1 // SPDX-License-Identifier: GPL-2.0
2 
3 #include <linux/ceph/ceph_debug.h>
4 
5 #include <linux/module.h>
6 #include <linux/err.h>
7 #include <linux/highmem.h>
8 #include <linux/mm.h>
9 #include <linux/pagemap.h>
10 #include <linux/slab.h>
11 #include <linux/uaccess.h>
12 #ifdef CONFIG_BLOCK
13 #include <linux/bio.h>
14 #endif
15 
16 #include <linux/ceph/ceph_features.h>
17 #include <linux/ceph/libceph.h>
18 #include <linux/ceph/osd_client.h>
19 #include <linux/ceph/messenger.h>
20 #include <linux/ceph/decode.h>
21 #include <linux/ceph/auth.h>
22 #include <linux/ceph/pagelist.h>
23 #include <linux/ceph/striper.h>
24 
25 #define OSD_OPREPLY_FRONT_LEN	512
26 
27 static struct kmem_cache	*ceph_osd_request_cache;
28 
29 static const struct ceph_connection_operations osd_con_ops;
30 
31 /*
32  * Implement client access to distributed object storage cluster.
33  *
34  * All data objects are stored within a cluster/cloud of OSDs, or
35  * "object storage devices."  (Note that Ceph OSDs have _nothing_ to
36  * do with the T10 OSD extensions to SCSI.)  Ceph OSDs are simply
37  * remote daemons serving up and coordinating consistent and safe
38  * access to storage.
39  *
40  * Cluster membership and the mapping of data objects onto storage devices
41  * are described by the osd map.
42  *
43  * We keep track of pending OSD requests (read, write), resubmit
44  * requests to different OSDs when the cluster topology/data layout
45  * change, or retry the affected requests when the communications
46  * channel with an OSD is reset.
47  */
48 
49 static void link_request(struct ceph_osd *osd, struct ceph_osd_request *req);
50 static void unlink_request(struct ceph_osd *osd, struct ceph_osd_request *req);
51 static void link_linger(struct ceph_osd *osd,
52 			struct ceph_osd_linger_request *lreq);
53 static void unlink_linger(struct ceph_osd *osd,
54 			  struct ceph_osd_linger_request *lreq);
55 static void clear_backoffs(struct ceph_osd *osd);
56 
57 #if 1
rwsem_is_wrlocked(struct rw_semaphore * sem)58 static inline bool rwsem_is_wrlocked(struct rw_semaphore *sem)
59 {
60 	bool wrlocked = true;
61 
62 	if (unlikely(down_read_trylock(sem))) {
63 		wrlocked = false;
64 		up_read(sem);
65 	}
66 
67 	return wrlocked;
68 }
verify_osdc_locked(struct ceph_osd_client * osdc)69 static inline void verify_osdc_locked(struct ceph_osd_client *osdc)
70 {
71 	WARN_ON(!rwsem_is_locked(&osdc->lock));
72 }
verify_osdc_wrlocked(struct ceph_osd_client * osdc)73 static inline void verify_osdc_wrlocked(struct ceph_osd_client *osdc)
74 {
75 	WARN_ON(!rwsem_is_wrlocked(&osdc->lock));
76 }
verify_osd_locked(struct ceph_osd * osd)77 static inline void verify_osd_locked(struct ceph_osd *osd)
78 {
79 	struct ceph_osd_client *osdc = osd->o_osdc;
80 
81 	WARN_ON(!(mutex_is_locked(&osd->lock) &&
82 		  rwsem_is_locked(&osdc->lock)) &&
83 		!rwsem_is_wrlocked(&osdc->lock));
84 }
verify_lreq_locked(struct ceph_osd_linger_request * lreq)85 static inline void verify_lreq_locked(struct ceph_osd_linger_request *lreq)
86 {
87 	WARN_ON(!mutex_is_locked(&lreq->lock));
88 }
89 #else
verify_osdc_locked(struct ceph_osd_client * osdc)90 static inline void verify_osdc_locked(struct ceph_osd_client *osdc) { }
verify_osdc_wrlocked(struct ceph_osd_client * osdc)91 static inline void verify_osdc_wrlocked(struct ceph_osd_client *osdc) { }
verify_osd_locked(struct ceph_osd * osd)92 static inline void verify_osd_locked(struct ceph_osd *osd) { }
verify_lreq_locked(struct ceph_osd_linger_request * lreq)93 static inline void verify_lreq_locked(struct ceph_osd_linger_request *lreq) { }
94 #endif
95 
96 /*
97  * calculate the mapping of a file extent onto an object, and fill out the
98  * request accordingly.  shorten extent as necessary if it crosses an
99  * object boundary.
100  *
101  * fill osd op in request message.
102  */
calc_layout(struct ceph_file_layout * layout,u64 off,u64 * plen,u64 * objnum,u64 * objoff,u64 * objlen)103 static int calc_layout(struct ceph_file_layout *layout, u64 off, u64 *plen,
104 			u64 *objnum, u64 *objoff, u64 *objlen)
105 {
106 	u64 orig_len = *plen;
107 	u32 xlen;
108 
109 	/* object extent? */
110 	ceph_calc_file_object_mapping(layout, off, orig_len, objnum,
111 					  objoff, &xlen);
112 	*objlen = xlen;
113 	if (*objlen < orig_len) {
114 		*plen = *objlen;
115 		dout(" skipping last %llu, final file extent %llu~%llu\n",
116 		     orig_len - *plen, off, *plen);
117 	}
118 
119 	dout("calc_layout objnum=%llx %llu~%llu\n", *objnum, *objoff, *objlen);
120 	return 0;
121 }
122 
ceph_osd_data_init(struct ceph_osd_data * osd_data)123 static void ceph_osd_data_init(struct ceph_osd_data *osd_data)
124 {
125 	memset(osd_data, 0, sizeof (*osd_data));
126 	osd_data->type = CEPH_OSD_DATA_TYPE_NONE;
127 }
128 
129 /*
130  * Consumes @pages if @own_pages is true.
131  */
ceph_osd_data_pages_init(struct ceph_osd_data * osd_data,struct page ** pages,u64 length,u32 alignment,bool pages_from_pool,bool own_pages)132 static void ceph_osd_data_pages_init(struct ceph_osd_data *osd_data,
133 			struct page **pages, u64 length, u32 alignment,
134 			bool pages_from_pool, bool own_pages)
135 {
136 	osd_data->type = CEPH_OSD_DATA_TYPE_PAGES;
137 	osd_data->pages = pages;
138 	osd_data->length = length;
139 	osd_data->alignment = alignment;
140 	osd_data->pages_from_pool = pages_from_pool;
141 	osd_data->own_pages = own_pages;
142 }
143 
144 /*
145  * Consumes a ref on @pagelist.
146  */
ceph_osd_data_pagelist_init(struct ceph_osd_data * osd_data,struct ceph_pagelist * pagelist)147 static void ceph_osd_data_pagelist_init(struct ceph_osd_data *osd_data,
148 			struct ceph_pagelist *pagelist)
149 {
150 	osd_data->type = CEPH_OSD_DATA_TYPE_PAGELIST;
151 	osd_data->pagelist = pagelist;
152 }
153 
154 #ifdef CONFIG_BLOCK
ceph_osd_data_bio_init(struct ceph_osd_data * osd_data,struct ceph_bio_iter * bio_pos,u32 bio_length)155 static void ceph_osd_data_bio_init(struct ceph_osd_data *osd_data,
156 				   struct ceph_bio_iter *bio_pos,
157 				   u32 bio_length)
158 {
159 	osd_data->type = CEPH_OSD_DATA_TYPE_BIO;
160 	osd_data->bio_pos = *bio_pos;
161 	osd_data->bio_length = bio_length;
162 }
163 #endif /* CONFIG_BLOCK */
164 
ceph_osd_data_bvecs_init(struct ceph_osd_data * osd_data,struct ceph_bvec_iter * bvec_pos,u32 num_bvecs)165 static void ceph_osd_data_bvecs_init(struct ceph_osd_data *osd_data,
166 				     struct ceph_bvec_iter *bvec_pos,
167 				     u32 num_bvecs)
168 {
169 	osd_data->type = CEPH_OSD_DATA_TYPE_BVECS;
170 	osd_data->bvec_pos = *bvec_pos;
171 	osd_data->num_bvecs = num_bvecs;
172 }
173 
174 static struct ceph_osd_data *
osd_req_op_raw_data_in(struct ceph_osd_request * osd_req,unsigned int which)175 osd_req_op_raw_data_in(struct ceph_osd_request *osd_req, unsigned int which)
176 {
177 	BUG_ON(which >= osd_req->r_num_ops);
178 
179 	return &osd_req->r_ops[which].raw_data_in;
180 }
181 
182 struct ceph_osd_data *
osd_req_op_extent_osd_data(struct ceph_osd_request * osd_req,unsigned int which)183 osd_req_op_extent_osd_data(struct ceph_osd_request *osd_req,
184 			unsigned int which)
185 {
186 	return osd_req_op_data(osd_req, which, extent, osd_data);
187 }
188 EXPORT_SYMBOL(osd_req_op_extent_osd_data);
189 
osd_req_op_raw_data_in_pages(struct ceph_osd_request * osd_req,unsigned int which,struct page ** pages,u64 length,u32 alignment,bool pages_from_pool,bool own_pages)190 void osd_req_op_raw_data_in_pages(struct ceph_osd_request *osd_req,
191 			unsigned int which, struct page **pages,
192 			u64 length, u32 alignment,
193 			bool pages_from_pool, bool own_pages)
194 {
195 	struct ceph_osd_data *osd_data;
196 
197 	osd_data = osd_req_op_raw_data_in(osd_req, which);
198 	ceph_osd_data_pages_init(osd_data, pages, length, alignment,
199 				pages_from_pool, own_pages);
200 }
201 EXPORT_SYMBOL(osd_req_op_raw_data_in_pages);
202 
osd_req_op_extent_osd_data_pages(struct ceph_osd_request * osd_req,unsigned int which,struct page ** pages,u64 length,u32 alignment,bool pages_from_pool,bool own_pages)203 void osd_req_op_extent_osd_data_pages(struct ceph_osd_request *osd_req,
204 			unsigned int which, struct page **pages,
205 			u64 length, u32 alignment,
206 			bool pages_from_pool, bool own_pages)
207 {
208 	struct ceph_osd_data *osd_data;
209 
210 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
211 	ceph_osd_data_pages_init(osd_data, pages, length, alignment,
212 				pages_from_pool, own_pages);
213 }
214 EXPORT_SYMBOL(osd_req_op_extent_osd_data_pages);
215 
osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request * osd_req,unsigned int which,struct ceph_pagelist * pagelist)216 void osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request *osd_req,
217 			unsigned int which, struct ceph_pagelist *pagelist)
218 {
219 	struct ceph_osd_data *osd_data;
220 
221 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
222 	ceph_osd_data_pagelist_init(osd_data, pagelist);
223 }
224 EXPORT_SYMBOL(osd_req_op_extent_osd_data_pagelist);
225 
226 #ifdef CONFIG_BLOCK
osd_req_op_extent_osd_data_bio(struct ceph_osd_request * osd_req,unsigned int which,struct ceph_bio_iter * bio_pos,u32 bio_length)227 void osd_req_op_extent_osd_data_bio(struct ceph_osd_request *osd_req,
228 				    unsigned int which,
229 				    struct ceph_bio_iter *bio_pos,
230 				    u32 bio_length)
231 {
232 	struct ceph_osd_data *osd_data;
233 
234 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
235 	ceph_osd_data_bio_init(osd_data, bio_pos, bio_length);
236 }
237 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bio);
238 #endif /* CONFIG_BLOCK */
239 
osd_req_op_extent_osd_data_bvecs(struct ceph_osd_request * osd_req,unsigned int which,struct bio_vec * bvecs,u32 num_bvecs,u32 bytes)240 void osd_req_op_extent_osd_data_bvecs(struct ceph_osd_request *osd_req,
241 				      unsigned int which,
242 				      struct bio_vec *bvecs, u32 num_bvecs,
243 				      u32 bytes)
244 {
245 	struct ceph_osd_data *osd_data;
246 	struct ceph_bvec_iter it = {
247 		.bvecs = bvecs,
248 		.iter = { .bi_size = bytes },
249 	};
250 
251 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
252 	ceph_osd_data_bvecs_init(osd_data, &it, num_bvecs);
253 }
254 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvecs);
255 
osd_req_op_extent_osd_data_bvec_pos(struct ceph_osd_request * osd_req,unsigned int which,struct ceph_bvec_iter * bvec_pos)256 void osd_req_op_extent_osd_data_bvec_pos(struct ceph_osd_request *osd_req,
257 					 unsigned int which,
258 					 struct ceph_bvec_iter *bvec_pos)
259 {
260 	struct ceph_osd_data *osd_data;
261 
262 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
263 	ceph_osd_data_bvecs_init(osd_data, bvec_pos, 0);
264 }
265 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvec_pos);
266 
osd_req_op_cls_request_info_pagelist(struct ceph_osd_request * osd_req,unsigned int which,struct ceph_pagelist * pagelist)267 static void osd_req_op_cls_request_info_pagelist(
268 			struct ceph_osd_request *osd_req,
269 			unsigned int which, struct ceph_pagelist *pagelist)
270 {
271 	struct ceph_osd_data *osd_data;
272 
273 	osd_data = osd_req_op_data(osd_req, which, cls, request_info);
274 	ceph_osd_data_pagelist_init(osd_data, pagelist);
275 }
276 
osd_req_op_cls_request_data_pagelist(struct ceph_osd_request * osd_req,unsigned int which,struct ceph_pagelist * pagelist)277 void osd_req_op_cls_request_data_pagelist(
278 			struct ceph_osd_request *osd_req,
279 			unsigned int which, struct ceph_pagelist *pagelist)
280 {
281 	struct ceph_osd_data *osd_data;
282 
283 	osd_data = osd_req_op_data(osd_req, which, cls, request_data);
284 	ceph_osd_data_pagelist_init(osd_data, pagelist);
285 	osd_req->r_ops[which].cls.indata_len += pagelist->length;
286 	osd_req->r_ops[which].indata_len += pagelist->length;
287 }
288 EXPORT_SYMBOL(osd_req_op_cls_request_data_pagelist);
289 
osd_req_op_cls_request_data_pages(struct ceph_osd_request * osd_req,unsigned int which,struct page ** pages,u64 length,u32 alignment,bool pages_from_pool,bool own_pages)290 void osd_req_op_cls_request_data_pages(struct ceph_osd_request *osd_req,
291 			unsigned int which, struct page **pages, u64 length,
292 			u32 alignment, bool pages_from_pool, bool own_pages)
293 {
294 	struct ceph_osd_data *osd_data;
295 
296 	osd_data = osd_req_op_data(osd_req, which, cls, request_data);
297 	ceph_osd_data_pages_init(osd_data, pages, length, alignment,
298 				pages_from_pool, own_pages);
299 	osd_req->r_ops[which].cls.indata_len += length;
300 	osd_req->r_ops[which].indata_len += length;
301 }
302 EXPORT_SYMBOL(osd_req_op_cls_request_data_pages);
303 
osd_req_op_cls_request_data_bvecs(struct ceph_osd_request * osd_req,unsigned int which,struct bio_vec * bvecs,u32 num_bvecs,u32 bytes)304 void osd_req_op_cls_request_data_bvecs(struct ceph_osd_request *osd_req,
305 				       unsigned int which,
306 				       struct bio_vec *bvecs, u32 num_bvecs,
307 				       u32 bytes)
308 {
309 	struct ceph_osd_data *osd_data;
310 	struct ceph_bvec_iter it = {
311 		.bvecs = bvecs,
312 		.iter = { .bi_size = bytes },
313 	};
314 
315 	osd_data = osd_req_op_data(osd_req, which, cls, request_data);
316 	ceph_osd_data_bvecs_init(osd_data, &it, num_bvecs);
317 	osd_req->r_ops[which].cls.indata_len += bytes;
318 	osd_req->r_ops[which].indata_len += bytes;
319 }
320 EXPORT_SYMBOL(osd_req_op_cls_request_data_bvecs);
321 
osd_req_op_cls_response_data_pages(struct ceph_osd_request * osd_req,unsigned int which,struct page ** pages,u64 length,u32 alignment,bool pages_from_pool,bool own_pages)322 void osd_req_op_cls_response_data_pages(struct ceph_osd_request *osd_req,
323 			unsigned int which, struct page **pages, u64 length,
324 			u32 alignment, bool pages_from_pool, bool own_pages)
325 {
326 	struct ceph_osd_data *osd_data;
327 
328 	osd_data = osd_req_op_data(osd_req, which, cls, response_data);
329 	ceph_osd_data_pages_init(osd_data, pages, length, alignment,
330 				pages_from_pool, own_pages);
331 }
332 EXPORT_SYMBOL(osd_req_op_cls_response_data_pages);
333 
ceph_osd_data_length(struct ceph_osd_data * osd_data)334 static u64 ceph_osd_data_length(struct ceph_osd_data *osd_data)
335 {
336 	switch (osd_data->type) {
337 	case CEPH_OSD_DATA_TYPE_NONE:
338 		return 0;
339 	case CEPH_OSD_DATA_TYPE_PAGES:
340 		return osd_data->length;
341 	case CEPH_OSD_DATA_TYPE_PAGELIST:
342 		return (u64)osd_data->pagelist->length;
343 #ifdef CONFIG_BLOCK
344 	case CEPH_OSD_DATA_TYPE_BIO:
345 		return (u64)osd_data->bio_length;
346 #endif /* CONFIG_BLOCK */
347 	case CEPH_OSD_DATA_TYPE_BVECS:
348 		return osd_data->bvec_pos.iter.bi_size;
349 	default:
350 		WARN(true, "unrecognized data type %d\n", (int)osd_data->type);
351 		return 0;
352 	}
353 }
354 
ceph_osd_data_release(struct ceph_osd_data * osd_data)355 static void ceph_osd_data_release(struct ceph_osd_data *osd_data)
356 {
357 	if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES && osd_data->own_pages) {
358 		int num_pages;
359 
360 		num_pages = calc_pages_for((u64)osd_data->alignment,
361 						(u64)osd_data->length);
362 		ceph_release_page_vector(osd_data->pages, num_pages);
363 	} else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
364 		ceph_pagelist_release(osd_data->pagelist);
365 	}
366 	ceph_osd_data_init(osd_data);
367 }
368 
osd_req_op_data_release(struct ceph_osd_request * osd_req,unsigned int which)369 static void osd_req_op_data_release(struct ceph_osd_request *osd_req,
370 			unsigned int which)
371 {
372 	struct ceph_osd_req_op *op;
373 
374 	BUG_ON(which >= osd_req->r_num_ops);
375 	op = &osd_req->r_ops[which];
376 
377 	switch (op->op) {
378 	case CEPH_OSD_OP_READ:
379 	case CEPH_OSD_OP_WRITE:
380 	case CEPH_OSD_OP_WRITEFULL:
381 		ceph_osd_data_release(&op->extent.osd_data);
382 		break;
383 	case CEPH_OSD_OP_CALL:
384 		ceph_osd_data_release(&op->cls.request_info);
385 		ceph_osd_data_release(&op->cls.request_data);
386 		ceph_osd_data_release(&op->cls.response_data);
387 		break;
388 	case CEPH_OSD_OP_SETXATTR:
389 	case CEPH_OSD_OP_CMPXATTR:
390 		ceph_osd_data_release(&op->xattr.osd_data);
391 		break;
392 	case CEPH_OSD_OP_STAT:
393 		ceph_osd_data_release(&op->raw_data_in);
394 		break;
395 	case CEPH_OSD_OP_NOTIFY_ACK:
396 		ceph_osd_data_release(&op->notify_ack.request_data);
397 		break;
398 	case CEPH_OSD_OP_NOTIFY:
399 		ceph_osd_data_release(&op->notify.request_data);
400 		ceph_osd_data_release(&op->notify.response_data);
401 		break;
402 	case CEPH_OSD_OP_LIST_WATCHERS:
403 		ceph_osd_data_release(&op->list_watchers.response_data);
404 		break;
405 	case CEPH_OSD_OP_COPY_FROM2:
406 		ceph_osd_data_release(&op->copy_from.osd_data);
407 		break;
408 	default:
409 		break;
410 	}
411 }
412 
413 /*
414  * Assumes @t is zero-initialized.
415  */
target_init(struct ceph_osd_request_target * t)416 static void target_init(struct ceph_osd_request_target *t)
417 {
418 	ceph_oid_init(&t->base_oid);
419 	ceph_oloc_init(&t->base_oloc);
420 	ceph_oid_init(&t->target_oid);
421 	ceph_oloc_init(&t->target_oloc);
422 
423 	ceph_osds_init(&t->acting);
424 	ceph_osds_init(&t->up);
425 	t->size = -1;
426 	t->min_size = -1;
427 
428 	t->osd = CEPH_HOMELESS_OSD;
429 }
430 
target_copy(struct ceph_osd_request_target * dest,const struct ceph_osd_request_target * src)431 static void target_copy(struct ceph_osd_request_target *dest,
432 			const struct ceph_osd_request_target *src)
433 {
434 	ceph_oid_copy(&dest->base_oid, &src->base_oid);
435 	ceph_oloc_copy(&dest->base_oloc, &src->base_oloc);
436 	ceph_oid_copy(&dest->target_oid, &src->target_oid);
437 	ceph_oloc_copy(&dest->target_oloc, &src->target_oloc);
438 
439 	dest->pgid = src->pgid; /* struct */
440 	dest->spgid = src->spgid; /* struct */
441 	dest->pg_num = src->pg_num;
442 	dest->pg_num_mask = src->pg_num_mask;
443 	ceph_osds_copy(&dest->acting, &src->acting);
444 	ceph_osds_copy(&dest->up, &src->up);
445 	dest->size = src->size;
446 	dest->min_size = src->min_size;
447 	dest->sort_bitwise = src->sort_bitwise;
448 	dest->recovery_deletes = src->recovery_deletes;
449 
450 	dest->flags = src->flags;
451 	dest->used_replica = src->used_replica;
452 	dest->paused = src->paused;
453 
454 	dest->epoch = src->epoch;
455 	dest->last_force_resend = src->last_force_resend;
456 
457 	dest->osd = src->osd;
458 }
459 
target_destroy(struct ceph_osd_request_target * t)460 static void target_destroy(struct ceph_osd_request_target *t)
461 {
462 	ceph_oid_destroy(&t->base_oid);
463 	ceph_oloc_destroy(&t->base_oloc);
464 	ceph_oid_destroy(&t->target_oid);
465 	ceph_oloc_destroy(&t->target_oloc);
466 }
467 
468 /*
469  * requests
470  */
request_release_checks(struct ceph_osd_request * req)471 static void request_release_checks(struct ceph_osd_request *req)
472 {
473 	WARN_ON(!RB_EMPTY_NODE(&req->r_node));
474 	WARN_ON(!RB_EMPTY_NODE(&req->r_mc_node));
475 	WARN_ON(!list_empty(&req->r_private_item));
476 	WARN_ON(req->r_osd);
477 }
478 
ceph_osdc_release_request(struct kref * kref)479 static void ceph_osdc_release_request(struct kref *kref)
480 {
481 	struct ceph_osd_request *req = container_of(kref,
482 					    struct ceph_osd_request, r_kref);
483 	unsigned int which;
484 
485 	dout("%s %p (r_request %p r_reply %p)\n", __func__, req,
486 	     req->r_request, req->r_reply);
487 	request_release_checks(req);
488 
489 	if (req->r_request)
490 		ceph_msg_put(req->r_request);
491 	if (req->r_reply)
492 		ceph_msg_put(req->r_reply);
493 
494 	for (which = 0; which < req->r_num_ops; which++)
495 		osd_req_op_data_release(req, which);
496 
497 	target_destroy(&req->r_t);
498 	ceph_put_snap_context(req->r_snapc);
499 
500 	if (req->r_mempool)
501 		mempool_free(req, req->r_osdc->req_mempool);
502 	else if (req->r_num_ops <= CEPH_OSD_SLAB_OPS)
503 		kmem_cache_free(ceph_osd_request_cache, req);
504 	else
505 		kfree(req);
506 }
507 
ceph_osdc_get_request(struct ceph_osd_request * req)508 void ceph_osdc_get_request(struct ceph_osd_request *req)
509 {
510 	dout("%s %p (was %d)\n", __func__, req,
511 	     kref_read(&req->r_kref));
512 	kref_get(&req->r_kref);
513 }
514 EXPORT_SYMBOL(ceph_osdc_get_request);
515 
ceph_osdc_put_request(struct ceph_osd_request * req)516 void ceph_osdc_put_request(struct ceph_osd_request *req)
517 {
518 	if (req) {
519 		dout("%s %p (was %d)\n", __func__, req,
520 		     kref_read(&req->r_kref));
521 		kref_put(&req->r_kref, ceph_osdc_release_request);
522 	}
523 }
524 EXPORT_SYMBOL(ceph_osdc_put_request);
525 
request_init(struct ceph_osd_request * req)526 static void request_init(struct ceph_osd_request *req)
527 {
528 	/* req only, each op is zeroed in osd_req_op_init() */
529 	memset(req, 0, sizeof(*req));
530 
531 	kref_init(&req->r_kref);
532 	init_completion(&req->r_completion);
533 	RB_CLEAR_NODE(&req->r_node);
534 	RB_CLEAR_NODE(&req->r_mc_node);
535 	INIT_LIST_HEAD(&req->r_private_item);
536 
537 	target_init(&req->r_t);
538 }
539 
540 /*
541  * This is ugly, but it allows us to reuse linger registration and ping
542  * requests, keeping the structure of the code around send_linger{_ping}()
543  * reasonable.  Setting up a min_nr=2 mempool for each linger request
544  * and dealing with copying ops (this blasts req only, watch op remains
545  * intact) isn't any better.
546  */
request_reinit(struct ceph_osd_request * req)547 static void request_reinit(struct ceph_osd_request *req)
548 {
549 	struct ceph_osd_client *osdc = req->r_osdc;
550 	bool mempool = req->r_mempool;
551 	unsigned int num_ops = req->r_num_ops;
552 	u64 snapid = req->r_snapid;
553 	struct ceph_snap_context *snapc = req->r_snapc;
554 	bool linger = req->r_linger;
555 	struct ceph_msg *request_msg = req->r_request;
556 	struct ceph_msg *reply_msg = req->r_reply;
557 
558 	dout("%s req %p\n", __func__, req);
559 	WARN_ON(kref_read(&req->r_kref) != 1);
560 	request_release_checks(req);
561 
562 	WARN_ON(kref_read(&request_msg->kref) != 1);
563 	WARN_ON(kref_read(&reply_msg->kref) != 1);
564 	target_destroy(&req->r_t);
565 
566 	request_init(req);
567 	req->r_osdc = osdc;
568 	req->r_mempool = mempool;
569 	req->r_num_ops = num_ops;
570 	req->r_snapid = snapid;
571 	req->r_snapc = snapc;
572 	req->r_linger = linger;
573 	req->r_request = request_msg;
574 	req->r_reply = reply_msg;
575 }
576 
ceph_osdc_alloc_request(struct ceph_osd_client * osdc,struct ceph_snap_context * snapc,unsigned int num_ops,bool use_mempool,gfp_t gfp_flags)577 struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
578 					       struct ceph_snap_context *snapc,
579 					       unsigned int num_ops,
580 					       bool use_mempool,
581 					       gfp_t gfp_flags)
582 {
583 	struct ceph_osd_request *req;
584 
585 	if (use_mempool) {
586 		BUG_ON(num_ops > CEPH_OSD_SLAB_OPS);
587 		req = mempool_alloc(osdc->req_mempool, gfp_flags);
588 	} else if (num_ops <= CEPH_OSD_SLAB_OPS) {
589 		req = kmem_cache_alloc(ceph_osd_request_cache, gfp_flags);
590 	} else {
591 		BUG_ON(num_ops > CEPH_OSD_MAX_OPS);
592 		req = kmalloc(struct_size(req, r_ops, num_ops), gfp_flags);
593 	}
594 	if (unlikely(!req))
595 		return NULL;
596 
597 	request_init(req);
598 	req->r_osdc = osdc;
599 	req->r_mempool = use_mempool;
600 	req->r_num_ops = num_ops;
601 	req->r_snapid = CEPH_NOSNAP;
602 	req->r_snapc = ceph_get_snap_context(snapc);
603 
604 	dout("%s req %p\n", __func__, req);
605 	return req;
606 }
607 EXPORT_SYMBOL(ceph_osdc_alloc_request);
608 
ceph_oloc_encoding_size(const struct ceph_object_locator * oloc)609 static int ceph_oloc_encoding_size(const struct ceph_object_locator *oloc)
610 {
611 	return 8 + 4 + 4 + 4 + (oloc->pool_ns ? oloc->pool_ns->len : 0);
612 }
613 
__ceph_osdc_alloc_messages(struct ceph_osd_request * req,gfp_t gfp,int num_request_data_items,int num_reply_data_items)614 static int __ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp,
615 				      int num_request_data_items,
616 				      int num_reply_data_items)
617 {
618 	struct ceph_osd_client *osdc = req->r_osdc;
619 	struct ceph_msg *msg;
620 	int msg_size;
621 
622 	WARN_ON(req->r_request || req->r_reply);
623 	WARN_ON(ceph_oid_empty(&req->r_base_oid));
624 	WARN_ON(ceph_oloc_empty(&req->r_base_oloc));
625 
626 	/* create request message */
627 	msg_size = CEPH_ENCODING_START_BLK_LEN +
628 			CEPH_PGID_ENCODING_LEN + 1; /* spgid */
629 	msg_size += 4 + 4 + 4; /* hash, osdmap_epoch, flags */
630 	msg_size += CEPH_ENCODING_START_BLK_LEN +
631 			sizeof(struct ceph_osd_reqid); /* reqid */
632 	msg_size += sizeof(struct ceph_blkin_trace_info); /* trace */
633 	msg_size += 4 + sizeof(struct ceph_timespec); /* client_inc, mtime */
634 	msg_size += CEPH_ENCODING_START_BLK_LEN +
635 			ceph_oloc_encoding_size(&req->r_base_oloc); /* oloc */
636 	msg_size += 4 + req->r_base_oid.name_len; /* oid */
637 	msg_size += 2 + req->r_num_ops * sizeof(struct ceph_osd_op);
638 	msg_size += 8; /* snapid */
639 	msg_size += 8; /* snap_seq */
640 	msg_size += 4 + 8 * (req->r_snapc ? req->r_snapc->num_snaps : 0);
641 	msg_size += 4 + 8; /* retry_attempt, features */
642 
643 	if (req->r_mempool)
644 		msg = ceph_msgpool_get(&osdc->msgpool_op, msg_size,
645 				       num_request_data_items);
646 	else
647 		msg = ceph_msg_new2(CEPH_MSG_OSD_OP, msg_size,
648 				    num_request_data_items, gfp, true);
649 	if (!msg)
650 		return -ENOMEM;
651 
652 	memset(msg->front.iov_base, 0, msg->front.iov_len);
653 	req->r_request = msg;
654 
655 	/* create reply message */
656 	msg_size = OSD_OPREPLY_FRONT_LEN;
657 	msg_size += req->r_base_oid.name_len;
658 	msg_size += req->r_num_ops * sizeof(struct ceph_osd_op);
659 
660 	if (req->r_mempool)
661 		msg = ceph_msgpool_get(&osdc->msgpool_op_reply, msg_size,
662 				       num_reply_data_items);
663 	else
664 		msg = ceph_msg_new2(CEPH_MSG_OSD_OPREPLY, msg_size,
665 				    num_reply_data_items, gfp, true);
666 	if (!msg)
667 		return -ENOMEM;
668 
669 	req->r_reply = msg;
670 
671 	return 0;
672 }
673 
osd_req_opcode_valid(u16 opcode)674 static bool osd_req_opcode_valid(u16 opcode)
675 {
676 	switch (opcode) {
677 #define GENERATE_CASE(op, opcode, str)	case CEPH_OSD_OP_##op: return true;
678 __CEPH_FORALL_OSD_OPS(GENERATE_CASE)
679 #undef GENERATE_CASE
680 	default:
681 		return false;
682 	}
683 }
684 
get_num_data_items(struct ceph_osd_request * req,int * num_request_data_items,int * num_reply_data_items)685 static void get_num_data_items(struct ceph_osd_request *req,
686 			       int *num_request_data_items,
687 			       int *num_reply_data_items)
688 {
689 	struct ceph_osd_req_op *op;
690 
691 	*num_request_data_items = 0;
692 	*num_reply_data_items = 0;
693 
694 	for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) {
695 		switch (op->op) {
696 		/* request */
697 		case CEPH_OSD_OP_WRITE:
698 		case CEPH_OSD_OP_WRITEFULL:
699 		case CEPH_OSD_OP_SETXATTR:
700 		case CEPH_OSD_OP_CMPXATTR:
701 		case CEPH_OSD_OP_NOTIFY_ACK:
702 		case CEPH_OSD_OP_COPY_FROM2:
703 			*num_request_data_items += 1;
704 			break;
705 
706 		/* reply */
707 		case CEPH_OSD_OP_STAT:
708 		case CEPH_OSD_OP_READ:
709 		case CEPH_OSD_OP_LIST_WATCHERS:
710 			*num_reply_data_items += 1;
711 			break;
712 
713 		/* both */
714 		case CEPH_OSD_OP_NOTIFY:
715 			*num_request_data_items += 1;
716 			*num_reply_data_items += 1;
717 			break;
718 		case CEPH_OSD_OP_CALL:
719 			*num_request_data_items += 2;
720 			*num_reply_data_items += 1;
721 			break;
722 
723 		default:
724 			WARN_ON(!osd_req_opcode_valid(op->op));
725 			break;
726 		}
727 	}
728 }
729 
730 /*
731  * oid, oloc and OSD op opcode(s) must be filled in before this function
732  * is called.
