xref: /linux/net/ceph/osd_client.c (revision 7684392f17b05a9881a5f05b6ca5684321598140)
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 
ceph_osd_iter_init(struct ceph_osd_data * osd_data,struct iov_iter * iter)174 static void ceph_osd_iter_init(struct ceph_osd_data *osd_data,
175 			       struct iov_iter *iter)
176 {
177 	osd_data->type = CEPH_OSD_DATA_TYPE_ITER;
178 	osd_data->iter = *iter;
179 }
180 
181 static struct ceph_osd_data *
osd_req_op_raw_data_in(struct ceph_osd_request * osd_req,unsigned int which)182 osd_req_op_raw_data_in(struct ceph_osd_request *osd_req, unsigned int which)
183 {
184 	BUG_ON(which >= osd_req->r_num_ops);
185 
186 	return &osd_req->r_ops[which].raw_data_in;
187 }
188 
189 struct ceph_osd_data *
osd_req_op_extent_osd_data(struct ceph_osd_request * osd_req,unsigned int which)190 osd_req_op_extent_osd_data(struct ceph_osd_request *osd_req,
191 			unsigned int which)
192 {
193 	return osd_req_op_data(osd_req, which, extent, osd_data);
194 }
195 EXPORT_SYMBOL(osd_req_op_extent_osd_data);
196 
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)197 void osd_req_op_raw_data_in_pages(struct ceph_osd_request *osd_req,
198 			unsigned int which, struct page **pages,
199 			u64 length, u32 alignment,
200 			bool pages_from_pool, bool own_pages)
201 {
202 	struct ceph_osd_data *osd_data;
203 
204 	osd_data = osd_req_op_raw_data_in(osd_req, which);
205 	ceph_osd_data_pages_init(osd_data, pages, length, alignment,
206 				pages_from_pool, own_pages);
207 }
208 EXPORT_SYMBOL(osd_req_op_raw_data_in_pages);
209 
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)210 void osd_req_op_extent_osd_data_pages(struct ceph_osd_request *osd_req,
211 			unsigned int which, struct page **pages,
212 			u64 length, u32 alignment,
213 			bool pages_from_pool, bool own_pages)
214 {
215 	struct ceph_osd_data *osd_data;
216 
217 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
218 	ceph_osd_data_pages_init(osd_data, pages, length, alignment,
219 				pages_from_pool, own_pages);
220 }
221 EXPORT_SYMBOL(osd_req_op_extent_osd_data_pages);
222 
osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request * osd_req,unsigned int which,struct ceph_pagelist * pagelist)223 void osd_req_op_extent_osd_data_pagelist(struct ceph_osd_request *osd_req,
224 			unsigned int which, struct ceph_pagelist *pagelist)
225 {
226 	struct ceph_osd_data *osd_data;
227 
228 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
229 	ceph_osd_data_pagelist_init(osd_data, pagelist);
230 }
231 EXPORT_SYMBOL(osd_req_op_extent_osd_data_pagelist);
232 
233 #ifdef CONFIG_BLOCK
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)234 void osd_req_op_extent_osd_data_bio(struct ceph_osd_request *osd_req,
235 				    unsigned int which,
236 				    struct ceph_bio_iter *bio_pos,
237 				    u32 bio_length)
238 {
239 	struct ceph_osd_data *osd_data;
240 
241 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
242 	ceph_osd_data_bio_init(osd_data, bio_pos, bio_length);
243 }
244 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bio);
245 #endif /* CONFIG_BLOCK */
246 
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)247 void osd_req_op_extent_osd_data_bvecs(struct ceph_osd_request *osd_req,
248 				      unsigned int which,
249 				      struct bio_vec *bvecs, u32 num_bvecs,
250 				      u32 bytes)
251 {
252 	struct ceph_osd_data *osd_data;
253 	struct ceph_bvec_iter it = {
254 		.bvecs = bvecs,
255 		.iter = { .bi_size = bytes },
256 	};
257 
258 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
259 	ceph_osd_data_bvecs_init(osd_data, &it, num_bvecs);
260 }
261 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvecs);
262 
osd_req_op_extent_osd_data_bvec_pos(struct ceph_osd_request * osd_req,unsigned int which,struct ceph_bvec_iter * bvec_pos)263 void osd_req_op_extent_osd_data_bvec_pos(struct ceph_osd_request *osd_req,
264 					 unsigned int which,
265 					 struct ceph_bvec_iter *bvec_pos)
266 {
267 	struct ceph_osd_data *osd_data;
268 
269 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
270 	ceph_osd_data_bvecs_init(osd_data, bvec_pos, 0);
271 }
272 EXPORT_SYMBOL(osd_req_op_extent_osd_data_bvec_pos);
273 
274 /**
275  * osd_req_op_extent_osd_iter - Set up an operation with an iterator buffer
276  * @osd_req: The request to set up
277  * @which: Index of the operation in which to set the iter
278  * @iter: The buffer iterator
279  */
osd_req_op_extent_osd_iter(struct ceph_osd_request * osd_req,unsigned int which,struct iov_iter * iter)280 void osd_req_op_extent_osd_iter(struct ceph_osd_request *osd_req,
281 				unsigned int which, struct iov_iter *iter)
282 {
283 	struct ceph_osd_data *osd_data;
284 
285 	osd_data = osd_req_op_data(osd_req, which, extent, osd_data);
286 	ceph_osd_iter_init(osd_data, iter);
287 }
288 EXPORT_SYMBOL(osd_req_op_extent_osd_iter);
289 
osd_req_op_cls_request_info_pagelist(struct ceph_osd_request * osd_req,unsigned int which,struct ceph_pagelist * pagelist)290 static void osd_req_op_cls_request_info_pagelist(
291 			struct ceph_osd_request *osd_req,
292 			unsigned int which, struct ceph_pagelist *pagelist)
293 {
294 	struct ceph_osd_data *osd_data;
295 
296 	osd_data = osd_req_op_data(osd_req, which, cls, request_info);
297 	ceph_osd_data_pagelist_init(osd_data, pagelist);
298 }
299 
osd_req_op_cls_request_data_pagelist(struct ceph_osd_request * osd_req,unsigned int which,struct ceph_pagelist * pagelist)300 void osd_req_op_cls_request_data_pagelist(
301 			struct ceph_osd_request *osd_req,
302 			unsigned int which, struct ceph_pagelist *pagelist)
303 {
304 	struct ceph_osd_data *osd_data;
305 
306 	osd_data = osd_req_op_data(osd_req, which, cls, request_data);
307 	ceph_osd_data_pagelist_init(osd_data, pagelist);
308 	osd_req->r_ops[which].cls.indata_len += pagelist->length;
309 	osd_req->r_ops[which].indata_len += pagelist->length;
310 }
311 EXPORT_SYMBOL(osd_req_op_cls_request_data_pagelist);
312 
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)313 void osd_req_op_cls_request_data_pages(struct ceph_osd_request *osd_req,
314 			unsigned int which, struct page **pages, u64 length,
315 			u32 alignment, bool pages_from_pool, bool own_pages)
316 {
317 	struct ceph_osd_data *osd_data;
318 
319 	osd_data = osd_req_op_data(osd_req, which, cls, request_data);
320 	ceph_osd_data_pages_init(osd_data, pages, length, alignment,
321 				pages_from_pool, own_pages);
322 	osd_req->r_ops[which].cls.indata_len += length;
323 	osd_req->r_ops[which].indata_len += length;
324 }
325 EXPORT_SYMBOL(osd_req_op_cls_request_data_pages);
326 
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)327 void osd_req_op_cls_request_data_bvecs(struct ceph_osd_request *osd_req,
328 				       unsigned int which,
329 				       struct bio_vec *bvecs, u32 num_bvecs,
330 				       u32 bytes)
331 {
332 	struct ceph_osd_data *osd_data;
333 	struct ceph_bvec_iter it = {
334 		.bvecs = bvecs,
335 		.iter = { .bi_size = bytes },
336 	};
337 
338 	osd_data = osd_req_op_data(osd_req, which, cls, request_data);
339 	ceph_osd_data_bvecs_init(osd_data, &it, num_bvecs);
340 	osd_req->r_ops[which].cls.indata_len += bytes;
341 	osd_req->r_ops[which].indata_len += bytes;
342 }
343 EXPORT_SYMBOL(osd_req_op_cls_request_data_bvecs);
344 
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)345 void osd_req_op_cls_response_data_pages(struct ceph_osd_request *osd_req,
346 			unsigned int which, struct page **pages, u64 length,
347 			u32 alignment, bool pages_from_pool, bool own_pages)
348 {
349 	struct ceph_osd_data *osd_data;
350 
351 	osd_data = osd_req_op_data(osd_req, which, cls, response_data);
352 	ceph_osd_data_pages_init(osd_data, pages, length, alignment,
353 				pages_from_pool, own_pages);
354 }
355 EXPORT_SYMBOL(osd_req_op_cls_response_data_pages);
356 
ceph_osd_data_length(struct ceph_osd_data * osd_data)357 static u64 ceph_osd_data_length(struct ceph_osd_data *osd_data)
358 {
359 	switch (osd_data->type) {
360 	case CEPH_OSD_DATA_TYPE_NONE:
361 		return 0;
362 	case CEPH_OSD_DATA_TYPE_PAGES:
363 		return osd_data->length;
364 	case CEPH_OSD_DATA_TYPE_PAGELIST:
365 		return (u64)osd_data->pagelist->length;
366 #ifdef CONFIG_BLOCK
367 	case CEPH_OSD_DATA_TYPE_BIO:
368 		return (u64)osd_data->bio_length;
369 #endif /* CONFIG_BLOCK */
370 	case CEPH_OSD_DATA_TYPE_BVECS:
371 		return osd_data->bvec_pos.iter.bi_size;
372 	case CEPH_OSD_DATA_TYPE_ITER:
373 		return iov_iter_count(&osd_data->iter);
374 	default:
375 		WARN(true, "unrecognized data type %d\n", (int)osd_data->type);
376 		return 0;
377 	}
378 }
379 
ceph_osd_data_release(struct ceph_osd_data * osd_data)380 static void ceph_osd_data_release(struct ceph_osd_data *osd_data)
381 {
382 	if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES && osd_data->own_pages) {
383 		int num_pages;
384 
385 		num_pages = calc_pages_for((u64)osd_data->alignment,
386 						(u64)osd_data->length);
387 		ceph_release_page_vector(osd_data->pages, num_pages);
388 	} else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
389 		ceph_pagelist_release(osd_data->pagelist);
390 	}
391 	ceph_osd_data_init(osd_data);
392 }
393 
osd_req_op_data_release(struct ceph_osd_request * osd_req,unsigned int which)394 static void osd_req_op_data_release(struct ceph_osd_request *osd_req,
395 			unsigned int which)
396 {
397 	struct ceph_osd_req_op *op;
398 
399 	BUG_ON(which >= osd_req->r_num_ops);
400 	op = &osd_req->r_ops[which];
401 
402 	switch (op->op) {
403 	case CEPH_OSD_OP_READ:
404 	case CEPH_OSD_OP_SPARSE_READ:
405 	case CEPH_OSD_OP_WRITE:
406 	case CEPH_OSD_OP_WRITEFULL:
407 		kfree(op->extent.sparse_ext);
408 		ceph_osd_data_release(&op->extent.osd_data);
409 		break;
410 	case CEPH_OSD_OP_CALL:
411 		ceph_osd_data_release(&op->cls.request_info);
412 		ceph_osd_data_release(&op->cls.request_data);
413 		ceph_osd_data_release(&op->cls.response_data);
414 		break;
415 	case CEPH_OSD_OP_SETXATTR:
416 	case CEPH_OSD_OP_CMPXATTR:
417 		ceph_osd_data_release(&op->xattr.osd_data);
418 		break;
419 	case CEPH_OSD_OP_STAT:
420 		ceph_osd_data_release(&op->raw_data_in);
421 		break;
422 	case CEPH_OSD_OP_NOTIFY_ACK:
423 		ceph_osd_data_release(&op->notify_ack.request_data);
424 		break;
425 	case CEPH_OSD_OP_NOTIFY:
426 		ceph_osd_data_release(&op->notify.request_data);
427 		ceph_osd_data_release(&op->notify.response_data);
428 		break;
429 	case CEPH_OSD_OP_LIST_WATCHERS:
430 		ceph_osd_data_release(&op->list_watchers.response_data);
431 		break;
432 	case CEPH_OSD_OP_COPY_FROM2:
433 		ceph_osd_data_release(&op->copy_from.osd_data);
434 		break;
435 	default:
436 		break;
437 	}
438 }
439 
440 /*
441  * Assumes @t is zero-initialized.
442  */
target_init(struct ceph_osd_request_target * t)443 static void target_init(struct ceph_osd_request_target *t)
444 {
445 	ceph_oid_init(&t->base_oid);
446 	ceph_oloc_init(&t->base_oloc);
447 	ceph_oid_init(&t->target_oid);
448 	ceph_oloc_init(&t->target_oloc);
449 
450 	ceph_osds_init(&t->acting);
451 	ceph_osds_init(&t->up);
452 	t->size = -1;
453 	t->min_size = -1;
454 
455 	t->osd = CEPH_HOMELESS_OSD;
456 }
457 
target_copy(struct ceph_osd_request_target * dest,const struct ceph_osd_request_target * src)458 static void target_copy(struct ceph_osd_request_target *dest,
459 			const struct ceph_osd_request_target *src)
460 {
461 	ceph_oid_copy(&dest->base_oid, &src->base_oid);
462 	ceph_oloc_copy(&dest->base_oloc, &src->base_oloc);
463 	ceph_oid_copy(&dest->target_oid, &src->target_oid);
464 	ceph_oloc_copy(&dest->target_oloc, &src->target_oloc);
465 
466 	dest->pgid = src->pgid; /* struct */
467 	dest->spgid = src->spgid; /* struct */
468 	dest->pg_num = src->pg_num;
469 	dest->pg_num_mask = src->pg_num_mask;
470 	ceph_osds_copy(&dest->acting, &src->acting);
471 	ceph_osds_copy(&dest->up, &src->up);
472 	dest->size = src->size;
473 	dest->min_size = src->min_size;
474 	dest->sort_bitwise = src->sort_bitwise;
475 	dest->recovery_deletes = src->recovery_deletes;
476 
477 	dest->flags = src->flags;
478 	dest->used_replica = src->used_replica;
479 	dest->paused = src->paused;
480 
481 	dest->epoch = src->epoch;
482 	dest->last_force_resend = src->last_force_resend;
483 
484 	dest->osd = src->osd;
485 }
486 
target_destroy(struct ceph_osd_request_target * t)487 static void target_destroy(struct ceph_osd_request_target *t)
488 {
489 	ceph_oid_destroy(&t->base_oid);
490 	ceph_oloc_destroy(&t->base_oloc);
491 	ceph_oid_destroy(&t->target_oid);
492 	ceph_oloc_destroy(&t->target_oloc);
493 }
494 
495 /*
496  * requests
497  */
request_release_checks(struct ceph_osd_request * req)498 static void request_release_checks(struct ceph_osd_request *req)
499 {
500 	WARN_ON(!RB_EMPTY_NODE(&req->r_node));
501 	WARN_ON(!RB_EMPTY_NODE(&req->r_mc_node));
502 	WARN_ON(!list_empty(&req->r_private_item));
503 	WARN_ON(req->r_osd);
504 }
505 
ceph_osdc_release_request(struct kref * kref)506 static void ceph_osdc_release_request(struct kref *kref)
507 {
508 	struct ceph_osd_request *req = container_of(kref,
509 					    struct ceph_osd_request, r_kref);
510 	unsigned int which;
511 
512 	dout("%s %p (r_request %p r_reply %p)\n", __func__, req,
513 	     req->r_request, req->r_reply);
514 	request_release_checks(req);
515 
516 	if (req->r_request)
517 		ceph_msg_put(req->r_request);
518 	if (req->r_reply)
519 		ceph_msg_put(req->r_reply);
520 
521 	for (which = 0; which < req->r_num_ops; which++)
522 		osd_req_op_data_release(req, which);
523 
524 	target_destroy(&req->r_t);
525 	ceph_put_snap_context(req->r_snapc);
526 
527 	if (req->r_mempool)
528 		mempool_free(req, req->r_osdc->req_mempool);
529 	else if (req->r_num_ops <= CEPH_OSD_SLAB_OPS)
530 		kmem_cache_free(ceph_osd_request_cache, req);
531 	else
532 		kfree(req);
533 }
534 
ceph_osdc_get_request(struct ceph_osd_request * req)535 void ceph_osdc_get_request(struct ceph_osd_request *req)
536 {
537 	dout("%s %p (was %d)\n", __func__, req,
538 	     kref_read(&req->r_kref));
539 	kref_get(&req->r_kref);
540 }
541 EXPORT_SYMBOL(ceph_osdc_get_request);
542 
ceph_osdc_put_request(struct ceph_osd_request * req)543 void ceph_osdc_put_request(struct ceph_osd_request *req)
544 {
545 	if (req) {
546 		dout("%s %p (was %d)\n", __func__, req,
547 		     kref_read(&req->r_kref));
548 		kref_put(&req->r_kref, ceph_osdc_release_request);
549 	}
550 }
551 EXPORT_SYMBOL(ceph_osdc_put_request);
552 
request_init(struct ceph_osd_request * req)553 static void request_init(struct ceph_osd_request *req)
554 {
555 	/* req only, each op is zeroed in osd_req_op_init() */
556 	memset(req, 0, sizeof(*req));
557 
558 	kref_init(&req->r_kref);
559 	init_completion(&req->r_completion);
560 	RB_CLEAR_NODE(&req->r_node);
561 	RB_CLEAR_NODE(&req->r_mc_node);
562 	INIT_LIST_HEAD(&req->r_private_item);
563 
564 	target_init(&req->r_t);
565 }
566 
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)567 struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
568 					       struct ceph_snap_context *snapc,
569 					       unsigned int num_ops,
570 					       bool use_mempool,
571 					       gfp_t gfp_flags)
572 {
573 	struct ceph_osd_request *req;
574 
575 	if (use_mempool) {
576 		BUG_ON(num_ops > CEPH_OSD_SLAB_OPS);
577 		req = mempool_alloc(osdc->req_mempool, gfp_flags);
578 	} else if (num_ops <= CEPH_OSD_SLAB_OPS) {
579 		req = kmem_cache_alloc(ceph_osd_request_cache, gfp_flags);
580 	} else {
581 		BUG_ON(num_ops > CEPH_OSD_MAX_OPS);
582 		req = kmalloc(struct_size(req, r_ops, num_ops), gfp_flags);
583 	}
584 	if (unlikely(!req))
585 		return NULL;
586 
587 	request_init(req);
588 	req->r_osdc = osdc;
589 	req->r_mempool = use_mempool;
590 	req->r_num_ops = num_ops;
591 	req->r_snapid = CEPH_NOSNAP;
592 	req->r_snapc = ceph_get_snap_context(snapc);
593 
594 	dout("%s req %p\n", __func__, req);
595 	return req;
596 }
597 EXPORT_SYMBOL(ceph_osdc_alloc_request);
598 
ceph_oloc_encoding_size(const struct ceph_object_locator * oloc)599 static int ceph_oloc_encoding_size(const struct ceph_object_locator *oloc)
600 {
601 	return 8 + 4 + 4 + 4 + (oloc->pool_ns ? oloc->pool_ns->len : 0);
602 }
603 
__ceph_osdc_alloc_messages(struct ceph_osd_request * req,gfp_t gfp,int num_request_data_items,int num_reply_data_items)604 static int __ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp,
605 				      int num_request_data_items,
606 				      int num_reply_data_items)
607 {
608 	struct ceph_osd_client *osdc = req->r_osdc;
609 	struct ceph_msg *msg;
610 	int msg_size;
611 
612 	WARN_ON(req->r_request || req->r_reply);
613 	WARN_ON(ceph_oid_empty(&req->r_base_oid));
614 	WARN_ON(ceph_oloc_empty(&req->r_base_oloc));
615 
616 	/* create request message */
617 	msg_size = CEPH_ENCODING_START_BLK_LEN +
618 			CEPH_PGID_ENCODING_LEN + 1; /* spgid */
619 	msg_size += 4 + 4 + 4; /* hash, osdmap_epoch, flags */
620 	msg_size += CEPH_ENCODING_START_BLK_LEN +
621 			sizeof(struct ceph_osd_reqid); /* reqid */
622 	msg_size += sizeof(struct ceph_blkin_trace_info); /* trace */
623 	msg_size += 4 + sizeof(struct ceph_timespec); /* client_inc, mtime */
624 	msg_size += CEPH_ENCODING_START_BLK_LEN +
625 			ceph_oloc_encoding_size(&req->r_base_oloc); /* oloc */
626 	msg_size += 4 + req->r_base_oid.name_len; /* oid */
627 	msg_size += 2 + req->r_num_ops * sizeof(struct ceph_osd_op);
628 	msg_size += 8; /* snapid */
629 	msg_size += 8; /* snap_seq */
630 	msg_size += 4 + 8 * (req->r_snapc ? req->r_snapc->num_snaps : 0);
631 	msg_size += 4 + 8; /* retry_attempt, features */
632 
633 	if (req->r_mempool)
634 		msg = ceph_msgpool_get(&osdc->msgpool_op, msg_size,
635 				       num_request_data_items);
636 	else
637 		msg = ceph_msg_new2(CEPH_MSG_OSD_OP, msg_size,
638 				    num_request_data_items, gfp, true);
639 	if (!msg)
640 		return -ENOMEM;
641 
642 	memset(msg->front.iov_base, 0, msg->front.iov_len);
643 	req->r_request = msg;
644 
645 	/* create reply message */
646 	msg_size = OSD_OPREPLY_FRONT_LEN;
647 	msg_size += req->r_base_oid.name_len;
648 	msg_size += req->r_num_ops * sizeof(struct ceph_osd_op);
649 
650 	if (req->r_mempool)
651 		msg = ceph_msgpool_get(&osdc->msgpool_op_reply, msg_size,
652 				       num_reply_data_items);
653 	else
654 		msg = ceph_msg_new2(CEPH_MSG_OSD_OPREPLY, msg_size,
655 				    num_reply_data_items, gfp, true);
656 	if (!msg)
657 		return -ENOMEM;
658 
659 	req->r_reply = msg;
660 
661 	return 0;
662 }
663 
osd_req_opcode_valid(u16 opcode)664 static bool osd_req_opcode_valid(u16 opcode)
665 {
666 	switch (opcode) {
667 #define GENERATE_CASE(op, opcode, str)	case CEPH_OSD_OP_##op: return true;
668 __CEPH_FORALL_OSD_OPS(GENERATE_CASE)
669 #undef GENERATE_CASE
670 	default:
671 		return false;
672 	}
673 }
674 
get_num_data_items(struct ceph_osd_request * req,int * num_request_data_items,int * num_reply_data_items)675 static void get_num_data_items(struct ceph_osd_request *req,
676 			       int *num_request_data_items,
677 			       int *num_reply_data_items)
678 {
679 	struct ceph_osd_req_op *op;
680 
681 	*num_request_data_items = 0;
682 	*num_reply_data_items = 0;
683 
684 	for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) {
685 		switch (op->op) {
686 		/* request */
687 		case CEPH_OSD_OP_WRITE:
688 		case CEPH_OSD_OP_WRITEFULL:
689 		case CEPH_OSD_OP_SETXATTR:
690 		case CEPH_OSD_OP_CMPXATTR:
691 		case CEPH_OSD_OP_NOTIFY_ACK:
692 		case CEPH_OSD_OP_COPY_FROM2:
693 			*num_request_data_items += 1;
694 			break;
695 
696 		/* reply */
697 		case CEPH_OSD_OP_STAT:
698 		case CEPH_OSD_OP_READ:
699 		case CEPH_OSD_OP_SPARSE_READ:
700 		case CEPH_OSD_OP_LIST_WATCHERS:
701 			*num_reply_data_items += 1;
702 			break;
703 
704 		/* both */
705 		case CEPH_OSD_OP_NOTIFY:
706 			*num_request_data_items += 1;
707 			*num_reply_data_items += 1;
708 			break;
709 		case CEPH_OSD_OP_CALL:
710 			*num_request_data_items += 2;
711 			*num_reply_data_items += 1;
712 			break;
713 
714 		default:
715 			WARN_ON(!osd_req_opcode_valid(op->op));
716 			break;
717 		}
718 	}
719 }
720 
721 /*
722  * oid, oloc and OSD op opcode(s) must be filled in before this function
723  * is called.
724  */
ceph_osdc_alloc_messages(struct ceph_osd_request * req,gfp_t gfp)725 int ceph_osdc_alloc_messages(struct ceph_osd_request *req, gfp_t gfp)
726 {
727 	int num_request_data_items, num_reply_data_items;
728 
729 	get_num_data_items(req, &num_request_data_items, &num_reply_data_items);
730 	return __ceph_osdc_alloc_messages(req, gfp, num_request_data_items,
731 					  num_reply_data_items);
732 }
733 EXPORT_SYMBOL(ceph_osdc_alloc_messages);
734 
735 /*
736  * This is an osd op init function for opcodes that have no data or
737  * other information associated with them.  It also serves as a
738  * common init routine for all the other init functions, below.
739  */
740 struct ceph_osd_req_op *
osd_req_op_init(struct ceph_osd_request * osd_req,unsigned int which,u16 opcode,u32 flags)741 osd_req_op_init(struct ceph_osd_request *osd_req, unsigned int which,
742 		 u16 opcode, u32 flags)
743 {
744 	struct ceph_osd_req_op *op;
745 
746 	BUG_ON(which >= osd_req->r_num_ops);
747 	BUG_ON(!osd_req_opcode_valid(opcode));
748 
749 	op = &osd_req->r_ops[which];
750 	memset(op, 0, sizeof (*op));
751 	op->op = opcode;
752 	op->flags = flags;
753 
754 	return op;
755 }
756 EXPORT_SYMBOL(osd_req_op_init);
757 
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)758 void osd_req_op_extent_init(struct ceph_osd_request *osd_req,
759 				unsigned int which, u16 opcode,
760 				u64 offset, u64 length,
761 				u64 truncate_size, u32 truncate_seq)
762 {
763 	struct ceph_osd_req_op *op = osd_req_op_init(osd_req, which,
764 						     opcode, 0);
765 	size_t payload_len = 0;
766 
767 	BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
768 	       opcode != CEPH_OSD_OP_WRITEFULL && opcode != CEPH_OSD_OP_ZERO &&
769 	       opcode != CEPH_OSD_OP_TRUNCATE && opcode != CEPH_OSD_OP_SPARSE_READ);
770 
771 	op->extent.offset = offset;
772 	op->extent.length = length;
773 	op->extent.truncate_size = truncate_size;
774 	op->extent.truncate_seq = truncate_seq;
775 	if (opcode == CEPH_OSD_OP_WRITE || opcode == CEPH_OSD_OP_WRITEFULL)
776 		payload_len += length;
777 
778 	op->indata_len = payload_len;
779 }
780 EXPORT_SYMBOL(osd_req_op_extent_init);
781 
osd_req_op_extent_update(struct ceph_osd_request * osd_req,unsigned int which,u64 length)782 void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
783 				unsigned int which, u64 length)
784 {
785 	struct ceph_osd_req_op *op;
786 	u64 previous;
787 
788 	BUG_ON(which >= osd_req->r_num_ops);
789 	op = &osd_req->r_ops[which];
790 	previous = op->extent.length;
791 
792 	if (length == previous)
793 		return;		/* Nothing to do */
794 	BUG_ON(length > previous);
795 
796 	op->extent.length = length;
797 	if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
798 		op->indata_len -= previous - length;
799 }
800 EXPORT_SYMBOL(osd_req_op_extent_update);
801 
osd_req_op_extent_dup_last(struct ceph_osd_request * osd_req,unsigned int which,u64 offset_inc)802 void osd_req_op_extent_dup_last(struct ceph_osd_request *osd_req,
803 				unsigned int which, u64 offset_inc)
804 {
805 	struct ceph_osd_req_op *op, *prev_op;
806 
807 	BUG_ON(which + 1 >= osd_req->r_num_ops);
808 
809 	prev_op = &osd_req->r_ops[which];
810 	op = osd_req_op_init(osd_req, which + 1, prev_op->op, prev_op->flags);
811 	/* dup previous one */
812 	op->indata_len = prev_op->indata_len;
813 	op->outdata_len = prev_op->outdata_len;
814 	op->extent = prev_op->extent;
815 	/* adjust offset */
816 	op->extent.offset += offset_inc;
817 	op->extent.length -= offset_inc;
818 
819 	if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
820 		op->indata_len -= offset_inc;
821 }
822 EXPORT_SYMBOL(osd_req_op_extent_dup_last);
823 
osd_req_op_cls_init(struct ceph_osd_request * osd_req,unsigned int which,const char * class,const char * method)824 int osd_req_op_cls_init(struct ceph_osd_request *osd_req, unsigned int which,
825 			const char *class, const char *method)
826 {
827 	struct ceph_osd_req_op *op;
828 	struct ceph_pagelist *pagelist;
829 	size_t payload_len = 0;
830 	size_t size;
831 	int ret;
832 
833 	op = osd_req_op_init(osd_req, which, CEPH_OSD_OP_CALL, 0);
834 
835 	pagelist = ceph_pagelist_alloc(GFP_NOFS);
836 	if (!pagelist)
837 		return -ENOMEM;
838 
839 	op->cls.class_name = class;
840 	size = strlen(class);
841 	BUG_ON(size > (size_t) U8_MAX);
842 	op->cls.class_len = size;
843 	ret = ceph_pagelist_append(pagelist, class, size);
844 	if (ret)
845 		goto err_pagelist_free;
846 	payload_len += size;
847 
848 	op->cls.method_name = method;
849 	size = strlen(method);
850 	BUG_ON(size > (size_t) U8_MAX);
851 	op->cls.method_len = size;
852 	ret = ceph_pagelist_append(pagelist, method, size);
853 	if (ret)
854 		goto err_pagelist_free;
855 	payload_len += size;
856 
857 	osd_req_op_cls_request_info_pagelist(osd_req, which, pagelist);
858 	op->indata_len = payload_len;
859 	return 0;
860 
861 err_pagelist_free:
862 	ceph_pagelist_release(pagelist);
863 	return ret;
864 }
865 EXPORT_SYMBOL(osd_req_op_cls_init);
866 
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)867 int osd_req_op_xattr_init(struct ceph_osd_request *osd_req, unsigned int which,
868 			  u16 opcode, const char *name, const void *value,
869 			  size_t size, u8 cmp_op, u8 cmp_mode)
870 {
871 	struct ceph_osd_req_op *op = osd_req_op_init(osd_req, which,
872 						     opcode, 0);
873 	struct ceph_pagelist *pagelist;
874 	size_t payload_len;
875 	int ret;
876 
877 	BUG_ON(opcode != CEPH_OSD_OP_SETXATTR && opcode != CEPH_OSD_OP_CMPXATTR);
878 
879 	pagelist = ceph_pagelist_alloc(GFP_NOFS);
880 	if (!pagelist)
881 		return -ENOMEM;
882 
883 	payload_len = strlen(name);
884 	op->xattr.name_len = payload_len;
885 	ret = ceph_pagelist_append(pagelist, name, payload_len);
886 	if (ret)
887 		goto err_pagelist_free;
888 
889 	op->xattr.value_len = size;
890 	ret = ceph_pagelist_append(pagelist, value, size);
891 	if (ret)
892 		goto err_pagelist_free;
893 	payload_len += size;
894 
895 	op->xattr.cmp_op = cmp_op;
896 	op->xattr.cmp_mode = cmp_mode;
897 
898 	ceph_osd_data_pagelist_init(&op->xattr.osd_data, pagelist);
899 	op->indata_len = payload_len;
900 	return 0;
901 
902 err_pagelist_free:
903 	ceph_pagelist_release(pagelist);
904 	return ret;
905 }
906 EXPORT_SYMBOL(osd_req_op_xattr_init);
907 
908 /*
909  * @watch_opcode: CEPH_OSD_WATCH_OP_*
910  */
osd_req_op_watch_init(struct ceph_osd_request * req,int which,u8 watch_opcode,u64 cookie,u32 gen)911 static void osd_req_op_watch_init(struct ceph_osd_request *req, int which,
912 				  u8 watch_opcode, u64 cookie, u32 gen)
913 {
914 	struct ceph_osd_req_op *op;
915 
916 	op = osd_req_op_init(req, which, CEPH_OSD_OP_WATCH, 0);
917 	op->watch.cookie = cookie;
918 	op->watch.op = watch_opcode;
919 	op->watch.gen = gen;
920 }
921 
922 /*
923  * prot_ver, timeout and notify payload (may be empty) should already be
924  * encoded in @request_pl
925  */
osd_req_op_notify_init(struct ceph_osd_request * req,int which,u64 cookie,struct ceph_pagelist * request_pl)926 static void osd_req_op_notify_init(struct ceph_osd_request *req, int which,
927 				   u64 cookie, struct ceph_pagelist *request_pl)
928 {
929 	struct ceph_osd_req_op *op;
930 
931 	op = osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY, 0);
932 	op->notify.cookie = cookie;
933 
934 	ceph_osd_data_pagelist_init(&op->notify.request_data, request_pl);
935 	op->indata_len = request_pl->length;
936 }
937 
938 /*
939  * @flags: CEPH_OSD_OP_ALLOC_HINT_FLAG_*
940  */
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)941 void osd_req_op_alloc_hint_init(struct ceph_osd_request *osd_req,
942 				unsigned int which,
943 				u64 expected_object_size,
944 				u64 expected_write_size,
945 				u32 flags)
946 {
947 	struct ceph_osd_req_op *op;
948 
949 	op = osd_req_op_init(osd_req, which, CEPH_OSD_OP_SETALLOCHINT, 0);
950 	op->alloc_hint.expected_object_size = expected_object_size;
951 	op->alloc_hint.expected_write_size = expected_write_size;
952 	op->alloc_hint.flags = flags;
953 
954 	/*
955 	 * CEPH_OSD_OP_SETALLOCHINT op is advisory and therefore deemed
956 	 * not worth a feature bit.  Set FAILOK per-op flag to make
957 	 * sure older osds don't trip over an unsupported opcode.
