projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
libceph: add spinlock around osd->o_requests
[linux-2.6-block.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
89486833
ID		 129/*
130 * Consumes @pages if @own_pages is true.
131 */
a4ce40a9 132static void ceph_osd_data_pages_init(struct ceph_osd_data *osd_data,
43bfe5de
AE		 133 struct page **pages, u64 length, u32 alignment,
134 bool pages_from_pool, bool own_pages)
135{
136 osd_data->type = CEPH_OSD_DATA_TYPE_PAGES;
137 osd_data->pages = pages;
138 osd_data->length = length;
139 osd_data->alignment = alignment;
140 osd_data->pages_from_pool = pages_from_pool;
141 osd_data->own_pages = own_pages;
142}
43bfe5de 143
89486833
ID		 144/*
145 * Consumes a ref on @pagelist.
146 */
a4ce40a9 147static void ceph_osd_data_pagelist_init(struct ceph_osd_data *osd_data,
43bfe5de
AE		 148 struct ceph_pagelist *pagelist)
149{
150 osd_data->type = CEPH_OSD_DATA_TYPE_PAGELIST;
151 osd_data->pagelist = pagelist;
152}
43bfe5de
AE		 153
154#ifdef CONFIG_BLOCK
a4ce40a9 155static void ceph_osd_data_bio_init(struct ceph_osd_data *osd_data,
5359a17d
ID		 156 struct ceph_bio_iter *bio_pos,
157 u32 bio_length)
43bfe5de
AE		 158{
159 osd_data->type = CEPH_OSD_DATA_TYPE_BIO;
5359a17d 160 osd_data->bio_pos = *bio_pos;
43bfe5de
AE		 161 osd_data->bio_length = bio_length;
162}
43bfe5de
AE		 163#endif /* CONFIG_BLOCK */
164
b9e281c2 165static void ceph_osd_data_bvecs_init(struct ceph_osd_data *osd_data,
0010f705
ID		 166 struct ceph_bvec_iter *bvec_pos,
167 u32 num_bvecs)
b9e281c2
ID		 168{
169 osd_data->type = CEPH_OSD_DATA_TYPE_BVECS;
170 osd_data->bvec_pos = *bvec_pos;
0010f705 171 osd_data->num_bvecs = num_bvecs;
b9e281c2
ID		 172}
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
0010f705
ID		 240void osd_req_op_extent_osd_data_bvecs(struct ceph_osd_request *osd_req,
241 unsigned int which,
242 struct bio_vec *bvecs, u32 num_bvecs,
243 u32 bytes)
244{
245 struct ceph_osd_data *osd_data;
246 struct ceph_bvec_iter it = {
247 .bvecs = bvecs,
248 .iter = { .bi_size = bytes },
249 };
250
251 osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
252 ceph_osd_data_bvecs_init(osd_data, &it, num_bvecs);
253}
254EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvecs);
255
b9e281c2
ID		 256void osd_req_op_extent_osd_data_bvec_pos(struct ceph_osd_request *osd_req,
257 unsigned int which,
258 struct ceph_bvec_iter *bvec_pos)
259{
260 struct ceph_osd_data *osd_data;
261
262 osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
0010f705 263 ceph_osd_data_bvecs_init(osd_data, bvec_pos, 0);
b9e281c2
ID		 264}
265EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvec_pos);
266
a4ce40a9
AE		 267static void osd_req_op_cls_request_info_pagelist(
268 struct ceph_osd_request *osd_req,
269 unsigned int which, struct ceph_pagelist *pagelist)
270{
271 struct ceph_osd_data *osd_data;
272
863c7eb5 273 osd_data = osd_req_op_data(osd_req, which, cls, request_info);
a4ce40a9 274 ceph_osd_data_pagelist_init(osd_data, pagelist);
a4ce40a9
AE		 275}
276
04017e29
AE		 277void osd_req_op_cls_request_data_pagelist(
278 struct ceph_osd_request *osd_req,
279 unsigned int which, struct ceph_pagelist *pagelist)
280{
281 struct ceph_osd_data *osd_data;
282
863c7eb5 283 osd_data = osd_req_op_data(osd_req, which, cls, request_data);
04017e29 284 ceph_osd_data_pagelist_init(osd_data, pagelist);
bb873b53
ID		 285 osd_req->r_ops[which].cls.indata_len += pagelist->length;
286 osd_req->r_ops[which].indata_len += pagelist->length;
04017e29
AE		 287}
288EXPORT_SYMBOL(osd_req_op_cls_request_data_pagelist);
289
6c57b554
AE		 290void osd_req_op_cls_request_data_pages(struct ceph_osd_request *osd_req,
291 unsigned int which, struct page **pages, u64 length,
292 u32 alignment, bool pages_from_pool, bool own_pages)
293{
294 struct ceph_osd_data *osd_data;
295
296 osd_data = osd_req_op_data(osd_req, which, cls, request_data);
297 ceph_osd_data_pages_init(osd_data, pages, length, alignment,
298 pages_from_pool, own_pages);
bb873b53
ID		 299 osd_req->r_ops[which].cls.indata_len += length;
300 osd_req->r_ops[which].indata_len += length;
6c57b554
AE		 301}
302EXPORT_SYMBOL(osd_req_op_cls_request_data_pages);
303
b9e281c2
ID		 304void osd_req_op_cls_request_data_bvecs(struct ceph_osd_request *osd_req,
305 unsigned int which,
0010f705
ID		 306 struct bio_vec *bvecs, u32 num_bvecs,
307 u32 bytes)
b9e281c2
ID		 308{
309 struct ceph_osd_data *osd_data;
310 struct ceph_bvec_iter it = {
311 .bvecs = bvecs,
312 .iter = { .bi_size = bytes },
313 };
314
315 osd_data = osd_req_op_data(osd_req, which, cls, request_data);
0010f705 316 ceph_osd_data_bvecs_init(osd_data, &it, num_bvecs);
b9e281c2
ID		 317 osd_req->r_ops[which].cls.indata_len += bytes;
318 osd_req->r_ops[which].indata_len += bytes;
319}
320EXPORT_SYMBOL(osd_req_op_cls_request_data_bvecs);
321
a4ce40a9
AE		 322void osd_req_op_cls_response_data_pages(struct ceph_osd_request *osd_req,
323 unsigned int which, struct page **pages, u64 length,
324 u32 alignment, bool pages_from_pool, bool own_pages)
325{
326 struct ceph_osd_data *osd_data;
327
863c7eb5 328 osd_data = osd_req_op_data(osd_req, which, cls, response_data);
a4ce40a9
AE		 329 ceph_osd_data_pages_init(osd_data, pages, length, alignment,
330 pages_from_pool, own_pages);
a4ce40a9
AE		 331}
332EXPORT_SYMBOL(osd_req_op_cls_response_data_pages);
333
23c08a9c
AE		 334static u64 ceph_osd_data_length(struct ceph_osd_data *osd_data)
335{
336 switch (osd_data->type) {
337 case CEPH_OSD_DATA_TYPE_NONE:
338 return 0;
339 case CEPH_OSD_DATA_TYPE_PAGES:
340 return osd_data->length;
341 case CEPH_OSD_DATA_TYPE_PAGELIST:
342 return (u64)osd_data->pagelist->length;
343#ifdef CONFIG_BLOCK
344 case CEPH_OSD_DATA_TYPE_BIO:
345 return (u64)osd_data->bio_length;
346#endif /* CONFIG_BLOCK */
b9e281c2
ID		 347 case CEPH_OSD_DATA_TYPE_BVECS:
348 return osd_data->bvec_pos.iter.bi_size;
23c08a9c
AE		 349 default:
350 WARN(true, "unrecognized data type %d\n", (int)osd_data->type);
351 return 0;
352 }
353}
354
c54d47bf
AE		 355static void ceph_osd_data_release(struct ceph_osd_data *osd_data)
356{
5476492f 357 if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES && osd_data->own_pages) {
c54d47bf
AE		 358 int num_pages;
359
360 num_pages = calc_pages_for((u64)osd_data->alignment,
361 (u64)osd_data->length);
362 ceph_release_page_vector(osd_data->pages, num_pages);
89486833
ID		 363 } else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
364 ceph_pagelist_release(osd_data->pagelist);
c54d47bf 365 }
5476492f
AE		 366 ceph_osd_data_init(osd_data);
367}
368
369static void osd_req_op_data_release(struct ceph_osd_request *osd_req,
370 unsigned int which)
371{
372 struct ceph_osd_req_op *op;
373
374 BUG_ON(which >= osd_req->r_num_ops);
375 op = &osd_req->r_ops[which];
376
377 switch (op->op) {
378 case CEPH_OSD_OP_READ:
379 case CEPH_OSD_OP_WRITE:
e30b7577 380 case CEPH_OSD_OP_WRITEFULL:
5476492f
AE		 381 ceph_osd_data_release(&op->extent.osd_data);
382 break;
383 case CEPH_OSD_OP_CALL:
384 ceph_osd_data_release(&op->cls.request_info);
04017e29 385 ceph_osd_data_release(&op->cls.request_data);
5476492f
AE		 386 ceph_osd_data_release(&op->cls.response_data);
387 break;
d74b50be
YZ		 388 case CEPH_OSD_OP_SETXATTR:
389 case CEPH_OSD_OP_CMPXATTR:
390 ceph_osd_data_release(&op->xattr.osd_data);
391 break;
66ba609f
YZ		 392 case CEPH_OSD_OP_STAT:
393 ceph_osd_data_release(&op->raw_data_in);
394 break;
922dab61
ID		 395 case CEPH_OSD_OP_NOTIFY_ACK:
396 ceph_osd_data_release(&op->notify_ack.request_data);
397 break;
19079203
ID		 398 case CEPH_OSD_OP_NOTIFY:
399 ceph_osd_data_release(&op->notify.request_data);
400 ceph_osd_data_release(&op->notify.response_data);
401 break;
a4ed38d7
DF		 402 case CEPH_OSD_OP_LIST_WATCHERS:
403 ceph_osd_data_release(&op->list_watchers.response_data);
404 break;
78beb0ff 405 case CEPH_OSD_OP_COPY_FROM2:
23ddf9be
LH		 406 ceph_osd_data_release(&op->copy_from.osd_data);
407 break;
5476492f
AE		 408 default:
409 break;
410 }
c54d47bf
AE		 411}
412
63244fa1
ID		 413/*
414 * Assumes @t is zero-initialized.
415 */
416static void target_init(struct ceph_osd_request_target *t)
417{
418 ceph_oid_init(&t->base_oid);
419 ceph_oloc_init(&t->base_oloc);
420 ceph_oid_init(&t->target_oid);
421 ceph_oloc_init(&t->target_oloc);
422
423 ceph_osds_init(&t->acting);
424 ceph_osds_init(&t->up);
425 t->size = -1;
426 t->min_size = -1;
427
428 t->osd = CEPH_HOMELESS_OSD;
429}
430
922dab61
ID		 431static void target_copy(struct ceph_osd_request_target *dest,
432 const struct ceph_osd_request_target *src)
433{
434 ceph_oid_copy(&dest->base_oid, &src->base_oid);
435 ceph_oloc_copy(&dest->base_oloc, &src->base_oloc);
436 ceph_oid_copy(&dest->target_oid, &src->target_oid);
437 ceph_oloc_copy(&dest->target_oloc, &src->target_oloc);
438
439 dest->pgid = src->pgid; /* struct */
dc98ff72 440 dest->spgid = src->spgid; /* struct */
922dab61
ID		 441 dest->pg_num = src->pg_num;
442 dest->pg_num_mask = src->pg_num_mask;
443 ceph_osds_copy(&dest->acting, &src->acting);
444 ceph_osds_copy(&dest->up, &src->up);
445 dest->size = src->size;
446 dest->min_size = src->min_size;
447 dest->sort_bitwise = src->sort_bitwise;
2f3fead6 448 dest->recovery_deletes = src->recovery_deletes;
922dab61
ID		 449
450 dest->flags = src->flags;
7ed286f3 451 dest->used_replica = src->used_replica;
922dab61
ID		 452 dest->paused = src->paused;
453
04c7d789 454 dest->epoch = src->epoch;
dc93e0e2
ID		 455 dest->last_force_resend = src->last_force_resend;
456
922dab61
ID		 457 dest->osd = src->osd;
458}
459
63244fa1
ID		 460static void target_destroy(struct ceph_osd_request_target *t)
461{
462 ceph_oid_destroy(&t->base_oid);
30c156d9 463 ceph_oloc_destroy(&t->base_oloc);
63244fa1 464 ceph_oid_destroy(&t->target_oid);
30c156d9 465 ceph_oloc_destroy(&t->target_oloc);
63244fa1
ID		 466}
467
f24e9980
SW		 468/*
469 * requests
470 */
3540bfdb
ID		 471static void request_release_checks(struct ceph_osd_request *req)
472{
473 WARN_ON(!RB_EMPTY_NODE(&req->r_node));
4609245e 474 WARN_ON(!RB_EMPTY_NODE(&req->r_mc_node));
94e85771 475 WARN_ON(!list_empty(&req->r_private_item));
3540bfdb
ID		 476 WARN_ON(req->r_osd);
477}
478
9e94af20 479static void ceph_osdc_release_request(struct kref *kref)
f24e9980 480{
9e94af20
ID		 481 struct ceph_osd_request *req = container_of(kref,
482 struct ceph_osd_request, r_kref);
5476492f 483 unsigned int which;
415e49a9 484
9e94af20
ID		 485 dout("%s %p (r_request %p r_reply %p)\n", __func__, req,
486 req->r_request, req->r_reply);
3540bfdb 487 request_release_checks(req);
9e94af20 488
415e49a9
SW		 489 if (req->r_request)
490 ceph_msg_put(req->r_request);
5aea3dcd 491 if (req->r_reply)
ab8cb34a 492 ceph_msg_put(req->r_reply);
0fff87ec 493
5476492f
AE		 494 for (which = 0; which < req->r_num_ops; which++)
495 osd_req_op_data_release(req, which);
0fff87ec 496
a66dd383 497 target_destroy(&req->r_t);
415e49a9 498 ceph_put_snap_context(req->r_snapc);
d30291b9 499
415e49a9
SW		 500 if (req->r_mempool)
501 mempool_free(req, req->r_osdc->req_mempool);
3f1af42a 502 else if (req->r_num_ops <= CEPH_OSD_SLAB_OPS)
5522ae0b 503 kmem_cache_free(ceph_osd_request_cache, req);
3f1af42a
ID		 504 else
505 kfree(req);
f24e9980 506}
9e94af20
ID		 507
508void ceph_osdc_get_request(struct ceph_osd_request *req)
509{
510 dout("%s %p (was %d)\n", __func__, req,
2c935bc5 511 kref_read(&req->r_kref));
9e94af20
ID		 512 kref_get(&req->r_kref);
513}
514EXPORT_SYMBOL(ceph_osdc_get_request);
515
516void ceph_osdc_put_request(struct ceph_osd_request *req)
517{
3ed97d63
ID		 518 if (req) {
519 dout("%s %p (was %d)\n", __func__, req,
2c935bc5 520 kref_read(&req->r_kref));
3ed97d63
ID		 521 kref_put(&req->r_kref, ceph_osdc_release_request);
522 }
9e94af20
ID		 523}
524EXPORT_SYMBOL(ceph_osdc_put_request);
68b4476b 525
3540bfdb
ID		 526static void request_init(struct ceph_osd_request *req)
527{
042f6498 528 /* req only, each op is zeroed in osd_req_op_init() */
3540bfdb
ID		 529 memset(req, 0, sizeof(*req));
530
531 kref_init(&req->r_kref);
532 init_completion(&req->r_completion);
3540bfdb 533 RB_CLEAR_NODE(&req->r_node);
4609245e 534 RB_CLEAR_NODE(&req->r_mc_node);
94e85771 535 INIT_LIST_HEAD(&req->r_private_item);
3540bfdb
ID		 536
537 target_init(&req->r_t);
538}
539
3499e8a5 540struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
f24e9980 541 struct ceph_snap_context *snapc,
1b83bef2 542 unsigned int num_ops,
3499e8a5 543 bool use_mempool,
54a54007 544 gfp_t gfp_flags)
f24e9980
SW		 545{
546 struct ceph_osd_request *req;
1b83bef2 547
f24e9980 548 if (use_mempool) {
3f1af42a 549 BUG_ON(num_ops > CEPH_OSD_SLAB_OPS);
3499e8a5 550 req = mempool_alloc(osdc->req_mempool, gfp_flags);
3f1af42a
ID		 551 } else if (num_ops <= CEPH_OSD_SLAB_OPS) {
552 req = kmem_cache_alloc(ceph_osd_request_cache, gfp_flags);
f24e9980 553 } else {
3f1af42a 554 BUG_ON(num_ops > CEPH_OSD_MAX_OPS);
acafe7e3 555 req = kmalloc(struct_size(req, r_ops, num_ops), gfp_flags);
f24e9980 556 }
3f1af42a 557 if (unlikely(!req))
a79832f2 558 return NULL;
f24e9980 559
3540bfdb 560 request_init(req);
f24e9980
SW		 561 req->r_osdc = osdc;
562 req->r_mempool = use_mempool;
79528734 563 req->r_num_ops = num_ops;
84127282
ID		 564 req->r_snapid = CEPH_NOSNAP;
565 req->r_snapc = ceph_get_snap_context(snapc);
68b4476b 566
13d1ad16
ID		 567 dout("%s req %p\n", __func__, req);
568 return req;
569}
570EXPORT_SYMBOL(ceph_osdc_alloc_request);
3f1af42a 571
2e59ffd1 572static int ceph_oloc_encoding_size(const struct ceph_object_locator *oloc)
30c156d9
YZ		 573{
574 return 8 + 4 + 4 + 4 + (oloc->pool_ns ? oloc->pool_ns->len : 0);
575}
576
0d9c1ab3
ID		 577static int __ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp,
578 int num_request_data_items,
579 int num_reply_data_items)
13d1ad16
ID		 580{
581 struct ceph_osd_client *osdc = req->r_osdc;
582 struct ceph_msg *msg;
583 int msg_size;
c16e7869 584
0d9c1ab3 585 WARN_ON(req->r_request || req->r_reply);
d30291b9 586 WARN_ON(ceph_oid_empty(&req->r_base_oid));
30c156d9 587 WARN_ON(ceph_oloc_empty(&req->r_base_oloc));
d30291b9 588
13d1ad16 589 /* create request message */
8cb441c0
ID		 590 msg_size = CEPH_ENCODING_START_BLK_LEN +
591 CEPH_PGID_ENCODING_LEN + 1; /* spgid */
592 msg_size += 4 + 4 + 4; /* hash, osdmap_epoch, flags */
593 msg_size += CEPH_ENCODING_START_BLK_LEN +
594 sizeof(struct ceph_osd_reqid); /* reqid */
595 msg_size += sizeof(struct ceph_blkin_trace_info); /* trace */
596 msg_size += 4 + sizeof(struct ceph_timespec); /* client_inc, mtime */
30c156d9
YZ		 597 msg_size += CEPH_ENCODING_START_BLK_LEN +
598 ceph_oloc_encoding_size(&req->r_base_oloc); /* oloc */
13d1ad16
ID		 599 msg_size += 4 + req->r_base_oid.name_len; /* oid */
600 msg_size += 2 + req->r_num_ops * sizeof(struct ceph_osd_op);
ae458f5a
ID		 601 msg_size += 8; /* snapid */
602 msg_size += 8; /* snap_seq */
13d1ad16 603 msg_size += 4 + 8 * (req->r_snapc ? req->r_snapc->num_snaps : 0);
8cb441c0 604 msg_size += 4 + 8; /* retry_attempt, features */
ae458f5a 605
13d1ad16 606 if (req->r_mempool)
0d9c1ab3
ID		 607 msg = ceph_msgpool_get(&osdc->msgpool_op, msg_size,
608 num_request_data_items);
f24e9980 609 else
0d9c1ab3
ID		 610 msg = ceph_msg_new2(CEPH_MSG_OSD_OP, msg_size,
611 num_request_data_items, gfp, true);
13d1ad16
ID		 612 if (!msg)
613 return -ENOMEM;
68b4476b 614
f24e9980 615 memset(msg->front.iov_base, 0, msg->front.iov_len);
3499e8a5 616 req->r_request = msg;
3499e8a5 617
13d1ad16
ID		 618 /* create reply message */
619 msg_size = OSD_OPREPLY_FRONT_LEN;
711da55d
ID		 620 msg_size += req->r_base_oid.name_len;
621 msg_size += req->r_num_ops * sizeof(struct ceph_osd_op);
13d1ad16
ID		 622
623 if (req->r_mempool)
0d9c1ab3
ID		 624 msg = ceph_msgpool_get(&osdc->msgpool_op_reply, msg_size,
625 num_reply_data_items);
13d1ad16 626 else
0d9c1ab3
ID		 627 msg = ceph_msg_new2(CEPH_MSG_OSD_OPREPLY, msg_size,
628 num_reply_data_items, gfp, true);
13d1ad16
ID		 629 if (!msg)
630 return -ENOMEM;
631
632 req->r_reply = msg;
633
634 return 0;
3499e8a5
YS		 635}
636
a8dd0a37 637static bool osd_req_opcode_valid(u16 opcode)
68b4476b 638{
a8dd0a37 639 switch (opcode) {
70b5bfa3
ID		 640#define GENERATE_CASE(op, opcode, str) case CEPH_OSD_OP_##op: return true;
641__CEPH_FORALL_OSD_OPS(GENERATE_CASE)
642#undef GENERATE_CASE
a8dd0a37
AE		 643 default:
644 return false;
645 }
646}
647
0d9c1ab3
ID		 648static void get_num_data_items(struct ceph_osd_request *req,
649 int *num_request_data_items,
650 int *num_reply_data_items)
651{
652 struct ceph_osd_req_op *op;
653
654 *num_request_data_items = 0;
655 *num_reply_data_items = 0;
656
657 for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) {
658 switch (op->op) {
659 /* request */
660 case CEPH_OSD_OP_WRITE:
661 case CEPH_OSD_OP_WRITEFULL:
662 case CEPH_OSD_OP_SETXATTR:
663 case CEPH_OSD_OP_CMPXATTR:
664 case CEPH_OSD_OP_NOTIFY_ACK:
78beb0ff 665 case CEPH_OSD_OP_COPY_FROM2:
0d9c1ab3
ID		 666 *num_request_data_items += 1;
667 break;
668
669 /* reply */
670 case CEPH_OSD_OP_STAT:
671 case CEPH_OSD_OP_READ:
672 case CEPH_OSD_OP_LIST_WATCHERS:
673 *num_reply_data_items += 1;
674 break;
675
676 /* both */
677 case CEPH_OSD_OP_NOTIFY:
678 *num_request_data_items += 1;
679 *num_reply_data_items += 1;
680 break;
681 case CEPH_OSD_OP_CALL:
682 *num_request_data_items += 2;
683 *num_reply_data_items += 1;
684 break;
685
686 default:
687 WARN_ON(!osd_req_opcode_valid(op->op));
688 break;
689 }
690 }
691}
692
693/*
694 * oid, oloc and OSD op opcode(s) must be filled in before this function
695 * is called.
696 */
697int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp)
698{
699 int num_request_data_items, num_reply_data_items;
700
701 get_num_data_items(req, &num_request_data_items, &num_reply_data_items);
702 return __ceph_osdc_alloc_messages(req, gfp, num_request_data_items,
703 num_reply_data_items);
704}
705EXPORT_SYMBOL(ceph_osdc_alloc_messages);
706
33803f33
AE		 707/*
708 * This is an osd op init function for opcodes that have no data or
709 * other information associated with them. It also serves as a
710 * common init routine for all the other init functions, below.
711 */
042f6498
JL		 712struct ceph_osd_req_op *
713osd_req_op_init(struct ceph_osd_request *osd_req, unsigned int which,
144cba14 714 u16 opcode, u32 flags)
33803f33 715{
c99d2d4a
AE		 716 struct ceph_osd_req_op *op;
717
718 BUG_ON(which >= osd_req->r_num_ops);
33803f33
AE		 719 BUG_ON(!osd_req_opcode_valid(opcode));
720
c99d2d4a 721 op = &osd_req->r_ops[which];
33803f33 722 memset(op, 0, sizeof (*op));
33803f33 723 op->op = opcode;
144cba14 724 op->flags = flags;
c99d2d4a
AE		 725
726 return op;
33803f33 727}
49719778
AE		 728EXPORT_SYMBOL(osd_req_op_init);
729
c99d2d4a
AE		 730void osd_req_op_extent_init(struct ceph_osd_request *osd_req,
731 unsigned int which, u16 opcode,
33803f33
AE		 732 u64 offset, u64 length,
733 u64 truncate_size, u32 truncate_seq)
734{
042f6498
JL		 735 struct ceph_osd_req_op *op = osd_req_op_init(osd_req, which,
736 opcode, 0);
33803f33
AE		 737 size_t payload_len = 0;
738
ad7a60de 739 BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
e30b7577
ID		 740 opcode != CEPH_OSD_OP_WRITEFULL && opcode != CEPH_OSD_OP_ZERO &&
741 opcode != CEPH_OSD_OP_TRUNCATE);
33803f33 742
33803f33
AE		 743 op->extent.offset = offset;
744 op->extent.length = length;
745 op->extent.truncate_size = truncate_size;
746 op->extent.truncate_seq = truncate_seq;
e30b7577 747 if (opcode == CEPH_OSD_OP_WRITE || opcode == CEPH_OSD_OP_WRITEFULL)
33803f33
AE		 748 payload_len += length;
749
de2aa102 750 op->indata_len = payload_len;
33803f33
AE		 751}
752EXPORT_SYMBOL(osd_req_op_extent_init);
753
c99d2d4a
AE		 754void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
755 unsigned int which, u64 length)
e5975c7c 756{
c99d2d4a
AE		 757 struct ceph_osd_req_op *op;
758 u64 previous;
759
760 BUG_ON(which >= osd_req->r_num_ops);
761 op = &osd_req->r_ops[which];
762 previous = op->extent.length;
e5975c7c
AE		 763
764 if (length == previous)
765 return; /* Nothing to do */
766 BUG_ON(length > previous);
767
768 op->extent.length = length;
d641df81
YZ		 769 if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
770 op->indata_len -= previous - length;
e5975c7c
AE		 771}
772EXPORT_SYMBOL(osd_req_op_extent_update);
773
2c63f49a
YZ		 774void osd_req_op_extent_dup_last(struct ceph_osd_request *osd_req,
775 unsigned int which, u64 offset_inc)
776{
777 struct ceph_osd_req_op *op, *prev_op;
778
779 BUG_ON(which + 1 >= osd_req->r_num_ops);
780
781 prev_op = &osd_req->r_ops[which];
042f6498 782 op = osd_req_op_init(osd_req, which + 1, prev_op->op, prev_op->flags);
2c63f49a
YZ		 783 /* dup previous one */
784 op->indata_len = prev_op->indata_len;
785 op->outdata_len = prev_op->outdata_len;
786 op->extent = prev_op->extent;
787 /* adjust offset */
788 op->extent.offset += offset_inc;
789 op->extent.length -= offset_inc;
790
791 if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
792 op->indata_len -= offset_inc;
793}
794EXPORT_SYMBOL(osd_req_op_extent_dup_last);
795
fe943d50 796int osd_req_op_cls_init(struct ceph_osd_request *osd_req, unsigned int which,
24639ce5 797 const char *class, const char *method)
33803f33 798{
24639ce5 799 struct ceph_osd_req_op *op;
5f562df5 800 struct ceph_pagelist *pagelist;
33803f33
AE		 801 size_t payload_len = 0;
802 size_t size;
4766815b 803 int ret;
33803f33 804
042f6498 805 op = osd_req_op_init(osd_req, which, CEPH_OSD_OP_CALL, 0);
33803f33 806
33165d47 807 pagelist = ceph_pagelist_alloc(GFP_NOFS);
fe943d50
CX		 808 if (!pagelist)
809 return -ENOMEM;
810
33803f33
AE		 811 op->cls.class_name = class;
812 size = strlen(class);
813 BUG_ON(size > (size_t) U8_MAX);
814 op->cls.class_len = size;
4766815b
DD		 815 ret = ceph_pagelist_append(pagelist, class, size);
816 if (ret)
817 goto err_pagelist_free;
33803f33
AE		 818 payload_len += size;
819
820 op->cls.method_name = method;
821 size = strlen(method);
822 BUG_ON(size > (size_t) U8_MAX);
823 op->cls.method_len = size;
4766815b
DD		 824 ret = ceph_pagelist_append(pagelist, method, size);
825 if (ret)
826 goto err_pagelist_free;
33803f33
AE		 827 payload_len += size;
828
a4ce40a9 829 osd_req_op_cls_request_info_pagelist(osd_req, which, pagelist);
de2aa102 830 op->indata_len = payload_len;
fe943d50 831 return 0;
4766815b
DD		 832
833err_pagelist_free:
834 ceph_pagelist_release(pagelist);
835 return ret;
33803f33
AE		 836}
837EXPORT_SYMBOL(osd_req_op_cls_init);
8c042b0d 838
d74b50be
YZ		 839int osd_req_op_xattr_init(struct ceph_osd_request *osd_req, unsigned int which,
840 u16 opcode, const char *name, const void *value,
841 size_t size, u8 cmp_op, u8 cmp_mode)
842{
042f6498
JL		 843 struct ceph_osd_req_op *op = osd_req_op_init(osd_req, which,
844 opcode, 0);
d74b50be
YZ		 845 struct ceph_pagelist *pagelist;
846 size_t payload_len;
4766815b 847 int ret;
d74b50be
YZ		 848
849 BUG_ON(opcode != CEPH_OSD_OP_SETXATTR && opcode != CEPH_OSD_OP_CMPXATTR);
850
33165d47 851 pagelist = ceph_pagelist_alloc(GFP_NOFS);
d74b50be
YZ		 852 if (!pagelist)
853 return -ENOMEM;
854
d74b50be
YZ		 855 payload_len = strlen(name);
856 op->xattr.name_len = payload_len;
4766815b
DD		 857 ret = ceph_pagelist_append(pagelist, name, payload_len);
858 if (ret)
859 goto err_pagelist_free;
d74b50be
YZ		 860
861 op->xattr.value_len = size;
4766815b
DD		 862 ret = ceph_pagelist_append(pagelist, value, size);
863 if (ret)
864 goto err_pagelist_free;
d74b50be
YZ		 865 payload_len += size;
866
867 op->xattr.cmp_op = cmp_op;
868 op->xattr.cmp_mode = cmp_mode;
869
870 ceph_osd_data_pagelist_init(&op->xattr.osd_data, pagelist);
de2aa102 871 op->indata_len = payload_len;
d74b50be 872 return 0;
4766815b
DD		 873
874err_pagelist_free:
875 ceph_pagelist_release(pagelist);
876 return ret;
d74b50be
YZ		 877}
878EXPORT_SYMBOL(osd_req_op_xattr_init);
879
922dab61
ID		 880/*
881 * @watch_opcode: CEPH_OSD_WATCH_OP_*
882 */
883static void osd_req_op_watch_init(struct ceph_osd_request *req, int which,
75dbb685 884 u8 watch_opcode, u64 cookie, u32 gen)
33803f33 885{
922dab61 886 struct ceph_osd_req_op *op;
33803f33 887
042f6498 888 op = osd_req_op_init(req, which, CEPH_OSD_OP_WATCH, 0);
33803f33 889 op->watch.cookie = cookie;
922dab61 890 op->watch.op = watch_opcode;
75dbb685
ID		 891 op->watch.gen = gen;
892}
893
894/*
895 * prot_ver, timeout and notify payload (may be empty) should already be
896 * encoded in @request_pl
897 */
898static void osd_req_op_notify_init(struct ceph_osd_request *req, int which,
899 u64 cookie, struct ceph_pagelist *request_pl)
900{
901 struct ceph_osd_req_op *op;
902
903 op = osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY, 0);
904 op->notify.cookie = cookie;
905
906 ceph_osd_data_pagelist_init(&op->notify.request_data, request_pl);
907 op->indata_len = request_pl->length;
33803f33 908}
33803f33 909
d3798acc
ID		 910/*
911 * @flags: CEPH_OSD_OP_ALLOC_HINT_FLAG_*
912 */
c647b8a8
ID		 913void osd_req_op_alloc_hint_init(struct ceph_osd_request *osd_req,
914 unsigned int which,
915 u64 expected_object_size,
d3798acc
ID		 916 u64 expected_write_size,
917 u32 flags)
c647b8a8 918{
042f6498 919 struct ceph_osd_req_op *op;
c647b8a8 920
042f6498 921 op = osd_req_op_init(osd_req, which, CEPH_OSD_OP_SETALLOCHINT, 0);
c647b8a8
ID		 922 op->alloc_hint.expected_object_size = expected_object_size;
923 op->alloc_hint.expected_write_size = expected_write_size;
d3798acc 924 op->alloc_hint.flags = flags;
c647b8a8
ID		 925
926 /*
927 * CEPH_OSD_OP_SETALLOCHINT op is advisory and therefore deemed
928 * not worth a feature bit. Set FAILOK per-op flag to make
929 * sure older osds don't trip over an unsupported opcode.
