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 27static struct kmem_cache *ceph_osd_request_cache; 28 29static 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 49static void link_request(struct ceph_osd *osd, struct ceph_osd_request *req); 50static void unlink_request(struct ceph_osd *osd, struct ceph_osd_request *req); 51static void link_linger(struct ceph_osd *osd, 52 struct ceph_osd_linger_request *lreq); 53static void unlink_linger(struct ceph_osd *osd, 54 struct ceph_osd_linger_request *lreq); 55static void clear_backoffs(struct ceph_osd *osd); 56 57#if 1 58static 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} 69static inline void verify_osdc_locked(struct ceph_osd_client *osdc) 70{ 71 WARN_ON(!rwsem_is_locked(&osdc->lock)); 72} 73static inline void verify_osdc_wrlocked(struct ceph_osd_client *osdc) 74{ 75 WARN_ON(!rwsem_is_wrlocked(&osdc->lock)); 76} 77static 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} 85static inline void verify_lreq_locked(struct ceph_osd_linger_request *lreq) 86{ 87 WARN_ON(!mutex_is_locked(&lreq->lock)); 88} 89#else 90static inline void verify_osdc_locked(struct ceph_osd_client *osdc) { } 91static inline void verify_osdc_wrlocked(struct ceph_osd_client *osdc) { } 92static inline void verify_osd_locked(struct ceph_osd *osd) { } 93static 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 */ 103static 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 123static 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 */ 132static 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 */ 147static 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 155static 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 165static 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 174static 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 181static struct ceph_osd_data * 182osd_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 189struct ceph_osd_data * 190osd_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} 195EXPORT_SYMBOL(osd_req_op_extent_osd_data); 196 197void 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} 208EXPORT_SYMBOL(osd_req_op_raw_data_in_pages); 209 210void 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} 221EXPORT_SYMBOL(osd_req_op_extent_osd_data_pages); 222 223void 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} 231EXPORT_SYMBOL(osd_req_op_extent_osd_data_pagelist); 232 233#ifdef CONFIG_BLOCK 234void 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} 244EXPORT_SYMBOL(osd_req_op_extent_osd_data_bio); 245#endif /* CONFIG_BLOCK */ 246 247void 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} 261EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvecs); 262 263void 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} 272EXPORT_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 */ 280void 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} 288EXPORT_SYMBOL(osd_req_op_extent_osd_iter); 289 290static 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 300void 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} 311EXPORT_SYMBOL(osd_req_op_cls_request_data_pagelist); 312 313void 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} 325EXPORT_SYMBOL(osd_req_op_cls_request_data_pages); 326 327void 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} 343EXPORT_SYMBOL(osd_req_op_cls_request_data_bvecs); 344 345void 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} 355EXPORT_SYMBOL(osd_req_op_cls_response_data_pages); 356 357static 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 380static 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 394static 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 */ 443static 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 458static 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 487static 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 */ 498static 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 506static 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 535void 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} 541EXPORT_SYMBOL(ceph_osdc_get_request); 542 543void 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} 551EXPORT_SYMBOL(ceph_osdc_put_request); 552 553static 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 567struct 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} 597EXPORT_SYMBOL(ceph_osdc_alloc_request); 598 599static 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 604static 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 664static 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 675static 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 */ 725int 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} 733EXPORT_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 */ 740struct ceph_osd_req_op * 741osd_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} 756EXPORT_SYMBOL(osd_req_op_init); 757 758void 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} 780EXPORT_SYMBOL(osd_req_op_extent_init); 781 782void 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} 800EXPORT_SYMBOL(osd_req_op_extent_update); 801 802void 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} 822EXPORT_SYMBOL(osd_req_op_extent_dup_last); 823 824int 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 861err_pagelist_free: 862 ceph_pagelist_release(pagelist); 863 return ret; 864} 865EXPORT_SYMBOL(osd_req_op_cls_init); 866 867int 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 902err_pagelist_free: 903 ceph_pagelist_release(pagelist); 904 return ret; 905} 906EXPORT_SYMBOL(osd_req_op_xattr_init); 907 908/* 909 * @watch_opcode: CEPH_OSD_WATCH_OP_* 910 */ 911static 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 */ 926static 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 */ 941void 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} 961EXPORT_SYMBOL(osd_req_op_alloc_hint_init); 962 963static 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 989static 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 */ 1078struct 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 1168fail: 1169 ceph_osdc_put_request(req); 1170 return ERR_PTR(r); 1171} 1172EXPORT_SYMBOL(ceph_osdc_new_request); 1173 1174int __ceph_alloc_sparse_ext_map(struct ceph_osd_req_op *op, int cnt) 1175{ 1176 op->extent.sparse_ext_cnt = cnt; 1177 op->extent.sparse_ext = kmalloc_array(cnt, 1178 sizeof(*op->extent.sparse_ext), 1179 GFP_NOFS); 1180 if (!op->extent.sparse_ext) 1181 return -ENOMEM; 1182 return 0; 1183} 1184EXPORT_SYMBOL(__ceph_alloc_sparse_ext_map); 1185 1186/* 1187 * We keep osd requests in an rbtree, sorted by ->r_tid. 1188 */ 1189DEFINE_RB_FUNCS(request, struct ceph_osd_request, r_tid, r_node) 1190DEFINE_RB_FUNCS(request_mc, struct ceph_osd_request, r_tid, r_mc_node) 1191 1192/* 1193 * Call @fn on each OSD request as long as @fn returns 0. 1194 */ 1195static void for_each_request(struct ceph_osd_client *osdc, 1196 int (*fn)(struct ceph_osd_request *req, void *arg), 1197 void *arg) 1198{ 1199 struct rb_node *n, *p; 1200 1201 for (n = rb_first(&osdc->osds); n; n = rb_next(n)) { 1202 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node); 1203 1204 for (p = rb_first(&osd->o_requests); p; ) { 1205 struct ceph_osd_request *req = 1206 rb_entry(p, struct ceph_osd_request, r_node); 1207 1208 p = rb_next(p); 1209 if (fn(req, arg)) 1210 return; 1211 } 1212 } 1213 1214 for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) { 1215 struct ceph_osd_request *req = 1216 rb_entry(p, struct ceph_osd_request, r_node); 1217 1218 p = rb_next(p); 1219 if (fn(req, arg)) 1220 return; 1221 } 1222} 1223 1224static bool osd_homeless(struct ceph_osd *osd) 1225{ 1226 return osd->o_osd == CEPH_HOMELESS_OSD; 1227} 1228 1229static bool osd_registered(struct ceph_osd *osd) 1230{ 1231 verify_osdc_locked(osd->o_osdc); 1232 1233 return !RB_EMPTY_NODE(&osd->o_node); 1234} 1235 1236/* 1237 * Assumes @osd is zero-initialized. 1238 */ 1239static void osd_init(struct ceph_osd *osd) 1240{ 1241 refcount_set(&osd->o_ref, 1); 1242 RB_CLEAR_NODE(&osd->o_node); 1243 spin_lock_init(&osd->o_requests_lock); 1244 osd->o_requests = RB_ROOT; 1245 osd->o_linger_requests = RB_ROOT; 1246 osd->o_backoff_mappings = RB_ROOT; 1247 osd->o_backoffs_by_id = RB_ROOT; 1248 INIT_LIST_HEAD(&osd->o_osd_lru); 1249 INIT_LIST_HEAD(&osd->o_keepalive_item); 1250 osd->o_incarnation = 1; 1251 mutex_init(&osd->lock); 1252} 1253 1254static void ceph_init_sparse_read(struct ceph_sparse_read *sr) 1255{ 1256 kfree(sr->sr_extent); 1257 memset(sr, '\0', sizeof(*sr)); 1258 sr->sr_state = CEPH_SPARSE_READ_HDR; 1259} 1260 1261static void osd_cleanup(struct ceph_osd *osd) 1262{ 1263 WARN_ON(!RB_EMPTY_NODE(&osd->o_node)); 1264 WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests)); 1265 WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests)); 1266 WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoff_mappings)); 1267 WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoffs_by_id)); 1268 WARN_ON(!list_empty(&osd->o_osd_lru)); 1269 WARN_ON(!list_empty(&osd->o_keepalive_item)); 1270 1271 ceph_init_sparse_read(&osd->o_sparse_read); 1272 1273 if (osd->o_auth.authorizer) { 1274 WARN_ON(osd_homeless(osd)); 1275 ceph_auth_destroy_authorizer(osd->o_auth.authorizer); 1276 } 1277} 1278 1279/* 1280 * Track open sessions with osds. 1281 */ 1282static struct ceph_osd *create_osd(struct ceph_osd_client *osdc, int onum) 1283{ 1284 struct ceph_osd *osd; 1285 1286 WARN_ON(onum == CEPH_HOMELESS_OSD); 1287 1288 osd = kzalloc(sizeof(*osd), GFP_NOIO | __GFP_NOFAIL); 1289 osd_init(osd); 1290 osd->o_osdc = osdc; 1291 osd->o_osd = onum; 1292 osd->o_sparse_op_idx = -1; 1293 1294 ceph_init_sparse_read(&osd->o_sparse_read); 1295 1296 ceph_con_init(&osd->o_con, osd, &osd_con_ops, &osdc->client->msgr); 1297 1298 return osd; 1299} 1300 1301static struct ceph_osd *get_osd(struct ceph_osd *osd) 1302{ 1303 if (refcount_inc_not_zero(&osd->o_ref)) { 1304 dout("get_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref)-1, 1305 refcount_read(&osd->o_ref)); 1306 return osd; 1307 } else { 1308 dout("get_osd %p FAIL\n", osd); 1309 return NULL; 1310 } 1311} 1312 1313static void put_osd(struct ceph_osd *osd) 1314{ 1315 dout("put_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref), 1316 refcount_read(&osd->o_ref) - 1); 1317 if (refcount_dec_and_test(&osd->o_ref)) { 1318 osd_cleanup(osd); 1319 kfree(osd); 1320 } 1321} 1322 1323DEFINE_RB_FUNCS(osd, struct ceph_osd, o_osd, o_node) 1324 1325static void __move_osd_to_lru(struct ceph_osd *osd) 1326{ 1327 struct ceph_osd_client *osdc = osd->o_osdc; 1328 1329 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd); 1330 BUG_ON(!list_empty(&osd->o_osd_lru)); 1331 1332 spin_lock(&osdc->osd_lru_lock); 1333 list_add_tail(&osd->o_osd_lru, &osdc->osd_lru); 1334 spin_unlock(&osdc->osd_lru_lock); 1335 1336 osd->lru_ttl = jiffies + osdc->client->options->osd_idle_ttl; 1337} 1338 1339static void maybe_move_osd_to_lru(struct ceph_osd *osd) 1340{ 1341 if (RB_EMPTY_ROOT(&osd->o_requests) && 1342 RB_EMPTY_ROOT(&osd->o_linger_requests)) 1343 __move_osd_to_lru(osd); 1344} 1345 1346static void __remove_osd_from_lru(struct ceph_osd *osd) 1347{ 1348 struct ceph_osd_client *osdc = osd->o_osdc; 1349 1350 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd); 1351 1352 spin_lock(&osdc->osd_lru_lock); 1353 if (!list_empty(&osd->o_osd_lru)) 1354 list_del_init(&osd->o_osd_lru); 1355 spin_unlock(&osdc->osd_lru_lock); 1356} 1357 1358/* 1359 * Close the connection and assign any leftover requests to the 1360 * homeless session. 1361 */ 1362static void close_osd(struct ceph_osd *osd) 1363{ 1364 struct ceph_osd_client *osdc = osd->o_osdc; 1365 struct rb_node *n; 1366 1367 verify_osdc_wrlocked(osdc); 1368 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd); 1369 1370 ceph_con_close(&osd->o_con); 1371 1372 for (n = rb_first(&osd->o_requests); n; ) { 1373 struct ceph_osd_request *req = 1374 rb_entry(n, struct ceph_osd_request, r_node); 1375 1376 n = rb_next(n); /* unlink_request() */ 1377 1378 dout(" reassigning req %p tid %llu\n", req, req->r_tid); 1379 unlink_request(osd, req); 1380 link_request(&osdc->homeless_osd, req); 1381 } 1382 for (n = rb_first(&osd->o_linger_requests); n; ) { 1383 struct ceph_osd_linger_request *lreq = 1384 rb_entry(n, struct ceph_osd_linger_request, node); 1385 1386 n = rb_next(n); /* unlink_linger() */ 1387 1388 dout(" reassigning lreq %p linger_id %llu\n", lreq, 1389 lreq->linger_id); 1390 unlink_linger(osd, lreq); 1391 link_linger(&osdc->homeless_osd, lreq); 1392 } 1393 clear_backoffs(osd); 1394 1395 __remove_osd_from_lru(osd); 1396 erase_osd(&osdc->osds, osd); 1397 put_osd(osd); 1398} 1399 1400/* 1401 * reset osd connect 1402 */ 1403static int reopen_osd(struct ceph_osd *osd) 1404{ 1405 struct ceph_entity_addr *peer_addr; 1406 1407 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd); 1408 1409 if (RB_EMPTY_ROOT(&osd->o_requests) && 1410 RB_EMPTY_ROOT(&osd->o_linger_requests)) { 1411 close_osd(osd); 1412 return -ENODEV; 1413 } 1414 1415 peer_addr = &osd->o_osdc->osdmap->osd_addr[osd->o_osd]; 1416 if (!memcmp(peer_addr, &osd->o_con.peer_addr, sizeof (*peer_addr)) && 1417 !ceph_con_opened(&osd->o_con)) { 1418 struct rb_node *n; 1419 1420 dout("osd addr hasn't changed and connection never opened, " 1421 "letting msgr retry\n"); 1422 /* touch each r_stamp for handle_timeout()'s benfit */ 1423 for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) { 1424 struct ceph_osd_request *req = 1425 rb_entry(n, struct ceph_osd_request, r_node); 1426 req->r_stamp = jiffies; 1427 } 1428 1429 return -EAGAIN; 1430 } 1431 1432 ceph_con_close(&osd->o_con); 1433 ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd, peer_addr); 1434 osd->o_incarnation++; 1435 1436 return 0; 1437} 1438 1439static struct ceph_osd *lookup_create_osd(struct ceph_osd_client *osdc, int o, 1440 bool wrlocked) 1441{ 1442 struct ceph_osd *osd; 1443 1444 if (wrlocked) 1445 verify_osdc_wrlocked(osdc); 1446 else 1447 verify_osdc_locked(osdc); 1448 1449 if (o != CEPH_HOMELESS_OSD) 1450 osd = lookup_osd(&osdc->osds, o); 1451 else 1452 osd = &osdc->homeless_osd; 1453 if (!osd) { 1454 if (!wrlocked) 1455 return ERR_PTR(-EAGAIN); 1456 1457 osd = create_osd(osdc, o); 1458 insert_osd(&osdc->osds, osd); 1459 ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd, 1460 &osdc->osdmap->osd_addr[osd->o_osd]); 1461 } 1462 1463 dout("%s osdc %p osd%d -> osd %p\n", __func__, osdc, o, osd); 1464 return osd; 1465} 1466 1467/* 1468 * Create request <-> OSD session relation. 1469 * 1470 * @req has to be assigned a tid, @osd may be homeless. 1471 */ 1472static void link_request(struct ceph_osd *osd, struct ceph_osd_request *req) 1473{ 1474 verify_osd_locked(osd); 1475 WARN_ON(!req->r_tid || req->r_osd); 1476 dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd, 1477 req, req->r_tid); 1478 1479 if (!osd_homeless(osd)) 1480 __remove_osd_from_lru(osd); 1481 else 1482 atomic_inc(&osd->o_osdc->num_homeless); 1483 1484 get_osd(osd); 1485 spin_lock(&osd->o_requests_lock); 1486 insert_request(&osd->o_requests, req); 1487 spin_unlock(&osd->o_requests_lock); 1488 req->r_osd = osd; 1489} 1490 1491static void unlink_request(struct ceph_osd *osd, struct ceph_osd_request *req) 1492{ 1493 verify_osd_locked(osd); 1494 WARN_ON(req->r_osd != osd); 1495 dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd, 1496 req, req->r_tid); 1497 1498 req->r_osd = NULL; 1499 spin_lock(&osd->o_requests_lock); 1500 erase_request(&osd->o_requests, req); 1501 spin_unlock(&osd->o_requests_lock); 1502 put_osd(osd); 1503 1504 if (!osd_homeless(osd)) 1505 maybe_move_osd_to_lru(osd); 1506 else 1507 atomic_dec(&osd->o_osdc->num_homeless); 1508} 1509 1510static bool __pool_full(struct ceph_pg_pool_info *pi) 1511{ 1512 return pi->flags & CEPH_POOL_FLAG_FULL; 1513} 1514 1515static bool have_pool_full(struct ceph_osd_client *osdc) 1516{ 1517 struct rb_node *n; 1518 1519 for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) { 1520 struct ceph_pg_pool_info *pi = 1521 rb_entry(n, struct ceph_pg_pool_info, node); 1522 1523 if (__pool_full(pi)) 1524 return true; 1525 } 1526 1527 return false; 1528} 1529 1530static bool pool_full(struct ceph_osd_client *osdc, s64 pool_id) 1531{ 1532 struct ceph_pg_pool_info *pi; 1533 1534 pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id); 1535 if (!pi) 1536 return false; 1537 1538 return __pool_full(pi); 1539} 1540 1541/* 1542 * Returns whether a request should be blocked from being sent 1543 * based on the current osdmap and osd_client settings. 