958 	 */
959 	op->flags |= CEPH_OSD_OP_FLAG_FAILOK;
960 }
961 EXPORT_SYMBOL(osd_req_op_alloc_hint_init);
962 
ceph_osdc_msg_data_add(struct ceph_msg * msg,struct ceph_osd_data * osd_data)963 static void ceph_osdc_msg_data_add(struct ceph_msg *msg,
964 				struct ceph_osd_data *osd_data)
965 {
966 	u64 length = ceph_osd_data_length(osd_data);
967 
968 	if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES) {
969 		BUG_ON(length > (u64) SIZE_MAX);
970 		if (length)
971 			ceph_msg_data_add_pages(msg, osd_data->pages,
972 					length, osd_data->alignment, false);
973 	} else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
974 		BUG_ON(!length);
975 		ceph_msg_data_add_pagelist(msg, osd_data->pagelist);
976 #ifdef CONFIG_BLOCK
977 	} else if (osd_data->type == CEPH_OSD_DATA_TYPE_BIO) {
978 		ceph_msg_data_add_bio(msg, &osd_data->bio_pos, length);
979 #endif
980 	} else if (osd_data->type == CEPH_OSD_DATA_TYPE_BVECS) {
981 		ceph_msg_data_add_bvecs(msg, &osd_data->bvec_pos);
982 	} else if (osd_data->type == CEPH_OSD_DATA_TYPE_ITER) {
983 		ceph_msg_data_add_iter(msg, &osd_data->iter);
984 	} else {
985 		BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_NONE);
986 	}
987 }
988 
osd_req_encode_op(struct ceph_osd_op * dst,const struct ceph_osd_req_op * src)989 static u32 osd_req_encode_op(struct ceph_osd_op *dst,
990 			     const struct ceph_osd_req_op *src)
991 {
992 	switch (src->op) {
993 	case CEPH_OSD_OP_STAT:
994 		break;
995 	case CEPH_OSD_OP_READ:
996 	case CEPH_OSD_OP_SPARSE_READ:
997 	case CEPH_OSD_OP_WRITE:
998 	case CEPH_OSD_OP_WRITEFULL:
999 	case CEPH_OSD_OP_ZERO:
1000 	case CEPH_OSD_OP_TRUNCATE:
1001 		dst->extent.offset = cpu_to_le64(src->extent.offset);
1002 		dst->extent.length = cpu_to_le64(src->extent.length);
1003 		dst->extent.truncate_size =
1004 			cpu_to_le64(src->extent.truncate_size);
1005 		dst->extent.truncate_seq =
1006 			cpu_to_le32(src->extent.truncate_seq);
1007 		break;
1008 	case CEPH_OSD_OP_CALL:
1009 		dst->cls.class_len = src->cls.class_len;
1010 		dst->cls.method_len = src->cls.method_len;
1011 		dst->cls.indata_len = cpu_to_le32(src->cls.indata_len);
1012 		break;
1013 	case CEPH_OSD_OP_WATCH:
1014 		dst->watch.cookie = cpu_to_le64(src->watch.cookie);
1015 		dst->watch.ver = cpu_to_le64(0);
1016 		dst->watch.op = src->watch.op;
1017 		dst->watch.gen = cpu_to_le32(src->watch.gen);
1018 		break;
1019 	case CEPH_OSD_OP_NOTIFY_ACK:
1020 		break;
1021 	case CEPH_OSD_OP_NOTIFY:
1022 		dst->notify.cookie = cpu_to_le64(src->notify.cookie);
1023 		break;
1024 	case CEPH_OSD_OP_LIST_WATCHERS:
1025 		break;
1026 	case CEPH_OSD_OP_SETALLOCHINT:
1027 		dst->alloc_hint.expected_object_size =
1028 		    cpu_to_le64(src->alloc_hint.expected_object_size);
1029 		dst->alloc_hint.expected_write_size =
1030 		    cpu_to_le64(src->alloc_hint.expected_write_size);
1031 		dst->alloc_hint.flags = cpu_to_le32(src->alloc_hint.flags);
1032 		break;
1033 	case CEPH_OSD_OP_SETXATTR:
1034 	case CEPH_OSD_OP_CMPXATTR:
1035 		dst->xattr.name_len = cpu_to_le32(src->xattr.name_len);
1036 		dst->xattr.value_len = cpu_to_le32(src->xattr.value_len);
1037 		dst->xattr.cmp_op = src->xattr.cmp_op;
1038 		dst->xattr.cmp_mode = src->xattr.cmp_mode;
1039 		break;
1040 	case CEPH_OSD_OP_CREATE:
1041 	case CEPH_OSD_OP_DELETE:
1042 		break;
1043 	case CEPH_OSD_OP_COPY_FROM2:
1044 		dst->copy_from.snapid = cpu_to_le64(src->copy_from.snapid);
1045 		dst->copy_from.src_version =
1046 			cpu_to_le64(src->copy_from.src_version);
1047 		dst->copy_from.flags = src->copy_from.flags;
1048 		dst->copy_from.src_fadvise_flags =
1049 			cpu_to_le32(src->copy_from.src_fadvise_flags);
1050 		break;
1051 	case CEPH_OSD_OP_ASSERT_VER:
1052 		dst->assert_ver.unused = cpu_to_le64(0);
1053 		dst->assert_ver.ver = cpu_to_le64(src->assert_ver.ver);
1054 		break;
1055 	default:
1056 		pr_err("unsupported osd opcode %s\n",
1057 			ceph_osd_op_name(src->op));
1058 		WARN_ON(1);
1059 
1060 		return 0;
1061 	}
1062 
1063 	dst->op = cpu_to_le16(src->op);
1064 	dst->flags = cpu_to_le32(src->flags);
1065 	dst->payload_len = cpu_to_le32(src->indata_len);
1066 
1067 	return src->indata_len;
1068 }
1069 
1070 /*
1071  * build new request AND message, calculate layout, and adjust file
1072  * extent as needed.
1073  *
1074  * if the file was recently truncated, we include information about its
1075  * old and new size so that the object can be updated appropriately.  (we
1076  * avoid synchronously deleting truncated objects because it's slow.)
1077  */
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)1078 struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *osdc,
1079 					       struct ceph_file_layout *layout,
1080 					       struct ceph_vino vino,
1081 					       u64 off, u64 *plen,
1082 					       unsigned int which, int num_ops,
1083 					       int opcode, int flags,
1084 					       struct ceph_snap_context *snapc,
1085 					       u32 truncate_seq,
1086 					       u64 truncate_size,
1087 					       bool use_mempool)
1088 {
1089 	struct ceph_osd_request *req;
1090 	u64 objnum = 0;
1091 	u64 objoff = 0;
1092 	u64 objlen = 0;
1093 	int r;
1094 
1095 	BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE &&
1096 	       opcode != CEPH_OSD_OP_ZERO && opcode != CEPH_OSD_OP_TRUNCATE &&
1097 	       opcode != CEPH_OSD_OP_CREATE && opcode != CEPH_OSD_OP_DELETE &&
1098 	       opcode != CEPH_OSD_OP_SPARSE_READ);
1099 
1100 	req = ceph_osdc_alloc_request(osdc, snapc, num_ops, use_mempool,
1101 					GFP_NOFS);
1102 	if (!req) {
1103 		r = -ENOMEM;
1104 		goto fail;
1105 	}
1106 
1107 	/* calculate max write size */
1108 	r = calc_layout(layout, off, plen, &objnum, &objoff, &objlen);
1109 	if (r)
1110 		goto fail;
1111 
1112 	if (opcode == CEPH_OSD_OP_CREATE || opcode == CEPH_OSD_OP_DELETE) {
1113 		osd_req_op_init(req, which, opcode, 0);
1114 	} else {
1115 		u32 object_size = layout->object_size;
1116 		u32 object_base = off - objoff;
1117 		if (!(truncate_seq == 1 && truncate_size == -1ULL)) {
1118 			if (truncate_size <= object_base) {
1119 				truncate_size = 0;
1120 			} else {
1121 				truncate_size -= object_base;
1122 				if (truncate_size > object_size)
1123 					truncate_size = object_size;
1124 			}
1125 		}
1126 		osd_req_op_extent_init(req, which, opcode, objoff, objlen,
1127 				       truncate_size, truncate_seq);
1128 	}
1129 
1130 	req->r_base_oloc.pool = layout->pool_id;
1131 	req->r_base_oloc.pool_ns = ceph_try_get_string(layout->pool_ns);
1132 	ceph_oid_printf(&req->r_base_oid, "%llx.%08llx", vino.ino, objnum);
1133 	req->r_flags = flags | osdc->client->options->read_from_replica;
1134 
1135 	req->r_snapid = vino.snap;
1136 	if (flags & CEPH_OSD_FLAG_WRITE)
1137 		req->r_data_offset = off;
1138 
1139 	if (num_ops > 1) {
1140 		int num_req_ops, num_rep_ops;
1141 
1142 		/*
1143 		 * If this is a multi-op write request, assume that we'll need
1144 		 * request ops. If it's a multi-op read then assume we'll need
1145 		 * reply ops. Anything else and call it -EINVAL.
1146 		 */
1147 		if (flags & CEPH_OSD_FLAG_WRITE) {
1148 			num_req_ops = num_ops;
1149 			num_rep_ops = 0;
1150 		} else if (flags & CEPH_OSD_FLAG_READ) {
1151 			num_req_ops = 0;
1152 			num_rep_ops = num_ops;
1153 		} else {
1154 			r = -EINVAL;
1155 			goto fail;
1156 		}
1157 
1158 		r = __ceph_osdc_alloc_messages(req, GFP_NOFS, num_req_ops,
1159 					       num_rep_ops);
1160 	} else {
1161 		r = ceph_osdc_alloc_messages(req, GFP_NOFS);
1162 	}
1163 	if (r)
1164 		goto fail;
1165 
1166 	return req;
1167 
1168 fail:
1169 	ceph_osdc_put_request(req);
1170 	return ERR_PTR(r);
1171 }
1172 EXPORT_SYMBOL(ceph_osdc_new_request);
1173 
__ceph_alloc_sparse_ext_map(struct ceph_osd_req_op * op,int cnt)1174 int __ceph_alloc_sparse_ext_map(struct ceph_osd_req_op *op, int cnt)
1175 {
1176 	WARN_ON(op->op != CEPH_OSD_OP_SPARSE_READ);
1177 
1178 	op->extent.sparse_ext_cnt = cnt;
1179 	op->extent.sparse_ext = kmalloc_array(cnt,
1180 					      sizeof(*op->extent.sparse_ext),
1181 					      GFP_NOFS);
1182 	if (!op->extent.sparse_ext)
1183 		return -ENOMEM;
1184 	return 0;
1185 }
1186 EXPORT_SYMBOL(__ceph_alloc_sparse_ext_map);
1187 
1188 /*
1189  * We keep osd requests in an rbtree, sorted by ->r_tid.
1190  */
DEFINE_RB_FUNCS(request,struct ceph_osd_request,r_tid,r_node)1191 DEFINE_RB_FUNCS(request, struct ceph_osd_request, r_tid, r_node)
1192 DEFINE_RB_FUNCS(request_mc, struct ceph_osd_request, r_tid, r_mc_node)
1193 
1194 /*
1195  * Call @fn on each OSD request as long as @fn returns 0.
1196  */
1197 static void for_each_request(struct ceph_osd_client *osdc,
1198 			int (*fn)(struct ceph_osd_request *req, void *arg),
1199 			void *arg)
1200 {
1201 	struct rb_node *n, *p;
1202 
1203 	for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
1204 		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
1205 
1206 		for (p = rb_first(&osd->o_requests); p; ) {
1207 			struct ceph_osd_request *req =
1208 			    rb_entry(p, struct ceph_osd_request, r_node);
1209 
1210 			p = rb_next(p);
1211 			if (fn(req, arg))
1212 				return;
1213 		}
1214 	}
1215 
1216 	for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
1217 		struct ceph_osd_request *req =
1218 		    rb_entry(p, struct ceph_osd_request, r_node);
1219 
1220 		p = rb_next(p);
1221 		if (fn(req, arg))
1222 			return;
1223 	}
1224 }
1225 
osd_homeless(struct ceph_osd * osd)1226 static bool osd_homeless(struct ceph_osd *osd)
1227 {
1228 	return osd->o_osd == CEPH_HOMELESS_OSD;
1229 }
1230 
osd_registered(struct ceph_osd * osd)1231 static bool osd_registered(struct ceph_osd *osd)
1232 {
1233 	verify_osdc_locked(osd->o_osdc);
1234 
1235 	return !RB_EMPTY_NODE(&osd->o_node);
1236 }
1237 
1238 /*
1239  * Assumes @osd is zero-initialized.
1240  */
osd_init(struct ceph_osd * osd)1241 static void osd_init(struct ceph_osd *osd)
1242 {
1243 	refcount_set(&osd->o_ref, 1);
1244 	RB_CLEAR_NODE(&osd->o_node);
1245 	spin_lock_init(&osd->o_requests_lock);
1246 	osd->o_requests = RB_ROOT;
1247 	osd->o_linger_requests = RB_ROOT;
1248 	osd->o_backoff_mappings = RB_ROOT;
1249 	osd->o_backoffs_by_id = RB_ROOT;
1250 	INIT_LIST_HEAD(&osd->o_osd_lru);
1251 	INIT_LIST_HEAD(&osd->o_keepalive_item);
1252 	osd->o_incarnation = 1;
1253 	mutex_init(&osd->lock);
1254 }
1255 
ceph_init_sparse_read(struct ceph_sparse_read * sr)1256 static void ceph_init_sparse_read(struct ceph_sparse_read *sr)
1257 {
1258 	kfree(sr->sr_extent);
1259 	memset(sr, '\0', sizeof(*sr));
1260 	sr->sr_state = CEPH_SPARSE_READ_HDR;
1261 }
1262 
osd_cleanup(struct ceph_osd * osd)1263 static void osd_cleanup(struct ceph_osd *osd)
1264 {
1265 	WARN_ON(!RB_EMPTY_NODE(&osd->o_node));
1266 	WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
1267 	WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
1268 	WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoff_mappings));
1269 	WARN_ON(!RB_EMPTY_ROOT(&osd->o_backoffs_by_id));
1270 	WARN_ON(!list_empty(&osd->o_osd_lru));
1271 	WARN_ON(!list_empty(&osd->o_keepalive_item));
1272 
1273 	ceph_init_sparse_read(&osd->o_sparse_read);
1274 
1275 	if (osd->o_auth.authorizer) {
1276 		WARN_ON(osd_homeless(osd));
1277 		ceph_auth_destroy_authorizer(osd->o_auth.authorizer);
1278 	}
1279 }
1280 
1281 /*
1282  * Track open sessions with osds.
1283  */
create_osd(struct ceph_osd_client * osdc,int onum)1284 static struct ceph_osd *create_osd(struct ceph_osd_client *osdc, int onum)
1285 {
1286 	struct ceph_osd *osd;
1287 
1288 	WARN_ON(onum == CEPH_HOMELESS_OSD);
1289 
1290 	osd = kzalloc(sizeof(*osd), GFP_NOIO | __GFP_NOFAIL);
1291 	osd_init(osd);
1292 	osd->o_osdc = osdc;
1293 	osd->o_osd = onum;
1294 	osd->o_sparse_op_idx = -1;
1295 
1296 	ceph_init_sparse_read(&osd->o_sparse_read);
1297 
1298 	ceph_con_init(&osd->o_con, osd, &osd_con_ops, &osdc->client->msgr);
1299 
1300 	return osd;
1301 }
1302 
get_osd(struct ceph_osd * osd)1303 static struct ceph_osd *get_osd(struct ceph_osd *osd)
1304 {
1305 	if (refcount_inc_not_zero(&osd->o_ref)) {
1306 		dout("get_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref)-1,
1307 		     refcount_read(&osd->o_ref));
1308 		return osd;
1309 	} else {
1310 		dout("get_osd %p FAIL\n", osd);
1311 		return NULL;
1312 	}
1313 }
1314 
put_osd(struct ceph_osd * osd)1315 static void put_osd(struct ceph_osd *osd)
1316 {
1317 	dout("put_osd %p %d -> %d\n", osd, refcount_read(&osd->o_ref),
1318 	     refcount_read(&osd->o_ref) - 1);
1319 	if (refcount_dec_and_test(&osd->o_ref)) {
1320 		osd_cleanup(osd);
1321 		kfree(osd);
1322 	}
1323 }
1324 
DEFINE_RB_FUNCS(osd,struct ceph_osd,o_osd,o_node)1325 DEFINE_RB_FUNCS(osd, struct ceph_osd, o_osd, o_node)
1326 
1327 static void __move_osd_to_lru(struct ceph_osd *osd)
1328 {
1329 	struct ceph_osd_client *osdc = osd->o_osdc;
1330 
1331 	dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1332 	BUG_ON(!list_empty(&osd->o_osd_lru));
1333 
1334 	spin_lock(&osdc->osd_lru_lock);
1335 	list_add_tail(&osd->o_osd_lru, &osdc->osd_lru);
1336 	spin_unlock(&osdc->osd_lru_lock);
1337 
1338 	osd->lru_ttl = jiffies + osdc->client->options->osd_idle_ttl;
1339 }
1340 
maybe_move_osd_to_lru(struct ceph_osd * osd)1341 static void maybe_move_osd_to_lru(struct ceph_osd *osd)
1342 {
1343 	if (RB_EMPTY_ROOT(&osd->o_requests) &&
1344 	    RB_EMPTY_ROOT(&osd->o_linger_requests))
1345 		__move_osd_to_lru(osd);
1346 }
1347 
__remove_osd_from_lru(struct ceph_osd * osd)1348 static void __remove_osd_from_lru(struct ceph_osd *osd)
1349 {
1350 	struct ceph_osd_client *osdc = osd->o_osdc;
1351 
1352 	dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1353 
1354 	spin_lock(&osdc->osd_lru_lock);
1355 	if (!list_empty(&osd->o_osd_lru))
1356 		list_del_init(&osd->o_osd_lru);
1357 	spin_unlock(&osdc->osd_lru_lock);
1358 }
1359 
1360 /*
1361  * Close the connection and assign any leftover requests to the
1362  * homeless session.
1363  */
close_osd(struct ceph_osd * osd)1364 static void close_osd(struct ceph_osd *osd)
1365 {
1366 	struct ceph_osd_client *osdc = osd->o_osdc;
1367 	struct rb_node *n;
1368 
1369 	verify_osdc_wrlocked(osdc);
1370 	dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1371 
1372 	ceph_con_close(&osd->o_con);
1373 
1374 	for (n = rb_first(&osd->o_requests); n; ) {
1375 		struct ceph_osd_request *req =
1376 		    rb_entry(n, struct ceph_osd_request, r_node);
1377 
1378 		n = rb_next(n); /* unlink_request() */
1379 
1380 		dout(" reassigning req %p tid %llu\n", req, req->r_tid);
1381 		unlink_request(osd, req);
1382 		link_request(&osdc->homeless_osd, req);
1383 	}
1384 	for (n = rb_first(&osd->o_linger_requests); n; ) {
1385 		struct ceph_osd_linger_request *lreq =
1386 		    rb_entry(n, struct ceph_osd_linger_request, node);
1387 
1388 		n = rb_next(n); /* unlink_linger() */
1389 
1390 		dout(" reassigning lreq %p linger_id %llu\n", lreq,
1391 		     lreq->linger_id);
1392 		unlink_linger(osd, lreq);
1393 		link_linger(&osdc->homeless_osd, lreq);
1394 	}
1395 	clear_backoffs(osd);
1396 
1397 	__remove_osd_from_lru(osd);
1398 	erase_osd(&osdc->osds, osd);
1399 	put_osd(osd);
1400 }
1401 
1402 /*
1403  * reset osd connect
1404  */
reopen_osd(struct ceph_osd * osd)1405 static int reopen_osd(struct ceph_osd *osd)
1406 {
1407 	struct ceph_entity_addr *peer_addr;
1408 
1409 	dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
1410 
1411 	if (RB_EMPTY_ROOT(&osd->o_requests) &&
1412 	    RB_EMPTY_ROOT(&osd->o_linger_requests)) {
1413 		close_osd(osd);
1414 		return -ENODEV;
1415 	}
1416 
1417 	peer_addr = &osd->o_osdc->osdmap->osd_addr[osd->o_osd];
1418 	if (!memcmp(peer_addr, &osd->o_con.peer_addr, sizeof (*peer_addr)) &&
1419 			!ceph_con_opened(&osd->o_con)) {
1420 		struct rb_node *n;
1421 
1422 		dout("osd addr hasn't changed and connection never opened, "
1423 		     "letting msgr retry\n");
1424 		/* touch each r_stamp for handle_timeout()'s benfit */
1425 		for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
1426 			struct ceph_osd_request *req =
1427 			    rb_entry(n, struct ceph_osd_request, r_node);
1428 			req->r_stamp = jiffies;
1429 		}
1430 
1431 		return -EAGAIN;
1432 	}
1433 
1434 	ceph_con_close(&osd->o_con);
1435 	ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd, peer_addr);
1436 	osd->o_incarnation++;
1437 
1438 	return 0;
1439 }
1440 
lookup_create_osd(struct ceph_osd_client * osdc,int o,bool wrlocked)1441 static struct ceph_osd *lookup_create_osd(struct ceph_osd_client *osdc, int o,
1442 					  bool wrlocked)
1443 {
1444 	struct ceph_osd *osd;
1445 
1446 	if (wrlocked)
1447 		verify_osdc_wrlocked(osdc);
1448 	else
1449 		verify_osdc_locked(osdc);
1450 
1451 	if (o != CEPH_HOMELESS_OSD)
1452 		osd = lookup_osd(&osdc->osds, o);
1453 	else
1454 		osd = &osdc->homeless_osd;
1455 	if (!osd) {
1456 		if (!wrlocked)
1457 			return ERR_PTR(-EAGAIN);
1458 
1459 		osd = create_osd(osdc, o);
1460 		insert_osd(&osdc->osds, osd);
1461 		ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd,
1462 			      &osdc->osdmap->osd_addr[osd->o_osd]);
1463 	}
1464 
1465 	dout("%s osdc %p osd%d -> osd %p\n", __func__, osdc, o, osd);
1466 	return osd;
1467 }
1468 
1469 /*
1470  * Create request <-> OSD session relation.
1471  *
1472  * @req has to be assigned a tid, @osd may be homeless.
1473  */
link_request(struct ceph_osd * osd,struct ceph_osd_request * req)1474 static void link_request(struct ceph_osd *osd, struct ceph_osd_request *req)
1475 {
1476 	verify_osd_locked(osd);
1477 	WARN_ON(!req->r_tid || req->r_osd);
1478 	dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1479 	     req, req->r_tid);
1480 
1481 	if (!osd_homeless(osd))
1482 		__remove_osd_from_lru(osd);
1483 	else
1484 		atomic_inc(&osd->o_osdc->num_homeless);
1485 
1486 	get_osd(osd);
1487 	spin_lock(&osd->o_requests_lock);
1488 	insert_request(&osd->o_requests, req);
1489 	spin_unlock(&osd->o_requests_lock);
1490 	req->r_osd = osd;
1491 }
1492 
unlink_request(struct ceph_osd * osd,struct ceph_osd_request * req)1493 static void unlink_request(struct ceph_osd *osd, struct ceph_osd_request *req)
1494 {
1495 	verify_osd_locked(osd);
1496 	WARN_ON(req->r_osd != osd);
1497 	dout("%s osd %p osd%d req %p tid %llu\n", __func__, osd, osd->o_osd,
1498 	     req, req->r_tid);
1499 
1500 	req->r_osd = NULL;
1501 	spin_lock(&osd->o_requests_lock);
1502 	erase_request(&osd->o_requests, req);
1503 	spin_unlock(&osd->o_requests_lock);
1504 	put_osd(osd);
1505 
1506 	if (!osd_homeless(osd))
1507 		maybe_move_osd_to_lru(osd);
1508 	else
1509 		atomic_dec(&osd->o_osdc->num_homeless);
1510 }
1511 
__pool_full(struct ceph_pg_pool_info * pi)1512 static bool __pool_full(struct ceph_pg_pool_info *pi)
1513 {
1514 	return pi->flags & CEPH_POOL_FLAG_FULL;
1515 }
1516 
have_pool_full(struct ceph_osd_client * osdc)1517 static bool have_pool_full(struct ceph_osd_client *osdc)
1518 {
1519 	struct rb_node *n;
1520 
1521 	for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
1522 		struct ceph_pg_pool_info *pi =
1523 		    rb_entry(n, struct ceph_pg_pool_info, node);
1524 
1525 		if (__pool_full(pi))
1526 			return true;
1527 	}
1528 
1529 	return false;
1530 }
1531 
pool_full(struct ceph_osd_client * osdc,s64 pool_id)1532 static bool pool_full(struct ceph_osd_client *osdc, s64 pool_id)
1533 {
1534 	struct ceph_pg_pool_info *pi;
1535 
1536 	pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
1537 	if (!pi)
1538 		return false;
1539 
1540 	return __pool_full(pi);
1541 }
1542 
1543 /*
1544  * Returns whether a request should be blocked from being sent
1545  * based on the current osdmap and osd_client settings.
1546  */
target_should_be_paused(struct ceph_osd_client * osdc,const struct ceph_osd_request_target * t,struct ceph_pg_pool_info * pi)1547 static bool target_should_be_paused(struct ceph_osd_client *osdc,
1548 				    const struct ceph_osd_request_target *t,
1549 				    struct ceph_pg_pool_info *pi)
1550 {
1551 	bool pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
1552 	bool pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
1553 		       ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
1554 		       __pool_full(pi);
1555 
1556 	WARN_ON(pi->id != t->target_oloc.pool);
1557 	return ((t->flags & CEPH_OSD_FLAG_READ) && pauserd) ||
1558 	       ((t->flags & CEPH_OSD_FLAG_WRITE) && pausewr) ||
1559 	       (osdc->osdmap->epoch < osdc->epoch_barrier);
1560 }
1561 
pick_random_replica(const struct ceph_osds * acting)1562 static int pick_random_replica(const struct ceph_osds *acting)
1563 {
1564 	int i = get_random_u32_below(acting->size);
1565 
1566 	dout("%s picked osd%d, primary osd%d\n", __func__,
1567 	     acting->osds[i], acting->primary);
1568 	return i;
1569 }
1570 
1571 /*
1572  * Picks the closest replica based on client's location given by
1573  * crush_location option.  Prefers the primary if the locality is
1574  * the same.
