spgid)) {
4516                         /*
4517                          * Match against @m, not @backoff -- the PG may
4518                          * have split on the OSD.
4519                          */
4520                         if (target_contained_by(&req->r_t, m->begin, m->end)) {
4521                                 /*
4522                                  * If no other installed backoff applies,
4523                                  * resend.
4524                                  */
4525                                 send_request(req);
4526                         }
4527                 }
4528         }
4529 }
4530
4531 static void handle_backoff(struct ceph_osd *osd, struct ceph_msg *msg)
4532 {
4533         struct ceph_osd_client *osdc = osd->o_osdc;
4534         struct MOSDBackoff m;
4535         int ret;
4536
4537         down_read(&osdc->lock);
4538         if (!osd_registered(osd)) {
4539                 dout("%s osd%d unknown\n", __func__, osd->o_osd);
4540                 up_read(&osdc->lock);
4541                 return;
4542         }
4543         WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
4544
4545         mutex_lock(&osd->lock);
4546         ret = decode_MOSDBackoff(msg, &m);
4547         if (ret) {
4548                 pr_err("failed to decode MOSDBackoff: %d\n", ret);
4549                 ceph_msg_dump(msg);
4550                 goto out_unlock;
4551         }
4552
4553         switch (m.op) {
4554         case CEPH_OSD_BACKOFF_OP_BLOCK:
4555                 handle_backoff_block(osd, &m);
4556                 break;
4557         case CEPH_OSD_BACKOFF_OP_UNBLOCK:
4558                 handle_backoff_unblock(osd, &m);
4559                 break;
4560         default:
4561                 pr_err("%s osd%d unknown op %d\n", __func__, osd->o_osd, m.op);
4562         }
4563
4564         free_hoid(m.begin);
4565         free_hoid(m.end);
4566
4567 out_unlock:
4568         mutex_unlock(&osd->lock);
4569         up_read(&osdc->lock);
4570 }
4571
4572 /*
4573  * Process osd watch notifications
4574  */
4575 static void handle_watch_notify(struct ceph_osd_client *osdc,
4576                                 struct ceph_msg *msg)
4577 {
4578         void *p = msg->front.iov_base;
4579         void *const end = p + msg->front.iov_len;
4580         struct ceph_osd_linger_request *lreq;
4581         struct linger_work *lwork;
4582         u8 proto_ver, opcode;
4583         u64 cookie, notify_id;
4584         u64 notifier_id = 0;
4585         s32 return_code = 0;
4586         void *payload = NULL;
4587         u32 payload_len = 0;
4588
4589         ceph_decode_8_safe(&p, end, proto_ver, bad);
4590         ceph_decode_8_safe(&p, end, opcode, bad);
4591         ceph_decode_64_safe(&p, end, cookie, bad);
4592         p += 8; /* skip ver */
4593         ceph_decode_64_safe(&p, end, notify_id, bad);
4594
4595         if (proto_ver >= 1) {
4596                 ceph_decode_32_safe(&p, end, payload_len, bad);
4597                 ceph_decode_need(&p, end, payload_len, bad);
4598                 payload = p;
4599                 p += payload_len;
4600         }
4601
4602         if (le16_to_cpu(msg->hdr.version) >= 2)
4603                 ceph_decode_32_safe(&p, end, return_code, bad);
4604
4605         if (le16_to_cpu(msg->hdr.version) >= 3)
4606                 ceph_decode_64_safe(&p, end, notifier_id, bad);
4607
4608         down_read(&osdc->lock);
4609         lreq = lookup_linger_osdc(&osdc->linger_requests, cookie);
4610         if (!lreq) {
4611                 dout("%s opcode %d cookie %llu dne\n", __func__, opcode,
4612                      cookie);
4613                 goto out_unlock_osdc;
4614         }
4615
4616         mutex_lock(&lreq->lock);
4617         dout("%s opcode %d cookie %llu lreq %p is_watch %d\n", __func__,
4618              opcode, cookie, lreq, lreq->is_watch);
4619         if (opcode == CEPH_WATCH_EVENT_DISCONNECT) {
4620                 if (!lreq->last_error) {
4621                         lreq->last_error = -ENOTCONN;
4622                         queue_watch_error(lreq);
4623                 }
4624         } else if (!lreq->is_watch) {
4625                 /* CEPH_WATCH_EVENT_NOTIFY_COMPLETE */
4626                 if (lreq->notify_id && lreq->notify_id != notify_id) {
4627                         dout("lreq %p notify_id %llu != %llu, ignoring\n", lreq,
4628                              lreq->notify_id, notify_id);
4629                 } else if (!completion_done(&lreq->notify_finish_wait)) {
4630                         struct ceph_msg_data *data =
4631                             msg->num_data_items ? &msg->data[0] : NULL;
4632
4633                         if (data) {
4634                                 if (lreq->preply_pages) {
4635                                         WARN_ON(data->type !=
4636                                                         CEPH_MSG_DATA_PAGES);
4637                                         *lreq->preply_pages = data->pages;
4638                                         *lreq->preply_len = data->length;
4639                                         data->own_pages = false;
4640                                 }
4641                         }
4642                         lreq->notify_finish_error = return_code;
4643                         complete_all(&lreq->notify_finish_wait);
4644                 }
4645         } else {
4646                 /* CEPH_WATCH_EVENT_NOTIFY */
4647                 lwork = lwork_alloc(lreq, do_watch_notify);
4648                 if (!lwork) {
4649                         pr_err("failed to allocate notify-lwork\n");
4650                         goto out_unlock_lreq;
4651                 }
4652
4653                 lwork->notify.notify_id = notify_id;
4654                 lwork->notify.notifier_id = notifier_id;
4655                 lwork->notify.payload = payload;
4656                 lwork->notify.payload_len = payload_len;
4657                 lwork->notify.msg = ceph_msg_get(msg);
4658                 lwork_queue(lwork);
4659         }
4660
4661 out_unlock_lreq:
4662         mutex_unlock(&lreq->lock);
4663 out_unlock_osdc:
4664         up_read(&osdc->lock);
4665         return;
4666
4667 bad:
4668         pr_err("osdc handle_watch_notify corrupt msg\n");
4669 }
4670
4671 /*
4672  * Register request, send initial attempt.
4673  */
4674 void ceph_osdc_start_request(struct ceph_osd_client *osdc,
4675                              struct ceph_osd_request *req)
4676 {
4677         down_read(&osdc->lock);
4678         submit_request(req, false);
4679         up_read(&osdc->lock);
4680 }
4681 EXPORT_SYMBOL(ceph_osdc_start_request);
4682
4683 /*
4684  * Unregister request.  If @req was registered, it isn't completed:
4685  * r_result isn't set and __complete_request() isn't invoked.
4686  *
4687  * If @req wasn't registered, this call may have raced with
4688  * handle_reply(), in which case r_result would already be set and
4689  * __complete_request() would be getting invoked, possibly even
4690  * concurrently with this call.