1544 */ 1545static bool target_should_be_paused(struct ceph_osd_client *osdc, 1546 const struct ceph_osd_request_target *t, 1547 struct ceph_pg_pool_info *pi) 1548{ 1549 bool pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD); 1550 bool pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) || 1551 ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || 1552 __pool_full(pi); 1553 1554 WARN_ON(pi->id != t->target_oloc.pool); 1555 return ((t->flags & CEPH_OSD_FLAG_READ) && pauserd) || 1556 ((t->flags & CEPH_OSD_FLAG_WRITE) && pausewr) || 1557 (osdc->osdmap->epoch < osdc->epoch_barrier); 1558} 1559 1560static int pick_random_replica(const struct ceph_osds *acting) 1561{ 1562 int i = get_random_u32_below(acting->size); 1563 1564 dout("%s picked osd%d, primary osd%d\n", __func__, 1565 acting->osds[i], acting->primary); 1566 return i; 1567} 1568 1569/* 1570 * Picks the closest replica based on client's location given by 1571 * crush_location option. Prefers the primary if the locality is 1572 * the same. 1573 */ 1574static int pick_closest_replica(struct ceph_osd_client *osdc, 1575 const struct ceph_osds *acting) 1576{ 1577 struct ceph_options *opt = osdc->client->options; 1578 int best_i, best_locality; 1579 int i = 0, locality; 1580 1581 do { 1582 locality = ceph_get_crush_locality(osdc->osdmap, 1583 acting->osds[i], 1584 &opt->crush_locs); 1585 if (i == 0 || 1586 (locality >= 0 && best_locality < 0) || 1587 (locality >= 0 && best_locality >= 0 && 1588 locality < best_locality)) { 1589 best_i = i; 1590 best_locality = locality; 1591 } 1592 } while (++i < acting->size); 1593 1594 dout("%s picked osd%d with locality %d, primary osd%d\n", __func__, 1595 acting->osds[best_i], best_locality, acting->primary); 1596 return best_i; 1597} 1598 1599enum calc_target_result { 1600 CALC_TARGET_NO_ACTION = 0, 1601 CALC_TARGET_NEED_RESEND, 1602 CALC_TARGET_POOL_DNE, 1603}; 1604 1605static enum calc_target_result calc_target(struct ceph_osd_client *osdc, 1606 struct ceph_osd_request_target *t, 1607 bool any_change) 1608{ 1609 struct ceph_pg_pool_info *pi; 1610 struct ceph_pg pgid, last_pgid; 1611 struct ceph_osds up, acting; 1612 bool is_read = t->flags & CEPH_OSD_FLAG_READ; 1613 bool is_write = t->flags & CEPH_OSD_FLAG_WRITE; 1614 bool force_resend = false; 1615 bool unpaused = false; 1616 bool legacy_change = false; 1617 bool split = false; 1618 bool sort_bitwise = ceph_osdmap_flag(osdc, CEPH_OSDMAP_SORTBITWISE); 1619 bool recovery_deletes = ceph_osdmap_flag(osdc, 1620 CEPH_OSDMAP_RECOVERY_DELETES); 1621 enum calc_target_result ct_res; 1622 1623 t->epoch = osdc->osdmap->epoch; 1624 pi = ceph_pg_pool_by_id(osdc->osdmap, t->base_oloc.pool); 1625 if (!pi) { 1626 t->osd = CEPH_HOMELESS_OSD; 1627 ct_res = CALC_TARGET_POOL_DNE; 1628 goto out; 1629 } 1630 1631 if (osdc->osdmap->epoch == pi->last_force_request_resend) { 1632 if (t->last_force_resend < pi->last_force_request_resend) { 1633 t->last_force_resend = pi->last_force_request_resend; 1634 force_resend = true; 1635 } else if (t->last_force_resend == 0) { 1636 force_resend = true; 1637 } 1638 } 1639 1640 /* apply tiering */ 1641 ceph_oid_copy(&t->target_oid, &t->base_oid); 1642 ceph_oloc_copy(&t->target_oloc, &t->base_oloc); 1643 if ((t->flags & CEPH_OSD_FLAG_IGNORE_OVERLAY) == 0) { 1644 if (is_read && pi->read_tier >= 0) 1645 t->target_oloc.pool = pi->read_tier; 1646 if (is_write && pi->write_tier >= 0) 1647 t->target_oloc.pool = pi->write_tier; 1648 1649 pi = ceph_pg_pool_by_id(osdc->osdmap, t->target_oloc.pool); 1650 if (!pi) { 1651 t->osd = CEPH_HOMELESS_OSD; 1652 ct_res = CALC_TARGET_POOL_DNE; 1653 goto out; 1654 } 1655 } 1656 1657 __ceph_object_locator_to_pg(pi, &t->target_oid, &t->target_oloc, &pgid); 1658 last_pgid.pool = pgid.pool; 1659 last_pgid.seed = ceph_stable_mod(pgid.seed, t->pg_num, t->pg_num_mask); 1660 1661 ceph_pg_to_up_acting_osds(osdc->osdmap, pi, &pgid, &up, &acting); 1662 if (any_change && 1663 ceph_is_new_interval(&t->acting, 1664 &acting, 1665 &t->up, 1666 &up, 1667 t->size, 1668 pi->size, 1669 t->min_size, 1670 pi->min_size, 1671 t->pg_num, 1672 pi->pg_num, 1673 t->sort_bitwise, 1674 sort_bitwise, 1675 t->recovery_deletes, 1676 recovery_deletes, 1677 &last_pgid)) 1678 force_resend = true; 1679 1680 if (t->paused && !target_should_be_paused(osdc, t, pi)) { 1681 t->paused = false; 1682 unpaused = true; 1683 } 1684 legacy_change = ceph_pg_compare(&t->pgid, &pgid) || 1685 ceph_osds_changed(&t->acting, &acting, 1686 t->used_replica || any_change); 1687 if (t->pg_num) 1688 split = ceph_pg_is_split(&last_pgid, t->pg_num, pi->pg_num); 1689 1690 if (legacy_change || force_resend || split) { 1691 t->pgid = pgid; /* struct */ 1692 ceph_pg_to_primary_shard(osdc->osdmap, pi, &pgid, &t->spgid); 1693 ceph_osds_copy(&t->acting, &acting); 1694 ceph_osds_copy(&t->up, &up); 1695 t->size = pi->size; 1696 t->min_size = pi->min_size; 1697 t->pg_num = pi->pg_num; 1698 t->pg_num_mask = pi->pg_num_mask; 1699 t->sort_bitwise = sort_bitwise; 1700 t->recovery_deletes = recovery_deletes; 1701 1702 if ((t->flags & (CEPH_OSD_FLAG_BALANCE_READS | 1703 CEPH_OSD_FLAG_LOCALIZE_READS)) && 1704 !is_write && pi->type == CEPH_POOL_TYPE_REP && 1705 acting.size > 1) { 1706 int pos; 1707 1708 WARN_ON(!is_read || acting.osds[0] != acting.primary); 1709 if (t->flags & CEPH_OSD_FLAG_BALANCE_READS) { 1710 pos = pick_random_replica(&acting); 1711 } else { 1712 pos = pick_closest_replica(osdc, &acting); 1713 } 1714 t->osd = acting.osds[pos]; 1715 t->used_replica = pos > 0; 1716 } else { 1717 t->osd = acting.primary; 1718 t->used_replica = false; 1719 } 1720 } 1721 1722 if (unpaused || legacy_change || force_resend || split) 1723 ct_res = CALC_TARGET_NEED_RESEND; 1724 else 1725 ct_res = CALC_TARGET_NO_ACTION; 1726 1727out: 1728 dout("%s t %p -> %d%d%d%d ct_res %d osd%d\n", __func__, t, unpaused, 1729 legacy_change, force_resend, split, ct_res, t->osd); 1730 return ct_res; 1731} 1732 1733static struct ceph_spg_mapping *alloc_spg_mapping(void) 1734{ 1735 struct ceph_spg_mapping *spg; 1736 1737 spg = kmalloc(sizeof(*spg), GFP_NOIO); 1738 if (!spg) 1739 return NULL; 1740 1741 RB_CLEAR_NODE(&spg->node); 1742 spg->backoffs = RB_ROOT; 1743 return spg; 1744} 1745 1746static void free_spg_mapping(struct ceph_spg_mapping *spg) 1747{ 1748 WARN_ON(!RB_EMPTY_NODE(&spg->node)); 1749 WARN_ON(!RB_EMPTY_ROOT(&spg->backoffs)); 1750 1751 kfree(spg); 1752} 1753 1754/* 1755 * rbtree of ceph_spg_mapping for handling map<spg_t, ...>, similar to 1756 * ceph_pg_mapping. Used to track OSD backoffs -- a backoff [range] is 1757 * defined only within a specific spgid; it does not pass anything to 1758 * children on split, or to another primary. 1759 */ 1760DEFINE_RB_FUNCS2(spg_mapping, struct ceph_spg_mapping, spgid, ceph_spg_compare, 1761 RB_BYPTR, const struct ceph_spg *, node) 1762 1763static u64 hoid_get_bitwise_key(const struct ceph_hobject_id *hoid) 1764{ 1765 return hoid->is_max ? 0x100000000ull : hoid->hash_reverse_bits; 1766} 1767 1768static void hoid_get_effective_key(const struct ceph_hobject_id *hoid, 1769 void **pkey, size_t *pkey_len) 1770{ 1771 if (hoid->key_len) { 1772 *pkey = hoid->key; 1773 *pkey_len = hoid->key_len; 1774 } else { 1775 *pkey = hoid->oid; 1776 *pkey_len = hoid->oid_len; 1777 } 1778} 1779 1780static int compare_names(const void *name1, size_t name1_len, 1781 const void *name2, size_t name2_len) 1782{ 1783 int ret; 1784 1785 ret = memcmp(name1, name2, min(name1_len, name2_len)); 1786 if (!ret) { 1787 if (name1_len < name2_len) 1788 ret = -1; 1789 else if (name1_len > name2_len) 1790 ret = 1; 1791 } 1792 return ret; 1793} 1794 1795static int hoid_compare(const struct ceph_hobject_id *lhs, 1796 const struct ceph_hobject_id *rhs) 1797{ 1798 void *effective_key1, *effective_key2; 1799 size_t effective_key1_len, effective_key2_len; 1800 int ret; 1801 1802 if (lhs->is_max < rhs->is_max) 1803 return -1; 1804 if (lhs->is_max > rhs->is_max) 1805 return 1; 1806 1807 if (lhs->pool < rhs->pool) 1808 return -1; 1809 if (lhs->pool > rhs->pool) 1810 return 1; 1811 1812 if (hoid_get_bitwise_key(lhs) < hoid_get_bitwise_key(rhs)) 1813 return -1; 1814 if (hoid_get_bitwise_key(lhs) > hoid_get_bitwise_key(rhs)) 1815 return 1; 1816 1817 ret = compare_names(lhs->nspace, lhs->nspace_len, 1818 rhs->nspace, rhs->nspace_len); 1819 if (ret) 1820 return ret; 1821 1822 hoid_get_effective_key(lhs, &effective_key1, &effective_key1_len); 1823 hoid_get_effective_key(rhs, &effective_key2, &effective_key2_len); 1824 ret = compare_names(effective_key1, effective_key1_len, 1825 effective_key2, effective_key2_len); 1826 if (ret) 1827 return ret; 1828 1829 ret = compare_names(lhs->oid, lhs->oid_len, rhs->oid, rhs->oid_len); 1830 if (ret) 1831 return ret; 1832 1833 if (lhs->snapid < rhs->snapid) 1834 return -1; 1835 if (lhs->snapid > rhs->snapid) 1836 return 1; 1837 1838 return 0; 1839} 1840 1841/* 1842 * For decoding ->begin and ->end of MOSDBackoff only -- no MIN/MAX 1843 * compat stuff here. 1844 * 1845 * Assumes @hoid is zero-initialized. 1846 */ 1847static int decode_hoid(void **p, void *end, struct ceph_hobject_id *hoid) 1848{ 1849 u8 struct_v; 1850 u32 struct_len; 1851 int ret; 1852 1853 ret = ceph_start_decoding(p, end, 4, "hobject_t", &struct_v, 1854 &struct_len); 1855 if (ret) 1856 return ret; 1857 1858 if (struct_v < 4) { 1859 pr_err("got struct_v %d < 4 of hobject_t\n", struct_v); 1860 goto e_inval; 1861 } 1862 1863 hoid->key = ceph_extract_encoded_string(p, end, &hoid->key_len, 1864 GFP_NOIO); 1865 if (IS_ERR(hoid->key)) { 1866 ret = PTR_ERR(hoid->key); 1867 hoid->key = NULL; 1868 return ret; 1869 } 1870 1871 hoid->oid = ceph_extract_encoded_string(p, end, &hoid->oid_len, 1872 GFP_NOIO); 1873 if (IS_ERR(hoid->oid)) { 1874 ret = PTR_ERR(hoid->oid); 1875 hoid->oid = NULL; 1876 return ret; 1877 } 1878 1879 ceph_decode_64_safe(p, end, hoid->snapid, e_inval); 1880 ceph_decode_32_safe(p, end, hoid->hash, e_inval); 1881 ceph_decode_8_safe(p, end, hoid->is_max, e_inval); 1882 1883 hoid->nspace = ceph_extract_encoded_string(p, end, &hoid->nspace_len, 1884 GFP_NOIO); 1885 if (IS_ERR(hoid->nspace)) { 1886 ret = PTR_ERR(hoid->nspace); 1887 hoid->nspace = NULL; 1888 return ret; 1889 } 1890 1891 ceph_decode_64_safe(p, end, hoid->pool, e_inval); 1892 1893 ceph_hoid_build_hash_cache(hoid); 1894 return 0; 1895 1896e_inval: 1897 return -EINVAL; 1898} 1899 1900static int hoid_encoding_size(const struct ceph_hobject_id *hoid) 1901{ 1902 return 8 + 4 + 1 + 8 + /* snapid, hash, is_max, pool */ 1903 4 + hoid->key_len + 4 + hoid->oid_len + 4 + hoid->nspace_len; 1904} 1905 1906static void encode_hoid(void **p, void *end, const struct ceph_hobject_id *hoid) 1907{ 1908 ceph_start_encoding(p, 4, 3, hoid_encoding_size(hoid)); 1909 ceph_encode_string(p, end, hoid->key, hoid->key_len); 1910 ceph_encode_string(p, end, hoid->oid, hoid->oid_len); 1911 ceph_encode_64(p, hoid->snapid); 1912 ceph_encode_32(p, hoid->hash); 1913 ceph_encode_8(p, hoid->is_max); 1914 ceph_encode_string(p, end, hoid->nspace, hoid->nspace_len); 1915 ceph_encode_64(p, hoid->pool); 1916} 1917 1918static void free_hoid(struct ceph_hobject_id *hoid) 1919{ 1920 if (hoid) { 1921 kfree(hoid->key); 1922 kfree(hoid->oid); 1923 kfree(hoid->nspace); 1924 kfree(hoid); 1925 } 1926} 1927 1928static struct ceph_osd_backoff *alloc_backoff(void) 1929{ 1930 struct ceph_osd_backoff *backoff; 1931 1932 backoff = kzalloc(sizeof(*backoff), GFP_NOIO); 1933 if (!backoff) 1934 return NULL; 1935 1936 RB_CLEAR_NODE(&backoff->spg_node); 1937 RB_CLEAR_NODE(&backoff->id_node); 1938 return backoff; 1939} 1940 1941static void free_backoff(struct ceph_osd_backoff *backoff) 1942{ 1943 WARN_ON(!RB_EMPTY_NODE(&backoff->spg_node)); 1944 WARN_ON(!RB_EMPTY_NODE(&backoff->id_node)); 1945 1946 free_hoid(backoff->begin); 1947 free_hoid(backoff->end); 1948 kfree(backoff); 1949} 1950 1951/* 1952 * Within a specific spgid, backoffs are managed by ->begin hoid. 1953 */ 1954DEFINE_RB_INSDEL_FUNCS2(backoff, struct ceph_osd_backoff, begin, hoid_compare, 1955 RB_BYVAL, spg_node); 1956 1957static struct ceph_osd_backoff *lookup_containing_backoff(struct rb_root *root, 1958 const struct ceph_hobject_id *hoid) 1959{ 1960 struct rb_node *n = root->rb_node; 1961 1962 while (n) { 1963 struct ceph_osd_backoff *cur = 1964 rb_entry(n, struct ceph_osd_backoff, spg_node); 1965 int cmp; 1966 1967 cmp = hoid_compare(hoid, cur->begin); 1968 if (cmp < 0) { 1969 n = n->rb_left; 1970 } else if (cmp > 0) { 1971 if (hoid_compare(hoid, cur->end) < 0) 1972 return cur; 1973 1974 n = n->rb_right; 1975 } else { 1976 return cur; 1977 } 1978 } 1979 1980 return NULL; 1981} 1982 1983/* 1984 * Each backoff has a unique id within its OSD session. 1985 */ 1986DEFINE_RB_FUNCS(backoff_by_id, struct ceph_osd_backoff, id, id_node) 1987 1988static void clear_backoffs(struct ceph_osd *osd) 1989{ 1990 while (!RB_EMPTY_ROOT(&osd->o_backoff_mappings)) { 1991 struct ceph_spg_mapping *spg = 1992 rb_entry(rb_first(&osd->o_backoff_mappings), 1993 struct ceph_spg_mapping, node); 1994 1995 while (!RB_EMPTY_ROOT(&spg->backoffs)) { 1996 struct ceph_osd_backoff *backoff = 1997 rb_entry(rb_first(&spg->backoffs), 1998 struct ceph_osd_backoff, spg_node); 1999 2000 erase_backoff(&spg->backoffs, backoff); 2001 erase_backoff_by_id(&osd->o_backoffs_by_id, backoff); 2002 free_backoff(backoff); 2003 } 2004 erase_spg_mapping(&osd->o_backoff_mappings, spg); 2005 free_spg_mapping(spg); 2006 } 2007} 2008 2009/* 2010 * Set up a temporary, non-owning view into @t. 2011 */ 2012static void hoid_fill_from_target(struct ceph_hobject_id *hoid, 2013 const struct ceph_osd_request_target *t) 2014{ 2015 hoid->key = NULL; 2016 hoid->key_len = 0; 2017 hoid->oid = t->target_oid.name; 2018 hoid->oid_len = t->target_oid.name_len; 2019 hoid->snapid = CEPH_NOSNAP; 2020 hoid->hash = t->pgid.seed; 2021 hoid->is_max = false; 2022 if (t->target_oloc.pool_ns) { 2023 hoid->nspace = t->target_oloc.pool_ns->str; 2024 hoid->nspace_len = t->target_oloc.pool_ns->len; 2025 } else { 2026 hoid->nspace = NULL; 2027 hoid->nspace_len = 0; 2028 } 2029 hoid->pool = t->target_oloc.pool; 2030 ceph_hoid_build_hash_cache(hoid); 2031} 2032 2033static bool should_plug_request(struct ceph_osd_request *req) 2034{ 2035 struct ceph_osd *osd = req->r_osd; 2036 struct ceph_spg_mapping *spg; 2037 struct ceph_osd_backoff *backoff; 2038 struct ceph_hobject_id hoid; 2039 2040 spg = lookup_spg_mapping(&osd->o_backoff_mappings, &req->r_t.spgid); 2041 if (!spg) 2042 return false; 2043 2044 hoid_fill_from_target(&hoid, &req->r_t); 2045 backoff = lookup_containing_backoff(&spg->backoffs, &hoid); 2046 if (!backoff) 2047 return false; 2048 2049 dout("%s req %p tid %llu backoff osd%d spgid %llu.%xs%d id %llu\n", 2050 __func__, req, req->r_tid, osd->o_osd, backoff->spgid.pgid.pool, 2051 backoff->spgid.pgid.seed, backoff->spgid.shard, backoff->id); 2052 return true; 2053} 2054 2055/* 2056 * Keep get_num_data_items() in sync with this function. 2057 */ 2058static void setup_request_data(struct ceph_osd_request *req) 2059{ 2060 struct ceph_msg *request_msg = req->r_request; 2061 struct ceph_msg *reply_msg = req->r_reply; 2062 struct ceph_osd_req_op *op; 2063 2064 if (req->r_request->num_data_items || req->r_reply->num_data_items) 2065 return; 2066 2067 WARN_ON(request_msg->data_length || reply_msg->data_length); 2068 for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) { 2069 switch (op->op) { 2070 /* request */ 2071 case CEPH_OSD_OP_WRITE: 2072 case CEPH_OSD_OP_WRITEFULL: 2073 WARN_ON(op->indata_len != op->extent.length); 2074 ceph_osdc_msg_data_add(request_msg, 2075 &op->extent.osd_data); 2076 break; 2077 case CEPH_OSD_OP_SETXATTR: 2078 case CEPH_OSD_OP_CMPXATTR: 2079 WARN_ON(op->indata_len != op->xattr.name_len + 2080 op->xattr.value_len); 2081 ceph_osdc_msg_data_add(request_msg, 2082 &op->xattr.osd_data); 2083 break; 2084 case CEPH_OSD_OP_NOTIFY_ACK: 2085 ceph_osdc_msg_data_add(request_msg, 2086 &op->notify_ack.request_data); 2087 break; 2088 case CEPH_OSD_OP_COPY_FROM2: 2089 ceph_osdc_msg_data_add(request_msg, 2090 &op->copy_from.osd_data); 2091 break; 2092 2093 /* reply */ 2094 case CEPH_OSD_OP_STAT: 2095 ceph_osdc_msg_data_add(reply_msg, 2096 &op->raw_data_in); 2097 break; 2098 case CEPH_OSD_OP_READ: 2099 case CEPH_OSD_OP_SPARSE_READ: 2100 ceph_osdc_msg_data_add(reply_msg, 2101 &op->extent.osd_data); 2102 break; 2103 case CEPH_OSD_OP_LIST_WATCHERS: 2104 ceph_osdc_msg_data_add(reply_msg, 2105 &op->list_watchers.response_data); 2106 break; 2107 2108 /* both */ 2109 case CEPH_OSD_OP_CALL: 2110 WARN_ON(op->indata_len != op->cls.class_len + 2111 op->cls.method_len + 2112 op->cls.indata_len); 2113 ceph_osdc_msg_data_add(request_msg, 2114 &op->cls.request_info); 2115 /* optional, can be NONE */ 2116 ceph_osdc_msg_data_add(request_msg, 2117 &op->cls.request_data); 2118 /* optional, can be NONE */ 2119 ceph_osdc_msg_data_add(reply_msg, 2120 &op->cls.response_data); 2121 break; 2122 case CEPH_OSD_OP_NOTIFY: 2123 ceph_osdc_msg_data_add(request_msg, 2124 &op->notify.request_data); 2125 ceph_osdc_msg_data_add(reply_msg, 2126 &op->notify.response_data); 2127 break; 2128 } 2129 } 2130} 2131 2132static void encode_pgid(void **p, const struct ceph_pg *pgid) 2133{ 2134 ceph_encode_8(p, 1); 2135 ceph_encode_64(p, pgid->pool); 2136 ceph_encode_32(p, pgid->seed); 2137 ceph_encode_32(p, -1); /* preferred */ 2138} 2139 2140static void encode_spgid(void **p, const struct ceph_spg *spgid) 2141{ 2142 ceph_start_encoding(p, 1, 1, CEPH_PGID_ENCODING_LEN + 1); 2143 encode_pgid(p, &spgid->pgid); 2144 ceph_encode_8(p, spgid->shard); 2145} 2146 2147static void encode_oloc(void **p, void *end, 2148 const struct ceph_object_locator *oloc) 2149{ 2150 ceph_start_encoding(p, 5, 4, ceph_oloc_encoding_size(oloc)); 2151 ceph_encode_64(p, oloc->pool); 2152 ceph_encode_32(p, -1); /* preferred */ 2153 ceph_encode_32(p, 0); /* key len */ 2154 if (oloc->pool_ns) 2155 ceph_encode_string(p, end, oloc->pool_ns->str, 2156 oloc->pool_ns->len); 2157 else 2158 ceph_encode_32(p, 0); 2159} 2160 2161static void encode_request_partial(struct ceph_osd_request *req, 2162 struct ceph_msg *msg) 2163{ 2164 void *p = msg->front.iov_base; 2165 void *const end = p + msg->front_alloc_len; 2166 u32 data_len = 0; 2167 int i; 2168 2169 if (req->r_flags & CEPH_OSD_FLAG_WRITE) { 2170 /* snapshots aren't writeable */ 2171 WARN_ON(req->r_snapid != CEPH_NOSNAP); 2172 } else { 2173 WARN_ON(req->r_mtime.tv_sec || req->r_mtime.tv_nsec || 2174 req->r_data_offset || req->r_snapc); 2175 } 2176 2177 setup_request_data(req); 2178 2179 encode_spgid(&p, &req->r_t.spgid); /* actual spg */ 2180 ceph_encode_32(&p, req->r_t.pgid.seed); /* raw hash */ 2181 ceph_encode_32(&p, req->r_osdc->osdmap->epoch); 2182 ceph_encode_32(&p, req->r_flags); 2183 2184 /* reqid */ 2185 ceph_start_encoding(&p, 2, 2, sizeof(struct ceph_osd_reqid)); 2186 memset(p, 0, sizeof(struct ceph_osd_reqid)); 2187 p += sizeof(struct ceph_osd_reqid); 2188 2189 /* trace */ 2190 memset(p, 0, sizeof(struct ceph_blkin_trace_info)); 2191 p += sizeof(struct ceph_blkin_trace_info); 2192 2193 ceph_encode_32(&p, 0); /* client_inc, always 0 */ 2194 ceph_encode_timespec64(p, &req->r_mtime); 2195 p += sizeof(struct ceph_timespec); 2196 2197 encode_oloc(&p, end, &req->r_t.target_oloc); 2198 ceph_encode_string(&p, end, req->r_t.target_oid.name, 2199 req->r_t.target_oid.name_len); 2200 2201 /* ops, can imply data */ 2202 ceph_encode_16(&p, req->r_num_ops); 2203 for (i = 0; i < req->r_num_ops; i++) { 2204 data_len += osd_req_encode_op(p, &req->r_ops[i]); 2205 p += sizeof(struct ceph_osd_op); 2206 } 2207 2208 ceph_encode_64(&p, req->r_snapid); /* snapid */ 2209 if (req->r_snapc) { 2210 ceph_encode_64(&p, req->r_snapc->seq); 2211 ceph_encode_32(&p, req->r_snapc->num_snaps); 2212 for (i = 0; i < req->r_snapc->num_snaps; i++) 2213 ceph_encode_64(&p, req->r_snapc->snaps[i]); 2214 } else { 2215 ceph_encode_64(&p, 0); /* snap_seq */ 2216 ceph_encode_32(&p, 0); /* snaps len */ 2217 } 2218 2219 ceph_encode_32(&p, req->r_attempts); /* retry_attempt */ 2220 BUG_ON(p > end - 8); /* space for features */ 2221 2222 msg->hdr.version = cpu_to_le16(8); /* MOSDOp v8 */ 2223 /* front_len is finalized in encode_request_finish() */ 2224 msg->front.iov_len = p - msg->front.iov_base; 2225 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len); 2226 msg->hdr.data_len = cpu_to_le32(data_len); 2227 /* 2228 * The header "data_off" is a hint to the receiver allowing it 2229 * to align received data into its buffers such that there's no 2230 * need to re-copy it before writing it to disk (direct I/O). 