733  */
ceph_osdc_alloc_messages(struct ceph_osd_request * req,gfp_t gfp)734 int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp)
735 {
736 	int num_request_data_items, num_reply_data_items;
737 
738 	get_num_data_items(req, &num_request_data_items, &num_reply_data_items);
739 	return __ceph_osdc_alloc_messages(req, gfp, num_request_data_items,
740 					  num_reply_data_items);
741 }
742 EXPORT_SYMBOL(ceph_osdc_alloc_messages);
743 
744 /*
745  * This is an osd op init function for opcodes that have no data or
746  * other information associated with them.  It also serves as a
747  * common init routine for all the other init functions, below.
748  */
749 struct ceph_osd_req_op *
osd_req_op_init(struct ceph_osd_request * osd_req,unsigned int which,u16 opcode,u32 flags)750 osd_req_op_init(struct ceph_osd_request *osd_req, unsigned int which,
751 		 u16 opcode, u32 flags)
752 {
753 	struct ceph_osd_req_op *op;
754 
755 	BUG_ON(which >= osd_req->r_num_ops);
756 	BUG_ON(!osd_req_opcode_valid(opcode));
757 
758 	op = &osd_req->r_ops[which];
759 	memset(op, 0, sizeof (*op));
760 	op->op = opcode;
761 	op->flags = flags;
762 
763 	return op;
764 }
765 EXPORT_SYMBOL(osd_req_op_init);
766 
osd_req_op_extent_init(struct ceph_osd_request * osd_req,unsigned int which,u16 opcode,u64 offset,u64 length,u64 truncate_size,u32 truncate_seq)767 void osd_req_op_extent_init(struct ceph_osd_request *osd_req,
768 				unsigned int which, u16 opcode,
769 				u64 offset, u64 length,
770 				u64 truncate_size, u32 truncate_seq)
771 {
772 	struct ceph_osd_req_op *op = osd_req_op_init(osd_req, which,
773 						     opcode, 0);
774 	size_t payload_len = 0;
775 
776 	BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
777 	       opcode != CEPH_OSD_OP_WRITEFULL && opcode != CEPH_OSD_OP_ZERO &&
778 	       opcode != CEPH_OSD_OP_TRUNCATE);
779 
780 	op->extent.offset = offset;
781 	op->extent.length = length;
782 	op->extent.truncate_size = truncate_size;
783 	op->extent.truncate_seq = truncate_seq;
784 	if (opcode == CEPH_OSD_OP_WRITE || opcode == CEPH_OSD_OP_WRITEFULL)
785 		payload_len += length;
786 
787 	op->indata_len = payload_len;
788 }
789 EXPORT_SYMBOL(osd_req_op_extent_init);
790 
osd_req_op_extent_update(struct ceph_osd_request * osd_req,unsigned int which,u64 length)791 void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
792 				unsigned int which, u64 length)
793 {
794 	struct ceph_osd_req_op *op;
795 	u64 previous;
796 
797 	BUG_ON(which >= osd_req->r_num_ops);
798 	op = &osd_req->r_ops[which];
799 	previous = op->extent.length;
800 
801 	if (length == previous)
802 		return;		/* Nothing to do */
803 	BUG_ON(length > previous);
804 
805 	op->extent.length = length;
806 	if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
807 		op->indata_len -= previous - length;
808 }
809 EXPORT_SYMBOL(osd_req_op_extent_update);
810 
osd_req_op_extent_dup_last(struct ceph_osd_request * osd_req,unsigned int which,u64 offset_inc)811 void osd_req_op_extent_dup_last(struct ceph_osd_request *osd_req,
812 				unsigned int which, u64 offset_inc)
813 {
814 	struct ceph_osd_req_op *op, *prev_op;
815 
816 	BUG_ON(which + 1 >= osd_req->r_num_ops);
817 
818 	prev_op = &osd_req->r_ops[which];
819 	op = osd_req_op_init(osd_req, which + 1, prev_op->op, prev_op->flags);
820 	/* dup previous one */
821 	op->indata_len = prev_op->indata_len;
822 	op->outdata_len = prev_op->outdata_len;
823 	op->extent = prev_op->extent;
824 	/* adjust offset */
825 	op->extent.offset += offset_inc;
826 	op->extent.length -= offset_inc;
827 
828 	if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
829 		op->indata_len -= offset_inc;
830 }
831 EXPORT_SYMBOL(osd_req_op_extent_dup_last);
832 
osd_req_op_cls_init(struct ceph_osd_request * osd_req,unsigned int which,const char * class,const char * method)833 int osd_req_op_cls_init(struct ceph_osd_request *osd_req, unsigned int which,
834 			const char *class, const char *method)
835 {
836 	struct ceph_osd_req_op *op;
837 	struct ceph_pagelist *pagelist;
838 	size_t payload_len = 0;
839 	size_t size;
840 	int ret;
841 
842 	op = osd_req_op_init(osd_req, which, CEPH_OSD_OP_CALL, 0);
843 
844 	pagelist = ceph_pagelist_alloc(GFP_NOFS);
845 	if (!pagelist)
846 		return -ENOMEM;
847 
848 	op->cls.class_name = class;
849 	size = strlen(class);
850 	BUG_ON(size > (size_t) U8_MAX);
851 	op->cls.class_len = size;
852 	ret = ceph_pagelist_append(pagelist, class, size);
853 	if (ret)
854 		goto err_pagelist_free;
855 	payload_len += size;
856 
857 	op->cls.method_name = method;
858 	size = strlen(method);
859 	BUG_ON(size > (size_t) U8_MAX);
860 	op->cls.method_len = size;
861 	ret = ceph_pagelist_append(pagelist, method, size);
862 	if (ret)
863 		goto err_pagelist_free;
864 	payload_len += size;
865 
866 	osd_req_op_cls_request_info_pagelist(osd_req, which, pagelist);
867 	op->indata_len = payload_len;
868 	return 0;
869 
870 err_pagelist_free:
871 	ceph_pagelist_release(pagelist);
872 	return ret;
873 }
874 EXPORT_SYMBOL(osd_req_op_cls_init);
875 
osd_req_op_xattr_init(struct ceph_osd_request * osd_req,unsigned int which,u16 opcode,const char * name,const void * value,size_t size,u8 cmp_op,u8 cmp_mode)876 int osd_req_op_xattr_init(struct ceph_osd_request *osd_req, unsigned int which,
877 			  u16 opcode, const char *name, const void *value,
878 			  size_t size, u8 cmp_op, u8 cmp_mode)
879 {
880 	struct ceph_osd_req_op *op = osd_req_op_init(osd_req, which,
881 						     opcode, 0);
882 	struct ceph_pagelist *pagelist;
883 	size_t payload_len;
884 	int ret;
885 
886 	BUG_ON(opcode != CEPH_OSD_OP_SETXATTR && opcode != CEPH_OSD_OP_CMPXATTR);
887 
888 	pagelist = ceph_pagelist_alloc(GFP_NOFS);
889 	if (!pagelist)
890 		return -ENOMEM;
891 
892 	payload_len = strlen(name);
893 	op->xattr.name_len = payload_len;
894 	ret = ceph_pagelist_append(pagelist, name, payload_len);
895 	if (ret)
896 		goto err_pagelist_free;
897 
898 	op->xattr.value_len = size;
899 	ret = ceph_pagelist_append(pagelist, value, size);
900 	if (ret)
901 		goto err_pagelist_free;
902 	payload_len += size;
903 
904 	op->xattr.cmp_op = cmp_op;
905 	op->xattr.cmp_mode = cmp_mode;
906 
907 	ceph_osd_data_pagelist_init(&op->xattr.osd_data, pagelist);
908 	op->indata_len = payload_len;
909 	return 0;
910 
911 err_pagelist_free:
912 	ceph_pagelist_release(pagelist);
913 	return ret;
914 }
915 EXPORT_SYMBOL(osd_req_op_xattr_init);
916 
917 /*
918  * @watch_opcode: CEPH_OSD_WATCH_OP_*
919  */
osd_req_op_watch_init(struct ceph_osd_request * req,int which,u64 cookie,u8 watch_opcode)920 static void osd_req_op_watch_init(struct ceph_osd_request *req, int which,
921 				  u64 cookie, u8 watch_opcode)
922 {
923 	struct ceph_osd_req_op *op;
924 
925 	op = osd_req_op_init(req, which, CEPH_OSD_OP_WATCH, 0);
926 	op->watch.cookie = cookie;
927 	op->watch.op = watch_opcode;
928 	op->watch.gen = 0;
929 }
930 
931 /*
932  * @flags: CEPH_OSD_OP_ALLOC_HINT_FLAG_*
933  */
osd_req_op_alloc_hint_init(struct ceph_osd_request * osd_req,unsigned int which,u64 expected_object_size,u64 expected_write_size,u32 flags)934 void osd_req_op_alloc_hint_init(struct ceph_osd_request *osd_req,
935 				unsigned int which,
936 				u64 expected_object_size,
937 				u64 expected_write_size,
938 				u32 flags)
939 {
940 	struct ceph_osd_req_op *op;
941 
942 	op = osd_req_op_init(osd_req, which, CEPH_OSD_OP_SETALLOCHINT, 0);
943 	op->alloc_hint.expected_object_size = expected_object_size;
944 	op->alloc_hint.expected_write_size = expected_write_size;
945 	op->alloc_hint.flags = flags;
946 
947 	/*
948 	 * CEPH_OSD_OP_SETALLOCHINT op is advisory and therefore deemed
949 	 * not worth a feature bit.  Set FAILOK per-op flag to make
950 	 * sure older osds don't trip over an unsupported opcode.
951 	 */
952 	op->flags |= CEPH_OSD_OP_FLAG_FAILOK;
953 }
954 EXPORT_SYMBOL(osd_req_op_alloc_hint_init);
955 
ceph_osdc_msg_data_add(struct ceph_msg * msg,struct ceph_osd_data * osd_data)956 static void ceph_osdc_msg_data_add(struct ceph_msg *msg,
957 				struct ceph_osd_data *osd_data)
958 {
959 	u64 length = ceph_osd_data_length(osd_data);
960 
961 	if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES) {
962 		BUG_ON(length > (u64) SIZE_MAX);
963 		if (length)
964 			ceph_msg_data_add_pages(msg, osd_data->pages,
965 					length, osd_data->alignment, false);
966 	} else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
967 		BUG_ON(!length);
968 		ceph_msg_data_add_pagelist(msg, osd_data->pagelist);
969 #ifdef CONFIG_BLOCK
970 	} else if (osd_data->type == CEPH_OSD_DATA_TYPE_BIO) {
971 		ceph_msg_data_add_bio(msg, &osd_data->bio_pos, length);
972 #endif
973 	} else if (osd_data->type == CEPH_OSD_DATA_TYPE_BVECS) {
974 		ceph_msg_data_add_bvecs(msg, &osd_data->bvec_pos);
975 	} else {
976 		BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_NONE);
977 	}
978 }
979 
osd_req_encode_op(struct ceph_osd_op * dst,const struct ceph_osd_req_op * src)980 static u32 osd_req_encode_op(struct ceph_osd_op *dst,
981 			     const struct ceph_osd_req_op *src)
982 {
983 	switch (src->op) {
984 	case CEPH_OSD_OP_STAT:
985 		break;
986 	case CEPH_OSD_OP_READ:
987 	case CEPH_OSD_OP_WRITE:
988 	case CEPH_OSD_OP_WRITEFULL:
989 	case CEPH_OSD_OP_ZERO:
990 	case CEPH_OSD_OP_TRUNCATE:
991 		dst->extent.offset = cpu_to_le64(src->extent.offset);
992 		dst->extent.length = cpu_to_le64(src->extent.length);
993 		dst->extent.truncate_size =
994 			cpu_to_le64(src->extent.truncate_size);
995 		dst->extent.truncate_seq =
996 			cpu_to_le32(src->extent.truncate_seq);
997 		break;
998 	case CEPH_OSD_OP_CALL:
999 		dst->cls.class_len = src->cls.class_len;
1000 		dst->cls.method_len = src->cls.method_len;
1001 		dst->cls.indata_len = cpu_to_le32(src->cls.indata_len);
1002 		break;
1003 	case CEPH_OSD_OP_WATCH:
1004 		dst->watch.cookie = cpu_to_le64(src->watch.cookie);
1005 		dst->watch.ver = cpu_to_le64(0);
1006 		dst->watch.op = src->watch.op;
1007 		dst->watch.gen = cpu_to_le32(src->watch.gen);
1008 		break;
1009 	case CEPH_OSD_OP_NOTIFY_ACK:
1010 		break;
1011 	case CEPH_OSD_OP_NOTIFY:
1012 		dst->notify.cookie = cpu_to_le64(src->notify.cookie);
1013 		break;
1014 	case CEPH_OSD_OP_LIST_WATCHERS:
1015 		break;
1016 	case CEPH_OSD_OP_SETALLOCHINT:
1017 		dst->alloc_hint.expected_object_size =
1018 		    cpu_to_le64(src->alloc_hint.expected_object_size);
1019 		dst->alloc_hint.expected_write_size =
1020 		    cpu_to_le64(src->alloc_hint.expected_write_size);
1021 		dst->alloc_hint.flags = cpu_to_le32(src->alloc_hint.flags);
1022 		break;
1023 	case CEPH_OSD_OP_SETXATTR:
1024 	case CEPH_OSD_OP_CMPXATTR:
1025 		dst->xattr.name_len = cpu_to_le32(src->xattr.name_len);
1026 		dst->xattr.value_len = cpu_to_le32(src->xattr.value_len);
1027 		dst->xattr.cmp_op = src->xattr.cmp_op;
1028 		dst->xattr.cmp_mode = src->xattr.cmp_mode;
1029 		break;
1030 	case CEPH_OSD_OP_CREATE:
1031 	case CEPH_OSD_OP_DELETE:
1032 		break;
1033 	case CEPH_OSD_OP_COPY_FROM2:
1034 		dst->copy_from.snapid = cpu_to_le64(src->copy_from.snapid);
1035 		dst->copy_from.src_version =
1036 			cpu_to_le64(src->copy_from.src_version);
1037 		dst->copy_from.flags = src->copy_from.flags;
1038 		dst->copy_from.src_fadvise_flags =
1039 			cpu_to_le32(src->copy_from.src_fadvise_flags);
1040 		break;
1041 	default:
1042 		pr_err("unsupported osd opcode %s\n",
1043 			ceph_osd_op_name(src->op));
1044 		WARN_ON(1);
1045 
1046 		return 0;
1047 	}
1048 
1049 	dst->op = cpu_to_le16(src->op);
1050 	dst->flags = cpu_to_le32(src->flags);
1051 	dst->payload_len = cpu_to_le32(src->indata_len);
1052 
1053 	return src->indata_len;
1054 }
1055 
1056 /*
1057  * build new request AND message, calculate layout, and adjust file
1058  * extent as needed.
1059  *
1060  * if the file was recently truncated, we include information about its
1061  * old and new size so that the object can be updated appropriately.  (we
1062  * avoid synchronously deleting truncated objects because it's slow.)
1063  */
ceph_osdc_new_request(struct ceph_osd_client * osdc,struct ceph_file_layout * layout,struct ceph_vino vino,u64 off,u64 * plen,unsigned int which,int num_ops,int opcode,int flags,struct ceph_snap_context * snapc,u32 truncate_seq,u64 truncate_size,bool use_mempool)1064 struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *osdc,
1065 					       struct ceph_file_layout *layout,
1066 					       struct ceph_vino vino,
1067 					       u64 off, u64 *plen,
1068 					       unsigned int which, int num_ops,
1069 					       int opcode, int flags,
1070 					       struct ceph_snap_context *snapc,
1071 					       u32 truncate_seq,
1072 					       u64 truncate_size,
1073 					       bool use_mempool)
1074 {
1075 	struct ceph_osd_request *req;
1076 	u64 objnum = 0;
1077 	u64 objoff = 0;
1078 	u64 objlen = 0;
1079 	int r;
1080 
1081 	BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
1082 	       opcode != CEPH_OSD_OP_ZERO && opcode != CEPH_OSD_OP_TRUNCATE &&
1083 	       opcode != CEPH_OSD_OP_CREATE && opcode != CEPH_OSD_OP_DELETE);
1084 
1085 	req = ceph_osdc_alloc_request(osdc, snapc, num_ops, use_mempool,
1086 					GFP_NOFS);
1087 	if (!req) {
1088 		r = -ENOMEM;
1089 		goto fail;
1090 	}
1091 
1092 	/* calculate max write size */
1093 	r = calc_layout(layout, off, plen, &objnum, &objoff, &objlen);
1094 	if (r)
1095 		goto fail;
1096 
1097 	if (opcode == CEPH_OSD_OP_CREATE || opcode == CEPH_OSD_OP_DELETE) {
1098 		osd_req_op_init(req, which, opcode, 0);
1099 	} else {
1100 		u32 object_size = layout->object_size;
1101 		u32 object_base = off - objoff;
1102 		if (!(truncate_seq == 1 && truncate_size == -1ULL)) {
1103 			if (truncate_size <= object_base) {
1104 				truncate_size = 0;
1105 			} else {
1106 				truncate_size -= object_base;
1107 				if (truncate_size > object_size)
1108 					truncate_size = object_size;
1109 			}
1110 		}
1111 		osd_req_op_extent_init(req, which, opcode, objoff, objlen,
1112 				       truncate_size, truncate_seq);
1113 	}
1114 
1115 	req->r_base_oloc.pool = layout->pool_id;
1116 	req->r_base_oloc.pool_ns = ceph_try_get_string(layout->pool_ns);
1117 	ceph_oid_printf(&req->r_base_oid, "%llx.%08llx", vino.ino, objnum);
1118 	req->r_flags = flags | osdc->client->options->read_from_replica;
1119 
1120 	req->r_snapid = vino.snap;
1121 	if (flags & CEPH_OSD_FLAG_WRITE)
1122 		req->r_data_offset = off;
1123 
1124 	if (num_ops > 1)
1125 		/*
1126 		 * This is a special case for ceph_writepages_start(), but it
1127 		 * also covers ceph_uninline_data().  If more multi-op request
1128 		 * use cases emerge, we will need a separate helper.
1129 		 */
1130 		r = __ceph_osdc_alloc_messages(req, GFP_NOFS, num_ops, 0);
1131 	else
1132 		r = ceph_osdc_alloc_messages(req, GFP_NOFS);
1133 	if (r)
1134 		goto fail;
1135 
1136 	return req;
1137 
1138 fail:
1139 	ceph_osdc_put_request(req);
1140 	return ERR_PTR(r);
1141 }
1142 EXPORT_SYMBOL(ceph_osdc_new_request);
1143 
1144 /*
1145  * We keep osd requests in an rbtree, sorted by ->r_tid.
1146  */
DEFINE_RB_FUNCS(request,struct ceph_osd_request,r_tid,r_node)1147 DEFINE_RB_FUNCS(request, struct ceph_osd_request, r_tid, r_node)
1148 DEFINE_RB_FUNCS(request_mc, struct ceph_osd_request, r_tid, r_mc_node)
1149 
1150 /*
1151  * Call @fn on each OSD request as long as @fn returns 0.
1152  */
1153 static void for_each_request(struct ceph_osd_client *osdc,
1154 			int (*fn)(struct ceph_osd_request *req, void *arg),
1155 			void *arg)
1156 {
1157 	struct rb_node *n, *p;
1158 
1159 	for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
1160 		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
1161 
1162 		for (p = rb_first(&osd->o_requests); p; ) {
1163 			struct ceph_osd_request *req =
1164 			    rb_entry(p, struct ceph_osd_request, r_node);
1165 
1166 			p = rb_next(p);
1167 			if (fn(req, arg))
1168 				return;
1169 		}
1170 	}
1171 
1172 	for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
1173 		struct ceph_osd_request *req =
1174 		    rb_entry(p, struct ceph_osd_request, r_node);
1175 
1176 		p = rb_next(p);
1177 		if (fn(req, arg))
1178 			return;
1179 	}
1180 }
1181 
osd_homeless(struct ceph_osd * osd)1182 static bool osd_homeless(struct ceph_osd *osd)
1183 {
1184 	return osd->o_osd == CEPH_HOMELESS_OSD;
1185 }
1186 
osd_registered(struct ceph_osd * osd)1187 static bool osd_registered(struct ceph_osd *osd)
1188 {
1189 	verify_osdc_locked(osd->o_osdc);
1190 
1191 	return !RB_EMPTY_NODE(&osd->o_node);
1192 }
1193 
1194 /*
1195  * Assumes @osd is zero-initialized.
1196  */
osd_init(struct ceph_osd * osd)1197 static void osd_init(struct ceph_osd *osd)
1198 {
1199 	refcount_set(&osd->o_ref, 1);
1200 	RB_CLEAR_NODE(&osd->o_node);
1201 	osd->o_requests = RB_ROOT;
1202 	osd->o_linger_requests = RB_ROOT;
1203 	osd->o_backoff_mappings = RB_ROOT;
1204 	osd->o_backoffs_by_id = RB_ROOT;
1205 	INIT_LIST_HEAD(&osd->o_osd_lru);
1206 	INIT_LIST_HEAD(&osd->o_keepalive_item);
1207 	osd->o_incarnation = 1;
1208 	mutex_init(&osd->lock);
1209 }
1210 
osd_cleanup(struct ceph_osd * osd)1211 static void osd_cleanup(struct ceph_osd *osd)
1212 {
1213 	WARN_ON(!RB_EMPTY_NODE(&osd->o_node));
1214 	WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
1215 	WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
1216 	WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoff_mappings));
1217 	WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoffs_by_id));
1218 	WARN_ON(!list_empty(&osd->o_osd_lru));
1219 	WARN_ON(!list_empty(&osd->o_keepalive_item));
1220 
1221 	if (osd->o_auth.authorizer) {
1222 		WARN_ON(osd_homeless(osd));
1223 		ceph_auth_destroy_authorizer(osd->o_auth.authorizer);
1224 	}
1225 }
1226 
1227 /*
1228  * Track open sessions with osds.
1229  */
create_osd(struct ceph_osd_client * osdc,int onum)1230 static struct ceph_osd *create_osd(struct ceph_osd_client *osdc, int onum)
1231 {
1232 	struct ceph_osd *osd;
1233 
1234 	WARN_ON(onum == CEPH_HOMELESS_OSD);
1235 
1236 	osd = kzalloc(sizeof(*osd), GFP_NOIO | __GFP_NOFAIL);
1237 	osd_init(osd);
1238 	osd->o_osdc = osdc;
1239 	osd->o_osd = onum;
1240 
1241 	ceph_con_init(&osd->o_con, osd, &osd_con_ops, &osdc->client->msgr);
1242 
1243 	return osd;
1244 }
1245 
get_osd(struct ceph_osd * osd)1246 static struct ceph_osd *get_osd(struct ceph_osd *osd)
1247 {
1248 	if (refcount_inc_not_zero(&osd->o_ref)) {
1249 		dout("get_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref)-1,
1250 		     refcount_read(&osd->o_ref));
1251 		return osd;
1252 	} else {
1253 		dout("get_osd %p FAIL\n", osd);
1254 		return NULL;
1255 	}
1256 }
1257 
put_osd(struct ceph_osd * osd)1258 static void put_osd(struct ceph_osd *osd)
1259 {
1260 	dout("put_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref),
1261 	     refcount_read(&osd->o_ref) - 1);
1262 	if (refcount_dec_and_test(&osd->o_ref)) {
1263 		osd_cleanup(osd);
1264 		kfree(osd);
1265 	}
1266 }
1267 
DEFINE_RB_FUNCS(osd,struct ceph_osd,o_osd,o_node)1268 DEFINE_RB_FUNCS(osd, struct ceph_osd, o_osd, o_node)
1269 
1270 static void __move_osd_to_lru(struct ceph_osd *osd)
1271 {
1272 	struct ceph_osd_client *osdc = osd->o_osdc;
1273 
1274 	dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1275 	BUG_ON(!list_empty(&osd->o_osd_lru));
1276 
1277 	spin_lock(&osdc->osd_lru_lock);
1278 	list_add_tail(&osd->o_osd_lru, &osdc->osd_lru);
1279 	spin_unlock(&osdc->osd_lru_lock);
1280 
1281 	osd->lru_ttl = jiffies + osdc->client->options->osd_idle_ttl;
1282 }
1283 
maybe_move_osd_to_lru(struct ceph_osd * osd)1284 static void maybe_move_osd_to_lru(struct ceph_osd *osd)
1285 {
1286 	if (RB_EMPTY_ROOT(&osd->o_requests) &&
1287 	    RB_EMPTY_ROOT(&osd->o_linger_requests))
1288 		__move_osd_to_lru(osd);
1289 }
1290 
__remove_osd_from_lru(struct ceph_osd * osd)1291 static void __remove_osd_from_lru(struct ceph_osd *osd)
1292 {
1293 	struct ceph_osd_client *osdc = osd->o_osdc;
1294 
1295 	dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1296 
1297 	spin_lock(&osdc->osd_lru_lock);
1298 	if (!list_empty(&osd->o_osd_lru))
1299 		list_del_init(&osd->o_osd_lru);
1300 	spin_unlock(&osdc->osd_lru_lock);
1301 }
1302 
1303 /*
1304  * Close the connection and assign any leftover requests to the
1305  * homeless session.
1306  */
close_osd(struct ceph_osd * osd)1307 static void close_osd(struct ceph_osd *osd)
1308 {
1309 	struct ceph_osd_client *osdc = osd->o_osdc;
1310 	struct rb_node *n;
1311 
1312 	verify_osdc_wrlocked(osdc);
1313 	dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1314 
1315 	ceph_con_close(&osd->o_con);
1316 
1317 	for (n = rb_first(&osd->o_requests); n; ) {
1318 		struct ceph_osd_request *req =
1319 		    rb_entry(n, struct ceph_osd_request, r_node);
1320 
1321 		n = rb_next(n); /* unlink_request() */
1322 
1323 		dout(" reassigning req %p tid %llu\n", req, req->r_tid);
1324 		unlink_request(osd, req);
1325 		link_request(&osdc->homeless_osd, req);
1326 	}
1327 	for (n = rb_first(&osd->o_linger_requests); n; ) {
1328 		struct ceph_osd_linger_request *lreq =
1329 		    rb_entry(n, struct ceph_osd_linger_request, node);
1330 
1331 		n = rb_next(n); /* unlink_linger() */
1332 
1333 		dout(" reassigning lreq %p linger_id %llu\n", lreq,
1334 		     lreq->linger_id);
1335 		unlink_linger(osd, lreq);
1336 		link_linger(&osdc->homeless_osd, lreq);
1337 	}
1338 	clear_backoffs(osd);
1339 
1340 	__remove_osd_from_lru(osd);
1341 	erase_osd(&osdc->osds, osd);
1342 	put_osd(osd);
1343 }
1344 
1345 /*
1346  * reset osd connect
1347  */
reopen_osd(struct ceph_osd * osd)1348 static int reopen_osd(struct ceph_osd *osd)
1349 {
1350 	struct ceph_entity_addr *peer_addr;
1351 
1352 	dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1353 
1354 	if (RB_EMPTY_ROOT(&osd->o_requests) &&
1355 	    RB_EMPTY_ROOT(&osd->o_linger_requests)) {
1356 		close_osd(osd);
1357 		return -ENODEV;
1358 	}
1359 
1360 	peer_addr = &osd->o_osdc->osdmap->osd_addr[osd->o_osd];
1361 	if (!memcmp(peer_addr, &osd->o_con.peer_addr, sizeof (*peer_addr)) &&
1362 			!ceph_con_opened(&osd->o_con)) {
1363 		struct rb_node *n;
1364 
1365 		dout("osd addr hasn't changed and connection never opened, "
1366 		     "letting msgr retry\n");
1367 		/* touch each r_stamp for handle_timeout()'s benfit */
1368 		for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
1369 			struct ceph_osd_request *req =
1370 			    rb_entry(n, struct ceph_osd_request, r_node);
1371 			req->r_stamp = jiffies;
1372 		}
1373 
1374 		return -EAGAIN;
1375 	}
1376 
1377 	ceph_con_close(&osd->o_con);
1378 	ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd, peer_addr);
1379 	osd->o_incarnation++;
1380 
1381 	return 0;
1382 }
1383 
lookup_create_osd(struct ceph_osd_client * osdc,int o,bool wrlocked)1384 static struct ceph_osd *lookup_create_osd(struct ceph_osd_client *osdc, int o,
1385 					  bool wrlocked)
1386 {
1387 	struct ceph_osd *osd;
1388 
1389 	if (wrlocked)
1390 		verify_osdc_wrlocked(osdc);
1391 	else
1392 		verify_osdc_locked(osdc);
1393 
1394 	if (o != CEPH_HOMELESS_OSD)
1395 		osd = lookup_osd(&osdc->osds, o);
1396 	else
1397 		osd = &osdc->homeless_osd;
1398 	if (!osd) {
1399 		if (!wrlocked)
1400 			return ERR_PTR(-EAGAIN);
1401 
1402 		osd = create_osd(osdc, o);
1403 		insert_osd(&osdc->osds, osd);
1404 		ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd,
1405 			      &osdc->osdmap->osd_addr[osd->o_osd]);
1406 	}
1407 
1408 	dout("%s osdc %p osd%d -> osd %p\n", __func__, osdc, o, osd);
1409 	return osd;
1410 }
1411 
1412 /*
1413  * Create request <-> OSD session relation.
