projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge remote-tracking branch 'usb-chipidea-next/ci-for-usb-next'
[deliverable/linux.git] / drivers / staging / lustre / lustre / osc / osc_request.c
1 /*
2 * GPL HEADER START
3 *
4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
19 *
20 * GPL HEADER END
21 */
22 /*
23 * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
25 *
26 * Copyright (c) 2011, 2015, Intel Corporation.
27 */
28 /*
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
31 */
32
33 #define DEBUG_SUBSYSTEM S_OSC
34
35 #include "../../include/linux/libcfs/libcfs.h"
36
37 #include "../include/lustre_dlm.h"
38 #include "../include/lustre_net.h"
39 #include "../include/lustre/lustre_user.h"
40 #include "../include/obd_cksum.h"
41
42 #include "../include/lustre_ha.h"
43 #include "../include/lprocfs_status.h"
44 #include "../include/lustre_debug.h"
45 #include "../include/lustre_param.h"
46 #include "../include/lustre_fid.h"
47 #include "../include/obd_class.h"
48 #include "../include/obd.h"
49 #include "osc_internal.h"
50 #include "osc_cl_internal.h"
51
52 atomic_t osc_pool_req_count;
53 unsigned int osc_reqpool_maxreqcount;
54 struct ptlrpc_request_pool *osc_rq_pool;
55
56 /* max memory used for request pool, unit is MB */
57 static unsigned int osc_reqpool_mem_max = 5;
58 module_param(osc_reqpool_mem_max, uint, 0444);
59
60 struct osc_brw_async_args {
61 struct obdo *aa_oa;
62 int aa_requested_nob;
63 int aa_nio_count;
64 u32 aa_page_count;
65 int aa_resends;
66 struct brw_page **aa_ppga;
67 struct client_obd *aa_cli;
68 struct list_head aa_oaps;
69 struct list_head aa_exts;
70 struct cl_req *aa_clerq;
71 };
72
73 struct osc_async_args {
74 struct obd_info *aa_oi;
75 };
76
77 struct osc_setattr_args {
78 struct obdo *sa_oa;
79 obd_enqueue_update_f sa_upcall;
80 void *sa_cookie;
81 };
82
83 struct osc_fsync_args {
84 struct obd_info *fa_oi;
85 obd_enqueue_update_f fa_upcall;
86 void *fa_cookie;
87 };
88
89 struct osc_enqueue_args {
90 struct obd_export *oa_exp;
91 enum ldlm_type oa_type;
92 enum ldlm_mode oa_mode;
93 __u64 *oa_flags;
94 osc_enqueue_upcall_f oa_upcall;
95 void *oa_cookie;
96 struct ost_lvb *oa_lvb;
97 struct lustre_handle oa_lockh;
98 unsigned int oa_agl:1;
99 };
100
101 static void osc_release_ppga(struct brw_page **ppga, u32 count);
102 static int brw_interpret(const struct lu_env *env,
103 struct ptlrpc_request *req, void *data, int rc);
104
105 /* Pack OSC object metadata for disk storage (LE byte order). */
106 static int osc_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
107 struct lov_stripe_md *lsm)
108 {
109 int lmm_size;
110
111 lmm_size = sizeof(**lmmp);
112 if (!lmmp)
113 return lmm_size;
114
115 if (*lmmp && !lsm) {
116 kfree(*lmmp);
117 *lmmp = NULL;
118 return 0;
119 } else if (unlikely(lsm && ostid_id(&lsm->lsm_oi) == 0)) {
120 return -EBADF;
121 }
122
123 if (!*lmmp) {
124 *lmmp = kzalloc(lmm_size, GFP_NOFS);
125 if (!*lmmp)
126 return -ENOMEM;
127 }
128
129 if (lsm)
130 ostid_cpu_to_le(&lsm->lsm_oi, &(*lmmp)->lmm_oi);
131
132 return lmm_size;
133 }
134
135 /* Unpack OSC object metadata from disk storage (LE byte order). */
136 static int osc_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
137 struct lov_mds_md *lmm, int lmm_bytes)
138 {
139 int lsm_size;
140 struct obd_import *imp = class_exp2cliimp(exp);
141
142 if (lmm) {
143 if (lmm_bytes < sizeof(*lmm)) {
144 CERROR("%s: lov_mds_md too small: %d, need %d\n",
145 exp->exp_obd->obd_name, lmm_bytes,
146 (int)sizeof(*lmm));
147 return -EINVAL;
148 }
149 /* XXX LOV_MAGIC etc check? */
150
151 if (unlikely(ostid_id(&lmm->lmm_oi) == 0)) {
152 CERROR("%s: zero lmm_object_id: rc = %d\n",
153 exp->exp_obd->obd_name, -EINVAL);
154 return -EINVAL;
155 }
156 }
157
158 lsm_size = lov_stripe_md_size(1);
159 if (!lsmp)
160 return lsm_size;
161
162 if (*lsmp && !lmm) {
163 kfree((*lsmp)->lsm_oinfo[0]);
164 kfree(*lsmp);
165 *lsmp = NULL;
166 return 0;
167 }
168
169 if (!*lsmp) {
170 *lsmp = kzalloc(lsm_size, GFP_NOFS);
171 if (unlikely(!*lsmp))
172 return -ENOMEM;
173 (*lsmp)->lsm_oinfo[0] = kzalloc(sizeof(struct lov_oinfo),
174 GFP_NOFS);
175 if (unlikely(!(*lsmp)->lsm_oinfo[0])) {
176 kfree(*lsmp);
177 return -ENOMEM;
178 }
179 loi_init((*lsmp)->lsm_oinfo[0]);
180 } else if (unlikely(ostid_id(&(*lsmp)->lsm_oi) == 0)) {
181 return -EBADF;
182 }
183
184 if (lmm)
185 /* XXX zero *lsmp? */
186 ostid_le_to_cpu(&lmm->lmm_oi, &(*lsmp)->lsm_oi);
187
188 if (imp &&
189 (imp->imp_connect_data.ocd_connect_flags & OBD_CONNECT_MAXBYTES))
190 (*lsmp)->lsm_maxbytes = imp->imp_connect_data.ocd_maxbytes;
191 else
192 (*lsmp)->lsm_maxbytes = LUSTRE_STRIPE_MAXBYTES;
193
194 return lsm_size;
195 }
196
197 static inline void osc_pack_req_body(struct ptlrpc_request *req,
198 struct obd_info *oinfo)
199 {
200 struct ost_body *body;
201
202 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
203 LASSERT(body);
204
205 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa,
206 oinfo->oi_oa);
207 }
208
209 static int osc_getattr_interpret(const struct lu_env *env,
210 struct ptlrpc_request *req,
211 struct osc_async_args *aa, int rc)
212 {
213 struct ost_body *body;
214
215 if (rc != 0)
216 goto out;
217
218 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
219 if (body) {
220 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
221 lustre_get_wire_obdo(&req->rq_import->imp_connect_data,
222 aa->aa_oi->oi_oa, &body->oa);
223
224 /* This should really be sent by the OST */
225 aa->aa_oi->oi_oa->o_blksize = DT_MAX_BRW_SIZE;
226 aa->aa_oi->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
227 } else {
228 CDEBUG(D_INFO, "can't unpack ost_body\n");
229 rc = -EPROTO;
230 aa->aa_oi->oi_oa->o_valid = 0;
231 }
232 out:
233 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
234 return rc;
235 }
236
237 static int osc_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
238 struct ptlrpc_request_set *set)
239 {
240 struct ptlrpc_request *req;
241 struct osc_async_args *aa;
242 int rc;
243
244 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
245 if (!req)
246 return -ENOMEM;
247
248 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
249 if (rc) {
250 ptlrpc_request_free(req);
251 return rc;
252 }
253
254 osc_pack_req_body(req, oinfo);
255
256 ptlrpc_request_set_replen(req);
257 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_getattr_interpret;
258
259 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
260 aa = ptlrpc_req_async_args(req);
261 aa->aa_oi = oinfo;
262
263 ptlrpc_set_add_req(set, req);
264 return 0;
265 }
266
267 static int osc_getattr(const struct lu_env *env, struct obd_export *exp,
268 struct obd_info *oinfo)
269 {
270 struct ptlrpc_request *req;
271 struct ost_body *body;
272 int rc;
273
274 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
275 if (!req)
276 return -ENOMEM;
277
278 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
279 if (rc) {
280 ptlrpc_request_free(req);
281 return rc;
282 }
283
284 osc_pack_req_body(req, oinfo);
285
286 ptlrpc_request_set_replen(req);
287
288 rc = ptlrpc_queue_wait(req);
289 if (rc)
290 goto out;
291
292 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
293 if (!body) {
294 rc = -EPROTO;
295 goto out;
296 }
297
298 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
299 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oinfo->oi_oa,
300 &body->oa);
301
302 oinfo->oi_oa->o_blksize = cli_brw_size(exp->exp_obd);
303 oinfo->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
304
305 out:
306 ptlrpc_req_finished(req);
307 return rc;
308 }
309
310 static int osc_setattr(const struct lu_env *env, struct obd_export *exp,
311 struct obd_info *oinfo, struct obd_trans_info *oti)
312 {
313 struct ptlrpc_request *req;
314 struct ost_body *body;
315 int rc;
316
317 LASSERT(oinfo->oi_oa->o_valid & OBD_MD_FLGROUP);
318
319 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
320 if (!req)
321 return -ENOMEM;
322
323 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
324 if (rc) {
325 ptlrpc_request_free(req);
326 return rc;
327 }
328
329 osc_pack_req_body(req, oinfo);
330
331 ptlrpc_request_set_replen(req);
332
333 rc = ptlrpc_queue_wait(req);
334 if (rc)
335 goto out;
336
337 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
338 if (!body) {
339 rc = -EPROTO;
340 goto out;
341 }
342
343 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oinfo->oi_oa,
344 &body->oa);
345
346 out:
347 ptlrpc_req_finished(req);
348 return rc;
349 }
350
351 static int osc_setattr_interpret(const struct lu_env *env,
352 struct ptlrpc_request *req,
353 struct osc_setattr_args *sa, int rc)
354 {
355 struct ost_body *body;
356
357 if (rc != 0)
358 goto out;
359
360 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
361 if (!body) {
362 rc = -EPROTO;
363 goto out;
364 }
365
366 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, sa->sa_oa,
367 &body->oa);
368 out:
369 rc = sa->sa_upcall(sa->sa_cookie, rc);
370 return rc;
371 }
372
373 int osc_setattr_async_base(struct obd_export *exp, struct obd_info *oinfo,
374 struct obd_trans_info *oti,
375 obd_enqueue_update_f upcall, void *cookie,
376 struct ptlrpc_request_set *rqset)
377 {
378 struct ptlrpc_request *req;
379 struct osc_setattr_args *sa;
380 int rc;
381
382 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
383 if (!req)
384 return -ENOMEM;
385
386 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
387 if (rc) {
388 ptlrpc_request_free(req);
389 return rc;
390 }
391
392 if (oti && oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE)
393 oinfo->oi_oa->o_lcookie = *oti->oti_logcookies;
394
395 osc_pack_req_body(req, oinfo);
396
397 ptlrpc_request_set_replen(req);
398
399 /* do mds to ost setattr asynchronously */
400 if (!rqset) {
401 /* Do not wait for response. */
402 ptlrpcd_add_req(req);
403 } else {
404 req->rq_interpret_reply =
405 (ptlrpc_interpterer_t)osc_setattr_interpret;
406
407 CLASSERT(sizeof(*sa) <= sizeof(req->rq_async_args));
408 sa = ptlrpc_req_async_args(req);
409 sa->sa_oa = oinfo->oi_oa;
410 sa->sa_upcall = upcall;
411 sa->sa_cookie = cookie;
412
413 if (rqset == PTLRPCD_SET)
414 ptlrpcd_add_req(req);
415 else
416 ptlrpc_set_add_req(rqset, req);
417 }
418
419 return 0;
420 }
421
422 static int osc_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
423 struct obd_trans_info *oti,
424 struct ptlrpc_request_set *rqset)
425 {
426 return osc_setattr_async_base(exp, oinfo, oti,
427 oinfo->oi_cb_up, oinfo, rqset);
428 }
429
430 static int osc_real_create(struct obd_export *exp, struct obdo *oa,
431 struct lov_stripe_md **ea,
432 struct obd_trans_info *oti)
433 {
434 struct ptlrpc_request *req;
435 struct ost_body *body;
436 struct lov_stripe_md *lsm;
437 int rc;
438
439 LASSERT(oa);
440 LASSERT(ea);
441
442 lsm = *ea;
443 if (!lsm) {
444 rc = obd_alloc_memmd(exp, &lsm);
445 if (rc < 0)
446 return rc;
447 }
448
449 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
450 if (!req) {
451 rc = -ENOMEM;
452 goto out;
453 }
454
455 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
456 if (rc) {
457 ptlrpc_request_free(req);
458 goto out;
459 }
460
461 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
462 LASSERT(body);
463
464 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
465
466 ptlrpc_request_set_replen(req);
467
468 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
469 oa->o_flags == OBD_FL_DELORPHAN) {
470 DEBUG_REQ(D_HA, req,
471 "delorphan from OST integration");
472 /* Don't resend the delorphan req */
473 req->rq_no_resend = 1;
474 req->rq_no_delay = 1;
475 }
476
477 rc = ptlrpc_queue_wait(req);
478 if (rc)
479 goto out_req;
480
481 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
482 if (!body) {
483 rc = -EPROTO;
484 goto out_req;
485 }
486
487 CDEBUG(D_INFO, "oa flags %x\n", oa->o_flags);
488 lustre_get_wire_obdo(&req->rq_import->imp_connect_data, oa, &body->oa);
489
490 oa->o_blksize = cli_brw_size(exp->exp_obd);
491 oa->o_valid |= OBD_MD_FLBLKSZ;
492
493 /* XXX LOV STACKING: the lsm that is passed to us from LOV does not
494 * have valid lsm_oinfo data structs, so don't go touching that.
495 * This needs to be fixed in a big way.