2231 */ 2232 msg->hdr.data_off = cpu_to_le16(req->r_data_offset); 2233 2234 dout("%s req %p msg %p oid %s oid_len %d\n", __func__, req, msg, 2235 req->r_t.target_oid.name, req->r_t.target_oid.name_len); 2236} 2237 2238static void encode_request_finish(struct ceph_msg *msg) 2239{ 2240 void *p = msg->front.iov_base; 2241 void *const partial_end = p + msg->front.iov_len; 2242 void *const end = p + msg->front_alloc_len; 2243 2244 if (CEPH_HAVE_FEATURE(msg->con->peer_features, RESEND_ON_SPLIT)) { 2245 /* luminous OSD -- encode features and be done */ 2246 p = partial_end; 2247 ceph_encode_64(&p, msg->con->peer_features); 2248 } else { 2249 struct { 2250 char spgid[CEPH_ENCODING_START_BLK_LEN + 2251 CEPH_PGID_ENCODING_LEN + 1]; 2252 __le32 hash; 2253 __le32 epoch; 2254 __le32 flags; 2255 char reqid[CEPH_ENCODING_START_BLK_LEN + 2256 sizeof(struct ceph_osd_reqid)]; 2257 char trace[sizeof(struct ceph_blkin_trace_info)]; 2258 __le32 client_inc; 2259 struct ceph_timespec mtime; 2260 } __packed head; 2261 struct ceph_pg pgid; 2262 void *oloc, *oid, *tail; 2263 int oloc_len, oid_len, tail_len; 2264 int len; 2265 2266 /* 2267 * Pre-luminous OSD -- reencode v8 into v4 using @head 2268 * as a temporary buffer. Encode the raw PG; the rest 2269 * is just a matter of moving oloc, oid and tail blobs 2270 * around. 2271 */ 2272 memcpy(&head, p, sizeof(head)); 2273 p += sizeof(head); 2274 2275 oloc = p; 2276 p += CEPH_ENCODING_START_BLK_LEN; 2277 pgid.pool = ceph_decode_64(&p); 2278 p += 4 + 4; /* preferred, key len */ 2279 len = ceph_decode_32(&p); 2280 p += len; /* nspace */ 2281 oloc_len = p - oloc; 2282 2283 oid = p; 2284 len = ceph_decode_32(&p); 2285 p += len; 2286 oid_len = p - oid; 2287 2288 tail = p; 2289 tail_len = partial_end - p; 2290 2291 p = msg->front.iov_base; 2292 ceph_encode_copy(&p, &head.client_inc, sizeof(head.client_inc)); 2293 ceph_encode_copy(&p, &head.epoch, sizeof(head.epoch)); 2294 ceph_encode_copy(&p, &head.flags, sizeof(head.flags)); 2295 ceph_encode_copy(&p, &head.mtime, sizeof(head.mtime)); 2296 2297 /* reassert_version */ 2298 memset(p, 0, sizeof(struct ceph_eversion)); 2299 p += sizeof(struct ceph_eversion); 2300 2301 BUG_ON(p >= oloc); 2302 memmove(p, oloc, oloc_len); 2303 p += oloc_len; 2304 2305 pgid.seed = le32_to_cpu(head.hash); 2306 encode_pgid(&p, &pgid); /* raw pg */ 2307 2308 BUG_ON(p >= oid); 2309 memmove(p, oid, oid_len); 2310 p += oid_len; 2311 2312 /* tail -- ops, snapid, snapc, retry_attempt */ 2313 BUG_ON(p >= tail); 2314 memmove(p, tail, tail_len); 2315 p += tail_len; 2316 2317 msg->hdr.version = cpu_to_le16(4); /* MOSDOp v4 */ 2318 } 2319 2320 BUG_ON(p > end); 2321 msg->front.iov_len = p - msg->front.iov_base; 2322 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len); 2323 2324 dout("%s msg %p tid %llu %u+%u+%u v%d\n", __func__, msg, 2325 le64_to_cpu(msg->hdr.tid), le32_to_cpu(msg->hdr.front_len), 2326 le32_to_cpu(msg->hdr.middle_len), le32_to_cpu(msg->hdr.data_len), 2327 le16_to_cpu(msg->hdr.version)); 2328} 2329 2330/* 2331 * @req has to be assigned a tid and registered. 2332 */ 2333static void send_request(struct ceph_osd_request *req) 2334{ 2335 struct ceph_osd *osd = req->r_osd; 2336 2337 verify_osd_locked(osd); 2338 WARN_ON(osd->o_osd != req->r_t.osd); 2339 2340 /* backoff? */ 2341 if (should_plug_request(req)) 2342 return; 2343 2344 /* 2345 * We may have a previously queued request message hanging 2346 * around. Cancel it to avoid corrupting the msgr. 2347 */ 2348 if (req->r_sent) 2349 ceph_msg_revoke(req->r_request); 2350 2351 req->r_flags |= CEPH_OSD_FLAG_KNOWN_REDIR; 2352 if (req->r_attempts) 2353 req->r_flags |= CEPH_OSD_FLAG_RETRY; 2354 else 2355 WARN_ON(req->r_flags & CEPH_OSD_FLAG_RETRY); 2356 2357 encode_request_partial(req, req->r_request); 2358 2359 dout("%s req %p tid %llu to pgid %llu.%x spgid %llu.%xs%d osd%d e%u flags 0x%x attempt %d\n", 2360 __func__, req, req->r_tid, req->r_t.pgid.pool, req->r_t.pgid.seed, 2361 req->r_t.spgid.pgid.pool, req->r_t.spgid.pgid.seed, 2362 req->r_t.spgid.shard, osd->o_osd, req->r_t.epoch, req->r_flags, 2363 req->r_attempts); 2364 2365 req->r_t.paused = false; 2366 req->r_stamp = jiffies; 2367 req->r_attempts++; 2368 2369 req->r_sent = osd->o_incarnation; 2370 req->r_request->hdr.tid = cpu_to_le64(req->r_tid); 2371 ceph_con_send(&osd->o_con, ceph_msg_get(req->r_request)); 2372} 2373 2374static void maybe_request_map(struct ceph_osd_client *osdc) 2375{ 2376 bool continuous = false; 2377 2378 verify_osdc_locked(osdc); 2379 WARN_ON(!osdc->osdmap->epoch); 2380 2381 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || 2382 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD) || 2383 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) { 2384 dout("%s osdc %p continuous\n", __func__, osdc); 2385 continuous = true; 2386 } else { 2387 dout("%s osdc %p onetime\n", __func__, osdc); 2388 } 2389 2390 if (ceph_monc_want_map(&osdc->client->monc, CEPH_SUB_OSDMAP, 2391 osdc->osdmap->epoch + 1, continuous)) 2392 ceph_monc_renew_subs(&osdc->client->monc); 2393} 2394 2395static void complete_request(struct ceph_osd_request *req, int err); 2396static void send_map_check(struct ceph_osd_request *req); 2397 2398static void __submit_request(struct ceph_osd_request *req, bool wrlocked) 2399{ 2400 struct ceph_osd_client *osdc = req->r_osdc; 2401 struct ceph_osd *osd; 2402 enum calc_target_result ct_res; 2403 int err = 0; 2404 bool need_send = false; 2405 bool promoted = false; 2406 2407 WARN_ON(req->r_tid); 2408 dout("%s req %p wrlocked %d\n", __func__, req, wrlocked); 2409 2410again: 2411 ct_res = calc_target(osdc, &req->r_t, false); 2412 if (ct_res == CALC_TARGET_POOL_DNE && !wrlocked) 2413 goto promote; 2414 2415 osd = lookup_create_osd(osdc, req->r_t.osd, wrlocked); 2416 if (IS_ERR(osd)) { 2417 WARN_ON(PTR_ERR(osd) != -EAGAIN || wrlocked); 2418 goto promote; 2419 } 2420 2421 if (osdc->abort_err) { 2422 dout("req %p abort_err %d\n", req, osdc->abort_err); 2423 err = osdc->abort_err; 2424 } else if (osdc->osdmap->epoch < osdc->epoch_barrier) { 2425 dout("req %p epoch %u barrier %u\n", req, osdc->osdmap->epoch, 2426 osdc->epoch_barrier); 2427 req->r_t.paused = true; 2428 maybe_request_map(osdc); 2429 } else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) && 2430 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) { 2431 dout("req %p pausewr\n", req); 2432 req->r_t.paused = true; 2433 maybe_request_map(osdc); 2434 } else if ((req->r_flags & CEPH_OSD_FLAG_READ) && 2435 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) { 2436 dout("req %p pauserd\n", req); 2437 req->r_t.paused = true; 2438 maybe_request_map(osdc); 2439 } else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) && 2440 !(req->r_flags & (CEPH_OSD_FLAG_FULL_TRY | 2441 CEPH_OSD_FLAG_FULL_FORCE)) && 2442 (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || 2443 pool_full(osdc, req->r_t.base_oloc.pool))) { 2444 dout("req %p full/pool_full\n", req); 2445 if (ceph_test_opt(osdc->client, ABORT_ON_FULL)) { 2446 err = -ENOSPC; 2447 } else { 2448 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL)) 2449 pr_warn_ratelimited("cluster is full (osdmap FULL)\n"); 2450 else 2451 pr_warn_ratelimited("pool %lld is full or reached quota\n", 2452 req->r_t.base_oloc.pool); 2453 req->r_t.paused = true; 2454 maybe_request_map(osdc); 2455 } 2456 } else if (!osd_homeless(osd)) { 2457 need_send = true; 2458 } else { 2459 maybe_request_map(osdc); 2460 } 2461 2462 mutex_lock(&osd->lock); 2463 /* 2464 * Assign the tid atomically with send_request() to protect 2465 * multiple writes to the same object from racing with each 2466 * other, resulting in out of order ops on the OSDs. 2467 */ 2468 req->r_tid = atomic64_inc_return(&osdc->last_tid); 2469 link_request(osd, req); 2470 if (need_send) 2471 send_request(req); 2472 else if (err) 2473 complete_request(req, err); 2474 mutex_unlock(&osd->lock); 2475 2476 if (!err && ct_res == CALC_TARGET_POOL_DNE) 2477 send_map_check(req); 2478 2479 if (promoted) 2480 downgrade_write(&osdc->lock); 2481 return; 2482 2483promote: 2484 up_read(&osdc->lock); 2485 down_write(&osdc->lock); 2486 wrlocked = true; 2487 promoted = true; 2488 goto again; 2489} 2490 2491static void account_request(struct ceph_osd_request *req) 2492{ 2493 WARN_ON(req->r_flags & (CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK)); 2494 WARN_ON(!(req->r_flags & (CEPH_OSD_FLAG_READ | CEPH_OSD_FLAG_WRITE))); 2495 2496 req->r_flags |= CEPH_OSD_FLAG_ONDISK; 2497 atomic_inc(&req->r_osdc->num_requests); 2498 2499 req->r_start_stamp = jiffies; 2500 req->r_start_latency = ktime_get(); 2501} 2502 2503static void submit_request(struct ceph_osd_request *req, bool wrlocked) 2504{ 2505 ceph_osdc_get_request(req); 2506 account_request(req); 2507 __submit_request(req, wrlocked); 2508} 2509 2510static void finish_request(struct ceph_osd_request *req) 2511{ 2512 struct ceph_osd_client *osdc = req->r_osdc; 2513 2514 WARN_ON(lookup_request_mc(&osdc->map_checks, req->r_tid)); 2515 dout("%s req %p tid %llu\n", __func__, req, req->r_tid); 2516 2517 req->r_end_latency = ktime_get(); 2518 2519 if (req->r_osd) { 2520 ceph_init_sparse_read(&req->r_osd->o_sparse_read); 2521 unlink_request(req->r_osd, req); 2522 } 2523 atomic_dec(&osdc->num_requests); 2524 2525 /* 2526 * If an OSD has failed or returned and a request has been sent 2527 * twice, it's possible to get a reply and end up here while the 2528 * request message is queued for delivery. We will ignore the 2529 * reply, so not a big deal, but better to try and catch it. 2530 */ 2531 ceph_msg_revoke(req->r_request); 2532 ceph_msg_revoke_incoming(req->r_reply); 2533} 2534 2535static void __complete_request(struct ceph_osd_request *req) 2536{ 2537 dout("%s req %p tid %llu cb %ps result %d\n", __func__, req, 2538 req->r_tid, req->r_callback, req->r_result); 2539 2540 if (req->r_callback) 2541 req->r_callback(req); 2542 complete_all(&req->r_completion); 2543 ceph_osdc_put_request(req); 2544} 2545 2546static void complete_request_workfn(struct work_struct *work) 2547{ 2548 struct ceph_osd_request *req = 2549 container_of(work, struct ceph_osd_request, r_complete_work); 2550 2551 __complete_request(req); 2552} 2553 2554/* 2555 * This is open-coded in handle_reply(). 2556 */ 2557static void complete_request(struct ceph_osd_request *req, int err) 2558{ 2559 dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err); 2560 2561 req->r_result = err; 2562 finish_request(req); 2563 2564 INIT_WORK(&req->r_complete_work, complete_request_workfn); 2565 queue_work(req->r_osdc->completion_wq, &req->r_complete_work); 2566} 2567 2568static void cancel_map_check(struct ceph_osd_request *req) 2569{ 2570 struct ceph_osd_client *osdc = req->r_osdc; 2571 struct ceph_osd_request *lookup_req; 2572 2573 verify_osdc_wrlocked(osdc); 2574 2575 lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid); 2576 if (!lookup_req) 2577 return; 2578 2579 WARN_ON(lookup_req != req); 2580 erase_request_mc(&osdc->map_checks, req); 2581 ceph_osdc_put_request(req); 2582} 2583 2584static void cancel_request(struct ceph_osd_request *req) 2585{ 2586 dout("%s req %p tid %llu\n", __func__, req, req->r_tid); 2587 2588 cancel_map_check(req); 2589 finish_request(req); 2590 complete_all(&req->r_completion); 2591 ceph_osdc_put_request(req); 2592} 2593 2594static void abort_request(struct ceph_osd_request *req, int err) 2595{ 2596 dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err); 2597 2598 cancel_map_check(req); 2599 complete_request(req, err); 2600} 2601 2602static int abort_fn(struct ceph_osd_request *req, void *arg) 2603{ 2604 int err = *(int *)arg; 2605 2606 abort_request(req, err); 2607 return 0; /* continue iteration */ 2608} 2609 2610/* 2611 * Abort all in-flight requests with @err and arrange for all future 2612 * requests to be failed immediately. 2613 */ 2614void ceph_osdc_abort_requests(struct ceph_osd_client *osdc, int err) 2615{ 2616 dout("%s osdc %p err %d\n", __func__, osdc, err); 2617 down_write(&osdc->lock); 2618 for_each_request(osdc, abort_fn, &err); 2619 osdc->abort_err = err; 2620 up_write(&osdc->lock); 2621} 2622EXPORT_SYMBOL(ceph_osdc_abort_requests); 2623 2624void ceph_osdc_clear_abort_err(struct ceph_osd_client *osdc) 2625{ 2626 down_write(&osdc->lock); 2627 osdc->abort_err = 0; 2628 up_write(&osdc->lock); 2629} 2630EXPORT_SYMBOL(ceph_osdc_clear_abort_err); 2631 2632static void update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb) 2633{ 2634 if (likely(eb > osdc->epoch_barrier)) { 2635 dout("updating epoch_barrier from %u to %u\n", 2636 osdc->epoch_barrier, eb); 2637 osdc->epoch_barrier = eb; 2638 /* Request map if we're not to the barrier yet */ 2639 if (eb > osdc->osdmap->epoch) 2640 maybe_request_map(osdc); 2641 } 2642} 2643 2644void ceph_osdc_update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb) 2645{ 2646 down_read(&osdc->lock); 2647 if (unlikely(eb > osdc->epoch_barrier)) { 2648 up_read(&osdc->lock); 2649 down_write(&osdc->lock); 2650 update_epoch_barrier(osdc, eb); 2651 up_write(&osdc->lock); 2652 } else { 2653 up_read(&osdc->lock); 2654 } 2655} 2656EXPORT_SYMBOL(ceph_osdc_update_epoch_barrier); 2657 2658/* 2659 * We can end up releasing caps as a result of abort_request(). 2660 * In that case, we probably want to ensure that the cap release message 2661 * has an updated epoch barrier in it, so set the epoch barrier prior to 2662 * aborting the first request. 2663 */ 2664static int abort_on_full_fn(struct ceph_osd_request *req, void *arg) 2665{ 2666 struct ceph_osd_client *osdc = req->r_osdc; 2667 bool *victims = arg; 2668 2669 if ((req->r_flags & CEPH_OSD_FLAG_WRITE) && 2670 (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || 2671 pool_full(osdc, req->r_t.base_oloc.pool))) { 2672 if (!*victims) { 2673 update_epoch_barrier(osdc, osdc->osdmap->epoch); 2674 *victims = true; 2675 } 2676 abort_request(req, -ENOSPC); 2677 } 2678 2679 return 0; /* continue iteration */ 2680} 2681 2682/* 2683 * Drop all pending requests that are stalled waiting on a full condition to 2684 * clear, and complete them with ENOSPC as the return code. Set the 2685 * osdc->epoch_barrier to the latest map epoch that we've seen if any were 2686 * cancelled. 2687 */ 2688static void ceph_osdc_abort_on_full(struct ceph_osd_client *osdc) 2689{ 2690 bool victims = false; 2691 2692 if (ceph_test_opt(osdc->client, ABORT_ON_FULL) && 2693 (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || have_pool_full(osdc))) 2694 for_each_request(osdc, abort_on_full_fn, &victims); 2695} 2696 2697static void check_pool_dne(struct ceph_osd_request *req) 2698{ 2699 struct ceph_osd_client *osdc = req->r_osdc; 2700 struct ceph_osdmap *map = osdc->osdmap; 2701 2702 verify_osdc_wrlocked(osdc); 2703 WARN_ON(!map->epoch); 2704 2705 if (req->r_attempts) { 2706 /* 2707 * We sent a request earlier, which means that 2708 * previously the pool existed, and now it does not 2709 * (i.e., it was deleted). 2710 */ 2711 req->r_map_dne_bound = map->epoch; 2712 dout("%s req %p tid %llu pool disappeared\n", __func__, req, 2713 req->r_tid); 2714 } else { 2715 dout("%s req %p tid %llu map_dne_bound %u have %u\n", __func__, 2716 req, req->r_tid, req->r_map_dne_bound, map->epoch); 2717 } 2718 2719 if (req->r_map_dne_bound) { 2720 if (map->epoch >= req->r_map_dne_bound) { 2721 /* we had a new enough map */ 2722 pr_info_ratelimited("tid %llu pool does not exist\n", 2723 req->r_tid); 2724 complete_request(req, -ENOENT); 2725 } 2726 } else { 2727 send_map_check(req); 2728 } 2729} 2730 2731static void map_check_cb(struct ceph_mon_generic_request *greq) 2732{ 2733 struct ceph_osd_client *osdc = &greq->monc->client->osdc; 2734 struct ceph_osd_request *req; 2735 u64 tid = greq->private_data; 2736 2737 WARN_ON(greq->result || !greq->u.newest); 2738 2739 down_write(&osdc->lock); 2740 req = lookup_request_mc(&osdc->map_checks, tid); 2741 if (!req) { 2742 dout("%s tid %llu dne\n", __func__, tid); 2743 goto out_unlock; 2744 } 2745 2746 dout("%s req %p tid %llu map_dne_bound %u newest %llu\n", __func__, 2747 req, req->r_tid, req->r_map_dne_bound, greq->u.newest); 2748 if (!req->r_map_dne_bound) 2749 req->r_map_dne_bound = greq->u.newest; 2750 erase_request_mc(&osdc->map_checks, req); 2751 check_pool_dne(req); 2752 2753 ceph_osdc_put_request(req); 2754out_unlock: 2755 up_write(&osdc->lock); 2756} 2757 2758static void send_map_check(struct ceph_osd_request *req) 2759{ 2760 struct ceph_osd_client *osdc = req->r_osdc; 2761 struct ceph_osd_request *lookup_req; 2762 int ret; 2763 2764 verify_osdc_wrlocked(osdc); 2765 2766 lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid); 2767 if (lookup_req) { 2768 WARN_ON(lookup_req != req); 2769 return; 2770 } 2771 2772 ceph_osdc_get_request(req); 2773 insert_request_mc(&osdc->map_checks, req); 2774 ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap", 2775 map_check_cb, req->r_tid); 2776 WARN_ON(ret); 2777} 2778 2779/* 2780 * lingering requests, watch/notify v2 infrastructure 2781 */ 2782static void linger_release(struct kref *kref) 2783{ 2784 struct ceph_osd_linger_request *lreq = 2785 container_of(kref, struct ceph_osd_linger_request, kref); 2786 2787 dout("%s lreq %p reg_req %p ping_req %p\n", __func__, lreq, 2788 lreq->reg_req, lreq->ping_req); 2789 WARN_ON(!RB_EMPTY_NODE(&lreq->node)); 2790 WARN_ON(!RB_EMPTY_NODE(&lreq->osdc_node)); 2791 WARN_ON(!RB_EMPTY_NODE(&lreq->mc_node)); 2792 WARN_ON(!list_empty(&lreq->scan_item)); 2793 WARN_ON(!list_empty(&lreq->pending_lworks)); 2794 WARN_ON(lreq->osd); 2795 2796 if (lreq->request_pl) 2797 ceph_pagelist_release(lreq->request_pl); 2798 if (lreq->notify_id_pages) 2799 ceph_release_page_vector(lreq->notify_id_pages, 1); 2800 2801 ceph_osdc_put_request(lreq->reg_req); 2802 ceph_osdc_put_request(lreq->ping_req); 2803 target_destroy(&lreq->t); 2804 kfree(lreq); 2805} 2806 2807static void linger_put(struct ceph_osd_linger_request *lreq) 2808{ 2809 if (lreq) 2810 kref_put(&lreq->kref, linger_release); 2811} 2812 2813static struct ceph_osd_linger_request * 2814linger_get(struct ceph_osd_linger_request *lreq) 2815{ 2816 kref_get(&lreq->kref); 2817 return lreq; 2818} 2819 2820static struct ceph_osd_linger_request * 2821linger_alloc(struct ceph_osd_client *osdc) 2822{ 2823 struct ceph_osd_linger_request *lreq; 2824 2825 lreq = kzalloc(sizeof(*lreq), GFP_NOIO); 2826 if (!lreq) 2827 return NULL; 2828 2829 kref_init(&lreq->kref); 2830 mutex_init(&lreq->lock); 2831 RB_CLEAR_NODE(&lreq->node); 2832 RB_CLEAR_NODE(&lreq->osdc_node); 2833 RB_CLEAR_NODE(&lreq->mc_node); 2834 INIT_LIST_HEAD(&lreq->scan_item); 2835 INIT_LIST_HEAD(&lreq->pending_lworks); 2836 init_completion(&lreq->reg_commit_wait); 2837 init_completion(&lreq->notify_finish_wait); 2838 2839 lreq->osdc = osdc; 2840 target_init(&lreq->t); 2841 2842 dout("%s lreq %p\n", __func__, lreq); 2843 return lreq; 2844} 2845 2846DEFINE_RB_INSDEL_FUNCS(linger, struct ceph_osd_linger_request, linger_id, node) 2847DEFINE_RB_FUNCS(linger_osdc, struct ceph_osd_linger_request, linger_id, osdc_node) 2848DEFINE_RB_FUNCS(linger_mc, struct ceph_osd_linger_request, linger_id, mc_node) 2849 2850/* 2851 * Create linger request <-> OSD session relation. 