1575  */
pick_closest_replica(struct ceph_osd_client * osdc,const struct ceph_osds * acting)1576 static int pick_closest_replica(struct ceph_osd_client *osdc,
1577 				const struct ceph_osds *acting)
1578 {
1579 	struct ceph_options *opt = osdc->client->options;
1580 	int best_i, best_locality;
1581 	int i = 0, locality;
1582 
1583 	do {
1584 		locality = ceph_get_crush_locality(osdc->osdmap,
1585 						   acting->osds[i],
1586 						   &opt->crush_locs);
1587 		if (i == 0 ||
1588 		    (locality >= 0 && best_locality < 0) ||
1589 		    (locality >= 0 && best_locality >= 0 &&
1590 		     locality < best_locality)) {
1591 			best_i = i;
1592 			best_locality = locality;
1593 		}
1594 	} while (++i < acting->size);
1595 
1596 	dout("%s picked osd%d with locality %d, primary osd%d\n", __func__,
1597 	     acting->osds[best_i], best_locality, acting->primary);
1598 	return best_i;
1599 }
1600 
1601 enum calc_target_result {
1602 	CALC_TARGET_NO_ACTION = 0,
1603 	CALC_TARGET_NEED_RESEND,
1604 	CALC_TARGET_POOL_DNE,
1605 };
1606 
calc_target(struct ceph_osd_client * osdc,struct ceph_osd_request_target * t,bool any_change)1607 static enum calc_target_result calc_target(struct ceph_osd_client *osdc,
1608 					   struct ceph_osd_request_target *t,
1609 					   bool any_change)
1610 {
1611 	struct ceph_pg_pool_info *pi;
1612 	struct ceph_pg pgid, last_pgid;
1613 	struct ceph_osds up, acting;
1614 	bool is_read = t->flags & CEPH_OSD_FLAG_READ;
1615 	bool is_write = t->flags & CEPH_OSD_FLAG_WRITE;
1616 	bool force_resend = false;
1617 	bool unpaused = false;
1618 	bool legacy_change = false;
1619 	bool split = false;
1620 	bool sort_bitwise = ceph_osdmap_flag(osdc, CEPH_OSDMAP_SORTBITWISE);
1621 	bool recovery_deletes = ceph_osdmap_flag(osdc,
1622 						 CEPH_OSDMAP_RECOVERY_DELETES);
1623 	enum calc_target_result ct_res;
1624 
1625 	t->epoch = osdc->osdmap->epoch;
1626 	pi = ceph_pg_pool_by_id(osdc->osdmap, t->base_oloc.pool);
1627 	if (!pi) {
1628 		t->osd = CEPH_HOMELESS_OSD;
1629 		ct_res = CALC_TARGET_POOL_DNE;
1630 		goto out;
1631 	}
1632 
1633 	if (osdc->osdmap->epoch == pi->last_force_request_resend) {
1634 		if (t->last_force_resend < pi->last_force_request_resend) {
1635 			t->last_force_resend = pi->last_force_request_resend;
1636 			force_resend = true;
1637 		} else if (t->last_force_resend == 0) {
1638 			force_resend = true;
1639 		}
1640 	}
1641 
1642 	/* apply tiering */
1643 	ceph_oid_copy(&t->target_oid, &t->base_oid);
1644 	ceph_oloc_copy(&t->target_oloc, &t->base_oloc);
1645 	if ((t->flags & CEPH_OSD_FLAG_IGNORE_OVERLAY) == 0) {
1646 		if (is_read && pi->read_tier >= 0)
1647 			t->target_oloc.pool = pi->read_tier;
1648 		if (is_write && pi->write_tier >= 0)
1649 			t->target_oloc.pool = pi->write_tier;
1650 
1651 		pi = ceph_pg_pool_by_id(osdc->osdmap, t->target_oloc.pool);
1652 		if (!pi) {
1653 			t->osd = CEPH_HOMELESS_OSD;
1654 			ct_res = CALC_TARGET_POOL_DNE;
1655 			goto out;
1656 		}
1657 	}
1658 
1659 	__ceph_object_locator_to_pg(pi, &t->target_oid, &t->target_oloc, &pgid);
1660 	last_pgid.pool = pgid.pool;
1661 	last_pgid.seed = ceph_stable_mod(pgid.seed, t->pg_num, t->pg_num_mask);
1662 
1663 	ceph_pg_to_up_acting_osds(osdc->osdmap, pi, &pgid, &up, &acting);
1664 	if (any_change &&
1665 	    ceph_is_new_interval(&t->acting,
1666 				 &acting,
1667 				 &t->up,
1668 				 &up,
1669 				 t->size,
1670 				 pi->size,
1671 				 t->min_size,
1672 				 pi->min_size,
1673 				 t->pg_num,
1674 				 pi->pg_num,
1675 				 t->sort_bitwise,
1676 				 sort_bitwise,
1677 				 t->recovery_deletes,
1678 				 recovery_deletes,
1679 				 &last_pgid))
1680 		force_resend = true;
1681 
1682 	if (t->paused && !target_should_be_paused(osdc, t, pi)) {
1683 		t->paused = false;
1684 		unpaused = true;
1685 	}
1686 	legacy_change = ceph_pg_compare(&t->pgid, &pgid) ||
1687 			ceph_osds_changed(&t->acting, &acting,
1688 					  t->used_replica || any_change);
1689 	if (t->pg_num)
1690 		split = ceph_pg_is_split(&last_pgid, t->pg_num, pi->pg_num);
1691 
1692 	if (legacy_change || force_resend || split) {
1693 		t->pgid = pgid; /* struct */
1694 		ceph_pg_to_primary_shard(osdc->osdmap, pi, &pgid, &t->spgid);
1695 		ceph_osds_copy(&t->acting, &acting);
1696 		ceph_osds_copy(&t->up, &up);
1697 		t->size = pi->size;
1698 		t->min_size = pi->min_size;
1699 		t->pg_num = pi->pg_num;
1700 		t->pg_num_mask = pi->pg_num_mask;
1701 		t->sort_bitwise = sort_bitwise;
1702 		t->recovery_deletes = recovery_deletes;
1703 
1704 		if ((t->flags & (CEPH_OSD_FLAG_BALANCE_READS |
1705 				 CEPH_OSD_FLAG_LOCALIZE_READS)) &&
1706 		    !is_write && pi->type == CEPH_POOL_TYPE_REP &&
1707 		    acting.size > 1) {
1708 			int pos;
1709 
1710 			WARN_ON(!is_read || acting.osds[0] != acting.primary);
1711 			if (t->flags & CEPH_OSD_FLAG_BALANCE_READS) {
1712 				pos = pick_random_replica(&acting);
1713 			} else {
1714 				pos = pick_closest_replica(osdc, &acting);
1715 			}
1716 			t->osd = acting.osds[pos];
1717 			t->used_replica = pos > 0;
1718 		} else {
1719 			t->osd = acting.primary;
1720 			t->used_replica = false;
1721 		}
1722 	}
1723 
1724 	if (unpaused || legacy_change || force_resend || split)
1725 		ct_res = CALC_TARGET_NEED_RESEND;
1726 	else
1727 		ct_res = CALC_TARGET_NO_ACTION;
1728 
1729 out:
1730 	dout("%s t %p -> %d%d%d%d ct_res %d osd%d\n", __func__, t, unpaused,
1731 	     legacy_change, force_resend, split, ct_res, t->osd);
1732 	return ct_res;
1733 }
1734 
alloc_spg_mapping(void)1735 static struct ceph_spg_mapping *alloc_spg_mapping(void)
1736 {
1737 	struct ceph_spg_mapping *spg;
1738 
1739 	spg = kmalloc(sizeof(*spg), GFP_NOIO);
1740 	if (!spg)
1741 		return NULL;
1742 
1743 	RB_CLEAR_NODE(&spg->node);
1744 	spg->backoffs = RB_ROOT;
1745 	return spg;
1746 }
1747 
free_spg_mapping(struct ceph_spg_mapping * spg)1748 static void free_spg_mapping(struct ceph_spg_mapping *spg)
1749 {
1750 	WARN_ON(!RB_EMPTY_NODE(&spg->node));
1751 	WARN_ON(!RB_EMPTY_ROOT(&spg->backoffs));
1752 
1753 	kfree(spg);
1754 }
1755 
1756 /*
1757  * rbtree of ceph_spg_mapping for handling map<spg_t, ...>, similar to
1758  * ceph_pg_mapping.  Used to track OSD backoffs -- a backoff [range] is
1759  * defined only within a specific spgid; it does not pass anything to
1760  * children on split, or to another primary.
1761  */
DEFINE_RB_FUNCS2(spg_mapping,struct ceph_spg_mapping,spgid,ceph_spg_compare,RB_BYPTR,const struct ceph_spg *,node)1762 DEFINE_RB_FUNCS2(spg_mapping, struct ceph_spg_mapping, spgid, ceph_spg_compare,
1763 		 RB_BYPTR, const struct ceph_spg *, node)
1764 
1765 static u64 hoid_get_bitwise_key(const struct ceph_hobject_id *hoid)
1766 {
1767 	return hoid->is_max ? 0x100000000ull : hoid->hash_reverse_bits;
1768 }
1769 
hoid_get_effective_key(const struct ceph_hobject_id * hoid,void ** pkey,size_t * pkey_len)1770 static void hoid_get_effective_key(const struct ceph_hobject_id *hoid,
1771 				   void **pkey, size_t *pkey_len)
1772 {
1773 	if (hoid->key_len) {
1774 		*pkey = hoid->key;
1775 		*pkey_len = hoid->key_len;
1776 	} else {
1777 		*pkey = hoid->oid;
1778 		*pkey_len = hoid->oid_len;
1779 	}
1780 }
1781 
compare_names(const void * name1,size_t name1_len,const void * name2,size_t name2_len)1782 static int compare_names(const void *name1, size_t name1_len,
1783 			 const void *name2, size_t name2_len)
1784 {
1785 	int ret;
1786 
1787 	ret = memcmp(name1, name2, min(name1_len, name2_len));
1788 	if (!ret) {
1789 		if (name1_len < name2_len)
1790 			ret = -1;
1791 		else if (name1_len > name2_len)
1792 			ret = 1;
1793 	}
1794 	return ret;
1795 }
1796 
hoid_compare(const struct ceph_hobject_id * lhs,const struct ceph_hobject_id * rhs)1797 static int hoid_compare(const struct ceph_hobject_id *lhs,
1798 			const struct ceph_hobject_id *rhs)
1799 {
1800 	void *effective_key1, *effective_key2;
1801 	size_t effective_key1_len, effective_key2_len;
1802 	int ret;
1803 
1804 	if (lhs->is_max < rhs->is_max)
1805 		return -1;
1806 	if (lhs->is_max > rhs->is_max)
1807 		return 1;
1808 
1809 	if (lhs->pool < rhs->pool)
1810 		return -1;
1811 	if (lhs->pool > rhs->pool)
1812 		return 1;
1813 
1814 	if (hoid_get_bitwise_key(lhs) < hoid_get_bitwise_key(rhs))
1815 		return -1;
1816 	if (hoid_get_bitwise_key(lhs) > hoid_get_bitwise_key(rhs))
1817 		return 1;
1818 
1819 	ret = compare_names(lhs->nspace, lhs->nspace_len,
1820 			    rhs->nspace, rhs->nspace_len);
1821 	if (ret)
1822 		return ret;
1823 
1824 	hoid_get_effective_key(lhs, &effective_key1, &effective_key1_len);
1825 	hoid_get_effective_key(rhs, &effective_key2, &effective_key2_len);
1826 	ret = compare_names(effective_key1, effective_key1_len,
1827 			    effective_key2, effective_key2_len);
1828 	if (ret)
1829 		return ret;
1830 
1831 	ret = compare_names(lhs->oid, lhs->oid_len, rhs->oid, rhs->oid_len);
1832 	if (ret)
1833 		return ret;
1834 
1835 	if (lhs->snapid < rhs->snapid)
1836 		return -1;
1837 	if (lhs->snapid > rhs->snapid)
1838 		return 1;
1839 
1840 	return 0;
1841 }
1842 
1843 /*
1844  * For decoding ->begin and ->end of MOSDBackoff only -- no MIN/MAX
1845  * compat stuff here.
1846  *
1847  * Assumes @hoid is zero-initialized.
1848  */
decode_hoid(void ** p,void * end,struct ceph_hobject_id * hoid)1849 static int decode_hoid(void **p, void *end, struct ceph_hobject_id *hoid)
1850 {
1851 	u8 struct_v;
1852 	u32 struct_len;
1853 	int ret;
1854 
1855 	ret = ceph_start_decoding(p, end, 4, "hobject_t", &struct_v,
1856 				  &struct_len);
1857 	if (ret)
1858 		return ret;
1859 
1860 	if (struct_v < 4) {
1861 		pr_err("got struct_v %d < 4 of hobject_t\n", struct_v);
1862 		goto e_inval;
1863 	}
1864 
1865 	hoid->key = ceph_extract_encoded_string(p, end, &hoid->key_len,
1866 						GFP_NOIO);
1867 	if (IS_ERR(hoid->key)) {
1868 		ret = PTR_ERR(hoid->key);
1869 		hoid->key = NULL;
1870 		return ret;
1871 	}
1872 
1873 	hoid->oid = ceph_extract_encoded_string(p, end, &hoid->oid_len,
1874 						GFP_NOIO);
1875 	if (IS_ERR(hoid->oid)) {
1876 		ret = PTR_ERR(hoid->oid);
1877 		hoid->oid = NULL;
1878 		return ret;
1879 	}
1880 
1881 	ceph_decode_64_safe(p, end, hoid->snapid, e_inval);
1882 	ceph_decode_32_safe(p, end, hoid->hash, e_inval);
1883 	ceph_decode_8_safe(p, end, hoid->is_max, e_inval);
1884 
1885 	hoid->nspace = ceph_extract_encoded_string(p, end, &hoid->nspace_len,
1886 						   GFP_NOIO);
1887 	if (IS_ERR(hoid->nspace)) {
1888 		ret = PTR_ERR(hoid->nspace);
1889 		hoid->nspace = NULL;
1890 		return ret;
1891 	}
1892 
1893 	ceph_decode_64_safe(p, end, hoid->pool, e_inval);
1894 
1895 	ceph_hoid_build_hash_cache(hoid);
1896 	return 0;
1897 
1898 e_inval:
1899 	return -EINVAL;
1900 }
1901 
hoid_encoding_size(const struct ceph_hobject_id * hoid)1902 static int hoid_encoding_size(const struct ceph_hobject_id *hoid)
1903 {
1904 	return 8 + 4 + 1 + 8 + /* snapid, hash, is_max, pool */
1905 	       4 + hoid->key_len + 4 + hoid->oid_len + 4 + hoid->nspace_len;
1906 }
1907 
encode_hoid(void ** p,void * end,const struct ceph_hobject_id * hoid)1908 static void encode_hoid(void **p, void *end, const struct ceph_hobject_id *hoid)
1909 {
1910 	ceph_start_encoding(p, 4, 3, hoid_encoding_size(hoid));
1911 	ceph_encode_string(p, end, hoid->key, hoid->key_len);
1912 	ceph_encode_string(p, end, hoid->oid, hoid->oid_len);
1913 	ceph_encode_64(p, hoid->snapid);
1914 	ceph_encode_32(p, hoid->hash);
1915 	ceph_encode_8(p, hoid->is_max);
1916 	ceph_encode_string(p, end, hoid->nspace, hoid->nspace_len);
1917 	ceph_encode_64(p, hoid->pool);
1918 }
1919 
free_hoid(struct ceph_hobject_id * hoid)1920 static void free_hoid(struct ceph_hobject_id *hoid)
1921 {
1922 	if (hoid) {
1923 		kfree(hoid->key);
1924 		kfree(hoid->oid);
1925 		kfree(hoid->nspace);
1926 		kfree(hoid);
1927 	}
1928 }
1929 
alloc_backoff(void)1930 static struct ceph_osd_backoff *alloc_backoff(void)
1931 {
1932 	struct ceph_osd_backoff *backoff;
1933 
1934 	backoff = kzalloc(sizeof(*backoff), GFP_NOIO);
1935 	if (!backoff)
1936 		return NULL;
1937 
1938 	RB_CLEAR_NODE(&backoff->spg_node);
1939 	RB_CLEAR_NODE(&backoff->id_node);
1940 	return backoff;
1941 }
1942 
free_backoff(struct ceph_osd_backoff * backoff)1943 static void free_backoff(struct ceph_osd_backoff *backoff)
1944 {
1945 	WARN_ON(!RB_EMPTY_NODE(&backoff->spg_node));
1946 	WARN_ON(!RB_EMPTY_NODE(&backoff->id_node));
1947 
1948 	free_hoid(backoff->begin);
1949 	free_hoid(backoff->end);
1950 	kfree(backoff);
1951 }
1952 
1953 /*
1954  * Within a specific spgid, backoffs are managed by ->begin hoid.
1955  */
1956 DEFINE_RB_INSDEL_FUNCS2(backoff, struct ceph_osd_backoff, begin, hoid_compare,
1957 			RB_BYVAL, spg_node);
1958 
lookup_containing_backoff(struct rb_root * root,const struct ceph_hobject_id * hoid)1959 static struct ceph_osd_backoff *lookup_containing_backoff(struct rb_root *root,
1960 					    const struct ceph_hobject_id *hoid)
1961 {
1962 	struct rb_node *n = root->rb_node;
1963 
1964 	while (n) {
1965 		struct ceph_osd_backoff *cur =
1966 		    rb_entry(n, struct ceph_osd_backoff, spg_node);
1967 		int cmp;
1968 
1969 		cmp = hoid_compare(hoid, cur->begin);
1970 		if (cmp < 0) {
1971 			n = n->rb_left;
1972 		} else if (cmp > 0) {
1973 			if (hoid_compare(hoid, cur->end) < 0)
1974 				return cur;
1975 
1976 			n = n->rb_right;
1977 		} else {
1978 			return cur;
1979 		}
1980 	}
1981 
1982 	return NULL;
1983 }
1984 
1985 /*
1986  * Each backoff has a unique id within its OSD session.
1987  */
DEFINE_RB_FUNCS(backoff_by_id,struct ceph_osd_backoff,id,id_node)1988 DEFINE_RB_FUNCS(backoff_by_id, struct ceph_osd_backoff, id, id_node)
1989 
1990 static void clear_backoffs(struct ceph_osd *osd)
1991 {
1992 	while (!RB_EMPTY_ROOT(&osd->o_backoff_mappings)) {
1993 		struct ceph_spg_mapping *spg =
1994 		    rb_entry(rb_first(&osd->o_backoff_mappings),
1995 			     struct ceph_spg_mapping, node);
1996 
1997 		while (!RB_EMPTY_ROOT(&spg->backoffs)) {
1998 			struct ceph_osd_backoff *backoff =
1999 			    rb_entry(rb_first(&spg->backoffs),
2000 				     struct ceph_osd_backoff, spg_node);
2001 
2002 			erase_backoff(&spg->backoffs, backoff);
2003 			erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
2004 			free_backoff(backoff);
2005 		}
2006 		erase_spg_mapping(&osd->o_backoff_mappings, spg);
2007 		free_spg_mapping(spg);
2008 	}
2009 }
2010 
2011 /*
2012  * Set up a temporary, non-owning view into @t.
2013  */
hoid_fill_from_target(struct ceph_hobject_id * hoid,const struct ceph_osd_request_target * t)2014 static void hoid_fill_from_target(struct ceph_hobject_id *hoid,
2015 				  const struct ceph_osd_request_target *t)
2016 {
2017 	hoid->key = NULL;
2018 	hoid->key_len = 0;
2019 	hoid->oid = t->target_oid.name;
2020 	hoid->oid_len = t->target_oid.name_len;
2021 	hoid->snapid = CEPH_NOSNAP;
2022 	hoid->hash = t->pgid.seed;
2023 	hoid->is_max = false;
2024 	if (t->target_oloc.pool_ns) {
2025 		hoid->nspace = t->target_oloc.pool_ns->str;
2026 		hoid->nspace_len = t->target_oloc.pool_ns->len;
2027 	} else {
2028 		hoid->nspace = NULL;
2029 		hoid->nspace_len = 0;
2030 	}
2031 	hoid->pool = t->target_oloc.pool;
2032 	ceph_hoid_build_hash_cache(hoid);
2033 }
2034 
should_plug_request(struct ceph_osd_request * req)2035 static bool should_plug_request(struct ceph_osd_request *req)
2036 {
2037 	struct ceph_osd *osd = req->r_osd;
2038 	struct ceph_spg_mapping *spg;
2039 	struct ceph_osd_backoff *backoff;
2040 	struct ceph_hobject_id hoid;
2041 
2042 	spg = lookup_spg_mapping(&osd->o_backoff_mappings, &req->r_t.spgid);
2043 	if (!spg)
2044 		return false;
2045 
2046 	hoid_fill_from_target(&hoid, &req->r_t);
2047 	backoff = lookup_containing_backoff(&spg->backoffs, &hoid);
2048 	if (!backoff)
2049 		return false;
2050 
2051 	dout("%s req %p tid %llu backoff osd%d spgid %llu.%xs%d id %llu\n",
2052 	     __func__, req, req->r_tid, osd->o_osd, backoff->spgid.pgid.pool,
2053 	     backoff->spgid.pgid.seed, backoff->spgid.shard, backoff->id);
2054 	return true;
2055 }
2056 
2057 /*
2058  * Keep get_num_data_items() in sync with this function.
2059  */
setup_request_data(struct ceph_osd_request * req)2060 static void setup_request_data(struct ceph_osd_request *req)
2061 {
2062 	struct ceph_msg *request_msg = req->r_request;
2063 	struct ceph_msg *reply_msg = req->r_reply;
2064 	struct ceph_osd_req_op *op;
2065 
2066 	if (req->r_request->num_data_items || req->r_reply->num_data_items)
2067 		return;
2068 
2069 	WARN_ON(request_msg->data_length || reply_msg->data_length);
2070 	for (op = req->r_ops; op != &req->r_ops[req->r_num_ops]; op++) {
2071 		switch (op->op) {
2072 		/* request */
2073 		case CEPH_OSD_OP_WRITE:
2074 		case CEPH_OSD_OP_WRITEFULL:
2075 			WARN_ON(op->indata_len != op->extent.length);
2076 			ceph_osdc_msg_data_add(request_msg,
2077 					       &op->extent.osd_data);
2078 			break;
2079 		case CEPH_OSD_OP_SETXATTR:
2080 		case CEPH_OSD_OP_CMPXATTR:
2081 			WARN_ON(op->indata_len != op->xattr.name_len +
2082 						  op->xattr.value_len);
2083 			ceph_osdc_msg_data_add(request_msg,
2084 					       &op->xattr.osd_data);
2085 			break;
2086 		case CEPH_OSD_OP_NOTIFY_ACK:
2087 			ceph_osdc_msg_data_add(request_msg,
2088 					       &op->notify_ack.request_data);
2089 			break;
2090 		case CEPH_OSD_OP_COPY_FROM2:
2091 			ceph_osdc_msg_data_add(request_msg,
2092 					       &op->copy_from.osd_data);
2093 			break;
2094 
2095 		/* reply */
2096 		case CEPH_OSD_OP_STAT:
2097 			ceph_osdc_msg_data_add(reply_msg,
2098 					       &op->raw_data_in);
2099 			break;
2100 		case CEPH_OSD_OP_READ:
2101 		case CEPH_OSD_OP_SPARSE_READ:
2102 			ceph_osdc_msg_data_add(reply_msg,
2103 					       &op->extent.osd_data);
2104 			break;
2105 		case CEPH_OSD_OP_LIST_WATCHERS:
2106 			ceph_osdc_msg_data_add(reply_msg,
2107 					       &op->list_watchers.response_data);
2108 			break;
2109 
2110 		/* both */
2111 		case CEPH_OSD_OP_CALL:
2112 			WARN_ON(op->indata_len != op->cls.class_len +
2113 						  op->cls.method_len +
2114 						  op->cls.indata_len);
2115 			ceph_osdc_msg_data_add(request_msg,
2116 					       &op->cls.request_info);
2117 			/* optional, can be NONE */
2118 			ceph_osdc_msg_data_add(request_msg,
2119 					       &op->cls.request_data);
2120 			/* optional, can be NONE */
2121 			ceph_osdc_msg_data_add(reply_msg,
2122 					       &op->cls.response_data);
2123 			break;
2124 		case CEPH_OSD_OP_NOTIFY:
2125 			ceph_osdc_msg_data_add(request_msg,
2126 					       &op->notify.request_data);
2127 			ceph_osdc_msg_data_add(reply_msg,
2128 					       &op->notify.response_data);
2129 			break;
2130 		}
2131 	}
2132 }
2133 
encode_pgid(void ** p,const struct ceph_pg * pgid)2134 static void encode_pgid(void **p, const struct ceph_pg *pgid)
2135 {
2136 	ceph_encode_8(p, 1);
2137 	ceph_encode_64(p, pgid->pool);
2138 	ceph_encode_32(p, pgid->seed);
2139 	ceph_encode_32(p, -1); /* preferred */
2140 }
2141 
encode_spgid(void ** p,const struct ceph_spg * spgid)2142 static void encode_spgid(void **p, const struct ceph_spg *spgid)
2143 {
2144 	ceph_start_encoding(p, 1, 1, CEPH_PGID_ENCODING_LEN + 1);
2145 	encode_pgid(p, &spgid->pgid);
2146 	ceph_encode_8(p, spgid->shard);
2147 }
2148 
encode_oloc(void ** p,void * end,const struct ceph_object_locator * oloc)2149 static void encode_oloc(void **p, void *end,
2150 			const struct ceph_object_locator *oloc)
2151 {
2152 	ceph_start_encoding(p, 5, 4, ceph_oloc_encoding_size(oloc));
2153 	ceph_encode_64(p, oloc->pool);
2154 	ceph_encode_32(p, -1); /* preferred */
2155 	ceph_encode_32(p, 0);  /* key len */
2156 	if (oloc->pool_ns)
2157 		ceph_encode_string(p, end, oloc->pool_ns->str,
2158 				   oloc->pool_ns->len);
2159 	else
2160 		ceph_encode_32(p, 0);
2161 }
2162 
encode_request_partial(struct ceph_osd_request * req,struct ceph_msg * msg)2163 static void encode_request_partial(struct ceph_osd_request *req,
2164 				   struct ceph_msg *msg)
2165 {
2166 	void *p = msg->front.iov_base;
2167 	void *const end = p + msg->front_alloc_len;
2168 	u32 data_len = 0;
2169 	int i;
2170 
2171 	if (req->r_flags & CEPH_OSD_FLAG_WRITE) {
2172 		/* snapshots aren't writeable */
2173 		WARN_ON(req->r_snapid != CEPH_NOSNAP);
2174 	} else {
2175 		WARN_ON(req->r_mtime.tv_sec || req->r_mtime.tv_nsec ||
2176 			req->r_data_offset || req->r_snapc);
2177 	}
2178 
2179 	setup_request_data(req);
2180 
2181 	encode_spgid(&p, &req->r_t.spgid); /* actual spg */
2182 	ceph_encode_32(&p, req->r_t.pgid.seed); /* raw hash */
2183 	ceph_encode_32(&p, req->r_osdc->osdmap->epoch);
2184 	ceph_encode_32(&p, req->r_flags);
2185 
2186 	/* reqid */
2187 	ceph_start_encoding(&p, 2, 2, sizeof(struct ceph_osd_reqid));
2188 	memset(p, 0, sizeof(struct ceph_osd_reqid));
2189 	p += sizeof(struct ceph_osd_reqid);
2190 
2191 	/* trace */
2192 	memset(p, 0, sizeof(struct ceph_blkin_trace_info));
2193 	p += sizeof(struct ceph_blkin_trace_info);
2194 
2195 	ceph_encode_32(&p, 0); /* client_inc, always 0 */
2196 	ceph_encode_timespec64(p, &req->r_mtime);
2197 	p += sizeof(struct ceph_timespec);
2198 
2199 	encode_oloc(&p, end, &req->r_t.target_oloc);
2200 	ceph_encode_string(&p, end, req->r_t.target_oid.name,
2201 			   req->r_t.target_oid.name_len);
2202 
2203 	/* ops, can imply data */
2204 	ceph_encode_16(&p, req->r_num_ops);
2205 	for (i = 0; i < req->r_num_ops; i++) {
2206 		data_len += osd_req_encode_op(p, &req->r_ops[i]);
2207 		p += sizeof(struct ceph_osd_op);
2208 	}
2209 
2210 	ceph_encode_64(&p, req->r_snapid); /* snapid */
2211 	if (req->r_snapc) {
2212 		ceph_encode_64(&p, req->r_snapc->seq);
2213 		ceph_encode_32(&p, req->r_snapc->num_snaps);
2214 		for (i = 0; i < req->r_snapc->num_snaps; i++)
2215 			ceph_encode_64(&p, req->r_snapc->snaps[i]);
2216 	} else {
2217 		ceph_encode_64(&p, 0); /* snap_seq */
2218 		ceph_encode_32(&p, 0); /* snaps len */
2219 	}
2220 
2221 	ceph_encode_32(&p, req->r_attempts); /* retry_attempt */
2222 	BUG_ON(p > end - 8); /* space for features */
2223 
2224 	msg->hdr.version = cpu_to_le16(8); /* MOSDOp v8 */
2225 	/* front_len is finalized in encode_request_finish() */
2226 	msg->front.iov_len = p - msg->front.iov_base;
2227 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2228 	msg->hdr.data_len = cpu_to_le32(data_len);
2229 	/*
2230 	 * The header "data_off" is a hint to the receiver allowing it
2231 	 * to align received data into its buffers such that there's no
2232 	 * need to re-copy it before writing it to disk (direct I/O).
2233 	 */
2234 	msg->hdr.data_off = cpu_to_le16(req->r_data_offset);
2235 
2236 	dout("%s req %p msg %p oid %s oid_len %d\n", __func__, req, msg,
2237 	     req->r_t.target_oid.name, req->r_t.target_oid.name_len);
2238 }
2239 
encode_request_finish(struct ceph_msg * msg)2240 static void encode_request_finish(struct ceph_msg *msg)
2241 {
2242 	void *p = msg->front.iov_base;
2243 	void *const partial_end = p + msg->front.iov_len;
2244 	void *const end = p + msg->front_alloc_len;
2245 
2246 	if (CEPH_HAVE_FEATURE(msg->con->peer_features, RESEND_ON_SPLIT)) {
2247 		/* luminous OSD -- encode features and be done */
2248 		p = partial_end;
2249 		ceph_encode_64(&p, msg->con->peer_features);
2250 	} else {
2251 		struct {
2252 			char spgid[CEPH_ENCODING_START_BLK_LEN +
2253 				   CEPH_PGID_ENCODING_LEN + 1];
2254 			__le32 hash;
2255 			__le32 epoch;
2256 			__le32 flags;
2257 			char reqid[CEPH_ENCODING_START_BLK_LEN +
2258 				   sizeof(struct ceph_osd_reqid)];
2259 			char trace[sizeof(struct ceph_blkin_trace_info)];
2260 			__le32 client_inc;
2261 			struct ceph_timespec mtime;
2262 		} __packed head;
2263 		struct ceph_pg pgid;
2264 		void *oloc, *oid, *tail;
2265 		int oloc_len, oid_len, tail_len;
2266 		int len;
2267 
2268 		/*
2269 		 * Pre-luminous OSD -- reencode v8 into v4 using @head
2270 		 * as a temporary buffer.  Encode the raw PG; the rest
2271 		 * is just a matter of moving oloc, oid and tail blobs
2272 		 * around.