496 */
497 lsm->lsm_oi = oa->o_oi;
498 *ea = lsm;
499
500 if (oti) {
501 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
502
503 if (oa->o_valid & OBD_MD_FLCOOKIE) {
504 if (!oti->oti_logcookies)
505 oti_alloc_cookies(oti, 1);
506 *oti->oti_logcookies = oa->o_lcookie;
507 }
508 }
509
510 CDEBUG(D_HA, "transno: %lld\n",
511 lustre_msg_get_transno(req->rq_repmsg));
512 out_req:
513 ptlrpc_req_finished(req);
514 out:
515 if (rc && !*ea)
516 obd_free_memmd(exp, &lsm);
517 return rc;
518 }
519
520 int osc_punch_base(struct obd_export *exp, struct obd_info *oinfo,
521 obd_enqueue_update_f upcall, void *cookie,
522 struct ptlrpc_request_set *rqset)
523 {
524 struct ptlrpc_request *req;
525 struct osc_setattr_args *sa;
526 struct ost_body *body;
527 int rc;
528
529 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_PUNCH);
530 if (!req)
531 return -ENOMEM;
532
533 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
534 if (rc) {
535 ptlrpc_request_free(req);
536 return rc;
537 }
538 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
539 ptlrpc_at_set_req_timeout(req);
540
541 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
542 LASSERT(body);
543 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa,
544 oinfo->oi_oa);
545
546 ptlrpc_request_set_replen(req);
547
548 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_setattr_interpret;
549 CLASSERT(sizeof(*sa) <= sizeof(req->rq_async_args));
550 sa = ptlrpc_req_async_args(req);
551 sa->sa_oa = oinfo->oi_oa;
552 sa->sa_upcall = upcall;
553 sa->sa_cookie = cookie;
554 if (rqset == PTLRPCD_SET)
555 ptlrpcd_add_req(req);
556 else
557 ptlrpc_set_add_req(rqset, req);
558
559 return 0;
560 }
561
562 static int osc_sync_interpret(const struct lu_env *env,
563 struct ptlrpc_request *req,
564 void *arg, int rc)
565 {
566 struct osc_fsync_args *fa = arg;
567 struct ost_body *body;
568
569 if (rc)
570 goto out;
571
572 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
573 if (!body) {
574 CERROR("can't unpack ost_body\n");
575 rc = -EPROTO;
576 goto out;
577 }
578
579 *fa->fa_oi->oi_oa = body->oa;
580 out:
581 rc = fa->fa_upcall(fa->fa_cookie, rc);
582 return rc;
583 }
584
585 int osc_sync_base(struct obd_export *exp, struct obd_info *oinfo,
586 obd_enqueue_update_f upcall, void *cookie,
587 struct ptlrpc_request_set *rqset)
588 {
589 struct ptlrpc_request *req;
590 struct ost_body *body;
591 struct osc_fsync_args *fa;
592 int rc;
593
594 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
595 if (!req)
596 return -ENOMEM;
597
598 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
599 if (rc) {
600 ptlrpc_request_free(req);
601 return rc;
602 }
603
604 /* overload the size and blocks fields in the oa with start/end */
605 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
606 LASSERT(body);
607 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa,
608 oinfo->oi_oa);
609
610 ptlrpc_request_set_replen(req);
611 req->rq_interpret_reply = osc_sync_interpret;
612
613 CLASSERT(sizeof(*fa) <= sizeof(req->rq_async_args));
614 fa = ptlrpc_req_async_args(req);
615 fa->fa_oi = oinfo;
616 fa->fa_upcall = upcall;
617 fa->fa_cookie = cookie;
618
619 if (rqset == PTLRPCD_SET)
620 ptlrpcd_add_req(req);
621 else
622 ptlrpc_set_add_req(rqset, req);
623
624 return 0;
625 }
626
627 /* Find and cancel locally locks matched by @mode in the resource found by
628 * @objid. Found locks are added into @cancel list. Returns the amount of
629 * locks added to @cancels list.
630 */
631 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
632 struct list_head *cancels,
633 enum ldlm_mode mode, __u64 lock_flags)
634 {
635 struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
636 struct ldlm_res_id res_id;
637 struct ldlm_resource *res;
638 int count;
639
640 /* Return, i.e. cancel nothing, only if ELC is supported (flag in
641 * export) but disabled through procfs (flag in NS).
642 *
643 * This distinguishes from a case when ELC is not supported originally,
644 * when we still want to cancel locks in advance and just cancel them
645 * locally, without sending any RPC.
646 */
647 if (exp_connect_cancelset(exp) && !ns_connect_cancelset(ns))
648 return 0;
649
650 ostid_build_res_name(&oa->o_oi, &res_id);
651 res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
652 if (!res)
653 return 0;
654
655 LDLM_RESOURCE_ADDREF(res);
656 count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
657 lock_flags, 0, NULL);
658 LDLM_RESOURCE_DELREF(res);
659 ldlm_resource_putref(res);
660 return count;
661 }
662
663 static int osc_destroy_interpret(const struct lu_env *env,
664 struct ptlrpc_request *req, void *data,
665 int rc)
666 {
667 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
668
669 atomic_dec(&cli->cl_destroy_in_flight);
670 wake_up(&cli->cl_destroy_waitq);
671 return 0;
672 }
673
674 static int osc_can_send_destroy(struct client_obd *cli)
675 {
676 if (atomic_inc_return(&cli->cl_destroy_in_flight) <=
677 cli->cl_max_rpcs_in_flight) {
678 /* The destroy request can be sent */
679 return 1;
680 }
681 if (atomic_dec_return(&cli->cl_destroy_in_flight) <
682 cli->cl_max_rpcs_in_flight) {
683 /*
684 * The counter has been modified between the two atomic
685 * operations.
686 */
687 wake_up(&cli->cl_destroy_waitq);
688 }
689 return 0;
690 }
691
692 static int osc_create(const struct lu_env *env, struct obd_export *exp,
693 struct obdo *oa, struct lov_stripe_md **ea,
694 struct obd_trans_info *oti)
695 {
696 int rc = 0;
697
698 LASSERT(oa);
699 LASSERT(ea);
700 LASSERT(oa->o_valid & OBD_MD_FLGROUP);
701
702 if ((oa->o_valid & OBD_MD_FLFLAGS) &&
703 oa->o_flags == OBD_FL_RECREATE_OBJS) {
704 return osc_real_create(exp, oa, ea, oti);
705 }
706
707 if (!fid_seq_is_mdt(ostid_seq(&oa->o_oi)))
708 return osc_real_create(exp, oa, ea, oti);
709
710 /* we should not get here anymore */
711 LBUG();
712
713 return rc;
714 }
715
716 /* Destroy requests can be async always on the client, and we don't even really
717 * care about the return code since the client cannot do anything at all about
718 * a destroy failure.
719 * When the MDS is unlinking a filename, it saves the file objects into a
720 * recovery llog, and these object records are cancelled when the OST reports
721 * they were destroyed and sync'd to disk (i.e. transaction committed).
722 * If the client dies, or the OST is down when the object should be destroyed,
723 * the records are not cancelled, and when the OST reconnects to the MDS next,
724 * it will retrieve the llog unlink logs and then sends the log cancellation
725 * cookies to the MDS after committing destroy transactions.
726 */
727 static int osc_destroy(const struct lu_env *env, struct obd_export *exp,
728 struct obdo *oa, struct lov_stripe_md *ea,
729 struct obd_trans_info *oti, struct obd_export *md_export)
730 {
731 struct client_obd *cli = &exp->exp_obd->u.cli;
732 struct ptlrpc_request *req;
733 struct ost_body *body;
734 LIST_HEAD(cancels);
735 int rc, count;
736
737 if (!oa) {
738 CDEBUG(D_INFO, "oa NULL\n");
739 return -EINVAL;
740 }
741
742 count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
743 LDLM_FL_DISCARD_DATA);
744
745 req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
746 if (!req) {
747 ldlm_lock_list_put(&cancels, l_bl_ast, count);
748 return -ENOMEM;
749 }
750
751 rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
752 0, &cancels, count);
753 if (rc) {
754 ptlrpc_request_free(req);
755 return rc;
756 }
757
758 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
759 ptlrpc_at_set_req_timeout(req);
760
761 if (oti && oa->o_valid & OBD_MD_FLCOOKIE)
762 oa->o_lcookie = *oti->oti_logcookies;
763 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
764 LASSERT(body);
765 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
766
767 ptlrpc_request_set_replen(req);
768
769 /* If osc_destroy is for destroying the unlink orphan,
770 * sent from MDT to OST, which should not be blocked here,
771 * because the process might be triggered by ptlrpcd, and
772 * it is not good to block ptlrpcd thread (b=16006
773 **/
774 if (!(oa->o_flags & OBD_FL_DELORPHAN)) {
775 req->rq_interpret_reply = osc_destroy_interpret;
776 if (!osc_can_send_destroy(cli)) {
777 struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP,
778 NULL);
779
780 /*
781 * Wait until the number of on-going destroy RPCs drops
782 * under max_rpc_in_flight
783 */
784 l_wait_event_exclusive(cli->cl_destroy_waitq,
785 osc_can_send_destroy(cli), &lwi);
786 }
787 }
788
789 /* Do not wait for response */
790 ptlrpcd_add_req(req);
791 return 0;
792 }
793
794 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
795 long writing_bytes)
796 {
797 u32 bits = OBD_MD_FLBLOCKS|OBD_MD_FLGRANT;
798
799 LASSERT(!(oa->o_valid & bits));
800
801 oa->o_valid |= bits;
802 spin_lock(&cli->cl_loi_list_lock);
803 oa->o_dirty = cli->cl_dirty;
804 if (unlikely(cli->cl_dirty - cli->cl_dirty_transit >
805 cli->cl_dirty_max)) {
806 CERROR("dirty %lu - %lu > dirty_max %lu\n",
807 cli->cl_dirty, cli->cl_dirty_transit, cli->cl_dirty_max);
808 oa->o_undirty = 0;
809 } else if (unlikely(atomic_read(&obd_unstable_pages) +
810 atomic_read(&obd_dirty_pages) -
811 atomic_read(&obd_dirty_transit_pages) >
812 (long)(obd_max_dirty_pages + 1))) {
813 /* The atomic_read() allowing the atomic_inc() are
814 * not covered by a lock thus they may safely race and trip
815 * this CERROR() unless we add in a small fudge factor (+1).
816 */
817 CERROR("%s: dirty %d + %d - %d > system dirty_max %d\n",
818 cli->cl_import->imp_obd->obd_name,
819 atomic_read(&obd_unstable_pages),
820 atomic_read(&obd_dirty_pages),
821 atomic_read(&obd_dirty_transit_pages),
822 obd_max_dirty_pages);
823 oa->o_undirty = 0;
824 } else if (unlikely(cli->cl_dirty_max - cli->cl_dirty > 0x7fffffff)) {
825 CERROR("dirty %lu - dirty_max %lu too big???\n",
826 cli->cl_dirty, cli->cl_dirty_max);
827 oa->o_undirty = 0;
828 } else {
829 long max_in_flight = (cli->cl_max_pages_per_rpc <<
830 PAGE_SHIFT)*
831 (cli->cl_max_rpcs_in_flight + 1);
832 oa->o_undirty = max(cli->cl_dirty_max, max_in_flight);
833 }
834 oa->o_grant = cli->cl_avail_grant + cli->cl_reserved_grant;
835 oa->o_dropped = cli->cl_lost_grant;
836 cli->cl_lost_grant = 0;
837 spin_unlock(&cli->cl_loi_list_lock);
838 CDEBUG(D_CACHE, "dirty: %llu undirty: %u dropped %u grant: %llu\n",
839 oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
840 }
841
842 void osc_update_next_shrink(struct client_obd *cli)
843 {
844 cli->cl_next_shrink_grant =
845 cfs_time_shift(cli->cl_grant_shrink_interval);
846 CDEBUG(D_CACHE, "next time %ld to shrink grant\n",
847 cli->cl_next_shrink_grant);
848 }
849
850 static void __osc_update_grant(struct client_obd *cli, u64 grant)
851 {
852 spin_lock(&cli->cl_loi_list_lock);
853 cli->cl_avail_grant += grant;
854 spin_unlock(&cli->cl_loi_list_lock);
855 }
856
857 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
858 {
859 if (body->oa.o_valid & OBD_MD_FLGRANT) {
860 CDEBUG(D_CACHE, "got %llu extra grant\n", body->oa.o_grant);
861 __osc_update_grant(cli, body->oa.o_grant);
862 }
863 }
864
865 static int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
866 u32 keylen, void *key, u32 vallen,
867 void *val, struct ptlrpc_request_set *set);
868
869 static int osc_shrink_grant_interpret(const struct lu_env *env,
870 struct ptlrpc_request *req,
871 void *aa, int rc)
872 {
873 struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
874 struct obdo *oa = ((struct osc_brw_async_args *)aa)->aa_oa;
875 struct ost_body *body;
876
877 if (rc != 0) {
878 __osc_update_grant(cli, oa->o_grant);
879 goto out;
880 }
881
882 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
883 LASSERT(body);
884 osc_update_grant(cli, body);
885 out:
886 kmem_cache_free(obdo_cachep, oa);
887 return rc;
888 }
889
890 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
891 {
892 spin_lock(&cli->cl_loi_list_lock);
893 oa->o_grant = cli->cl_avail_grant / 4;
894 cli->cl_avail_grant -= oa->o_grant;
895 spin_unlock(&cli->cl_loi_list_lock);
896 if (!(oa->o_valid & OBD_MD_FLFLAGS)) {
897 oa->o_valid |= OBD_MD_FLFLAGS;
898 oa->o_flags = 0;
899 }
900 oa->o_flags |= OBD_FL_SHRINK_GRANT;
901 osc_update_next_shrink(cli);
902 }
903
904 /* Shrink the current grant, either from some large amount to enough for a
905 * full set of in-flight RPCs, or if we have already shrunk to that limit
906 * then to enough for a single RPC. This avoids keeping more grant than
907 * needed, and avoids shrinking the grant piecemeal.
908 */
909 static int osc_shrink_grant(struct client_obd *cli)
910 {
911 __u64 target_bytes = (cli->cl_max_rpcs_in_flight + 1) *
912 (cli->cl_max_pages_per_rpc << PAGE_SHIFT);
913
914 spin_lock(&cli->cl_loi_list_lock);
915 if (cli->cl_avail_grant <= target_bytes)
916 target_bytes = cli->cl_max_pages_per_rpc << PAGE_SHIFT;
917 spin_unlock(&cli->cl_loi_list_lock);
918
919 return osc_shrink_grant_to_target(cli, target_bytes);
920 }
921
922 int osc_shrink_grant_to_target(struct client_obd *cli, __u64 target_bytes)
923 {
924 int rc = 0;
925 struct ost_body *body;
926
927 spin_lock(&cli->cl_loi_list_lock);
928 /* Don't shrink if we are already above or below the desired limit
929 * We don't want to shrink below a single RPC, as that will negatively
930 * impact block allocation and long-term performance.
931 */
932 if (target_bytes < cli->cl_max_pages_per_rpc << PAGE_SHIFT)
933 target_bytes = cli->cl_max_pages_per_rpc << PAGE_SHIFT;
934
935 if (target_bytes >= cli->cl_avail_grant) {
936 spin_unlock(&cli->cl_loi_list_lock);
937 return 0;
938 }
939 spin_unlock(&cli->cl_loi_list_lock);
940
941 body = kzalloc(sizeof(*body), GFP_NOFS);
942 if (!body)
943 return -ENOMEM;
944
945 osc_announce_cached(cli, &body->oa, 0);
946
947 spin_lock(&cli->cl_loi_list_lock);
948 body->oa.o_grant = cli->cl_avail_grant - target_bytes;
949 cli->cl_avail_grant = target_bytes;
950 spin_unlock(&cli->cl_loi_list_lock);
951 if (!(body->oa.o_valid & OBD_MD_FLFLAGS)) {
952 body->oa.o_valid |= OBD_MD_FLFLAGS;
953 body->oa.o_flags = 0;
954 }
955 body->oa.o_flags |= OBD_FL_SHRINK_GRANT;
956 osc_update_next_shrink(cli);
957
958 rc = osc_set_info_async(NULL, cli->cl_import->imp_obd->obd_self_export,
959 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
960 sizeof(*body), body, NULL);
961 if (rc != 0)
962 __osc_update_grant(cli, body->oa.o_grant);
963 kfree(body);
964 return rc;
965 }
966
967 static int osc_should_shrink_grant(struct client_obd *client)
968 {
969 unsigned long time = cfs_time_current();
970 unsigned long next_shrink = client->cl_next_shrink_grant;
971
972 if ((client->cl_import->imp_connect_data.ocd_connect_flags &
973 OBD_CONNECT_GRANT_SHRINK) == 0)
974 return 0;
975
976 if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
977 /* Get the current RPC size directly, instead of going via:
978 * cli_brw_size(obd->u.cli.cl_import->imp_obd->obd_self_export)
979 * Keep comment here so that it can be found by searching.