1414  *
1415  * @req has to be assigned a tid, @osd may be homeless.
1416  */
link_request(struct ceph_osd * osd,struct ceph_osd_request * req)1417 static void link_request(struct ceph_osd *osd, struct ceph_osd_request *req)
1418 {
1419 	verify_osd_locked(osd);
1420 	WARN_ON(!req->r_tid || req->r_osd);
1421 	dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1422 	     req, req->r_tid);
1423 
1424 	if (!osd_homeless(osd))
1425 		__remove_osd_from_lru(osd);
1426 	else
1427 		atomic_inc(&osd->o_osdc->num_homeless);
1428 
1429 	get_osd(osd);
1430 	insert_request(&osd->o_requests, req);
1431 	req->r_osd = osd;
1432 }
1433 
unlink_request(struct ceph_osd * osd,struct ceph_osd_request * req)1434 static void unlink_request(struct ceph_osd *osd, struct ceph_osd_request *req)
1435 {
1436 	verify_osd_locked(osd);
1437 	WARN_ON(req->r_osd != osd);
1438 	dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1439 	     req, req->r_tid);
1440 
1441 	req->r_osd = NULL;
1442 	erase_request(&osd->o_requests, req);
1443 	put_osd(osd);
1444 
1445 	if (!osd_homeless(osd))
1446 		maybe_move_osd_to_lru(osd);
1447 	else
1448 		atomic_dec(&osd->o_osdc->num_homeless);
1449 }
1450 
__pool_full(struct ceph_pg_pool_info * pi)1451 static bool __pool_full(struct ceph_pg_pool_info *pi)
1452 {
1453 	return pi->flags & CEPH_POOL_FLAG_FULL;
1454 }
1455 
have_pool_full(struct ceph_osd_client * osdc)1456 static bool have_pool_full(struct ceph_osd_client *osdc)
1457 {
1458 	struct rb_node *n;
1459 
1460 	for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
1461 		struct ceph_pg_pool_info *pi =
1462 		    rb_entry(n, struct ceph_pg_pool_info, node);
1463 
1464 		if (__pool_full(pi))
1465 			return true;
1466 	}
1467 
1468 	return false;
1469 }
1470 
pool_full(struct ceph_osd_client * osdc,s64 pool_id)1471 static bool pool_full(struct ceph_osd_client *osdc, s64 pool_id)
1472 {
1473 	struct ceph_pg_pool_info *pi;
1474 
1475 	pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
1476 	if (!pi)
1477 		return false;
1478 
1479 	return __pool_full(pi);
1480 }
1481 
1482 /*
1483  * Returns whether a request should be blocked from being sent
1484  * based on the current osdmap and osd_client settings.
1485  */
target_should_be_paused(struct ceph_osd_client * osdc,const struct ceph_osd_request_target * t,struct ceph_pg_pool_info * pi)1486 static bool target_should_be_paused(struct ceph_osd_client *osdc,
1487 				    const struct ceph_osd_request_target *t,
1488 				    struct ceph_pg_pool_info *pi)
1489 {
1490 	bool pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
1491 	bool pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
1492 		       ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
1493 		       __pool_full(pi);
1494 
1495 	WARN_ON(pi->id != t->target_oloc.pool);
1496 	return ((t->flags & CEPH_OSD_FLAG_READ) && pauserd) ||
1497 	       ((t->flags & CEPH_OSD_FLAG_WRITE) && pausewr) ||
1498 	       (osdc->osdmap->epoch < osdc->epoch_barrier);
1499 }
1500 
pick_random_replica(const struct ceph_osds * acting)1501 static int pick_random_replica(const struct ceph_osds *acting)
1502 {
1503 	int i = prandom_u32() % acting->size;
1504 
1505 	dout("%s picked osd%d, primary osd%d\n", __func__,
1506 	     acting->osds[i], acting->primary);
1507 	return i;
1508 }
1509 
1510 /*
1511  * Picks the closest replica based on client's location given by
1512  * crush_location option.  Prefers the primary if the locality is
1513  * the same.
1514  */
pick_closest_replica(struct ceph_osd_client * osdc,const struct ceph_osds * acting)1515 static int pick_closest_replica(struct ceph_osd_client *osdc,
1516 				const struct ceph_osds *acting)
1517 {
1518 	struct ceph_options *opt = osdc->client->options;
1519 	int best_i, best_locality;
1520 	int i = 0, locality;
1521 
1522 	do {
1523 		locality = ceph_get_crush_locality(osdc->osdmap,
1524 						   acting->osds[i],
1525 						   &opt->crush_locs);
1526 		if (i == 0 ||
1527 		    (locality >= 0 && best_locality < 0) ||
1528 		    (locality >= 0 && best_locality >= 0 &&
1529 		     locality < best_locality)) {
1530 			best_i = i;
1531 			best_locality = locality;
1532 		}
1533 	} while (++i < acting->size);
1534 
1535 	dout("%s picked osd%d with locality %d, primary osd%d\n", __func__,
1536 	     acting->osds[best_i], best_locality, acting->primary);
1537 	return best_i;
1538 }
1539 
1540 enum calc_target_result {
1541 	CALC_TARGET_NO_ACTION = 0,
1542 	CALC_TARGET_NEED_RESEND,
1543 	CALC_TARGET_POOL_DNE,
1544 };
1545 
calc_target(struct ceph_osd_client * osdc,struct ceph_osd_request_target * t,bool any_change)1546 static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
1547 					   struct ceph_osd_request_target *t,
1548 					   bool any_change)
1549 {
1550 	struct ceph_pg_pool_info *pi;
1551 	struct ceph_pg pgid, last_pgid;
1552 	struct ceph_osds up, acting;
1553 	bool is_read = t->flags & CEPH_OSD_FLAG_READ;
1554 	bool is_write = t->flags & CEPH_OSD_FLAG_WRITE;
1555 	bool force_resend = false;
1556 	bool unpaused = false;
1557 	bool legacy_change = false;
1558 	bool split = false;
1559 	bool sort_bitwise = ceph_osdmap_flag(osdc, CEPH_OSDMAP_SORTBITWISE);
1560 	bool recovery_deletes = ceph_osdmap_flag(osdc,
1561 						 CEPH_OSDMAP_RECOVERY_DELETES);
1562 	enum calc_target_result ct_res;
1563 
1564 	t->epoch = osdc->osdmap->epoch;
1565 	pi = ceph_pg_pool_by_id(osdc->osdmap, t->base_oloc.pool);
1566 	if (!pi) {
1567 		t->osd = CEPH_HOMELESS_OSD;
1568 		ct_res = CALC_TARGET_POOL_DNE;
1569 		goto out;
1570 	}
1571 
1572 	if (osdc->osdmap->epoch == pi->last_force_request_resend) {
1573 		if (t->last_force_resend < pi->last_force_request_resend) {
1574 			t->last_force_resend = pi->last_force_request_resend;
1575 			force_resend = true;
1576 		} else if (t->last_force_resend == 0) {
1577 			force_resend = true;
1578 		}
1579 	}
1580 
1581 	/* apply tiering */
1582 	ceph_oid_copy(&t->target_oid, &t->base_oid);
1583 	ceph_oloc_copy(&t->target_oloc, &t->base_oloc);
1584 	if ((t->flags & CEPH_OSD_FLAG_IGNORE_OVERLAY) == 0) {
1585 		if (is_read && pi->read_tier >= 0)
1586 			t->target_oloc.pool = pi->read_tier;
1587 		if (is_write && pi->write_tier >= 0)
1588 			t->target_oloc.pool = pi->write_tier;
1589 
1590 		pi = ceph_pg_pool_by_id(osdc->osdmap, t->target_oloc.pool);
1591 		if (!pi) {
1592 			t->osd = CEPH_HOMELESS_OSD;
1593 			ct_res = CALC_TARGET_POOL_DNE;
1594 			goto out;
1595 		}
1596 	}
1597 
1598 	__ceph_object_locator_to_pg(pi, &t->target_oid, &t->target_oloc, &pgid);
1599 	last_pgid.pool = pgid.pool;
1600 	last_pgid.seed = ceph_stable_mod(pgid.seed, t->pg_num, t->pg_num_mask);
1601 
1602 	ceph_pg_to_up_acting_osds(osdc->osdmap, pi, &pgid, &up, &acting);
1603 	if (any_change &&
1604 	    ceph_is_new_interval(&t->acting,
1605 				 &acting,
1606 				 &t->up,
1607 				 &up,
1608 				 t->size,
1609 				 pi->size,
1610 				 t->min_size,
1611 				 pi->min_size,
1612 				 t->pg_num,
1613 				 pi->pg_num,
1614 				 t->sort_bitwise,
1615 				 sort_bitwise,
1616 				 t->recovery_deletes,
1617 				 recovery_deletes,
1618 				 &last_pgid))
1619 		force_resend = true;
1620 
1621 	if (t->paused && !target_should_be_paused(osdc, t, pi)) {
1622 		t->paused = false;
1623 		unpaused = true;
1624 	}
1625 	legacy_change = ceph_pg_compare(&t->pgid, &pgid) ||
1626 			ceph_osds_changed(&t->acting, &acting,
1627 					  t->used_replica || any_change);
1628 	if (t->pg_num)
1629 		split = ceph_pg_is_split(&last_pgid, t->pg_num, pi->pg_num);
1630 
1631 	if (legacy_change || force_resend || split) {
1632 		t->pgid = pgid; /* struct */
1633 		ceph_pg_to_primary_shard(osdc->osdmap, pi, &pgid, &t->spgid);
1634 		ceph_osds_copy(&t->acting, &acting);
1635 		ceph_osds_copy(&t->up, &up);
1636 		t->size = pi->size;
1637 		t->min_size = pi->min_size;
1638 		t->pg_num = pi->pg_num;
1639 		t->pg_num_mask = pi->pg_num_mask;
1640 		t->sort_bitwise = sort_bitwise;
1641 		t->recovery_deletes = recovery_deletes;
1642 
1643 		if ((t->flags & (CEPH_OSD_FLAG_BALANCE_READS |
1644 				 CEPH_OSD_FLAG_LOCALIZE_READS)) &&
1645 		    !is_write && pi->type == CEPH_POOL_TYPE_REP &&
1646 		    acting.size > 1) {
1647 			int pos;
1648 
1649 			WARN_ON(!is_read || acting.osds[0] != acting.primary);
1650 			if (t->flags & CEPH_OSD_FLAG_BALANCE_READS) {
1651 				pos = pick_random_replica(&acting);
1652 			} else {
1653 				pos = pick_closest_replica(osdc, &acting);
1654 			}
1655 			t->osd = acting.osds[pos];
1656 			t->used_replica = pos > 0;
1657 		} else {
1658 			t->osd = acting.primary;
1659 			t->used_replica = false;
1660 		}
1661 	}
1662 
1663 	if (unpaused || legacy_change || force_resend || split)
1664 		ct_res = CALC_TARGET_NEED_RESEND;
1665 	else
1666 		ct_res = CALC_TARGET_NO_ACTION;
1667 
1668 out:
1669 	dout("%s t %p -> %d%d%d%d ct_res %d osd%d\n", __func__, t, unpaused,
1670 	     legacy_change, force_resend, split, ct_res, t->osd);
1671 	return ct_res;
1672 }
1673 
alloc_spg_mapping(void)1674 static struct ceph_spg_mapping *alloc_spg_mapping(void)
1675 {
1676 	struct ceph_spg_mapping *spg;
1677 
1678 	spg = kmalloc(sizeof(*spg), GFP_NOIO);
1679 	if (!spg)
1680 		return NULL;
1681 
1682 	RB_CLEAR_NODE(&spg->node);
1683 	spg->backoffs = RB_ROOT;
1684 	return spg;
1685 }
1686 
free_spg_mapping(struct ceph_spg_mapping * spg)1687 static void free_spg_mapping(struct ceph_spg_mapping *spg)
1688 {
1689 	WARN_ON(!RB_EMPTY_NODE(&spg->node));
1690 	WARN_ON(!RB_EMPTY_ROOT(&spg->backoffs));
1691 
1692 	kfree(spg);
1693 }
1694 
1695 /*
1696  * rbtree of ceph_spg_mapping for handling map<spg_t, ...>, similar to
1697  * ceph_pg_mapping.  Used to track OSD backoffs -- a backoff [range] is
1698  * defined only within a specific spgid; it does not pass anything to
1699  * children on split, or to another primary.
1700  */
DEFINE_RB_FUNCS2(spg_mapping,struct ceph_spg_mapping,spgid,ceph_spg_compare,RB_BYPTR,const struct ceph_spg *,node)1701 DEFINE_RB_FUNCS2(spg_mapping, struct ceph_spg_mapping, spgid, ceph_spg_compare,
1702 		 RB_BYPTR, const struct ceph_spg *, node)
1703 
1704 static u64 hoid_get_bitwise_key(const struct ceph_hobject_id *hoid)
1705 {
1706 	return hoid->is_max ? 0x100000000ull : hoid->hash_reverse_bits;
1707 }
1708 
hoid_get_effective_key(const struct ceph_hobject_id * hoid,void ** pkey,size_t * pkey_len)1709 static void hoid_get_effective_key(const struct ceph_hobject_id *hoid,
1710 				   void **pkey, size_t *pkey_len)
1711 {
1712 	if (hoid->key_len) {
1713 		*pkey = hoid->key;
1714 		*pkey_len = hoid->key_len;
1715 	} else {
1716 		*pkey = hoid->oid;
1717 		*pkey_len = hoid->oid_len;
1718 	}
1719 }
1720 
compare_names(const void * name1,size_t name1_len,const void * name2,size_t name2_len)1721 static int compare_names(const void *name1, size_t name1_len,
1722 			 const void *name2, size_t name2_len)
1723 {
1724 	int ret;
1725 
1726 	ret = memcmp(name1, name2, min(name1_len, name2_len));
1727 	if (!ret) {
1728 		if (name1_len < name2_len)
1729 			ret = -1;
1730 		else if (name1_len > name2_len)
1731 			ret = 1;
1732 	}
1733 	return ret;
1734 }
1735 
hoid_compare(const struct ceph_hobject_id * lhs,const struct ceph_hobject_id * rhs)1736 static int hoid_compare(const struct ceph_hobject_id *lhs,
1737 			const struct ceph_hobject_id *rhs)
1738 {
1739 	void *effective_key1, *effective_key2;
1740 	size_t effective_key1_len, effective_key2_len;
1741 	int ret;
1742 
1743 	if (lhs->is_max < rhs->is_max)
1744 		return -1;
1745 	if (lhs->is_max > rhs->is_max)
1746 		return 1;
1747 
1748 	if (lhs->pool < rhs->pool)
1749 		return -1;
1750 	if (lhs->pool > rhs->pool)
1751 		return 1;
1752 
1753 	if (hoid_get_bitwise_key(lhs) < hoid_get_bitwise_key(rhs))
1754 		return -1;
1755 	if (hoid_get_bitwise_key(lhs) > hoid_get_bitwise_key(rhs))
1756 		return 1;
1757 
1758 	ret = compare_names(lhs->nspace, lhs->nspace_len,
1759 			    rhs->nspace, rhs->nspace_len);
1760 	if (ret)
1761 		return ret;
1762 
1763 	hoid_get_effective_key(lhs, &effective_key1, &effective_key1_len);
1764 	hoid_get_effective_key(rhs, &effective_key2, &effective_key2_len);
1765 	ret = compare_names(effective_key1, effective_key1_len,
1766 			    effective_key2, effective_key2_len);
1767 	if (ret)
1768 		return ret;
1769 
1770 	ret = compare_names(lhs->oid, lhs->oid_len, rhs->oid, rhs->oid_len);
1771 	if (ret)
1772 		return ret;
1773 
1774 	if (lhs->snapid < rhs->snapid)
1775 		return -1;
1776 	if (lhs->snapid > rhs->snapid)
1777 		return 1;
1778 
1779 	return 0;
1780 }
1781 
1782 /*
1783  * For decoding ->begin and ->end of MOSDBackoff only -- no MIN/MAX
1784  * compat stuff here.
1785  *
1786  * Assumes @hoid is zero-initialized.
1787  */
decode_hoid(void ** p,void * end,struct ceph_hobject_id * hoid)1788 static int decode_hoid(void **p, void *end, struct ceph_hobject_id *hoid)
1789 {
1790 	u8 struct_v;
1791 	u32 struct_len;
1792 	int ret;
1793 
1794 	ret = ceph_start_decoding(p, end, 4, "hobject_t", &struct_v,
1795 				  &struct_len);
1796 	if (ret)
1797 		return ret;
1798 
1799 	if (struct_v < 4) {
1800 		pr_err("got struct_v %d < 4 of hobject_t\n", struct_v);
1801 		goto e_inval;
1802 	}
1803 
1804 	hoid->key = ceph_extract_encoded_string(p, end, &hoid->key_len,
1805 						GFP_NOIO);
1806 	if (IS_ERR(hoid->key)) {
1807 		ret = PTR_ERR(hoid->key);
1808 		hoid->key = NULL;
1809 		return ret;
1810 	}
1811 
1812 	hoid->oid = ceph_extract_encoded_string(p, end, &hoid->oid_len,
1813 						GFP_NOIO);
1814 	if (IS_ERR(hoid->oid)) {
1815 		ret = PTR_ERR(hoid->oid);
1816 		hoid->oid = NULL;
1817 		return ret;
1818 	}
1819 
1820 	ceph_decode_64_safe(p, end, hoid->snapid, e_inval);
1821 	ceph_decode_32_safe(p, end, hoid->hash, e_inval);
1822 	ceph_decode_8_safe(p, end, hoid->is_max, e_inval);
1823 
1824 	hoid->nspace = ceph_extract_encoded_string(p, end, &hoid->nspace_len,
1825 						   GFP_NOIO);
1826 	if (IS_ERR(hoid->nspace)) {
1827 		ret = PTR_ERR(hoid->nspace);
1828 		hoid->nspace = NULL;
1829 		return ret;
1830 	}
1831 
1832 	ceph_decode_64_safe(p, end, hoid->pool, e_inval);
1833 
1834 	ceph_hoid_build_hash_cache(hoid);
1835 	return 0;
1836 
1837 e_inval:
1838 	return -EINVAL;
1839 }
1840 
hoid_encoding_size(const struct ceph_hobject_id * hoid)1841 static int hoid_encoding_size(const struct ceph_hobject_id *hoid)
1842 {
1843 	return 8 + 4 + 1 + 8 + /* snapid, hash, is_max, pool */
1844 	       4 + hoid->key_len + 4 + hoid->oid_len + 4 + hoid->nspace_len;
1845 }
1846 
encode_hoid(void ** p,void * end,const struct ceph_hobject_id * hoid)1847 static void encode_hoid(void **p, void *end, const struct ceph_hobject_id *hoid)
1848 {
1849 	ceph_start_encoding(p, 4, 3, hoid_encoding_size(hoid));
1850 	ceph_encode_string(p, end, hoid->key, hoid->key_len);
1851 	ceph_encode_string(p, end, hoid->oid, hoid->oid_len);
1852 	ceph_encode_64(p, hoid->snapid);
1853 	ceph_encode_32(p, hoid->hash);
1854 	ceph_encode_8(p, hoid->is_max);
1855 	ceph_encode_string(p, end, hoid->nspace, hoid->nspace_len);
1856 	ceph_encode_64(p, hoid->pool);
1857 }
1858 
free_hoid(struct ceph_hobject_id * hoid)1859 static void free_hoid(struct ceph_hobject_id *hoid)
1860 {
1861 	if (hoid) {
1862 		kfree(hoid->key);
1863 		kfree(hoid->oid);
1864 		kfree(hoid->nspace);
1865 		kfree(hoid);
1866 	}
1867 }
1868 
alloc_backoff(void)1869 static struct ceph_osd_backoff *alloc_backoff(void)
1870 {
1871 	struct ceph_osd_backoff *backoff;
1872 
1873 	backoff = kzalloc(sizeof(*backoff), GFP_NOIO);
1874 	if (!backoff)
1875 		return NULL;
1876 
1877 	RB_CLEAR_NODE(&backoff->spg_node);
1878 	RB_CLEAR_NODE(&backoff->id_node);
1879 	return backoff;
1880 }
1881 
free_backoff(struct ceph_osd_backoff * backoff)1882 static void free_backoff(struct ceph_osd_backoff *backoff)
1883 {
1884 	WARN_ON(!RB_EMPTY_NODE(&backoff->spg_node));
1885 	WARN_ON(!RB_EMPTY_NODE(&backoff->id_node));
1886 
1887 	free_hoid(backoff->begin);
1888 	free_hoid(backoff->end);
1889 	kfree(backoff);
1890 }
1891 
1892 /*
1893  * Within a specific spgid, backoffs are managed by ->begin hoid.
1894  */
1895 DEFINE_RB_INSDEL_FUNCS2(backoff, struct ceph_osd_backoff, begin, hoid_compare,
1896 			RB_BYVAL, spg_node);
1897 
lookup_containing_backoff(struct rb_root * root,const struct ceph_hobject_id * hoid)1898 static struct ceph_osd_backoff *lookup_containing_backoff(struct rb_root *root,
1899 					    const struct ceph_hobject_id *hoid)
1900 {
1901 	struct rb_node *n = root->rb_node;
1902 
1903 	while (n) {
1904 		struct ceph_osd_backoff *cur =
1905 		    rb_entry(n, struct ceph_osd_backoff, spg_node);
1906 		int cmp;
1907 
1908 		cmp = hoid_compare(hoid, cur->begin);
1909 		if (cmp < 0) {
1910 			n = n->rb_left;
1911 		} else if (cmp > 0) {
1912 			if (hoid_compare(hoid, cur->end) < 0)
1913 				return cur;
1914 
1915 			n = n->rb_right;
1916 		} else {
1917 			return cur;
1918 		}
1919 	}
1920 
1921 	return NULL;
1922 }
1923 
1924 /*
1925  * Each backoff has a unique id within its OSD session.
1926  */
DEFINE_RB_FUNCS(backoff_by_id,struct ceph_osd_backoff,id,id_node)1927 DEFINE_RB_FUNCS(backoff_by_id, struct ceph_osd_backoff, id, id_node)
1928 
1929 static void clear_backoffs(struct ceph_osd *osd)
1930 {
1931 	while (!RB_EMPTY_ROOT(&osd->o_backoff_mappings)) {
1932 		struct ceph_spg_mapping *spg =
1933 		    rb_entry(rb_first(&osd->o_backoff_mappings),
1934 			     struct ceph_spg_mapping, node);
1935 
1936 		while (!RB_EMPTY_ROOT(&spg->backoffs)) {
1937 			struct ceph_osd_backoff *backoff =
1938 			    rb_entry(rb_first(&spg->backoffs),
1939 				     struct ceph_osd_backoff, spg_node);
1940 
1941 			erase_backoff(&spg->backoffs, backoff);
1942 			erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
1943 			free_backoff(backoff);
1944 		}
1945 		erase_spg_mapping(&osd->o_backoff_mappings, spg);
1946 		free_spg_mapping(spg);
1947 	}
1948 }
1949 
1950 /*
1951  * Set up a temporary, non-owning view into @t.
1952  */
hoid_fill_from_target(struct ceph_hobject_id * hoid,const struct ceph_osd_request_target * t)1953 static void hoid_fill_from_target(struct ceph_hobject_id *hoid,
1954 				  const struct ceph_osd_request_target *t)
1955 {
1956 	hoid->key = NULL;
1957 	hoid->key_len = 0;
1958 	hoid->oid = t->target_oid.name;
1959 	hoid->oid_len = t->target_oid.name_len;
1960 	hoid->snapid = CEPH_NOSNAP;
1961 	hoid->hash = t->pgid.seed;
1962 	hoid->is_max = false;
1963 	if (t->target_oloc.pool_ns) {
1964 		hoid->nspace = t->target_oloc.pool_ns->str;
1965 		hoid->nspace_len = t->target_oloc.pool_ns->len;
1966 	} else {
1967 		hoid->nspace = NULL;
1968 		hoid->nspace_len = 0;
1969 	}
1970 	hoid->pool = t->target_oloc.pool;
1971 	ceph_hoid_build_hash_cache(hoid);
1972 }
1973 
should_plug_request(struct ceph_osd_request * req)1974 static bool should_plug_request(struct ceph_osd_request *req)
1975 {
1976 	struct ceph_osd *osd = req->r_osd;
1977 	struct ceph_spg_mapping *spg;
1978 	struct ceph_osd_backoff *backoff;
1979 	struct ceph_hobject_id hoid;
1980 
1981 	spg = lookup_spg_mapping(&osd->o_backoff_mappings, &req->r_t.spgid);
1982 	if (!spg)
1983 		return false;
1984 
1985 	hoid_fill_from_target(&hoid, &req->r_t);
1986 	backoff = lookup_containing_backoff(&spg->backoffs, &hoid);
1987 	if (!backoff)
1988 		return false;
1989 
1990 	dout("%s req %p tid %llu backoff osd%d spgid %llu.%xs%d id %llu\n",
1991 	     __func__, req, req->r_tid, osd->o_osd, backoff->spgid.pgid.pool,
1992 	     backoff->spgid.pgid.seed, backoff->spgid.shard, backoff->id);
1993 	return true;
1994 }
1995 
1996 /*
1997  * Keep get_num_data_items() in sync with this function.
