]> git.proxmox.com Git - mirror_ubuntu-disco-kernel.git/blame - net/ceph/osd_client.c
projects / mirror_ubuntu-disco-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
treewide: Use struct_size() for kmalloc()-family
[mirror_ubuntu-disco-kernel.git] / net / ceph / osd_client.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
a4ce40a9 2
3d14c5d2 3#include <linux/ceph/ceph_debug.h>
f24e9980 4
3d14c5d2 5#include <linux/module.h>
f24e9980
SW		 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>
68b4476b
YS		 12#ifdef CONFIG_BLOCK
13#include <linux/bio.h>
14#endif
f24e9980 15
8cb441c0 16#include <linux/ceph/ceph_features.h>
3d14c5d2
YS		 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>
08c1ac50 23#include <linux/ceph/striper.h>
f24e9980 24
c16e7869 25#define OSD_OPREPLY_FRONT_LEN 512
0d59ab81 26
5522ae0b
AE		 27static struct kmem_cache *ceph_osd_request_cache;
28
9e32789f 29static const struct ceph_connection_operations osd_con_ops;
f24e9980 30
f24e9980
SW		 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
5aea3dcd
ID		 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);
922dab61
ID		 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);
a02a946d 55static void clear_backoffs(struct ceph_osd *osd);
5aea3dcd
ID		 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}
922dab61
ID		 85static inline void verify_lreq_locked(struct ceph_osd_linger_request *lreq)
86{
87 WARN_ON(!mutex_is_locked(&lreq->lock));
88}
5aea3dcd
ID		 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) { }
922dab61 93static inline void verify_lreq_locked(struct ceph_osd_linger_request *lreq) { }
5aea3dcd
ID		 94#endif
95
f24e9980
SW		 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 */
dbe0fc41 103static int calc_layout(struct ceph_file_layout *layout, u64 off, u64 *plen,
a19dadfb 104 u64 *objnum, u64 *objoff, u64 *objlen)
f24e9980 105{
60e56f13 106 u64 orig_len = *plen;
dccbf080 107 u32 xlen;
f24e9980 108
60e56f13 109 /* object extent? */
dccbf080
ID		 110 ceph_calc_file_object_mapping(layout, off, orig_len, objnum,
111 objoff, &xlen);
112 *objlen = xlen;
75d1c941
AE		 113 if (*objlen < orig_len) {
114 *plen = *objlen;
60e56f13
AE		 115 dout(" skipping last %llu, final file extent %llu~%llu\n",
116 orig_len - *plen, off, *plen);
117 }
118
75d1c941 119 dout("calc_layout objnum=%llx %llu~%llu\n", *objnum, *objoff, *objlen);
3ff5f385 120 return 0;
f24e9980
SW		 121}
122
c54d47bf
AE		 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
a4ce40a9 129static void ceph_osd_data_pages_init(struct ceph_osd_data *osd_data,
43bfe5de
AE		 130 struct page **pages, u64 length, u32 alignment,
131 bool pages_from_pool, bool own_pages)
132{
133 osd_data->type = CEPH_OSD_DATA_TYPE_PAGES;
134 osd_data->pages = pages;
135 osd_data->length = length;
136 osd_data->alignment = alignment;
137 osd_data->pages_from_pool = pages_from_pool;
138 osd_data->own_pages = own_pages;
139}
43bfe5de 140
a4ce40a9 141static void ceph_osd_data_pagelist_init(struct ceph_osd_data *osd_data,
43bfe5de
AE		 142 struct ceph_pagelist *pagelist)
143{
144 osd_data->type = CEPH_OSD_DATA_TYPE_PAGELIST;
145 osd_data->pagelist = pagelist;
146}
43bfe5de
AE		 147
148#ifdef CONFIG_BLOCK
a4ce40a9 149static void ceph_osd_data_bio_init(struct ceph_osd_data *osd_data,
5359a17d
ID		 150 struct ceph_bio_iter *bio_pos,
151 u32 bio_length)
43bfe5de
AE		 152{
153 osd_data->type = CEPH_OSD_DATA_TYPE_BIO;
5359a17d 154 osd_data->bio_pos = *bio_pos;
43bfe5de
AE		 155 osd_data->bio_length = bio_length;
156}
43bfe5de
AE		 157#endif /* CONFIG_BLOCK */
158
b9e281c2
ID		 159static void ceph_osd_data_bvecs_init(struct ceph_osd_data *osd_data,
160 struct ceph_bvec_iter *bvec_pos)
161{
162 osd_data->type = CEPH_OSD_DATA_TYPE_BVECS;
163 osd_data->bvec_pos = *bvec_pos;
164}
165
8a703a38
IC		 166#define osd_req_op_data(oreq, whch, typ, fld) \
167({ \
168 struct ceph_osd_request *__oreq = (oreq); \
169 unsigned int __whch = (whch); \
170 BUG_ON(__whch >= __oreq->r_num_ops); \
171 &__oreq->r_ops[__whch].typ.fld; \
172})
863c7eb5 173
49719778
AE		 174static struct ceph_osd_data *
175osd_req_op_raw_data_in(struct ceph_osd_request *osd_req, unsigned int which)
176{
177 BUG_ON(which >= osd_req->r_num_ops);
178
179 return &osd_req->r_ops[which].raw_data_in;
180}
181
a4ce40a9
AE		 182struct ceph_osd_data *
183osd_req_op_extent_osd_data(struct ceph_osd_request *osd_req,
406e2c9f 184 unsigned int which)
a4ce40a9 185{
863c7eb5 186 return osd_req_op_data(osd_req, which, extent, osd_data);
a4ce40a9
AE		 187}
188EXPORT_SYMBOL(osd_req_op_extent_osd_data);
189
49719778
AE		 190void osd_req_op_raw_data_in_pages(struct ceph_osd_request *osd_req,
191 unsigned int which, struct page **pages,
192 u64 length, u32 alignment,
193 bool pages_from_pool, bool own_pages)
194{
195 struct ceph_osd_data *osd_data;
196
197 osd_data = osd_req_op_raw_data_in(osd_req, which);
198 ceph_osd_data_pages_init(osd_data, pages, length, alignment,
199 pages_from_pool, own_pages);
200}
201EXPORT_SYMBOL(osd_req_op_raw_data_in_pages);
202
a4ce40a9 203void osd_req_op_extent_osd_data_pages(struct ceph_osd_request *osd_req,
406e2c9f
AE		 204 unsigned int which, struct page **pages,
205 u64 length, u32 alignment,
a4ce40a9
AE		 206 bool pages_from_pool, bool own_pages)
207{
208 struct ceph_osd_data *osd_data;
209
863c7eb5 210 osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
a4ce40a9
AE		 211 ceph_osd_data_pages_init(osd_data, pages, length, alignment,
212 pages_from_pool, own_pages);
a4ce40a9
AE		 213}
214EXPORT_SYMBOL(osd_req_op_extent_osd_data_pages);
215
216void osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request *osd_req,
406e2c9f 217 unsigned int which, struct ceph_pagelist *pagelist)
a4ce40a9
AE		 218{
219 struct ceph_osd_data *osd_data;
220
863c7eb5 221 osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
a4ce40a9 222 ceph_osd_data_pagelist_init(osd_data, pagelist);
a4ce40a9
AE		 223}
224EXPORT_SYMBOL(osd_req_op_extent_osd_data_pagelist);
225
226#ifdef CONFIG_BLOCK
227void osd_req_op_extent_osd_data_bio(struct ceph_osd_request *osd_req,
5359a17d
ID		 228 unsigned int which,
229 struct ceph_bio_iter *bio_pos,
230 u32 bio_length)
a4ce40a9
AE		 231{
232 struct ceph_osd_data *osd_data;
863c7eb5
AE		 233
234 osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
5359a17d 235 ceph_osd_data_bio_init(osd_data, bio_pos, bio_length);
a4ce40a9
AE		 236}
237EXPORT_SYMBOL(osd_req_op_extent_osd_data_bio);
238#endif /* CONFIG_BLOCK */
239
b9e281c2
ID		 240void osd_req_op_extent_osd_data_bvec_pos(struct ceph_osd_request *osd_req,
241 unsigned int which,
242 struct ceph_bvec_iter *bvec_pos)
243{
244 struct ceph_osd_data *osd_data;
245
246 osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
247 ceph_osd_data_bvecs_init(osd_data, bvec_pos);
248}
249EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvec_pos);
250
a4ce40a9
AE		 251static void osd_req_op_cls_request_info_pagelist(
252 struct ceph_osd_request *osd_req,
253 unsigned int which, struct ceph_pagelist *pagelist)
254{
255 struct ceph_osd_data *osd_data;
256
863c7eb5 257 osd_data = osd_req_op_data(osd_req, which, cls, request_info);
a4ce40a9 258 ceph_osd_data_pagelist_init(osd_data, pagelist);
a4ce40a9
AE		 259}
260
04017e29
AE		 261void osd_req_op_cls_request_data_pagelist(
262 struct ceph_osd_request *osd_req,
263 unsigned int which, struct ceph_pagelist *pagelist)
264{
265 struct ceph_osd_data *osd_data;
266
863c7eb5 267 osd_data = osd_req_op_data(osd_req, which, cls, request_data);
04017e29 268 ceph_osd_data_pagelist_init(osd_data, pagelist);
bb873b53
ID		 269 osd_req->r_ops[which].cls.indata_len += pagelist->length;
270 osd_req->r_ops[which].indata_len += pagelist->length;
04017e29
AE		 271}
272EXPORT_SYMBOL(osd_req_op_cls_request_data_pagelist);
273
6c57b554
AE		 274void osd_req_op_cls_request_data_pages(struct ceph_osd_request *osd_req,
275 unsigned int which, struct page **pages, u64 length,
276 u32 alignment, bool pages_from_pool, bool own_pages)
277{
278 struct ceph_osd_data *osd_data;
279
280 osd_data = osd_req_op_data(osd_req, which, cls, request_data);
281 ceph_osd_data_pages_init(osd_data, pages, length, alignment,
282 pages_from_pool, own_pages);
bb873b53
ID		 283 osd_req->r_ops[which].cls.indata_len += length;
284 osd_req->r_ops[which].indata_len += length;
6c57b554
AE		 285}
286EXPORT_SYMBOL(osd_req_op_cls_request_data_pages);
287
b9e281c2
ID		 288void osd_req_op_cls_request_data_bvecs(struct ceph_osd_request *osd_req,
289 unsigned int which,
290 struct bio_vec *bvecs, u32 bytes)
291{
292 struct ceph_osd_data *osd_data;
293 struct ceph_bvec_iter it = {
294 .bvecs = bvecs,
295 .iter = { .bi_size = bytes },
296 };
297
298 osd_data = osd_req_op_data(osd_req, which, cls, request_data);
299 ceph_osd_data_bvecs_init(osd_data, &it);
300 osd_req->r_ops[which].cls.indata_len += bytes;
301 osd_req->r_ops[which].indata_len += bytes;
302}
303EXPORT_SYMBOL(osd_req_op_cls_request_data_bvecs);
304
a4ce40a9
AE		 305void osd_req_op_cls_response_data_pages(struct ceph_osd_request *osd_req,
306 unsigned int which, struct page **pages, u64 length,
307 u32 alignment, bool pages_from_pool, bool own_pages)
308{
309 struct ceph_osd_data *osd_data;
310
863c7eb5 311 osd_data = osd_req_op_data(osd_req, which, cls, response_data);
a4ce40a9
AE		 312 ceph_osd_data_pages_init(osd_data, pages, length, alignment,
313 pages_from_pool, own_pages);
a4ce40a9
AE		 314}
315EXPORT_SYMBOL(osd_req_op_cls_response_data_pages);
316
23c08a9c
AE		 317static u64 ceph_osd_data_length(struct ceph_osd_data *osd_data)
318{
319 switch (osd_data->type) {
320 case CEPH_OSD_DATA_TYPE_NONE:
321 return 0;
322 case CEPH_OSD_DATA_TYPE_PAGES:
323 return osd_data->length;
324 case CEPH_OSD_DATA_TYPE_PAGELIST:
325 return (u64)osd_data->pagelist->length;
326#ifdef CONFIG_BLOCK
327 case CEPH_OSD_DATA_TYPE_BIO:
328 return (u64)osd_data->bio_length;
329#endif /* CONFIG_BLOCK */
b9e281c2
ID		 330 case CEPH_OSD_DATA_TYPE_BVECS:
331 return osd_data->bvec_pos.iter.bi_size;
23c08a9c
AE		 332 default:
333 WARN(true, "unrecognized data type %d\n", (int)osd_data->type);
334 return 0;
335 }
336}
337
c54d47bf
AE		 338static void ceph_osd_data_release(struct ceph_osd_data *osd_data)
339{
5476492f 340 if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES && osd_data->own_pages) {
c54d47bf
AE		 341 int num_pages;
342
343 num_pages = calc_pages_for((u64)osd_data->alignment,
344 (u64)osd_data->length);
345 ceph_release_page_vector(osd_data->pages, num_pages);
346 }
5476492f
AE		 347 ceph_osd_data_init(osd_data);
348}
349
350static void osd_req_op_data_release(struct ceph_osd_request *osd_req,
351 unsigned int which)
352{
353 struct ceph_osd_req_op *op;
354
355 BUG_ON(which >= osd_req->r_num_ops);
356 op = &osd_req->r_ops[which];
357
358 switch (op->op) {
359 case CEPH_OSD_OP_READ:
360 case CEPH_OSD_OP_WRITE:
e30b7577 361 case CEPH_OSD_OP_WRITEFULL:
5476492f
AE		 362 ceph_osd_data_release(&op->extent.osd_data);
363 break;
364 case CEPH_OSD_OP_CALL:
365 ceph_osd_data_release(&op->cls.request_info);
04017e29 366 ceph_osd_data_release(&op->cls.request_data);
5476492f
AE		 367 ceph_osd_data_release(&op->cls.response_data);
368 break;
d74b50be
YZ		 369 case CEPH_OSD_OP_SETXATTR:
370 case CEPH_OSD_OP_CMPXATTR:
371 ceph_osd_data_release(&op->xattr.osd_data);
372 break;
66ba609f
YZ		 373 case CEPH_OSD_OP_STAT:
374 ceph_osd_data_release(&op->raw_data_in);
375 break;
922dab61
ID		 376 case CEPH_OSD_OP_NOTIFY_ACK:
377 ceph_osd_data_release(&op->notify_ack.request_data);
378 break;
19079203
ID		 379 case CEPH_OSD_OP_NOTIFY:
380 ceph_osd_data_release(&op->notify.request_data);
381 ceph_osd_data_release(&op->notify.response_data);
382 break;
a4ed38d7
DF		 383 case CEPH_OSD_OP_LIST_WATCHERS:
384 ceph_osd_data_release(&op->list_watchers.response_data);
385 break;
5476492f
AE		 386 default:
387 break;
388 }
c54d47bf
AE		 389}
390
63244fa1
ID		 391/*
392 * Assumes @t is zero-initialized.
393 */
394static void target_init(struct ceph_osd_request_target *t)
395{
396 ceph_oid_init(&t->base_oid);
397 ceph_oloc_init(&t->base_oloc);
398 ceph_oid_init(&t->target_oid);
399 ceph_oloc_init(&t->target_oloc);
400
401 ceph_osds_init(&t->acting);
402 ceph_osds_init(&t->up);
403 t->size = -1;
404 t->min_size = -1;
405
406 t->osd = CEPH_HOMELESS_OSD;
407}
408
922dab61
ID		 409static void target_copy(struct ceph_osd_request_target *dest,
410 const struct ceph_osd_request_target *src)
411{
412 ceph_oid_copy(&dest->base_oid, &src->base_oid);
413 ceph_oloc_copy(&dest->base_oloc, &src->base_oloc);
414 ceph_oid_copy(&dest->target_oid, &src->target_oid);
415 ceph_oloc_copy(&dest->target_oloc, &src->target_oloc);
416
417 dest->pgid = src->pgid; /* struct */
dc98ff72 418 dest->spgid = src->spgid; /* struct */
922dab61
ID		 419 dest->pg_num = src->pg_num;
420 dest->pg_num_mask = src->pg_num_mask;
421 ceph_osds_copy(&dest->acting, &src->acting);
422 ceph_osds_copy(&dest->up, &src->up);
423 dest->size = src->size;
424 dest->min_size = src->min_size;
425 dest->sort_bitwise = src->sort_bitwise;
426
427 dest->flags = src->flags;
428 dest->paused = src->paused;
429
04c7d789 430 dest->epoch = src->epoch;
dc93e0e2
ID		 431 dest->last_force_resend = src->last_force_resend;
432
922dab61
ID		 433 dest->osd = src->osd;
434}
435
63244fa1
ID		 436static void target_destroy(struct ceph_osd_request_target *t)
437{
438 ceph_oid_destroy(&t->base_oid);
30c156d9 439 ceph_oloc_destroy(&t->base_oloc);
63244fa1 440 ceph_oid_destroy(&t->target_oid);
30c156d9 441 ceph_oloc_destroy(&t->target_oloc);
63244fa1
ID		 442}
443
f24e9980
SW		 444/*
445 * requests
446 */
3540bfdb
ID		 447static void request_release_checks(struct ceph_osd_request *req)
448{
449 WARN_ON(!RB_EMPTY_NODE(&req->r_node));
4609245e 450 WARN_ON(!RB_EMPTY_NODE(&req->r_mc_node));
3540bfdb
ID		 451 WARN_ON(!list_empty(&req->r_unsafe_item));
452 WARN_ON(req->r_osd);
453}
454
9e94af20 455static void ceph_osdc_release_request(struct kref *kref)
f24e9980 456{
9e94af20
ID		 457 struct ceph_osd_request *req = container_of(kref,
458 struct ceph_osd_request, r_kref);
5476492f 459 unsigned int which;
415e49a9 460
9e94af20
ID		 461 dout("%s %p (r_request %p r_reply %p)\n", __func__, req,
462 req->r_request, req->r_reply);
3540bfdb 463 request_release_checks(req);
9e94af20 464
415e49a9
SW		 465 if (req->r_request)
466 ceph_msg_put(req->r_request);
5aea3dcd 467 if (req->r_reply)
ab8cb34a 468 ceph_msg_put(req->r_reply);
0fff87ec 469
5476492f
AE		 470 for (which = 0; which < req->r_num_ops; which++)
471 osd_req_op_data_release(req, which);
0fff87ec 472
a66dd383 473 target_destroy(&req->r_t);
415e49a9 474 ceph_put_snap_context(req->r_snapc);
d30291b9 475
415e49a9
SW		 476 if (req->r_mempool)
477 mempool_free(req, req->r_osdc->req_mempool);
3f1af42a 478 else if (req->r_num_ops <= CEPH_OSD_SLAB_OPS)
5522ae0b 479 kmem_cache_free(ceph_osd_request_cache, req);
3f1af42a
ID		 480 else
481 kfree(req);
f24e9980 482}
9e94af20
ID		 483
484void ceph_osdc_get_request(struct ceph_osd_request *req)
485{
486 dout("%s %p (was %d)\n", __func__, req,
2c935bc5 487 kref_read(&req->r_kref));
9e94af20
ID		 488 kref_get(&req->r_kref);
489}
490EXPORT_SYMBOL(ceph_osdc_get_request);
491
492void ceph_osdc_put_request(struct ceph_osd_request *req)
493{
3ed97d63
ID		 494 if (req) {
495 dout("%s %p (was %d)\n", __func__, req,
2c935bc5 496 kref_read(&req->r_kref));
3ed97d63
ID		 497 kref_put(&req->r_kref, ceph_osdc_release_request);
498 }
9e94af20
ID		 499}
500EXPORT_SYMBOL(ceph_osdc_put_request);
68b4476b 501
3540bfdb
ID		 502static void request_init(struct ceph_osd_request *req)
503{
504 /* req only, each op is zeroed in _osd_req_op_init() */
505 memset(req, 0, sizeof(*req));
506
507 kref_init(&req->r_kref);
508 init_completion(&req->r_completion);
3540bfdb 509 RB_CLEAR_NODE(&req->r_node);
4609245e 510 RB_CLEAR_NODE(&req->r_mc_node);
3540bfdb
ID		 511 INIT_LIST_HEAD(&req->r_unsafe_item);
512
513 target_init(&req->r_t);
514}
515
922dab61
ID		 516/*
517 * This is ugly, but it allows us to reuse linger registration and ping
518 * requests, keeping the structure of the code around send_linger{_ping}()
519 * reasonable. Setting up a min_nr=2 mempool for each linger request
520 * and dealing with copying ops (this blasts req only, watch op remains
521 * intact) isn't any better.
522 */
523static void request_reinit(struct ceph_osd_request *req)
524{
525 struct ceph_osd_client *osdc = req->r_osdc;
526 bool mempool = req->r_mempool;
527 unsigned int num_ops = req->r_num_ops;
528 u64 snapid = req->r_snapid;
529 struct ceph_snap_context *snapc = req->r_snapc;
530 bool linger = req->r_linger;
531 struct ceph_msg *request_msg = req->r_request;
532 struct ceph_msg *reply_msg = req->r_reply;
533
534 dout("%s req %p\n", __func__, req);
2c935bc5 535 WARN_ON(kref_read(&req->r_kref) != 1);
922dab61
ID		 536 request_release_checks(req);
537
2c935bc5
PZ		 538 WARN_ON(kref_read(&request_msg->kref) != 1);
539 WARN_ON(kref_read(&reply_msg->kref) != 1);
922dab61
ID		 540 target_destroy(&req->r_t);
541
542 request_init(req);
543 req->r_osdc = osdc;
544 req->r_mempool = mempool;
545 req->r_num_ops = num_ops;
546 req->r_snapid = snapid;
547 req->r_snapc = snapc;
548 req->r_linger = linger;
549 req->r_request = request_msg;
550 req->r_reply = reply_msg;
551}
552
3499e8a5 553struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
f24e9980 554 struct ceph_snap_context *snapc,
1b83bef2 555 unsigned int num_ops,
3499e8a5 556 bool use_mempool,
54a54007 557 gfp_t gfp_flags)
f24e9980
SW		 558{
559 struct ceph_osd_request *req;
1b83bef2 560
f24e9980 561 if (use_mempool) {
3f1af42a 562 BUG_ON(num_ops > CEPH_OSD_SLAB_OPS);
3499e8a5 563 req = mempool_alloc(osdc->req_mempool, gfp_flags);
3f1af42a
ID		 564 } else if (num_ops <= CEPH_OSD_SLAB_OPS) {
565 req = kmem_cache_alloc(ceph_osd_request_cache, gfp_flags);
f24e9980 566 } else {
3f1af42a 567 BUG_ON(num_ops > CEPH_OSD_MAX_OPS);
acafe7e3 568 req = kmalloc(struct_size(req, r_ops, num_ops), gfp_flags);
f24e9980 569 }
3f1af42a 570 if (unlikely(!req))
a79832f2 571 return NULL;
f24e9980 572
3540bfdb 573 request_init(req);
f24e9980
SW		 574 req->r_osdc = osdc;
575 req->r_mempool = use_mempool;
79528734 576 req->r_num_ops = num_ops;
84127282
ID		 577 req->r_snapid = CEPH_NOSNAP;
578 req->r_snapc = ceph_get_snap_context(snapc);
68b4476b 579
13d1ad16
ID		 580 dout("%s req %p\n", __func__, req);
581 return req;
582}
583EXPORT_SYMBOL(ceph_osdc_alloc_request);
3f1af42a 584
2e59ffd1 585static int ceph_oloc_encoding_size(const struct ceph_object_locator *oloc)
30c156d9
YZ		 586{
587 return 8 + 4 + 4 + 4 + (oloc->pool_ns ? oloc->pool_ns->len : 0);
588}
589
13d1ad16
ID		 590int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp)
591{
592 struct ceph_osd_client *osdc = req->r_osdc;
593 struct ceph_msg *msg;
594 int msg_size;
c16e7869 595
d30291b9 596 WARN_ON(ceph_oid_empty(&req->r_base_oid));
30c156d9 597 WARN_ON(ceph_oloc_empty(&req->r_base_oloc));
d30291b9 598
13d1ad16 599 /* create request message */
8cb441c0
ID		 600 msg_size = CEPH_ENCODING_START_BLK_LEN +
601 CEPH_PGID_ENCODING_LEN + 1; /* spgid */
602 msg_size += 4 + 4 + 4; /* hash, osdmap_epoch, flags */
603 msg_size += CEPH_ENCODING_START_BLK_LEN +
604 sizeof(struct ceph_osd_reqid); /* reqid */
605 msg_size += sizeof(struct ceph_blkin_trace_info); /* trace */
606 msg_size += 4 + sizeof(struct ceph_timespec); /* client_inc, mtime */
30c156d9
YZ		 607 msg_size += CEPH_ENCODING_START_BLK_LEN +
608 ceph_oloc_encoding_size(&req->r_base_oloc); /* oloc */
13d1ad16
ID		 609 msg_size += 4 + req->r_base_oid.name_len; /* oid */
610 msg_size += 2 + req->r_num_ops * sizeof(struct ceph_osd_op);
ae458f5a
ID		 611 msg_size += 8; /* snapid */
612 msg_size += 8; /* snap_seq */
13d1ad16 613 msg_size += 4 + 8 * (req->r_snapc ? req->r_snapc->num_snaps : 0);
8cb441c0 614 msg_size += 4 + 8; /* retry_attempt, features */
ae458f5a 615
13d1ad16 616 if (req->r_mempool)
8f3bc053 617 msg = ceph_msgpool_get(&osdc->msgpool_op, 0);
f24e9980 618 else
13d1ad16
ID		 619 msg = ceph_msg_new(CEPH_MSG_OSD_OP, msg_size, gfp, true);
620 if (!msg)
621 return -ENOMEM;
68b4476b 622
f24e9980 623 memset(msg->front.iov_base, 0, msg->front.iov_len);
3499e8a5 624 req->r_request = msg;
3499e8a5 625
13d1ad16
ID		 626 /* create reply message */
627 msg_size = OSD_OPREPLY_FRONT_LEN;
711da55d
ID		 628 msg_size += req->r_base_oid.name_len;
629 msg_size += req->r_num_ops * sizeof(struct ceph_osd_op);
13d1ad16
ID		 630
631 if (req->r_mempool)
632 msg = ceph_msgpool_get(&osdc->msgpool_op_reply, 0);
633 else
634 msg = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, msg_size, gfp, true);
635 if (!msg)
636 return -ENOMEM;
637
638 req->r_reply = msg;
639
640 return 0;
3499e8a5 641}
13d1ad16 642EXPORT_SYMBOL(ceph_osdc_alloc_messages);
3499e8a5 643
a8dd0a37 644static bool osd_req_opcode_valid(u16 opcode)
68b4476b 645{
a8dd0a37 646 switch (opcode) {
70b5bfa3
ID		 647#define GENERATE_CASE(op, opcode, str) case CEPH_OSD_OP_##op: return true;
648__CEPH_FORALL_OSD_OPS(GENERATE_CASE)
649#undef GENERATE_CASE
a8dd0a37
AE		 650 default:
651 return false;
652 }
653}
654
33803f33
AE		 655/*
656 * This is an osd op init function for opcodes that have no data or
657 * other information associated with them. It also serves as a
658 * common init routine for all the other init functions, below.