980 */
981 int brw_size = client->cl_max_pages_per_rpc << PAGE_SHIFT;
982
983 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
984 client->cl_avail_grant > brw_size)
985 return 1;
986
987 osc_update_next_shrink(client);
988 }
989 return 0;
990 }
991
992 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
993 {
994 struct client_obd *client;
995
996 list_for_each_entry(client, &item->ti_obd_list, cl_grant_shrink_list) {
997 if (osc_should_shrink_grant(client))
998 osc_shrink_grant(client);
999 }
1000 return 0;
1001 }
1002
1003 static int osc_add_shrink_grant(struct client_obd *client)
1004 {
1005 int rc;
1006
1007 rc = ptlrpc_add_timeout_client(client->cl_grant_shrink_interval,
1008 TIMEOUT_GRANT,
1009 osc_grant_shrink_grant_cb, NULL,
1010 &client->cl_grant_shrink_list);
1011 if (rc) {
1012 CERROR("add grant client %s error %d\n",
1013 client->cl_import->imp_obd->obd_name, rc);
1014 return rc;
1015 }
1016 CDEBUG(D_CACHE, "add grant client %s\n",
1017 client->cl_import->imp_obd->obd_name);
1018 osc_update_next_shrink(client);
1019 return 0;
1020 }
1021
1022 static int osc_del_shrink_grant(struct client_obd *client)
1023 {
1024 return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list,
1025 TIMEOUT_GRANT);
1026 }
1027
1028 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1029 {
1030 /*
1031 * ocd_grant is the total grant amount we're expect to hold: if we've
1032 * been evicted, it's the new avail_grant amount, cl_dirty will drop
1033 * to 0 as inflight RPCs fail out; otherwise, it's avail_grant + dirty.
1034 *
1035 * race is tolerable here: if we're evicted, but imp_state already
1036 * left EVICTED state, then cl_dirty must be 0 already.
1037 */
1038 spin_lock(&cli->cl_loi_list_lock);
1039 if (cli->cl_import->imp_state == LUSTRE_IMP_EVICTED)
1040 cli->cl_avail_grant = ocd->ocd_grant;
1041 else
1042 cli->cl_avail_grant = ocd->ocd_grant - cli->cl_dirty;
1043
1044 if (cli->cl_avail_grant < 0) {
1045 CWARN("%s: available grant < 0: avail/ocd/dirty %ld/%u/%ld\n",
1046 cli->cl_import->imp_obd->obd_name, cli->cl_avail_grant,
1047 ocd->ocd_grant, cli->cl_dirty);
1048 /* workaround for servers which do not have the patch from
1049 * LU-2679
1050 */
1051 cli->cl_avail_grant = ocd->ocd_grant;
1052 }
1053
1054 /* determine the appropriate chunk size used by osc_extent. */
1055 cli->cl_chunkbits = max_t(int, PAGE_SHIFT, ocd->ocd_blocksize);
1056 spin_unlock(&cli->cl_loi_list_lock);
1057
1058 CDEBUG(D_CACHE, "%s, setting cl_avail_grant: %ld cl_lost_grant: %ld chunk bits: %d\n",
1059 cli->cl_import->imp_obd->obd_name,
1060 cli->cl_avail_grant, cli->cl_lost_grant, cli->cl_chunkbits);
1061
1062 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
1063 list_empty(&cli->cl_grant_shrink_list))
1064 osc_add_shrink_grant(cli);
1065 }
1066
1067 /* We assume that the reason this OSC got a short read is because it read
1068 * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1069 * via the LOV, and it _knows_ it's reading inside the file, it's just that
1070 * this stripe never got written at or beyond this stripe offset yet.
1071 */
1072 static void handle_short_read(int nob_read, u32 page_count,
1073 struct brw_page **pga)
1074 {
1075 char *ptr;
1076 int i = 0;
1077
1078 /* skip bytes read OK */
1079 while (nob_read > 0) {
1080 LASSERT(page_count > 0);
1081
1082 if (pga[i]->count > nob_read) {
1083 /* EOF inside this page */
1084 ptr = kmap(pga[i]->pg) +
1085 (pga[i]->off & ~PAGE_MASK);
1086 memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1087 kunmap(pga[i]->pg);
1088 page_count--;
1089 i++;
1090 break;
1091 }
1092
1093 nob_read -= pga[i]->count;
1094 page_count--;
1095 i++;
1096 }
1097
1098 /* zero remaining pages */
1099 while (page_count-- > 0) {
1100 ptr = kmap(pga[i]->pg) + (pga[i]->off & ~PAGE_MASK);
1101 memset(ptr, 0, pga[i]->count);
1102 kunmap(pga[i]->pg);
1103 i++;
1104 }
1105 }
1106
1107 static int check_write_rcs(struct ptlrpc_request *req,
1108 int requested_nob, int niocount,
1109 u32 page_count, struct brw_page **pga)
1110 {
1111 int i;
1112 __u32 *remote_rcs;
1113
1114 remote_rcs = req_capsule_server_sized_get(&req->rq_pill, &RMF_RCS,
1115 sizeof(*remote_rcs) *
1116 niocount);
1117 if (!remote_rcs) {
1118 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1119 return -EPROTO;
1120 }
1121
1122 /* return error if any niobuf was in error */
1123 for (i = 0; i < niocount; i++) {
1124 if ((int)remote_rcs[i] < 0)
1125 return remote_rcs[i];
1126
1127 if (remote_rcs[i] != 0) {
1128 CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1129 i, remote_rcs[i], req);
1130 return -EPROTO;
1131 }
1132 }
1133
1134 if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1135 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1136 req->rq_bulk->bd_nob_transferred, requested_nob);
1137 return -EPROTO;
1138 }
1139
1140 return 0;
1141 }
1142
1143 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1144 {
1145 if (p1->flag != p2->flag) {
1146 unsigned mask = ~(OBD_BRW_FROM_GRANT | OBD_BRW_NOCACHE |
1147 OBD_BRW_SYNC | OBD_BRW_ASYNC |
1148 OBD_BRW_NOQUOTA | OBD_BRW_SOFT_SYNC);
1149
1150 /* warn if we try to combine flags that we don't know to be
1151 * safe to combine
1152 */
1153 if (unlikely((p1->flag & mask) != (p2->flag & mask))) {
1154 CWARN("Saw flags 0x%x and 0x%x in the same brw, please report this at http://bugs.whamcloud.com/\n",
1155 p1->flag, p2->flag);
1156 }
1157 return 0;
1158 }
1159
1160 return (p1->off + p1->count == p2->off);
1161 }
1162
1163 static u32 osc_checksum_bulk(int nob, u32 pg_count,
1164 struct brw_page **pga, int opc,
1165 enum cksum_type cksum_type)
1166 {
1167 __u32 cksum;
1168 int i = 0;
1169 struct cfs_crypto_hash_desc *hdesc;
1170 unsigned int bufsize;
1171 int err;
1172 unsigned char cfs_alg = cksum_obd2cfs(cksum_type);
1173
1174 LASSERT(pg_count > 0);
1175
1176 hdesc = cfs_crypto_hash_init(cfs_alg, NULL, 0);
1177 if (IS_ERR(hdesc)) {
1178 CERROR("Unable to initialize checksum hash %s\n",
1179 cfs_crypto_hash_name(cfs_alg));
1180 return PTR_ERR(hdesc);
1181 }
1182
1183 while (nob > 0 && pg_count > 0) {
1184 int count = pga[i]->count > nob ? nob : pga[i]->count;
1185
1186 /* corrupt the data before we compute the checksum, to
1187 * simulate an OST->client data error
1188 */
1189 if (i == 0 && opc == OST_READ &&
1190 OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE)) {
1191 unsigned char *ptr = kmap(pga[i]->pg);
1192 int off = pga[i]->off & ~PAGE_MASK;
1193
1194 memcpy(ptr + off, "bad1", min(4, nob));
1195 kunmap(pga[i]->pg);
1196 }
1197 cfs_crypto_hash_update_page(hdesc, pga[i]->pg,
1198 pga[i]->off & ~PAGE_MASK,
1199 count);
1200 CDEBUG(D_PAGE,
1201 "page %p map %p index %lu flags %lx count %u priv %0lx: off %d\n",
1202 pga[i]->pg, pga[i]->pg->mapping, pga[i]->pg->index,
1203 (long)pga[i]->pg->flags, page_count(pga[i]->pg),
1204 page_private(pga[i]->pg),
1205 (int)(pga[i]->off & ~PAGE_MASK));
1206
1207 nob -= pga[i]->count;
1208 pg_count--;
1209 i++;
1210 }
1211
1212 bufsize = sizeof(cksum);
1213 err = cfs_crypto_hash_final(hdesc, (unsigned char *)&cksum, &bufsize);
1214
1215 /* For sending we only compute the wrong checksum instead
1216 * of corrupting the data so it is still correct on a redo
1217 */
1218 if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1219 cksum++;
1220
1221 return cksum;
1222 }
1223
1224 static int osc_brw_prep_request(int cmd, struct client_obd *cli,
1225 struct obdo *oa,
1226 struct lov_stripe_md *lsm, u32 page_count,
1227 struct brw_page **pga,
1228 struct ptlrpc_request **reqp,
1229 int reserve,
1230 int resend)
1231 {
1232 struct ptlrpc_request *req;
1233 struct ptlrpc_bulk_desc *desc;
1234 struct ost_body *body;
1235 struct obd_ioobj *ioobj;
1236 struct niobuf_remote *niobuf;
1237 int niocount, i, requested_nob, opc, rc;
1238 struct osc_brw_async_args *aa;
1239 struct req_capsule *pill;
1240 struct brw_page *pg_prev;
1241
1242 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1243 return -ENOMEM; /* Recoverable */
1244 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1245 return -EINVAL; /* Fatal */
1246
1247 if ((cmd & OBD_BRW_WRITE) != 0) {
1248 opc = OST_WRITE;
1249 req = ptlrpc_request_alloc_pool(cli->cl_import,
1250 osc_rq_pool,
1251 &RQF_OST_BRW_WRITE);
1252 } else {
1253 opc = OST_READ;
1254 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW_READ);
1255 }
1256 if (!req)
1257 return -ENOMEM;
1258
1259 for (niocount = i = 1; i < page_count; i++) {
1260 if (!can_merge_pages(pga[i - 1], pga[i]))
1261 niocount++;
1262 }
1263
1264 pill = &req->rq_pill;
1265 req_capsule_set_size(pill, &RMF_OBD_IOOBJ, RCL_CLIENT,
1266 sizeof(*ioobj));
1267 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1268 niocount * sizeof(*niobuf));
1269
1270 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1271 if (rc) {
1272 ptlrpc_request_free(req);
1273 return rc;
1274 }
1275 req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1276 ptlrpc_at_set_req_timeout(req);
1277 /* ask ptlrpc not to resend on EINPROGRESS since BRWs have their own
1278 * retry logic
1279 */
1280 req->rq_no_retry_einprogress = 1;
1281
1282 desc = ptlrpc_prep_bulk_imp(req, page_count,
1283 cli->cl_import->imp_connect_data.ocd_brw_size >> LNET_MTU_BITS,
1284 opc == OST_WRITE ? BULK_GET_SOURCE : BULK_PUT_SINK,
1285 OST_BULK_PORTAL);
1286
1287 if (!desc) {
1288 rc = -ENOMEM;
1289 goto out;
1290 }
1291 /* NB request now owns desc and will free it when it gets freed */
1292
1293 body = req_capsule_client_get(pill, &RMF_OST_BODY);
1294 ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1295 niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1296 LASSERT(body && ioobj && niobuf);
1297
1298 lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
1299
1300 obdo_to_ioobj(oa, ioobj);
1301 ioobj->ioo_bufcnt = niocount;
1302 /* The high bits of ioo_max_brw tells server _maximum_ number of bulks
1303 * that might be send for this request. The actual number is decided
1304 * when the RPC is finally sent in ptlrpc_register_bulk(). It sends
1305 * "max - 1" for old client compatibility sending "0", and also so the
1306 * the actual maximum is a power-of-two number, not one less. LU-1431
1307 */
1308 ioobj_max_brw_set(ioobj, desc->bd_md_max_brw);
1309 LASSERT(page_count > 0);
1310 pg_prev = pga[0];
1311 for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1312 struct brw_page *pg = pga[i];
1313 int poff = pg->off & ~PAGE_MASK;
1314
1315 LASSERT(pg->count > 0);
1316 /* make sure there is no gap in the middle of page array */
1317 LASSERTF(page_count == 1 ||
1318 (ergo(i == 0, poff + pg->count == PAGE_SIZE) &&
1319 ergo(i > 0 && i < page_count - 1,
1320 poff == 0 && pg->count == PAGE_SIZE) &&
1321 ergo(i == page_count - 1, poff == 0)),
1322 "i: %d/%d pg: %p off: %llu, count: %u\n",
1323 i, page_count, pg, pg->off, pg->count);
1324 LASSERTF(i == 0 || pg->off > pg_prev->off,
1325 "i %d p_c %u pg %p [pri %lu ind %lu] off %llu prev_pg %p [pri %lu ind %lu] off %llu\n",
1326 i, page_count,
1327 pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1328 pg_prev->pg, page_private(pg_prev->pg),
1329 pg_prev->pg->index, pg_prev->off);
1330 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1331 (pg->flag & OBD_BRW_SRVLOCK));
1332
1333 ptlrpc_prep_bulk_page_pin(desc, pg->pg, poff, pg->count);
1334 requested_nob += pg->count;
1335
1336 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1337 niobuf--;
1338 niobuf->len += pg->count;
1339 } else {
1340 niobuf->offset = pg->off;
1341 niobuf->len = pg->count;
1342 niobuf->flags = pg->flag;
1343 }
1344 pg_prev = pg;
1345 }
1346
1347 LASSERTF((void *)(niobuf - niocount) ==
1348 req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE),
1349 "want %p - real %p\n", req_capsule_client_get(&req->rq_pill,
1350 &RMF_NIOBUF_REMOTE), (void *)(niobuf - niocount));
1351
1352 osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1353 if (resend) {
1354 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1355 body->oa.o_valid |= OBD_MD_FLFLAGS;
1356 body->oa.o_flags = 0;
1357 }
1358 body->oa.o_flags |= OBD_FL_RECOV_RESEND;
1359 }
1360
1361 if (osc_should_shrink_grant(cli))
1362 osc_shrink_grant_local(cli, &body->oa);
1363
1364 /* size[REQ_REC_OFF] still sizeof (*body) */
1365 if (opc == OST_WRITE) {
1366 if (cli->cl_checksum &&
1367 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1368 /* store cl_cksum_type in a local variable since
1369 * it can be changed via lprocfs
1370 */
1371 enum cksum_type cksum_type = cli->cl_cksum_type;
1372
1373 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1374 oa->o_flags &= OBD_FL_LOCAL_MASK;
1375 body->oa.o_flags = 0;
1376 }
1377 body->oa.o_flags |= cksum_type_pack(cksum_type);
1378 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1379 body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1380 page_count, pga,
1381 OST_WRITE,
1382 cksum_type);
1383 CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1384 body->oa.o_cksum);
1385 /* save this in 'oa', too, for later checking */
1386 oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1387 oa->o_flags |= cksum_type_pack(cksum_type);
1388 } else {
1389 /* clear out the checksum flag, in case this is a
1390 * resend but cl_checksum is no longer set. b=11238
1391 */
1392 oa->o_valid &= ~OBD_MD_FLCKSUM;
1393 }
1394 oa->o_cksum = body->oa.o_cksum;
1395 /* 1 RC per niobuf */
1396 req_capsule_set_size(pill, &RMF_RCS, RCL_SERVER,
1397 sizeof(__u32) * niocount);
1398 } else {
1399 if (cli->cl_checksum &&
1400 !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1401 if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1402 body->oa.o_flags = 0;
1403 body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1404 body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1405 }
1406 }
1407 ptlrpc_request_set_replen(req);
1408
1409 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1410 aa = ptlrpc_req_async_args(req);
1411 aa->aa_oa = oa;
1412 aa->aa_requested_nob = requested_nob;
1413 aa->aa_nio_count = niocount;
1414 aa->aa_page_count = page_count;
1415 aa->aa_resends = 0;
1416 aa->aa_ppga = pga;
1417 aa->aa_cli = cli;
1418 INIT_LIST_HEAD(&aa->aa_oaps);
1419
1420 *reqp = req;
1421 return 0;
1422
1423 out:
1424 ptlrpc_req_finished(req);
1425 return rc;
1426 }
1427
1428 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1429 __u32 client_cksum, __u32 server_cksum, int nob,
1430 u32 page_count, struct brw_page **pga,
1431 enum cksum_type client_cksum_type)
1432 {
1433 __u32 new_cksum;
1434 char *msg;
1435 enum cksum_type cksum_type;
1436
1437 if (server_cksum == client_cksum) {
1438 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1439 return 0;
1440 }
1441
1442 cksum_type = cksum_type_unpack(oa->o_valid & OBD_MD_FLFLAGS ?