1998  */
setup_request_data(struct ceph_osd_request * req)1999 static void setup_request_data(struct ceph_osd_request *req)
2000 {
2001 	struct ceph_msg *request_msg = req->r_request;
2002 	struct ceph_msg *reply_msg = req->r_reply;
2003 	struct ceph_osd_req_op *op;
2004 
2005 	if (req->r_request->num_data_items || req->r_reply->num_data_items)
2006 		return;
2007 
2008 	WARN_ON(request_msg->data_length || reply_msg->data_length);
2009 	for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) {
2010 		switch (op->op) {
2011 		/* request */
2012 		case CEPH_OSD_OP_WRITE:
2013 		case CEPH_OSD_OP_WRITEFULL:
2014 			WARN_ON(op->indata_len != op->extent.length);
2015 			ceph_osdc_msg_data_add(request_msg,
2016 					       &op->extent.osd_data);
2017 			break;
2018 		case CEPH_OSD_OP_SETXATTR:
2019 		case CEPH_OSD_OP_CMPXATTR:
2020 			WARN_ON(op->indata_len != op->xattr.name_len +
2021 						  op->xattr.value_len);
2022 			ceph_osdc_msg_data_add(request_msg,
2023 					       &op->xattr.osd_data);
2024 			break;
2025 		case CEPH_OSD_OP_NOTIFY_ACK:
2026 			ceph_osdc_msg_data_add(request_msg,
2027 					       &op->notify_ack.request_data);
2028 			break;
2029 		case CEPH_OSD_OP_COPY_FROM2:
2030 			ceph_osdc_msg_data_add(request_msg,
2031 					       &op->copy_from.osd_data);
2032 			break;
2033 
2034 		/* reply */
2035 		case CEPH_OSD_OP_STAT:
2036 			ceph_osdc_msg_data_add(reply_msg,
2037 					       &op->raw_data_in);
2038 			break;
2039 		case CEPH_OSD_OP_READ:
2040 			ceph_osdc_msg_data_add(reply_msg,
2041 					       &op->extent.osd_data);
2042 			break;
2043 		case CEPH_OSD_OP_LIST_WATCHERS:
2044 			ceph_osdc_msg_data_add(reply_msg,
2045 					       &op->list_watchers.response_data);
2046 			break;
2047 
2048 		/* both */
2049 		case CEPH_OSD_OP_CALL:
2050 			WARN_ON(op->indata_len != op->cls.class_len +
2051 						  op->cls.method_len +
2052 						  op->cls.indata_len);
2053 			ceph_osdc_msg_data_add(request_msg,
2054 					       &op->cls.request_info);
2055 			/* optional, can be NONE */
2056 			ceph_osdc_msg_data_add(request_msg,
2057 					       &op->cls.request_data);
2058 			/* optional, can be NONE */
2059 			ceph_osdc_msg_data_add(reply_msg,
2060 					       &op->cls.response_data);
2061 			break;
2062 		case CEPH_OSD_OP_NOTIFY:
2063 			ceph_osdc_msg_data_add(request_msg,
2064 					       &op->notify.request_data);
2065 			ceph_osdc_msg_data_add(reply_msg,
2066 					       &op->notify.response_data);
2067 			break;
2068 		}
2069 	}
2070 }
2071 
encode_pgid(void ** p,const struct ceph_pg * pgid)2072 static void encode_pgid(void **p, const struct ceph_pg *pgid)
2073 {
2074 	ceph_encode_8(p, 1);
2075 	ceph_encode_64(p, pgid->pool);
2076 	ceph_encode_32(p, pgid->seed);
2077 	ceph_encode_32(p, -1); /* preferred */
2078 }
2079 
encode_spgid(void ** p,const struct ceph_spg * spgid)2080 static void encode_spgid(void **p, const struct ceph_spg *spgid)
2081 {
2082 	ceph_start_encoding(p, 1, 1, CEPH_PGID_ENCODING_LEN + 1);
2083 	encode_pgid(p, &spgid->pgid);
2084 	ceph_encode_8(p, spgid->shard);
2085 }
2086 
encode_oloc(void ** p,void * end,const struct ceph_object_locator * oloc)2087 static void encode_oloc(void **p, void *end,
2088 			const struct ceph_object_locator *oloc)
2089 {
2090 	ceph_start_encoding(p, 5, 4, ceph_oloc_encoding_size(oloc));
2091 	ceph_encode_64(p, oloc->pool);
2092 	ceph_encode_32(p, -1); /* preferred */
2093 	ceph_encode_32(p, 0);  /* key len */
2094 	if (oloc->pool_ns)
2095 		ceph_encode_string(p, end, oloc->pool_ns->str,
2096 				   oloc->pool_ns->len);
2097 	else
2098 		ceph_encode_32(p, 0);
2099 }
2100 
encode_request_partial(struct ceph_osd_request * req,struct ceph_msg * msg)2101 static void encode_request_partial(struct ceph_osd_request *req,
2102 				   struct ceph_msg *msg)
2103 {
2104 	void *p = msg->front.iov_base;
2105 	void *const end = p + msg->front_alloc_len;
2106 	u32 data_len = 0;
2107 	int i;
2108 
2109 	if (req->r_flags & CEPH_OSD_FLAG_WRITE) {
2110 		/* snapshots aren't writeable */
2111 		WARN_ON(req->r_snapid != CEPH_NOSNAP);
2112 	} else {
2113 		WARN_ON(req->r_mtime.tv_sec || req->r_mtime.tv_nsec ||
2114 			req->r_data_offset || req->r_snapc);
2115 	}
2116 
2117 	setup_request_data(req);
2118 
2119 	encode_spgid(&p, &req->r_t.spgid); /* actual spg */
2120 	ceph_encode_32(&p, req->r_t.pgid.seed); /* raw hash */
2121 	ceph_encode_32(&p, req->r_osdc->osdmap->epoch);
2122 	ceph_encode_32(&p, req->r_flags);
2123 
2124 	/* reqid */
2125 	ceph_start_encoding(&p, 2, 2, sizeof(struct ceph_osd_reqid));
2126 	memset(p, 0, sizeof(struct ceph_osd_reqid));
2127 	p += sizeof(struct ceph_osd_reqid);
2128 
2129 	/* trace */
2130 	memset(p, 0, sizeof(struct ceph_blkin_trace_info));
2131 	p += sizeof(struct ceph_blkin_trace_info);
2132 
2133 	ceph_encode_32(&p, 0); /* client_inc, always 0 */
2134 	ceph_encode_timespec64(p, &req->r_mtime);
2135 	p += sizeof(struct ceph_timespec);
2136 
2137 	encode_oloc(&p, end, &req->r_t.target_oloc);
2138 	ceph_encode_string(&p, end, req->r_t.target_oid.name,
2139 			   req->r_t.target_oid.name_len);
2140 
2141 	/* ops, can imply data */
2142 	ceph_encode_16(&p, req->r_num_ops);
2143 	for (i = 0; i < req->r_num_ops; i++) {
2144 		data_len += osd_req_encode_op(p, &req->r_ops[i]);
2145 		p += sizeof(struct ceph_osd_op);
2146 	}
2147 
2148 	ceph_encode_64(&p, req->r_snapid); /* snapid */
2149 	if (req->r_snapc) {
2150 		ceph_encode_64(&p, req->r_snapc->seq);
2151 		ceph_encode_32(&p, req->r_snapc->num_snaps);
2152 		for (i = 0; i < req->r_snapc->num_snaps; i++)
2153 			ceph_encode_64(&p, req->r_snapc->snaps[i]);
2154 	} else {
2155 		ceph_encode_64(&p, 0); /* snap_seq */
2156 		ceph_encode_32(&p, 0); /* snaps len */
2157 	}
2158 
2159 	ceph_encode_32(&p, req->r_attempts); /* retry_attempt */
2160 	BUG_ON(p > end - 8); /* space for features */
2161 
2162 	msg->hdr.version = cpu_to_le16(8); /* MOSDOp v8 */
2163 	/* front_len is finalized in encode_request_finish() */
2164 	msg->front.iov_len = p - msg->front.iov_base;
2165 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2166 	msg->hdr.data_len = cpu_to_le32(data_len);
2167 	/*
2168 	 * The header "data_off" is a hint to the receiver allowing it
2169 	 * to align received data into its buffers such that there's no
2170 	 * need to re-copy it before writing it to disk (direct I/O).
2171 	 */
2172 	msg->hdr.data_off = cpu_to_le16(req->r_data_offset);
2173 
2174 	dout("%s req %p msg %p oid %s oid_len %d\n", __func__, req, msg,
2175 	     req->r_t.target_oid.name, req->r_t.target_oid.name_len);
2176 }
2177 
encode_request_finish(struct ceph_msg * msg)2178 static void encode_request_finish(struct ceph_msg *msg)
2179 {
2180 	void *p = msg->front.iov_base;
2181 	void *const partial_end = p + msg->front.iov_len;
2182 	void *const end = p + msg->front_alloc_len;
2183 
2184 	if (CEPH_HAVE_FEATURE(msg->con->peer_features, RESEND_ON_SPLIT)) {
2185 		/* luminous OSD -- encode features and be done */
2186 		p = partial_end;
2187 		ceph_encode_64(&p, msg->con->peer_features);
2188 	} else {
2189 		struct {
2190 			char spgid[CEPH_ENCODING_START_BLK_LEN +
2191 				   CEPH_PGID_ENCODING_LEN + 1];
2192 			__le32 hash;
2193 			__le32 epoch;
2194 			__le32 flags;
2195 			char reqid[CEPH_ENCODING_START_BLK_LEN +
2196 				   sizeof(struct ceph_osd_reqid)];
2197 			char trace[sizeof(struct ceph_blkin_trace_info)];
2198 			__le32 client_inc;
2199 			struct ceph_timespec mtime;
2200 		} __packed head;
2201 		struct ceph_pg pgid;
2202 		void *oloc, *oid, *tail;
2203 		int oloc_len, oid_len, tail_len;
2204 		int len;
2205 
2206 		/*
2207 		 * Pre-luminous OSD -- reencode v8 into v4 using @head
2208 		 * as a temporary buffer.  Encode the raw PG; the rest
2209 		 * is just a matter of moving oloc, oid and tail blobs
2210 		 * around.
2211 		 */
2212 		memcpy(&head, p, sizeof(head));
2213 		p += sizeof(head);
2214 
2215 		oloc = p;
2216 		p += CEPH_ENCODING_START_BLK_LEN;
2217 		pgid.pool = ceph_decode_64(&p);
2218 		p += 4 + 4; /* preferred, key len */
2219 		len = ceph_decode_32(&p);
2220 		p += len;   /* nspace */
2221 		oloc_len = p - oloc;
2222 
2223 		oid = p;
2224 		len = ceph_decode_32(&p);
2225 		p += len;
2226 		oid_len = p - oid;
2227 
2228 		tail = p;
2229 		tail_len = partial_end - p;
2230 
2231 		p = msg->front.iov_base;
2232 		ceph_encode_copy(&p, &head.client_inc, sizeof(head.client_inc));
2233 		ceph_encode_copy(&p, &head.epoch, sizeof(head.epoch));
2234 		ceph_encode_copy(&p, &head.flags, sizeof(head.flags));
2235 		ceph_encode_copy(&p, &head.mtime, sizeof(head.mtime));
2236 
2237 		/* reassert_version */
2238 		memset(p, 0, sizeof(struct ceph_eversion));
2239 		p += sizeof(struct ceph_eversion);
2240 
2241 		BUG_ON(p >= oloc);
2242 		memmove(p, oloc, oloc_len);
2243 		p += oloc_len;
2244 
2245 		pgid.seed = le32_to_cpu(head.hash);
2246 		encode_pgid(&p, &pgid); /* raw pg */
2247 
2248 		BUG_ON(p >= oid);
2249 		memmove(p, oid, oid_len);
2250 		p += oid_len;
2251 
2252 		/* tail -- ops, snapid, snapc, retry_attempt */
2253 		BUG_ON(p >= tail);
2254 		memmove(p, tail, tail_len);
2255 		p += tail_len;
2256 
2257 		msg->hdr.version = cpu_to_le16(4); /* MOSDOp v4 */
2258 	}
2259 
2260 	BUG_ON(p > end);
2261 	msg->front.iov_len = p - msg->front.iov_base;
2262 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2263 
2264 	dout("%s msg %p tid %llu %u+%u+%u v%d\n", __func__, msg,
2265 	     le64_to_cpu(msg->hdr.tid), le32_to_cpu(msg->hdr.front_len),
2266 	     le32_to_cpu(msg->hdr.middle_len), le32_to_cpu(msg->hdr.data_len),
2267 	     le16_to_cpu(msg->hdr.version));
2268 }
2269 
2270 /*
2271  * @req has to be assigned a tid and registered.
2272  */
send_request(struct ceph_osd_request * req)2273 static void send_request(struct ceph_osd_request *req)
2274 {
2275 	struct ceph_osd *osd = req->r_osd;
2276 
2277 	verify_osd_locked(osd);
2278 	WARN_ON(osd->o_osd != req->r_t.osd);
2279 
2280 	/* backoff? */
2281 	if (should_plug_request(req))
2282 		return;
2283 
2284 	/*
2285 	 * We may have a previously queued request message hanging
2286 	 * around.  Cancel it to avoid corrupting the msgr.
2287 	 */
2288 	if (req->r_sent)
2289 		ceph_msg_revoke(req->r_request);
2290 
2291 	req->r_flags |= CEPH_OSD_FLAG_KNOWN_REDIR;
2292 	if (req->r_attempts)
2293 		req->r_flags |= CEPH_OSD_FLAG_RETRY;
2294 	else
2295 		WARN_ON(req->r_flags & CEPH_OSD_FLAG_RETRY);
2296 
2297 	encode_request_partial(req, req->r_request);
2298 
2299 	dout("%s req %p tid %llu to pgid %llu.%x spgid %llu.%xs%d osd%d e%u flags 0x%x attempt %d\n",
2300 	     __func__, req, req->r_tid, req->r_t.pgid.pool, req->r_t.pgid.seed,
2301 	     req->r_t.spgid.pgid.pool, req->r_t.spgid.pgid.seed,
2302 	     req->r_t.spgid.shard, osd->o_osd, req->r_t.epoch, req->r_flags,
2303 	     req->r_attempts);
2304 
2305 	req->r_t.paused = false;
2306 	req->r_stamp = jiffies;
2307 	req->r_attempts++;
2308 
2309 	req->r_sent = osd->o_incarnation;
2310 	req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
2311 	ceph_con_send(&osd->o_con, ceph_msg_get(req->r_request));
2312 }
2313 
maybe_request_map(struct ceph_osd_client * osdc)2314 static void maybe_request_map(struct ceph_osd_client *osdc)
2315 {
2316 	bool continuous = false;
2317 
2318 	verify_osdc_locked(osdc);
2319 	WARN_ON(!osdc->osdmap->epoch);
2320 
2321 	if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2322 	    ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD) ||
2323 	    ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
2324 		dout("%s osdc %p continuous\n", __func__, osdc);
2325 		continuous = true;
2326 	} else {
2327 		dout("%s osdc %p onetime\n", __func__, osdc);
2328 	}
2329 
2330 	if (ceph_monc_want_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
2331 			       osdc->osdmap->epoch + 1, continuous))
2332 		ceph_monc_renew_subs(&osdc->client->monc);
2333 }
2334 
2335 static void complete_request(struct ceph_osd_request *req, int err);
2336 static void send_map_check(struct ceph_osd_request *req);
2337 
__submit_request(struct ceph_osd_request * req,bool wrlocked)2338 static void __submit_request(struct ceph_osd_request *req, bool wrlocked)
2339 {
2340 	struct ceph_osd_client *osdc = req->r_osdc;
2341 	struct ceph_osd *osd;
2342 	enum calc_target_result ct_res;
2343 	int err = 0;
2344 	bool need_send = false;
2345 	bool promoted = false;
2346 
2347 	WARN_ON(req->r_tid);
2348 	dout("%s req %p wrlocked %d\n", __func__, req, wrlocked);
2349 
2350 again:
2351 	ct_res = calc_target(osdc, &req->r_t, false);
2352 	if (ct_res == CALC_TARGET_POOL_DNE && !wrlocked)
2353 		goto promote;
2354 
2355 	osd = lookup_create_osd(osdc, req->r_t.osd, wrlocked);
2356 	if (IS_ERR(osd)) {
2357 		WARN_ON(PTR_ERR(osd) != -EAGAIN || wrlocked);
2358 		goto promote;
2359 	}
2360 
2361 	if (osdc->abort_err) {
2362 		dout("req %p abort_err %d\n", req, osdc->abort_err);
2363 		err = osdc->abort_err;
2364 	} else if (osdc->osdmap->epoch < osdc->epoch_barrier) {
2365 		dout("req %p epoch %u barrier %u\n", req, osdc->osdmap->epoch,
2366 		     osdc->epoch_barrier);
2367 		req->r_t.paused = true;
2368 		maybe_request_map(osdc);
2369 	} else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2370 		   ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
2371 		dout("req %p pausewr\n", req);
2372 		req->r_t.paused = true;
2373 		maybe_request_map(osdc);
2374 	} else if ((req->r_flags & CEPH_OSD_FLAG_READ) &&
2375 		   ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
2376 		dout("req %p pauserd\n", req);
2377 		req->r_t.paused = true;
2378 		maybe_request_map(osdc);
2379 	} else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2380 		   !(req->r_flags & (CEPH_OSD_FLAG_FULL_TRY |
2381 				     CEPH_OSD_FLAG_FULL_FORCE)) &&
2382 		   (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2383 		    pool_full(osdc, req->r_t.base_oloc.pool))) {
2384 		dout("req %p full/pool_full\n", req);
2385 		if (ceph_test_opt(osdc->client, ABORT_ON_FULL)) {
2386 			err = -ENOSPC;
2387 		} else {
2388 			pr_warn_ratelimited("FULL or reached pool quota\n");
2389 			req->r_t.paused = true;
2390 			maybe_request_map(osdc);
2391 		}
2392 	} else if (!osd_homeless(osd)) {
2393 		need_send = true;
2394 	} else {
2395 		maybe_request_map(osdc);
2396 	}
2397 
2398 	mutex_lock(&osd->lock);
2399 	/*
2400 	 * Assign the tid atomically with send_request() to protect
2401 	 * multiple writes to the same object from racing with each
2402 	 * other, resulting in out of order ops on the OSDs.
2403 	 */
2404 	req->r_tid = atomic64_inc_return(&osdc->last_tid);
2405 	link_request(osd, req);
2406 	if (need_send)
2407 		send_request(req);
2408 	else if (err)
2409 		complete_request(req, err);
2410 	mutex_unlock(&osd->lock);
2411 
2412 	if (!err && ct_res == CALC_TARGET_POOL_DNE)
2413 		send_map_check(req);
2414 
2415 	if (promoted)
2416 		downgrade_write(&osdc->lock);
2417 	return;
2418 
2419 promote:
2420 	up_read(&osdc->lock);
2421 	down_write(&osdc->lock);
2422 	wrlocked = true;
2423 	promoted = true;
2424 	goto again;
2425 }
2426 
account_request(struct ceph_osd_request * req)2427 static void account_request(struct ceph_osd_request *req)
2428 {
2429 	WARN_ON(req->r_flags & (CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK));
2430 	WARN_ON(!(req->r_flags & (CEPH_OSD_FLAG_READ | CEPH_OSD_FLAG_WRITE)));
2431 
2432 	req->r_flags |= CEPH_OSD_FLAG_ONDISK;
2433 	atomic_inc(&req->r_osdc->num_requests);
2434 
2435 	req->r_start_stamp = jiffies;
2436 	req->r_start_latency = ktime_get();
2437 }
2438 
submit_request(struct ceph_osd_request * req,bool wrlocked)2439 static void submit_request(struct ceph_osd_request *req, bool wrlocked)
2440 {
2441 	ceph_osdc_get_request(req);
2442 	account_request(req);
2443 	__submit_request(req, wrlocked);
2444 }
2445 
finish_request(struct ceph_osd_request * req)2446 static void finish_request(struct ceph_osd_request *req)
2447 {
2448 	struct ceph_osd_client *osdc = req->r_osdc;
2449 
2450 	WARN_ON(lookup_request_mc(&osdc->map_checks, req->r_tid));
2451 	dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2452 
2453 	req->r_end_latency = ktime_get();
2454 
2455 	if (req->r_osd)
2456 		unlink_request(req->r_osd, req);
2457 	atomic_dec(&osdc->num_requests);
2458 
2459 	/*
2460 	 * If an OSD has failed or returned and a request has been sent
2461 	 * twice, it's possible to get a reply and end up here while the
2462 	 * request message is queued for delivery.  We will ignore the
2463 	 * reply, so not a big deal, but better to try and catch it.
2464 	 */
2465 	ceph_msg_revoke(req->r_request);
2466 	ceph_msg_revoke_incoming(req->r_reply);
2467 }
2468 
__complete_request(struct ceph_osd_request * req)2469 static void __complete_request(struct ceph_osd_request *req)
2470 {
2471 	dout("%s req %p tid %llu cb %ps result %d\n", __func__, req,
2472 	     req->r_tid, req->r_callback, req->r_result);
2473 
2474 	if (req->r_callback)
2475 		req->r_callback(req);
2476 	complete_all(&req->r_completion);
2477 	ceph_osdc_put_request(req);
2478 }
2479 
complete_request_workfn(struct work_struct * work)2480 static void complete_request_workfn(struct work_struct *work)
2481 {
2482 	struct ceph_osd_request *req =
2483 	    container_of(work, struct ceph_osd_request, r_complete_work);
2484 
2485 	__complete_request(req);
2486 }
2487 
2488 /*
2489  * This is open-coded in handle_reply().
2490  */
complete_request(struct ceph_osd_request * req,int err)2491 static void complete_request(struct ceph_osd_request *req, int err)
2492 {
2493 	dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2494 
2495 	req->r_result = err;
2496 	finish_request(req);
2497 
2498 	INIT_WORK(&req->r_complete_work, complete_request_workfn);
2499 	queue_work(req->r_osdc->completion_wq, &req->r_complete_work);
2500 }
2501 
cancel_map_check(struct ceph_osd_request * req)2502 static void cancel_map_check(struct ceph_osd_request *req)
2503 {
2504 	struct ceph_osd_client *osdc = req->r_osdc;
2505 	struct ceph_osd_request *lookup_req;
2506 
2507 	verify_osdc_wrlocked(osdc);
2508 
2509 	lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2510 	if (!lookup_req)
2511 		return;
2512 
2513 	WARN_ON(lookup_req != req);
2514 	erase_request_mc(&osdc->map_checks, req);
2515 	ceph_osdc_put_request(req);
2516 }
2517 
cancel_request(struct ceph_osd_request * req)2518 static void cancel_request(struct ceph_osd_request *req)
2519 {
2520 	dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2521 
2522 	cancel_map_check(req);
2523 	finish_request(req);
2524 	complete_all(&req->r_completion);
2525 	ceph_osdc_put_request(req);
2526 }
2527 
abort_request(struct ceph_osd_request * req,int err)2528 static void abort_request(struct ceph_osd_request *req, int err)
2529 {
2530 	dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2531 
2532 	cancel_map_check(req);
2533 	complete_request(req, err);
2534 }
2535 
abort_fn(struct ceph_osd_request * req,void * arg)2536 static int abort_fn(struct ceph_osd_request *req, void *arg)
2537 {
2538 	int err = *(int *)arg;
2539 
2540 	abort_request(req, err);
2541 	return 0; /* continue iteration */
2542 }
2543 
2544 /*
2545  * Abort all in-flight requests with @err and arrange for all future
2546  * requests to be failed immediately.
2547  */
ceph_osdc_abort_requests(struct ceph_osd_client * osdc,int err)2548 void ceph_osdc_abort_requests(struct ceph_osd_client *osdc, int err)
2549 {
2550 	dout("%s osdc %p err %d\n", __func__, osdc, err);
2551 	down_write(&osdc->lock);
2552 	for_each_request(osdc, abort_fn, &err);
2553 	osdc->abort_err = err;
2554 	up_write(&osdc->lock);
2555 }
2556 EXPORT_SYMBOL(ceph_osdc_abort_requests);
2557 
ceph_osdc_clear_abort_err(struct ceph_osd_client * osdc)2558 void ceph_osdc_clear_abort_err(struct ceph_osd_client *osdc)
2559 {
2560 	down_write(&osdc->lock);
2561 	osdc->abort_err = 0;
2562 	up_write(&osdc->lock);
2563 }
2564 EXPORT_SYMBOL(ceph_osdc_clear_abort_err);
2565 
update_epoch_barrier(struct ceph_osd_client * osdc,u32 eb)2566 static void update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2567 {
2568 	if (likely(eb > osdc->epoch_barrier)) {
2569 		dout("updating epoch_barrier from %u to %u\n",
2570 				osdc->epoch_barrier, eb);
2571 		osdc->epoch_barrier = eb;
2572 		/* Request map if we're not to the barrier yet */
2573 		if (eb > osdc->osdmap->epoch)
2574 			maybe_request_map(osdc);
2575 	}
2576 }
2577 
ceph_osdc_update_epoch_barrier(struct ceph_osd_client * osdc,u32 eb)2578 void ceph_osdc_update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2579 {
2580 	down_read(&osdc->lock);
2581 	if (unlikely(eb > osdc->epoch_barrier)) {
2582 		up_read(&osdc->lock);
2583 		down_write(&osdc->lock);
2584 		update_epoch_barrier(osdc, eb);
2585 		up_write(&osdc->lock);
2586 	} else {
2587 		up_read(&osdc->lock);
2588 	}
2589 }
2590 EXPORT_SYMBOL(ceph_osdc_update_epoch_barrier);
2591 
2592 /*
2593  * We can end up releasing caps as a result of abort_request().
2594  * In that case, we probably want to ensure that the cap release message
2595  * has an updated epoch barrier in it, so set the epoch barrier prior to
2596  * aborting the first request.
2597  */
abort_on_full_fn(struct ceph_osd_request * req,void * arg)2598 static int abort_on_full_fn(struct ceph_osd_request *req, void *arg)
2599 {
2600 	struct ceph_osd_client *osdc = req->r_osdc;
2601 	bool *victims = arg;
2602 
2603 	if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2604 	    (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2605 	     pool_full(osdc, req->r_t.base_oloc.pool))) {
2606 		if (!*victims) {
2607 			update_epoch_barrier(osdc, osdc->osdmap->epoch);
2608 			*victims = true;
2609 		}
2610 		abort_request(req, -ENOSPC);
2611 	}
2612 
2613 	return 0; /* continue iteration */
2614 }
2615 
2616 /*
2617  * Drop all pending requests that are stalled waiting on a full condition to
2618  * clear, and complete them with ENOSPC as the return code. Set the
2619  * osdc->epoch_barrier to the latest map epoch that we've seen if any were
2620  * cancelled.
2621  */
ceph_osdc_abort_on_full(struct ceph_osd_client * osdc)2622 static void ceph_osdc_abort_on_full(struct ceph_osd_client *osdc)
2623 {
2624 	bool victims = false;
2625 
2626 	if (ceph_test_opt(osdc->client, ABORT_ON_FULL) &&
2627 	    (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || have_pool_full(osdc)))
2628 		for_each_request(osdc, abort_on_full_fn, &victims);
2629 }
2630 
check_pool_dne(struct ceph_osd_request * req)2631 static void check_pool_dne(struct ceph_osd_request *req)
2632 {
2633 	struct ceph_osd_client *osdc = req->r_osdc;
2634 	struct ceph_osdmap *map = osdc->osdmap;
2635 
2636 	verify_osdc_wrlocked(osdc);
2637 	WARN_ON(!map->epoch);
2638 
2639 	if (req->r_attempts) {
2640 		/*
2641 		 * We sent a request earlier, which means that
2642 		 * previously the pool existed, and now it does not
2643 		 * (i.e., it was deleted).
2644 		 */
2645 		req->r_map_dne_bound = map->epoch;
2646 		dout("%s req %p tid %llu pool disappeared\n", __func__, req,
2647 		     req->r_tid);
2648 	} else {
2649 		dout("%s req %p tid %llu map_dne_bound %u have %u\n", __func__,
2650 		     req, req->r_tid, req->r_map_dne_bound, map->epoch);
2651 	}
2652 
2653 	if (req->r_map_dne_bound) {
2654 		if (map->epoch >= req->r_map_dne_bound) {
2655 			/* we had a new enough map */
2656 			pr_info_ratelimited("tid %llu pool does not exist\n",
2657 					    req->r_tid);
2658 			complete_request(req, -ENOENT);
2659 		}
2660 	} else {
2661 		send_map_check(req);
2662 	}
2663 }
2664 
map_check_cb(struct ceph_mon_generic_request * greq)2665 static void map_check_cb(struct ceph_mon_generic_request *greq)
2666 {
2667 	struct ceph_osd_client *osdc = &greq->monc->client->osdc;
2668 	struct ceph_osd_request *req;
2669 	u64 tid = greq->private_data;
2670 
2671 	WARN_ON(greq->result || !greq->u.newest);
2672 
2673 	down_write(&osdc->lock);
2674 	req = lookup_request_mc(&osdc->map_checks, tid);
2675 	if (!req) {
2676 		dout("%s tid %llu dne\n", __func__, tid);
2677 		goto out_unlock;
2678 	}
2679 
2680 	dout("%s req %p tid %llu map_dne_bound %u newest %llu\n", __func__,
2681 	     req, req->r_tid, req->r_map_dne_bound, greq->u.newest);
2682 	if (!req->r_map_dne_bound)
2683 		req->r_map_dne_bound = greq->u.newest;
2684 	erase_request_mc(&osdc->map_checks, req);
2685 	check_pool_dne(req);
2686 
2687 	ceph_osdc_put_request(req);
2688 out_unlock:
2689 	up_write(&osdc->lock);
2690 }
2691 
send_map_check(struct ceph_osd_request * req)2692 static void send_map_check(struct ceph_osd_request *req)
2693 {
2694 	struct ceph_osd_client *osdc = req->r_osdc;
2695 	struct ceph_osd_request *lookup_req;
2696 	int ret;
2697 
2698 	verify_osdc_wrlocked(osdc);
2699 
2700 	lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2701 	if (lookup_req) {
2702 		WARN_ON(lookup_req != req);
2703 		return;
2704 	}
2705 
2706 	ceph_osdc_get_request(req);
2707 	insert_request_mc(&osdc->map_checks, req);
2708 	ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
2709 					  map_check_cb, req->r_tid);
2710 	WARN_ON(ret);
2711 }
2712 
2713 /*
2714  * lingering requests, watch/notify v2 infrastructure
2715  */
linger_release(struct kref * kref)2716 static void linger_release(struct kref *kref)
2717 {
2718 	struct ceph_osd_linger_request *lreq =
2719 	    container_of(kref, struct ceph_osd_linger_request, kref);
2720 
2721 	dout("%s lreq %p reg_req %p ping_req %p\n", __func__, lreq,
2722 	     lreq->reg_req, lreq->ping_req);
2723 	WARN_ON(!RB_EMPTY_NODE(&lreq->node));
2724 	WARN_ON(!RB_EMPTY_NODE(&lreq->osdc_node));
2725 	WARN_ON(!RB_EMPTY_NODE(&lreq->mc_node));
2726 	WARN_ON(!list_empty(&lreq->scan_item));
2727 	WARN_ON(!list_empty(&lreq->pending_lworks));
2728 	WARN_ON(lreq->osd);
2729 
2730 	if (lreq->reg_req)
2731 		ceph_osdc_put_request(lreq->reg_req);
2732 	if (lreq->ping_req)
2733 		ceph_osdc_put_request(lreq->ping_req);
2734 	target_destroy(&lreq->t);
2735 	kfree(lreq);
2736 }
2737 
linger_put(struct ceph_osd_linger_request * lreq)2738 static void linger_put(struct ceph_osd_linger_request *lreq)
2739 {
2740 	if (lreq)
2741 		kref_put(&lreq->kref, linger_release);
2742 }
2743 
2744 static struct ceph_osd_linger_request *
linger_get(struct ceph_osd_linger_request * lreq)2745 linger_get(struct ceph_osd_linger_request *lreq)
2746 {
2747 	kref_get(&lreq->kref);
2748 	return lreq;
2749 }
2750 
2751 static struct ceph_osd_linger_request *
linger_alloc(struct ceph_osd_client * osdc)2752 linger_alloc(struct ceph_osd_client *osdc)
2753 {
2754 	struct ceph_osd_linger_request *lreq;
2755 
2756 	lreq = kzalloc(sizeof(*lreq), GFP_NOIO);
2757 	if (!lreq)
2758 		return NULL;
2759 
2760 	kref_init(&lreq->kref);
2761 	mutex_init(&lreq->lock);
2762 	RB_CLEAR_NODE(&lreq->node);
2763 	RB_CLEAR_NODE(&lreq->osdc_node);
2764 	RB_CLEAR_NODE(&lreq->mc_node);
2765 	INIT_LIST_HEAD(&lreq->scan_item);
2766 	INIT_LIST_HEAD(&lreq->pending_lworks);
2767 	init_completion(&lreq->reg_commit_wait);
2768 	init_completion(&lreq->notify_finish_wait);
2769 
2770 	lreq->osdc = osdc;
2771 	target_init(&lreq->t);
2772 
2773 	dout("%s lreq %p\n", __func__, lreq);
2774 	return lreq;
2775 }
2776 
DEFINE_RB_INSDEL_FUNCS(linger,struct ceph_osd_linger_request,linger_id,node)2777 DEFINE_RB_INSDEL_FUNCS(linger, struct ceph_osd_linger_request, linger_id, node)
2778 DEFINE_RB_FUNCS(linger_osdc, struct ceph_osd_linger_request, linger_id, osdc_node)
2779 DEFINE_RB_FUNCS(linger_mc, struct ceph_osd_linger_request, linger_id, mc_node)
2780 
2781 /*
2782  * Create linger request <-> OSD session relation.
2783  *
2784  * @lreq has to be registered, @osd may be homeless.