2273 		 */
2274 		memcpy(&head, p, sizeof(head));
2275 		p += sizeof(head);
2276 
2277 		oloc = p;
2278 		p += CEPH_ENCODING_START_BLK_LEN;
2279 		pgid.pool = ceph_decode_64(&p);
2280 		p += 4 + 4; /* preferred, key len */
2281 		len = ceph_decode_32(&p);
2282 		p += len;   /* nspace */
2283 		oloc_len = p - oloc;
2284 
2285 		oid = p;
2286 		len = ceph_decode_32(&p);
2287 		p += len;
2288 		oid_len = p - oid;
2289 
2290 		tail = p;
2291 		tail_len = partial_end - p;
2292 
2293 		p = msg->front.iov_base;
2294 		ceph_encode_copy(&p, &head.client_inc, sizeof(head.client_inc));
2295 		ceph_encode_copy(&p, &head.epoch, sizeof(head.epoch));
2296 		ceph_encode_copy(&p, &head.flags, sizeof(head.flags));
2297 		ceph_encode_copy(&p, &head.mtime, sizeof(head.mtime));
2298 
2299 		/* reassert_version */
2300 		memset(p, 0, sizeof(struct ceph_eversion));
2301 		p += sizeof(struct ceph_eversion);
2302 
2303 		BUG_ON(p >= oloc);
2304 		memmove(p, oloc, oloc_len);
2305 		p += oloc_len;
2306 
2307 		pgid.seed = le32_to_cpu(head.hash);
2308 		encode_pgid(&p, &pgid); /* raw pg */
2309 
2310 		BUG_ON(p >= oid);
2311 		memmove(p, oid, oid_len);
2312 		p += oid_len;
2313 
2314 		/* tail -- ops, snapid, snapc, retry_attempt */
2315 		BUG_ON(p >= tail);
2316 		memmove(p, tail, tail_len);
2317 		p += tail_len;
2318 
2319 		msg->hdr.version = cpu_to_le16(4); /* MOSDOp v4 */
2320 	}
2321 
2322 	BUG_ON(p > end);
2323 	msg->front.iov_len = p - msg->front.iov_base;
2324 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2325 
2326 	dout("%s msg %p tid %llu %u+%u+%u v%d\n", __func__, msg,
2327 	     le64_to_cpu(msg->hdr.tid), le32_to_cpu(msg->hdr.front_len),
2328 	     le32_to_cpu(msg->hdr.middle_len), le32_to_cpu(msg->hdr.data_len),
2329 	     le16_to_cpu(msg->hdr.version));
2330 }
2331 
2332 /*
2333  * @req has to be assigned a tid and registered.
2334  */
send_request(struct ceph_osd_request * req)2335 static void send_request(struct ceph_osd_request *req)
2336 {
2337 	struct ceph_osd *osd = req->r_osd;
2338 
2339 	verify_osd_locked(osd);
2340 	WARN_ON(osd->o_osd != req->r_t.osd);
2341 
2342 	/* backoff? */
2343 	if (should_plug_request(req))
2344 		return;
2345 
2346 	/*
2347 	 * We may have a previously queued request message hanging
2348 	 * around.  Cancel it to avoid corrupting the msgr.
2349 	 */
2350 	if (req->r_sent)
2351 		ceph_msg_revoke(req->r_request);
2352 
2353 	req->r_flags |= CEPH_OSD_FLAG_KNOWN_REDIR;
2354 	if (req->r_attempts)
2355 		req->r_flags |= CEPH_OSD_FLAG_RETRY;
2356 	else
2357 		WARN_ON(req->r_flags & CEPH_OSD_FLAG_RETRY);
2358 
2359 	encode_request_partial(req, req->r_request);
2360 
2361 	dout("%s req %p tid %llu to pgid %llu.%x spgid %llu.%xs%d osd%d e%u flags 0x%x attempt %d\n",
2362 	     __func__, req, req->r_tid, req->r_t.pgid.pool, req->r_t.pgid.seed,
2363 	     req->r_t.spgid.pgid.pool, req->r_t.spgid.pgid.seed,
2364 	     req->r_t.spgid.shard, osd->o_osd, req->r_t.epoch, req->r_flags,
2365 	     req->r_attempts);
2366 
2367 	req->r_t.paused = false;
2368 	req->r_stamp = jiffies;
2369 	req->r_attempts++;
2370 
2371 	req->r_sent = osd->o_incarnation;
2372 	req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
2373 	ceph_con_send(&osd->o_con, ceph_msg_get(req->r_request));
2374 }
2375 
maybe_request_map(struct ceph_osd_client * osdc)2376 static void maybe_request_map(struct ceph_osd_client *osdc)
2377 {
2378 	bool continuous = false;
2379 
2380 	verify_osdc_locked(osdc);
2381 	WARN_ON(!osdc->osdmap->epoch);
2382 
2383 	if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2384 	    ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD) ||
2385 	    ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
2386 		dout("%s osdc %p continuous\n", __func__, osdc);
2387 		continuous = true;
2388 	} else {
2389 		dout("%s osdc %p onetime\n", __func__, osdc);
2390 	}
2391 
2392 	if (ceph_monc_want_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
2393 			       osdc->osdmap->epoch + 1, continuous))
2394 		ceph_monc_renew_subs(&osdc->client->monc);
2395 }
2396 
2397 static void complete_request(struct ceph_osd_request *req, int err);
2398 static void send_map_check(struct ceph_osd_request *req);
2399 
__submit_request(struct ceph_osd_request * req,bool wrlocked)2400 static void __submit_request(struct ceph_osd_request *req, bool wrlocked)
2401 {
2402 	struct ceph_osd_client *osdc = req->r_osdc;
2403 	struct ceph_osd *osd;
2404 	enum calc_target_result ct_res;
2405 	int err = 0;
2406 	bool need_send = false;
2407 	bool promoted = false;
2408 
2409 	WARN_ON(req->r_tid);
2410 	dout("%s req %p wrlocked %d\n", __func__, req, wrlocked);
2411 
2412 again:
2413 	ct_res = calc_target(osdc, &req->r_t, false);
2414 	if (ct_res == CALC_TARGET_POOL_DNE && !wrlocked)
2415 		goto promote;
2416 
2417 	osd = lookup_create_osd(osdc, req->r_t.osd, wrlocked);
2418 	if (IS_ERR(osd)) {
2419 		WARN_ON(PTR_ERR(osd) != -EAGAIN || wrlocked);
2420 		goto promote;
2421 	}
2422 
2423 	if (osdc->abort_err) {
2424 		dout("req %p abort_err %d\n", req, osdc->abort_err);
2425 		err = osdc->abort_err;
2426 	} else if (osdc->osdmap->epoch < osdc->epoch_barrier) {
2427 		dout("req %p epoch %u barrier %u\n", req, osdc->osdmap->epoch,
2428 		     osdc->epoch_barrier);
2429 		req->r_t.paused = true;
2430 		maybe_request_map(osdc);
2431 	} else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2432 		   ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR)) {
2433 		dout("req %p pausewr\n", req);
2434 		req->r_t.paused = true;
2435 		maybe_request_map(osdc);
2436 	} else if ((req->r_flags & CEPH_OSD_FLAG_READ) &&
2437 		   ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
2438 		dout("req %p pauserd\n", req);
2439 		req->r_t.paused = true;
2440 		maybe_request_map(osdc);
2441 	} else if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2442 		   !(req->r_flags & (CEPH_OSD_FLAG_FULL_TRY |
2443 				     CEPH_OSD_FLAG_FULL_FORCE)) &&
2444 		   (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2445 		    pool_full(osdc, req->r_t.base_oloc.pool))) {
2446 		dout("req %p full/pool_full\n", req);
2447 		if (ceph_test_opt(osdc->client, ABORT_ON_FULL)) {
2448 			err = -ENOSPC;
2449 		} else {
2450 			if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL))
2451 				pr_warn_ratelimited("cluster is full (osdmap FULL)\n");
2452 			else
2453 				pr_warn_ratelimited("pool %lld is full or reached quota\n",
2454 						    req->r_t.base_oloc.pool);
2455 			req->r_t.paused = true;
2456 			maybe_request_map(osdc);
2457 		}
2458 	} else if (!osd_homeless(osd)) {
2459 		need_send = true;
2460 	} else {
2461 		maybe_request_map(osdc);
2462 	}
2463 
2464 	mutex_lock(&osd->lock);
2465 	/*
2466 	 * Assign the tid atomically with send_request() to protect
2467 	 * multiple writes to the same object from racing with each
2468 	 * other, resulting in out of order ops on the OSDs.
2469 	 */
2470 	req->r_tid = atomic64_inc_return(&osdc->last_tid);
2471 	link_request(osd, req);
2472 	if (need_send)
2473 		send_request(req);
2474 	else if (err)
2475 		complete_request(req, err);
2476 	mutex_unlock(&osd->lock);
2477 
2478 	if (!err && ct_res == CALC_TARGET_POOL_DNE)
2479 		send_map_check(req);
2480 
2481 	if (promoted)
2482 		downgrade_write(&osdc->lock);
2483 	return;
2484 
2485 promote:
2486 	up_read(&osdc->lock);
2487 	down_write(&osdc->lock);
2488 	wrlocked = true;
2489 	promoted = true;
2490 	goto again;
2491 }
2492 
account_request(struct ceph_osd_request * req)2493 static void account_request(struct ceph_osd_request *req)
2494 {
2495 	WARN_ON(req->r_flags & (CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK));
2496 	WARN_ON(!(req->r_flags & (CEPH_OSD_FLAG_READ | CEPH_OSD_FLAG_WRITE)));
2497 
2498 	req->r_flags |= CEPH_OSD_FLAG_ONDISK;
2499 	atomic_inc(&req->r_osdc->num_requests);
2500 
2501 	req->r_start_stamp = jiffies;
2502 	req->r_start_latency = ktime_get();
2503 }
2504 
submit_request(struct ceph_osd_request * req,bool wrlocked)2505 static void submit_request(struct ceph_osd_request *req, bool wrlocked)
2506 {
2507 	ceph_osdc_get_request(req);
2508 	account_request(req);
2509 	__submit_request(req, wrlocked);
2510 }
2511 
finish_request(struct ceph_osd_request * req)2512 static void finish_request(struct ceph_osd_request *req)
2513 {
2514 	struct ceph_osd_client *osdc = req->r_osdc;
2515 
2516 	WARN_ON(lookup_request_mc(&osdc->map_checks, req->r_tid));
2517 	dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2518 
2519 	req->r_end_latency = ktime_get();
2520 
2521 	if (req->r_osd) {
2522 		ceph_init_sparse_read(&req->r_osd->o_sparse_read);
2523 		unlink_request(req->r_osd, req);
2524 	}
2525 	atomic_dec(&osdc->num_requests);
2526 
2527 	/*
2528 	 * If an OSD has failed or returned and a request has been sent
2529 	 * twice, it's possible to get a reply and end up here while the
2530 	 * request message is queued for delivery.  We will ignore the
2531 	 * reply, so not a big deal, but better to try and catch it.
2532 	 */
2533 	ceph_msg_revoke(req->r_request);
2534 	ceph_msg_revoke_incoming(req->r_reply);
2535 }
2536 
__complete_request(struct ceph_osd_request * req)2537 static void __complete_request(struct ceph_osd_request *req)
2538 {
2539 	dout("%s req %p tid %llu cb %ps result %d\n", __func__, req,
2540 	     req->r_tid, req->r_callback, req->r_result);
2541 
2542 	if (req->r_callback)
2543 		req->r_callback(req);
2544 	complete_all(&req->r_completion);
2545 	ceph_osdc_put_request(req);
2546 }
2547 
complete_request_workfn(struct work_struct * work)2548 static void complete_request_workfn(struct work_struct *work)
2549 {
2550 	struct ceph_osd_request *req =
2551 	    container_of(work, struct ceph_osd_request, r_complete_work);
2552 
2553 	__complete_request(req);
2554 }
2555 
2556 /*
2557  * This is open-coded in handle_reply().
2558  */
complete_request(struct ceph_osd_request * req,int err)2559 static void complete_request(struct ceph_osd_request *req, int err)
2560 {
2561 	dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2562 
2563 	req->r_result = err;
2564 	finish_request(req);
2565 
2566 	INIT_WORK(&req->r_complete_work, complete_request_workfn);
2567 	queue_work(req->r_osdc->completion_wq, &req->r_complete_work);
2568 }
2569 
cancel_map_check(struct ceph_osd_request * req)2570 static void cancel_map_check(struct ceph_osd_request *req)
2571 {
2572 	struct ceph_osd_client *osdc = req->r_osdc;
2573 	struct ceph_osd_request *lookup_req;
2574 
2575 	verify_osdc_wrlocked(osdc);
2576 
2577 	lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2578 	if (!lookup_req)
2579 		return;
2580 
2581 	WARN_ON(lookup_req != req);
2582 	erase_request_mc(&osdc->map_checks, req);
2583 	ceph_osdc_put_request(req);
2584 }
2585 
cancel_request(struct ceph_osd_request * req)2586 static void cancel_request(struct ceph_osd_request *req)
2587 {
2588 	dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
2589 
2590 	cancel_map_check(req);
2591 	finish_request(req);
2592 	complete_all(&req->r_completion);
2593 	ceph_osdc_put_request(req);
2594 }
2595 
abort_request(struct ceph_osd_request * req,int err)2596 static void abort_request(struct ceph_osd_request *req, int err)
2597 {
2598 	dout("%s req %p tid %llu err %d\n", __func__, req, req->r_tid, err);
2599 
2600 	cancel_map_check(req);
2601 	complete_request(req, err);
2602 }
2603 
abort_fn(struct ceph_osd_request * req,void * arg)2604 static int abort_fn(struct ceph_osd_request *req, void *arg)
2605 {
2606 	int err = *(int *)arg;
2607 
2608 	abort_request(req, err);
2609 	return 0; /* continue iteration */
2610 }
2611 
2612 /*
2613  * Abort all in-flight requests with @err and arrange for all future
2614  * requests to be failed immediately.
2615  */
ceph_osdc_abort_requests(struct ceph_osd_client * osdc,int err)2616 void ceph_osdc_abort_requests(struct ceph_osd_client *osdc, int err)
2617 {
2618 	dout("%s osdc %p err %d\n", __func__, osdc, err);
2619 	down_write(&osdc->lock);
2620 	for_each_request(osdc, abort_fn, &err);
2621 	osdc->abort_err = err;
2622 	up_write(&osdc->lock);
2623 }
2624 EXPORT_SYMBOL(ceph_osdc_abort_requests);
2625 
ceph_osdc_clear_abort_err(struct ceph_osd_client * osdc)2626 void ceph_osdc_clear_abort_err(struct ceph_osd_client *osdc)
2627 {
2628 	down_write(&osdc->lock);
2629 	osdc->abort_err = 0;
2630 	up_write(&osdc->lock);
2631 }
2632 EXPORT_SYMBOL(ceph_osdc_clear_abort_err);
2633 
update_epoch_barrier(struct ceph_osd_client * osdc,u32 eb)2634 static void update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2635 {
2636 	if (likely(eb > osdc->epoch_barrier)) {
2637 		dout("updating epoch_barrier from %u to %u\n",
2638 				osdc->epoch_barrier, eb);
2639 		osdc->epoch_barrier = eb;
2640 		/* Request map if we're not to the barrier yet */
2641 		if (eb > osdc->osdmap->epoch)
2642 			maybe_request_map(osdc);
2643 	}
2644 }
2645 
ceph_osdc_update_epoch_barrier(struct ceph_osd_client * osdc,u32 eb)2646 void ceph_osdc_update_epoch_barrier(struct ceph_osd_client *osdc, u32 eb)
2647 {
2648 	down_read(&osdc->lock);
2649 	if (unlikely(eb > osdc->epoch_barrier)) {
2650 		up_read(&osdc->lock);
2651 		down_write(&osdc->lock);
2652 		update_epoch_barrier(osdc, eb);
2653 		up_write(&osdc->lock);
2654 	} else {
2655 		up_read(&osdc->lock);
2656 	}
2657 }
2658 EXPORT_SYMBOL(ceph_osdc_update_epoch_barrier);
2659 
2660 /*
2661  * We can end up releasing caps as a result of abort_request().
2662  * In that case, we probably want to ensure that the cap release message
2663  * has an updated epoch barrier in it, so set the epoch barrier prior to
2664  * aborting the first request.
2665  */
abort_on_full_fn(struct ceph_osd_request * req,void * arg)2666 static int abort_on_full_fn(struct ceph_osd_request *req, void *arg)
2667 {
2668 	struct ceph_osd_client *osdc = req->r_osdc;
2669 	bool *victims = arg;
2670 
2671 	if ((req->r_flags & CEPH_OSD_FLAG_WRITE) &&
2672 	    (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
2673 	     pool_full(osdc, req->r_t.base_oloc.pool))) {
2674 		if (!*victims) {
2675 			update_epoch_barrier(osdc, osdc->osdmap->epoch);
2676 			*victims = true;
2677 		}
2678 		abort_request(req, -ENOSPC);
2679 	}
2680 
2681 	return 0; /* continue iteration */
2682 }
2683 
2684 /*
2685  * Drop all pending requests that are stalled waiting on a full condition to
2686  * clear, and complete them with ENOSPC as the return code. Set the
2687  * osdc->epoch_barrier to the latest map epoch that we've seen if any were
2688  * cancelled.
2689  */
ceph_osdc_abort_on_full(struct ceph_osd_client * osdc)2690 static void ceph_osdc_abort_on_full(struct ceph_osd_client *osdc)
2691 {
2692 	bool victims = false;
2693 
2694 	if (ceph_test_opt(osdc->client, ABORT_ON_FULL) &&
2695 	    (ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) || have_pool_full(osdc)))
2696 		for_each_request(osdc, abort_on_full_fn, &victims);
2697 }
2698 
check_pool_dne(struct ceph_osd_request * req)2699 static void check_pool_dne(struct ceph_osd_request *req)
2700 {
2701 	struct ceph_osd_client *osdc = req->r_osdc;
2702 	struct ceph_osdmap *map = osdc->osdmap;
2703 
2704 	verify_osdc_wrlocked(osdc);
2705 	WARN_ON(!map->epoch);
2706 
2707 	if (req->r_attempts) {
2708 		/*
2709 		 * We sent a request earlier, which means that
2710 		 * previously the pool existed, and now it does not
2711 		 * (i.e., it was deleted).
2712 		 */
2713 		req->r_map_dne_bound = map->epoch;
2714 		dout("%s req %p tid %llu pool disappeared\n", __func__, req,
2715 		     req->r_tid);
2716 	} else {
2717 		dout("%s req %p tid %llu map_dne_bound %u have %u\n", __func__,
2718 		     req, req->r_tid, req->r_map_dne_bound, map->epoch);
2719 	}
2720 
2721 	if (req->r_map_dne_bound) {
2722 		if (map->epoch >= req->r_map_dne_bound) {
2723 			/* we had a new enough map */
2724 			pr_info_ratelimited("tid %llu pool does not exist\n",
2725 					    req->r_tid);
2726 			complete_request(req, -ENOENT);
2727 		}
2728 	} else {
2729 		send_map_check(req);
2730 	}
2731 }
2732 
map_check_cb(struct ceph_mon_generic_request * greq)2733 static void map_check_cb(struct ceph_mon_generic_request *greq)
2734 {
2735 	struct ceph_osd_client *osdc = &greq->monc->client->osdc;
2736 	struct ceph_osd_request *req;
2737 	u64 tid = greq->private_data;
2738 
2739 	WARN_ON(greq->result || !greq->u.newest);
2740 
2741 	down_write(&osdc->lock);
2742 	req = lookup_request_mc(&osdc->map_checks, tid);
2743 	if (!req) {
2744 		dout("%s tid %llu dne\n", __func__, tid);
2745 		goto out_unlock;
2746 	}
2747 
2748 	dout("%s req %p tid %llu map_dne_bound %u newest %llu\n", __func__,
2749 	     req, req->r_tid, req->r_map_dne_bound, greq->u.newest);
2750 	if (!req->r_map_dne_bound)
2751 		req->r_map_dne_bound = greq->u.newest;
2752 	erase_request_mc(&osdc->map_checks, req);
2753 	check_pool_dne(req);
2754 
2755 	ceph_osdc_put_request(req);
2756 out_unlock:
2757 	up_write(&osdc->lock);
2758 }
2759 
send_map_check(struct ceph_osd_request * req)2760 static void send_map_check(struct ceph_osd_request *req)
2761 {
2762 	struct ceph_osd_client *osdc = req->r_osdc;
2763 	struct ceph_osd_request *lookup_req;
2764 	int ret;
2765 
2766 	verify_osdc_wrlocked(osdc);
2767 
2768 	lookup_req = lookup_request_mc(&osdc->map_checks, req->r_tid);
2769 	if (lookup_req) {
2770 		WARN_ON(lookup_req != req);
2771 		return;
2772 	}
2773 
2774 	ceph_osdc_get_request(req);
2775 	insert_request_mc(&osdc->map_checks, req);
2776 	ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
2777 					  map_check_cb, req->r_tid);
2778 	WARN_ON(ret);
2779 }
2780 
2781 /*
2782  * lingering requests, watch/notify v2 infrastructure
2783  */
linger_release(struct kref * kref)2784 static void linger_release(struct kref *kref)
2785 {
2786 	struct ceph_osd_linger_request *lreq =
2787 	    container_of(kref, struct ceph_osd_linger_request, kref);
2788 
2789 	dout("%s lreq %p reg_req %p ping_req %p\n", __func__, lreq,
2790 	     lreq->reg_req, lreq->ping_req);
2791 	WARN_ON(!RB_EMPTY_NODE(&lreq->node));
2792 	WARN_ON(!RB_EMPTY_NODE(&lreq->osdc_node));
2793 	WARN_ON(!RB_EMPTY_NODE(&lreq->mc_node));
2794 	WARN_ON(!list_empty(&lreq->scan_item));
2795 	WARN_ON(!list_empty(&lreq->pending_lworks));
2796 	WARN_ON(lreq->osd);
2797 
2798 	if (lreq->request_pl)
2799 		ceph_pagelist_release(lreq->request_pl);
2800 	if (lreq->notify_id_pages)
2801 		ceph_release_page_vector(lreq->notify_id_pages, 1);
2802 
2803 	ceph_osdc_put_request(lreq->reg_req);
2804 	ceph_osdc_put_request(lreq->ping_req);
2805 	target_destroy(&lreq->t);
2806 	kfree(lreq);
2807 }
2808 
linger_put(struct ceph_osd_linger_request * lreq)2809 static void linger_put(struct ceph_osd_linger_request *lreq)
2810 {
2811 	if (lreq)
2812 		kref_put(&lreq->kref, linger_release);
2813 }
2814 
2815 static struct ceph_osd_linger_request *
linger_get(struct ceph_osd_linger_request * lreq)2816 linger_get(struct ceph_osd_linger_request *lreq)
2817 {
2818 	kref_get(&lreq->kref);
2819 	return lreq;
2820 }
2821 
2822 static struct ceph_osd_linger_request *
linger_alloc(struct ceph_osd_client * osdc)2823 linger_alloc(struct ceph_osd_client *osdc)
2824 {
2825 	struct ceph_osd_linger_request *lreq;
2826 
2827 	lreq = kzalloc(sizeof(*lreq), GFP_NOIO);
2828 	if (!lreq)
2829 		return NULL;
2830 
2831 	kref_init(&lreq->kref);
2832 	mutex_init(&lreq->lock);
2833 	RB_CLEAR_NODE(&lreq->node);
2834 	RB_CLEAR_NODE(&lreq->osdc_node);
2835 	RB_CLEAR_NODE(&lreq->mc_node);
2836 	INIT_LIST_HEAD(&lreq->scan_item);
2837 	INIT_LIST_HEAD(&lreq->pending_lworks);
2838 	init_completion(&lreq->reg_commit_wait);
2839 	init_completion(&lreq->notify_finish_wait);
2840 
2841 	lreq->osdc = osdc;
2842 	target_init(&lreq->t);
2843 
2844 	dout("%s lreq %p\n", __func__, lreq);
2845 	return lreq;
2846 }
2847 
DEFINE_RB_INSDEL_FUNCS(linger,struct ceph_osd_linger_request,linger_id,node)2848 DEFINE_RB_INSDEL_FUNCS(linger, struct ceph_osd_linger_request, linger_id, node)
2849 DEFINE_RB_FUNCS(linger_osdc, struct ceph_osd_linger_request, linger_id, osdc_node)
2850 DEFINE_RB_FUNCS(linger_mc, struct ceph_osd_linger_request, linger_id, mc_node)
2851 
2852 /*
2853  * Create linger request <-> OSD session relation.
2854  *
2855  * @lreq has to be registered, @osd may be homeless.
2856  */
2857 static void link_linger(struct ceph_osd *osd,
2858 			struct ceph_osd_linger_request *lreq)
2859 {
2860 	verify_osd_locked(osd);
2861 	WARN_ON(!lreq->linger_id || lreq->osd);
2862 	dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2863 	     osd->o_osd, lreq, lreq->linger_id);
2864 
2865 	if (!osd_homeless(osd))
2866 		__remove_osd_from_lru(osd);
2867 	else
2868 		atomic_inc(&osd->o_osdc->num_homeless);
2869 
2870 	get_osd(osd);
2871 	insert_linger(&osd->o_linger_requests, lreq);
2872 	lreq->osd = osd;
2873 }
2874 
unlink_linger(struct ceph_osd * osd,struct ceph_osd_linger_request * lreq)2875 static void unlink_linger(struct ceph_osd *osd,
2876 			  struct ceph_osd_linger_request *lreq)
2877 {
2878 	verify_osd_locked(osd);
2879 	WARN_ON(lreq->osd != osd);
2880 	dout("%s osd %p osd%d lreq %p linger_id %llu\n", __func__, osd,
2881 	     osd->o_osd, lreq, lreq->linger_id);
2882 
2883 	lreq->osd = NULL;
2884 	erase_linger(&osd->o_linger_requests, lreq);
2885 	put_osd(osd);
2886 
2887 	if (!osd_homeless(osd))
2888 		maybe_move_osd_to_lru(osd);
2889 	else
2890 		atomic_dec(&osd->o_osdc->num_homeless);
2891 }
2892 
__linger_registered(struct ceph_osd_linger_request * lreq)2893 static bool __linger_registered(struct ceph_osd_linger_request *lreq)
2894 {
2895 	verify_osdc_locked(lreq->osdc);
2896 
2897 	return !RB_EMPTY_NODE(&lreq->osdc_node);
2898 }
2899 
linger_registered(struct ceph_osd_linger_request * lreq)2900 static bool linger_registered(struct ceph_osd_linger_request *lreq)
2901 {
2902 	struct ceph_osd_client *osdc = lreq->osdc;
2903 	bool registered;
2904 
2905 	down_read(&osdc->lock);
2906 	registered = __linger_registered(lreq);
2907 	up_read(&osdc->lock);
2908 
2909 	return registered;
2910 }
2911 
linger_register(struct ceph_osd_linger_request * lreq)2912 static void linger_register(struct ceph_osd_linger_request *lreq)
2913 {
2914 	struct ceph_osd_client *osdc = lreq->osdc;
2915 
2916 	verify_osdc_wrlocked(osdc);
2917 	WARN_ON(lreq->linger_id);
2918 
2919 	linger_get(lreq);
2920 	lreq->linger_id = ++osdc->last_linger_id;
2921 	insert_linger_osdc(&osdc->linger_requests, lreq);
2922 }
2923 
linger_unregister(struct ceph_osd_linger_request * lreq)2924 static void linger_unregister(struct ceph_osd_linger_request *lreq)
2925 {
2926 	struct ceph_osd_client *osdc = lreq->osdc;
2927 
2928 	verify_osdc_wrlocked(osdc);
2929 
2930 	erase_linger_osdc(&osdc->linger_requests, lreq);
2931 	linger_put(lreq);
2932 }
2933 
cancel_linger_request(struct ceph_osd_request * req)2934 static void cancel_linger_request(struct ceph_osd_request *req)
2935 {
2936 	struct ceph_osd_linger_request *lreq = req->r_priv;
2937 
2938 	WARN_ON(!req->r_linger);
2939 	cancel_request(req);
2940 	linger_put(lreq);
2941 }
2942 
2943 struct linger_work {
2944 	struct work_struct work;
2945 	struct ceph_osd_linger_request *lreq;
2946 	struct list_head pending_item;
2947 	unsigned long queued_stamp;
2948 
2949 	union {
2950 		struct {
2951 			u64 notify_id;
2952 			u64 notifier_id;
2953 			void *payload; /* points into @msg front */
2954 			size_t payload_len;
2955 
2956 			struct ceph_msg *msg; /* for ceph_msg_put() */
2957 		} notify;
2958 		struct {
2959 			int err;
2960 		} error;
2961 	};
2962 };
2963 
lwork_alloc(struct ceph_osd_linger_request * lreq,work_func_t workfn)2964 static struct linger_work *lwork_alloc(struct ceph_osd_linger_request *lreq,
2965 				       work_func_t workfn)
2966 {
2967 	struct linger_work *lwork;
2968 
2969 	lwork = kzalloc(sizeof(*lwork), GFP_NOIO);
2970 	if (!lwork)
2971 		return NULL;
2972 
2973 	INIT_WORK(&lwork->work, workfn);
2974 	INIT_LIST_HEAD(&lwork->pending_item);
2975 	lwork->lreq = linger_get(lreq);
2976 
2977 	return lwork;
2978 }
2979 
lwork_free(struct linger_work * lwork)2980 static void lwork_free(struct linger_work *lwork)
2981 {
2982 	struct ceph_osd_linger_request *lreq = lwork->lreq;
2983 
2984 	mutex_lock(&lreq->lock);
2985 	list_del(&lwork->pending_item);
2986 	mutex_unlock(&lreq->lock);
2987 
2988 	linger_put(lreq);
2989 	kfree(lwork);
2990 }
2991 
lwork_queue(struct linger_work * lwork)2992 static void lwork_queue(struct linger_work *lwork)
2993 {
2994 	struct ceph_osd_linger_request *lreq = lwork->lreq;
2995 	struct ceph_osd_client *osdc = lreq->osdc;
2996 
2997 	verify_lreq_locked(lreq);
2998 	WARN_ON(!list_empty(&lwork->pending_item));
2999 
3000 	lwork->queued_stamp = jiffies;
3001 	list_add_tail(&lwork->pending_item, &lreq->pending_lworks);
3002 	queue_work(osdc->notify_wq, &lwork->work);
3003 }
3004 
do_watch_notify(struct work_struct * w)3005 static void do_watch_notify(struct work_struct *w)
3006 {
3007 	struct linger_work *lwork = container_of(w, struct linger_work, work);
3008 	struct ceph_osd_linger_request *lreq = lwork->lreq;
3009 
3010 	if (!linger_registered(lreq)) {
3011 		dout("%s lreq %p not registered\n", __func__, lreq);
3012 		goto out;
3013 	}
3014 
3015 	WARN_ON(!lreq->is_watch);
3016 	dout("%s lreq %p notify_id %llu notifier_id %llu payload_len %zu\n",
3017 	     __func__, lreq, lwork->notify.notify_id, lwork->notify.notifier_id,
3018 	     lwork->notify.payload_len);
3019 	lreq->wcb(lreq->data, lwork->notify.notify_id, lreq->linger_id,
3020 		  lwork->notify.notifier_id, lwork->notify.payload,
3021 		  lwork->notify.payload_len);
3022 
3023 out:
3024 	ceph_msg_put(lwork->notify.msg);
3025 	lwork_free(lwork);
3026 }
3027 
do_watch_error(struct work_struct * w)3028 static void do_watch_error(struct work_struct *w)
3029 {
3030 	struct linger_work *lwork = container_of(w, struct linger_work, work);
3031 	struct ceph_osd_linger_request *lreq = lwork->lreq;
3032 
3033 	if (!linger_registered(lreq)) {
3034 		dout("%s lreq %p not registered\n", __func__, lreq);
3035 		goto out;
3036 	}
3037 
3038 	dout("%s lreq %p err %d\n", __func__, lreq, lwork->error.err);
3039 	lreq->errcb(lreq->data, lreq->linger_id, lwork->error.err);
3040 
3041 out:
3042 	lwork_free(lwork);
3043 }
3044 
queue_watch_error(struct ceph_osd_linger_request * lreq)3045 static void queue_watch_error(struct ceph_osd_linger_request *lreq)
3046 {
3047 	struct linger_work *lwork;
3048 
3049 	lwork = lwork_alloc(lreq, do_watch_error);
3050 	if (!lwork) {
3051 		pr_err("failed to allocate error-lwork\n");
3052 		return;
3053 	}
3054 
3055 	lwork->error.err = lreq->last_error;
3056 	lwork_queue(lwork);
3057 }
3058 
linger_reg_commit_complete(struct ceph_osd_linger_request * lreq,int result)3059 static void linger_reg_commit_complete(struct ceph_osd_linger_request *lreq,
3060 				       int result)
3061 {
3062 	if (!completion_done(&lreq->reg_commit_wait)) {
3063 		lreq->reg_commit_error = (result <= 0 ? result : 0);
3064 		complete_all(&lreq->reg_commit_wait);
3065 	}
3066 }
3067 
linger_commit_cb(struct ceph_osd_request * req)3068 static void linger_commit_cb(struct ceph_osd_request *req)
3069 {
3070 	struct ceph_osd_linger_request *lreq = req->r_priv;
3071 
3072 	mutex_lock(&lreq->lock);
3073 	if (req != lreq->reg_req) {
3074 		dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n",
3075 		     __func__, lreq, lreq->linger_id, req, lreq->reg_req);
3076 		goto out;
3077 	}
3078 
3079 	dout("%s lreq %p linger_id %llu result %d\n", __func__, lreq,
3080 	     lreq->linger_id, req->r_result);
3081 	linger_reg_commit_complete(lreq, req->r_result);
3082 	lreq->committed = true;
3083 
3084 	if (!lreq->is_watch) {
3085 		struct ceph_osd_data *osd_data =
3086 		    osd_req_op_data(req, 0, notify, response_data);
3087 		void *p = page_address(osd_data->pages[0]);
3088 
3089 		WARN_ON(req->r_ops[0].op != CEPH_OSD_OP_NOTIFY ||
3090 			osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
3091 
3092 		/* make note of the notify_id */
3093 		if (req->r_ops[0].outdata_len >= sizeof(u64)) {
3094 			lreq->notify_id = ceph_decode_64(&p);
3095 			dout("lreq %p notify_id %llu\n", lreq,
3096 			     lreq->notify_id);
3097 		} else {
3098 			dout("lreq %p no notify_id\n", lreq);
3099 		}
3100 	}
3101 
3102 out:
3103 	mutex_unlock(&lreq->lock);
3104 	linger_put(lreq);
3105 }
3106 
normalize_watch_error(int err)3107 static int normalize_watch_error(int err)
3108 {
3109 	/*
3110 	 * Translate ENOENT -> ENOTCONN so that a delete->disconnection
3111 	 * notification and a failure to reconnect because we raced with
3112 	 * the delete appear the same to the user.