1443 oa->o_flags : 0);
1444 new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1445 cksum_type);
1446
1447 if (cksum_type != client_cksum_type)
1448 msg = "the server did not use the checksum type specified in the original request - likely a protocol problem"
1449 ;
1450 else if (new_cksum == server_cksum)
1451 msg = "changed on the client after we checksummed it - likely false positive due to mmap IO (bug 11742)"
1452 ;
1453 else if (new_cksum == client_cksum)
1454 msg = "changed in transit before arrival at OST";
1455 else
1456 msg = "changed in transit AND doesn't match the original - likely false positive due to mmap IO (bug 11742)"
1457 ;
1458
1459 LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inode "DFID
1460 " object "DOSTID" extent [%llu-%llu]\n",
1461 msg, libcfs_nid2str(peer->nid),
1462 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : (__u64)0,
1463 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1464 oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1465 POSTID(&oa->o_oi), pga[0]->off,
1466 pga[page_count-1]->off + pga[page_count-1]->count - 1);
1467 CERROR("original client csum %x (type %x), server csum %x (type %x), client csum now %x\n",
1468 client_cksum, client_cksum_type,
1469 server_cksum, cksum_type, new_cksum);
1470 return 1;
1471 }
1472
1473 /* Note rc enters this function as number of bytes transferred */
1474 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1475 {
1476 struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1477 const lnet_process_id_t *peer =
1478 &req->rq_import->imp_connection->c_peer;
1479 struct client_obd *cli = aa->aa_cli;
1480 struct ost_body *body;
1481 __u32 client_cksum = 0;
1482
1483 if (rc < 0 && rc != -EDQUOT) {
1484 DEBUG_REQ(D_INFO, req, "Failed request with rc = %d\n", rc);
1485 return rc;
1486 }
1487
1488 LASSERTF(req->rq_repmsg, "rc = %d\n", rc);
1489 body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
1490 if (!body) {
1491 DEBUG_REQ(D_INFO, req, "Can't unpack body\n");
1492 return -EPROTO;
1493 }
1494
1495 /* set/clear over quota flag for a uid/gid */
1496 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1497 body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA)) {
1498 unsigned int qid[MAXQUOTAS] = { body->oa.o_uid, body->oa.o_gid };
1499
1500 CDEBUG(D_QUOTA, "setdq for [%u %u] with valid %#llx, flags %x\n",
1501 body->oa.o_uid, body->oa.o_gid, body->oa.o_valid,
1502 body->oa.o_flags);
1503 osc_quota_setdq(cli, qid, body->oa.o_valid, body->oa.o_flags);
1504 }
1505
1506 osc_update_grant(cli, body);
1507
1508 if (rc < 0)
1509 return rc;
1510
1511 if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1512 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1513
1514 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1515 if (rc > 0) {
1516 CERROR("Unexpected +ve rc %d\n", rc);
1517 return -EPROTO;
1518 }
1519 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1520
1521 if (sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1522 return -EAGAIN;
1523
1524 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1525 check_write_checksum(&body->oa, peer, client_cksum,
1526 body->oa.o_cksum, aa->aa_requested_nob,
1527 aa->aa_page_count, aa->aa_ppga,
1528 cksum_type_unpack(aa->aa_oa->o_flags)))
1529 return -EAGAIN;
1530
1531 rc = check_write_rcs(req, aa->aa_requested_nob,
1532 aa->aa_nio_count,
1533 aa->aa_page_count, aa->aa_ppga);
1534 goto out;
1535 }
1536
1537 /* The rest of this function executes only for OST_READs */
1538
1539 /* if unwrap_bulk failed, return -EAGAIN to retry */
1540 rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1541 if (rc < 0) {
1542 rc = -EAGAIN;
1543 goto out;
1544 }
1545
1546 if (rc > aa->aa_requested_nob) {
1547 CERROR("Unexpected rc %d (%d requested)\n", rc,
1548 aa->aa_requested_nob);
1549 return -EPROTO;
1550 }
1551
1552 if (rc != req->rq_bulk->bd_nob_transferred) {
1553 CERROR("Unexpected rc %d (%d transferred)\n",
1554 rc, req->rq_bulk->bd_nob_transferred);
1555 return -EPROTO;
1556 }
1557
1558 if (rc < aa->aa_requested_nob)
1559 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1560
1561 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1562 static int cksum_counter;
1563 __u32 server_cksum = body->oa.o_cksum;
1564 char *via = "";
1565 char *router = "";
1566 enum cksum_type cksum_type;
1567
1568 cksum_type = cksum_type_unpack(body->oa.o_valid&OBD_MD_FLFLAGS ?
1569 body->oa.o_flags : 0);
1570 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1571 aa->aa_ppga, OST_READ,
1572 cksum_type);
1573
1574 if (peer->nid != req->rq_bulk->bd_sender) {
1575 via = " via ";
1576 router = libcfs_nid2str(req->rq_bulk->bd_sender);
1577 }
1578
1579 if (server_cksum != client_cksum) {
1580 LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from %s%s%s inode " DFID " object " DOSTID " extent [%llu-%llu]\n",
1581 req->rq_import->imp_obd->obd_name,
1582 libcfs_nid2str(peer->nid),
1583 via, router,
1584 body->oa.o_valid & OBD_MD_FLFID ?
1585 body->oa.o_parent_seq : (__u64)0,
1586 body->oa.o_valid & OBD_MD_FLFID ?
1587 body->oa.o_parent_oid : 0,
1588 body->oa.o_valid & OBD_MD_FLFID ?
1589 body->oa.o_parent_ver : 0,
1590 POSTID(&body->oa.o_oi),
1591 aa->aa_ppga[0]->off,
1592 aa->aa_ppga[aa->aa_page_count-1]->off +
1593 aa->aa_ppga[aa->aa_page_count-1]->count -
1594 1);
1595 CERROR("client %x, server %x, cksum_type %x\n",
1596 client_cksum, server_cksum, cksum_type);
1597 cksum_counter = 0;
1598 aa->aa_oa->o_cksum = client_cksum;
1599 rc = -EAGAIN;
1600 } else {
1601 cksum_counter++;
1602 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1603 rc = 0;
1604 }
1605 } else if (unlikely(client_cksum)) {
1606 static int cksum_missed;
1607
1608 cksum_missed++;
1609 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1610 CERROR("Checksum %u requested from %s but not sent\n",
1611 cksum_missed, libcfs_nid2str(peer->nid));
1612 } else {
1613 rc = 0;
1614 }
1615 out:
1616 if (rc >= 0)
1617 lustre_get_wire_obdo(&req->rq_import->imp_connect_data,
1618 aa->aa_oa, &body->oa);
1619
1620 return rc;
1621 }
1622
1623 static int osc_brw_redo_request(struct ptlrpc_request *request,
1624 struct osc_brw_async_args *aa, int rc)
1625 {
1626 struct ptlrpc_request *new_req;
1627 struct osc_brw_async_args *new_aa;
1628 struct osc_async_page *oap;
1629
1630 DEBUG_REQ(rc == -EINPROGRESS ? D_RPCTRACE : D_ERROR, request,
1631 "redo for recoverable error %d", rc);
1632
1633 rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1634 OST_WRITE ? OBD_BRW_WRITE : OBD_BRW_READ,
1635 aa->aa_cli, aa->aa_oa,
1636 NULL /* lsm unused by osc currently */,
1637 aa->aa_page_count, aa->aa_ppga,
1638 &new_req, 0, 1);
1639 if (rc)
1640 return rc;
1641
1642 list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1643 if (oap->oap_request) {
1644 LASSERTF(request == oap->oap_request,
1645 "request %p != oap_request %p\n",
1646 request, oap->oap_request);
1647 if (oap->oap_interrupted) {
1648 ptlrpc_req_finished(new_req);
1649 return -EINTR;
1650 }
1651 }
1652 }
1653 /* New request takes over pga and oaps from old request.
1654 * Note that copying a list_head doesn't work, need to move it...
1655 */
1656 aa->aa_resends++;
1657 new_req->rq_interpret_reply = request->rq_interpret_reply;
1658 new_req->rq_async_args = request->rq_async_args;
1659 new_req->rq_commit_cb = request->rq_commit_cb;
1660 /* cap resend delay to the current request timeout, this is similar to
1661 * what ptlrpc does (see after_reply())
1662 */
1663 if (aa->aa_resends > new_req->rq_timeout)
1664 new_req->rq_sent = ktime_get_real_seconds() + new_req->rq_timeout;
1665 else
1666 new_req->rq_sent = ktime_get_real_seconds() + aa->aa_resends;
1667 new_req->rq_generation_set = 1;
1668 new_req->rq_import_generation = request->rq_import_generation;
1669
1670 new_aa = ptlrpc_req_async_args(new_req);
1671
1672 INIT_LIST_HEAD(&new_aa->aa_oaps);
1673 list_splice_init(&aa->aa_oaps, &new_aa->aa_oaps);
1674 INIT_LIST_HEAD(&new_aa->aa_exts);
1675 list_splice_init(&aa->aa_exts, &new_aa->aa_exts);
1676 new_aa->aa_resends = aa->aa_resends;
1677
1678 list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1679 if (oap->oap_request) {
1680 ptlrpc_req_finished(oap->oap_request);
1681 oap->oap_request = ptlrpc_request_addref(new_req);
1682 }
1683 }
1684
1685 /* XXX: This code will run into problem if we're going to support
1686 * to add a series of BRW RPCs into a self-defined ptlrpc_request_set
1687 * and wait for all of them to be finished. We should inherit request
1688 * set from old request.
1689 */
1690 ptlrpcd_add_req(new_req);
1691
1692 DEBUG_REQ(D_INFO, new_req, "new request");
1693 return 0;
1694 }
1695
1696 /*
1697 * ugh, we want disk allocation on the target to happen in offset order. we'll
1698 * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1699 * fine for our small page arrays and doesn't require allocation. its an
1700 * insertion sort that swaps elements that are strides apart, shrinking the
1701 * stride down until its '1' and the array is sorted.
1702 */
1703 static void sort_brw_pages(struct brw_page **array, int num)
1704 {
1705 int stride, i, j;
1706 struct brw_page *tmp;
1707
1708 if (num == 1)
1709 return;
1710 for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1711 ;
1712
1713 do {
1714 stride /= 3;
1715 for (i = stride ; i < num ; i++) {
1716 tmp = array[i];
1717 j = i;
1718 while (j >= stride && array[j - stride]->off > tmp->off) {
1719 array[j] = array[j - stride];
1720 j -= stride;
1721 }
1722 array[j] = tmp;
1723 }
1724 } while (stride > 1);
1725 }
1726
1727 static void osc_release_ppga(struct brw_page **ppga, u32 count)
1728 {
1729 LASSERT(ppga);
1730 kfree(ppga);
1731 }
1732
1733 static int brw_interpret(const struct lu_env *env,
1734 struct ptlrpc_request *req, void *data, int rc)
1735 {
1736 struct osc_brw_async_args *aa = data;
1737 struct osc_extent *ext;
1738 struct osc_extent *tmp;
1739 struct client_obd *cli = aa->aa_cli;
1740
1741 rc = osc_brw_fini_request(req, rc);
1742 CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
1743 /* When server return -EINPROGRESS, client should always retry
1744 * regardless of the number of times the bulk was resent already.
1745 */
1746 if (osc_recoverable_error(rc)) {
1747 if (req->rq_import_generation !=
1748 req->rq_import->imp_generation) {
1749 CDEBUG(D_HA, "%s: resend cross eviction for object: " DOSTID ", rc = %d.\n",
1750 req->rq_import->imp_obd->obd_name,
1751 POSTID(&aa->aa_oa->o_oi), rc);
1752 } else if (rc == -EINPROGRESS ||
1753 client_should_resend(aa->aa_resends, aa->aa_cli)) {
1754 rc = osc_brw_redo_request(req, aa, rc);
1755 } else {
1756 CERROR("%s: too many resent retries for object: %llu:%llu, rc = %d.\n",
1757 req->rq_import->imp_obd->obd_name,
1758 POSTID(&aa->aa_oa->o_oi), rc);
1759 }
1760
1761 if (rc == 0)
1762 return 0;
1763 else if (rc == -EAGAIN || rc == -EINPROGRESS)
1764 rc = -EIO;
1765 }
1766
1767 if (rc == 0) {
1768 struct obdo *oa = aa->aa_oa;
1769 struct cl_attr *attr = &osc_env_info(env)->oti_attr;
1770 unsigned long valid = 0;
1771 struct cl_object *obj;
1772 struct osc_async_page *last;
1773
1774 last = brw_page2oap(aa->aa_ppga[aa->aa_page_count - 1]);
1775 obj = osc2cl(last->oap_obj);
1776
1777 cl_object_attr_lock(obj);
1778 if (oa->o_valid & OBD_MD_FLBLOCKS) {
1779 attr->cat_blocks = oa->o_blocks;
1780 valid |= CAT_BLOCKS;
1781 }
1782 if (oa->o_valid & OBD_MD_FLMTIME) {
1783 attr->cat_mtime = oa->o_mtime;
1784 valid |= CAT_MTIME;
1785 }
1786 if (oa->o_valid & OBD_MD_FLATIME) {
1787 attr->cat_atime = oa->o_atime;
1788 valid |= CAT_ATIME;
1789 }
1790 if (oa->o_valid & OBD_MD_FLCTIME) {
1791 attr->cat_ctime = oa->o_ctime;
1792 valid |= CAT_CTIME;
1793 }
1794
1795 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1796 struct lov_oinfo *loi = cl2osc(obj)->oo_oinfo;
1797 loff_t last_off = last->oap_count + last->oap_obj_off;
1798
1799 /* Change file size if this is an out of quota or
1800 * direct IO write and it extends the file size
1801 */
1802 if (loi->loi_lvb.lvb_size < last_off) {
1803 attr->cat_size = last_off;
1804 valid |= CAT_SIZE;
1805 }
1806 /* Extend KMS if it's not a lockless write */
1807 if (loi->loi_kms < last_off &&
1808 oap2osc_page(last)->ops_srvlock == 0) {
1809 attr->cat_kms = last_off;
1810 valid |= CAT_KMS;
1811 }
1812 }
1813
1814 if (valid != 0)
1815 cl_object_attr_set(env, obj, attr, valid);
1816 cl_object_attr_unlock(obj);
1817 }
1818 kmem_cache_free(obdo_cachep, aa->aa_oa);
1819
1820 list_for_each_entry_safe(ext, tmp, &aa->aa_exts, oe_link) {
1821 list_del_init(&ext->oe_link);
1822 osc_extent_finish(env, ext, 1, rc);
1823 }
1824 LASSERT(list_empty(&aa->aa_exts));
1825 LASSERT(list_empty(&aa->aa_oaps));
1826
1827 cl_req_completion(env, aa->aa_clerq, rc < 0 ? rc :
1828 req->rq_bulk->bd_nob_transferred);
1829 osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
1830 ptlrpc_lprocfs_brw(req, req->rq_bulk->bd_nob_transferred);
1831
1832 spin_lock(&cli->cl_loi_list_lock);
1833 /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
1834 * is called so we know whether to go to sync BRWs or wait for more
1835 * RPCs to complete
1836 */
1837 if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
1838 cli->cl_w_in_flight--;
1839 else
1840 cli->cl_r_in_flight--;
1841 osc_wake_cache_waiters(cli);
1842 spin_unlock(&cli->cl_loi_list_lock);
1843
1844 osc_io_unplug(env, cli, NULL);
1845 return rc;
1846 }
1847
1848 static void brw_commit(struct ptlrpc_request *req)
1849 {
1850 spin_lock(&req->rq_lock);
1851 /*
1852 * If osc_inc_unstable_pages (via osc_extent_finish) races with
1853 * this called via the rq_commit_cb, I need to ensure
1854 * osc_dec_unstable_pages is still called. Otherwise unstable
1855 * pages may be leaked.