4691  */
4692 void ceph_osdc_cancel_request(struct ceph_osd_request *req)
4693 {
4694         struct ceph_osd_client *osdc = req->r_osdc;
4695
4696         down_write(&osdc->lock);
4697         if (req->r_osd)
4698                 cancel_request(req);
4699         up_write(&osdc->lock);
4700 }
4701 EXPORT_SYMBOL(ceph_osdc_cancel_request);
4702
4703 /*
4704  * @timeout: in jiffies, 0 means "wait forever"
4705  */
4706 static int wait_request_timeout(struct ceph_osd_request *req,
4707                                 unsigned long timeout)
4708 {
4709         long left;
4710
4711         dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
4712         left = wait_for_completion_killable_timeout(&req->r_completion,
4713                                                 ceph_timeout_jiffies(timeout));
4714         if (left <= 0) {
4715                 left = left ?: -ETIMEDOUT;
4716                 ceph_osdc_cancel_request(req);
4717         } else {
4718                 left = req->r_result; /* completed */
4719         }
4720
4721         return left;
4722 }
4723
4724 /*
4725  * wait for a request to complete
4726  */
4727 int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
4728                            struct ceph_osd_request *req)
4729 {
4730         return wait_request_timeout(req, 0);
4731 }
4732 EXPORT_SYMBOL(ceph_osdc_wait_request);
4733
4734 /*
4735  * sync - wait for all in-flight requests to flush.  avoid starvation.
4736  */
4737 void ceph_osdc_sync(struct ceph_osd_client *osdc)
4738 {
4739         struct rb_node *n, *p;
4740         u64 last_tid = atomic64_read(&osdc->last_tid);
4741
4742 again:
4743         down_read(&osdc->lock);
4744         for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
4745                 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
4746
4747                 mutex_lock(&osd->lock);
4748                 for (p = rb_first(&osd->o_requests); p; p = rb_next(p)) {
4749                         struct ceph_osd_request *req =
4750                             rb_entry(p, struct ceph_osd_request, r_node);
4751
4752                         if (req->r_tid > last_tid)
4753                                 break;
4754
4755                         if (!(req->r_flags & CEPH_OSD_FLAG_WRITE))
4756                                 continue;
4757
4758                         ceph_osdc_get_request(req);
4759                         mutex_unlock(&osd->lock);
4760                         up_read(&osdc->lock);
4761                         dout("%s waiting on req %p tid %llu last_tid %llu\n",
4762                              __func__, req, req->r_tid, last_tid);
4763                         wait_for_completion(&req->r_completion);
4764                         ceph_osdc_put_request(req);
4765                         goto again;
4766                 }
4767
4768                 mutex_unlock(&osd->lock);
4769         }
4770
4771         up_read(&osdc->lock);
4772         dout("%s done last_tid %llu\n", __func__, last_tid);
4773 }
4774 EXPORT_SYMBOL(ceph_osdc_sync);
4775
4776 /*
4777  * Returns a handle, caller owns a ref.
4778  */
4779 struct ceph_osd_linger_request *
4780 ceph_osdc_watch(struct ceph_osd_client *osdc,
4781                 struct ceph_object_id *oid,
4782                 struct ceph_object_locator *oloc,
4783                 rados_watchcb2_t wcb,
4784                 rados_watcherrcb_t errcb,
4785                 void *data)
4786 {
4787         struct ceph_osd_linger_request *lreq;
4788         int ret;
4789
4790         lreq = linger_alloc(osdc);
4791         if (!lreq)
4792                 return ERR_PTR(-ENOMEM);
4793
4794         lreq->is_watch = true;
4795         lreq->wcb = wcb;
4796         lreq->errcb = errcb;
4797         lreq->data = data;
4798         lreq->watch_valid_thru = jiffies;
4799
4800         ceph_oid_copy(&lreq->t.base_oid, oid);
4801         ceph_oloc_copy(&lreq->t.base_oloc, oloc);
4802         lreq->t.flags = CEPH_OSD_FLAG_WRITE;
4803         ktime_get_real_ts64(&lreq->mtime);
4804
4805         linger_submit(lreq);
4806         ret = linger_reg_commit_wait(lreq);
4807         if (ret) {
4808                 linger_cancel(lreq);
4809                 goto err_put_lreq;
4810         }
4811
4812         return lreq;
4813
4814 err_put_lreq:
4815         linger_put(lreq);
4816         return ERR_PTR(ret);
4817 }
4818 EXPORT_SYMBOL(ceph_osdc_watch);
4819
4820 /*
4821  * Releases a ref.
4822  *
4823  * Times out after mount_timeout to preserve rbd unmap behaviour
4824  * introduced in 2894e1d76974 ("rbd: timeout watch teardown on unmap
4825  * with mount_timeout").
4826  */
4827 int ceph_osdc_unwatch(struct ceph_osd_client *osdc,
4828                       struct ceph_osd_linger_request *lreq)
4829 {
4830         struct ceph_options *opts = osdc->client->options;
4831         struct ceph_osd_request *req;
4832         int ret;
4833
4834         req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4835         if (!req)
4836                 return -ENOMEM;
4837
4838         ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
4839         ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
4840         req->r_flags = CEPH_OSD_FLAG_WRITE;
4841         ktime_get_real_ts64(&req->r_mtime);
4842         osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_UNWATCH,
4843                               lreq->linger_id, 0);
4844
4845         ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4846         if (ret)
4847                 goto out_put_req;
4848
4849         ceph_osdc_start_request(osdc, req);
4850         linger_cancel(lreq);
4851         linger_put(lreq);
4852         ret = wait_request_timeout(req, opts->mount_timeout);
4853
4854 out_put_req:
4855         ceph_osdc_put_request(req);
4856         return ret;
4857 }
4858 EXPORT_SYMBOL(ceph_osdc_unwatch);
4859
4860 static int osd_req_op_notify_ack_init(struct ceph_osd_request *req, int which,
4861                                       u64 notify_id, u64 cookie, void *payload,
4862                                       u32 payload_len)
4863 {
4864         struct ceph_osd_req_op *op;
4865         struct ceph_pagelist *pl;
4866         int ret;
4867
4868         op = osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY_ACK, 0);
4869
4870         pl = ceph_pagelist_alloc(GFP_NOIO);
4871         if (!pl)
4872                 return -ENOMEM;
4873
4874         ret = ceph_pagelist_encode_64(pl, notify_id);
4875         ret |= ceph_pagelist_encode_64(pl, cookie);
4876         if (payload) {
4877                 ret |= ceph_pagelist_encode_32(pl, payload_len);
4878                 ret |= ceph_pagelist_append(pl, payload, payload_len);
4879         } else {
4880                 ret |= ceph_pagelist_encode_32(pl, 0);
4881         }
4882         if (ret) {
4883                 ceph_pagelist_release(pl);
4884                 return -ENOMEM;
4885         }
4886
4887         ceph_osd_data_pagelist_init(&op->notify_ack.request_data, pl);
4888         op->indata_len = pl->length;
4889         return 0;
4890 }
4891
4892 int ceph_osdc_notify_ack(struct ceph_osd_client *osdc,
4893                          struct ceph_object_id *oid,
4894                          struct ceph_object_locator *oloc,
4895                          u64 notify_id,
4896                          u64 cookie,
4897                          void *payload,
4898                          u32 payload_len)
4899 {
4900         struct ceph_osd_request *req;
4901         int ret;
4902
4903         req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4904         if (!req)
4905                 return -ENOMEM;
4906
4907         ceph_oid_copy(&req->r_base_oid, oid);
4908         ceph_oloc_copy(&req->r_base_oloc, oloc);
4909         req->r_flags = CEPH_OSD_FLAG_READ;
4910
4911         ret = osd_req_op_notify_ack_init(req, 0, notify_id, cookie, payload,
4912                                          payload_len);
4913         if (ret)
4914                 goto out_put_req;
4915
4916         ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4917         if (ret)
4918                 goto out_put_req;
4919
4920         ceph_osdc_start_request(osdc, req);
4921         ret = ceph_osdc_wait_request(osdc, req);
4922
4923 out_put_req:
4924         ceph_osdc_put_request(req);
4925         return ret;
4926 }
4927 EXPORT_SYMBOL(ceph_osdc_notify_ack);
4928
4929 /*
4930  * @timeout: in seconds
4931  *
4932  * @preply_{pages,len} are initialized both on success and error.
