projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
libceph: pass object number back to calc_layout() caller
[linux-2.6-block.git] / net / ceph / osd_client.c
CommitLineData
3d14c5d2 1#include <linux/ceph/ceph_debug.h>
f24e9980 2
3d14c5d2 3#include <linux/module.h>
f24e9980
SW		 4#include <linux/err.h>
5#include <linux/highmem.h>
6#include <linux/mm.h>
7#include <linux/pagemap.h>
8#include <linux/slab.h>
9#include <linux/uaccess.h>
68b4476b
YS		 10#ifdef CONFIG_BLOCK
11#include <linux/bio.h>
12#endif
f24e9980 13
3d14c5d2
YS		 14#include <linux/ceph/libceph.h>
15#include <linux/ceph/osd_client.h>
16#include <linux/ceph/messenger.h>
17#include <linux/ceph/decode.h>
18#include <linux/ceph/auth.h>
19#include <linux/ceph/pagelist.h>
f24e9980 20
c16e7869
SW		 21#define OSD_OP_FRONT_LEN 4096
22#define OSD_OPREPLY_FRONT_LEN 512
0d59ab81 23
9e32789f 24static const struct ceph_connection_operations osd_con_ops;
f24e9980 25
f9d25199 26static void __send_queued(struct ceph_osd_client *osdc);
6f6c7006 27static int __reset_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd);
a40c4f10
YS		 28static void __register_request(struct ceph_osd_client *osdc,
29 struct ceph_osd_request *req);
30static void __unregister_linger_request(struct ceph_osd_client *osdc,
31 struct ceph_osd_request *req);
56e925b6
SW		 32static void __send_request(struct ceph_osd_client *osdc,
33 struct ceph_osd_request *req);
f24e9980 34
68b4476b
YS		 35static int op_has_extent(int op)
36{
37 return (op == CEPH_OSD_OP_READ ||
38 op == CEPH_OSD_OP_WRITE);
39}
40
f24e9980
SW		 41/*
42 * Implement client access to distributed object storage cluster.
43 *
44 * All data objects are stored within a cluster/cloud of OSDs, or
45 * "object storage devices." (Note that Ceph OSDs have _nothing_ to
46 * do with the T10 OSD extensions to SCSI.) Ceph OSDs are simply
47 * remote daemons serving up and coordinating consistent and safe
48 * access to storage.
49 *
50 * Cluster membership and the mapping of data objects onto storage devices
51 * are described by the osd map.
52 *
53 * We keep track of pending OSD requests (read, write), resubmit
54 * requests to different OSDs when the cluster topology/data layout
55 * change, or retry the affected requests when the communications
56 * channel with an OSD is reset.
57 */
58
59/*
60 * calculate the mapping of a file extent onto an object, and fill out the
61 * request accordingly. shorten extent as necessary if it crosses an
62 * object boundary.
63 *
64 * fill osd op in request message.
65 */
e75b45cf 66static int calc_layout(struct ceph_vino vino,
d63b77f4
SW		 67 struct ceph_file_layout *layout,
68 u64 off, u64 *plen,
69 struct ceph_osd_request *req,
47a05811 70 struct ceph_osd_req_op *op, u64 *bno)
f24e9980 71{
60e56f13 72 u64 orig_len = *plen;
60e56f13
AE		 73 u64 objoff = 0;
74 u64 objlen = 0;
d63b77f4 75 int r;
f24e9980 76
60e56f13 77 /* object extent? */
47a05811 78 r = ceph_calc_file_object_mapping(layout, off, orig_len, bno,
60e56f13 79 &objoff, &objlen);
d63b77f4
SW		 80 if (r < 0)
81 return r;
60e56f13
AE		 82 if (objlen < orig_len) {
83 *plen = objlen;
84 dout(" skipping last %llu, final file extent %llu~%llu\n",
85 orig_len - *plen, off, *plen);
86 }
87
88 if (op_has_extent(op->op)) {
89 u32 osize = le32_to_cpu(layout->fl_object_size);
90 op->extent.offset = objoff;
91 op->extent.length = objlen;
92 if (op->extent.truncate_size <= off - objoff) {
93 op->extent.truncate_size = 0;
94 } else {
95 op->extent.truncate_size -= off - objoff;
96 if (op->extent.truncate_size > osize)
97 op->extent.truncate_size = osize;
98 }
99 }
100 req->r_num_pages = calc_pages_for(off, *plen);
101 req->r_page_alignment = off & ~PAGE_MASK;
102 if (op->op == CEPH_OSD_OP_WRITE)
103 op->payload_len = *plen;
104
105 dout("calc_layout bno=%llx %llu~%llu (%d pages)\n",
47a05811 106 *bno, objoff, objlen, req->r_num_pages);
f24e9980 107
47a05811 108 snprintf(req->r_oid, sizeof(req->r_oid), "%llx.%08llx", vino.ino, *bno);
f24e9980 109 req->r_oid_len = strlen(req->r_oid);
d63b77f4 110
3ff5f385 111 return 0;
f24e9980
SW		 112}
113
f24e9980
SW		 114/*
115 * requests
116 */
415e49a9 117void ceph_osdc_release_request(struct kref *kref)
f24e9980 118{
415e49a9
SW		 119 struct ceph_osd_request *req = container_of(kref,
120 struct ceph_osd_request,
121 r_kref);
122
123 if (req->r_request)
124 ceph_msg_put(req->r_request);
0d59ab81 125 if (req->r_con_filling_msg) {
9cbb1d72
AE		 126 dout("%s revoking msg %p from con %p\n", __func__,
127 req->r_reply, req->r_con_filling_msg);
8921d114 128 ceph_msg_revoke_incoming(req->r_reply);
0d47766f 129 req->r_con_filling_msg->ops->put(req->r_con_filling_msg);
9cbb1d72 130 req->r_con_filling_msg = NULL;
350b1c32 131 }
ab8cb34a
AE		 132 if (req->r_reply)
133 ceph_msg_put(req->r_reply);
415e49a9
SW		 134 if (req->r_own_pages)
135 ceph_release_page_vector(req->r_pages,
136 req->r_num_pages);
137 ceph_put_snap_context(req->r_snapc);
c885837f 138 ceph_pagelist_release(&req->r_trail);
415e49a9
SW		 139 if (req->r_mempool)
140 mempool_free(req, req->r_osdc->req_mempool);
141 else
142 kfree(req);
f24e9980 143}
3d14c5d2 144EXPORT_SYMBOL(ceph_osdc_release_request);
68b4476b 145
3499e8a5 146struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
f24e9980 147 struct ceph_snap_context *snapc,
1b83bef2 148 unsigned int num_ops,
3499e8a5 149 bool use_mempool,
54a54007 150 gfp_t gfp_flags)
f24e9980
SW		 151{
152 struct ceph_osd_request *req;
153 struct ceph_msg *msg;
1b83bef2
SW		 154 size_t msg_size;
155
156 msg_size = 4 + 4 + 8 + 8 + 4+8;
157 msg_size += 2 + 4 + 8 + 4 + 4; /* oloc */
158 msg_size += 1 + 8 + 4 + 4; /* pg_t */
159 msg_size += 4 + MAX_OBJ_NAME_SIZE;
160 msg_size += 2 + num_ops*sizeof(struct ceph_osd_op);
161 msg_size += 8; /* snapid */
162 msg_size += 8; /* snap_seq */
163 msg_size += 8 * (snapc ? snapc->num_snaps : 0); /* snaps */
164 msg_size += 4;
f24e9980
SW		 165
166 if (use_mempool) {
3499e8a5 167 req = mempool_alloc(osdc->req_mempool, gfp_flags);
f24e9980
SW		 168 memset(req, 0, sizeof(*req));
169 } else {
3499e8a5 170 req = kzalloc(sizeof(*req), gfp_flags);
f24e9980
SW		 171 }
172 if (req == NULL)
a79832f2 173 return NULL;
f24e9980 174
f24e9980
SW		 175 req->r_osdc = osdc;
176 req->r_mempool = use_mempool;
68b4476b 177
415e49a9 178 kref_init(&req->r_kref);
f24e9980
SW		 179 init_completion(&req->r_completion);
180 init_completion(&req->r_safe_completion);
a978fa20 181 RB_CLEAR_NODE(&req->r_node);
f24e9980 182 INIT_LIST_HEAD(&req->r_unsafe_item);
a40c4f10
YS		 183 INIT_LIST_HEAD(&req->r_linger_item);
184 INIT_LIST_HEAD(&req->r_linger_osd);
935b639a 185 INIT_LIST_HEAD(&req->r_req_lru_item);
cd43045c
SW		 186 INIT_LIST_HEAD(&req->r_osd_item);
187
c16e7869
SW		 188 /* create reply message */
189 if (use_mempool)
190 msg = ceph_msgpool_get(&osdc->msgpool_op_reply, 0);
191 else
192 msg = ceph_msg_new(CEPH_MSG_OSD_OPREPLY,
b61c2763 193 OSD_OPREPLY_FRONT_LEN, gfp_flags, true);
a79832f2 194 if (!msg) {
c16e7869 195 ceph_osdc_put_request(req);
a79832f2 196 return NULL;
c16e7869
SW		 197 }
198 req->r_reply = msg;
199
c885837f 200 ceph_pagelist_init(&req->r_trail);
d50b409f 201
c16e7869 202 /* create request message; allow space for oid */
f24e9980 203 if (use_mempool)
8f3bc053 204 msg = ceph_msgpool_get(&osdc->msgpool_op, 0);
f24e9980 205 else
b61c2763 206 msg = ceph_msg_new(CEPH_MSG_OSD_OP, msg_size, gfp_flags, true);
a79832f2 207 if (!msg) {
f24e9980 208 ceph_osdc_put_request(req);
a79832f2 209 return NULL;
f24e9980 210 }
68b4476b 211
f24e9980 212 memset(msg->front.iov_base, 0, msg->front.iov_len);
3499e8a5
YS		 213
214 req->r_request = msg;
3499e8a5
YS		 215
216 return req;
217}
3d14c5d2 218EXPORT_SYMBOL(ceph_osdc_alloc_request);
3499e8a5 219
68b4476b
YS		 220static void osd_req_encode_op(struct ceph_osd_request *req,
221 struct ceph_osd_op *dst,
222 struct ceph_osd_req_op *src)
223{
224 dst->op = cpu_to_le16(src->op);
225
065a68f9 226 switch (src->op) {
fbfab539
AE		 227 case CEPH_OSD_OP_STAT:
228 break;
68b4476b
YS		 229 case CEPH_OSD_OP_READ:
230 case CEPH_OSD_OP_WRITE:
231 dst->extent.offset =
232 cpu_to_le64(src->extent.offset);
233 dst->extent.length =
234 cpu_to_le64(src->extent.length);
235 dst->extent.truncate_size =
236 cpu_to_le64(src->extent.truncate_size);
237 dst->extent.truncate_seq =
238 cpu_to_le32(src->extent.truncate_seq);
239 break;
ae1533b6 240 case CEPH_OSD_OP_CALL:
ae1533b6
YS		 241 dst->cls.class_len = src->cls.class_len;
242 dst->cls.method_len = src->cls.method_len;
243 dst->cls.indata_len = cpu_to_le32(src->cls.indata_len);
244
c885837f 245 ceph_pagelist_append(&req->r_trail, src->cls.class_name,
ae1533b6 246 src->cls.class_len);
c885837f 247 ceph_pagelist_append(&req->r_trail, src->cls.method_name,
ae1533b6 248 src->cls.method_len);
c885837f 249 ceph_pagelist_append(&req->r_trail, src->cls.indata,
ae1533b6
YS		 250 src->cls.indata_len);
251 break;
68b4476b
YS		 252 case CEPH_OSD_OP_STARTSYNC:
253 break;
a40c4f10
YS		 254 case CEPH_OSD_OP_NOTIFY_ACK:
255 case CEPH_OSD_OP_WATCH:
256 dst->watch.cookie = cpu_to_le64(src->watch.cookie);
257 dst->watch.ver = cpu_to_le64(src->watch.ver);
258 dst->watch.flag = src->watch.flag;
259 break;
68b4476b
YS		 260 default:
261 pr_err("unrecognized osd opcode %d\n", dst->op);