659 */
c99d2d4a 660static struct ceph_osd_req_op *
49719778 661_osd_req_op_init(struct ceph_osd_request *osd_req, unsigned int which,
144cba14 662 u16 opcode, u32 flags)
33803f33 663{
c99d2d4a
AE		 664 struct ceph_osd_req_op *op;
665
666 BUG_ON(which >= osd_req->r_num_ops);
33803f33
AE		 667 BUG_ON(!osd_req_opcode_valid(opcode));
668
c99d2d4a 669 op = &osd_req->r_ops[which];
33803f33 670 memset(op, 0, sizeof (*op));
33803f33 671 op->op = opcode;
144cba14 672 op->flags = flags;
c99d2d4a
AE		 673
674 return op;
33803f33
AE		 675}
676
49719778 677void osd_req_op_init(struct ceph_osd_request *osd_req,
144cba14 678 unsigned int which, u16 opcode, u32 flags)
49719778 679{
144cba14 680 (void)_osd_req_op_init(osd_req, which, opcode, flags);
49719778
AE		 681}
682EXPORT_SYMBOL(osd_req_op_init);
683
c99d2d4a
AE		 684void osd_req_op_extent_init(struct ceph_osd_request *osd_req,
685 unsigned int which, u16 opcode,
33803f33
AE		 686 u64 offset, u64 length,
687 u64 truncate_size, u32 truncate_seq)
688{
144cba14
YZ		 689 struct ceph_osd_req_op *op = _osd_req_op_init(osd_req, which,
690 opcode, 0);
33803f33
AE		 691 size_t payload_len = 0;
692
ad7a60de 693 BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
e30b7577
ID		 694 opcode != CEPH_OSD_OP_WRITEFULL && opcode != CEPH_OSD_OP_ZERO &&
695 opcode != CEPH_OSD_OP_TRUNCATE);
33803f33 696
33803f33
AE		 697 op->extent.offset = offset;
698 op->extent.length = length;
699 op->extent.truncate_size = truncate_size;
700 op->extent.truncate_seq = truncate_seq;
e30b7577 701 if (opcode == CEPH_OSD_OP_WRITE || opcode == CEPH_OSD_OP_WRITEFULL)
33803f33
AE		 702 payload_len += length;
703
de2aa102 704 op->indata_len = payload_len;
33803f33
AE		 705}
706EXPORT_SYMBOL(osd_req_op_extent_init);
707
c99d2d4a
AE		 708void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
709 unsigned int which, u64 length)
e5975c7c 710{
c99d2d4a
AE		 711 struct ceph_osd_req_op *op;
712 u64 previous;
713
714 BUG_ON(which >= osd_req->r_num_ops);
715 op = &osd_req->r_ops[which];
716 previous = op->extent.length;
e5975c7c
AE		 717
718 if (length == previous)
719 return; /* Nothing to do */
720 BUG_ON(length > previous);
721
722 op->extent.length = length;
d641df81
YZ		 723 if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
724 op->indata_len -= previous - length;
e5975c7c
AE		 725}
726EXPORT_SYMBOL(osd_req_op_extent_update);
727
2c63f49a
YZ		 728void osd_req_op_extent_dup_last(struct ceph_osd_request *osd_req,
729 unsigned int which, u64 offset_inc)
730{
731 struct ceph_osd_req_op *op, *prev_op;
732
733 BUG_ON(which + 1 >= osd_req->r_num_ops);
734
735 prev_op = &osd_req->r_ops[which];
736 op = _osd_req_op_init(osd_req, which + 1, prev_op->op, prev_op->flags);
737 /* dup previous one */
738 op->indata_len = prev_op->indata_len;
739 op->outdata_len = prev_op->outdata_len;
740 op->extent = prev_op->extent;
741 /* adjust offset */
742 op->extent.offset += offset_inc;
743 op->extent.length -= offset_inc;
744
745 if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
746 op->indata_len -= offset_inc;
747}
748EXPORT_SYMBOL(osd_req_op_extent_dup_last);
749
c99d2d4a 750void osd_req_op_cls_init(struct ceph_osd_request *osd_req, unsigned int which,
04017e29 751 u16 opcode, const char *class, const char *method)
33803f33 752{
144cba14
YZ		 753 struct ceph_osd_req_op *op = _osd_req_op_init(osd_req, which,
754 opcode, 0);
5f562df5 755 struct ceph_pagelist *pagelist;
33803f33
AE		 756 size_t payload_len = 0;
757 size_t size;
758
759 BUG_ON(opcode != CEPH_OSD_OP_CALL);
760
5f562df5
AE		 761 pagelist = kmalloc(sizeof (*pagelist), GFP_NOFS);
762 BUG_ON(!pagelist);
763 ceph_pagelist_init(pagelist);
764
33803f33
AE		 765 op->cls.class_name = class;
766 size = strlen(class);
767 BUG_ON(size > (size_t) U8_MAX);
768 op->cls.class_len = size;
5f562df5 769 ceph_pagelist_append(pagelist, class, size);
33803f33
AE		 770 payload_len += size;
771
772 op->cls.method_name = method;
773 size = strlen(method);
774 BUG_ON(size > (size_t) U8_MAX);
775 op->cls.method_len = size;
5f562df5 776 ceph_pagelist_append(pagelist, method, size);
33803f33
AE		 777 payload_len += size;
778
a4ce40a9 779 osd_req_op_cls_request_info_pagelist(osd_req, which, pagelist);
5f562df5 780
de2aa102 781 op->indata_len = payload_len;
33803f33
AE		 782}
783EXPORT_SYMBOL(osd_req_op_cls_init);
8c042b0d 784
d74b50be
YZ		 785int osd_req_op_xattr_init(struct ceph_osd_request *osd_req, unsigned int which,
786 u16 opcode, const char *name, const void *value,
787 size_t size, u8 cmp_op, u8 cmp_mode)
788{
144cba14
YZ		 789 struct ceph_osd_req_op *op = _osd_req_op_init(osd_req, which,
790 opcode, 0);
d74b50be
YZ		 791 struct ceph_pagelist *pagelist;
792 size_t payload_len;
793
794 BUG_ON(opcode != CEPH_OSD_OP_SETXATTR && opcode != CEPH_OSD_OP_CMPXATTR);
795
796 pagelist = kmalloc(sizeof(*pagelist), GFP_NOFS);
797 if (!pagelist)
798 return -ENOMEM;
799
800 ceph_pagelist_init(pagelist);
801
802 payload_len = strlen(name);
803 op->xattr.name_len = payload_len;
804 ceph_pagelist_append(pagelist, name, payload_len);
805
806 op->xattr.value_len = size;
807 ceph_pagelist_append(pagelist, value, size);
808 payload_len += size;
809
810 op->xattr.cmp_op = cmp_op;
811 op->xattr.cmp_mode = cmp_mode;
812
813 ceph_osd_data_pagelist_init(&op->xattr.osd_data, pagelist);
de2aa102 814 op->indata_len = payload_len;
d74b50be
YZ		 815 return 0;
816}
817EXPORT_SYMBOL(osd_req_op_xattr_init);
818
922dab61
ID		 819/*
820 * @watch_opcode: CEPH_OSD_WATCH_OP_*
821 */
822static void osd_req_op_watch_init(struct ceph_osd_request *req, int which,
823 u64 cookie, u8 watch_opcode)
33803f33 824{
922dab61 825 struct ceph_osd_req_op *op;
33803f33 826
922dab61 827 op = _osd_req_op_init(req, which, CEPH_OSD_OP_WATCH, 0);
33803f33 828 op->watch.cookie = cookie;
922dab61
ID		 829 op->watch.op = watch_opcode;
830 op->watch.gen = 0;
33803f33 831}
33803f33 832
c647b8a8
ID		 833void osd_req_op_alloc_hint_init(struct ceph_osd_request *osd_req,
834 unsigned int which,
835 u64 expected_object_size,
836 u64 expected_write_size)
837{
838 struct ceph_osd_req_op *op = _osd_req_op_init(osd_req, which,
144cba14
YZ		 839 CEPH_OSD_OP_SETALLOCHINT,
840 0);
c647b8a8
ID		 841
842 op->alloc_hint.expected_object_size = expected_object_size;
843 op->alloc_hint.expected_write_size = expected_write_size;
844
845 /*
846 * CEPH_OSD_OP_SETALLOCHINT op is advisory and therefore deemed
847 * not worth a feature bit. Set FAILOK per-op flag to make
848 * sure older osds don't trip over an unsupported opcode.
849 */
850 op->flags |= CEPH_OSD_OP_FLAG_FAILOK;
851}
852EXPORT_SYMBOL(osd_req_op_alloc_hint_init);
853
90af3602 854static void ceph_osdc_msg_data_add(struct ceph_msg *msg,
ec9123c5
AE		 855 struct ceph_osd_data *osd_data)
856{
857 u64 length = ceph_osd_data_length(osd_data);
858
859 if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES) {
860 BUG_ON(length > (u64) SIZE_MAX);
861 if (length)
90af3602 862 ceph_msg_data_add_pages(msg, osd_data->pages,
ec9123c5
AE		 863 length, osd_data->alignment);
864 } else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
865 BUG_ON(!length);
90af3602 866 ceph_msg_data_add_pagelist(msg, osd_data->pagelist);
ec9123c5
AE		 867#ifdef CONFIG_BLOCK
868 } else if (osd_data->type == CEPH_OSD_DATA_TYPE_BIO) {
5359a17d 869 ceph_msg_data_add_bio(msg, &osd_data->bio_pos, length);
ec9123c5 870#endif
b9e281c2
ID		 871 } else if (osd_data->type == CEPH_OSD_DATA_TYPE_BVECS) {
872 ceph_msg_data_add_bvecs(msg, &osd_data->bvec_pos);
ec9123c5
AE		 873 } else {
874 BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_NONE);
875 }
876}
877
bb873b53
ID		 878static u32 osd_req_encode_op(struct ceph_osd_op *dst,
879 const struct ceph_osd_req_op *src)
a8dd0a37 880{
a8dd0a37
AE		 881 if (WARN_ON(!osd_req_opcode_valid(src->op))) {
882 pr_err("unrecognized osd opcode %d\n", src->op);
883
884 return 0;
885 }
886
887 switch (src->op) {
888 case CEPH_OSD_OP_STAT:
889 break;
890 case CEPH_OSD_OP_READ:
891 case CEPH_OSD_OP_WRITE:
e30b7577 892 case CEPH_OSD_OP_WRITEFULL:
ad7a60de 893 case CEPH_OSD_OP_ZERO:
ad7a60de 894 case CEPH_OSD_OP_TRUNCATE:
a8dd0a37
AE		 895 dst->extent.offset = cpu_to_le64(src->extent.offset);
896 dst->extent.length = cpu_to_le64(src->extent.length);
897 dst->extent.truncate_size =
898 cpu_to_le64(src->extent.truncate_size);
899 dst->extent.truncate_seq =
900 cpu_to_le32(src->extent.truncate_seq);
901 break;
902 case CEPH_OSD_OP_CALL:
a8dd0a37
AE		 903 dst->cls.class_len = src->cls.class_len;
904 dst->cls.method_len = src->cls.method_len;
bb873b53 905 dst->cls.indata_len = cpu_to_le32(src->cls.indata_len);
a8dd0a37 906 break;
a8dd0a37
AE		 907 case CEPH_OSD_OP_WATCH:
908 dst->watch.cookie = cpu_to_le64(src->watch.cookie);
922dab61
ID		 909 dst->watch.ver = cpu_to_le64(0);
910 dst->watch.op = src->watch.op;
911 dst->watch.gen = cpu_to_le32(src->watch.gen);
912 break;
913 case CEPH_OSD_OP_NOTIFY_ACK:
a8dd0a37 914 break;
19079203
ID		 915 case CEPH_OSD_OP_NOTIFY:
916 dst->notify.cookie = cpu_to_le64(src->notify.cookie);
917 break;
a4ed38d7
DF		 918 case CEPH_OSD_OP_LIST_WATCHERS:
919 break;
c647b8a8
ID		 920 case CEPH_OSD_OP_SETALLOCHINT:
921 dst->alloc_hint.expected_object_size =
922 cpu_to_le64(src->alloc_hint.expected_object_size);
923 dst->alloc_hint.expected_write_size =
924 cpu_to_le64(src->alloc_hint.expected_write_size);
925 break;
d74b50be
YZ		 926 case CEPH_OSD_OP_SETXATTR:
927 case CEPH_OSD_OP_CMPXATTR:
928 dst->xattr.name_len = cpu_to_le32(src->xattr.name_len);
929 dst->xattr.value_len = cpu_to_le32(src->xattr.value_len);
930 dst->xattr.cmp_op = src->xattr.cmp_op;
931 dst->xattr.cmp_mode = src->xattr.cmp_mode;
d74b50be 932 break;
864e9197
YZ		 933 case CEPH_OSD_OP_CREATE:
934 case CEPH_OSD_OP_DELETE:
935 break;
a8dd0a37 936 default:
4c46459c 937 pr_err("unsupported osd opcode %s\n",
8f63ca2d 938 ceph_osd_op_name(src->op));
4c46459c 939 WARN_ON(1);
a8dd0a37
AE		 940
941 return 0;
68b4476b 942 }
7b25bf5f 943
a8dd0a37 944 dst->op = cpu_to_le16(src->op);
7b25bf5f 945 dst->flags = cpu_to_le32(src->flags);
de2aa102 946 dst->payload_len = cpu_to_le32(src->indata_len);
175face2 947
bb873b53 948 return src->indata_len;
68b4476b
YS		 949}
950
3499e8a5
YS		 951/*
952 * build new request AND message, calculate layout, and adjust file
953 * extent as needed.
954 *
955 * if the file was recently truncated, we include information about its
956 * old and new size so that the object can be updated appropriately. (we
957 * avoid synchronously deleting truncated objects because it's slow.)
3499e8a5
YS		 958 */
959struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *osdc,
960 struct ceph_file_layout *layout,
961 struct ceph_vino vino,
715e4cd4
YZ		 962 u64 off, u64 *plen,
963 unsigned int which, int num_ops,
3499e8a5
YS		 964 int opcode, int flags,
965 struct ceph_snap_context *snapc,
3499e8a5
YS		 966 u32 truncate_seq,
967 u64 truncate_size,
153e5167 968 bool use_mempool)
3499e8a5 969{
68b4476b 970 struct ceph_osd_request *req;
75d1c941
AE		 971 u64 objnum = 0;
972 u64 objoff = 0;
973 u64 objlen = 0;
6816282d 974 int r;
68b4476b 975
ad7a60de 976 BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
864e9197
YZ		 977 opcode != CEPH_OSD_OP_ZERO && opcode != CEPH_OSD_OP_TRUNCATE &&
978 opcode != CEPH_OSD_OP_CREATE && opcode != CEPH_OSD_OP_DELETE);
68b4476b 979
acead002 980 req = ceph_osdc_alloc_request(osdc, snapc, num_ops, use_mempool,
ae7ca4a3 981 GFP_NOFS);
13d1ad16
ID		 982 if (!req) {
983 r = -ENOMEM;
984 goto fail;
985 }
79528734 986
3499e8a5 987 /* calculate max write size */
a19dadfb 988 r = calc_layout(layout, off, plen, &objnum, &objoff, &objlen);
13d1ad16
ID		 989 if (r)
990 goto fail;
a19dadfb 991
864e9197 992 if (opcode == CEPH_OSD_OP_CREATE || opcode == CEPH_OSD_OP_DELETE) {
144cba14 993 osd_req_op_init(req, which, opcode, 0);
864e9197 994 } else {
7627151e 995 u32 object_size = layout->object_size;
864e9197
YZ		 996 u32 object_base = off - objoff;
997 if (!(truncate_seq == 1 && truncate_size == -1ULL)) {
998 if (truncate_size <= object_base) {
999 truncate_size = 0;
1000 } else {
1001 truncate_size -= object_base;
1002 if (truncate_size > object_size)
1003 truncate_size = object_size;
1004 }
ccca4e37 1005 }
715e4cd4 1006 osd_req_op_extent_init(req, which, opcode, objoff, objlen,
864e9197
YZ		 1007 truncate_size, truncate_seq);
1008 }
d18d1e28 1009
a1f4020a 1010 req->r_abort_on_full = true;
bb873b53 1011 req->r_flags = flags;
7627151e 1012 req->r_base_oloc.pool = layout->pool_id;
30c156d9 1013 req->r_base_oloc.pool_ns = ceph_try_get_string(layout->pool_ns);
d30291b9 1014 ceph_oid_printf(&req->r_base_oid, "%llx.%08llx", vino.ino, objnum);
dbe0fc41 1015
bb873b53
ID		 1016 req->r_snapid = vino.snap;
1017 if (flags & CEPH_OSD_FLAG_WRITE)
1018 req->r_data_offset = off;
1019
13d1ad16
ID		 1020 r = ceph_osdc_alloc_messages(req, GFP_NOFS);
1021 if (r)
1022 goto fail;
1023
f24e9980 1024 return req;
13d1ad16
ID		 1025
1026fail:
1027 ceph_osdc_put_request(req);
1028 return ERR_PTR(r);
f24e9980 1029}
3d14c5d2 1030EXPORT_SYMBOL(ceph_osdc_new_request);
f24e9980
SW		 1031
1032/*
1033 * We keep osd requests in an rbtree, sorted by ->r_tid.
1034 */
fcd00b68 1035DEFINE_RB_FUNCS(request, struct ceph_osd_request, r_tid, r_node)
4609245e 1036DEFINE_RB_FUNCS(request_mc, struct ceph_osd_request, r_tid, r_mc_node)
f24e9980 1037
0247a0cf
ID		 1038static bool osd_homeless(struct ceph_osd *osd)
1039{
1040 return osd->o_osd == CEPH_HOMELESS_OSD;
1041}
1042
5aea3dcd 1043static bool osd_registered(struct ceph_osd *osd)
f24e9980 1044{
5aea3dcd 1045 verify_osdc_locked(osd->o_osdc);
f24e9980 1046
5aea3dcd 1047 return !RB_EMPTY_NODE(&osd->o_node);
f24e9980
SW		 1048}
1049
0247a0cf
ID		 1050/*
1051 * Assumes @osd is zero-initialized.
1052 */
1053static void osd_init(struct ceph_osd *osd)
1054{
02113a0f 1055 refcount_set(&osd->o_ref, 1);
0247a0cf 1056 RB_CLEAR_NODE(&osd->o_node);
5aea3dcd 1057 osd->o_requests = RB_ROOT;
922dab61 1058 osd->o_linger_requests = RB_ROOT;
a02a946d
ID		 1059 osd->o_backoff_mappings = RB_ROOT;
1060 osd->o_backoffs_by_id = RB_ROOT;
0247a0cf
ID		 1061 INIT_LIST_HEAD(&osd->o_osd_lru);
1062 INIT_LIST_HEAD(&osd->o_keepalive_item);
1063 osd->o_incarnation = 1;
5aea3dcd 1064 mutex_init(&osd->lock);
0247a0cf
ID		 1065}
1066
1067static void osd_cleanup(struct ceph_osd *osd)
1068{
1069 WARN_ON(!RB_EMPTY_NODE(&osd->o_node));
5aea3dcd 1070 WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
922dab61 1071 WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
a02a946d
ID		 1072 WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoff_mappings));
1073 WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoffs_by_id));
0247a0cf
ID		 1074 WARN_ON(!list_empty(&osd->o_osd_lru));
1075 WARN_ON(!list_empty(&osd->o_keepalive_item));
1076
1077 if (osd->o_auth.authorizer) {
1078 WARN_ON(osd_homeless(osd));
1079 ceph_auth_destroy_authorizer(osd->o_auth.authorizer);
1080 }
1081}
1082
f24e9980
SW		 1083/*
1084 * Track open sessions with osds.
1085 */
e10006f8 1086static struct ceph_osd *create_osd(struct ceph_osd_client *osdc, int onum)
f24e9980
SW		 1087{
1088 struct ceph_osd *osd;
1089
0247a0cf
ID		 1090 WARN_ON(onum == CEPH_HOMELESS_OSD);
1091
7a28f59b 1092 osd = kzalloc(sizeof(*osd), GFP_NOIO | __GFP_NOFAIL);
0247a0cf 1093 osd_init(osd);
f24e9980 1094 osd->o_osdc = osdc;
e10006f8 1095 osd->o_osd = onum;
f24e9980 1096
b7a9e5dd 1097 ceph_con_init(&osd->o_con, osd, &osd_con_ops, &osdc->client->msgr);
4e7a5dcd 1098
f24e9980
SW		 1099 return osd;
1100}
1101
1102static struct ceph_osd *get_osd(struct ceph_osd *osd)
1103{
02113a0f
ER		 1104 if (refcount_inc_not_zero(&osd->o_ref)) {
1105 dout("get_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref)-1,
1106 refcount_read(&osd->o_ref));
f24e9980
SW		 1107 return osd;
1108 } else {
1109 dout("get_osd %p FAIL\n", osd);
1110 return NULL;
1111 }
1112}
1113
1114static void put_osd(struct ceph_osd *osd)
1115{
02113a0f
ER		 1116 dout("put_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref),
1117 refcount_read(&osd->o_ref) - 1);
1118 if (refcount_dec_and_test(&osd->o_ref)) {
0247a0cf 1119 osd_cleanup(osd);
f24e9980 1120 kfree(osd);
79494d1b 1121 }
f24e9980
SW		 1122}
1123
fcd00b68
ID		 1124DEFINE_RB_FUNCS(osd, struct ceph_osd, o_osd, o_node)
1125
9dd2845c 1126static void __move_osd_to_lru(struct ceph_osd *osd)
f5a2041b 1127{
9dd2845c
ID		 1128 struct ceph_osd_client *osdc = osd->o_osdc;
1129
1130 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
f5a2041b 1131 BUG_ON(!list_empty(&osd->o_osd_lru));
bbf37ec3 1132
9dd2845c 1133 spin_lock(&osdc->osd_lru_lock);
f5a2041b 1134 list_add_tail(&osd->o_osd_lru, &osdc->osd_lru);
9dd2845c
ID		 1135 spin_unlock(&osdc->osd_lru_lock);
1136
a319bf56 1137 osd->lru_ttl = jiffies + osdc->client->options->osd_idle_ttl;
f5a2041b
YS		 1138}
1139
9dd2845c 1140static void maybe_move_osd_to_lru(struct ceph_osd *osd)
bbf37ec3 1141{
5aea3dcd 1142 if (RB_EMPTY_ROOT(&osd->o_requests) &&
922dab61 1143 RB_EMPTY_ROOT(&osd->o_linger_requests))
9dd2845c 1144 __move_osd_to_lru(osd);
bbf37ec3
ID		 1145}
1146
f5a2041b
YS		 1147static void __remove_osd_from_lru(struct ceph_osd *osd)
1148{
9dd2845c
ID		 1149 struct ceph_osd_client *osdc = osd->o_osdc;
1150
1151 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1152
1153 spin_lock(&osdc->osd_lru_lock);
f5a2041b
YS		 1154 if (!list_empty(&osd->o_osd_lru))
1155 list_del_init(&osd->o_osd_lru);
9dd2845c 1156 spin_unlock(&osdc->osd_lru_lock);
f5a2041b
YS		 1157}
1158
5aea3dcd
ID		 1159/*
1160 * Close the connection and assign any leftover requests to the
1161 * homeless session.
1162 */
1163static void close_osd(struct ceph_osd *osd)
1164{
1165 struct ceph_osd_client *osdc = osd->o_osdc;
1166 struct rb_node *n;
1167
1168 verify_osdc_wrlocked(osdc);
1169 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1170
1171 ceph_con_close(&osd->o_con);
1172
1173 for (n = rb_first(&osd->o_requests); n; ) {
1174 struct ceph_osd_request *req =
1175 rb_entry(n, struct ceph_osd_request, r_node);
1176
1177 n = rb_next(n); /* unlink_request() */
1178
1179 dout(" reassigning req %p tid %llu\n", req, req->r_tid);
1180 unlink_request(osd, req);
1181 link_request(&osdc->homeless_osd, req);
1182 }
922dab61
ID		 1183 for (n = rb_first(&osd->o_linger_requests); n; ) {
1184 struct ceph_osd_linger_request *lreq =
1185 rb_entry(n, struct ceph_osd_linger_request, node);
1186
1187 n = rb_next(n); /* unlink_linger() */
1188
1189 dout(" reassigning lreq %p linger_id %llu\n", lreq,
1190 lreq->linger_id);
1191 unlink_linger(osd, lreq);
1192 link_linger(&osdc->homeless_osd, lreq);
1193 }
a02a946d 1194 clear_backoffs(osd);
5aea3dcd
ID		 1195
1196 __remove_osd_from_lru(osd);
1197 erase_osd(&osdc->osds, osd);
1198 put_osd(osd);
1199}
1200
f24e9980
SW		 1201/*
1202 * reset osd connect
1203 */
5aea3dcd 1204static int reopen_osd(struct ceph_osd *osd)
f24e9980 1205{
c3acb181 1206 struct ceph_entity_addr *peer_addr;
f24e9980 1207
5aea3dcd
ID		 1208 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1209
1210 if (RB_EMPTY_ROOT(&osd->o_requests) &&
922dab61 1211 RB_EMPTY_ROOT(&osd->o_linger_requests)) {
5aea3dcd 1212 close_osd(osd);
c3acb181
AE		 1213 return -ENODEV;
1214 }
1215
5aea3dcd 1216 peer_addr = &osd->o_osdc->osdmap->osd_addr[osd->o_osd];
c3acb181
AE		 1217 if (!memcmp(peer_addr, &osd->o_con.peer_addr, sizeof (*peer_addr)) &&
1218 !ceph_con_opened(&osd->o_con)) {
5aea3dcd 1219 struct rb_node *n;
c3acb181 1220
0b4af2e8
ID		 1221 dout("osd addr hasn't changed and connection never opened, "
1222 "letting msgr retry\n");
87b315a5 1223 /* touch each r_stamp for handle_timeout()'s benfit */
5aea3dcd
ID		 1224 for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
1225 struct ceph_osd_request *req =
1226 rb_entry(n, struct ceph_osd_request, r_node);
87b315a5 1227 req->r_stamp = jiffies;
5aea3dcd 1228 }
c3acb181
AE		 1229
1230 return -EAGAIN;
f24e9980 1231 }
c3acb181
AE		 1232
1233 ceph_con_close(&osd->o_con);
1234 ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd, peer_addr);
1235 osd->o_incarnation++;
1236
1237 return 0;
f24e9980
SW		 1238}
1239
5aea3dcd
ID		 1240static struct ceph_osd *lookup_create_osd(struct ceph_osd_client *osdc, int o,
1241 bool wrlocked)
f24e9980 1242{
5aea3dcd 1243 struct ceph_osd *osd;
35f9f8a0 1244
5aea3dcd
ID		 1245 if (wrlocked)
1246 verify_osdc_wrlocked(osdc);
1247 else
1248 verify_osdc_locked(osdc);
f24e9980 1249
5aea3dcd
ID		 1250 if (o != CEPH_HOMELESS_OSD)
1251 osd = lookup_osd(&osdc->osds, o);
1252 else
1253 osd = &osdc->homeless_osd;
1254 if (!osd) {
1255 if (!wrlocked)
1256 return ERR_PTR(-EAGAIN);
0ba6478d 1257
5aea3dcd
ID		 1258 osd = create_osd(osdc, o);
1259 insert_osd(&osdc->osds, osd);
1260 ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd,
1261 &osdc->osdmap->osd_addr[osd->o_osd]);
0ba6478d 1262 }
f24e9980 1263
5aea3dcd
ID		 1264 dout("%s osdc %p osd%d -> osd %p\n", __func__, osdc, o, osd);
1265 return osd;
f24e9980
SW		 1266}
1267
1268/*
5aea3dcd
ID		 1269 * Create request <-> OSD session relation.
