Whamcloud - gitweb
Branch: HEAD
[fs/lustre-release.git] / lustre / osc / osc_request.c
1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2  * vim:expandtab:shiftwidth=8:tabstop=8:
3  *
4  * GPL HEADER START
5  *
6  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
7  *
8  * This program is free software; you can redistribute it and/or modify
9  * it under the terms of the GNU General Public License version 2 only,
10  * as published by the Free Software Foundation.
11  *
12  * This program is distributed in the hope that it will be useful, but
13  * WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15  * General Public License version 2 for more details (a copy is included
16  * in the LICENSE file that accompanied this code).
17  *
18  * You should have received a copy of the GNU General Public License
19  * version 2 along with this program; If not, see
20  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
21  *
22  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23  * CA 95054 USA or visit www.sun.com if you need additional information or
24  * have any questions.
25  *
26  * GPL HEADER END
27  */
28 /*
29  * Copyright  2008 Sun Microsystems, Inc. All rights reserved
30  * Use is subject to license terms.
31  */
32 /*
33  * This file is part of Lustre, http://www.lustre.org/
34  * Lustre is a trademark of Sun Microsystems, Inc.
35  */
36
37 #ifndef EXPORT_SYMTAB
38 # define EXPORT_SYMTAB
39 #endif
40 #define DEBUG_SUBSYSTEM S_OSC
41
42 #include <libcfs/libcfs.h>
43
44 #ifndef __KERNEL__
45 # include <liblustre.h>
46 #endif
47
48 #include <lustre_dlm.h>
49 #include <lustre_net.h>
50 #include <lustre/lustre_user.h>
51 #include <obd_cksum.h>
52 #include <obd_ost.h>
53 #include <obd_lov.h>
54
55 #ifdef  __CYGWIN__
56 # include <ctype.h>
57 #endif
58
59 #include <lustre_ha.h>
60 #include <lprocfs_status.h>
61 #include <lustre_log.h>
62 #include <lustre_debug.h>
63 #include <lustre_param.h>
64 #include "osc_internal.h"
65
66 static quota_interface_t *quota_interface = NULL;
67 extern quota_interface_t osc_quota_interface;
68
69 static void osc_release_ppga(struct brw_page **ppga, obd_count count);
70 static int brw_interpret(const struct lu_env *env,
71                          struct ptlrpc_request *req, void *data, int rc);
72 int osc_cleanup(struct obd_device *obd);
73
74 /* Pack OSC object metadata for disk storage (LE byte order). */
75 static int osc_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
76                       struct lov_stripe_md *lsm)
77 {
78         int lmm_size;
79         ENTRY;
80
81         lmm_size = sizeof(**lmmp);
82         if (!lmmp)
83                 RETURN(lmm_size);
84
85         if (*lmmp && !lsm) {
86                 OBD_FREE(*lmmp, lmm_size);
87                 *lmmp = NULL;
88                 RETURN(0);
89         }
90
91         if (!*lmmp) {
92                 OBD_ALLOC(*lmmp, lmm_size);
93                 if (!*lmmp)
94                         RETURN(-ENOMEM);
95         }
96
97         if (lsm) {
98                 LASSERT(lsm->lsm_object_id);
99                 LASSERT_MDS_GROUP(lsm->lsm_object_gr);
100                 (*lmmp)->lmm_object_id = cpu_to_le64(lsm->lsm_object_id);
101                 (*lmmp)->lmm_object_gr = cpu_to_le64(lsm->lsm_object_gr);
102         }
103
104         RETURN(lmm_size);
105 }
106
107 /* Unpack OSC object metadata from disk storage (LE byte order). */
108 static int osc_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
109                         struct lov_mds_md *lmm, int lmm_bytes)
110 {
111         int lsm_size;
112         ENTRY;
113
114         if (lmm != NULL) {
115                 if (lmm_bytes < sizeof (*lmm)) {
116                         CERROR("lov_mds_md too small: %d, need %d\n",
117                                lmm_bytes, (int)sizeof(*lmm));
118                         RETURN(-EINVAL);
119                 }
120                 /* XXX LOV_MAGIC etc check? */
121
122                 if (lmm->lmm_object_id == 0) {
123                         CERROR("lov_mds_md: zero lmm_object_id\n");
124                         RETURN(-EINVAL);
125                 }
126         }
127
128         lsm_size = lov_stripe_md_size(1);
129         if (lsmp == NULL)
130                 RETURN(lsm_size);
131
132         if (*lsmp != NULL && lmm == NULL) {
133                 OBD_FREE((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
134                 OBD_FREE(*lsmp, lsm_size);
135                 *lsmp = NULL;
136                 RETURN(0);
137         }
138
139         if (*lsmp == NULL) {
140                 OBD_ALLOC(*lsmp, lsm_size);
141                 if (*lsmp == NULL)
142                         RETURN(-ENOMEM);
143                 OBD_ALLOC((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
144                 if ((*lsmp)->lsm_oinfo[0] == NULL) {
145                         OBD_FREE(*lsmp, lsm_size);
146                         RETURN(-ENOMEM);
147                 }
148                 loi_init((*lsmp)->lsm_oinfo[0]);
149         }
150
151         if (lmm != NULL) {
152                 /* XXX zero *lsmp? */
153                 (*lsmp)->lsm_object_id = le64_to_cpu (lmm->lmm_object_id);
154                 (*lsmp)->lsm_object_gr = le64_to_cpu (lmm->lmm_object_gr);
155                 LASSERT((*lsmp)->lsm_object_id);