262 WARN_ON(1);
263 break;
4c46459c
AE		 264 case CEPH_OSD_OP_MAPEXT:
265 case CEPH_OSD_OP_MASKTRUNC:
266 case CEPH_OSD_OP_SPARSE_READ:
a9f36c3e 267 case CEPH_OSD_OP_NOTIFY:
4c46459c
AE		 268 case CEPH_OSD_OP_ASSERT_VER:
269 case CEPH_OSD_OP_WRITEFULL:
270 case CEPH_OSD_OP_TRUNCATE:
271 case CEPH_OSD_OP_ZERO:
272 case CEPH_OSD_OP_DELETE:
273 case CEPH_OSD_OP_APPEND:
274 case CEPH_OSD_OP_SETTRUNC:
275 case CEPH_OSD_OP_TRIMTRUNC:
276 case CEPH_OSD_OP_TMAPUP:
277 case CEPH_OSD_OP_TMAPPUT:
278 case CEPH_OSD_OP_TMAPGET:
279 case CEPH_OSD_OP_CREATE:
a9f36c3e 280 case CEPH_OSD_OP_ROLLBACK:
4c46459c
AE		 281 case CEPH_OSD_OP_OMAPGETKEYS:
282 case CEPH_OSD_OP_OMAPGETVALS:
283 case CEPH_OSD_OP_OMAPGETHEADER:
284 case CEPH_OSD_OP_OMAPGETVALSBYKEYS:
285 case CEPH_OSD_OP_MODE_RD:
286 case CEPH_OSD_OP_OMAPSETVALS:
287 case CEPH_OSD_OP_OMAPSETHEADER:
288 case CEPH_OSD_OP_OMAPCLEAR:
289 case CEPH_OSD_OP_OMAPRMKEYS:
290 case CEPH_OSD_OP_OMAP_CMP:
291 case CEPH_OSD_OP_CLONERANGE:
292 case CEPH_OSD_OP_ASSERT_SRC_VERSION:
293 case CEPH_OSD_OP_SRC_CMPXATTR:
a9f36c3e 294 case CEPH_OSD_OP_GETXATTR:
4c46459c 295 case CEPH_OSD_OP_GETXATTRS:
a9f36c3e
AE		 296 case CEPH_OSD_OP_CMPXATTR:
297 case CEPH_OSD_OP_SETXATTR:
4c46459c
AE		 298 case CEPH_OSD_OP_SETXATTRS:
299 case CEPH_OSD_OP_RESETXATTRS:
300 case CEPH_OSD_OP_RMXATTR:
301 case CEPH_OSD_OP_PULL:
302 case CEPH_OSD_OP_PUSH:
303 case CEPH_OSD_OP_BALANCEREADS:
304 case CEPH_OSD_OP_UNBALANCEREADS:
305 case CEPH_OSD_OP_SCRUB:
306 case CEPH_OSD_OP_SCRUB_RESERVE:
307 case CEPH_OSD_OP_SCRUB_UNRESERVE:
308 case CEPH_OSD_OP_SCRUB_STOP:
309 case CEPH_OSD_OP_SCRUB_MAP:
310 case CEPH_OSD_OP_WRLOCK:
311 case CEPH_OSD_OP_WRUNLOCK:
312 case CEPH_OSD_OP_RDLOCK:
313 case CEPH_OSD_OP_RDUNLOCK:
314 case CEPH_OSD_OP_UPLOCK:
315 case CEPH_OSD_OP_DNLOCK:
316 case CEPH_OSD_OP_PGLS:
317 case CEPH_OSD_OP_PGLS_FILTER:
318 pr_err("unsupported osd opcode %s\n",
319 ceph_osd_op_name(dst->op));
320 WARN_ON(1);
321 break;
68b4476b
YS		 322 }
323 dst->payload_len = cpu_to_le32(src->payload_len);
324}
325
3499e8a5
YS		 326/*
327 * build new request AND message
328 *
329 */
330void ceph_osdc_build_request(struct ceph_osd_request *req,
1b83bef2 331 u64 off, u64 len, unsigned int num_ops,
68b4476b 332 struct ceph_osd_req_op *src_ops,
4d6b250b 333 struct ceph_snap_context *snapc, u64 snap_id,
af77f26c 334 struct timespec *mtime)
3499e8a5
YS		 335{
336 struct ceph_msg *msg = req->r_request;
68b4476b 337 struct ceph_osd_req_op *src_op;
3499e8a5 338 void *p;
1b83bef2 339 size_t msg_size;
3499e8a5 340 int flags = req->r_flags;
f44246e3 341 u64 data_len;
68b4476b 342 int i;
3499e8a5 343
1b83bef2
SW		 344 req->r_num_ops = num_ops;
345 req->r_snapid = snap_id;
f24e9980
SW		 346 req->r_snapc = ceph_get_snap_context(snapc);
347
1b83bef2
SW		 348 /* encode request */
349 msg->hdr.version = cpu_to_le16(4);
350
351 p = msg->front.iov_base;
352 ceph_encode_32(&p, 1); /* client_inc is always 1 */
353 req->r_request_osdmap_epoch = p;
354 p += 4;
355 req->r_request_flags = p;
356 p += 4;
357 if (req->r_flags & CEPH_OSD_FLAG_WRITE)
358 ceph_encode_timespec(p, mtime);
359 p += sizeof(struct ceph_timespec);
360 req->r_request_reassert_version = p;
361 p += sizeof(struct ceph_eversion); /* will get filled in */
362
363 /* oloc */
364 ceph_encode_8(&p, 4);
365 ceph_encode_8(&p, 4);
366 ceph_encode_32(&p, 8 + 4 + 4);
367 req->r_request_pool = p;
368 p += 8;
369 ceph_encode_32(&p, -1); /* preferred */
370 ceph_encode_32(&p, 0); /* key len */
371
372 ceph_encode_8(&p, 1);
373 req->r_request_pgid = p;
374 p += 8 + 4;
375 ceph_encode_32(&p, -1); /* preferred */
376
377 /* oid */
378 ceph_encode_32(&p, req->r_oid_len);
af77f26c 379 memcpy(p, req->r_oid, req->r_oid_len);
1b83bef2 380 dout("oid '%.*s' len %d\n", req->r_oid_len, req->r_oid, req->r_oid_len);
af77f26c 381 p += req->r_oid_len;
f24e9980 382
1b83bef2
SW		 383 /* ops */
384 ceph_encode_16(&p, num_ops);
68b4476b 385 src_op = src_ops;
1b83bef2
SW		 386 req->r_request_ops = p;
387 for (i = 0; i < num_ops; i++, src_op++) {
388 osd_req_encode_op(req, p, src_op);
389 p += sizeof(struct ceph_osd_op);
390 }
68b4476b 391
1b83bef2
SW		 392 /* snaps */
393 ceph_encode_64(&p, req->r_snapid);
394 ceph_encode_64(&p, req->r_snapc ? req->r_snapc->seq : 0);
395 ceph_encode_32(&p, req->r_snapc ? req->r_snapc->num_snaps : 0);
396 if (req->r_snapc) {
f24e9980 397 for (i = 0; i < snapc->num_snaps; i++) {
1b83bef2 398 ceph_encode_64(&p, req->r_snapc->snaps[i]);
f24e9980
SW		 399 }
400 }
401
1b83bef2
SW		 402 req->r_request_attempts = p;
403 p += 4;
404
f44246e3 405 data_len = req->r_trail.length;
68b4476b
YS		 406 if (flags & CEPH_OSD_FLAG_WRITE) {
407 req->r_request->hdr.data_off = cpu_to_le16(off);
f44246e3 408 data_len += len;
68b4476b 409 }
f44246e3 410 req->r_request->hdr.data_len = cpu_to_le32(data_len);
c5c6b19d
SW		 411 req->r_request->page_alignment = req->r_page_alignment;
412
f24e9980 413 BUG_ON(p > msg->front.iov_base + msg->front.iov_len);
6f863e71
SW		 414 msg_size = p - msg->front.iov_base;
415 msg->front.iov_len = msg_size;
416 msg->hdr.front_len = cpu_to_le32(msg_size);
1b83bef2
SW		 417
418 dout("build_request msg_size was %d num_ops %d\n", (int)msg_size,
419 num_ops);
3499e8a5
YS		 420 return;
421}
3d14c5d2 422EXPORT_SYMBOL(ceph_osdc_build_request);
3499e8a5
YS		 423
424/*
425 * build new request AND message, calculate layout, and adjust file
426 * extent as needed.
427 *
428 * if the file was recently truncated, we include information about its
429 * old and new size so that the object can be updated appropriately. (we
430 * avoid synchronously deleting truncated objects because it's slow.)
431 *
432 * if @do_sync, include a 'startsync' command so that the osd will flush
433 * data quickly.
434 */
435struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *osdc,
436 struct ceph_file_layout *layout,
437 struct ceph_vino vino,
438 u64 off, u64 *plen,
439 int opcode, int flags,
440 struct ceph_snap_context *snapc,
441 int do_sync,
442 u32 truncate_seq,
443 u64 truncate_size,
444 struct timespec *mtime,
a3bea47e 445 bool use_mempool,
b7495fc2 446 int page_align)
3499e8a5 447{
ae7ca4a3 448 struct ceph_osd_req_op ops[2];
68b4476b 449 struct ceph_osd_request *req;
ae7ca4a3 450 unsigned int num_op = 1;
47a05811 451 u64 bno = 0;
6816282d 452 int r;
68b4476b 453
ae7ca4a3
AE		 454 memset(&ops, 0, sizeof ops);
455
68b4476b
YS		 456 ops[0].op = opcode;
457 ops[0].extent.truncate_seq = truncate_seq;
458 ops[0].extent.truncate_size = truncate_size;
68b4476b
YS		 459
460 if (do_sync) {
461 ops[1].op = CEPH_OSD_OP_STARTSYNC;
ae7ca4a3
AE		 462 num_op++;
463 }
68b4476b 464
ae7ca4a3
AE		 465 req = ceph_osdc_alloc_request(osdc, snapc, num_op, use_mempool,
466 GFP_NOFS);
4ad12621 467 if (!req)
6816282d 468 return ERR_PTR(-ENOMEM);
d178a9e7 469 req->r_flags = flags;
3499e8a5
YS		 470
471 /* calculate max write size */
47a05811 472 r = calc_layout(vino, layout, off, plen, req, ops, &bno);
3ff5f385
AE		 473 if (r < 0) {
474 ceph_osdc_put_request(req);
6816282d 475 return ERR_PTR(r);
3ff5f385 476 }
47a05811 477
3499e8a5
YS		 478 req->r_file_layout = *layout; /* keep a copy */
479
9bb0ce2b
SW		 480 /* in case it differs from natural (file) alignment that
481 calc_layout filled in for us */
482 req->r_num_pages = calc_pages_for(page_align, *plen);
b7495fc2
SW		 483 req->r_page_alignment = page_align;
484
ae7ca4a3
AE		 485 ceph_osdc_build_request(req, off, *plen, num_op, ops,
486 snapc, vino.snap, mtime);
3499e8a5 487
f24e9980
SW		 488 return req;
489}
3d14c5d2 490EXPORT_SYMBOL(ceph_osdc_new_request);
f24e9980
SW		 491
492/*
493 * We keep osd requests in an rbtree, sorted by ->r_tid.
494 */
495static void __insert_request(struct ceph_osd_client *osdc,
496 struct ceph_osd_request *new)
497{
498 struct rb_node **p = &osdc->requests.rb_node;
499 struct rb_node *parent = NULL;
500 struct ceph_osd_request *req = NULL;
501
502 while (*p) {
503 parent = *p;
504 req = rb_entry(parent, struct ceph_osd_request, r_node);
505 if (new->r_tid < req->r_tid)
506 p = &(*p)->rb_left;
507 else if (new->r_tid > req->r_tid)
508 p = &(*p)->rb_right;
509 else
510 BUG();
511 }
512
513 rb_link_node(&new->r_node, parent, p);
514 rb_insert_color(&new->r_node, &osdc->requests);
515}
516
517static struct ceph_osd_request *__lookup_request(struct ceph_osd_client *osdc,
518 u64 tid)
519{
520 struct ceph_osd_request *req;
521 struct rb_node *n = osdc->requests.rb_node;
522
523 while (n) {
524 req = rb_entry(n, struct ceph_osd_request, r_node);
525 if (tid < req->r_tid)
526 n = n->rb_left;
527 else if (tid > req->r_tid)
528 n = n->rb_right;
529 else
530 return req;
531 }
532 return NULL;
533}
534
535static struct ceph_osd_request *
536__lookup_request_ge(struct ceph_osd_client *osdc,
537 u64 tid)
538{
539 struct ceph_osd_request *req;
540 struct rb_node *n = osdc->requests.rb_node;
541
542 while (n) {
543 req = rb_entry(n, struct ceph_osd_request, r_node);
544 if (tid < req->r_tid) {
545 if (!n->rb_left)
546 return req;
547 n = n->rb_left;
548 } else if (tid > req->r_tid) {
549 n = n->rb_right;
550 } else {
551 return req;
552 }
553 }
554 return NULL;
555}
556
6f6c7006
SW		 557/*
558 * Resubmit requests pending on the given osd.