4933  * The caller is responsible for:
4934  *
4935  *     ceph_release_page_vector(reply_pages, calc_pages_for(0, reply_len))
4936  */
4937 int ceph_osdc_notify(struct ceph_osd_client *osdc,
4938                      struct ceph_object_id *oid,
4939                      struct ceph_object_locator *oloc,
4940                      void *payload,
4941                      u32 payload_len,
4942                      u32 timeout,
4943                      struct page ***preply_pages,
4944                      size_t *preply_len)
4945 {
4946         struct ceph_osd_linger_request *lreq;
4947         int ret;
4948
4949         WARN_ON(!timeout);
4950         if (preply_pages) {
4951                 *preply_pages = NULL;
4952                 *preply_len = 0;
4953         }
4954
4955         lreq = linger_alloc(osdc);
4956         if (!lreq)
4957                 return -ENOMEM;
4958
4959         lreq->request_pl = ceph_pagelist_alloc(GFP_NOIO);
4960         if (!lreq->request_pl) {
4961                 ret = -ENOMEM;
4962                 goto out_put_lreq;
4963         }
4964
4965         ret = ceph_pagelist_encode_32(lreq->request_pl, 1); /* prot_ver */
4966         ret |= ceph_pagelist_encode_32(lreq->request_pl, timeout);
4967         ret |= ceph_pagelist_encode_32(lreq->request_pl, payload_len);
4968         ret |= ceph_pagelist_append(lreq->request_pl, payload, payload_len);
4969         if (ret) {
4970                 ret = -ENOMEM;
4971                 goto out_put_lreq;
4972         }
4973
4974         /* for notify_id */
4975         lreq->notify_id_pages = ceph_alloc_page_vector(1, GFP_NOIO);
4976         if (IS_ERR(lreq->notify_id_pages)) {
4977                 ret = PTR_ERR(lreq->notify_id_pages);
4978                 lreq->notify_id_pages = NULL;
4979                 goto out_put_lreq;
4980         }
4981
4982         lreq->preply_pages = preply_pages;
4983         lreq->preply_len = preply_len;
4984
4985         ceph_oid_copy(&lreq->t.base_oid, oid);
4986         ceph_oloc_copy(&lreq->t.base_oloc, oloc);
4987         lreq->t.flags = CEPH_OSD_FLAG_READ;
4988
4989         linger_submit(lreq);
4990         ret = linger_reg_commit_wait(lreq);
4991         if (!ret)
4992                 ret = linger_notify_finish_wait(lreq,
4993                                  msecs_to_jiffies(2 * timeout * MSEC_PER_SEC));
4994         else
4995                 dout("lreq %p failed to initiate notify %d\n", lreq, ret);
4996
4997         linger_cancel(lreq);
4998 out_put_lreq:
4999         linger_put(lreq);
5000         return ret;
5001 }
5002 EXPORT_SYMBOL(ceph_osdc_notify);
5003
5004 static int decode_watcher(void **p, void *end, struct ceph_watch_item *item)
5005 {
5006         u8 struct_v;
5007         u32 struct_len;
5008         int ret;
5009
5010         ret = ceph_start_decoding(p, end, 2, "watch_item_t",
5011                                   &struct_v, &struct_len);
5012         if (ret)
5013                 goto bad;
5014
5015         ret = -EINVAL;
5016         ceph_decode_copy_safe(p, end, &item->name, sizeof(item->name), bad);
5017         ceph_decode_64_safe(p, end, item->cookie, bad);
5018         ceph_decode_skip_32(p, end, bad); /* skip timeout seconds */
5019
5020         if (struct_v >= 2) {
5021                 ret = ceph_decode_entity_addr(p, end, &item->addr);
5022                 if (ret)
5023                         goto bad;
5024         } else {
5025                 ret = 0;
5026         }
5027
5028         dout("%s %s%llu cookie %llu addr %s\n", __func__,
5029              ENTITY_NAME(item->name), item->cookie,
5030              ceph_pr_addr(&item->addr));
5031 bad:
5032         return ret;
5033 }
5034
5035 static int decode_watchers(void **p, void *end,
5036                            struct ceph_watch_item **watchers,
5037                            u32 *num_watchers)
5038 {
5039         u8 struct_v;
5040         u32 struct_len;
5041         int i;
5042         int ret;
5043
5044         ret = ceph_start_decoding(p, end, 1, "obj_list_watch_response_t",
5045                                   &struct_v, &struct_len);
5046         if (ret)
5047                 return ret;
5048
5049         *num_watchers = ceph_decode_32(p);
5050         *watchers = kcalloc(*num_watchers, sizeof(**watchers), GFP_NOIO);
5051         if (!*watchers)
5052                 return -ENOMEM;
5053
5054         for (i = 0; i < *num_watchers; i++) {
5055                 ret = decode_watcher(p, end, *watchers + i);
5056                 if (ret) {
5057                         kfree(*watchers);
5058                         return ret;
5059                 }
5060         }
5061
5062         return 0;
5063 }
5064
5065 /*
5066  * On success, the caller is responsible for:
5067  *
5068  *     kfree(watchers);
5069  */
5070 int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
5071                             struct ceph_object_id *oid,
5072                             struct ceph_object_locator *oloc,
5073                             struct ceph_watch_item **watchers,
5074                             u32 *num_watchers)
5075 {
5076         struct ceph_osd_request *req;
5077         struct page **pages;
5078         int ret;
5079
5080         req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
5081         if (!req)
5082                 return -ENOMEM;
5083
5084         ceph_oid_copy(&req->r_base_oid, oid);
5085         ceph_oloc_copy(&req->r_base_oloc, oloc);
5086         req->r_flags = CEPH_OSD_FLAG_READ;
5087
5088         pages = ceph_alloc_page_vector(1, GFP_NOIO);
5089         if (IS_ERR(pages)) {
5090                 ret = PTR_ERR(pages);
5091                 goto out_put_req;
5092         }
5093
5094         osd_req_op_init(req, 0, CEPH_OSD_OP_LIST_WATCHERS, 0);
5095         ceph_osd_data_pages_init(osd_req_op_data(req, 0, list_watchers,
5096                                                  response_data),
5097                                  pages, PAGE_SIZE, 0, false, true);
5098
5099         ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
5100         if (ret)