156                 LASSERT_MDS_GROUP((*lsmp)->lsm_object_gr);
157         }
158
159         (*lsmp)->lsm_maxbytes = LUSTRE_STRIPE_MAXBYTES;
160
161         RETURN(lsm_size);
162 }
163
164 static inline void osc_pack_capa(struct ptlrpc_request *req,
165                                  struct ost_body *body, void *capa)
166 {
167         struct obd_capa *oc = (struct obd_capa *)capa;
168         struct lustre_capa *c;
169
170         if (!capa)
171                 return;
172
173         c = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
174         LASSERT(c);
175         capa_cpy(c, oc);
176         body->oa.o_valid |= OBD_MD_FLOSSCAPA;
177         DEBUG_CAPA(D_SEC, c, "pack");
178 }
179
180 static inline void osc_pack_req_body(struct ptlrpc_request *req,
181                                      struct obd_info *oinfo)
182 {
183         struct ost_body *body;
184
185         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
186         LASSERT(body);
187
188         body->oa = *oinfo->oi_oa;
189         osc_pack_capa(req, body, oinfo->oi_capa);
190 }
191
192 static inline void osc_set_capa_size(struct ptlrpc_request *req,
193                                      const struct req_msg_field *field,
194                                      struct obd_capa *oc)
195 {
196         if (oc == NULL)
197                 req_capsule_set_size(&req->rq_pill, field, RCL_CLIENT, 0);
198         else
199                 /* it is already calculated as sizeof struct obd_capa */
200                 ;
201 }
202
203 static int osc_getattr_interpret(const struct lu_env *env,
204                                  struct ptlrpc_request *req,
205                                  struct osc_async_args *aa, int rc)
206 {
207         struct ost_body *body;
208         ENTRY;
209
210         if (rc != 0)
211                 GOTO(out, rc);
212
213         body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
214                                   lustre_swab_ost_body);
215         if (body) {
216                 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
217                 memcpy(aa->aa_oi->oi_oa, &body->oa, sizeof(*aa->aa_oi->oi_oa));
218
219                 /* This should really be sent by the OST */
220                 aa->aa_oi->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
221                 aa->aa_oi->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
222         } else {
223                 CDEBUG(D_INFO, "can't unpack ost_body\n");
224                 rc = -EPROTO;
225                 aa->aa_oi->oi_oa->o_valid = 0;
226         }
227 out:
228         rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
229         RETURN(rc);
230 }
231
232 static int osc_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
233                              struct ptlrpc_request_set *set)
234 {
235         struct ptlrpc_request *req;
236         struct osc_async_args *aa;
237         int                    rc;
238         ENTRY;
239
240         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
241         if (req == NULL)
242                 RETURN(-ENOMEM);
243
244         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
245         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
246         if (rc) {
247                 ptlrpc_request_free(req);
248                 RETURN(rc);
249         }
250
251         osc_pack_req_body(req, oinfo);
252
253         ptlrpc_request_set_replen(req);
254         req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_getattr_interpret;
255
256         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
257         aa = ptlrpc_req_async_args(req);
258         aa->aa_oi = oinfo;
259
260         ptlrpc_set_add_req(set, req);
261         RETURN(0);
262 }
263
264 static int osc_getattr(struct obd_export *exp, struct obd_info *oinfo)
265 {
266         struct ptlrpc_request *req;
267         struct ost_body       *body;
268         int                    rc;
269         ENTRY;
270
271         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
272         if (req == NULL)
273                 RETURN(-ENOMEM);
274
275         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
276         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
277         if (rc) {
278                 ptlrpc_request_free(req);
279                 RETURN(rc);
280         }
281
282         osc_pack_req_body(req, oinfo);
283
284         ptlrpc_request_set_replen(req);
285
286         rc = ptlrpc_queue_wait(req);
287         if (rc)
288                 GOTO(out, rc);
289
290         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
291         if (body == NULL)
292                 GOTO(out, rc = -EPROTO);
293
294         CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
295         *oinfo->oi_oa = body->oa;
296
297         /* This should really be sent by the OST */
298         oinfo->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
299         oinfo->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
300
301         EXIT;
302  out:
303         ptlrpc_req_finished(req);
304         return rc;
305 }
306
307 static int osc_setattr(struct obd_export *exp, struct obd_info *oinfo,
308                        struct obd_trans_info *oti)
309 {
310         struct ptlrpc_request *req;