559 */
560static void __kick_osd_requests(struct ceph_osd_client *osdc,
561 struct ceph_osd *osd)
562{
a40c4f10 563 struct ceph_osd_request *req, *nreq;
6f6c7006
SW		 564 int err;
565
566 dout("__kick_osd_requests osd%d\n", osd->o_osd);
567 err = __reset_osd(osdc, osd);
685a7555 568 if (err)
6f6c7006
SW		 569 return;
570
571 list_for_each_entry(req, &osd->o_requests, r_osd_item) {
572 list_move(&req->r_req_lru_item, &osdc->req_unsent);
573 dout("requeued %p tid %llu osd%d\n", req, req->r_tid,
574 osd->o_osd);
a40c4f10
YS		 575 if (!req->r_linger)
576 req->r_flags |= CEPH_OSD_FLAG_RETRY;
577 }
578
579 list_for_each_entry_safe(req, nreq, &osd->o_linger_requests,
580 r_linger_osd) {
77f38e0e
SW		 581 /*
582 * reregister request prior to unregistering linger so
583 * that r_osd is preserved.
584 */
585 BUG_ON(!list_empty(&req->r_req_lru_item));
a40c4f10 586 __register_request(osdc, req);
77f38e0e
SW		 587 list_add(&req->r_req_lru_item, &osdc->req_unsent);
588 list_add(&req->r_osd_item, &req->r_osd->o_requests);
589 __unregister_linger_request(osdc, req);
a40c4f10
YS		 590 dout("requeued lingering %p tid %llu osd%d\n", req, req->r_tid,
591 osd->o_osd);
6f6c7006
SW		 592 }
593}
594
f24e9980 595/*
81b024e7 596 * If the osd connection drops, we need to resubmit all requests.
f24e9980
SW		 597 */
598static void osd_reset(struct ceph_connection *con)
599{
600 struct ceph_osd *osd = con->private;
601 struct ceph_osd_client *osdc;
602
603 if (!osd)
604 return;
605 dout("osd_reset osd%d\n", osd->o_osd);
606 osdc = osd->o_osdc;
f24e9980 607 down_read(&osdc->map_sem);
83aff95e
SW		 608 mutex_lock(&osdc->request_mutex);
609 __kick_osd_requests(osdc, osd);
f9d25199 610 __send_queued(osdc);
83aff95e 611 mutex_unlock(&osdc->request_mutex);
f24e9980
SW		 612 up_read(&osdc->map_sem);
613}
614
615/*
616 * Track open sessions with osds.
617 */
e10006f8 618static struct ceph_osd *create_osd(struct ceph_osd_client *osdc, int onum)
f24e9980
SW		 619{
620 struct ceph_osd *osd;
621
622 osd = kzalloc(sizeof(*osd), GFP_NOFS);
623 if (!osd)
624 return NULL;
625
626 atomic_set(&osd->o_ref, 1);
627 osd->o_osdc = osdc;
e10006f8 628 osd->o_osd = onum;
f407731d 629 RB_CLEAR_NODE(&osd->o_node);
f24e9980 630 INIT_LIST_HEAD(&osd->o_requests);
a40c4f10 631 INIT_LIST_HEAD(&osd->o_linger_requests);
f5a2041b 632 INIT_LIST_HEAD(&osd->o_osd_lru);
f24e9980
SW		 633 osd->o_incarnation = 1;
634
b7a9e5dd 635 ceph_con_init(&osd->o_con, osd, &osd_con_ops, &osdc->client->msgr);
4e7a5dcd 636
422d2cb8 637 INIT_LIST_HEAD(&osd->o_keepalive_item);
f24e9980
SW		 638 return osd;
639}
640
641static struct ceph_osd *get_osd(struct ceph_osd *osd)
642{
643 if (atomic_inc_not_zero(&osd->o_ref)) {
644 dout("get_osd %p %d -> %d\n", osd, atomic_read(&osd->o_ref)-1,
645 atomic_read(&osd->o_ref));
646 return osd;
647 } else {
648 dout("get_osd %p FAIL\n", osd);
649 return NULL;
650 }
651}
652
653static void put_osd(struct ceph_osd *osd)
654{
655 dout("put_osd %p %d -> %d\n", osd, atomic_read(&osd->o_ref),
656 atomic_read(&osd->o_ref) - 1);
a255651d 657 if (atomic_dec_and_test(&osd->o_ref) && osd->o_auth.authorizer) {
79494d1b
SW		 658 struct ceph_auth_client *ac = osd->o_osdc->client->monc.auth;
659
a255651d 660 if (ac->ops && ac->ops->destroy_authorizer)
6c4a1915 661 ac->ops->destroy_authorizer(ac, osd->o_auth.authorizer);
f24e9980 662 kfree(osd);
79494d1b 663 }
f24e9980
SW		 664}
665
666/*
667 * remove an osd from our map
668 */
f5a2041b 669static void __remove_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd)
f24e9980 670{
f5a2041b 671 dout("__remove_osd %p\n", osd);
f24e9980
SW		 672 BUG_ON(!list_empty(&osd->o_requests));
673 rb_erase(&osd->o_node, &osdc->osds);
f5a2041b 674 list_del_init(&osd->o_osd_lru);
f24e9980
SW		 675 ceph_con_close(&osd->o_con);
676 put_osd(osd);
677}
678
aca420bc
SW		 679static void remove_all_osds(struct ceph_osd_client *osdc)
680{
048a9d2d 681 dout("%s %p\n", __func__, osdc);
aca420bc
SW		 682 mutex_lock(&osdc->request_mutex);
683 while (!RB_EMPTY_ROOT(&osdc->osds)) {
684 struct ceph_osd *osd = rb_entry(rb_first(&osdc->osds),
685 struct ceph_osd, o_node);
686 __remove_osd(osdc, osd);
687 }
688 mutex_unlock(&osdc->request_mutex);
689}
690
f5a2041b
YS		 691static void __move_osd_to_lru(struct ceph_osd_client *osdc,
692 struct ceph_osd *osd)
693{
694 dout("__move_osd_to_lru %p\n", osd);
695 BUG_ON(!list_empty(&osd->o_osd_lru));
696 list_add_tail(&osd->o_osd_lru, &osdc->osd_lru);
3d14c5d2 697 osd->lru_ttl = jiffies + osdc->client->options->osd_idle_ttl * HZ;
f5a2041b
YS		 698}
699
700static void __remove_osd_from_lru(struct ceph_osd *osd)
701{
702 dout("__remove_osd_from_lru %p\n", osd);
703 if (!list_empty(&osd->o_osd_lru))
704 list_del_init(&osd->o_osd_lru);
705}
706
aca420bc 707static void remove_old_osds(struct ceph_osd_client *osdc)
f5a2041b
YS		 708{
709 struct ceph_osd *osd, *nosd;
710
711 dout("__remove_old_osds %p\n", osdc);
712 mutex_lock(&osdc->request_mutex);
713 list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) {
aca420bc 714 if (time_before(jiffies, osd->lru_ttl))
f5a2041b
YS		 715 break;
716 __remove_osd(osdc, osd);
717 }
718 mutex_unlock(&osdc->request_mutex);
719}
720
f24e9980
SW		 721/*
722 * reset osd connect
723 */
f5a2041b 724static int __reset_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd)
f24e9980 725{
c3acb181 726 struct ceph_entity_addr *peer_addr;
f24e9980 727
f5a2041b 728 dout("__reset_osd %p osd%d\n", osd, osd->o_osd);
a40c4f10
YS		 729 if (list_empty(&osd->o_requests) &&
730 list_empty(&osd->o_linger_requests)) {
f5a2041b 731 __remove_osd(osdc, osd);
c3acb181
AE		 732
733 return -ENODEV;
734 }
735
736 peer_addr = &osdc->osdmap->osd_addr[osd->o_osd];
737 if (!memcmp(peer_addr, &osd->o_con.peer_addr, sizeof (*peer_addr)) &&
738 !ceph_con_opened(&osd->o_con)) {
739 struct ceph_osd_request *req;
740
87b315a5
SW		 741 dout(" osd addr hasn't changed and connection never opened,"
742 " letting msgr retry");
743 /* touch each r_stamp for handle_timeout()'s benfit */
744 list_for_each_entry(req, &osd->o_requests, r_osd_item)
745 req->r_stamp = jiffies;
c3acb181
AE		 746
747 return -EAGAIN;
f24e9980 748 }
c3acb181
AE		 749
750 ceph_con_close(&osd->o_con);
751 ceph_con_open(&osd->o_con, CEPH_ENTITY_TYPE_OSD, osd->o_osd, peer_addr);
752 osd->o_incarnation++;
753
754 return 0;
f24e9980
SW		 755}
756
757static void __insert_osd(struct ceph_osd_client *osdc, struct ceph_osd *new)
758{
759 struct rb_node **p = &osdc->osds.rb_node;
760 struct rb_node *parent = NULL;
761 struct ceph_osd *osd = NULL;
762
aca420bc 763 dout("__insert_osd %p osd%d\n", new, new->o_osd);
f24e9980
SW		 764 while (*p) {
765 parent = *p;
766 osd = rb_entry(parent, struct ceph_osd, o_node);
767 if (new->o_osd < osd->o_osd)
768 p = &(*p)->rb_left;
769 else if (new->o_osd > osd->o_osd)
770 p = &(*p)->rb_right;
771 else
772 BUG();
773 }
774
775 rb_link_node(&new->o_node, parent, p);
776 rb_insert_color(&new->o_node, &osdc->osds);
777}
778
779static struct ceph_osd *__lookup_osd(struct ceph_osd_client *osdc, int o)
780{
781 struct ceph_osd *osd;
782 struct rb_node *n = osdc->osds.rb_node;
783
784 while (n) {
785 osd = rb_entry(n, struct ceph_osd, o_node);
786 if (o < osd->o_osd)
787 n = n->rb_left;
788 else if (o > osd->o_osd)
789 n = n->rb_right;
790 else
791 return osd;
792 }
793 return NULL;
794}
795
422d2cb8
YS		 796static void __schedule_osd_timeout(struct ceph_osd_client *osdc)
797{
798 schedule_delayed_work(&osdc->timeout_work,
3d14c5d2 799 osdc->client->options->osd_keepalive_timeout * HZ);
422d2cb8
YS		 800}
801
802static void __cancel_osd_timeout(struct ceph_osd_client *osdc)
803{
804 cancel_delayed_work(&osdc->timeout_work);
805}
f24e9980
SW		 806
807/*
808 * Register request, assign tid. If this is the first request, set up
809 * the timeout event.