2785  */
2786 static void link_linger(struct ceph_osd *osd,
2787 			struct ceph_osd_linger_request *lreq)
2788 {
2789 	verify_osd_locked(osd);
2790 	WARN_ON(!lreq->linger_id || lreq->osd);
2791 	dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2792 	     osd->o_osd, lreq, lreq->linger_id);
2793 
2794 	if (!osd_homeless(osd))
2795 		__remove_osd_from_lru(osd);
2796 	else
2797 		atomic_inc(&osd->o_osdc->num_homeless);
2798 
2799 	get_osd(osd);
2800 	insert_linger(&osd->o_linger_requests, lreq);
2801 	lreq->osd = osd;
2802 }
2803 
unlink_linger(struct ceph_osd * osd,struct ceph_osd_linger_request * lreq)2804 static void unlink_linger(struct ceph_osd *osd,
2805 			  struct ceph_osd_linger_request *lreq)
2806 {
2807 	verify_osd_locked(osd);
2808 	WARN_ON(lreq->osd != osd);
2809 	dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2810 	     osd->o_osd, lreq, lreq->linger_id);
2811 
2812 	lreq->osd = NULL;
2813 	erase_linger(&osd->o_linger_requests, lreq);
2814 	put_osd(osd);
2815 
2816 	if (!osd_homeless(osd))
2817 		maybe_move_osd_to_lru(osd);
2818 	else
2819 		atomic_dec(&osd->o_osdc->num_homeless);
2820 }
2821 
__linger_registered(struct ceph_osd_linger_request * lreq)2822 static bool __linger_registered(struct ceph_osd_linger_request *lreq)
2823 {
2824 	verify_osdc_locked(lreq->osdc);
2825 
2826 	return !RB_EMPTY_NODE(&lreq->osdc_node);
2827 }
2828 
linger_registered(struct ceph_osd_linger_request * lreq)2829 static bool linger_registered(struct ceph_osd_linger_request *lreq)
2830 {
2831 	struct ceph_osd_client *osdc = lreq->osdc;
2832 	bool registered;
2833 
2834 	down_read(&osdc->lock);
2835 	registered = __linger_registered(lreq);
2836 	up_read(&osdc->lock);
2837 
2838 	return registered;
2839 }
2840 
linger_register(struct ceph_osd_linger_request * lreq)2841 static void linger_register(struct ceph_osd_linger_request *lreq)
2842 {
2843 	struct ceph_osd_client *osdc = lreq->osdc;
2844 
2845 	verify_osdc_wrlocked(osdc);
2846 	WARN_ON(lreq->linger_id);
2847 
2848 	linger_get(lreq);
2849 	lreq->linger_id = ++osdc->last_linger_id;
2850 	insert_linger_osdc(&osdc->linger_requests, lreq);
2851 }
2852 
linger_unregister(struct ceph_osd_linger_request * lreq)2853 static void linger_unregister(struct ceph_osd_linger_request *lreq)
2854 {
2855 	struct ceph_osd_client *osdc = lreq->osdc;
2856 
2857 	verify_osdc_wrlocked(osdc);
2858 
2859 	erase_linger_osdc(&osdc->linger_requests, lreq);
2860 	linger_put(lreq);
2861 }
2862 
cancel_linger_request(struct ceph_osd_request * req)2863 static void cancel_linger_request(struct ceph_osd_request *req)
2864 {
2865 	struct ceph_osd_linger_request *lreq = req->r_priv;
2866 
2867 	WARN_ON(!req->r_linger);
2868 	cancel_request(req);
2869 	linger_put(lreq);
2870 }
2871 
2872 struct linger_work {
2873 	struct work_struct work;
2874 	struct ceph_osd_linger_request *lreq;
2875 	struct list_head pending_item;
2876 	unsigned long queued_stamp;
2877 
2878 	union {
2879 		struct {
2880 			u64 notify_id;
2881 			u64 notifier_id;
2882 			void *payload; /* points into @msg front */
2883 			size_t payload_len;
2884 
2885 			struct ceph_msg *msg; /* for ceph_msg_put() */
2886 		} notify;
2887 		struct {
2888 			int err;
2889 		} error;
2890 	};
2891 };
2892 
lwork_alloc(struct ceph_osd_linger_request * lreq,work_func_t workfn)2893 static struct linger_work *lwork_alloc(struct ceph_osd_linger_request *lreq,
2894 				       work_func_t workfn)
2895 {
2896 	struct linger_work *lwork;
2897 
2898 	lwork = kzalloc(sizeof(*lwork), GFP_NOIO);
2899 	if (!lwork)
2900 		return NULL;
2901 
2902 	INIT_WORK(&lwork->work, workfn);
2903 	INIT_LIST_HEAD(&lwork->pending_item);
2904 	lwork->lreq = linger_get(lreq);
2905 
2906 	return lwork;
2907 }
2908 
lwork_free(struct linger_work * lwork)2909 static void lwork_free(struct linger_work *lwork)
2910 {
2911 	struct ceph_osd_linger_request *lreq = lwork->lreq;
2912 
2913 	mutex_lock(&lreq->lock);
2914 	list_del(&lwork->pending_item);
2915 	mutex_unlock(&lreq->lock);
2916 
2917 	linger_put(lreq);
2918 	kfree(lwork);
2919 }
2920 
lwork_queue(struct linger_work * lwork)2921 static void lwork_queue(struct linger_work *lwork)
2922 {
2923 	struct ceph_osd_linger_request *lreq = lwork->lreq;
2924 	struct ceph_osd_client *osdc = lreq->osdc;
2925 
2926 	verify_lreq_locked(lreq);
2927 	WARN_ON(!list_empty(&lwork->pending_item));
2928 
2929 	lwork->queued_stamp = jiffies;
2930 	list_add_tail(&lwork->pending_item, &lreq->pending_lworks);
2931 	queue_work(osdc->notify_wq, &lwork->work);
2932 }
2933 
do_watch_notify(struct work_struct * w)2934 static void do_watch_notify(struct work_struct *w)
2935 {
2936 	struct linger_work *lwork = container_of(w, struct linger_work, work);
2937 	struct ceph_osd_linger_request *lreq = lwork->lreq;
2938 
2939 	if (!linger_registered(lreq)) {
2940 		dout("%s lreq %p not registered\n", __func__, lreq);
2941 		goto out;
2942 	}
2943 
2944 	WARN_ON(!lreq->is_watch);
2945 	dout("%s lreq %p notify_id %llu notifier_id %llu payload_len %zu\n",
2946 	     __func__, lreq, lwork->notify.notify_id, lwork->notify.notifier_id,
2947 	     lwork->notify.payload_len);
2948 	lreq->wcb(lreq->data, lwork->notify.notify_id, lreq->linger_id,
2949 		  lwork->notify.notifier_id, lwork->notify.payload,
2950 		  lwork->notify.payload_len);
2951 
2952 out:
2953 	ceph_msg_put(lwork->notify.msg);
2954 	lwork_free(lwork);
2955 }
2956 
do_watch_error(struct work_struct * w)2957 static void do_watch_error(struct work_struct *w)
2958 {
2959 	struct linger_work *lwork = container_of(w, struct linger_work, work);
2960 	struct ceph_osd_linger_request *lreq = lwork->lreq;
2961 
2962 	if (!linger_registered(lreq)) {
2963 		dout("%s lreq %p not registered\n", __func__, lreq);
2964 		goto out;
2965 	}
2966 
2967 	dout("%s lreq %p err %d\n", __func__, lreq, lwork->error.err);
2968 	lreq->errcb(lreq->data, lreq->linger_id, lwork->error.err);
2969 
2970 out:
2971 	lwork_free(lwork);
2972 }
2973 
queue_watch_error(struct ceph_osd_linger_request * lreq)2974 static void queue_watch_error(struct ceph_osd_linger_request *lreq)
2975 {
2976 	struct linger_work *lwork;
2977 
2978 	lwork = lwork_alloc(lreq, do_watch_error);
2979 	if (!lwork) {
2980 		pr_err("failed to allocate error-lwork\n");
2981 		return;
2982 	}
2983 
2984 	lwork->error.err = lreq->last_error;
2985 	lwork_queue(lwork);
2986 }
2987 
linger_reg_commit_complete(struct ceph_osd_linger_request * lreq,int result)2988 static void linger_reg_commit_complete(struct ceph_osd_linger_request *lreq,
2989 				       int result)
2990 {
2991 	if (!completion_done(&lreq->reg_commit_wait)) {
2992 		lreq->reg_commit_error = (result <= 0 ? result : 0);
2993 		complete_all(&lreq->reg_commit_wait);
2994 	}
2995 }
2996 
linger_commit_cb(struct ceph_osd_request * req)2997 static void linger_commit_cb(struct ceph_osd_request *req)
2998 {
2999 	struct ceph_osd_linger_request *lreq = req->r_priv;
3000 
3001 	mutex_lock(&lreq->lock);
3002 	dout("%s lreq %p linger_id %llu result %d\n", __func__, lreq,
3003 	     lreq->linger_id, req->r_result);
3004 	linger_reg_commit_complete(lreq, req->r_result);
3005 	lreq->committed = true;
3006 
3007 	if (!lreq->is_watch) {
3008 		struct ceph_osd_data *osd_data =
3009 		    osd_req_op_data(req, 0, notify, response_data);
3010 		void *p = page_address(osd_data->pages[0]);
3011 
3012 		WARN_ON(req->r_ops[0].op != CEPH_OSD_OP_NOTIFY ||
3013 			osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
3014 
3015 		/* make note of the notify_id */
3016 		if (req->r_ops[0].outdata_len >= sizeof(u64)) {
3017 			lreq->notify_id = ceph_decode_64(&p);
3018 			dout("lreq %p notify_id %llu\n", lreq,
3019 			     lreq->notify_id);
3020 		} else {
3021 			dout("lreq %p no notify_id\n", lreq);
3022 		}
3023 	}
3024 
3025 	mutex_unlock(&lreq->lock);
3026 	linger_put(lreq);
3027 }
3028 
normalize_watch_error(int err)3029 static int normalize_watch_error(int err)
3030 {
3031 	/*
3032 	 * Translate ENOENT -> ENOTCONN so that a delete->disconnection
3033 	 * notification and a failure to reconnect because we raced with
3034 	 * the delete appear the same to the user.
3035 	 */
3036 	if (err == -ENOENT)
3037 		err = -ENOTCONN;
3038 
3039 	return err;
3040 }
3041 
linger_reconnect_cb(struct ceph_osd_request * req)3042 static void linger_reconnect_cb(struct ceph_osd_request *req)
3043 {
3044 	struct ceph_osd_linger_request *lreq = req->r_priv;
3045 
3046 	mutex_lock(&lreq->lock);
3047 	dout("%s lreq %p linger_id %llu result %d last_error %d\n", __func__,
3048 	     lreq, lreq->linger_id, req->r_result, lreq->last_error);
3049 	if (req->r_result < 0) {
3050 		if (!lreq->last_error) {
3051 			lreq->last_error = normalize_watch_error(req->r_result);
3052 			queue_watch_error(lreq);
3053 		}
3054 	}
3055 
3056 	mutex_unlock(&lreq->lock);
3057 	linger_put(lreq);
3058 }
3059 
send_linger(struct ceph_osd_linger_request * lreq)3060 static void send_linger(struct ceph_osd_linger_request *lreq)
3061 {
3062 	struct ceph_osd_request *req = lreq->reg_req;
3063 	struct ceph_osd_req_op *op = &req->r_ops[0];
3064 
3065 	verify_osdc_wrlocked(req->r_osdc);
3066 	dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3067 
3068 	if (req->r_osd)
3069 		cancel_linger_request(req);
3070 
3071 	request_reinit(req);
3072 	target_copy(&req->r_t, &lreq->t);
3073 	req->r_mtime = lreq->mtime;
3074 
3075 	mutex_lock(&lreq->lock);
3076 	if (lreq->is_watch && lreq->committed) {
3077 		WARN_ON(op->op != CEPH_OSD_OP_WATCH ||
3078 			op->watch.cookie != lreq->linger_id);
3079 		op->watch.op = CEPH_OSD_WATCH_OP_RECONNECT;
3080 		op->watch.gen = ++lreq->register_gen;
3081 		dout("lreq %p reconnect register_gen %u\n", lreq,
3082 		     op->watch.gen);
3083 		req->r_callback = linger_reconnect_cb;
3084 	} else {
3085 		if (!lreq->is_watch)
3086 			lreq->notify_id = 0;
3087 		else
3088 			WARN_ON(op->watch.op != CEPH_OSD_WATCH_OP_WATCH);
3089 		dout("lreq %p register\n", lreq);
3090 		req->r_callback = linger_commit_cb;
3091 	}
3092 	mutex_unlock(&lreq->lock);
3093 
3094 	req->r_priv = linger_get(lreq);
3095 	req->r_linger = true;
3096 
3097 	submit_request(req, true);
3098 }
3099 
linger_ping_cb(struct ceph_osd_request * req)3100 static void linger_ping_cb(struct ceph_osd_request *req)
3101 {
3102 	struct ceph_osd_linger_request *lreq = req->r_priv;
3103 
3104 	mutex_lock(&lreq->lock);
3105 	dout("%s lreq %p linger_id %llu result %d ping_sent %lu last_error %d\n",
3106 	     __func__, lreq, lreq->linger_id, req->r_result, lreq->ping_sent,
3107 	     lreq->last_error);
3108 	if (lreq->register_gen == req->r_ops[0].watch.gen) {
3109 		if (!req->r_result) {
3110 			lreq->watch_valid_thru = lreq->ping_sent;
3111 		} else if (!lreq->last_error) {
3112 			lreq->last_error = normalize_watch_error(req->r_result);
3113 			queue_watch_error(lreq);
3114 		}
3115 	} else {
3116 		dout("lreq %p register_gen %u ignoring old pong %u\n", lreq,
3117 		     lreq->register_gen, req->r_ops[0].watch.gen);
3118 	}
3119 
3120 	mutex_unlock(&lreq->lock);
3121 	linger_put(lreq);
3122 }
3123 
send_linger_ping(struct ceph_osd_linger_request * lreq)3124 static void send_linger_ping(struct ceph_osd_linger_request *lreq)
3125 {
3126 	struct ceph_osd_client *osdc = lreq->osdc;
3127 	struct ceph_osd_request *req = lreq->ping_req;
3128 	struct ceph_osd_req_op *op = &req->r_ops[0];
3129 
3130 	if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
3131 		dout("%s PAUSERD\n", __func__);
3132 		return;
3133 	}
3134 
3135 	lreq->ping_sent = jiffies;
3136 	dout("%s lreq %p linger_id %llu ping_sent %lu register_gen %u\n",
3137 	     __func__, lreq, lreq->linger_id, lreq->ping_sent,
3138 	     lreq->register_gen);
3139 
3140 	if (req->r_osd)
3141 		cancel_linger_request(req);
3142 
3143 	request_reinit(req);
3144 	target_copy(&req->r_t, &lreq->t);
3145 
3146 	WARN_ON(op->op != CEPH_OSD_OP_WATCH ||
3147 		op->watch.cookie != lreq->linger_id ||
3148 		op->watch.op != CEPH_OSD_WATCH_OP_PING);
3149 	op->watch.gen = lreq->register_gen;
3150 	req->r_callback = linger_ping_cb;
3151 	req->r_priv = linger_get(lreq);
3152 	req->r_linger = true;
3153 
3154 	ceph_osdc_get_request(req);
3155 	account_request(req);
3156 	req->r_tid = atomic64_inc_return(&osdc->last_tid);
3157 	link_request(lreq->osd, req);
3158 	send_request(req);
3159 }
3160 
linger_submit(struct ceph_osd_linger_request * lreq)3161 static void linger_submit(struct ceph_osd_linger_request *lreq)
3162 {
3163 	struct ceph_osd_client *osdc = lreq->osdc;
3164 	struct ceph_osd *osd;
3165 
3166 	down_write(&osdc->lock);
3167 	linger_register(lreq);
3168 	if (lreq->is_watch) {
3169 		lreq->reg_req->r_ops[0].watch.cookie = lreq->linger_id;
3170 		lreq->ping_req->r_ops[0].watch.cookie = lreq->linger_id;
3171 	} else {
3172 		lreq->reg_req->r_ops[0].notify.cookie = lreq->linger_id;
3173 	}
3174 
3175 	calc_target(osdc, &lreq->t, false);
3176 	osd = lookup_create_osd(osdc, lreq->t.osd, true);
3177 	link_linger(osd, lreq);
3178 
3179 	send_linger(lreq);
3180 	up_write(&osdc->lock);
3181 }
3182 
cancel_linger_map_check(struct ceph_osd_linger_request * lreq)3183 static void cancel_linger_map_check(struct ceph_osd_linger_request *lreq)
3184 {
3185 	struct ceph_osd_client *osdc = lreq->osdc;
3186 	struct ceph_osd_linger_request *lookup_lreq;
3187 
3188 	verify_osdc_wrlocked(osdc);
3189 
3190 	lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3191 				       lreq->linger_id);
3192 	if (!lookup_lreq)
3193 		return;
3194 
3195 	WARN_ON(lookup_lreq != lreq);
3196 	erase_linger_mc(&osdc->linger_map_checks, lreq);
3197 	linger_put(lreq);
3198 }
3199 
3200 /*
3201  * @lreq has to be both registered and linked.
3202  */
__linger_cancel(struct ceph_osd_linger_request * lreq)3203 static void __linger_cancel(struct ceph_osd_linger_request *lreq)
3204 {
3205 	if (lreq->is_watch && lreq->ping_req->r_osd)
3206 		cancel_linger_request(lreq->ping_req);
3207 	if (lreq->reg_req->r_osd)
3208 		cancel_linger_request(lreq->reg_req);
3209 	cancel_linger_map_check(lreq);
3210 	unlink_linger(lreq->osd, lreq);
3211 	linger_unregister(lreq);
3212 }
3213 
linger_cancel(struct ceph_osd_linger_request * lreq)3214 static void linger_cancel(struct ceph_osd_linger_request *lreq)
3215 {
3216 	struct ceph_osd_client *osdc = lreq->osdc;
3217 
3218 	down_write(&osdc->lock);
3219 	if (__linger_registered(lreq))
3220 		__linger_cancel(lreq);
3221 	up_write(&osdc->lock);
3222 }
3223 
3224 static void send_linger_map_check(struct ceph_osd_linger_request *lreq);
3225 
check_linger_pool_dne(struct ceph_osd_linger_request * lreq)3226 static void check_linger_pool_dne(struct ceph_osd_linger_request *lreq)
3227 {
3228 	struct ceph_osd_client *osdc = lreq->osdc;
3229 	struct ceph_osdmap *map = osdc->osdmap;
3230 
3231 	verify_osdc_wrlocked(osdc);
3232 	WARN_ON(!map->epoch);
3233 
3234 	if (lreq->register_gen) {
3235 		lreq->map_dne_bound = map->epoch;
3236 		dout("%s lreq %p linger_id %llu pool disappeared\n", __func__,
3237 		     lreq, lreq->linger_id);
3238 	} else {
3239 		dout("%s lreq %p linger_id %llu map_dne_bound %u have %u\n",
3240 		     __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3241 		     map->epoch);
3242 	}
3243 
3244 	if (lreq->map_dne_bound) {
3245 		if (map->epoch >= lreq->map_dne_bound) {
3246 			/* we had a new enough map */
3247 			pr_info("linger_id %llu pool does not exist\n",
3248 				lreq->linger_id);
3249 			linger_reg_commit_complete(lreq, -ENOENT);
3250 			__linger_cancel(lreq);
3251 		}
3252 	} else {
3253 		send_linger_map_check(lreq);
3254 	}
3255 }
3256 
linger_map_check_cb(struct ceph_mon_generic_request * greq)3257 static void linger_map_check_cb(struct ceph_mon_generic_request *greq)
3258 {
3259 	struct ceph_osd_client *osdc = &greq->monc->client->osdc;
3260 	struct ceph_osd_linger_request *lreq;
3261 	u64 linger_id = greq->private_data;
3262 
3263 	WARN_ON(greq->result || !greq->u.newest);
3264 
3265 	down_write(&osdc->lock);
3266 	lreq = lookup_linger_mc(&osdc->linger_map_checks, linger_id);
3267 	if (!lreq) {
3268 		dout("%s linger_id %llu dne\n", __func__, linger_id);
3269 		goto out_unlock;
3270 	}
3271 
3272 	dout("%s lreq %p linger_id %llu map_dne_bound %u newest %llu\n",
3273 	     __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3274 	     greq->u.newest);
3275 	if (!lreq->map_dne_bound)
3276 		lreq->map_dne_bound = greq->u.newest;
3277 	erase_linger_mc(&osdc->linger_map_checks, lreq);
3278 	check_linger_pool_dne(lreq);
3279 
3280 	linger_put(lreq);
3281 out_unlock:
3282 	up_write(&osdc->lock);
3283 }
3284 
send_linger_map_check(struct ceph_osd_linger_request * lreq)3285 static void send_linger_map_check(struct ceph_osd_linger_request *lreq)
3286 {
3287 	struct ceph_osd_client *osdc = lreq->osdc;
3288 	struct ceph_osd_linger_request *lookup_lreq;
3289 	int ret;
3290 
3291 	verify_osdc_wrlocked(osdc);
3292 
3293 	lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3294 				       lreq->linger_id);
3295 	if (lookup_lreq) {
3296 		WARN_ON(lookup_lreq != lreq);
3297 		return;
3298 	}
3299 
3300 	linger_get(lreq);
3301 	insert_linger_mc(&osdc->linger_map_checks, lreq);
3302 	ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
3303 					  linger_map_check_cb, lreq->linger_id);
3304 	WARN_ON(ret);
3305 }
3306 
linger_reg_commit_wait(struct ceph_osd_linger_request * lreq)3307 static int linger_reg_commit_wait(struct ceph_osd_linger_request *lreq)
3308 {
3309 	int ret;
3310 
3311 	dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3312 	ret = wait_for_completion_interruptible(&lreq->reg_commit_wait);
3313 	return ret ?: lreq->reg_commit_error;
3314 }
3315 
linger_notify_finish_wait(struct ceph_osd_linger_request * lreq)3316 static int linger_notify_finish_wait(struct ceph_osd_linger_request *lreq)
3317 {
3318 	int ret;
3319 
3320 	dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3321 	ret = wait_for_completion_interruptible(&lreq->notify_finish_wait);
3322 	return ret ?: lreq->notify_finish_error;
3323 }
3324 
3325 /*
3326  * Timeout callback, called every N seconds.  When 1 or more OSD
3327  * requests has been active for more than N seconds, we send a keepalive
3328  * (tag + timestamp) to its OSD to ensure any communications channel
3329  * reset is detected.
3330  */
handle_timeout(struct work_struct * work)3331 static void handle_timeout(struct work_struct *work)
3332 {
3333 	struct ceph_osd_client *osdc =
3334 		container_of(work, struct ceph_osd_client, timeout_work.work);
3335 	struct ceph_options *opts = osdc->client->options;
3336 	unsigned long cutoff = jiffies - opts->osd_keepalive_timeout;
3337 	unsigned long expiry_cutoff = jiffies - opts->osd_request_timeout;
3338 	LIST_HEAD(slow_osds);
3339 	struct rb_node *n, *p;
3340 
3341 	dout("%s osdc %p\n", __func__, osdc);
3342 	down_write(&osdc->lock);
3343 
3344 	/*
3345 	 * ping osds that are a bit slow.  this ensures that if there
3346 	 * is a break in the TCP connection we will notice, and reopen
3347 	 * a connection with that osd (from the fault callback).
3348 	 */
3349 	for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
3350 		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
3351 		bool found = false;
3352 
3353 		for (p = rb_first(&osd->o_requests); p; ) {
3354 			struct ceph_osd_request *req =
3355 			    rb_entry(p, struct ceph_osd_request, r_node);
3356 
3357 			p = rb_next(p); /* abort_request() */
3358 
3359 			if (time_before(req->r_stamp, cutoff)) {
3360 				dout(" req %p tid %llu on osd%d is laggy\n",
3361 				     req, req->r_tid, osd->o_osd);
3362 				found = true;
3363 			}
3364 			if (opts->osd_request_timeout &&
3365 			    time_before(req->r_start_stamp, expiry_cutoff)) {
3366 				pr_err_ratelimited("tid %llu on osd%d timeout\n",
3367 				       req->r_tid, osd->o_osd);
3368 				abort_request(req, -ETIMEDOUT);
3369 			}
3370 		}
3371 		for (p = rb_first(&osd->o_linger_requests); p; p = rb_next(p)) {
3372 			struct ceph_osd_linger_request *lreq =
3373 			    rb_entry(p, struct ceph_osd_linger_request, node);
3374 
3375 			dout(" lreq %p linger_id %llu is served by osd%d\n",
3376 			     lreq, lreq->linger_id, osd->o_osd);
3377 			found = true;
3378 
3379 			mutex_lock(&lreq->lock);
3380 			if (lreq->is_watch && lreq->committed && !lreq->last_error)
3381 				send_linger_ping(lreq);
3382 			mutex_unlock(&lreq->lock);
3383 		}
3384 
3385 		if (found)
3386 			list_move_tail(&osd->o_keepalive_item, &slow_osds);
3387 	}
3388 
3389 	if (opts->osd_request_timeout) {
3390 		for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
3391 			struct ceph_osd_request *req =
3392 			    rb_entry(p, struct ceph_osd_request, r_node);
3393 
3394 			p = rb_next(p); /* abort_request() */
3395 
3396 			if (time_before(req->r_start_stamp, expiry_cutoff)) {
3397 				pr_err_ratelimited("tid %llu on osd%d timeout\n",
3398 				       req->r_tid, osdc->homeless_osd.o_osd);
3399 				abort_request(req, -ETIMEDOUT);
3400 			}
3401 		}
3402 	}
3403 
3404 	if (atomic_read(&osdc->num_homeless) || !list_empty(&slow_osds))
3405 		maybe_request_map(osdc);
3406 
3407 	while (!list_empty(&slow_osds)) {
3408 		struct ceph_osd *osd = list_first_entry(&slow_osds,
3409 							struct ceph_osd,
3410 							o_keepalive_item);
3411 		list_del_init(&osd->o_keepalive_item);
3412 		ceph_con_keepalive(&osd->o_con);
3413 	}
3414 
3415 	up_write(&osdc->lock);
3416 	schedule_delayed_work(&osdc->timeout_work,
3417 			      osdc->client->options->osd_keepalive_timeout);
3418 }
3419 
handle_osds_timeout(struct work_struct * work)3420 static void handle_osds_timeout(struct work_struct *work)
3421 {
3422 	struct ceph_osd_client *osdc =
3423 		container_of(work, struct ceph_osd_client,
3424 			     osds_timeout_work.work);
3425 	unsigned long delay = osdc->client->options->osd_idle_ttl / 4;
3426 	struct ceph_osd *osd, *nosd;
3427 
3428 	dout("%s osdc %p\n", __func__, osdc);
3429 	down_write(&osdc->lock);
3430 	list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) {
3431 		if (time_before(jiffies, osd->lru_ttl))
3432 			break;
3433 
3434 		WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
3435 		WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
3436 		close_osd(osd);
3437 	}
3438 
3439 	up_write(&osdc->lock);
3440 	schedule_delayed_work(&osdc->osds_timeout_work,
3441 			      round_jiffies_relative(delay));
3442 }
3443 
ceph_oloc_decode(void ** p,void * end,struct ceph_object_locator * oloc)3444 static int ceph_oloc_decode(void **p, void *end,
3445 			    struct ceph_object_locator *oloc)
3446 {
3447 	u8 struct_v, struct_cv;
3448 	u32 len;
3449 	void *struct_end;
3450 	int ret = 0;
3451 
3452 	ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3453 	struct_v = ceph_decode_8(p);
3454 	struct_cv = ceph_decode_8(p);
3455 	if (struct_v < 3) {
3456 		pr_warn("got v %d < 3 cv %d of ceph_object_locator\n",
3457 			struct_v, struct_cv);
3458 		goto e_inval;
3459 	}
3460 	if (struct_cv > 6) {
3461 		pr_warn("got v %d cv %d > 6 of ceph_object_locator\n",
3462 			struct_v, struct_cv);
3463 		goto e_inval;
3464 	}
3465 	len = ceph_decode_32(p);
3466 	ceph_decode_need(p, end, len, e_inval);
3467 	struct_end = *p + len;
3468 
3469 	oloc->pool = ceph_decode_64(p);
3470 	*p += 4; /* skip preferred */
3471 
3472 	len = ceph_decode_32(p);
3473 	if (len > 0) {
3474 		pr_warn("ceph_object_locator::key is set\n");
3475 		goto e_inval;
3476 	}
3477 
3478 	if (struct_v >= 5) {
3479 		bool changed = false;
3480 
3481 		len = ceph_decode_32(p);
3482 		if (len > 0) {
3483 			ceph_decode_need(p, end, len, e_inval);
3484 			if (!oloc->pool_ns ||
3485 			    ceph_compare_string(oloc->pool_ns, *p, len))
3486 				changed = true;
3487 			*p += len;
3488 		} else {
3489 			if (oloc->pool_ns)
3490 				changed = true;
3491 		}
3492 		if (changed) {
3493 			/* redirect changes namespace */
3494 			pr_warn("ceph_object_locator::nspace is changed\n");
3495 			goto e_inval;
3496 		}
3497 	}
3498 
3499 	if (struct_v >= 6) {
3500 		s64 hash = ceph_decode_64(p);
3501 		if (hash != -1) {
3502 			pr_warn("ceph_object_locator::hash is set\n");
3503 			goto e_inval;
3504 		}
3505 	}
3506 
3507 	/* skip the rest */
3508 	*p = struct_end;
3509 out:
3510 	return ret;
3511 
3512 e_inval:
3513 	ret = -EINVAL;
3514 	goto out;
3515 }
3516 
ceph_redirect_decode(void ** p,void * end,struct ceph_request_redirect * redir)3517 static int ceph_redirect_decode(void **p, void *end,
3518 				struct ceph_request_redirect *redir)
3519 {
3520 	u8 struct_v, struct_cv;
3521 	u32 len;
3522 	void *struct_end;
3523 	int ret;
3524 
3525 	ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3526 	struct_v = ceph_decode_8(p);
3527 	struct_cv = ceph_decode_8(p);
3528 	if (struct_cv > 1) {
3529 		pr_warn("got v %d cv %d > 1 of ceph_request_redirect\n",
3530 			struct_v, struct_cv);
3531 		goto e_inval;
3532 	}
3533 	len = ceph_decode_32(p);
3534 	ceph_decode_need(p, end, len, e_inval);
3535 	struct_end = *p + len;
3536 
3537 	ret = ceph_oloc_decode(p, end, &redir->oloc);
3538 	if (ret)
3539 		goto out;
3540 
3541 	len = ceph_decode_32(p);
3542 	if (len > 0) {
3543 		pr_warn("ceph_request_redirect::object_name is set\n");
3544 		goto e_inval;
3545 	}
3546 
3547 	/* skip the rest */
3548 	*p = struct_end;
3549 out:
3550 	return ret;
3551 
3552 e_inval:
3553 	ret = -EINVAL;
3554 	goto out;
3555 }
3556 
3557 struct MOSDOpReply {
3558 	struct ceph_pg pgid;
3559 	u64 flags;
3560 	int result;
3561 	u32 epoch;
3562 	int num_ops;
3563 	u32 outdata_len[CEPH_OSD_MAX_OPS];
3564 	s32 rval[CEPH_OSD_MAX_OPS];
3565 	int retry_attempt;
3566 	struct ceph_eversion replay_version;
3567 	u64 user_version;
3568 	struct ceph_request_redirect redirect;
3569 };
3570 
decode_MOSDOpReply(const struct ceph_msg * msg,struct MOSDOpReply * m)3571 static int decode_MOSDOpReply(const struct ceph_msg *msg, struct MOSDOpReply *m)
3572 {
3573 	void *p = msg->front.iov_base;
3574 	void *const end = p + msg->front.iov_len;
3575 	u16 version = le16_to_cpu(msg->hdr.version);
3576 	struct ceph_eversion bad_replay_version;
3577 	u8 decode_redir;
3578 	u32 len;
3579 	int ret;
3580 	int i;
3581 
3582 	ceph_decode_32_safe(&p, end, len, e_inval);
3583 	ceph_decode_need(&p, end, len, e_inval);
3584 	p += len; /* skip oid */
3585 
3586 	ret = ceph_decode_pgid(&p, end, &m->pgid);
3587 	if (ret)
3588 		return ret;
3589 
3590 	ceph_decode_64_safe(&p, end, m->flags, e_inval);
3591 	ceph_decode_32_safe(&p, end, m->result, e_inval);
3592 	ceph_decode_need(&p, end, sizeof(bad_replay_version), e_inval);
3593 	memcpy(&bad_replay_version, p, sizeof(bad_replay_version));
3594 	p += sizeof(bad_replay_version);
3595 	ceph_decode_32_safe(&p, end, m->epoch, e_inval);
3596 
3597 	ceph_decode_32_safe(&p, end, m->num_ops, e_inval);
3598 	if (m->num_ops > ARRAY_SIZE(m->outdata_len))
3599 		goto e_inval;
3600 
3601 	ceph_decode_need(&p, end, m->num_ops * sizeof(struct ceph_osd_op),
3602 			 e_inval);
3603 	for (i = 0; i < m->num_ops; i++) {
3604 		struct ceph_osd_op *op = p;
3605 
3606 		m->outdata_len[i] = le32_to_cpu(op->payload_len);
3607 		p += sizeof(*op);
3608 	}
3609 
3610 	ceph_decode_32_safe(&p, end, m->retry_attempt, e_inval);
3611 	for (i = 0; i < m->num_ops; i++)
3612 		ceph_decode_32_safe(&p, end, m->rval[i], e_inval);
3613 
3614 	if (version >= 5) {
3615 		ceph_decode_need(&p, end, sizeof(m->replay_version), e_inval);
3616 		memcpy(&m->replay_version, p, sizeof(m->replay_version));
3617 		p += sizeof(m->replay_version);
3618 		ceph_decode_64_safe(&p, end, m->user_version, e_inval);
3619 	} else {
3620 		m->replay_version = bad_replay_version; /* struct */
3621 		m->user_version = le64_to_cpu(m->replay_version.version);
3622 	}
3623 
3624 	if (version >= 6) {
3625 		if (version >= 7)
3626 			ceph_decode_8_safe(&p, end, decode_redir, e_inval);
3627 		else
3628 			decode_redir = 1;
3629 	} else {
3630 		decode_redir = 0;
3631 	}
3632 
3633 	if (decode_redir) {
3634 		ret = ceph_redirect_decode(&p, end, &m->redirect);
3635 		if (ret)
3636 			return ret;
3637 	} else {
3638 		ceph_oloc_init(&m->redirect.oloc);
3639 	}
3640 
3641 	return 0;
3642 
3643 e_inval:
3644 	return -EINVAL;
3645 }
3646 
3647 /*
3648  * Handle MOSDOpReply.  Set ->r_result and call the callback if it is
3649  * specified.