930 */
931 op->flags |= CEPH_OSD_OP_FLAG_FAILOK;
932}
933EXPORT_SYMBOL(osd_req_op_alloc_hint_init);
934
90af3602 935static void ceph_osdc_msg_data_add(struct ceph_msg *msg,
ec9123c5
AE		 936 struct ceph_osd_data *osd_data)
937{
938 u64 length = ceph_osd_data_length(osd_data);
939
940 if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES) {
941 BUG_ON(length > (u64) SIZE_MAX);
942 if (length)
90af3602 943 ceph_msg_data_add_pages(msg, osd_data->pages,
e8862740 944 length, osd_data->alignment, false);
ec9123c5
AE		 945 } else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
946 BUG_ON(!length);
90af3602 947 ceph_msg_data_add_pagelist(msg, osd_data->pagelist);
ec9123c5
AE		 948#ifdef CONFIG_BLOCK
949 } else if (osd_data->type == CEPH_OSD_DATA_TYPE_BIO) {
5359a17d 950 ceph_msg_data_add_bio(msg, &osd_data->bio_pos, length);
ec9123c5 951#endif
b9e281c2
ID		 952 } else if (osd_data->type == CEPH_OSD_DATA_TYPE_BVECS) {
953 ceph_msg_data_add_bvecs(msg, &osd_data->bvec_pos);
ec9123c5
AE		 954 } else {
955 BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_NONE);
956 }
957}
958
bb873b53
ID		 959static u32 osd_req_encode_op(struct ceph_osd_op *dst,
960 const struct ceph_osd_req_op *src)
a8dd0a37 961{
a8dd0a37
AE		 962 switch (src->op) {
963 case CEPH_OSD_OP_STAT:
964 break;
965 case CEPH_OSD_OP_READ:
966 case CEPH_OSD_OP_WRITE:
e30b7577 967 case CEPH_OSD_OP_WRITEFULL:
ad7a60de 968 case CEPH_OSD_OP_ZERO:
ad7a60de 969 case CEPH_OSD_OP_TRUNCATE:
a8dd0a37
AE		 970 dst->extent.offset = cpu_to_le64(src->extent.offset);
971 dst->extent.length = cpu_to_le64(src->extent.length);
972 dst->extent.truncate_size =
973 cpu_to_le64(src->extent.truncate_size);
974 dst->extent.truncate_seq =
975 cpu_to_le32(src->extent.truncate_seq);
976 break;
977 case CEPH_OSD_OP_CALL:
a8dd0a37
AE		 978 dst->cls.class_len = src->cls.class_len;
979 dst->cls.method_len = src->cls.method_len;
bb873b53 980 dst->cls.indata_len = cpu_to_le32(src->cls.indata_len);
a8dd0a37 981 break;
a8dd0a37
AE		 982 case CEPH_OSD_OP_WATCH:
983 dst->watch.cookie = cpu_to_le64(src->watch.cookie);
922dab61
ID		 984 dst->watch.ver = cpu_to_le64(0);
985 dst->watch.op = src->watch.op;
986 dst->watch.gen = cpu_to_le32(src->watch.gen);
987 break;
988 case CEPH_OSD_OP_NOTIFY_ACK:
a8dd0a37 989 break;
19079203
ID		 990 case CEPH_OSD_OP_NOTIFY:
991 dst->notify.cookie = cpu_to_le64(src->notify.cookie);
992 break;
a4ed38d7
DF		 993 case CEPH_OSD_OP_LIST_WATCHERS:
994 break;
c647b8a8
ID		 995 case CEPH_OSD_OP_SETALLOCHINT:
996 dst->alloc_hint.expected_object_size =
997 cpu_to_le64(src->alloc_hint.expected_object_size);
998 dst->alloc_hint.expected_write_size =
999 cpu_to_le64(src->alloc_hint.expected_write_size);
d3798acc 1000 dst->alloc_hint.flags = cpu_to_le32(src->alloc_hint.flags);
c647b8a8 1001 break;
d74b50be
YZ		 1002 case CEPH_OSD_OP_SETXATTR:
1003 case CEPH_OSD_OP_CMPXATTR:
1004 dst->xattr.name_len = cpu_to_le32(src->xattr.name_len);
1005 dst->xattr.value_len = cpu_to_le32(src->xattr.value_len);
1006 dst->xattr.cmp_op = src->xattr.cmp_op;
1007 dst->xattr.cmp_mode = src->xattr.cmp_mode;
d74b50be 1008 break;
864e9197
YZ		 1009 case CEPH_OSD_OP_CREATE:
1010 case CEPH_OSD_OP_DELETE:
1011 break;
78beb0ff 1012 case CEPH_OSD_OP_COPY_FROM2:
23ddf9be
LH		 1013 dst->copy_from.snapid = cpu_to_le64(src->copy_from.snapid);
1014 dst->copy_from.src_version =
1015 cpu_to_le64(src->copy_from.src_version);
1016 dst->copy_from.flags = src->copy_from.flags;
1017 dst->copy_from.src_fadvise_flags =
1018 cpu_to_le32(src->copy_from.src_fadvise_flags);
1019 break;
a8dd0a37 1020 default:
4c46459c 1021 pr_err("unsupported osd opcode %s\n",
8f63ca2d 1022 ceph_osd_op_name(src->op));
4c46459c 1023 WARN_ON(1);
a8dd0a37
AE		 1024
1025 return 0;
68b4476b 1026 }
7b25bf5f 1027
a8dd0a37 1028 dst->op = cpu_to_le16(src->op);
7b25bf5f 1029 dst->flags = cpu_to_le32(src->flags);
de2aa102 1030 dst->payload_len = cpu_to_le32(src->indata_len);
175face2 1031
bb873b53 1032 return src->indata_len;
68b4476b
YS		 1033}
1034
3499e8a5
YS		 1035/*
1036 * build new request AND message, calculate layout, and adjust file
1037 * extent as needed.
1038 *
1039 * if the file was recently truncated, we include information about its
1040 * old and new size so that the object can be updated appropriately. (we
1041 * avoid synchronously deleting truncated objects because it's slow.)
3499e8a5
YS		 1042 */
1043struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *osdc,
1044 struct ceph_file_layout *layout,
1045 struct ceph_vino vino,
715e4cd4
YZ		 1046 u64 off, u64 *plen,
1047 unsigned int which, int num_ops,
3499e8a5
YS		 1048 int opcode, int flags,
1049 struct ceph_snap_context *snapc,
3499e8a5
YS		 1050 u32 truncate_seq,
1051 u64 truncate_size,
153e5167 1052 bool use_mempool)
3499e8a5 1053{
68b4476b 1054 struct ceph_osd_request *req;
75d1c941
AE		 1055 u64 objnum = 0;
1056 u64 objoff = 0;
1057 u64 objlen = 0;
6816282d 1058 int r;
68b4476b 1059
ad7a60de 1060 BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
864e9197
YZ		 1061 opcode != CEPH_OSD_OP_ZERO && opcode != CEPH_OSD_OP_TRUNCATE &&
1062 opcode != CEPH_OSD_OP_CREATE && opcode != CEPH_OSD_OP_DELETE);
68b4476b 1063
acead002 1064 req = ceph_osdc_alloc_request(osdc, snapc, num_ops, use_mempool,
ae7ca4a3 1065 GFP_NOFS);
13d1ad16
ID		 1066 if (!req) {
1067 r = -ENOMEM;
1068 goto fail;
1069 }
79528734 1070
3499e8a5 1071 /* calculate max write size */
a19dadfb 1072 r = calc_layout(layout, off, plen, &objnum, &objoff, &objlen);
13d1ad16
ID		 1073 if (r)
1074 goto fail;
a19dadfb 1075
864e9197 1076 if (opcode == CEPH_OSD_OP_CREATE || opcode == CEPH_OSD_OP_DELETE) {
144cba14 1077 osd_req_op_init(req, which, opcode, 0);
864e9197 1078 } else {
7627151e 1079 u32 object_size = layout->object_size;
864e9197
YZ		 1080 u32 object_base = off - objoff;
1081 if (!(truncate_seq == 1 && truncate_size == -1ULL)) {
1082 if (truncate_size <= object_base) {
1083 truncate_size = 0;
1084 } else {
1085 truncate_size -= object_base;
1086 if (truncate_size > object_size)
1087 truncate_size = object_size;
1088 }
ccca4e37 1089 }
715e4cd4 1090 osd_req_op_extent_init(req, which, opcode, objoff, objlen,
864e9197
YZ		 1091 truncate_size, truncate_seq);
1092 }
d18d1e28 1093
7627151e 1094 req->r_base_oloc.pool = layout->pool_id;
30c156d9 1095 req->r_base_oloc.pool_ns = ceph_try_get_string(layout->pool_ns);
d30291b9 1096 ceph_oid_printf(&req->r_base_oid, "%llx.%08llx", vino.ino, objnum);
22d2cfdf 1097 req->r_flags = flags | osdc->client->options->read_from_replica;
dbe0fc41 1098
bb873b53
ID		 1099 req->r_snapid = vino.snap;
1100 if (flags & CEPH_OSD_FLAG_WRITE)
1101 req->r_data_offset = off;
1102
0d9c1ab3
ID		 1103 if (num_ops > 1)
1104 /*
1105 * This is a special case for ceph_writepages_start(), but it
1106 * also covers ceph_uninline_data(). If more multi-op request
1107 * use cases emerge, we will need a separate helper.
1108 */
1109 r = __ceph_osdc_alloc_messages(req, GFP_NOFS, num_ops, 0);
1110 else
1111 r = ceph_osdc_alloc_messages(req, GFP_NOFS);
13d1ad16
ID		 1112 if (r)
1113 goto fail;
1114
f24e9980 1115 return req;
13d1ad16
ID		 1116
1117fail:
1118 ceph_osdc_put_request(req);
1119 return ERR_PTR(r);
f24e9980 1120}
3d14c5d2 1121EXPORT_SYMBOL(ceph_osdc_new_request);
f24e9980
SW		 1122
1123/*
1124 * We keep osd requests in an rbtree, sorted by ->r_tid.
1125 */
fcd00b68 1126DEFINE_RB_FUNCS(request, struct ceph_osd_request, r_tid, r_node)
4609245e 1127DEFINE_RB_FUNCS(request_mc, struct ceph_osd_request, r_tid, r_mc_node)
f24e9980 1128
66850df5
ID		 1129/*
1130 * Call @fn on each OSD request as long as @fn returns 0.
1131 */
1132static void for_each_request(struct ceph_osd_client *osdc,
1133 int (*fn)(struct ceph_osd_request *req, void *arg),
1134 void *arg)
1135{
1136 struct rb_node *n, *p;
1137
1138 for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
1139 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
1140
1141 for (p = rb_first(&osd->o_requests); p; ) {
1142 struct ceph_osd_request *req =
1143 rb_entry(p, struct ceph_osd_request, r_node);
1144
1145 p = rb_next(p);
1146 if (fn(req, arg))
1147 return;
1148 }
1149 }
1150
1151 for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
1152 struct ceph_osd_request *req =
1153 rb_entry(p, struct ceph_osd_request, r_node);
1154
1155 p = rb_next(p);
1156 if (fn(req, arg))
1157 return;
1158 }
1159}
1160
0247a0cf
ID		 1161static bool osd_homeless(struct ceph_osd *osd)
1162{
1163 return osd->o_osd == CEPH_HOMELESS_OSD;
1164}
1165
5aea3dcd 1166static bool osd_registered(struct ceph_osd *osd)
f24e9980 1167{
5aea3dcd 1168 verify_osdc_locked(osd->o_osdc);
f24e9980 1169
5aea3dcd 1170 return !RB_EMPTY_NODE(&osd->o_node);
f24e9980
SW		 1171}
1172
0247a0cf
ID		 1173/*
1174 * Assumes @osd is zero-initialized.
1175 */
1176static void osd_init(struct ceph_osd *osd)
1177{
02113a0f 1178 refcount_set(&osd->o_ref, 1);
0247a0cf 1179 RB_CLEAR_NODE(&osd->o_node);
08b8a044 1180 spin_lock_init(&osd->o_requests_lock);
5aea3dcd 1181 osd->o_requests = RB_ROOT;
922dab61 1182 osd->o_linger_requests = RB_ROOT;
a02a946d
ID		 1183 osd->o_backoff_mappings = RB_ROOT;
1184 osd->o_backoffs_by_id = RB_ROOT;
0247a0cf
ID		 1185 INIT_LIST_HEAD(&osd->o_osd_lru);
1186 INIT_LIST_HEAD(&osd->o_keepalive_item);
1187 osd->o_incarnation = 1;
5aea3dcd 1188 mutex_init(&osd->lock);
0247a0cf
ID		 1189}
1190
1191static void osd_cleanup(struct ceph_osd *osd)
1192{
1193 WARN_ON(!RB_EMPTY_NODE(&osd->o_node));
5aea3dcd 1194 WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
922dab61 1195 WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
a02a946d
ID		 1196 WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoff_mappings));
1197 WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoffs_by_id));
0247a0cf
ID		 1198 WARN_ON(!list_empty(&osd->o_osd_lru));
1199 WARN_ON(!list_empty(&osd->o_keepalive_item));
1200
1201 if (osd->o_auth.authorizer) {
1202 WARN_ON(osd_homeless(osd));
1203 ceph_auth_destroy_authorizer(osd->o_auth.authorizer);
1204 }
1205}
1206
f24e9980
SW		 1207/*
1208 * Track open sessions with osds.
1209 */
e10006f8 1210static struct ceph_osd *create_osd(struct ceph_osd_client *osdc, int onum)
f24e9980
SW		 1211{
1212 struct ceph_osd *osd;
1213
0247a0cf
ID		 1214 WARN_ON(onum == CEPH_HOMELESS_OSD);
1215
7a28f59b 1216 osd = kzalloc(sizeof(*osd), GFP_NOIO | __GFP_NOFAIL);
0247a0cf 1217 osd_init(osd);
f24e9980 1218 osd->o_osdc = osdc;
e10006f8 1219 osd->o_osd = onum;
f24e9980 1220
b7a9e5dd 1221 ceph_con_init(&osd->o_con, osd, &osd_con_ops, &osdc->client->msgr);
4e7a5dcd 1222
f24e9980
SW		 1223 return osd;
1224}
1225
1226static struct ceph_osd *get_osd(struct ceph_osd *osd)
1227{
02113a0f
ER		 1228 if (refcount_inc_not_zero(&osd->o_ref)) {
1229 dout("get_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref)-1,
1230 refcount_read(&osd->o_ref));
f24e9980
SW		 1231 return osd;
1232 } else {
1233 dout("get_osd %p FAIL\n", osd);
1234 return NULL;
1235 }
1236}
1237
1238static void put_osd(struct ceph_osd *osd)
1239{
02113a0f
ER		 1240 dout("put_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref),
1241 refcount_read(&osd->o_ref) - 1);
1242 if (refcount_dec_and_test(&osd->o_ref)) {
0247a0cf 1243 osd_cleanup(osd);
f24e9980 1244 kfree(osd);
79494d1b 1245 }
f24e9980
SW		 1246}
1247
fcd00b68
ID		 1248DEFINE_RB_FUNCS(osd, struct ceph_osd, o_osd, o_node)
1249
9dd2845c 1250static void __move_osd_to_lru(struct ceph_osd *osd)
f5a2041b 1251{
9dd2845c
ID		 1252 struct ceph_osd_client *osdc = osd->o_osdc;
1253
1254 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
f5a2041b 1255 BUG_ON(!list_empty(&osd->o_osd_lru));
bbf37ec3 1256
9dd2845c 1257 spin_lock(&osdc->osd_lru_lock);
f5a2041b 1258 list_add_tail(&osd->o_osd_lru, &osdc->osd_lru);
9dd2845c
ID		 1259 spin_unlock(&osdc->osd_lru_lock);
1260
a319bf56 1261 osd->lru_ttl = jiffies + osdc->client->options->osd_idle_ttl;
f5a2041b
YS		 1262}
1263
9dd2845c 1264static void maybe_move_osd_to_lru(struct ceph_osd *osd)
bbf37ec3 1265{
5aea3dcd 1266 if (RB_EMPTY_ROOT(&osd->o_requests) &&
922dab61 1267 RB_EMPTY_ROOT(&osd->o_linger_requests))
9dd2845c 1268 __move_osd_to_lru(osd);
bbf37ec3
ID		 1269}
1270
f5a2041b
YS		 1271static void __remove_osd_from_lru(struct ceph_osd *osd)
1272{
9dd2845c
ID		 1273 struct ceph_osd_client *osdc = osd->o_osdc;
1274
1275 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1276
1277 spin_lock(&osdc->osd_lru_lock);
f5a2041b
YS		 1278 if (!list_empty(&osd->o_osd_lru))
1279 list_del_init(&osd->o_osd_lru);
9dd2845c 1280 spin_unlock(&osdc->osd_lru_lock);
f5a2041b
YS		 1281}
1282
5aea3dcd
ID		 1283/*
1284 * Close the connection and assign any leftover requests to the
1285 * homeless session.
1286 */
1287static void close_osd(struct ceph_osd *osd)
1288{
1289 struct ceph_osd_client *osdc = osd->o_osdc;
1290 struct rb_node *n;
1291
1292 verify_osdc_wrlocked(osdc);
1293 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1294
1295 ceph_con_close(&osd->o_con);
1296
1297 for (n = rb_first(&osd->o_requests); n; ) {
1298 struct ceph_osd_request *req =
1299 rb_entry(n, struct ceph_osd_request, r_node);
1300
1301 n = rb_next(n); /* unlink_request() */
1302
1303 dout(" reassigning req %p tid %llu\n", req, req->r_tid);
1304 unlink_request(osd, req);
1305 link_request(&osdc->homeless_osd, req);
1306 }
922dab61
ID		 1307 for (n = rb_first(&osd->o_linger_requests); n; ) {
1308 struct ceph_osd_linger_request *lreq =
1309 rb_entry(n, struct ceph_osd_linger_request, node);
1310
1311 n = rb_next(n); /* unlink_linger() */
1312
1313 dout(" reassigning lreq %p linger_id %llu\n", lreq,
1314 lreq->linger_id);
1315 unlink_linger(osd, lreq);
1316 link_linger(&osdc->homeless_osd, lreq);
1317 }
a02a946d 1318 clear_backoffs(osd);
5aea3dcd
ID		 1319
1320 __remove_osd_from_lru(osd);
1321 erase_osd(&osdc->osds, osd);
1322 put_osd(osd);
1323}
1324
f24e9980
SW		 1325/*
1326 * reset osd connect
1327 */
5aea3dcd 1328static int reopen_osd(struct ceph_osd *osd)
f24e9980 1329{
c3acb181 1330 struct ceph_entity_addr *peer_addr;
f24e9980 1331
5aea3dcd
ID		 1332 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1333
1334 if (RB_EMPTY_ROOT(&osd->o_requests) &&
922dab61 1335 RB_EMPTY_ROOT(&osd->o_linger_requests)) {
5aea3dcd 1336 close_osd(osd);
c3acb181
AE		 1337 return -ENODEV;
1338 }
1339
5aea3dcd 1340 peer_addr = &osd->o_osdc->osdmap->osd_addr[osd->o_osd];
c3acb181
AE		 1341 if (!memcmp(peer_addr, &osd->o_con.peer_addr, sizeof (*peer_addr)) &&
1342 !ceph_con_opened(&osd->o_con)) {
5aea3dcd 1343 struct rb_node *n;
c3acb181 1344
0b4af2e8
ID		 1345 dout("osd addr hasn't changed and connection never opened, "
1346 "letting msgr retry\n");
87b315a5 1347 /* touch each r_stamp for handle_timeout()'s benfit */
5aea3dcd
ID		 1348 for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
1349 struct ceph_osd_request *req =
1350 rb_entry(n, struct ceph_osd_request, r_node);
87b315a5 1351 req->r_stamp = jiffies;
5aea3dcd 1352 }
c3acb181
AE		 1353
1354 return -EAGAIN;
f24e9980 1355 }
c3acb181
AE		 1356
1357 ceph_con_close(&osd->o_con);
1358 ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd, peer_addr);
1359 osd->o_incarnation++;
1360
1361 return 0;
f24e9980
SW		 1362}
1363
5aea3dcd
ID		 1364static struct ceph_osd *lookup_create_osd(struct ceph_osd_client *osdc, int o,
1365 bool wrlocked)
f24e9980 1366{
5aea3dcd 1367 struct ceph_osd *osd;
35f9f8a0 1368
5aea3dcd
ID		 1369 if (wrlocked)
1370 verify_osdc_wrlocked(osdc);
1371 else
1372 verify_osdc_locked(osdc);
f24e9980 1373
5aea3dcd
ID		 1374 if (o != CEPH_HOMELESS_OSD)
1375 osd = lookup_osd(&osdc->osds, o);
1376 else
1377 osd = &osdc->homeless_osd;
1378 if (!osd) {
1379 if (!wrlocked)
1380 return ERR_PTR(-EAGAIN);
0ba6478d 1381
5aea3dcd
ID		 1382 osd = create_osd(osdc, o);
1383 insert_osd(&osdc->osds, osd);
1384 ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd,
1385 &osdc->osdmap->osd_addr[osd->o_osd]);
0ba6478d 1386 }
f24e9980 1387
5aea3dcd
ID		 1388 dout("%s osdc %p osd%d -> osd %p\n", __func__, osdc, o, osd);
1389 return osd;
f24e9980
SW		 1390}
1391
1392/*
5aea3dcd
ID		 1393 * Create request <-> OSD session relation.
1394 *
1395 * @req has to be assigned a tid, @osd may be homeless.
f24e9980 1396 */
5aea3dcd 1397static void link_request(struct ceph_osd *osd, struct ceph_osd_request *req)
f24e9980 1398{
5aea3dcd
ID		 1399 verify_osd_locked(osd);
1400 WARN_ON(!req->r_tid || req->r_osd);
1401 dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1402 req, req->r_tid);
1403
1404 if (!osd_homeless(osd))
1405 __remove_osd_from_lru(osd);
1406 else
1407 atomic_inc(&osd->o_osdc->num_homeless);
1408
1409 get_osd(osd);
08b8a044 1410 spin_lock(&osd->o_requests_lock);
5aea3dcd 1411 insert_request(&osd->o_requests, req);
08b8a044 1412 spin_unlock(&osd->o_requests_lock);
5aea3dcd 1413 req->r_osd = osd;
f24e9980
SW		 1414}
1415
5aea3dcd
ID		 1416static void unlink_request(struct ceph_osd *osd, struct ceph_osd_request *req)
1417{
1418 verify_osd_locked(osd);
1419 WARN_ON(req->r_osd != osd);
1420 dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1421 req, req->r_tid);
1422
1423 req->r_osd = NULL;
08b8a044 1424 spin_lock(&osd->o_requests_lock);
5aea3dcd 1425 erase_request(&osd->o_requests, req);
08b8a044 1426 spin_unlock(&osd->o_requests_lock);
5aea3dcd
ID		 1427 put_osd(osd);
1428
1429 if (!osd_homeless(osd))
1430 maybe_move_osd_to_lru(osd);
1431 else
1432 atomic_dec(&osd->o_osdc->num_homeless);
1433}
1434
63244fa1
ID		 1435static bool __pool_full(struct ceph_pg_pool_info *pi)
1436{
1437 return pi->flags & CEPH_POOL_FLAG_FULL;
1438}
1439
42c1b124
ID		 1440static bool have_pool_full(struct ceph_osd_client *osdc)
1441{
1442 struct rb_node *n;
1443
1444 for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
1445 struct ceph_pg_pool_info *pi =
1446 rb_entry(n, struct ceph_pg_pool_info, node);
1447
1448 if (__pool_full(pi))
1449 return true;
1450 }
1451
1452 return false;
1453}
1454
5aea3dcd
ID		 1455static bool pool_full(struct ceph_osd_client *osdc, s64 pool_id)
1456{
1457 struct ceph_pg_pool_info *pi;
1458
1459 pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
1460 if (!pi)
1461 return false;
1462
1463 return __pool_full(pi);
1464}
1465
d29adb34
JD		 1466/*
1467 * Returns whether a request should be blocked from being sent
1468 * based on the current osdmap and osd_client settings.
d29adb34 1469 */
63244fa1
ID		 1470static bool target_should_be_paused(struct ceph_osd_client *osdc,
1471 const struct ceph_osd_request_target *t,
1472 struct ceph_pg_pool_info *pi)
1473{
b7ec35b3
ID		 1474 bool pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
1475 bool pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
1476 ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
63244fa1
ID		 1477 __pool_full(pi);
1478
6d637a54 1479 WARN_ON(pi->id != t->target_oloc.pool);
58eb7932
JL		 1480 return ((t->flags & CEPH_OSD_FLAG_READ) && pauserd) ||
1481 ((t->flags & CEPH_OSD_FLAG_WRITE) && pausewr) ||
1482 (osdc->osdmap->epoch < osdc->epoch_barrier);
63244fa1
ID		 1483}
1484
117d96a0
ID		 1485static int pick_random_replica(const struct ceph_osds *acting)
1486{
8032bf12 1487 int i = get_random_u32_below(acting->size);
117d96a0
ID		 1488
1489 dout("%s picked osd%d, primary osd%d\n", __func__,
1490 acting->osds[i], acting->primary);
1491 return i;
1492}
1493
1494/*
1495 * Picks the closest replica based on client's location given by
1496 * crush_location option. Prefers the primary if the locality is
1497 * the same.