3113 	 */
3114 	if (err == -ENOENT)
3115 		err = -ENOTCONN;
3116 
3117 	return err;
3118 }
3119 
linger_reconnect_cb(struct ceph_osd_request * req)3120 static void linger_reconnect_cb(struct ceph_osd_request *req)
3121 {
3122 	struct ceph_osd_linger_request *lreq = req->r_priv;
3123 
3124 	mutex_lock(&lreq->lock);
3125 	if (req != lreq->reg_req) {
3126 		dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n",
3127 		     __func__, lreq, lreq->linger_id, req, lreq->reg_req);
3128 		goto out;
3129 	}
3130 
3131 	dout("%s lreq %p linger_id %llu result %d last_error %d\n", __func__,
3132 	     lreq, lreq->linger_id, req->r_result, lreq->last_error);
3133 	if (req->r_result < 0) {
3134 		if (!lreq->last_error) {
3135 			lreq->last_error = normalize_watch_error(req->r_result);
3136 			queue_watch_error(lreq);
3137 		}
3138 	}
3139 
3140 out:
3141 	mutex_unlock(&lreq->lock);
3142 	linger_put(lreq);
3143 }
3144 
send_linger(struct ceph_osd_linger_request * lreq)3145 static void send_linger(struct ceph_osd_linger_request *lreq)
3146 {
3147 	struct ceph_osd_client *osdc = lreq->osdc;
3148 	struct ceph_osd_request *req;
3149 	int ret;
3150 
3151 	verify_osdc_wrlocked(osdc);
3152 	mutex_lock(&lreq->lock);
3153 	dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3154 
3155 	if (lreq->reg_req) {
3156 		if (lreq->reg_req->r_osd)
3157 			cancel_linger_request(lreq->reg_req);
3158 		ceph_osdc_put_request(lreq->reg_req);
3159 	}
3160 
3161 	req = ceph_osdc_alloc_request(osdc, NULL, 1, true, GFP_NOIO);
3162 	BUG_ON(!req);
3163 
3164 	target_copy(&req->r_t, &lreq->t);
3165 	req->r_mtime = lreq->mtime;
3166 
3167 	if (lreq->is_watch && lreq->committed) {
3168 		osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_RECONNECT,
3169 				      lreq->linger_id, ++lreq->register_gen);
3170 		dout("lreq %p reconnect register_gen %u\n", lreq,
3171 		     req->r_ops[0].watch.gen);
3172 		req->r_callback = linger_reconnect_cb;
3173 	} else {
3174 		if (lreq->is_watch) {
3175 			osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_WATCH,
3176 					      lreq->linger_id, 0);
3177 		} else {
3178 			lreq->notify_id = 0;
3179 
3180 			refcount_inc(&lreq->request_pl->refcnt);
3181 			osd_req_op_notify_init(req, 0, lreq->linger_id,
3182 					       lreq->request_pl);
3183 			ceph_osd_data_pages_init(
3184 			    osd_req_op_data(req, 0, notify, response_data),
3185 			    lreq->notify_id_pages, PAGE_SIZE, 0, false, false);
3186 		}
3187 		dout("lreq %p register\n", lreq);
3188 		req->r_callback = linger_commit_cb;
3189 	}
3190 
3191 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
3192 	BUG_ON(ret);
3193 
3194 	req->r_priv = linger_get(lreq);
3195 	req->r_linger = true;
3196 	lreq->reg_req = req;
3197 	mutex_unlock(&lreq->lock);
3198 
3199 	submit_request(req, true);
3200 }
3201 
linger_ping_cb(struct ceph_osd_request * req)3202 static void linger_ping_cb(struct ceph_osd_request *req)
3203 {
3204 	struct ceph_osd_linger_request *lreq = req->r_priv;
3205 
3206 	mutex_lock(&lreq->lock);
3207 	if (req != lreq->ping_req) {
3208 		dout("%s lreq %p linger_id %llu unknown req (%p != %p)\n",
3209 		     __func__, lreq, lreq->linger_id, req, lreq->ping_req);
3210 		goto out;
3211 	}
3212 
3213 	dout("%s lreq %p linger_id %llu result %d ping_sent %lu last_error %d\n",
3214 	     __func__, lreq, lreq->linger_id, req->r_result, lreq->ping_sent,
3215 	     lreq->last_error);
3216 	if (lreq->register_gen == req->r_ops[0].watch.gen) {
3217 		if (!req->r_result) {
3218 			lreq->watch_valid_thru = lreq->ping_sent;
3219 		} else if (!lreq->last_error) {
3220 			lreq->last_error = normalize_watch_error(req->r_result);
3221 			queue_watch_error(lreq);
3222 		}
3223 	} else {
3224 		dout("lreq %p register_gen %u ignoring old pong %u\n", lreq,
3225 		     lreq->register_gen, req->r_ops[0].watch.gen);
3226 	}
3227 
3228 out:
3229 	mutex_unlock(&lreq->lock);
3230 	linger_put(lreq);
3231 }
3232 
send_linger_ping(struct ceph_osd_linger_request * lreq)3233 static void send_linger_ping(struct ceph_osd_linger_request *lreq)
3234 {
3235 	struct ceph_osd_client *osdc = lreq->osdc;
3236 	struct ceph_osd_request *req;
3237 	int ret;
3238 
3239 	if (ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD)) {
3240 		dout("%s PAUSERD\n", __func__);
3241 		return;
3242 	}
3243 
3244 	lreq->ping_sent = jiffies;
3245 	dout("%s lreq %p linger_id %llu ping_sent %lu register_gen %u\n",
3246 	     __func__, lreq, lreq->linger_id, lreq->ping_sent,
3247 	     lreq->register_gen);
3248 
3249 	if (lreq->ping_req) {
3250 		if (lreq->ping_req->r_osd)
3251 			cancel_linger_request(lreq->ping_req);
3252 		ceph_osdc_put_request(lreq->ping_req);
3253 	}
3254 
3255 	req = ceph_osdc_alloc_request(osdc, NULL, 1, true, GFP_NOIO);
3256 	BUG_ON(!req);
3257 
3258 	target_copy(&req->r_t, &lreq->t);
3259 	osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_PING, lreq->linger_id,
3260 			      lreq->register_gen);
3261 	req->r_callback = linger_ping_cb;
3262 
3263 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
3264 	BUG_ON(ret);
3265 
3266 	req->r_priv = linger_get(lreq);
3267 	req->r_linger = true;
3268 	lreq->ping_req = req;
3269 
3270 	ceph_osdc_get_request(req);
3271 	account_request(req);
3272 	req->r_tid = atomic64_inc_return(&osdc->last_tid);
3273 	link_request(lreq->osd, req);
3274 	send_request(req);
3275 }
3276 
linger_submit(struct ceph_osd_linger_request * lreq)3277 static void linger_submit(struct ceph_osd_linger_request *lreq)
3278 {
3279 	struct ceph_osd_client *osdc = lreq->osdc;
3280 	struct ceph_osd *osd;
3281 
3282 	down_write(&osdc->lock);
3283 	linger_register(lreq);
3284 
3285 	calc_target(osdc, &lreq->t, false);
3286 	osd = lookup_create_osd(osdc, lreq->t.osd, true);
3287 	link_linger(osd, lreq);
3288 
3289 	send_linger(lreq);
3290 	up_write(&osdc->lock);
3291 }
3292 
cancel_linger_map_check(struct ceph_osd_linger_request * lreq)3293 static void cancel_linger_map_check(struct ceph_osd_linger_request *lreq)
3294 {
3295 	struct ceph_osd_client *osdc = lreq->osdc;
3296 	struct ceph_osd_linger_request *lookup_lreq;
3297 
3298 	verify_osdc_wrlocked(osdc);
3299 
3300 	lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3301 				       lreq->linger_id);
3302 	if (!lookup_lreq)
3303 		return;
3304 
3305 	WARN_ON(lookup_lreq != lreq);
3306 	erase_linger_mc(&osdc->linger_map_checks, lreq);
3307 	linger_put(lreq);
3308 }
3309 
3310 /*
3311  * @lreq has to be both registered and linked.
3312  */
__linger_cancel(struct ceph_osd_linger_request * lreq)3313 static void __linger_cancel(struct ceph_osd_linger_request *lreq)
3314 {
3315 	if (lreq->ping_req && lreq->ping_req->r_osd)
3316 		cancel_linger_request(lreq->ping_req);
3317 	if (lreq->reg_req && lreq->reg_req->r_osd)
3318 		cancel_linger_request(lreq->reg_req);
3319 	cancel_linger_map_check(lreq);
3320 	unlink_linger(lreq->osd, lreq);
3321 	linger_unregister(lreq);
3322 }
3323 
linger_cancel(struct ceph_osd_linger_request * lreq)3324 static void linger_cancel(struct ceph_osd_linger_request *lreq)
3325 {
3326 	struct ceph_osd_client *osdc = lreq->osdc;
3327 
3328 	down_write(&osdc->lock);
3329 	if (__linger_registered(lreq))
3330 		__linger_cancel(lreq);
3331 	up_write(&osdc->lock);
3332 }
3333 
3334 static void send_linger_map_check(struct ceph_osd_linger_request *lreq);
3335 
check_linger_pool_dne(struct ceph_osd_linger_request * lreq)3336 static void check_linger_pool_dne(struct ceph_osd_linger_request *lreq)
3337 {
3338 	struct ceph_osd_client *osdc = lreq->osdc;
3339 	struct ceph_osdmap *map = osdc->osdmap;
3340 
3341 	verify_osdc_wrlocked(osdc);
3342 	WARN_ON(!map->epoch);
3343 
3344 	if (lreq->register_gen) {
3345 		lreq->map_dne_bound = map->epoch;
3346 		dout("%s lreq %p linger_id %llu pool disappeared\n", __func__,
3347 		     lreq, lreq->linger_id);
3348 	} else {
3349 		dout("%s lreq %p linger_id %llu map_dne_bound %u have %u\n",
3350 		     __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3351 		     map->epoch);
3352 	}
3353 
3354 	if (lreq->map_dne_bound) {
3355 		if (map->epoch >= lreq->map_dne_bound) {
3356 			/* we had a new enough map */
3357 			pr_info("linger_id %llu pool does not exist\n",
3358 				lreq->linger_id);
3359 			linger_reg_commit_complete(lreq, -ENOENT);
3360 			__linger_cancel(lreq);
3361 		}
3362 	} else {
3363 		send_linger_map_check(lreq);
3364 	}
3365 }
3366 
linger_map_check_cb(struct ceph_mon_generic_request * greq)3367 static void linger_map_check_cb(struct ceph_mon_generic_request *greq)
3368 {
3369 	struct ceph_osd_client *osdc = &greq->monc->client->osdc;
3370 	struct ceph_osd_linger_request *lreq;
3371 	u64 linger_id = greq->private_data;
3372 
3373 	WARN_ON(greq->result || !greq->u.newest);
3374 
3375 	down_write(&osdc->lock);
3376 	lreq = lookup_linger_mc(&osdc->linger_map_checks, linger_id);
3377 	if (!lreq) {
3378 		dout("%s linger_id %llu dne\n", __func__, linger_id);
3379 		goto out_unlock;
3380 	}
3381 
3382 	dout("%s lreq %p linger_id %llu map_dne_bound %u newest %llu\n",
3383 	     __func__, lreq, lreq->linger_id, lreq->map_dne_bound,
3384 	     greq->u.newest);
3385 	if (!lreq->map_dne_bound)
3386 		lreq->map_dne_bound = greq->u.newest;
3387 	erase_linger_mc(&osdc->linger_map_checks, lreq);
3388 	check_linger_pool_dne(lreq);
3389 
3390 	linger_put(lreq);
3391 out_unlock:
3392 	up_write(&osdc->lock);
3393 }
3394 
send_linger_map_check(struct ceph_osd_linger_request * lreq)3395 static void send_linger_map_check(struct ceph_osd_linger_request *lreq)
3396 {
3397 	struct ceph_osd_client *osdc = lreq->osdc;
3398 	struct ceph_osd_linger_request *lookup_lreq;
3399 	int ret;
3400 
3401 	verify_osdc_wrlocked(osdc);
3402 
3403 	lookup_lreq = lookup_linger_mc(&osdc->linger_map_checks,
3404 				       lreq->linger_id);
3405 	if (lookup_lreq) {
3406 		WARN_ON(lookup_lreq != lreq);
3407 		return;
3408 	}
3409 
3410 	linger_get(lreq);
3411 	insert_linger_mc(&osdc->linger_map_checks, lreq);
3412 	ret = ceph_monc_get_version_async(&osdc->client->monc, "osdmap",
3413 					  linger_map_check_cb, lreq->linger_id);
3414 	WARN_ON(ret);
3415 }
3416 
linger_reg_commit_wait(struct ceph_osd_linger_request * lreq)3417 static int linger_reg_commit_wait(struct ceph_osd_linger_request *lreq)
3418 {
3419 	int ret;
3420 
3421 	dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3422 	ret = wait_for_completion_killable(&lreq->reg_commit_wait);
3423 	return ret ?: lreq->reg_commit_error;
3424 }
3425 
linger_notify_finish_wait(struct ceph_osd_linger_request * lreq,unsigned long timeout)3426 static int linger_notify_finish_wait(struct ceph_osd_linger_request *lreq,
3427 				     unsigned long timeout)
3428 {
3429 	long left;
3430 
3431 	dout("%s lreq %p linger_id %llu\n", __func__, lreq, lreq->linger_id);
3432 	left = wait_for_completion_killable_timeout(&lreq->notify_finish_wait,
3433 						ceph_timeout_jiffies(timeout));
3434 	if (left <= 0)
3435 		left = left ?: -ETIMEDOUT;
3436 	else
3437 		left = lreq->notify_finish_error; /* completed */
3438 
3439 	return left;
3440 }
3441 
3442 /*
3443  * Timeout callback, called every N seconds.  When 1 or more OSD
3444  * requests has been active for more than N seconds, we send a keepalive
3445  * (tag + timestamp) to its OSD to ensure any communications channel
3446  * reset is detected.
3447  */
handle_timeout(struct work_struct * work)3448 static void handle_timeout(struct work_struct *work)
3449 {
3450 	struct ceph_osd_client *osdc =
3451 		container_of(work, struct ceph_osd_client, timeout_work.work);
3452 	struct ceph_options *opts = osdc->client->options;
3453 	unsigned long cutoff = jiffies - opts->osd_keepalive_timeout;
3454 	unsigned long expiry_cutoff = jiffies - opts->osd_request_timeout;
3455 	LIST_HEAD(slow_osds);
3456 	struct rb_node *n, *p;
3457 
3458 	dout("%s osdc %p\n", __func__, osdc);
3459 	down_write(&osdc->lock);
3460 
3461 	/*
3462 	 * ping osds that are a bit slow.  this ensures that if there
3463 	 * is a break in the TCP connection we will notice, and reopen
3464 	 * a connection with that osd (from the fault callback).
3465 	 */
3466 	for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
3467 		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
3468 		bool found = false;
3469 
3470 		for (p = rb_first(&osd->o_requests); p; ) {
3471 			struct ceph_osd_request *req =
3472 			    rb_entry(p, struct ceph_osd_request, r_node);
3473 
3474 			p = rb_next(p); /* abort_request() */
3475 
3476 			if (time_before(req->r_stamp, cutoff)) {
3477 				dout(" req %p tid %llu on osd%d is laggy\n",
3478 				     req, req->r_tid, osd->o_osd);
3479 				found = true;
3480 			}
3481 			if (opts->osd_request_timeout &&
3482 			    time_before(req->r_start_stamp, expiry_cutoff)) {
3483 				pr_err_ratelimited("tid %llu on osd%d timeout\n",
3484 				       req->r_tid, osd->o_osd);
3485 				abort_request(req, -ETIMEDOUT);
3486 			}
3487 		}
3488 		for (p = rb_first(&osd->o_linger_requests); p; p = rb_next(p)) {
3489 			struct ceph_osd_linger_request *lreq =
3490 			    rb_entry(p, struct ceph_osd_linger_request, node);
3491 
3492 			dout(" lreq %p linger_id %llu is served by osd%d\n",
3493 			     lreq, lreq->linger_id, osd->o_osd);
3494 			found = true;
3495 
3496 			mutex_lock(&lreq->lock);
3497 			if (lreq->is_watch && lreq->committed && !lreq->last_error)
3498 				send_linger_ping(lreq);
3499 			mutex_unlock(&lreq->lock);
3500 		}
3501 
3502 		if (found)
3503 			list_move_tail(&osd->o_keepalive_item, &slow_osds);
3504 	}
3505 
3506 	if (opts->osd_request_timeout) {
3507 		for (p = rb_first(&osdc->homeless_osd.o_requests); p; ) {
3508 			struct ceph_osd_request *req =
3509 			    rb_entry(p, struct ceph_osd_request, r_node);
3510 
3511 			p = rb_next(p); /* abort_request() */
3512 
3513 			if (time_before(req->r_start_stamp, expiry_cutoff)) {
3514 				pr_err_ratelimited("tid %llu on osd%d timeout\n",
3515 				       req->r_tid, osdc->homeless_osd.o_osd);
3516 				abort_request(req, -ETIMEDOUT);
3517 			}
3518 		}
3519 	}
3520 
3521 	if (atomic_read(&osdc->num_homeless) || !list_empty(&slow_osds))
3522 		maybe_request_map(osdc);
3523 
3524 	while (!list_empty(&slow_osds)) {
3525 		struct ceph_osd *osd = list_first_entry(&slow_osds,
3526 							struct ceph_osd,
3527 							o_keepalive_item);
3528 		list_del_init(&osd->o_keepalive_item);
3529 		ceph_con_keepalive(&osd->o_con);
3530 	}
3531 
3532 	up_write(&osdc->lock);
3533 	schedule_delayed_work(&osdc->timeout_work,
3534 			      osdc->client->options->osd_keepalive_timeout);
3535 }
3536 
handle_osds_timeout(struct work_struct * work)3537 static void handle_osds_timeout(struct work_struct *work)
3538 {
3539 	struct ceph_osd_client *osdc =
3540 		container_of(work, struct ceph_osd_client,
3541 			     osds_timeout_work.work);
3542 	unsigned long delay = osdc->client->options->osd_idle_ttl / 4;
3543 	struct ceph_osd *osd, *nosd;
3544 
3545 	dout("%s osdc %p\n", __func__, osdc);
3546 	down_write(&osdc->lock);
3547 	list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) {
3548 		if (time_before(jiffies, osd->lru_ttl))
3549 			break;
3550 
3551 		WARN_ON(!RB_EMPTY_ROOT(&osd->o_requests));
3552 		WARN_ON(!RB_EMPTY_ROOT(&osd->o_linger_requests));
3553 		close_osd(osd);
3554 	}
3555 
3556 	up_write(&osdc->lock);
3557 	schedule_delayed_work(&osdc->osds_timeout_work,
3558 			      round_jiffies_relative(delay));
3559 }
3560 
ceph_oloc_decode(void ** p,void * end,struct ceph_object_locator * oloc)3561 static int ceph_oloc_decode(void **p, void *end,
3562 			    struct ceph_object_locator *oloc)
3563 {
3564 	u8 struct_v, struct_cv;
3565 	u32 len;
3566 	void *struct_end;
3567 	int ret = 0;
3568 
3569 	ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3570 	struct_v = ceph_decode_8(p);
3571 	struct_cv = ceph_decode_8(p);
3572 	if (struct_v < 3) {
3573 		pr_warn("got v %d < 3 cv %d of ceph_object_locator\n",
3574 			struct_v, struct_cv);
3575 		goto e_inval;
3576 	}
3577 	if (struct_cv > 6) {
3578 		pr_warn("got v %d cv %d > 6 of ceph_object_locator\n",
3579 			struct_v, struct_cv);
3580 		goto e_inval;
3581 	}
3582 	len = ceph_decode_32(p);
3583 	ceph_decode_need(p, end, len, e_inval);
3584 	struct_end = *p + len;
3585 
3586 	oloc->pool = ceph_decode_64(p);
3587 	*p += 4; /* skip preferred */
3588 
3589 	len = ceph_decode_32(p);
3590 	if (len > 0) {
3591 		pr_warn("ceph_object_locator::key is set\n");
3592 		goto e_inval;
3593 	}
3594 
3595 	if (struct_v >= 5) {
3596 		bool changed = false;
3597 
3598 		len = ceph_decode_32(p);
3599 		if (len > 0) {
3600 			ceph_decode_need(p, end, len, e_inval);
3601 			if (!oloc->pool_ns ||
3602 			    ceph_compare_string(oloc->pool_ns, *p, len))
3603 				changed = true;
3604 			*p += len;
3605 		} else {
3606 			if (oloc->pool_ns)
3607 				changed = true;
3608 		}
3609 		if (changed) {
3610 			/* redirect changes namespace */
3611 			pr_warn("ceph_object_locator::nspace is changed\n");
3612 			goto e_inval;
3613 		}
3614 	}
3615 
3616 	if (struct_v >= 6) {
3617 		s64 hash = ceph_decode_64(p);
3618 		if (hash != -1) {
3619 			pr_warn("ceph_object_locator::hash is set\n");
3620 			goto e_inval;
3621 		}
3622 	}
3623 
3624 	/* skip the rest */
3625 	*p = struct_end;
3626 out:
3627 	return ret;
3628 
3629 e_inval:
3630 	ret = -EINVAL;
3631 	goto out;
3632 }
3633 
ceph_redirect_decode(void ** p,void * end,struct ceph_request_redirect * redir)3634 static int ceph_redirect_decode(void **p, void *end,
3635 				struct ceph_request_redirect *redir)
3636 {
3637 	u8 struct_v, struct_cv;
3638 	u32 len;
3639 	void *struct_end;
3640 	int ret;
3641 
3642 	ceph_decode_need(p, end, 1 + 1 + 4, e_inval);
3643 	struct_v = ceph_decode_8(p);
3644 	struct_cv = ceph_decode_8(p);
3645 	if (struct_cv > 1) {
3646 		pr_warn("got v %d cv %d > 1 of ceph_request_redirect\n",
3647 			struct_v, struct_cv);
3648 		goto e_inval;
3649 	}
3650 	len = ceph_decode_32(p);
3651 	ceph_decode_need(p, end, len, e_inval);
3652 	struct_end = *p + len;
3653 
3654 	ret = ceph_oloc_decode(p, end, &redir->oloc);
3655 	if (ret)
3656 		goto out;
3657 
3658 	len = ceph_decode_32(p);
3659 	if (len > 0) {
3660 		pr_warn("ceph_request_redirect::object_name is set\n");
3661 		goto e_inval;
3662 	}
3663 
3664 	/* skip the rest */
3665 	*p = struct_end;
3666 out:
3667 	return ret;
3668 
3669 e_inval:
3670 	ret = -EINVAL;
3671 	goto out;
3672 }
3673 
3674 struct MOSDOpReply {
3675 	struct ceph_pg pgid;
3676 	u64 flags;
3677 	int result;
3678 	u32 epoch;
3679 	int num_ops;
3680 	u32 outdata_len[CEPH_OSD_MAX_OPS];
3681 	s32 rval[CEPH_OSD_MAX_OPS];
3682 	int retry_attempt;
3683 	struct ceph_eversion replay_version;
3684 	u64 user_version;
3685 	struct ceph_request_redirect redirect;
3686 };
3687 
decode_MOSDOpReply(const struct ceph_msg * msg,struct MOSDOpReply * m)3688 static int decode_MOSDOpReply(const struct ceph_msg *msg, struct MOSDOpReply *m)
3689 {
3690 	void *p = msg->front.iov_base;
3691 	void *const end = p + msg->front.iov_len;
3692 	u16 version = le16_to_cpu(msg->hdr.version);
3693 	struct ceph_eversion bad_replay_version;
3694 	u8 decode_redir;
3695 	u32 len;
3696 	int ret;
3697 	int i;
3698 
3699 	ceph_decode_32_safe(&p, end, len, e_inval);
3700 	ceph_decode_need(&p, end, len, e_inval);
3701 	p += len; /* skip oid */
3702 
3703 	ret = ceph_decode_pgid(&p, end, &m->pgid);
3704 	if (ret)
3705 		return ret;
3706 
3707 	ceph_decode_64_safe(&p, end, m->flags, e_inval);
3708 	ceph_decode_32_safe(&p, end, m->result, e_inval);
3709 	ceph_decode_need(&p, end, sizeof(bad_replay_version), e_inval);
3710 	memcpy(&bad_replay_version, p, sizeof(bad_replay_version));
3711 	p += sizeof(bad_replay_version);
3712 	ceph_decode_32_safe(&p, end, m->epoch, e_inval);
3713 
3714 	ceph_decode_32_safe(&p, end, m->num_ops, e_inval);
3715 	if (m->num_ops > ARRAY_SIZE(m->outdata_len))
3716 		goto e_inval;
3717 
3718 	ceph_decode_need(&p, end, m->num_ops * sizeof(struct ceph_osd_op),
3719 			 e_inval);
3720 	for (i = 0; i < m->num_ops; i++) {
3721 		struct ceph_osd_op *op = p;
3722 
3723 		m->outdata_len[i] = le32_to_cpu(op->payload_len);
3724 		p += sizeof(*op);
3725 	}
3726 
3727 	ceph_decode_32_safe(&p, end, m->retry_attempt, e_inval);
3728 	for (i = 0; i < m->num_ops; i++)
3729 		ceph_decode_32_safe(&p, end, m->rval[i], e_inval);
3730 
3731 	if (version >= 5) {
3732 		ceph_decode_need(&p, end, sizeof(m->replay_version), e_inval);
3733 		memcpy(&m->replay_version, p, sizeof(m->replay_version));
3734 		p += sizeof(m->replay_version);
3735 		ceph_decode_64_safe(&p, end, m->user_version, e_inval);
3736 	} else {
3737 		m->replay_version = bad_replay_version; /* struct */
3738 		m->user_version = le64_to_cpu(m->replay_version.version);
3739 	}
3740 
3741 	if (version >= 6) {
3742 		if (version >= 7)
3743 			ceph_decode_8_safe(&p, end, decode_redir, e_inval);
3744 		else
3745 			decode_redir = 1;
3746 	} else {
3747 		decode_redir = 0;
3748 	}
3749 
3750 	if (decode_redir) {
3751 		ret = ceph_redirect_decode(&p, end, &m->redirect);
3752 		if (ret)
3753 			return ret;
3754 	} else {
3755 		ceph_oloc_init(&m->redirect.oloc);
3756 	}
3757 
3758 	return 0;
3759 
3760 e_inval:
3761 	return -EINVAL;
3762 }
3763 
3764 /*
3765  * Handle MOSDOpReply.  Set ->r_result and call the callback if it is
3766  * specified.