3650  */
handle_reply(struct ceph_osd * osd,struct ceph_msg * msg)3651 static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg)
3652 {
3653 	struct ceph_osd_client *osdc = osd->o_osdc;
3654 	struct ceph_osd_request *req;
3655 	struct MOSDOpReply m;
3656 	u64 tid = le64_to_cpu(msg->hdr.tid);
3657 	u32 data_len = 0;
3658 	int ret;
3659 	int i;
3660 
3661 	dout("%s msg %p tid %llu\n", __func__, msg, tid);
3662 
3663 	down_read(&osdc->lock);
3664 	if (!osd_registered(osd)) {
3665 		dout("%s osd%d unknown\n", __func__, osd->o_osd);
3666 		goto out_unlock_osdc;
3667 	}
3668 	WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
3669 
3670 	mutex_lock(&osd->lock);
3671 	req = lookup_request(&osd->o_requests, tid);
3672 	if (!req) {
3673 		dout("%s osd%d tid %llu unknown\n", __func__, osd->o_osd, tid);
3674 		goto out_unlock_session;
3675 	}
3676 
3677 	m.redirect.oloc.pool_ns = req->r_t.target_oloc.pool_ns;
3678 	ret = decode_MOSDOpReply(msg, &m);
3679 	m.redirect.oloc.pool_ns = NULL;
3680 	if (ret) {
3681 		pr_err("failed to decode MOSDOpReply for tid %llu: %d\n",
3682 		       req->r_tid, ret);
3683 		ceph_msg_dump(msg);
3684 		goto fail_request;
3685 	}
3686 	dout("%s req %p tid %llu flags 0x%llx pgid %llu.%x epoch %u attempt %d v %u'%llu uv %llu\n",
3687 	     __func__, req, req->r_tid, m.flags, m.pgid.pool, m.pgid.seed,
3688 	     m.epoch, m.retry_attempt, le32_to_cpu(m.replay_version.epoch),
3689 	     le64_to_cpu(m.replay_version.version), m.user_version);
3690 
3691 	if (m.retry_attempt >= 0) {
3692 		if (m.retry_attempt != req->r_attempts - 1) {
3693 			dout("req %p tid %llu retry_attempt %d != %d, ignoring\n",
3694 			     req, req->r_tid, m.retry_attempt,
3695 			     req->r_attempts - 1);
3696 			goto out_unlock_session;
3697 		}
3698 	} else {
3699 		WARN_ON(1); /* MOSDOpReply v4 is assumed */
3700 	}
3701 
3702 	if (!ceph_oloc_empty(&m.redirect.oloc)) {
3703 		dout("req %p tid %llu redirect pool %lld\n", req, req->r_tid,
3704 		     m.redirect.oloc.pool);
3705 		unlink_request(osd, req);
3706 		mutex_unlock(&osd->lock);
3707 
3708 		/*
3709 		 * Not ceph_oloc_copy() - changing pool_ns is not
3710 		 * supported.
3711 		 */
3712 		req->r_t.target_oloc.pool = m.redirect.oloc.pool;
3713 		req->r_flags |= CEPH_OSD_FLAG_REDIRECTED |
3714 				CEPH_OSD_FLAG_IGNORE_OVERLAY |
3715 				CEPH_OSD_FLAG_IGNORE_CACHE;
3716 		req->r_tid = 0;
3717 		__submit_request(req, false);
3718 		goto out_unlock_osdc;
3719 	}
3720 
3721 	if (m.result == -EAGAIN) {
3722 		dout("req %p tid %llu EAGAIN\n", req, req->r_tid);
3723 		unlink_request(osd, req);
3724 		mutex_unlock(&osd->lock);
3725 
3726 		/*
3727 		 * The object is missing on the replica or not (yet)
3728 		 * readable.  Clear pgid to force a resend to the primary
3729 		 * via legacy_change.
3730 		 */
3731 		req->r_t.pgid.pool = 0;
3732 		req->r_t.pgid.seed = 0;
3733 		WARN_ON(!req->r_t.used_replica);
3734 		req->r_flags &= ~(CEPH_OSD_FLAG_BALANCE_READS |
3735 				  CEPH_OSD_FLAG_LOCALIZE_READS);
3736 		req->r_tid = 0;
3737 		__submit_request(req, false);
3738 		goto out_unlock_osdc;
3739 	}
3740 
3741 	if (m.num_ops != req->r_num_ops) {
3742 		pr_err("num_ops %d != %d for tid %llu\n", m.num_ops,
3743 		       req->r_num_ops, req->r_tid);
3744 		goto fail_request;
3745 	}
3746 	for (i = 0; i < req->r_num_ops; i++) {
3747 		dout(" req %p tid %llu op %d rval %d len %u\n", req,
3748 		     req->r_tid, i, m.rval[i], m.outdata_len[i]);
3749 		req->r_ops[i].rval = m.rval[i];
3750 		req->r_ops[i].outdata_len = m.outdata_len[i];
3751 		data_len += m.outdata_len[i];
3752 	}
3753 	if (data_len != le32_to_cpu(msg->hdr.data_len)) {
3754 		pr_err("sum of lens %u != %u for tid %llu\n", data_len,
3755 		       le32_to_cpu(msg->hdr.data_len), req->r_tid);
3756 		goto fail_request;
3757 	}
3758 	dout("%s req %p tid %llu result %d data_len %u\n", __func__,
3759 	     req, req->r_tid, m.result, data_len);
3760 
3761 	/*
3762 	 * Since we only ever request ONDISK, we should only ever get
3763 	 * one (type of) reply back.
3764 	 */
3765 	WARN_ON(!(m.flags & CEPH_OSD_FLAG_ONDISK));
3766 	req->r_result = m.result ?: data_len;
3767 	finish_request(req);
3768 	mutex_unlock(&osd->lock);
3769 	up_read(&osdc->lock);
3770 
3771 	__complete_request(req);
3772 	return;
3773 
3774 fail_request:
3775 	complete_request(req, -EIO);
3776 out_unlock_session:
3777 	mutex_unlock(&osd->lock);
3778 out_unlock_osdc:
3779 	up_read(&osdc->lock);
3780 }
3781 
set_pool_was_full(struct ceph_osd_client * osdc)3782 static void set_pool_was_full(struct ceph_osd_client *osdc)
3783 {
3784 	struct rb_node *n;
3785 
3786 	for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
3787 		struct ceph_pg_pool_info *pi =
3788 		    rb_entry(n, struct ceph_pg_pool_info, node);
3789 
3790 		pi->was_full = __pool_full(pi);
3791 	}
3792 }
3793 
pool_cleared_full(struct ceph_osd_client * osdc,s64 pool_id)3794 static bool pool_cleared_full(struct ceph_osd_client *osdc, s64 pool_id)
3795 {
3796 	struct ceph_pg_pool_info *pi;
3797 
3798 	pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
3799 	if (!pi)
3800 		return false;
3801 
3802 	return pi->was_full && !__pool_full(pi);
3803 }
3804 
3805 static enum calc_target_result
recalc_linger_target(struct ceph_osd_linger_request * lreq)3806 recalc_linger_target(struct ceph_osd_linger_request *lreq)
3807 {
3808 	struct ceph_osd_client *osdc = lreq->osdc;
3809 	enum calc_target_result ct_res;
3810 
3811 	ct_res = calc_target(osdc, &lreq->t, true);
3812 	if (ct_res == CALC_TARGET_NEED_RESEND) {
3813 		struct ceph_osd *osd;
3814 
3815 		osd = lookup_create_osd(osdc, lreq->t.osd, true);
3816 		if (osd != lreq->osd) {
3817 			unlink_linger(lreq->osd, lreq);
3818 			link_linger(osd, lreq);
3819 		}
3820 	}
3821 
3822 	return ct_res;
3823 }
3824 
3825 /*
3826  * Requeue requests whose mapping to an OSD has changed.
3827  */
scan_requests(struct ceph_osd * osd,bool force_resend,bool cleared_full,bool check_pool_cleared_full,struct rb_root * need_resend,struct list_head * need_resend_linger)3828 static void scan_requests(struct ceph_osd *osd,
3829 			  bool force_resend,
3830 			  bool cleared_full,
3831 			  bool check_pool_cleared_full,
3832 			  struct rb_root *need_resend,
3833 			  struct list_head *need_resend_linger)
3834 {
3835 	struct ceph_osd_client *osdc = osd->o_osdc;
3836 	struct rb_node *n;
3837 	bool force_resend_writes;
3838 
3839 	for (n = rb_first(&osd->o_linger_requests); n; ) {
3840 		struct ceph_osd_linger_request *lreq =
3841 		    rb_entry(n, struct ceph_osd_linger_request, node);
3842 		enum calc_target_result ct_res;
3843 
3844 		n = rb_next(n); /* recalc_linger_target() */
3845 
3846 		dout("%s lreq %p linger_id %llu\n", __func__, lreq,
3847 		     lreq->linger_id);
3848 		ct_res = recalc_linger_target(lreq);
3849 		switch (ct_res) {
3850 		case CALC_TARGET_NO_ACTION:
3851 			force_resend_writes = cleared_full ||
3852 			    (check_pool_cleared_full &&
3853 			     pool_cleared_full(osdc, lreq->t.base_oloc.pool));
3854 			if (!force_resend && !force_resend_writes)
3855 				break;
3856 
3857 			fallthrough;
3858 		case CALC_TARGET_NEED_RESEND:
3859 			cancel_linger_map_check(lreq);
3860 			/*
3861 			 * scan_requests() for the previous epoch(s)
3862 			 * may have already added it to the list, since
3863 			 * it's not unlinked here.
3864 			 */
3865 			if (list_empty(&lreq->scan_item))
3866 				list_add_tail(&lreq->scan_item, need_resend_linger);
3867 			break;
3868 		case CALC_TARGET_POOL_DNE:
3869 			list_del_init(&lreq->scan_item);
3870 			check_linger_pool_dne(lreq);
3871 			break;
3872 		}
3873 	}
3874 
3875 	for (n = rb_first(&osd->o_requests); n; ) {
3876 		struct ceph_osd_request *req =
3877 		    rb_entry(n, struct ceph_osd_request, r_node);
3878 		enum calc_target_result ct_res;
3879 
3880 		n = rb_next(n); /* unlink_request(), check_pool_dne() */
3881 
3882 		dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
3883 		ct_res = calc_target(osdc, &req->r_t, false);
3884 		switch (ct_res) {
3885 		case CALC_TARGET_NO_ACTION:
3886 			force_resend_writes = cleared_full ||
3887 			    (check_pool_cleared_full &&
3888 			     pool_cleared_full(osdc, req->r_t.base_oloc.pool));
3889 			if (!force_resend &&
3890 			    (!(req->r_flags & CEPH_OSD_FLAG_WRITE) ||
3891 			     !force_resend_writes))
3892 				break;
3893 
3894 			fallthrough;
3895 		case CALC_TARGET_NEED_RESEND:
3896 			cancel_map_check(req);
3897 			unlink_request(osd, req);
3898 			insert_request(need_resend, req);
3899 			break;
3900 		case CALC_TARGET_POOL_DNE:
3901 			check_pool_dne(req);
3902 			break;
3903 		}
3904 	}
3905 }
3906 
handle_one_map(struct ceph_osd_client * osdc,void * p,void * end,bool incremental,struct rb_root * need_resend,struct list_head * need_resend_linger)3907 static int handle_one_map(struct ceph_osd_client *osdc,
3908 			  void *p, void *end, bool incremental,
3909 			  struct rb_root *need_resend,
3910 			  struct list_head *need_resend_linger)
3911 {
3912 	struct ceph_osdmap *newmap;
3913 	struct rb_node *n;
3914 	bool skipped_map = false;
3915 	bool was_full;
3916 
3917 	was_full = ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
3918 	set_pool_was_full(osdc);
3919 
3920 	if (incremental)
3921 		newmap = osdmap_apply_incremental(&p, end,
3922 						  ceph_msgr2(osdc->client),
3923 						  osdc->osdmap);
3924 	else
3925 		newmap = ceph_osdmap_decode(&p, end, ceph_msgr2(osdc->client));
3926 	if (IS_ERR(newmap))
3927 		return PTR_ERR(newmap);
3928 
3929 	if (newmap != osdc->osdmap) {
3930 		/*
3931 		 * Preserve ->was_full before destroying the old map.
3932 		 * For pools that weren't in the old map, ->was_full
3933 		 * should be false.
3934 		 */
3935 		for (n = rb_first(&newmap->pg_pools); n; n = rb_next(n)) {
3936 			struct ceph_pg_pool_info *pi =
3937 			    rb_entry(n, struct ceph_pg_pool_info, node);
3938 			struct ceph_pg_pool_info *old_pi;
3939 
3940 			old_pi = ceph_pg_pool_by_id(osdc->osdmap, pi->id);
3941 			if (old_pi)
3942 				pi->was_full = old_pi->was_full;
3943 			else
3944 				WARN_ON(pi->was_full);
3945 		}
3946 
3947 		if (osdc->osdmap->epoch &&
3948 		    osdc->osdmap->epoch + 1 < newmap->epoch) {
3949 			WARN_ON(incremental);
3950 			skipped_map = true;
3951 		}
3952 
3953 		ceph_osdmap_destroy(osdc->osdmap);
3954 		osdc->osdmap = newmap;
3955 	}
3956 
3957 	was_full &= !ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
3958 	scan_requests(&osdc->homeless_osd, skipped_map, was_full, true,
3959 		      need_resend, need_resend_linger);
3960 
3961 	for (n = rb_first(&osdc->osds); n; ) {
3962 		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
3963 
3964 		n = rb_next(n); /* close_osd() */
3965 
3966 		scan_requests(osd, skipped_map, was_full, true, need_resend,
3967 			      need_resend_linger);
3968 		if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) ||
3969 		    memcmp(&osd->o_con.peer_addr,
3970 			   ceph_osd_addr(osdc->osdmap, osd->o_osd),
3971 			   sizeof(struct ceph_entity_addr)))
3972 			close_osd(osd);
3973 	}
3974 
3975 	return 0;
3976 }
3977 
kick_requests(struct ceph_osd_client * osdc,struct rb_root * need_resend,struct list_head * need_resend_linger)3978 static void kick_requests(struct ceph_osd_client *osdc,
3979 			  struct rb_root *need_resend,
3980 			  struct list_head *need_resend_linger)
3981 {
3982 	struct ceph_osd_linger_request *lreq, *nlreq;
3983 	enum calc_target_result ct_res;
3984 	struct rb_node *n;
3985 
3986 	/* make sure need_resend targets reflect latest map */
3987 	for (n = rb_first(need_resend); n; ) {
3988 		struct ceph_osd_request *req =
3989 		    rb_entry(n, struct ceph_osd_request, r_node);
3990 
3991 		n = rb_next(n);
3992 
3993 		if (req->r_t.epoch < osdc->osdmap->epoch) {
3994 			ct_res = calc_target(osdc, &req->r_t, false);
3995 			if (ct_res == CALC_TARGET_POOL_DNE) {
3996 				erase_request(need_resend, req);
3997 				check_pool_dne(req);
3998 			}
3999 		}
4000 	}
4001 
4002 	for (n = rb_first(need_resend); n; ) {
4003 		struct ceph_osd_request *req =
4004 		    rb_entry(n, struct ceph_osd_request, r_node);
4005 		struct ceph_osd *osd;
4006 
4007 		n = rb_next(n);
4008 		erase_request(need_resend, req); /* before link_request() */
4009 
4010 		osd = lookup_create_osd(osdc, req->r_t.osd, true);
4011 		link_request(osd, req);
4012 		if (!req->r_linger) {
4013 			if (!osd_homeless(osd) && !req->r_t.paused)
4014 				send_request(req);
4015 		} else {
4016 			cancel_linger_request(req);
4017 		}
4018 	}
4019 
4020 	list_for_each_entry_safe(lreq, nlreq, need_resend_linger, scan_item) {
4021 		if (!osd_homeless(lreq->osd))
4022 			send_linger(lreq);
4023 
4024 		list_del_init(&lreq->scan_item);
4025 	}
4026 }
4027 
4028 /*
4029  * Process updated osd map.
4030  *
4031  * The message contains any number of incremental and full maps, normally
4032  * indicating some sort of topology change in the cluster.  Kick requests
4033  * off to different OSDs as needed.
4034  */
ceph_osdc_handle_map(struct ceph_osd_client * osdc,struct ceph_msg * msg)4035 void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
4036 {
4037 	void *p = msg->front.iov_base;
4038 	void *const end = p + msg->front.iov_len;
4039 	u32 nr_maps, maplen;
4040 	u32 epoch;
4041 	struct ceph_fsid fsid;
4042 	struct rb_root need_resend = RB_ROOT;
4043 	LIST_HEAD(need_resend_linger);
4044 	bool handled_incremental = false;
4045 	bool was_pauserd, was_pausewr;
4046 	bool pauserd, pausewr;
4047 	int err;
4048 
4049 	dout("%s have %u\n", __func__, osdc->osdmap->epoch);
4050 	down_write(&osdc->lock);
4051 
4052 	/* verify fsid */
4053 	ceph_decode_need(&p, end, sizeof(fsid), bad);
4054 	ceph_decode_copy(&p, &fsid, sizeof(fsid));
4055 	if (ceph_check_fsid(osdc->client, &fsid) < 0)
4056 		goto bad;
4057 
4058 	was_pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
4059 	was_pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
4060 		      ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
4061 		      have_pool_full(osdc);
4062 
4063 	/* incremental maps */
4064 	ceph_decode_32_safe(&p, end, nr_maps, bad);
4065 	dout(" %d inc maps\n", nr_maps);
4066 	while (nr_maps > 0) {
4067 		ceph_decode_need(&p, end, 2*sizeof(u32), bad);
4068 		epoch = ceph_decode_32(&p);
4069 		maplen = ceph_decode_32(&p);
4070 		ceph_decode_need(&p, end, maplen, bad);
4071 		if (osdc->osdmap->epoch &&
4072 		    osdc->osdmap->epoch + 1 == epoch) {
4073 			dout("applying incremental map %u len %d\n",
4074 			     epoch, maplen);
4075 			err = handle_one_map(osdc, p, p + maplen, true,
4076 					     &need_resend, &need_resend_linger);
4077 			if (err)
4078 				goto bad;
4079 			handled_incremental = true;
4080 		} else {
4081 			dout("ignoring incremental map %u len %d\n",
4082 			     epoch, maplen);
4083 		}
4084 		p += maplen;
4085 		nr_maps--;
4086 	}
4087 	if (handled_incremental)
4088 		goto done;
4089 
4090 	/* full maps */
4091 	ceph_decode_32_safe(&p, end, nr_maps, bad);
4092 	dout(" %d full maps\n", nr_maps);
4093 	while (nr_maps) {
4094 		ceph_decode_need(&p, end, 2*sizeof(u32), bad);
4095 		epoch = ceph_decode_32(&p);
4096 		maplen = ceph_decode_32(&p);
4097 		ceph_decode_need(&p, end, maplen, bad);
4098 		if (nr_maps > 1) {
4099 			dout("skipping non-latest full map %u len %d\n",
4100 			     epoch, maplen);
4101 		} else if (osdc->osdmap->epoch >= epoch) {
4102 			dout("skipping full map %u len %d, "
4103 			     "older than our %u\n", epoch, maplen,
4104 			     osdc->osdmap->epoch);
4105 		} else {
4106 			dout("taking full map %u len %d\n", epoch, maplen);
4107 			err = handle_one_map(osdc, p, p + maplen, false,
4108 					     &need_resend, &need_resend_linger);
4109 			if (err)
4110 				goto bad;
4111 		}
4112 		p += maplen;
4113 		nr_maps--;
4114 	}
4115 
4116 done:
4117 	/*
4118 	 * subscribe to subsequent osdmap updates if full to ensure
4119 	 * we find out when we are no longer full and stop returning
4120 	 * ENOSPC.
4121 	 */
4122 	pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
4123 	pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
4124 		  ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
4125 		  have_pool_full(osdc);
4126 	if (was_pauserd || was_pausewr || pauserd || pausewr ||
4127 	    osdc->osdmap->epoch < osdc->epoch_barrier)
4128 		maybe_request_map(osdc);
4129 
4130 	kick_requests(osdc, &need_resend, &need_resend_linger);
4131 
4132 	ceph_osdc_abort_on_full(osdc);
4133 	ceph_monc_got_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
4134 			  osdc->osdmap->epoch);
4135 	up_write(&osdc->lock);
4136 	wake_up_all(&osdc->client->auth_wq);
4137 	return;
4138 
4139 bad:
4140 	pr_err("osdc handle_map corrupt msg\n");
4141 	ceph_msg_dump(msg);
4142 	up_write(&osdc->lock);
4143 }
4144 
4145 /*
4146  * Resubmit requests pending on the given osd.
4147  */
kick_osd_requests(struct ceph_osd * osd)4148 static void kick_osd_requests(struct ceph_osd *osd)
4149 {
4150 	struct rb_node *n;
4151 
4152 	clear_backoffs(osd);
4153 
4154 	for (n = rb_first(&osd->o_requests); n; ) {
4155 		struct ceph_osd_request *req =
4156 		    rb_entry(n, struct ceph_osd_request, r_node);
4157 
4158 		n = rb_next(n); /* cancel_linger_request() */
4159 
4160 		if (!req->r_linger) {
4161 			if (!req->r_t.paused)
4162 				send_request(req);
4163 		} else {
4164 			cancel_linger_request(req);
4165 		}
4166 	}
4167 	for (n = rb_first(&osd->o_linger_requests); n; n = rb_next(n)) {
4168 		struct ceph_osd_linger_request *lreq =
4169 		    rb_entry(n, struct ceph_osd_linger_request, node);
4170 
4171 		send_linger(lreq);
4172 	}
4173 }
4174 
4175 /*
4176  * If the osd connection drops, we need to resubmit all requests.
4177  */
osd_fault(struct ceph_connection * con)4178 static void osd_fault(struct ceph_connection *con)
4179 {
4180 	struct ceph_osd *osd = con->private;
4181 	struct ceph_osd_client *osdc = osd->o_osdc;
4182 
4183 	dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
4184 
4185 	down_write(&osdc->lock);
4186 	if (!osd_registered(osd)) {
4187 		dout("%s osd%d unknown\n", __func__, osd->o_osd);
4188 		goto out_unlock;
4189 	}
4190 
4191 	if (!reopen_osd(osd))
4192 		kick_osd_requests(osd);
4193 	maybe_request_map(osdc);
4194 
4195 out_unlock:
4196 	up_write(&osdc->lock);
4197 }
4198 
4199 struct MOSDBackoff {
4200 	struct ceph_spg spgid;
4201 	u32 map_epoch;
4202 	u8 op;
4203 	u64 id;
4204 	struct ceph_hobject_id *begin;
4205 	struct ceph_hobject_id *end;
4206 };
4207 
decode_MOSDBackoff(const struct ceph_msg * msg,struct MOSDBackoff * m)4208 static int decode_MOSDBackoff(const struct ceph_msg *msg, struct MOSDBackoff *m)
4209 {
4210 	void *p = msg->front.iov_base;
4211 	void *const end = p + msg->front.iov_len;
4212 	u8 struct_v;
4213 	u32 struct_len;
4214 	int ret;
4215 
4216 	ret = ceph_start_decoding(&p, end, 1, "spg_t", &struct_v, &struct_len);
4217 	if (ret)
4218 		return ret;
4219 
4220 	ret = ceph_decode_pgid(&p, end, &m->spgid.pgid);
4221 	if (ret)
4222 		return ret;
4223 
4224 	ceph_decode_8_safe(&p, end, m->spgid.shard, e_inval);
4225 	ceph_decode_32_safe(&p, end, m->map_epoch, e_inval);
4226 	ceph_decode_8_safe(&p, end, m->op, e_inval);
4227 	ceph_decode_64_safe(&p, end, m->id, e_inval);
4228 
4229 	m->begin = kzalloc(sizeof(*m->begin), GFP_NOIO);
4230 	if (!m->begin)
4231 		return -ENOMEM;
4232 
4233 	ret = decode_hoid(&p, end, m->begin);
4234 	if (ret) {
4235 		free_hoid(m->begin);
4236 		return ret;
4237 	}
4238 
4239 	m->end = kzalloc(sizeof(*m->end), GFP_NOIO);
4240 	if (!m->end) {
4241 		free_hoid(m->begin);
4242 		return -ENOMEM;
4243 	}
4244 
4245 	ret = decode_hoid(&p, end, m->end);
4246 	if (ret) {
4247 		free_hoid(m->begin);
4248 		free_hoid(m->end);
4249 		return ret;
4250 	}
4251 
4252 	return 0;
4253 
4254 e_inval:
4255 	return -EINVAL;
4256 }
4257 
create_backoff_message(const struct ceph_osd_backoff * backoff,u32 map_epoch)4258 static struct ceph_msg *create_backoff_message(
4259 				const struct ceph_osd_backoff *backoff,
4260 				u32 map_epoch)
4261 {
4262 	struct ceph_msg *msg;
4263 	void *p, *end;
4264 	int msg_size;
4265 
4266 	msg_size = CEPH_ENCODING_START_BLK_LEN +
4267 			CEPH_PGID_ENCODING_LEN + 1; /* spgid */
4268 	msg_size += 4 + 1 + 8; /* map_epoch, op, id */
4269 	msg_size += CEPH_ENCODING_START_BLK_LEN +
4270 			hoid_encoding_size(backoff->begin);
4271 	msg_size += CEPH_ENCODING_START_BLK_LEN +
4272 			hoid_encoding_size(backoff->end);
4273 
4274 	msg = ceph_msg_new(CEPH_MSG_OSD_BACKOFF, msg_size, GFP_NOIO, true);
4275 	if (!msg)
4276 		return NULL;
4277 
4278 	p = msg->front.iov_base;
4279 	end = p + msg->front_alloc_len;
4280 
4281 	encode_spgid(&p, &backoff->spgid);
4282 	ceph_encode_32(&p, map_epoch);
4283 	ceph_encode_8(&p, CEPH_OSD_BACKOFF_OP_ACK_BLOCK);
4284 	ceph_encode_64(&p, backoff->id);
4285 	encode_hoid(&p, end, backoff->begin);
4286 	encode_hoid(&p, end, backoff->end);
4287 	BUG_ON(p != end);
4288 
4289 	msg->front.iov_len = p - msg->front.iov_base;
4290 	msg->hdr.version = cpu_to_le16(1); /* MOSDBackoff v1 */
4291 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
4292 
4293 	return msg;
4294 }
4295 
handle_backoff_block(struct ceph_osd * osd,struct MOSDBackoff * m)4296 static void handle_backoff_block(struct ceph_osd *osd, struct MOSDBackoff *m)
4297 {
4298 	struct ceph_spg_mapping *spg;
4299 	struct ceph_osd_backoff *backoff;
4300 	struct ceph_msg *msg;
4301 
4302 	dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4303 	     m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4304 
4305 	spg = lookup_spg_mapping(&osd->o_backoff_mappings, &m->spgid);
4306 	if (!spg) {
4307 		spg = alloc_spg_mapping();
4308 		if (!spg) {
4309 			pr_err("%s failed to allocate spg\n", __func__);
4310 			return;
4311 		}
4312 		spg->spgid = m->spgid; /* struct */
4313 		insert_spg_mapping(&osd->o_backoff_mappings, spg);
4314 	}
4315 
4316 	backoff = alloc_backoff();
4317 	if (!backoff) {
4318 		pr_err("%s failed to allocate backoff\n", __func__);
4319 		return;
4320 	}
4321 	backoff->spgid = m->spgid; /* struct */
4322 	backoff->id = m->id;
4323 	backoff->begin = m->begin;
4324 	m->begin = NULL; /* backoff now owns this */
4325 	backoff->end = m->end;
4326 	m->end = NULL;   /* ditto */
4327 
4328 	insert_backoff(&spg->backoffs, backoff);
4329 	insert_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4330 
4331 	/*
4332 	 * Ack with original backoff's epoch so that the OSD can
4333 	 * discard this if there was a PG split.