1498 */
1499static int pick_closest_replica(struct ceph_osd_client *osdc,
1500 const struct ceph_osds *acting)
1501{
1502 struct ceph_options *opt = osdc->client->options;
1503 int best_i, best_locality;
1504 int i = 0, locality;
1505
1506 do {
1507 locality = ceph_get_crush_locality(osdc->osdmap,
1508 acting->osds[i],
1509 &opt->crush_locs);
1510 if (i == 0 ||
1511 (locality >= 0 && best_locality < 0) ||
1512 (locality >= 0 && best_locality >= 0 &&
1513 locality < best_locality)) {
1514 best_i = i;
1515 best_locality = locality;
1516 }
1517 } while (++i < acting->size);
1518
1519 dout("%s picked osd%d with locality %d, primary osd%d\n", __func__,
1520 acting->osds[best_i], best_locality, acting->primary);
1521 return best_i;
1522}
1523
63244fa1
ID		 1524enum calc_target_result {
1525 CALC_TARGET_NO_ACTION = 0,
1526 CALC_TARGET_NEED_RESEND,
1527 CALC_TARGET_POOL_DNE,
1528};
1529
1530static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
1531 struct ceph_osd_request_target *t,
63244fa1
ID		 1532 bool any_change)
1533{
1534 struct ceph_pg_pool_info *pi;
1535 struct ceph_pg pgid, last_pgid;
1536 struct ceph_osds up, acting;
117d96a0
ID		 1537 bool is_read = t->flags & CEPH_OSD_FLAG_READ;
1538 bool is_write = t->flags & CEPH_OSD_FLAG_WRITE;
63244fa1 1539 bool force_resend = false;
84ed45df 1540 bool unpaused = false;
a5613724 1541 bool legacy_change = false;
7de030d6 1542 bool split = false;
b7ec35b3 1543 bool sort_bitwise = ceph_osdmap_flag(osdc, CEPH_OSDMAP_SORTBITWISE);
ae78dd81
ID		 1544 bool recovery_deletes = ceph_osdmap_flag(osdc,
1545 CEPH_OSDMAP_RECOVERY_DELETES);
63244fa1 1546 enum calc_target_result ct_res;
63244fa1 1547
04c7d789 1548 t->epoch = osdc->osdmap->epoch;
63244fa1
ID		 1549 pi = ceph_pg_pool_by_id(osdc->osdmap, t->base_oloc.pool);
1550 if (!pi) {
1551 t->osd = CEPH_HOMELESS_OSD;
1552 ct_res = CALC_TARGET_POOL_DNE;
1553 goto out;
1554 }
1555
1556 if (osdc->osdmap->epoch == pi->last_force_request_resend) {
dc93e0e2
ID		 1557 if (t->last_force_resend < pi->last_force_request_resend) {
1558 t->last_force_resend = pi->last_force_request_resend;
63244fa1 1559 force_resend = true;
dc93e0e2 1560 } else if (t->last_force_resend == 0) {
63244fa1
ID		 1561 force_resend = true;
1562 }
1563 }
63244fa1 1564
db098ec4
ID		 1565 /* apply tiering */
1566 ceph_oid_copy(&t->target_oid, &t->base_oid);
1567 ceph_oloc_copy(&t->target_oloc, &t->base_oloc);
1568 if ((t->flags & CEPH_OSD_FLAG_IGNORE_OVERLAY) == 0) {
117d96a0 1569 if (is_read && pi->read_tier >= 0)
63244fa1 1570 t->target_oloc.pool = pi->read_tier;
117d96a0 1571 if (is_write && pi->write_tier >= 0)
63244fa1 1572 t->target_oloc.pool = pi->write_tier;
6d637a54
ID		 1573
1574 pi = ceph_pg_pool_by_id(osdc->osdmap, t->target_oloc.pool);
1575 if (!pi) {
1576 t->osd = CEPH_HOMELESS_OSD;
1577 ct_res = CALC_TARGET_POOL_DNE;
1578 goto out;
1579 }
63244fa1
ID		 1580 }
1581
a86f009f 1582 __ceph_object_locator_to_pg(pi, &t->target_oid, &t->target_oloc, &pgid);
63244fa1
ID		 1583 last_pgid.pool = pgid.pool;
1584 last_pgid.seed = ceph_stable_mod(pgid.seed, t->pg_num, t->pg_num_mask);
1585
df28152d 1586 ceph_pg_to_up_acting_osds(osdc->osdmap, pi, &pgid, &up, &acting);
63244fa1
ID		 1587 if (any_change &&
1588 ceph_is_new_interval(&t->acting,
1589 &acting,
1590 &t->up,
1591 &up,
1592 t->size,
1593 pi->size,
1594 t->min_size,
1595 pi->min_size,
1596 t->pg_num,
1597 pi->pg_num,
1598 t->sort_bitwise,
1599 sort_bitwise,
ae78dd81
ID		 1600 t->recovery_deletes,
1601 recovery_deletes,
63244fa1
ID		 1602 &last_pgid))
1603 force_resend = true;
1604
1605 if (t->paused && !target_should_be_paused(osdc, t, pi)) {
1606 t->paused = false;
84ed45df 1607 unpaused = true;
63244fa1 1608 }
84ed45df 1609 legacy_change = ceph_pg_compare(&t->pgid, &pgid) ||
117d96a0
ID		 1610 ceph_osds_changed(&t->acting, &acting,
1611 t->used_replica || any_change);
7de030d6
ID		 1612 if (t->pg_num)
1613 split = ceph_pg_is_split(&last_pgid, t->pg_num, pi->pg_num);
63244fa1 1614
7de030d6 1615 if (legacy_change || force_resend || split) {
63244fa1 1616 t->pgid = pgid; /* struct */
df28152d 1617 ceph_pg_to_primary_shard(osdc->osdmap, pi, &pgid, &t->spgid);
63244fa1
ID		 1618 ceph_osds_copy(&t->acting, &acting);
1619 ceph_osds_copy(&t->up, &up);
1620 t->size = pi->size;
1621 t->min_size = pi->min_size;
1622 t->pg_num = pi->pg_num;
1623 t->pg_num_mask = pi->pg_num_mask;
1624 t->sort_bitwise = sort_bitwise;
ae78dd81 1625 t->recovery_deletes = recovery_deletes;
63244fa1 1626
117d96a0
ID		 1627 if ((t->flags & (CEPH_OSD_FLAG_BALANCE_READS |
1628 CEPH_OSD_FLAG_LOCALIZE_READS)) &&
1629 !is_write && pi->type == CEPH_POOL_TYPE_REP &&
1630 acting.size > 1) {
1631 int pos;
1632
1633 WARN_ON(!is_read || acting.osds[0] != acting.primary);
1634 if (t->flags & CEPH_OSD_FLAG_BALANCE_READS) {
1635 pos = pick_random_replica(&acting);
1636 } else {
1637 pos = pick_closest_replica(osdc, &acting);
1638 }
1639 t->osd = acting.osds[pos];
1640 t->used_replica = pos > 0;
1641 } else {
1642 t->osd = acting.primary;
1643 t->used_replica = false;
1644 }
63244fa1
ID		 1645 }
1646
a5613724 1647 if (unpaused || legacy_change || force_resend || split)
84ed45df
ID		 1648 ct_res = CALC_TARGET_NEED_RESEND;
1649 else
1650 ct_res = CALC_TARGET_NO_ACTION;
1651
63244fa1 1652out:
a5613724
ID		 1653 dout("%s t %p -> %d%d%d%d ct_res %d osd%d\n", __func__, t, unpaused,
1654 legacy_change, force_resend, split, ct_res, t->osd);
63244fa1
ID		 1655 return ct_res;
1656}
1657
a02a946d
ID		 1658static struct ceph_spg_mapping *alloc_spg_mapping(void)
1659{
1660 struct ceph_spg_mapping *spg;
1661
1662 spg = kmalloc(sizeof(*spg), GFP_NOIO);
1663 if (!spg)
1664 return NULL;
1665
1666 RB_CLEAR_NODE(&spg->node);
1667 spg->backoffs = RB_ROOT;
1668 return spg;
1669}
1670
1671static void free_spg_mapping(struct ceph_spg_mapping *spg)
1672{
1673 WARN_ON(!RB_EMPTY_NODE(&spg->node));
1674 WARN_ON(!RB_EMPTY_ROOT(&spg->backoffs));
1675
1676 kfree(spg);
1677}
1678
1679/*
1680 * rbtree of ceph_spg_mapping for handling map<spg_t, ...>, similar to
1681 * ceph_pg_mapping. Used to track OSD backoffs -- a backoff [range] is
1682 * defined only within a specific spgid; it does not pass anything to
1683 * children on split, or to another primary.
1684 */
1685DEFINE_RB_FUNCS2(spg_mapping, struct ceph_spg_mapping, spgid, ceph_spg_compare,
1686 RB_BYPTR, const struct ceph_spg *, node)
1687
1688static u64 hoid_get_bitwise_key(const struct ceph_hobject_id *hoid)
1689{
1690 return hoid->is_max ? 0x100000000ull : hoid->hash_reverse_bits;
1691}
1692
1693static void hoid_get_effective_key(const struct ceph_hobject_id *hoid,
1694 void **pkey, size_t *pkey_len)
1695{
1696 if (hoid->key_len) {
1697 *pkey = hoid->key;
1698 *pkey_len = hoid->key_len;
1699 } else {
1700 *pkey = hoid->oid;
1701 *pkey_len = hoid->oid_len;
1702 }
1703}
1704
1705static int compare_names(const void *name1, size_t name1_len,
1706 const void *name2, size_t name2_len)
1707{
1708 int ret;
1709
1710 ret = memcmp(name1, name2, min(name1_len, name2_len));
1711 if (!ret) {
1712 if (name1_len < name2_len)
1713 ret = -1;
1714 else if (name1_len > name2_len)
1715 ret = 1;
1716 }
1717 return ret;
1718}
1719
1720static int hoid_compare(const struct ceph_hobject_id *lhs,
1721 const struct ceph_hobject_id *rhs)
1722{
1723 void *effective_key1, *effective_key2;
1724 size_t effective_key1_len, effective_key2_len;
1725 int ret;
1726
1727 if (lhs->is_max < rhs->is_max)
1728 return -1;
1729 if (lhs->is_max > rhs->is_max)
1730 return 1;
1731
1732 if (lhs->pool < rhs->pool)
1733 return -1;
1734 if (lhs->pool > rhs->pool)
1735 return 1;
1736
1737 if (hoid_get_bitwise_key(lhs) < hoid_get_bitwise_key(rhs))
1738 return -1;
1739 if (hoid_get_bitwise_key(lhs) > hoid_get_bitwise_key(rhs))
1740 return 1;
1741
1742 ret = compare_names(lhs->nspace, lhs->nspace_len,
1743 rhs->nspace, rhs->nspace_len);
1744 if (ret)
1745 return ret;
1746
1747 hoid_get_effective_key(lhs, &effective_key1, &effective_key1_len);
1748 hoid_get_effective_key(rhs, &effective_key2, &effective_key2_len);
1749 ret = compare_names(effective_key1, effective_key1_len,
1750 effective_key2, effective_key2_len);
1751 if (ret)
1752 return ret;
1753
1754 ret = compare_names(lhs->oid, lhs->oid_len, rhs->oid, rhs->oid_len);
1755 if (ret)
1756 return ret;
1757
1758 if (lhs->snapid < rhs->snapid)
1759 return -1;
1760 if (lhs->snapid > rhs->snapid)
1761 return 1;
1762
1763 return 0;
1764}
1765
1766/*
1767 * For decoding ->begin and ->end of MOSDBackoff only -- no MIN/MAX
1768 * compat stuff here.
1769 *
1770 * Assumes @hoid is zero-initialized.
1771 */
1772static int decode_hoid(void **p, void *end, struct ceph_hobject_id *hoid)
1773{
1774 u8 struct_v;
1775 u32 struct_len;
1776 int ret;
1777
1778 ret = ceph_start_decoding(p, end, 4, "hobject_t", &struct_v,
1779 &struct_len);
1780 if (ret)
1781 return ret;
1782
1783 if (struct_v < 4) {
1784 pr_err("got struct_v %d < 4 of hobject_t\n", struct_v);
1785 goto e_inval;
1786 }
1787
1788 hoid->key = ceph_extract_encoded_string(p, end, &hoid->key_len,
1789 GFP_NOIO);
1790 if (IS_ERR(hoid->key)) {
1791 ret = PTR_ERR(hoid->key);
1792 hoid->key = NULL;
1793 return ret;
1794 }
1795
1796 hoid->oid = ceph_extract_encoded_string(p, end, &hoid->oid_len,
1797 GFP_NOIO);
1798 if (IS_ERR(hoid->oid)) {
1799 ret = PTR_ERR(hoid->oid);
1800 hoid->oid = NULL;
1801 return ret;
1802 }
1803
1804 ceph_decode_64_safe(p, end, hoid->snapid, e_inval);
1805 ceph_decode_32_safe(p, end, hoid->hash, e_inval);
1806 ceph_decode_8_safe(p, end, hoid->is_max, e_inval);
1807
1808 hoid->nspace = ceph_extract_encoded_string(p, end, &hoid->nspace_len,
1809 GFP_NOIO);
1810 if (IS_ERR(hoid->nspace)) {
1811 ret = PTR_ERR(hoid->nspace);
1812 hoid->nspace = NULL;
1813 return ret;
1814 }
1815
1816 ceph_decode_64_safe(p, end, hoid->pool, e_inval);
1817
1818 ceph_hoid_build_hash_cache(hoid);
1819 return 0;
1820
1821e_inval:
1822 return -EINVAL;
1823}
1824
1825static int hoid_encoding_size(const struct ceph_hobject_id *hoid)
1826{
1827 return 8 + 4 + 1 + 8 + /* snapid, hash, is_max, pool */
1828 4 + hoid->key_len + 4 + hoid->oid_len + 4 + hoid->nspace_len;
1829}
1830
1831static void encode_hoid(void **p, void *end, const struct ceph_hobject_id *hoid)
1832{
1833 ceph_start_encoding(p, 4, 3, hoid_encoding_size(hoid));
1834 ceph_encode_string(p, end, hoid->key, hoid->key_len);
1835 ceph_encode_string(p, end, hoid->oid, hoid->oid_len);
1836 ceph_encode_64(p, hoid->snapid);
1837 ceph_encode_32(p, hoid->hash);
1838 ceph_encode_8(p, hoid->is_max);
1839 ceph_encode_string(p, end, hoid->nspace, hoid->nspace_len);
1840 ceph_encode_64(p, hoid->pool);
1841}
1842
1843static void free_hoid(struct ceph_hobject_id *hoid)
1844{
1845 if (hoid) {
1846 kfree(hoid->key);
1847 kfree(hoid->oid);
1848 kfree(hoid->nspace);
1849 kfree(hoid);
1850 }
1851}
1852
1853static struct ceph_osd_backoff *alloc_backoff(void)
1854{
1855 struct ceph_osd_backoff *backoff;
1856
1857 backoff = kzalloc(sizeof(*backoff), GFP_NOIO);
1858 if (!backoff)
1859 return NULL;
1860
1861 RB_CLEAR_NODE(&backoff->spg_node);
1862 RB_CLEAR_NODE(&backoff->id_node);
1863 return backoff;
1864}
1865
1866static void free_backoff(struct ceph_osd_backoff *backoff)
1867{
1868 WARN_ON(!RB_EMPTY_NODE(&backoff->spg_node));
1869 WARN_ON(!RB_EMPTY_NODE(&backoff->id_node));
1870
1871 free_hoid(backoff->begin);
1872 free_hoid(backoff->end);
1873 kfree(backoff);
1874}
1875
1876/*
1877 * Within a specific spgid, backoffs are managed by ->begin hoid.
1878 */
1879DEFINE_RB_INSDEL_FUNCS2(backoff, struct ceph_osd_backoff, begin, hoid_compare,
1880 RB_BYVAL, spg_node);
1881
1882static struct ceph_osd_backoff *lookup_containing_backoff(struct rb_root *root,
1883 const struct ceph_hobject_id *hoid)
1884{
1885 struct rb_node *n = root->rb_node;
1886
1887 while (n) {
1888 struct ceph_osd_backoff *cur =
1889 rb_entry(n, struct ceph_osd_backoff, spg_node);
1890 int cmp;
1891
1892 cmp = hoid_compare(hoid, cur->begin);
1893 if (cmp < 0) {
1894 n = n->rb_left;
1895 } else if (cmp > 0) {
1896 if (hoid_compare(hoid, cur->end) < 0)
1897 return cur;
1898
1899 n = n->rb_right;
1900 } else {
1901 return cur;
1902 }
1903 }
1904
1905 return NULL;
1906}
1907
1908/*
1909 * Each backoff has a unique id within its OSD session.
1910 */
1911DEFINE_RB_FUNCS(backoff_by_id, struct ceph_osd_backoff, id, id_node)
1912
1913static void clear_backoffs(struct ceph_osd *osd)
1914{
1915 while (!RB_EMPTY_ROOT(&osd->o_backoff_mappings)) {
1916 struct ceph_spg_mapping *spg =
1917 rb_entry(rb_first(&osd->o_backoff_mappings),
1918 struct ceph_spg_mapping, node);
1919
1920 while (!RB_EMPTY_ROOT(&spg->backoffs)) {
1921 struct ceph_osd_backoff *backoff =
1922 rb_entry(rb_first(&spg->backoffs),
1923 struct ceph_osd_backoff, spg_node);
1924
1925 erase_backoff(&spg->backoffs, backoff);
1926 erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
1927 free_backoff(backoff);
1928 }
1929 erase_spg_mapping(&osd->o_backoff_mappings, spg);
1930 free_spg_mapping(spg);
1931 }
1932}
1933
1934/*
1935 * Set up a temporary, non-owning view into @t.
1936 */
1937static void hoid_fill_from_target(struct ceph_hobject_id *hoid,
1938 const struct ceph_osd_request_target *t)
1939{
1940 hoid->key = NULL;
1941 hoid->key_len = 0;
1942 hoid->oid = t->target_oid.name;
1943 hoid->oid_len = t->target_oid.name_len;
1944 hoid->snapid = CEPH_NOSNAP;
1945 hoid->hash = t->pgid.seed;
1946 hoid->is_max = false;
1947 if (t->target_oloc.pool_ns) {
1948 hoid->nspace = t->target_oloc.pool_ns->str;
1949 hoid->nspace_len = t->target_oloc.pool_ns->len;
1950 } else {
1951 hoid->nspace = NULL;
1952 hoid->nspace_len = 0;
1953 }
1954 hoid->pool = t->target_oloc.pool;
1955 ceph_hoid_build_hash_cache(hoid);
1956}
1957
1958static bool should_plug_request(struct ceph_osd_request *req)
1959{
1960 struct ceph_osd *osd = req->r_osd;
1961 struct ceph_spg_mapping *spg;
1962 struct ceph_osd_backoff *backoff;
1963 struct ceph_hobject_id hoid;
1964
1965 spg = lookup_spg_mapping(&osd->o_backoff_mappings, &req->r_t.spgid);
1966 if (!spg)
1967 return false;
1968
1969 hoid_fill_from_target(&hoid, &req->r_t);
1970 backoff = lookup_containing_backoff(&spg->backoffs, &hoid);
1971 if (!backoff)
1972 return false;
1973
1974 dout("%s req %p tid %llu backoff osd%d spgid %llu.%xs%d id %llu\n",
1975 __func__, req, req->r_tid, osd->o_osd, backoff->spgid.pgid.pool,
1976 backoff->spgid.pgid.seed, backoff->spgid.shard, backoff->id);
1977 return true;
1978}
1979
0d9c1ab3
ID		 1980/*
1981 * Keep get_num_data_items() in sync with this function.
1982 */
98c4bfe9 1983static void setup_request_data(struct ceph_osd_request *req)
f24e9980 1984{
98c4bfe9
ID		 1985 struct ceph_msg *request_msg = req->r_request;
1986 struct ceph_msg *reply_msg = req->r_reply;
1987 struct ceph_osd_req_op *op;
bb873b53 1988
98c4bfe9 1989 if (req->r_request->num_data_items || req->r_reply->num_data_items)
bb873b53 1990 return;
f24e9980 1991
98c4bfe9
ID		 1992 WARN_ON(request_msg->data_length || reply_msg->data_length);
1993 for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) {
bb873b53
ID		 1994 switch (op->op) {
1995 /* request */
1996 case CEPH_OSD_OP_WRITE:
1997 case CEPH_OSD_OP_WRITEFULL:
1998 WARN_ON(op->indata_len != op->extent.length);
98c4bfe9
ID		 1999 ceph_osdc_msg_data_add(request_msg,
2000 &op->extent.osd_data);
bb873b53
ID		 2001 break;
2002 case CEPH_OSD_OP_SETXATTR:
2003 case CEPH_OSD_OP_CMPXATTR:
2004 WARN_ON(op->indata_len != op->xattr.name_len +
2005 op->xattr.value_len);
98c4bfe9
ID		 2006 ceph_osdc_msg_data_add(request_msg,
2007 &op->xattr.osd_data);
bb873b53 2008 break;
922dab61 2009 case CEPH_OSD_OP_NOTIFY_ACK:
98c4bfe9 2010 ceph_osdc_msg_data_add(request_msg,
922dab61
ID		 2011 &op->notify_ack.request_data);
2012 break;
78beb0ff 2013 case CEPH_OSD_OP_COPY_FROM2:
23ddf9be
LH		 2014 ceph_osdc_msg_data_add(request_msg,
2015 &op->copy_from.osd_data);
2016 break;
bb873b53
ID		 2017
2018 /* reply */
2019 case CEPH_OSD_OP_STAT:
98c4bfe9 2020 ceph_osdc_msg_data_add(reply_msg,
bb873b53
ID		 2021 &op->raw_data_in);
2022 break;
2023 case CEPH_OSD_OP_READ:
98c4bfe9 2024 ceph_osdc_msg_data_add(reply_msg,
bb873b53
ID		 2025 &op->extent.osd_data);
2026 break;
a4ed38d7 2027 case CEPH_OSD_OP_LIST_WATCHERS:
98c4bfe9 2028 ceph_osdc_msg_data_add(reply_msg,
a4ed38d7
DF		 2029 &op->list_watchers.response_data);
2030 break;
bb873b53
ID		 2031
2032 /* both */
2033 case CEPH_OSD_OP_CALL:
2034 WARN_ON(op->indata_len != op->cls.class_len +
2035 op->cls.method_len +
2036 op->cls.indata_len);
98c4bfe9
ID		 2037 ceph_osdc_msg_data_add(request_msg,
2038 &op->cls.request_info);
bb873b53 2039 /* optional, can be NONE */
98c4bfe9
ID		 2040 ceph_osdc_msg_data_add(request_msg,
2041 &op->cls.request_data);
bb873b53 2042 /* optional, can be NONE */
98c4bfe9 2043 ceph_osdc_msg_data_add(reply_msg,
bb873b53
ID		 2044 &op->cls.response_data);
2045 break;
19079203 2046 case CEPH_OSD_OP_NOTIFY:
98c4bfe9 2047 ceph_osdc_msg_data_add(request_msg,
19079203 2048 &op->notify.request_data);
98c4bfe9 2049 ceph_osdc_msg_data_add(reply_msg,
19079203
ID		 2050 &op->notify.response_data);
2051 break;
bb873b53 2052 }
bb873b53 2053 }
bb873b53
ID		 2054}
2055
2e59ffd1
ID		 2056static void encode_pgid(void **p, const struct ceph_pg *pgid)
2057{
2058 ceph_encode_8(p, 1);
2059 ceph_encode_64(p, pgid->pool);
2060 ceph_encode_32(p, pgid->seed);
2061 ceph_encode_32(p, -1); /* preferred */
2062}
2063
8cb441c0
ID		 2064static void encode_spgid(void **p, const struct ceph_spg *spgid)
2065{
2066 ceph_start_encoding(p, 1, 1, CEPH_PGID_ENCODING_LEN + 1);
2067 encode_pgid(p, &spgid->pgid);
2068 ceph_encode_8(p, spgid->shard);
2069}
2070
2e59ffd1
ID		 2071static void encode_oloc(void **p, void *end,
2072 const struct ceph_object_locator *oloc)
2073{
2074 ceph_start_encoding(p, 5, 4, ceph_oloc_encoding_size(oloc));
2075 ceph_encode_64(p, oloc->pool);
2076 ceph_encode_32(p, -1); /* preferred */
2077 ceph_encode_32(p, 0); /* key len */
2078 if (oloc->pool_ns)
2079 ceph_encode_string(p, end, oloc->pool_ns->str,
2080 oloc->pool_ns->len);
2081 else
2082 ceph_encode_32(p, 0);
2083}
2084
8cb441c0
ID		 2085static void encode_request_partial(struct ceph_osd_request *req,
2086 struct ceph_msg *msg)
bb873b53
ID		 2087{
2088 void *p = msg->front.iov_base;
2089 void *const end = p + msg->front_alloc_len;
2090 u32 data_len = 0;
2091 int i;
2092
2093 if (req->r_flags & CEPH_OSD_FLAG_WRITE) {
2094 /* snapshots aren't writeable */
2095 WARN_ON(req->r_snapid != CEPH_NOSNAP);
2096 } else {
2097 WARN_ON(req->r_mtime.tv_sec || req->r_mtime.tv_nsec ||
2098 req->r_data_offset || req->r_snapc);
2099 }
2100
98c4bfe9 2101 setup_request_data(req);
bb873b53 2102
8cb441c0
ID		 2103 encode_spgid(&p, &req->r_t.spgid); /* actual spg */
2104 ceph_encode_32(&p, req->r_t.pgid.seed); /* raw hash */
bb873b53
ID		 2105 ceph_encode_32(&p, req->r_osdc->osdmap->epoch);
2106 ceph_encode_32(&p, req->r_flags);
8cb441c0
ID		 2107
2108 /* reqid */
2109 ceph_start_encoding(&p, 2, 2, sizeof(struct ceph_osd_reqid));
2110 memset(p, 0, sizeof(struct ceph_osd_reqid));
2111 p += sizeof(struct ceph_osd_reqid);
2112
2113 /* trace */
2114 memset(p, 0, sizeof(struct ceph_blkin_trace_info));
2115 p += sizeof(struct ceph_blkin_trace_info);
2116
2117 ceph_encode_32(&p, 0); /* client_inc, always 0 */
fac02ddf 2118 ceph_encode_timespec64(p, &req->r_mtime);
bb873b53 2119 p += sizeof(struct ceph_timespec);
aa26d662 2120
2e59ffd1 2121 encode_oloc(&p, end, &req->r_t.target_oloc);
2e59ffd1
ID		 2122 ceph_encode_string(&p, end, req->r_t.target_oid.name,
2123 req->r_t.target_oid.name_len);
f24e9980 2124
bb873b53
ID		 2125 /* ops, can imply data */
2126 ceph_encode_16(&p, req->r_num_ops);
2127 for (i = 0; i < req->r_num_ops; i++) {
2128 data_len += osd_req_encode_op(p, &req->r_ops[i]);
2129 p += sizeof(struct ceph_osd_op);
2130 }
26be8808 2131
bb873b53
ID		 2132 ceph_encode_64(&p, req->r_snapid); /* snapid */
2133 if (req->r_snapc) {
2134 ceph_encode_64(&p, req->r_snapc->seq);
2135 ceph_encode_32(&p, req->r_snapc->num_snaps);
2136 for (i = 0; i < req->r_snapc->num_snaps; i++)
2137 ceph_encode_64(&p, req->r_snapc->snaps[i]);
2138 } else {
2139 ceph_encode_64(&p, 0); /* snap_seq */
2140 ceph_encode_32(&p, 0); /* snaps len */
2141 }
2142
2143 ceph_encode_32(&p, req->r_attempts); /* retry_attempt */
986e8989 2144 BUG_ON(p > end - 8); /* space for features */
bb873b53 2145
8cb441c0
ID		 2146 msg->hdr.version = cpu_to_le16(8); /* MOSDOp v8 */
2147 /* front_len is finalized in encode_request_finish() */
986e8989
ID		 2148 msg->front.iov_len = p - msg->front.iov_base;
2149 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
bb873b53
ID		 2150 msg->hdr.data_len = cpu_to_le32(data_len);
2151 /*
2152 * The header "data_off" is a hint to the receiver allowing it
2153 * to align received data into its buffers such that there's no
2154 * need to re-copy it before writing it to disk (direct I/O).