1270 *
1271 * @req has to be assigned a tid, @osd may be homeless.
f24e9980 1272 */
5aea3dcd 1273static void link_request(struct ceph_osd *osd, struct ceph_osd_request *req)
f24e9980 1274{
5aea3dcd
ID		 1275 verify_osd_locked(osd);
1276 WARN_ON(!req->r_tid || req->r_osd);
1277 dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1278 req, req->r_tid);
1279
1280 if (!osd_homeless(osd))
1281 __remove_osd_from_lru(osd);
1282 else
1283 atomic_inc(&osd->o_osdc->num_homeless);
1284
1285 get_osd(osd);
1286 insert_request(&osd->o_requests, req);
1287 req->r_osd = osd;
f24e9980
SW		 1288}
1289
5aea3dcd
ID		 1290static void unlink_request(struct ceph_osd *osd, struct ceph_osd_request *req)
1291{
1292 verify_osd_locked(osd);
1293 WARN_ON(req->r_osd != osd);
1294 dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1295 req, req->r_tid);
1296
1297 req->r_osd = NULL;
1298 erase_request(&osd->o_requests, req);
1299 put_osd(osd);
1300
1301 if (!osd_homeless(osd))
1302 maybe_move_osd_to_lru(osd);
1303 else
1304 atomic_dec(&osd->o_osdc->num_homeless);
1305}
1306
63244fa1
ID		 1307static bool __pool_full(struct ceph_pg_pool_info *pi)
1308{
1309 return pi->flags & CEPH_POOL_FLAG_FULL;
1310}
1311
42c1b124
ID		 1312static bool have_pool_full(struct ceph_osd_client *osdc)
1313{
1314 struct rb_node *n;
1315
1316 for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
1317 struct ceph_pg_pool_info *pi =
1318 rb_entry(n, struct ceph_pg_pool_info, node);
1319
1320 if (__pool_full(pi))
1321 return true;
1322 }
1323
1324 return false;
1325}
1326
5aea3dcd
ID		 1327static bool pool_full(struct ceph_osd_client *osdc, s64 pool_id)
1328{
1329 struct ceph_pg_pool_info *pi;
1330
1331 pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
1332 if (!pi)
1333 return false;
1334
1335 return __pool_full(pi);
1336}
1337
d29adb34
JD		 1338/*
1339 * Returns whether a request should be blocked from being sent
1340 * based on the current osdmap and osd_client settings.
d29adb34 1341 */
63244fa1
ID		 1342static bool target_should_be_paused(struct ceph_osd_client *osdc,
1343 const struct ceph_osd_request_target *t,
1344 struct ceph_pg_pool_info *pi)
1345{
b7ec35b3
ID		 1346 bool pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
1347 bool pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
1348 ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
63244fa1
ID		 1349 __pool_full(pi);
1350
6d637a54 1351 WARN_ON(pi->id != t->target_oloc.pool);
58eb7932
JL		 1352 return ((t->flags & CEPH_OSD_FLAG_READ) && pauserd) ||
1353 ((t->flags & CEPH_OSD_FLAG_WRITE) && pausewr) ||
1354 (osdc->osdmap->epoch < osdc->epoch_barrier);
63244fa1
ID		 1355}
1356
63244fa1
ID		 1357enum calc_target_result {
1358 CALC_TARGET_NO_ACTION = 0,
1359 CALC_TARGET_NEED_RESEND,
1360 CALC_TARGET_POOL_DNE,
1361};
1362
1363static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
1364 struct ceph_osd_request_target *t,
7de030d6 1365 struct ceph_connection *con,
63244fa1
ID		 1366 bool any_change)
1367{
1368 struct ceph_pg_pool_info *pi;
1369 struct ceph_pg pgid, last_pgid;
1370 struct ceph_osds up, acting;
1371 bool force_resend = false;
84ed45df
ID		 1372 bool unpaused = false;
1373 bool legacy_change;
7de030d6 1374 bool split = false;
b7ec35b3 1375 bool sort_bitwise = ceph_osdmap_flag(osdc, CEPH_OSDMAP_SORTBITWISE);
ae78dd81
ID		 1376 bool recovery_deletes = ceph_osdmap_flag(osdc,
1377 CEPH_OSDMAP_RECOVERY_DELETES);
63244fa1
ID		 1378 enum calc_target_result ct_res;
1379 int ret;
1380
04c7d789 1381 t->epoch = osdc->osdmap->epoch;
63244fa1
ID		 1382 pi = ceph_pg_pool_by_id(osdc->osdmap, t->base_oloc.pool);
1383 if (!pi) {
1384 t->osd = CEPH_HOMELESS_OSD;
1385 ct_res = CALC_TARGET_POOL_DNE;
1386 goto out;
1387 }
1388
1389 if (osdc->osdmap->epoch == pi->last_force_request_resend) {
dc93e0e2
ID		 1390 if (t->last_force_resend < pi->last_force_request_resend) {
1391 t->last_force_resend = pi->last_force_request_resend;
63244fa1 1392 force_resend = true;
dc93e0e2 1393 } else if (t->last_force_resend == 0) {
63244fa1
ID		 1394 force_resend = true;
1395 }
1396 }
63244fa1 1397
db098ec4
ID		 1398 /* apply tiering */
1399 ceph_oid_copy(&t->target_oid, &t->base_oid);
1400 ceph_oloc_copy(&t->target_oloc, &t->base_oloc);
1401 if ((t->flags & CEPH_OSD_FLAG_IGNORE_OVERLAY) == 0) {
63244fa1
ID		 1402 if (t->flags & CEPH_OSD_FLAG_READ && pi->read_tier >= 0)
1403 t->target_oloc.pool = pi->read_tier;
1404 if (t->flags & CEPH_OSD_FLAG_WRITE && pi->write_tier >= 0)
1405 t->target_oloc.pool = pi->write_tier;
6d637a54
ID		 1406
1407 pi = ceph_pg_pool_by_id(osdc->osdmap, t->target_oloc.pool);
1408 if (!pi) {
1409 t->osd = CEPH_HOMELESS_OSD;
1410 ct_res = CALC_TARGET_POOL_DNE;
1411 goto out;
1412 }
63244fa1
ID		 1413 }
1414
df28152d
ID		 1415 ret = __ceph_object_locator_to_pg(pi, &t->target_oid, &t->target_oloc,
1416 &pgid);
63244fa1
ID		 1417 if (ret) {
1418 WARN_ON(ret != -ENOENT);
1419 t->osd = CEPH_HOMELESS_OSD;
1420 ct_res = CALC_TARGET_POOL_DNE;
1421 goto out;
1422 }
1423 last_pgid.pool = pgid.pool;
1424 last_pgid.seed = ceph_stable_mod(pgid.seed, t->pg_num, t->pg_num_mask);
1425
df28152d 1426 ceph_pg_to_up_acting_osds(osdc->osdmap, pi, &pgid, &up, &acting);
63244fa1
ID		 1427 if (any_change &&
1428 ceph_is_new_interval(&t->acting,
1429 &acting,
1430 &t->up,
1431 &up,
1432 t->size,
1433 pi->size,
1434 t->min_size,
1435 pi->min_size,
1436 t->pg_num,
1437 pi->pg_num,
1438 t->sort_bitwise,
1439 sort_bitwise,
ae78dd81
ID		 1440 t->recovery_deletes,
1441 recovery_deletes,
63244fa1
ID		 1442 &last_pgid))
1443 force_resend = true;
1444
1445 if (t->paused && !target_should_be_paused(osdc, t, pi)) {
1446 t->paused = false;
84ed45df 1447 unpaused = true;
63244fa1 1448 }
84ed45df
ID		 1449 legacy_change = ceph_pg_compare(&t->pgid, &pgid) ||
1450 ceph_osds_changed(&t->acting, &acting, any_change);
7de030d6
ID		 1451 if (t->pg_num)
1452 split = ceph_pg_is_split(&last_pgid, t->pg_num, pi->pg_num);
63244fa1 1453
7de030d6 1454 if (legacy_change || force_resend || split) {
63244fa1 1455 t->pgid = pgid; /* struct */
df28152d 1456 ceph_pg_to_primary_shard(osdc->osdmap, pi, &pgid, &t->spgid);
63244fa1
ID		 1457 ceph_osds_copy(&t->acting, &acting);
1458 ceph_osds_copy(&t->up, &up);
1459 t->size = pi->size;
1460 t->min_size = pi->min_size;
1461 t->pg_num = pi->pg_num;
1462 t->pg_num_mask = pi->pg_num_mask;
1463 t->sort_bitwise = sort_bitwise;
ae78dd81 1464 t->recovery_deletes = recovery_deletes;
63244fa1
ID		 1465
1466 t->osd = acting.primary;
63244fa1
ID		 1467 }
1468
7de030d6
ID		 1469 if (unpaused || legacy_change || force_resend ||
1470 (split && con && CEPH_HAVE_FEATURE(con->peer_features,
1471 RESEND_ON_SPLIT)))
84ed45df
ID		 1472 ct_res = CALC_TARGET_NEED_RESEND;
1473 else
1474 ct_res = CALC_TARGET_NO_ACTION;
1475
63244fa1
ID		 1476out:
1477 dout("%s t %p -> ct_res %d osd %d\n", __func__, t, ct_res, t->osd);
1478 return ct_res;
1479}
1480
a02a946d
ID		 1481static struct ceph_spg_mapping *alloc_spg_mapping(void)
1482{
1483 struct ceph_spg_mapping *spg;
1484
1485 spg = kmalloc(sizeof(*spg), GFP_NOIO);
1486 if (!spg)
1487 return NULL;
1488
1489 RB_CLEAR_NODE(&spg->node);
1490 spg->backoffs = RB_ROOT;
1491 return spg;
1492}
1493
1494static void free_spg_mapping(struct ceph_spg_mapping *spg)
1495{
1496 WARN_ON(!RB_EMPTY_NODE(&spg->node));
1497 WARN_ON(!RB_EMPTY_ROOT(&spg->backoffs));
1498
1499 kfree(spg);
1500}
1501
1502/*
1503 * rbtree of ceph_spg_mapping for handling map<spg_t, ...>, similar to
1504 * ceph_pg_mapping. Used to track OSD backoffs -- a backoff [range] is
1505 * defined only within a specific spgid; it does not pass anything to
1506 * children on split, or to another primary.
1507 */
1508DEFINE_RB_FUNCS2(spg_mapping, struct ceph_spg_mapping, spgid, ceph_spg_compare,
1509 RB_BYPTR, const struct ceph_spg *, node)
1510
1511static u64 hoid_get_bitwise_key(const struct ceph_hobject_id *hoid)
1512{
1513 return hoid->is_max ? 0x100000000ull : hoid->hash_reverse_bits;
1514}
1515
1516static void hoid_get_effective_key(const struct ceph_hobject_id *hoid,
1517 void **pkey, size_t *pkey_len)
1518{
1519 if (hoid->key_len) {
1520 *pkey = hoid->key;
1521 *pkey_len = hoid->key_len;
1522 } else {
1523 *pkey = hoid->oid;
1524 *pkey_len = hoid->oid_len;
1525 }
1526}
1527
1528static int compare_names(const void *name1, size_t name1_len,
1529 const void *name2, size_t name2_len)
1530{
1531 int ret;
1532
1533 ret = memcmp(name1, name2, min(name1_len, name2_len));
1534 if (!ret) {
1535 if (name1_len < name2_len)
1536 ret = -1;
1537 else if (name1_len > name2_len)
1538 ret = 1;
1539 }
1540 return ret;
1541}
1542
1543static int hoid_compare(const struct ceph_hobject_id *lhs,
1544 const struct ceph_hobject_id *rhs)
1545{
1546 void *effective_key1, *effective_key2;
1547 size_t effective_key1_len, effective_key2_len;
1548 int ret;
1549
1550 if (lhs->is_max < rhs->is_max)
1551 return -1;
1552 if (lhs->is_max > rhs->is_max)
1553 return 1;
1554
1555 if (lhs->pool < rhs->pool)
1556 return -1;
1557 if (lhs->pool > rhs->pool)
1558 return 1;
1559
1560 if (hoid_get_bitwise_key(lhs) < hoid_get_bitwise_key(rhs))
1561 return -1;
1562 if (hoid_get_bitwise_key(lhs) > hoid_get_bitwise_key(rhs))
1563 return 1;
1564
1565 ret = compare_names(lhs->nspace, lhs->nspace_len,
1566 rhs->nspace, rhs->nspace_len);
1567 if (ret)
1568 return ret;
1569
1570 hoid_get_effective_key(lhs, &effective_key1, &effective_key1_len);
1571 hoid_get_effective_key(rhs, &effective_key2, &effective_key2_len);
1572 ret = compare_names(effective_key1, effective_key1_len,
1573 effective_key2, effective_key2_len);
1574 if (ret)
1575 return ret;
1576
1577 ret = compare_names(lhs->oid, lhs->oid_len, rhs->oid, rhs->oid_len);
1578 if (ret)
1579 return ret;
1580
1581 if (lhs->snapid < rhs->snapid)
1582 return -1;
1583 if (lhs->snapid > rhs->snapid)
1584 return 1;
1585
1586 return 0;
1587}
1588
1589/*
1590 * For decoding ->begin and ->end of MOSDBackoff only -- no MIN/MAX
1591 * compat stuff here.
1592 *
1593 * Assumes @hoid is zero-initialized.
1594 */
1595static int decode_hoid(void **p, void *end, struct ceph_hobject_id *hoid)
1596{
1597 u8 struct_v;
1598 u32 struct_len;
1599 int ret;
1600
1601 ret = ceph_start_decoding(p, end, 4, "hobject_t", &struct_v,
1602 &struct_len);
1603 if (ret)
1604 return ret;
1605
1606 if (struct_v < 4) {
1607 pr_err("got struct_v %d < 4 of hobject_t\n", struct_v);
1608 goto e_inval;
1609 }
1610
1611 hoid->key = ceph_extract_encoded_string(p, end, &hoid->key_len,
1612 GFP_NOIO);
1613 if (IS_ERR(hoid->key)) {
1614 ret = PTR_ERR(hoid->key);
1615 hoid->key = NULL;
1616 return ret;
1617 }
1618
1619 hoid->oid = ceph_extract_encoded_string(p, end, &hoid->oid_len,
1620 GFP_NOIO);
1621 if (IS_ERR(hoid->oid)) {
1622 ret = PTR_ERR(hoid->oid);
1623 hoid->oid = NULL;
1624 return ret;
1625 }
1626
1627 ceph_decode_64_safe(p, end, hoid->snapid, e_inval);
1628 ceph_decode_32_safe(p, end, hoid->hash, e_inval);
1629 ceph_decode_8_safe(p, end, hoid->is_max, e_inval);
1630
1631 hoid->nspace = ceph_extract_encoded_string(p, end, &hoid->nspace_len,
1632 GFP_NOIO);
1633 if (IS_ERR(hoid->nspace)) {
1634 ret = PTR_ERR(hoid->nspace);
1635 hoid->nspace = NULL;
1636 return ret;
1637 }
1638
1639 ceph_decode_64_safe(p, end, hoid->pool, e_inval);
1640
1641 ceph_hoid_build_hash_cache(hoid);
1642 return 0;
1643
1644e_inval:
1645 return -EINVAL;
1646}
1647
1648static int hoid_encoding_size(const struct ceph_hobject_id *hoid)
1649{
1650 return 8 + 4 + 1 + 8 + /* snapid, hash, is_max, pool */
1651 4 + hoid->key_len + 4 + hoid->oid_len + 4 + hoid->nspace_len;
1652}
1653
1654static void encode_hoid(void **p, void *end, const struct ceph_hobject_id *hoid)
1655{
1656 ceph_start_encoding(p, 4, 3, hoid_encoding_size(hoid));
1657 ceph_encode_string(p, end, hoid->key, hoid->key_len);
1658 ceph_encode_string(p, end, hoid->oid, hoid->oid_len);
1659 ceph_encode_64(p, hoid->snapid);
1660 ceph_encode_32(p, hoid->hash);
1661 ceph_encode_8(p, hoid->is_max);
1662 ceph_encode_string(p, end, hoid->nspace, hoid->nspace_len);
1663 ceph_encode_64(p, hoid->pool);
1664}
1665
1666static void free_hoid(struct ceph_hobject_id *hoid)
1667{
1668 if (hoid) {
1669 kfree(hoid->key);
1670 kfree(hoid->oid);
1671 kfree(hoid->nspace);
1672 kfree(hoid);
1673 }
1674}
1675
1676static struct ceph_osd_backoff *alloc_backoff(void)
1677{
1678 struct ceph_osd_backoff *backoff;
1679
1680 backoff = kzalloc(sizeof(*backoff), GFP_NOIO);
1681 if (!backoff)
1682 return NULL;
1683
1684 RB_CLEAR_NODE(&backoff->spg_node);
1685 RB_CLEAR_NODE(&backoff->id_node);
1686 return backoff;
1687}
1688
1689static void free_backoff(struct ceph_osd_backoff *backoff)
1690{
1691 WARN_ON(!RB_EMPTY_NODE(&backoff->spg_node));
1692 WARN_ON(!RB_EMPTY_NODE(&backoff->id_node));
1693
1694 free_hoid(backoff->begin);
1695 free_hoid(backoff->end);
1696 kfree(backoff);
1697}
1698
1699/*
1700 * Within a specific spgid, backoffs are managed by ->begin hoid.
1701 */
1702DEFINE_RB_INSDEL_FUNCS2(backoff, struct ceph_osd_backoff, begin, hoid_compare,
1703 RB_BYVAL, spg_node);
1704
1705static struct ceph_osd_backoff *lookup_containing_backoff(struct rb_root *root,
1706 const struct ceph_hobject_id *hoid)
1707{
1708 struct rb_node *n = root->rb_node;
1709
1710 while (n) {
1711 struct ceph_osd_backoff *cur =
1712 rb_entry(n, struct ceph_osd_backoff, spg_node);
1713 int cmp;
1714
1715 cmp = hoid_compare(hoid, cur->begin);
1716 if (cmp < 0) {
1717 n = n->rb_left;
1718 } else if (cmp > 0) {
1719 if (hoid_compare(hoid, cur->end) < 0)
1720 return cur;
1721
1722 n = n->rb_right;
1723 } else {
1724 return cur;
1725 }
1726 }
1727
1728 return NULL;
1729}
1730
1731/*
1732 * Each backoff has a unique id within its OSD session.
1733 */
1734DEFINE_RB_FUNCS(backoff_by_id, struct ceph_osd_backoff, id, id_node)
1735
1736static void clear_backoffs(struct ceph_osd *osd)
1737{
1738 while (!RB_EMPTY_ROOT(&osd->o_backoff_mappings)) {
1739 struct ceph_spg_mapping *spg =
1740 rb_entry(rb_first(&osd->o_backoff_mappings),
1741 struct ceph_spg_mapping, node);
1742
1743 while (!RB_EMPTY_ROOT(&spg->backoffs)) {
1744 struct ceph_osd_backoff *backoff =
1745 rb_entry(rb_first(&spg->backoffs),
1746 struct ceph_osd_backoff, spg_node);
1747
1748 erase_backoff(&spg->backoffs, backoff);
1749 erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
1750 free_backoff(backoff);
1751 }
1752 erase_spg_mapping(&osd->o_backoff_mappings, spg);
1753 free_spg_mapping(spg);
1754 }
1755}
1756
1757/*
1758 * Set up a temporary, non-owning view into @t.
1759 */
1760static void hoid_fill_from_target(struct ceph_hobject_id *hoid,
1761 const struct ceph_osd_request_target *t)
1762{
1763 hoid->key = NULL;
1764 hoid->key_len = 0;
1765 hoid->oid = t->target_oid.name;
1766 hoid->oid_len = t->target_oid.name_len;
1767 hoid->snapid = CEPH_NOSNAP;
1768 hoid->hash = t->pgid.seed;
1769 hoid->is_max = false;
1770 if (t->target_oloc.pool_ns) {
1771 hoid->nspace = t->target_oloc.pool_ns->str;
1772 hoid->nspace_len = t->target_oloc.pool_ns->len;
1773 } else {
1774 hoid->nspace = NULL;
1775 hoid->nspace_len = 0;
1776 }
1777 hoid->pool = t->target_oloc.pool;
1778 ceph_hoid_build_hash_cache(hoid);
1779}
1780
1781static bool should_plug_request(struct ceph_osd_request *req)
1782{
1783 struct ceph_osd *osd = req->r_osd;
1784 struct ceph_spg_mapping *spg;
1785 struct ceph_osd_backoff *backoff;
1786 struct ceph_hobject_id hoid;
1787
1788 spg = lookup_spg_mapping(&osd->o_backoff_mappings, &req->r_t.spgid);
1789 if (!spg)
1790 return false;
1791
1792 hoid_fill_from_target(&hoid, &req->r_t);
1793 backoff = lookup_containing_backoff(&spg->backoffs, &hoid);
1794 if (!backoff)
1795 return false;
1796
1797 dout("%s req %p tid %llu backoff osd%d spgid %llu.%xs%d id %llu\n",
1798 __func__, req, req->r_tid, osd->o_osd, backoff->spgid.pgid.pool,
1799 backoff->spgid.pgid.seed, backoff->spgid.shard, backoff->id);
1800 return true;
1801}
1802
bb873b53
ID		 1803static void setup_request_data(struct ceph_osd_request *req,
1804 struct ceph_msg *msg)
f24e9980 1805{
bb873b53
ID		 1806 u32 data_len = 0;
1807 int i;
1808
1809 if (!list_empty(&msg->data))
1810 return;
f24e9980 1811
bb873b53
ID		 1812 WARN_ON(msg->data_length);
1813 for (i = 0; i < req->r_num_ops; i++) {
1814 struct ceph_osd_req_op *op = &req->r_ops[i];
1815
1816 switch (op->op) {
1817 /* request */
1818 case CEPH_OSD_OP_WRITE:
1819 case CEPH_OSD_OP_WRITEFULL:
1820 WARN_ON(op->indata_len != op->extent.length);
1821 ceph_osdc_msg_data_add(msg, &op->extent.osd_data);
1822 break;
1823 case CEPH_OSD_OP_SETXATTR:
1824 case CEPH_OSD_OP_CMPXATTR:
1825 WARN_ON(op->indata_len != op->xattr.name_len +
1826 op->xattr.value_len);
1827 ceph_osdc_msg_data_add(msg, &op->xattr.osd_data);
1828 break;
922dab61
ID		 1829 case CEPH_OSD_OP_NOTIFY_ACK:
1830 ceph_osdc_msg_data_add(msg,
1831 &op->notify_ack.request_data);
1832 break;
bb873b53
ID		 1833
1834 /* reply */
1835 case CEPH_OSD_OP_STAT:
1836 ceph_osdc_msg_data_add(req->r_reply,
1837 &op->raw_data_in);
1838 break;
1839 case CEPH_OSD_OP_READ:
1840 ceph_osdc_msg_data_add(req->r_reply,
1841 &op->extent.osd_data);
1842 break;
a4ed38d7
DF		 1843 case CEPH_OSD_OP_LIST_WATCHERS:
1844 ceph_osdc_msg_data_add(req->r_reply,
1845 &op->list_watchers.response_data);
1846 break;
bb873b53
ID		 1847
1848 /* both */
1849 case CEPH_OSD_OP_CALL:
1850 WARN_ON(op->indata_len != op->cls.class_len +
1851 op->cls.method_len +
1852 op->cls.indata_len);
1853 ceph_osdc_msg_data_add(msg, &op->cls.request_info);
1854 /* optional, can be NONE */
1855 ceph_osdc_msg_data_add(msg, &op->cls.request_data);
1856 /* optional, can be NONE */
1857 ceph_osdc_msg_data_add(req->r_reply,
1858 &op->cls.response_data);
1859 break;
19079203
ID		 1860 case CEPH_OSD_OP_NOTIFY:
1861 ceph_osdc_msg_data_add(msg,
1862 &op->notify.request_data);
1863 ceph_osdc_msg_data_add(req->r_reply,
1864 &op->notify.response_data);
1865 break;
bb873b53
ID		 1866 }
1867
1868 data_len += op->indata_len;
1869 }
1b83bef2 1870
bb873b53
ID		 1871 WARN_ON(data_len != msg->data_length);
1872}
1873
2e59ffd1
ID		 1874static void encode_pgid(void **p, const struct ceph_pg *pgid)
1875{
1876 ceph_encode_8(p, 1);
1877 ceph_encode_64(p, pgid->pool);
1878 ceph_encode_32(p, pgid->seed);
1879 ceph_encode_32(p, -1); /* preferred */
1880}
1881
8cb441c0
ID		 1882static void encode_spgid(void **p, const struct ceph_spg *spgid)
1883{
1884 ceph_start_encoding(p, 1, 1, CEPH_PGID_ENCODING_LEN + 1);
1885 encode_pgid(p, &spgid->pgid);
1886 ceph_encode_8(p, spgid->shard);
1887}
1888
2e59ffd1
ID		 1889static void encode_oloc(void **p, void *end,
1890 const struct ceph_object_locator *oloc)
1891{
1892 ceph_start_encoding(p, 5, 4, ceph_oloc_encoding_size(oloc));
1893 ceph_encode_64(p, oloc->pool);
1894 ceph_encode_32(p, -1); /* preferred */
1895 ceph_encode_32(p, 0); /* key len */
1896 if (oloc->pool_ns)
1897 ceph_encode_string(p, end, oloc->pool_ns->str,
1898 oloc->pool_ns->len);
1899 else
1900 ceph_encode_32(p, 0);
1901}
1902
8cb441c0
ID		 1903static void encode_request_partial(struct ceph_osd_request *req,
1904 struct ceph_msg *msg)
bb873b53
ID		 1905{
1906 void *p = msg->front.iov_base;
1907 void *const end = p + msg->front_alloc_len;
1908 u32 data_len = 0;
1909 int i;
1910
1911 if (req->r_flags & CEPH_OSD_FLAG_WRITE) {
1912 /* snapshots aren't writeable */
1913 WARN_ON(req->r_snapid != CEPH_NOSNAP);
1914 } else {
1915 WARN_ON(req->r_mtime.tv_sec || req->r_mtime.tv_nsec ||
1916 req->r_data_offset || req->r_snapc);
1917 }
1918
1919 setup_request_data(req, msg);
1920
8cb441c0
ID		 1921 encode_spgid(&p, &req->r_t.spgid); /* actual spg */
1922 ceph_encode_32(&p, req->r_t.pgid.seed); /* raw hash */
bb873b53
ID		 1923 ceph_encode_32(&p, req->r_osdc->osdmap->epoch);
1924 ceph_encode_32(&p, req->r_flags);
8cb441c0
ID		 1925
1926 /* reqid */
1927 ceph_start_encoding(&p, 2, 2, sizeof(struct ceph_osd_reqid));
1928 memset(p, 0, sizeof(struct ceph_osd_reqid));
1929 p += sizeof(struct ceph_osd_reqid);
1930
1931 /* trace */
1932 memset(p, 0, sizeof(struct ceph_blkin_trace_info));
1933 p += sizeof(struct ceph_blkin_trace_info);
1934
1935 ceph_encode_32(&p, 0); /* client_inc, always 0 */
bb873b53
ID		 1936 ceph_encode_timespec(p, &req->r_mtime);
1937 p += sizeof(struct ceph_timespec);
aa26d662 1938
2e59ffd1 1939 encode_oloc(&p, end, &req->r_t.target_oloc);
2e59ffd1
ID		 1940 ceph_encode_string(&p, end, req->r_t.target_oid.name,
1941 req->r_t.target_oid.name_len);
f24e9980 1942
bb873b53
ID		 1943 /* ops, can imply data */
1944 ceph_encode_16(&p, req->r_num_ops);
1945 for (i = 0; i < req->r_num_ops; i++) {
1946 data_len += osd_req_encode_op(p, &req->r_ops[i]);
1947 p += sizeof(struct ceph_osd_op);
1948 }
26be8808 1949
bb873b53
ID		 1950 ceph_encode_64(&p, req->r_snapid); /* snapid */
1951 if (req->r_snapc) {
1952 ceph_encode_64(&p, req->r_snapc->seq);
1953 ceph_encode_32(&p, req->r_snapc->num_snaps);
1954 for (i = 0; i < req->r_snapc->num_snaps; i++)
1955 ceph_encode_64(&p, req->r_snapc->snaps[i]);
1956 } else {
1957 ceph_encode_64(&p, 0); /* snap_seq */
1958 ceph_encode_32(&p, 0); /* snaps len */
1959 }
1960
1961 ceph_encode_32(&p, req->r_attempts); /* retry_attempt */
986e8989 1962 BUG_ON(p > end - 8); /* space for features */
bb873b53 1963
8cb441c0
ID		 1964 msg->hdr.version = cpu_to_le16(8); /* MOSDOp v8 */
1965 /* front_len is finalized in encode_request_finish() */
986e8989
ID		 1966 msg->front.iov_len = p - msg->front.iov_base;
1967 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
bb873b53
ID		 1968 msg->hdr.data_len = cpu_to_le32(data_len);
1969 /*
1970 * The header "data_off" is a hint to the receiver allowing it
1971 * to align received data into its buffers such that there's no
1972 * need to re-copy it before writing it to disk (direct I/O).