5101                 goto out_put_req;
5102
5103         ceph_osdc_start_request(osdc, req);
5104         ret = ceph_osdc_wait_request(osdc, req);
5105         if (ret >= 0) {
5106                 void *p = page_address(pages[0]);
5107                 void *const end = p + req->r_ops[0].outdata_len;
5108
5109                 ret = decode_watchers(&p, end, watchers, num_watchers);
5110         }
5111
5112 out_put_req:
5113         ceph_osdc_put_request(req);
5114         return ret;
5115 }
5116 EXPORT_SYMBOL(ceph_osdc_list_watchers);
5117
5118 /*
5119  * Call all pending notify callbacks - for use after a watch is
5120  * unregistered, to make sure no more callbacks for it will be invoked
5121  */
5122 void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc)
5123 {
5124         dout("%s osdc %p\n", __func__, osdc);
5125         flush_workqueue(osdc->notify_wq);
5126 }
5127 EXPORT_SYMBOL(ceph_osdc_flush_notifies);
5128
5129 void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc)
5130 {
5131         down_read(&osdc->lock);
5132         maybe_request_map(osdc);
5133         up_read(&osdc->lock);
5134 }
5135 EXPORT_SYMBOL(ceph_osdc_maybe_request_map);
5136
5137 /*
5138  * Execute an OSD class method on an object.
5139  *
5140  * @flags: CEPH_OSD_FLAG_*
5141  * @resp_len: in/out param for reply length
5142  */
5143 int ceph_osdc_call(struct ceph_osd_client *osdc,
5144                    struct ceph_object_id *oid,
5145                    struct ceph_object_locator *oloc,
5146                    const char *class, const char *method,
5147                    unsigned int flags,
5148                    struct page *req_page, size_t req_len,
5149                    struct page **resp_pages, size_t *resp_len)
5150 {
5151         struct ceph_osd_request *req;
5152         int ret;
5153
5154         if (req_len > PAGE_SIZE)
5155                 return -E2BIG;
5156
5157         req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
5158         if (!req)
5159                 return -ENOMEM;
5160
5161         ceph_oid_copy(&req->r_base_oid, oid);
5162         ceph_oloc_copy(&req->r_base_oloc, oloc);
5163         req->r_flags = flags;
5164
5165         ret = osd_req_op_cls_init(req, 0, class, method);
5166         if (ret)
5167                 goto out_put_req;
5168
5169         if (req_page)
5170                 osd_req_op_cls_request_data_pages(req, 0, &req_page, req_len,
5171                                                   0, false, false);
5172         if (resp_pages)
5173                 osd_req_op_cls_response_data_pages(req, 0, resp_pages,
5174                                                    *resp_len, 0, false, false);
5175
5176         ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
5177         if (ret)
5178                 goto out_put_req;
5179
5180         ceph_osdc_start_request(osdc, req);
5181         ret = ceph_osdc_wait_request(osdc, req);
5182         if (ret >= 0) {
5183                 ret = req->r_ops[0].rval;
5184                 if (resp_pages)
5185                         *resp_len = req->r_ops[0].outdata_len;
5186         }
5187
5188 out_put_req:
5189         ceph_osdc_put_request(req);
5190         return ret;
5191 }
5192 EXPORT_SYMBOL(ceph_osdc_call);
5193
5194 /*
5195  * reset all osd connections
5196  */
5197 void ceph_osdc_reopen_osds(struct ceph_osd_client *osdc)
5198 {
5199         struct rb_node *n;
5200
5201         down_write(&osdc->lock);
5202         for (n = rb_first(&osdc->osds); n; ) {
5203                 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
5204
5205                 n = rb_next(n);
5206                 if (!reopen_osd(osd))
5207                         kick_osd_requests(osd);
5208         }
5209         up_write(&osdc->lock);
5210 }
5211
5212 /*
5213  * init, shutdown
5214  */
5215 int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client)
5216 {
5217         int err;
5218
5219         dout("init\n");
5220         osdc->client = client;
5221         init_rwsem(&osdc->lock);
5222         osdc->osds = RB_ROOT;
5223         INIT_LIST_HEAD(&osdc->osd_lru);
5224         spin_lock_init(&osdc->osd_lru_lock);
5225         osd_init(&osdc->homeless_osd);
5226         osdc->homeless_osd.o_osdc = osdc;
5227         osdc->homeless_osd.o_osd = CEPH_HOMELESS_OSD;
5228         osdc->last_linger_id = CEPH_LINGER_ID_START;
5229         osdc->linger_requests = RB_ROOT;
5230         osdc->map_checks = RB_ROOT;
5231         osdc->linger_map_checks = RB_ROOT;
5232         INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout);
5233         INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout);
5234
5235         err = -ENOMEM;
5236         osdc->osdmap = ceph_osdmap_alloc();
5237         if (!osdc->osdmap)
5238                 goto out;
5239
5240         osdc->req_mempool = mempool_create_slab_pool(10,
5241                                                      ceph_osd_request_cache);
5242         if (!osdc->req_mempool)
5243                 goto out_map;
5244
5245         err = ceph_msgpool_init(&osdc->msgpool_op, CEPH_MSG_OSD_OP,
5246                                 PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10, "osd_op");
5247         if (err < 0)
5248                 goto out_mempool;
5249         err = ceph_msgpool_init(&osdc->msgpool_op_reply, CEPH_MSG_OSD_OPREPLY,
5250                                 PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10,
5251                                 "osd_op_reply");
5252         if (err < 0)
5253                 goto out_msgpool;
5254
5255         err = -ENOMEM;
5256         osdc->notify_wq = create_singlethread_workqueue("ceph-watch-notify");
5257         if (!osdc->notify_wq)
5258                 goto out_msgpool_reply;
5259
5260         osdc->completion_wq = create_singlethread_workqueue("ceph-completion");
5261         if (!osdc->completion_wq)
5262                 goto out_notify_wq;
5263
5264         schedule_delayed_work(&osdc->timeout_work,
5265                               osdc->client->options->osd_keepalive_timeout);