810 */
a40c4f10
YS		 811static void __register_request(struct ceph_osd_client *osdc,
812 struct ceph_osd_request *req)
f24e9980 813{
f24e9980 814 req->r_tid = ++osdc->last_tid;
6df058c0 815 req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
77f38e0e 816 dout("__register_request %p tid %lld\n", req, req->r_tid);
f24e9980
SW		 817 __insert_request(osdc, req);
818 ceph_osdc_get_request(req);
819 osdc->num_requests++;
f24e9980 820 if (osdc->num_requests == 1) {
422d2cb8
YS		 821 dout(" first request, scheduling timeout\n");
822 __schedule_osd_timeout(osdc);
f24e9980 823 }
a40c4f10
YS		 824}
825
826static void register_request(struct ceph_osd_client *osdc,
827 struct ceph_osd_request *req)
828{
829 mutex_lock(&osdc->request_mutex);
830 __register_request(osdc, req);
f24e9980
SW		 831 mutex_unlock(&osdc->request_mutex);
832}
833
834/*
835 * called under osdc->request_mutex
836 */
837static void __unregister_request(struct ceph_osd_client *osdc,
838 struct ceph_osd_request *req)
839{
35f9f8a0
SW		 840 if (RB_EMPTY_NODE(&req->r_node)) {
841 dout("__unregister_request %p tid %lld not registered\n",
842 req, req->r_tid);
843 return;
844 }
845
f24e9980
SW		 846 dout("__unregister_request %p tid %lld\n", req, req->r_tid);
847 rb_erase(&req->r_node, &osdc->requests);
848 osdc->num_requests--;
849
0ba6478d
SW		 850 if (req->r_osd) {
851 /* make sure the original request isn't in flight. */
6740a845 852 ceph_msg_revoke(req->r_request);
0ba6478d
SW		 853
854 list_del_init(&req->r_osd_item);
a40c4f10
YS		 855 if (list_empty(&req->r_osd->o_requests) &&
856 list_empty(&req->r_osd->o_linger_requests)) {
857 dout("moving osd to %p lru\n", req->r_osd);
f5a2041b 858 __move_osd_to_lru(osdc, req->r_osd);
a40c4f10 859 }
fbdb9190 860 if (list_empty(&req->r_linger_item))
a40c4f10 861 req->r_osd = NULL;
0ba6478d 862 }
f24e9980 863
7d5f2481 864 list_del_init(&req->r_req_lru_item);
f24e9980
SW		 865 ceph_osdc_put_request(req);
866
422d2cb8
YS		 867 if (osdc->num_requests == 0) {
868 dout(" no requests, canceling timeout\n");
869 __cancel_osd_timeout(osdc);
f24e9980
SW		 870 }
871}
872
873/*
874 * Cancel a previously queued request message
875 */
876static void __cancel_request(struct ceph_osd_request *req)
877{
6bc18876 878 if (req->r_sent && req->r_osd) {
6740a845 879 ceph_msg_revoke(req->r_request);
f24e9980
SW		 880 req->r_sent = 0;
881 }
882}
883
a40c4f10
YS		 884static void __register_linger_request(struct ceph_osd_client *osdc,
885 struct ceph_osd_request *req)
886{
887 dout("__register_linger_request %p\n", req);
888 list_add_tail(&req->r_linger_item, &osdc->req_linger);
6194ea89
SW		 889 if (req->r_osd)
890 list_add_tail(&req->r_linger_osd,
891 &req->r_osd->o_linger_requests);
a40c4f10
YS		 892}
893
894static void __unregister_linger_request(struct ceph_osd_client *osdc,
895 struct ceph_osd_request *req)
896{
897 dout("__unregister_linger_request %p\n", req);
61c74035 898 list_del_init(&req->r_linger_item);
a40c4f10 899 if (req->r_osd) {
a40c4f10
YS		 900 list_del_init(&req->r_linger_osd);
901
902 if (list_empty(&req->r_osd->o_requests) &&
903 list_empty(&req->r_osd->o_linger_requests)) {
904 dout("moving osd to %p lru\n", req->r_osd);
905 __move_osd_to_lru(osdc, req->r_osd);
906 }
fbdb9190
SW		 907 if (list_empty(&req->r_osd_item))
908 req->r_osd = NULL;
a40c4f10
YS		 909 }
910}
911
912void ceph_osdc_unregister_linger_request(struct ceph_osd_client *osdc,
913 struct ceph_osd_request *req)
914{
915 mutex_lock(&osdc->request_mutex);
916 if (req->r_linger) {
917 __unregister_linger_request(osdc, req);
918 ceph_osdc_put_request(req);
919 }
920 mutex_unlock(&osdc->request_mutex);
921}
922EXPORT_SYMBOL(ceph_osdc_unregister_linger_request);
923
924void ceph_osdc_set_request_linger(struct ceph_osd_client *osdc,
925 struct ceph_osd_request *req)
926{
927 if (!req->r_linger) {
928 dout("set_request_linger %p\n", req);
929 req->r_linger = 1;
930 /*
931 * caller is now responsible for calling
932 * unregister_linger_request
933 */
934 ceph_osdc_get_request(req);
935 }
936}
937EXPORT_SYMBOL(ceph_osdc_set_request_linger);
938
f24e9980
SW		 939/*
940 * Pick an osd (the first 'up' osd in the pg), allocate the osd struct
941 * (as needed), and set the request r_osd appropriately. If there is
25985edc 942 * no up osd, set r_osd to NULL. Move the request to the appropriate list
6f6c7006 943 * (unsent, homeless) or leave on in-flight lru.
f24e9980
SW		 944 *
945 * Return 0 if unchanged, 1 if changed, or negative on error.
946 *
947 * Caller should hold map_sem for read and request_mutex.
948 */
6f6c7006 949static int __map_request(struct ceph_osd_client *osdc,
38d6453c 950 struct ceph_osd_request *req, int force_resend)
f24e9980 951{
5b191d99 952 struct ceph_pg pgid;
d85b7056
SW		 953 int acting[CEPH_PG_MAX_SIZE];
954 int o = -1, num = 0;
f24e9980 955 int err;
f24e9980 956
6f6c7006 957 dout("map_request %p tid %lld\n", req, req->r_tid);
2169aea6 958 err = ceph_calc_object_layout(&pgid, req->r_oid,
f24e9980 959 &req->r_file_layout, osdc->osdmap);
6f6c7006
SW		 960 if (err) {
961 list_move(&req->r_req_lru_item, &osdc->req_notarget);
f24e9980 962 return err;
6f6c7006 963 }
7740a42f
SW		 964 req->r_pgid = pgid;
965
d85b7056
SW		 966 err = ceph_calc_pg_acting(osdc->osdmap, pgid, acting);
967 if (err > 0) {
968 o = acting[0];
969 num = err;
970 }
f24e9980 971
38d6453c
SW		 972 if ((!force_resend &&
973 req->r_osd && req->r_osd->o_osd == o &&
d85b7056
SW		 974 req->r_sent >= req->r_osd->o_incarnation &&
975 req->r_num_pg_osds == num &&
976 memcmp(req->r_pg_osds, acting, sizeof(acting[0])*num) == 0) ||
f24e9980
SW		 977 (req->r_osd == NULL && o == -1))
978 return 0; /* no change */
979
5b191d99
SW		 980 dout("map_request tid %llu pgid %lld.%x osd%d (was osd%d)\n",
981 req->r_tid, pgid.pool, pgid.seed, o,
f24e9980
SW		 982 req->r_osd ? req->r_osd->o_osd : -1);
983
d85b7056
SW		 984 /* record full pg acting set */
985 memcpy(req->r_pg_osds, acting, sizeof(acting[0]) * num);
986 req->r_num_pg_osds = num;
987
f24e9980
SW		 988 if (req->r_osd) {
989 __cancel_request(req);
990 list_del_init(&req->r_osd_item);
f24e9980
SW		 991 req->r_osd = NULL;
992 }
993
994 req->r_osd = __lookup_osd(osdc, o);
995 if (!req->r_osd && o >= 0) {
c99eb1c7 996 err = -ENOMEM;
e10006f8 997 req->r_osd = create_osd(osdc, o);
6f6c7006
SW		 998 if (!req->r_osd) {
999 list_move(&req->r_req_lru_item, &osdc->req_notarget);
c99eb1c7 1000 goto out;
6f6c7006 1001 }
f24e9980 1002
6f6c7006 1003 dout("map_request osd %p is osd%d\n", req->r_osd, o);
f24e9980
SW		 1004 __insert_osd(osdc, req->r_osd);
1005
b7a9e5dd
SW		 1006 ceph_con_open(&req->r_osd->o_con,
1007 CEPH_ENTITY_TYPE_OSD, o,
1008 &osdc->osdmap->osd_addr[o]);
f24e9980
SW		 1009 }
1010
f5a2041b
YS		 1011 if (req->r_osd) {
1012 __remove_osd_from_lru(req->r_osd);
f24e9980 1013 list_add(&req->r_osd_item, &req->r_osd->o_requests);
6f6c7006
SW		 1014 list_move(&req->r_req_lru_item, &osdc->req_unsent);
1015 } else {
1016 list_move(&req->r_req_lru_item, &osdc->req_notarget);
f5a2041b 1017 }
d85b7056 1018 err = 1; /* osd or pg changed */
f24e9980
SW		 1019
1020out:
f24e9980
SW		 1021 return err;
1022}
1023
1024/*
1025 * caller should hold map_sem (for read) and request_mutex
1026 */
56e925b6
SW		 1027static void __send_request(struct ceph_osd_client *osdc,
1028 struct ceph_osd_request *req)
f24e9980 1029{
1b83bef2 1030 void *p;
f24e9980 1031
1b83bef2
SW		 1032 dout("send_request %p tid %llu to osd%d flags %d pg %lld.%x\n",
1033 req, req->r_tid, req->r_osd->o_osd, req->r_flags,
1034 (unsigned long long)req->r_pgid.pool, req->r_pgid.seed);
1035
1036 /* fill in message content that changes each time we send it */
1037 put_unaligned_le32(osdc->osdmap->epoch, req->r_request_osdmap_epoch);
1038 put_unaligned_le32(req->r_flags, req->r_request_flags);
1039 put_unaligned_le64(req->r_pgid.pool, req->r_request_pool);
1040 p = req->r_request_pgid;
1041 ceph_encode_64(&p, req->r_pgid.pool);
1042 ceph_encode_32(&p, req->r_pgid.seed);
1043 put_unaligned_le64(1, req->r_request_attempts); /* FIXME */
1044 memcpy(req->r_request_reassert_version, &req->r_reassert_version,
1045 sizeof(req->r_reassert_version));
2169aea6 1046
3dd72fc0 1047 req->r_stamp = jiffies;
07a27e22 1048 list_move_tail(&req->r_req_lru_item, &osdc->req_lru);
f24e9980
SW		 1049
1050 ceph_msg_get(req->r_request); /* send consumes a ref */
1051 ceph_con_send(&req->r_osd->o_con, req->r_request);
1052 req->r_sent = req->r_osd->o_incarnation;
f24e9980
SW		 1053}
1054
6f6c7006
SW		 1055/*
1056 * Send any requests in the queue (req_unsent).
1057 */
f9d25199 1058static void __send_queued(struct ceph_osd_client *osdc)
6f6c7006
SW		 1059{
1060 struct ceph_osd_request *req, *tmp;
1061
f9d25199
AE		 1062 dout("__send_queued\n");
1063 list_for_each_entry_safe(req, tmp, &osdc->req_unsent, r_req_lru_item)
6f6c7006 1064 __send_request(osdc, req);
6f6c7006
SW		 1065}
1066
f24e9980
SW		 1067/*
1068 * Timeout callback, called every N seconds when 1 or more osd
1069 * requests has been active for more than N seconds. When this
1070 * happens, we ping all OSDs with requests who have timed out to
1071 * ensure any communications channel reset is detected. Reset the
1072 * request timeouts another N seconds in the future as we go.
1073 * Reschedule the timeout event another N seconds in future (unless
1074 * there are no open requests).
1075 */
1076static void handle_timeout(struct work_struct *work)
1077{
1078 struct ceph_osd_client *osdc =
1079 container_of(work, struct ceph_osd_client, timeout_work.work);
83aff95e 1080 struct ceph_osd_request *req;
f24e9980 1081 struct ceph_osd *osd;
422d2cb8 1082 unsigned long keepalive =
3d14c5d2 1083 osdc->client->options->osd_keepalive_timeout * HZ;
422d2cb8 1084 struct list_head slow_osds;
f24e9980
SW		 1085 dout("timeout\n");
1086 down_read(&osdc->map_sem);
1087
1088 ceph_monc_request_next_osdmap(&osdc->client->monc);
1089
1090 mutex_lock(&osdc->request_mutex);
f24e9980 1091
422d2cb8
YS		 1092 /*
1093 * ping osds that are a bit slow. this ensures that if there
1094 * is a break in the TCP connection we will notice, and reopen
1095 * a connection with that osd (from the fault callback).