2852 * 2853 * @lreq has to be registered, @osd may be homeless. 2854 */ 2855static void link_linger(struct ceph_osd *osd, 2856 struct ceph_osd_linger_request *lreq) 2857{ 2858 verify_osd_locked(osd); 2859 WARN_ON(!lreq->linger_id || lreq->osd); 2860 dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd, 2861 osd->o_osd, lreq, lreq->linger_id); 2862 2863 if (!osd_homeless(osd)) 2864 __remove_osd_from_lru(osd); 2865 else 2866 atomic_inc(&osd->o_osdc->num_homeless); 2867 2868 get_osd(osd); 2869 insert_linger(&osd->o_linger_requests, lreq); 2870 lreq->osd = osd; 2871} 2872 2873static void unlink_linger(struct ceph_osd *osd, 2874 struct ceph_osd_linger_request *lreq) 2875{ 2876 verify_osd_locked(osd); 2877 WARN_ON(lreq->osd != osd); 2878 dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd, 2879 osd->o_osd, lreq, lreq->linger_id); 2880 2881 lreq->osd = NULL; 2882 erase_linger(&osd->o_linger_requests, lreq); 2883 put_osd(osd); 2884 2885 if (!osd_homeless(osd)) 2886 maybe_move_osd_to_lru(osd); 2887 else 2888 atomic_dec(&osd->o_osdc->num_homeless); 2889} 2890 2891static bool __linger_registered(struct ceph_osd_linger_request *lreq) 2892{ 2893 verify_osdc_locked(lreq->osdc); 2894 2895 return !RB_EMPTY_NODE(&lreq->osdc_node); 2896} 2897 2898static bool linger_registered(struct ceph_osd_linger_request *lreq) 2899{ 2900 struct ceph_osd_client *osdc = lreq->osdc; 2901 bool registered; 2902 2903 down_read(&osdc->lock); 2904 registered = __linger_registered(lreq); 2905 up_read(&osdc->lock); 2906 2907 return registered; 2908} 2909 2910static void linger_register(struct ceph_osd_linger_request *lreq) 2911{ 2912 struct ceph_osd_client *osdc = lreq->osdc; 2913 2914 verify_osdc_wrlocked(osdc); 2915 WARN_ON(lreq->linger_id); 2916 2917 linger_get(lreq); 2918 lreq->linger_id = ++osdc->last_linger_id; 2919 insert_linger_osdc(&osdc->linger_requests, lreq); 2920} 2921 2922static void linger_unregister(struct ceph_osd_linger_request *lreq) 2923{ 2924 struct ceph_osd_client *osdc = lreq->osdc; 2925 2926 verify_osdc_wrlocked(osdc); 2927 2928 erase_linger_osdc(&osdc->linger_requests, lreq); 2929 linger_put(lreq); 2930} 2931 2932static void cancel_linger_request(struct ceph_osd_request *req) 2933{ 2934 struct ceph_osd_linger_request *lreq = req->r_priv; 2935 2936 WARN_ON(!req->r_linger); 2937 cancel_request(req); 2938 linger_put(lreq); 2939} 2940 2941struct linger_work { 2942 struct work_struct work; 2943 struct ceph_osd_linger_request *lreq; 2944 struct list_head pending_item; 2945 unsigned long queued_stamp; 2946 2947 union { 2948 struct { 2949 u64 notify_id; 2950 u64 notifier_id; 2951 void *payload; /* points into @msg front */ 2952 size_t payload_len; 2953 2954 struct ceph_msg *msg; /* for ceph_msg_put() */ 2955 } notify; 2956 struct { 2957 int err; 2958 } error; 2959 }; 2960}; 2961 2962static struct linger_work *lwork_alloc(struct ceph_osd_linger_request *lreq, 2963 work_func_t workfn) 2964{ 2965 struct linger_work *lwork; 2966 2967 lwork = kzalloc(sizeof(*lwork), GFP_NOIO); 2968 if (!lwork) 2969 return NULL; 2970 2971 INIT_WORK(&lwork->work, workfn); 2972 INIT_LIST_HEAD(&lwork->pending_item); 2973 lwork->lreq = linger_get(lreq); 2974 2975 return lwork; 2976} 2977 2978static void lwork_free(struct linger_work *lwork) 2979{ 2980 struct ceph_osd_linger_request *lreq = lwork->lreq; 2981 2982 mutex_lock(&lreq->lock); 2983 list_del(&lwork->pending_item); 2984 mutex_unlock(&lreq->lock); 2985 2986 linger_put(lreq); 2987 kfree(lwork); 2988} 2989 2990static void lwork_queue(struct linger_work *lwork) 2991{ 2992 struct ceph_osd_linger_request *lreq = lwork->lreq; 2993 struct ceph_osd_client *osdc = lreq->osdc; 2994 2995 verify_lreq_locked(lreq); 2996 WARN_ON(!list_empty(&lwork->pending_item)); 2997 2998 lwork->queued_stamp = jiffies; 2999 list_add_tail(&lwork->pending_item, &lreq->pending_lworks); 3000 queue_work(osdc->notify_wq, &lwork->work); 3001} 3002 3003static void do_watch_notify(struct work_struct *w) 3004{ 3005 struct linger_work *lwork = container_of(w, struct linger_work, work); 3006 struct ceph_osd_linger_request *lreq = lwork->lreq; 3007 3008 if (!linger_registered(lreq)) { 3009 dout("%s lreq %p not registered\n", __func__, lreq); 3010 goto out; 3011 } 3012 3013 WARN_ON(!lreq->is_watch); 3014 dout("%s lreq %p notify_id %llu notifier_id %llu payload_len %zu\n", 3015 __func__, lreq, lwork->notify.notify_id, lwork->notify.notifier_id, 3016 lwork->notify.payload_len); 3017 lreq->wcb(lreq->data, lwork->notify.notify_id, lreq->linger_id, 3018 lwork->notify.notifier_id, lwork->notify.payload, 3019 lwork->notify.payload_len); 3020 3021out: 3022 ceph_msg_put(lwork->notify.msg); 3023 lwork_free(lwork); 3024} 3025 3026static void do_watch_error(struct work_struct *w) 3027{ 3028 struct linger_work *lwork = container_of(w, struct linger_work, work); 3029 struct ceph_osd_linger_request *lreq = lwork->lreq; 3030 3031 if (!linger_registered(lreq)) { 3032 dout("%s lreq %p not registered\n", __func__, lreq); 3033 goto out; 3034 } 3035 3036 dout("%s lreq %p err %d\n", __func__, lreq, lwork->error.err); 3037 lreq->errcb(lreq->data, lreq->linger_id, lwork->error.err); 3038 3039out: 3040 lwork_free(lwork); 3041} 3042 3043static void queue_watch_error(struct ceph_osd_linger_request *lreq) 3044{ 3045 struct linger_work *lwork; 3046 3047 lwork = lwork_alloc(lreq, do_watch_error); 3048 if (!lwork) { 3049 pr_err("failed to allocate error-lwork\n"); 3050 return; 3051 } 3052 3053 lwork->error.err = lreq->last_error; 3054 lwork_queue(lwork); 3055} 3056 3057static void linger_reg_commit_complete(struct ceph_osd_linger_request *lreq, 3058 int result) 3059{ 3060 if (!completion_done(&lreq->reg_commit_wait)) { 3061 lreq->reg_commit_error = (result <= 0 ? result : 0); 3062 complete_all(&lreq->reg_commit_wait); 3063 } 3064} 3065 3066static void linger_commit_cb(struct ceph_osd_request *req) 3067{ 3068 struct ceph_osd_linger_request *lreq = req->r_priv; 3069 3070 mutex_lock(&lreq->lock); 3071 if (req != lreq->reg_req) { 3072 dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n", 3073 __func__, lreq, lreq->linger_id, req, lreq->reg_req); 3074 goto out; 3075 } 3076 3077 dout("%s lreq %p linger_id %llu result %d\n", __func__, lreq, 3078 lreq->linger_id, req->r_result); 3079 linger_reg_commit_complete(lreq, req->r_result); 3080 lreq->committed = true; 3081 3082 if (!lreq->is_watch) { 3083 struct ceph_osd_data *osd_data = 3084 osd_req_op_data(req, 0, notify, response_data); 3085 void *p = page_address(osd_data->pages[0]); 3086 3087 WARN_ON(req->r_ops[0].op != CEPH_OSD_OP_NOTIFY || 3088 osd_data->type != CEPH_OSD_DATA_TYPE_PAGES); 3089 3090 /* make note of the notify_id */ 3091 if (req->r_ops[0].outdata_len >= sizeof(u64)) { 3092 lreq->notify_id = ceph_decode_64(&p); 3093 dout("lreq %p notify_id %llu\n", lreq, 3094 lreq->notify_id); 3095 } else { 3096 dout("lreq %p no notify_id\n", lreq); 3097 } 3098 } 3099 3100out: 3101 mutex_unlock(&lreq->lock); 3102 linger_put(lreq); 3103} 3104 3105static int normalize_watch_error(int err) 3106{ 3107 /* 3108 * Translate ENOENT -> ENOTCONN so that a delete->disconnection 3109 * notification and a failure to reconnect because we raced with 3110 * the delete appear the same to the user. 3111 */ 3112 if (err == -ENOENT) 3113 err = -ENOTCONN; 3114 3115 return err; 3116} 3117 3118static void linger_reconnect_cb(struct ceph_osd_request *req) 3119{ 3120 struct ceph_osd_linger_request *lreq = req->r_priv; 3121 3122 mutex_lock(&lreq->lock); 3123 if (req != lreq->reg_req) { 3124 dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n", 3125 __func__, lreq, lreq->linger_id, req, lreq->reg_req); 3126 goto out; 3127 } 3128 3129 dout("%s lreq %p linger_id %llu result %d last_error %d\n", __func__, 3130 lreq, lreq->linger_id, req->r_result, lreq->last_error); 3131 if (req->r_result < 0) { 3132 if (!lreq->last_error) { 3133 lreq->last_error = normalize_watch_error(req->r_result); 3134 queue_watch_error(lreq); 3135 } 3136 } 3137 3138out: 3139 mutex_unlock(&lreq->lock); 3140 linger_put(lreq); 3141} 3142 3143static void send_linger(struct ceph_osd_linger_request *lreq) 3144{ 3145 struct ceph_osd_client *osdc = lreq->osdc; 3146 struct ceph_osd_request *req; 3147 int ret; 3148 3149 verify_osdc_wrlocked(osdc); 3150 mutex_lock(&lreq->lock); 3151 dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id); 3152 3153 if (lreq->reg_req) { 3154 if (lreq->reg_req->r_osd) 3155 cancel_linger_request(lreq->reg_req); 3156 ceph_osdc_put_request(lreq->reg_req); 3157 } 3158 3159 req = ceph_osdc_alloc_request(osdc, NULL, 1, true, GFP_NOIO); 3160 BUG_ON(!req); 3161 3162 target_copy(&req->r_t, &lreq->t); 3163 req->r_mtime = lreq->mtime; 3164 3165 if (lreq->is_watch && lreq->committed) { 3166 osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_RECONNECT, 3167 lreq->linger_id, ++lreq->register_gen); 3168 dout("lreq %p reconnect register_gen %u\n", lreq, 3169 req->r_ops[0].watch.gen); 3170 req->r_callback = linger_reconnect_cb; 3171 } else { 3172 if (lreq->is_watch) { 3173 osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_WATCH, 3174 lreq->linger_id, 0); 3175 } else { 3176 lreq->notify_id = 0; 3177 3178 refcount_inc(&lreq->request_pl->refcnt); 3179 osd_req_op_notify_init(req, 0, lreq->linger_id, 3180 lreq->request_pl); 3181 ceph_osd_data_pages_init( 3182 osd_req_op_data(req, 0, notify, response_data), 3183 lreq->notify_id_pages, PAGE_SIZE, 0, false, false); 3184 } 3185 dout("lreq %p register\n", lreq); 3186 req->r_callback = linger_commit_cb; 3187 } 3188 3189 ret = ceph_osdc_alloc_messages(req, GFP_NOIO); 3190 BUG_ON(ret); 3191 3192 req->r_priv = linger_get(lreq); 3193 req->r_linger = true; 3194 lreq->reg_req = req; 3195 mutex_unlock(&lreq->lock); 3196 3197 submit_request(req, true); 3198} 3199 3200static void linger_ping_cb(struct ceph_osd_request *req) 3201{ 3202 struct ceph_osd_linger_request *lreq = req->r_priv; 3203 3204 mutex_lock(&lreq->lock); 3205 if (req != lreq->ping_req) { 3206 dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n", 3207 __func__, lreq, lreq->linger_id, req, lreq->ping_req); 3208 goto out; 3209 } 3210 3211 dout("%s lreq %p linger_id %llu result %d ping_sent %lu last_error %d\n", 3212 __func__, lreq, lreq->linger_id, req->r_result, lreq->ping_sent, 3213 lreq->last_error); 3214 if (lreq->register_gen == req->r_ops[0].watch.gen) { 3215 if (!req->r_result) { 3216 lreq->watch_valid_thru = lreq->ping_sent; 3217 } else if (!lreq->last_error) { 3218 lreq->last_error = normalize_watch_error(req->r_result); 3219 queue_watch_error(lreq); 3220 } 3221 } else { 3222 dout("lreq %p register_gen %u ignoring old pong %u\n", lreq, 3223 lreq->register_gen, req->r_ops[0].watch.gen); 3224 } 3225 3226out: 3227 mutex_unlock(&lreq->lock); 3228 linger_put(lreq); 3229} 3230 3231static void send_linger_ping(struct ceph_osd_linger_request *lreq) 3232{ 3233 struct ceph_osd_client *osdc = lreq->osdc; 3234 struct ceph_osd_request *req; 3235 int ret; 3236 3237 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) { 3238 dout("%s PAUSERD\n", __func__); 3239 return; 3240 } 3241 3242 lreq->ping_sent = jiffies; 3243 dout("%s lreq %p linger_id %llu ping_sent %lu register_gen %u\n", 3244 __func__, lreq, lreq->linger_id, lreq->ping_sent, 3245 lreq->register_gen); 3246 3247 if (lreq->ping_req) { 3248 if (lreq->ping_req->r_osd) 3249 cancel_linger_request(lreq->ping_req); 3250 ceph_osdc_put_request(lreq->ping_req); 3251 } 3252 3253 req = ceph_osdc_alloc_request(osdc, NULL, 1, true, GFP_NOIO); 3254 BUG_ON(!req); 3255 3256 target_copy(&req->r_t, &lreq->t); 3257 osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_PING, lreq->linger_id, 3258 lreq->register_gen); 3259 req->r_callback = linger_ping_cb; 3260 3261 ret = ceph_osdc_alloc_messages(req, GFP_NOIO); 3262 BUG_ON(ret); 3263 3264 req->r_priv = linger_get(lreq); 3265 req->r_linger = true; 3266 lreq->ping_req = req; 3267 3268 ceph_osdc_get_request(req); 3269 account_request(req); 3270 req->r_tid = atomic64_inc_return(&osdc->last_tid); 3271 link_request(lreq->osd, req); 3272 send_request(req); 3273} 3274 3275static void linger_submit(struct ceph_osd_linger_request *lreq) 3276{ 3277 struct ceph_osd_client *osdc = lreq->osdc; 3278 struct ceph_osd *osd; 3279 3280 down_write(&osdc->lock); 3281 linger_register(lreq); 3282 3283 calc_target(osdc, &lreq->t, false); 3284 osd = lookup_create_osd(osdc, lreq->t.osd, true); 3285 link_linger(osd, lreq); 3286 3287 send_linger(lreq); 3288 up_write(&osdc->lock); 3289} 3290 3291static void cancel_linger_map_check(struct ceph_osd_linger_request *lreq) 3292{ 3293 struct ceph_osd_client *osdc = lreq->osdc; 3294 struct ceph_osd_linger_request *lookup_lreq; 3295 3296 verify_osdc_wrlocked(osdc); 3297 3298 lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks, 3299 lreq->linger_id); 3300 if (!lookup_lreq) 3301 return; 3302 3303 WARN_ON(lookup_lreq != lreq); 3304 erase_linger_mc(&osdc->linger_map_checks, lreq); 3305 linger_put(lreq); 3306} 3307 3308/* 3309 * @lreq has to be both registered and linked. 3310 */ 3311static void __linger_cancel(struct ceph_osd_linger_request *lreq) 3312{ 3313 if (lreq->ping_req && lreq->ping_req->r_osd) 3314 cancel_linger_request(lreq->ping_req); 3315 if (lreq->reg_req && lreq->reg_req->r_osd) 3316 cancel_linger_request(lreq->reg_req); 3317 cancel_linger_map_check(lreq); 3318 unlink_linger(lreq->osd, lreq); 3319 linger_unregister(lreq); 3320} 3321 3322static void linger_cancel(struct ceph_osd_linger_request *lreq) 3323{ 3324 struct ceph_osd_client *osdc = lreq->osdc; 3325 3326 down_write(&osdc->lock); 3327 if (__linger_registered(lreq)) 3328 __linger_cancel(lreq); 3329 up_write(&osdc->lock); 3330} 3331 3332static void send_linger_map_check(struct ceph_osd_linger_request *lreq); 3333 3334static void check_linger_pool_dne(struct ceph_osd_linger_request *lreq) 3335{ 3336 struct ceph_osd_client *osdc = lreq->osdc; 3337 struct ceph_osdmap *map = osdc->osdmap; 3338 3339 verify_osdc_wrlocked(osdc); 3340 WARN_ON(!map->epoch); 3341 3342 if (lreq->register_gen) { 3343 lreq->map_dne_bound = map->epoch; 3344 dout("%s lreq %p linger_id %llu pool disappeared\n", __func__, 3345 lreq, lreq->linger_id); 3346 } else { 3347 dout("%s lreq %p linger_id %llu map_dne_bound %u have %u\n", 3348 __func__, lreq, lreq->linger_id, lreq->map_dne_bound, 3349 map->epoch); 3350 } 3351 3352 if (lreq->map_dne_bound) { 3353 if (map->epoch >= lreq->map_dne_bound) { 3354 /* we had a new enough map */ 3355 pr_info("linger_id %llu pool does not exist\n", 3356 lreq->linger_id); 3357 linger_reg_commit_complete(lreq, -ENOENT); 3358 __linger_cancel(lreq); 3359 } 3360 } else { 3361 send_linger_map_check(lreq); 3362 } 3363} 3364 3365static void linger_map_check_cb(struct ceph_mon_generic_request *greq) 3366{ 3367 struct ceph_osd_client *osdc = &greq->monc->client->osdc; 3368 struct ceph_osd_linger_request *lreq; 3369 u64 linger_id = greq->private_data; 3370 3371 WARN_ON(greq->result || !greq->u.newest); 3372 3373 down_write(&osdc->lock); 3374 lreq = lookup_linger_mc(&osdc->linger_map_checks, linger_id); 3375 if (!lreq) { 3376 dout("%s linger_id %llu dne\n", __func__, linger_id); 3377 goto out_unlock; 3378 } 3379 3380 dout("%s lreq %p linger_id %llu map_dne_bound %u newest %llu\n", 3381 __func__, lreq, lreq->linger_id, lreq->map_dne_bound, 3382 greq->u.newest); 3383 if (!lreq->map_dne_bound) 3384 lreq->map_dne_bound = greq->u.newest; 3385 erase_linger_mc(&osdc->linger_map_checks, lreq); 3386 check_linger_pool_dne(lreq); 3387 3388 linger_put(lreq); 3389out_unlock: 3390 up_write(&osdc->lock); 3391} 3392 3393static void send_linger_map_check(struct ceph_osd_linger_request *lreq) 3394{ 3395 struct ceph_osd_client *osdc = lreq->osdc; 3396 struct ceph_osd_linger_request *lookup_lreq; 3397 int ret; 3398 3399 verify_osdc_wrlocked(osdc); 3400 3401 lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks, 3402 lreq->linger_id); 3403 if (lookup_lreq) { 3404 WARN_ON(lookup_lreq != lreq); 3405 return; 3406 } 3407 3408 linger_get(lreq); 3409 insert_linger_mc(&osdc->linger_map_checks, lreq); 3410 ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap", 3411 linger_map_check_cb, lreq->linger_id); 3412 WARN_ON(ret); 3413} 3414 3415static int linger_reg_commit_wait(struct ceph_osd_linger_request *lreq) 3416{ 3417 int ret; 3418 3419 dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id); 3420 ret = wait_for_completion_killable(&lreq->reg_commit_wait); 3421 return ret ?: lreq->reg_commit_error; 3422} 3423 3424static int linger_notify_finish_wait(struct ceph_osd_linger_request *lreq, 3425 unsigned long timeout) 3426{ 3427 long left; 3428 3429 dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id); 3430 left = wait_for_completion_killable_timeout(&lreq->notify_finish_wait, 3431 ceph_timeout_jiffies(timeout)); 3432 if (left <= 0) 3433 left = left ?: -ETIMEDOUT; 3434 else 3435 left = lreq->notify_finish_error; /* completed */ 3436 3437 return left; 3438} 3439 3440/* 3441 * Timeout callback, called every N seconds. When 1 or more OSD 3442 * requests has been active for more than N seconds, we send a keepalive 3443 * (tag + timestamp) to its OSD to ensure any communications channel 3444 * reset is detected. 