311         struct ost_body       *body;
312         int                    rc;
313         ENTRY;
314
315         LASSERTF(!(oinfo->oi_oa->o_valid & OBD_MD_FLGROUP) ||
316                  CHECK_MDS_GROUP(oinfo->oi_oa->o_gr),
317                  "oinfo->oi_oa->o_valid="LPU64" oinfo->oi_oa->o_gr="LPU64"\n",
318                  oinfo->oi_oa->o_valid, oinfo->oi_oa->o_gr);
319
320         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
321         if (req == NULL)
322                 RETURN(-ENOMEM);
323
324         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
325         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
326         if (rc) {
327                 ptlrpc_request_free(req);
328                 RETURN(rc);
329         }
330
331         osc_pack_req_body(req, oinfo);
332
333         ptlrpc_request_set_replen(req);
334
335         rc = ptlrpc_queue_wait(req);
336         if (rc)
337                 GOTO(out, rc);
338
339         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
340         if (body == NULL)
341                 GOTO(out, rc = -EPROTO);
342
343         *oinfo->oi_oa = body->oa;
344
345         EXIT;
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_async_args *aa, int rc)
354 {
355         struct ost_body *body;
356         ENTRY;
357
358         if (rc != 0)
359                 GOTO(out, rc);
360
361         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
362         if (body == NULL)
363                 GOTO(out, rc = -EPROTO);
364
365         *aa->aa_oi->oi_oa = body->oa;
366 out:
367         rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
368         RETURN(rc);
369 }
370
371 static int osc_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
372                              struct obd_trans_info *oti,
373                              struct ptlrpc_request_set *rqset)
374 {
375         struct ptlrpc_request *req;
376         struct osc_async_args *aa;
377         int                    rc;
378         ENTRY;
379
380         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
381         if (req == NULL)
382                 RETURN(-ENOMEM);
383
384         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
385         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
386         if (rc) {
387                 ptlrpc_request_free(req);
388                 RETURN(rc);
389         }
390
391         osc_pack_req_body(req, oinfo);
392
393         ptlrpc_request_set_replen(req);
394
395         if (oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE) {
396                 LASSERT(oti);
397                 oinfo->oi_oa->o_lcookie = *oti->oti_logcookies;
398         }
399
400         /* do mds to ost setattr asynchronously */
401         if (!rqset) {
402                 /* Do not wait for response. */
403                 ptlrpcd_add_req(req, PSCOPE_OTHER);
404         } else {
405                 req->rq_interpret_reply =
406                         (ptlrpc_interpterer_t)osc_setattr_interpret;
407
408                 CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
409                 aa = ptlrpc_req_async_args(req);
410                 aa->aa_oi = oinfo;
411
412                 ptlrpc_set_add_req(rqset, req);
413         }
414
415         RETURN(0);
416 }
417
418 int osc_real_create(struct obd_export *exp, struct obdo *oa,
419                     struct lov_stripe_md **ea, struct obd_trans_info *oti)
420 {
421         struct ptlrpc_request *req;
422         struct ost_body       *body;
423         struct lov_stripe_md  *lsm;
424         int                    rc;
425         ENTRY;
426
427         LASSERT(oa);
428         LASSERT(ea);
429
430         lsm = *ea;
431         if (!lsm) {
432                 rc = obd_alloc_memmd(exp, &lsm);
433                 if (rc < 0)
434                         RETURN(rc);
435         }
436
437         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
438         if (req == NULL)
439                 GOTO(out, rc = -ENOMEM);
440
441         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
442         if (rc) {
443                 ptlrpc_request_free(req);
444                 GOTO(out, rc);
445         }
446
447         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
448         LASSERT(body);
449         body->oa = *oa;
450
451         ptlrpc_request_set_replen(req);
452
453         if ((oa->o_valid & OBD_MD_FLFLAGS) &&
454             oa->o_flags == OBD_FL_DELORPHAN) {
455                 DEBUG_REQ(D_HA, req,
456                           "delorphan from OST integration");
457                 /* Don't resend the delorphan req */
458                 req->rq_no_resend = req->rq_no_delay = 1;
459         }
460
461         rc = ptlrpc_queue_wait(req);
462         if (rc)
463                 GOTO(out_req, rc);
464
465         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
466         if (body == NULL)
467                 GOTO(out_req, rc = -EPROTO);
468
469         *oa = body->oa;
470
471         /* This should really be sent by the OST */
472         oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
473         oa->o_valid |= OBD_MD_FLBLKSZ;
474
475         /* XXX LOV STACKING: the lsm that is passed to us from LOV does not
476          * have valid lsm_oinfo data structs, so don't go touching that.