5266         schedule_delayed_work(&osdc->osds_timeout_work,
5267             round_jiffies_relative(osdc->client->options->osd_idle_ttl));
5268
5269         return 0;
5270
5271 out_notify_wq:
5272         destroy_workqueue(osdc->notify_wq);
5273 out_msgpool_reply:
5274         ceph_msgpool_destroy(&osdc->msgpool_op_reply);
5275 out_msgpool:
5276         ceph_msgpool_destroy(&osdc->msgpool_op);
5277 out_mempool:
5278         mempool_destroy(osdc->req_mempool);
5279 out_map:
5280         ceph_osdmap_destroy(osdc->osdmap);
5281 out:
5282         return err;
5283 }
5284
5285 void ceph_osdc_stop(struct ceph_osd_client *osdc)
5286 {
5287         destroy_workqueue(osdc->completion_wq);
5288         destroy_workqueue(osdc->notify_wq);
5289         cancel_delayed_work_sync(&osdc->timeout_work);
5290         cancel_delayed_work_sync(&osdc->osds_timeout_work);
5291
5292         down_write(&osdc->lock);
5293         while (!RB_EMPTY_ROOT(&osdc->osds)) {
5294                 struct ceph_osd *osd = rb_entry(rb_first(&osdc->osds),
5295                                                 struct ceph_osd, o_node);
5296                 close_osd(osd);
5297         }
5298         up_write(&osdc->lock);
5299         WARN_ON(refcount_read(&osdc->homeless_osd.o_ref) != 1);
5300         osd_cleanup(&osdc->homeless_osd);
5301
5302         WARN_ON(!list_empty(&osdc->osd_lru));
5303         WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_requests));
5304         WARN_ON(!RB_EMPTY_ROOT(&osdc->map_checks));
5305         WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_map_checks));
5306         WARN_ON(atomic_read(&osdc->num_requests));
5307         WARN_ON(atomic_read(&osdc->num_homeless));
5308
5309         ceph_osdmap_destroy(osdc->osdmap);
5310         mempool_destroy(osdc->req_mempool);
5311         ceph_msgpool_destroy(&osdc->msgpool_op);
5312         ceph_msgpool_destroy(&osdc->msgpool_op_reply);
5313 }
5314
5315 int osd_req_op_copy_from_init(struct ceph_osd_request *req,
5316                               u64 src_snapid, u64 src_version,
5317                               struct ceph_object_id *src_oid,
5318                               struct ceph_object_locator *src_oloc,
5319                               u32 src_fadvise_flags,
5320                               u32 dst_fadvise_flags,
5321                               u32 truncate_seq, u64 truncate_size,
5322                               u8 copy_from_flags)
5323 {
5324         struct ceph_osd_req_op *op;
5325         struct page **pages;
5326         void *p, *end;
5327
5328         pages = ceph_alloc_page_vector(1, GFP_KERNEL);
5329         if (IS_ERR(pages))
5330                 return PTR_ERR(pages);
5331
5332         op = osd_req_op_init(req, 0, CEPH_OSD_OP_COPY_FROM2,
5333                              dst_fadvise_flags);
5334         op->copy_from.snapid = src_snapid;
5335         op->copy_from.src_version = src_version;
5336         op->copy_from.flags = copy_from_flags;
5337         op->copy_from.src_fadvise_flags = src_fadvise_flags;
5338
5339         p = page_address(pages[0]);
5340         end = p + PAGE_SIZE;
5341         ceph_encode_string(&p, end, src_oid->name, src_oid->name_len);
5342         encode_oloc(&p, end, src_oloc);
5343         ceph_encode_32(&p, truncate_seq);
5344         ceph_encode_64(&p, truncate_size);
5345         op->indata_len = PAGE_SIZE - (end - p);
5346
5347         ceph_osd_data_pages_init(&op->copy_from.osd_data, pages,
5348                                  op->indata_len, 0, false, true);
5349         return 0;
5350 }
5351 EXPORT_SYMBOL(osd_req_op_copy_from_init);
5352
5353 int __init ceph_osdc_setup(void)
5354 {
5355         size_t size = sizeof(struct ceph_osd_request) +
5356             CEPH_OSD_SLAB_OPS * sizeof(struct ceph_osd_req_op);
5357
5358         BUG_ON(ceph_osd_request_cache);
5359         ceph_osd_request_cache = kmem_cache_create("ceph_osd_request", size,
5360                                                    0, 0, NULL);
5361
5362         return ceph_osd_request_cache ? 0 : -ENOMEM;
5363 }
5364
5365 void ceph_osdc_cleanup(void)
5366 {
5367         BUG_ON(!ceph_osd_request_cache);
5368         kmem_cache_destroy(ceph_osd_request_cache);
5369         ceph_osd_request_cache = NULL;
5370 }
5371
5372 /*
5373  * handle incoming message
5374  */
5375 static void osd_dispatch(struct ceph_connection *con, struct ceph_msg *msg)
5376 {
5377         struct ceph_osd *osd = con->private;
5378         struct ceph_osd_client *osdc = osd->o_osdc;
5379         int type = le16_to_cpu(msg->hdr.type);
5380
5381         switch (type) {
5382         case CEPH_MSG_OSD_MAP:
5383                 ceph_osdc_handle_map(osdc, msg);
5384                 break;
5385         case CEPH_MSG_OSD_OPREPLY:
5386                 handle_reply(osd, msg);
5387                 break;
5388         case CEPH_MSG_OSD_BACKOFF:
5389                 handle_backoff(osd, msg);
5390                 break;
5391         case CEPH_MSG_WATCH_NOTIFY:
5392                 handle_watch_notify(osdc, msg);
5393                 break;
5394
5395         default:
5396                 pr_err("received unknown message type %d %s\n", type,
5397                        ceph_msg_type_name(type));
5398         }
5399
5400         ceph_msg_put(msg);
5401 }
5402
5403 /* How much sparse data was requested? */
5404 static u64 sparse_data_requested(struct ceph_osd_request *req)
5405 {
5406         u64 len = 0;
5407
5408         if (req->r_flags & CEPH_OSD_FLAG_READ) {
5409                 int i;
5410
5411                 for (i = 0; i < req->r_num_ops; ++i) {
5412                         struct ceph_osd_req_op *op = &req->r_ops[i];
5413
5414                         if (op->op == CEPH_OSD_OP_SPARSE_READ)
5415                                 len += op->extent.length;
5416                 }
5417         }
5418         return len;
5419 }
5420
5421 /*
5422  * Lookup and return message for incoming reply.  Don't try to do
5423  * anything about a larger than preallocated data portion of the
5424  * message at the moment - for now, just skip the message.