1096 */
1097 INIT_LIST_HEAD(&slow_osds);
1098 list_for_each_entry(req, &osdc->req_lru, r_req_lru_item) {
3dd72fc0 1099 if (time_before(jiffies, req->r_stamp + keepalive))
422d2cb8
YS		 1100 break;
1101
1102 osd = req->r_osd;
1103 BUG_ON(!osd);
1104 dout(" tid %llu is slow, will send keepalive on osd%d\n",
f24e9980 1105 req->r_tid, osd->o_osd);
422d2cb8
YS		 1106 list_move_tail(&osd->o_keepalive_item, &slow_osds);
1107 }
1108 while (!list_empty(&slow_osds)) {
1109 osd = list_entry(slow_osds.next, struct ceph_osd,
1110 o_keepalive_item);
1111 list_del_init(&osd->o_keepalive_item);
f24e9980
SW		 1112 ceph_con_keepalive(&osd->o_con);
1113 }
1114
422d2cb8 1115 __schedule_osd_timeout(osdc);
f9d25199 1116 __send_queued(osdc);
f24e9980 1117 mutex_unlock(&osdc->request_mutex);
f24e9980
SW		 1118 up_read(&osdc->map_sem);
1119}
1120
f5a2041b
YS		 1121static void handle_osds_timeout(struct work_struct *work)
1122{
1123 struct ceph_osd_client *osdc =
1124 container_of(work, struct ceph_osd_client,
1125 osds_timeout_work.work);
1126 unsigned long delay =
3d14c5d2 1127 osdc->client->options->osd_idle_ttl * HZ >> 2;
f5a2041b
YS		 1128
1129 dout("osds timeout\n");
1130 down_read(&osdc->map_sem);
aca420bc 1131 remove_old_osds(osdc);
f5a2041b
YS		 1132 up_read(&osdc->map_sem);
1133
1134 schedule_delayed_work(&osdc->osds_timeout_work,
1135 round_jiffies_relative(delay));
1136}
1137
25845472
SW		 1138static void complete_request(struct ceph_osd_request *req)
1139{
1140 if (req->r_safe_callback)
1141 req->r_safe_callback(req, NULL);
1142 complete_all(&req->r_safe_completion); /* fsync waiter */
1143}
1144
1b83bef2
SW		 1145static int __decode_pgid(void **p, void *end, struct ceph_pg *pgid)
1146{
1147 __u8 v;
1148
1149 ceph_decode_need(p, end, 1 + 8 + 4 + 4, bad);
1150 v = ceph_decode_8(p);
1151 if (v > 1) {
1152 pr_warning("do not understand pg encoding %d > 1", v);
1153 return -EINVAL;
1154 }
1155 pgid->pool = ceph_decode_64(p);
1156 pgid->seed = ceph_decode_32(p);
1157 *p += 4;
1158 return 0;
1159
1160bad:
1161 pr_warning("incomplete pg encoding");
1162 return -EINVAL;
1163}
1164
f24e9980
SW		 1165/*
1166 * handle osd op reply. either call the callback if it is specified,
1167 * or do the completion to wake up the waiting thread.
1168 */
350b1c32
SW		 1169static void handle_reply(struct ceph_osd_client *osdc, struct ceph_msg *msg,
1170 struct ceph_connection *con)
f24e9980 1171{
1b83bef2 1172 void *p, *end;
f24e9980
SW		 1173 struct ceph_osd_request *req;
1174 u64 tid;
1b83bef2
SW		 1175 int object_len;
1176 int numops, payload_len, flags;
0ceed5db 1177 s32 result;
1b83bef2
SW		 1178 s32 retry_attempt;
1179 struct ceph_pg pg;
1180 int err;
1181 u32 reassert_epoch;
1182 u64 reassert_version;
1183 u32 osdmap_epoch;
1184 int i;
f24e9980 1185
6df058c0 1186 tid = le64_to_cpu(msg->hdr.tid);
1b83bef2
SW		 1187 dout("handle_reply %p tid %llu\n", msg, tid);
1188
1189 p = msg->front.iov_base;
1190 end = p + msg->front.iov_len;
1191
1192 ceph_decode_need(&p, end, 4, bad);
1193 object_len = ceph_decode_32(&p);
1194 ceph_decode_need(&p, end, object_len, bad);
1195 p += object_len;
1196
1197 err = __decode_pgid(&p, end, &pg);
1198 if (err)
f24e9980 1199 goto bad;
1b83bef2
SW		 1200
1201 ceph_decode_need(&p, end, 8 + 4 + 4 + 8 + 4, bad);
1202 flags = ceph_decode_64(&p);
1203 result = ceph_decode_32(&p);
1204 reassert_epoch = ceph_decode_32(&p);
1205 reassert_version = ceph_decode_64(&p);
1206 osdmap_epoch = ceph_decode_32(&p);
1207
f24e9980
SW		 1208 /* lookup */
1209 mutex_lock(&osdc->request_mutex);
1210 req = __lookup_request(osdc, tid);
1211 if (req == NULL) {
1212 dout("handle_reply tid %llu dne\n", tid);
1213 mutex_unlock(&osdc->request_mutex);
1214 return;
1215 }
1216 ceph_osdc_get_request(req);
1b83bef2
SW		 1217
1218 dout("handle_reply %p tid %llu req %p result %d\n", msg, tid,
1219 req, result);
1220
1221 ceph_decode_need(&p, end, 4, bad);
1222 numops = ceph_decode_32(&p);
1223 if (numops > CEPH_OSD_MAX_OP)
1224 goto bad_put;
1225 if (numops != req->r_num_ops)
1226 goto bad_put;
1227 payload_len = 0;
1228 ceph_decode_need(&p, end, numops * sizeof(struct ceph_osd_op), bad);
1229 for (i = 0; i < numops; i++) {
1230 struct ceph_osd_op *op = p;
1231 int len;
1232
1233 len = le32_to_cpu(op->payload_len);
1234 req->r_reply_op_len[i] = len;
1235 dout(" op %d has %d bytes\n", i, len);
1236 payload_len += len;
1237 p += sizeof(*op);
1238 }
1239 if (payload_len != le32_to_cpu(msg->hdr.data_len)) {
1240 pr_warning("sum of op payload lens %d != data_len %d",
1241 payload_len, le32_to_cpu(msg->hdr.data_len));
1242 goto bad_put;
1243 }
1244
1245 ceph_decode_need(&p, end, 4 + numops * 4, bad);
1246 retry_attempt = ceph_decode_32(&p);
1247 for (i = 0; i < numops; i++)
1248 req->r_reply_op_result[i] = ceph_decode_32(&p);
f24e9980 1249
350b1c32 1250 /*
0d59ab81 1251 * if this connection filled our message, drop our reference now, to
350b1c32
SW		 1252 * avoid a (safe but slower) revoke later.
1253 */
0d59ab81 1254 if (req->r_con_filling_msg == con && req->r_reply == msg) {
c16e7869 1255 dout(" dropping con_filling_msg ref %p\n", con);
0d59ab81 1256 req->r_con_filling_msg = NULL;
0d47766f 1257 con->ops->put(con);
350b1c32
SW		 1258 }
1259
f24e9980 1260 if (!req->r_got_reply) {
95c96174 1261 unsigned int bytes;
f24e9980 1262
1b83bef2 1263 req->r_result = result;
f24e9980
SW		 1264 bytes = le32_to_cpu(msg->hdr.data_len);
1265 dout("handle_reply result %d bytes %d\n", req->r_result,
1266 bytes);
1267 if (req->r_result == 0)
1268 req->r_result = bytes;
1269
1270 /* in case this is a write and we need to replay, */
1b83bef2
SW		 1271 req->r_reassert_version.epoch = cpu_to_le32(reassert_epoch);
1272 req->r_reassert_version.version = cpu_to_le64(reassert_version);
f24e9980
SW		 1273
1274 req->r_got_reply = 1;
1275 } else if ((flags & CEPH_OSD_FLAG_ONDISK) == 0) {
1276 dout("handle_reply tid %llu dup ack\n", tid);
34b43a56 1277 mutex_unlock(&osdc->request_mutex);
f24e9980
SW		 1278 goto done;
1279 }
1280
1281 dout("handle_reply tid %llu flags %d\n", tid, flags);
1282
a40c4f10
YS		 1283 if (req->r_linger && (flags & CEPH_OSD_FLAG_ONDISK))
1284 __register_linger_request(osdc, req);
1285
f24e9980 1286 /* either this is a read, or we got the safe response */
0ceed5db
SW		 1287 if (result < 0 ||
1288 (flags & CEPH_OSD_FLAG_ONDISK) ||
f24e9980
SW		 1289 ((flags & CEPH_OSD_FLAG_WRITE) == 0))
1290 __unregister_request(osdc, req);
1291
1292 mutex_unlock(&osdc->request_mutex);
1293
1294 if (req->r_callback)
1295 req->r_callback(req, msg);
1296 else
03066f23 1297 complete_all(&req->r_completion);
f24e9980 1298
25845472
SW		 1299 if (flags & CEPH_OSD_FLAG_ONDISK)
1300 complete_request(req);
f24e9980
SW		 1301
1302done:
a40c4f10 1303 dout("req=%p req->r_linger=%d\n", req, req->r_linger);
f24e9980
SW		 1304 ceph_osdc_put_request(req);
1305 return;
1306
1b83bef2
SW		 1307bad_put:
1308 ceph_osdc_put_request(req);
f24e9980 1309bad:
1b83bef2
SW		 1310 pr_err("corrupt osd_op_reply got %d %d\n",
1311 (int)msg->front.iov_len, le32_to_cpu(msg->hdr.front_len));
9ec7cab1 1312 ceph_msg_dump(msg);
f24e9980
SW		 1313}
1314
6f6c7006 1315static void reset_changed_osds(struct ceph_osd_client *osdc)
f24e9980 1316{
f24e9980 1317 struct rb_node *p, *n;
f24e9980 1318
6f6c7006
SW		 1319 for (p = rb_first(&osdc->osds); p; p = n) {
1320 struct ceph_osd *osd = rb_entry(p, struct ceph_osd, o_node);
f24e9980 1321
6f6c7006
SW		 1322 n = rb_next(p);
1323 if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) ||
1324 memcmp(&osd->o_con.peer_addr,
1325 ceph_osd_addr(osdc->osdmap,
1326 osd->o_osd),
1327 sizeof(struct ceph_entity_addr)) != 0)
1328 __reset_osd(osdc, osd);
f24e9980 1329 }
422d2cb8
YS		 1330}
1331
1332/*
6f6c7006
SW		 1333 * Requeue requests whose mapping to an OSD has changed. If requests map to
1334 * no osd, request a new map.
422d2cb8 1335 *
e6d50f67 1336 * Caller should hold map_sem for read.
422d2cb8 1337 */
38d6453c 1338static void kick_requests(struct ceph_osd_client *osdc, int force_resend)
422d2cb8 1339{
a40c4f10 1340 struct ceph_osd_request *req, *nreq;
6f6c7006
SW		 1341 struct rb_node *p;
1342 int needmap = 0;
1343 int err;
422d2cb8 1344
38d6453c 1345 dout("kick_requests %s\n", force_resend ? " (force resend)" : "");
422d2cb8 1346 mutex_lock(&osdc->request_mutex);
6194ea89 1347 for (p = rb_first(&osdc->requests); p; ) {
6f6c7006 1348 req = rb_entry(p, struct ceph_osd_request, r_node);
6194ea89 1349 p = rb_next(p);
ab60b16d
AE		 1350
1351 /*
1352 * For linger requests that have not yet been
1353 * registered, move them to the linger list; they'll
1354 * be sent to the osd in the loop below. Unregister
1355 * the request before re-registering it as a linger
1356 * request to ensure the __map_request() below
1357 * will decide it needs to be sent.