3445 */ 3446static void handle_timeout(struct work_struct *work) 3447{ 3448 struct ceph_osd_client *osdc = 3449 container_of(work, struct ceph_osd_client, timeout_work.work); 3450 struct ceph_options *opts = osdc->client->options; 3451 unsigned long cutoff = jiffies - opts->osd_keepalive_timeout; 3452 unsigned long expiry_cutoff = jiffies - opts->osd_request_timeout; 3453 LIST_HEAD(slow_osds); 3454 struct rb_node *n, *p; 3455 3456 dout("%s osdc %p\n", __func__, osdc); 3457 down_write(&osdc->lock); 3458 3459 /* 3460 * ping osds that are a bit slow. this ensures that if there 3461 * is a break in the TCP connection we will notice, and reopen 3462 * a connection with that osd (from the fault callback). 3463 */ 3464 for (n = rb_first(&osdc->osds); n; n = rb_next(n)) { 3465 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node); 3466 bool found = false; 3467 3468 for (p = rb_first(&osd->o_requests); p; ) { 3469 struct ceph_osd_request *req = 3470 rb_entry(p, struct ceph_osd_request, r_node); 3471 3472 p = rb_next(p); /* abort_request() */ 3473 3474 if (time_before(req->r_stamp, cutoff)) { 3475 dout(" req %p tid %llu on osd%d is laggy\n", 3476 req, req->r_tid, osd->o_osd); 3477 found = true; 3478 } 3479 if (opts->osd_request_timeout && 3480 time_before(req->r_start_stamp, expiry_cutoff)) { 3481 pr_err_ratelimited("tid %llu on osd%d timeout\n", 3482 req->r_tid, osd->o_osd); 3483 abort_request(req, -ETIMEDOUT); 3484 } 3485 } 3486 for (p = rb_first(&osd->o_linger_requests); p; p = rb_next(p)) { 3487 struct ceph_osd_linger_request *lreq = 3488 rb_entry(p, struct ceph_osd_linger_request, node); 3489 3490 dout(" lreq %p linger_id %llu is served by osd%d\n", 3491 lreq, lreq->linger_id, osd->o_osd); 3492 found = true; 3493 3494 mutex_lock(&lreq->lock); 3495 if (lreq->is_watch && lreq->committed && !lreq->last_error) 3496 send_linger_ping(lreq); 3497 mutex_unlock(&lreq->lock); 3498 } 3499 3500 if (found) 3501 list_move_tail(&osd->o_keepalive_item, &slow_osds); 3502 } 3503 3504 if (opts->osd_request_timeout) { 3505 for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) { 3506 struct ceph_osd_request *req = 3507 rb_entry(p, struct ceph_osd_request, r_node); 3508 3509 p = rb_next(p); /* abort_request() */ 3510 3511 if (time_before(req->r_start_stamp, expiry_cutoff)) { 3512 pr_err_ratelimited("tid %llu on osd%d timeout\n", 3513 req->r_tid, osdc->homeless_osd.o_osd); 3514 abort_request(req, -ETIMEDOUT); 3515 } 3516 } 3517 } 3518 3519 if (atomic_read(&osdc->num_homeless) || !list_empty(&slow_osds)) 3520 maybe_request_map(osdc); 3521 3522 while (!list_empty(&slow_osds)) { 3523 struct ceph_osd *osd = list_first_entry(&slow_osds, 3524 struct ceph_osd, 3525 o_keepalive_item); 3526 list_del_init(&osd->o_keepalive_item); 3527 ceph_con_keepalive(&osd->o_con); 3528 } 3529 3530 up_write(&osdc->lock); 3531 schedule_delayed_work(&osdc->timeout_work, 3532 osdc->client->options->osd_keepalive_timeout); 3533} 3534 3535static void handle_osds_timeout(struct work_struct *work) 3536{ 3537 struct ceph_osd_client *osdc = 3538 container_of(work, struct ceph_osd_client, 3539 osds_timeout_work.work); 3540 unsigned long delay = osdc->client->options->osd_idle_ttl / 4; 3541 struct ceph_osd *osd, *nosd; 3542 3543 dout("%s osdc %p\n", __func__, osdc); 3544 down_write(&osdc->lock); 3545 list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) { 3546 if (time_before(jiffies, osd->lru_ttl)) 3547 break; 3548 3549 WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests)); 3550 WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests)); 3551 close_osd(osd); 3552 } 3553 3554 up_write(&osdc->lock); 3555 schedule_delayed_work(&osdc->osds_timeout_work, 3556 round_jiffies_relative(delay)); 3557} 3558 3559static int ceph_oloc_decode(void **p, void *end, 3560 struct ceph_object_locator *oloc) 3561{ 3562 u8 struct_v, struct_cv; 3563 u32 len; 3564 void *struct_end; 3565 int ret = 0; 3566 3567 ceph_decode_need(p, end, 1 + 1 + 4, e_inval); 3568 struct_v = ceph_decode_8(p); 3569 struct_cv = ceph_decode_8(p); 3570 if (struct_v < 3) { 3571 pr_warn("got v %d < 3 cv %d of ceph_object_locator\n", 3572 struct_v, struct_cv); 3573 goto e_inval; 3574 } 3575 if (struct_cv > 6) { 3576 pr_warn("got v %d cv %d > 6 of ceph_object_locator\n", 3577 struct_v, struct_cv); 3578 goto e_inval; 3579 } 3580 len = ceph_decode_32(p); 3581 ceph_decode_need(p, end, len, e_inval); 3582 struct_end = *p + len; 3583 3584 oloc->pool = ceph_decode_64(p); 3585 *p += 4; /* skip preferred */ 3586 3587 len = ceph_decode_32(p); 3588 if (len > 0) { 3589 pr_warn("ceph_object_locator::key is set\n"); 3590 goto e_inval; 3591 } 3592 3593 if (struct_v >= 5) { 3594 bool changed = false; 3595 3596 len = ceph_decode_32(p); 3597 if (len > 0) { 3598 ceph_decode_need(p, end, len, e_inval); 3599 if (!oloc->pool_ns || 3600 ceph_compare_string(oloc->pool_ns, *p, len)) 3601 changed = true; 3602 *p += len; 3603 } else { 3604 if (oloc->pool_ns) 3605 changed = true; 3606 } 3607 if (changed) { 3608 /* redirect changes namespace */ 3609 pr_warn("ceph_object_locator::nspace is changed\n"); 3610 goto e_inval; 3611 } 3612 } 3613 3614 if (struct_v >= 6) { 3615 s64 hash = ceph_decode_64(p); 3616 if (hash != -1) { 3617 pr_warn("ceph_object_locator::hash is set\n"); 3618 goto e_inval; 3619 } 3620 } 3621 3622 /* skip the rest */ 3623 *p = struct_end; 3624out: 3625 return ret; 3626 3627e_inval: 3628 ret = -EINVAL; 3629 goto out; 3630} 3631 3632static int ceph_redirect_decode(void **p, void *end, 3633 struct ceph_request_redirect *redir) 3634{ 3635 u8 struct_v, struct_cv; 3636 u32 len; 3637 void *struct_end; 3638 int ret; 3639 3640 ceph_decode_need(p, end, 1 + 1 + 4, e_inval); 3641 struct_v = ceph_decode_8(p); 3642 struct_cv = ceph_decode_8(p); 3643 if (struct_cv > 1) { 3644 pr_warn("got v %d cv %d > 1 of ceph_request_redirect\n", 3645 struct_v, struct_cv); 3646 goto e_inval; 3647 } 3648 len = ceph_decode_32(p); 3649 ceph_decode_need(p, end, len, e_inval); 3650 struct_end = *p + len; 3651 3652 ret = ceph_oloc_decode(p, end, &redir->oloc); 3653 if (ret) 3654 goto out; 3655 3656 len = ceph_decode_32(p); 3657 if (len > 0) { 3658 pr_warn("ceph_request_redirect::object_name is set\n"); 3659 goto e_inval; 3660 } 3661 3662 /* skip the rest */ 3663 *p = struct_end; 3664out: 3665 return ret; 3666 3667e_inval: 3668 ret = -EINVAL; 3669 goto out; 3670} 3671 3672struct MOSDOpReply { 3673 struct ceph_pg pgid; 3674 u64 flags; 3675 int result; 3676 u32 epoch; 3677 int num_ops; 3678 u32 outdata_len[CEPH_OSD_MAX_OPS]; 3679 s32 rval[CEPH_OSD_MAX_OPS]; 3680 int retry_attempt; 3681 struct ceph_eversion replay_version; 3682 u64 user_version; 3683 struct ceph_request_redirect redirect; 3684}; 3685 3686static int decode_MOSDOpReply(const struct ceph_msg *msg, struct MOSDOpReply *m) 3687{ 3688 void *p = msg->front.iov_base; 3689 void *const end = p + msg->front.iov_len; 3690 u16 version = le16_to_cpu(msg->hdr.version); 3691 struct ceph_eversion bad_replay_version; 3692 u8 decode_redir; 3693 u32 len; 3694 int ret; 3695 int i; 3696 3697 ceph_decode_32_safe(&p, end, len, e_inval); 3698 ceph_decode_need(&p, end, len, e_inval); 3699 p += len; /* skip oid */ 3700 3701 ret = ceph_decode_pgid(&p, end, &m->pgid); 3702 if (ret) 3703 return ret; 3704 3705 ceph_decode_64_safe(&p, end, m->flags, e_inval); 3706 ceph_decode_32_safe(&p, end, m->result, e_inval); 3707 ceph_decode_need(&p, end, sizeof(bad_replay_version), e_inval); 3708 memcpy(&bad_replay_version, p, sizeof(bad_replay_version)); 3709 p += sizeof(bad_replay_version); 3710 ceph_decode_32_safe(&p, end, m->epoch, e_inval); 3711 3712 ceph_decode_32_safe(&p, end, m->num_ops, e_inval); 3713 if (m->num_ops > ARRAY_SIZE(m->outdata_len)) 3714 goto e_inval; 3715 3716 ceph_decode_need(&p, end, m->num_ops * sizeof(struct ceph_osd_op), 3717 e_inval); 3718 for (i = 0; i < m->num_ops; i++) { 3719 struct ceph_osd_op *op = p; 3720 3721 m->outdata_len[i] = le32_to_cpu(op->payload_len); 3722 p += sizeof(*op); 3723 } 3724 3725 ceph_decode_32_safe(&p, end, m->retry_attempt, e_inval); 3726 for (i = 0; i < m->num_ops; i++) 3727 ceph_decode_32_safe(&p, end, m->rval[i], e_inval); 3728 3729 if (version >= 5) { 3730 ceph_decode_need(&p, end, sizeof(m->replay_version), e_inval); 3731 memcpy(&m->replay_version, p, sizeof(m->replay_version)); 3732 p += sizeof(m->replay_version); 3733 ceph_decode_64_safe(&p, end, m->user_version, e_inval); 3734 } else { 3735 m->replay_version = bad_replay_version; /* struct */ 3736 m->user_version = le64_to_cpu(m->replay_version.version); 3737 } 3738 3739 if (version >= 6) { 3740 if (version >= 7) 3741 ceph_decode_8_safe(&p, end, decode_redir, e_inval); 3742 else 3743 decode_redir = 1; 3744 } else { 3745 decode_redir = 0; 3746 } 3747 3748 if (decode_redir) { 3749 ret = ceph_redirect_decode(&p, end, &m->redirect); 3750 if (ret) 3751 return ret; 3752 } else { 3753 ceph_oloc_init(&m->redirect.oloc); 3754 } 3755 3756 return 0; 3757 3758e_inval: 3759 return -EINVAL; 3760} 3761 3762/* 3763 * Handle MOSDOpReply. Set ->r_result and call the callback if it is 3764 * specified. 3765 */ 3766static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg) 3767{ 3768 struct ceph_osd_client *osdc = osd->o_osdc; 3769 struct ceph_osd_request *req; 3770 struct MOSDOpReply m; 3771 u64 tid = le64_to_cpu(msg->hdr.tid); 3772 u32 data_len = 0; 3773 int ret; 3774 int i; 3775 3776 dout("%s msg %p tid %llu\n", __func__, msg, tid); 3777 3778 down_read(&osdc->lock); 3779 if (!osd_registered(osd)) { 3780 dout("%s osd%d unknown\n", __func__, osd->o_osd); 3781 goto out_unlock_osdc; 3782 } 3783 WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num)); 3784 3785 mutex_lock(&osd->lock); 3786 req = lookup_request(&osd->o_requests, tid); 3787 if (!req) { 3788 dout("%s osd%d tid %llu unknown\n", __func__, osd->o_osd, tid); 3789 goto out_unlock_session; 3790 } 3791 3792 m.redirect.oloc.pool_ns = req->r_t.target_oloc.pool_ns; 3793 ret = decode_MOSDOpReply(msg, &m); 3794 m.redirect.oloc.pool_ns = NULL; 3795 if (ret) { 3796 pr_err("failed to decode MOSDOpReply for tid %llu: %d\n", 3797 req->r_tid, ret); 3798 ceph_msg_dump(msg); 3799 goto fail_request; 3800 } 3801 dout("%s req %p tid %llu flags 0x%llx pgid %llu.%x epoch %u attempt %d v %u'%llu uv %llu\n", 3802 __func__, req, req->r_tid, m.flags, m.pgid.pool, m.pgid.seed, 3803 m.epoch, m.retry_attempt, le32_to_cpu(m.replay_version.epoch), 3804 le64_to_cpu(m.replay_version.version), m.user_version); 3805 3806 if (m.retry_attempt >= 0) { 3807 if (m.retry_attempt != req->r_attempts - 1) { 3808 dout("req %p tid %llu retry_attempt %d != %d, ignoring\n", 3809 req, req->r_tid, m.retry_attempt, 3810 req->r_attempts - 1); 3811 goto out_unlock_session; 3812 } 3813 } else { 3814 WARN_ON(1); /* MOSDOpReply v4 is assumed */ 3815 } 3816 3817 if (!ceph_oloc_empty(&m.redirect.oloc)) { 3818 dout("req %p tid %llu redirect pool %lld\n", req, req->r_tid, 3819 m.redirect.oloc.pool); 3820 unlink_request(osd, req); 3821 mutex_unlock(&osd->lock); 3822 3823 /* 3824 * Not ceph_oloc_copy() - changing pool_ns is not 3825 * supported. 3826 */ 3827 req->r_t.target_oloc.pool = m.redirect.oloc.pool; 3828 req->r_flags |= CEPH_OSD_FLAG_REDIRECTED | 3829 CEPH_OSD_FLAG_IGNORE_OVERLAY | 3830 CEPH_OSD_FLAG_IGNORE_CACHE; 3831 req->r_tid = 0; 3832 __submit_request(req, false); 3833 goto out_unlock_osdc; 3834 } 3835 3836 if (m.result == -EAGAIN) { 3837 dout("req %p tid %llu EAGAIN\n", req, req->r_tid); 3838 unlink_request(osd, req); 3839 mutex_unlock(&osd->lock); 3840 3841 /* 3842 * The object is missing on the replica or not (yet) 3843 * readable. Clear pgid to force a resend to the primary 3844 * via legacy_change. 3845 */ 3846 req->r_t.pgid.pool = 0; 3847 req->r_t.pgid.seed = 0; 3848 WARN_ON(!req->r_t.used_replica); 3849 req->r_flags &= ~(CEPH_OSD_FLAG_BALANCE_READS | 3850 CEPH_OSD_FLAG_LOCALIZE_READS); 3851 req->r_tid = 0; 3852 __submit_request(req, false); 3853 goto out_unlock_osdc; 3854 } 3855 3856 if (m.num_ops != req->r_num_ops) { 3857 pr_err("num_ops %d != %d for tid %llu\n", m.num_ops, 3858 req->r_num_ops, req->r_tid); 3859 goto fail_request; 3860 } 3861 for (i = 0; i < req->r_num_ops; i++) { 3862 dout(" req %p tid %llu op %d rval %d len %u\n", req, 3863 req->r_tid, i, m.rval[i], m.outdata_len[i]); 3864 req->r_ops[i].rval = m.rval[i]; 3865 req->r_ops[i].outdata_len = m.outdata_len[i]; 3866 data_len += m.outdata_len[i]; 3867 } 3868 if (data_len != le32_to_cpu(msg->hdr.data_len)) { 3869 pr_err("sum of lens %u != %u for tid %llu\n", data_len, 3870 le32_to_cpu(msg->hdr.data_len), req->r_tid); 3871 goto fail_request; 3872 } 3873 dout("%s req %p tid %llu result %d data_len %u\n", __func__, 3874 req, req->r_tid, m.result, data_len); 3875 3876 /* 3877 * Since we only ever request ONDISK, we should only ever get 3878 * one (type of) reply back. 3879 */ 3880 WARN_ON(!(m.flags & CEPH_OSD_FLAG_ONDISK)); 3881 req->r_version = m.user_version; 3882 req->r_result = m.result ?: data_len; 3883 finish_request(req); 3884 mutex_unlock(&osd->lock); 3885 up_read(&osdc->lock); 3886 3887 __complete_request(req); 3888 return; 3889 3890fail_request: 3891 complete_request(req, -EIO); 3892out_unlock_session: 3893 mutex_unlock(&osd->lock); 3894out_unlock_osdc: 3895 up_read(&osdc->lock); 3896} 3897 3898static void set_pool_was_full(struct ceph_osd_client *osdc) 3899{ 3900 struct rb_node *n; 3901 3902 for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) { 3903 struct ceph_pg_pool_info *pi = 3904 rb_entry(n, struct ceph_pg_pool_info, node); 3905 3906 pi->was_full = __pool_full(pi); 3907 } 3908} 3909 3910static bool pool_cleared_full(struct ceph_osd_client *osdc, s64 pool_id) 3911{ 3912 struct ceph_pg_pool_info *pi; 3913 3914 pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id); 3915 if (!pi) 3916 return false; 3917 3918 return pi->was_full && !__pool_full(pi); 3919} 3920 3921static enum calc_target_result 3922recalc_linger_target(struct ceph_osd_linger_request *lreq) 3923{ 3924 struct ceph_osd_client *osdc = lreq->osdc; 3925 enum calc_target_result ct_res; 3926 3927 ct_res = calc_target(osdc, &lreq->t, true); 3928 if (ct_res == CALC_TARGET_NEED_RESEND) { 3929 struct ceph_osd *osd; 3930 3931 osd = lookup_create_osd(osdc, lreq->t.osd, true); 3932 if (osd != lreq->osd) { 3933 unlink_linger(lreq->osd, lreq); 3934 link_linger(osd, lreq); 3935 } 3936 } 3937 3938 return ct_res; 3939} 3940 3941/* 3942 * Requeue requests whose mapping to an OSD has changed. 3943 */ 3944static void scan_requests(struct ceph_osd *osd, 3945 bool force_resend, 3946 bool cleared_full, 3947 bool check_pool_cleared_full, 3948 struct rb_root *need_resend, 3949 struct list_head *need_resend_linger) 3950{ 3951 struct ceph_osd_client *osdc = osd->o_osdc; 3952 struct rb_node *n; 3953 bool force_resend_writes; 3954 3955 for (n = rb_first(&osd->o_linger_requests); n; ) { 3956 struct ceph_osd_linger_request *lreq = 3957 rb_entry(n, struct ceph_osd_linger_request, node); 3958 enum calc_target_result ct_res; 3959 3960 n = rb_next(n); /* recalc_linger_target() */ 3961 3962 dout("%s lreq %p linger_id %llu\n", __func__, lreq, 3963 lreq->linger_id); 3964 ct_res = recalc_linger_target(lreq); 3965 switch (ct_res) { 3966 case CALC_TARGET_NO_ACTION: 3967 force_resend_writes = cleared_full || 3968 (check_pool_cleared_full && 3969 pool_cleared_full(osdc, lreq->t.base_oloc.pool)); 3970 if (!force_resend && !force_resend_writes) 3971 break; 3972 3973 fallthrough; 3974 case CALC_TARGET_NEED_RESEND: 3975 cancel_linger_map_check(lreq); 3976 /* 3977 * scan_requests() for the previous epoch(s) 3978 * may have already added it to the list, since 3979 * it's not unlinked here. 3980 */ 3981 if (list_empty(&lreq->scan_item)) 3982 list_add_tail(&lreq->scan_item, need_resend_linger); 3983 break; 3984 case CALC_TARGET_POOL_DNE: 3985 list_del_init(&lreq->scan_item); 3986 check_linger_pool_dne(lreq); 3987 break; 3988 } 3989 } 3990 3991 for (n = rb_first(&osd->o_requests); n; ) { 3992 struct ceph_osd_request *req = 3993 rb_entry(n, struct ceph_osd_request, r_node); 3994 enum calc_target_result ct_res; 3995 3996 n = rb_next(n); /* unlink_request(), check_pool_dne() */ 3997 3998 dout("%s req %p tid %llu\n", __func__, req, req->r_tid); 3999 ct_res = calc_target(osdc, &req->r_t, false); 4000 switch (ct_res) { 4001 case CALC_TARGET_NO_ACTION: 4002 force_resend_writes = cleared_full || 4003 (check_pool_cleared_full && 4004 pool_cleared_full(osdc, req->r_t.base_oloc.pool)); 4005 if (!force_resend && 4006 (!(req->r_flags & CEPH_OSD_FLAG_WRITE) || 4007 !force_resend_writes)) 4008 break; 4009 4010 fallthrough; 4011 case CALC_TARGET_NEED_RESEND: 4012 cancel_map_check(req); 4013 unlink_request(osd, req); 4014 insert_request(need_resend, req); 4015 break; 4016 case CALC_TARGET_POOL_DNE: 4017 check_pool_dne(req); 4018 break; 4019 } 4020 } 4021} 4022 4023static int handle_one_map(struct ceph_osd_client *osdc, 4024 void *p, void *end, bool incremental, 4025 struct rb_root *need_resend, 4026 struct list_head *need_resend_linger) 4027{ 4028 struct ceph_osdmap *newmap; 4029 struct rb_node *n; 4030 bool skipped_map = false; 4031 bool was_full; 4032 4033 was_full = ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL); 4034 set_pool_was_full(osdc); 4035 4036 if (incremental) 4037 newmap = osdmap_apply_incremental(&p, end, 4038 ceph_msgr2(osdc->client), 4039 osdc->osdmap); 4040 else 4041 newmap = ceph_osdmap_decode(&p, end, ceph_msgr2(osdc->client)); 4042 if (IS_ERR(newmap)) 4043 return PTR_ERR(newmap); 4044 4045 if (newmap != osdc->osdmap) { 4046 /* 4047 * Preserve ->was_full before destroying the old map. 4048 * For pools that weren't in the old map, ->was_full 4049 * should be false. 4050 */ 4051 for (n = rb_first(&newmap->pg_pools); n; n = rb_next(n)) { 4052 struct ceph_pg_pool_info *pi = 4053 rb_entry(n, struct ceph_pg_pool_info, node); 4054 struct ceph_pg_pool_info *old_pi; 4055 4056 old_pi = ceph_pg_pool_by_id(osdc->osdmap, pi->id); 4057 if (old_pi) 4058 pi->was_full = old_pi->was_full; 4059 else 4060 WARN_ON(pi->was_full); 4061 } 4062 4063 if (osdc->osdmap->epoch && 4064 osdc->osdmap->epoch + 1 < newmap->epoch) { 4065 WARN_ON(incremental); 4066 skipped_map = true; 4067 } 4068 4069 ceph_osdmap_destroy(osdc->osdmap); 4070 osdc->osdmap = newmap; 4071 } 4072 4073 was_full &= !ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL); 4074 scan_requests(&osdc->homeless_osd, skipped_map, was_full, true, 4075 need_resend, need_resend_linger); 4076 4077 for (n = rb_first(&osdc->osds); n; ) { 4078 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node); 4079 4080 n = rb_next(n); /* close_osd() */ 4081 4082 scan_requests(osd, skipped_map, was_full, true, need_resend, 4083 need_resend_linger); 4084 if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) || 4085 memcmp(&osd->o_con.peer_addr, 4086 ceph_osd_addr(osdc->osdmap, osd->o_osd), 4087 sizeof(struct ceph_entity_addr))) 4088 close_osd(osd); 4089 } 4090 4091 return 0; 4092} 4093 4094static void kick_requests(struct ceph_osd_client *osdc, 4095 struct rb_root *need_resend, 4096 struct list_head *need_resend_linger) 4097{ 4098 struct ceph_osd_linger_request *lreq, *nlreq; 4099 enum calc_target_result ct_res; 4100 struct rb_node *n; 4101 4102 /* make sure need_resend targets reflect latest map */ 4103 for (n = rb_first(need_resend); n; ) { 4104 struct ceph_osd_request *req = 4105 rb_entry(n, struct ceph_osd_request, r_node); 4106 4107 n = rb_next(n); 4108 4109 if (req->r_t.epoch < osdc->osdmap->epoch) { 4110 ct_res = calc_target(osdc, &req->r_t, false); 4111 if (ct_res == CALC_TARGET_POOL_DNE) { 4112 erase_request(need_resend, req); 4113 check_pool_dne(req); 4114 } 4115 } 4116 } 4117 4118 for (n = rb_first(need_resend); n; ) { 4119 struct ceph_osd_request *req = 4120 rb_entry(n, struct ceph_osd_request, r_node); 4121 struct ceph_osd *osd; 4122 4123 n = rb_next(n); 4124 erase_request(need_resend, req); /* before link_request() */ 4125 4126 osd = lookup_create_osd(osdc, req->r_t.osd, true); 4127 link_request(osd, req); 4128 if (!req->r_linger) { 4129 if (!osd_homeless(osd) && !req->r_t.paused) 4130 send_request(req); 4131 } else { 4132 cancel_linger_request(req); 4133 } 4134 } 4135 4136 list_for_each_entry_safe(lreq, nlreq, need_resend_linger, scan_item) { 4137 if (!osd_homeless(lreq->osd)) 4138 send_linger(lreq); 4139 4140 list_del_init(&lreq->scan_item); 4141 } 4142} 4143 4144/* 4145 * Process updated osd map. 4146 * 4147 * The message contains any number of incremental and full maps, normally 4148 * indicating some sort of topology change in the cluster. Kick requests 4149 * off to different OSDs as needed. 4150 */ 4151void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg) 4152{ 4153 void *p = msg->front.iov_base; 4154 void *const end = p + msg->front.iov_len; 4155 u32 nr_maps, maplen; 4156 u32 epoch; 4157 struct ceph_fsid fsid; 4158 struct rb_root need_resend = RB_ROOT; 4159 LIST_HEAD(need_resend_linger); 4160 bool handled_incremental = false; 4161 bool was_pauserd, was_pausewr; 4162 bool pauserd, pausewr; 4163 int err; 4164 4165 dout("%s have %u\n", __func__, osdc->osdmap->epoch); 4166 down_write(&osdc->lock); 4167 4168 /* verify fsid */ 4169 ceph_decode_need(&p, end, sizeof(fsid), bad); 4170 ceph_decode_copy(&p, &fsid, sizeof(fsid)); 4171 if (ceph_check_fsid(osdc->client, &fsid) < 0) 4172 goto bad; 4173 4174 was_pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD); 4175 was_pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) || 4176 ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || 4177 have_pool_full(osdc); 4178 4179 /* incremental maps */ 4180 ceph_decode_32_safe(&p, end, nr_maps, bad); 4181 dout(" %d inc maps\n", nr_maps); 4182 while (nr_maps > 0) { 4183 ceph_decode_need(&p, end, 2*sizeof(u32), bad); 4184 epoch = ceph_decode_32(&p); 4185 maplen = ceph_decode_32(&p); 4186 ceph_decode_need(&p, end, maplen, bad); 4187 if (osdc->osdmap->epoch && 4188 osdc->osdmap->epoch + 1 == epoch) { 4189 dout("applying incremental map %u len %d\n", 4190 epoch, maplen); 4191 err = handle_one_map(osdc, p, p + maplen, true, 4192 &need_resend, &need_resend_linger); 4193 if (err) 4194 goto bad; 4195 handled_incremental = true; 4196 } else { 4197 dout("ignoring incremental map %u len %d\n", 4198 epoch, maplen); 4199 } 4200 p += maplen; 4201 nr_maps--; 4202 } 4203 if (handled_incremental) 4204 goto done; 4205 4206 /* full maps */ 4207 ceph_decode_32_safe(&p, end, nr_maps, bad); 4208 dout(" %d full maps\n", nr_maps); 4209 while (nr_maps) { 4210 ceph_decode_need(&p, end, 2*sizeof(u32), bad); 4211 epoch = ceph_decode_32(&p); 4212 maplen = ceph_decode_32(&p); 4213 ceph_decode_need(&p, end, maplen, bad); 4214 if (nr_maps > 1) { 4215 dout("skipping non-latest full map %u len %d\n", 4216 epoch, maplen); 4217 } else if (osdc->osdmap->epoch >= epoch) { 4218 dout("skipping full map %u len %d, " 4219 "older than our %u\n", epoch, maplen, 4220 osdc->osdmap->epoch); 4221 } else { 4222 dout("taking full map %u len %d\n", epoch, maplen); 4223 err = handle_one_map(osdc, p, p + maplen, false, 4224 &need_resend, &need_resend_linger); 4225 if (err) 4226 goto bad; 4227 } 4228 p += maplen; 4229 nr_maps--; 4230 } 4231 4232done: 4233 /* 4234 * subscribe to subsequent osdmap updates if full to ensure 4235 * we find out when we are no longer full and stop returning 4236 * ENOSPC. 4237 */ 4238 pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD); 4239 pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) || 4240 ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || 4241 have_pool_full(osdc); 4242 if (was_pauserd || was_pausewr || pauserd || pausewr || 4243 osdc->osdmap->epoch < osdc->epoch_barrier) 4244 maybe_request_map(osdc); 4245 4246 kick_requests(osdc, &need_resend, &need_resend_linger); 4247 4248 ceph_osdc_abort_on_full(osdc); 4249 ceph_monc_got_map(&osdc->client->monc, CEPH_SUB_OSDMAP, 4250 osdc->osdmap->epoch); 4251 up_write(&osdc->lock); 4252 wake_up_all(&osdc->client->auth_wq); 4253 return; 4254 4255bad: 4256 pr_err("osdc handle_map corrupt msg\n"); 4257 ceph_msg_dump(msg); 4258 up_write(&osdc->lock); 4259} 4260 4261/* 4262 * Resubmit requests pending on the given osd. 4263 */ 4264static void kick_osd_requests(struct ceph_osd *osd) 4265{ 4266 struct rb_node *n; 4267 4268 clear_backoffs(osd); 4269 4270 for (n = rb_first(&osd->o_requests); n; ) { 4271 struct ceph_osd_request *req = 4272 rb_entry(n, struct ceph_osd_request, r_node); 4273 4274 n = rb_next(n); /* cancel_linger_request() */ 4275 4276 if (!req->r_linger) { 4277 if (!req->r_t.paused) 4278 send_request(req); 4279 } else { 4280 cancel_linger_request(req); 4281 } 4282 } 4283 for (n = rb_first(&osd->o_linger_requests); n; n = rb_next(n)) { 4284 struct ceph_osd_linger_request *lreq = 4285 rb_entry(n, struct ceph_osd_linger_request, node); 4286 4287 send_linger(lreq); 4288 } 4289} 4290 4291/* 4292 * If the osd connection drops, we need to resubmit all requests. 4293 */ 4294static void osd_fault(struct ceph_connection *con) 4295{ 4296 struct ceph_osd *osd = con->private; 4297 struct ceph_osd_client *osdc = osd->o_osdc; 4298 4299 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd); 4300 4301 down_write(&osdc->lock); 4302 if (!osd_registered(osd)) { 4303 dout("%s osd%d unknown\n", __func__, osd->o_osd); 4304 goto out_unlock; 4305 } 4306 4307 if (!reopen_osd(osd)) 4308 kick_osd_requests(osd); 4309 maybe_request_map(osdc); 4310 4311out_unlock: 4312 up_write(&osdc->lock); 4313} 4314 4315struct MOSDBackoff { 4316 struct ceph_spg spgid; 4317 u32 map_epoch; 4318 u8 op; 4319 u64 id; 4320 struct ceph_hobject_id *begin; 4321 struct ceph_hobject_id *end; 4322}; 4323 4324static int decode_MOSDBackoff(const struct ceph_msg *msg, struct MOSDBackoff *m) 4325{ 4326 void *p = msg->front.iov_base; 4327 void *const end = p + msg->front.iov_len; 4328 u8 struct_v; 4329 u32 struct_len; 4330 int ret; 4331 4332 ret = ceph_start_decoding(&p, end, 1, "spg_t", &struct_v, &struct_len); 4333 if (ret) 4334 return ret; 4335 4336 ret = ceph_decode_pgid(&p, end, &m->spgid.pgid); 4337 if (ret) 4338 return ret; 4339 4340 ceph_decode_8_safe(&p, end, m->spgid.shard, e_inval); 4341 ceph_decode_32_safe(&p, end, m->map_epoch, e_inval); 4342 ceph_decode_8_safe(&p, end, m->op, e_inval); 4343 ceph_decode_64_safe(&p, end, m->id, e_inval); 4344 4345 m->begin = kzalloc(sizeof(*m->begin), GFP_NOIO); 4346 if (!m->begin) 4347 return -ENOMEM; 4348 4349 ret = decode_hoid(&p, end, m->begin); 4350 if (ret) { 4351 free_hoid(m->begin); 4352 return ret; 4353 } 4354 4355 m->end = kzalloc(sizeof(*m->end), GFP_NOIO); 4356 if (!m->end) { 4357 free_hoid(m->begin); 4358 return -ENOMEM; 4359 } 4360 4361 ret = decode_hoid(&p, end, m->end); 4362 if (ret) { 4363 free_hoid(m->begin); 4364 free_hoid(m->end); 4365 return ret; 4366 } 4367 4368 return 0; 4369 4370e_inval: 4371 return -EINVAL; 4372} 4373 4374static struct ceph_msg *create_backoff_message( 4375 const struct ceph_osd_backoff *backoff, 4376 u32 map_epoch) 4377{ 4378 struct ceph_msg *msg; 4379 void *p, *end; 4380 int msg_size; 4381 4382 msg_size = CEPH_ENCODING_START_BLK_LEN + 4383 CEPH_PGID_ENCODING_LEN + 1; /* spgid */ 4384 msg_size += 4 + 1 + 8; /* map_epoch, op, id */ 4385 msg_size += CEPH_ENCODING_START_BLK_LEN + 4386 hoid_encoding_size(backoff->begin); 4387 msg_size += CEPH_ENCODING_START_BLK_LEN + 4388 hoid_encoding_size(backoff->end); 4389 4390 msg = ceph_msg_new(CEPH_MSG_OSD_BACKOFF, msg_size, GFP_NOIO, true); 4391 if (!msg) 4392 return NULL; 4393 4394 p = msg->front.iov_base; 4395 end = p + msg->front_alloc_len; 4396 4397 encode_spgid(&p, &backoff->spgid); 4398 ceph_encode_32(&p, map_epoch); 4399 ceph_encode_8(&p, CEPH_OSD_BACKOFF_OP_ACK_BLOCK); 4400 ceph_encode_64(&p, backoff->id); 4401 encode_hoid(&p, end, backoff->begin); 4402 encode_hoid(&p, end, backoff->end); 4403 BUG_ON(p != end); 4404 4405 msg->front.iov_len = p - msg->front.iov_base; 4406 msg->hdr.version = cpu_to_le16(1); /* MOSDBackoff v1 */ 4407 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len); 4408 4409 return msg; 4410} 4411 4412static void handle_backoff_block(struct ceph_osd *osd, struct MOSDBackoff *m) 4413{ 4414 struct ceph_spg_mapping *spg; 4415 struct ceph_osd_backoff *backoff; 4416 struct ceph_msg *msg; 4417 4418 dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd, 4419 m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id); 4420 4421 spg = lookup_spg_mapping(&osd->o_backoff_mappings, &m->spgid); 4422 if (!spg) { 4423 spg = alloc_spg_mapping(); 4424 if (!spg) { 4425 pr_err("%s failed to allocate spg\n", __func__); 4426 return; 4427 } 4428 spg->spgid = m->spgid; /* struct */ 4429 insert_spg_mapping(&osd->o_backoff_mappings, spg); 4430 } 4431 4432 backoff = alloc_backoff(); 4433 if (!backoff) { 4434 pr_err("%s failed to allocate backoff\n", __func__); 4435 return; 4436 } 4437 backoff->spgid = m->spgid; /* struct */ 4438 backoff->id = m->id; 4439 backoff->begin = m->begin; 4440 m->begin = NULL; /* backoff now owns this */ 4441 backoff->end = m->end; 4442 m->end = NULL; /* ditto */ 4443 4444 insert_backoff(&spg->backoffs, backoff); 4445 insert_backoff_by_id(&osd->o_backoffs_by_id, backoff); 4446 4447 /* 4448 * Ack with original backoff's epoch so that the OSD can 4449 * discard this if there was a PG split. 4450 */ 4451 msg = create_backoff_message(backoff, m->map_epoch); 4452 if (!msg) { 4453 pr_err("%s failed to allocate msg\n", __func__); 4454 return; 4455 } 4456 ceph_con_send(&osd->o_con, msg); 4457} 4458 4459static bool target_contained_by(const struct ceph_osd_request_target *t, 4460 const struct ceph_hobject_id *begin, 4461 const struct ceph_hobject_id *end) 4462{ 4463 struct ceph_hobject_id hoid; 4464 int cmp; 4465 4466 hoid_fill_from_target(&hoid, t); 4467 cmp = hoid_compare(&hoid, begin); 4468 return !cmp || (cmp > 0 && hoid_compare(&hoid, end) < 0); 4469} 4470 4471static void handle_backoff_unblock(struct ceph_osd *osd, 4472 const struct MOSDBackoff *m) 4473{ 4474 struct ceph_spg_mapping *spg; 4475 struct ceph_osd_backoff *backoff; 4476 struct rb_node *n; 4477 4478 dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd, 4479 m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id); 4480 4481 backoff = lookup_backoff_by_id(&osd->o_backoffs_by_id, m->id); 4482 if (!backoff) { 4483 pr_err("%s osd%d spgid %llu.%xs%d id %llu backoff dne\n", 4484 __func__, osd->o_osd, m->spgid.pgid.pool, 4485 m->spgid.pgid.seed, m->spgid.shard, m->id); 4486 return; 4487 } 4488 4489 if (hoid_compare(backoff->begin, m->begin) && 4490 hoid_compare(backoff->end, m->end)) { 4491 pr_err("%s osd%d spgid %llu.%xs%d id %llu bad range?\n", 4492 __func__, osd->o_osd, m->spgid.pgid.pool, 4493 m->spgid.pgid.seed, m->spgid.shard, m->id); 4494 /* unblock it anyway... */ 4495 } 4496 4497 spg = lookup_spg_mapping(&osd->o_backoff_mappings, &backoff->spgid); 4498 BUG_ON(!spg); 4499 4500 erase_backoff(&spg->backoffs, backoff); 4501 erase_backoff_by_id(&osd->o_backoffs_by_id, backoff); 4502 free_backoff(backoff); 4503 4504 if (RB_EMPTY_ROOT(&spg->backoffs)) { 4505 erase_spg_mapping(&osd->o_backoff_mappings, spg); 4506 free_spg_mapping(spg); 4507 } 4508 4509 for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) { 4510 struct ceph_osd_request *req = 4511 rb_entry(n, struct ceph_osd_request, r_node); 4512 4513 if (!ceph_spg_compare(&req->r_t.spgid, &m->spgid)) { 4514 /* 4515 * Match against @m, not @backoff -- the PG may 4516 * have split on the OSD. 4517 */ 4518 if (target_contained_by(&req->r_t, m->begin, m->end)) { 4519 /* 4520 * If no other installed backoff applies, 4521 * resend. 4522 */ 4523 send_request(req); 4524 } 4525 } 4526 } 4527} 4528 4529static void handle_backoff(struct ceph_osd *osd, struct ceph_msg *msg) 4530{ 4531 struct ceph_osd_client *osdc = osd->o_osdc; 4532 struct MOSDBackoff m; 4533 int ret; 4534 4535 down_read(&osdc->lock); 4536 if (!osd_registered(osd)) { 4537 dout("%s osd%d unknown\n", __func__, osd->o_osd); 4538 up_read(&osdc->lock); 4539 return; 4540 } 4541 WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num)); 4542 4543 mutex_lock(&osd->lock); 4544 ret = decode_MOSDBackoff(msg, &m); 4545 if (ret) { 4546 pr_err("failed to decode MOSDBackoff: %d\n", ret); 4547 ceph_msg_dump(msg); 4548 goto out_unlock; 4549 } 4550 4551 switch (m.op) { 4552 case CEPH_OSD_BACKOFF_OP_BLOCK: 4553 handle_backoff_block(osd, &m); 4554 break; 4555 case CEPH_OSD_BACKOFF_OP_UNBLOCK: 4556 handle_backoff_unblock(osd, &m); 4557 break; 4558 default: 4559 pr_err("%s osd%d unknown op %d\n", __func__, osd->o_osd, m.op); 4560 } 4561 4562 free_hoid(m.begin); 4563 free_hoid(m.end); 4564 4565out_unlock: 4566 mutex_unlock(&osd->lock); 4567 up_read(&osdc->lock); 4568} 4569 4570/* 4571 * Process osd watch notifications 4572 */ 4573static void handle_watch_notify(struct ceph_osd_client *osdc, 4574 struct ceph_msg *msg) 4575{ 4576 void *p = msg->front.iov_base; 4577 void *const end = p + msg->front.iov_len; 4578 struct ceph_osd_linger_request *lreq; 4579 struct linger_work *lwork; 4580 u8 proto_ver, opcode; 4581 u64 cookie, notify_id; 4582 u64 notifier_id = 0; 4583 s32 return_code = 0; 4584 void *payload = NULL; 4585 u32 payload_len = 0; 4586 4587 ceph_decode_8_safe(&p, end, proto_ver, bad); 4588 ceph_decode_8_safe(&p, end, opcode, bad); 4589 ceph_decode_64_safe(&p, end, cookie, bad); 4590 p += 8; /* skip ver */ 4591 ceph_decode_64_safe(&p, end, notify_id, bad); 4592 4593 if (proto_ver >= 1) { 4594 ceph_decode_32_safe(&p, end, payload_len, bad); 4595 ceph_decode_need(&p, end, payload_len, bad); 4596 payload = p; 4597 p += payload_len; 4598 } 4599 4600 if (le16_to_cpu(msg->hdr.version) >= 2) 4601 ceph_decode_32_safe(&p, end, return_code, bad); 4602 4603 if (le16_to_cpu(msg->hdr.version) >= 3) 4604 ceph_decode_64_safe(&p, end, notifier_id, bad); 4605 4606 down_read(&osdc->lock); 4607 lreq = lookup_linger_osdc(&osdc->linger_requests, cookie); 4608 if (!lreq) { 4609 dout("%s opcode %d cookie %llu dne\n", __func__, opcode, 4610 cookie); 4611 goto out_unlock_osdc; 4612 } 4613 4614 mutex_lock(&lreq->lock); 4615 dout("%s opcode %d cookie %llu lreq %p is_watch %d\n", __func__, 4616 opcode, cookie, lreq, lreq->is_watch); 4617 if (opcode == CEPH_WATCH_EVENT_DISCONNECT) { 4618 if (!lreq->last_error) { 4619 lreq->last_error = -ENOTCONN; 4620 queue_watch_error(lreq); 4621 } 4622 } else if (!lreq->is_watch) { 4623 /* CEPH_WATCH_EVENT_NOTIFY_COMPLETE */ 4624 if (lreq->notify_id && lreq->notify_id != notify_id) { 4625 dout("lreq %p notify_id %llu != %llu, ignoring\n", lreq, 4626 lreq->notify_id, notify_id); 4627 } else if (!completion_done(&lreq->notify_finish_wait)) { 4628 struct ceph_msg_data *data = 4629 msg->num_data_items ? &msg->data[0] : NULL; 4630 4631 if (data) { 4632 if (lreq->preply_pages) { 4633 WARN_ON(data->type != 4634 CEPH_MSG_DATA_PAGES); 4635 *lreq->preply_pages = data->pages; 4636 *lreq->preply_len = data->length; 4637 data->own_pages = false; 4638 } 4639 } 4640 lreq->notify_finish_error = return_code; 4641 complete_all(&lreq->notify_finish_wait); 4642 } 4643 } else { 4644 /* CEPH_WATCH_EVENT_NOTIFY */ 4645 lwork = lwork_alloc(lreq, do_watch_notify); 4646 if (!lwork) { 4647 pr_err("failed to allocate notify-lwork\n"); 4648 goto out_unlock_lreq; 4649 } 4650 4651 lwork->notify.notify_id = notify_id; 4652 lwork->notify.notifier_id = notifier_id; 4653 lwork->notify.payload = payload; 4654 lwork->notify.payload_len = payload_len; 4655 lwork->notify.msg = ceph_msg_get(msg); 4656 lwork_queue(lwork); 4657 } 4658 4659out_unlock_lreq: 4660 mutex_unlock(&lreq->lock); 4661out_unlock_osdc: 4662 up_read(&osdc->lock); 4663 return; 4664 4665bad: 4666 pr_err("osdc handle_watch_notify corrupt msg\n"); 4667} 4668 4669/* 4670 * Register request, send initial attempt. 4671 */ 4672void ceph_osdc_start_request(struct ceph_osd_client *osdc, 4673 struct ceph_osd_request *req) 4674{ 4675 down_read(&osdc->lock); 4676 submit_request(req, false); 4677 up_read(&osdc->lock); 4678} 4679EXPORT_SYMBOL(ceph_osdc_start_request); 4680 4681/* 4682 * Unregister request. If @req was registered, it isn't completed: 4683 * r_result isn't set and __complete_request() isn't invoked. 