5425  */
5426 static struct ceph_msg *get_reply(struct ceph_connection *con,
5427                                   struct ceph_msg_header *hdr,
5428                                   int *skip)
5429 {
5430         struct ceph_osd *osd = con->private;
5431         struct ceph_osd_client *osdc = osd->o_osdc;
5432         struct ceph_msg *m = NULL;
5433         struct ceph_osd_request *req;
5434         int front_len = le32_to_cpu(hdr->front_len);
5435         int data_len = le32_to_cpu(hdr->data_len);
5436         u64 tid = le64_to_cpu(hdr->tid);
5437         u64 srlen;
5438
5439         down_read(&osdc->lock);
5440         if (!osd_registered(osd)) {
5441                 dout("%s osd%d unknown, skipping\n", __func__, osd->o_osd);
5442                 *skip = 1;
5443                 goto out_unlock_osdc;
5444         }
5445         WARN_ON(osd->o_osd != le64_to_cpu(hdr->src.num));
5446
5447         mutex_lock(&osd->lock);
5448         req = lookup_request(&osd->o_requests, tid);
5449         if (!req) {
5450                 dout("%s osd%d tid %llu unknown, skipping\n", __func__,
5451                      osd->o_osd, tid);
5452                 *skip = 1;
5453                 goto out_unlock_session;
5454         }
5455
5456         ceph_msg_revoke_incoming(req->r_reply);
5457
5458         if (front_len > req->r_reply->front_alloc_len) {
5459                 pr_warn("%s osd%d tid %llu front %d > preallocated %d\n",
5460                         __func__, osd->o_osd, req->r_tid, front_len,
5461                         req->r_reply->front_alloc_len);
5462                 m = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, front_len, GFP_NOFS,
5463                                  false);
5464                 if (!m)
5465                         goto out_unlock_session;
5466                 ceph_msg_put(req->r_reply);
5467                 req->r_reply = m;
5468         }
5469
5470         srlen = sparse_data_requested(req);
5471         if (!srlen && data_len > req->r_reply->data_length) {
5472                 pr_warn("%s osd%d tid %llu data %d > preallocated %zu, skipping\n",
5473                         __func__, osd->o_osd, req->r_tid, data_len,
5474                         req->r_reply->data_length);
5475                 m = NULL;
5476                 *skip = 1;
5477                 goto out_unlock_session;
5478         }
5479
5480         m = ceph_msg_get(req->r_reply);
5481         m->sparse_read_total = srlen;
5482
5483         dout("get_reply tid %lld %p\n", tid, m);
5484
5485 out_unlock_session:
5486         mutex_unlock(&osd->lock);
5487 out_unlock_osdc:
5488         up_read(&osdc->lock);
5489         return m;
5490 }
5491
5492 static struct ceph_msg *alloc_msg_with_page_vector(struct ceph_msg_header *hdr)
5493 {
5494         struct ceph_msg *m;
5495         int type = le16_to_cpu(hdr->type);
5496         u32 front_len = le32_to_cpu(hdr->front_len);
5497         u32 data_len = le32_to_cpu(hdr->data_len);
5498
5499         m = ceph_msg_new2(type, front_len, 1, GFP_NOIO, false);
5500         if (!m)
5501                 return NULL;
5502
5503         if (data_len) {
5504                 struct page **pages;
5505
5506                 pages = ceph_alloc_page_vector(calc_pages_for(0, data_len),
5507                                                GFP_NOIO);
5508                 if (IS_ERR(pages)) {
5509                         ceph_msg_put(m);
5510                         return NULL;
5511                 }
5512
5513                 ceph_msg_data_add_pages(m, pages, data_len, 0, true);
5514         }
5515
5516         return m;
5517 }
5518
5519 static struct ceph_msg *osd_alloc_msg(struct ceph_connection *con,
5520                                       struct ceph_msg_header *hdr,
5521                                       int *skip)
5522 {
5523         struct ceph_osd *osd = con->private;
5524         int type = le16_to_cpu(hdr->type);
5525
5526         *skip = 0;
5527         switch (type) {
5528         case CEPH_MSG_OSD_MAP:
5529         case CEPH_MSG_OSD_BACKOFF:
5530         case CEPH_MSG_WATCH_NOTIFY:
5531                 return alloc_msg_with_page_vector(hdr);
5532         case CEPH_MSG_OSD_OPREPLY:
5533                 return get_reply(con, hdr, skip);
5534         default:
5535                 pr_warn("%s osd%d unknown msg type %d, skipping\n", __func__,
5536                         osd->o_osd, type);
5537                 *skip = 1;
5538                 return NULL;
5539         }
5540 }
5541
5542 /*
5543  * Wrappers to refcount containing ceph_osd struct
5544  */
5545 static struct ceph_connection *osd_get_con(struct ceph_connection *con)
5546 {
5547         struct ceph_osd *osd = con->private;
5548         if (get_osd(osd))
5549                 return con;
5550         return NULL;
5551 }
5552
5553 static void osd_put_con(struct ceph_connection *con)
5554 {
5555         struct ceph_osd *osd = con->private;
5556         put_osd(osd);
5557 }
5558
5559 /*
5560  * authentication
5561  */
5562
5563 /*
5564  * Note: returned pointer is the address of a structure that's
5565  * managed separately.  Caller must *not* attempt to free it.