1358 */
1359 if (req->r_linger && list_empty(&req->r_linger_item)) {
1360 dout("%p tid %llu restart on osd%d\n",
1361 req, req->r_tid,
1362 req->r_osd ? req->r_osd->o_osd : -1);
1363 __unregister_request(osdc, req);
1364 __register_linger_request(osdc, req);
1365 continue;
1366 }
1367
38d6453c 1368 err = __map_request(osdc, req, force_resend);
6f6c7006
SW		 1369 if (err < 0)
1370 continue; /* error */
1371 if (req->r_osd == NULL) {
1372 dout("%p tid %llu maps to no osd\n", req, req->r_tid);
1373 needmap++; /* request a newer map */
1374 } else if (err > 0) {
6194ea89
SW		 1375 if (!req->r_linger) {
1376 dout("%p tid %llu requeued on osd%d\n", req,
1377 req->r_tid,
1378 req->r_osd ? req->r_osd->o_osd : -1);
a40c4f10 1379 req->r_flags |= CEPH_OSD_FLAG_RETRY;
6194ea89
SW		 1380 }
1381 }
a40c4f10
YS		 1382 }
1383
1384 list_for_each_entry_safe(req, nreq, &osdc->req_linger,
1385 r_linger_item) {
1386 dout("linger req=%p req->r_osd=%p\n", req, req->r_osd);
1387
38d6453c 1388 err = __map_request(osdc, req, force_resend);
ab60b16d 1389 dout("__map_request returned %d\n", err);
a40c4f10
YS		 1390 if (err == 0)
1391 continue; /* no change and no osd was specified */
1392 if (err < 0)
1393 continue; /* hrm! */
1394 if (req->r_osd == NULL) {
1395 dout("tid %llu maps to no valid osd\n", req->r_tid);
1396 needmap++; /* request a newer map */
1397 continue;
6f6c7006 1398 }
a40c4f10
YS		 1399
1400 dout("kicking lingering %p tid %llu osd%d\n", req, req->r_tid,
1401 req->r_osd ? req->r_osd->o_osd : -1);
a40c4f10 1402 __register_request(osdc, req);
c89ce05e 1403 __unregister_linger_request(osdc, req);
6f6c7006 1404 }
f24e9980
SW		 1405 mutex_unlock(&osdc->request_mutex);
1406
1407 if (needmap) {
1408 dout("%d requests for down osds, need new map\n", needmap);
1409 ceph_monc_request_next_osdmap(&osdc->client->monc);
1410 }
e6d50f67 1411 reset_changed_osds(osdc);
422d2cb8 1412}
6f6c7006
SW		 1413
1414
f24e9980
SW		 1415/*
1416 * Process updated osd map.
1417 *
1418 * The message contains any number of incremental and full maps, normally
1419 * indicating some sort of topology change in the cluster. Kick requests
1420 * off to different OSDs as needed.
1421 */
1422void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
1423{
1424 void *p, *end, *next;
1425 u32 nr_maps, maplen;
1426 u32 epoch;
1427 struct ceph_osdmap *newmap = NULL, *oldmap;
1428 int err;
1429 struct ceph_fsid fsid;
1430
1431 dout("handle_map have %u\n", osdc->osdmap ? osdc->osdmap->epoch : 0);
1432 p = msg->front.iov_base;
1433 end = p + msg->front.iov_len;
1434
1435 /* verify fsid */
1436 ceph_decode_need(&p, end, sizeof(fsid), bad);
1437 ceph_decode_copy(&p, &fsid, sizeof(fsid));
0743304d
SW		 1438 if (ceph_check_fsid(osdc->client, &fsid) < 0)
1439 return;
f24e9980
SW		 1440
1441 down_write(&osdc->map_sem);
1442
1443 /* incremental maps */
1444 ceph_decode_32_safe(&p, end, nr_maps, bad);
1445 dout(" %d inc maps\n", nr_maps);
1446 while (nr_maps > 0) {
1447 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
c89136ea
SW		 1448 epoch = ceph_decode_32(&p);
1449 maplen = ceph_decode_32(&p);
f24e9980
SW		 1450 ceph_decode_need(&p, end, maplen, bad);
1451 next = p + maplen;
1452 if (osdc->osdmap && osdc->osdmap->epoch+1 == epoch) {
1453 dout("applying incremental map %u len %d\n",
1454 epoch, maplen);
1455 newmap = osdmap_apply_incremental(&p, next,
1456 osdc->osdmap,
15d9882c 1457 &osdc->client->msgr);
f24e9980
SW		 1458 if (IS_ERR(newmap)) {
1459 err = PTR_ERR(newmap);
1460 goto bad;
1461 }
30dc6381 1462 BUG_ON(!newmap);
f24e9980
SW		 1463 if (newmap != osdc->osdmap) {
1464 ceph_osdmap_destroy(osdc->osdmap);
1465 osdc->osdmap = newmap;
1466 }
38d6453c 1467 kick_requests(osdc, 0);
f24e9980
SW		 1468 } else {
1469 dout("ignoring incremental map %u len %d\n",
1470 epoch, maplen);
1471 }
1472 p = next;
1473 nr_maps--;
1474 }
1475 if (newmap)
1476 goto done;
1477
1478 /* full maps */
1479 ceph_decode_32_safe(&p, end, nr_maps, bad);
1480 dout(" %d full maps\n", nr_maps);
1481 while (nr_maps) {
1482 ceph_decode_need(&p, end, 2*sizeof(u32), bad);
c89136ea
SW		 1483 epoch = ceph_decode_32(&p);
1484 maplen = ceph_decode_32(&p);
f24e9980
SW		 1485 ceph_decode_need(&p, end, maplen, bad);
1486 if (nr_maps > 1) {
1487 dout("skipping non-latest full map %u len %d\n",
1488 epoch, maplen);
1489 } else if (osdc->osdmap && osdc->osdmap->epoch >= epoch) {
1490 dout("skipping full map %u len %d, "
1491 "older than our %u\n", epoch, maplen,
1492 osdc->osdmap->epoch);
1493 } else {
38d6453c
SW		 1494 int skipped_map = 0;
1495
f24e9980
SW		 1496 dout("taking full map %u len %d\n", epoch, maplen);
1497 newmap = osdmap_decode(&p, p+maplen);
1498 if (IS_ERR(newmap)) {
1499 err = PTR_ERR(newmap);
1500 goto bad;
1501 }
30dc6381 1502 BUG_ON(!newmap);
f24e9980
SW		 1503 oldmap = osdc->osdmap;
1504 osdc->osdmap = newmap;
38d6453c
SW		 1505 if (oldmap) {
1506 if (oldmap->epoch + 1 < newmap->epoch)
1507 skipped_map = 1;
f24e9980 1508 ceph_osdmap_destroy(oldmap);
38d6453c
SW		 1509 }
1510 kick_requests(osdc, skipped_map);
f24e9980
SW		 1511 }
1512 p += maplen;
1513 nr_maps--;
1514 }
1515
1516done:
1517 downgrade_write(&osdc->map_sem);
1518 ceph_monc_got_osdmap(&osdc->client->monc, osdc->osdmap->epoch);
cd634fb6
SW		 1519
1520 /*
1521 * subscribe to subsequent osdmap updates if full to ensure
1522 * we find out when we are no longer full and stop returning
1523 * ENOSPC.
1524 */
1525 if (ceph_osdmap_flag(osdc->osdmap, CEPH_OSDMAP_FULL))
1526 ceph_monc_request_next_osdmap(&osdc->client->monc);
1527
f9d25199
AE		 1528 mutex_lock(&osdc->request_mutex);
1529 __send_queued(osdc);
1530 mutex_unlock(&osdc->request_mutex);
f24e9980 1531 up_read(&osdc->map_sem);
03066f23 1532 wake_up_all(&osdc->client->auth_wq);
f24e9980
SW		 1533 return;
1534
1535bad:
1536 pr_err("osdc handle_map corrupt msg\n");
9ec7cab1 1537 ceph_msg_dump(msg);
f24e9980
SW		 1538 up_write(&osdc->map_sem);
1539 return;
1540}
1541
a40c4f10
YS		 1542/*
1543 * watch/notify callback event infrastructure
1544 *
1545 * These callbacks are used both for watch and notify operations.
1546 */
1547static void __release_event(struct kref *kref)
1548{
1549 struct ceph_osd_event *event =
1550 container_of(kref, struct ceph_osd_event, kref);
1551
1552 dout("__release_event %p\n", event);
1553 kfree(event);
1554}
1555
1556static void get_event(struct ceph_osd_event *event)
1557{
1558 kref_get(&event->kref);
1559}
1560
1561void ceph_osdc_put_event(struct ceph_osd_event *event)
1562{
1563 kref_put(&event->kref, __release_event);
1564}
1565EXPORT_SYMBOL(ceph_osdc_put_event);
1566
1567static void __insert_event(struct ceph_osd_client *osdc,
1568 struct ceph_osd_event *new)
1569{
1570 struct rb_node **p = &osdc->event_tree.rb_node;
1571 struct rb_node *parent = NULL;
1572 struct ceph_osd_event *event = NULL;
1573
1574 while (*p) {
1575 parent = *p;
1576 event = rb_entry(parent, struct ceph_osd_event, node);
1577 if (new->cookie < event->cookie)
1578 p = &(*p)->rb_left;
1579 else if (new->cookie > event->cookie)
1580 p = &(*p)->rb_right;
1581 else
1582 BUG();
1583 }
1584
1585 rb_link_node(&new->node, parent, p);
1586 rb_insert_color(&new->node, &osdc->event_tree);
1587}
1588
1589static struct ceph_osd_event *__find_event(struct ceph_osd_client *osdc,
1590 u64 cookie)
1591{
1592 struct rb_node **p = &osdc->event_tree.rb_node;
1593 struct rb_node *parent = NULL;
1594 struct ceph_osd_event *event = NULL;
1595
1596 while (*p) {
1597 parent = *p;
1598 event = rb_entry(parent, struct ceph_osd_event, node);
1599 if (cookie < event->cookie)
1600 p = &(*p)->rb_left;
1601 else if (cookie > event->cookie)
1602 p = &(*p)->rb_right;
1603 else
1604 return event;
1605 }
1606 return NULL;
1607}
1608
1609static void __remove_event(struct ceph_osd_event *event)
1610{
1611 struct ceph_osd_client *osdc = event->osdc;
1612
1613 if (!RB_EMPTY_NODE(&event->node)) {
1614 dout("__remove_event removed %p\n", event);
1615 rb_erase(&event->node, &osdc->event_tree);
1616 ceph_osdc_put_event(event);
1617 } else {
1618 dout("__remove_event didn't remove %p\n", event);
1619 }
1620}
1621
1622int ceph_osdc_create_event(struct ceph_osd_client *osdc,
1623 void (*event_cb)(u64, u64, u8, void *),
3c663bbd 1624 void *data, struct ceph_osd_event **pevent)
a40c4f10
YS		 1625{
1626 struct ceph_osd_event *event;
1627
1628 event = kmalloc(sizeof(*event), GFP_NOIO);
1629 if (!event)
1630 return -ENOMEM;
1631
1632 dout("create_event %p\n", event);
1633 event->cb = event_cb;
3c663bbd 1634 event->one_shot = 0;
a40c4f10
YS		 1635 event->data = data;
1636 event->osdc = osdc;
1637 INIT_LIST_HEAD(&event->osd_node);
3ee5234d 1638 RB_CLEAR_NODE(&event->node);
a40c4f10
YS		 1639 kref_init(&event->kref); /* one ref for us */
1640 kref_get(&event->kref); /* one ref for the caller */
a40c4f10
YS		 1641
1642 spin_lock(&osdc->event_lock);
1643 event->cookie = ++osdc->event_count;
1644 __insert_event(osdc, event);
1645 spin_unlock(&osdc->event_lock);
1646
1647 *pevent = event;
1648 return 0;
1649}
1650EXPORT_SYMBOL(ceph_osdc_create_event);
1651
1652void ceph_osdc_cancel_event(struct ceph_osd_event *event)
1653{
1654 struct ceph_osd_client *osdc = event->osdc;
1655
1656 dout("cancel_event %p\n", event);
1657 spin_lock(&osdc->event_lock);
1658 __remove_event(event);
1659 spin_unlock(&osdc->event_lock);
1660 ceph_osdc_put_event(event); /* caller's */
1661}
1662EXPORT_SYMBOL(ceph_osdc_cancel_event);
1663
1664
1665static void do_event_work(struct work_struct *work)
1666{
1667 struct ceph_osd_event_work *event_work =
1668 container_of(work, struct ceph_osd_event_work, work);
1669 struct ceph_osd_event *event = event_work->event;
1670 u64 ver = event_work->ver;
1671 u64 notify_id = event_work->notify_id;
1672 u8 opcode = event_work->opcode;
1673
1674 dout("do_event_work completing %p\n", event);
1675 event->cb(ver, notify_id, opcode, event->data);
a40c4f10
YS		 1676 dout("do_event_work completed %p\n", event);
1677 ceph_osdc_put_event(event);
1678 kfree(event_work);
1679}
1680
1681
1682/*
1683 * Process osd watch notifications
1684 */
3c663bbd
AE		 1685static void handle_watch_notify(struct ceph_osd_client *osdc,
1686 struct ceph_msg *msg)
a40c4f10
YS		 1687{
1688 void *p, *end;
1689 u8 proto_ver;
1690 u64 cookie, ver, notify_id;
1691 u8 opcode;
1692 struct ceph_osd_event *event;
1693 struct ceph_osd_event_work *event_work;
1694
1695 p = msg->front.iov_base;
1696 end = p + msg->front.iov_len;
1697
1698 ceph_decode_8_safe(&p, end, proto_ver, bad);
1699 ceph_decode_8_safe(&p, end, opcode, bad);
1700 ceph_decode_64_safe(&p, end, cookie, bad);
1701 ceph_decode_64_safe(&p, end, ver, bad);
1702 ceph_decode_64_safe(&p, end, notify_id, bad);
1703
1704 spin_lock(&osdc->event_lock);
1705 event = __find_event(osdc, cookie);
1706 if (event) {
3c663bbd 1707 BUG_ON(event->one_shot);
a40c4f10 1708 get_event(event);
a40c4f10
YS		 1709 }
1710 spin_unlock(&osdc->event_lock);
1711 dout("handle_watch_notify cookie %lld ver %lld event %p\n",
1712 cookie, ver, event);
1713 if (event) {
1714 event_work = kmalloc(sizeof(*event_work), GFP_NOIO);
a40c4f10
YS		 1715 if (!event_work) {
1716 dout("ERROR: could not allocate event_work\n");
1717 goto done_err;
1718 }
6b0ae409 1719 INIT_WORK(&event_work->work, do_event_work);
a40c4f10
YS		 1720 event_work->event = event;
1721 event_work->ver = ver;
1722 event_work->notify_id = notify_id;
1723 event_work->opcode = opcode;
1724 if (!queue_work(osdc->notify_wq, &event_work->work)) {
1725 dout("WARNING: failed to queue notify event work\n");
1726 goto done_err;
1727 }
1728 }
1729
1730 return;
1731
1732done_err:
a40c4f10
YS		 1733 ceph_osdc_put_event(event);
1734 return;
1735
1736bad:
1737 pr_err("osdc handle_watch_notify corrupt msg\n");
1738 return;
1739}
1740
f24e9980
SW		 1741/*
1742 * Register request, send initial attempt.