1856 */
1857 if (req->rq_unstable) {
1858 spin_unlock(&req->rq_lock);
1859 osc_dec_unstable_pages(req);
1860 spin_lock(&req->rq_lock);
1861 } else {
1862 req->rq_committed = 1;
1863 }
1864 spin_unlock(&req->rq_lock);
1865 }
1866
1867 /**
1868 * Build an RPC by the list of extent @ext_list. The caller must ensure
1869 * that the total pages in this list are NOT over max pages per RPC.
1870 * Extents in the list must be in OES_RPC state.
1871 */
1872 int osc_build_rpc(const struct lu_env *env, struct client_obd *cli,
1873 struct list_head *ext_list, int cmd)
1874 {
1875 struct ptlrpc_request *req = NULL;
1876 struct osc_extent *ext;
1877 struct brw_page **pga = NULL;
1878 struct osc_brw_async_args *aa = NULL;
1879 struct obdo *oa = NULL;
1880 struct osc_async_page *oap;
1881 struct osc_async_page *tmp;
1882 struct cl_req *clerq = NULL;
1883 enum cl_req_type crt = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
1884 struct ldlm_lock *lock = NULL;
1885 struct cl_req_attr *crattr = NULL;
1886 u64 starting_offset = OBD_OBJECT_EOF;
1887 u64 ending_offset = 0;
1888 int mpflag = 0;
1889 int mem_tight = 0;
1890 int page_count = 0;
1891 int i;
1892 int rc;
1893 struct ost_body *body;
1894 LIST_HEAD(rpc_list);
1895
1896 LASSERT(!list_empty(ext_list));
1897
1898 /* add pages into rpc_list to build BRW rpc */
1899 list_for_each_entry(ext, ext_list, oe_link) {
1900 LASSERT(ext->oe_state == OES_RPC);
1901 mem_tight |= ext->oe_memalloc;
1902 list_for_each_entry(oap, &ext->oe_pages, oap_pending_item) {
1903 ++page_count;
1904 list_add_tail(&oap->oap_rpc_item, &rpc_list);
1905 if (starting_offset > oap->oap_obj_off)
1906 starting_offset = oap->oap_obj_off;
1907 else
1908 LASSERT(oap->oap_page_off == 0);
1909 if (ending_offset < oap->oap_obj_off + oap->oap_count)
1910 ending_offset = oap->oap_obj_off +
1911 oap->oap_count;
1912 else
1913 LASSERT(oap->oap_page_off + oap->oap_count ==
1914 PAGE_SIZE);
1915 }
1916 }
1917
1918 if (mem_tight)
1919 mpflag = cfs_memory_pressure_get_and_set();
1920
1921 crattr = kzalloc(sizeof(*crattr), GFP_NOFS);
1922 if (!crattr) {
1923 rc = -ENOMEM;
1924 goto out;
1925 }
1926
1927 pga = kcalloc(page_count, sizeof(*pga), GFP_NOFS);
1928 if (!pga) {
1929 rc = -ENOMEM;
1930 goto out;
1931 }
1932
1933 oa = kmem_cache_zalloc(obdo_cachep, GFP_NOFS);
1934 if (!oa) {
1935 rc = -ENOMEM;
1936 goto out;
1937 }
1938
1939 i = 0;
1940 list_for_each_entry(oap, &rpc_list, oap_rpc_item) {
1941 struct cl_page *page = oap2cl_page(oap);
1942
1943 if (!clerq) {
1944 clerq = cl_req_alloc(env, page, crt,
1945 1 /* only 1-object rpcs for now */);
1946 if (IS_ERR(clerq)) {
1947 rc = PTR_ERR(clerq);
1948 goto out;
1949 }
1950 lock = oap->oap_ldlm_lock;
1951 }
1952 if (mem_tight)
1953 oap->oap_brw_flags |= OBD_BRW_MEMALLOC;
1954 pga[i] = &oap->oap_brw_page;
1955 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
1956 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
1957 pga[i]->pg, oap->oap_page->index, oap,
1958 pga[i]->flag);
1959 i++;
1960 cl_req_page_add(env, clerq, page);
1961 }
1962
1963 /* always get the data for the obdo for the rpc */
1964 LASSERT(clerq);
1965 crattr->cra_oa = oa;
1966 cl_req_attr_set(env, clerq, crattr, ~0ULL);
1967 if (lock) {
1968 oa->o_handle = lock->l_remote_handle;
1969 oa->o_valid |= OBD_MD_FLHANDLE;
1970 }
1971
1972 rc = cl_req_prep(env, clerq);
1973 if (rc != 0) {
1974 CERROR("cl_req_prep failed: %d\n", rc);
1975 goto out;
1976 }
1977
1978 sort_brw_pages(pga, page_count);
1979 rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count,
1980 pga, &req, 1, 0);
1981 if (rc != 0) {
1982 CERROR("prep_req failed: %d\n", rc);
1983 goto out;
1984 }
1985
1986 req->rq_commit_cb = brw_commit;
1987 req->rq_interpret_reply = brw_interpret;
1988
1989 if (mem_tight != 0)
1990 req->rq_memalloc = 1;
1991
1992 /* Need to update the timestamps after the request is built in case
1993 * we race with setattr (locally or in queue at OST). If OST gets
1994 * later setattr before earlier BRW (as determined by the request xid),
1995 * the OST will not use BRW timestamps. Sadly, there is no obvious
1996 * way to do this in a single call. bug 10150
1997 */
1998 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
1999 crattr->cra_oa = &body->oa;
2000 cl_req_attr_set(env, clerq, crattr,
2001 OBD_MD_FLMTIME|OBD_MD_FLCTIME|OBD_MD_FLATIME);
2002
2003 lustre_msg_set_jobid(req->rq_reqmsg, crattr->cra_jobid);
2004
2005 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2006 aa = ptlrpc_req_async_args(req);
2007 INIT_LIST_HEAD(&aa->aa_oaps);
2008 list_splice_init(&rpc_list, &aa->aa_oaps);
2009 INIT_LIST_HEAD(&aa->aa_exts);
2010 list_splice_init(ext_list, &aa->aa_exts);
2011 aa->aa_clerq = clerq;
2012
2013 /* queued sync pages can be torn down while the pages
2014 * were between the pending list and the rpc
2015 */
2016 tmp = NULL;
2017 list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2018 /* only one oap gets a request reference */
2019 if (!tmp)
2020 tmp = oap;
2021 if (oap->oap_interrupted && !req->rq_intr) {
2022 CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2023 oap, req);
2024 ptlrpc_mark_interrupted(req);
2025 }
2026 }
2027 if (tmp)
2028 tmp->oap_request = ptlrpc_request_addref(req);
2029
2030 spin_lock(&cli->cl_loi_list_lock);
2031 starting_offset >>= PAGE_SHIFT;
2032 if (cmd == OBD_BRW_READ) {
2033 cli->cl_r_in_flight++;
2034 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2035 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2036 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2037 starting_offset + 1);
2038 } else {
2039 cli->cl_w_in_flight++;
2040 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2041 lprocfs_oh_tally(&cli->cl_write_rpc_hist, cli->cl_w_in_flight);
2042 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2043 starting_offset + 1);
2044 }
2045 spin_unlock(&cli->cl_loi_list_lock);
2046
2047 DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2048 page_count, aa, cli->cl_r_in_flight,
2049 cli->cl_w_in_flight);
2050
2051 ptlrpcd_add_req(req);
2052 rc = 0;
2053
2054 out:
2055 if (mem_tight != 0)
2056 cfs_memory_pressure_restore(mpflag);
2057
2058 kfree(crattr);
2059
2060 if (rc != 0) {
2061 LASSERT(!req);
2062
2063 if (oa)
2064 kmem_cache_free(obdo_cachep, oa);
2065 kfree(pga);
2066 /* this should happen rarely and is pretty bad, it makes the
2067 * pending list not follow the dirty order
2068 */
2069 while (!list_empty(ext_list)) {
2070 ext = list_entry(ext_list->next, struct osc_extent,
2071 oe_link);
2072 list_del_init(&ext->oe_link);
2073 osc_extent_finish(env, ext, 0, rc);
2074 }
2075 if (clerq && !IS_ERR(clerq))
2076 cl_req_completion(env, clerq, rc);
2077 }
2078 return rc;
2079 }
2080
2081 static int osc_set_lock_data_with_check(struct ldlm_lock *lock,
2082 struct ldlm_enqueue_info *einfo)
2083 {
2084 void *data = einfo->ei_cbdata;
2085 int set = 0;
2086
2087 LASSERT(lock->l_blocking_ast == einfo->ei_cb_bl);
2088 LASSERT(lock->l_resource->lr_type == einfo->ei_type);
2089 LASSERT(lock->l_completion_ast == einfo->ei_cb_cp);
2090 LASSERT(lock->l_glimpse_ast == einfo->ei_cb_gl);
2091
2092 lock_res_and_lock(lock);
2093
2094 if (!lock->l_ast_data)
2095 lock->l_ast_data = data;
2096 if (lock->l_ast_data == data)
2097 set = 1;
2098
2099 unlock_res_and_lock(lock);
2100
2101 return set;
2102 }
2103
2104 static int osc_set_data_with_check(struct lustre_handle *lockh,
2105 struct ldlm_enqueue_info *einfo)
2106 {
2107 struct ldlm_lock *lock = ldlm_handle2lock(lockh);
2108 int set = 0;
2109
2110 if (lock) {
2111 set = osc_set_lock_data_with_check(lock, einfo);
2112 LDLM_LOCK_PUT(lock);
2113 } else
2114 CERROR("lockh %p, data %p - client evicted?\n",
2115 lockh, einfo->ei_cbdata);
2116 return set;
2117 }
2118
2119 /* find any ldlm lock of the inode in osc
2120 * return 0 not find
2121 * 1 find one
2122 * < 0 error
2123 */
2124 static int osc_find_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
2125 ldlm_iterator_t replace, void *data)
2126 {
2127 struct ldlm_res_id res_id;
2128 struct obd_device *obd = class_exp2obd(exp);
2129 int rc = 0;
2130
2131 ostid_build_res_name(&lsm->lsm_oi, &res_id);
2132 rc = ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
2133 if (rc == LDLM_ITER_STOP)
2134 return 1;
2135 if (rc == LDLM_ITER_CONTINUE)
2136 return 0;
2137 return rc;
2138 }
2139
2140 static int osc_enqueue_fini(struct ptlrpc_request *req,
2141 osc_enqueue_upcall_f upcall, void *cookie,
2142 struct lustre_handle *lockh, enum ldlm_mode mode,
2143 __u64 *flags, int agl, int errcode)
2144 {
2145 bool intent = *flags & LDLM_FL_HAS_INTENT;
2146 int rc;
2147
2148 /* The request was created before ldlm_cli_enqueue call. */
2149 if (intent && errcode == ELDLM_LOCK_ABORTED) {
2150 struct ldlm_reply *rep;
2151
2152 rep = req_capsule_server_get(&req->rq_pill, &RMF_DLM_REP);
2153
2154 rep->lock_policy_res1 =
2155 ptlrpc_status_ntoh(rep->lock_policy_res1);
2156 if (rep->lock_policy_res1)
2157 errcode = rep->lock_policy_res1;
2158 if (!agl)
2159 *flags |= LDLM_FL_LVB_READY;
2160 } else if (errcode == ELDLM_OK) {
2161 *flags |= LDLM_FL_LVB_READY;
2162 }
2163
2164 /* Call the update callback. */
2165 rc = (*upcall)(cookie, lockh, errcode);
2166 /* release the reference taken in ldlm_cli_enqueue() */
2167 if (errcode == ELDLM_LOCK_MATCHED)
2168 errcode = ELDLM_OK;
2169 if (errcode == ELDLM_OK && lustre_handle_is_used(lockh))
2170 ldlm_lock_decref(lockh, mode);
2171
2172 return rc;
2173 }
2174
2175 static int osc_enqueue_interpret(const struct lu_env *env,
2176 struct ptlrpc_request *req,
2177 struct osc_enqueue_args *aa, int rc)
2178 {
2179 struct ldlm_lock *lock;
2180 struct lustre_handle *lockh = &aa->oa_lockh;
2181 enum ldlm_mode mode = aa->oa_mode;
2182 struct ost_lvb *lvb = aa->oa_lvb;
2183 __u32 lvb_len = sizeof(*lvb);
2184 __u64 flags = 0;
2185
2186
2187 /* ldlm_cli_enqueue is holding a reference on the lock, so it must
2188 * be valid.
2189 */
2190 lock = ldlm_handle2lock(lockh);
2191 LASSERTF(lock, "lockh %llx, req %p, aa %p - client evicted?\n",
2192 lockh->cookie, req, aa);
2193
2194 /* Take an additional reference so that a blocking AST that
2195 * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
2196 * to arrive after an upcall has been executed by
2197 * osc_enqueue_fini().