2155 */
2156 msg->hdr.data_off = cpu_to_le16(req->r_data_offset);
26be8808 2157
8cb441c0
ID		 2158 dout("%s req %p msg %p oid %s oid_len %d\n", __func__, req, msg,
2159 req->r_t.target_oid.name, req->r_t.target_oid.name_len);
2160}
2161
2162static void encode_request_finish(struct ceph_msg *msg)
2163{
2164 void *p = msg->front.iov_base;
986e8989 2165 void *const partial_end = p + msg->front.iov_len;
8cb441c0
ID		 2166 void *const end = p + msg->front_alloc_len;
2167
2168 if (CEPH_HAVE_FEATURE(msg->con->peer_features, RESEND_ON_SPLIT)) {
2169 /* luminous OSD -- encode features and be done */
986e8989 2170 p = partial_end;
8cb441c0
ID		 2171 ceph_encode_64(&p, msg->con->peer_features);
2172 } else {
2173 struct {
2174 char spgid[CEPH_ENCODING_START_BLK_LEN +
2175 CEPH_PGID_ENCODING_LEN + 1];
2176 __le32 hash;
2177 __le32 epoch;
2178 __le32 flags;
2179 char reqid[CEPH_ENCODING_START_BLK_LEN +
2180 sizeof(struct ceph_osd_reqid)];
2181 char trace[sizeof(struct ceph_blkin_trace_info)];
2182 __le32 client_inc;
2183 struct ceph_timespec mtime;
2184 } __packed head;
2185 struct ceph_pg pgid;
2186 void *oloc, *oid, *tail;
2187 int oloc_len, oid_len, tail_len;
2188 int len;
2189
2190 /*
2191 * Pre-luminous OSD -- reencode v8 into v4 using @head
2192 * as a temporary buffer. Encode the raw PG; the rest
2193 * is just a matter of moving oloc, oid and tail blobs
2194 * around.
2195 */
2196 memcpy(&head, p, sizeof(head));
2197 p += sizeof(head);
2198
2199 oloc = p;
2200 p += CEPH_ENCODING_START_BLK_LEN;
2201 pgid.pool = ceph_decode_64(&p);
2202 p += 4 + 4; /* preferred, key len */
2203 len = ceph_decode_32(&p);
2204 p += len; /* nspace */
2205 oloc_len = p - oloc;
2206
2207 oid = p;
2208 len = ceph_decode_32(&p);
2209 p += len;
2210 oid_len = p - oid;
2211
2212 tail = p;
986e8989 2213 tail_len = partial_end - p;
8cb441c0
ID		 2214
2215 p = msg->front.iov_base;
2216 ceph_encode_copy(&p, &head.client_inc, sizeof(head.client_inc));
2217 ceph_encode_copy(&p, &head.epoch, sizeof(head.epoch));
2218 ceph_encode_copy(&p, &head.flags, sizeof(head.flags));
2219 ceph_encode_copy(&p, &head.mtime, sizeof(head.mtime));
2220
2221 /* reassert_version */
2222 memset(p, 0, sizeof(struct ceph_eversion));
2223 p += sizeof(struct ceph_eversion);
2224
2225 BUG_ON(p >= oloc);
2226 memmove(p, oloc, oloc_len);
2227 p += oloc_len;
2228
2229 pgid.seed = le32_to_cpu(head.hash);
2230 encode_pgid(&p, &pgid); /* raw pg */
2231
2232 BUG_ON(p >= oid);
2233 memmove(p, oid, oid_len);
2234 p += oid_len;
2235
2236 /* tail -- ops, snapid, snapc, retry_attempt */
2237 BUG_ON(p >= tail);
2238 memmove(p, tail, tail_len);
2239 p += tail_len;
2240
2241 msg->hdr.version = cpu_to_le16(4); /* MOSDOp v4 */
2242 }
2243
2244 BUG_ON(p > end);
2245 msg->front.iov_len = p - msg->front.iov_base;
2246 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2247
2248 dout("%s msg %p tid %llu %u+%u+%u v%d\n", __func__, msg,
2249 le64_to_cpu(msg->hdr.tid), le32_to_cpu(msg->hdr.front_len),
2250 le32_to_cpu(msg->hdr.middle_len), le32_to_cpu(msg->hdr.data_len),
2251 le16_to_cpu(msg->hdr.version));
bb873b53
ID		 2252}
2253
2254/*
2255 * @req has to be assigned a tid and registered.
2256 */
2257static void send_request(struct ceph_osd_request *req)
2258{
2259 struct ceph_osd *osd = req->r_osd;
2260
5aea3dcd 2261 verify_osd_locked(osd);
bb873b53
ID		 2262 WARN_ON(osd->o_osd != req->r_t.osd);
2263
a02a946d
ID		 2264 /* backoff? */
2265 if (should_plug_request(req))
2266 return;
2267
5aea3dcd
ID		 2268 /*
2269 * We may have a previously queued request message hanging
2270 * around. Cancel it to avoid corrupting the msgr.
2271 */
2272 if (req->r_sent)
2273 ceph_msg_revoke(req->r_request);
2274
bb873b53
ID		 2275 req->r_flags |= CEPH_OSD_FLAG_KNOWN_REDIR;
2276 if (req->r_attempts)
2277 req->r_flags |= CEPH_OSD_FLAG_RETRY;
2278 else
2279 WARN_ON(req->r_flags & CEPH_OSD_FLAG_RETRY);
2280
8cb441c0 2281 encode_request_partial(req, req->r_request);
bb873b53 2282
04c7d789 2283 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 2284 __func__, req, req->r_tid, req->r_t.pgid.pool, req->r_t.pgid.seed,
dc98ff72 2285 req->r_t.spgid.pgid.pool, req->r_t.spgid.pgid.seed,
04c7d789
ID		 2286 req->r_t.spgid.shard, osd->o_osd, req->r_t.epoch, req->r_flags,
2287 req->r_attempts);
bb873b53
ID		 2288
2289 req->r_t.paused = false;
2290 req->r_stamp = jiffies;
2291 req->r_attempts++;
2292
2293 req->r_sent = osd->o_incarnation;
2294 req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
2295 ceph_con_send(&osd->o_con, ceph_msg_get(req->r_request));
f24e9980
SW		 2296}
2297
42c1b124
ID		 2298static void maybe_request_map(struct ceph_osd_client *osdc)
2299{
2300 bool continuous = false;
2301
5aea3dcd 2302 verify_osdc_locked(osdc);
42c1b124
ID		 2303 WARN_ON(!osdc->osdmap->epoch);
2304
b7ec35b3
ID		 2305 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2306 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD) ||
2307 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
42c1b124
ID		 2308 dout("%s osdc %p continuous\n", __func__, osdc);
2309 continuous = true;
2310 } else {
2311 dout("%s osdc %p onetime\n", __func__, osdc);
2312 }
2313
2314 if (ceph_monc_want_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
2315 osdc->osdmap->epoch + 1, continuous))
2316 ceph_monc_renew_subs(&osdc->client->monc);
2317}
2318
a1f4020a 2319static void complete_request(struct ceph_osd_request *req, int err);
4609245e
ID		 2320static void send_map_check(struct ceph_osd_request *req);
2321
5aea3dcd 2322static void __submit_request(struct ceph_osd_request *req, bool wrlocked)
0bbfdfe8 2323{
5aea3dcd
ID		 2324 struct ceph_osd_client *osdc = req->r_osdc;
2325 struct ceph_osd *osd;
4609245e 2326 enum calc_target_result ct_res;
66850df5 2327 int err = 0;
5aea3dcd
ID		 2328 bool need_send = false;
2329 bool promoted = false;
0bbfdfe8 2330
b18b9550 2331 WARN_ON(req->r_tid);
5aea3dcd
ID		 2332 dout("%s req %p wrlocked %d\n", __func__, req, wrlocked);
2333
2334again:
8edf84ba 2335 ct_res = calc_target(osdc, &req->r_t, false);
4609245e
ID		 2336 if (ct_res == CALC_TARGET_POOL_DNE && !wrlocked)
2337 goto promote;
2338
5aea3dcd
ID		 2339 osd = lookup_create_osd(osdc, req->r_t.osd, wrlocked);
2340 if (IS_ERR(osd)) {
2341 WARN_ON(PTR_ERR(osd) != -EAGAIN || wrlocked);
2342 goto promote;
0bbfdfe8
ID		 2343 }
2344
66850df5
ID		 2345 if (osdc->abort_err) {
2346 dout("req %p abort_err %d\n", req, osdc->abort_err);
2347 err = osdc->abort_err;
2348 } else if (osdc->osdmap->epoch < osdc->epoch_barrier) {
58eb7932
JL		 2349 dout("req %p epoch %u barrier %u\n", req, osdc->osdmap->epoch,
2350 osdc->epoch_barrier);
2351 req->r_t.paused = true;
2352 maybe_request_map(osdc);
2353 } else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2354 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
5aea3dcd
ID		 2355 dout("req %p pausewr\n", req);
2356 req->r_t.paused = true;
2357 maybe_request_map(osdc);
2358 } else if ((req->r_flags & CEPH_OSD_FLAG_READ) &&
b7ec35b3 2359 ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
5aea3dcd
ID		 2360 dout("req %p pauserd\n", req);
2361 req->r_t.paused = true;
2362 maybe_request_map(osdc);
2363 } else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2364 !(req->r_flags & (CEPH_OSD_FLAG_FULL_TRY |
2365 CEPH_OSD_FLAG_FULL_FORCE)) &&
b7ec35b3 2366 (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
5aea3dcd
ID		 2367 pool_full(osdc, req->r_t.base_oloc.pool))) {
2368 dout("req %p full/pool_full\n", req);
02b2f549 2369 if (ceph_test_opt(osdc->client, ABORT_ON_FULL)) {
66850df5 2370 err = -ENOSPC;
29e87820 2371 } else {
dc9b0dc4
ID		 2372 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL))
2373 pr_warn_ratelimited("cluster is full (osdmap FULL)\n");
2374 else
2375 pr_warn_ratelimited("pool %lld is full or reached quota\n",
2376 req->r_t.base_oloc.pool);
29e87820
ID		 2377 req->r_t.paused = true;
2378 maybe_request_map(osdc);
2379 }
5aea3dcd
ID		 2380 } else if (!osd_homeless(osd)) {
2381 need_send = true;
0bbfdfe8 2382 } else {
5aea3dcd 2383 maybe_request_map(osdc);
0bbfdfe8
ID		 2384 }
2385
5aea3dcd
ID		 2386 mutex_lock(&osd->lock);
2387 /*
2388 * Assign the tid atomically with send_request() to protect
2389 * multiple writes to the same object from racing with each
2390 * other, resulting in out of order ops on the OSDs.
2391 */
2392 req->r_tid = atomic64_inc_return(&osdc->last_tid);
2393 link_request(osd, req);
2394 if (need_send)
2395 send_request(req);
66850df5
ID		 2396 else if (err)
2397 complete_request(req, err);
5aea3dcd
ID		 2398 mutex_unlock(&osd->lock);
2399
6001567c 2400 if (!err && ct_res == CALC_TARGET_POOL_DNE)
4609245e
ID		 2401 send_map_check(req);
2402
5aea3dcd
ID		 2403 if (promoted)
2404 downgrade_write(&osdc->lock);
2405 return;
2406
2407promote:
2408 up_read(&osdc->lock);
2409 down_write(&osdc->lock);
2410 wrlocked = true;
2411 promoted = true;
2412 goto again;
2413}
2414
2415static void account_request(struct ceph_osd_request *req)
2416{
54ea0046 2417 WARN_ON(req->r_flags & (CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK));
b18b9550 2418 WARN_ON(!(req->r_flags & (CEPH_OSD_FLAG_READ | CEPH_OSD_FLAG_WRITE)));
5aea3dcd 2419
b18b9550 2420 req->r_flags |= CEPH_OSD_FLAG_ONDISK;
22d2cfdf 2421 atomic_inc(&req->r_osdc->num_requests);
7cc5e38f
ID		 2422
2423 req->r_start_stamp = jiffies;
97e27aaa 2424 req->r_start_latency = ktime_get();
5aea3dcd
ID		 2425}
2426
2427static void submit_request(struct ceph_osd_request *req, bool wrlocked)
2428{
2429 ceph_osdc_get_request(req);
2430 account_request(req);
2431 __submit_request(req, wrlocked);
2432}
2433
45ee2c1d 2434static void finish_request(struct ceph_osd_request *req)
5aea3dcd
ID		 2435{
2436 struct ceph_osd_client *osdc = req->r_osdc;
5aea3dcd 2437
04c7d789 2438 WARN_ON(lookup_request_mc(&osdc->map_checks, req->r_tid));
5aea3dcd
ID		 2439 dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2440
97e27aaa
XL		 2441 req->r_end_latency = ktime_get();
2442
04c7d789
ID		 2443 if (req->r_osd)
2444 unlink_request(req->r_osd, req);
5aea3dcd
ID		 2445 atomic_dec(&osdc->num_requests);
2446
2447 /*
2448 * If an OSD has failed or returned and a request has been sent
2449 * twice, it's possible to get a reply and end up here while the
2450 * request message is queued for delivery. We will ignore the
2451 * reply, so not a big deal, but better to try and catch it.
2452 */
2453 ceph_msg_revoke(req->r_request);
2454 ceph_msg_revoke_incoming(req->r_reply);
2455}
2456
fe5da05e
ID		 2457static void __complete_request(struct ceph_osd_request *req)
2458{
d75f773c 2459 dout("%s req %p tid %llu cb %ps result %d\n", __func__, req,
26df726b
ID		 2460 req->r_tid, req->r_callback, req->r_result);
2461
2462 if (req->r_callback)
fe5da05e 2463 req->r_callback(req);
26df726b
ID		 2464 complete_all(&req->r_completion);
2465 ceph_osdc_put_request(req);
fe5da05e
ID		 2466}
2467
88bc1922
ID		 2468static void complete_request_workfn(struct work_struct *work)
2469{
2470 struct ceph_osd_request *req =
2471 container_of(work, struct ceph_osd_request, r_complete_work);
2472
2473 __complete_request(req);
fe5da05e
ID		 2474}
2475
4609245e 2476/*
b18b9550 2477 * This is open-coded in handle_reply().
4609245e
ID		 2478 */
2479static void complete_request(struct ceph_osd_request *req, int err)
2480{
2481 dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2482
2483 req->r_result = err;
45ee2c1d 2484 finish_request(req);
88bc1922
ID		 2485
2486 INIT_WORK(&req->r_complete_work, complete_request_workfn);
2487 queue_work(req->r_osdc->completion_wq, &req->r_complete_work);
4609245e
ID		 2488}
2489
2490static void cancel_map_check(struct ceph_osd_request *req)
2491{
2492 struct ceph_osd_client *osdc = req->r_osdc;
2493 struct ceph_osd_request *lookup_req;
2494
2495 verify_osdc_wrlocked(osdc);
2496
2497 lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2498 if (!lookup_req)
2499 return;
2500
2501 WARN_ON(lookup_req != req);
2502 erase_request_mc(&osdc->map_checks, req);
2503 ceph_osdc_put_request(req);
2504}
2505
5aea3dcd
ID		 2506static void cancel_request(struct ceph_osd_request *req)
2507{
2508 dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2509
4609245e 2510 cancel_map_check(req);
45ee2c1d 2511 finish_request(req);
b18b9550 2512 complete_all(&req->r_completion);
c297eb42 2513 ceph_osdc_put_request(req);
5aea3dcd
ID		 2514}
2515
7cc5e38f
ID		 2516static void abort_request(struct ceph_osd_request *req, int err)
2517{
2518 dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2519
2520 cancel_map_check(req);
2521 complete_request(req, err);
2522}
2523
66850df5
ID		 2524static int abort_fn(struct ceph_osd_request *req, void *arg)
2525{
2526 int err = *(int *)arg;
2527
2528 abort_request(req, err);
2529 return 0; /* continue iteration */
2530}
2531
2532/*
2533 * Abort all in-flight requests with @err and arrange for all future
2534 * requests to be failed immediately.
2535 */
2536void ceph_osdc_abort_requests(struct ceph_osd_client *osdc, int err)
2537{
2538 dout("%s osdc %p err %d\n", __func__, osdc, err);
2539 down_write(&osdc->lock);
2540 for_each_request(osdc, abort_fn, &err);
2541 osdc->abort_err = err;
2542 up_write(&osdc->lock);
2543}
2544EXPORT_SYMBOL(ceph_osdc_abort_requests);
2545
2cef0ba8
YZ		 2546void ceph_osdc_clear_abort_err(struct ceph_osd_client *osdc)
2547{
2548 down_write(&osdc->lock);
2549 osdc->abort_err = 0;
2550 up_write(&osdc->lock);
2551}
2552EXPORT_SYMBOL(ceph_osdc_clear_abort_err);
2553
58eb7932
JL		 2554static void update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2555{
2556 if (likely(eb > osdc->epoch_barrier)) {
2557 dout("updating epoch_barrier from %u to %u\n",
2558 osdc->epoch_barrier, eb);
2559 osdc->epoch_barrier = eb;
2560 /* Request map if we're not to the barrier yet */
2561 if (eb > osdc->osdmap->epoch)
2562 maybe_request_map(osdc);
2563 }
2564}
2565
2566void ceph_osdc_update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2567{
2568 down_read(&osdc->lock);
2569 if (unlikely(eb > osdc->epoch_barrier)) {
2570 up_read(&osdc->lock);
2571 down_write(&osdc->lock);
2572 update_epoch_barrier(osdc, eb);
2573 up_write(&osdc->lock);
2574 } else {
2575 up_read(&osdc->lock);
2576 }
2577}
2578EXPORT_SYMBOL(ceph_osdc_update_epoch_barrier);
2579
4eea0fef
ID		 2580/*
2581 * We can end up releasing caps as a result of abort_request().
2582 * In that case, we probably want to ensure that the cap release message
2583 * has an updated epoch barrier in it, so set the epoch barrier prior to
2584 * aborting the first request.
2585 */
2586static int abort_on_full_fn(struct ceph_osd_request *req, void *arg)
2587{
2588 struct ceph_osd_client *osdc = req->r_osdc;
2589 bool *victims = arg;
2590
c843d13c 2591 if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
4eea0fef 2592 (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
690f951d 2593 pool_full(osdc, req->r_t.base_oloc.pool))) {
4eea0fef
ID		 2594 if (!*victims) {
2595 update_epoch_barrier(osdc, osdc->osdmap->epoch);
2596 *victims = true;
2597 }
2598 abort_request(req, -ENOSPC);
2599 }
2600
2601 return 0; /* continue iteration */
2602}
2603
fc36d0a4
JL		 2604/*
2605 * Drop all pending requests that are stalled waiting on a full condition to
58eb7932
JL		 2606 * clear, and complete them with ENOSPC as the return code. Set the
2607 * osdc->epoch_barrier to the latest map epoch that we've seen if any were
2608 * cancelled.
fc36d0a4
JL		 2609 */
2610static void ceph_osdc_abort_on_full(struct ceph_osd_client *osdc)
2611{
58eb7932 2612 bool victims = false;
fc36d0a4 2613
02b2f549 2614 if (ceph_test_opt(osdc->client, ABORT_ON_FULL) &&
c843d13c 2615 (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || have_pool_full(osdc)))
4eea0fef 2616 for_each_request(osdc, abort_on_full_fn, &victims);
fc36d0a4
JL		 2617}
2618
4609245e
ID		 2619static void check_pool_dne(struct ceph_osd_request *req)
2620{
2621 struct ceph_osd_client *osdc = req->r_osdc;
2622 struct ceph_osdmap *map = osdc->osdmap;
2623
2624 verify_osdc_wrlocked(osdc);
2625 WARN_ON(!map->epoch);
2626
2627 if (req->r_attempts) {
2628 /*
2629 * We sent a request earlier, which means that
2630 * previously the pool existed, and now it does not
2631 * (i.e., it was deleted).
2632 */
2633 req->r_map_dne_bound = map->epoch;
2634 dout("%s req %p tid %llu pool disappeared\n", __func__, req,
2635 req->r_tid);
2636 } else {
2637 dout("%s req %p tid %llu map_dne_bound %u have %u\n", __func__,
2638 req, req->r_tid, req->r_map_dne_bound, map->epoch);
2639 }
2640
2641 if (req->r_map_dne_bound) {
2642 if (map->epoch >= req->r_map_dne_bound) {
2643 /* we had a new enough map */
2644 pr_info_ratelimited("tid %llu pool does not exist\n",
2645 req->r_tid);
2646 complete_request(req, -ENOENT);
2647 }
2648 } else {
2649 send_map_check(req);
2650 }
2651}
2652
2653static void map_check_cb(struct ceph_mon_generic_request *greq)
2654{
2655 struct ceph_osd_client *osdc = &greq->monc->client->osdc;
2656 struct ceph_osd_request *req;
2657 u64 tid = greq->private_data;
2658
2659 WARN_ON(greq->result || !greq->u.newest);
2660
2661 down_write(&osdc->lock);
2662 req = lookup_request_mc(&osdc->map_checks, tid);
2663 if (!req) {
2664 dout("%s tid %llu dne\n", __func__, tid);
2665 goto out_unlock;
2666 }
2667
2668 dout("%s req %p tid %llu map_dne_bound %u newest %llu\n", __func__,
2669 req, req->r_tid, req->r_map_dne_bound, greq->u.newest);
2670 if (!req->r_map_dne_bound)
2671 req->r_map_dne_bound = greq->u.newest;
2672 erase_request_mc(&osdc->map_checks, req);
2673 check_pool_dne(req);
2674
2675 ceph_osdc_put_request(req);
2676out_unlock:
2677 up_write(&osdc->lock);
2678}
2679
2680static void send_map_check(struct ceph_osd_request *req)
2681{
2682 struct ceph_osd_client *osdc = req->r_osdc;
2683 struct ceph_osd_request *lookup_req;
2684 int ret;
2685
2686 verify_osdc_wrlocked(osdc);
2687
2688 lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2689 if (lookup_req) {
2690 WARN_ON(lookup_req != req);
2691 return;
2692 }
2693
2694 ceph_osdc_get_request(req);
2695 insert_request_mc(&osdc->map_checks, req);
2696 ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
2697 map_check_cb, req->r_tid);
2698 WARN_ON(ret);
2699}
2700
922dab61
ID		 2701/*
2702 * lingering requests, watch/notify v2 infrastructure
2703 */
2704static void linger_release(struct kref *kref)
2705{
2706 struct ceph_osd_linger_request *lreq =
2707 container_of(kref, struct ceph_osd_linger_request, kref);
2708
2709 dout("%s lreq %p reg_req %p ping_req %p\n", __func__, lreq,
2710 lreq->reg_req, lreq->ping_req);
2711 WARN_ON(!RB_EMPTY_NODE(&lreq->node));
2712 WARN_ON(!RB_EMPTY_NODE(&lreq->osdc_node));
4609245e 2713 WARN_ON(!RB_EMPTY_NODE(&lreq->mc_node));
922dab61 2714 WARN_ON(!list_empty(&lreq->scan_item));
b07d3c4b 2715 WARN_ON(!list_empty(&lreq->pending_lworks));
922dab61
ID		 2716 WARN_ON(lreq->osd);
2717
75dbb685
ID		 2718 if (lreq->request_pl)
2719 ceph_pagelist_release(lreq->request_pl);
2720 if (lreq->notify_id_pages)
2721 ceph_release_page_vector(lreq->notify_id_pages, 1);
2722
2723 ceph_osdc_put_request(lreq->reg_req);
2724 ceph_osdc_put_request(lreq->ping_req);
922dab61
ID		 2725 target_destroy(&lreq->t);
2726 kfree(lreq);
2727}
2728
2729static void linger_put(struct ceph_osd_linger_request *lreq)
2730{
2731 if (lreq)
2732 kref_put(&lreq->kref, linger_release);
2733}
2734
2735static struct ceph_osd_linger_request *
2736linger_get(struct ceph_osd_linger_request *lreq)
2737{
2738 kref_get(&lreq->kref);
2739 return lreq;
2740}
2741
2742static struct ceph_osd_linger_request *
2743linger_alloc(struct ceph_osd_client *osdc)
2744{
2745 struct ceph_osd_linger_request *lreq;
2746
2747 lreq = kzalloc(sizeof(*lreq), GFP_NOIO);
2748 if (!lreq)
2749 return NULL;
2750
2751 kref_init(&lreq->kref);
2752 mutex_init(&lreq->lock);
2753 RB_CLEAR_NODE(&lreq->node);
2754 RB_CLEAR_NODE(&lreq->osdc_node);
4609245e 2755 RB_CLEAR_NODE(&lreq->mc_node);
922dab61 2756 INIT_LIST_HEAD(&lreq->scan_item);
b07d3c4b 2757 INIT_LIST_HEAD(&lreq->pending_lworks);
922dab61 2758 init_completion(&lreq->reg_commit_wait);
19079203 2759 init_completion(&lreq->notify_finish_wait);
922dab61
ID		 2760
2761 lreq->osdc = osdc;
2762 target_init(&lreq->t);
2763
2764 dout("%s lreq %p\n", __func__, lreq);
2765 return lreq;
2766}
2767
2768DEFINE_RB_INSDEL_FUNCS(linger, struct ceph_osd_linger_request, linger_id, node)
2769DEFINE_RB_FUNCS(linger_osdc, struct ceph_osd_linger_request, linger_id, osdc_node)
4609245e 2770DEFINE_RB_FUNCS(linger_mc, struct ceph_osd_linger_request, linger_id, mc_node)
922dab61
ID		 2771
2772/*
2773 * Create linger request <-> OSD session relation.
2774 *
2775 * @lreq has to be registered, @osd may be homeless.