1973 */
1974 msg->hdr.data_off = cpu_to_le16(req->r_data_offset);
26be8808 1975
8cb441c0
ID		 1976 dout("%s req %p msg %p oid %s oid_len %d\n", __func__, req, msg,
1977 req->r_t.target_oid.name, req->r_t.target_oid.name_len);
1978}
1979
1980static void encode_request_finish(struct ceph_msg *msg)
1981{
1982 void *p = msg->front.iov_base;
986e8989 1983 void *const partial_end = p + msg->front.iov_len;
8cb441c0
ID		 1984 void *const end = p + msg->front_alloc_len;
1985
1986 if (CEPH_HAVE_FEATURE(msg->con->peer_features, RESEND_ON_SPLIT)) {
1987 /* luminous OSD -- encode features and be done */
986e8989 1988 p = partial_end;
8cb441c0
ID		 1989 ceph_encode_64(&p, msg->con->peer_features);
1990 } else {
1991 struct {
1992 char spgid[CEPH_ENCODING_START_BLK_LEN +
1993 CEPH_PGID_ENCODING_LEN + 1];
1994 __le32 hash;
1995 __le32 epoch;
1996 __le32 flags;
1997 char reqid[CEPH_ENCODING_START_BLK_LEN +
1998 sizeof(struct ceph_osd_reqid)];
1999 char trace[sizeof(struct ceph_blkin_trace_info)];
2000 __le32 client_inc;
2001 struct ceph_timespec mtime;
2002 } __packed head;
2003 struct ceph_pg pgid;
2004 void *oloc, *oid, *tail;
2005 int oloc_len, oid_len, tail_len;
2006 int len;
2007
2008 /*
2009 * Pre-luminous OSD -- reencode v8 into v4 using @head
2010 * as a temporary buffer. Encode the raw PG; the rest
2011 * is just a matter of moving oloc, oid and tail blobs
2012 * around.
2013 */
2014 memcpy(&head, p, sizeof(head));
2015 p += sizeof(head);
2016
2017 oloc = p;
2018 p += CEPH_ENCODING_START_BLK_LEN;
2019 pgid.pool = ceph_decode_64(&p);
2020 p += 4 + 4; /* preferred, key len */
2021 len = ceph_decode_32(&p);
2022 p += len; /* nspace */
2023 oloc_len = p - oloc;
2024
2025 oid = p;
2026 len = ceph_decode_32(&p);
2027 p += len;
2028 oid_len = p - oid;
2029
2030 tail = p;
986e8989 2031 tail_len = partial_end - p;
8cb441c0
ID		 2032
2033 p = msg->front.iov_base;
2034 ceph_encode_copy(&p, &head.client_inc, sizeof(head.client_inc));
2035 ceph_encode_copy(&p, &head.epoch, sizeof(head.epoch));
2036 ceph_encode_copy(&p, &head.flags, sizeof(head.flags));
2037 ceph_encode_copy(&p, &head.mtime, sizeof(head.mtime));
2038
2039 /* reassert_version */
2040 memset(p, 0, sizeof(struct ceph_eversion));
2041 p += sizeof(struct ceph_eversion);
2042
2043 BUG_ON(p >= oloc);
2044 memmove(p, oloc, oloc_len);
2045 p += oloc_len;
2046
2047 pgid.seed = le32_to_cpu(head.hash);
2048 encode_pgid(&p, &pgid); /* raw pg */
2049
2050 BUG_ON(p >= oid);
2051 memmove(p, oid, oid_len);
2052 p += oid_len;
2053
2054 /* tail -- ops, snapid, snapc, retry_attempt */
2055 BUG_ON(p >= tail);
2056 memmove(p, tail, tail_len);
2057 p += tail_len;
2058
2059 msg->hdr.version = cpu_to_le16(4); /* MOSDOp v4 */
2060 }
2061
2062 BUG_ON(p > end);
2063 msg->front.iov_len = p - msg->front.iov_base;
2064 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2065
2066 dout("%s msg %p tid %llu %u+%u+%u v%d\n", __func__, msg,
2067 le64_to_cpu(msg->hdr.tid), le32_to_cpu(msg->hdr.front_len),
2068 le32_to_cpu(msg->hdr.middle_len), le32_to_cpu(msg->hdr.data_len),
2069 le16_to_cpu(msg->hdr.version));
bb873b53
ID		 2070}
2071
2072/*
2073 * @req has to be assigned a tid and registered.
2074 */
2075static void send_request(struct ceph_osd_request *req)
2076{
2077 struct ceph_osd *osd = req->r_osd;
2078
5aea3dcd 2079 verify_osd_locked(osd);
bb873b53
ID		 2080 WARN_ON(osd->o_osd != req->r_t.osd);
2081
a02a946d
ID		 2082 /* backoff? */
2083 if (should_plug_request(req))
2084 return;
2085
5aea3dcd
ID		 2086 /*
2087 * We may have a previously queued request message hanging
2088 * around. Cancel it to avoid corrupting the msgr.
2089 */
2090 if (req->r_sent)
2091 ceph_msg_revoke(req->r_request);
2092
bb873b53
ID		 2093 req->r_flags |= CEPH_OSD_FLAG_KNOWN_REDIR;
2094 if (req->r_attempts)
2095 req->r_flags |= CEPH_OSD_FLAG_RETRY;
2096 else
2097 WARN_ON(req->r_flags & CEPH_OSD_FLAG_RETRY);
2098
8cb441c0 2099 encode_request_partial(req, req->r_request);
bb873b53 2100
04c7d789 2101 dout("%s req %p tid %llu to pgid %llu.%x spgid %llu.%xs%d osd%d e%u flags 0x%x attempt %d\n",
bb873b53 2102 __func__, req, req->r_tid, req->r_t.pgid.pool, req->r_t.pgid.seed,
dc98ff72 2103 req->r_t.spgid.pgid.pool, req->r_t.spgid.pgid.seed,
04c7d789
ID		 2104 req->r_t.spgid.shard, osd->o_osd, req->r_t.epoch, req->r_flags,
2105 req->r_attempts);
bb873b53
ID		 2106
2107 req->r_t.paused = false;
2108 req->r_stamp = jiffies;
2109 req->r_attempts++;
2110
2111 req->r_sent = osd->o_incarnation;
2112 req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
2113 ceph_con_send(&osd->o_con, ceph_msg_get(req->r_request));
f24e9980
SW		 2114}
2115
42c1b124
ID		 2116static void maybe_request_map(struct ceph_osd_client *osdc)
2117{
2118 bool continuous = false;
2119
5aea3dcd 2120 verify_osdc_locked(osdc);
42c1b124
ID		 2121 WARN_ON(!osdc->osdmap->epoch);
2122
b7ec35b3
ID		 2123 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2124 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD) ||
2125 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
42c1b124
ID		 2126 dout("%s osdc %p continuous\n", __func__, osdc);
2127 continuous = true;
2128 } else {
2129 dout("%s osdc %p onetime\n", __func__, osdc);
2130 }
2131
2132 if (ceph_monc_want_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
2133 osdc->osdmap->epoch + 1, continuous))
2134 ceph_monc_renew_subs(&osdc->client->monc);
2135}
2136
a1f4020a 2137static void complete_request(struct ceph_osd_request *req, int err);
4609245e
ID		 2138static void send_map_check(struct ceph_osd_request *req);
2139
5aea3dcd 2140static void __submit_request(struct ceph_osd_request *req, bool wrlocked)
0bbfdfe8 2141{
5aea3dcd
ID		 2142 struct ceph_osd_client *osdc = req->r_osdc;
2143 struct ceph_osd *osd;
4609245e 2144 enum calc_target_result ct_res;
5aea3dcd
ID		 2145 bool need_send = false;
2146 bool promoted = false;
a1f4020a 2147 bool need_abort = false;
0bbfdfe8 2148
b18b9550 2149 WARN_ON(req->r_tid);
5aea3dcd
ID		 2150 dout("%s req %p wrlocked %d\n", __func__, req, wrlocked);
2151
2152again:
7de030d6 2153 ct_res = calc_target(osdc, &req->r_t, NULL, false);
4609245e
ID		 2154 if (ct_res == CALC_TARGET_POOL_DNE && !wrlocked)
2155 goto promote;
2156
5aea3dcd
ID		 2157 osd = lookup_create_osd(osdc, req->r_t.osd, wrlocked);
2158 if (IS_ERR(osd)) {
2159 WARN_ON(PTR_ERR(osd) != -EAGAIN || wrlocked);
2160 goto promote;
0bbfdfe8
ID		 2161 }
2162
58eb7932
JL		 2163 if (osdc->osdmap->epoch < osdc->epoch_barrier) {
2164 dout("req %p epoch %u barrier %u\n", req, osdc->osdmap->epoch,
2165 osdc->epoch_barrier);
2166 req->r_t.paused = true;
2167 maybe_request_map(osdc);
2168 } else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2169 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
5aea3dcd
ID		 2170 dout("req %p pausewr\n", req);
2171 req->r_t.paused = true;
2172 maybe_request_map(osdc);
2173 } else if ((req->r_flags & CEPH_OSD_FLAG_READ) &&
b7ec35b3 2174 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
5aea3dcd
ID		 2175 dout("req %p pauserd\n", req);
2176 req->r_t.paused = true;
2177 maybe_request_map(osdc);
2178 } else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2179 !(req->r_flags & (CEPH_OSD_FLAG_FULL_TRY |
2180 CEPH_OSD_FLAG_FULL_FORCE)) &&
b7ec35b3 2181 (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
5aea3dcd
ID		 2182 pool_full(osdc, req->r_t.base_oloc.pool))) {
2183 dout("req %p full/pool_full\n", req);
2184 pr_warn_ratelimited("FULL or reached pool quota\n");
2185 req->r_t.paused = true;
2186 maybe_request_map(osdc);
a1f4020a
JL		 2187 if (req->r_abort_on_full)
2188 need_abort = true;
5aea3dcd
ID		 2189 } else if (!osd_homeless(osd)) {
2190 need_send = true;
0bbfdfe8 2191 } else {
5aea3dcd 2192 maybe_request_map(osdc);
0bbfdfe8
ID		 2193 }
2194
5aea3dcd
ID		 2195 mutex_lock(&osd->lock);
2196 /*
2197 * Assign the tid atomically with send_request() to protect
2198 * multiple writes to the same object from racing with each
2199 * other, resulting in out of order ops on the OSDs.
2200 */
2201 req->r_tid = atomic64_inc_return(&osdc->last_tid);
2202 link_request(osd, req);
2203 if (need_send)
2204 send_request(req);
a1f4020a
JL		 2205 else if (need_abort)
2206 complete_request(req, -ENOSPC);
5aea3dcd
ID		 2207 mutex_unlock(&osd->lock);
2208
4609245e
ID		 2209 if (ct_res == CALC_TARGET_POOL_DNE)
2210 send_map_check(req);
2211
5aea3dcd
ID		 2212 if (promoted)
2213 downgrade_write(&osdc->lock);
2214 return;
2215
2216promote:
2217 up_read(&osdc->lock);
2218 down_write(&osdc->lock);
2219 wrlocked = true;
2220 promoted = true;
2221 goto again;
2222}
2223
2224static void account_request(struct ceph_osd_request *req)
2225{
54ea0046 2226 WARN_ON(req->r_flags & (CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK));
b18b9550 2227 WARN_ON(!(req->r_flags & (CEPH_OSD_FLAG_READ | CEPH_OSD_FLAG_WRITE)));
5aea3dcd 2228
b18b9550 2229 req->r_flags |= CEPH_OSD_FLAG_ONDISK;
5aea3dcd 2230 atomic_inc(&req->r_osdc->num_requests);
7cc5e38f
ID		 2231
2232 req->r_start_stamp = jiffies;
5aea3dcd
ID		 2233}
2234
2235static void submit_request(struct ceph_osd_request *req, bool wrlocked)
2236{
2237 ceph_osdc_get_request(req);
2238 account_request(req);
2239 __submit_request(req, wrlocked);
2240}
2241
45ee2c1d 2242static void finish_request(struct ceph_osd_request *req)
5aea3dcd
ID		 2243{
2244 struct ceph_osd_client *osdc = req->r_osdc;
5aea3dcd 2245
04c7d789 2246 WARN_ON(lookup_request_mc(&osdc->map_checks, req->r_tid));
5aea3dcd
ID		 2247 dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2248
04c7d789
ID		 2249 if (req->r_osd)
2250 unlink_request(req->r_osd, req);
5aea3dcd
ID		 2251 atomic_dec(&osdc->num_requests);
2252
2253 /*
2254 * If an OSD has failed or returned and a request has been sent
2255 * twice, it's possible to get a reply and end up here while the
2256 * request message is queued for delivery. We will ignore the
2257 * reply, so not a big deal, but better to try and catch it.
2258 */
2259 ceph_msg_revoke(req->r_request);
2260 ceph_msg_revoke_incoming(req->r_reply);
2261}
2262
fe5da05e
ID		 2263static void __complete_request(struct ceph_osd_request *req)
2264{
b18b9550
ID		 2265 if (req->r_callback) {
2266 dout("%s req %p tid %llu cb %pf result %d\n", __func__, req,
2267 req->r_tid, req->r_callback, req->r_result);
fe5da05e 2268 req->r_callback(req);
b18b9550 2269 }
fe5da05e
ID		 2270}
2271
4609245e 2272/*
b18b9550 2273 * This is open-coded in handle_reply().
4609245e
ID		 2274 */
2275static void complete_request(struct ceph_osd_request *req, int err)
2276{
2277 dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2278
2279 req->r_result = err;
45ee2c1d 2280 finish_request(req);
4609245e 2281 __complete_request(req);
b18b9550 2282 complete_all(&req->r_completion);
4609245e
ID		 2283 ceph_osdc_put_request(req);
2284}
2285
2286static void cancel_map_check(struct ceph_osd_request *req)
2287{
2288 struct ceph_osd_client *osdc = req->r_osdc;
2289 struct ceph_osd_request *lookup_req;
2290
2291 verify_osdc_wrlocked(osdc);
2292
2293 lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2294 if (!lookup_req)
2295 return;
2296
2297 WARN_ON(lookup_req != req);
2298 erase_request_mc(&osdc->map_checks, req);
2299 ceph_osdc_put_request(req);
2300}
2301
5aea3dcd
ID		 2302static void cancel_request(struct ceph_osd_request *req)
2303{
2304 dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2305
4609245e 2306 cancel_map_check(req);
45ee2c1d 2307 finish_request(req);
b18b9550 2308 complete_all(&req->r_completion);
c297eb42 2309 ceph_osdc_put_request(req);
5aea3dcd
ID		 2310}
2311
7cc5e38f
ID		 2312static void abort_request(struct ceph_osd_request *req, int err)
2313{
2314 dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2315
2316 cancel_map_check(req);
2317 complete_request(req, err);
2318}
2319
58eb7932
JL		 2320static void update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2321{
2322 if (likely(eb > osdc->epoch_barrier)) {
2323 dout("updating epoch_barrier from %u to %u\n",
2324 osdc->epoch_barrier, eb);
2325 osdc->epoch_barrier = eb;
2326 /* Request map if we're not to the barrier yet */
2327 if (eb > osdc->osdmap->epoch)
2328 maybe_request_map(osdc);
2329 }
2330}
2331
2332void ceph_osdc_update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2333{
2334 down_read(&osdc->lock);
2335 if (unlikely(eb > osdc->epoch_barrier)) {
2336 up_read(&osdc->lock);
2337 down_write(&osdc->lock);
2338 update_epoch_barrier(osdc, eb);
2339 up_write(&osdc->lock);
2340 } else {
2341 up_read(&osdc->lock);
2342 }
2343}
2344EXPORT_SYMBOL(ceph_osdc_update_epoch_barrier);
2345
fc36d0a4
JL		 2346/*
2347 * Drop all pending requests that are stalled waiting on a full condition to
58eb7932
JL		 2348 * clear, and complete them with ENOSPC as the return code. Set the
2349 * osdc->epoch_barrier to the latest map epoch that we've seen if any were
2350 * cancelled.
fc36d0a4
JL		 2351 */
2352static void ceph_osdc_abort_on_full(struct ceph_osd_client *osdc)
2353{
2354 struct rb_node *n;
58eb7932 2355 bool victims = false;
fc36d0a4
JL		 2356
2357 dout("enter abort_on_full\n");
2358
2359 if (!ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) && !have_pool_full(osdc))
2360 goto out;
2361
58eb7932
JL		 2362 /* Scan list and see if there is anything to abort */
2363 for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
2364 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
2365 struct rb_node *m;
2366
2367 m = rb_first(&osd->o_requests);
2368 while (m) {
2369 struct ceph_osd_request *req = rb_entry(m,
2370 struct ceph_osd_request, r_node);
2371 m = rb_next(m);
2372
2373 if (req->r_abort_on_full) {
2374 victims = true;
2375 break;
2376 }
2377 }
2378 if (victims)
2379 break;
2380 }
2381
2382 if (!victims)
2383 goto out;
2384
2385 /*
2386 * Update the barrier to current epoch if it's behind that point,
2387 * since we know we have some calls to be aborted in the tree.
2388 */
2389 update_epoch_barrier(osdc, osdc->osdmap->epoch);
2390
fc36d0a4
JL		 2391 for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
2392 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
2393 struct rb_node *m;
2394
2395 m = rb_first(&osd->o_requests);
2396 while (m) {
2397 struct ceph_osd_request *req = rb_entry(m,
2398 struct ceph_osd_request, r_node);
2399 m = rb_next(m);
2400
2401 if (req->r_abort_on_full &&
2402 (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2403 pool_full(osdc, req->r_t.target_oloc.pool)))
2404 abort_request(req, -ENOSPC);
2405 }
2406 }
2407out:
58eb7932 2408 dout("return abort_on_full barrier=%u\n", osdc->epoch_barrier);
fc36d0a4
JL		 2409}
2410
4609245e
ID		 2411static void check_pool_dne(struct ceph_osd_request *req)
2412{
2413 struct ceph_osd_client *osdc = req->r_osdc;
2414 struct ceph_osdmap *map = osdc->osdmap;
2415
2416 verify_osdc_wrlocked(osdc);
2417 WARN_ON(!map->epoch);
2418
2419 if (req->r_attempts) {
2420 /*
2421 * We sent a request earlier, which means that
2422 * previously the pool existed, and now it does not
2423 * (i.e., it was deleted).
2424 */
2425 req->r_map_dne_bound = map->epoch;
2426 dout("%s req %p tid %llu pool disappeared\n", __func__, req,
2427 req->r_tid);
2428 } else {
2429 dout("%s req %p tid %llu map_dne_bound %u have %u\n", __func__,
2430 req, req->r_tid, req->r_map_dne_bound, map->epoch);
2431 }
2432
2433 if (req->r_map_dne_bound) {
2434 if (map->epoch >= req->r_map_dne_bound) {
2435 /* we had a new enough map */
2436 pr_info_ratelimited("tid %llu pool does not exist\n",
2437 req->r_tid);
2438 complete_request(req, -ENOENT);
2439 }
2440 } else {
2441 send_map_check(req);
2442 }
2443}
2444
2445static void map_check_cb(struct ceph_mon_generic_request *greq)
2446{
2447 struct ceph_osd_client *osdc = &greq->monc->client->osdc;
2448 struct ceph_osd_request *req;
2449 u64 tid = greq->private_data;
2450
2451 WARN_ON(greq->result || !greq->u.newest);
2452
2453 down_write(&osdc->lock);
2454 req = lookup_request_mc(&osdc->map_checks, tid);
2455 if (!req) {
2456 dout("%s tid %llu dne\n", __func__, tid);
2457 goto out_unlock;
2458 }
2459
2460 dout("%s req %p tid %llu map_dne_bound %u newest %llu\n", __func__,
2461 req, req->r_tid, req->r_map_dne_bound, greq->u.newest);
2462 if (!req->r_map_dne_bound)
2463 req->r_map_dne_bound = greq->u.newest;
2464 erase_request_mc(&osdc->map_checks, req);
2465 check_pool_dne(req);
2466
2467 ceph_osdc_put_request(req);
2468out_unlock:
2469 up_write(&osdc->lock);
2470}
2471
2472static void send_map_check(struct ceph_osd_request *req)
2473{
2474 struct ceph_osd_client *osdc = req->r_osdc;
2475 struct ceph_osd_request *lookup_req;
2476 int ret;
2477
2478 verify_osdc_wrlocked(osdc);
2479
2480 lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2481 if (lookup_req) {
2482 WARN_ON(lookup_req != req);
2483 return;
2484 }
2485
2486 ceph_osdc_get_request(req);
2487 insert_request_mc(&osdc->map_checks, req);
2488 ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
2489 map_check_cb, req->r_tid);
2490 WARN_ON(ret);
2491}
2492
922dab61
ID		 2493/*
2494 * lingering requests, watch/notify v2 infrastructure
2495 */
2496static void linger_release(struct kref *kref)
2497{
2498 struct ceph_osd_linger_request *lreq =
2499 container_of(kref, struct ceph_osd_linger_request, kref);
2500
2501 dout("%s lreq %p reg_req %p ping_req %p\n", __func__, lreq,
2502 lreq->reg_req, lreq->ping_req);
2503 WARN_ON(!RB_EMPTY_NODE(&lreq->node));
2504 WARN_ON(!RB_EMPTY_NODE(&lreq->osdc_node));
4609245e 2505 WARN_ON(!RB_EMPTY_NODE(&lreq->mc_node));
922dab61 2506 WARN_ON(!list_empty(&lreq->scan_item));
b07d3c4b 2507 WARN_ON(!list_empty(&lreq->pending_lworks));
922dab61
ID		 2508 WARN_ON(lreq->osd);
2509
2510 if (lreq->reg_req)
2511 ceph_osdc_put_request(lreq->reg_req);
2512 if (lreq->ping_req)
2513 ceph_osdc_put_request(lreq->ping_req);
2514 target_destroy(&lreq->t);
2515 kfree(lreq);
2516}
2517
2518static void linger_put(struct ceph_osd_linger_request *lreq)
2519{
2520 if (lreq)
2521 kref_put(&lreq->kref, linger_release);
2522}
2523
2524static struct ceph_osd_linger_request *
2525linger_get(struct ceph_osd_linger_request *lreq)
2526{
2527 kref_get(&lreq->kref);
2528 return lreq;
2529}
2530
2531static struct ceph_osd_linger_request *
2532linger_alloc(struct ceph_osd_client *osdc)
2533{
2534 struct ceph_osd_linger_request *lreq;
2535
2536 lreq = kzalloc(sizeof(*lreq), GFP_NOIO);
2537 if (!lreq)
2538 return NULL;
2539
2540 kref_init(&lreq->kref);
2541 mutex_init(&lreq->lock);
2542 RB_CLEAR_NODE(&lreq->node);
2543 RB_CLEAR_NODE(&lreq->osdc_node);
4609245e 2544 RB_CLEAR_NODE(&lreq->mc_node);
922dab61 2545 INIT_LIST_HEAD(&lreq->scan_item);
b07d3c4b 2546 INIT_LIST_HEAD(&lreq->pending_lworks);
922dab61 2547 init_completion(&lreq->reg_commit_wait);
19079203 2548 init_completion(&lreq->notify_finish_wait);
922dab61
ID		 2549
2550 lreq->osdc = osdc;
2551 target_init(&lreq->t);
2552
2553 dout("%s lreq %p\n", __func__, lreq);
2554 return lreq;
2555}
2556
2557DEFINE_RB_INSDEL_FUNCS(linger, struct ceph_osd_linger_request, linger_id, node)
2558DEFINE_RB_FUNCS(linger_osdc, struct ceph_osd_linger_request, linger_id, osdc_node)
4609245e 2559DEFINE_RB_FUNCS(linger_mc, struct ceph_osd_linger_request, linger_id, mc_node)
922dab61
ID		 2560
2561/*
2562 * Create linger request <-> OSD session relation.
2563 *
2564 * @lreq has to be registered, @osd may be homeless.