5566  */
5567 static struct ceph_auth_handshake *
5568 osd_get_authorizer(struct ceph_connection *con, int *proto, int force_new)
5569 {
5570         struct ceph_osd *o = con->private;
5571         struct ceph_osd_client *osdc = o->o_osdc;
5572         struct ceph_auth_client *ac = osdc->client->monc.auth;
5573         struct ceph_auth_handshake *auth = &o->o_auth;
5574         int ret;
5575
5576         ret = __ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_OSD,
5577                                          force_new, proto, NULL, NULL);
5578         if (ret)
5579                 return ERR_PTR(ret);
5580
5581         return auth;
5582 }
5583
5584 static int osd_add_authorizer_challenge(struct ceph_connection *con,
5585                                     void *challenge_buf, int challenge_buf_len)
5586 {
5587         struct ceph_osd *o = con->private;
5588         struct ceph_osd_client *osdc = o->o_osdc;
5589         struct ceph_auth_client *ac = osdc->client->monc.auth;
5590
5591         return ceph_auth_add_authorizer_challenge(ac, o->o_auth.authorizer,
5592                                             challenge_buf, challenge_buf_len);
5593 }
5594
5595 static int osd_verify_authorizer_reply(struct ceph_connection *con)
5596 {
5597         struct ceph_osd *o = con->private;
5598         struct ceph_osd_client *osdc = o->o_osdc;
5599         struct ceph_auth_client *ac = osdc->client->monc.auth;
5600         struct ceph_auth_handshake *auth = &o->o_auth;
5601
5602         return ceph_auth_verify_authorizer_reply(ac, auth->authorizer,
5603                 auth->authorizer_reply_buf, auth->authorizer_reply_buf_len,
5604                 NULL, NULL, NULL, NULL);
5605 }
5606
5607 static int osd_invalidate_authorizer(struct ceph_connection *con)
5608 {
5609         struct ceph_osd *o = con->private;
5610         struct ceph_osd_client *osdc = o->o_osdc;
5611         struct ceph_auth_client *ac = osdc->client->monc.auth;
5612
5613         ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_OSD);
5614         return ceph_monc_validate_auth(&osdc->client->monc);
5615 }
5616
5617 static int osd_get_auth_request(struct ceph_connection *con,
5618                                 void *buf, int *buf_len,
5619                                 void **authorizer, int *authorizer_len)
5620 {
5621         struct ceph_osd *o = con->private;
5622         struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5623         struct ceph_auth_handshake *auth = &o->o_auth;
5624         int ret;
5625
5626         ret = ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_OSD,
5627                                        buf, buf_len);
5628         if (ret)
5629                 return ret;
5630
5631         *authorizer = auth->authorizer_buf;
5632         *authorizer_len = auth->authorizer_buf_len;
5633         return 0;
5634 }
5635
5636 static int osd_handle_auth_reply_more(struct ceph_connection *con,
5637                                       void *reply, int reply_len,
5638                                       void *buf, int *buf_len,
5639                                       void **authorizer, int *authorizer_len)
5640 {
5641         struct ceph_osd *o = con->private;
5642         struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5643         struct ceph_auth_handshake *auth = &o->o_auth;
5644         int ret;
5645
5646         ret = ceph_auth_handle_svc_reply_more(ac, auth, reply, reply_len,
5647                                               buf, buf_len);
5648         if (ret)
5649                 return ret;
5650
5651         *authorizer = auth->authorizer_buf;
5652         *authorizer_len = auth->authorizer_buf_len;
5653         return 0;
5654 }
5655
5656 static int osd_handle_auth_done(struct ceph_connection *con,
5657                                 u64 global_id, void *reply, int reply_len,
5658                                 u8 *session_key, int *session_key_len,
5659                                 u8 *con_secret, int *con_secret_len)
5660 {
5661         struct ceph_osd *o = con->private;
5662         struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5663         struct ceph_auth_handshake *auth = &o->o_auth;
5664
5665         return ceph_auth_handle_svc_reply_done(ac, auth, reply, reply_len,
5666                                                session_key, session_key_len,
5667                                                con_secret, con_secret_len);
5668 }
5669
5670 static int osd_handle_auth_bad_method(struct ceph_connection *con,
5671                                       int used_proto, int result,
5672                                       const int *allowed_protos, int proto_cnt,
5673                                       const int *allowed_modes, int mode_cnt)
5674 {
5675         struct ceph_osd *o = con->private;
5676         struct ceph_mon_client *monc = &o->o_osdc->client->monc;
5677         int ret;
5678
5679         if (ceph_auth_handle_bad_authorizer(monc->auth, CEPH_ENTITY_TYPE_OSD,
5680                                             used_proto, result,
5681                                             allowed_protos, proto_cnt,
5682                                             allowed_modes, mode_cnt)) {
5683                 ret = ceph_monc_validate_auth(monc);
5684                 if (ret)
5685                         return ret;
5686         }
5687
5688         return -EACCES;
5689 }
5690
5691 static void osd_reencode_message(struct ceph_msg *msg)
5692 {
5693         int type = le16_to_cpu(msg->hdr.type);
5694
5695         if (type == CEPH_MSG_OSD_OP)
5696                 encode_request_finish(msg);
5697 }
5698
5699 static int osd_sign_message(struct ceph_msg *msg)
5700 {
5701         struct ceph_osd *o = msg->con->private;
5702         struct ceph_auth_handshake *auth = &o->o_auth;
5703
5704         return ceph_auth_sign_message(auth, msg);
5705 }
5706
5707 static int osd_check_message_signature(struct ceph_msg *msg)
5708 {
5709         struct ceph_osd *o = msg->con->private;
5710         struct ceph_auth_handshake *auth = &o->o_auth;
5711
5712         return ceph_auth_check_message_signature(auth, msg);
5713 }
5714
5715 static void advance_cursor(struct ceph_msg_data_cursor *cursor, size_t len,
5716                            bool zero)
5717 {
5718         while (len) {
5719                 struct page *page;
5720                 size_t poff, plen;
5721
5722                 page = ceph_msg_data_next(cursor, &poff, &plen);
5723                 if (plen > len)
5724                         plen = len;
5725                 if (zero)
5726                         zero_user_segment(page, poff, poff + plen);
5727                 len -= plen;
5728                 ceph_msg_data_advance(cursor, plen);
5729         }
5730 }
5731
5732 static int prep_next_sparse_read(struct ceph_connection *con,
5733                                  struct ceph_msg_data_cursor *cursor)
5734 {
5735         struct ceph_osd *o = con->private;
5736         struct ceph_sparse_read *sr = &o->o_sparse_read;
5737         struct ceph_osd_request *req;
5738         struct ceph_osd_req_op *op;
5739
5740         spin_lock(&o->o_requests_lock);
5741         req = lookup_request(&o->o_requests, le64_to_cpu(con->in_msg->hdr.tid));
5742         if (!req) {
5743                 spin_unlock(&o->o_requests_lock);
5744                 return -EBADR;
5745         }
5746
5747         if (o->o_sparse_op_idx < 0) {
5748                 dout("%s: [%d] starting new sparse read req\n",
5749                      __func__, o->o_osd);
5750         } else {
5751                 u64 end;
5752
5753                 op = &req->r_ops[o->o_sparse_op_idx];
5754
5755                 WARN_ON_ONCE(op->extent.sparse_ext);
5756
5757                 /* hand back buffer we took earlier */
5758                 op->extent.sparse_ext = sr->sr_extent;
5759                 sr->sr_extent = NULL;
5760                 op->extent.sparse_ext_cnt = sr->sr_count;