2776 */
2777static void link_linger(struct ceph_osd *osd,
2778 struct ceph_osd_linger_request *lreq)
2779{
2780 verify_osd_locked(osd);
2781 WARN_ON(!lreq->linger_id || lreq->osd);
2782 dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2783 osd->o_osd, lreq, lreq->linger_id);
2784
2785 if (!osd_homeless(osd))
2786 __remove_osd_from_lru(osd);
2787 else
2788 atomic_inc(&osd->o_osdc->num_homeless);
2789
2790 get_osd(osd);
2791 insert_linger(&osd->o_linger_requests, lreq);
2792 lreq->osd = osd;
2793}
2794
2795static void unlink_linger(struct ceph_osd *osd,
2796 struct ceph_osd_linger_request *lreq)
2797{
2798 verify_osd_locked(osd);
2799 WARN_ON(lreq->osd != osd);
2800 dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2801 osd->o_osd, lreq, lreq->linger_id);
2802
2803 lreq->osd = NULL;
2804 erase_linger(&osd->o_linger_requests, lreq);
2805 put_osd(osd);
2806
2807 if (!osd_homeless(osd))
2808 maybe_move_osd_to_lru(osd);
2809 else
2810 atomic_dec(&osd->o_osdc->num_homeless);
2811}
2812
2813static bool __linger_registered(struct ceph_osd_linger_request *lreq)
2814{
2815 verify_osdc_locked(lreq->osdc);
2816
2817 return !RB_EMPTY_NODE(&lreq->osdc_node);
2818}
2819
2820static bool linger_registered(struct ceph_osd_linger_request *lreq)
2821{
2822 struct ceph_osd_client *osdc = lreq->osdc;
2823 bool registered;
2824
2825 down_read(&osdc->lock);
2826 registered = __linger_registered(lreq);
2827 up_read(&osdc->lock);
2828
2829 return registered;
2830}
2831
2832static void linger_register(struct ceph_osd_linger_request *lreq)
2833{
2834 struct ceph_osd_client *osdc = lreq->osdc;
2835
2836 verify_osdc_wrlocked(osdc);
2837 WARN_ON(lreq->linger_id);
2838
2839 linger_get(lreq);
2840 lreq->linger_id = ++osdc->last_linger_id;
2841 insert_linger_osdc(&osdc->linger_requests, lreq);
2842}
2843
2844static void linger_unregister(struct ceph_osd_linger_request *lreq)
2845{
2846 struct ceph_osd_client *osdc = lreq->osdc;
2847
2848 verify_osdc_wrlocked(osdc);
2849
2850 erase_linger_osdc(&osdc->linger_requests, lreq);
2851 linger_put(lreq);
2852}
2853
2854static void cancel_linger_request(struct ceph_osd_request *req)
2855{
2856 struct ceph_osd_linger_request *lreq = req->r_priv;
2857
2858 WARN_ON(!req->r_linger);
2859 cancel_request(req);
2860 linger_put(lreq);
2861}
2862
2863struct linger_work {
2864 struct work_struct work;
2865 struct ceph_osd_linger_request *lreq;
b07d3c4b
ID		 2866 struct list_head pending_item;
2867 unsigned long queued_stamp;
922dab61
ID		 2868
2869 union {
2870 struct {
2871 u64 notify_id;
2872 u64 notifier_id;
2873 void *payload; /* points into @msg front */
2874 size_t payload_len;
2875
2876 struct ceph_msg *msg; /* for ceph_msg_put() */
2877 } notify;
2878 struct {
2879 int err;
2880 } error;
2881 };
2882};
2883
2884static struct linger_work *lwork_alloc(struct ceph_osd_linger_request *lreq,
2885 work_func_t workfn)
2886{
2887 struct linger_work *lwork;
2888
2889 lwork = kzalloc(sizeof(*lwork), GFP_NOIO);
2890 if (!lwork)
2891 return NULL;
2892
2893 INIT_WORK(&lwork->work, workfn);
b07d3c4b 2894 INIT_LIST_HEAD(&lwork->pending_item);
922dab61
ID		 2895 lwork->lreq = linger_get(lreq);
2896
2897 return lwork;
2898}
2899
2900static void lwork_free(struct linger_work *lwork)
2901{
2902 struct ceph_osd_linger_request *lreq = lwork->lreq;
2903
b07d3c4b
ID		 2904 mutex_lock(&lreq->lock);
2905 list_del(&lwork->pending_item);
2906 mutex_unlock(&lreq->lock);
2907
922dab61
ID		 2908 linger_put(lreq);
2909 kfree(lwork);
2910}
2911
2912static void lwork_queue(struct linger_work *lwork)
2913{
2914 struct ceph_osd_linger_request *lreq = lwork->lreq;
2915 struct ceph_osd_client *osdc = lreq->osdc;
2916
2917 verify_lreq_locked(lreq);
b07d3c4b
ID		 2918 WARN_ON(!list_empty(&lwork->pending_item));
2919
2920 lwork->queued_stamp = jiffies;
2921 list_add_tail(&lwork->pending_item, &lreq->pending_lworks);
922dab61
ID		 2922 queue_work(osdc->notify_wq, &lwork->work);
2923}
2924
2925static void do_watch_notify(struct work_struct *w)
2926{
2927 struct linger_work *lwork = container_of(w, struct linger_work, work);
2928 struct ceph_osd_linger_request *lreq = lwork->lreq;
2929
2930 if (!linger_registered(lreq)) {
2931 dout("%s lreq %p not registered\n", __func__, lreq);
2932 goto out;
2933 }
2934
19079203 2935 WARN_ON(!lreq->is_watch);
922dab61
ID		 2936 dout("%s lreq %p notify_id %llu notifier_id %llu payload_len %zu\n",
2937 __func__, lreq, lwork->notify.notify_id, lwork->notify.notifier_id,
2938 lwork->notify.payload_len);
2939 lreq->wcb(lreq->data, lwork->notify.notify_id, lreq->linger_id,
2940 lwork->notify.notifier_id, lwork->notify.payload,
2941 lwork->notify.payload_len);
2942
2943out:
2944 ceph_msg_put(lwork->notify.msg);
2945 lwork_free(lwork);
2946}
2947
2948static void do_watch_error(struct work_struct *w)
2949{
2950 struct linger_work *lwork = container_of(w, struct linger_work, work);
2951 struct ceph_osd_linger_request *lreq = lwork->lreq;
2952
2953 if (!linger_registered(lreq)) {
2954 dout("%s lreq %p not registered\n", __func__, lreq);
2955 goto out;
2956 }
2957
2958 dout("%s lreq %p err %d\n", __func__, lreq, lwork->error.err);
2959 lreq->errcb(lreq->data, lreq->linger_id, lwork->error.err);
2960
2961out:
2962 lwork_free(lwork);
2963}
2964
2965static void queue_watch_error(struct ceph_osd_linger_request *lreq)
2966{
2967 struct linger_work *lwork;
2968
2969 lwork = lwork_alloc(lreq, do_watch_error);
2970 if (!lwork) {
2971 pr_err("failed to allocate error-lwork\n");
2972 return;
2973 }
2974
2975 lwork->error.err = lreq->last_error;
2976 lwork_queue(lwork);
2977}
2978
2979static void linger_reg_commit_complete(struct ceph_osd_linger_request *lreq,
2980 int result)
2981{
2982 if (!completion_done(&lreq->reg_commit_wait)) {
2983 lreq->reg_commit_error = (result <= 0 ? result : 0);
2984 complete_all(&lreq->reg_commit_wait);
2985 }
2986}
2987
2988static void linger_commit_cb(struct ceph_osd_request *req)
2989{
2990 struct ceph_osd_linger_request *lreq = req->r_priv;
2991
2992 mutex_lock(&lreq->lock);
75dbb685
ID		 2993 if (req != lreq->reg_req) {
2994 dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n",
2995 __func__, lreq, lreq->linger_id, req, lreq->reg_req);
2996 goto out;
2997 }
2998
922dab61
ID		 2999 dout("%s lreq %p linger_id %llu result %d\n", __func__, lreq,
3000 lreq->linger_id, req->r_result);
922dab61
ID		 3001 linger_reg_commit_complete(lreq, req->r_result);
3002 lreq->committed = true;
3003
19079203
ID		 3004 if (!lreq->is_watch) {
3005 struct ceph_osd_data *osd_data =
3006 osd_req_op_data(req, 0, notify, response_data);
3007 void *p = page_address(osd_data->pages[0]);
3008
3009 WARN_ON(req->r_ops[0].op != CEPH_OSD_OP_NOTIFY ||
3010 osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
3011
3012 /* make note of the notify_id */
3013 if (req->r_ops[0].outdata_len >= sizeof(u64)) {
3014 lreq->notify_id = ceph_decode_64(&p);
3015 dout("lreq %p notify_id %llu\n", lreq,
3016 lreq->notify_id);
3017 } else {
3018 dout("lreq %p no notify_id\n", lreq);
3019 }
3020 }
3021
75dbb685 3022out:
922dab61
ID		 3023 mutex_unlock(&lreq->lock);
3024 linger_put(lreq);
3025}
3026
3027static int normalize_watch_error(int err)
3028{
3029 /*
3030 * Translate ENOENT -> ENOTCONN so that a delete->disconnection
3031 * notification and a failure to reconnect because we raced with
3032 * the delete appear the same to the user.
3033 */
3034 if (err == -ENOENT)
3035 err = -ENOTCONN;
3036
3037 return err;
3038}
3039
3040static void linger_reconnect_cb(struct ceph_osd_request *req)
3041{
3042 struct ceph_osd_linger_request *lreq = req->r_priv;
3043
3044 mutex_lock(&lreq->lock);
75dbb685
ID		 3045 if (req != lreq->reg_req) {
3046 dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n",
3047 __func__, lreq, lreq->linger_id, req, lreq->reg_req);
3048 goto out;
3049 }
3050
922dab61
ID		 3051 dout("%s lreq %p linger_id %llu result %d last_error %d\n", __func__,
3052 lreq, lreq->linger_id, req->r_result, lreq->last_error);
3053 if (req->r_result < 0) {
3054 if (!lreq->last_error) {
3055 lreq->last_error = normalize_watch_error(req->r_result);
3056 queue_watch_error(lreq);
3057 }
3058 }
3059
75dbb685 3060out:
922dab61
ID		 3061 mutex_unlock(&lreq->lock);
3062 linger_put(lreq);
3063}
3064
3065static void send_linger(struct ceph_osd_linger_request *lreq)
3066{
75dbb685
ID		 3067 struct ceph_osd_client *osdc = lreq->osdc;
3068 struct ceph_osd_request *req;
3069 int ret;
922dab61 3070
75dbb685
ID		 3071 verify_osdc_wrlocked(osdc);
3072 mutex_lock(&lreq->lock);
922dab61
ID		 3073 dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3074
75dbb685
ID		 3075 if (lreq->reg_req) {
3076 if (lreq->reg_req->r_osd)
3077 cancel_linger_request(lreq->reg_req);
3078 ceph_osdc_put_request(lreq->reg_req);
3079 }
3080
3081 req = ceph_osdc_alloc_request(osdc, NULL, 1, true, GFP_NOIO);
3082 BUG_ON(!req);
922dab61 3083
5133ba8f 3084 target_copy(&req->r_t, &lreq->t);
922dab61
ID		 3085 req->r_mtime = lreq->mtime;
3086
19079203 3087 if (lreq->is_watch && lreq->committed) {
75dbb685
ID		 3088 osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_RECONNECT,
3089 lreq->linger_id, ++lreq->register_gen);
922dab61 3090 dout("lreq %p reconnect register_gen %u\n", lreq,
75dbb685 3091 req->r_ops[0].watch.gen);
922dab61
ID		 3092 req->r_callback = linger_reconnect_cb;
3093 } else {
75dbb685
ID		 3094 if (lreq->is_watch) {
3095 osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_WATCH,
3096 lreq->linger_id, 0);
3097 } else {
19079203 3098 lreq->notify_id = 0;
75dbb685
ID		 3099
3100 refcount_inc(&lreq->request_pl->refcnt);
3101 osd_req_op_notify_init(req, 0, lreq->linger_id,
3102 lreq->request_pl);
3103 ceph_osd_data_pages_init(
3104 osd_req_op_data(req, 0, notify, response_data),
3105 lreq->notify_id_pages, PAGE_SIZE, 0, false, false);
3106 }
922dab61
ID		 3107 dout("lreq %p register\n", lreq);
3108 req->r_callback = linger_commit_cb;
3109 }
75dbb685
ID		 3110
3111 ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
3112 BUG_ON(ret);
922dab61
ID		 3113
3114 req->r_priv = linger_get(lreq);
3115 req->r_linger = true;
75dbb685
ID		 3116 lreq->reg_req = req;
3117 mutex_unlock(&lreq->lock);
922dab61
ID		 3118
3119 submit_request(req, true);
3120}
3121
3122static void linger_ping_cb(struct ceph_osd_request *req)
3123{
3124 struct ceph_osd_linger_request *lreq = req->r_priv;
3125
3126 mutex_lock(&lreq->lock);
75dbb685
ID		 3127 if (req != lreq->ping_req) {
3128 dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n",
3129 __func__, lreq, lreq->linger_id, req, lreq->ping_req);
3130 goto out;
3131 }
3132
922dab61
ID		 3133 dout("%s lreq %p linger_id %llu result %d ping_sent %lu last_error %d\n",
3134 __func__, lreq, lreq->linger_id, req->r_result, lreq->ping_sent,
3135 lreq->last_error);
3136 if (lreq->register_gen == req->r_ops[0].watch.gen) {
b07d3c4b
ID		 3137 if (!req->r_result) {
3138 lreq->watch_valid_thru = lreq->ping_sent;
3139 } else if (!lreq->last_error) {
922dab61
ID		 3140 lreq->last_error = normalize_watch_error(req->r_result);
3141 queue_watch_error(lreq);
3142 }
3143 } else {
3144 dout("lreq %p register_gen %u ignoring old pong %u\n", lreq,
3145 lreq->register_gen, req->r_ops[0].watch.gen);
3146 }
3147
75dbb685 3148out:
922dab61
ID		 3149 mutex_unlock(&lreq->lock);
3150 linger_put(lreq);
3151}
3152
3153static void send_linger_ping(struct ceph_osd_linger_request *lreq)
3154{
3155 struct ceph_osd_client *osdc = lreq->osdc;
75dbb685
ID		 3156 struct ceph_osd_request *req;
3157 int ret;
922dab61 3158
b7ec35b3 3159 if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
922dab61
ID		 3160 dout("%s PAUSERD\n", __func__);
3161 return;
3162 }
3163
3164 lreq->ping_sent = jiffies;
3165 dout("%s lreq %p linger_id %llu ping_sent %lu register_gen %u\n",
3166 __func__, lreq, lreq->linger_id, lreq->ping_sent,
3167 lreq->register_gen);
3168
75dbb685
ID		 3169 if (lreq->ping_req) {
3170 if (lreq->ping_req->r_osd)
3171 cancel_linger_request(lreq->ping_req);
3172 ceph_osdc_put_request(lreq->ping_req);
3173 }
922dab61 3174
75dbb685
ID		 3175 req = ceph_osdc_alloc_request(osdc, NULL, 1, true, GFP_NOIO);
3176 BUG_ON(!req);
922dab61 3177
75dbb685
ID		 3178 target_copy(&req->r_t, &lreq->t);
3179 osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_PING, lreq->linger_id,
3180 lreq->register_gen);
922dab61 3181 req->r_callback = linger_ping_cb;
75dbb685
ID		 3182
3183 ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
3184 BUG_ON(ret);
3185
922dab61
ID		 3186 req->r_priv = linger_get(lreq);
3187 req->r_linger = true;
75dbb685 3188 lreq->ping_req = req;
922dab61
ID		 3189
3190 ceph_osdc_get_request(req);
3191 account_request(req);
3192 req->r_tid = atomic64_inc_return(&osdc->last_tid);
3193 link_request(lreq->osd, req);
3194 send_request(req);
3195}
3196
3197static void linger_submit(struct ceph_osd_linger_request *lreq)
3198{
3199 struct ceph_osd_client *osdc = lreq->osdc;
3200 struct ceph_osd *osd;
3201
81c65213
ID		 3202 down_write(&osdc->lock);
3203 linger_register(lreq);
81c65213 3204
8edf84ba 3205 calc_target(osdc, &lreq->t, false);
922dab61
ID		 3206 osd = lookup_create_osd(osdc, lreq->t.osd, true);
3207 link_linger(osd, lreq);
3208
3209 send_linger(lreq);
81c65213 3210 up_write(&osdc->lock);
922dab61
ID		 3211}
3212
4609245e
ID		 3213static void cancel_linger_map_check(struct ceph_osd_linger_request *lreq)
3214{
3215 struct ceph_osd_client *osdc = lreq->osdc;
3216 struct ceph_osd_linger_request *lookup_lreq;
3217
3218 verify_osdc_wrlocked(osdc);
3219
3220 lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3221 lreq->linger_id);
3222 if (!lookup_lreq)
3223 return;
3224
3225 WARN_ON(lookup_lreq != lreq);
3226 erase_linger_mc(&osdc->linger_map_checks, lreq);
3227 linger_put(lreq);
3228}
3229
922dab61
ID		 3230/*
3231 * @lreq has to be both registered and linked.
3232 */
3233static void __linger_cancel(struct ceph_osd_linger_request *lreq)
3234{
75dbb685 3235 if (lreq->ping_req && lreq->ping_req->r_osd)
922dab61 3236 cancel_linger_request(lreq->ping_req);
75dbb685 3237 if (lreq->reg_req && lreq->reg_req->r_osd)
922dab61 3238 cancel_linger_request(lreq->reg_req);
4609245e 3239 cancel_linger_map_check(lreq);
922dab61
ID		 3240 unlink_linger(lreq->osd, lreq);
3241 linger_unregister(lreq);
3242}
3243
3244static void linger_cancel(struct ceph_osd_linger_request *lreq)
3245{
3246 struct ceph_osd_client *osdc = lreq->osdc;
3247
3248 down_write(&osdc->lock);
3249 if (__linger_registered(lreq))
3250 __linger_cancel(lreq);
3251 up_write(&osdc->lock);
3252}
3253
4609245e
ID		 3254static void send_linger_map_check(struct ceph_osd_linger_request *lreq);
3255
3256static void check_linger_pool_dne(struct ceph_osd_linger_request *lreq)
3257{
3258 struct ceph_osd_client *osdc = lreq->osdc;
3259 struct ceph_osdmap *map = osdc->osdmap;
3260
3261 verify_osdc_wrlocked(osdc);
3262 WARN_ON(!map->epoch);
3263
3264 if (lreq->register_gen) {
3265 lreq->map_dne_bound = map->epoch;
3266 dout("%s lreq %p linger_id %llu pool disappeared\n", __func__,
3267 lreq, lreq->linger_id);
3268 } else {
3269 dout("%s lreq %p linger_id %llu map_dne_bound %u have %u\n",
3270 __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3271 map->epoch);
3272 }
3273
3274 if (lreq->map_dne_bound) {
3275 if (map->epoch >= lreq->map_dne_bound) {
3276 /* we had a new enough map */
3277 pr_info("linger_id %llu pool does not exist\n",
3278 lreq->linger_id);
3279 linger_reg_commit_complete(lreq, -ENOENT);
3280 __linger_cancel(lreq);
3281 }
3282 } else {
3283 send_linger_map_check(lreq);
3284 }
3285}
3286
3287static void linger_map_check_cb(struct ceph_mon_generic_request *greq)
3288{
3289 struct ceph_osd_client *osdc = &greq->monc->client->osdc;
3290 struct ceph_osd_linger_request *lreq;
3291 u64 linger_id = greq->private_data;
3292
3293 WARN_ON(greq->result || !greq->u.newest);
3294
3295 down_write(&osdc->lock);
3296 lreq = lookup_linger_mc(&osdc->linger_map_checks, linger_id);
3297 if (!lreq) {
3298 dout("%s linger_id %llu dne\n", __func__, linger_id);
3299 goto out_unlock;
3300 }
3301
3302 dout("%s lreq %p linger_id %llu map_dne_bound %u newest %llu\n",
3303 __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3304 greq->u.newest);
3305 if (!lreq->map_dne_bound)
3306 lreq->map_dne_bound = greq->u.newest;
3307 erase_linger_mc(&osdc->linger_map_checks, lreq);
3308 check_linger_pool_dne(lreq);
3309
3310 linger_put(lreq);
3311out_unlock:
3312 up_write(&osdc->lock);
3313}
3314
3315static void send_linger_map_check(struct ceph_osd_linger_request *lreq)
3316{
3317 struct ceph_osd_client *osdc = lreq->osdc;
3318 struct ceph_osd_linger_request *lookup_lreq;
3319 int ret;
3320
3321 verify_osdc_wrlocked(osdc);
3322
3323 lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3324 lreq->linger_id);
3325 if (lookup_lreq) {
3326 WARN_ON(lookup_lreq != lreq);
3327 return;
3328 }
3329
3330 linger_get(lreq);
3331 insert_linger_mc(&osdc->linger_map_checks, lreq);
3332 ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
3333 linger_map_check_cb, lreq->linger_id);
3334 WARN_ON(ret);
3335}
3336
922dab61
ID		 3337static int linger_reg_commit_wait(struct ceph_osd_linger_request *lreq)
3338{
3339 int ret;
3340
3341 dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
e6e28432 3342 ret = wait_for_completion_killable(&lreq->reg_commit_wait);
922dab61
ID		 3343 return ret ?: lreq->reg_commit_error;
3344}
3345
e6e28432
ID		 3346static int linger_notify_finish_wait(struct ceph_osd_linger_request *lreq,
3347 unsigned long timeout)
19079203 3348{
e6e28432 3349 long left;
19079203
ID		 3350
3351 dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
e6e28432
ID		 3352 left = wait_for_completion_killable_timeout(&lreq->notify_finish_wait,
3353 ceph_timeout_jiffies(timeout));
3354 if (left <= 0)
3355 left = left ?: -ETIMEDOUT;
3356 else
3357 left = lreq->notify_finish_error; /* completed */
3358
3359 return left;
19079203
ID		 3360}
3361
f24e9980 3362/*
fbca9635
ID		 3363 * Timeout callback, called every N seconds. When 1 or more OSD
3364 * requests has been active for more than N seconds, we send a keepalive
3365 * (tag + timestamp) to its OSD to ensure any communications channel
3366 * reset is detected.
f24e9980
SW		 3367 */
3368static void handle_timeout(struct work_struct *work)
3369{
3370 struct ceph_osd_client *osdc =
3371 container_of(work, struct ceph_osd_client, timeout_work.work);
a319bf56 3372 struct ceph_options *opts = osdc->client->options;
5aea3dcd 3373 unsigned long cutoff = jiffies - opts->osd_keepalive_timeout;
7cc5e38f 3374 unsigned long expiry_cutoff = jiffies - opts->osd_request_timeout;
5aea3dcd
ID		 3375 LIST_HEAD(slow_osds);
3376 struct rb_node *n, *p;
f24e9980 3377
5aea3dcd
ID		 3378 dout("%s osdc %p\n", __func__, osdc);
3379 down_write(&osdc->lock);
f24e9980 3380
422d2cb8
YS		 3381 /*
3382 * ping osds that are a bit slow. this ensures that if there
3383 * is a break in the TCP connection we will notice, and reopen
3384 * a connection with that osd (from the fault callback).
3385 */
5aea3dcd
ID		 3386 for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
3387 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
3388 bool found = false;
3389
7cc5e38f 3390 for (p = rb_first(&osd->o_requests); p; ) {
5aea3dcd
ID		 3391 struct ceph_osd_request *req =
3392 rb_entry(p, struct ceph_osd_request, r_node);
3393
7cc5e38f
ID		 3394 p = rb_next(p); /* abort_request() */
3395
5aea3dcd
ID		 3396 if (time_before(req->r_stamp, cutoff)) {
3397 dout(" req %p tid %llu on osd%d is laggy\n",
3398 req, req->r_tid, osd->o_osd);
3399 found = true;
3400 }
7cc5e38f
ID		 3401 if (opts->osd_request_timeout &&
3402 time_before(req->r_start_stamp, expiry_cutoff)) {
3403 pr_err_ratelimited("tid %llu on osd%d timeout\n",
3404 req->r_tid, osd->o_osd);
3405 abort_request(req, -ETIMEDOUT);
3406 }
5aea3dcd 3407 }
922dab61
ID		 3408 for (p = rb_first(&osd->o_linger_requests); p; p = rb_next(p)) {
3409 struct ceph_osd_linger_request *lreq =
3410 rb_entry(p, struct ceph_osd_linger_request, node);
3411
3412 dout(" lreq %p linger_id %llu is served by osd%d\n",
3413 lreq, lreq->linger_id, osd->o_osd);
3414 found = true;
3415
3416 mutex_lock(&lreq->lock);
19079203 3417 if (lreq->is_watch && lreq->committed && !lreq->last_error)
922dab61
ID		 3418 send_linger_ping(lreq);
3419 mutex_unlock(&lreq->lock);
3420 }
422d2cb8 3421
5aea3dcd
ID		 3422 if (found)
3423 list_move_tail(&osd->o_keepalive_item, &slow_osds);
422d2cb8 3424 }
5aea3dcd 3425
7cc5e38f
ID		 3426 if (opts->osd_request_timeout) {
3427 for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
3428 struct ceph_osd_request *req =
3429 rb_entry(p, struct ceph_osd_request, r_node);
3430
3431 p = rb_next(p); /* abort_request() */
3432
3433 if (time_before(req->r_start_stamp, expiry_cutoff)) {
3434 pr_err_ratelimited("tid %llu on osd%d timeout\n",
3435 req->r_tid, osdc->homeless_osd.o_osd);
3436 abort_request(req, -ETIMEDOUT);
3437 }
3438 }
3439 }
3440
5aea3dcd
ID		 3441 if (atomic_read(&osdc->num_homeless) || !list_empty(&slow_osds))
3442 maybe_request_map(osdc);
3443
422d2cb8 3444 while (!list_empty(&slow_osds)) {
5aea3dcd
ID		 3445 struct ceph_osd *osd = list_first_entry(&slow_osds,
3446 struct ceph_osd,
3447 o_keepalive_item);
422d2cb8 3448 list_del_init(&osd->o_keepalive_item);
f24e9980
SW		 3449 ceph_con_keepalive(&osd->o_con);
3450 }
3451
5aea3dcd 3452 up_write(&osdc->lock);
fbca9635
ID		 3453 schedule_delayed_work(&osdc->timeout_work,
3454 osdc->client->options->osd_keepalive_timeout);
f24e9980
SW		 3455}
3456
f5a2041b
YS		 3457static void handle_osds_timeout(struct work_struct *work)
3458{
3459 struct ceph_osd_client *osdc =
3460 container_of(work, struct ceph_osd_client,
3461 osds_timeout_work.work);
a319bf56 3462 unsigned long delay = osdc->client->options->osd_idle_ttl / 4;
42a2c09f 3463 struct ceph_osd *osd, *nosd;
f5a2041b 3464
42a2c09f 3465 dout("%s osdc %p\n", __func__, osdc);
5aea3dcd 3466 down_write(&osdc->lock);
42a2c09f
ID		 3467 list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) {
3468 if (time_before(jiffies, osd->lru_ttl))
3469 break;
3470
5aea3dcd 3471 WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
922dab61 3472 WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
5aea3dcd 3473 close_osd(osd);
42a2c09f 3474 }
f5a2041b 3475
5aea3dcd 3476 up_write(&osdc->lock);
f5a2041b
YS		 3477 schedule_delayed_work(&osdc->osds_timeout_work,
3478 round_jiffies_relative(delay));
3479}
3480
205ee118
ID		 3481static int ceph_oloc_decode(void **p, void *end,
3482 struct ceph_object_locator *oloc)
3483{
3484 u8 struct_v, struct_cv;
3485 u32 len;
3486 void *struct_end;
3487 int ret = 0;
3488
3489 ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3490 struct_v = ceph_decode_8(p);
3491 struct_cv = ceph_decode_8(p);
3492 if (struct_v < 3) {
3493 pr_warn("got v %d < 3 cv %d of ceph_object_locator\n",
3494 struct_v, struct_cv);
3495 goto e_inval;
3496 }
3497 if (struct_cv > 6) {
3498 pr_warn("got v %d cv %d > 6 of ceph_object_locator\n",
3499 struct_v, struct_cv);
3500 goto e_inval;
3501 }
3502 len = ceph_decode_32(p);
3503 ceph_decode_need(p, end, len, e_inval);
3504 struct_end = *p + len;
3505
3506 oloc->pool = ceph_decode_64(p);
3507 *p += 4; /* skip preferred */
3508
3509 len = ceph_decode_32(p);
3510 if (len > 0) {
3511 pr_warn("ceph_object_locator::key is set\n");
3512 goto e_inval;
3513 }
3514
3515 if (struct_v >= 5) {
cd08e0a2
YZ		 3516 bool changed = false;
3517
205ee118
ID		 3518 len = ceph_decode_32(p);
3519 if (len > 0) {
30c156d9 3520 ceph_decode_need(p, end, len, e_inval);
cd08e0a2
YZ		 3521 if (!oloc->pool_ns ||
3522 ceph_compare_string(oloc->pool_ns, *p, len))
3523 changed = true;
30c156d9 3524 *p += len;
cd08e0a2
YZ		 3525 } else {
3526 if (oloc->pool_ns)
3527 changed = true;
3528 }
3529 if (changed) {
3530 /* redirect changes namespace */
3531 pr_warn("ceph_object_locator::nspace is changed\n");
3532 goto e_inval;
205ee118
ID		 3533 }
3534 }
3535
3536 if (struct_v >= 6) {
3537 s64 hash = ceph_decode_64(p);
3538 if (hash != -1) {
3539 pr_warn("ceph_object_locator::hash is set\n");
3540 goto e_inval;
3541 }
3542 }
3543
3544 /* skip the rest */
3545 *p = struct_end;
3546out:
3547 return ret;
3548
3549e_inval:
3550 ret = -EINVAL;
3551 goto out;
3552}
3553
3554static int ceph_redirect_decode(void **p, void *end,
3555 struct ceph_request_redirect *redir)
3556{
3557 u8 struct_v, struct_cv;
3558 u32 len;
3559 void *struct_end;
3560 int ret;
3561
3562 ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3563 struct_v = ceph_decode_8(p);
3564 struct_cv = ceph_decode_8(p);
3565 if (struct_cv > 1) {
3566 pr_warn("got v %d cv %d > 1 of ceph_request_redirect\n",
3567 struct_v, struct_cv);
3568 goto e_inval;
3569 }
3570 len = ceph_decode_32(p);
3571 ceph_decode_need(p, end, len, e_inval);
3572 struct_end = *p + len;
3573
3574 ret = ceph_oloc_decode(p, end, &redir->oloc);
3575 if (ret)
3576 goto out;
3577
3578 len = ceph_decode_32(p);
3579 if (len > 0) {
3580 pr_warn("ceph_request_redirect::object_name is set\n");
3581 goto e_inval;
3582 }
3583
205ee118
ID		 3584 /* skip the rest */
3585 *p = struct_end;
3586out:
3587 return ret;
3588
3589e_inval:
3590 ret = -EINVAL;
3591 goto out;
3592}
3593
fe5da05e
ID		 3594struct MOSDOpReply {
3595 struct ceph_pg pgid;
3596 u64 flags;
3597 int result;
3598 u32 epoch;
3599 int num_ops;
3600 u32 outdata_len[CEPH_OSD_MAX_OPS];
3601 s32 rval[CEPH_OSD_MAX_OPS];
3602 int retry_attempt;
3603 struct ceph_eversion replay_version;
3604 u64 user_version;
3605 struct ceph_request_redirect redirect;
3606};
25845472 3607
fe5da05e 3608static int decode_MOSDOpReply(const struct ceph_msg *msg, struct MOSDOpReply *m)
f24e9980 3609{
fe5da05e
ID		 3610 void *p = msg->front.iov_base;
3611 void *const end = p + msg->front.iov_len;
3612 u16 version = le16_to_cpu(msg->hdr.version);
3613 struct ceph_eversion bad_replay_version;
b0b31a8f 3614 u8 decode_redir;
fe5da05e
ID		 3615 u32 len;
3616 int ret;
3617 int i;
1b83bef2 3618
fe5da05e
ID		 3619 ceph_decode_32_safe(&p, end, len, e_inval);
3620 ceph_decode_need(&p, end, len, e_inval);
3621 p += len; /* skip oid */
1b83bef2 3622
fe5da05e
ID		 3623 ret = ceph_decode_pgid(&p, end, &m->pgid);
3624 if (ret)
3625 return ret;
1b83bef2 3626
fe5da05e
ID		 3627 ceph_decode_64_safe(&p, end, m->flags, e_inval);
3628 ceph_decode_32_safe(&p, end, m->result, e_inval);
3629 ceph_decode_need(&p, end, sizeof(bad_replay_version), e_inval);
3630 memcpy(&bad_replay_version, p, sizeof(bad_replay_version));
3631 p += sizeof(bad_replay_version);
3632 ceph_decode_32_safe(&p, end, m->epoch, e_inval);
1b83bef2 3633
fe5da05e
ID		 3634 ceph_decode_32_safe(&p, end, m->num_ops, e_inval);
3635 if (m->num_ops > ARRAY_SIZE(m->outdata_len))
3636 goto e_inval;
1b83bef2 3637
fe5da05e
ID		 3638 ceph_decode_need(&p, end, m->num_ops * sizeof(struct ceph_osd_op),
3639 e_inval);
3640 for (i = 0; i < m->num_ops; i++) {
1b83bef2 3641 struct ceph_osd_op *op = p;
1b83bef2 3642
fe5da05e 3643 m->outdata_len[i] = le32_to_cpu(op->payload_len);
1b83bef2
SW		 3644 p += sizeof(*op);
3645 }
1b83bef2 3646
fe5da05e
ID		 3647 ceph_decode_32_safe(&p, end, m->retry_attempt, e_inval);
3648 for (i = 0; i < m->num_ops; i++)
3649 ceph_decode_32_safe(&p, end, m->rval[i], e_inval);
f24e9980 3650
fe5da05e
ID		 3651 if (version >= 5) {
3652 ceph_decode_need(&p, end, sizeof(m->replay_version), e_inval);
3653 memcpy(&m->replay_version, p, sizeof(m->replay_version));
3654 p += sizeof(m->replay_version);
3655 ceph_decode_64_safe(&p, end, m->user_version, e_inval);
3656 } else {
3657 m->replay_version = bad_replay_version; /* struct */
3658 m->user_version = le64_to_cpu(m->replay_version.version);
3659 }
eb845ff1 3660
fe5da05e
ID		 3661 if (version >= 6) {
3662 if (version >= 7)
3663 ceph_decode_8_safe(&p, end, decode_redir, e_inval);
b0b31a8f
ID		 3664 else
3665 decode_redir = 1;
3666 } else {
3667 decode_redir = 0;
3668 }
3669
3670 if (decode_redir) {
fe5da05e
ID		 3671 ret = ceph_redirect_decode(&p, end, &m->redirect);
3672 if (ret)
3673 return ret;
205ee118 3674 } else {
fe5da05e 3675 ceph_oloc_init(&m->redirect.oloc);
205ee118 3676 }
f24e9980 3677
fe5da05e
ID		 3678 return 0;
3679
3680e_inval:
3681 return -EINVAL;
3682}
3683
3684/*
b18b9550
ID		 3685 * Handle MOSDOpReply. Set ->r_result and call the callback if it is
3686 * specified.