2565 */
2566static void link_linger(struct ceph_osd *osd,
2567 struct ceph_osd_linger_request *lreq)
2568{
2569 verify_osd_locked(osd);
2570 WARN_ON(!lreq->linger_id || lreq->osd);
2571 dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2572 osd->o_osd, lreq, lreq->linger_id);
2573
2574 if (!osd_homeless(osd))
2575 __remove_osd_from_lru(osd);
2576 else
2577 atomic_inc(&osd->o_osdc->num_homeless);
2578
2579 get_osd(osd);
2580 insert_linger(&osd->o_linger_requests, lreq);
2581 lreq->osd = osd;
2582}
2583
2584static void unlink_linger(struct ceph_osd *osd,
2585 struct ceph_osd_linger_request *lreq)
2586{
2587 verify_osd_locked(osd);
2588 WARN_ON(lreq->osd != osd);
2589 dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2590 osd->o_osd, lreq, lreq->linger_id);
2591
2592 lreq->osd = NULL;
2593 erase_linger(&osd->o_linger_requests, lreq);
2594 put_osd(osd);
2595
2596 if (!osd_homeless(osd))
2597 maybe_move_osd_to_lru(osd);
2598 else
2599 atomic_dec(&osd->o_osdc->num_homeless);
2600}
2601
2602static bool __linger_registered(struct ceph_osd_linger_request *lreq)
2603{
2604 verify_osdc_locked(lreq->osdc);
2605
2606 return !RB_EMPTY_NODE(&lreq->osdc_node);
2607}
2608
2609static bool linger_registered(struct ceph_osd_linger_request *lreq)
2610{
2611 struct ceph_osd_client *osdc = lreq->osdc;
2612 bool registered;
2613
2614 down_read(&osdc->lock);
2615 registered = __linger_registered(lreq);
2616 up_read(&osdc->lock);
2617
2618 return registered;
2619}
2620
2621static void linger_register(struct ceph_osd_linger_request *lreq)
2622{
2623 struct ceph_osd_client *osdc = lreq->osdc;
2624
2625 verify_osdc_wrlocked(osdc);
2626 WARN_ON(lreq->linger_id);
2627
2628 linger_get(lreq);
2629 lreq->linger_id = ++osdc->last_linger_id;
2630 insert_linger_osdc(&osdc->linger_requests, lreq);
2631}
2632
2633static void linger_unregister(struct ceph_osd_linger_request *lreq)
2634{
2635 struct ceph_osd_client *osdc = lreq->osdc;
2636
2637 verify_osdc_wrlocked(osdc);
2638
2639 erase_linger_osdc(&osdc->linger_requests, lreq);
2640 linger_put(lreq);
2641}
2642
2643static void cancel_linger_request(struct ceph_osd_request *req)
2644{
2645 struct ceph_osd_linger_request *lreq = req->r_priv;
2646
2647 WARN_ON(!req->r_linger);
2648 cancel_request(req);
2649 linger_put(lreq);
2650}
2651
2652struct linger_work {
2653 struct work_struct work;
2654 struct ceph_osd_linger_request *lreq;
b07d3c4b
ID		 2655 struct list_head pending_item;
2656 unsigned long queued_stamp;
922dab61
ID		 2657
2658 union {
2659 struct {
2660 u64 notify_id;
2661 u64 notifier_id;
2662 void *payload; /* points into @msg front */
2663 size_t payload_len;
2664
2665 struct ceph_msg *msg; /* for ceph_msg_put() */
2666 } notify;
2667 struct {
2668 int err;
2669 } error;
2670 };
2671};
2672
2673static struct linger_work *lwork_alloc(struct ceph_osd_linger_request *lreq,
2674 work_func_t workfn)
2675{
2676 struct linger_work *lwork;
2677
2678 lwork = kzalloc(sizeof(*lwork), GFP_NOIO);
2679 if (!lwork)
2680 return NULL;
2681
2682 INIT_WORK(&lwork->work, workfn);
b07d3c4b 2683 INIT_LIST_HEAD(&lwork->pending_item);
922dab61
ID		 2684 lwork->lreq = linger_get(lreq);
2685
2686 return lwork;
2687}
2688
2689static void lwork_free(struct linger_work *lwork)
2690{
2691 struct ceph_osd_linger_request *lreq = lwork->lreq;
2692
b07d3c4b
ID		 2693 mutex_lock(&lreq->lock);
2694 list_del(&lwork->pending_item);
2695 mutex_unlock(&lreq->lock);
2696
922dab61
ID		 2697 linger_put(lreq);
2698 kfree(lwork);
2699}
2700
2701static void lwork_queue(struct linger_work *lwork)
2702{
2703 struct ceph_osd_linger_request *lreq = lwork->lreq;
2704 struct ceph_osd_client *osdc = lreq->osdc;
2705
2706 verify_lreq_locked(lreq);
b07d3c4b
ID		 2707 WARN_ON(!list_empty(&lwork->pending_item));
2708
2709 lwork->queued_stamp = jiffies;
2710 list_add_tail(&lwork->pending_item, &lreq->pending_lworks);
922dab61
ID		 2711 queue_work(osdc->notify_wq, &lwork->work);
2712}
2713
2714static void do_watch_notify(struct work_struct *w)
2715{
2716 struct linger_work *lwork = container_of(w, struct linger_work, work);
2717 struct ceph_osd_linger_request *lreq = lwork->lreq;
2718
2719 if (!linger_registered(lreq)) {
2720 dout("%s lreq %p not registered\n", __func__, lreq);
2721 goto out;
2722 }
2723
19079203 2724 WARN_ON(!lreq->is_watch);
922dab61
ID		 2725 dout("%s lreq %p notify_id %llu notifier_id %llu payload_len %zu\n",
2726 __func__, lreq, lwork->notify.notify_id, lwork->notify.notifier_id,
2727 lwork->notify.payload_len);
2728 lreq->wcb(lreq->data, lwork->notify.notify_id, lreq->linger_id,
2729 lwork->notify.notifier_id, lwork->notify.payload,
2730 lwork->notify.payload_len);
2731
2732out:
2733 ceph_msg_put(lwork->notify.msg);
2734 lwork_free(lwork);
2735}
2736
2737static void do_watch_error(struct work_struct *w)
2738{
2739 struct linger_work *lwork = container_of(w, struct linger_work, work);
2740 struct ceph_osd_linger_request *lreq = lwork->lreq;
2741
2742 if (!linger_registered(lreq)) {
2743 dout("%s lreq %p not registered\n", __func__, lreq);
2744 goto out;
2745 }
2746
2747 dout("%s lreq %p err %d\n", __func__, lreq, lwork->error.err);
2748 lreq->errcb(lreq->data, lreq->linger_id, lwork->error.err);
2749
2750out:
2751 lwork_free(lwork);
2752}
2753
2754static void queue_watch_error(struct ceph_osd_linger_request *lreq)
2755{
2756 struct linger_work *lwork;
2757
2758 lwork = lwork_alloc(lreq, do_watch_error);
2759 if (!lwork) {
2760 pr_err("failed to allocate error-lwork\n");
2761 return;
2762 }
2763
2764 lwork->error.err = lreq->last_error;
2765 lwork_queue(lwork);
2766}
2767
2768static void linger_reg_commit_complete(struct ceph_osd_linger_request *lreq,
2769 int result)
2770{
2771 if (!completion_done(&lreq->reg_commit_wait)) {
2772 lreq->reg_commit_error = (result <= 0 ? result : 0);
2773 complete_all(&lreq->reg_commit_wait);
2774 }
2775}
2776
2777static void linger_commit_cb(struct ceph_osd_request *req)
2778{
2779 struct ceph_osd_linger_request *lreq = req->r_priv;
2780
2781 mutex_lock(&lreq->lock);
2782 dout("%s lreq %p linger_id %llu result %d\n", __func__, lreq,
2783 lreq->linger_id, req->r_result);
922dab61
ID		 2784 linger_reg_commit_complete(lreq, req->r_result);
2785 lreq->committed = true;
2786
19079203
ID		 2787 if (!lreq->is_watch) {
2788 struct ceph_osd_data *osd_data =
2789 osd_req_op_data(req, 0, notify, response_data);
2790 void *p = page_address(osd_data->pages[0]);
2791
2792 WARN_ON(req->r_ops[0].op != CEPH_OSD_OP_NOTIFY ||
2793 osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
2794
2795 /* make note of the notify_id */
2796 if (req->r_ops[0].outdata_len >= sizeof(u64)) {
2797 lreq->notify_id = ceph_decode_64(&p);
2798 dout("lreq %p notify_id %llu\n", lreq,
2799 lreq->notify_id);
2800 } else {
2801 dout("lreq %p no notify_id\n", lreq);
2802 }
2803 }
2804
922dab61
ID		 2805 mutex_unlock(&lreq->lock);
2806 linger_put(lreq);
2807}
2808
2809static int normalize_watch_error(int err)
2810{
2811 /*
2812 * Translate ENOENT -> ENOTCONN so that a delete->disconnection
2813 * notification and a failure to reconnect because we raced with
2814 * the delete appear the same to the user.
2815 */
2816 if (err == -ENOENT)
2817 err = -ENOTCONN;
2818
2819 return err;
2820}
2821
2822static void linger_reconnect_cb(struct ceph_osd_request *req)
2823{
2824 struct ceph_osd_linger_request *lreq = req->r_priv;
2825
2826 mutex_lock(&lreq->lock);
2827 dout("%s lreq %p linger_id %llu result %d last_error %d\n", __func__,
2828 lreq, lreq->linger_id, req->r_result, lreq->last_error);
2829 if (req->r_result < 0) {
2830 if (!lreq->last_error) {
2831 lreq->last_error = normalize_watch_error(req->r_result);
2832 queue_watch_error(lreq);
2833 }
2834 }
2835
2836 mutex_unlock(&lreq->lock);
2837 linger_put(lreq);
2838}
2839
2840static void send_linger(struct ceph_osd_linger_request *lreq)
2841{
2842 struct ceph_osd_request *req = lreq->reg_req;
2843 struct ceph_osd_req_op *op = &req->r_ops[0];
2844
2845 verify_osdc_wrlocked(req->r_osdc);
2846 dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
2847
2848 if (req->r_osd)
2849 cancel_linger_request(req);
2850
2851 request_reinit(req);
2852 ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
2853 ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
2854 req->r_flags = lreq->t.flags;
2855 req->r_mtime = lreq->mtime;
2856
2857 mutex_lock(&lreq->lock);
19079203 2858 if (lreq->is_watch && lreq->committed) {
922dab61
ID		 2859 WARN_ON(op->op != CEPH_OSD_OP_WATCH ||
2860 op->watch.cookie != lreq->linger_id);
2861 op->watch.op = CEPH_OSD_WATCH_OP_RECONNECT;
2862 op->watch.gen = ++lreq->register_gen;
2863 dout("lreq %p reconnect register_gen %u\n", lreq,
2864 op->watch.gen);
2865 req->r_callback = linger_reconnect_cb;
2866 } else {
19079203
ID		 2867 if (!lreq->is_watch)
2868 lreq->notify_id = 0;
2869 else
2870 WARN_ON(op->watch.op != CEPH_OSD_WATCH_OP_WATCH);
922dab61
ID		 2871 dout("lreq %p register\n", lreq);
2872 req->r_callback = linger_commit_cb;
2873 }
2874 mutex_unlock(&lreq->lock);
2875
2876 req->r_priv = linger_get(lreq);
2877 req->r_linger = true;
2878
2879 submit_request(req, true);
2880}
2881
2882static void linger_ping_cb(struct ceph_osd_request *req)
2883{
2884 struct ceph_osd_linger_request *lreq = req->r_priv;
2885
2886 mutex_lock(&lreq->lock);
2887 dout("%s lreq %p linger_id %llu result %d ping_sent %lu last_error %d\n",
2888 __func__, lreq, lreq->linger_id, req->r_result, lreq->ping_sent,
2889 lreq->last_error);
2890 if (lreq->register_gen == req->r_ops[0].watch.gen) {
b07d3c4b
ID		 2891 if (!req->r_result) {
2892 lreq->watch_valid_thru = lreq->ping_sent;
2893 } else if (!lreq->last_error) {
922dab61
ID		 2894 lreq->last_error = normalize_watch_error(req->r_result);
2895 queue_watch_error(lreq);
2896 }
2897 } else {
2898 dout("lreq %p register_gen %u ignoring old pong %u\n", lreq,
2899 lreq->register_gen, req->r_ops[0].watch.gen);
2900 }
2901
2902 mutex_unlock(&lreq->lock);
2903 linger_put(lreq);
2904}
2905
2906static void send_linger_ping(struct ceph_osd_linger_request *lreq)
2907{
2908 struct ceph_osd_client *osdc = lreq->osdc;
2909 struct ceph_osd_request *req = lreq->ping_req;
2910 struct ceph_osd_req_op *op = &req->r_ops[0];
2911
b7ec35b3 2912 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
922dab61
ID		 2913 dout("%s PAUSERD\n", __func__);
2914 return;
2915 }
2916
2917 lreq->ping_sent = jiffies;
2918 dout("%s lreq %p linger_id %llu ping_sent %lu register_gen %u\n",
2919 __func__, lreq, lreq->linger_id, lreq->ping_sent,
2920 lreq->register_gen);
2921
2922 if (req->r_osd)
2923 cancel_linger_request(req);
2924
2925 request_reinit(req);
2926 target_copy(&req->r_t, &lreq->t);
2927
2928 WARN_ON(op->op != CEPH_OSD_OP_WATCH ||
2929 op->watch.cookie != lreq->linger_id ||
2930 op->watch.op != CEPH_OSD_WATCH_OP_PING);
2931 op->watch.gen = lreq->register_gen;
2932 req->r_callback = linger_ping_cb;
2933 req->r_priv = linger_get(lreq);
2934 req->r_linger = true;
2935
2936 ceph_osdc_get_request(req);
2937 account_request(req);
2938 req->r_tid = atomic64_inc_return(&osdc->last_tid);
2939 link_request(lreq->osd, req);
2940 send_request(req);
2941}
2942
2943static void linger_submit(struct ceph_osd_linger_request *lreq)
2944{
2945 struct ceph_osd_client *osdc = lreq->osdc;
2946 struct ceph_osd *osd;
2947
7de030d6 2948 calc_target(osdc, &lreq->t, NULL, false);
922dab61
ID		 2949 osd = lookup_create_osd(osdc, lreq->t.osd, true);
2950 link_linger(osd, lreq);
2951
2952 send_linger(lreq);
2953}
2954
4609245e
ID		 2955static void cancel_linger_map_check(struct ceph_osd_linger_request *lreq)
2956{
2957 struct ceph_osd_client *osdc = lreq->osdc;
2958 struct ceph_osd_linger_request *lookup_lreq;
2959
2960 verify_osdc_wrlocked(osdc);
2961
2962 lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
2963 lreq->linger_id);
2964 if (!lookup_lreq)
2965 return;
2966
2967 WARN_ON(lookup_lreq != lreq);
2968 erase_linger_mc(&osdc->linger_map_checks, lreq);
2969 linger_put(lreq);
2970}
2971
922dab61
ID		 2972/*
2973 * @lreq has to be both registered and linked.
2974 */
2975static void __linger_cancel(struct ceph_osd_linger_request *lreq)
2976{
19079203 2977 if (lreq->is_watch && lreq->ping_req->r_osd)
922dab61
ID		 2978 cancel_linger_request(lreq->ping_req);
2979 if (lreq->reg_req->r_osd)
2980 cancel_linger_request(lreq->reg_req);
4609245e 2981 cancel_linger_map_check(lreq);
922dab61
ID		 2982 unlink_linger(lreq->osd, lreq);
2983 linger_unregister(lreq);
2984}
2985
2986static void linger_cancel(struct ceph_osd_linger_request *lreq)
2987{
2988 struct ceph_osd_client *osdc = lreq->osdc;
2989
2990 down_write(&osdc->lock);
2991 if (__linger_registered(lreq))
2992 __linger_cancel(lreq);
2993 up_write(&osdc->lock);
2994}
2995
4609245e
ID		 2996static void send_linger_map_check(struct ceph_osd_linger_request *lreq);
2997
2998static void check_linger_pool_dne(struct ceph_osd_linger_request *lreq)
2999{
3000 struct ceph_osd_client *osdc = lreq->osdc;
3001 struct ceph_osdmap *map = osdc->osdmap;
3002
3003 verify_osdc_wrlocked(osdc);
3004 WARN_ON(!map->epoch);
3005
3006 if (lreq->register_gen) {
3007 lreq->map_dne_bound = map->epoch;
3008 dout("%s lreq %p linger_id %llu pool disappeared\n", __func__,
3009 lreq, lreq->linger_id);
3010 } else {
3011 dout("%s lreq %p linger_id %llu map_dne_bound %u have %u\n",
3012 __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3013 map->epoch);
3014 }
3015
3016 if (lreq->map_dne_bound) {
3017 if (map->epoch >= lreq->map_dne_bound) {
3018 /* we had a new enough map */
3019 pr_info("linger_id %llu pool does not exist\n",
3020 lreq->linger_id);
3021 linger_reg_commit_complete(lreq, -ENOENT);
3022 __linger_cancel(lreq);
3023 }
3024 } else {
3025 send_linger_map_check(lreq);
3026 }
3027}
3028
3029static void linger_map_check_cb(struct ceph_mon_generic_request *greq)
3030{
3031 struct ceph_osd_client *osdc = &greq->monc->client->osdc;
3032 struct ceph_osd_linger_request *lreq;
3033 u64 linger_id = greq->private_data;
3034
3035 WARN_ON(greq->result || !greq->u.newest);
3036
3037 down_write(&osdc->lock);
3038 lreq = lookup_linger_mc(&osdc->linger_map_checks, linger_id);
3039 if (!lreq) {
3040 dout("%s linger_id %llu dne\n", __func__, linger_id);
3041 goto out_unlock;
3042 }
3043
3044 dout("%s lreq %p linger_id %llu map_dne_bound %u newest %llu\n",
3045 __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3046 greq->u.newest);
3047 if (!lreq->map_dne_bound)
3048 lreq->map_dne_bound = greq->u.newest;
3049 erase_linger_mc(&osdc->linger_map_checks, lreq);
3050 check_linger_pool_dne(lreq);
3051
3052 linger_put(lreq);
3053out_unlock:
3054 up_write(&osdc->lock);
3055}
3056
3057static void send_linger_map_check(struct ceph_osd_linger_request *lreq)
3058{
3059 struct ceph_osd_client *osdc = lreq->osdc;
3060 struct ceph_osd_linger_request *lookup_lreq;
3061 int ret;
3062
3063 verify_osdc_wrlocked(osdc);
3064
3065 lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3066 lreq->linger_id);
3067 if (lookup_lreq) {
3068 WARN_ON(lookup_lreq != lreq);
3069 return;
3070 }
3071
3072 linger_get(lreq);
3073 insert_linger_mc(&osdc->linger_map_checks, lreq);
3074 ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
3075 linger_map_check_cb, lreq->linger_id);
3076 WARN_ON(ret);
3077}
3078
922dab61
ID		 3079static int linger_reg_commit_wait(struct ceph_osd_linger_request *lreq)
3080{
3081 int ret;
3082
3083 dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3084 ret = wait_for_completion_interruptible(&lreq->reg_commit_wait);
3085 return ret ?: lreq->reg_commit_error;
3086}
3087
19079203
ID		 3088static int linger_notify_finish_wait(struct ceph_osd_linger_request *lreq)
3089{
3090 int ret;
3091
3092 dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3093 ret = wait_for_completion_interruptible(&lreq->notify_finish_wait);
3094 return ret ?: lreq->notify_finish_error;
3095}
3096
f24e9980 3097/*
fbca9635
ID		 3098 * Timeout callback, called every N seconds. When 1 or more OSD
3099 * requests has been active for more than N seconds, we send a keepalive
3100 * (tag + timestamp) to its OSD to ensure any communications channel
3101 * reset is detected.
f24e9980
SW		 3102 */
3103static void handle_timeout(struct work_struct *work)
3104{
3105 struct ceph_osd_client *osdc =
3106 container_of(work, struct ceph_osd_client, timeout_work.work);
a319bf56 3107 struct ceph_options *opts = osdc->client->options;
5aea3dcd 3108 unsigned long cutoff = jiffies - opts->osd_keepalive_timeout;
7cc5e38f 3109 unsigned long expiry_cutoff = jiffies - opts->osd_request_timeout;
5aea3dcd
ID		 3110 LIST_HEAD(slow_osds);
3111 struct rb_node *n, *p;
f24e9980 3112
5aea3dcd
ID		 3113 dout("%s osdc %p\n", __func__, osdc);
3114 down_write(&osdc->lock);
f24e9980 3115
422d2cb8
YS		 3116 /*
3117 * ping osds that are a bit slow. this ensures that if there
3118 * is a break in the TCP connection we will notice, and reopen
3119 * a connection with that osd (from the fault callback).
3120 */
5aea3dcd
ID		 3121 for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
3122 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
3123 bool found = false;
3124
7cc5e38f 3125 for (p = rb_first(&osd->o_requests); p; ) {
5aea3dcd
ID		 3126 struct ceph_osd_request *req =
3127 rb_entry(p, struct ceph_osd_request, r_node);
3128
7cc5e38f
ID		 3129 p = rb_next(p); /* abort_request() */
3130
5aea3dcd
ID		 3131 if (time_before(req->r_stamp, cutoff)) {
3132 dout(" req %p tid %llu on osd%d is laggy\n",
3133 req, req->r_tid, osd->o_osd);
3134 found = true;
3135 }
7cc5e38f
ID		 3136 if (opts->osd_request_timeout &&
3137 time_before(req->r_start_stamp, expiry_cutoff)) {
3138 pr_err_ratelimited("tid %llu on osd%d timeout\n",
3139 req->r_tid, osd->o_osd);
3140 abort_request(req, -ETIMEDOUT);
3141 }
5aea3dcd 3142 }
922dab61
ID		 3143 for (p = rb_first(&osd->o_linger_requests); p; p = rb_next(p)) {
3144 struct ceph_osd_linger_request *lreq =
3145 rb_entry(p, struct ceph_osd_linger_request, node);
3146
3147 dout(" lreq %p linger_id %llu is served by osd%d\n",
3148 lreq, lreq->linger_id, osd->o_osd);
3149 found = true;
3150
3151 mutex_lock(&lreq->lock);
19079203 3152 if (lreq->is_watch && lreq->committed && !lreq->last_error)
922dab61
ID		 3153 send_linger_ping(lreq);
3154 mutex_unlock(&lreq->lock);
3155 }
422d2cb8 3156
5aea3dcd
ID		 3157 if (found)
3158 list_move_tail(&osd->o_keepalive_item, &slow_osds);
422d2cb8 3159 }
5aea3dcd 3160
7cc5e38f
ID		 3161 if (opts->osd_request_timeout) {
3162 for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
3163 struct ceph_osd_request *req =
3164 rb_entry(p, struct ceph_osd_request, r_node);
3165
3166 p = rb_next(p); /* abort_request() */
3167
3168 if (time_before(req->r_start_stamp, expiry_cutoff)) {
3169 pr_err_ratelimited("tid %llu on osd%d timeout\n",
3170 req->r_tid, osdc->homeless_osd.o_osd);
3171 abort_request(req, -ETIMEDOUT);
3172 }
3173 }
3174 }
3175
5aea3dcd
ID		 3176 if (atomic_read(&osdc->num_homeless) || !list_empty(&slow_osds))
3177 maybe_request_map(osdc);
3178
422d2cb8 3179 while (!list_empty(&slow_osds)) {
5aea3dcd
ID		 3180 struct ceph_osd *osd = list_first_entry(&slow_osds,
3181 struct ceph_osd,
3182 o_keepalive_item);
422d2cb8 3183 list_del_init(&osd->o_keepalive_item);
f24e9980
SW		 3184 ceph_con_keepalive(&osd->o_con);
3185 }
3186
5aea3dcd 3187 up_write(&osdc->lock);
fbca9635
ID		 3188 schedule_delayed_work(&osdc->timeout_work,
3189 osdc->client->options->osd_keepalive_timeout);
f24e9980
SW		 3190}
3191
f5a2041b
YS		 3192static void handle_osds_timeout(struct work_struct *work)
3193{
3194 struct ceph_osd_client *osdc =
3195 container_of(work, struct ceph_osd_client,
3196 osds_timeout_work.work);
a319bf56 3197 unsigned long delay = osdc->client->options->osd_idle_ttl / 4;
42a2c09f 3198 struct ceph_osd *osd, *nosd;
f5a2041b 3199
42a2c09f 3200 dout("%s osdc %p\n", __func__, osdc);
5aea3dcd 3201 down_write(&osdc->lock);
42a2c09f
ID		 3202 list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) {
3203 if (time_before(jiffies, osd->lru_ttl))
3204 break;
3205
5aea3dcd 3206 WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
922dab61 3207 WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
5aea3dcd 3208 close_osd(osd);
42a2c09f 3209 }
f5a2041b 3210
5aea3dcd 3211 up_write(&osdc->lock);
f5a2041b
YS		 3212 schedule_delayed_work(&osdc->osds_timeout_work,
3213 round_jiffies_relative(delay));
3214}
3215
205ee118
ID		 3216static int ceph_oloc_decode(void **p, void *end,
3217 struct ceph_object_locator *oloc)
3218{
3219 u8 struct_v, struct_cv;
3220 u32 len;
3221 void *struct_end;
3222 int ret = 0;
3223
3224 ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3225 struct_v = ceph_decode_8(p);
3226 struct_cv = ceph_decode_8(p);
3227 if (struct_v < 3) {
3228 pr_warn("got v %d < 3 cv %d of ceph_object_locator\n",
3229 struct_v, struct_cv);
3230 goto e_inval;
3231 }
3232 if (struct_cv > 6) {
3233 pr_warn("got v %d cv %d > 6 of ceph_object_locator\n",
3234 struct_v, struct_cv);
3235 goto e_inval;
3236 }
3237 len = ceph_decode_32(p);
3238 ceph_decode_need(p, end, len, e_inval);
3239 struct_end = *p + len;
3240
3241 oloc->pool = ceph_decode_64(p);
3242 *p += 4; /* skip preferred */
3243
3244 len = ceph_decode_32(p);
3245 if (len > 0) {
3246 pr_warn("ceph_object_locator::key is set\n");
3247 goto e_inval;
3248 }
3249
3250 if (struct_v >= 5) {
cd08e0a2
YZ		 3251 bool changed = false;
3252
205ee118
ID		 3253 len = ceph_decode_32(p);
3254 if (len > 0) {
30c156d9 3255 ceph_decode_need(p, end, len, e_inval);
cd08e0a2
YZ		 3256 if (!oloc->pool_ns ||
3257 ceph_compare_string(oloc->pool_ns, *p, len))
3258 changed = true;
30c156d9 3259 *p += len;
cd08e0a2
YZ		 3260 } else {
3261 if (oloc->pool_ns)
3262 changed = true;
3263 }
3264 if (changed) {
3265 /* redirect changes namespace */
3266 pr_warn("ceph_object_locator::nspace is changed\n");
3267 goto e_inval;
205ee118
ID		 3268 }
3269 }
3270
3271 if (struct_v >= 6) {
3272 s64 hash = ceph_decode_64(p);
3273 if (hash != -1) {
3274 pr_warn("ceph_object_locator::hash is set\n");
3275 goto e_inval;
3276 }
3277 }
3278
3279 /* skip the rest */
3280 *p = struct_end;
3281out:
3282 return ret;
3283
3284e_inval:
3285 ret = -EINVAL;
3286 goto out;
3287}
3288
3289static int ceph_redirect_decode(void **p, void *end,
3290 struct ceph_request_redirect *redir)
3291{
3292 u8 struct_v, struct_cv;
3293 u32 len;
3294 void *struct_end;
3295 int ret;
3296
3297 ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3298 struct_v = ceph_decode_8(p);
3299 struct_cv = ceph_decode_8(p);
3300 if (struct_cv > 1) {
3301 pr_warn("got v %d cv %d > 1 of ceph_request_redirect\n",
3302 struct_v, struct_cv);
3303 goto e_inval;
3304 }
3305 len = ceph_decode_32(p);
3306 ceph_decode_need(p, end, len, e_inval);
3307 struct_end = *p + len;
3308
3309 ret = ceph_oloc_decode(p, end, &redir->oloc);
3310 if (ret)
3311 goto out;
3312
3313 len = ceph_decode_32(p);
3314 if (len > 0) {
3315 pr_warn("ceph_request_redirect::object_name is set\n");
3316 goto e_inval;
3317 }
3318
3319 len = ceph_decode_32(p);
3320 *p += len; /* skip osd_instructions */
3321
3322 /* skip the rest */
3323 *p = struct_end;
3324out:
3325 return ret;
3326
3327e_inval:
3328 ret = -EINVAL;
3329 goto out;
3330}
3331
fe5da05e
ID		 3332struct MOSDOpReply {
3333 struct ceph_pg pgid;
3334 u64 flags;
3335 int result;
3336 u32 epoch;
3337 int num_ops;
3338 u32 outdata_len[CEPH_OSD_MAX_OPS];
3339 s32 rval[CEPH_OSD_MAX_OPS];
3340 int retry_attempt;
3341 struct ceph_eversion replay_version;
3342 u64 user_version;
3343 struct ceph_request_redirect redirect;
3344};
25845472 3345
fe5da05e 3346static int decode_MOSDOpReply(const struct ceph_msg *msg, struct MOSDOpReply *m)
f24e9980 3347{
fe5da05e
ID		 3348 void *p = msg->front.iov_base;
3349 void *const end = p + msg->front.iov_len;
3350 u16 version = le16_to_cpu(msg->hdr.version);
3351 struct ceph_eversion bad_replay_version;
b0b31a8f 3352 u8 decode_redir;
fe5da05e
ID		 3353 u32 len;
3354 int ret;
3355 int i;
1b83bef2 3356
fe5da05e
ID		 3357 ceph_decode_32_safe(&p, end, len, e_inval);
3358 ceph_decode_need(&p, end, len, e_inval);
3359 p += len; /* skip oid */
1b83bef2 3360
fe5da05e
ID		 3361 ret = ceph_decode_pgid(&p, end, &m->pgid);
3362 if (ret)
3363 return ret;
1b83bef2 3364
fe5da05e
ID		 3365 ceph_decode_64_safe(&p, end, m->flags, e_inval);
3366 ceph_decode_32_safe(&p, end, m->result, e_inval);
3367 ceph_decode_need(&p, end, sizeof(bad_replay_version), e_inval);
3368 memcpy(&bad_replay_version, p, sizeof(bad_replay_version));
3369 p += sizeof(bad_replay_version);
3370 ceph_decode_32_safe(&p, end, m->epoch, e_inval);
1b83bef2 3371
fe5da05e
ID		 3372 ceph_decode_32_safe(&p, end, m->num_ops, e_inval);
3373 if (m->num_ops > ARRAY_SIZE(m->outdata_len))
3374 goto e_inval;
1b83bef2 3375
fe5da05e
ID		 3376 ceph_decode_need(&p, end, m->num_ops * sizeof(struct ceph_osd_op),
3377 e_inval);
3378 for (i = 0; i < m->num_ops; i++) {
1b83bef2 3379 struct ceph_osd_op *op = p;
1b83bef2 3380
fe5da05e 3381 m->outdata_len[i] = le32_to_cpu(op->payload_len);
1b83bef2
SW		 3382 p += sizeof(*op);
3383 }
1b83bef2 3384
fe5da05e
ID		 3385 ceph_decode_32_safe(&p, end, m->retry_attempt, e_inval);
3386 for (i = 0; i < m->num_ops; i++)
3387 ceph_decode_32_safe(&p, end, m->rval[i], e_inval);
f24e9980 3388
fe5da05e
ID		 3389 if (version >= 5) {
3390 ceph_decode_need(&p, end, sizeof(m->replay_version), e_inval);
3391 memcpy(&m->replay_version, p, sizeof(m->replay_version));
3392 p += sizeof(m->replay_version);
3393 ceph_decode_64_safe(&p, end, m->user_version, e_inval);
3394 } else {
3395 m->replay_version = bad_replay_version; /* struct */
3396 m->user_version = le64_to_cpu(m->replay_version.version);
3397 }
eb845ff1 3398
fe5da05e
ID		 3399 if (version >= 6) {
3400 if (version >= 7)
3401 ceph_decode_8_safe(&p, end, decode_redir, e_inval);
b0b31a8f
ID		 3402 else
3403 decode_redir = 1;
3404 } else {
3405 decode_redir = 0;
3406 }
3407
3408 if (decode_redir) {
fe5da05e
ID		 3409 ret = ceph_redirect_decode(&p, end, &m->redirect);
3410 if (ret)
3411 return ret;
205ee118 3412 } else {
fe5da05e 3413 ceph_oloc_init(&m->redirect.oloc);
205ee118 3414 }
f24e9980 3415
fe5da05e
ID		 3416 return 0;
3417
3418e_inval:
3419 return -EINVAL;
3420}
3421
3422/*
b18b9550
ID		 3423 * Handle MOSDOpReply. Set ->r_result and call the callback if it is
3424 * specified.