3767  */
handle_reply(struct ceph_osd * osd,struct ceph_msg * msg)3768 static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg)
3769 {
3770 	struct ceph_osd_client *osdc = osd->o_osdc;
3771 	struct ceph_osd_request *req;
3772 	struct MOSDOpReply m;
3773 	u64 tid = le64_to_cpu(msg->hdr.tid);
3774 	u32 data_len = 0;
3775 	int ret;
3776 	int i;
3777 
3778 	dout("%s msg %p tid %llu\n", __func__, msg, tid);
3779 
3780 	down_read(&osdc->lock);
3781 	if (!osd_registered(osd)) {
3782 		dout("%s osd%d unknown\n", __func__, osd->o_osd);
3783 		goto out_unlock_osdc;
3784 	}
3785 	WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
3786 
3787 	mutex_lock(&osd->lock);
3788 	req = lookup_request(&osd->o_requests, tid);
3789 	if (!req) {
3790 		dout("%s osd%d tid %llu unknown\n", __func__, osd->o_osd, tid);
3791 		goto out_unlock_session;
3792 	}
3793 
3794 	m.redirect.oloc.pool_ns = req->r_t.target_oloc.pool_ns;
3795 	ret = decode_MOSDOpReply(msg, &m);
3796 	m.redirect.oloc.pool_ns = NULL;
3797 	if (ret) {
3798 		pr_err("failed to decode MOSDOpReply for tid %llu: %d\n",
3799 		       req->r_tid, ret);
3800 		ceph_msg_dump(msg);
3801 		goto fail_request;
3802 	}
3803 	dout("%s req %p tid %llu flags 0x%llx pgid %llu.%x epoch %u attempt %d v %u'%llu uv %llu\n",
3804 	     __func__, req, req->r_tid, m.flags, m.pgid.pool, m.pgid.seed,
3805 	     m.epoch, m.retry_attempt, le32_to_cpu(m.replay_version.epoch),
3806 	     le64_to_cpu(m.replay_version.version), m.user_version);
3807 
3808 	if (m.retry_attempt >= 0) {
3809 		if (m.retry_attempt != req->r_attempts - 1) {
3810 			dout("req %p tid %llu retry_attempt %d != %d, ignoring\n",
3811 			     req, req->r_tid, m.retry_attempt,
3812 			     req->r_attempts - 1);
3813 			goto out_unlock_session;
3814 		}
3815 	} else {
3816 		WARN_ON(1); /* MOSDOpReply v4 is assumed */
3817 	}
3818 
3819 	if (!ceph_oloc_empty(&m.redirect.oloc)) {
3820 		dout("req %p tid %llu redirect pool %lld\n", req, req->r_tid,
3821 		     m.redirect.oloc.pool);
3822 		unlink_request(osd, req);
3823 		mutex_unlock(&osd->lock);
3824 
3825 		/*
3826 		 * Not ceph_oloc_copy() - changing pool_ns is not
3827 		 * supported.
3828 		 */
3829 		req->r_t.target_oloc.pool = m.redirect.oloc.pool;
3830 		req->r_flags |= CEPH_OSD_FLAG_REDIRECTED |
3831 				CEPH_OSD_FLAG_IGNORE_OVERLAY |
3832 				CEPH_OSD_FLAG_IGNORE_CACHE;
3833 		req->r_tid = 0;
3834 		__submit_request(req, false);
3835 		goto out_unlock_osdc;
3836 	}
3837 
3838 	if (m.result == -EAGAIN) {
3839 		dout("req %p tid %llu EAGAIN\n", req, req->r_tid);
3840 		unlink_request(osd, req);
3841 		mutex_unlock(&osd->lock);
3842 
3843 		/*
3844 		 * The object is missing on the replica or not (yet)
3845 		 * readable.  Clear pgid to force a resend to the primary
3846 		 * via legacy_change.
3847 		 */
3848 		req->r_t.pgid.pool = 0;
3849 		req->r_t.pgid.seed = 0;
3850 		WARN_ON(!req->r_t.used_replica);
3851 		req->r_flags &= ~(CEPH_OSD_FLAG_BALANCE_READS |
3852 				  CEPH_OSD_FLAG_LOCALIZE_READS);
3853 		req->r_tid = 0;
3854 		__submit_request(req, false);
3855 		goto out_unlock_osdc;
3856 	}
3857 
3858 	if (m.num_ops != req->r_num_ops) {
3859 		pr_err("num_ops %d != %d for tid %llu\n", m.num_ops,
3860 		       req->r_num_ops, req->r_tid);
3861 		goto fail_request;
3862 	}
3863 	for (i = 0; i < req->r_num_ops; i++) {
3864 		dout(" req %p tid %llu op %d rval %d len %u\n", req,
3865 		     req->r_tid, i, m.rval[i], m.outdata_len[i]);
3866 		req->r_ops[i].rval = m.rval[i];
3867 		req->r_ops[i].outdata_len = m.outdata_len[i];
3868 		data_len += m.outdata_len[i];
3869 	}
3870 	if (data_len != le32_to_cpu(msg->hdr.data_len)) {
3871 		pr_err("sum of lens %u != %u for tid %llu\n", data_len,
3872 		       le32_to_cpu(msg->hdr.data_len), req->r_tid);
3873 		goto fail_request;
3874 	}
3875 	dout("%s req %p tid %llu result %d data_len %u\n", __func__,
3876 	     req, req->r_tid, m.result, data_len);
3877 
3878 	/*
3879 	 * Since we only ever request ONDISK, we should only ever get
3880 	 * one (type of) reply back.
3881 	 */
3882 	WARN_ON(!(m.flags & CEPH_OSD_FLAG_ONDISK));
3883 	req->r_version = m.user_version;
3884 	req->r_result = m.result ?: data_len;
3885 	finish_request(req);
3886 	mutex_unlock(&osd->lock);
3887 	up_read(&osdc->lock);
3888 
3889 	__complete_request(req);
3890 	return;
3891 
3892 fail_request:
3893 	complete_request(req, -EIO);
3894 out_unlock_session:
3895 	mutex_unlock(&osd->lock);
3896 out_unlock_osdc:
3897 	up_read(&osdc->lock);
3898 }
3899 
set_pool_was_full(struct ceph_osd_client * osdc)3900 static void set_pool_was_full(struct ceph_osd_client *osdc)
3901 {
3902 	struct rb_node *n;
3903 
3904 	for (n = rb_first(&osdc->osdmap->pg_pools); n; n = rb_next(n)) {
3905 		struct ceph_pg_pool_info *pi =
3906 		    rb_entry(n, struct ceph_pg_pool_info, node);
3907 
3908 		pi->was_full = __pool_full(pi);
3909 	}
3910 }
3911 
pool_cleared_full(struct ceph_osd_client * osdc,s64 pool_id)3912 static bool pool_cleared_full(struct ceph_osd_client *osdc, s64 pool_id)
3913 {
3914 	struct ceph_pg_pool_info *pi;
3915 
3916 	pi = ceph_pg_pool_by_id(osdc->osdmap, pool_id);
3917 	if (!pi)
3918 		return false;
3919 
3920 	return pi->was_full && !__pool_full(pi);
3921 }
3922 
3923 static enum calc_target_result
recalc_linger_target(struct ceph_osd_linger_request * lreq)3924 recalc_linger_target(struct ceph_osd_linger_request *lreq)
3925 {
3926 	struct ceph_osd_client *osdc = lreq->osdc;
3927 	enum calc_target_result ct_res;
3928 
3929 	ct_res = calc_target(osdc, &lreq->t, true);
3930 	if (ct_res == CALC_TARGET_NEED_RESEND) {
3931 		struct ceph_osd *osd;
3932 
3933 		osd = lookup_create_osd(osdc, lreq->t.osd, true);
3934 		if (osd != lreq->osd) {
3935 			unlink_linger(lreq->osd, lreq);
3936 			link_linger(osd, lreq);
3937 		}
3938 	}
3939 
3940 	return ct_res;
3941 }
3942 
3943 /*
3944  * Requeue requests whose mapping to an OSD has changed.
3945  */
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)3946 static void scan_requests(struct ceph_osd *osd,
3947 			  bool force_resend,
3948 			  bool cleared_full,
3949 			  bool check_pool_cleared_full,
3950 			  struct rb_root *need_resend,
3951 			  struct list_head *need_resend_linger)
3952 {
3953 	struct ceph_osd_client *osdc = osd->o_osdc;
3954 	struct rb_node *n;
3955 	bool force_resend_writes;
3956 
3957 	for (n = rb_first(&osd->o_linger_requests); n; ) {
3958 		struct ceph_osd_linger_request *lreq =
3959 		    rb_entry(n, struct ceph_osd_linger_request, node);
3960 		enum calc_target_result ct_res;
3961 
3962 		n = rb_next(n); /* recalc_linger_target() */
3963 
3964 		dout("%s lreq %p linger_id %llu\n", __func__, lreq,
3965 		     lreq->linger_id);
3966 		ct_res = recalc_linger_target(lreq);
3967 		switch (ct_res) {
3968 		case CALC_TARGET_NO_ACTION:
3969 			force_resend_writes = cleared_full ||
3970 			    (check_pool_cleared_full &&
3971 			     pool_cleared_full(osdc, lreq->t.base_oloc.pool));
3972 			if (!force_resend && !force_resend_writes)
3973 				break;
3974 
3975 			fallthrough;
3976 		case CALC_TARGET_NEED_RESEND:
3977 			cancel_linger_map_check(lreq);
3978 			/*
3979 			 * scan_requests() for the previous epoch(s)
3980 			 * may have already added it to the list, since
3981 			 * it's not unlinked here.
3982 			 */
3983 			if (list_empty(&lreq->scan_item))
3984 				list_add_tail(&lreq->scan_item, need_resend_linger);
3985 			break;
3986 		case CALC_TARGET_POOL_DNE:
3987 			list_del_init(&lreq->scan_item);
3988 			check_linger_pool_dne(lreq);
3989 			break;
3990 		}
3991 	}
3992 
3993 	for (n = rb_first(&osd->o_requests); n; ) {
3994 		struct ceph_osd_request *req =
3995 		    rb_entry(n, struct ceph_osd_request, r_node);
3996 		enum calc_target_result ct_res;
3997 
3998 		n = rb_next(n); /* unlink_request(), check_pool_dne() */
3999 
4000 		dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
4001 		ct_res = calc_target(osdc, &req->r_t, false);
4002 		switch (ct_res) {
4003 		case CALC_TARGET_NO_ACTION:
4004 			force_resend_writes = cleared_full ||
4005 			    (check_pool_cleared_full &&
4006 			     pool_cleared_full(osdc, req->r_t.base_oloc.pool));
4007 			if (!force_resend &&
4008 			    (!(req->r_flags & CEPH_OSD_FLAG_WRITE) ||
4009 			     !force_resend_writes))
4010 				break;
4011 
4012 			fallthrough;
4013 		case CALC_TARGET_NEED_RESEND:
4014 			cancel_map_check(req);
4015 			unlink_request(osd, req);
4016 			insert_request(need_resend, req);
4017 			break;
4018 		case CALC_TARGET_POOL_DNE:
4019 			check_pool_dne(req);
4020 			break;
4021 		}
4022 	}
4023 }
4024 
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)4025 static int handle_one_map(struct ceph_osd_client *osdc,
4026 			  void *p, void *end, bool incremental,
4027 			  struct rb_root *need_resend,
4028 			  struct list_head *need_resend_linger)
4029 {
4030 	struct ceph_osdmap *newmap;
4031 	struct rb_node *n;
4032 	bool skipped_map = false;
4033 	bool was_full;
4034 
4035 	was_full = ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
4036 	set_pool_was_full(osdc);
4037 
4038 	if (incremental)
4039 		newmap = osdmap_apply_incremental(&p, end,
4040 						  ceph_msgr2(osdc->client),
4041 						  osdc->osdmap);
4042 	else
4043 		newmap = ceph_osdmap_decode(&p, end, ceph_msgr2(osdc->client));
4044 	if (IS_ERR(newmap))
4045 		return PTR_ERR(newmap);
4046 
4047 	if (newmap != osdc->osdmap) {
4048 		/*
4049 		 * Preserve ->was_full before destroying the old map.
4050 		 * For pools that weren't in the old map, ->was_full
4051 		 * should be false.
4052 		 */
4053 		for (n = rb_first(&newmap->pg_pools); n; n = rb_next(n)) {
4054 			struct ceph_pg_pool_info *pi =
4055 			    rb_entry(n, struct ceph_pg_pool_info, node);
4056 			struct ceph_pg_pool_info *old_pi;
4057 
4058 			old_pi = ceph_pg_pool_by_id(osdc->osdmap, pi->id);
4059 			if (old_pi)
4060 				pi->was_full = old_pi->was_full;
4061 			else
4062 				WARN_ON(pi->was_full);
4063 		}
4064 
4065 		if (osdc->osdmap->epoch &&
4066 		    osdc->osdmap->epoch + 1 < newmap->epoch) {
4067 			WARN_ON(incremental);
4068 			skipped_map = true;
4069 		}
4070 
4071 		ceph_osdmap_destroy(osdc->osdmap);
4072 		osdc->osdmap = newmap;
4073 	}
4074 
4075 	was_full &= !ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL);
4076 	scan_requests(&osdc->homeless_osd, skipped_map, was_full, true,
4077 		      need_resend, need_resend_linger);
4078 
4079 	for (n = rb_first(&osdc->osds); n; ) {
4080 		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
4081 
4082 		n = rb_next(n); /* close_osd() */
4083 
4084 		scan_requests(osd, skipped_map, was_full, true, need_resend,
4085 			      need_resend_linger);
4086 		if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) ||
4087 		    memcmp(&osd->o_con.peer_addr,
4088 			   ceph_osd_addr(osdc->osdmap, osd->o_osd),
4089 			   sizeof(struct ceph_entity_addr)))
4090 			close_osd(osd);
4091 	}
4092 
4093 	return 0;
4094 }
4095 
kick_requests(struct ceph_osd_client * osdc,struct rb_root * need_resend,struct list_head * need_resend_linger)4096 static void kick_requests(struct ceph_osd_client *osdc,
4097 			  struct rb_root *need_resend,
4098 			  struct list_head *need_resend_linger)
4099 {
4100 	struct ceph_osd_linger_request *lreq, *nlreq;
4101 	enum calc_target_result ct_res;
4102 	struct rb_node *n;
4103 
4104 	/* make sure need_resend targets reflect latest map */
4105 	for (n = rb_first(need_resend); n; ) {
4106 		struct ceph_osd_request *req =
4107 		    rb_entry(n, struct ceph_osd_request, r_node);
4108 
4109 		n = rb_next(n);
4110 
4111 		if (req->r_t.epoch < osdc->osdmap->epoch) {
4112 			ct_res = calc_target(osdc, &req->r_t, false);
4113 			if (ct_res == CALC_TARGET_POOL_DNE) {
4114 				erase_request(need_resend, req);
4115 				check_pool_dne(req);
4116 			}
4117 		}
4118 	}
4119 
4120 	for (n = rb_first(need_resend); n; ) {
4121 		struct ceph_osd_request *req =
4122 		    rb_entry(n, struct ceph_osd_request, r_node);
4123 		struct ceph_osd *osd;
4124 
4125 		n = rb_next(n);
4126 		erase_request(need_resend, req); /* before link_request() */
4127 
4128 		osd = lookup_create_osd(osdc, req->r_t.osd, true);
4129 		link_request(osd, req);
4130 		if (!req->r_linger) {
4131 			if (!osd_homeless(osd) && !req->r_t.paused)
4132 				send_request(req);
4133 		} else {
4134 			cancel_linger_request(req);
4135 		}
4136 	}
4137 
4138 	list_for_each_entry_safe(lreq, nlreq, need_resend_linger, scan_item) {
4139 		if (!osd_homeless(lreq->osd))
4140 			send_linger(lreq);
4141 
4142 		list_del_init(&lreq->scan_item);
4143 	}
4144 }
4145 
4146 /*
4147  * Process updated osd map.
4148  *
4149  * The message contains any number of incremental and full maps, normally
4150  * indicating some sort of topology change in the cluster.  Kick requests
4151  * off to different OSDs as needed.
4152  */
ceph_osdc_handle_map(struct ceph_osd_client * osdc,struct ceph_msg * msg)4153 void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
4154 {
4155 	void *p = msg->front.iov_base;
4156 	void *const end = p + msg->front.iov_len;
4157 	u32 nr_maps, maplen;
4158 	u32 epoch;
4159 	struct ceph_fsid fsid;
4160 	struct rb_root need_resend = RB_ROOT;
4161 	LIST_HEAD(need_resend_linger);
4162 	bool handled_incremental = false;
4163 	bool was_pauserd, was_pausewr;
4164 	bool pauserd, pausewr;
4165 	int err;
4166 
4167 	dout("%s have %u\n", __func__, osdc->osdmap->epoch);
4168 	down_write(&osdc->lock);
4169 
4170 	/* verify fsid */
4171 	ceph_decode_need(&p, end, sizeof(fsid), bad);
4172 	ceph_decode_copy(&p, &fsid, sizeof(fsid));
4173 	if (ceph_check_fsid(osdc->client, &fsid) < 0)
4174 		goto bad;
4175 
4176 	was_pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
4177 	was_pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
4178 		      ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
4179 		      have_pool_full(osdc);
4180 
4181 	/* incremental maps */
4182 	ceph_decode_32_safe(&p, end, nr_maps, bad);
4183 	dout(" %d inc maps\n", nr_maps);
4184 	while (nr_maps > 0) {
4185 		ceph_decode_need(&p, end, 2*sizeof(u32), bad);
4186 		epoch = ceph_decode_32(&p);
4187 		maplen = ceph_decode_32(&p);
4188 		ceph_decode_need(&p, end, maplen, bad);
4189 		if (osdc->osdmap->epoch &&
4190 		    osdc->osdmap->epoch + 1 == epoch) {
4191 			dout("applying incremental map %u len %d\n",
4192 			     epoch, maplen);
4193 			err = handle_one_map(osdc, p, p + maplen, true,
4194 					     &need_resend, &need_resend_linger);
4195 			if (err)
4196 				goto bad;
4197 			handled_incremental = true;
4198 		} else {
4199 			dout("ignoring incremental map %u len %d\n",
4200 			     epoch, maplen);
4201 		}
4202 		p += maplen;
4203 		nr_maps--;
4204 	}
4205 	if (handled_incremental)
4206 		goto done;
4207 
4208 	/* full maps */
4209 	ceph_decode_32_safe(&p, end, nr_maps, bad);
4210 	dout(" %d full maps\n", nr_maps);
4211 	while (nr_maps) {
4212 		ceph_decode_need(&p, end, 2*sizeof(u32), bad);
4213 		epoch = ceph_decode_32(&p);
4214 		maplen = ceph_decode_32(&p);
4215 		ceph_decode_need(&p, end, maplen, bad);
4216 		if (nr_maps > 1) {
4217 			dout("skipping non-latest full map %u len %d\n",
4218 			     epoch, maplen);
4219 		} else if (osdc->osdmap->epoch >= epoch) {
4220 			dout("skipping full map %u len %d, "
4221 			     "older than our %u\n", epoch, maplen,
4222 			     osdc->osdmap->epoch);
4223 		} else {
4224 			dout("taking full map %u len %d\n", epoch, maplen);
4225 			err = handle_one_map(osdc, p, p + maplen, false,
4226 					     &need_resend, &need_resend_linger);
4227 			if (err)
4228 				goto bad;
4229 		}
4230 		p += maplen;
4231 		nr_maps--;
4232 	}
4233 
4234 done:
4235 	/*
4236 	 * subscribe to subsequent osdmap updates if full to ensure
4237 	 * we find out when we are no longer full and stop returning
4238 	 * ENOSPC.
4239 	 */
4240 	pauserd = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSERD);
4241 	pausewr = ceph_osdmap_flag(osdc, CEPH_OSDMAP_PAUSEWR) ||
4242 		  ceph_osdmap_flag(osdc, CEPH_OSDMAP_FULL) ||
4243 		  have_pool_full(osdc);
4244 	if (was_pauserd || was_pausewr || pauserd || pausewr ||
4245 	    osdc->osdmap->epoch < osdc->epoch_barrier)
4246 		maybe_request_map(osdc);
4247 
4248 	kick_requests(osdc, &need_resend, &need_resend_linger);
4249 
4250 	ceph_osdc_abort_on_full(osdc);
4251 	ceph_monc_got_map(&osdc->client->monc, CEPH_SUB_OSDMAP,
4252 			  osdc->osdmap->epoch);
4253 	up_write(&osdc->lock);
4254 	wake_up_all(&osdc->client->auth_wq);
4255 	return;
4256 
4257 bad:
4258 	pr_err("osdc handle_map corrupt msg\n");
4259 	ceph_msg_dump(msg);
4260 	up_write(&osdc->lock);
4261 }
4262 
4263 /*
4264  * Resubmit requests pending on the given osd.
4265  */
kick_osd_requests(struct ceph_osd * osd)4266 static void kick_osd_requests(struct ceph_osd *osd)
4267 {
4268 	struct rb_node *n;
4269 
4270 	clear_backoffs(osd);
4271 
4272 	for (n = rb_first(&osd->o_requests); n; ) {
4273 		struct ceph_osd_request *req =
4274 		    rb_entry(n, struct ceph_osd_request, r_node);
4275 
4276 		n = rb_next(n); /* cancel_linger_request() */
4277 
4278 		if (!req->r_linger) {
4279 			if (!req->r_t.paused)
4280 				send_request(req);
4281 		} else {
4282 			cancel_linger_request(req);
4283 		}
4284 	}
4285 	for (n = rb_first(&osd->o_linger_requests); n; n = rb_next(n)) {
4286 		struct ceph_osd_linger_request *lreq =
4287 		    rb_entry(n, struct ceph_osd_linger_request, node);
4288 
4289 		send_linger(lreq);
4290 	}
4291 }
4292 
4293 /*
4294  * If the osd connection drops, we need to resubmit all requests.
4295  */
osd_fault(struct ceph_connection * con)4296 static void osd_fault(struct ceph_connection *con)
4297 {
4298 	struct ceph_osd *osd = con->private;
4299 	struct ceph_osd_client *osdc = osd->o_osdc;
4300 
4301 	dout("%s osd %p osd%d\n", __func__, osd, osd->o_osd);
4302 
4303 	down_write(&osdc->lock);
4304 	if (!osd_registered(osd)) {
4305 		dout("%s osd%d unknown\n", __func__, osd->o_osd);
4306 		goto out_unlock;
4307 	}
4308 
4309 	if (!reopen_osd(osd))
4310 		kick_osd_requests(osd);
4311 	maybe_request_map(osdc);
4312 
4313 out_unlock:
4314 	up_write(&osdc->lock);
4315 }
4316 
4317 struct MOSDBackoff {
4318 	struct ceph_spg spgid;
4319 	u32 map_epoch;
4320 	u8 op;
4321 	u64 id;
4322 	struct ceph_hobject_id *begin;
4323 	struct ceph_hobject_id *end;
4324 };
4325 
decode_MOSDBackoff(const struct ceph_msg * msg,struct MOSDBackoff * m)4326 static int decode_MOSDBackoff(const struct ceph_msg *msg, struct MOSDBackoff *m)
4327 {
4328 	void *p = msg->front.iov_base;
4329 	void *const end = p + msg->front.iov_len;
4330 	u8 struct_v;
4331 	u32 struct_len;
4332 	int ret;
4333 
4334 	ret = ceph_start_decoding(&p, end, 1, "spg_t", &struct_v, &struct_len);
4335 	if (ret)
4336 		return ret;
4337 
4338 	ret = ceph_decode_pgid(&p, end, &m->spgid.pgid);
4339 	if (ret)
4340 		return ret;
4341 
4342 	ceph_decode_8_safe(&p, end, m->spgid.shard, e_inval);
4343 	ceph_decode_32_safe(&p, end, m->map_epoch, e_inval);
4344 	ceph_decode_8_safe(&p, end, m->op, e_inval);
4345 	ceph_decode_64_safe(&p, end, m->id, e_inval);
4346 
4347 	m->begin = kzalloc(sizeof(*m->begin), GFP_NOIO);
4348 	if (!m->begin)
4349 		return -ENOMEM;
4350 
4351 	ret = decode_hoid(&p, end, m->begin);
4352 	if (ret) {
4353 		free_hoid(m->begin);
4354 		return ret;
4355 	}
4356 
4357 	m->end = kzalloc(sizeof(*m->end), GFP_NOIO);
4358 	if (!m->end) {
4359 		free_hoid(m->begin);
4360 		return -ENOMEM;
4361 	}
4362 
4363 	ret = decode_hoid(&p, end, m->end);
4364 	if (ret) {
4365 		free_hoid(m->begin);
4366 		free_hoid(m->end);
4367 		return ret;
4368 	}
4369 
4370 	return 0;
4371 
4372 e_inval:
4373 	return -EINVAL;
4374 }
4375 
create_backoff_message(const struct ceph_osd_backoff * backoff,u32 map_epoch)4376 static struct ceph_msg *create_backoff_message(
4377 				const struct ceph_osd_backoff *backoff,
4378 				u32 map_epoch)
4379 {
4380 	struct ceph_msg *msg;
4381 	void *p, *end;
4382 	int msg_size;
4383 
4384 	msg_size = CEPH_ENCODING_START_BLK_LEN +
4385 			CEPH_PGID_ENCODING_LEN + 1; /* spgid */
4386 	msg_size += 4 + 1 + 8; /* map_epoch, op, id */
4387 	msg_size += CEPH_ENCODING_START_BLK_LEN +
4388 			hoid_encoding_size(backoff->begin);
4389 	msg_size += CEPH_ENCODING_START_BLK_LEN +
4390 			hoid_encoding_size(backoff->end);
4391 
4392 	msg = ceph_msg_new(CEPH_MSG_OSD_BACKOFF, msg_size, GFP_NOIO, true);
4393 	if (!msg)
4394 		return NULL;
4395 
4396 	p = msg->front.iov_base;
4397 	end = p + msg->front_alloc_len;
4398 
4399 	encode_spgid(&p, &backoff->spgid);
4400 	ceph_encode_32(&p, map_epoch);
4401 	ceph_encode_8(&p, CEPH_OSD_BACKOFF_OP_ACK_BLOCK);
4402 	ceph_encode_64(&p, backoff->id);
4403 	encode_hoid(&p, end, backoff->begin);
4404 	encode_hoid(&p, end, backoff->end);
4405 	BUG_ON(p != end);
4406 
4407 	msg->front.iov_len = p - msg->front.iov_base;
4408 	msg->hdr.version = cpu_to_le16(1); /* MOSDBackoff v1 */
4409 	msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
4410 
4411 	return msg;
4412 }
4413 
handle_backoff_block(struct ceph_osd * osd,struct MOSDBackoff * m)4414 static void handle_backoff_block(struct ceph_osd *osd, struct MOSDBackoff *m)
4415 {
4416 	struct ceph_spg_mapping *spg;
4417 	struct ceph_osd_backoff *backoff;
4418 	struct ceph_msg *msg;
4419 
4420 	dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4421 	     m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4422 
4423 	spg = lookup_spg_mapping(&osd->o_backoff_mappings, &m->spgid);
4424 	if (!spg) {
4425 		spg = alloc_spg_mapping();
4426 		if (!spg) {
4427 			pr_err("%s failed to allocate spg\n", __func__);
4428 			return;
4429 		}
4430 		spg->spgid = m->spgid; /* struct */
4431 		insert_spg_mapping(&osd->o_backoff_mappings, spg);
4432 	}
4433 
4434 	backoff = alloc_backoff();
4435 	if (!backoff) {
4436 		pr_err("%s failed to allocate backoff\n", __func__);
4437 		return;
4438 	}
4439 	backoff->spgid = m->spgid; /* struct */
4440 	backoff->id = m->id;
4441 	backoff->begin = m->begin;
4442 	m->begin = NULL; /* backoff now owns this */
4443 	backoff->end = m->end;
4444 	m->end = NULL;   /* ditto */
4445 
4446 	insert_backoff(&spg->backoffs, backoff);
4447 	insert_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4448 
4449 	/*
4450 	 * Ack with original backoff's epoch so that the OSD can
4451 	 * discard this if there was a PG split.