fe5da05e 3687 */
5aea3dcd 3688static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg)
fe5da05e 3689{
5aea3dcd 3690 struct ceph_osd_client *osdc = osd->o_osdc;
fe5da05e
ID		 3691 struct ceph_osd_request *req;
3692 struct MOSDOpReply m;
3693 u64 tid = le64_to_cpu(msg->hdr.tid);
3694 u32 data_len = 0;
fe5da05e
ID		 3695 int ret;
3696 int i;
3697
3698 dout("%s msg %p tid %llu\n", __func__, msg, tid);
3699
5aea3dcd
ID		 3700 down_read(&osdc->lock);
3701 if (!osd_registered(osd)) {
3702 dout("%s osd%d unknown\n", __func__, osd->o_osd);
3703 goto out_unlock_osdc;
3704 }
3705 WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
3706
3707 mutex_lock(&osd->lock);
3708 req = lookup_request(&osd->o_requests, tid);
fe5da05e 3709 if (!req) {
5aea3dcd
ID		 3710 dout("%s osd%d tid %llu unknown\n", __func__, osd->o_osd, tid);
3711 goto out_unlock_session;
fe5da05e 3712 }
fe5da05e 3713
cd08e0a2 3714 m.redirect.oloc.pool_ns = req->r_t.target_oloc.pool_ns;
fe5da05e 3715 ret = decode_MOSDOpReply(msg, &m);
cd08e0a2 3716 m.redirect.oloc.pool_ns = NULL;
fe5da05e
ID		 3717 if (ret) {
3718 pr_err("failed to decode MOSDOpReply for tid %llu: %d\n",
3719 req->r_tid, ret);
3720 ceph_msg_dump(msg);
3721 goto fail_request;
3722 }
3723 dout("%s req %p tid %llu flags 0x%llx pgid %llu.%x epoch %u attempt %d v %u'%llu uv %llu\n",
3724 __func__, req, req->r_tid, m.flags, m.pgid.pool, m.pgid.seed,
3725 m.epoch, m.retry_attempt, le32_to_cpu(m.replay_version.epoch),
3726 le64_to_cpu(m.replay_version.version), m.user_version);
3727
3728 if (m.retry_attempt >= 0) {
3729 if (m.retry_attempt != req->r_attempts - 1) {
3730 dout("req %p tid %llu retry_attempt %d != %d, ignoring\n",
3731 req, req->r_tid, m.retry_attempt,
3732 req->r_attempts - 1);
5aea3dcd 3733 goto out_unlock_session;
fe5da05e
ID		 3734 }
3735 } else {
3736 WARN_ON(1); /* MOSDOpReply v4 is assumed */
3737 }
3738
3739 if (!ceph_oloc_empty(&m.redirect.oloc)) {
3740 dout("req %p tid %llu redirect pool %lld\n", req, req->r_tid,
3741 m.redirect.oloc.pool);
5aea3dcd
ID		 3742 unlink_request(osd, req);
3743 mutex_unlock(&osd->lock);
205ee118 3744
30c156d9
YZ		 3745 /*
3746 * Not ceph_oloc_copy() - changing pool_ns is not
3747 * supported.
3748 */
3749 req->r_t.target_oloc.pool = m.redirect.oloc.pool;
890bd0f8
JL		 3750 req->r_flags |= CEPH_OSD_FLAG_REDIRECTED |
3751 CEPH_OSD_FLAG_IGNORE_OVERLAY |
3752 CEPH_OSD_FLAG_IGNORE_CACHE;
5aea3dcd
ID		 3753 req->r_tid = 0;
3754 __submit_request(req, false);
3755 goto out_unlock_osdc;
205ee118
ID		 3756 }
3757
117d96a0
ID		 3758 if (m.result == -EAGAIN) {
3759 dout("req %p tid %llu EAGAIN\n", req, req->r_tid);
3760 unlink_request(osd, req);
3761 mutex_unlock(&osd->lock);
3762
3763 /*
3764 * The object is missing on the replica or not (yet)
3765 * readable. Clear pgid to force a resend to the primary
3766 * via legacy_change.
3767 */
3768 req->r_t.pgid.pool = 0;
3769 req->r_t.pgid.seed = 0;
3770 WARN_ON(!req->r_t.used_replica);
3771 req->r_flags &= ~(CEPH_OSD_FLAG_BALANCE_READS |
3772 CEPH_OSD_FLAG_LOCALIZE_READS);
3773 req->r_tid = 0;
3774 __submit_request(req, false);
3775 goto out_unlock_osdc;
3776 }
3777
fe5da05e
ID		 3778 if (m.num_ops != req->r_num_ops) {
3779 pr_err("num_ops %d != %d for tid %llu\n", m.num_ops,
3780 req->r_num_ops, req->r_tid);
3781 goto fail_request;
f24e9980 3782 }
fe5da05e
ID		 3783 for (i = 0; i < req->r_num_ops; i++) {
3784 dout(" req %p tid %llu op %d rval %d len %u\n", req,
3785 req->r_tid, i, m.rval[i], m.outdata_len[i]);
3786 req->r_ops[i].rval = m.rval[i];
3787 req->r_ops[i].outdata_len = m.outdata_len[i];
3788 data_len += m.outdata_len[i];
3789 }
3790 if (data_len != le32_to_cpu(msg->hdr.data_len)) {
3791 pr_err("sum of lens %u != %u for tid %llu\n", data_len,
3792 le32_to_cpu(msg->hdr.data_len), req->r_tid);
3793 goto fail_request;
3794 }
b18b9550
ID		 3795 dout("%s req %p tid %llu result %d data_len %u\n", __func__,
3796 req, req->r_tid, m.result, data_len);
f24e9980 3797
b18b9550
ID		 3798 /*
3799 * Since we only ever request ONDISK, we should only ever get
3800 * one (type of) reply back.
3801 */
3802 WARN_ON(!(m.flags & CEPH_OSD_FLAG_ONDISK));
3803 req->r_result = m.result ?: data_len;
3804 finish_request(req);
5aea3dcd
ID		 3805 mutex_unlock(&osd->lock);
3806 up_read(&osdc->lock);
f24e9980 3807
b18b9550 3808 __complete_request(req);
f24e9980
SW		 3809 return;
3810
fe5da05e 3811fail_request:
4609245e 3812 complete_request(req, -EIO);
5aea3dcd
ID		 3813out_unlock_session:
3814 mutex_unlock(&osd->lock);
3815out_unlock_osdc:
3816 up_read(&osdc->lock);
f24e9980
SW		 3817}
3818
42c1b124
ID		 3819static void set_pool_was_full(struct ceph_osd_client *osdc)
3820{
3821 struct rb_node *n;
3822
3823 for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
3824 struct ceph_pg_pool_info *pi =
3825 rb_entry(n, struct ceph_pg_pool_info, node);
3826
3827 pi->was_full = __pool_full(pi);
3828 }
3829}
3830
5aea3dcd 3831static bool pool_cleared_full(struct ceph_osd_client *osdc, s64 pool_id)
f24e9980 3832{
5aea3dcd 3833 struct ceph_pg_pool_info *pi;
f24e9980 3834
5aea3dcd
ID		 3835 pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
3836 if (!pi)
3837 return false;
f24e9980 3838
5aea3dcd 3839 return pi->was_full && !__pool_full(pi);
422d2cb8
YS		 3840}
3841
922dab61
ID		 3842static enum calc_target_result
3843recalc_linger_target(struct ceph_osd_linger_request *lreq)
3844{
3845 struct ceph_osd_client *osdc = lreq->osdc;
3846 enum calc_target_result ct_res;
3847
8edf84ba 3848 ct_res = calc_target(osdc, &lreq->t, true);
922dab61
ID		 3849 if (ct_res == CALC_TARGET_NEED_RESEND) {
3850 struct ceph_osd *osd;
3851
3852 osd = lookup_create_osd(osdc, lreq->t.osd, true);
3853 if (osd != lreq->osd) {
3854 unlink_linger(lreq->osd, lreq);
3855 link_linger(osd, lreq);
3856 }
3857 }
3858
3859 return ct_res;
3860}
3861
422d2cb8 3862/*
5aea3dcd 3863 * Requeue requests whose mapping to an OSD has changed.
422d2cb8 3864 */
5aea3dcd
ID		 3865static void scan_requests(struct ceph_osd *osd,
3866 bool force_resend,
3867 bool cleared_full,
3868 bool check_pool_cleared_full,
3869 struct rb_root *need_resend,
3870 struct list_head *need_resend_linger)
422d2cb8 3871{
5aea3dcd
ID		 3872 struct ceph_osd_client *osdc = osd->o_osdc;
3873 struct rb_node *n;
3874 bool force_resend_writes;
3875
922dab61
ID		 3876 for (n = rb_first(&osd->o_linger_requests); n; ) {
3877 struct ceph_osd_linger_request *lreq =
3878 rb_entry(n, struct ceph_osd_linger_request, node);
3879 enum calc_target_result ct_res;
3880
3881 n = rb_next(n); /* recalc_linger_target() */
3882
3883 dout("%s lreq %p linger_id %llu\n", __func__, lreq,
3884 lreq->linger_id);
3885 ct_res = recalc_linger_target(lreq);
3886 switch (ct_res) {
3887 case CALC_TARGET_NO_ACTION:
3888 force_resend_writes = cleared_full ||
3889 (check_pool_cleared_full &&
3890 pool_cleared_full(osdc, lreq->t.base_oloc.pool));
3891 if (!force_resend && !force_resend_writes)
3892 break;
3893
df561f66 3894 fallthrough;
922dab61 3895 case CALC_TARGET_NEED_RESEND:
4609245e 3896 cancel_linger_map_check(lreq);
922dab61
ID		 3897 /*
3898 * scan_requests() for the previous epoch(s)
3899 * may have already added it to the list, since
3900 * it's not unlinked here.
3901 */
3902 if (list_empty(&lreq->scan_item))
3903 list_add_tail(&lreq->scan_item, need_resend_linger);
3904 break;
3905 case CALC_TARGET_POOL_DNE:
a10bcb19 3906 list_del_init(&lreq->scan_item);
4609245e 3907 check_linger_pool_dne(lreq);
922dab61
ID		 3908 break;
3909 }
3910 }
3911
5aea3dcd
ID		 3912 for (n = rb_first(&osd->o_requests); n; ) {
3913 struct ceph_osd_request *req =
3914 rb_entry(n, struct ceph_osd_request, r_node);
3915 enum calc_target_result ct_res;
3916
4609245e 3917 n = rb_next(n); /* unlink_request(), check_pool_dne() */
5aea3dcd
ID		 3918
3919 dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
8edf84ba 3920 ct_res = calc_target(osdc, &req->r_t, false);
5aea3dcd
ID		 3921 switch (ct_res) {
3922 case CALC_TARGET_NO_ACTION:
3923 force_resend_writes = cleared_full ||
3924 (check_pool_cleared_full &&
3925 pool_cleared_full(osdc, req->r_t.base_oloc.pool));
3926 if (!force_resend &&
3927 (!(req->r_flags & CEPH_OSD_FLAG_WRITE) ||
3928 !force_resend_writes))
3929 break;
3930
df561f66 3931 fallthrough;
5aea3dcd 3932 case CALC_TARGET_NEED_RESEND:
4609245e 3933 cancel_map_check(req);
5aea3dcd
ID		 3934 unlink_request(osd, req);
3935 insert_request(need_resend, req);
3936 break;
3937 case CALC_TARGET_POOL_DNE:
4609245e 3938 check_pool_dne(req);
5aea3dcd 3939 break;
b0494532 3940 }
6f6c7006 3941 }
422d2cb8 3942}
6f6c7006 3943
42c1b124 3944static int handle_one_map(struct ceph_osd_client *osdc,
5aea3dcd
ID		 3945 void *p, void *end, bool incremental,
3946 struct rb_root *need_resend,
3947 struct list_head *need_resend_linger)
42c1b124
ID		 3948{
3949 struct ceph_osdmap *newmap;
3950 struct rb_node *n;
3951 bool skipped_map = false;
3952 bool was_full;
3953
b7ec35b3 3954 was_full = ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
42c1b124
ID		 3955 set_pool_was_full(osdc);
3956
3957 if (incremental)
cd1a677c
ID		 3958 newmap = osdmap_apply_incremental(&p, end,
3959 ceph_msgr2(osdc->client),
3960 osdc->osdmap);
42c1b124 3961 else
cd1a677c 3962 newmap = ceph_osdmap_decode(&p, end, ceph_msgr2(osdc->client));
42c1b124
ID		 3963 if (IS_ERR(newmap))
3964 return PTR_ERR(newmap);
3965
3966 if (newmap != osdc->osdmap) {
3967 /*
3968 * Preserve ->was_full before destroying the old map.
3969 * For pools that weren't in the old map, ->was_full
3970 * should be false.
3971 */
3972 for (n = rb_first(&newmap->pg_pools); n; n = rb_next(n)) {
3973 struct ceph_pg_pool_info *pi =
3974 rb_entry(n, struct ceph_pg_pool_info, node);
3975 struct ceph_pg_pool_info *old_pi;
3976
3977 old_pi = ceph_pg_pool_by_id(osdc->osdmap, pi->id);
3978 if (old_pi)
3979 pi->was_full = old_pi->was_full;
3980 else
3981 WARN_ON(pi->was_full);
3982 }
3983
3984 if (osdc->osdmap->epoch &&
3985 osdc->osdmap->epoch + 1 < newmap->epoch) {
3986 WARN_ON(incremental);
3987 skipped_map = true;
3988 }
3989
3990 ceph_osdmap_destroy(osdc->osdmap);
3991 osdc->osdmap = newmap;
3992 }
3993
b7ec35b3 3994 was_full &= !ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
5aea3dcd
ID		 3995 scan_requests(&osdc->homeless_osd, skipped_map, was_full, true,
3996 need_resend, need_resend_linger);
3997
3998 for (n = rb_first(&osdc->osds); n; ) {
3999 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
4000
4001 n = rb_next(n); /* close_osd() */
4002
4003 scan_requests(osd, skipped_map, was_full, true, need_resend,
4004 need_resend_linger);
4005 if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) ||
4006 memcmp(&osd->o_con.peer_addr,
4007 ceph_osd_addr(osdc->osdmap, osd->o_osd),
4008 sizeof(struct ceph_entity_addr)))
4009 close_osd(osd);
4010 }
42c1b124
ID		 4011
4012 return 0;
4013}
6f6c7006 4014
5aea3dcd
ID		 4015static void kick_requests(struct ceph_osd_client *osdc,
4016 struct rb_root *need_resend,
4017 struct list_head *need_resend_linger)
4018{
922dab61 4019 struct ceph_osd_linger_request *lreq, *nlreq;
04c7d789 4020 enum calc_target_result ct_res;
5aea3dcd
ID		 4021 struct rb_node *n;
4022
04c7d789
ID		 4023 /* make sure need_resend targets reflect latest map */
4024 for (n = rb_first(need_resend); n; ) {
4025 struct ceph_osd_request *req =
4026 rb_entry(n, struct ceph_osd_request, r_node);
4027
4028 n = rb_next(n);
4029
4030 if (req->r_t.epoch < osdc->osdmap->epoch) {
8edf84ba 4031 ct_res = calc_target(osdc, &req->r_t, false);
04c7d789
ID		 4032 if (ct_res == CALC_TARGET_POOL_DNE) {
4033 erase_request(need_resend, req);
4034 check_pool_dne(req);
4035 }
4036 }
4037 }
4038
5aea3dcd
ID		 4039 for (n = rb_first(need_resend); n; ) {
4040 struct ceph_osd_request *req =
4041 rb_entry(n, struct ceph_osd_request, r_node);
4042 struct ceph_osd *osd;
4043
4044 n = rb_next(n);
4045 erase_request(need_resend, req); /* before link_request() */
4046
5aea3dcd
ID		 4047 osd = lookup_create_osd(osdc, req->r_t.osd, true);
4048 link_request(osd, req);
4049 if (!req->r_linger) {
4050 if (!osd_homeless(osd) && !req->r_t.paused)
4051 send_request(req);
922dab61
ID		 4052 } else {
4053 cancel_linger_request(req);
5aea3dcd
ID		 4054 }
4055 }
922dab61
ID		 4056
4057 list_for_each_entry_safe(lreq, nlreq, need_resend_linger, scan_item) {
4058 if (!osd_homeless(lreq->osd))
4059 send_linger(lreq);
4060
4061 list_del_init(&lreq->scan_item);
4062 }
5aea3dcd
ID		 4063}
4064
f24e9980
SW		 4065/*
4066 * Process updated osd map.
4067 *
4068 * The message contains any number of incremental and full maps, normally
4069 * indicating some sort of topology change in the cluster. Kick requests
4070 * off to different OSDs as needed.
4071 */
4072void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
4073{
42c1b124
ID		 4074 void *p = msg->front.iov_base;
4075 void *const end = p + msg->front.iov_len;
f24e9980
SW		 4076 u32 nr_maps, maplen;
4077 u32 epoch;
f24e9980 4078 struct ceph_fsid fsid;
5aea3dcd
ID		 4079 struct rb_root need_resend = RB_ROOT;
4080 LIST_HEAD(need_resend_linger);
42c1b124
ID		 4081 bool handled_incremental = false;
4082 bool was_pauserd, was_pausewr;
4083 bool pauserd, pausewr;
4084 int err;
f24e9980 4085
42c1b124 4086 dout("%s have %u\n", __func__, osdc->osdmap->epoch);
5aea3dcd 4087 down_write(&osdc->lock);
f24e9980
SW		 4088
4089 /* verify fsid */
4090 ceph_decode_need(&p, end, sizeof(fsid), bad);
4091 ceph_decode_copy(&p, &fsid, sizeof(fsid));
0743304d 4092 if (ceph_check_fsid(osdc->client, &fsid) < 0)
42c1b124 4093 goto bad;
f24e9980 4094
b7ec35b3
ID		 4095 was_pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
4096 was_pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
4097 ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
42c1b124 4098 have_pool_full(osdc);
9a1ea2db 4099
f24e9980
SW		 4100 /* incremental maps */
4101 ceph_decode_32_safe(&p, end, nr_maps, bad);
4102 dout(" %d inc maps\n", nr_maps);
4103 while (nr_maps > 0) {
4104 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
c89136ea
SW		 4105 epoch = ceph_decode_32(&p);
4106 maplen = ceph_decode_32(&p);
f24e9980 4107 ceph_decode_need(&p, end, maplen, bad);
42c1b124
ID		 4108 if (osdc->osdmap->epoch &&
4109 osdc->osdmap->epoch + 1 == epoch) {
f24e9980
SW		 4110 dout("applying incremental map %u len %d\n",
4111 epoch, maplen);
5aea3dcd
ID		 4112 err = handle_one_map(osdc, p, p + maplen, true,
4113 &need_resend, &need_resend_linger);
42c1b124 4114 if (err)
f24e9980 4115 goto bad;
42c1b124 4116 handled_incremental = true;
f24e9980
SW		 4117 } else {
4118 dout("ignoring incremental map %u len %d\n",
4119 epoch, maplen);
4120 }
42c1b124 4121 p += maplen;
f24e9980
SW		 4122 nr_maps--;
4123 }
42c1b124 4124 if (handled_incremental)
f24e9980
SW		 4125 goto done;
4126
4127 /* full maps */
4128 ceph_decode_32_safe(&p, end, nr_maps, bad);
4129 dout(" %d full maps\n", nr_maps);
4130 while (nr_maps) {
4131 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
c89136ea
SW		 4132 epoch = ceph_decode_32(&p);
4133 maplen = ceph_decode_32(&p);
f24e9980
SW		 4134 ceph_decode_need(&p, end, maplen, bad);
4135 if (nr_maps > 1) {
4136 dout("skipping non-latest full map %u len %d\n",
4137 epoch, maplen);
e5253a7b 4138 } else if (osdc->osdmap->epoch >= epoch) {
f24e9980
SW		 4139 dout("skipping full map %u len %d, "
4140 "older than our %u\n", epoch, maplen,
4141 osdc->osdmap->epoch);
4142 } else {
4143 dout("taking full map %u len %d\n", epoch, maplen);
5aea3dcd
ID		 4144 err = handle_one_map(osdc, p, p + maplen, false,
4145 &need_resend, &need_resend_linger);
42c1b124 4146 if (err)
f24e9980 4147 goto bad;
f24e9980
SW		 4148 }
4149 p += maplen;
4150 nr_maps--;
4151 }
4152
4153done:
cd634fb6
SW		 4154 /*
4155 * subscribe to subsequent osdmap updates if full to ensure
4156 * we find out when we are no longer full and stop returning
4157 * ENOSPC.
4158 */
b7ec35b3
ID		 4159 pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
4160 pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
4161 ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
42c1b124 4162 have_pool_full(osdc);
58eb7932
JL		 4163 if (was_pauserd || was_pausewr || pauserd || pausewr ||
4164 osdc->osdmap->epoch < osdc->epoch_barrier)
42c1b124 4165 maybe_request_map(osdc);
cd634fb6 4166
5aea3dcd 4167 kick_requests(osdc, &need_resend, &need_resend_linger);
42c1b124 4168
fc36d0a4 4169 ceph_osdc_abort_on_full(osdc);
42c1b124
ID		 4170 ceph_monc_got_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
4171 osdc->osdmap->epoch);
5aea3dcd 4172 up_write(&osdc->lock);
03066f23 4173 wake_up_all(&osdc->client->auth_wq);
f24e9980
SW		 4174 return;
4175
4176bad:
4177 pr_err("osdc handle_map corrupt msg\n");
9ec7cab1 4178 ceph_msg_dump(msg);
5aea3dcd
ID		 4179 up_write(&osdc->lock);
4180}
4181
4182/*
4183 * Resubmit requests pending on the given osd.
4184 */
4185static void kick_osd_requests(struct ceph_osd *osd)
4186{
4187 struct rb_node *n;
4188
a02a946d
ID		 4189 clear_backoffs(osd);
4190
922dab61 4191 for (n = rb_first(&osd->o_requests); n; ) {
5aea3dcd
ID		 4192 struct ceph_osd_request *req =
4193 rb_entry(n, struct ceph_osd_request, r_node);
4194
922dab61
ID		 4195 n = rb_next(n); /* cancel_linger_request() */
4196
5aea3dcd
ID		 4197 if (!req->r_linger) {
4198 if (!req->r_t.paused)
4199 send_request(req);
922dab61
ID		 4200 } else {
4201 cancel_linger_request(req);
5aea3dcd
ID		 4202 }
4203 }
922dab61
ID		 4204 for (n = rb_first(&osd->o_linger_requests); n; n = rb_next(n)) {
4205 struct ceph_osd_linger_request *lreq =
4206 rb_entry(n, struct ceph_osd_linger_request, node);
4207
4208 send_linger(lreq);
4209 }
5aea3dcd
ID		 4210}
4211
4212/*
4213 * If the osd connection drops, we need to resubmit all requests.