4684 * 4685 * If @req wasn't registered, this call may have raced with 4686 * handle_reply(), in which case r_result would already be set and 4687 * __complete_request() would be getting invoked, possibly even 4688 * concurrently with this call. 4689 */ 4690void ceph_osdc_cancel_request(struct ceph_osd_request *req) 4691{ 4692 struct ceph_osd_client *osdc = req->r_osdc; 4693 4694 down_write(&osdc->lock); 4695 if (req->r_osd) 4696 cancel_request(req); 4697 up_write(&osdc->lock); 4698} 4699EXPORT_SYMBOL(ceph_osdc_cancel_request); 4700 4701/* 4702 * @timeout: in jiffies, 0 means "wait forever" 4703 */ 4704static int wait_request_timeout(struct ceph_osd_request *req, 4705 unsigned long timeout) 4706{ 4707 long left; 4708 4709 dout("%s req %p tid %llu\n", __func__, req, req->r_tid); 4710 left = wait_for_completion_killable_timeout(&req->r_completion, 4711 ceph_timeout_jiffies(timeout)); 4712 if (left <= 0) { 4713 left = left ?: -ETIMEDOUT; 4714 ceph_osdc_cancel_request(req); 4715 } else { 4716 left = req->r_result; /* completed */ 4717 } 4718 4719 return left; 4720} 4721 4722/* 4723 * wait for a request to complete 4724 */ 4725int ceph_osdc_wait_request(struct ceph_osd_client *osdc, 4726 struct ceph_osd_request *req) 4727{ 4728 return wait_request_timeout(req, 0); 4729} 4730EXPORT_SYMBOL(ceph_osdc_wait_request); 4731 4732/* 4733 * sync - wait for all in-flight requests to flush. avoid starvation. 4734 */ 4735void ceph_osdc_sync(struct ceph_osd_client *osdc) 4736{ 4737 struct rb_node *n, *p; 4738 u64 last_tid = atomic64_read(&osdc->last_tid); 4739 4740again: 4741 down_read(&osdc->lock); 4742 for (n = rb_first(&osdc->osds); n; n = rb_next(n)) { 4743 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node); 4744 4745 mutex_lock(&osd->lock); 4746 for (p = rb_first(&osd->o_requests); p; p = rb_next(p)) { 4747 struct ceph_osd_request *req = 4748 rb_entry(p, struct ceph_osd_request, r_node); 4749 4750 if (req->r_tid > last_tid) 4751 break; 4752 4753 if (!(req->r_flags & CEPH_OSD_FLAG_WRITE)) 4754 continue; 4755 4756 ceph_osdc_get_request(req); 4757 mutex_unlock(&osd->lock); 4758 up_read(&osdc->lock); 4759 dout("%s waiting on req %p tid %llu last_tid %llu\n", 4760 __func__, req, req->r_tid, last_tid); 4761 wait_for_completion(&req->r_completion); 4762 ceph_osdc_put_request(req); 4763 goto again; 4764 } 4765 4766 mutex_unlock(&osd->lock); 4767 } 4768 4769 up_read(&osdc->lock); 4770 dout("%s done last_tid %llu\n", __func__, last_tid); 4771} 4772EXPORT_SYMBOL(ceph_osdc_sync); 4773 4774/* 4775 * Returns a handle, caller owns a ref. 4776 */ 4777struct ceph_osd_linger_request * 4778ceph_osdc_watch(struct ceph_osd_client *osdc, 4779 struct ceph_object_id *oid, 4780 struct ceph_object_locator *oloc, 4781 rados_watchcb2_t wcb, 4782 rados_watcherrcb_t errcb, 4783 void *data) 4784{ 4785 struct ceph_osd_linger_request *lreq; 4786 int ret; 4787 4788 lreq = linger_alloc(osdc); 4789 if (!lreq) 4790 return ERR_PTR(-ENOMEM); 4791 4792 lreq->is_watch = true; 4793 lreq->wcb = wcb; 4794 lreq->errcb = errcb; 4795 lreq->data = data; 4796 lreq->watch_valid_thru = jiffies; 4797 4798 ceph_oid_copy(&lreq->t.base_oid, oid); 4799 ceph_oloc_copy(&lreq->t.base_oloc, oloc); 4800 lreq->t.flags = CEPH_OSD_FLAG_WRITE; 4801 ktime_get_real_ts64(&lreq->mtime); 4802 4803 linger_submit(lreq); 4804 ret = linger_reg_commit_wait(lreq); 4805 if (ret) { 4806 linger_cancel(lreq); 4807 goto err_put_lreq; 4808 } 4809 4810 return lreq; 4811 4812err_put_lreq: 4813 linger_put(lreq); 4814 return ERR_PTR(ret); 4815} 4816EXPORT_SYMBOL(ceph_osdc_watch); 4817 4818/* 4819 * Releases a ref. 4820 * 4821 * Times out after mount_timeout to preserve rbd unmap behaviour 4822 * introduced in 2894e1d76974 ("rbd: timeout watch teardown on unmap 4823 * with mount_timeout"). 4824 */ 4825int ceph_osdc_unwatch(struct ceph_osd_client *osdc, 4826 struct ceph_osd_linger_request *lreq) 4827{ 4828 struct ceph_options *opts = osdc->client->options; 4829 struct ceph_osd_request *req; 4830 int ret; 4831 4832 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO); 4833 if (!req) 4834 return -ENOMEM; 4835 4836 ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid); 4837 ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc); 4838 req->r_flags = CEPH_OSD_FLAG_WRITE; 4839 ktime_get_real_ts64(&req->r_mtime); 4840 osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_UNWATCH, 4841 lreq->linger_id, 0); 4842 4843 ret = ceph_osdc_alloc_messages(req, GFP_NOIO); 4844 if (ret) 4845 goto out_put_req; 4846 4847 ceph_osdc_start_request(osdc, req); 4848 linger_cancel(lreq); 4849 linger_put(lreq); 4850 ret = wait_request_timeout(req, opts->mount_timeout); 4851 4852out_put_req: 4853 ceph_osdc_put_request(req); 4854 return ret; 4855} 4856EXPORT_SYMBOL(ceph_osdc_unwatch); 4857 4858static int osd_req_op_notify_ack_init(struct ceph_osd_request *req, int which, 4859 u64 notify_id, u64 cookie, void *payload, 4860 u32 payload_len) 4861{ 4862 struct ceph_osd_req_op *op; 4863 struct ceph_pagelist *pl; 4864 int ret; 4865 4866 op = osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY_ACK, 0); 4867 4868 pl = ceph_pagelist_alloc(GFP_NOIO); 4869 if (!pl) 4870 return -ENOMEM; 4871 4872 ret = ceph_pagelist_encode_64(pl, notify_id); 4873 ret |= ceph_pagelist_encode_64(pl, cookie); 4874 if (payload) { 4875 ret |= ceph_pagelist_encode_32(pl, payload_len); 4876 ret |= ceph_pagelist_append(pl, payload, payload_len); 4877 } else { 4878 ret |= ceph_pagelist_encode_32(pl, 0); 4879 } 4880 if (ret) { 4881 ceph_pagelist_release(pl); 4882 return -ENOMEM; 4883 } 4884 4885 ceph_osd_data_pagelist_init(&op->notify_ack.request_data, pl); 4886 op->indata_len = pl->length; 4887 return 0; 4888} 4889 4890int ceph_osdc_notify_ack(struct ceph_osd_client *osdc, 4891 struct ceph_object_id *oid, 4892 struct ceph_object_locator *oloc, 4893 u64 notify_id, 4894 u64 cookie, 4895 void *payload, 4896 u32 payload_len) 4897{ 4898 struct ceph_osd_request *req; 4899 int ret; 4900 4901 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO); 4902 if (!req) 4903 return -ENOMEM; 4904 4905 ceph_oid_copy(&req->r_base_oid, oid); 4906 ceph_oloc_copy(&req->r_base_oloc, oloc); 4907 req->r_flags = CEPH_OSD_FLAG_READ; 4908 4909 ret = osd_req_op_notify_ack_init(req, 0, notify_id, cookie, payload, 4910 payload_len); 4911 if (ret) 4912 goto out_put_req; 4913 4914 ret = ceph_osdc_alloc_messages(req, GFP_NOIO); 4915 if (ret) 4916 goto out_put_req; 4917 4918 ceph_osdc_start_request(osdc, req); 4919 ret = ceph_osdc_wait_request(osdc, req); 4920 4921out_put_req: 4922 ceph_osdc_put_request(req); 4923 return ret; 4924} 4925EXPORT_SYMBOL(ceph_osdc_notify_ack); 4926 4927/* 4928 * @timeout: in seconds 4929 * 4930 * @preply_{pages,len} are initialized both on success and error. 4931 * The caller is responsible for: 4932 * 4933 * ceph_release_page_vector(reply_pages, calc_pages_for(0, reply_len)) 4934 */ 4935int ceph_osdc_notify(struct ceph_osd_client *osdc, 4936 struct ceph_object_id *oid, 4937 struct ceph_object_locator *oloc, 4938 void *payload, 4939 u32 payload_len, 4940 u32 timeout, 4941 struct page ***preply_pages, 4942 size_t *preply_len) 4943{ 4944 struct ceph_osd_linger_request *lreq; 4945 int ret; 4946 4947 WARN_ON(!timeout); 4948 if (preply_pages) { 4949 *preply_pages = NULL; 4950 *preply_len = 0; 4951 } 4952 4953 lreq = linger_alloc(osdc); 4954 if (!lreq) 4955 return -ENOMEM; 4956 4957 lreq->request_pl = ceph_pagelist_alloc(GFP_NOIO); 4958 if (!lreq->request_pl) { 4959 ret = -ENOMEM; 4960 goto out_put_lreq; 4961 } 4962 4963 ret = ceph_pagelist_encode_32(lreq->request_pl, 1); /* prot_ver */ 4964 ret |= ceph_pagelist_encode_32(lreq->request_pl, timeout); 4965 ret |= ceph_pagelist_encode_32(lreq->request_pl, payload_len); 4966 ret |= ceph_pagelist_append(lreq->request_pl, payload, payload_len); 4967 if (ret) { 4968 ret = -ENOMEM; 4969 goto out_put_lreq; 4970 } 4971 4972 /* for notify_id */ 4973 lreq->notify_id_pages = ceph_alloc_page_vector(1, GFP_NOIO); 4974 if (IS_ERR(lreq->notify_id_pages)) { 4975 ret = PTR_ERR(lreq->notify_id_pages); 4976 lreq->notify_id_pages = NULL; 4977 goto out_put_lreq; 4978 } 4979 4980 lreq->preply_pages = preply_pages; 4981 lreq->preply_len = preply_len; 4982 4983 ceph_oid_copy(&lreq->t.base_oid, oid); 4984 ceph_oloc_copy(&lreq->t.base_oloc, oloc); 4985 lreq->t.flags = CEPH_OSD_FLAG_READ; 4986 4987 linger_submit(lreq); 4988 ret = linger_reg_commit_wait(lreq); 4989 if (!ret) 4990 ret = linger_notify_finish_wait(lreq, 4991 msecs_to_jiffies(2 * timeout * MSEC_PER_SEC)); 4992 else 4993 dout("lreq %p failed to initiate notify %d\n", lreq, ret); 4994 4995 linger_cancel(lreq); 4996out_put_lreq: 4997 linger_put(lreq); 4998 return ret; 4999} 5000EXPORT_SYMBOL(ceph_osdc_notify); 5001 5002/* 5003 * Return the number of milliseconds since the watch was last 5004 * confirmed, or an error. If there is an error, the watch is no 5005 * longer valid, and should be destroyed with ceph_osdc_unwatch(). 5006 */ 5007int ceph_osdc_watch_check(struct ceph_osd_client *osdc, 5008 struct ceph_osd_linger_request *lreq) 5009{ 5010 unsigned long stamp, age; 5011 int ret; 5012 5013 down_read(&osdc->lock); 5014 mutex_lock(&lreq->lock); 5015 stamp = lreq->watch_valid_thru; 5016 if (!list_empty(&lreq->pending_lworks)) { 5017 struct linger_work *lwork = 5018 list_first_entry(&lreq->pending_lworks, 5019 struct linger_work, 5020 pending_item); 5021 5022 if (time_before(lwork->queued_stamp, stamp)) 5023 stamp = lwork->queued_stamp; 5024 } 5025 age = jiffies - stamp; 5026 dout("%s lreq %p linger_id %llu age %lu last_error %d\n", __func__, 5027 lreq, lreq->linger_id, age, lreq->last_error); 5028 /* we are truncating to msecs, so return a safe upper bound */ 5029 ret = lreq->last_error ?: 1 + jiffies_to_msecs(age); 5030 5031 mutex_unlock(&lreq->lock); 5032 up_read(&osdc->lock); 5033 return ret; 5034} 5035 5036static int decode_watcher(void **p, void *end, struct ceph_watch_item *item) 5037{ 5038 u8 struct_v; 5039 u32 struct_len; 5040 int ret; 5041 5042 ret = ceph_start_decoding(p, end, 2, "watch_item_t", 5043 &struct_v, &struct_len); 5044 if (ret) 5045 goto bad; 5046 5047 ret = -EINVAL; 5048 ceph_decode_copy_safe(p, end, &item->name, sizeof(item->name), bad); 5049 ceph_decode_64_safe(p, end, item->cookie, bad); 5050 ceph_decode_skip_32(p, end, bad); /* skip timeout seconds */ 5051 5052 if (struct_v >= 2) { 5053 ret = ceph_decode_entity_addr(p, end, &item->addr); 5054 if (ret) 5055 goto bad; 5056 } else { 5057 ret = 0; 5058 } 5059 5060 dout("%s %s%llu cookie %llu addr %s\n", __func__, 5061 ENTITY_NAME(item->name), item->cookie, 5062 ceph_pr_addr(&item->addr)); 5063bad: 5064 return ret; 5065} 5066 5067static int decode_watchers(void **p, void *end, 5068 struct ceph_watch_item **watchers, 5069 u32 *num_watchers) 5070{ 5071 u8 struct_v; 5072 u32 struct_len; 5073 int i; 5074 int ret; 5075 5076 ret = ceph_start_decoding(p, end, 1, "obj_list_watch_response_t", 5077 &struct_v, &struct_len); 5078 if (ret) 5079 return ret; 5080 5081 *num_watchers = ceph_decode_32(p); 5082 *watchers = kcalloc(*num_watchers, sizeof(**watchers), GFP_NOIO); 5083 if (!*watchers) 5084 return -ENOMEM; 5085 5086 for (i = 0; i < *num_watchers; i++) { 5087 ret = decode_watcher(p, end, *watchers + i); 5088 if (ret) { 5089 kfree(*watchers); 5090 return ret; 5091 } 5092 } 5093 5094 return 0; 5095} 5096 5097/* 5098 * On success, the caller is responsible for: 5099 * 5100 * kfree(watchers); 5101 */ 5102int ceph_osdc_list_watchers(struct ceph_osd_client *osdc, 5103 struct ceph_object_id *oid, 5104 struct ceph_object_locator *oloc, 5105 struct ceph_watch_item **watchers, 5106 u32 *num_watchers) 5107{ 5108 struct ceph_osd_request *req; 5109 struct page **pages; 5110 int ret; 5111 5112 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO); 5113 if (!req) 5114 return -ENOMEM; 5115 5116 ceph_oid_copy(&req->r_base_oid, oid); 5117 ceph_oloc_copy(&req->r_base_oloc, oloc); 5118 req->r_flags = CEPH_OSD_FLAG_READ; 5119 5120 pages = ceph_alloc_page_vector(1, GFP_NOIO); 5121 if (IS_ERR(pages)) { 5122 ret = PTR_ERR(pages); 5123 goto out_put_req; 5124 } 5125 5126 osd_req_op_init(req, 0, CEPH_OSD_OP_LIST_WATCHERS, 0); 5127 ceph_osd_data_pages_init(osd_req_op_data(req, 0, list_watchers, 5128 response_data), 5129 pages, PAGE_SIZE, 0, false, true); 5130 5131 ret = ceph_osdc_alloc_messages(req, GFP_NOIO); 5132 if (ret) 5133 goto out_put_req; 5134 5135 ceph_osdc_start_request(osdc, req); 5136 ret = ceph_osdc_wait_request(osdc, req); 5137 if (ret >= 0) { 5138 void *p = page_address(pages[0]); 5139 void *const end = p + req->r_ops[0].outdata_len; 5140 5141 ret = decode_watchers(&p, end, watchers, num_watchers); 5142 } 5143 5144out_put_req: 5145 ceph_osdc_put_request(req); 5146 return ret; 5147} 5148EXPORT_SYMBOL(ceph_osdc_list_watchers); 5149 5150/* 5151 * Call all pending notify callbacks - for use after a watch is 5152 * unregistered, to make sure no more callbacks for it will be invoked 5153 */ 5154void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc) 5155{ 5156 dout("%s osdc %p\n", __func__, osdc); 5157 flush_workqueue(osdc->notify_wq); 5158} 5159EXPORT_SYMBOL(ceph_osdc_flush_notifies); 5160 5161void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc) 5162{ 5163 down_read(&osdc->lock); 5164 maybe_request_map(osdc); 5165 up_read(&osdc->lock); 5166} 5167EXPORT_SYMBOL(ceph_osdc_maybe_request_map); 5168 5169/* 5170 * Execute an OSD class method on an object. 5171 * 5172 * @flags: CEPH_OSD_FLAG_* 5173 * @resp_len: in/out param for reply length 5174 */ 5175int ceph_osdc_call(struct ceph_osd_client *osdc, 5176 struct ceph_object_id *oid, 5177 struct ceph_object_locator *oloc, 5178 const char *class, const char *method, 5179 unsigned int flags, 5180 struct page *req_page, size_t req_len, 5181 struct page **resp_pages, size_t *resp_len) 5182{ 5183 struct ceph_osd_request *req; 5184 int ret; 5185 5186 if (req_len > PAGE_SIZE) 5187 return -E2BIG; 5188 5189 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO); 5190 if (!req) 5191 return -ENOMEM; 5192 5193 ceph_oid_copy(&req->r_base_oid, oid); 5194 ceph_oloc_copy(&req->r_base_oloc, oloc); 5195 req->r_flags = flags; 5196 5197 ret = osd_req_op_cls_init(req, 0, class, method); 5198 if (ret) 5199 goto out_put_req; 5200 5201 if (req_page) 5202 osd_req_op_cls_request_data_pages(req, 0, &req_page, req_len, 5203 0, false, false); 5204 if (resp_pages) 5205 osd_req_op_cls_response_data_pages(req, 0, resp_pages, 5206 *resp_len, 0, false, false); 5207 5208 ret = ceph_osdc_alloc_messages(req, GFP_NOIO); 5209 if (ret) 5210 goto out_put_req; 5211 5212 ceph_osdc_start_request(osdc, req); 5213 ret = ceph_osdc_wait_request(osdc, req); 5214 if (ret >= 0) { 5215 ret = req->r_ops[0].rval; 5216 if (resp_pages) 5217 *resp_len = req->r_ops[0].outdata_len; 5218 } 5219 5220out_put_req: 5221 ceph_osdc_put_request(req); 5222 return ret; 5223} 5224EXPORT_SYMBOL(ceph_osdc_call); 5225 5226/* 5227 * reset all osd connections 5228 */ 5229void ceph_osdc_reopen_osds(struct ceph_osd_client *osdc) 5230{ 5231 struct rb_node *n; 5232 5233 down_write(&osdc->lock); 5234 for (n = rb_first(&osdc->osds); n; ) { 5235 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node); 5236 5237 n = rb_next(n); 5238 if (!reopen_osd(osd)) 5239 kick_osd_requests(osd); 5240 } 5241 up_write(&osdc->lock); 5242} 5243 5244/* 5245 * init, shutdown 5246 */ 5247int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client) 5248{ 5249 int err; 5250 5251 dout("init\n"); 5252 osdc->client = client; 5253 init_rwsem(&osdc->lock); 5254 osdc->osds = RB_ROOT; 5255 INIT_LIST_HEAD(&osdc->osd_lru); 5256 spin_lock_init(&osdc->osd_lru_lock); 5257 osd_init(&osdc->homeless_osd); 5258 osdc->homeless_osd.o_osdc = osdc; 5259 osdc->homeless_osd.o_osd = CEPH_HOMELESS_OSD; 5260 osdc->last_linger_id = CEPH_LINGER_ID_START; 5261 osdc->linger_requests = RB_ROOT; 5262 osdc->map_checks = RB_ROOT; 5263 osdc->linger_map_checks = RB_ROOT; 5264 INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout); 5265 INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout); 5266 5267 err = -ENOMEM; 5268 osdc->osdmap = ceph_osdmap_alloc(); 5269 if (!osdc->osdmap) 5270 goto out; 5271 5272 osdc->req_mempool = mempool_create_slab_pool(10, 5273 ceph_osd_request_cache); 5274 if (!osdc->req_mempool) 5275 goto out_map; 5276 5277 err = ceph_msgpool_init(&osdc->msgpool_op, CEPH_MSG_OSD_OP, 5278 PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10, "osd_op"); 5279 if (err < 0) 5280 goto out_mempool; 5281 err = ceph_msgpool_init(&osdc->msgpool_op_reply, CEPH_MSG_OSD_OPREPLY, 5282 PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10, 5283 "osd_op_reply"); 5284 if (err < 0) 5285 goto out_msgpool; 5286 5287 err = -ENOMEM; 5288 osdc->notify_wq = create_singlethread_workqueue("ceph-watch-notify"); 5289 if (!osdc->notify_wq) 5290 goto out_msgpool_reply; 5291 5292 osdc->completion_wq = create_singlethread_workqueue("ceph-completion"); 5293 if (!osdc->completion_wq) 5294 goto out_notify_wq; 5295 5296 schedule_delayed_work(&osdc->timeout_work, 5297 osdc->client->options->osd_keepalive_timeout); 5298 schedule_delayed_work(&osdc->osds_timeout_work, 5299 round_jiffies_relative(osdc->client->options->osd_idle_ttl)); 5300 5301 return 0; 5302 5303out_notify_wq: 5304 destroy_workqueue(osdc->notify_wq); 5305out_msgpool_reply: 5306 ceph_msgpool_destroy(&osdc->msgpool_op_reply); 5307out_msgpool: 5308 ceph_msgpool_destroy(&osdc->msgpool_op); 5309out_mempool: 5310 mempool_destroy(osdc->req_mempool); 5311out_map: 5312 ceph_osdmap_destroy(osdc->osdmap); 5313out: 5314 return err; 5315} 5316 5317void ceph_osdc_stop(struct ceph_osd_client *osdc) 5318{ 5319 destroy_workqueue(osdc->completion_wq); 5320 destroy_workqueue(osdc->notify_wq); 5321 cancel_delayed_work_sync(&osdc->timeout_work); 5322 cancel_delayed_work_sync(&osdc->osds_timeout_work); 5323 5324 down_write(&osdc->lock); 5325 while (!RB_EMPTY_ROOT(&osdc->osds)) { 5326 struct ceph_osd *osd = rb_entry(rb_first(&osdc->osds), 5327 struct ceph_osd, o_node); 5328 close_osd(osd); 5329 } 5330 up_write(&osdc->lock); 5331 WARN_ON(refcount_read(&osdc->homeless_osd.o_ref) != 1); 5332 osd_cleanup(&osdc->homeless_osd); 5333 5334 WARN_ON(!list_empty(&osdc->osd_lru)); 5335 WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_requests)); 5336 WARN_ON(!RB_EMPTY_ROOT(&osdc->map_checks)); 5337 WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_map_checks)); 5338 WARN_ON(atomic_read(&osdc->num_requests)); 5339 WARN_ON(atomic_read(&osdc->num_homeless)); 5340 5341 ceph_osdmap_destroy(osdc->osdmap); 5342 mempool_destroy(osdc->req_mempool); 5343 ceph_msgpool_destroy(&osdc->msgpool_op); 5344 ceph_msgpool_destroy(&osdc->msgpool_op_reply); 5345} 5346 5347int osd_req_op_copy_from_init(struct ceph_osd_request *req, 5348 u64 src_snapid, u64 src_version, 5349 struct ceph_object_id *src_oid, 5350 struct ceph_object_locator *src_oloc, 5351 u32 src_fadvise_flags, 5352 u32 dst_fadvise_flags, 5353 u32 truncate_seq, u64 truncate_size, 5354 u8 copy_from_flags) 5355{ 5356 struct ceph_osd_req_op *op; 5357 struct page **pages; 5358 void *p, *end; 5359 5360 pages = ceph_alloc_page_vector(1, GFP_KERNEL); 5361 if (IS_ERR(pages)) 5362 return PTR_ERR(pages); 5363 5364 op = osd_req_op_init(req, 0, CEPH_OSD_OP_COPY_FROM2, 5365 dst_fadvise_flags); 5366 op->copy_from.snapid = src_snapid; 5367 op->copy_from.src_version = src_version; 5368 op->copy_from.flags = copy_from_flags; 5369 op->copy_from.src_fadvise_flags = src_fadvise_flags; 5370 5371 p = page_address(pages[0]); 5372 end = p + PAGE_SIZE; 5373 ceph_encode_string(&p, end, src_oid->name, src_oid->name_len); 5374 encode_oloc(&p, end, src_oloc); 5375 ceph_encode_32(&p, truncate_seq); 5376 ceph_encode_64(&p, truncate_size); 5377 op->indata_len = PAGE_SIZE - (end - p); 5378 5379 ceph_osd_data_pages_init(&op->copy_from.osd_data, pages, 5380 op->indata_len, 0, false, true); 5381 return 0; 5382} 5383EXPORT_SYMBOL(osd_req_op_copy_from_init); 5384 5385int __init ceph_osdc_setup(void) 5386{ 5387 size_t size = sizeof(struct ceph_osd_request) + 5388 CEPH_OSD_SLAB_OPS * sizeof(struct ceph_osd_req_op); 5389 5390 BUG_ON(ceph_osd_request_cache); 5391 ceph_osd_request_cache = kmem_cache_create("ceph_osd_request", size, 5392 0, 0, NULL); 5393 5394 return ceph_osd_request_cache ? 