fe5da05e 3425 */
5aea3dcd 3426static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg)
fe5da05e 3427{
5aea3dcd 3428 struct ceph_osd_client *osdc = osd->o_osdc;
fe5da05e
ID		 3429 struct ceph_osd_request *req;
3430 struct MOSDOpReply m;
3431 u64 tid = le64_to_cpu(msg->hdr.tid);
3432 u32 data_len = 0;
fe5da05e
ID		 3433 int ret;
3434 int i;
3435
3436 dout("%s msg %p tid %llu\n", __func__, msg, tid);
3437
5aea3dcd
ID		 3438 down_read(&osdc->lock);
3439 if (!osd_registered(osd)) {
3440 dout("%s osd%d unknown\n", __func__, osd->o_osd);
3441 goto out_unlock_osdc;
3442 }
3443 WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
3444
3445 mutex_lock(&osd->lock);
3446 req = lookup_request(&osd->o_requests, tid);
fe5da05e 3447 if (!req) {
5aea3dcd
ID		 3448 dout("%s osd%d tid %llu unknown\n", __func__, osd->o_osd, tid);
3449 goto out_unlock_session;
fe5da05e 3450 }
fe5da05e 3451
cd08e0a2 3452 m.redirect.oloc.pool_ns = req->r_t.target_oloc.pool_ns;
fe5da05e 3453 ret = decode_MOSDOpReply(msg, &m);
cd08e0a2 3454 m.redirect.oloc.pool_ns = NULL;
fe5da05e
ID		 3455 if (ret) {
3456 pr_err("failed to decode MOSDOpReply for tid %llu: %d\n",
3457 req->r_tid, ret);
3458 ceph_msg_dump(msg);
3459 goto fail_request;
3460 }
3461 dout("%s req %p tid %llu flags 0x%llx pgid %llu.%x epoch %u attempt %d v %u'%llu uv %llu\n",
3462 __func__, req, req->r_tid, m.flags, m.pgid.pool, m.pgid.seed,
3463 m.epoch, m.retry_attempt, le32_to_cpu(m.replay_version.epoch),
3464 le64_to_cpu(m.replay_version.version), m.user_version);
3465
3466 if (m.retry_attempt >= 0) {
3467 if (m.retry_attempt != req->r_attempts - 1) {
3468 dout("req %p tid %llu retry_attempt %d != %d, ignoring\n",
3469 req, req->r_tid, m.retry_attempt,
3470 req->r_attempts - 1);
5aea3dcd 3471 goto out_unlock_session;
fe5da05e
ID		 3472 }
3473 } else {
3474 WARN_ON(1); /* MOSDOpReply v4 is assumed */
3475 }
3476
3477 if (!ceph_oloc_empty(&m.redirect.oloc)) {
3478 dout("req %p tid %llu redirect pool %lld\n", req, req->r_tid,
3479 m.redirect.oloc.pool);
5aea3dcd
ID		 3480 unlink_request(osd, req);
3481 mutex_unlock(&osd->lock);
205ee118 3482
30c156d9
YZ		 3483 /*
3484 * Not ceph_oloc_copy() - changing pool_ns is not
3485 * supported.
3486 */
3487 req->r_t.target_oloc.pool = m.redirect.oloc.pool;
5aea3dcd
ID		 3488 req->r_flags |= CEPH_OSD_FLAG_REDIRECTED;
3489 req->r_tid = 0;
3490 __submit_request(req, false);
3491 goto out_unlock_osdc;
205ee118
ID		 3492 }
3493
fe5da05e
ID		 3494 if (m.num_ops != req->r_num_ops) {
3495 pr_err("num_ops %d != %d for tid %llu\n", m.num_ops,
3496 req->r_num_ops, req->r_tid);
3497 goto fail_request;
f24e9980 3498 }
fe5da05e
ID		 3499 for (i = 0; i < req->r_num_ops; i++) {
3500 dout(" req %p tid %llu op %d rval %d len %u\n", req,
3501 req->r_tid, i, m.rval[i], m.outdata_len[i]);
3502 req->r_ops[i].rval = m.rval[i];
3503 req->r_ops[i].outdata_len = m.outdata_len[i];
3504 data_len += m.outdata_len[i];
3505 }
3506 if (data_len != le32_to_cpu(msg->hdr.data_len)) {
3507 pr_err("sum of lens %u != %u for tid %llu\n", data_len,
3508 le32_to_cpu(msg->hdr.data_len), req->r_tid);
3509 goto fail_request;
3510 }
b18b9550
ID		 3511 dout("%s req %p tid %llu result %d data_len %u\n", __func__,
3512 req, req->r_tid, m.result, data_len);
f24e9980 3513
b18b9550
ID		 3514 /*
3515 * Since we only ever request ONDISK, we should only ever get
3516 * one (type of) reply back.
3517 */
3518 WARN_ON(!(m.flags & CEPH_OSD_FLAG_ONDISK));
3519 req->r_result = m.result ?: data_len;
3520 finish_request(req);
5aea3dcd
ID		 3521 mutex_unlock(&osd->lock);
3522 up_read(&osdc->lock);
f24e9980 3523
b18b9550
ID		 3524 __complete_request(req);
3525 complete_all(&req->r_completion);
3526 ceph_osdc_put_request(req);
f24e9980
SW		 3527 return;
3528
fe5da05e 3529fail_request:
4609245e 3530 complete_request(req, -EIO);
5aea3dcd
ID		 3531out_unlock_session:
3532 mutex_unlock(&osd->lock);
3533out_unlock_osdc:
3534 up_read(&osdc->lock);
f24e9980
SW		 3535}
3536
42c1b124
ID		 3537static void set_pool_was_full(struct ceph_osd_client *osdc)
3538{
3539 struct rb_node *n;
3540
3541 for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
3542 struct ceph_pg_pool_info *pi =
3543 rb_entry(n, struct ceph_pg_pool_info, node);
3544
3545 pi->was_full = __pool_full(pi);
3546 }
3547}
3548
5aea3dcd 3549static bool pool_cleared_full(struct ceph_osd_client *osdc, s64 pool_id)
f24e9980 3550{
5aea3dcd 3551 struct ceph_pg_pool_info *pi;
f24e9980 3552
5aea3dcd
ID		 3553 pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
3554 if (!pi)
3555 return false;
f24e9980 3556
5aea3dcd 3557 return pi->was_full && !__pool_full(pi);
422d2cb8
YS		 3558}
3559
922dab61
ID		 3560static enum calc_target_result
3561recalc_linger_target(struct ceph_osd_linger_request *lreq)
3562{
3563 struct ceph_osd_client *osdc = lreq->osdc;
3564 enum calc_target_result ct_res;
3565
7de030d6 3566 ct_res = calc_target(osdc, &lreq->t, NULL, true);
922dab61
ID		 3567 if (ct_res == CALC_TARGET_NEED_RESEND) {
3568 struct ceph_osd *osd;
3569
3570 osd = lookup_create_osd(osdc, lreq->t.osd, true);
3571 if (osd != lreq->osd) {
3572 unlink_linger(lreq->osd, lreq);
3573 link_linger(osd, lreq);
3574 }
3575 }
3576
3577 return ct_res;
3578}
3579
422d2cb8 3580/*
5aea3dcd 3581 * Requeue requests whose mapping to an OSD has changed.
422d2cb8 3582 */
5aea3dcd
ID		 3583static void scan_requests(struct ceph_osd *osd,
3584 bool force_resend,
3585 bool cleared_full,
3586 bool check_pool_cleared_full,
3587 struct rb_root *need_resend,
3588 struct list_head *need_resend_linger)
422d2cb8 3589{
5aea3dcd
ID		 3590 struct ceph_osd_client *osdc = osd->o_osdc;
3591 struct rb_node *n;
3592 bool force_resend_writes;
3593
922dab61
ID		 3594 for (n = rb_first(&osd->o_linger_requests); n; ) {
3595 struct ceph_osd_linger_request *lreq =
3596 rb_entry(n, struct ceph_osd_linger_request, node);
3597 enum calc_target_result ct_res;
3598
3599 n = rb_next(n); /* recalc_linger_target() */
3600
3601 dout("%s lreq %p linger_id %llu\n", __func__, lreq,
3602 lreq->linger_id);
3603 ct_res = recalc_linger_target(lreq);
3604 switch (ct_res) {
3605 case CALC_TARGET_NO_ACTION:
3606 force_resend_writes = cleared_full ||
3607 (check_pool_cleared_full &&
3608 pool_cleared_full(osdc, lreq->t.base_oloc.pool));
3609 if (!force_resend && !force_resend_writes)
3610 break;
3611
3612 /* fall through */
3613 case CALC_TARGET_NEED_RESEND:
4609245e 3614 cancel_linger_map_check(lreq);
922dab61
ID		 3615 /*
3616 * scan_requests() for the previous epoch(s)
3617 * may have already added it to the list, since
3618 * it's not unlinked here.
3619 */
3620 if (list_empty(&lreq->scan_item))
3621 list_add_tail(&lreq->scan_item, need_resend_linger);
3622 break;
3623 case CALC_TARGET_POOL_DNE:
a10bcb19 3624 list_del_init(&lreq->scan_item);
4609245e 3625 check_linger_pool_dne(lreq);
922dab61
ID		 3626 break;
3627 }
3628 }
3629
5aea3dcd
ID		 3630 for (n = rb_first(&osd->o_requests); n; ) {
3631 struct ceph_osd_request *req =
3632 rb_entry(n, struct ceph_osd_request, r_node);
3633 enum calc_target_result ct_res;
3634
4609245e 3635 n = rb_next(n); /* unlink_request(), check_pool_dne() */
5aea3dcd
ID		 3636
3637 dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
7de030d6
ID		 3638 ct_res = calc_target(osdc, &req->r_t, &req->r_osd->o_con,
3639 false);
5aea3dcd
ID		 3640 switch (ct_res) {
3641 case CALC_TARGET_NO_ACTION:
3642 force_resend_writes = cleared_full ||
3643 (check_pool_cleared_full &&
3644 pool_cleared_full(osdc, req->r_t.base_oloc.pool));
3645 if (!force_resend &&
3646 (!(req->r_flags & CEPH_OSD_FLAG_WRITE) ||
3647 !force_resend_writes))
3648 break;
3649
3650 /* fall through */
3651 case CALC_TARGET_NEED_RESEND:
4609245e 3652 cancel_map_check(req);
5aea3dcd
ID		 3653 unlink_request(osd, req);
3654 insert_request(need_resend, req);
3655 break;
3656 case CALC_TARGET_POOL_DNE:
4609245e 3657 check_pool_dne(req);
5aea3dcd 3658 break;
b0494532 3659 }
6f6c7006 3660 }
422d2cb8 3661}
6f6c7006 3662
42c1b124 3663static int handle_one_map(struct ceph_osd_client *osdc,
5aea3dcd
ID		 3664 void *p, void *end, bool incremental,
3665 struct rb_root *need_resend,
3666 struct list_head *need_resend_linger)
42c1b124
ID		 3667{
3668 struct ceph_osdmap *newmap;
3669 struct rb_node *n;
3670 bool skipped_map = false;
3671 bool was_full;
3672
b7ec35b3 3673 was_full = ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
42c1b124
ID		 3674 set_pool_was_full(osdc);
3675
3676 if (incremental)
3677 newmap = osdmap_apply_incremental(&p, end, osdc->osdmap);
3678 else
3679 newmap = ceph_osdmap_decode(&p, end);
3680 if (IS_ERR(newmap))
3681 return PTR_ERR(newmap);
3682
3683 if (newmap != osdc->osdmap) {
3684 /*
3685 * Preserve ->was_full before destroying the old map.
3686 * For pools that weren't in the old map, ->was_full
3687 * should be false.
3688 */
3689 for (n = rb_first(&newmap->pg_pools); n; n = rb_next(n)) {
3690 struct ceph_pg_pool_info *pi =
3691 rb_entry(n, struct ceph_pg_pool_info, node);
3692 struct ceph_pg_pool_info *old_pi;
3693
3694 old_pi = ceph_pg_pool_by_id(osdc->osdmap, pi->id);
3695 if (old_pi)
3696 pi->was_full = old_pi->was_full;
3697 else
3698 WARN_ON(pi->was_full);
3699 }
3700
3701 if (osdc->osdmap->epoch &&
3702 osdc->osdmap->epoch + 1 < newmap->epoch) {
3703 WARN_ON(incremental);
3704 skipped_map = true;
3705 }
3706
3707 ceph_osdmap_destroy(osdc->osdmap);
3708 osdc->osdmap = newmap;
3709 }
3710
b7ec35b3 3711 was_full &= !ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
5aea3dcd
ID		 3712 scan_requests(&osdc->homeless_osd, skipped_map, was_full, true,
3713 need_resend, need_resend_linger);
3714
3715 for (n = rb_first(&osdc->osds); n; ) {
3716 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
3717
3718 n = rb_next(n); /* close_osd() */
3719
3720 scan_requests(osd, skipped_map, was_full, true, need_resend,
3721 need_resend_linger);
3722 if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) ||
3723 memcmp(&osd->o_con.peer_addr,
3724 ceph_osd_addr(osdc->osdmap, osd->o_osd),
3725 sizeof(struct ceph_entity_addr)))
3726 close_osd(osd);
3727 }
42c1b124
ID		 3728
3729 return 0;
3730}
6f6c7006 3731
5aea3dcd
ID		 3732static void kick_requests(struct ceph_osd_client *osdc,
3733 struct rb_root *need_resend,
3734 struct list_head *need_resend_linger)
3735{
922dab61 3736 struct ceph_osd_linger_request *lreq, *nlreq;
04c7d789 3737 enum calc_target_result ct_res;
5aea3dcd
ID		 3738 struct rb_node *n;
3739
04c7d789
ID		 3740 /* make sure need_resend targets reflect latest map */
3741 for (n = rb_first(need_resend); n; ) {
3742 struct ceph_osd_request *req =
3743 rb_entry(n, struct ceph_osd_request, r_node);
3744
3745 n = rb_next(n);
3746
3747 if (req->r_t.epoch < osdc->osdmap->epoch) {
3748 ct_res = calc_target(osdc, &req->r_t, NULL, false);
3749 if (ct_res == CALC_TARGET_POOL_DNE) {
3750 erase_request(need_resend, req);
3751 check_pool_dne(req);
3752 }
3753 }
3754 }
3755
5aea3dcd
ID		 3756 for (n = rb_first(need_resend); n; ) {
3757 struct ceph_osd_request *req =
3758 rb_entry(n, struct ceph_osd_request, r_node);
3759 struct ceph_osd *osd;
3760
3761 n = rb_next(n);
3762 erase_request(need_resend, req); /* before link_request() */
3763
5aea3dcd
ID		 3764 osd = lookup_create_osd(osdc, req->r_t.osd, true);
3765 link_request(osd, req);
3766 if (!req->r_linger) {
3767 if (!osd_homeless(osd) && !req->r_t.paused)
3768 send_request(req);
922dab61
ID		 3769 } else {
3770 cancel_linger_request(req);
5aea3dcd
ID		 3771 }
3772 }
922dab61
ID		 3773
3774 list_for_each_entry_safe(lreq, nlreq, need_resend_linger, scan_item) {
3775 if (!osd_homeless(lreq->osd))
3776 send_linger(lreq);
3777
3778 list_del_init(&lreq->scan_item);
3779 }
5aea3dcd
ID		 3780}
3781
f24e9980
SW		 3782/*
3783 * Process updated osd map.
3784 *
3785 * The message contains any number of incremental and full maps, normally
3786 * indicating some sort of topology change in the cluster. Kick requests
3787 * off to different OSDs as needed.
3788 */
3789void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
3790{
42c1b124
ID		 3791 void *p = msg->front.iov_base;
3792 void *const end = p + msg->front.iov_len;
f24e9980
SW		 3793 u32 nr_maps, maplen;
3794 u32 epoch;
f24e9980 3795 struct ceph_fsid fsid;
5aea3dcd
ID		 3796 struct rb_root need_resend = RB_ROOT;
3797 LIST_HEAD(need_resend_linger);
42c1b124
ID		 3798 bool handled_incremental = false;
3799 bool was_pauserd, was_pausewr;
3800 bool pauserd, pausewr;
3801 int err;
f24e9980 3802
42c1b124 3803 dout("%s have %u\n", __func__, osdc->osdmap->epoch);
5aea3dcd 3804 down_write(&osdc->lock);
f24e9980
SW		 3805
3806 /* verify fsid */
3807 ceph_decode_need(&p, end, sizeof(fsid), bad);
3808 ceph_decode_copy(&p, &fsid, sizeof(fsid));
0743304d 3809 if (ceph_check_fsid(osdc->client, &fsid) < 0)
42c1b124 3810 goto bad;
f24e9980 3811
b7ec35b3
ID		 3812 was_pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
3813 was_pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
3814 ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
42c1b124 3815 have_pool_full(osdc);
9a1ea2db 3816
f24e9980
SW		 3817 /* incremental maps */
3818 ceph_decode_32_safe(&p, end, nr_maps, bad);
3819 dout(" %d inc maps\n", nr_maps);
3820 while (nr_maps > 0) {
3821 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
c89136ea
SW		 3822 epoch = ceph_decode_32(&p);
3823 maplen = ceph_decode_32(&p);
f24e9980 3824 ceph_decode_need(&p, end, maplen, bad);
42c1b124
ID		 3825 if (osdc->osdmap->epoch &&
3826 osdc->osdmap->epoch + 1 == epoch) {
f24e9980
SW		 3827 dout("applying incremental map %u len %d\n",
3828 epoch, maplen);
5aea3dcd
ID		 3829 err = handle_one_map(osdc, p, p + maplen, true,
3830 &need_resend, &need_resend_linger);
42c1b124 3831 if (err)
f24e9980 3832 goto bad;
42c1b124 3833 handled_incremental = true;
f24e9980
SW		 3834 } else {
3835 dout("ignoring incremental map %u len %d\n",
3836 epoch, maplen);
3837 }
42c1b124 3838 p += maplen;
f24e9980
SW		 3839 nr_maps--;
3840 }
42c1b124 3841 if (handled_incremental)
f24e9980
SW		 3842 goto done;
3843
3844 /* full maps */
3845 ceph_decode_32_safe(&p, end, nr_maps, bad);
3846 dout(" %d full maps\n", nr_maps);
3847 while (nr_maps) {
3848 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
c89136ea
SW		 3849 epoch = ceph_decode_32(&p);
3850 maplen = ceph_decode_32(&p);
f24e9980
SW		 3851 ceph_decode_need(&p, end, maplen, bad);
3852 if (nr_maps > 1) {
3853 dout("skipping non-latest full map %u len %d\n",
3854 epoch, maplen);
e5253a7b 3855 } else if (osdc->osdmap->epoch >= epoch) {
f24e9980
SW		 3856 dout("skipping full map %u len %d, "
3857 "older than our %u\n", epoch, maplen,
3858 osdc->osdmap->epoch);
3859 } else {
3860 dout("taking full map %u len %d\n", epoch, maplen);
5aea3dcd
ID		 3861 err = handle_one_map(osdc, p, p + maplen, false,
3862 &need_resend, &need_resend_linger);
42c1b124 3863 if (err)
f24e9980 3864 goto bad;
f24e9980
SW		 3865 }
3866 p += maplen;
3867 nr_maps--;
3868 }
3869
3870done:
cd634fb6
SW		 3871 /*
3872 * subscribe to subsequent osdmap updates if full to ensure
3873 * we find out when we are no longer full and stop returning
3874 * ENOSPC.
3875 */
b7ec35b3
ID		 3876 pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
3877 pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
3878 ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
42c1b124 3879 have_pool_full(osdc);
58eb7932
JL		 3880 if (was_pauserd || was_pausewr || pauserd || pausewr ||
3881 osdc->osdmap->epoch < osdc->epoch_barrier)
42c1b124 3882 maybe_request_map(osdc);
cd634fb6 3883
5aea3dcd 3884 kick_requests(osdc, &need_resend, &need_resend_linger);
42c1b124 3885
fc36d0a4 3886 ceph_osdc_abort_on_full(osdc);
42c1b124
ID		 3887 ceph_monc_got_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
3888 osdc->osdmap->epoch);
5aea3dcd 3889 up_write(&osdc->lock);
03066f23 3890 wake_up_all(&osdc->client->auth_wq);
f24e9980
SW		 3891 return;
3892
3893bad:
3894 pr_err("osdc handle_map corrupt msg\n");
9ec7cab1 3895 ceph_msg_dump(msg);
5aea3dcd
ID		 3896 up_write(&osdc->lock);
3897}
3898
3899/*
3900 * Resubmit requests pending on the given osd.
3901 */
3902static void kick_osd_requests(struct ceph_osd *osd)
3903{
3904 struct rb_node *n;
3905
a02a946d
ID		 3906 clear_backoffs(osd);
3907
922dab61 3908 for (n = rb_first(&osd->o_requests); n; ) {
5aea3dcd
ID		 3909 struct ceph_osd_request *req =
3910 rb_entry(n, struct ceph_osd_request, r_node);
3911
922dab61
ID		 3912 n = rb_next(n); /* cancel_linger_request() */
3913
5aea3dcd
ID		 3914 if (!req->r_linger) {
3915 if (!req->r_t.paused)
3916 send_request(req);
922dab61
ID		 3917 } else {
3918 cancel_linger_request(req);
5aea3dcd
ID		 3919 }
3920 }
922dab61
ID		 3921 for (n = rb_first(&osd->o_linger_requests); n; n = rb_next(n)) {
3922 struct ceph_osd_linger_request *lreq =
3923 rb_entry(n, struct ceph_osd_linger_request, node);
3924
3925 send_linger(lreq);
3926 }
5aea3dcd
ID		 3927}
3928
3929/*
3930 * If the osd connection drops, we need to resubmit all requests.
3931 */
3932static void osd_fault(struct ceph_connection *con)
3933{
3934 struct ceph_osd *osd = con->private;
3935 struct ceph_osd_client *osdc = osd->o_osdc;
3936
3937 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
3938
3939 down_write(&osdc->lock);
3940 if (!osd_registered(osd)) {
3941 dout("%s osd%d unknown\n", __func__, osd->o_osd);
3942 goto out_unlock;
3943 }
3944
3945 if (!reopen_osd(osd))
3946 kick_osd_requests(osd);
3947 maybe_request_map(osdc);
3948
3949out_unlock:
3950 up_write(&osdc->lock);
f24e9980
SW		 3951}
3952
a02a946d
ID		 3953struct MOSDBackoff {
3954 struct ceph_spg spgid;
3955 u32 map_epoch;
3956 u8 op;
3957 u64 id;
3958 struct ceph_hobject_id *begin;
3959 struct ceph_hobject_id *end;
3960};
3961
3962static int decode_MOSDBackoff(const struct ceph_msg *msg, struct MOSDBackoff *m)
3963{
3964 void *p = msg->front.iov_base;
3965 void *const end = p + msg->front.iov_len;
3966 u8 struct_v;
3967 u32 struct_len;
3968 int ret;
3969
3970 ret = ceph_start_decoding(&p, end, 1, "spg_t", &struct_v, &struct_len);
3971 if (ret)
3972 return ret;
3973
3974 ret = ceph_decode_pgid(&p, end, &m->spgid.pgid);
3975 if (ret)
3976 return ret;
3977
3978 ceph_decode_8_safe(&p, end, m->spgid.shard, e_inval);
3979 ceph_decode_32_safe(&p, end, m->map_epoch, e_inval);
3980 ceph_decode_8_safe(&p, end, m->op, e_inval);
3981 ceph_decode_64_safe(&p, end, m->id, e_inval);
3982
3983 m->begin = kzalloc(sizeof(*m->begin), GFP_NOIO);
3984 if (!m->begin)
3985 return -ENOMEM;
3986
3987 ret = decode_hoid(&p, end, m->begin);
3988 if (ret) {
3989 free_hoid(m->begin);
3990 return ret;
3991 }
3992
3993 m->end = kzalloc(sizeof(*m->end), GFP_NOIO);
3994 if (!m->end) {
3995 free_hoid(m->begin);
3996 return -ENOMEM;
3997 }
3998
3999 ret = decode_hoid(&p, end, m->end);
4000 if (ret) {
4001 free_hoid(m->begin);
4002 free_hoid(m->end);
4003 return ret;
4004 }
4005
4006 return 0;
4007
4008e_inval:
4009 return -EINVAL;
4010}
4011
4012static struct ceph_msg *create_backoff_message(
4013 const struct ceph_osd_backoff *backoff,
4014 u32 map_epoch)
4015{
4016 struct ceph_msg *msg;
4017 void *p, *end;
4018 int msg_size;
4019
4020 msg_size = CEPH_ENCODING_START_BLK_LEN +
4021 CEPH_PGID_ENCODING_LEN + 1; /* spgid */
4022 msg_size += 4 + 1 + 8; /* map_epoch, op, id */
4023 msg_size += CEPH_ENCODING_START_BLK_LEN +
4024 hoid_encoding_size(backoff->begin);
4025 msg_size += CEPH_ENCODING_START_BLK_LEN +
4026 hoid_encoding_size(backoff->end);
4027
4028 msg = ceph_msg_new(CEPH_MSG_OSD_BACKOFF, msg_size, GFP_NOIO, true);
4029 if (!msg)
4030 return NULL;
4031
4032 p = msg->front.iov_base;
4033 end = p + msg->front_alloc_len;
4034
4035 encode_spgid(&p, &backoff->spgid);
4036 ceph_encode_32(&p, map_epoch);
4037 ceph_encode_8(&p, CEPH_OSD_BACKOFF_OP_ACK_BLOCK);
4038 ceph_encode_64(&p, backoff->id);
4039 encode_hoid(&p, end, backoff->begin);
4040 encode_hoid(&p, end, backoff->end);
4041 BUG_ON(p != end);
4042
4043 msg->front.iov_len = p - msg->front.iov_base;
4044 msg->hdr.version = cpu_to_le16(1); /* MOSDBackoff v1 */
4045 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
4046
4047 return msg;
4048}
4049
4050static void handle_backoff_block(struct ceph_osd *osd, struct MOSDBackoff *m)
4051{
4052 struct ceph_spg_mapping *spg;
4053 struct ceph_osd_backoff *backoff;
4054 struct ceph_msg *msg;
4055
4056 dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4057 m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4058
4059 spg = lookup_spg_mapping(&osd->o_backoff_mappings, &m->spgid);
4060 if (!spg) {
4061 spg = alloc_spg_mapping();
4062 if (!spg) {
4063 pr_err("%s failed to allocate spg\n", __func__);
4064 return;
4065 }
4066 spg->spgid = m->spgid; /* struct */
4067 insert_spg_mapping(&osd->o_backoff_mappings, spg);
4068 }
4069
4070 backoff = alloc_backoff();
4071 if (!backoff) {
4072 pr_err("%s failed to allocate backoff\n", __func__);
4073 return;
4074 }
4075 backoff->spgid = m->spgid; /* struct */
4076 backoff->id = m->id;
4077 backoff->begin = m->begin;
4078 m->begin = NULL; /* backoff now owns this */
4079 backoff->end = m->end;
4080 m->end = NULL; /* ditto */
4081
4082 insert_backoff(&spg->backoffs, backoff);
4083 insert_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4084
4085 /*
4086 * Ack with original backoff's epoch so that the OSD can
4087 * discard this if there was a PG split.