4452 	 */
4453 	msg = create_backoff_message(backoff, m->map_epoch);
4454 	if (!msg) {
4455 		pr_err("%s failed to allocate msg\n", __func__);
4456 		return;
4457 	}
4458 	ceph_con_send(&osd->o_con, msg);
4459 }
4460 
target_contained_by(const struct ceph_osd_request_target * t,const struct ceph_hobject_id * begin,const struct ceph_hobject_id * end)4461 static bool target_contained_by(const struct ceph_osd_request_target *t,
4462 				const struct ceph_hobject_id *begin,
4463 				const struct ceph_hobject_id *end)
4464 {
4465 	struct ceph_hobject_id hoid;
4466 	int cmp;
4467 
4468 	hoid_fill_from_target(&hoid, t);
4469 	cmp = hoid_compare(&hoid, begin);
4470 	return !cmp || (cmp > 0 && hoid_compare(&hoid, end) < 0);
4471 }
4472 
handle_backoff_unblock(struct ceph_osd * osd,const struct MOSDBackoff * m)4473 static void handle_backoff_unblock(struct ceph_osd *osd,
4474 				   const struct MOSDBackoff *m)
4475 {
4476 	struct ceph_spg_mapping *spg;
4477 	struct ceph_osd_backoff *backoff;
4478 	struct rb_node *n;
4479 
4480 	dout("%s osd%d spgid %llu.%xs%d id %llu\n", __func__, osd->o_osd,
4481 	     m->spgid.pgid.pool, m->spgid.pgid.seed, m->spgid.shard, m->id);
4482 
4483 	backoff = lookup_backoff_by_id(&osd->o_backoffs_by_id, m->id);
4484 	if (!backoff) {
4485 		pr_err("%s osd%d spgid %llu.%xs%d id %llu backoff dne\n",
4486 		       __func__, osd->o_osd, m->spgid.pgid.pool,
4487 		       m->spgid.pgid.seed, m->spgid.shard, m->id);
4488 		return;
4489 	}
4490 
4491 	if (hoid_compare(backoff->begin, m->begin) &&
4492 	    hoid_compare(backoff->end, m->end)) {
4493 		pr_err("%s osd%d spgid %llu.%xs%d id %llu bad range?\n",
4494 		       __func__, osd->o_osd, m->spgid.pgid.pool,
4495 		       m->spgid.pgid.seed, m->spgid.shard, m->id);
4496 		/* unblock it anyway... */
4497 	}
4498 
4499 	spg = lookup_spg_mapping(&osd->o_backoff_mappings, &backoff->spgid);
4500 	BUG_ON(!spg);
4501 
4502 	erase_backoff(&spg->backoffs, backoff);
4503 	erase_backoff_by_id(&osd->o_backoffs_by_id, backoff);
4504 	free_backoff(backoff);
4505 
4506 	if (RB_EMPTY_ROOT(&spg->backoffs)) {
4507 		erase_spg_mapping(&osd->o_backoff_mappings, spg);
4508 		free_spg_mapping(spg);
4509 	}
4510 
4511 	for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
4512 		struct ceph_osd_request *req =
4513 		    rb_entry(n, struct ceph_osd_request, r_node);
4514 
4515 		if (!ceph_spg_compare(&req->r_t.spgid, &m->spgid)) {
4516 			/*
4517 			 * Match against @m, not @backoff -- the PG may
4518 			 * have split on the OSD.
4519 			 */
4520 			if (target_contained_by(&req->r_t, m->begin, m->end)) {
4521 				/*
4522 				 * If no other installed backoff applies,
4523 				 * resend.
4524 				 */
4525 				send_request(req);
4526 			}
4527 		}
4528 	}
4529 }
4530 
handle_backoff(struct ceph_osd * osd,struct ceph_msg * msg)4531 static void handle_backoff(struct ceph_osd *osd, struct ceph_msg *msg)
4532 {
4533 	struct ceph_osd_client *osdc = osd->o_osdc;
4534 	struct MOSDBackoff m;
4535 	int ret;
4536 
4537 	down_read(&osdc->lock);
4538 	if (!osd_registered(osd)) {
4539 		dout("%s osd%d unknown\n", __func__, osd->o_osd);
4540 		up_read(&osdc->lock);
4541 		return;
4542 	}
4543 	WARN_ON(osd->o_osd != le64_to_cpu(msg->hdr.src.num));
4544 
4545 	mutex_lock(&osd->lock);
4546 	ret = decode_MOSDBackoff(msg, &m);
4547 	if (ret) {
4548 		pr_err("failed to decode MOSDBackoff: %d\n", ret);
4549 		ceph_msg_dump(msg);
4550 		goto out_unlock;
4551 	}
4552 
4553 	switch (m.op) {
4554 	case CEPH_OSD_BACKOFF_OP_BLOCK:
4555 		handle_backoff_block(osd, &m);
4556 		break;
4557 	case CEPH_OSD_BACKOFF_OP_UNBLOCK:
4558 		handle_backoff_unblock(osd, &m);
4559 		break;
4560 	default:
4561 		pr_err("%s osd%d unknown op %d\n", __func__, osd->o_osd, m.op);
4562 	}
4563 
4564 	free_hoid(m.begin);
4565 	free_hoid(m.end);
4566 
4567 out_unlock:
4568 	mutex_unlock(&osd->lock);
4569 	up_read(&osdc->lock);
4570 }
4571 
4572 /*
4573  * Process osd watch notifications
4574  */
handle_watch_notify(struct ceph_osd_client * osdc,struct ceph_msg * msg)4575 static void handle_watch_notify(struct ceph_osd_client *osdc,
4576 				struct ceph_msg *msg)
4577 {
4578 	void *p = msg->front.iov_base;
4579 	void *const end = p + msg->front.iov_len;
4580 	struct ceph_osd_linger_request *lreq;
4581 	struct linger_work *lwork;
4582 	u8 proto_ver, opcode;
4583 	u64 cookie, notify_id;
4584 	u64 notifier_id = 0;
4585 	s32 return_code = 0;
4586 	void *payload = NULL;
4587 	u32 payload_len = 0;
4588 
4589 	ceph_decode_8_safe(&p, end, proto_ver, bad);
4590 	ceph_decode_8_safe(&p, end, opcode, bad);
4591 	ceph_decode_64_safe(&p, end, cookie, bad);
4592 	p += 8; /* skip ver */
4593 	ceph_decode_64_safe(&p, end, notify_id, bad);
4594 
4595 	if (proto_ver >= 1) {
4596 		ceph_decode_32_safe(&p, end, payload_len, bad);
4597 		ceph_decode_need(&p, end, payload_len, bad);
4598 		payload = p;
4599 		p += payload_len;
4600 	}
4601 
4602 	if (le16_to_cpu(msg->hdr.version) >= 2)
4603 		ceph_decode_32_safe(&p, end, return_code, bad);
4604 
4605 	if (le16_to_cpu(msg->hdr.version) >= 3)
4606 		ceph_decode_64_safe(&p, end, notifier_id, bad);
4607 
4608 	down_read(&osdc->lock);
4609 	lreq = lookup_linger_osdc(&osdc->linger_requests, cookie);
4610 	if (!lreq) {
4611 		dout("%s opcode %d cookie %llu dne\n", __func__, opcode,
4612 		     cookie);
4613 		goto out_unlock_osdc;
4614 	}
4615 
4616 	mutex_lock(&lreq->lock);
4617 	dout("%s opcode %d cookie %llu lreq %p is_watch %d\n", __func__,
4618 	     opcode, cookie, lreq, lreq->is_watch);
4619 	if (opcode == CEPH_WATCH_EVENT_DISCONNECT) {
4620 		if (!lreq->last_error) {
4621 			lreq->last_error = -ENOTCONN;
4622 			queue_watch_error(lreq);
4623 		}
4624 	} else if (!lreq->is_watch) {
4625 		/* CEPH_WATCH_EVENT_NOTIFY_COMPLETE */
4626 		if (lreq->notify_id && lreq->notify_id != notify_id) {
4627 			dout("lreq %p notify_id %llu != %llu, ignoring\n", lreq,
4628 			     lreq->notify_id, notify_id);
4629 		} else if (!completion_done(&lreq->notify_finish_wait)) {
4630 			struct ceph_msg_data *data =
4631 			    msg->num_data_items ? &msg->data[0] : NULL;
4632 
4633 			if (data) {
4634 				if (lreq->preply_pages) {
4635 					WARN_ON(data->type !=
4636 							CEPH_MSG_DATA_PAGES);
4637 					*lreq->preply_pages = data->pages;
4638 					*lreq->preply_len = data->length;
4639 					data->own_pages = false;
4640 				}
4641 			}
4642 			lreq->notify_finish_error = return_code;
4643 			complete_all(&lreq->notify_finish_wait);
4644 		}
4645 	} else {
4646 		/* CEPH_WATCH_EVENT_NOTIFY */
4647 		lwork = lwork_alloc(lreq, do_watch_notify);
4648 		if (!lwork) {
4649 			pr_err("failed to allocate notify-lwork\n");
4650 			goto out_unlock_lreq;
4651 		}
4652 
4653 		lwork->notify.notify_id = notify_id;
4654 		lwork->notify.notifier_id = notifier_id;
4655 		lwork->notify.payload = payload;
4656 		lwork->notify.payload_len = payload_len;
4657 		lwork->notify.msg = ceph_msg_get(msg);
4658 		lwork_queue(lwork);
4659 	}
4660 
4661 out_unlock_lreq:
4662 	mutex_unlock(&lreq->lock);
4663 out_unlock_osdc:
4664 	up_read(&osdc->lock);
4665 	return;
4666 
4667 bad:
4668 	pr_err("osdc handle_watch_notify corrupt msg\n");
4669 }
4670 
4671 /*
4672  * Register request, send initial attempt.
4673  */
ceph_osdc_start_request(struct ceph_osd_client * osdc,struct ceph_osd_request * req)4674 void ceph_osdc_start_request(struct ceph_osd_client *osdc,
4675 			     struct ceph_osd_request *req)
4676 {
4677 	down_read(&osdc->lock);
4678 	submit_request(req, false);
4679 	up_read(&osdc->lock);
4680 }
4681 EXPORT_SYMBOL(ceph_osdc_start_request);
4682 
4683 /*
4684  * Unregister request.  If @req was registered, it isn't completed:
4685  * r_result isn't set and __complete_request() isn't invoked.
4686  *
4687  * If @req wasn't registered, this call may have raced with
4688  * handle_reply(), in which case r_result would already be set and
4689  * __complete_request() would be getting invoked, possibly even
4690  * concurrently with this call.
4691  */
ceph_osdc_cancel_request(struct ceph_osd_request * req)4692 void ceph_osdc_cancel_request(struct ceph_osd_request *req)
4693 {
4694 	struct ceph_osd_client *osdc = req->r_osdc;
4695 
4696 	down_write(&osdc->lock);
4697 	if (req->r_osd)
4698 		cancel_request(req);
4699 	up_write(&osdc->lock);
4700 }
4701 EXPORT_SYMBOL(ceph_osdc_cancel_request);
4702 
4703 /*
4704  * @timeout: in jiffies, 0 means "wait forever"
4705  */
wait_request_timeout(struct ceph_osd_request * req,unsigned long timeout)4706 static int wait_request_timeout(struct ceph_osd_request *req,
4707 				unsigned long timeout)
4708 {
4709 	long left;
4710 
4711 	dout("%s req %p tid %llu\n", __func__, req, req->r_tid);
4712 	left = wait_for_completion_killable_timeout(&req->r_completion,
4713 						ceph_timeout_jiffies(timeout));
4714 	if (left <= 0) {
4715 		left = left ?: -ETIMEDOUT;
4716 		ceph_osdc_cancel_request(req);
4717 	} else {
4718 		left = req->r_result; /* completed */
4719 	}
4720 
4721 	return left;
4722 }
4723 
4724 /*
4725  * wait for a request to complete
4726  */
ceph_osdc_wait_request(struct ceph_osd_client * osdc,struct ceph_osd_request * req)4727 int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
4728 			   struct ceph_osd_request *req)
4729 {
4730 	return wait_request_timeout(req, 0);
4731 }
4732 EXPORT_SYMBOL(ceph_osdc_wait_request);
4733 
4734 /*
4735  * sync - wait for all in-flight requests to flush.  avoid starvation.
4736  */
ceph_osdc_sync(struct ceph_osd_client * osdc)4737 void ceph_osdc_sync(struct ceph_osd_client *osdc)
4738 {
4739 	struct rb_node *n, *p;
4740 	u64 last_tid = atomic64_read(&osdc->last_tid);
4741 
4742 again:
4743 	down_read(&osdc->lock);
4744 	for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
4745 		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
4746 
4747 		mutex_lock(&osd->lock);
4748 		for (p = rb_first(&osd->o_requests); p; p = rb_next(p)) {
4749 			struct ceph_osd_request *req =
4750 			    rb_entry(p, struct ceph_osd_request, r_node);
4751 
4752 			if (req->r_tid > last_tid)
4753 				break;
4754 
4755 			if (!(req->r_flags & CEPH_OSD_FLAG_WRITE))
4756 				continue;
4757 
4758 			ceph_osdc_get_request(req);
4759 			mutex_unlock(&osd->lock);
4760 			up_read(&osdc->lock);
4761 			dout("%s waiting on req %p tid %llu last_tid %llu\n",
4762 			     __func__, req, req->r_tid, last_tid);
4763 			wait_for_completion(&req->r_completion);
4764 			ceph_osdc_put_request(req);
4765 			goto again;
4766 		}
4767 
4768 		mutex_unlock(&osd->lock);
4769 	}
4770 
4771 	up_read(&osdc->lock);
4772 	dout("%s done last_tid %llu\n", __func__, last_tid);
4773 }
4774 EXPORT_SYMBOL(ceph_osdc_sync);
4775 
4776 /*
4777  * Returns a handle, caller owns a ref.
4778  */
4779 struct ceph_osd_linger_request *
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)4780 ceph_osdc_watch(struct ceph_osd_client *osdc,
4781 		struct ceph_object_id *oid,
4782 		struct ceph_object_locator *oloc,
4783 		rados_watchcb2_t wcb,
4784 		rados_watcherrcb_t errcb,
4785 		void *data)
4786 {
4787 	struct ceph_osd_linger_request *lreq;
4788 	int ret;
4789 
4790 	lreq = linger_alloc(osdc);
4791 	if (!lreq)
4792 		return ERR_PTR(-ENOMEM);
4793 
4794 	lreq->is_watch = true;
4795 	lreq->wcb = wcb;
4796 	lreq->errcb = errcb;
4797 	lreq->data = data;
4798 	lreq->watch_valid_thru = jiffies;
4799 
4800 	ceph_oid_copy(&lreq->t.base_oid, oid);
4801 	ceph_oloc_copy(&lreq->t.base_oloc, oloc);
4802 	lreq->t.flags = CEPH_OSD_FLAG_WRITE;
4803 	ktime_get_real_ts64(&lreq->mtime);
4804 
4805 	linger_submit(lreq);
4806 	ret = linger_reg_commit_wait(lreq);
4807 	if (ret) {
4808 		linger_cancel(lreq);
4809 		goto err_put_lreq;
4810 	}
4811 
4812 	return lreq;
4813 
4814 err_put_lreq:
4815 	linger_put(lreq);
4816 	return ERR_PTR(ret);
4817 }
4818 EXPORT_SYMBOL(ceph_osdc_watch);
4819 
4820 /*
4821  * Releases a ref.
4822  *
4823  * Times out after mount_timeout to preserve rbd unmap behaviour
4824  * introduced in 2894e1d76974 ("rbd: timeout watch teardown on unmap
4825  * with mount_timeout").
4826  */
ceph_osdc_unwatch(struct ceph_osd_client * osdc,struct ceph_osd_linger_request * lreq)4827 int ceph_osdc_unwatch(struct ceph_osd_client *osdc,
4828 		      struct ceph_osd_linger_request *lreq)
4829 {
4830 	struct ceph_options *opts = osdc->client->options;
4831 	struct ceph_osd_request *req;
4832 	int ret;
4833 
4834 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4835 	if (!req)
4836 		return -ENOMEM;
4837 
4838 	ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
4839 	ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
4840 	req->r_flags = CEPH_OSD_FLAG_WRITE;
4841 	ktime_get_real_ts64(&req->r_mtime);
4842 	osd_req_op_watch_init(req, 0, CEPH_OSD_WATCH_OP_UNWATCH,
4843 			      lreq->linger_id, 0);
4844 
4845 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4846 	if (ret)
4847 		goto out_put_req;
4848 
4849 	ceph_osdc_start_request(osdc, req);
4850 	linger_cancel(lreq);
4851 	linger_put(lreq);
4852 	ret = wait_request_timeout(req, opts->mount_timeout);
4853 
4854 out_put_req:
4855 	ceph_osdc_put_request(req);
4856 	return ret;
4857 }
4858 EXPORT_SYMBOL(ceph_osdc_unwatch);
4859 
osd_req_op_notify_ack_init(struct ceph_osd_request * req,int which,u64 notify_id,u64 cookie,void * payload,u32 payload_len)4860 static int osd_req_op_notify_ack_init(struct ceph_osd_request *req, int which,
4861 				      u64 notify_id, u64 cookie, void *payload,
4862 				      u32 payload_len)
4863 {
4864 	struct ceph_osd_req_op *op;
4865 	struct ceph_pagelist *pl;
4866 	int ret;
4867 
4868 	op = osd_req_op_init(req, which, CEPH_OSD_OP_NOTIFY_ACK, 0);
4869 
4870 	pl = ceph_pagelist_alloc(GFP_NOIO);
4871 	if (!pl)
4872 		return -ENOMEM;
4873 
4874 	ret = ceph_pagelist_encode_64(pl, notify_id);
4875 	ret |= ceph_pagelist_encode_64(pl, cookie);
4876 	if (payload) {
4877 		ret |= ceph_pagelist_encode_32(pl, payload_len);
4878 		ret |= ceph_pagelist_append(pl, payload, payload_len);
4879 	} else {
4880 		ret |= ceph_pagelist_encode_32(pl, 0);
4881 	}
4882 	if (ret) {
4883 		ceph_pagelist_release(pl);
4884 		return -ENOMEM;
4885 	}
4886 
4887 	ceph_osd_data_pagelist_init(&op->notify_ack.request_data, pl);
4888 	op->indata_len = pl->length;
4889 	return 0;
4890 }
4891 
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)4892 int ceph_osdc_notify_ack(struct ceph_osd_client *osdc,
4893 			 struct ceph_object_id *oid,
4894 			 struct ceph_object_locator *oloc,
4895 			 u64 notify_id,
4896 			 u64 cookie,
4897 			 void *payload,
4898 			 u32 payload_len)
4899 {
4900 	struct ceph_osd_request *req;
4901 	int ret;
4902 
4903 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
4904 	if (!req)
4905 		return -ENOMEM;
4906 
4907 	ceph_oid_copy(&req->r_base_oid, oid);
4908 	ceph_oloc_copy(&req->r_base_oloc, oloc);
4909 	req->r_flags = CEPH_OSD_FLAG_READ;
4910 
4911 	ret = osd_req_op_notify_ack_init(req, 0, notify_id, cookie, payload,
4912 					 payload_len);
4913 	if (ret)
4914 		goto out_put_req;
4915 
4916 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
4917 	if (ret)
4918 		goto out_put_req;
4919 
4920 	ceph_osdc_start_request(osdc, req);
4921 	ret = ceph_osdc_wait_request(osdc, req);
4922 
4923 out_put_req:
4924 	ceph_osdc_put_request(req);
4925 	return ret;
4926 }
4927 EXPORT_SYMBOL(ceph_osdc_notify_ack);
4928 
4929 /*
4930  * @timeout: in seconds
4931  *
4932  * @preply_{pages,len} are initialized both on success and error.
4933  * The caller is responsible for:
4934  *
4935  *     ceph_release_page_vector(reply_pages, calc_pages_for(0, reply_len))
4936  */
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)4937 int ceph_osdc_notify(struct ceph_osd_client *osdc,
4938 		     struct ceph_object_id *oid,
4939 		     struct ceph_object_locator *oloc,
4940 		     void *payload,
4941 		     u32 payload_len,
4942 		     u32 timeout,
4943 		     struct page ***preply_pages,
4944 		     size_t *preply_len)
4945 {
4946 	struct ceph_osd_linger_request *lreq;
4947 	int ret;
4948 
4949 	WARN_ON(!timeout);
4950 	if (preply_pages) {
4951 		*preply_pages = NULL;
4952 		*preply_len = 0;
4953 	}
4954 
4955 	lreq = linger_alloc(osdc);
4956 	if (!lreq)
4957 		return -ENOMEM;
4958 
4959 	lreq->request_pl = ceph_pagelist_alloc(GFP_NOIO);
4960 	if (!lreq->request_pl) {
4961 		ret = -ENOMEM;
4962 		goto out_put_lreq;
4963 	}
4964 
4965 	ret = ceph_pagelist_encode_32(lreq->request_pl, 1); /* prot_ver */
4966 	ret |= ceph_pagelist_encode_32(lreq->request_pl, timeout);
4967 	ret |= ceph_pagelist_encode_32(lreq->request_pl, payload_len);
4968 	ret |= ceph_pagelist_append(lreq->request_pl, payload, payload_len);
4969 	if (ret) {
4970 		ret = -ENOMEM;
4971 		goto out_put_lreq;
4972 	}
4973 
4974 	/* for notify_id */
4975 	lreq->notify_id_pages = ceph_alloc_page_vector(1, GFP_NOIO);
4976 	if (IS_ERR(lreq->notify_id_pages)) {
4977 		ret = PTR_ERR(lreq->notify_id_pages);
4978 		lreq->notify_id_pages = NULL;
4979 		goto out_put_lreq;
4980 	}
4981 
4982 	lreq->preply_pages = preply_pages;
4983 	lreq->preply_len = preply_len;
4984 
4985 	ceph_oid_copy(&lreq->t.base_oid, oid);
4986 	ceph_oloc_copy(&lreq->t.base_oloc, oloc);
4987 	lreq->t.flags = CEPH_OSD_FLAG_READ;
4988 
4989 	linger_submit(lreq);
4990 	ret = linger_reg_commit_wait(lreq);
4991 	if (!ret)
4992 		ret = linger_notify_finish_wait(lreq,
4993 				 msecs_to_jiffies(2 * timeout * MSEC_PER_SEC));
4994 	else
4995 		dout("lreq %p failed to initiate notify %d\n", lreq, ret);
4996 
4997 	linger_cancel(lreq);
4998 out_put_lreq:
4999 	linger_put(lreq);
5000 	return ret;
5001 }
5002 EXPORT_SYMBOL(ceph_osdc_notify);
5003 
decode_watcher(void ** p,void * end,struct ceph_watch_item * item)5004 static int decode_watcher(void **p, void *end, struct ceph_watch_item *item)
5005 {
5006 	u8 struct_v;
5007 	u32 struct_len;
5008 	int ret;
5009 
5010 	ret = ceph_start_decoding(p, end, 2, "watch_item_t",
5011 				  &struct_v, &struct_len);
5012 	if (ret)
5013 		goto bad;
5014 
5015 	ret = -EINVAL;
5016 	ceph_decode_copy_safe(p, end, &item->name, sizeof(item->name), bad);
5017 	ceph_decode_64_safe(p, end, item->cookie, bad);
5018 	ceph_decode_skip_32(p, end, bad); /* skip timeout seconds */
5019 
5020 	if (struct_v >= 2) {
5021 		ret = ceph_decode_entity_addr(p, end, &item->addr);
5022 		if (ret)
5023 			goto bad;
5024 	} else {
5025 		ret = 0;
5026 	}
5027 
5028 	dout("%s %s%llu cookie %llu addr %s\n", __func__,
5029 	     ENTITY_NAME(item->name), item->cookie,
5030 	     ceph_pr_addr(&item->addr));
5031 bad:
5032 	return ret;
5033 }
5034 
decode_watchers(void ** p,void * end,struct ceph_watch_item ** watchers,u32 * num_watchers)5035 static int decode_watchers(void **p, void *end,
5036 			   struct ceph_watch_item **watchers,
5037 			   u32 *num_watchers)
5038 {
5039 	u8 struct_v;
5040 	u32 struct_len;
5041 	int i;
5042 	int ret;
5043 
5044 	ret = ceph_start_decoding(p, end, 1, "obj_list_watch_response_t",
5045 				  &struct_v, &struct_len);
5046 	if (ret)
5047 		return ret;
5048 
5049 	*num_watchers = ceph_decode_32(p);
5050 	*watchers = kcalloc(*num_watchers, sizeof(**watchers), GFP_NOIO);
5051 	if (!*watchers)
5052 		return -ENOMEM;
5053 
5054 	for (i = 0; i < *num_watchers; i++) {
5055 		ret = decode_watcher(p, end, *watchers + i);
5056 		if (ret) {
5057 			kfree(*watchers);
5058 			return ret;
5059 		}
5060 	}
5061 
5062 	return 0;
5063 }
5064 
5065 /*
5066  * On success, the caller is responsible for:
5067  *
5068  *     kfree(watchers);
5069  */
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)5070 int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
5071 			    struct ceph_object_id *oid,
5072 			    struct ceph_object_locator *oloc,
5073 			    struct ceph_watch_item **watchers,
5074 			    u32 *num_watchers)
5075 {
5076 	struct ceph_osd_request *req;
5077 	struct page **pages;
5078 	int ret;
5079 
5080 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
5081 	if (!req)
5082 		return -ENOMEM;
5083 
5084 	ceph_oid_copy(&req->r_base_oid, oid);
5085 	ceph_oloc_copy(&req->r_base_oloc, oloc);
5086 	req->r_flags = CEPH_OSD_FLAG_READ;
5087 
5088 	pages = ceph_alloc_page_vector(1, GFP_NOIO);
5089 	if (IS_ERR(pages)) {
5090 		ret = PTR_ERR(pages);
5091 		goto out_put_req;
5092 	}
5093 
5094 	osd_req_op_init(req, 0, CEPH_OSD_OP_LIST_WATCHERS, 0);
5095 	ceph_osd_data_pages_init(osd_req_op_data(req, 0, list_watchers,
5096 						 response_data),
5097 				 pages, PAGE_SIZE, 0, false, true);
5098 
5099 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
5100 	if (ret)
5101 		goto out_put_req;
5102 
5103 	ceph_osdc_start_request(osdc, req);
5104 	ret = ceph_osdc_wait_request(osdc, req);
5105 	if (ret >= 0) {
5106 		void *p = page_address(pages[0]);
5107 		void *const end = p + req->r_ops[0].outdata_len;
5108 
5109 		ret = decode_watchers(&p, end, watchers, num_watchers);
5110 	}
5111 
5112 out_put_req:
5113 	ceph_osdc_put_request(req);
5114 	return ret;
5115 }
5116 EXPORT_SYMBOL(ceph_osdc_list_watchers);
5117 
5118 /*
5119  * Call all pending notify callbacks - for use after a watch is
5120  * unregistered, to make sure no more callbacks for it will be invoked
5121  */
ceph_osdc_flush_notifies(struct ceph_osd_client * osdc)5122 void ceph_osdc_flush_notifies(struct ceph_osd_client *osdc)
5123 {
5124 	dout("%s osdc %p\n", __func__, osdc);
5125 	flush_workqueue(osdc->notify_wq);
5126 }
5127 EXPORT_SYMBOL(ceph_osdc_flush_notifies);
5128 
ceph_osdc_maybe_request_map(struct ceph_osd_client * osdc)5129 void ceph_osdc_maybe_request_map(struct ceph_osd_client *osdc)
5130 {
5131 	down_read(&osdc->lock);
5132 	maybe_request_map(osdc);
5133 	up_read(&osdc->lock);
5134 }
5135 EXPORT_SYMBOL(ceph_osdc_maybe_request_map);
5136 
5137 /*
5138  * Execute an OSD class method on an object.