1743 */
1744int ceph_osdc_start_request(struct ceph_osd_client *osdc,
1745 struct ceph_osd_request *req,
1746 bool nofail)
1747{
c1ea8823 1748 int rc = 0;
f24e9980
SW		 1749
1750 req->r_request->pages = req->r_pages;
1751 req->r_request->nr_pages = req->r_num_pages;
68b4476b
YS		 1752#ifdef CONFIG_BLOCK
1753 req->r_request->bio = req->r_bio;
1754#endif
c885837f 1755 req->r_request->trail = &req->r_trail;
f24e9980
SW		 1756
1757 register_request(osdc, req);
1758
1759 down_read(&osdc->map_sem);
1760 mutex_lock(&osdc->request_mutex);
c1ea8823
SW		 1761 /*
1762 * a racing kick_requests() may have sent the message for us
1763 * while we dropped request_mutex above, so only send now if
1764 * the request still han't been touched yet.
1765 */
1766 if (req->r_sent == 0) {
38d6453c 1767 rc = __map_request(osdc, req, 0);
9d6fcb08
SW		 1768 if (rc < 0) {
1769 if (nofail) {
1770 dout("osdc_start_request failed map, "
1771 " will retry %lld\n", req->r_tid);
1772 rc = 0;
1773 }
234af26f 1774 goto out_unlock;
9d6fcb08 1775 }
6f6c7006
SW		 1776 if (req->r_osd == NULL) {
1777 dout("send_request %p no up osds in pg\n", req);
1778 ceph_monc_request_next_osdmap(&osdc->client->monc);
1779 } else {
56e925b6 1780 __send_request(osdc, req);
f24e9980 1781 }
56e925b6 1782 rc = 0;
f24e9980 1783 }
234af26f
DC		 1784
1785out_unlock:
f24e9980
SW		 1786 mutex_unlock(&osdc->request_mutex);
1787 up_read(&osdc->map_sem);
1788 return rc;
1789}
3d14c5d2 1790EXPORT_SYMBOL(ceph_osdc_start_request);
f24e9980
SW		 1791
1792/*
1793 * wait for a request to complete
1794 */
1795int ceph_osdc_wait_request(struct ceph_osd_client *osdc,
1796 struct ceph_osd_request *req)
1797{
1798 int rc;
1799
1800 rc = wait_for_completion_interruptible(&req->r_completion);
1801 if (rc < 0) {
1802 mutex_lock(&osdc->request_mutex);
1803 __cancel_request(req);
529cfcc4 1804 __unregister_request(osdc, req);
f24e9980 1805 mutex_unlock(&osdc->request_mutex);
25845472 1806 complete_request(req);
529cfcc4 1807 dout("wait_request tid %llu canceled/timed out\n", req->r_tid);
f24e9980
SW		 1808 return rc;
1809 }
1810
1811 dout("wait_request tid %llu result %d\n", req->r_tid, req->r_result);
1812 return req->r_result;
1813}
3d14c5d2 1814EXPORT_SYMBOL(ceph_osdc_wait_request);
f24e9980
SW		 1815
1816/*
1817 * sync - wait for all in-flight requests to flush. avoid starvation.
1818 */
1819void ceph_osdc_sync(struct ceph_osd_client *osdc)
1820{
1821 struct ceph_osd_request *req;
1822 u64 last_tid, next_tid = 0;
1823
1824 mutex_lock(&osdc->request_mutex);
1825 last_tid = osdc->last_tid;
1826 while (1) {
1827 req = __lookup_request_ge(osdc, next_tid);
1828 if (!req)
1829 break;
1830 if (req->r_tid > last_tid)
1831 break;
1832
1833 next_tid = req->r_tid + 1;
1834 if ((req->r_flags & CEPH_OSD_FLAG_WRITE) == 0)
1835 continue;
1836
1837 ceph_osdc_get_request(req);
1838 mutex_unlock(&osdc->request_mutex);
1839 dout("sync waiting on tid %llu (last is %llu)\n",
1840 req->r_tid, last_tid);
1841 wait_for_completion(&req->r_safe_completion);
1842 mutex_lock(&osdc->request_mutex);
1843 ceph_osdc_put_request(req);
1844 }
1845 mutex_unlock(&osdc->request_mutex);
1846 dout("sync done (thru tid %llu)\n", last_tid);
1847}
3d14c5d2 1848EXPORT_SYMBOL(ceph_osdc_sync);
f24e9980
SW		 1849
1850/*
1851 * init, shutdown
1852 */
1853int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client)
1854{
1855 int err;
1856
1857 dout("init\n");
1858 osdc->client = client;
1859 osdc->osdmap = NULL;
1860 init_rwsem(&osdc->map_sem);
1861 init_completion(&osdc->map_waiters);
1862 osdc->last_requested_map = 0;
1863 mutex_init(&osdc->request_mutex);
f24e9980
SW		 1864 osdc->last_tid = 0;
1865 osdc->osds = RB_ROOT;
f5a2041b 1866 INIT_LIST_HEAD(&osdc->osd_lru);
f24e9980 1867 osdc->requests = RB_ROOT;
422d2cb8 1868 INIT_LIST_HEAD(&osdc->req_lru);
6f6c7006
SW		 1869 INIT_LIST_HEAD(&osdc->req_unsent);
1870 INIT_LIST_HEAD(&osdc->req_notarget);
a40c4f10 1871 INIT_LIST_HEAD(&osdc->req_linger);
f24e9980
SW		 1872 osdc->num_requests = 0;
1873 INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout);
f5a2041b 1874 INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout);
a40c4f10
YS		 1875 spin_lock_init(&osdc->event_lock);
1876 osdc->event_tree = RB_ROOT;
1877 osdc->event_count = 0;
f5a2041b
YS		 1878
1879 schedule_delayed_work(&osdc->osds_timeout_work,
3d14c5d2 1880 round_jiffies_relative(osdc->client->options->osd_idle_ttl * HZ));
f24e9980 1881
5f44f142 1882 err = -ENOMEM;
f24e9980
SW		 1883 osdc->req_mempool = mempool_create_kmalloc_pool(10,
1884 sizeof(struct ceph_osd_request));
1885 if (!osdc->req_mempool)
5f44f142 1886 goto out;
f24e9980 1887
d50b409f
SW		 1888 err = ceph_msgpool_init(&osdc->msgpool_op, CEPH_MSG_OSD_OP,
1889 OSD_OP_FRONT_LEN, 10, true,
4f48280e 1890 "osd_op");
f24e9980 1891 if (err < 0)
5f44f142 1892 goto out_mempool;
d50b409f 1893 err = ceph_msgpool_init(&osdc->msgpool_op_reply, CEPH_MSG_OSD_OPREPLY,
4f48280e
SW		 1894 OSD_OPREPLY_FRONT_LEN, 10, true,
1895 "osd_op_reply");
c16e7869
SW		 1896 if (err < 0)
1897 goto out_msgpool;
a40c4f10
YS		 1898
1899 osdc->notify_wq = create_singlethread_workqueue("ceph-watch-notify");
1900 if (IS_ERR(osdc->notify_wq)) {
1901 err = PTR_ERR(osdc->notify_wq);
1902 osdc->notify_wq = NULL;
1903 goto out_msgpool;
1904 }
f24e9980 1905 return 0;
5f44f142 1906
c16e7869
SW		 1907out_msgpool:
1908 ceph_msgpool_destroy(&osdc->msgpool_op);
5f44f142
SW		 1909out_mempool:
1910 mempool_destroy(osdc->req_mempool);
1911out:
1912 return err;
f24e9980
SW		 1913}
1914
1915void ceph_osdc_stop(struct ceph_osd_client *osdc)
1916{
a40c4f10
YS		 1917 flush_workqueue(osdc->notify_wq);
1918 destroy_workqueue(osdc->notify_wq);
f24e9980 1919 cancel_delayed_work_sync(&osdc->timeout_work);
f5a2041b 1920 cancel_delayed_work_sync(&osdc->osds_timeout_work);
f24e9980
SW		 1921 if (osdc->osdmap) {
1922 ceph_osdmap_destroy(osdc->osdmap);
1923 osdc->osdmap = NULL;
1924 }
aca420bc 1925 remove_all_osds(osdc);
f24e9980
SW		 1926 mempool_destroy(osdc->req_mempool);
1927 ceph_msgpool_destroy(&osdc->msgpool_op);
c16e7869 1928 ceph_msgpool_destroy(&osdc->msgpool_op_reply);
f24e9980
SW		 1929}
1930
1931/*
1932 * Read some contiguous pages. If we cross a stripe boundary, shorten
1933 * *plen. Return number of bytes read, or error.