4088 */
4089 msg = create_backoff_message(backoff, m->map_epoch);
4090 if (!msg) {
4091 pr_err("%s failed to allocate msg\n", __func__);
4092 return;
4093 }
4094 ceph_con_send(&osd->o_con, msg);
4095}
4096
4097static bool target_contained_by(const struct ceph_osd_request_target *t,
4098 const struct ceph_hobject_id *begin,
4099 const struct ceph_hobject_id *end)
4100{
4101 struct ceph_hobject_id hoid;
4102 int cmp;
4103
4104 hoid_fill_from_target(&hoid, t);
4105 cmp = hoid_compare(&hoid, begin);
4106 return !cmp || (cmp > 0 && hoid_compare(&hoid, end) < 0);
4107}
4108
4109static void handle_backoff_unblock(struct ceph_osd *osd,
4110 const struct MOSDBackoff *m)
4111{
4112 struct ceph_spg_mapping *spg;
4113 struct ceph_osd_backoff *backoff;
4114 struct rb_node *n;
4115
4116 dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4117 m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4118
4119 backoff = lookup_backoff_by_id(&osd->o_backoffs_by_id, m->id);
4120 if (!backoff) {
4121 pr_err("%s osd%d spgid %llu.%xs%d id %llu backoff dne\n",
4122 __func__, osd->o_osd, m->spgid.pgid.pool,
4123 m->spgid.pgid.seed, m->spgid.shard, m->id);
4124 return;
4125 }
4126
4127 if (hoid_compare(backoff->begin, m->begin) &&
4128 hoid_compare(backoff->end, m->end)) {
4129 pr_err("%s osd%d spgid %llu.%xs%d id %llu bad range?\n",
4130 __func__, osd->o_osd, m->spgid.pgid.pool,
4131 m->spgid.pgid.seed, m->spgid.shard, m->id);
4132 /* unblock it anyway... */
4133 }
4134
4135 spg = lookup_spg_mapping(&osd->o_backoff_mappings, &backoff->spgid);
4136 BUG_ON(!spg);
4137
4138 erase_backoff(&spg->backoffs, backoff);
4139 erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4140 free_backoff(backoff);
4141
4142 if (RB_EMPTY_ROOT(&spg->backoffs)) {
4143 erase_spg_mapping(&osd->o_backoff_mappings, spg);
4144 free_spg_mapping(spg);
4145 }
4146
4147 for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
4148 struct ceph_osd_request *req =
4149 rb_entry(n, struct ceph_osd_request, r_node);
4150
4151 if (!ceph_spg_compare(&req->r_t.spgid, &m->spgid)) {
4152 /*
4153 * Match against @m, not @backoff -- the PG may
4154 * have split on the OSD.
4155 */
4156 if (target_contained_by(&req->r_t, m->begin, m->end)) {
4157 /*
4158 * If no other installed backoff applies,
4159 * resend.
4160 */
4161 send_request(req);
4162 }
4163 }
4164 }
4165}
4166
4167static void handle_backoff(struct ceph_osd *osd, struct ceph_msg *msg)
4168{
4169 struct ceph_osd_client *osdc = osd->o_osdc;
4170 struct MOSDBackoff m;
4171 int ret;
4172
4173 down_read(&osdc->lock);
4174 if (!osd_registered(osd)) {
4175 dout("%s osd%d unknown\n", __func__, osd->o_osd);
4176 up_read(&osdc->lock);
4177 return;
4178 }
4179 WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
4180
4181 mutex_lock(&osd->lock);
4182 ret = decode_MOSDBackoff(msg, &m);
4183 if (ret) {
4184 pr_err("failed to decode MOSDBackoff: %d\n", ret);
4185 ceph_msg_dump(msg);
4186 goto out_unlock;
4187 }
4188
4189 switch (m.op) {
4190 case CEPH_OSD_BACKOFF_OP_BLOCK:
4191 handle_backoff_block(osd, &m);
4192 break;
4193 case CEPH_OSD_BACKOFF_OP_UNBLOCK:
4194 handle_backoff_unblock(osd, &m);
4195 break;
4196 default:
4197 pr_err("%s osd%d unknown op %d\n", __func__, osd->o_osd, m.op);
4198 }
4199
4200 free_hoid(m.begin);
4201 free_hoid(m.end);
4202
4203out_unlock:
4204 mutex_unlock(&osd->lock);
4205 up_read(&osdc->lock);
4206}
4207
a40c4f10
YS		 4208/*
4209 * Process osd watch notifications
4210 */
3c663bbd
AE		 4211static void handle_watch_notify(struct ceph_osd_client *osdc,
4212 struct ceph_msg *msg)
a40c4f10 4213{
922dab61
ID		 4214 void *p = msg->front.iov_base;
4215 void *const end = p + msg->front.iov_len;
4216 struct ceph_osd_linger_request *lreq;
4217 struct linger_work *lwork;
4218 u8 proto_ver, opcode;
4219 u64 cookie, notify_id;
4220 u64 notifier_id = 0;
19079203 4221 s32 return_code = 0;
922dab61
ID		 4222 void *payload = NULL;
4223 u32 payload_len = 0;
a40c4f10
YS		 4224
4225 ceph_decode_8_safe(&p, end, proto_ver, bad);
4226 ceph_decode_8_safe(&p, end, opcode, bad);
4227 ceph_decode_64_safe(&p, end, cookie, bad);
922dab61 4228 p += 8; /* skip ver */
a40c4f10
YS		 4229 ceph_decode_64_safe(&p, end, notify_id, bad);
4230
922dab61
ID		 4231 if (proto_ver >= 1) {
4232 ceph_decode_32_safe(&p, end, payload_len, bad);
4233 ceph_decode_need(&p, end, payload_len, bad);
4234 payload = p;
4235 p += payload_len;
4236 }
4237
4238 if (le16_to_cpu(msg->hdr.version) >= 2)
19079203 4239 ceph_decode_32_safe(&p, end, return_code, bad);
922dab61
ID		 4240
4241 if (le16_to_cpu(msg->hdr.version) >= 3)
4242 ceph_decode_64_safe(&p, end, notifier_id, bad);
4243
4244 down_read(&osdc->lock);
4245 lreq = lookup_linger_osdc(&osdc->linger_requests, cookie);
4246 if (!lreq) {
4247 dout("%s opcode %d cookie %llu dne\n", __func__, opcode,
4248 cookie);
4249 goto out_unlock_osdc;
4250 }
4251
4252 mutex_lock(&lreq->lock);
19079203
ID		 4253 dout("%s opcode %d cookie %llu lreq %p is_watch %d\n", __func__,
4254 opcode, cookie, lreq, lreq->is_watch);
922dab61
ID		 4255 if (opcode == CEPH_WATCH_EVENT_DISCONNECT) {
4256 if (!lreq->last_error) {
4257 lreq->last_error = -ENOTCONN;
4258 queue_watch_error(lreq);
4259 }
19079203
ID		 4260 } else if (!lreq->is_watch) {
4261 /* CEPH_WATCH_EVENT_NOTIFY_COMPLETE */
4262 if (lreq->notify_id && lreq->notify_id != notify_id) {
4263 dout("lreq %p notify_id %llu != %llu, ignoring\n", lreq,
4264 lreq->notify_id, notify_id);
4265 } else if (!completion_done(&lreq->notify_finish_wait)) {
4266 struct ceph_msg_data *data =
4267 list_first_entry_or_null(&msg->data,
4268 struct ceph_msg_data,
4269 links);
4270
4271 if (data) {
4272 if (lreq->preply_pages) {
4273 WARN_ON(data->type !=
4274 CEPH_MSG_DATA_PAGES);
4275 *lreq->preply_pages = data->pages;
4276 *lreq->preply_len = data->length;
4277 } else {
4278 ceph_release_page_vector(data->pages,
4279 calc_pages_for(0, data->length));
4280 }
4281 }
4282 lreq->notify_finish_error = return_code;
4283 complete_all(&lreq->notify_finish_wait);
4284 }
922dab61
ID		 4285 } else {
4286 /* CEPH_WATCH_EVENT_NOTIFY */
4287 lwork = lwork_alloc(lreq, do_watch_notify);
4288 if (!lwork) {
4289 pr_err("failed to allocate notify-lwork\n");
4290 goto out_unlock_lreq;
a40c4f10 4291 }
a40c4f10 4292
922dab61
ID		 4293 lwork->notify.notify_id = notify_id;
4294 lwork->notify.notifier_id = notifier_id;
4295 lwork->notify.payload = payload;
4296 lwork->notify.payload_len = payload_len;
4297 lwork->notify.msg = ceph_msg_get(msg);
4298 lwork_queue(lwork);
91883cd2 4299 }
a40c4f10 4300
922dab61
ID		 4301out_unlock_lreq:
4302 mutex_unlock(&lreq->lock);
4303out_unlock_osdc:
4304 up_read(&osdc->lock);
a40c4f10
YS		 4305 return;
4306
4307bad:
4308 pr_err("osdc handle_watch_notify corrupt msg\n");
a40c4f10
YS		 4309}
4310
70636773
AE		 4311/*
4312 * Register request, send initial attempt.
4313 */
4314int ceph_osdc_start_request(struct ceph_osd_client *osdc,
4315 struct ceph_osd_request *req,
4316 bool nofail)
4317{
5aea3dcd
ID		 4318 down_read(&osdc->lock);
4319 submit_request(req, false);
4320 up_read(&osdc->lock);
0bbfdfe8 4321
5aea3dcd 4322 return 0;
f24e9980 4323}
3d14c5d2 4324EXPORT_SYMBOL(ceph_osdc_start_request);
f24e9980 4325
c9f9b93d 4326/*
c297eb42
ID		 4327 * Unregister a registered request. The request is not completed:
4328 * ->r_result isn't set and __complete_request() isn't called.
c9f9b93d
ID		 4329 */
4330void ceph_osdc_cancel_request(struct ceph_osd_request *req)
4331{
4332 struct ceph_osd_client *osdc = req->r_osdc;
4333
5aea3dcd 4334 down_write(&osdc->lock);
5aea3dcd
ID		 4335 if (req->r_osd)
4336 cancel_request(req);
4337 up_write(&osdc->lock);
c9f9b93d
ID		 4338}
4339EXPORT_SYMBOL(ceph_osdc_cancel_request);
4340
f24e9980 4341/*
42b06965 4342 * @timeout: in jiffies, 0 means "wait forever"
f24e9980 4343 */
42b06965
ID		 4344static int wait_request_timeout(struct ceph_osd_request *req,
4345 unsigned long timeout)
f24e9980 4346{
42b06965 4347 long left;
c9f9b93d 4348
42b06965 4349 dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
0e76abf2 4350 left = wait_for_completion_killable_timeout(&req->r_completion,
42b06965
ID		 4351 ceph_timeout_jiffies(timeout));
4352 if (left <= 0) {
4353 left = left ?: -ETIMEDOUT;
c9f9b93d 4354 ceph_osdc_cancel_request(req);
42b06965
ID		 4355 } else {
4356 left = req->r_result; /* completed */
f24e9980
SW		 4357 }
4358
42b06965
ID		 4359 return left;
4360}
4361
4362/*
4363 * wait for a request to complete
4364 */
4365int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
4366 struct ceph_osd_request *req)
4367{
4368 return wait_request_timeout(req, 0);
f24e9980 4369}
3d14c5d2 4370EXPORT_SYMBOL(ceph_osdc_wait_request);
f24e9980
SW		 4371
4372/*
4373 * sync - wait for all in-flight requests to flush. avoid starvation.
4374 */
4375void ceph_osdc_sync(struct ceph_osd_client *osdc)
4376{
5aea3dcd
ID		 4377 struct rb_node *n, *p;
4378 u64 last_tid = atomic64_read(&osdc->last_tid);
f24e9980 4379
5aea3dcd
ID		 4380again:
4381 down_read(&osdc->lock);
4382 for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
4383 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
4384
4385 mutex_lock(&osd->lock);
4386 for (p = rb_first(&osd->o_requests); p; p = rb_next(p)) {
4387 struct ceph_osd_request *req =
4388 rb_entry(p, struct ceph_osd_request, r_node);
4389
4390 if (req->r_tid > last_tid)
4391 break;
4392
4393 if (!(req->r_flags & CEPH_OSD_FLAG_WRITE))
4394 continue;
f24e9980 4395
5aea3dcd
ID		 4396 ceph_osdc_get_request(req);
4397 mutex_unlock(&osd->lock);
4398 up_read(&osdc->lock);
4399 dout("%s waiting on req %p tid %llu last_tid %llu\n",
4400 __func__, req, req->r_tid, last_tid);
b18b9550 4401 wait_for_completion(&req->r_completion);
5aea3dcd
ID		 4402 ceph_osdc_put_request(req);
4403 goto again;
4404 }
f24e9980 4405
5aea3dcd 4406 mutex_unlock(&osd->lock);
f24e9980 4407 }
5aea3dcd
ID		 4408
4409 up_read(&osdc->lock);
4410 dout("%s done last_tid %llu\n", __func__, last_tid);
f24e9980 4411}
3d14c5d2 4412EXPORT_SYMBOL(ceph_osdc_sync);
f24e9980 4413
922dab61
ID		 4414static struct ceph_osd_request *
4415alloc_linger_request(struct ceph_osd_linger_request *lreq)
4416{
4417 struct ceph_osd_request *req;
4418
4419 req = ceph_osdc_alloc_request(lreq->osdc, NULL, 1, false, GFP_NOIO);
4420 if (!req)
4421 return NULL;
4422
4423 ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
4424 ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
4425
4426 if (ceph_osdc_alloc_messages(req, GFP_NOIO)) {
4427 ceph_osdc_put_request(req);
4428 return NULL;
4429 }
4430
4431 return req;
4432}
4433
4434/*
4435 * Returns a handle, caller owns a ref.
4436 */
4437struct ceph_osd_linger_request *
4438ceph_osdc_watch(struct ceph_osd_client *osdc,
4439 struct ceph_object_id *oid,
4440 struct ceph_object_locator *oloc,
4441 rados_watchcb2_t wcb,
4442 rados_watcherrcb_t errcb,
4443 void *data)
4444{
4445 struct ceph_osd_linger_request *lreq;
4446 int ret;
4447
4448 lreq = linger_alloc(osdc);
4449 if (!lreq)
4450 return ERR_PTR(-ENOMEM);
4451
19079203 4452 lreq->is_watch = true;
922dab61
ID		 4453 lreq->wcb = wcb;
4454 lreq->errcb = errcb;
4455 lreq->data = data;
b07d3c4b 4456 lreq->watch_valid_thru = jiffies;
922dab61
ID		 4457
4458 ceph_oid_copy(&lreq->t.base_oid, oid);
4459 ceph_oloc_copy(&lreq->t.base_oloc, oloc);
54ea0046 4460 lreq->t.flags = CEPH_OSD_FLAG_WRITE;
1134e091 4461 ktime_get_real_ts(&lreq->mtime);
922dab61
ID		 4462
4463 lreq->reg_req = alloc_linger_request(lreq);
4464 if (!lreq->reg_req) {
4465 ret = -ENOMEM;
4466 goto err_put_lreq;
4467 }
4468
4469 lreq->ping_req = alloc_linger_request(lreq);
4470 if (!lreq->ping_req) {
4471 ret = -ENOMEM;
4472 goto err_put_lreq;
4473 }
4474
4475 down_write(&osdc->lock);
4476 linger_register(lreq); /* before osd_req_op_* */
4477 osd_req_op_watch_init(lreq->reg_req, 0, lreq->linger_id,
4478 CEPH_OSD_WATCH_OP_WATCH);
4479 osd_req_op_watch_init(lreq->ping_req, 0, lreq->linger_id,
4480 CEPH_OSD_WATCH_OP_PING);
4481 linger_submit(lreq);
4482 up_write(&osdc->lock);
4483
4484 ret = linger_reg_commit_wait(lreq);
4485 if (ret) {
4486 linger_cancel(lreq);
4487 goto err_put_lreq;
4488 }
4489
4490 return lreq;
4491
4492err_put_lreq:
4493 linger_put(lreq);
4494 return ERR_PTR(ret);
4495}
4496EXPORT_SYMBOL(ceph_osdc_watch);
4497
4498/*
4499 * Releases a ref.
4500 *
4501 * Times out after mount_timeout to preserve rbd unmap behaviour
4502 * introduced in 2894e1d76974 ("rbd: timeout watch teardown on unmap
4503 * with mount_timeout").
4504 */
4505int ceph_osdc_unwatch(struct ceph_osd_client *osdc,
4506 struct ceph_osd_linger_request *lreq)
4507{
4508 struct ceph_options *opts = osdc->client->options;
4509 struct ceph_osd_request *req;
4510 int ret;
4511
4512 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4513 if (!req)
4514 return -ENOMEM;
4515
4516 ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
4517 ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
54ea0046 4518 req->r_flags = CEPH_OSD_FLAG_WRITE;
1134e091 4519 ktime_get_real_ts(&req->r_mtime);
922dab61
ID		 4520 osd_req_op_watch_init(req, 0, lreq->linger_id,
4521 CEPH_OSD_WATCH_OP_UNWATCH);
4522
4523 ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4524 if (ret)
4525 goto out_put_req;
4526
4527 ceph_osdc_start_request(osdc, req, false);
4528 linger_cancel(lreq);
4529 linger_put(lreq);
4530 ret = wait_request_timeout(req, opts->mount_timeout);
4531
4532out_put_req:
4533 ceph_osdc_put_request(req);
4534 return ret;
4535}
4536EXPORT_SYMBOL(ceph_osdc_unwatch);
4537
4538static int osd_req_op_notify_ack_init(struct ceph_osd_request *req, int which,
4539 u64 notify_id, u64 cookie, void *payload,
4540 size_t payload_len)
4541{
4542 struct ceph_osd_req_op *op;
4543 struct ceph_pagelist *pl;
4544 int ret;
4545
4546 op = _osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY_ACK, 0);
4547
4548 pl = kmalloc(sizeof(*pl), GFP_NOIO);
4549 if (!pl)
4550 return -ENOMEM;
4551
4552 ceph_pagelist_init(pl);
4553 ret = ceph_pagelist_encode_64(pl, notify_id);
4554 ret |= ceph_pagelist_encode_64(pl, cookie);
4555 if (payload) {
4556 ret |= ceph_pagelist_encode_32(pl, payload_len);
4557 ret |= ceph_pagelist_append(pl, payload, payload_len);
4558 } else {
4559 ret |= ceph_pagelist_encode_32(pl, 0);
4560 }
4561 if (ret) {
4562 ceph_pagelist_release(pl);
4563 return -ENOMEM;
4564 }
4565
4566 ceph_osd_data_pagelist_init(&op->notify_ack.request_data, pl);
4567 op->indata_len = pl->length;
4568 return 0;
4569}
4570
4571int ceph_osdc_notify_ack(struct ceph_osd_client *osdc,
4572 struct ceph_object_id *oid,
4573 struct ceph_object_locator *oloc,
4574 u64 notify_id,
4575 u64 cookie,
4576 void *payload,
4577 size_t payload_len)
4578{
4579 struct ceph_osd_request *req;
4580 int ret;
4581
4582 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4583 if (!req)
4584 return -ENOMEM;
4585
4586 ceph_oid_copy(&req->r_base_oid, oid);
4587 ceph_oloc_copy(&req->r_base_oloc, oloc);
4588 req->r_flags = CEPH_OSD_FLAG_READ;
4589
4590 ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4591 if (ret)
4592 goto out_put_req;
4593
4594 ret = osd_req_op_notify_ack_init(req, 0, notify_id, cookie, payload,
4595 payload_len);
4596 if (ret)
4597 goto out_put_req;
4598
4599 ceph_osdc_start_request(osdc, req, false);
4600 ret = ceph_osdc_wait_request(osdc, req);
4601
4602out_put_req:
4603 ceph_osdc_put_request(req);
4604 return ret;
4605}
4606EXPORT_SYMBOL(ceph_osdc_notify_ack);
4607
19079203
ID		 4608static int osd_req_op_notify_init(struct ceph_osd_request *req, int which,
4609 u64 cookie, u32 prot_ver, u32 timeout,
4610 void *payload, size_t payload_len)
4611{
4612 struct ceph_osd_req_op *op;
4613 struct ceph_pagelist *pl;
4614 int ret;
4615
4616 op = _osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY, 0);
4617 op->notify.cookie = cookie;
4618
4619 pl = kmalloc(sizeof(*pl), GFP_NOIO);
4620 if (!pl)
4621 return -ENOMEM;
4622
4623 ceph_pagelist_init(pl);
4624 ret = ceph_pagelist_encode_32(pl, 1); /* prot_ver */
4625 ret |= ceph_pagelist_encode_32(pl, timeout);
4626 ret |= ceph_pagelist_encode_32(pl, payload_len);
4627 ret |= ceph_pagelist_append(pl, payload, payload_len);
4628 if (ret) {
4629 ceph_pagelist_release(pl);
4630 return -ENOMEM;
4631 }
4632
4633 ceph_osd_data_pagelist_init(&op->notify.request_data, pl);
4634 op->indata_len = pl->length;
4635 return 0;
4636}
4637
4638/*
4639 * @timeout: in seconds
4640 *
4641 * @preply_{pages,len} are initialized both on success and error.
4642 * The caller is responsible for:
4643 *
4644 * ceph_release_page_vector(reply_pages, calc_pages_for(0, reply_len))
4645 */
4646int ceph_osdc_notify(struct ceph_osd_client *osdc,
4647 struct ceph_object_id *oid,
4648 struct ceph_object_locator *oloc,
4649 void *payload,
4650 size_t payload_len,
4651 u32 timeout,
4652 struct page ***preply_pages,
4653 size_t *preply_len)
4654{
4655 struct ceph_osd_linger_request *lreq;
4656 struct page **pages;
4657 int ret;
4658
4659 WARN_ON(!timeout);
4660 if (preply_pages) {
4661 *preply_pages = NULL;
4662 *preply_len = 0;
4663 }
4664
4665 lreq = linger_alloc(osdc);
4666 if (!lreq)
4667 return -ENOMEM;
4668
4669 lreq->preply_pages = preply_pages;
4670 lreq->preply_len = preply_len;
4671
4672 ceph_oid_copy(&lreq->t.base_oid, oid);
4673 ceph_oloc_copy(&lreq->t.base_oloc, oloc);
4674 lreq->t.flags = CEPH_OSD_FLAG_READ;
4675
4676 lreq->reg_req = alloc_linger_request(lreq);
4677 if (!lreq->reg_req) {
4678 ret = -ENOMEM;
4679 goto out_put_lreq;
4680 }
4681
4682 /* for notify_id */
4683 pages = ceph_alloc_page_vector(1, GFP_NOIO);
4684 if (IS_ERR(pages)) {
4685 ret = PTR_ERR(pages);
4686 goto out_put_lreq;
4687 }
4688
4689 down_write(&osdc->lock);
4690 linger_register(lreq); /* before osd_req_op_* */
4691 ret = osd_req_op_notify_init(lreq->reg_req, 0, lreq->linger_id, 1,
4692 timeout, payload, payload_len);
4693 if (ret) {
4694 linger_unregister(lreq);
4695 up_write(&osdc->lock);
4696 ceph_release_page_vector(pages, 1);
4697 goto out_put_lreq;
4698 }
4699 ceph_osd_data_pages_init(osd_req_op_data(lreq->reg_req, 0, notify,
4700 response_data),
4701 pages, PAGE_SIZE, 0, false, true);
4702 linger_submit(lreq);
4703 up_write(&osdc->lock);
4704
4705 ret = linger_reg_commit_wait(lreq);
4706 if (!ret)
4707 ret = linger_notify_finish_wait(lreq);
4708 else
4709 dout("lreq %p failed to initiate notify %d\n", lreq, ret);
4710
4711 linger_cancel(lreq);
4712out_put_lreq:
4713 linger_put(lreq);
4714 return ret;
4715}
4716EXPORT_SYMBOL(ceph_osdc_notify);
4717
b07d3c4b
ID		 4718/*
4719 * Return the number of milliseconds since the watch was last
4720 * confirmed, or an error. If there is an error, the watch is no
4721 * longer valid, and should be destroyed with ceph_osdc_unwatch().