5761                 sr->sr_ext_len = 0;
5762                 dout("%s: [%d] completed extent array len %d cursor->resid %zd\n",
5763                      __func__, o->o_osd, op->extent.sparse_ext_cnt, cursor->resid);
5764                 /* Advance to end of data for this operation */
5765                 end = ceph_sparse_ext_map_end(op);
5766                 if (end < sr->sr_req_len)
5767                         advance_cursor(cursor, sr->sr_req_len - end, false);
5768         }
5769
5770         ceph_init_sparse_read(sr);
5771
5772         /* find next op in this request (if any) */
5773         while (++o->o_sparse_op_idx < req->r_num_ops) {
5774                 op = &req->r_ops[o->o_sparse_op_idx];
5775                 if (op->op == CEPH_OSD_OP_SPARSE_READ)
5776                         goto found;
5777         }
5778
5779         /* reset for next sparse read request */
5780         spin_unlock(&o->o_requests_lock);
5781         o->o_sparse_op_idx = -1;
5782         return 0;
5783 found:
5784         sr->sr_req_off = op->extent.offset;
5785         sr->sr_req_len = op->extent.length;
5786         sr->sr_pos = sr->sr_req_off;
5787         dout("%s: [%d] new sparse read op at idx %d 0x%llx~0x%llx\n", __func__,
5788              o->o_osd, o->o_sparse_op_idx, sr->sr_req_off, sr->sr_req_len);
5789
5790         /* hand off request's sparse extent map buffer */
5791         sr->sr_ext_len = op->extent.sparse_ext_cnt;
5792         op->extent.sparse_ext_cnt = 0;
5793         sr->sr_extent = op->extent.sparse_ext;
5794         op->extent.sparse_ext = NULL;
5795
5796         spin_unlock(&o->o_requests_lock);
5797         return 1;
5798 }
5799
5800 #ifdef __BIG_ENDIAN
5801 static inline void convert_extent_map(struct ceph_sparse_read *sr)
5802 {
5803         int i;
5804
5805         for (i = 0; i < sr->sr_count; i++) {
5806                 struct ceph_sparse_extent *ext = &sr->sr_extent[i];
5807
5808                 ext->off = le64_to_cpu((__force __le64)ext->off);
5809                 ext->len = le64_to_cpu((__force __le64)ext->len);
5810         }
5811 }
5812 #else
5813 static inline void convert_extent_map(struct ceph_sparse_read *sr)
5814 {
5815 }
5816 #endif
5817
5818 static int osd_sparse_read(struct ceph_connection *con,
5819                            struct ceph_msg_data_cursor *cursor,
5820                            char **pbuf)
5821 {
5822         struct ceph_osd *o = con->private;
5823         struct ceph_sparse_read *sr = &o->o_sparse_read;
5824         u32 count = sr->sr_count;
5825         u64 eoff, elen, len = 0;
5826         int i, ret;
5827
5828         switch (sr->sr_state) {
5829         case CEPH_SPARSE_READ_HDR:
5830 next_op:
5831                 ret = prep_next_sparse_read(con, cursor);
5832                 if (ret <= 0)
5833                         return ret;
5834
5835                 /* number of extents */
5836                 ret = sizeof(sr->sr_count);
5837                 *pbuf = (char *)&sr->sr_count;
5838                 sr->sr_state = CEPH_SPARSE_READ_EXTENTS;
5839                 break;
5840         case CEPH_SPARSE_READ_EXTENTS:
5841                 /* Convert sr_count to host-endian */
5842                 count = le32_to_cpu((__force __le32)sr->sr_count);
5843                 sr->sr_count = count;
5844                 dout("[%d] got %u extents\n", o->o_osd, count);
5845
5846                 if (count > 0) {
5847                         if (!sr->sr_extent || count > sr->sr_ext_len) {
5848                                 /* no extent array provided, or too short */
5849                                 kfree(sr->sr_extent);
5850                                 sr->sr_extent = kmalloc_array(count,
5851                                                               sizeof(*sr->sr_extent),
5852                                                               GFP_NOIO);
5853                                 if (!sr->sr_extent) {
5854                                         pr_err("%s: failed to allocate %u extents\n",
5855                                                __func__, count);
5856                                         return -ENOMEM;
5857                                 }
5858                                 sr->sr_ext_len = count;
5859                         }
5860                         ret = count * sizeof(*sr->sr_extent);
5861                         *pbuf = (char *)sr->sr_extent;
5862                         sr->sr_state = CEPH_SPARSE_READ_DATA_LEN;
5863                         break;
5864                 }
5865                 /* No extents? Read data len */
5866                 fallthrough;
5867         case CEPH_SPARSE_READ_DATA_LEN:
5868                 convert_extent_map(sr);
5869                 ret = sizeof(sr->sr_datalen);
5870                 *pbuf = (char *)&sr->sr_datalen;
5871                 sr->sr_state = CEPH_SPARSE_READ_DATA_PRE;
5872                 break;
5873         case CEPH_SPARSE_READ_DATA_PRE:
5874                 /* Convert sr_datalen to host-endian */
5875                 sr->sr_datalen = le32_to_cpu((__force __le32)sr->sr_datalen);
5876                 for (i = 0; i < count; i++)
5877                         len += sr->sr_extent[i].len;
5878                 if (sr->sr_datalen != len) {
5879                         pr_warn_ratelimited("data len %u != extent len %llu\n",
5880                                             sr->sr_datalen, len);
5881                         return -EREMOTEIO;
5882                 }
5883                 sr->sr_state = CEPH_SPARSE_READ_DATA;
5884                 fallthrough;
5885         case CEPH_SPARSE_READ_DATA:
5886                 if (sr->sr_index >= count) {
5887                         sr->sr_state = CEPH_SPARSE_READ_HDR;
5888                         goto next_op;
5889                 }
5890
5891                 eoff = sr->sr_extent[sr->sr_index].off;
5892                 elen = sr->sr_extent[sr->sr_index].len;
5893
5894                 dout("[%d] ext %d off 0x%llx len 0x%llx\n",
5895                      o->o_osd, sr->sr_index, eoff, elen);
5896
5897                 if (elen > INT_MAX) {
5898                         dout("Sparse read extent length too long (0x%llx)\n",
5899                              elen);
5900                         return -EREMOTEIO;
5901                 }
5902
5903                 /* zero out anything from sr_pos to start of extent */
5904                 if (sr->sr_pos < eoff)
5905                         advance_cursor(cursor, eoff - sr->sr_pos, true);
5906
5907                 /* Set position to end of extent */
5908                 sr->sr_pos = eoff + elen;
5909
5910                 /* send back the new length and nullify the ptr */
5911                 cursor->sr_resid = elen;
5912                 ret = elen;
5913                 *pbuf = NULL;
5914
5915                 /* Bump the array index */
5916                 ++sr->sr_index;
5917                 break;
5918         }
5919         return ret;
5920 }
5921
5922 static const struct ceph_connection_operations osd_con_ops = {
5923         .get = osd_get_con,
5924         .put = osd_put_con,
5925         .sparse_read = osd_sparse_read,
5926         .alloc_msg = osd_alloc_msg,
5927         .dispatch = osd_dispatch,
5928         .fault = osd_fault,
5929         .reencode_message = osd_reencode_message,
5930         .get_authorizer = osd_get_authorizer,
5931         .add_authorizer_challenge = osd_add_authorizer_challenge,
5932         .verify_authorizer_reply = osd_verify_authorizer_reply,
5933         .invalidate_authorizer = osd_invalidate_authorizer,
5934         .sign_message = osd_sign_message,
5935         .check_message_signature = osd_check_message_signature,
5936         .get_auth_request = osd_get_auth_request,
5937         .handle_auth_reply_more = osd_handle_auth_reply_more,
5938         .handle_auth_done = osd_handle_auth_done,
5939         .handle_auth_bad_method = osd_handle_auth_bad_method,
5940 };