4214 */
4215static void osd_fault(struct ceph_connection *con)
4216{
4217 struct ceph_osd *osd = con->private;
4218 struct ceph_osd_client *osdc = osd->o_osdc;
4219
4220 dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
4221
4222 down_write(&osdc->lock);
4223 if (!osd_registered(osd)) {
4224 dout("%s osd%d unknown\n", __func__, osd->o_osd);
4225 goto out_unlock;
4226 }
4227
4228 if (!reopen_osd(osd))
4229 kick_osd_requests(osd);
4230 maybe_request_map(osdc);
4231
4232out_unlock:
4233 up_write(&osdc->lock);
f24e9980
SW		 4234}
4235
a02a946d
ID		 4236struct MOSDBackoff {
4237 struct ceph_spg spgid;
4238 u32 map_epoch;
4239 u8 op;
4240 u64 id;
4241 struct ceph_hobject_id *begin;
4242 struct ceph_hobject_id *end;
4243};
4244
4245static int decode_MOSDBackoff(const struct ceph_msg *msg, struct MOSDBackoff *m)
4246{
4247 void *p = msg->front.iov_base;
4248 void *const end = p + msg->front.iov_len;
4249 u8 struct_v;
4250 u32 struct_len;
4251 int ret;
4252
4253 ret = ceph_start_decoding(&p, end, 1, "spg_t", &struct_v, &struct_len);
4254 if (ret)
4255 return ret;
4256
4257 ret = ceph_decode_pgid(&p, end, &m->spgid.pgid);
4258 if (ret)
4259 return ret;
4260
4261 ceph_decode_8_safe(&p, end, m->spgid.shard, e_inval);
4262 ceph_decode_32_safe(&p, end, m->map_epoch, e_inval);
4263 ceph_decode_8_safe(&p, end, m->op, e_inval);
4264 ceph_decode_64_safe(&p, end, m->id, e_inval);
4265
4266 m->begin = kzalloc(sizeof(*m->begin), GFP_NOIO);
4267 if (!m->begin)
4268 return -ENOMEM;
4269
4270 ret = decode_hoid(&p, end, m->begin);
4271 if (ret) {
4272 free_hoid(m->begin);
4273 return ret;
4274 }
4275
4276 m->end = kzalloc(sizeof(*m->end), GFP_NOIO);
4277 if (!m->end) {
4278 free_hoid(m->begin);
4279 return -ENOMEM;
4280 }
4281
4282 ret = decode_hoid(&p, end, m->end);
4283 if (ret) {
4284 free_hoid(m->begin);
4285 free_hoid(m->end);
4286 return ret;
4287 }
4288
4289 return 0;
4290
4291e_inval:
4292 return -EINVAL;
4293}
4294
4295static struct ceph_msg *create_backoff_message(
4296 const struct ceph_osd_backoff *backoff,
4297 u32 map_epoch)
4298{
4299 struct ceph_msg *msg;
4300 void *p, *end;
4301 int msg_size;
4302
4303 msg_size = CEPH_ENCODING_START_BLK_LEN +
4304 CEPH_PGID_ENCODING_LEN + 1; /* spgid */
4305 msg_size += 4 + 1 + 8; /* map_epoch, op, id */
4306 msg_size += CEPH_ENCODING_START_BLK_LEN +
4307 hoid_encoding_size(backoff->begin);
4308 msg_size += CEPH_ENCODING_START_BLK_LEN +
4309 hoid_encoding_size(backoff->end);
4310
4311 msg = ceph_msg_new(CEPH_MSG_OSD_BACKOFF, msg_size, GFP_NOIO, true);
4312 if (!msg)
4313 return NULL;
4314
4315 p = msg->front.iov_base;
4316 end = p + msg->front_alloc_len;
4317
4318 encode_spgid(&p, &backoff->spgid);
4319 ceph_encode_32(&p, map_epoch);
4320 ceph_encode_8(&p, CEPH_OSD_BACKOFF_OP_ACK_BLOCK);
4321 ceph_encode_64(&p, backoff->id);
4322 encode_hoid(&p, end, backoff->begin);
4323 encode_hoid(&p, end, backoff->end);
4324 BUG_ON(p != end);
4325
4326 msg->front.iov_len = p - msg->front.iov_base;
4327 msg->hdr.version = cpu_to_le16(1); /* MOSDBackoff v1 */
4328 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
4329
4330 return msg;
4331}
4332
4333static void handle_backoff_block(struct ceph_osd *osd, struct MOSDBackoff *m)
4334{
4335 struct ceph_spg_mapping *spg;
4336 struct ceph_osd_backoff *backoff;
4337 struct ceph_msg *msg;
4338
4339 dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4340 m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4341
4342 spg = lookup_spg_mapping(&osd->o_backoff_mappings, &m->spgid);
4343 if (!spg) {
4344 spg = alloc_spg_mapping();
4345 if (!spg) {
4346 pr_err("%s failed to allocate spg\n", __func__);
4347 return;
4348 }
4349 spg->spgid = m->spgid; /* struct */
4350 insert_spg_mapping(&osd->o_backoff_mappings, spg);
4351 }
4352
4353 backoff = alloc_backoff();
4354 if (!backoff) {
4355 pr_err("%s failed to allocate backoff\n", __func__);
4356 return;
4357 }
4358 backoff->spgid = m->spgid; /* struct */
4359 backoff->id = m->id;
4360 backoff->begin = m->begin;
4361 m->begin = NULL; /* backoff now owns this */
4362 backoff->end = m->end;
4363 m->end = NULL; /* ditto */
4364
4365 insert_backoff(&spg->backoffs, backoff);
4366 insert_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4367
4368 /*
4369 * Ack with original backoff's epoch so that the OSD can
4370 * discard this if there was a PG split.
4371 */
4372 msg = create_backoff_message(backoff, m->map_epoch);
4373 if (!msg) {
4374 pr_err("%s failed to allocate msg\n", __func__);
4375 return;
4376 }
4377 ceph_con_send(&osd->o_con, msg);
4378}
4379
4380static bool target_contained_by(const struct ceph_osd_request_target *t,
4381 const struct ceph_hobject_id *begin,
4382 const struct ceph_hobject_id *end)
4383{
4384 struct ceph_hobject_id hoid;
4385 int cmp;
4386
4387 hoid_fill_from_target(&hoid, t);
4388 cmp = hoid_compare(&hoid, begin);
4389 return !cmp || (cmp > 0 && hoid_compare(&hoid, end) < 0);
4390}
4391
4392static void handle_backoff_unblock(struct ceph_osd *osd,
4393 const struct MOSDBackoff *m)
4394{
4395 struct ceph_spg_mapping *spg;
4396 struct ceph_osd_backoff *backoff;
4397 struct rb_node *n;
4398
4399 dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4400 m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4401
4402 backoff = lookup_backoff_by_id(&osd->o_backoffs_by_id, m->id);
4403 if (!backoff) {
4404 pr_err("%s osd%d spgid %llu.%xs%d id %llu backoff dne\n",
4405 __func__, osd->o_osd, m->spgid.pgid.pool,
4406 m->spgid.pgid.seed, m->spgid.shard, m->id);
4407 return;
4408 }
4409
4410 if (hoid_compare(backoff->begin, m->begin) &&
4411 hoid_compare(backoff->end, m->end)) {
4412 pr_err("%s osd%d spgid %llu.%xs%d id %llu bad range?\n",
4413 __func__, osd->o_osd, m->spgid.pgid.pool,
4414 m->spgid.pgid.seed, m->spgid.shard, m->id);
4415 /* unblock it anyway... */
4416 }
4417
4418 spg = lookup_spg_mapping(&osd->o_backoff_mappings, &backoff->spgid);
4419 BUG_ON(!spg);
4420
4421 erase_backoff(&spg->backoffs, backoff);
4422 erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4423 free_backoff(backoff);
4424
4425 if (RB_EMPTY_ROOT(&spg->backoffs)) {
4426 erase_spg_mapping(&osd->o_backoff_mappings, spg);
4427 free_spg_mapping(spg);
4428 }
4429
4430 for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
4431 struct ceph_osd_request *req =
4432 rb_entry(n, struct ceph_osd_request, r_node);
4433
4434 if (!ceph_spg_compare(&req->r_t.spgid, &m->spgid)) {
4435 /*
4436 * Match against @m, not @backoff -- the PG may
4437 * have split on the OSD.
4438 */
4439 if (target_contained_by(&req->r_t, m->begin, m->end)) {
4440 /*
4441 * If no other installed backoff applies,
4442 * resend.
4443 */
4444 send_request(req);
4445 }
4446 }
4447 }
4448}
4449
4450static void handle_backoff(struct ceph_osd *osd, struct ceph_msg *msg)
4451{
4452 struct ceph_osd_client *osdc = osd->o_osdc;
4453 struct MOSDBackoff m;
4454 int ret;
4455
4456 down_read(&osdc->lock);
4457 if (!osd_registered(osd)) {
4458 dout("%s osd%d unknown\n", __func__, osd->o_osd);
4459 up_read(&osdc->lock);
4460 return;
4461 }
4462 WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
4463
4464 mutex_lock(&osd->lock);
4465 ret = decode_MOSDBackoff(msg, &m);
4466 if (ret) {
4467 pr_err("failed to decode MOSDBackoff: %d\n", ret);
4468 ceph_msg_dump(msg);
4469 goto out_unlock;
4470 }
4471
4472 switch (m.op) {
4473 case CEPH_OSD_BACKOFF_OP_BLOCK:
4474 handle_backoff_block(osd, &m);
4475 break;
4476 case CEPH_OSD_BACKOFF_OP_UNBLOCK:
4477 handle_backoff_unblock(osd, &m);
4478 break;
4479 default:
4480 pr_err("%s osd%d unknown op %d\n", __func__, osd->o_osd, m.op);
4481 }
4482
4483 free_hoid(m.begin);
4484 free_hoid(m.end);
4485
4486out_unlock:
4487 mutex_unlock(&osd->lock);
4488 up_read(&osdc->lock);
4489}
4490
a40c4f10
YS		 4491/*
4492 * Process osd watch notifications
4493 */
3c663bbd
AE		 4494static void handle_watch_notify(struct ceph_osd_client *osdc,
4495 struct ceph_msg *msg)
a40c4f10 4496{
922dab61
ID		 4497 void *p = msg->front.iov_base;
4498 void *const end = p + msg->front.iov_len;
4499 struct ceph_osd_linger_request *lreq;
4500 struct linger_work *lwork;
4501 u8 proto_ver, opcode;
4502 u64 cookie, notify_id;
4503 u64 notifier_id = 0;
19079203 4504 s32 return_code = 0;
922dab61
ID		 4505 void *payload = NULL;
4506 u32 payload_len = 0;
a40c4f10
YS		 4507
4508 ceph_decode_8_safe(&p, end, proto_ver, bad);
4509 ceph_decode_8_safe(&p, end, opcode, bad);
4510 ceph_decode_64_safe(&p, end, cookie, bad);
922dab61 4511 p += 8; /* skip ver */
a40c4f10
YS		 4512 ceph_decode_64_safe(&p, end, notify_id, bad);
4513
922dab61
ID		 4514 if (proto_ver >= 1) {
4515 ceph_decode_32_safe(&p, end, payload_len, bad);
4516 ceph_decode_need(&p, end, payload_len, bad);
4517 payload = p;
4518 p += payload_len;
4519 }
4520
4521 if (le16_to_cpu(msg->hdr.version) >= 2)
19079203 4522 ceph_decode_32_safe(&p, end, return_code, bad);
922dab61
ID		 4523
4524 if (le16_to_cpu(msg->hdr.version) >= 3)
4525 ceph_decode_64_safe(&p, end, notifier_id, bad);
4526
4527 down_read(&osdc->lock);
4528 lreq = lookup_linger_osdc(&osdc->linger_requests, cookie);
4529 if (!lreq) {
4530 dout("%s opcode %d cookie %llu dne\n", __func__, opcode,
4531 cookie);
4532 goto out_unlock_osdc;
4533 }
4534
4535 mutex_lock(&lreq->lock);
19079203
ID		 4536 dout("%s opcode %d cookie %llu lreq %p is_watch %d\n", __func__,
4537 opcode, cookie, lreq, lreq->is_watch);
922dab61
ID		 4538 if (opcode == CEPH_WATCH_EVENT_DISCONNECT) {
4539 if (!lreq->last_error) {
4540 lreq->last_error = -ENOTCONN;
4541 queue_watch_error(lreq);
4542 }
19079203
ID		 4543 } else if (!lreq->is_watch) {
4544 /* CEPH_WATCH_EVENT_NOTIFY_COMPLETE */
4545 if (lreq->notify_id && lreq->notify_id != notify_id) {
4546 dout("lreq %p notify_id %llu != %llu, ignoring\n", lreq,
4547 lreq->notify_id, notify_id);
4548 } else if (!completion_done(&lreq->notify_finish_wait)) {
4549 struct ceph_msg_data *data =
0d9c1ab3 4550 msg->num_data_items ? &msg->data[0] : NULL;
19079203
ID		 4551
4552 if (data) {
4553 if (lreq->preply_pages) {
4554 WARN_ON(data->type !=
4555 CEPH_MSG_DATA_PAGES);
4556 *lreq->preply_pages = data->pages;
4557 *lreq->preply_len = data->length;
e8862740 4558 data->own_pages = false;
19079203
ID		 4559 }
4560 }
4561 lreq->notify_finish_error = return_code;
4562 complete_all(&lreq->notify_finish_wait);
4563 }
922dab61
ID		 4564 } else {
4565 /* CEPH_WATCH_EVENT_NOTIFY */
4566 lwork = lwork_alloc(lreq, do_watch_notify);
4567 if (!lwork) {
4568 pr_err("failed to allocate notify-lwork\n");
4569 goto out_unlock_lreq;
a40c4f10 4570 }
a40c4f10 4571
922dab61
ID		 4572 lwork->notify.notify_id = notify_id;
4573 lwork->notify.notifier_id = notifier_id;
4574 lwork->notify.payload = payload;
4575 lwork->notify.payload_len = payload_len;
4576 lwork->notify.msg = ceph_msg_get(msg);
4577 lwork_queue(lwork);
91883cd2 4578 }
a40c4f10 4579
922dab61
ID		 4580out_unlock_lreq:
4581 mutex_unlock(&lreq->lock);
4582out_unlock_osdc:
4583 up_read(&osdc->lock);
a40c4f10
YS		 4584 return;
4585
4586bad:
4587 pr_err("osdc handle_watch_notify corrupt msg\n");
a40c4f10
YS		 4588}
4589
70636773
AE		 4590/*
4591 * Register request, send initial attempt.
4592 */
a8af0d68
JL		 4593void ceph_osdc_start_request(struct ceph_osd_client *osdc,
4594 struct ceph_osd_request *req)
70636773 4595{
5aea3dcd
ID		 4596 down_read(&osdc->lock);
4597 submit_request(req, false);
4598 up_read(&osdc->lock);
f24e9980 4599}
3d14c5d2 4600EXPORT_SYMBOL(ceph_osdc_start_request);
f24e9980 4601
c9f9b93d 4602/*
d0bb883c
ID		 4603 * Unregister request. If @req was registered, it isn't completed:
4604 * r_result isn't set and __complete_request() isn't invoked.
4605 *
4606 * If @req wasn't registered, this call may have raced with
4607 * handle_reply(), in which case r_result would already be set and
4608 * __complete_request() would be getting invoked, possibly even
4609 * concurrently with this call.
c9f9b93d
ID		 4610 */
4611void ceph_osdc_cancel_request(struct ceph_osd_request *req)
4612{
4613 struct ceph_osd_client *osdc = req->r_osdc;
4614
5aea3dcd 4615 down_write(&osdc->lock);
5aea3dcd
ID		 4616 if (req->r_osd)
4617 cancel_request(req);
4618 up_write(&osdc->lock);
c9f9b93d
ID		 4619}
4620EXPORT_SYMBOL(ceph_osdc_cancel_request);
4621
f24e9980 4622/*
42b06965 4623 * @timeout: in jiffies, 0 means "wait forever"
f24e9980 4624 */
42b06965
ID		 4625static int wait_request_timeout(struct ceph_osd_request *req,
4626 unsigned long timeout)
f24e9980 4627{
42b06965 4628 long left;
c9f9b93d 4629
42b06965 4630 dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
0e76abf2 4631 left = wait_for_completion_killable_timeout(&req->r_completion,
42b06965
ID		 4632 ceph_timeout_jiffies(timeout));
4633 if (left <= 0) {
4634 left = left ?: -ETIMEDOUT;
c9f9b93d 4635 ceph_osdc_cancel_request(req);
42b06965
ID		 4636 } else {
4637 left = req->r_result; /* completed */
f24e9980
SW		 4638 }
4639
42b06965
ID		 4640 return left;
4641}
4642
4643/*
4644 * wait for a request to complete
4645 */
4646int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
4647 struct ceph_osd_request *req)
4648{
4649 return wait_request_timeout(req, 0);
f24e9980 4650}
3d14c5d2 4651EXPORT_SYMBOL(ceph_osdc_wait_request);
f24e9980
SW		 4652
4653/*
4654 * sync - wait for all in-flight requests to flush. avoid starvation.
4655 */
4656void ceph_osdc_sync(struct ceph_osd_client *osdc)
4657{
5aea3dcd
ID		 4658 struct rb_node *n, *p;
4659 u64 last_tid = atomic64_read(&osdc->last_tid);
f24e9980 4660
5aea3dcd
ID		 4661again:
4662 down_read(&osdc->lock);
4663 for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
4664 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
4665
4666 mutex_lock(&osd->lock);
4667 for (p = rb_first(&osd->o_requests); p; p = rb_next(p)) {
4668 struct ceph_osd_request *req =
4669 rb_entry(p, struct ceph_osd_request, r_node);
4670
4671 if (req->r_tid > last_tid)
4672 break;
4673
4674 if (!(req->r_flags & CEPH_OSD_FLAG_WRITE))
4675 continue;
f24e9980 4676
5aea3dcd
ID		 4677 ceph_osdc_get_request(req);
4678 mutex_unlock(&osd->lock);
4679 up_read(&osdc->lock);
4680 dout("%s waiting on req %p tid %llu last_tid %llu\n",
4681 __func__, req, req->r_tid, last_tid);
b18b9550 4682 wait_for_completion(&req->r_completion);
5aea3dcd
ID		 4683 ceph_osdc_put_request(req);
4684 goto again;
4685 }
f24e9980 4686
5aea3dcd 4687 mutex_unlock(&osd->lock);
f24e9980 4688 }
5aea3dcd
ID		 4689
4690 up_read(&osdc->lock);
4691 dout("%s done last_tid %llu\n", __func__, last_tid);
f24e9980 4692}
3d14c5d2 4693EXPORT_SYMBOL(ceph_osdc_sync);
f24e9980 4694
922dab61
ID		 4695/*
4696 * Returns a handle, caller owns a ref.
4697 */
4698struct ceph_osd_linger_request *
4699ceph_osdc_watch(struct ceph_osd_client *osdc,
4700 struct ceph_object_id *oid,
4701 struct ceph_object_locator *oloc,
4702 rados_watchcb2_t wcb,
4703 rados_watcherrcb_t errcb,
4704 void *data)
4705{
4706 struct ceph_osd_linger_request *lreq;
4707 int ret;
4708
4709 lreq = linger_alloc(osdc);
4710 if (!lreq)
4711 return ERR_PTR(-ENOMEM);
4712
19079203 4713 lreq->is_watch = true;
922dab61
ID		 4714 lreq->wcb = wcb;
4715 lreq->errcb = errcb;
4716 lreq->data = data;
b07d3c4b 4717 lreq->watch_valid_thru = jiffies;
922dab61
ID		 4718
4719 ceph_oid_copy(&lreq->t.base_oid, oid);
4720 ceph_oloc_copy(&lreq->t.base_oloc, oloc);
54ea0046 4721 lreq->t.flags = CEPH_OSD_FLAG_WRITE;
fac02ddf 4722 ktime_get_real_ts64(&lreq->mtime);
922dab61 4723
81c65213 4724 linger_submit(lreq);
922dab61
ID		 4725 ret = linger_reg_commit_wait(lreq);
4726 if (ret) {
4727 linger_cancel(lreq);
4728 goto err_put_lreq;
4729 }
4730
4731 return lreq;
4732
4733err_put_lreq:
4734 linger_put(lreq);
4735 return ERR_PTR(ret);
4736}
4737EXPORT_SYMBOL(ceph_osdc_watch);
4738
4739/*
4740 * Releases a ref.
4741 *
4742 * Times out after mount_timeout to preserve rbd unmap behaviour
4743 * introduced in 2894e1d76974 ("rbd: timeout watch teardown on unmap
4744 * with mount_timeout").
4745 */
4746int ceph_osdc_unwatch(struct ceph_osd_client *osdc,
4747 struct ceph_osd_linger_request *lreq)
4748{
4749 struct ceph_options *opts = osdc->client->options;
4750 struct ceph_osd_request *req;
4751 int ret;
4752
4753 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4754 if (!req)
4755 return -ENOMEM;
4756
4757 ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
4758 ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
54ea0046 4759 req->r_flags = CEPH_OSD_FLAG_WRITE;
fac02ddf 4760 ktime_get_real_ts64(&req->r_mtime);
75dbb685
ID		 4761 osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_UNWATCH,
4762 lreq->linger_id, 0);
922dab61
ID		 4763
4764 ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4765 if (ret)
4766 goto out_put_req;
4767
a8af0d68 4768 ceph_osdc_start_request(osdc, req);
922dab61
ID		 4769 linger_cancel(lreq);
4770 linger_put(lreq);
4771 ret = wait_request_timeout(req, opts->mount_timeout);
4772
4773out_put_req:
4774 ceph_osdc_put_request(req);
4775 return ret;
4776}
4777EXPORT_SYMBOL(ceph_osdc_unwatch);
4778
4779static int osd_req_op_notify_ack_init(struct ceph_osd_request *req, int which,
4780 u64 notify_id, u64 cookie, void *payload,
6d54228f 4781 u32 payload_len)
922dab61
ID		 4782{
4783 struct ceph_osd_req_op *op;
4784 struct ceph_pagelist *pl;
4785 int ret;
4786
042f6498 4787 op = osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY_ACK, 0);
922dab61 4788
33165d47 4789 pl = ceph_pagelist_alloc(GFP_NOIO);
922dab61
ID		 4790 if (!pl)
4791 return -ENOMEM;
4792
922dab61
ID		 4793 ret = ceph_pagelist_encode_64(pl, notify_id);
4794 ret |= ceph_pagelist_encode_64(pl, cookie);
4795 if (payload) {
4796 ret |= ceph_pagelist_encode_32(pl, payload_len);
4797 ret |= ceph_pagelist_append(pl, payload, payload_len);
4798 } else {
4799 ret |= ceph_pagelist_encode_32(pl, 0);
4800 }
4801 if (ret) {
4802 ceph_pagelist_release(pl);
4803 return -ENOMEM;
4804 }
4805
4806 ceph_osd_data_pagelist_init(&op->notify_ack.request_data, pl);
4807 op->indata_len = pl->length;
4808 return 0;
4809}
4810
4811int ceph_osdc_notify_ack(struct ceph_osd_client *osdc,
4812 struct ceph_object_id *oid,
4813 struct ceph_object_locator *oloc,
4814 u64 notify_id,
4815 u64 cookie,
4816 void *payload,
6d54228f 4817 u32 payload_len)
922dab61
ID		 4818{
4819 struct ceph_osd_request *req;
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
26f887e0
ID		 4830 ret = osd_req_op_notify_ack_init(req, 0, notify_id, cookie, payload,
4831 payload_len);
922dab61
ID		 4832 if (ret)
4833 goto out_put_req;
4834
26f887e0 4835 ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
922dab61
ID		 4836 if (ret)
4837 goto out_put_req;
4838
a8af0d68 4839 ceph_osdc_start_request(osdc, req);
922dab61
ID		 4840 ret = ceph_osdc_wait_request(osdc, req);
4841
4842out_put_req:
4843 ceph_osdc_put_request(req);
4844 return ret;
4845}
4846EXPORT_SYMBOL(ceph_osdc_notify_ack);
4847
19079203
ID		 4848/*
4849 * @timeout: in seconds
4850 *
4851 * @preply_{pages,len} are initialized both on success and error.
4852 * The caller is responsible for:
4853 *
4854 * ceph_release_page_vector(reply_pages, calc_pages_for(0, reply_len))
4855 */
4856int ceph_osdc_notify(struct ceph_osd_client *osdc,
4857 struct ceph_object_id *oid,
4858 struct ceph_object_locator *oloc,
4859 void *payload,
6d54228f 4860 u32 payload_len,
19079203
ID		 4861 u32 timeout,
4862 struct page ***preply_pages,
4863 size_t *preply_len)
4864{
4865 struct ceph_osd_linger_request *lreq;
19079203
ID		 4866 int ret;
4867
4868 WARN_ON(!timeout);
4869 if (preply_pages) {
4870 *preply_pages = NULL;
4871 *preply_len = 0;
4872 }
4873
4874 lreq = linger_alloc(osdc);
4875 if (!lreq)
4876 return -ENOMEM;
4877
75dbb685
ID		 4878 lreq->request_pl = ceph_pagelist_alloc(GFP_NOIO);
4879 if (!lreq->request_pl) {
19079203
ID		 4880 ret = -ENOMEM;
4881 goto out_put_lreq;
4882 }
4883
75dbb685
ID		 4884 ret = ceph_pagelist_encode_32(lreq->request_pl, 1); /* prot_ver */
4885 ret |= ceph_pagelist_encode_32(lreq->request_pl, timeout);
4886 ret |= ceph_pagelist_encode_32(lreq->request_pl, payload_len);
4887 ret |= ceph_pagelist_append(lreq->request_pl, payload, payload_len);
4888 if (ret) {
4889 ret = -ENOMEM;
81c65213 4890 goto out_put_lreq;
75dbb685 4891 }
81c65213 4892
19079203 4893 /* for notify_id */
75dbb685
ID		 4894 lreq->notify_id_pages = ceph_alloc_page_vector(1, GFP_NOIO);
4895 if (IS_ERR(lreq->notify_id_pages)) {
4896 ret = PTR_ERR(lreq->notify_id_pages);
4897 lreq->notify_id_pages = NULL;
19079203
ID		 4898 goto out_put_lreq;
4899 }
19079203 4900
75dbb685
ID		 4901 lreq->preply_pages = preply_pages;
4902 lreq->preply_len = preply_len;
4903
4904 ceph_oid_copy(&lreq->t.base_oid, oid);
4905 ceph_oloc_copy(&lreq->t.base_oloc, oloc);
4906 lreq->t.flags = CEPH_OSD_FLAG_READ;
26f887e0 4907
81c65213 4908 linger_submit(lreq);
19079203
ID		 4909 ret = linger_reg_commit_wait(lreq);
4910 if (!ret)
e6e28432
ID		 4911 ret = linger_notify_finish_wait(lreq,
4912 msecs_to_jiffies(2 * timeout * MSEC_PER_SEC));
19079203
ID		 4913 else
4914 dout("lreq %p failed to initiate notify %d\n", lreq, ret);
4915
4916 linger_cancel(lreq);
4917out_put_lreq:
4918 linger_put(lreq);
4919 return ret;
4920}
4921EXPORT_SYMBOL(ceph_osdc_notify);
4922
b07d3c4b
ID		 4923/*
4924 * Return the number of milliseconds since the watch was last
4925 * confirmed, or an error. If there is an error, the watch is no
4926 * longer valid, and should be destroyed with ceph_osdc_unwatch().
4927 */
4928int ceph_osdc_watch_check(struct ceph_osd_client *osdc,
4929 struct ceph_osd_linger_request *lreq)
4930{
4931 unsigned long stamp, age;
4932 int ret;
4933
4934 down_read(&osdc->lock);
4935 mutex_lock(&lreq->lock);
4936 stamp = lreq->watch_valid_thru;
4937 if (!list_empty(&lreq->pending_lworks)) {
4938 struct linger_work *lwork =
4939 list_first_entry(&lreq->pending_lworks,
4940 struct linger_work,
4941 pending_item);
4942
4943 if (time_before(lwork->queued_stamp, stamp))
4944 stamp = lwork->queued_stamp;
4945 }
4946 age = jiffies - stamp;
4947 dout("%s lreq %p linger_id %llu age %lu last_error %d\n", __func__,
4948 lreq, lreq->linger_id, age, lreq->last_error);
4949 /* we are truncating to msecs, so return a safe upper bound */
4950 ret = lreq->last_error ?: 1 + jiffies_to_msecs(age);
4951
4952 mutex_unlock(&lreq->lock);
4953 up_read(&osdc->lock);
4954 return ret;
4955}
4956
a4ed38d7
DF		 4957static int decode_watcher(void **p, void *end, struct ceph_watch_item *item)
4958{
4959 u8 struct_v;
4960 u32 struct_len;
4961 int ret;
4962
4963 ret = ceph_start_decoding(p, end, 2, "watch_item_t",
4964 &struct_v, &struct_len);
4965 if (ret)
51fc7ab4
JL		 4966 goto bad;
4967
4968 ret = -EINVAL;
4969 ceph_decode_copy_safe(p, end, &item->name, sizeof(item->name), bad);
4970 ceph_decode_64_safe(p, end, item->cookie, bad);
4971 ceph_decode_skip_32(p, end, bad); /* skip timeout seconds */
a4ed38d7 4972
a4ed38d7 4973 if (struct_v >= 2) {
51fc7ab4
JL		 4974 ret = ceph_decode_entity_addr(p, end, &item->addr);
4975 if (ret)
4976 goto bad;
4977 } else {
4978 ret = 0;
a4ed38d7
DF		 4979 }
4980
4981 dout("%s %s%llu cookie %llu addr %s\n", __func__,
4982 ENTITY_NAME(item->name), item->cookie,
b726ec97 4983 ceph_pr_addr(&item->addr));
51fc7ab4
JL		 4984bad:
4985 return ret;
a4ed38d7
DF		 4986}
4987
4988static int decode_watchers(void **p, void *end,
4989 struct ceph_watch_item **watchers,
4990 u32 *num_watchers)
4991{
4992 u8 struct_v;
4993 u32 struct_len;
4994 int i;
4995 int ret;
4996
4997 ret = ceph_start_decoding(p, end, 1, "obj_list_watch_response_t",
4998 &struct_v, &struct_len);
4999 if (ret)
5000 return ret;
5001
5002 *num_watchers = ceph_decode_32(p);
5003 *watchers = kcalloc(*num_watchers, sizeof(**watchers), GFP_NOIO);
5004 if (!*watchers)
5005 return -ENOMEM;
5006
5007 for (i = 0; i < *num_watchers; i++) {
5008 ret = decode_watcher(p, end, *watchers + i);
5009 if (ret) {
5010 kfree(*watchers);
5011 return ret;
5012 }
5013 }
5014
5015 return 0;
5016}
5017
5018/*
5019 * On success, the caller is responsible for:
5020 *
5021 * kfree(watchers);
5022 */
5023int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
5024 struct ceph_object_id *oid,
5025 struct ceph_object_locator *oloc,
5026 struct ceph_watch_item **watchers,
5027 u32 *num_watchers)
5028{
5029 struct ceph_osd_request *req;
5030 struct page **pages;
5031 int ret;
5032
5033 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
5034 if (!req)
5035 return -ENOMEM;
5036
5037 ceph_oid_copy(&req->r_base_oid, oid);
5038 ceph_oloc_copy(&req->r_base_oloc, oloc);
5039 req->r_flags = CEPH_OSD_FLAG_READ;
5040
a4ed38d7
DF		 5041 pages = ceph_alloc_page_vector(1, GFP_NOIO);
5042 if (IS_ERR(pages)) {
5043 ret = PTR_ERR(pages);
5044 goto out_put_req;
5045 }
5046
5047 osd_req_op_init(req, 0, CEPH_OSD_OP_LIST_WATCHERS, 0);
5048 ceph_osd_data_pages_init(osd_req_op_data(req, 0, list_watchers,
5049 response_data),
5050 pages, PAGE_SIZE, 0, false, true);
5051
26f887e0
ID		 5052 ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
5053 if (ret)
5054 goto out_put_req;
5055
a8af0d68 5056 ceph_osdc_start_request(osdc, req);
a4ed38d7
DF		 5057 ret = ceph_osdc_wait_request(osdc, req);
5058 if (ret >= 0) {
5059 void *p = page_address(pages[0]);
5060 void *const end = p + req->r_ops[0].outdata_len;
5061
5062 ret = decode_watchers(&p, end, watchers, num_watchers);
5063 }
5064
5065out_put_req:
5066 ceph_osdc_put_request(req);
5067 return ret;
5068}
5069EXPORT_SYMBOL(ceph_osdc_list_watchers);
5070
dd935f44
JD		 5071/*
5072 * Call all pending notify callbacks - for use after a watch is
5073 * unregistered, to make sure no more callbacks for it will be invoked
5074 */
f6479449 5075void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc)
dd935f44 5076{
99d16943 5077 dout("%s osdc %p\n", __func__, osdc);
dd935f44
JD		 5078 flush_workqueue(osdc->notify_wq);
5079}
5080EXPORT_SYMBOL(ceph_osdc_flush_notifies);
5081
7cca78c9
ID		 5082void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc)
5083{
5084 down_read(&osdc->lock);
5085 maybe_request_map(osdc);
5086 up_read(&osdc->lock);
5087}
5088EXPORT_SYMBOL(ceph_osdc_maybe_request_map);
dd935f44 5089
428a7158
DF		 5090/*
5091 * Execute an OSD class method on an object.