4334 	 */
4335 	msg = create_backoff_message(backoff, m->map_epoch);
4336 	if (!msg) {
4337 		pr_err("%s failed to allocate msg\n", __func__);
4338 		return;
4339 	}
4340 	ceph_con_send(&osd->o_con, msg);
4341 }
4342 
target_contained_by(const struct ceph_osd_request_target * t,const struct ceph_hobject_id * begin,const struct ceph_hobject_id * end)4343 static bool target_contained_by(const struct ceph_osd_request_target *t,
4344 				const struct ceph_hobject_id *begin,
4345 				const struct ceph_hobject_id *end)
4346 {
4347 	struct ceph_hobject_id hoid;
4348 	int cmp;
4349 
4350 	hoid_fill_from_target(&hoid, t);
4351 	cmp = hoid_compare(&hoid, begin);
4352 	return !cmp || (cmp > 0 && hoid_compare(&hoid, end) < 0);
4353 }
4354 
handle_backoff_unblock(struct ceph_osd * osd,const struct MOSDBackoff * m)4355 static void handle_backoff_unblock(struct ceph_osd *osd,
4356 				   const struct MOSDBackoff *m)
4357 {
4358 	struct ceph_spg_mapping *spg;
4359 	struct ceph_osd_backoff *backoff;
4360 	struct rb_node *n;
4361 
4362 	dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4363 	     m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4364 
4365 	backoff = lookup_backoff_by_id(&osd->o_backoffs_by_id, m->id);
4366 	if (!backoff) {
4367 		pr_err("%s osd%d spgid %llu.%xs%d id %llu backoff dne\n",
4368 		       __func__, osd->o_osd, m->spgid.pgid.pool,
4369 		       m->spgid.pgid.seed, m->spgid.shard, m->id);
4370 		return;
4371 	}
4372 
4373 	if (hoid_compare(backoff->begin, m->begin) &&
4374 	    hoid_compare(backoff->end, m->end)) {
4375 		pr_err("%s osd%d spgid %llu.%xs%d id %llu bad range?\n",
4376 		       __func__, osd->o_osd, m->spgid.pgid.pool,
4377 		       m->spgid.pgid.seed, m->spgid.shard, m->id);
4378 		/* unblock it anyway... */
4379 	}
4380 
4381 	spg = lookup_spg_mapping(&osd->o_backoff_mappings, &backoff->spgid);
4382 	BUG_ON(!spg);
4383 
4384 	erase_backoff(&spg->backoffs, backoff);
4385 	erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4386 	free_backoff(backoff);
4387 
4388 	if (RB_EMPTY_ROOT(&spg->backoffs)) {
4389 		erase_spg_mapping(&osd->o_backoff_mappings, spg);
4390 		free_spg_mapping(spg);
4391 	}
4392 
4393 	for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
4394 		struct ceph_osd_request *req =
4395 		    rb_entry(n, struct ceph_osd_request, r_node);
4396 
4397 		if (!ceph_spg_compare(&req->r_t.spgid, &m->spgid)) {
4398 			/*
4399 			 * Match against @m, not @backoff -- the PG may
4400 			 * have split on the OSD.
4401 			 */
4402 			if (target_contained_by(&req->r_t, m->begin, m->end)) {
4403 				/*
4404 				 * If no other installed backoff applies,
4405 				 * resend.
4406 				 */
4407 				send_request(req);
4408 			}
4409 		}
4410 	}
4411 }
4412 
handle_backoff(struct ceph_osd * osd,struct ceph_msg * msg)4413 static void handle_backoff(struct ceph_osd *osd, struct ceph_msg *msg)
4414 {
4415 	struct ceph_osd_client *osdc = osd->o_osdc;
4416 	struct MOSDBackoff m;
4417 	int ret;
4418 
4419 	down_read(&osdc->lock);
4420 	if (!osd_registered(osd)) {
4421 		dout("%s osd%d unknown\n", __func__, osd->o_osd);
4422 		up_read(&osdc->lock);
4423 		return;
4424 	}
4425 	WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
4426 
4427 	mutex_lock(&osd->lock);
4428 	ret = decode_MOSDBackoff(msg, &m);
4429 	if (ret) {
4430 		pr_err("failed to decode MOSDBackoff: %d\n", ret);
4431 		ceph_msg_dump(msg);
4432 		goto out_unlock;
4433 	}
4434 
4435 	switch (m.op) {
4436 	case CEPH_OSD_BACKOFF_OP_BLOCK:
4437 		handle_backoff_block(osd, &m);
4438 		break;
4439 	case CEPH_OSD_BACKOFF_OP_UNBLOCK:
4440 		handle_backoff_unblock(osd, &m);
4441 		break;
4442 	default:
4443 		pr_err("%s osd%d unknown op %d\n", __func__, osd->o_osd, m.op);
4444 	}
4445 
4446 	free_hoid(m.begin);
4447 	free_hoid(m.end);
4448 
4449 out_unlock:
4450 	mutex_unlock(&osd->lock);
4451 	up_read(&osdc->lock);
4452 }
4453 
4454 /*
4455  * Process osd watch notifications
4456  */
handle_watch_notify(struct ceph_osd_client * osdc,struct ceph_msg * msg)4457 static void handle_watch_notify(struct ceph_osd_client *osdc,
4458 				struct ceph_msg *msg)
4459 {
4460 	void *p = msg->front.iov_base;
4461 	void *const end = p + msg->front.iov_len;
4462 	struct ceph_osd_linger_request *lreq;
4463 	struct linger_work *lwork;
4464 	u8 proto_ver, opcode;
4465 	u64 cookie, notify_id;
4466 	u64 notifier_id = 0;
4467 	s32 return_code = 0;
4468 	void *payload = NULL;
4469 	u32 payload_len = 0;
4470 
4471 	ceph_decode_8_safe(&p, end, proto_ver, bad);
4472 	ceph_decode_8_safe(&p, end, opcode, bad);
4473 	ceph_decode_64_safe(&p, end, cookie, bad);
4474 	p += 8; /* skip ver */
4475 	ceph_decode_64_safe(&p, end, notify_id, bad);
4476 
4477 	if (proto_ver >= 1) {
4478 		ceph_decode_32_safe(&p, end, payload_len, bad);
4479 		ceph_decode_need(&p, end, payload_len, bad);
4480 		payload = p;
4481 		p += payload_len;
4482 	}
4483 
4484 	if (le16_to_cpu(msg->hdr.version) >= 2)
4485 		ceph_decode_32_safe(&p, end, return_code, bad);
4486 
4487 	if (le16_to_cpu(msg->hdr.version) >= 3)
4488 		ceph_decode_64_safe(&p, end, notifier_id, bad);
4489 
4490 	down_read(&osdc->lock);
4491 	lreq = lookup_linger_osdc(&osdc->linger_requests, cookie);
4492 	if (!lreq) {
4493 		dout("%s opcode %d cookie %llu dne\n", __func__, opcode,
4494 		     cookie);
4495 		goto out_unlock_osdc;
4496 	}
4497 
4498 	mutex_lock(&lreq->lock);
4499 	dout("%s opcode %d cookie %llu lreq %p is_watch %d\n", __func__,
4500 	     opcode, cookie, lreq, lreq->is_watch);
4501 	if (opcode == CEPH_WATCH_EVENT_DISCONNECT) {
4502 		if (!lreq->last_error) {
4503 			lreq->last_error = -ENOTCONN;
4504 			queue_watch_error(lreq);
4505 		}
4506 	} else if (!lreq->is_watch) {
4507 		/* CEPH_WATCH_EVENT_NOTIFY_COMPLETE */
4508 		if (lreq->notify_id && lreq->notify_id != notify_id) {
4509 			dout("lreq %p notify_id %llu != %llu, ignoring\n", lreq,
4510 			     lreq->notify_id, notify_id);
4511 		} else if (!completion_done(&lreq->notify_finish_wait)) {
4512 			struct ceph_msg_data *data =
4513 			    msg->num_data_items ? &msg->data[0] : NULL;
4514 
4515 			if (data) {
4516 				if (lreq->preply_pages) {
4517 					WARN_ON(data->type !=
4518 							CEPH_MSG_DATA_PAGES);
4519 					*lreq->preply_pages = data->pages;
4520 					*lreq->preply_len = data->length;
4521 					data->own_pages = false;
4522 				}
4523 			}
4524 			lreq->notify_finish_error = return_code;
4525 			complete_all(&lreq->notify_finish_wait);
4526 		}
4527 	} else {
4528 		/* CEPH_WATCH_EVENT_NOTIFY */
4529 		lwork = lwork_alloc(lreq, do_watch_notify);
4530 		if (!lwork) {
4531 			pr_err("failed to allocate notify-lwork\n");
4532 			goto out_unlock_lreq;
4533 		}
4534 
4535 		lwork->notify.notify_id = notify_id;
4536 		lwork->notify.notifier_id = notifier_id;
4537 		lwork->notify.payload = payload;
4538 		lwork->notify.payload_len = payload_len;
4539 		lwork->notify.msg = ceph_msg_get(msg);
4540 		lwork_queue(lwork);
4541 	}
4542 
4543 out_unlock_lreq:
4544 	mutex_unlock(&lreq->lock);
4545 out_unlock_osdc:
4546 	up_read(&osdc->lock);
4547 	return;
4548 
4549 bad:
4550 	pr_err("osdc handle_watch_notify corrupt msg\n");
4551 }
4552 
4553 /*
4554  * Register request, send initial attempt.
4555  */
ceph_osdc_start_request(struct ceph_osd_client * osdc,struct ceph_osd_request * req,bool nofail)4556 int ceph_osdc_start_request(struct ceph_osd_client *osdc,
4557 			    struct ceph_osd_request *req,
4558 			    bool nofail)
4559 {
4560 	down_read(&osdc->lock);
4561 	submit_request(req, false);
4562 	up_read(&osdc->lock);
4563 
4564 	return 0;
4565 }
4566 EXPORT_SYMBOL(ceph_osdc_start_request);
4567 
4568 /*
4569  * Unregister a registered request.  The request is not completed:
4570  * ->r_result isn't set and __complete_request() isn't called.
4571  */
ceph_osdc_cancel_request(struct ceph_osd_request * req)4572 void ceph_osdc_cancel_request(struct ceph_osd_request *req)
4573 {
4574 	struct ceph_osd_client *osdc = req->r_osdc;
4575 
4576 	down_write(&osdc->lock);
4577 	if (req->r_osd)
4578 		cancel_request(req);
4579 	up_write(&osdc->lock);
4580 }
4581 EXPORT_SYMBOL(ceph_osdc_cancel_request);
4582 
4583 /*
4584  * @timeout: in jiffies, 0 means "wait forever"
4585  */
wait_request_timeout(struct ceph_osd_request * req,unsigned long timeout)4586 static int wait_request_timeout(struct ceph_osd_request *req,
4587 				unsigned long timeout)
4588 {
4589 	long left;
4590 
4591 	dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
4592 	left = wait_for_completion_killable_timeout(&req->r_completion,
4593 						ceph_timeout_jiffies(timeout));
4594 	if (left <= 0) {
4595 		left = left ?: -ETIMEDOUT;
4596 		ceph_osdc_cancel_request(req);
4597 	} else {
4598 		left = req->r_result; /* completed */
4599 	}
4600 
4601 	return left;
4602 }
4603 
4604 /*
4605  * wait for a request to complete
4606  */
ceph_osdc_wait_request(struct ceph_osd_client * osdc,struct ceph_osd_request * req)4607 int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
4608 			   struct ceph_osd_request *req)
4609 {
4610 	return wait_request_timeout(req, 0);
4611 }
4612 EXPORT_SYMBOL(ceph_osdc_wait_request);
4613 
4614 /*
4615  * sync - wait for all in-flight requests to flush.  avoid starvation.
4616  */
ceph_osdc_sync(struct ceph_osd_client * osdc)4617 void ceph_osdc_sync(struct ceph_osd_client *osdc)
4618 {
4619 	struct rb_node *n, *p;
4620 	u64 last_tid = atomic64_read(&osdc->last_tid);
4621 
4622 again:
4623 	down_read(&osdc->lock);
4624 	for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
4625 		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
4626 
4627 		mutex_lock(&osd->lock);
4628 		for (p = rb_first(&osd->o_requests); p; p = rb_next(p)) {
4629 			struct ceph_osd_request *req =
4630 			    rb_entry(p, struct ceph_osd_request, r_node);
4631 
4632 			if (req->r_tid > last_tid)
4633 				break;
4634 
4635 			if (!(req->r_flags & CEPH_OSD_FLAG_WRITE))
4636 				continue;
4637 
4638 			ceph_osdc_get_request(req);
4639 			mutex_unlock(&osd->lock);
4640 			up_read(&osdc->lock);
4641 			dout("%s waiting on req %p tid %llu last_tid %llu\n",
4642 			     __func__, req, req->r_tid, last_tid);
4643 			wait_for_completion(&req->r_completion);
4644 			ceph_osdc_put_request(req);
4645 			goto again;
4646 		}
4647 
4648 		mutex_unlock(&osd->lock);
4649 	}
4650 
4651 	up_read(&osdc->lock);
4652 	dout("%s done last_tid %llu\n", __func__, last_tid);
4653 }
4654 EXPORT_SYMBOL(ceph_osdc_sync);
4655 
4656 static struct ceph_osd_request *
alloc_linger_request(struct ceph_osd_linger_request * lreq)4657 alloc_linger_request(struct ceph_osd_linger_request *lreq)
4658 {
4659 	struct ceph_osd_request *req;
4660 
4661 	req = ceph_osdc_alloc_request(lreq->osdc, NULL, 1, false, GFP_NOIO);
4662 	if (!req)
4663 		return NULL;
4664 
4665 	ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
4666 	ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
4667 	return req;
4668 }
4669 
4670 static struct ceph_osd_request *
alloc_watch_request(struct ceph_osd_linger_request * lreq,u8 watch_opcode)4671 alloc_watch_request(struct ceph_osd_linger_request *lreq, u8 watch_opcode)
4672 {
4673 	struct ceph_osd_request *req;
4674 
4675 	req = alloc_linger_request(lreq);
4676 	if (!req)
4677 		return NULL;
4678 
4679 	/*
4680 	 * Pass 0 for cookie because we don't know it yet, it will be
4681 	 * filled in by linger_submit().
4682 	 */
4683 	osd_req_op_watch_init(req, 0, 0, watch_opcode);
4684 
4685 	if (ceph_osdc_alloc_messages(req, GFP_NOIO)) {
4686 		ceph_osdc_put_request(req);
4687 		return NULL;
4688 	}
4689 
4690 	return req;
4691 }
4692 
4693 /*
4694  * Returns a handle, caller owns a ref.
4695  */
4696 struct ceph_osd_linger_request *
ceph_osdc_watch(struct ceph_osd_client * osdc,struct ceph_object_id * oid,struct ceph_object_locator * oloc,rados_watchcb2_t wcb,rados_watcherrcb_t errcb,void * data)4697 ceph_osdc_watch(struct ceph_osd_client *osdc,
4698 		struct ceph_object_id *oid,
4699 		struct ceph_object_locator *oloc,
4700 		rados_watchcb2_t wcb,
4701 		rados_watcherrcb_t errcb,
4702 		void *data)
4703 {
4704 	struct ceph_osd_linger_request *lreq;
4705 	int ret;
4706 
4707 	lreq = linger_alloc(osdc);
4708 	if (!lreq)
4709 		return ERR_PTR(-ENOMEM);
4710 
4711 	lreq->is_watch = true;
4712 	lreq->wcb = wcb;
4713 	lreq->errcb = errcb;
4714 	lreq->data = data;
4715 	lreq->watch_valid_thru = jiffies;
4716 
4717 	ceph_oid_copy(&lreq->t.base_oid, oid);
4718 	ceph_oloc_copy(&lreq->t.base_oloc, oloc);
4719 	lreq->t.flags = CEPH_OSD_FLAG_WRITE;
4720 	ktime_get_real_ts64(&lreq->mtime);
4721 
4722 	lreq->reg_req = alloc_watch_request(lreq, CEPH_OSD_WATCH_OP_WATCH);
4723 	if (!lreq->reg_req) {
4724 		ret = -ENOMEM;
4725 		goto err_put_lreq;
4726 	}
4727 
4728 	lreq->ping_req = alloc_watch_request(lreq, CEPH_OSD_WATCH_OP_PING);
4729 	if (!lreq->ping_req) {
4730 		ret = -ENOMEM;
4731 		goto err_put_lreq;
4732 	}
4733 
4734 	linger_submit(lreq);
4735 	ret = linger_reg_commit_wait(lreq);
4736 	if (ret) {
4737 		linger_cancel(lreq);
4738 		goto err_put_lreq;
4739 	}
4740 
4741 	return lreq;
4742 
4743 err_put_lreq:
4744 	linger_put(lreq);
4745 	return ERR_PTR(ret);
4746 }
4747 EXPORT_SYMBOL(ceph_osdc_watch);
4748 
4749 /*
4750  * Releases a ref.
4751  *
4752  * Times out after mount_timeout to preserve rbd unmap behaviour
4753  * introduced in 2894e1d76974 ("rbd: timeout watch teardown on unmap
4754  * with mount_timeout").
4755  */
ceph_osdc_unwatch(struct ceph_osd_client * osdc,struct ceph_osd_linger_request * lreq)4756 int ceph_osdc_unwatch(struct ceph_osd_client *osdc,
4757 		      struct ceph_osd_linger_request *lreq)
4758 {
4759 	struct ceph_options *opts = osdc->client->options;
4760 	struct ceph_osd_request *req;
4761 	int ret;
4762 
4763 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4764 	if (!req)
4765 		return -ENOMEM;
4766 
4767 	ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
4768 	ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
4769 	req->r_flags = CEPH_OSD_FLAG_WRITE;
4770 	ktime_get_real_ts64(&req->r_mtime);
4771 	osd_req_op_watch_init(req, 0, lreq->linger_id,
4772 			      CEPH_OSD_WATCH_OP_UNWATCH);
4773 
4774 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4775 	if (ret)
4776 		goto out_put_req;
4777 
4778 	ceph_osdc_start_request(osdc, req, false);
4779 	linger_cancel(lreq);
4780 	linger_put(lreq);
4781 	ret = wait_request_timeout(req, opts->mount_timeout);
4782 
4783 out_put_req:
4784 	ceph_osdc_put_request(req);
4785 	return ret;
4786 }
4787 EXPORT_SYMBOL(ceph_osdc_unwatch);
4788 
osd_req_op_notify_ack_init(struct ceph_osd_request * req,int which,u64 notify_id,u64 cookie,void * payload,u32 payload_len)4789 static int osd_req_op_notify_ack_init(struct ceph_osd_request *req, int which,
4790 				      u64 notify_id, u64 cookie, void *payload,
4791 				      u32 payload_len)
4792 {
4793 	struct ceph_osd_req_op *op;
4794 	struct ceph_pagelist *pl;
4795 	int ret;
4796 
4797 	op = osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY_ACK, 0);
4798 
4799 	pl = ceph_pagelist_alloc(GFP_NOIO);
4800 	if (!pl)
4801 		return -ENOMEM;
4802 
4803 	ret = ceph_pagelist_encode_64(pl, notify_id);
4804 	ret |= ceph_pagelist_encode_64(pl, cookie);
4805 	if (payload) {
4806 		ret |= ceph_pagelist_encode_32(pl, payload_len);
4807 		ret |= ceph_pagelist_append(pl, payload, payload_len);
4808 	} else {
4809 		ret |= ceph_pagelist_encode_32(pl, 0);
4810 	}
4811 	if (ret) {
4812 		ceph_pagelist_release(pl);
4813 		return -ENOMEM;
4814 	}
4815 
4816 	ceph_osd_data_pagelist_init(&op->notify_ack.request_data, pl);
4817 	op->indata_len = pl->length;
4818 	return 0;
4819 }
4820 
ceph_osdc_notify_ack(struct ceph_osd_client * osdc,struct ceph_object_id * oid,struct ceph_object_locator * oloc,u64 notify_id,u64 cookie,void * payload,u32 payload_len)4821 int ceph_osdc_notify_ack(struct ceph_osd_client *osdc,
4822 			 struct ceph_object_id *oid,
4823 			 struct ceph_object_locator *oloc,
4824 			 u64 notify_id,
4825 			 u64 cookie,
4826 			 void *payload,
4827 			 u32 payload_len)
4828 {
4829 	struct ceph_osd_request *req;
4830 	int ret;
4831 
4832 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4833 	if (!req)
4834 		return -ENOMEM;
4835 
4836 	ceph_oid_copy(&req->r_base_oid, oid);
4837 	ceph_oloc_copy(&req->r_base_oloc, oloc);
4838 	req->r_flags = CEPH_OSD_FLAG_READ;
4839 
4840 	ret = osd_req_op_notify_ack_init(req, 0, notify_id, cookie, payload,
4841 					 payload_len);
4842 	if (ret)
4843 		goto out_put_req;
4844 
4845 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4846 	if (ret)
4847 		goto out_put_req;
4848 
4849 	ceph_osdc_start_request(osdc, req, false);
4850 	ret = ceph_osdc_wait_request(osdc, req);
4851 
4852 out_put_req:
4853 	ceph_osdc_put_request(req);
4854 	return ret;
4855 }
4856 EXPORT_SYMBOL(ceph_osdc_notify_ack);
4857 
osd_req_op_notify_init(struct ceph_osd_request * req,int which,u64 cookie,u32 prot_ver,u32 timeout,void * payload,u32 payload_len)4858 static int osd_req_op_notify_init(struct ceph_osd_request *req, int which,
4859 				  u64 cookie, u32 prot_ver, u32 timeout,
4860 				  void *payload, u32 payload_len)
4861 {
4862 	struct ceph_osd_req_op *op;
4863 	struct ceph_pagelist *pl;
4864 	int ret;
4865 
4866 	op = osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY, 0);
4867 	op->notify.cookie = cookie;
4868 
4869 	pl = ceph_pagelist_alloc(GFP_NOIO);
4870 	if (!pl)
4871 		return -ENOMEM;
4872 
4873 	ret = ceph_pagelist_encode_32(pl, 1); /* prot_ver */
4874 	ret |= ceph_pagelist_encode_32(pl, timeout);
4875 	ret |= ceph_pagelist_encode_32(pl, payload_len);
4876 	ret |= ceph_pagelist_append(pl, payload, payload_len);
4877 	if (ret) {
4878 		ceph_pagelist_release(pl);
4879 		return -ENOMEM;
4880 	}
4881 
4882 	ceph_osd_data_pagelist_init(&op->notify.request_data, pl);
4883 	op->indata_len = pl->length;
4884 	return 0;
4885 }
4886 
4887 /*
4888  * @timeout: in seconds
4889  *
4890  * @preply_{pages,len} are initialized both on success and error.
4891  * The caller is responsible for:
4892  *
4893  *     ceph_release_page_vector(reply_pages, calc_pages_for(0, reply_len))
4894  */
ceph_osdc_notify(struct ceph_osd_client * osdc,struct ceph_object_id * oid,struct ceph_object_locator * oloc,void * payload,u32 payload_len,u32 timeout,struct page *** preply_pages,size_t * preply_len)4895 int ceph_osdc_notify(struct ceph_osd_client *osdc,
4896 		     struct ceph_object_id *oid,
4897 		     struct ceph_object_locator *oloc,
4898 		     void *payload,
4899 		     u32 payload_len,
4900 		     u32 timeout,
4901 		     struct page ***preply_pages,
4902 		     size_t *preply_len)
4903 {
4904 	struct ceph_osd_linger_request *lreq;
4905 	struct page **pages;
4906 	int ret;
4907 
4908 	WARN_ON(!timeout);
4909 	if (preply_pages) {
4910 		*preply_pages = NULL;
4911 		*preply_len = 0;
4912 	}
4913 
4914 	lreq = linger_alloc(osdc);
4915 	if (!lreq)
4916 		return -ENOMEM;
4917 
4918 	lreq->preply_pages = preply_pages;
4919 	lreq->preply_len = preply_len;
4920 
4921 	ceph_oid_copy(&lreq->t.base_oid, oid);
4922 	ceph_oloc_copy(&lreq->t.base_oloc, oloc);
4923 	lreq->t.flags = CEPH_OSD_FLAG_READ;
4924 
4925 	lreq->reg_req = alloc_linger_request(lreq);
4926 	if (!lreq->reg_req) {
4927 		ret = -ENOMEM;
4928 		goto out_put_lreq;
4929 	}
4930 
4931 	/*
4932 	 * Pass 0 for cookie because we don't know it yet, it will be
4933 	 * filled in by linger_submit().
4934 	 */
4935 	ret = osd_req_op_notify_init(lreq->reg_req, 0, 0, 1, timeout,
4936 				     payload, payload_len);
4937 	if (ret)
4938 		goto out_put_lreq;
4939 
4940 	/* for notify_id */
4941 	pages = ceph_alloc_page_vector(1, GFP_NOIO);
4942 	if (IS_ERR(pages)) {
4943 		ret = PTR_ERR(pages);
4944 		goto out_put_lreq;
4945 	}
4946 	ceph_osd_data_pages_init(osd_req_op_data(lreq->reg_req, 0, notify,
4947 						 response_data),
4948 				 pages, PAGE_SIZE, 0, false, true);
4949 
4950 	ret = ceph_osdc_alloc_messages(lreq->reg_req, GFP_NOIO);
4951 	if (ret)
4952 		goto out_put_lreq;
4953 
4954 	linger_submit(lreq);
4955 	ret = linger_reg_commit_wait(lreq);
4956 	if (!ret)
4957 		ret = linger_notify_finish_wait(lreq);
4958 	else
4959 		dout("lreq %p failed to initiate notify %d\n", lreq, ret);
4960 
4961 	linger_cancel(lreq);
4962 out_put_lreq:
4963 	linger_put(lreq);
4964 	return ret;
4965 }
4966 EXPORT_SYMBOL(ceph_osdc_notify);
4967 
4968 /*
4969  * Return the number of milliseconds since the watch was last
4970  * confirmed, or an error.  If there is an error, the watch is no
4971  * longer valid, and should be destroyed with ceph_osdc_unwatch().