5139  *
5140  * @flags: CEPH_OSD_FLAG_*
5141  * @resp_len: in/out param for reply length
5142  */
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)5143 int ceph_osdc_call(struct ceph_osd_client *osdc,
5144 		   struct ceph_object_id *oid,
5145 		   struct ceph_object_locator *oloc,
5146 		   const char *class, const char *method,
5147 		   unsigned int flags,
5148 		   struct page *req_page, size_t req_len,
5149 		   struct page **resp_pages, size_t *resp_len)
5150 {
5151 	struct ceph_osd_request *req;
5152 	int ret;
5153 
5154 	if (req_len > PAGE_SIZE)
5155 		return -E2BIG;
5156 
5157 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
5158 	if (!req)
5159 		return -ENOMEM;
5160 
5161 	ceph_oid_copy(&req->r_base_oid, oid);
5162 	ceph_oloc_copy(&req->r_base_oloc, oloc);
5163 	req->r_flags = flags;
5164 
5165 	ret = osd_req_op_cls_init(req, 0, class, method);
5166 	if (ret)
5167 		goto out_put_req;
5168 
5169 	if (req_page)
5170 		osd_req_op_cls_request_data_pages(req, 0, &req_page, req_len,
5171 						  0, false, false);
5172 	if (resp_pages)
5173 		osd_req_op_cls_response_data_pages(req, 0, resp_pages,
5174 						   *resp_len, 0, false, false);
5175 
5176 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
5177 	if (ret)
5178 		goto out_put_req;
5179 
5180 	ceph_osdc_start_request(osdc, req);
5181 	ret = ceph_osdc_wait_request(osdc, req);
5182 	if (ret >= 0) {
5183 		ret = req->r_ops[0].rval;
5184 		if (resp_pages)
5185 			*resp_len = req->r_ops[0].outdata_len;
5186 	}
5187 
5188 out_put_req:
5189 	ceph_osdc_put_request(req);
5190 	return ret;
5191 }
5192 EXPORT_SYMBOL(ceph_osdc_call);
5193 
5194 /*
5195  * reset all osd connections
5196  */
ceph_osdc_reopen_osds(struct ceph_osd_client * osdc)5197 void ceph_osdc_reopen_osds(struct ceph_osd_client *osdc)
5198 {
5199 	struct rb_node *n;
5200 
5201 	down_write(&osdc->lock);
5202 	for (n = rb_first(&osdc->osds); n; ) {
5203 		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
5204 
5205 		n = rb_next(n);
5206 		if (!reopen_osd(osd))
5207 			kick_osd_requests(osd);
5208 	}
5209 	up_write(&osdc->lock);
5210 }
5211 
5212 /*
5213  * init, shutdown
5214  */
ceph_osdc_init(struct ceph_osd_client * osdc,struct ceph_client * client)5215 int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client)
5216 {
5217 	int err;
5218 
5219 	dout("init\n");
5220 	osdc->client = client;
5221 	init_rwsem(&osdc->lock);
5222 	osdc->osds = RB_ROOT;
5223 	INIT_LIST_HEAD(&osdc->osd_lru);
5224 	spin_lock_init(&osdc->osd_lru_lock);
5225 	osd_init(&osdc->homeless_osd);
5226 	osdc->homeless_osd.o_osdc = osdc;
5227 	osdc->homeless_osd.o_osd = CEPH_HOMELESS_OSD;
5228 	osdc->last_linger_id = CEPH_LINGER_ID_START;
5229 	osdc->linger_requests = RB_ROOT;
5230 	osdc->map_checks = RB_ROOT;
5231 	osdc->linger_map_checks = RB_ROOT;
5232 	INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout);
5233 	INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout);
5234 
5235 	err = -ENOMEM;
5236 	osdc->osdmap = ceph_osdmap_alloc();
5237 	if (!osdc->osdmap)
5238 		goto out;
5239 
5240 	osdc->req_mempool = mempool_create_slab_pool(10,
5241 						     ceph_osd_request_cache);
5242 	if (!osdc->req_mempool)
5243 		goto out_map;
5244 
5245 	err = ceph_msgpool_init(&osdc->msgpool_op, CEPH_MSG_OSD_OP,
5246 				PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10, "osd_op");
5247 	if (err < 0)
5248 		goto out_mempool;
5249 	err = ceph_msgpool_init(&osdc->msgpool_op_reply, CEPH_MSG_OSD_OPREPLY,
5250 				PAGE_SIZE, CEPH_OSD_SLAB_OPS, 10,
5251 				"osd_op_reply");
5252 	if (err < 0)
5253 		goto out_msgpool;
5254 
5255 	err = -ENOMEM;
5256 	osdc->notify_wq = create_singlethread_workqueue("ceph-watch-notify");
5257 	if (!osdc->notify_wq)
5258 		goto out_msgpool_reply;
5259 
5260 	osdc->completion_wq = create_singlethread_workqueue("ceph-completion");
5261 	if (!osdc->completion_wq)
5262 		goto out_notify_wq;
5263 
5264 	schedule_delayed_work(&osdc->timeout_work,
5265 			      osdc->client->options->osd_keepalive_timeout);
5266 	schedule_delayed_work(&osdc->osds_timeout_work,
5267 	    round_jiffies_relative(osdc->client->options->osd_idle_ttl));
5268 
5269 	return 0;
5270 
5271 out_notify_wq:
5272 	destroy_workqueue(osdc->notify_wq);
5273 out_msgpool_reply:
5274 	ceph_msgpool_destroy(&osdc->msgpool_op_reply);
5275 out_msgpool:
5276 	ceph_msgpool_destroy(&osdc->msgpool_op);
5277 out_mempool:
5278 	mempool_destroy(osdc->req_mempool);
5279 out_map:
5280 	ceph_osdmap_destroy(osdc->osdmap);
5281 out:
5282 	return err;
5283 }
5284 
ceph_osdc_stop(struct ceph_osd_client * osdc)5285 void ceph_osdc_stop(struct ceph_osd_client *osdc)
5286 {
5287 	destroy_workqueue(osdc->completion_wq);
5288 	destroy_workqueue(osdc->notify_wq);
5289 	cancel_delayed_work_sync(&osdc->timeout_work);
5290 	cancel_delayed_work_sync(&osdc->osds_timeout_work);
5291 
5292 	down_write(&osdc->lock);
5293 	while (!RB_EMPTY_ROOT(&osdc->osds)) {
5294 		struct ceph_osd *osd = rb_entry(rb_first(&osdc->osds),
5295 						struct ceph_osd, o_node);
5296 		close_osd(osd);
5297 	}
5298 	up_write(&osdc->lock);
5299 	WARN_ON(refcount_read(&osdc->homeless_osd.o_ref) != 1);
5300 	osd_cleanup(&osdc->homeless_osd);
5301 
5302 	WARN_ON(!list_empty(&osdc->osd_lru));
5303 	WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_requests));
5304 	WARN_ON(!RB_EMPTY_ROOT(&osdc->map_checks));
5305 	WARN_ON(!RB_EMPTY_ROOT(&osdc->linger_map_checks));
5306 	WARN_ON(atomic_read(&osdc->num_requests));
5307 	WARN_ON(atomic_read(&osdc->num_homeless));
5308 
5309 	ceph_osdmap_destroy(osdc->osdmap);
5310 	mempool_destroy(osdc->req_mempool);
5311 	ceph_msgpool_destroy(&osdc->msgpool_op);
5312 	ceph_msgpool_destroy(&osdc->msgpool_op_reply);
5313 }
5314 
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)5315 int osd_req_op_copy_from_init(struct ceph_osd_request *req,
5316 			      u64 src_snapid, u64 src_version,
5317 			      struct ceph_object_id *src_oid,
5318 			      struct ceph_object_locator *src_oloc,
5319 			      u32 src_fadvise_flags,
5320 			      u32 dst_fadvise_flags,
5321 			      u32 truncate_seq, u64 truncate_size,
5322 			      u8 copy_from_flags)
5323 {
5324 	struct ceph_osd_req_op *op;
5325 	struct page **pages;
5326 	void *p, *end;
5327 
5328 	pages = ceph_alloc_page_vector(1, GFP_KERNEL);
5329 	if (IS_ERR(pages))
5330 		return PTR_ERR(pages);
5331 
5332 	op = osd_req_op_init(req, 0, CEPH_OSD_OP_COPY_FROM2,
5333 			     dst_fadvise_flags);
5334 	op->copy_from.snapid = src_snapid;
5335 	op->copy_from.src_version = src_version;
5336 	op->copy_from.flags = copy_from_flags;
5337 	op->copy_from.src_fadvise_flags = src_fadvise_flags;
5338 
5339 	p = page_address(pages[0]);
5340 	end = p + PAGE_SIZE;
5341 	ceph_encode_string(&p, end, src_oid->name, src_oid->name_len);
5342 	encode_oloc(&p, end, src_oloc);
5343 	ceph_encode_32(&p, truncate_seq);
5344 	ceph_encode_64(&p, truncate_size);
5345 	op->indata_len = PAGE_SIZE - (end - p);
5346 
5347 	ceph_osd_data_pages_init(&op->copy_from.osd_data, pages,
5348 				 op->indata_len, 0, false, true);
5349 	return 0;
5350 }
5351 EXPORT_SYMBOL(osd_req_op_copy_from_init);
5352 
ceph_osdc_setup(void)5353 int __init ceph_osdc_setup(void)
5354 {
5355 	size_t size = sizeof(struct ceph_osd_request) +
5356 	    CEPH_OSD_SLAB_OPS * sizeof(struct ceph_osd_req_op);
5357 
5358 	BUG_ON(ceph_osd_request_cache);
5359 	ceph_osd_request_cache = kmem_cache_create("ceph_osd_request", size,
5360 						   0, 0, NULL);
5361 
5362 	return ceph_osd_request_cache ? 0 : -ENOMEM;
5363 }
5364 
ceph_osdc_cleanup(void)5365 void ceph_osdc_cleanup(void)
5366 {
5367 	BUG_ON(!ceph_osd_request_cache);
5368 	kmem_cache_destroy(ceph_osd_request_cache);
5369 	ceph_osd_request_cache = NULL;
5370 }
5371 
5372 /*
5373  * handle incoming message
5374  */
osd_dispatch(struct ceph_connection * con,struct ceph_msg * msg)5375 static void osd_dispatch(struct ceph_connection *con, struct ceph_msg *msg)
5376 {
5377 	struct ceph_osd *osd = con->private;
5378 	struct ceph_osd_client *osdc = osd->o_osdc;
5379 	int type = le16_to_cpu(msg->hdr.type);
5380 
5381 	switch (type) {
5382 	case CEPH_MSG_OSD_MAP:
5383 		ceph_osdc_handle_map(osdc, msg);
5384 		break;
5385 	case CEPH_MSG_OSD_OPREPLY:
5386 		handle_reply(osd, msg);
5387 		break;
5388 	case CEPH_MSG_OSD_BACKOFF:
5389 		handle_backoff(osd, msg);
5390 		break;
5391 	case CEPH_MSG_WATCH_NOTIFY:
5392 		handle_watch_notify(osdc, msg);
5393 		break;
5394 
5395 	default:
5396 		pr_err("received unknown message type %d %s\n", type,
5397 		       ceph_msg_type_name(type));
5398 	}
5399 
5400 	ceph_msg_put(msg);
5401 }
5402 
5403 /* How much sparse data was requested? */
sparse_data_requested(struct ceph_osd_request * req)5404 static u64 sparse_data_requested(struct ceph_osd_request *req)
5405 {
5406 	u64 len = 0;
5407 
5408 	if (req->r_flags & CEPH_OSD_FLAG_READ) {
5409 		int i;
5410 
5411 		for (i = 0; i < req->r_num_ops; ++i) {
5412 			struct ceph_osd_req_op *op = &req->r_ops[i];
5413 
5414 			if (op->op == CEPH_OSD_OP_SPARSE_READ)
5415 				len += op->extent.length;
5416 		}
5417 	}
5418 	return len;
5419 }
5420 
5421 /*
5422  * Lookup and return message for incoming reply.  Don't try to do
5423  * anything about a larger than preallocated data portion of the
5424  * message at the moment - for now, just skip the message.
5425  */
get_reply(struct ceph_connection * con,struct ceph_msg_header * hdr,int * skip)5426 static struct ceph_msg *get_reply(struct ceph_connection *con,
5427 				  struct ceph_msg_header *hdr,
5428 				  int *skip)
5429 {
5430 	struct ceph_osd *osd = con->private;
5431 	struct ceph_osd_client *osdc = osd->o_osdc;
5432 	struct ceph_msg *m = NULL;
5433 	struct ceph_osd_request *req;
5434 	int front_len = le32_to_cpu(hdr->front_len);
5435 	int data_len = le32_to_cpu(hdr->data_len);
5436 	u64 tid = le64_to_cpu(hdr->tid);
5437 	u64 srlen;
5438 
5439 	down_read(&osdc->lock);
5440 	if (!osd_registered(osd)) {
5441 		dout("%s osd%d unknown, skipping\n", __func__, osd->o_osd);
5442 		*skip = 1;
5443 		goto out_unlock_osdc;
5444 	}
5445 	WARN_ON(osd->o_osd != le64_to_cpu(hdr->src.num));
5446 
5447 	mutex_lock(&osd->lock);
5448 	req = lookup_request(&osd->o_requests, tid);
5449 	if (!req) {
5450 		dout("%s osd%d tid %llu unknown, skipping\n", __func__,
5451 		     osd->o_osd, tid);
5452 		*skip = 1;
5453 		goto out_unlock_session;
5454 	}
5455 
5456 	ceph_msg_revoke_incoming(req->r_reply);
5457 
5458 	if (front_len > req->r_reply->front_alloc_len) {
5459 		pr_warn("%s osd%d tid %llu front %d > preallocated %d\n",
5460 			__func__, osd->o_osd, req->r_tid, front_len,
5461 			req->r_reply->front_alloc_len);
5462 		m = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, front_len, GFP_NOFS,
5463 				 false);
5464 		if (!m)
5465 			goto out_unlock_session;
5466 		ceph_msg_put(req->r_reply);
5467 		req->r_reply = m;
5468 	}
5469 
5470 	srlen = sparse_data_requested(req);
5471 	if (!srlen && data_len > req->r_reply->data_length) {
5472 		pr_warn("%s osd%d tid %llu data %d > preallocated %zu, skipping\n",
5473 			__func__, osd->o_osd, req->r_tid, data_len,
5474 			req->r_reply->data_length);
5475 		m = NULL;
5476 		*skip = 1;
5477 		goto out_unlock_session;
5478 	}
5479 
5480 	m = ceph_msg_get(req->r_reply);
5481 	m->sparse_read_total = srlen;
5482 
5483 	dout("get_reply tid %lld %p\n", tid, m);
5484 
5485 out_unlock_session:
5486 	mutex_unlock(&osd->lock);
5487 out_unlock_osdc:
5488 	up_read(&osdc->lock);
5489 	return m;
5490 }
5491 
alloc_msg_with_page_vector(struct ceph_msg_header * hdr)5492 static struct ceph_msg *alloc_msg_with_page_vector(struct ceph_msg_header *hdr)
5493 {
5494 	struct ceph_msg *m;
5495 	int type = le16_to_cpu(hdr->type);
5496 	u32 front_len = le32_to_cpu(hdr->front_len);
5497 	u32 data_len = le32_to_cpu(hdr->data_len);
5498 
5499 	m = ceph_msg_new2(type, front_len, 1, GFP_NOIO, false);
5500 	if (!m)
5501 		return NULL;
5502 
5503 	if (data_len) {
5504 		struct page **pages;
5505 
5506 		pages = ceph_alloc_page_vector(calc_pages_for(0, data_len),
5507 					       GFP_NOIO);
5508 		if (IS_ERR(pages)) {
5509 			ceph_msg_put(m);
5510 			return NULL;
5511 		}
5512 
5513 		ceph_msg_data_add_pages(m, pages, data_len, 0, true);
5514 	}
5515 
5516 	return m;
5517 }
5518 
osd_alloc_msg(struct ceph_connection * con,struct ceph_msg_header * hdr,int * skip)5519 static struct ceph_msg *osd_alloc_msg(struct ceph_connection *con,
5520 				      struct ceph_msg_header *hdr,
5521 				      int *skip)
5522 {
5523 	struct ceph_osd *osd = con->private;
5524 	int type = le16_to_cpu(hdr->type);
5525 
5526 	*skip = 0;
5527 	switch (type) {
5528 	case CEPH_MSG_OSD_MAP:
5529 	case CEPH_MSG_OSD_BACKOFF:
5530 	case CEPH_MSG_WATCH_NOTIFY:
5531 		return alloc_msg_with_page_vector(hdr);
5532 	case CEPH_MSG_OSD_OPREPLY:
5533 		return get_reply(con, hdr, skip);
5534 	default:
5535 		pr_warn("%s osd%d unknown msg type %d, skipping\n", __func__,
5536 			osd->o_osd, type);
5537 		*skip = 1;
5538 		return NULL;
5539 	}
5540 }
5541 
5542 /*
5543  * Wrappers to refcount containing ceph_osd struct
5544  */
osd_get_con(struct ceph_connection * con)5545 static struct ceph_connection *osd_get_con(struct ceph_connection *con)
5546 {
5547 	struct ceph_osd *osd = con->private;
5548 	if (get_osd(osd))
5549 		return con;
5550 	return NULL;
5551 }
5552 
osd_put_con(struct ceph_connection * con)5553 static void osd_put_con(struct ceph_connection *con)
5554 {
5555 	struct ceph_osd *osd = con->private;
5556 	put_osd(osd);
5557 }
5558 
5559 /*
5560  * authentication
5561  */
5562 
5563 /*
5564  * Note: returned pointer is the address of a structure that's
5565  * managed separately.  Caller must *not* attempt to free it.
5566  */
5567 static struct ceph_auth_handshake *
osd_get_authorizer(struct ceph_connection * con,int * proto,int force_new)5568 osd_get_authorizer(struct ceph_connection *con, int *proto, int force_new)
5569 {
5570 	struct ceph_osd *o = con->private;
5571 	struct ceph_osd_client *osdc = o->o_osdc;
5572 	struct ceph_auth_client *ac = osdc->client->monc.auth;
5573 	struct ceph_auth_handshake *auth = &o->o_auth;
5574 	int ret;
5575 
5576 	ret = __ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_OSD,
5577 					 force_new, proto, NULL, NULL);
5578 	if (ret)
5579 		return ERR_PTR(ret);
5580 
5581 	return auth;
5582 }
5583 
osd_add_authorizer_challenge(struct ceph_connection * con,void * challenge_buf,int challenge_buf_len)5584 static int osd_add_authorizer_challenge(struct ceph_connection *con,
5585 				    void *challenge_buf, int challenge_buf_len)
5586 {
5587 	struct ceph_osd *o = con->private;
5588 	struct ceph_osd_client *osdc = o->o_osdc;
5589 	struct ceph_auth_client *ac = osdc->client->monc.auth;
5590 
5591 	return ceph_auth_add_authorizer_challenge(ac, o->o_auth.authorizer,
5592 					    challenge_buf, challenge_buf_len);
5593 }
5594 
osd_verify_authorizer_reply(struct ceph_connection * con)5595 static int osd_verify_authorizer_reply(struct ceph_connection *con)
5596 {
5597 	struct ceph_osd *o = con->private;
5598 	struct ceph_osd_client *osdc = o->o_osdc;
5599 	struct ceph_auth_client *ac = osdc->client->monc.auth;
5600 	struct ceph_auth_handshake *auth = &o->o_auth;
5601 
5602 	return ceph_auth_verify_authorizer_reply(ac, auth->authorizer,
5603 		auth->authorizer_reply_buf, auth->authorizer_reply_buf_len,
5604 		NULL, NULL, NULL, NULL);
5605 }
5606 
osd_invalidate_authorizer(struct ceph_connection * con)5607 static int osd_invalidate_authorizer(struct ceph_connection *con)
5608 {
5609 	struct ceph_osd *o = con->private;
5610 	struct ceph_osd_client *osdc = o->o_osdc;
5611 	struct ceph_auth_client *ac = osdc->client->monc.auth;
5612 
5613 	ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_OSD);
5614 	return ceph_monc_validate_auth(&osdc->client->monc);
5615 }
5616 
osd_get_auth_request(struct ceph_connection * con,void * buf,int * buf_len,void ** authorizer,int * authorizer_len)5617 static int osd_get_auth_request(struct ceph_connection *con,
5618 				void *buf, int *buf_len,
5619 				void **authorizer, int *authorizer_len)
5620 {
5621 	struct ceph_osd *o = con->private;
5622 	struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5623 	struct ceph_auth_handshake *auth = &o->o_auth;
5624 	int ret;
5625 
5626 	ret = ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_OSD,
5627 				       buf, buf_len);
5628 	if (ret)
5629 		return ret;
5630 
5631 	*authorizer = auth->authorizer_buf;
5632 	*authorizer_len = auth->authorizer_buf_len;
5633 	return 0;
5634 }
5635 
osd_handle_auth_reply_more(struct ceph_connection * con,void * reply,int reply_len,void * buf,int * buf_len,void ** authorizer,int * authorizer_len)5636 static int osd_handle_auth_reply_more(struct ceph_connection *con,
5637 				      void *reply, int reply_len,
5638 				      void *buf, int *buf_len,
5639 				      void **authorizer, int *authorizer_len)
5640 {
5641 	struct ceph_osd *o = con->private;
5642 	struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5643 	struct ceph_auth_handshake *auth = &o->o_auth;
5644 	int ret;
5645 
5646 	ret = ceph_auth_handle_svc_reply_more(ac, auth, reply, reply_len,
5647 					      buf, buf_len);
5648 	if (ret)
5649 		return ret;
5650 
5651 	*authorizer = auth->authorizer_buf;
5652 	*authorizer_len = auth->authorizer_buf_len;
5653 	return 0;
5654 }
5655 
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)5656 static int osd_handle_auth_done(struct ceph_connection *con,
5657 				u64 global_id, void *reply, int reply_len,
5658 				u8 *session_key, int *session_key_len,
5659 				u8 *con_secret, int *con_secret_len)
5660 {
5661 	struct ceph_osd *o = con->private;
5662 	struct ceph_auth_client *ac = o->o_osdc->client->monc.auth;
5663 	struct ceph_auth_handshake *auth = &o->o_auth;
5664 
5665 	return ceph_auth_handle_svc_reply_done(ac, auth, reply, reply_len,
5666 					       session_key, session_key_len,
5667 					       con_secret, con_secret_len);
5668 }
5669 
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)5670 static int osd_handle_auth_bad_method(struct ceph_connection *con,
5671 				      int used_proto, int result,
5672 				      const int *allowed_protos, int proto_cnt,
5673 				      const int *allowed_modes, int mode_cnt)
5674 {
5675 	struct ceph_osd *o = con->private;
5676 	struct ceph_mon_client *monc = &o->o_osdc->client->monc;
5677 	int ret;
5678 
5679 	if (ceph_auth_handle_bad_authorizer(monc->auth, CEPH_ENTITY_TYPE_OSD,
5680 					    used_proto, result,
5681 					    allowed_protos, proto_cnt,
5682 					    allowed_modes, mode_cnt)) {
5683 		ret = ceph_monc_validate_auth(monc);
5684 		if (ret)
5685 			return ret;
5686 	}
5687 
5688 	return -EACCES;
5689 }
5690 
osd_reencode_message(struct ceph_msg * msg)5691 static void osd_reencode_message(struct ceph_msg *msg)
5692 {
5693 	int type = le16_to_cpu(msg->hdr.type);
5694 
5695 	if (type == CEPH_MSG_OSD_OP)
5696 		encode_request_finish(msg);
5697 }
5698 
osd_sign_message(struct ceph_msg * msg)5699 static int osd_sign_message(struct ceph_msg *msg)
5700 {
5701 	struct ceph_osd *o = msg->con->private;
5702 	struct ceph_auth_handshake *auth = &o->o_auth;
5703 
5704 	return ceph_auth_sign_message(auth, msg);
5705 }
5706 
osd_check_message_signature(struct ceph_msg * msg)5707 static int osd_check_message_signature(struct ceph_msg *msg)
5708 {
5709 	struct ceph_osd *o = msg->con->private;
5710 	struct ceph_auth_handshake *auth = &o->o_auth;
5711 
5712 	return ceph_auth_check_message_signature(auth, msg);
5713 }
5714 
advance_cursor(struct ceph_msg_data_cursor * cursor,size_t len,bool zero)5715 static void advance_cursor(struct ceph_msg_data_cursor *cursor, size_t len,
5716 			   bool zero)
5717 {
5718 	while (len) {
5719 		struct page *page;
5720 		size_t poff, plen;
5721 
5722 		page = ceph_msg_data_next(cursor, &poff, &plen);
5723 		if (plen > len)
5724 			plen = len;
5725 		if (zero)
5726 			zero_user_segment(page, poff, poff + plen);
5727 		len -= plen;
5728 		ceph_msg_data_advance(cursor, plen);
5729 	}
5730 }
5731 
prep_next_sparse_read(struct ceph_connection * con,struct ceph_msg_data_cursor * cursor)5732 static int prep_next_sparse_read(struct ceph_connection *con,
5733 				 struct ceph_msg_data_cursor *cursor)
5734 {
5735 	struct ceph_osd *o = con->private;
5736 	struct ceph_sparse_read *sr = &o->o_sparse_read;
5737 	struct ceph_osd_request *req;
5738 	struct ceph_osd_req_op *op;
5739 
5740 	spin_lock(&o->o_requests_lock);
5741 	req = lookup_request(&o->o_requests, le64_to_cpu(con->in_msg->hdr.tid));
5742 	if (!req) {
5743 		spin_unlock(&o->o_requests_lock);
5744 		return -EBADR;
5745 	}
5746 
5747 	if (o->o_sparse_op_idx < 0) {
5748 		dout("%s: [%d] starting new sparse read req\n",
5749 		     __func__, o->o_osd);
5750 	} else {
5751 		u64 end;
5752 
5753 		op = &req->r_ops[o->o_sparse_op_idx];
5754 
5755 		WARN_ON_ONCE(op->extent.sparse_ext);
5756 
5757 		/* hand back buffer we took earlier */
5758 		op->extent.sparse_ext = sr->sr_extent;
5759 		sr->sr_extent = NULL;
5760 		op->extent.sparse_ext_cnt = sr->sr_count;
5761 		sr->sr_ext_len = 0;
5762 		dout("%s: [%d] completed extent array len %d cursor->resid %zd\n",
5763 		     __func__, o->o_osd, op->extent.sparse_ext_cnt, cursor->resid);
5764 		/* Advance to end of data for this operation */
5765 		end = ceph_sparse_ext_map_end(op);
5766 		if (end < sr->sr_req_len)
5767 			advance_cursor(cursor, sr->sr_req_len - end, false);
5768 	}
5769 
5770 	ceph_init_sparse_read(sr);
5771 
5772 	/* find next op in this request (if any) */
5773 	while (++o->o_sparse_op_idx < req->r_num_ops) {
5774 		op = &req->r_ops[o->o_sparse_op_idx];
5775 		if (op->op == CEPH_OSD_OP_SPARSE_READ)
5776 			goto found;
5777 	}
5778 
5779 	/* reset for next sparse read request */
5780 	spin_unlock(&o->o_requests_lock);
5781 	o->o_sparse_op_idx = -1;
5782 	return 0;
5783 found:
5784 	sr->sr_req_off = op->extent.offset;
5785 	sr->sr_req_len = op->extent.length;
5786 	sr->sr_pos = sr->sr_req_off;
5787 	dout("%s: [%d] new sparse read op at idx %d 0x%llx~0x%llx\n", __func__,
5788 	     o->o_osd, o->o_sparse_op_idx, sr->sr_req_off, sr->sr_req_len);
5789 
5790 	/* hand off request's sparse extent map buffer */
5791 	sr->sr_ext_len = op->extent.sparse_ext_cnt;
5792 	op->extent.sparse_ext_cnt = 0;
5793 	sr->sr_extent = op->extent.sparse_ext;
5794 	op->extent.sparse_ext = NULL;
5795 
5796 	spin_unlock(&o->o_requests_lock);
5797 	return 1;
5798 }
5799 
5800 #ifdef __BIG_ENDIAN
convert_extent_map(struct ceph_sparse_read * sr)5801 static inline void convert_extent_map(struct ceph_sparse_read *sr)
5802 {
5803 	int i;
5804 
5805 	for (i = 0; i < sr->sr_count; i++) {
5806 		struct ceph_sparse_extent *ext = &sr->sr_extent[i];
5807 
5808 		ext->off = le64_to_cpu((__force __le64)ext->off);
5809 		ext->len = le64_to_cpu((__force __le64)ext->len);
5810 	}
5811 }
5812 #else
convert_extent_map(struct ceph_sparse_read * sr)5813 static inline void convert_extent_map(struct ceph_sparse_read *sr)
5814 {
5815 }
5816 #endif
5817 
osd_sparse_read(struct ceph_connection * con,struct ceph_msg_data_cursor * cursor,char ** pbuf)5818 static int osd_sparse_read(struct ceph_connection *con,
5819 			   struct ceph_msg_data_cursor *cursor,
5820 			   char **pbuf)
5821 {
5822 	struct ceph_osd *o = con->private;
5823 	struct ceph_sparse_read *sr = &o->o_sparse_read;
5824 	u32 count = sr->sr_count;
5825 	u64 eoff, elen, len = 0;
5826 	int i, ret;
5827 
5828 	switch (sr->sr_state) {
5829 	case CEPH_SPARSE_READ_HDR:
5830 next_op:
5831 		ret = prep_next_sparse_read(con, cursor);
5832 		if (ret <= 0)
5833 			return ret;
5834 
5835 		/* number of extents */
5836 		ret = sizeof(sr->sr_count);
5837 		*pbuf = (char *)&sr->sr_count;
5838 		sr->sr_state = CEPH_SPARSE_READ_EXTENTS;
5839 		break;
5840 	case CEPH_SPARSE_READ_EXTENTS:
5841 		/* Convert sr_count to host-endian */
5842 		count = le32_to_cpu((__force __le32)sr->sr_count);
5843 		sr->sr_count = count;
5844 		dout("[%d] got %u extents\n", o->o_osd, count);
5845 
5846 		if (count > 0) {
5847 			if (!sr->sr_extent || count > sr->sr_ext_len) {
5848 				/* no extent array provided, or too short */
5849 				kfree(sr->sr_extent);
5850 				sr->sr_extent = kmalloc_array(count,
5851 							      sizeof(*sr->sr_extent),
5852 							      GFP_NOIO);
5853 				if (!sr->sr_extent) {
5854 					pr_err("%s: failed to allocate %u extents\n",
5855 					       __func__, count);
5856 					return -ENOMEM;
5857 				}
5858 				sr->sr_ext_len = count;
5859 			}
5860 			ret = count * sizeof(*sr->sr_extent);
5861 			*pbuf = (char *)sr->sr_extent;
5862 			sr->sr_state = CEPH_SPARSE_READ_DATA_LEN;
5863 			break;
5864 		}
5865 		/* No extents? Read data len */
5866 		fallthrough;
5867 	case CEPH_SPARSE_READ_DATA_LEN:
5868 		convert_extent_map(sr);
5869 		ret = sizeof(sr->sr_datalen);
5870 		*pbuf = (char *)&sr->sr_datalen;
5871 		sr->sr_state = CEPH_SPARSE_READ_DATA_PRE;
5872 		break;
5873 	case CEPH_SPARSE_READ_DATA_PRE:
5874 		/* Convert sr_datalen to host-endian */
5875 		sr->sr_datalen = le32_to_cpu((__force __le32)sr->sr_datalen);
5876 		for (i = 0; i < count; i++)
5877 			len += sr->sr_extent[i].len;
5878 		if (sr->sr_datalen != len) {
5879 			pr_warn_ratelimited("data len %u != extent len %llu\n",
5880 					    sr->sr_datalen, len);
5881 			return -EREMOTEIO;
5882 		}
5883 		sr->sr_state = CEPH_SPARSE_READ_DATA;
5884 		fallthrough;
5885 	case CEPH_SPARSE_READ_DATA:
5886 		if (sr->sr_index >= count) {
5887 			sr->sr_state = CEPH_SPARSE_READ_HDR;
5888 			goto next_op;
5889 		}
5890 
5891 		eoff = sr->sr_extent[sr->sr_index].off;
5892 		elen = sr->sr_extent[sr->sr_index].len;
5893 
5894 		dout("[%d] ext %d off 0x%llx len 0x%llx\n",
5895 		     o->o_osd, sr->sr_index, eoff, elen);
5896 
5897 		if (elen > INT_MAX) {
5898 			dout("Sparse read extent length too long (0x%llx)\n",
5899 			     elen);
5900 			return -EREMOTEIO;
5901 		}
5902 
5903 		/* zero out anything from sr_pos to start of extent */
5904 		if (sr->sr_pos < eoff)
5905 			advance_cursor(cursor, eoff - sr->sr_pos, true);
5906 
5907 		/* Set position to end of extent */
5908 		sr->sr_pos = eoff + elen;
5909 
5910 		/* send back the new length and nullify the ptr */
5911 		cursor->sr_resid = elen;
5912 		ret = elen;
5913 		*pbuf = NULL;
5914 
5915 		/* Bump the array index */
5916 		++sr->sr_index;
5917 		break;
5918 	}
5919 	return ret;
5920 }
5921 
5922 static const struct ceph_connection_operations osd_con_ops = {
5923 	.get = osd_get_con,
5924 	.put = osd_put_con,
5925 	.sparse_read = osd_sparse_read,
5926 	.alloc_msg = osd_alloc_msg,
5927 	.dispatch = osd_dispatch,
5928 	.fault = osd_fault,
5929 	.reencode_message = osd_reencode_message,
5930 	.get_authorizer = osd_get_authorizer,
5931 	.add_authorizer_challenge = osd_add_authorizer_challenge,
5932 	.verify_authorizer_reply = osd_verify_authorizer_reply,
5933 	.invalidate_authorizer = osd_invalidate_authorizer,
5934 	.sign_message = osd_sign_message,
5935 	.check_message_signature = osd_check_message_signature,
5936 	.get_auth_request = osd_get_auth_request,
5937 	.handle_auth_reply_more = osd_handle_auth_reply_more,
5938 	.handle_auth_done = osd_handle_auth_done,
5939 	.handle_auth_bad_method = osd_handle_auth_bad_method,
5940 };
5941