5092 *
5093 * @flags: CEPH_OSD_FLAG_*
2544a020 5094 * @resp_len: in/out param for reply length
428a7158
DF		 5095 */
5096int ceph_osdc_call(struct ceph_osd_client *osdc,
5097 struct ceph_object_id *oid,
5098 struct ceph_object_locator *oloc,
5099 const char *class, const char *method,
5100 unsigned int flags,
5101 struct page *req_page, size_t req_len,
68ada915 5102 struct page **resp_pages, size_t *resp_len)
428a7158
DF		 5103{
5104 struct ceph_osd_request *req;
5105 int ret;
5106
68ada915 5107 if (req_len > PAGE_SIZE)
2544a020
ID		 5108 return -E2BIG;
5109
428a7158
DF		 5110 req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
5111 if (!req)
5112 return -ENOMEM;
5113
5114 ceph_oid_copy(&req->r_base_oid, oid);
5115 ceph_oloc_copy(&req->r_base_oloc, oloc);
5116 req->r_flags = flags;
5117
24639ce5 5118 ret = osd_req_op_cls_init(req, 0, class, method);
fe943d50
CX		 5119 if (ret)
5120 goto out_put_req;
5121
428a7158
DF		 5122 if (req_page)
5123 osd_req_op_cls_request_data_pages(req, 0, &req_page, req_len,
5124 0, false, false);
68ada915
ID		 5125 if (resp_pages)
5126 osd_req_op_cls_response_data_pages(req, 0, resp_pages,
2544a020 5127 *resp_len, 0, false, false);
428a7158 5128
26f887e0
ID		 5129 ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
5130 if (ret)
5131 goto out_put_req;
5132
a8af0d68 5133 ceph_osdc_start_request(osdc, req);
428a7158
DF		 5134 ret = ceph_osdc_wait_request(osdc, req);
5135 if (ret >= 0) {
5136 ret = req->r_ops[0].rval;
68ada915 5137 if (resp_pages)
428a7158
DF		 5138 *resp_len = req->r_ops[0].outdata_len;
5139 }
5140
5141out_put_req:
5142 ceph_osdc_put_request(req);
5143 return ret;
5144}
5145EXPORT_SYMBOL(ceph_osdc_call);
5146
120a75ea
YZ		 5147/*
5148 * reset all osd connections
5149 */
5150void ceph_osdc_reopen_osds(struct ceph_osd_client *osdc)
5151{
5152 struct rb_node *n;
5153
5154 down_write(&osdc->lock);
5155 for (n = rb_first(&osdc->osds); n; ) {
5156 struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
5157
5158 n = rb_next(n);
5159 if (!reopen_osd(osd))
5160 kick_osd_requests(osd);
5161 }
5162 up_write(&osdc->lock);
5163}
5164
f24e9980
SW		 5165/*
5166 * init, shutdown
5167 */
5168int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client)
5169{
5170 int err;
5171
5172 dout("init\n");
5173 osdc->client = client;
5aea3dcd 5174 init_rwsem(&osdc->lock);
f24e9980 5175 osdc->osds = RB_ROOT;
f5a2041b 5176 INIT_LIST_HEAD(&osdc->osd_lru);
9dd2845c 5177 spin_lock_init(&osdc->osd_lru_lock);
5aea3dcd
ID		 5178 osd_init(&osdc->homeless_osd);
5179 osdc->homeless_osd.o_osdc = osdc;
5180 osdc->homeless_osd.o_osd = CEPH_HOMELESS_OSD;
264048af 5181 osdc->last_linger_id = CEPH_LINGER_ID_START;
922dab61 5182 osdc->linger_requests = RB_ROOT;
4609245e
ID		 5183 osdc->map_checks = RB_ROOT;
5184 osdc->linger_map_checks = RB_ROOT;
f24e9980 5185 INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout);
f5a2041b
YS		 5186 INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout);
5187
5f44f142 5188 err = -ENOMEM;
e5253a7b
ID		 5189 osdc->osdmap = ceph_osdmap_alloc();
5190 if (!osdc->osdmap)
5191 goto out;
5192
9e767adb
ID		 5193 osdc->req_mempool = mempool_create_slab_pool(10,
5194 ceph_osd_request_cache);
f24e9980 5195 if (!osdc->req_mempool)
e5253a7b 5196 goto out_map;
f24e9980 5197
d50b409f 5198 err = ceph_msgpool_init(&osdc->msgpool_op, CEPH_MSG_OSD_OP,
0d9c1ab3 5199 PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10, "osd_op");
f24e9980 5200 if (err < 0)
5f44f142 5201 goto out_mempool;
d50b409f 5202 err = ceph_msgpool_init(&osdc->msgpool_op_reply, CEPH_MSG_OSD_OPREPLY,
0d9c1ab3
ID		 5203 PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10,
5204 "osd_op_reply");
c16e7869
SW		 5205 if (err < 0)
5206 goto out_msgpool;
a40c4f10 5207
dbcae088 5208 err = -ENOMEM;
a40c4f10 5209 osdc->notify_wq = create_singlethread_workqueue("ceph-watch-notify");
dbcae088 5210 if (!osdc->notify_wq)
c172ec5c
ID		 5211 goto out_msgpool_reply;
5212
88bc1922
ID		 5213 osdc->completion_wq = create_singlethread_workqueue("ceph-completion");
5214 if (!osdc->completion_wq)
5215 goto out_notify_wq;
5216
fbca9635
ID		 5217 schedule_delayed_work(&osdc->timeout_work,
5218 osdc->client->options->osd_keepalive_timeout);
b37ee1b9
ID		 5219 schedule_delayed_work(&osdc->osds_timeout_work,
5220 round_jiffies_relative(osdc->client->options->osd_idle_ttl));
5221
f24e9980 5222 return 0;
5f44f142 5223
88bc1922
ID		 5224out_notify_wq:
5225 destroy_workqueue(osdc->notify_wq);
c172ec5c
ID		 5226out_msgpool_reply:
5227 ceph_msgpool_destroy(&osdc->msgpool_op_reply);
c16e7869
SW		 5228out_msgpool:
5229 ceph_msgpool_destroy(&osdc->msgpool_op);
5f44f142
SW		 5230out_mempool:
5231 mempool_destroy(osdc->req_mempool);
e5253a7b
ID		 5232out_map:
5233 ceph_osdmap_destroy(osdc->osdmap);
5f44f142
SW		 5234out:
5235 return err;
f24e9980
SW		 5236}
5237
5238void ceph_osdc_stop(struct ceph_osd_client *osdc)
5239{
88bc1922 5240 destroy_workqueue(osdc->completion_wq);
a40c4f10 5241 destroy_workqueue(osdc->notify_wq);
f24e9980 5242 cancel_delayed_work_sync(&osdc->timeout_work);
f5a2041b 5243 cancel_delayed_work_sync(&osdc->osds_timeout_work);
42a2c09f 5244
5aea3dcd 5245 down_write(&osdc->lock);
42a2c09f
ID		 5246 while (!RB_EMPTY_ROOT(&osdc->osds)) {
5247 struct ceph_osd *osd = rb_entry(rb_first(&osdc->osds),
5248 struct ceph_osd, o_node);
5aea3dcd 5249 close_osd(osd);
42a2c09f 5250 }
5aea3dcd 5251 up_write(&osdc->lock);
02113a0f 5252 WARN_ON(refcount_read(&osdc->homeless_osd.o_ref) != 1);
5aea3dcd
ID		 5253 osd_cleanup(&osdc->homeless_osd);
5254
5255 WARN_ON(!list_empty(&osdc->osd_lru));
922dab61 5256 WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_requests));
4609245e
ID		 5257 WARN_ON(!RB_EMPTY_ROOT(&osdc->map_checks));
5258 WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_map_checks));
5aea3dcd
ID		 5259 WARN_ON(atomic_read(&osdc->num_requests));
5260 WARN_ON(atomic_read(&osdc->num_homeless));
42a2c09f 5261
e5253a7b 5262 ceph_osdmap_destroy(osdc->osdmap);
f24e9980
SW		 5263 mempool_destroy(osdc->req_mempool);
5264 ceph_msgpool_destroy(&osdc->msgpool_op);
c16e7869 5265 ceph_msgpool_destroy(&osdc->msgpool_op_reply);
f24e9980
SW		 5266}
5267
aca39d9e
LH		 5268int osd_req_op_copy_from_init(struct ceph_osd_request *req,
5269 u64 src_snapid, u64 src_version,
5270 struct ceph_object_id *src_oid,
5271 struct ceph_object_locator *src_oloc,
5272 u32 src_fadvise_flags,
5273 u32 dst_fadvise_flags,
5274 u32 truncate_seq, u64 truncate_size,
5275 u8 copy_from_flags)
23ddf9be
LH		 5276{
5277 struct ceph_osd_req_op *op;
5278 struct page **pages;
5279 void *p, *end;
5280
5281 pages = ceph_alloc_page_vector(1, GFP_KERNEL);
5282 if (IS_ERR(pages))
5283 return PTR_ERR(pages);
5284
042f6498
JL		 5285 op = osd_req_op_init(req, 0, CEPH_OSD_OP_COPY_FROM2,
5286 dst_fadvise_flags);
23ddf9be
LH		 5287 op->copy_from.snapid = src_snapid;
5288 op->copy_from.src_version = src_version;
5289 op->copy_from.flags = copy_from_flags;
5290 op->copy_from.src_fadvise_flags = src_fadvise_flags;
5291
5292 p = page_address(pages[0]);
5293 end = p + PAGE_SIZE;
5294 ceph_encode_string(&p, end, src_oid->name, src_oid->name_len);
5295 encode_oloc(&p, end, src_oloc);
78beb0ff
LH		 5296 ceph_encode_32(&p, truncate_seq);
5297 ceph_encode_64(&p, truncate_size);
23ddf9be
LH		 5298 op->indata_len = PAGE_SIZE - (end - p);
5299
5300 ceph_osd_data_pages_init(&op->copy_from.osd_data, pages,
5301 op->indata_len, 0, false, true);
5302 return 0;
5303}
aca39d9e 5304EXPORT_SYMBOL(osd_req_op_copy_from_init);
23ddf9be 5305
57a35dfb 5306int __init ceph_osdc_setup(void)
5522ae0b 5307{
3f1af42a
ID		 5308 size_t size = sizeof(struct ceph_osd_request) +
5309 CEPH_OSD_SLAB_OPS * sizeof(struct ceph_osd_req_op);
5310
5522ae0b 5311 BUG_ON(ceph_osd_request_cache);
3f1af42a
ID		 5312 ceph_osd_request_cache = kmem_cache_create("ceph_osd_request", size,
5313 0, 0, NULL);
5522ae0b
AE		 5314
5315 return ceph_osd_request_cache ? 0 : -ENOMEM;
5316}
5522ae0b
AE		 5317
5318void ceph_osdc_cleanup(void)
5319{
5320 BUG_ON(!ceph_osd_request_cache);
5321 kmem_cache_destroy(ceph_osd_request_cache);
5322 ceph_osd_request_cache = NULL;
5323}
5522ae0b 5324
f24e9980
SW		 5325/*
5326 * handle incoming message
5327 */
4972cf60 5328static void osd_dispatch(struct ceph_connection *con, struct ceph_msg *msg)
f24e9980
SW		 5329{
5330 struct ceph_osd *osd = con->private;
5aea3dcd 5331 struct ceph_osd_client *osdc = osd->o_osdc;
f24e9980
SW		 5332 int type = le16_to_cpu(msg->hdr.type);
5333
f24e9980
SW		 5334 switch (type) {
5335 case CEPH_MSG_OSD_MAP:
5336 ceph_osdc_handle_map(osdc, msg);
5337 break;
5338 case CEPH_MSG_OSD_OPREPLY:
5aea3dcd 5339 handle_reply(osd, msg);
f24e9980 5340 break;
a02a946d
ID		 5341 case CEPH_MSG_OSD_BACKOFF:
5342 handle_backoff(osd, msg);
5343 break;
a40c4f10
YS		 5344 case CEPH_MSG_WATCH_NOTIFY:
5345 handle_watch_notify(osdc, msg);
5346 break;
f24e9980
SW		 5347
5348 default:
5349 pr_err("received unknown message type %d %s\n", type,
5350 ceph_msg_type_name(type));
5351 }
5aea3dcd 5352
f24e9980
SW		 5353 ceph_msg_put(msg);
5354}
5355
5b3a4db3 5356/*
d15f9d69
ID		 5357 * Lookup and return message for incoming reply. Don't try to do
5358 * anything about a larger than preallocated data portion of the
5359 * message at the moment - for now, just skip the message.
5b3a4db3
SW		 5360 */
5361static struct ceph_msg *get_reply(struct ceph_connection *con,
2450418c
YS		 5362 struct ceph_msg_header *hdr,
5363 int *skip)
f24e9980
SW		 5364{
5365 struct ceph_osd *osd = con->private;
5366 struct ceph_osd_client *osdc = osd->o_osdc;
5aea3dcd 5367 struct ceph_msg *m = NULL;
0547a9b3 5368 struct ceph_osd_request *req;
3f0a4ac5 5369 int front_len = le32_to_cpu(hdr->front_len);
5b3a4db3 5370 int data_len = le32_to_cpu(hdr->data_len);
5aea3dcd 5371 u64 tid = le64_to_cpu(hdr->tid);
f24e9980 5372
5aea3dcd
ID		 5373 down_read(&osdc->lock);
5374 if (!osd_registered(osd)) {
5375 dout("%s osd%d unknown, skipping\n", __func__, osd->o_osd);
5376 *skip = 1;
5377 goto out_unlock_osdc;
5378 }
5379 WARN_ON(osd->o_osd != le64_to_cpu(hdr->src.num));
5380
5381 mutex_lock(&osd->lock);
5382 req = lookup_request(&osd->o_requests, tid);
0547a9b3 5383 if (!req) {
cd8140c6
ID		 5384 dout("%s osd%d tid %llu unknown, skipping\n", __func__,
5385 osd->o_osd, tid);
d15f9d69 5386 *skip = 1;
5aea3dcd 5387 goto out_unlock_session;
0547a9b3 5388 }
c16e7869 5389
ace6d3a9 5390 ceph_msg_revoke_incoming(req->r_reply);
0547a9b3 5391
f2be82b0 5392 if (front_len > req->r_reply->front_alloc_len) {
d15f9d69
ID		 5393 pr_warn("%s osd%d tid %llu front %d > preallocated %d\n",
5394 __func__, osd->o_osd, req->r_tid, front_len,
5395 req->r_reply->front_alloc_len);
3f0a4ac5
ID		 5396 m = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, front_len, GFP_NOFS,
5397 false);
a79832f2 5398 if (!m)
5aea3dcd 5399 goto out_unlock_session;
c16e7869
SW		 5400 ceph_msg_put(req->r_reply);
5401 req->r_reply = m;
5402 }
0fff87ec 5403
d15f9d69
ID		 5404 if (data_len > req->r_reply->data_length) {
5405 pr_warn("%s osd%d tid %llu data %d > preallocated %zu, skipping\n",
5406 __func__, osd->o_osd, req->r_tid, data_len,
5407 req->r_reply->data_length);
5408 m = NULL;
5409 *skip = 1;
5aea3dcd 5410 goto out_unlock_session;
0547a9b3 5411 }
d15f9d69
ID		 5412
5413 m = ceph_msg_get(req->r_reply);
c16e7869 5414 dout("get_reply tid %lld %p\n", tid, m);
0547a9b3 5415
5aea3dcd
ID		 5416out_unlock_session:
5417 mutex_unlock(&osd->lock);
5418out_unlock_osdc:
5419 up_read(&osdc->lock);
2450418c 5420 return m;
5b3a4db3
SW		 5421}
5422
19079203
ID		 5423static struct ceph_msg *alloc_msg_with_page_vector(struct ceph_msg_header *hdr)
5424{
5425 struct ceph_msg *m;
5426 int type = le16_to_cpu(hdr->type);
5427 u32 front_len = le32_to_cpu(hdr->front_len);
5428 u32 data_len = le32_to_cpu(hdr->data_len);
5429
0d9c1ab3 5430 m = ceph_msg_new2(type, front_len, 1, GFP_NOIO, false);
19079203
ID		 5431 if (!m)
5432 return NULL;
5433
5434 if (data_len) {
5435 struct page **pages;
19079203
ID		 5436
5437 pages = ceph_alloc_page_vector(calc_pages_for(0, data_len),
5438 GFP_NOIO);
c22e853a 5439 if (IS_ERR(pages)) {
19079203
ID		 5440 ceph_msg_put(m);
5441 return NULL;
5442 }
5443
e8862740 5444 ceph_msg_data_add_pages(m, pages, data_len, 0, true);
19079203
ID		 5445 }
5446
5447 return m;
5448}
5449
4972cf60
ID		 5450static struct ceph_msg *osd_alloc_msg(struct ceph_connection *con,
5451 struct ceph_msg_header *hdr,
5452 int *skip)
5b3a4db3
SW		 5453{
5454 struct ceph_osd *osd = con->private;
5455 int type = le16_to_cpu(hdr->type);
5b3a4db3 5456
1c20f2d2 5457 *skip = 0;
5b3a4db3
SW		 5458 switch (type) {
5459 case CEPH_MSG_OSD_MAP:
a02a946d 5460 case CEPH_MSG_OSD_BACKOFF:
a40c4f10 5461 case CEPH_MSG_WATCH_NOTIFY:
19079203 5462 return alloc_msg_with_page_vector(hdr);
5b3a4db3
SW		 5463 case CEPH_MSG_OSD_OPREPLY:
5464 return get_reply(con, hdr, skip);
5465 default:
5aea3dcd
ID		 5466 pr_warn("%s osd%d unknown msg type %d, skipping\n", __func__,
5467 osd->o_osd, type);
5b3a4db3
SW		 5468 *skip = 1;
5469 return NULL;
5470 }
f24e9980
SW		 5471}
5472
5473/*
5474 * Wrappers to refcount containing ceph_osd struct
5475 */
4972cf60 5476static struct ceph_connection *osd_get_con(struct ceph_connection *con)
f24e9980
SW		 5477{
5478 struct ceph_osd *osd = con->private;
5479 if (get_osd(osd))
5480 return con;
5481 return NULL;
5482}
5483
4972cf60 5484static void osd_put_con(struct ceph_connection *con)
f24e9980
SW		 5485{
5486 struct ceph_osd *osd = con->private;
5487 put_osd(osd);
5488}
5489
4e7a5dcd
SW		 5490/*
5491 * authentication
5492 */
cd1a677c 5493
a3530df3
AE		 5494/*
5495 * Note: returned pointer is the address of a structure that's
5496 * managed separately. Caller must *not* attempt to free it.
5497 */
4972cf60
ID		 5498static struct ceph_auth_handshake *
5499osd_get_authorizer(struct ceph_connection *con, int *proto, int force_new)
4e7a5dcd
SW		 5500{
5501 struct ceph_osd *o = con->private;
5502 struct ceph_osd_client *osdc = o->o_osdc;
5503 struct ceph_auth_client *ac = osdc->client->monc.auth;
74f1869f 5504 struct ceph_auth_handshake *auth = &o->o_auth;
ce287162 5505 int ret;
4e7a5dcd 5506
ce287162
ID		 5507 ret = __ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_OSD,
5508 force_new, proto, NULL, NULL);
5509 if (ret)
5510 return ERR_PTR(ret);
74f1869f 5511
a3530df3 5512 return auth;
4e7a5dcd
SW		 5513}
5514
4972cf60 5515static int osd_add_authorizer_challenge(struct ceph_connection *con,
6daca13d
ID		 5516 void *challenge_buf, int challenge_buf_len)
5517{
5518 struct ceph_osd *o = con->private;
5519 struct ceph_osd_client *osdc = o->o_osdc;
5520 struct ceph_auth_client *ac = osdc->client->monc.auth;
5521
5522 return ceph_auth_add_authorizer_challenge(ac, o->o_auth.authorizer,
5523 challenge_buf, challenge_buf_len);
5524}
4e7a5dcd 5525
4972cf60 5526static int osd_verify_authorizer_reply(struct ceph_connection *con)
4e7a5dcd
SW		 5527{
5528 struct ceph_osd *o = con->private;
5529 struct ceph_osd_client *osdc = o->o_osdc;
5530 struct ceph_auth_client *ac = osdc->client->monc.auth;
285ea34f 5531 struct ceph_auth_handshake *auth = &o->o_auth;
4e7a5dcd 5532
285ea34f
ID		 5533 return ceph_auth_verify_authorizer_reply(ac, auth->authorizer,
5534 auth->authorizer_reply_buf, auth->authorizer_reply_buf_len,
5535 NULL, NULL, NULL, NULL);
4e7a5dcd
SW		 5536}
5537
4972cf60 5538static int osd_invalidate_authorizer(struct ceph_connection *con)
9bd2e6f8
SW		 5539{
5540 struct ceph_osd *o = con->private;
5541 struct ceph_osd_client *osdc = o->o_osdc;
5542 struct ceph_auth_client *ac = osdc->client->monc.auth;
5543
27859f97 5544 ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_OSD);
9bd2e6f8
SW		 5545 return ceph_monc_validate_auth(&osdc->client->monc);
5546}
4e7a5dcd 5547
cd1a677c
ID		 5548static int osd_get_auth_request(struct ceph_connection *con,
5549 void *buf, int *buf_len,
5550 void **authorizer, int *authorizer_len)
5551{
5552 struct ceph_osd *o = con->private;
5553 struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5554 struct ceph_auth_handshake *auth = &o->o_auth;
5555 int ret;
5556
5557 ret = ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_OSD,
5558 buf, buf_len);
5559 if (ret)
5560 return ret;
5561
5562 *authorizer = auth->authorizer_buf;
5563 *authorizer_len = auth->authorizer_buf_len;
5564 return 0;
5565}
5566
5567static int osd_handle_auth_reply_more(struct ceph_connection *con,
5568 void *reply, int reply_len,
5569 void *buf, int *buf_len,
5570 void **authorizer, int *authorizer_len)
5571{
5572 struct ceph_osd *o = con->private;
5573 struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5574 struct ceph_auth_handshake *auth = &o->o_auth;
5575 int ret;
5576
5577 ret = ceph_auth_handle_svc_reply_more(ac, auth, reply, reply_len,
5578 buf, buf_len);
5579 if (ret)
5580 return ret;
5581
5582 *authorizer = auth->authorizer_buf;
5583 *authorizer_len = auth->authorizer_buf_len;
5584 return 0;
5585}
5586
5587static int osd_handle_auth_done(struct ceph_connection *con,
5588 u64 global_id, void *reply, int reply_len,
5589 u8 *session_key, int *session_key_len,
5590 u8 *con_secret, int *con_secret_len)
5591{
5592 struct ceph_osd *o = con->private;
5593 struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5594 struct ceph_auth_handshake *auth = &o->o_auth;
5595
5596 return ceph_auth_handle_svc_reply_done(ac, auth, reply, reply_len,
5597 session_key, session_key_len,
5598 con_secret, con_secret_len);
5599}
5600
5601static int osd_handle_auth_bad_method(struct ceph_connection *con,
5602 int used_proto, int result,
5603 const int *allowed_protos, int proto_cnt,
5604 const int *allowed_modes, int mode_cnt)
5605{
5606 struct ceph_osd *o = con->private;
5607 struct ceph_mon_client *monc = &o->o_osdc->client->monc;
5608 int ret;
5609
5610 if (ceph_auth_handle_bad_authorizer(monc->auth, CEPH_ENTITY_TYPE_OSD,
5611 used_proto, result,
5612 allowed_protos, proto_cnt,
5613 allowed_modes, mode_cnt)) {
5614 ret = ceph_monc_validate_auth(monc);
5615 if (ret)
5616 return ret;
5617 }
5618
5619 return -EACCES;
5620}
5621
8cb441c0
ID		 5622static void osd_reencode_message(struct ceph_msg *msg)
5623{
914902af
ID		 5624 int type = le16_to_cpu(msg->hdr.type);
5625
5626 if (type == CEPH_MSG_OSD_OP)
5627 encode_request_finish(msg);
8cb441c0
ID		 5628}
5629
79dbd1ba 5630static int osd_sign_message(struct ceph_msg *msg)
33d07337 5631{
79dbd1ba 5632 struct ceph_osd *o = msg->con->private;
33d07337 5633 struct ceph_auth_handshake *auth = &o->o_auth;
79dbd1ba 5634
33d07337
YZ		 5635 return ceph_auth_sign_message(auth, msg);
5636}
5637
79dbd1ba 5638static int osd_check_message_signature(struct ceph_msg *msg)
33d07337 5639{
79dbd1ba 5640 struct ceph_osd *o = msg->con->private;
33d07337 5641 struct ceph_auth_handshake *auth = &o->o_auth;
79dbd1ba 5642
33d07337
YZ		 5643 return ceph_auth_check_message_signature(auth, msg);
5644}
5645
9e32789f 5646static const struct ceph_connection_operations osd_con_ops = {
4972cf60
ID		 5647 .get = osd_get_con,
5648 .put = osd_put_con,
5649 .alloc_msg = osd_alloc_msg,
5650 .dispatch = osd_dispatch,
5651 .fault = osd_fault,
8cb441c0 5652 .reencode_message = osd_reencode_message,
4972cf60
ID		 5653 .get_authorizer = osd_get_authorizer,
5654 .add_authorizer_challenge = osd_add_authorizer_challenge,
5655 .verify_authorizer_reply = osd_verify_authorizer_reply,
5656 .invalidate_authorizer = osd_invalidate_authorizer,
79dbd1ba
ID		 5657 .sign_message = osd_sign_message,
5658 .check_message_signature = osd_check_message_signature,
cd1a677c
ID		 5659 .get_auth_request = osd_get_auth_request,
5660 .handle_auth_reply_more = osd_handle_auth_reply_more,
5661 .handle_auth_done = osd_handle_auth_done,
5662 .handle_auth_bad_method = osd_handle_auth_bad_method,
f24e9980 5663};
Linux 6.x block layer and io_uring tree(s)