2198 */
2199 ldlm_lock_addref(lockh, mode);
2200
2201 /* Let cl_lock_state_wait fail with -ERESTARTSYS to unuse sublocks. */
2202 OBD_FAIL_TIMEOUT(OBD_FAIL_LDLM_ENQUEUE_HANG, 2);
2203
2204 /* Let CP AST to grant the lock first. */
2205 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_ENQ_RACE, 1);
2206
2207 if (aa->oa_agl) {
2208 LASSERT(!aa->oa_lvb);
2209 LASSERT(!aa->oa_flags);
2210 aa->oa_flags = &flags;
2211 }
2212
2213 /* Complete obtaining the lock procedure. */
2214 rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_type, 1,
2215 aa->oa_mode, aa->oa_flags, lvb, lvb_len,
2216 lockh, rc);
2217 /* Complete osc stuff. */
2218 rc = osc_enqueue_fini(req, aa->oa_upcall, aa->oa_cookie, lockh, mode,
2219 aa->oa_flags, aa->oa_agl, rc);
2220
2221 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_CANCEL_RACE, 10);
2222
2223 ldlm_lock_decref(lockh, mode);
2224 LDLM_LOCK_PUT(lock);
2225 return rc;
2226 }
2227
2228 struct ptlrpc_request_set *PTLRPCD_SET = (void *)1;
2229
2230 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
2231 * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
2232 * other synchronous requests, however keeping some locks and trying to obtain
2233 * others may take a considerable amount of time in a case of ost failure; and
2234 * when other sync requests do not get released lock from a client, the client
2235 * is evicted from the cluster -- such scenaries make the life difficult, so
2236 * release locks just after they are obtained.
2237 */
2238 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2239 __u64 *flags, ldlm_policy_data_t *policy,
2240 struct ost_lvb *lvb, int kms_valid,
2241 osc_enqueue_upcall_f upcall, void *cookie,
2242 struct ldlm_enqueue_info *einfo,
2243 struct ptlrpc_request_set *rqset, int async, int agl)
2244 {
2245 struct obd_device *obd = exp->exp_obd;
2246 struct lustre_handle lockh = { 0 };
2247 struct ptlrpc_request *req = NULL;
2248 int intent = *flags & LDLM_FL_HAS_INTENT;
2249 __u64 match_flags = *flags;
2250 enum ldlm_mode mode;
2251 int rc;
2252
2253 /* Filesystem lock extents are extended to page boundaries so that
2254 * dealing with the page cache is a little smoother.
2255 */
2256 policy->l_extent.start -= policy->l_extent.start & ~PAGE_MASK;
2257 policy->l_extent.end |= ~PAGE_MASK;
2258
2259 /*
2260 * kms is not valid when either object is completely fresh (so that no
2261 * locks are cached), or object was evicted. In the latter case cached
2262 * lock cannot be used, because it would prime inode state with
2263 * potentially stale LVB.
2264 */
2265 if (!kms_valid)
2266 goto no_match;
2267
2268 /* Next, search for already existing extent locks that will cover us */
2269 /* If we're trying to read, we also search for an existing PW lock. The
2270 * VFS and page cache already protect us locally, so lots of readers/
2271 * writers can share a single PW lock.
2272 *
2273 * There are problems with conversion deadlocks, so instead of
2274 * converting a read lock to a write lock, we'll just enqueue a new
2275 * one.
2276 *
2277 * At some point we should cancel the read lock instead of making them
2278 * send us a blocking callback, but there are problems with canceling
2279 * locks out from other users right now, too.
2280 */
2281 mode = einfo->ei_mode;
2282 if (einfo->ei_mode == LCK_PR)
2283 mode |= LCK_PW;
2284 if (agl == 0)
2285 match_flags |= LDLM_FL_LVB_READY;
2286 if (intent != 0)
2287 match_flags |= LDLM_FL_BLOCK_GRANTED;
2288 mode = ldlm_lock_match(obd->obd_namespace, match_flags, res_id,
2289 einfo->ei_type, policy, mode, &lockh, 0);
2290 if (mode) {
2291 struct ldlm_lock *matched;
2292
2293 if (*flags & LDLM_FL_TEST_LOCK)
2294 return ELDLM_OK;
2295
2296 matched = ldlm_handle2lock(&lockh);
2297 if (agl) {
2298 /* AGL enqueues DLM locks speculatively. Therefore if
2299 * it already exists a DLM lock, it wll just inform the
2300 * caller to cancel the AGL process for this stripe.
2301 */
2302 ldlm_lock_decref(&lockh, mode);
2303 LDLM_LOCK_PUT(matched);
2304 return -ECANCELED;
2305 } else if (osc_set_lock_data_with_check(matched, einfo)) {
2306 *flags |= LDLM_FL_LVB_READY;
2307 /* We already have a lock, and it's referenced. */
2308 (*upcall)(cookie, &lockh, ELDLM_LOCK_MATCHED);
2309
2310 ldlm_lock_decref(&lockh, mode);
2311 LDLM_LOCK_PUT(matched);
2312 return ELDLM_OK;
2313 } else {
2314 ldlm_lock_decref(&lockh, mode);
2315 LDLM_LOCK_PUT(matched);
2316 }
2317 }
2318
2319 no_match:
2320 if (*flags & LDLM_FL_TEST_LOCK)
2321 return -ENOLCK;
2322 if (intent) {
2323 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
2324 &RQF_LDLM_ENQUEUE_LVB);
2325 if (!req)
2326 return -ENOMEM;
2327
2328 rc = ldlm_prep_enqueue_req(exp, req, NULL, 0);
2329 if (rc) {
2330 ptlrpc_request_free(req);
2331 return rc;
2332 }
2333
2334 req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
2335 sizeof(*lvb));
2336 ptlrpc_request_set_replen(req);
2337 }
2338
2339 /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
2340 *flags &= ~LDLM_FL_BLOCK_GRANTED;
2341
2342 rc = ldlm_cli_enqueue(exp, &req, einfo, res_id, policy, flags, lvb,
2343 sizeof(*lvb), LVB_T_OST, &lockh, async);
2344 if (async) {
2345 if (!rc) {
2346 struct osc_enqueue_args *aa;
2347
2348 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2349 aa = ptlrpc_req_async_args(req);
2350 aa->oa_exp = exp;
2351 aa->oa_mode = einfo->ei_mode;
2352 aa->oa_type = einfo->ei_type;
2353 lustre_handle_copy(&aa->oa_lockh, &lockh);
2354 aa->oa_upcall = upcall;
2355 aa->oa_cookie = cookie;
2356 aa->oa_agl = !!agl;
2357 if (!agl) {
2358 aa->oa_flags = flags;
2359 aa->oa_lvb = lvb;
2360 } else {
2361 /* AGL is essentially to enqueue an DLM lock
2362 * in advance, so we don't care about the
2363 * result of AGL enqueue.
2364 */
2365 aa->oa_lvb = NULL;
2366 aa->oa_flags = NULL;
2367 }
2368
2369 req->rq_interpret_reply =
2370 (ptlrpc_interpterer_t)osc_enqueue_interpret;
2371 if (rqset == PTLRPCD_SET)
2372 ptlrpcd_add_req(req);
2373 else
2374 ptlrpc_set_add_req(rqset, req);
2375 } else if (intent) {
2376 ptlrpc_req_finished(req);
2377 }
2378 return rc;
2379 }
2380
2381 rc = osc_enqueue_fini(req, upcall, cookie, &lockh, einfo->ei_mode,
2382 flags, agl, rc);
2383 if (intent)
2384 ptlrpc_req_finished(req);
2385
2386 return rc;
2387 }
2388
2389 int osc_match_base(struct obd_export *exp, struct ldlm_res_id *res_id,
2390 __u32 type, ldlm_policy_data_t *policy, __u32 mode,
2391 __u64 *flags, void *data, struct lustre_handle *lockh,
2392 int unref)
2393 {
2394 struct obd_device *obd = exp->exp_obd;
2395 __u64 lflags = *flags;
2396 enum ldlm_mode rc;
2397
2398 if (OBD_FAIL_CHECK(OBD_FAIL_OSC_MATCH))
2399 return -EIO;
2400
2401 /* Filesystem lock extents are extended to page boundaries so that
2402 * dealing with the page cache is a little smoother
2403 */
2404 policy->l_extent.start -= policy->l_extent.start & ~PAGE_MASK;
2405 policy->l_extent.end |= ~PAGE_MASK;
2406
2407 /* Next, search for already existing extent locks that will cover us */
2408 /* If we're trying to read, we also search for an existing PW lock. The
2409 * VFS and page cache already protect us locally, so lots of readers/
2410 * writers can share a single PW lock.
2411 */
2412 rc = mode;
2413 if (mode == LCK_PR)
2414 rc |= LCK_PW;
2415 rc = ldlm_lock_match(obd->obd_namespace, lflags,
2416 res_id, type, policy, rc, lockh, unref);
2417 if (rc) {
2418 if (data) {
2419 if (!osc_set_data_with_check(lockh, data)) {
2420 if (!(lflags & LDLM_FL_TEST_LOCK))
2421 ldlm_lock_decref(lockh, rc);
2422 return 0;
2423 }
2424 }
2425 if (!(lflags & LDLM_FL_TEST_LOCK) && mode != rc) {
2426 ldlm_lock_addref(lockh, LCK_PR);
2427 ldlm_lock_decref(lockh, LCK_PW);
2428 }
2429 return rc;
2430 }
2431 return rc;
2432 }
2433
2434 int osc_cancel_base(struct lustre_handle *lockh, __u32 mode)
2435 {
2436 if (unlikely(mode == LCK_GROUP))
2437 ldlm_lock_decref_and_cancel(lockh, mode);
2438 else
2439 ldlm_lock_decref(lockh, mode);
2440
2441 return 0;
2442 }
2443
2444 static int osc_statfs_interpret(const struct lu_env *env,
2445 struct ptlrpc_request *req,
2446 struct osc_async_args *aa, int rc)
2447 {
2448 struct obd_statfs *msfs;
2449
2450 if (rc == -EBADR)
2451 /* The request has in fact never been sent
2452 * due to issues at a higher level (LOV).
2453 * Exit immediately since the caller is
2454 * aware of the problem and takes care
2455 * of the clean up
2456 */
2457 return rc;
2458
2459 if ((rc == -ENOTCONN || rc == -EAGAIN) &&
2460 (aa->aa_oi->oi_flags & OBD_STATFS_NODELAY)) {
2461 rc = 0;
2462 goto out;
2463 }
2464
2465 if (rc != 0)
2466 goto out;
2467
2468 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2469 if (!msfs) {
2470 rc = -EPROTO;
2471 goto out;
2472 }
2473
2474 *aa->aa_oi->oi_osfs = *msfs;
2475 out:
2476 rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
2477 return rc;
2478 }
2479
2480 static int osc_statfs_async(struct obd_export *exp,
2481 struct obd_info *oinfo, __u64 max_age,
2482 struct ptlrpc_request_set *rqset)
2483 {
2484 struct obd_device *obd = class_exp2obd(exp);
2485 struct ptlrpc_request *req;
2486 struct osc_async_args *aa;
2487 int rc;
2488
2489 /* We could possibly pass max_age in the request (as an absolute
2490 * timestamp or a "seconds.usec ago") so the target can avoid doing
2491 * extra calls into the filesystem if that isn't necessary (e.g.
2492 * during mount that would help a bit). Having relative timestamps
2493 * is not so great if request processing is slow, while absolute
2494 * timestamps are not ideal because they need time synchronization.
2495 */
2496 req = ptlrpc_request_alloc(obd->u.cli.cl_import, &RQF_OST_STATFS);
2497 if (!req)
2498 return -ENOMEM;
2499
2500 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2501 if (rc) {
2502 ptlrpc_request_free(req);
2503 return rc;
2504 }
2505 ptlrpc_request_set_replen(req);
2506 req->rq_request_portal = OST_CREATE_PORTAL;
2507 ptlrpc_at_set_req_timeout(req);
2508
2509 if (oinfo->oi_flags & OBD_STATFS_NODELAY) {
2510 /* procfs requests not want stat in wait for avoid deadlock */
2511 req->rq_no_resend = 1;
2512 req->rq_no_delay = 1;
2513 }
2514
2515 req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_statfs_interpret;
2516 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2517 aa = ptlrpc_req_async_args(req);
2518 aa->aa_oi = oinfo;
2519
2520 ptlrpc_set_add_req(rqset, req);
2521 return 0;
2522 }
2523
2524 static int osc_statfs(const struct lu_env *env, struct obd_export *exp,
2525 struct obd_statfs *osfs, __u64 max_age, __u32 flags)
2526 {
2527 struct obd_device *obd = class_exp2obd(exp);
2528 struct obd_statfs *msfs;
2529 struct ptlrpc_request *req;
2530 struct obd_import *imp = NULL;
2531 int rc;
2532
2533 /* Since the request might also come from lprocfs, so we need
2534 * sync this with client_disconnect_export Bug15684
2535 */
2536 down_read(&obd->u.cli.cl_sem);
2537 if (obd->u.cli.cl_import)
2538 imp = class_import_get(obd->u.cli.cl_import);
2539 up_read(&obd->u.cli.cl_sem);
2540 if (!imp)
2541 return -ENODEV;
2542
2543 /* We could possibly pass max_age in the request (as an absolute
2544 * timestamp or a "seconds.usec ago") so the target can avoid doing
2545 * extra calls into the filesystem if that isn't necessary (e.g.
2546 * during mount that would help a bit). Having relative timestamps
2547 * is not so great if request processing is slow, while absolute
2548 * timestamps are not ideal because they need time synchronization.
2549 */
2550 req = ptlrpc_request_alloc(imp, &RQF_OST_STATFS);
2551
2552 class_import_put(imp);
2553
2554 if (!req)
2555 return -ENOMEM;
2556
2557 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
2558 if (rc) {
2559 ptlrpc_request_free(req);
2560 return rc;
2561 }
2562 ptlrpc_request_set_replen(req);
2563 req->rq_request_portal = OST_CREATE_PORTAL;
2564 ptlrpc_at_set_req_timeout(req);
2565
2566 if (flags & OBD_STATFS_NODELAY) {
2567 /* procfs requests not want stat in wait for avoid deadlock */
2568 req->rq_no_resend = 1;
2569 req->rq_no_delay = 1;
2570 }
2571
2572 rc = ptlrpc_queue_wait(req);
2573 if (rc)
2574 goto out;
2575
2576 msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
2577 if (!msfs) {
2578 rc = -EPROTO;
2579 goto out;
2580 }
2581
2582 *osfs = *msfs;
2583
2584 out:
2585 ptlrpc_req_finished(req);
2586 return rc;
2587 }
2588
2589 /* Retrieve object striping information.
2590 *
2591 * @lmmu is a pointer to an in-core struct with lmm_ost_count indicating
2592 * the maximum number of OST indices which will fit in the user buffer.
2593 * lmm_magic must be LOV_MAGIC (we only use 1 slot here).