1934 */
1935int ceph_osdc_readpages(struct ceph_osd_client *osdc,
1936 struct ceph_vino vino, struct ceph_file_layout *layout,
1937 u64 off, u64 *plen,
1938 u32 truncate_seq, u64 truncate_size,
b7495fc2 1939 struct page **pages, int num_pages, int page_align)
f24e9980
SW		 1940{
1941 struct ceph_osd_request *req;
1942 int rc = 0;
1943
1944 dout("readpages on ino %llx.%llx on %llu~%llu\n", vino.ino,
1945 vino.snap, off, *plen);
1946 req = ceph_osdc_new_request(osdc, layout, vino, off, plen,
1947 CEPH_OSD_OP_READ, CEPH_OSD_FLAG_READ,
1948 NULL, 0, truncate_seq, truncate_size, NULL,
a3bea47e 1949 false, page_align);
6816282d
SW		 1950 if (IS_ERR(req))
1951 return PTR_ERR(req);
f24e9980
SW		 1952
1953 /* it may be a short read due to an object boundary */
1954 req->r_pages = pages;
f24e9980 1955
b7495fc2
SW		 1956 dout("readpages final extent is %llu~%llu (%d pages align %d)\n",
1957 off, *plen, req->r_num_pages, page_align);
f24e9980
SW		 1958
1959 rc = ceph_osdc_start_request(osdc, req, false);
1960 if (!rc)
1961 rc = ceph_osdc_wait_request(osdc, req);
1962
1963 ceph_osdc_put_request(req);
1964 dout("readpages result %d\n", rc);
1965 return rc;
1966}
3d14c5d2 1967EXPORT_SYMBOL(ceph_osdc_readpages);
f24e9980
SW		 1968
1969/*
1970 * do a synchronous write on N pages
1971 */
1972int ceph_osdc_writepages(struct ceph_osd_client *osdc, struct ceph_vino vino,
1973 struct ceph_file_layout *layout,
1974 struct ceph_snap_context *snapc,
1975 u64 off, u64 len,
1976 u32 truncate_seq, u64 truncate_size,
1977 struct timespec *mtime,
24808826 1978 struct page **pages, int num_pages)
f24e9980
SW		 1979{
1980 struct ceph_osd_request *req;
1981 int rc = 0;
b7495fc2 1982 int page_align = off & ~PAGE_MASK;
f24e9980
SW		 1983
1984 BUG_ON(vino.snap != CEPH_NOSNAP);
1985 req = ceph_osdc_new_request(osdc, layout, vino, off, &len,
1986 CEPH_OSD_OP_WRITE,
24808826 1987 CEPH_OSD_FLAG_ONDISK | CEPH_OSD_FLAG_WRITE,
fbf8685f 1988 snapc, 0,
f24e9980 1989 truncate_seq, truncate_size, mtime,
a3bea47e 1990 true, page_align);
6816282d
SW		 1991 if (IS_ERR(req))
1992 return PTR_ERR(req);
f24e9980
SW		 1993
1994 /* it may be a short write due to an object boundary */
1995 req->r_pages = pages;
f24e9980
SW		 1996 dout("writepages %llu~%llu (%d pages)\n", off, len,
1997 req->r_num_pages);
1998
87f979d3 1999 rc = ceph_osdc_start_request(osdc, req, true);
f24e9980
SW		 2000 if (!rc)
2001 rc = ceph_osdc_wait_request(osdc, req);
2002
2003 ceph_osdc_put_request(req);
2004 if (rc == 0)
2005 rc = len;
2006 dout("writepages result %d\n", rc);
2007 return rc;
2008}
3d14c5d2 2009EXPORT_SYMBOL(ceph_osdc_writepages);
f24e9980
SW		 2010
2011/*
2012 * handle incoming message
2013 */
2014static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
2015{
2016 struct ceph_osd *osd = con->private;
32c895e7 2017 struct ceph_osd_client *osdc;
f24e9980
SW		 2018 int type = le16_to_cpu(msg->hdr.type);
2019
2020 if (!osd)
4a32f93d 2021 goto out;
32c895e7 2022 osdc = osd->o_osdc;
f24e9980
SW		 2023
2024 switch (type) {
2025 case CEPH_MSG_OSD_MAP:
2026 ceph_osdc_handle_map(osdc, msg);
2027 break;
2028 case CEPH_MSG_OSD_OPREPLY:
350b1c32 2029 handle_reply(osdc, msg, con);
f24e9980 2030 break;
a40c4f10
YS		 2031 case CEPH_MSG_WATCH_NOTIFY:
2032 handle_watch_notify(osdc, msg);
2033 break;
f24e9980
SW		 2034
2035 default:
2036 pr_err("received unknown message type %d %s\n", type,
2037 ceph_msg_type_name(type));
2038 }
4a32f93d 2039out:
f24e9980
SW		 2040 ceph_msg_put(msg);
2041}
2042
5b3a4db3 2043/*
21b667f6
SW		 2044 * lookup and return message for incoming reply. set up reply message
2045 * pages.
5b3a4db3
SW		 2046 */
2047static struct ceph_msg *get_reply(struct ceph_connection *con,
2450418c
YS		 2048 struct ceph_msg_header *hdr,
2049 int *skip)
f24e9980
SW		 2050{
2051 struct ceph_osd *osd = con->private;
2052 struct ceph_osd_client *osdc = osd->o_osdc;
2450418c 2053 struct ceph_msg *m;
0547a9b3 2054 struct ceph_osd_request *req;
5b3a4db3
SW		 2055 int front = le32_to_cpu(hdr->front_len);
2056 int data_len = le32_to_cpu(hdr->data_len);
0547a9b3 2057 u64 tid;
f24e9980 2058
0547a9b3
YS		 2059 tid = le64_to_cpu(hdr->tid);
2060 mutex_lock(&osdc->request_mutex);
2061 req = __lookup_request(osdc, tid);
2062 if (!req) {
2063 *skip = 1;
2064 m = NULL;
756a16a5
SW		 2065 dout("get_reply unknown tid %llu from osd%d\n", tid,
2066 osd->o_osd);
0547a9b3
YS		 2067 goto out;
2068 }
c16e7869
SW		 2069
2070 if (req->r_con_filling_msg) {
8921d114 2071 dout("%s revoking msg %p from old con %p\n", __func__,
c16e7869 2072 req->r_reply, req->r_con_filling_msg);
8921d114 2073 ceph_msg_revoke_incoming(req->r_reply);
0d47766f 2074 req->r_con_filling_msg->ops->put(req->r_con_filling_msg);
6f46cb29 2075 req->r_con_filling_msg = NULL;
0547a9b3
YS		 2076 }
2077
c16e7869
SW		 2078 if (front > req->r_reply->front.iov_len) {
2079 pr_warning("get_reply front %d > preallocated %d\n",
2080 front, (int)req->r_reply->front.iov_len);
b61c2763 2081 m = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, front, GFP_NOFS, false);
a79832f2 2082 if (!m)
c16e7869
SW		 2083 goto out;
2084 ceph_msg_put(req->r_reply);
2085 req->r_reply = m;
2086 }
2087 m = ceph_msg_get(req->r_reply);
2088
0547a9b3 2089 if (data_len > 0) {
b7495fc2 2090 int want = calc_pages_for(req->r_page_alignment, data_len);
21b667f6 2091
9cbb1d72 2092 if (req->r_pages && unlikely(req->r_num_pages < want)) {
9bb0ce2b
SW		 2093 pr_warning("tid %lld reply has %d bytes %d pages, we"
2094 " had only %d pages ready\n", tid, data_len,
2095 want, req->r_num_pages);
0547a9b3
YS		 2096 *skip = 1;
2097 ceph_msg_put(m);
a79832f2 2098 m = NULL;
21b667f6 2099 goto out;
0547a9b3 2100 }
21b667f6
SW		 2101 m->pages = req->r_pages;
2102 m->nr_pages = req->r_num_pages;
c5c6b19d 2103 m->page_alignment = req->r_page_alignment;
68b4476b
YS		 2104#ifdef CONFIG_BLOCK
2105 m->bio = req->r_bio;
2106#endif
0547a9b3 2107 }
5b3a4db3 2108 *skip = 0;
0d47766f 2109 req->r_con_filling_msg = con->ops->get(con);
c16e7869 2110 dout("get_reply tid %lld %p\n", tid, m);
0547a9b3
YS		 2111
2112out:
2113 mutex_unlock(&osdc->request_mutex);
2450418c 2114 return m;
5b3a4db3
SW		 2115
2116}
2117
2118static struct ceph_msg *alloc_msg(struct ceph_connection *con,
2119 struct ceph_msg_header *hdr,
2120 int *skip)
2121{
2122 struct ceph_osd *osd = con->private;
2123 int type = le16_to_cpu(hdr->type);
2124 int front = le32_to_cpu(hdr->front_len);
2125
1c20f2d2 2126 *skip = 0;
5b3a4db3
SW		 2127 switch (type) {
2128 case CEPH_MSG_OSD_MAP:
a40c4f10 2129 case CEPH_MSG_WATCH_NOTIFY:
b61c2763 2130 return ceph_msg_new(type, front, GFP_NOFS, false);
5b3a4db3
SW		 2131 case CEPH_MSG_OSD_OPREPLY:
2132 return get_reply(con, hdr, skip);
2133 default:
2134 pr_info("alloc_msg unexpected msg type %d from osd%d\n", type,
2135 osd->o_osd);
2136 *skip = 1;
2137 return NULL;
2138 }
f24e9980
SW		 2139}
2140
2141/*
2142 * Wrappers to refcount containing ceph_osd struct
2143 */
2144static struct ceph_connection *get_osd_con(struct ceph_connection *con)
2145{
2146 struct ceph_osd *osd = con->private;
2147 if (get_osd(osd))
2148 return con;
2149 return NULL;
2150}
2151
2152static void put_osd_con(struct ceph_connection *con)
2153{
2154 struct ceph_osd *osd = con->private;
2155 put_osd(osd);
2156}
2157
4e7a5dcd
SW		 2158/*
2159 * authentication
2160 */
a3530df3
AE		 2161/*
2162 * Note: returned pointer is the address of a structure that's
2163 * managed separately. Caller must *not* attempt to free it.
2164 */
2165static struct ceph_auth_handshake *get_authorizer(struct ceph_connection *con,
8f43fb53 2166 int *proto, int force_new)
4e7a5dcd
SW		 2167{
2168 struct ceph_osd *o = con->private;
2169 struct ceph_osd_client *osdc = o->o_osdc;
2170 struct ceph_auth_client *ac = osdc->client->monc.auth;
74f1869f 2171 struct ceph_auth_handshake *auth = &o->o_auth;
4e7a5dcd 2172
74f1869f 2173 if (force_new && auth->authorizer) {
a255651d
AE		 2174 if (ac->ops && ac->ops->destroy_authorizer)
2175 ac->ops->destroy_authorizer(ac, auth->authorizer);
74f1869f
AE		 2176 auth->authorizer = NULL;
2177 }
a255651d 2178 if (!auth->authorizer && ac->ops && ac->ops->create_authorizer) {
a3530df3
AE		 2179 int ret = ac->ops->create_authorizer(ac, CEPH_ENTITY_TYPE_OSD,
2180 auth);
4e7a5dcd 2181 if (ret)
a3530df3 2182 return ERR_PTR(ret);
4e7a5dcd 2183 }
4e7a5dcd 2184 *proto = ac->protocol;
74f1869f 2185
a3530df3 2186 return auth;
4e7a5dcd
SW		 2187}
2188
2189
2190static int verify_authorizer_reply(struct ceph_connection *con, int len)
2191{
2192 struct ceph_osd *o = con->private;
2193 struct ceph_osd_client *osdc = o->o_osdc;
2194 struct ceph_auth_client *ac = osdc->client->monc.auth;
2195
a255651d
AE		 2196 /*
2197 * XXX If ac->ops or ac->ops->verify_authorizer_reply is null,
2198 * XXX which do we do: succeed or fail?
2199 */
6c4a1915 2200 return ac->ops->verify_authorizer_reply(ac, o->o_auth.authorizer, len);
4e7a5dcd
SW		 2201}
2202
9bd2e6f8
SW		 2203static int invalidate_authorizer(struct ceph_connection *con)
2204{
2205 struct ceph_osd *o = con->private;
2206 struct ceph_osd_client *osdc = o->o_osdc;
2207 struct ceph_auth_client *ac = osdc->client->monc.auth;
2208
a255651d 2209 if (ac->ops && ac->ops->invalidate_authorizer)
9bd2e6f8
SW		 2210 ac->ops->invalidate_authorizer(ac, CEPH_ENTITY_TYPE_OSD);
2211
2212 return ceph_monc_validate_auth(&osdc->client->monc);
2213}
4e7a5dcd 2214
9e32789f 2215static const struct ceph_connection_operations osd_con_ops = {
f24e9980
SW		 2216 .get = get_osd_con,
2217 .put = put_osd_con,
2218 .dispatch = dispatch,
4e7a5dcd
SW		 2219 .get_authorizer = get_authorizer,
2220 .verify_authorizer_reply = verify_authorizer_reply,
9bd2e6f8 2221 .invalidate_authorizer = invalidate_authorizer,
f24e9980 2222 .alloc_msg = alloc_msg,
81b024e7 2223 .fault = osd_reset,
f24e9980 2224};
Linux 6.x block layer and io_uring tree(s)