4972  */
ceph_osdc_watch_check(struct ceph_osd_client * osdc,struct ceph_osd_linger_request * lreq)4973 int ceph_osdc_watch_check(struct ceph_osd_client *osdc,
4974 			  struct ceph_osd_linger_request *lreq)
4975 {
4976 	unsigned long stamp, age;
4977 	int ret;
4978 
4979 	down_read(&osdc->lock);
4980 	mutex_lock(&lreq->lock);
4981 	stamp = lreq->watch_valid_thru;
4982 	if (!list_empty(&lreq->pending_lworks)) {
4983 		struct linger_work *lwork =
4984 		    list_first_entry(&lreq->pending_lworks,
4985 				     struct linger_work,
4986 				     pending_item);
4987 
4988 		if (time_before(lwork->queued_stamp, stamp))
4989 			stamp = lwork->queued_stamp;
4990 	}
4991 	age = jiffies - stamp;
4992 	dout("%s lreq %p linger_id %llu age %lu last_error %d\n", __func__,
4993 	     lreq, lreq->linger_id, age, lreq->last_error);
4994 	/* we are truncating to msecs, so return a safe upper bound */
4995 	ret = lreq->last_error ?: 1 + jiffies_to_msecs(age);
4996 
4997 	mutex_unlock(&lreq->lock);
4998 	up_read(&osdc->lock);
4999 	return ret;
5000 }
5001 
decode_watcher(void ** p,void * end,struct ceph_watch_item * item)5002 static int decode_watcher(void **p, void *end, struct ceph_watch_item *item)
5003 {
5004 	u8 struct_v;
5005 	u32 struct_len;
5006 	int ret;
5007 
5008 	ret = ceph_start_decoding(p, end, 2, "watch_item_t",
5009 				  &struct_v, &struct_len);
5010 	if (ret)
5011 		goto bad;
5012 
5013 	ret = -EINVAL;
5014 	ceph_decode_copy_safe(p, end, &item->name, sizeof(item->name), bad);
5015 	ceph_decode_64_safe(p, end, item->cookie, bad);
5016 	ceph_decode_skip_32(p, end, bad); /* skip timeout seconds */
5017 
5018 	if (struct_v >= 2) {
5019 		ret = ceph_decode_entity_addr(p, end, &item->addr);
5020 		if (ret)
5021 			goto bad;
5022 	} else {
5023 		ret = 0;
5024 	}
5025 
5026 	dout("%s %s%llu cookie %llu addr %s\n", __func__,
5027 	     ENTITY_NAME(item->name), item->cookie,
5028 	     ceph_pr_addr(&item->addr));
5029 bad:
5030 	return ret;
5031 }
5032 
decode_watchers(void ** p,void * end,struct ceph_watch_item ** watchers,u32 * num_watchers)5033 static int decode_watchers(void **p, void *end,
5034 			   struct ceph_watch_item **watchers,
5035 			   u32 *num_watchers)
5036 {
5037 	u8 struct_v;
5038 	u32 struct_len;
5039 	int i;
5040 	int ret;
5041 
5042 	ret = ceph_start_decoding(p, end, 1, "obj_list_watch_response_t",
5043 				  &struct_v, &struct_len);
5044 	if (ret)
5045 		return ret;
5046 
5047 	*num_watchers = ceph_decode_32(p);
5048 	*watchers = kcalloc(*num_watchers, sizeof(**watchers), GFP_NOIO);
5049 	if (!*watchers)
5050 		return -ENOMEM;
5051 
5052 	for (i = 0; i < *num_watchers; i++) {
5053 		ret = decode_watcher(p, end, *watchers + i);
5054 		if (ret) {
5055 			kfree(*watchers);
5056 			return ret;
5057 		}
5058 	}
5059 
5060 	return 0;
5061 }
5062 
5063 /*
5064  * On success, the caller is responsible for:
5065  *
5066  *     kfree(watchers);
5067  */
ceph_osdc_list_watchers(struct ceph_osd_client * osdc,struct ceph_object_id * oid,struct ceph_object_locator * oloc,struct ceph_watch_item ** watchers,u32 * num_watchers)5068 int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
5069 			    struct ceph_object_id *oid,
5070 			    struct ceph_object_locator *oloc,
5071 			    struct ceph_watch_item **watchers,
5072 			    u32 *num_watchers)
5073 {
5074 	struct ceph_osd_request *req;
5075 	struct page **pages;
5076 	int ret;
5077 
5078 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
5079 	if (!req)
5080 		return -ENOMEM;
5081 
5082 	ceph_oid_copy(&req->r_base_oid, oid);
5083 	ceph_oloc_copy(&req->r_base_oloc, oloc);
5084 	req->r_flags = CEPH_OSD_FLAG_READ;
5085 
5086 	pages = ceph_alloc_page_vector(1, GFP_NOIO);
5087 	if (IS_ERR(pages)) {
5088 		ret = PTR_ERR(pages);
5089 		goto out_put_req;
5090 	}
5091 
5092 	osd_req_op_init(req, 0, CEPH_OSD_OP_LIST_WATCHERS, 0);
5093 	ceph_osd_data_pages_init(osd_req_op_data(req, 0, list_watchers,
5094 						 response_data),
5095 				 pages, PAGE_SIZE, 0, false, true);
5096 
5097 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
5098 	if (ret)
5099 		goto out_put_req;
5100 
5101 	ceph_osdc_start_request(osdc, req, false);
5102 	ret = ceph_osdc_wait_request(osdc, req);
5103 	if (ret >= 0) {
5104 		void *p = page_address(pages[0]);
5105 		void *const end = p + req->r_ops[0].outdata_len;
5106 
5107 		ret = decode_watchers(&p, end, watchers, num_watchers);
5108 	}
5109 
5110 out_put_req:
5111 	ceph_osdc_put_request(req);
5112 	return ret;
5113 }
5114 EXPORT_SYMBOL(ceph_osdc_list_watchers);
5115 
5116 /*
5117  * Call all pending notify callbacks - for use after a watch is
5118  * unregistered, to make sure no more callbacks for it will be invoked
5119  */
ceph_osdc_flush_notifies(struct ceph_osd_client * osdc)5120 void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc)
5121 {
5122 	dout("%s osdc %p\n", __func__, osdc);
5123 	flush_workqueue(osdc->notify_wq);
5124 }
5125 EXPORT_SYMBOL(ceph_osdc_flush_notifies);
5126 
ceph_osdc_maybe_request_map(struct ceph_osd_client * osdc)5127 void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc)
5128 {
5129 	down_read(&osdc->lock);
5130 	maybe_request_map(osdc);
5131 	up_read(&osdc->lock);
5132 }
5133 EXPORT_SYMBOL(ceph_osdc_maybe_request_map);
5134 
5135 /*
5136  * Execute an OSD class method on an object.
5137  *
5138  * @flags: CEPH_OSD_FLAG_*
5139  * @resp_len: in/out param for reply length
5140  */
ceph_osdc_call(struct ceph_osd_client * osdc,struct ceph_object_id * oid,struct ceph_object_locator * oloc,const char * class,const char * method,unsigned int flags,struct page * req_page,size_t req_len,struct page ** resp_pages,size_t * resp_len)5141 int ceph_osdc_call(struct ceph_osd_client *osdc,
5142 		   struct ceph_object_id *oid,
5143 		   struct ceph_object_locator *oloc,
5144 		   const char *class, const char *method,
5145 		   unsigned int flags,
5146 		   struct page *req_page, size_t req_len,
5147 		   struct page **resp_pages, size_t *resp_len)
5148 {
5149 	struct ceph_osd_request *req;
5150 	int ret;
5151 
5152 	if (req_len > PAGE_SIZE)
5153 		return -E2BIG;
5154 
5155 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
5156 	if (!req)
5157 		return -ENOMEM;
5158 
5159 	ceph_oid_copy(&req->r_base_oid, oid);
5160 	ceph_oloc_copy(&req->r_base_oloc, oloc);
5161 	req->r_flags = flags;
5162 
5163 	ret = osd_req_op_cls_init(req, 0, class, method);
5164 	if (ret)
5165 		goto out_put_req;
5166 
5167 	if (req_page)
5168 		osd_req_op_cls_request_data_pages(req, 0, &req_page, req_len,
5169 						  0, false, false);
5170 	if (resp_pages)
5171 		osd_req_op_cls_response_data_pages(req, 0, resp_pages,
5172 						   *resp_len, 0, false, false);
5173 
5174 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
5175 	if (ret)
5176 		goto out_put_req;
5177 
5178 	ceph_osdc_start_request(osdc, req, false);
5179 	ret = ceph_osdc_wait_request(osdc, req);
5180 	if (ret >= 0) {
5181 		ret = req->r_ops[0].rval;
5182 		if (resp_pages)
5183 			*resp_len = req->r_ops[0].outdata_len;
5184 	}
5185 
5186 out_put_req:
5187 	ceph_osdc_put_request(req);
5188 	return ret;
5189 }
5190 EXPORT_SYMBOL(ceph_osdc_call);
5191 
5192 /*
5193  * reset all osd connections
5194  */
ceph_osdc_reopen_osds(struct ceph_osd_client * osdc)5195 void ceph_osdc_reopen_osds(struct ceph_osd_client *osdc)
5196 {
5197 	struct rb_node *n;
5198 
5199 	down_write(&osdc->lock);
5200 	for (n = rb_first(&osdc->osds); n; ) {
5201 		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
5202 
5203 		n = rb_next(n);
5204 		if (!reopen_osd(osd))
5205 			kick_osd_requests(osd);
5206 	}
5207 	up_write(&osdc->lock);
5208 }
5209 
5210 /*
5211  * init, shutdown
5212  */
ceph_osdc_init(struct ceph_osd_client * osdc,struct ceph_client * client)5213 int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client)
5214 {
5215 	int err;
5216 
5217 	dout("init\n");
5218 	osdc->client = client;
5219 	init_rwsem(&osdc->lock);
5220 	osdc->osds = RB_ROOT;
5221 	INIT_LIST_HEAD(&osdc->osd_lru);
5222 	spin_lock_init(&osdc->osd_lru_lock);
5223 	osd_init(&osdc->homeless_osd);
5224 	osdc->homeless_osd.o_osdc = osdc;
5225 	osdc->homeless_osd.o_osd = CEPH_HOMELESS_OSD;
5226 	osdc->last_linger_id = CEPH_LINGER_ID_START;
5227 	osdc->linger_requests = RB_ROOT;
5228 	osdc->map_checks = RB_ROOT;
5229 	osdc->linger_map_checks = RB_ROOT;
5230 	INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout);
5231 	INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout);
5232 
5233 	err = -ENOMEM;
5234 	osdc->osdmap = ceph_osdmap_alloc();
5235 	if (!osdc->osdmap)
5236 		goto out;
5237 
5238 	osdc->req_mempool = mempool_create_slab_pool(10,
5239 						     ceph_osd_request_cache);
5240 	if (!osdc->req_mempool)
5241 		goto out_map;
5242 
5243 	err = ceph_msgpool_init(&osdc->msgpool_op, CEPH_MSG_OSD_OP,
5244 				PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10, "osd_op");
5245 	if (err < 0)
5246 		goto out_mempool;
5247 	err = ceph_msgpool_init(&osdc->msgpool_op_reply, CEPH_MSG_OSD_OPREPLY,
5248 				PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10,
5249 				"osd_op_reply");
5250 	if (err < 0)
5251 		goto out_msgpool;
5252 
5253 	err = -ENOMEM;
5254 	osdc->notify_wq = create_singlethread_workqueue("ceph-watch-notify");
5255 	if (!osdc->notify_wq)
5256 		goto out_msgpool_reply;
5257 
5258 	osdc->completion_wq = create_singlethread_workqueue("ceph-completion");
5259 	if (!osdc->completion_wq)
5260 		goto out_notify_wq;
5261 
5262 	schedule_delayed_work(&osdc->timeout_work,
5263 			      osdc->client->options->osd_keepalive_timeout);
5264 	schedule_delayed_work(&osdc->osds_timeout_work,
5265 	    round_jiffies_relative(osdc->client->options->osd_idle_ttl));
5266 
5267 	return 0;
5268 
5269 out_notify_wq:
5270 	destroy_workqueue(osdc->notify_wq);
5271 out_msgpool_reply:
5272 	ceph_msgpool_destroy(&osdc->msgpool_op_reply);
5273 out_msgpool:
5274 	ceph_msgpool_destroy(&osdc->msgpool_op);
5275 out_mempool:
5276 	mempool_destroy(osdc->req_mempool);
5277 out_map:
5278 	ceph_osdmap_destroy(osdc->osdmap);
5279 out:
5280 	return err;
5281 }
5282 
ceph_osdc_stop(struct ceph_osd_client * osdc)5283 void ceph_osdc_stop(struct ceph_osd_client *osdc)
5284 {
5285 	destroy_workqueue(osdc->completion_wq);
5286 	destroy_workqueue(osdc->notify_wq);
5287 	cancel_delayed_work_sync(&osdc->timeout_work);
5288 	cancel_delayed_work_sync(&osdc->osds_timeout_work);
5289 
5290 	down_write(&osdc->lock);
5291 	while (!RB_EMPTY_ROOT(&osdc->osds)) {
5292 		struct ceph_osd *osd = rb_entry(rb_first(&osdc->osds),
5293 						struct ceph_osd, o_node);
5294 		close_osd(osd);
5295 	}
5296 	up_write(&osdc->lock);
5297 	WARN_ON(refcount_read(&osdc->homeless_osd.o_ref) != 1);
5298 	osd_cleanup(&osdc->homeless_osd);
5299 
5300 	WARN_ON(!list_empty(&osdc->osd_lru));
5301 	WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_requests));
5302 	WARN_ON(!RB_EMPTY_ROOT(&osdc->map_checks));
5303 	WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_map_checks));
5304 	WARN_ON(atomic_read(&osdc->num_requests));
5305 	WARN_ON(atomic_read(&osdc->num_homeless));
5306 
5307 	ceph_osdmap_destroy(osdc->osdmap);
5308 	mempool_destroy(osdc->req_mempool);
5309 	ceph_msgpool_destroy(&osdc->msgpool_op);
5310 	ceph_msgpool_destroy(&osdc->msgpool_op_reply);
5311 }
5312 
osd_req_op_copy_from_init(struct ceph_osd_request * req,u64 src_snapid,u64 src_version,struct ceph_object_id * src_oid,struct ceph_object_locator * src_oloc,u32 src_fadvise_flags,u32 dst_fadvise_flags,u32 truncate_seq,u64 truncate_size,u8 copy_from_flags)5313 static int osd_req_op_copy_from_init(struct ceph_osd_request *req,
5314 				     u64 src_snapid, u64 src_version,
5315 				     struct ceph_object_id *src_oid,
5316 				     struct ceph_object_locator *src_oloc,
5317 				     u32 src_fadvise_flags,
5318 				     u32 dst_fadvise_flags,
5319 				     u32 truncate_seq, u64 truncate_size,
5320 				     u8 copy_from_flags)
5321 {
5322 	struct ceph_osd_req_op *op;
5323 	struct page **pages;
5324 	void *p, *end;
5325 
5326 	pages = ceph_alloc_page_vector(1, GFP_KERNEL);
5327 	if (IS_ERR(pages))
5328 		return PTR_ERR(pages);
5329 
5330 	op = osd_req_op_init(req, 0, CEPH_OSD_OP_COPY_FROM2,
5331 			     dst_fadvise_flags);
5332 	op->copy_from.snapid = src_snapid;
5333 	op->copy_from.src_version = src_version;
5334 	op->copy_from.flags = copy_from_flags;
5335 	op->copy_from.src_fadvise_flags = src_fadvise_flags;
5336 
5337 	p = page_address(pages[0]);
5338 	end = p + PAGE_SIZE;
5339 	ceph_encode_string(&p, end, src_oid->name, src_oid->name_len);
5340 	encode_oloc(&p, end, src_oloc);
5341 	ceph_encode_32(&p, truncate_seq);
5342 	ceph_encode_64(&p, truncate_size);
5343 	op->indata_len = PAGE_SIZE - (end - p);
5344 
5345 	ceph_osd_data_pages_init(&op->copy_from.osd_data, pages,
5346 				 op->indata_len, 0, false, true);
5347 	return 0;
5348 }
5349 
ceph_osdc_copy_from(struct ceph_osd_client * osdc,u64 src_snapid,u64 src_version,struct ceph_object_id * src_oid,struct ceph_object_locator * src_oloc,u32 src_fadvise_flags,struct ceph_object_id * dst_oid,struct ceph_object_locator * dst_oloc,u32 dst_fadvise_flags,u32 truncate_seq,u64 truncate_size,u8 copy_from_flags)5350 int ceph_osdc_copy_from(struct ceph_osd_client *osdc,
5351 			u64 src_snapid, u64 src_version,
5352 			struct ceph_object_id *src_oid,
5353 			struct ceph_object_locator *src_oloc,
5354 			u32 src_fadvise_flags,
5355 			struct ceph_object_id *dst_oid,
5356 			struct ceph_object_locator *dst_oloc,
5357 			u32 dst_fadvise_flags,
5358 			u32 truncate_seq, u64 truncate_size,
5359 			u8 copy_from_flags)
5360 {
5361 	struct ceph_osd_request *req;
5362 	int ret;
5363 
5364 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_KERNEL);
5365 	if (!req)
5366 		return -ENOMEM;
5367 
5368 	req->r_flags = CEPH_OSD_FLAG_WRITE;
5369 
5370 	ceph_oloc_copy(&req->r_t.base_oloc, dst_oloc);
5371 	ceph_oid_copy(&req->r_t.base_oid, dst_oid);
5372 
5373 	ret = osd_req_op_copy_from_init(req, src_snapid, src_version, src_oid,
5374 					src_oloc, src_fadvise_flags,
5375 					dst_fadvise_flags, truncate_seq,
5376 					truncate_size, copy_from_flags);
5377 	if (ret)
5378 		goto out;
5379 
5380 	ret = ceph_osdc_alloc_messages(req, GFP_KERNEL);
5381 	if (ret)
5382 		goto out;
5383 
5384 	ceph_osdc_start_request(osdc, req, false);
5385 	ret = ceph_osdc_wait_request(osdc, req);
5386 
5387 out:
5388 	ceph_osdc_put_request(req);
5389 	return ret;
5390 }
5391 EXPORT_SYMBOL(ceph_osdc_copy_from);
5392 
ceph_osdc_setup(void)5393 int __init ceph_osdc_setup(void)
5394 {
5395 	size_t size = sizeof(struct ceph_osd_request) +
5396 	    CEPH_OSD_SLAB_OPS * sizeof(struct ceph_osd_req_op);
5397 
5398 	BUG_ON(ceph_osd_request_cache);
5399 	ceph_osd_request_cache = kmem_cache_create("ceph_osd_request", size,
5400 						   0, 0, NULL);
5401 
5402 	return ceph_osd_request_cache ? 0 : -ENOMEM;
5403 }
5404 
ceph_osdc_cleanup(void)5405 void ceph_osdc_cleanup(void)
5406 {
5407 	BUG_ON(!ceph_osd_request_cache);
5408 	kmem_cache_destroy(ceph_osd_request_cache);
5409 	ceph_osd_request_cache = NULL;
5410 }
5411 
5412 /*
5413  * handle incoming message
5414  */
osd_dispatch(struct ceph_connection * con,struct ceph_msg * msg)5415 static void osd_dispatch(struct ceph_connection *con, struct ceph_msg *msg)
5416 {
5417 	struct ceph_osd *osd = con->private;
5418 	struct ceph_osd_client *osdc = osd->o_osdc;
5419 	int type = le16_to_cpu(msg->hdr.type);
5420 
5421 	switch (type) {
5422 	case CEPH_MSG_OSD_MAP:
5423 		ceph_osdc_handle_map(osdc, msg);
5424 		break;
5425 	case CEPH_MSG_OSD_OPREPLY:
5426 		handle_reply(osd, msg);
5427 		break;
5428 	case CEPH_MSG_OSD_BACKOFF:
5429 		handle_backoff(osd, msg);
5430 		break;
5431 	case CEPH_MSG_WATCH_NOTIFY:
5432 		handle_watch_notify(osdc, msg);
5433 		break;
5434 
5435 	default:
5436 		pr_err("received unknown message type %d %s\n", type,
5437 		       ceph_msg_type_name(type));
5438 	}
5439 
5440 	ceph_msg_put(msg);
5441 }
5442 
5443 /*
5444  * Lookup and return message for incoming reply.  Don't try to do
5445  * anything about a larger than preallocated data portion of the
5446  * message at the moment - for now, just skip the message.
5447  */
get_reply(struct ceph_connection * con,struct ceph_msg_header * hdr,int * skip)5448 static struct ceph_msg *get_reply(struct ceph_connection *con,
5449 				  struct ceph_msg_header *hdr,
5450 				  int *skip)
5451 {
5452 	struct ceph_osd *osd = con->private;
5453 	struct ceph_osd_client *osdc = osd->o_osdc;
5454 	struct ceph_msg *m = NULL;
5455 	struct ceph_osd_request *req;
5456 	int front_len = le32_to_cpu(hdr->front_len);
5457 	int data_len = le32_to_cpu(hdr->data_len);
5458 	u64 tid = le64_to_cpu(hdr->tid);
5459 
5460 	down_read(&osdc->lock);
5461 	if (!osd_registered(osd)) {
5462 		dout("%s osd%d unknown, skipping\n", __func__, osd->o_osd);
5463 		*skip = 1;
5464 		goto out_unlock_osdc;
5465 	}
5466 	WARN_ON(osd->o_osd != le64_to_cpu(hdr->src.num));
5467 
5468 	mutex_lock(&osd->lock);
5469 	req = lookup_request(&osd->o_requests, tid);
5470 	if (!req) {
5471 		dout("%s osd%d tid %llu unknown, skipping\n", __func__,
5472 		     osd->o_osd, tid);
5473 		*skip = 1;
5474 		goto out_unlock_session;
5475 	}
5476 
5477 	ceph_msg_revoke_incoming(req->r_reply);
5478 
5479 	if (front_len > req->r_reply->front_alloc_len) {
5480 		pr_warn("%s osd%d tid %llu front %d > preallocated %d\n",
5481 			__func__, osd->o_osd, req->r_tid, front_len,
5482 			req->r_reply->front_alloc_len);
5483 		m = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, front_len, GFP_NOFS,
5484 				 false);
5485 		if (!m)
5486 			goto out_unlock_session;
5487 		ceph_msg_put(req->r_reply);
5488 		req->r_reply = m;
5489 	}
5490 
5491 	if (data_len > req->r_reply->data_length) {
5492 		pr_warn("%s osd%d tid %llu data %d > preallocated %zu, skipping\n",
5493 			__func__, osd->o_osd, req->r_tid, data_len,
5494 			req->r_reply->data_length);
5495 		m = NULL;
5496 		*skip = 1;
5497 		goto out_unlock_session;
5498 	}
5499 
5500 	m = ceph_msg_get(req->r_reply);
5501 	dout("get_reply tid %lld %p\n", tid, m);
5502 
5503 out_unlock_session:
5504 	mutex_unlock(&osd->lock);
5505 out_unlock_osdc:
5506 	up_read(&osdc->lock);
5507 	return m;
5508 }
5509 
alloc_msg_with_page_vector(struct ceph_msg_header * hdr)5510 static struct ceph_msg *alloc_msg_with_page_vector(struct ceph_msg_header *hdr)
5511 {
5512 	struct ceph_msg *m;
5513 	int type = le16_to_cpu(hdr->type);
5514 	u32 front_len = le32_to_cpu(hdr->front_len);
5515 	u32 data_len = le32_to_cpu(hdr->data_len);
5516 
5517 	m = ceph_msg_new2(type, front_len, 1, GFP_NOIO, false);
5518 	if (!m)
5519 		return NULL;
5520 
5521 	if (data_len) {
5522 		struct page **pages;
5523 
5524 		pages = ceph_alloc_page_vector(calc_pages_for(0, data_len),
5525 					       GFP_NOIO);
5526 		if (IS_ERR(pages)) {
5527 			ceph_msg_put(m);
5528 			return NULL;
5529 		}
5530 
5531 		ceph_msg_data_add_pages(m, pages, data_len, 0, true);
5532 	}
5533 
5534 	return m;
5535 }
5536 
osd_alloc_msg(struct ceph_connection * con,struct ceph_msg_header * hdr,int * skip)5537 static struct ceph_msg *osd_alloc_msg(struct ceph_connection *con,
5538 				      struct ceph_msg_header *hdr,
5539 				      int *skip)
5540 {
5541 	struct ceph_osd *osd = con->private;
5542 	int type = le16_to_cpu(hdr->type);
5543 
5544 	*skip = 0;
5545 	switch (type) {
5546 	case CEPH_MSG_OSD_MAP:
5547 	case CEPH_MSG_OSD_BACKOFF:
5548 	case CEPH_MSG_WATCH_NOTIFY:
5549 		return alloc_msg_with_page_vector(hdr);
5550 	case CEPH_MSG_OSD_OPREPLY:
5551 		return get_reply(con, hdr, skip);
5552 	default:
5553 		pr_warn("%s osd%d unknown msg type %d, skipping\n", __func__,
5554 			osd->o_osd, type);
5555 		*skip = 1;
5556 		return NULL;
5557 	}
5558 }
5559 
5560 /*
5561  * Wrappers to refcount containing ceph_osd struct
5562  */
osd_get_con(struct ceph_connection * con)5563 static struct ceph_connection *osd_get_con(struct ceph_connection *con)
5564 {
5565 	struct ceph_osd *osd = con->private;
5566 	if (get_osd(osd))
5567 		return con;
5568 	return NULL;
5569 }
5570 
osd_put_con(struct ceph_connection * con)5571 static void osd_put_con(struct ceph_connection *con)
5572 {
5573 	struct ceph_osd *osd = con->private;
5574 	put_osd(osd);
5575 }
5576 
5577 /*
5578  * authentication
5579  */
5580 
5581 /*
5582  * Note: returned pointer is the address of a structure that's
5583  * managed separately.  Caller must *not* attempt to free it.
5584  */
5585 static struct ceph_auth_handshake *
osd_get_authorizer(struct ceph_connection * con,int * proto,int force_new)5586 osd_get_authorizer(struct ceph_connection *con, int *proto, int force_new)
5587 {
5588 	struct ceph_osd *o = con->private;
5589 	struct ceph_osd_client *osdc = o->o_osdc;
5590 	struct ceph_auth_client *ac = osdc->client->monc.auth;
5591 	struct ceph_auth_handshake *auth = &o->o_auth;
5592 	int ret;
5593 
5594 	ret = __ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_OSD,
5595 					 force_new, proto, NULL, NULL);
5596 	if (ret)
5597 		return ERR_PTR(ret);
5598 
5599 	return auth;
5600 }
5601 
osd_add_authorizer_challenge(struct ceph_connection * con,void * challenge_buf,int challenge_buf_len)5602 static int osd_add_authorizer_challenge(struct ceph_connection *con,
5603 				    void *challenge_buf, int challenge_buf_len)
5604 {
5605 	struct ceph_osd *o = con->private;
5606 	struct ceph_osd_client *osdc = o->o_osdc;
5607 	struct ceph_auth_client *ac = osdc->client->monc.auth;
5608 
5609 	return ceph_auth_add_authorizer_challenge(ac, o->o_auth.authorizer,
5610 					    challenge_buf, challenge_buf_len);
5611 }
5612 
osd_verify_authorizer_reply(struct ceph_connection * con)5613 static int osd_verify_authorizer_reply(struct ceph_connection *con)
5614 {
5615 	struct ceph_osd *o = con->private;
5616 	struct ceph_osd_client *osdc = o->o_osdc;
5617 	struct ceph_auth_client *ac = osdc->client->monc.auth;
5618 	struct ceph_auth_handshake *auth = &o->o_auth;
5619 
5620 	return ceph_auth_verify_authorizer_reply(ac, auth->authorizer,
5621 		auth->authorizer_reply_buf, auth->authorizer_reply_buf_len,
5622 		NULL, NULL, NULL, NULL);
5623 }
5624 
osd_invalidate_authorizer(struct ceph_connection * con)5625 static int osd_invalidate_authorizer(struct ceph_connection *con)
5626 {
5627 	struct ceph_osd *o = con->private;
5628 	struct ceph_osd_client *osdc = o->o_osdc;
5629 	struct ceph_auth_client *ac = osdc->client->monc.auth;
5630 
5631 	ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_OSD);
5632 	return ceph_monc_validate_auth(&osdc->client->monc);
5633 }
5634 
osd_get_auth_request(struct ceph_connection * con,void * buf,int * buf_len,void ** authorizer,int * authorizer_len)5635 static int osd_get_auth_request(struct ceph_connection *con,
5636 				void *buf, int *buf_len,
5637 				void **authorizer, int *authorizer_len)
5638 {
5639 	struct ceph_osd *o = con->private;
5640 	struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5641 	struct ceph_auth_handshake *auth = &o->o_auth;
5642 	int ret;
5643 
5644 	ret = ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_OSD,
5645 				       buf, buf_len);
5646 	if (ret)
5647 		return ret;
5648 
5649 	*authorizer = auth->authorizer_buf;
5650 	*authorizer_len = auth->authorizer_buf_len;
5651 	return 0;
5652 }
5653 
osd_handle_auth_reply_more(struct ceph_connection * con,void * reply,int reply_len,void * buf,int * buf_len,void ** authorizer,int * authorizer_len)5654 static int osd_handle_auth_reply_more(struct ceph_connection *con,
5655 				      void *reply, int reply_len,
5656 				      void *buf, int *buf_len,
5657 				      void **authorizer, int *authorizer_len)
5658 {
5659 	struct ceph_osd *o = con->private;
5660 	struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5661 	struct ceph_auth_handshake *auth = &o->o_auth;
5662 	int ret;
5663 
5664 	ret = ceph_auth_handle_svc_reply_more(ac, auth, reply, reply_len,
5665 					      buf, buf_len);
5666 	if (ret)
5667 		return ret;
5668 
5669 	*authorizer = auth->authorizer_buf;
5670 	*authorizer_len = auth->authorizer_buf_len;
5671 	return 0;
5672 }
5673 
osd_handle_auth_done(struct ceph_connection * con,u64 global_id,void * reply,int reply_len,u8 * session_key,int * session_key_len,u8 * con_secret,int * con_secret_len)5674 static int osd_handle_auth_done(struct ceph_connection *con,
5675 				u64 global_id, void *reply, int reply_len,
5676 				u8 *session_key, int *session_key_len,
5677 				u8 *con_secret, int *con_secret_len)
5678 {
5679 	struct ceph_osd *o = con->private;
5680 	struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5681 	struct ceph_auth_handshake *auth = &o->o_auth;
5682 
5683 	return ceph_auth_handle_svc_reply_done(ac, auth, reply, reply_len,
5684 					       session_key, session_key_len,
5685 					       con_secret, con_secret_len);
5686 }
5687 
osd_handle_auth_bad_method(struct ceph_connection * con,int used_proto,int result,const int * allowed_protos,int proto_cnt,const int * allowed_modes,int mode_cnt)5688 static int osd_handle_auth_bad_method(struct ceph_connection *con,
5689 				      int used_proto, int result,
5690 				      const int *allowed_protos, int proto_cnt,
5691 				      const int *allowed_modes, int mode_cnt)
5692 {
5693 	struct ceph_osd *o = con->private;
5694 	struct ceph_mon_client *monc = &o->o_osdc->client->monc;
5695 	int ret;
5696 
5697 	if (ceph_auth_handle_bad_authorizer(monc->auth, CEPH_ENTITY_TYPE_OSD,
5698 					    used_proto, result,
5699 					    allowed_protos, proto_cnt,
5700 					    allowed_modes, mode_cnt)) {
5701 		ret = ceph_monc_validate_auth(monc);
5702 		if (ret)
5703 			return ret;
5704 	}
5705 
5706 	return -EACCES;
5707 }
5708 
osd_reencode_message(struct ceph_msg * msg)5709 static void osd_reencode_message(struct ceph_msg *msg)
5710 {
5711 	int type = le16_to_cpu(msg->hdr.type);
5712 
5713 	if (type == CEPH_MSG_OSD_OP)
5714 		encode_request_finish(msg);
5715 }
5716 
osd_sign_message(struct ceph_msg * msg)5717 static int osd_sign_message(struct ceph_msg *msg)
5718 {
5719 	struct ceph_osd *o = msg->con->private;
5720 	struct ceph_auth_handshake *auth = &o->o_auth;
5721 
5722 	return ceph_auth_sign_message(auth, msg);
5723 }
5724 
osd_check_message_signature(struct ceph_msg * msg)5725 static int osd_check_message_signature(struct ceph_msg *msg)
5726 {
5727 	struct ceph_osd *o = msg->con->private;
5728 	struct ceph_auth_handshake *auth = &o->o_auth;
5729 
5730 	return ceph_auth_check_message_signature(auth, msg);
5731 }
5732 
5733 static const struct ceph_connection_operations osd_con_ops = {
5734 	.get = osd_get_con,
5735 	.put = osd_put_con,
5736 	.alloc_msg = osd_alloc_msg,
5737 	.dispatch = osd_dispatch,
5738 	.fault = osd_fault,
5739 	.reencode_message = osd_reencode_message,
5740 	.get_authorizer = osd_get_authorizer,
5741 	.add_authorizer_challenge = osd_add_authorizer_challenge,
5742 	.verify_authorizer_reply = osd_verify_authorizer_reply,
5743 	.invalidate_authorizer = osd_invalidate_authorizer,
5744 	.sign_message = osd_sign_message,
5745 	.check_message_signature = osd_check_message_signature,
5746 	.get_auth_request = osd_get_auth_request,
5747 	.handle_auth_reply_more = osd_handle_auth_reply_more,
5748 	.handle_auth_done = osd_handle_auth_done,
5749 	.handle_auth_bad_method = osd_handle_auth_bad_method,
5750 };
5751