2594 */
2595 static int osc_getstripe(struct lov_stripe_md *lsm,
2596 struct lov_user_md __user *lump)
2597 {
2598 /* we use lov_user_md_v3 because it is larger than lov_user_md_v1 */
2599 struct lov_user_md_v3 lum, *lumk;
2600 struct lov_user_ost_data_v1 *lmm_objects;
2601 int rc = 0, lum_size;
2602
2603 if (!lsm)
2604 return -ENODATA;
2605
2606 /* we only need the header part from user space to get lmm_magic and
2607 * lmm_stripe_count, (the header part is common to v1 and v3)
2608 */
2609 lum_size = sizeof(struct lov_user_md_v1);
2610 if (copy_from_user(&lum, lump, lum_size))
2611 return -EFAULT;
2612
2613 if ((lum.lmm_magic != LOV_USER_MAGIC_V1) &&
2614 (lum.lmm_magic != LOV_USER_MAGIC_V3))
2615 return -EINVAL;
2616
2617 /* lov_user_md_vX and lov_mds_md_vX must have the same size */
2618 LASSERT(sizeof(struct lov_user_md_v1) == sizeof(struct lov_mds_md_v1));
2619 LASSERT(sizeof(struct lov_user_md_v3) == sizeof(struct lov_mds_md_v3));
2620 LASSERT(sizeof(lum.lmm_objects[0]) == sizeof(lumk->lmm_objects[0]));
2621
2622 /* we can use lov_mds_md_size() to compute lum_size
2623 * because lov_user_md_vX and lov_mds_md_vX have the same size
2624 */
2625 if (lum.lmm_stripe_count > 0) {
2626 lum_size = lov_mds_md_size(lum.lmm_stripe_count, lum.lmm_magic);
2627 lumk = kzalloc(lum_size, GFP_NOFS);
2628 if (!lumk)
2629 return -ENOMEM;
2630
2631 if (lum.lmm_magic == LOV_USER_MAGIC_V1)
2632 lmm_objects =
2633 &(((struct lov_user_md_v1 *)lumk)->lmm_objects[0]);
2634 else
2635 lmm_objects = &(lumk->lmm_objects[0]);
2636 lmm_objects->l_ost_oi = lsm->lsm_oi;
2637 } else {
2638 lum_size = lov_mds_md_size(0, lum.lmm_magic);
2639 lumk = &lum;
2640 }
2641
2642 lumk->lmm_oi = lsm->lsm_oi;
2643 lumk->lmm_stripe_count = 1;
2644
2645 if (copy_to_user(lump, lumk, lum_size))
2646 rc = -EFAULT;
2647
2648 if (lumk != &lum)
2649 kfree(lumk);
2650
2651 return rc;
2652 }
2653
2654 static int osc_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
2655 void *karg, void __user *uarg)
2656 {
2657 struct obd_device *obd = exp->exp_obd;
2658 struct obd_ioctl_data *data = karg;
2659 int err = 0;
2660
2661 if (!try_module_get(THIS_MODULE)) {
2662 CERROR("%s: cannot get module '%s'\n", obd->obd_name,
2663 module_name(THIS_MODULE));
2664 return -EINVAL;
2665 }
2666 switch (cmd) {
2667 case OBD_IOC_LOV_GET_CONFIG: {
2668 char *buf;
2669 struct lov_desc *desc;
2670 struct obd_uuid uuid;
2671
2672 buf = NULL;
2673 len = 0;
2674 if (obd_ioctl_getdata(&buf, &len, uarg)) {
2675 err = -EINVAL;
2676 goto out;
2677 }
2678
2679 data = (struct obd_ioctl_data *)buf;
2680
2681 if (sizeof(*desc) > data->ioc_inllen1) {
2682 obd_ioctl_freedata(buf, len);
2683 err = -EINVAL;
2684 goto out;
2685 }
2686
2687 if (data->ioc_inllen2 < sizeof(uuid)) {
2688 obd_ioctl_freedata(buf, len);
2689 err = -EINVAL;
2690 goto out;
2691 }
2692
2693 desc = (struct lov_desc *)data->ioc_inlbuf1;
2694 desc->ld_tgt_count = 1;
2695 desc->ld_active_tgt_count = 1;
2696 desc->ld_default_stripe_count = 1;
2697 desc->ld_default_stripe_size = 0;
2698 desc->ld_default_stripe_offset = 0;
2699 desc->ld_pattern = 0;
2700 memcpy(&desc->ld_uuid, &obd->obd_uuid, sizeof(uuid));
2701
2702 memcpy(data->ioc_inlbuf2, &obd->obd_uuid, sizeof(uuid));
2703
2704 err = copy_to_user(uarg, buf, len);
2705 if (err)
2706 err = -EFAULT;
2707 obd_ioctl_freedata(buf, len);
2708 goto out;
2709 }
2710 case LL_IOC_LOV_SETSTRIPE:
2711 err = obd_alloc_memmd(exp, karg);
2712 if (err > 0)
2713 err = 0;
2714 goto out;
2715 case LL_IOC_LOV_GETSTRIPE:
2716 err = osc_getstripe(karg, uarg);
2717 goto out;
2718 case OBD_IOC_CLIENT_RECOVER:
2719 err = ptlrpc_recover_import(obd->u.cli.cl_import,
2720 data->ioc_inlbuf1, 0);
2721 if (err > 0)
2722 err = 0;
2723 goto out;
2724 case IOC_OSC_SET_ACTIVE:
2725 err = ptlrpc_set_import_active(obd->u.cli.cl_import,
2726 data->ioc_offset);
2727 goto out;
2728 case OBD_IOC_POLL_QUOTACHECK:
2729 err = osc_quota_poll_check(exp, karg);
2730 goto out;
2731 case OBD_IOC_PING_TARGET:
2732 err = ptlrpc_obd_ping(obd);
2733 goto out;
2734 default:
2735 CDEBUG(D_INODE, "unrecognised ioctl %#x by %s\n",
2736 cmd, current_comm());
2737 err = -ENOTTY;
2738 goto out;
2739 }
2740 out:
2741 module_put(THIS_MODULE);
2742 return err;
2743 }
2744
2745 static int osc_get_info(const struct lu_env *env, struct obd_export *exp,
2746 u32 keylen, void *key, __u32 *vallen, void *val,
2747 struct lov_stripe_md *lsm)
2748 {
2749 if (!vallen || !val)
2750 return -EFAULT;
2751
2752 if (KEY_IS(KEY_LOCK_TO_STRIPE)) {
2753 __u32 *stripe = val;
2754 *vallen = sizeof(*stripe);
2755 *stripe = 0;
2756 return 0;
2757 } else if (KEY_IS(KEY_LAST_ID)) {
2758 struct ptlrpc_request *req;
2759 u64 *reply;
2760 char *tmp;
2761 int rc;
2762
2763 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
2764 &RQF_OST_GET_INFO_LAST_ID);
2765 if (!req)
2766 return -ENOMEM;
2767
2768 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
2769 RCL_CLIENT, keylen);
2770 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
2771 if (rc) {
2772 ptlrpc_request_free(req);
2773 return rc;
2774 }
2775
2776 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
2777 memcpy(tmp, key, keylen);
2778
2779 req->rq_no_delay = 1;
2780 req->rq_no_resend = 1;
2781 ptlrpc_request_set_replen(req);
2782 rc = ptlrpc_queue_wait(req);
2783 if (rc)
2784 goto out;
2785
2786 reply = req_capsule_server_get(&req->rq_pill, &RMF_OBD_ID);
2787 if (!reply) {
2788 rc = -EPROTO;
2789 goto out;
2790 }
2791
2792 *((u64 *)val) = *reply;
2793 out:
2794 ptlrpc_req_finished(req);
2795 return rc;
2796 } else if (KEY_IS(KEY_FIEMAP)) {
2797 struct ll_fiemap_info_key *fm_key = key;
2798 struct ldlm_res_id res_id;
2799 ldlm_policy_data_t policy;
2800 struct lustre_handle lockh;
2801 enum ldlm_mode mode = 0;
2802 struct ptlrpc_request *req;
2803 struct ll_user_fiemap *reply;
2804 char *tmp;
2805 int rc;
2806
2807 if (!(fm_key->fiemap.fm_flags & FIEMAP_FLAG_SYNC))
2808 goto skip_locking;
2809
2810 policy.l_extent.start = fm_key->fiemap.fm_start &
2811 PAGE_MASK;
2812
2813 if (OBD_OBJECT_EOF - fm_key->fiemap.fm_length <=
2814 fm_key->fiemap.fm_start + PAGE_SIZE - 1)
2815 policy.l_extent.end = OBD_OBJECT_EOF;
2816 else
2817 policy.l_extent.end = (fm_key->fiemap.fm_start +
2818 fm_key->fiemap.fm_length +
2819 PAGE_SIZE - 1) & PAGE_MASK;
2820
2821 ostid_build_res_name(&fm_key->oa.o_oi, &res_id);
2822 mode = ldlm_lock_match(exp->exp_obd->obd_namespace,
2823 LDLM_FL_BLOCK_GRANTED |
2824 LDLM_FL_LVB_READY,
2825 &res_id, LDLM_EXTENT, &policy,
2826 LCK_PR | LCK_PW, &lockh, 0);
2827 if (mode) { /* lock is cached on client */
2828 if (mode != LCK_PR) {
2829 ldlm_lock_addref(&lockh, LCK_PR);
2830 ldlm_lock_decref(&lockh, LCK_PW);
2831 }
2832 } else { /* no cached lock, needs acquire lock on server side */
2833 fm_key->oa.o_valid |= OBD_MD_FLFLAGS;
2834 fm_key->oa.o_flags |= OBD_FL_SRVLOCK;
2835 }
2836
2837 skip_locking:
2838 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
2839 &RQF_OST_GET_INFO_FIEMAP);
2840 if (!req) {
2841 rc = -ENOMEM;
2842 goto drop_lock;
2843 }
2844
2845 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_KEY,
2846 RCL_CLIENT, keylen);
2847 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
2848 RCL_CLIENT, *vallen);
2849 req_capsule_set_size(&req->rq_pill, &RMF_FIEMAP_VAL,
2850 RCL_SERVER, *vallen);
2851
2852 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
2853 if (rc) {
2854 ptlrpc_request_free(req);
2855 goto drop_lock;
2856 }
2857
2858 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_KEY);
2859 memcpy(tmp, key, keylen);
2860 tmp = req_capsule_client_get(&req->rq_pill, &RMF_FIEMAP_VAL);
2861 memcpy(tmp, val, *vallen);
2862
2863 ptlrpc_request_set_replen(req);
2864 rc = ptlrpc_queue_wait(req);
2865 if (rc)
2866 goto fini_req;
2867
2868 reply = req_capsule_server_get(&req->rq_pill, &RMF_FIEMAP_VAL);
2869 if (!reply) {
2870 rc = -EPROTO;
2871 goto fini_req;
2872 }
2873
2874 memcpy(val, reply, *vallen);
2875 fini_req:
2876 ptlrpc_req_finished(req);
2877 drop_lock:
2878 if (mode)
2879 ldlm_lock_decref(&lockh, LCK_PR);
2880 return rc;
2881 }
2882
2883 return -EINVAL;
2884 }
2885
2886 static int osc_set_info_async(const struct lu_env *env, struct obd_export *exp,
2887 u32 keylen, void *key, u32 vallen,
2888 void *val, struct ptlrpc_request_set *set)
2889 {
2890 struct ptlrpc_request *req;
2891 struct obd_device *obd = exp->exp_obd;
2892 struct obd_import *imp = class_exp2cliimp(exp);
2893 char *tmp;
2894 int rc;
2895
2896 OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_SHUTDOWN, 10);
2897
2898 if (KEY_IS(KEY_CHECKSUM)) {
2899 if (vallen != sizeof(int))
2900 return -EINVAL;
2901 exp->exp_obd->u.cli.cl_checksum = (*(int *)val) ? 1 : 0;
2902 return 0;
2903 }
2904
2905 if (KEY_IS(KEY_SPTLRPC_CONF)) {
2906 sptlrpc_conf_client_adapt(obd);
2907 return 0;
2908 }
2909
2910 if (KEY_IS(KEY_FLUSH_CTX)) {
2911 sptlrpc_import_flush_my_ctx(imp);
2912 return 0;
2913 }
2914
2915 if (KEY_IS(KEY_CACHE_SET)) {
2916 struct client_obd *cli = &obd->u.cli;
2917
2918 LASSERT(!cli->cl_cache); /* only once */
2919 cli->cl_cache = val;
2920 cl_cache_incref(cli->cl_cache);
2921 cli->cl_lru_left = &cli->cl_cache->ccc_lru_left;
2922
2923 /* add this osc into entity list */
2924 LASSERT(list_empty(&cli->cl_lru_osc));
2925 spin_lock(&cli->cl_cache->ccc_lru_lock);
2926 list_add(&cli->cl_lru_osc, &cli->cl_cache->ccc_lru);
2927 spin_unlock(&cli->cl_cache->ccc_lru_lock);
2928
2929 return 0;
2930 }
2931
2932 if (KEY_IS(KEY_CACHE_LRU_SHRINK)) {
2933 struct client_obd *cli = &obd->u.cli;
2934 int nr = atomic_read(&cli->cl_lru_in_list) >> 1;
2935 int target = *(int *)val;
2936
2937 nr = osc_lru_shrink(env, cli, min(nr, target), true);
2938 *(int *)val -= nr;
2939 return 0;
2940 }
2941
2942 if (!set && !KEY_IS(KEY_GRANT_SHRINK))
2943 return -EINVAL;
2944
2945 /* We pass all other commands directly to OST. Since nobody calls osc
2946 * methods directly and everybody is supposed to go through LOV, we
2947 * assume lov checked invalid values for us.
2948 * The only recognised values so far are evict_by_nid and mds_conn.
2949 * Even if something bad goes through, we'd get a -EINVAL from OST
2950 * anyway.
2951 */
2952
2953 req = ptlrpc_request_alloc(imp, KEY_IS(KEY_GRANT_SHRINK) ?
2954 &RQF_OST_SET_GRANT_INFO :
2955 &RQF_OBD_SET_INFO);
2956 if (!req)
2957 return -ENOMEM;
2958
2959 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_KEY,
2960 RCL_CLIENT, keylen);
2961 if (!KEY_IS(KEY_GRANT_SHRINK))
2962 req_capsule_set_size(&req->rq_pill, &RMF_SETINFO_VAL,
2963 RCL_CLIENT, vallen);
2964 rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SET_INFO);
2965 if (rc) {
2966 ptlrpc_request_free(req);
2967 return rc;
2968 }
2969
2970 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
2971 memcpy(tmp, key, keylen);
2972 tmp = req_capsule_client_get(&req->rq_pill, KEY_IS(KEY_GRANT_SHRINK) ?
2973 &RMF_OST_BODY :
2974 &RMF_SETINFO_VAL);
2975 memcpy(tmp, val, vallen);
2976
2977 if (KEY_IS(KEY_GRANT_SHRINK)) {
2978 struct osc_brw_async_args *aa;
2979 struct obdo *oa;
2980
2981 CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2982 aa = ptlrpc_req_async_args(req);
2983 oa = kmem_cache_zalloc(obdo_cachep, GFP_NOFS);
2984 if (!oa) {
2985 ptlrpc_req_finished(req);
2986 return -ENOMEM;
2987 }
2988 *oa = ((struct ost_body *)val)->oa;
2989 aa->aa_oa = oa;
2990 req->rq_interpret_reply = osc_shrink_grant_interpret;
2991 }
2992
2993 ptlrpc_request_set_replen(req);
2994 if (!KEY_IS(KEY_GRANT_SHRINK)) {
2995 LASSERT(set);
2996 ptlrpc_set_add_req(set, req);
2997 ptlrpc_check_set(NULL, set);
2998 } else {
2999 ptlrpcd_add_req(req);
3000 }
3001
3002 return 0;
3003 }
3004
3005 static int osc_reconnect(const struct lu_env *env,
3006 struct obd_export *exp, struct obd_device *obd,
3007 struct obd_uuid *cluuid,
3008 struct obd_connect_data *data,
3009 void *localdata)
3010 {
3011 struct client_obd *cli = &obd->u.cli;
3012
3013 if (data && (data->ocd_connect_flags & OBD_CONNECT_GRANT)) {
3014 long lost_grant;
3015
3016 spin_lock(&cli->cl_loi_list_lock);
3017 data->ocd_grant = (cli->cl_avail_grant + cli->cl_dirty) ?:
3018 2 * cli_brw_size(obd);
3019 lost_grant = cli->cl_lost_grant;
3020 cli->cl_lost_grant = 0;
3021 spin_unlock(&cli->cl_loi_list_lock);
3022
3023 CDEBUG(D_RPCTRACE, "ocd_connect_flags: %#llx ocd_version: %d ocd_grant: %d, lost: %ld.\n",
3024 data->ocd_connect_flags,
3025 data->ocd_version, data->ocd_grant, lost_grant);
3026 }
3027
3028 return 0;
3029 }
3030
3031 static int osc_disconnect(struct obd_export *exp)
3032 {
3033 struct obd_device *obd = class_exp2obd(exp);
3034 int rc;
3035
3036 rc = client_disconnect_export(exp);
3037 /**
3038 * Initially we put del_shrink_grant before disconnect_export, but it
3039 * causes the following problem if setup (connect) and cleanup
3040 * (disconnect) are tangled together.