4722 */
4723int ceph_osdc_watch_check(struct ceph_osd_client *osdc,
4724 struct ceph_osd_linger_request *lreq)
4725{
4726 unsigned long stamp, age;
4727 int ret;
4728
4729 down_read(&osdc->lock);
4730 mutex_lock(&lreq->lock);
4731 stamp = lreq->watch_valid_thru;
4732 if (!list_empty(&lreq->pending_lworks)) {
4733 struct linger_work *lwork =
4734 list_first_entry(&lreq->pending_lworks,
4735 struct linger_work,
4736 pending_item);
4737
4738 if (time_before(lwork->queued_stamp, stamp))
4739 stamp = lwork->queued_stamp;
4740 }
4741 age = jiffies - stamp;
4742 dout("%s lreq %p linger_id %llu age %lu last_error %d\n", __func__,
4743 lreq, lreq->linger_id, age, lreq->last_error);
4744 /* we are truncating to msecs, so return a safe upper bound */
4745 ret = lreq->last_error ?: 1 + jiffies_to_msecs(age);
4746
4747 mutex_unlock(&lreq->lock);
4748 up_read(&osdc->lock);
4749 return ret;
4750}
4751
a4ed38d7
DF		 4752static int decode_watcher(void **p, void *end, struct ceph_watch_item *item)
4753{
4754 u8 struct_v;
4755 u32 struct_len;
4756 int ret;
4757
4758 ret = ceph_start_decoding(p, end, 2, "watch_item_t",
4759 &struct_v, &struct_len);
4760 if (ret)
4761 return ret;
4762
4763 ceph_decode_copy(p, &item->name, sizeof(item->name));
4764 item->cookie = ceph_decode_64(p);
4765 *p += 4; /* skip timeout_seconds */
4766 if (struct_v >= 2) {
4767 ceph_decode_copy(p, &item->addr, sizeof(item->addr));
4768 ceph_decode_addr(&item->addr);
4769 }
4770
4771 dout("%s %s%llu cookie %llu addr %s\n", __func__,
4772 ENTITY_NAME(item->name), item->cookie,
4773 ceph_pr_addr(&item->addr.in_addr));
4774 return 0;
4775}
4776
4777static int decode_watchers(void **p, void *end,
4778 struct ceph_watch_item **watchers,
4779 u32 *num_watchers)
4780{
4781 u8 struct_v;
4782 u32 struct_len;
4783 int i;
4784 int ret;
4785
4786 ret = ceph_start_decoding(p, end, 1, "obj_list_watch_response_t",
4787 &struct_v, &struct_len);
4788 if (ret)
4789 return ret;
4790
4791 *num_watchers = ceph_decode_32(p);
4792 *watchers = kcalloc(*num_watchers, sizeof(**watchers), GFP_NOIO);
4793 if (!*watchers)
4794 return -ENOMEM;
4795
4796 for (i = 0; i < *num_watchers; i++) {
4797 ret = decode_watcher(p, end, *watchers + i);
4798 if (ret) {
4799 kfree(*watchers);
4800 return ret;
4801 }
4802 }
4803
4804 return 0;
4805}
4806
4807/*
4808 * On success, the caller is responsible for:
4809 *
4810 * kfree(watchers);
4811 */
4812int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
4813 struct ceph_object_id *oid,
4814 struct ceph_object_locator *oloc,
4815 struct ceph_watch_item **watchers,
4816 u32 *num_watchers)
4817{
4818 struct ceph_osd_request *req;
4819 struct page **pages;
4820 int ret;
4821
4822 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4823 if (!req)
4824 return -ENOMEM;
4825
4826 ceph_oid_copy(&req->r_base_oid, oid);
4827 ceph_oloc_copy(&req->r_base_oloc, oloc);
4828 req->r_flags = CEPH_OSD_FLAG_READ;
4829
4830 ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4831 if (ret)
4832 goto out_put_req;
4833
4834 pages = ceph_alloc_page_vector(1, GFP_NOIO);
4835 if (IS_ERR(pages)) {
4836 ret = PTR_ERR(pages);
4837 goto out_put_req;
4838 }
4839
4840 osd_req_op_init(req, 0, CEPH_OSD_OP_LIST_WATCHERS, 0);
4841 ceph_osd_data_pages_init(osd_req_op_data(req, 0, list_watchers,
4842 response_data),
4843 pages, PAGE_SIZE, 0, false, true);
4844
4845 ceph_osdc_start_request(osdc, req, false);
4846 ret = ceph_osdc_wait_request(osdc, req);
4847 if (ret >= 0) {
4848 void *p = page_address(pages[0]);
4849 void *const end = p + req->r_ops[0].outdata_len;
4850
4851 ret = decode_watchers(&p, end, watchers, num_watchers);
4852 }
4853
4854out_put_req:
4855 ceph_osdc_put_request(req);
4856 return ret;
4857}
4858EXPORT_SYMBOL(ceph_osdc_list_watchers);
4859
dd935f44
JD		 4860/*
4861 * Call all pending notify callbacks - for use after a watch is
4862 * unregistered, to make sure no more callbacks for it will be invoked
4863 */
f6479449 4864void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc)
dd935f44 4865{
99d16943 4866 dout("%s osdc %p\n", __func__, osdc);
dd935f44
JD		 4867 flush_workqueue(osdc->notify_wq);
4868}
4869EXPORT_SYMBOL(ceph_osdc_flush_notifies);
4870
7cca78c9
ID		 4871void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc)
4872{
4873 down_read(&osdc->lock);
4874 maybe_request_map(osdc);
4875 up_read(&osdc->lock);
4876}
4877EXPORT_SYMBOL(ceph_osdc_maybe_request_map);
dd935f44 4878
428a7158
DF		 4879/*
4880 * Execute an OSD class method on an object.
4881 *
4882 * @flags: CEPH_OSD_FLAG_*
2544a020 4883 * @resp_len: in/out param for reply length
428a7158
DF		 4884 */
4885int ceph_osdc_call(struct ceph_osd_client *osdc,
4886 struct ceph_object_id *oid,
4887 struct ceph_object_locator *oloc,
4888 const char *class, const char *method,
4889 unsigned int flags,
4890 struct page *req_page, size_t req_len,
4891 struct page *resp_page, size_t *resp_len)
4892{
4893 struct ceph_osd_request *req;
4894 int ret;
4895
2544a020
ID		 4896 if (req_len > PAGE_SIZE || (resp_page && *resp_len > PAGE_SIZE))
4897 return -E2BIG;
4898
428a7158
DF		 4899 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4900 if (!req)
4901 return -ENOMEM;
4902
4903 ceph_oid_copy(&req->r_base_oid, oid);
4904 ceph_oloc_copy(&req->r_base_oloc, oloc);
4905 req->r_flags = flags;
4906
4907 ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4908 if (ret)
4909 goto out_put_req;
4910
4911 osd_req_op_cls_init(req, 0, CEPH_OSD_OP_CALL, class, method);
4912 if (req_page)
4913 osd_req_op_cls_request_data_pages(req, 0, &req_page, req_len,
4914 0, false, false);
4915 if (resp_page)
4916 osd_req_op_cls_response_data_pages(req, 0, &resp_page,
2544a020 4917 *resp_len, 0, false, false);
428a7158
DF		 4918
4919 ceph_osdc_start_request(osdc, req, false);
4920 ret = ceph_osdc_wait_request(osdc, req);
4921 if (ret >= 0) {
4922 ret = req->r_ops[0].rval;
4923 if (resp_page)
4924 *resp_len = req->r_ops[0].outdata_len;
4925 }
4926
4927out_put_req:
4928 ceph_osdc_put_request(req);
4929 return ret;
4930}
4931EXPORT_SYMBOL(ceph_osdc_call);
4932
f24e9980
SW		 4933/*
4934 * init, shutdown
4935 */
4936int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client)
4937{
4938 int err;
4939
4940 dout("init\n");
4941 osdc->client = client;
5aea3dcd 4942 init_rwsem(&osdc->lock);
f24e9980 4943 osdc->osds = RB_ROOT;
f5a2041b 4944 INIT_LIST_HEAD(&osdc->osd_lru);
9dd2845c 4945 spin_lock_init(&osdc->osd_lru_lock);
5aea3dcd
ID		 4946 osd_init(&osdc->homeless_osd);
4947 osdc->homeless_osd.o_osdc = osdc;
4948 osdc->homeless_osd.o_osd = CEPH_HOMELESS_OSD;
264048af 4949 osdc->last_linger_id = CEPH_LINGER_ID_START;
922dab61 4950 osdc->linger_requests = RB_ROOT;
4609245e
ID		 4951 osdc->map_checks = RB_ROOT;
4952 osdc->linger_map_checks = RB_ROOT;
f24e9980 4953 INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout);
f5a2041b
YS		 4954 INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout);
4955
5f44f142 4956 err = -ENOMEM;
e5253a7b
ID		 4957 osdc->osdmap = ceph_osdmap_alloc();
4958 if (!osdc->osdmap)
4959 goto out;
4960
9e767adb
ID		 4961 osdc->req_mempool = mempool_create_slab_pool(10,
4962 ceph_osd_request_cache);
f24e9980 4963 if (!osdc->req_mempool)
e5253a7b 4964 goto out_map;
f24e9980 4965
d50b409f 4966 err = ceph_msgpool_init(&osdc->msgpool_op, CEPH_MSG_OSD_OP,
711da55d 4967 PAGE_SIZE, 10, true, "osd_op");
f24e9980 4968 if (err < 0)
5f44f142 4969 goto out_mempool;
d50b409f 4970 err = ceph_msgpool_init(&osdc->msgpool_op_reply, CEPH_MSG_OSD_OPREPLY,
711da55d 4971 PAGE_SIZE, 10, true, "osd_op_reply");
c16e7869
SW		 4972 if (err < 0)
4973 goto out_msgpool;
a40c4f10 4974
dbcae088 4975 err = -ENOMEM;
a40c4f10 4976 osdc->notify_wq = create_singlethread_workqueue("ceph-watch-notify");
dbcae088 4977 if (!osdc->notify_wq)
c172ec5c
ID		 4978 goto out_msgpool_reply;
4979
fbca9635
ID		 4980 schedule_delayed_work(&osdc->timeout_work,
4981 osdc->client->options->osd_keepalive_timeout);
b37ee1b9
ID		 4982 schedule_delayed_work(&osdc->osds_timeout_work,
4983 round_jiffies_relative(osdc->client->options->osd_idle_ttl));
4984
f24e9980 4985 return 0;
5f44f142 4986
c172ec5c
ID		 4987out_msgpool_reply:
4988 ceph_msgpool_destroy(&osdc->msgpool_op_reply);
c16e7869
SW		 4989out_msgpool:
4990 ceph_msgpool_destroy(&osdc->msgpool_op);
5f44f142
SW		 4991out_mempool:
4992 mempool_destroy(osdc->req_mempool);
e5253a7b
ID		 4993out_map:
4994 ceph_osdmap_destroy(osdc->osdmap);
5f44f142
SW		 4995out:
4996 return err;
f24e9980
SW		 4997}
4998
4999void ceph_osdc_stop(struct ceph_osd_client *osdc)
5000{
a40c4f10
YS		 5001 flush_workqueue(osdc->notify_wq);
5002 destroy_workqueue(osdc->notify_wq);
f24e9980 5003 cancel_delayed_work_sync(&osdc->timeout_work);
f5a2041b 5004 cancel_delayed_work_sync(&osdc->osds_timeout_work);
42a2c09f 5005
5aea3dcd 5006 down_write(&osdc->lock);
42a2c09f
ID		 5007 while (!RB_EMPTY_ROOT(&osdc->osds)) {
5008 struct ceph_osd *osd = rb_entry(rb_first(&osdc->osds),
5009 struct ceph_osd, o_node);
5aea3dcd 5010 close_osd(osd);
42a2c09f 5011 }
5aea3dcd 5012 up_write(&osdc->lock);
02113a0f 5013 WARN_ON(refcount_read(&osdc->homeless_osd.o_ref) != 1);
5aea3dcd
ID		 5014 osd_cleanup(&osdc->homeless_osd);
5015
5016 WARN_ON(!list_empty(&osdc->osd_lru));
922dab61 5017 WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_requests));
4609245e
ID		 5018 WARN_ON(!RB_EMPTY_ROOT(&osdc->map_checks));
5019 WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_map_checks));
5aea3dcd
ID		 5020 WARN_ON(atomic_read(&osdc->num_requests));
5021 WARN_ON(atomic_read(&osdc->num_homeless));
42a2c09f 5022
e5253a7b 5023 ceph_osdmap_destroy(osdc->osdmap);
f24e9980
SW		 5024 mempool_destroy(osdc->req_mempool);
5025 ceph_msgpool_destroy(&osdc->msgpool_op);
c16e7869 5026 ceph_msgpool_destroy(&osdc->msgpool_op_reply);
f24e9980
SW		 5027}
5028
5029/*
5030 * Read some contiguous pages. If we cross a stripe boundary, shorten
5031 * *plen. Return number of bytes read, or error.
5032 */
5033int ceph_osdc_readpages(struct ceph_osd_client *osdc,
5034 struct ceph_vino vino, struct ceph_file_layout *layout,
5035 u64 off, u64 *plen,
5036 u32 truncate_seq, u64 truncate_size,
b7495fc2 5037 struct page **pages, int num_pages, int page_align)
f24e9980
SW		 5038{
5039 struct ceph_osd_request *req;
5040 int rc = 0;
5041
5042 dout("readpages on ino %llx.%llx on %llu~%llu\n", vino.ino,
5043 vino.snap, off, *plen);
715e4cd4 5044 req = ceph_osdc_new_request(osdc, layout, vino, off, plen, 0, 1,
f24e9980 5045 CEPH_OSD_OP_READ, CEPH_OSD_FLAG_READ,
acead002 5046 NULL, truncate_seq, truncate_size,
153e5167 5047 false);
6816282d
SW		 5048 if (IS_ERR(req))
5049 return PTR_ERR(req);
f24e9980
SW		 5050
5051 /* it may be a short read due to an object boundary */
406e2c9f 5052 osd_req_op_extent_osd_data_pages(req, 0,
a4ce40a9 5053 pages, *plen, page_align, false, false);
f24e9980 5054
e0c59487 5055 dout("readpages final extent is %llu~%llu (%llu bytes align %d)\n",
43bfe5de 5056 off, *plen, *plen, page_align);
f24e9980
SW		 5057
5058 rc = ceph_osdc_start_request(osdc, req, false);
5059 if (!rc)
5060 rc = ceph_osdc_wait_request(osdc, req);
5061
5062 ceph_osdc_put_request(req);
5063 dout("readpages result %d\n", rc);
5064 return rc;
5065}
3d14c5d2 5066EXPORT_SYMBOL(ceph_osdc_readpages);
f24e9980
SW		 5067
5068/*
5069 * do a synchronous write on N pages
5070 */
5071int ceph_osdc_writepages(struct ceph_osd_client *osdc, struct ceph_vino vino,
5072 struct ceph_file_layout *layout,
5073 struct ceph_snap_context *snapc,
5074 u64 off, u64 len,
5075 u32 truncate_seq, u64 truncate_size,
5076 struct timespec *mtime,
24808826 5077 struct page **pages, int num_pages)
f24e9980
SW		 5078{
5079 struct ceph_osd_request *req;
5080 int rc = 0;
b7495fc2 5081 int page_align = off & ~PAGE_MASK;
f24e9980 5082
715e4cd4 5083 req = ceph_osdc_new_request(osdc, layout, vino, off, &len, 0, 1,
54ea0046 5084 CEPH_OSD_OP_WRITE, CEPH_OSD_FLAG_WRITE,
acead002 5085 snapc, truncate_seq, truncate_size,
153e5167 5086 true);
6816282d
SW		 5087 if (IS_ERR(req))
5088 return PTR_ERR(req);
f24e9980
SW		 5089
5090 /* it may be a short write due to an object boundary */
406e2c9f 5091 osd_req_op_extent_osd_data_pages(req, 0, pages, len, page_align,
43bfe5de
AE		 5092 false, false);
5093 dout("writepages %llu~%llu (%llu bytes)\n", off, len, len);
f24e9980 5094
bb873b53 5095 req->r_mtime = *mtime;
87f979d3 5096 rc = ceph_osdc_start_request(osdc, req, true);
f24e9980
SW		 5097 if (!rc)
5098 rc = ceph_osdc_wait_request(osdc, req);
5099
5100 ceph_osdc_put_request(req);
5101 if (rc == 0)
5102 rc = len;
5103 dout("writepages result %d\n", rc);
5104 return rc;
5105}
3d14c5d2 5106EXPORT_SYMBOL(ceph_osdc_writepages);
f24e9980 5107
57a35dfb 5108int __init ceph_osdc_setup(void)
5522ae0b 5109{
3f1af42a
ID		 5110 size_t size = sizeof(struct ceph_osd_request) +
5111 CEPH_OSD_SLAB_OPS * sizeof(struct ceph_osd_req_op);
5112
5522ae0b 5113 BUG_ON(ceph_osd_request_cache);
3f1af42a
ID		 5114 ceph_osd_request_cache = kmem_cache_create("ceph_osd_request", size,
5115 0, 0, NULL);
5522ae0b
AE		 5116
5117 return ceph_osd_request_cache ? 0 : -ENOMEM;
5118}
5522ae0b
AE		 5119
5120void ceph_osdc_cleanup(void)
5121{
5122 BUG_ON(!ceph_osd_request_cache);
5123 kmem_cache_destroy(ceph_osd_request_cache);
5124 ceph_osd_request_cache = NULL;
5125}
5522ae0b 5126
f24e9980
SW		 5127/*
5128 * handle incoming message
5129 */
5130static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
5131{
5132 struct ceph_osd *osd = con->private;
5aea3dcd 5133 struct ceph_osd_client *osdc = osd->o_osdc;
f24e9980
SW		 5134 int type = le16_to_cpu(msg->hdr.type);
5135
f24e9980
SW		 5136 switch (type) {
5137 case CEPH_MSG_OSD_MAP:
5138 ceph_osdc_handle_map(osdc, msg);
5139 break;
5140 case CEPH_MSG_OSD_OPREPLY:
5aea3dcd 5141 handle_reply(osd, msg);
f24e9980 5142 break;
a02a946d
ID		 5143 case CEPH_MSG_OSD_BACKOFF:
5144 handle_backoff(osd, msg);
5145 break;
a40c4f10
YS		 5146 case CEPH_MSG_WATCH_NOTIFY:
5147 handle_watch_notify(osdc, msg);
5148 break;
f24e9980
SW		 5149
5150 default:
5151 pr_err("received unknown message type %d %s\n", type,
5152 ceph_msg_type_name(type));
5153 }
5aea3dcd 5154
f24e9980
SW		 5155 ceph_msg_put(msg);
5156}
5157
5b3a4db3 5158/*
d15f9d69
ID		 5159 * Lookup and return message for incoming reply. Don't try to do
5160 * anything about a larger than preallocated data portion of the
5161 * message at the moment - for now, just skip the message.
5b3a4db3
SW		 5162 */
5163static struct ceph_msg *get_reply(struct ceph_connection *con,
2450418c
YS		 5164 struct ceph_msg_header *hdr,
5165 int *skip)
f24e9980
SW		 5166{
5167 struct ceph_osd *osd = con->private;
5168 struct ceph_osd_client *osdc = osd->o_osdc;
5aea3dcd 5169 struct ceph_msg *m = NULL;
0547a9b3 5170 struct ceph_osd_request *req;
3f0a4ac5 5171 int front_len = le32_to_cpu(hdr->front_len);
5b3a4db3 5172 int data_len = le32_to_cpu(hdr->data_len);
5aea3dcd 5173 u64 tid = le64_to_cpu(hdr->tid);
f24e9980 5174
5aea3dcd
ID		 5175 down_read(&osdc->lock);
5176 if (!osd_registered(osd)) {
5177 dout("%s osd%d unknown, skipping\n", __func__, osd->o_osd);
5178 *skip = 1;
5179 goto out_unlock_osdc;
5180 }
5181 WARN_ON(osd->o_osd != le64_to_cpu(hdr->src.num));
5182
5183 mutex_lock(&osd->lock);
5184 req = lookup_request(&osd->o_requests, tid);
0547a9b3 5185 if (!req) {
cd8140c6
ID		 5186 dout("%s osd%d tid %llu unknown, skipping\n", __func__,
5187 osd->o_osd, tid);
d15f9d69 5188 *skip = 1;
5aea3dcd 5189 goto out_unlock_session;
0547a9b3 5190 }
c16e7869 5191
ace6d3a9 5192 ceph_msg_revoke_incoming(req->r_reply);
0547a9b3 5193
f2be82b0 5194 if (front_len > req->r_reply->front_alloc_len) {
d15f9d69
ID		 5195 pr_warn("%s osd%d tid %llu front %d > preallocated %d\n",
5196 __func__, osd->o_osd, req->r_tid, front_len,
5197 req->r_reply->front_alloc_len);
3f0a4ac5
ID		 5198 m = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, front_len, GFP_NOFS,
5199 false);
a79832f2 5200 if (!m)
5aea3dcd 5201 goto out_unlock_session;
c16e7869
SW		 5202 ceph_msg_put(req->r_reply);
5203 req->r_reply = m;
5204 }
0fff87ec 5205
d15f9d69
ID		 5206 if (data_len > req->r_reply->data_length) {
5207 pr_warn("%s osd%d tid %llu data %d > preallocated %zu, skipping\n",
5208 __func__, osd->o_osd, req->r_tid, data_len,
5209 req->r_reply->data_length);
5210 m = NULL;
5211 *skip = 1;
5aea3dcd 5212 goto out_unlock_session;
0547a9b3 5213 }
d15f9d69
ID		 5214
5215 m = ceph_msg_get(req->r_reply);
c16e7869 5216 dout("get_reply tid %lld %p\n", tid, m);
0547a9b3 5217
5aea3dcd
ID		 5218out_unlock_session:
5219 mutex_unlock(&osd->lock);
5220out_unlock_osdc:
5221 up_read(&osdc->lock);
2450418c 5222 return m;
5b3a4db3
SW		 5223}
5224
19079203
ID		 5225/*
5226 * TODO: switch to a msg-owned pagelist
5227 */
5228static struct ceph_msg *alloc_msg_with_page_vector(struct ceph_msg_header *hdr)
5229{
5230 struct ceph_msg *m;
5231 int type = le16_to_cpu(hdr->type);
5232 u32 front_len = le32_to_cpu(hdr->front_len);
5233 u32 data_len = le32_to_cpu(hdr->data_len);
5234
5235 m = ceph_msg_new(type, front_len, GFP_NOIO, false);
5236 if (!m)
5237 return NULL;
5238
5239 if (data_len) {
5240 struct page **pages;
5241 struct ceph_osd_data osd_data;
5242
5243 pages = ceph_alloc_page_vector(calc_pages_for(0, data_len),
5244 GFP_NOIO);
c22e853a 5245 if (IS_ERR(pages)) {
19079203
ID		 5246 ceph_msg_put(m);
5247 return NULL;
5248 }
5249
5250 ceph_osd_data_pages_init(&osd_data, pages, data_len, 0, false,
5251 false);
5252 ceph_osdc_msg_data_add(m, &osd_data);
5253 }
5254
5255 return m;
5256}
5257
5b3a4db3
SW		 5258static struct ceph_msg *alloc_msg(struct ceph_connection *con,
5259 struct ceph_msg_header *hdr,
5260 int *skip)
5261{
5262 struct ceph_osd *osd = con->private;
5263 int type = le16_to_cpu(hdr->type);
5b3a4db3 5264
1c20f2d2 5265 *skip = 0;
5b3a4db3
SW		 5266 switch (type) {
5267 case CEPH_MSG_OSD_MAP:
a02a946d 5268 case CEPH_MSG_OSD_BACKOFF:
a40c4f10 5269 case CEPH_MSG_WATCH_NOTIFY:
19079203 5270 return alloc_msg_with_page_vector(hdr);
5b3a4db3
SW		 5271 case CEPH_MSG_OSD_OPREPLY:
5272 return get_reply(con, hdr, skip);
5273 default:
5aea3dcd
ID		 5274 pr_warn("%s osd%d unknown msg type %d, skipping\n", __func__,
5275 osd->o_osd, type);
5b3a4db3
SW		 5276 *skip = 1;
5277 return NULL;
5278 }
f24e9980
SW		 5279}
5280
5281/*
5282 * Wrappers to refcount containing ceph_osd struct
5283 */
5284static struct ceph_connection *get_osd_con(struct ceph_connection *con)
5285{
5286 struct ceph_osd *osd = con->private;
5287 if (get_osd(osd))
5288 return con;
5289 return NULL;
5290}
5291
5292static void put_osd_con(struct ceph_connection *con)
5293{
5294 struct ceph_osd *osd = con->private;
5295 put_osd(osd);
5296}
5297
4e7a5dcd
SW		 5298/*
5299 * authentication
5300 */
a3530df3
AE		 5301/*
5302 * Note: returned pointer is the address of a structure that's
5303 * managed separately. Caller must *not* attempt to free it.
5304 */
5305static struct ceph_auth_handshake *get_authorizer(struct ceph_connection *con,
8f43fb53 5306 int *proto, int force_new)
4e7a5dcd
SW		 5307{
5308 struct ceph_osd *o = con->private;
5309 struct ceph_osd_client *osdc = o->o_osdc;
5310 struct ceph_auth_client *ac = osdc->client->monc.auth;
74f1869f 5311 struct ceph_auth_handshake *auth = &o->o_auth;
4e7a5dcd 5312
74f1869f 5313 if (force_new && auth->authorizer) {
6c1ea260 5314 ceph_auth_destroy_authorizer(auth->authorizer);
74f1869f
AE		 5315 auth->authorizer = NULL;
5316 }
27859f97
SW		 5317 if (!auth->authorizer) {
5318 int ret = ceph_auth_create_authorizer(ac, CEPH_ENTITY_TYPE_OSD,
5319 auth);
4e7a5dcd 5320 if (ret)
a3530df3 5321 return ERR_PTR(ret);
27859f97
SW		 5322 } else {
5323 int ret = ceph_auth_update_authorizer(ac, CEPH_ENTITY_TYPE_OSD,
0bed9b5c
SW		 5324 auth);
5325 if (ret)
5326 return ERR_PTR(ret);
4e7a5dcd 5327 }
4e7a5dcd 5328 *proto = ac->protocol;
74f1869f 5329
a3530df3 5330 return auth;
4e7a5dcd
SW		 5331}
5332
5333
0dde5848 5334static int verify_authorizer_reply(struct ceph_connection *con)
4e7a5dcd
SW		 5335{
5336 struct ceph_osd *o = con->private;
5337 struct ceph_osd_client *osdc = o->o_osdc;
5338 struct ceph_auth_client *ac = osdc->client->monc.auth;
5339
0dde5848 5340 return ceph_auth_verify_authorizer_reply(ac, o->o_auth.authorizer);
4e7a5dcd
SW		 5341}
5342
9bd2e6f8
SW		 5343static int invalidate_authorizer(struct ceph_connection *con)
5344{
5345 struct ceph_osd *o = con->private;
5346 struct ceph_osd_client *osdc = o->o_osdc;
5347 struct ceph_auth_client *ac = osdc->client->monc.auth;
5348
27859f97 5349 ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_OSD);
9bd2e6f8
SW		 5350 return ceph_monc_validate_auth(&osdc->client->monc);
5351}
4e7a5dcd 5352
8cb441c0
ID		 5353static void osd_reencode_message(struct ceph_msg *msg)
5354{
914902af
ID		 5355 int type = le16_to_cpu(msg->hdr.type);
5356
5357 if (type == CEPH_MSG_OSD_OP)
5358 encode_request_finish(msg);
8cb441c0
ID		 5359}
5360
79dbd1ba 5361static int osd_sign_message(struct ceph_msg *msg)
33d07337 5362{
79dbd1ba 5363 struct ceph_osd *o = msg->con->private;
33d07337 5364 struct ceph_auth_handshake *auth = &o->o_auth;
79dbd1ba 5365
33d07337
YZ		 5366 return ceph_auth_sign_message(auth, msg);
5367}
5368
79dbd1ba 5369static int osd_check_message_signature(struct ceph_msg *msg)
33d07337 5370{
79dbd1ba 5371 struct ceph_osd *o = msg->con->private;
33d07337 5372 struct ceph_auth_handshake *auth = &o->o_auth;
79dbd1ba 5373
33d07337
YZ		 5374 return ceph_auth_check_message_signature(auth, msg);
5375}
5376
9e32789f 5377static const struct ceph_connection_operations osd_con_ops = {
f24e9980
SW		 5378 .get = get_osd_con,
5379 .put = put_osd_con,
5380 .dispatch = dispatch,
4e7a5dcd
SW		 5381 .get_authorizer = get_authorizer,
5382 .verify_authorizer_reply = verify_authorizer_reply,
9bd2e6f8 5383 .invalidate_authorizer = invalidate_authorizer,
f24e9980 5384 .alloc_msg = alloc_msg,
8cb441c0 5385 .reencode_message = osd_reencode_message,
79dbd1ba
ID		 5386 .sign_message = osd_sign_message,
5387 .check_message_signature = osd_check_message_signature,
5aea3dcd 5388 .fault = osd_fault,
f24e9980 5389};
Ubuntu Disco kernel mirror for Proxmox VE and PMG