0 : -ENOMEM; 5395} 5396 5397void ceph_osdc_cleanup(void) 5398{ 5399 BUG_ON(!ceph_osd_request_cache); 5400 kmem_cache_destroy(ceph_osd_request_cache); 5401 ceph_osd_request_cache = NULL; 5402} 5403 5404/* 5405 * handle incoming message 5406 */ 5407static void osd_dispatch(struct ceph_connection *con, struct ceph_msg *msg) 5408{ 5409 struct ceph_osd *osd = con->private; 5410 struct ceph_osd_client *osdc = osd->o_osdc; 5411 int type = le16_to_cpu(msg->hdr.type); 5412 5413 switch (type) { 5414 case CEPH_MSG_OSD_MAP: 5415 ceph_osdc_handle_map(osdc, msg); 5416 break; 5417 case CEPH_MSG_OSD_OPREPLY: 5418 handle_reply(osd, msg); 5419 break; 5420 case CEPH_MSG_OSD_BACKOFF: 5421 handle_backoff(osd, msg); 5422 break; 5423 case CEPH_MSG_WATCH_NOTIFY: 5424 handle_watch_notify(osdc, msg); 5425 break; 5426 5427 default: 5428 pr_err("received unknown message type %d %s\n", type, 5429 ceph_msg_type_name(type)); 5430 } 5431 5432 ceph_msg_put(msg); 5433} 5434 5435/* How much sparse data was requested? */ 5436static u64 sparse_data_requested(struct ceph_osd_request *req) 5437{ 5438 u64 len = 0; 5439 5440 if (req->r_flags & CEPH_OSD_FLAG_READ) { 5441 int i; 5442 5443 for (i = 0; i < req->r_num_ops; ++i) { 5444 struct ceph_osd_req_op *op = &req->r_ops[i]; 5445 5446 if (op->op == CEPH_OSD_OP_SPARSE_READ) 5447 len += op->extent.length; 5448 } 5449 } 5450 return len; 5451} 5452 5453/* 5454 * Lookup and return message for incoming reply. Don't try to do 5455 * anything about a larger than preallocated data portion of the 5456 * message at the moment - for now, just skip the message. 5457 */ 5458static struct ceph_msg *get_reply(struct ceph_connection *con, 5459 struct ceph_msg_header *hdr, 5460 int *skip) 5461{ 5462 struct ceph_osd *osd = con->private; 5463 struct ceph_osd_client *osdc = osd->o_osdc; 5464 struct ceph_msg *m = NULL; 5465 struct ceph_osd_request *req; 5466 int front_len = le32_to_cpu(hdr->front_len); 5467 int data_len = le32_to_cpu(hdr->data_len); 5468 u64 tid = le64_to_cpu(hdr->tid); 5469 u64 srlen; 5470 5471 down_read(&osdc->lock); 5472 if (!osd_registered(osd)) { 5473 dout("%s osd%d unknown, skipping\n", __func__, osd->o_osd); 5474 *skip = 1; 5475 goto out_unlock_osdc; 5476 } 5477 WARN_ON(osd->o_osd != le64_to_cpu(hdr->src.num)); 5478 5479 mutex_lock(&osd->lock); 5480 req = lookup_request(&osd->o_requests, tid); 5481 if (!req) { 5482 dout("%s osd%d tid %llu unknown, skipping\n", __func__, 5483 osd->o_osd, tid); 5484 *skip = 1; 5485 goto out_unlock_session; 5486 } 5487 5488 ceph_msg_revoke_incoming(req->r_reply); 5489 5490 if (front_len > req->r_reply->front_alloc_len) { 5491 pr_warn("%s osd%d tid %llu front %d > preallocated %d\n", 5492 __func__, osd->o_osd, req->r_tid, front_len, 5493 req->r_reply->front_alloc_len); 5494 m = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, front_len, GFP_NOFS, 5495 false); 5496 if (!m) 5497 goto out_unlock_session; 5498 ceph_msg_put(req->r_reply); 5499 req->r_reply = m; 5500 } 5501 5502 srlen = sparse_data_requested(req); 5503 if (!srlen && data_len > req->r_reply->data_length) { 5504 pr_warn("%s osd%d tid %llu data %d > preallocated %zu, skipping\n", 5505 __func__, osd->o_osd, req->r_tid, data_len, 5506 req->r_reply->data_length); 5507 m = NULL; 5508 *skip = 1; 5509 goto out_unlock_session; 5510 } 5511 5512 m = ceph_msg_get(req->r_reply); 5513 m->sparse_read_total = srlen; 5514 5515 dout("get_reply tid %lld %p\n", tid, m); 5516 5517out_unlock_session: 5518 mutex_unlock(&osd->lock); 5519out_unlock_osdc: 5520 up_read(&osdc->lock); 5521 return m; 5522} 5523 5524static struct ceph_msg *alloc_msg_with_page_vector(struct ceph_msg_header *hdr) 5525{ 5526 struct ceph_msg *m; 5527 int type = le16_to_cpu(hdr->type); 5528 u32 front_len = le32_to_cpu(hdr->front_len); 5529 u32 data_len = le32_to_cpu(hdr->data_len); 5530 5531 m = ceph_msg_new2(type, front_len, 1, GFP_NOIO, false); 5532 if (!m) 5533 return NULL; 5534 5535 if (data_len) { 5536 struct page **pages; 5537 5538 pages = ceph_alloc_page_vector(calc_pages_for(0, data_len), 5539 GFP_NOIO); 5540 if (IS_ERR(pages)) { 5541 ceph_msg_put(m); 5542 return NULL; 5543 } 5544 5545 ceph_msg_data_add_pages(m, pages, data_len, 0, true); 5546 } 5547 5548 return m; 5549} 5550 5551static struct ceph_msg *osd_alloc_msg(struct ceph_connection *con, 5552 struct ceph_msg_header *hdr, 5553 int *skip) 5554{ 5555 struct ceph_osd *osd = con->private; 5556 int type = le16_to_cpu(hdr->type); 5557 5558 *skip = 0; 5559 switch (type) { 5560 case CEPH_MSG_OSD_MAP: 5561 case CEPH_MSG_OSD_BACKOFF: 5562 case CEPH_MSG_WATCH_NOTIFY: 5563 return alloc_msg_with_page_vector(hdr); 5564 case CEPH_MSG_OSD_OPREPLY: 5565 return get_reply(con, hdr, skip); 5566 default: 5567 pr_warn("%s osd%d unknown msg type %d, skipping\n", __func__, 5568 osd->o_osd, type); 5569 *skip = 1; 5570 return NULL; 5571 } 5572} 5573 5574/* 5575 * Wrappers to refcount containing ceph_osd struct 5576 */ 5577static struct ceph_connection *osd_get_con(struct ceph_connection *con) 5578{ 5579 struct ceph_osd *osd = con->private; 5580 if (get_osd(osd)) 5581 return con; 5582 return NULL; 5583} 5584 5585static void osd_put_con(struct ceph_connection *con) 5586{ 5587 struct ceph_osd *osd = con->private; 5588 put_osd(osd); 5589} 5590 5591/* 5592 * authentication 5593 */ 5594 5595/* 5596 * Note: returned pointer is the address of a structure that's 5597 * managed separately. Caller must *not* attempt to free it. 5598 */ 5599static struct ceph_auth_handshake * 5600osd_get_authorizer(struct ceph_connection *con, int *proto, int force_new) 5601{ 5602 struct ceph_osd *o = con->private; 5603 struct ceph_osd_client *osdc = o->o_osdc; 5604 struct ceph_auth_client *ac = osdc->client->monc.auth; 5605 struct ceph_auth_handshake *auth = &o->o_auth; 5606 int ret; 5607 5608 ret = __ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_OSD, 5609 force_new, proto, NULL, NULL); 5610 if (ret) 5611 return ERR_PTR(ret); 5612 5613 return auth; 5614} 5615 5616static int osd_add_authorizer_challenge(struct ceph_connection *con, 5617 void *challenge_buf, int challenge_buf_len) 5618{ 5619 struct ceph_osd *o = con->private; 5620 struct ceph_osd_client *osdc = o->o_osdc; 5621 struct ceph_auth_client *ac = osdc->client->monc.auth; 5622 5623 return ceph_auth_add_authorizer_challenge(ac, o->o_auth.authorizer, 5624 challenge_buf, challenge_buf_len); 5625} 5626 5627static int osd_verify_authorizer_reply(struct ceph_connection *con) 5628{ 5629 struct ceph_osd *o = con->private; 5630 struct ceph_osd_client *osdc = o->o_osdc; 5631 struct ceph_auth_client *ac = osdc->client->monc.auth; 5632 struct ceph_auth_handshake *auth = &o->o_auth; 5633 5634 return ceph_auth_verify_authorizer_reply(ac, auth->authorizer, 5635 auth->authorizer_reply_buf, auth->authorizer_reply_buf_len, 5636 NULL, NULL, NULL, NULL); 5637} 5638 5639static int osd_invalidate_authorizer(struct ceph_connection *con) 5640{ 5641 struct ceph_osd *o = con->private; 5642 struct ceph_osd_client *osdc = o->o_osdc; 5643 struct ceph_auth_client *ac = osdc->client->monc.auth; 5644 5645 ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_OSD); 5646 return ceph_monc_validate_auth(&osdc->client->monc); 5647} 5648 5649static int osd_get_auth_request(struct ceph_connection *con, 5650 void *buf, int *buf_len, 5651 void **authorizer, int *authorizer_len) 5652{ 5653 struct ceph_osd *o = con->private; 5654 struct ceph_auth_client *ac = o->o_osdc->client->monc.auth; 5655 struct ceph_auth_handshake *auth = &o->o_auth; 5656 int ret; 5657 5658 ret = ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_OSD, 5659 buf, buf_len); 5660 if (ret) 5661 return ret; 5662 5663 *authorizer = auth->authorizer_buf; 5664 *authorizer_len = auth->authorizer_buf_len; 5665 return 0; 5666} 5667 5668static int osd_handle_auth_reply_more(struct ceph_connection *con, 5669 void *reply, int reply_len, 5670 void *buf, int *buf_len, 5671 void **authorizer, int *authorizer_len) 5672{ 5673 struct ceph_osd *o = con->private; 5674 struct ceph_auth_client *ac = o->o_osdc->client->monc.auth; 5675 struct ceph_auth_handshake *auth = &o->o_auth; 5676 int ret; 5677 5678 ret = ceph_auth_handle_svc_reply_more(ac, auth, reply, reply_len, 5679 buf, buf_len); 5680 if (ret) 5681 return ret; 5682 5683 *authorizer = auth->authorizer_buf; 5684 *authorizer_len = auth->authorizer_buf_len; 5685 return 0; 5686} 5687 5688static int osd_handle_auth_done(struct ceph_connection *con, 5689 u64 global_id, void *reply, int reply_len, 5690 u8 *session_key, int *session_key_len, 5691 u8 *con_secret, int *con_secret_len) 5692{ 5693 struct ceph_osd *o = con->private; 5694 struct ceph_auth_client *ac = o->o_osdc->client->monc.auth; 5695 struct ceph_auth_handshake *auth = &o->o_auth; 5696 5697 return ceph_auth_handle_svc_reply_done(ac, auth, reply, reply_len, 5698 session_key, session_key_len, 5699 con_secret, con_secret_len); 5700} 5701 5702static int osd_handle_auth_bad_method(struct ceph_connection *con, 5703 int used_proto, int result, 5704 const int *allowed_protos, int proto_cnt, 5705 const int *allowed_modes, int mode_cnt) 5706{ 5707 struct ceph_osd *o = con->private; 5708 struct ceph_mon_client *monc = &o->o_osdc->client->monc; 5709 int ret; 5710 5711 if (ceph_auth_handle_bad_authorizer(monc->auth, CEPH_ENTITY_TYPE_OSD, 5712 used_proto, result, 5713 allowed_protos, proto_cnt, 5714 allowed_modes, mode_cnt)) { 5715 ret = ceph_monc_validate_auth(monc); 5716 if (ret) 5717 return ret; 5718 } 5719 5720 return -EACCES; 5721} 5722 5723static void osd_reencode_message(struct ceph_msg *msg) 5724{ 5725 int type = le16_to_cpu(msg->hdr.type); 5726 5727 if (type == CEPH_MSG_OSD_OP) 5728 encode_request_finish(msg); 5729} 5730 5731static int osd_sign_message(struct ceph_msg *msg) 5732{ 5733 struct ceph_osd *o = msg->con->private; 5734 struct ceph_auth_handshake *auth = &o->o_auth; 5735 5736 return ceph_auth_sign_message(auth, msg); 5737} 5738 5739static int osd_check_message_signature(struct ceph_msg *msg) 5740{ 5741 struct ceph_osd *o = msg->con->private; 5742 struct ceph_auth_handshake *auth = &o->o_auth; 5743 5744 return ceph_auth_check_message_signature(auth, msg); 5745} 5746 5747static void advance_cursor(struct ceph_msg_data_cursor *cursor, size_t len, 5748 bool zero) 5749{ 5750 while (len) { 5751 struct page *page; 5752 size_t poff, plen; 5753 5754 page = ceph_msg_data_next(cursor, &poff, &plen); 5755 if (plen > len) 5756 plen = len; 5757 if (zero) 5758 zero_user_segment(page, poff, poff + plen); 5759 len -= plen; 5760 ceph_msg_data_advance(cursor, plen); 5761 } 5762} 5763 5764static int prep_next_sparse_read(struct ceph_connection *con, 5765 struct ceph_msg_data_cursor *cursor) 5766{ 5767 struct ceph_osd *o = con->private; 5768 struct ceph_sparse_read *sr = &o->o_sparse_read; 5769 struct ceph_osd_request *req; 5770 struct ceph_osd_req_op *op; 5771 5772 spin_lock(&o->o_requests_lock); 5773 req = lookup_request(&o->o_requests, le64_to_cpu(con->in_msg->hdr.tid)); 5774 if (!req) { 5775 spin_unlock(&o->o_requests_lock); 5776 return -EBADR; 5777 } 5778 5779 if (o->o_sparse_op_idx < 0) { 5780 dout("%s: [%d] starting new sparse read req\n", 5781 __func__, o->o_osd); 5782 } else { 5783 u64 end; 5784 5785 op = &req->r_ops[o->o_sparse_op_idx]; 5786 5787 WARN_ON_ONCE(op->extent.sparse_ext); 5788 5789 /* hand back buffer we took earlier */ 5790 op->extent.sparse_ext = sr->sr_extent; 5791 sr->sr_extent = NULL; 5792 op->extent.sparse_ext_cnt = sr->sr_count; 5793 sr->sr_ext_len = 0; 5794 dout("%s: [%d] completed extent array len %d cursor->resid %zd\n", 5795 __func__, o->o_osd, op->extent.sparse_ext_cnt, cursor->resid); 5796 /* Advance to end of data for this operation */ 5797 end = ceph_sparse_ext_map_end(op); 5798 if (end < sr->sr_req_len) 5799 advance_cursor(cursor, sr->sr_req_len - end, false); 5800 } 5801 5802 ceph_init_sparse_read(sr); 5803 5804 /* find next op in this request (if any) */ 5805 while (++o->o_sparse_op_idx < req->r_num_ops) { 5806 op = &req->r_ops[o->o_sparse_op_idx]; 5807 if (op->op == CEPH_OSD_OP_SPARSE_READ) 5808 goto found; 5809 } 5810 5811 /* reset for next sparse read request */ 5812 spin_unlock(&o->o_requests_lock); 5813 o->o_sparse_op_idx = -1; 5814 return 0; 5815found: 5816 sr->sr_req_off = op->extent.offset; 5817 sr->sr_req_len = op->extent.length; 5818 sr->sr_pos = sr->sr_req_off; 5819 dout("%s: [%d] new sparse read op at idx %d 0x%llx~0x%llx\n", __func__, 5820 o->o_osd, o->o_sparse_op_idx, sr->sr_req_off, sr->sr_req_len); 5821 5822 /* hand off request's sparse extent map buffer */ 5823 sr->sr_ext_len = op->extent.sparse_ext_cnt; 5824 op->extent.sparse_ext_cnt = 0; 5825 sr->sr_extent = op->extent.sparse_ext; 5826 op->extent.sparse_ext = NULL; 5827 5828 spin_unlock(&o->o_requests_lock); 5829 return 1; 5830} 5831 5832#ifdef __BIG_ENDIAN 5833static inline void convert_extent_map(struct ceph_sparse_read *sr) 5834{ 5835 int i; 5836 5837 for (i = 0; i < sr->sr_count; i++) { 5838 struct ceph_sparse_extent *ext = &sr->sr_extent[i]; 5839 5840 ext->off = le64_to_cpu((__force __le64)ext->off); 5841 ext->len = le64_to_cpu((__force __le64)ext->len); 5842 } 5843} 5844#else 5845static inline void convert_extent_map(struct ceph_sparse_read *sr) 5846{ 5847} 5848#endif 5849 5850static int osd_sparse_read(struct ceph_connection *con, 5851 struct ceph_msg_data_cursor *cursor, 5852 char **pbuf) 5853{ 5854 struct ceph_osd *o = con->private; 5855 struct ceph_sparse_read *sr = &o->o_sparse_read; 5856 u32 count = sr->sr_count; 5857 u64 eoff, elen, len = 0; 5858 int i, ret; 5859 5860 switch (sr->sr_state) { 5861 case CEPH_SPARSE_READ_HDR: 5862next_op: 5863 ret = prep_next_sparse_read(con, cursor); 5864 if (ret <= 0) 5865 return ret; 5866 5867 /* number of extents */ 5868 ret = sizeof(sr->sr_count); 5869 *pbuf = (char *)&sr->sr_count; 5870 sr->sr_state = CEPH_SPARSE_READ_EXTENTS; 5871 break; 5872 case CEPH_SPARSE_READ_EXTENTS: 5873 /* Convert sr_count to host-endian */ 5874 count = le32_to_cpu((__force __le32)sr->sr_count); 5875 sr->sr_count = count; 5876 dout("[%d] got %u extents\n", o->o_osd, count); 5877 5878 if (count > 0) { 5879 if (!sr->sr_extent || count > sr->sr_ext_len) { 5880 /* no extent array provided, or too short */ 5881 kfree(sr->sr_extent); 5882 sr->sr_extent = kmalloc_array(count, 5883 sizeof(*sr->sr_extent), 5884 GFP_NOIO); 5885 if (!sr->sr_extent) { 5886 pr_err("%s: failed to allocate %u extents\n", 5887 __func__, count); 5888 return -ENOMEM; 5889 } 5890 sr->sr_ext_len = count; 5891 } 5892 ret = count * sizeof(*sr->sr_extent); 5893 *pbuf = (char *)sr->sr_extent; 5894 sr->sr_state = CEPH_SPARSE_READ_DATA_LEN; 5895 break; 5896 } 5897 /* No extents? Read data len */ 5898 fallthrough; 5899 case CEPH_SPARSE_READ_DATA_LEN: 5900 convert_extent_map(sr); 5901 ret = sizeof(sr->sr_datalen); 5902 *pbuf = (char *)&sr->sr_datalen; 5903 sr->sr_state = CEPH_SPARSE_READ_DATA_PRE; 5904 break; 5905 case CEPH_SPARSE_READ_DATA_PRE: 5906 /* Convert sr_datalen to host-endian */ 5907 sr->sr_datalen = le32_to_cpu((__force __le32)sr->sr_datalen); 5908 for (i = 0; i < count; i++) 5909 len += sr->sr_extent[i].len; 5910 if (sr->sr_datalen != len) { 5911 pr_warn_ratelimited("data len %u != extent len %llu\n", 5912 sr->sr_datalen, len); 5913 return -EREMOTEIO; 5914 } 5915 sr->sr_state = CEPH_SPARSE_READ_DATA; 5916 fallthrough; 5917 case CEPH_SPARSE_READ_DATA: 5918 if (sr->sr_index >= count) { 5919 sr->sr_state = CEPH_SPARSE_READ_HDR; 5920 goto next_op; 5921 } 5922 5923 eoff = sr->sr_extent[sr->sr_index].off; 5924 elen = sr->sr_extent[sr->sr_index].len; 5925 5926 dout("[%d] ext %d off 0x%llx len 0x%llx\n", 5927 o->o_osd, sr->sr_index, eoff, elen); 5928 5929 if (elen > INT_MAX) { 5930 dout("Sparse read extent length too long (0x%llx)\n", 5931 elen); 5932 return -EREMOTEIO; 5933 } 5934 5935 /* zero out anything from sr_pos to start of extent */ 5936 if (sr->sr_pos < eoff) 5937 advance_cursor(cursor, eoff - sr->sr_pos, true); 5938 5939 /* Set position to end of extent */ 5940 sr->sr_pos = eoff + elen; 5941 5942 /* send back the new length and nullify the ptr */ 5943 cursor->sr_resid = elen; 5944 ret = elen; 5945 *pbuf = NULL; 5946 5947 /* Bump the array index */ 5948 ++sr->sr_index; 5949 break; 5950 } 5951 return ret; 5952} 5953 5954static const struct ceph_connection_operations osd_con_ops = { 5955 .get = osd_get_con, 5956 .put = osd_put_con, 5957 .sparse_read = osd_sparse_read, 5958 .alloc_msg = osd_alloc_msg, 5959 .dispatch = osd_dispatch, 5960 .fault = osd_fault, 5961 .reencode_message = osd_reencode_message, 5962 .get_authorizer = osd_get_authorizer, 5963 .add_authorizer_challenge = osd_add_authorizer_challenge, 5964 .verify_authorizer_reply = osd_verify_authorizer_reply, 5965 .invalidate_authorizer = osd_invalidate_authorizer, 5966 .sign_message = osd_sign_message, 5967 .check_message_signature = osd_check_message_signature, 5968 .get_auth_request = osd_get_auth_request, 5969 .handle_auth_reply_more = osd_handle_auth_reply_more, 5970 .handle_auth_done = osd_handle_auth_done, 5971 .handle_auth_bad_method = osd_handle_auth_bad_method, 5972}; 5973