3041 * connect p1 disconnect p2
3042 * ptlrpc_connect_import
3043 * ............... class_manual_cleanup
3044 * osc_disconnect
3045 * del_shrink_grant
3046 * ptlrpc_connect_interrupt
3047 * init_grant_shrink
3048 * add this client to shrink list
3049 * cleanup_osc
3050 * Bang! pinger trigger the shrink.
3051 * So the osc should be disconnected from the shrink list, after we
3052 * are sure the import has been destroyed. BUG18662
3053 */
3054 if (!obd->u.cli.cl_import)
3055 osc_del_shrink_grant(&obd->u.cli);
3056 return rc;
3057 }
3058
3059 static int osc_import_event(struct obd_device *obd,
3060 struct obd_import *imp,
3061 enum obd_import_event event)
3062 {
3063 struct client_obd *cli;
3064 int rc = 0;
3065
3066 LASSERT(imp->imp_obd == obd);
3067
3068 switch (event) {
3069 case IMP_EVENT_DISCON: {
3070 cli = &obd->u.cli;
3071 spin_lock(&cli->cl_loi_list_lock);
3072 cli->cl_avail_grant = 0;
3073 cli->cl_lost_grant = 0;
3074 spin_unlock(&cli->cl_loi_list_lock);
3075 break;
3076 }
3077 case IMP_EVENT_INACTIVE: {
3078 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_INACTIVE, NULL);
3079 break;
3080 }
3081 case IMP_EVENT_INVALIDATE: {
3082 struct ldlm_namespace *ns = obd->obd_namespace;
3083 struct lu_env *env;
3084 int refcheck;
3085
3086 env = cl_env_get(&refcheck);
3087 if (!IS_ERR(env)) {
3088 /* Reset grants */
3089 cli = &obd->u.cli;
3090 /* all pages go to failing rpcs due to the invalid
3091 * import
3092 */
3093 osc_io_unplug(env, cli, NULL);
3094
3095 ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
3096 cl_env_put(env, &refcheck);
3097 } else {
3098 rc = PTR_ERR(env);
3099 }
3100 break;
3101 }
3102 case IMP_EVENT_ACTIVE: {
3103 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVE, NULL);
3104 break;
3105 }
3106 case IMP_EVENT_OCD: {
3107 struct obd_connect_data *ocd = &imp->imp_connect_data;
3108
3109 if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT)
3110 osc_init_grant(&obd->u.cli, ocd);
3111
3112 /* See bug 7198 */
3113 if (ocd->ocd_connect_flags & OBD_CONNECT_REQPORTAL)
3114 imp->imp_client->cli_request_portal = OST_REQUEST_PORTAL;
3115
3116 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_OCD, NULL);
3117 break;
3118 }
3119 case IMP_EVENT_DEACTIVATE: {
3120 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_DEACTIVATE, NULL);
3121 break;
3122 }
3123 case IMP_EVENT_ACTIVATE: {
3124 rc = obd_notify_observer(obd, obd, OBD_NOTIFY_ACTIVATE, NULL);
3125 break;
3126 }
3127 default:
3128 CERROR("Unknown import event %d\n", event);
3129 LBUG();
3130 }
3131 return rc;
3132 }
3133
3134 /**
3135 * Determine whether the lock can be canceled before replaying the lock
3136 * during recovery, see bug16774 for detailed information.
3137 *
3138 * \retval zero the lock can't be canceled
3139 * \retval other ok to cancel
3140 */
3141 static int osc_cancel_weight(struct ldlm_lock *lock)
3142 {
3143 /*
3144 * Cancel all unused and granted extent lock.
3145 */
3146 if (lock->l_resource->lr_type == LDLM_EXTENT &&
3147 lock->l_granted_mode == lock->l_req_mode &&
3148 osc_ldlm_weigh_ast(lock) == 0)
3149 return 1;
3150
3151 return 0;
3152 }
3153
3154 static int brw_queue_work(const struct lu_env *env, void *data)
3155 {
3156 struct client_obd *cli = data;
3157
3158 CDEBUG(D_CACHE, "Run writeback work for client obd %p.\n", cli);
3159
3160 osc_io_unplug(env, cli, NULL);
3161 return 0;
3162 }
3163
3164 int osc_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
3165 {
3166 struct lprocfs_static_vars lvars = { NULL };
3167 struct client_obd *cli = &obd->u.cli;
3168 void *handler;
3169 int rc;
3170 int adding;
3171 int added;
3172 int req_count;
3173
3174 rc = ptlrpcd_addref();
3175 if (rc)
3176 return rc;
3177
3178 rc = client_obd_setup(obd, lcfg);
3179 if (rc)
3180 goto out_ptlrpcd;
3181
3182 handler = ptlrpcd_alloc_work(cli->cl_import, brw_queue_work, cli);
3183 if (IS_ERR(handler)) {
3184 rc = PTR_ERR(handler);
3185 goto out_client_setup;
3186 }
3187 cli->cl_writeback_work = handler;
3188
3189 handler = ptlrpcd_alloc_work(cli->cl_import, lru_queue_work, cli);
3190 if (IS_ERR(handler)) {
3191 rc = PTR_ERR(handler);
3192 goto out_ptlrpcd_work;
3193 }
3194
3195 cli->cl_lru_work = handler;
3196
3197 rc = osc_quota_setup(obd);
3198 if (rc)
3199 goto out_ptlrpcd_work;
3200
3201 cli->cl_grant_shrink_interval = GRANT_SHRINK_INTERVAL;
3202 lprocfs_osc_init_vars(&lvars);
3203 if (lprocfs_obd_setup(obd, lvars.obd_vars, lvars.sysfs_vars) == 0) {
3204 lproc_osc_attach_seqstat(obd);
3205 sptlrpc_lprocfs_cliobd_attach(obd);
3206 ptlrpc_lprocfs_register_obd(obd);
3207 }
3208
3209 /*
3210 * We try to control the total number of requests with a upper limit
3211 * osc_reqpool_maxreqcount. There might be some race which will cause
3212 * over-limit allocation, but it is fine.
3213 */
3214 req_count = atomic_read(&osc_pool_req_count);
3215 if (req_count < osc_reqpool_maxreqcount) {
3216 adding = cli->cl_max_rpcs_in_flight + 2;
3217 if (req_count + adding > osc_reqpool_maxreqcount)
3218 adding = osc_reqpool_maxreqcount - req_count;
3219
3220 added = ptlrpc_add_rqs_to_pool(osc_rq_pool, adding);
3221 atomic_add(added, &osc_pool_req_count);
3222 }
3223
3224 INIT_LIST_HEAD(&cli->cl_grant_shrink_list);
3225 ns_register_cancel(obd->obd_namespace, osc_cancel_weight);
3226 return rc;
3227
3228 out_ptlrpcd_work:
3229 if (cli->cl_writeback_work) {
3230 ptlrpcd_destroy_work(cli->cl_writeback_work);
3231 cli->cl_writeback_work = NULL;
3232 }
3233 if (cli->cl_lru_work) {
3234 ptlrpcd_destroy_work(cli->cl_lru_work);
3235 cli->cl_lru_work = NULL;
3236 }
3237 out_client_setup:
3238 client_obd_cleanup(obd);
3239 out_ptlrpcd:
3240 ptlrpcd_decref();
3241 return rc;
3242 }
3243
3244 static int osc_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
3245 {
3246 switch (stage) {
3247 case OBD_CLEANUP_EARLY: {
3248 struct obd_import *imp;
3249
3250 imp = obd->u.cli.cl_import;
3251 CDEBUG(D_HA, "Deactivating import %s\n", obd->obd_name);
3252 /* ptlrpc_abort_inflight to stop an mds_lov_synchronize */
3253 ptlrpc_deactivate_import(imp);
3254 spin_lock(&imp->imp_lock);
3255 imp->imp_pingable = 0;
3256 spin_unlock(&imp->imp_lock);
3257 break;
3258 }
3259 case OBD_CLEANUP_EXPORTS: {
3260 struct client_obd *cli = &obd->u.cli;
3261 /* LU-464
3262 * for echo client, export may be on zombie list, wait for
3263 * zombie thread to cull it, because cli.cl_import will be
3264 * cleared in client_disconnect_export():
3265 * class_export_destroy() -> obd_cleanup() ->
3266 * echo_device_free() -> echo_client_cleanup() ->
3267 * obd_disconnect() -> osc_disconnect() ->
3268 * client_disconnect_export()
3269 */
3270 obd_zombie_barrier();
3271 if (cli->cl_writeback_work) {
3272 ptlrpcd_destroy_work(cli->cl_writeback_work);
3273 cli->cl_writeback_work = NULL;
3274 }
3275 if (cli->cl_lru_work) {
3276 ptlrpcd_destroy_work(cli->cl_lru_work);
3277 cli->cl_lru_work = NULL;
3278 }
3279 obd_cleanup_client_import(obd);
3280 ptlrpc_lprocfs_unregister_obd(obd);
3281 lprocfs_obd_cleanup(obd);
3282 break;
3283 }
3284 }
3285 return 0;
3286 }
3287
3288 static int osc_cleanup(struct obd_device *obd)
3289 {
3290 struct client_obd *cli = &obd->u.cli;
3291 int rc;
3292
3293 /* lru cleanup */
3294 if (cli->cl_cache) {
3295 LASSERT(atomic_read(&cli->cl_cache->ccc_users) > 0);
3296 spin_lock(&cli->cl_cache->ccc_lru_lock);
3297 list_del_init(&cli->cl_lru_osc);
3298 spin_unlock(&cli->cl_cache->ccc_lru_lock);
3299 cli->cl_lru_left = NULL;
3300 cl_cache_decref(cli->cl_cache);
3301 cli->cl_cache = NULL;
3302 }
3303
3304 /* free memory of osc quota cache */
3305 osc_quota_cleanup(obd);
3306
3307 rc = client_obd_cleanup(obd);
3308
3309 ptlrpcd_decref();
3310 return rc;
3311 }
3312
3313 int osc_process_config_base(struct obd_device *obd, struct lustre_cfg *lcfg)
3314 {
3315 struct lprocfs_static_vars lvars = { NULL };
3316 int rc = 0;
3317
3318 lprocfs_osc_init_vars(&lvars);
3319
3320 switch (lcfg->lcfg_command) {
3321 default:
3322 rc = class_process_proc_param(PARAM_OSC, lvars.obd_vars,
3323 lcfg, obd);
3324 if (rc > 0)
3325 rc = 0;
3326 break;
3327 }
3328
3329 return rc;
3330 }
3331
3332 static int osc_process_config(struct obd_device *obd, u32 len, void *buf)
3333 {
3334 return osc_process_config_base(obd, buf);
3335 }
3336
3337 static struct obd_ops osc_obd_ops = {
3338 .owner = THIS_MODULE,
3339 .setup = osc_setup,
3340 .precleanup = osc_precleanup,
3341 .cleanup = osc_cleanup,
3342 .add_conn = client_import_add_conn,
3343 .del_conn = client_import_del_conn,
3344 .connect = client_connect_import,
3345 .reconnect = osc_reconnect,
3346 .disconnect = osc_disconnect,
3347 .statfs = osc_statfs,
3348 .statfs_async = osc_statfs_async,
3349 .packmd = osc_packmd,
3350 .unpackmd = osc_unpackmd,
3351 .create = osc_create,
3352 .destroy = osc_destroy,
3353 .getattr = osc_getattr,
3354 .getattr_async = osc_getattr_async,
3355 .setattr = osc_setattr,
3356 .setattr_async = osc_setattr_async,
3357 .find_cbdata = osc_find_cbdata,
3358 .iocontrol = osc_iocontrol,
3359 .get_info = osc_get_info,
3360 .set_info_async = osc_set_info_async,
3361 .import_event = osc_import_event,
3362 .process_config = osc_process_config,
3363 .quotactl = osc_quotactl,
3364 .quotacheck = osc_quotacheck,
3365 };
3366
3367 extern struct lu_kmem_descr osc_caches[];
3368 extern struct lock_class_key osc_ast_guard_class;
3369
3370 static int __init osc_init(void)
3371 {
3372 struct lprocfs_static_vars lvars = { NULL };
3373 unsigned int reqpool_size;
3374 unsigned int reqsize;
3375 int rc;
3376
3377 /* print an address of _any_ initialized kernel symbol from this
3378 * module, to allow debugging with gdb that doesn't support data
3379 * symbols from modules.
3380 */
3381 CDEBUG(D_INFO, "Lustre OSC module (%p).\n", &osc_caches);
3382
3383 rc = lu_kmem_init(osc_caches);
3384 if (rc)
3385 return rc;
3386
3387 lprocfs_osc_init_vars(&lvars);
3388
3389 rc = class_register_type(&osc_obd_ops, NULL,
3390 LUSTRE_OSC_NAME, &osc_device_type);
3391 if (rc)
3392 goto out_kmem;
3393
3394 /* This is obviously too much memory, only prevent overflow here */
3395 if (osc_reqpool_mem_max >= 1 << 12 || osc_reqpool_mem_max == 0) {
3396 rc = -EINVAL;
3397 goto out_type;
3398 }
3399
3400 reqpool_size = osc_reqpool_mem_max << 20;
3401
3402 reqsize = 1;
3403 while (reqsize < OST_MAXREQSIZE)
3404 reqsize = reqsize << 1;
3405
3406 /*
3407 * We don't enlarge the request count in OSC pool according to
3408 * cl_max_rpcs_in_flight. The allocation from the pool will only be
3409 * tried after normal allocation failed. So a small OSC pool won't
3410 * cause much performance degression in most of cases.
3411 */
3412 osc_reqpool_maxreqcount = reqpool_size / reqsize;
3413
3414 atomic_set(&osc_pool_req_count, 0);
3415 osc_rq_pool = ptlrpc_init_rq_pool(0, OST_MAXREQSIZE,
3416 ptlrpc_add_rqs_to_pool);
3417
3418 if (osc_rq_pool)
3419 return 0;
3420
3421 rc = -ENOMEM;
3422
3423 out_type:
3424 class_unregister_type(LUSTRE_OSC_NAME);
3425 out_kmem:
3426 lu_kmem_fini(osc_caches);
3427 return rc;
3428 }
3429
3430 static void /*__exit*/ osc_exit(void)
3431 {
3432 class_unregister_type(LUSTRE_OSC_NAME);
3433 lu_kmem_fini(osc_caches);
3434 ptlrpc_free_rq_pool(osc_rq_pool);
3435 }
3436
3437 MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
3438 MODULE_DESCRIPTION("Lustre Object Storage Client (OSC)");
3439 MODULE_LICENSE("GPL");
3440 MODULE_VERSION(LUSTRE_VERSION_STRING);
3441
3442 module_init(osc_init);
3443 module_exit(osc_exit);
Linux kernel - Deliverables
This page took 0.172483 seconds and 5 git commands to generate.