477          * This needs to be fixed in a big way.
478          */
479         lsm->lsm_object_id = oa->o_id;
480         lsm->lsm_object_gr = oa->o_gr;
481         *ea = lsm;
482
483         if (oti != NULL) {
484                 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
485
486                 if (oa->o_valid & OBD_MD_FLCOOKIE) {
487                         if (!oti->oti_logcookies)
488                                 oti_alloc_cookies(oti, 1);
489                         *oti->oti_logcookies = oa->o_lcookie;
490                 }
491         }
492
493         CDEBUG(D_HA, "transno: "LPD64"\n",
494                lustre_msg_get_transno(req->rq_repmsg));
495 out_req:
496         ptlrpc_req_finished(req);
497 out:
498         if (rc && !*ea)
499                 obd_free_memmd(exp, &lsm);
500         RETURN(rc);
501 }
502
503 static int osc_punch_interpret(const struct lu_env *env,
504                                struct ptlrpc_request *req,
505                                struct osc_punch_args *aa, int rc)
506 {
507         struct ost_body *body;
508         ENTRY;
509
510         if (rc != 0)
511                 GOTO(out, rc);
512
513         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
514         if (body == NULL)
515                 GOTO(out, rc = -EPROTO);
516
517         *aa->pa_oa = body->oa;
518 out:
519         rc = aa->pa_upcall(aa->pa_cookie, rc);
520         RETURN(rc);
521 }
522
523 int osc_punch_base(struct obd_export *exp, struct obdo *oa,
524                    struct obd_capa *capa,
525                    obd_enqueue_update_f upcall, void *cookie,
526                    struct ptlrpc_request_set *rqset)
527 {
528         struct ptlrpc_request *req;
529         struct osc_punch_args *aa;
530         struct ost_body       *body;
531         int                    rc;
532         ENTRY;
533
534         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_PUNCH);
535         if (req == NULL)
536                 RETURN(-ENOMEM);
537
538         osc_set_capa_size(req, &RMF_CAPA1, capa);
539         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
540         if (rc) {
541                 ptlrpc_request_free(req);
542                 RETURN(rc);
543         }
544         req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
545         ptlrpc_at_set_req_timeout(req);
546
547         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
548         LASSERT(body);
549         body->oa = *oa;
550         osc_pack_capa(req, body, capa);
551
552         ptlrpc_request_set_replen(req);
553
554
555         req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_punch_interpret;
556         CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
557         aa = ptlrpc_req_async_args(req);
558         aa->pa_oa     = oa;
559         aa->pa_upcall = upcall;
560         aa->pa_cookie = cookie;
561         if (rqset == PTLRPCD_SET)
562                 ptlrpcd_add_req(req, PSCOPE_OTHER);
563         else
564                 ptlrpc_set_add_req(rqset, req);
565
566         RETURN(0);
567 }
568
569 static int osc_punch(struct obd_export *exp, struct obd_info *oinfo,
570                      struct obd_trans_info *oti,
571                      struct ptlrpc_request_set *rqset)
572 {
573         oinfo->oi_oa->o_size   = oinfo->oi_policy.l_extent.start;
574         oinfo->oi_oa->o_blocks = oinfo->oi_policy.l_extent.end;
575         oinfo->oi_oa->o_valid |= OBD_MD_FLSIZE | OBD_MD_FLBLOCKS;
576         return osc_punch_base(exp, oinfo->oi_oa, oinfo->oi_capa,
577                               oinfo->oi_cb_up, oinfo, rqset);
578 }
579
580 static int osc_sync(struct obd_export *exp, struct obdo *oa,
581                     struct lov_stripe_md *md, obd_size start, obd_size end,
582                     void *capa)
583 {
584         struct ptlrpc_request *req;
585         struct ost_body       *body;
586         int                    rc;
587         ENTRY;
588
589         if (!oa) {
590                 CDEBUG(D_INFO, "oa NULL\n");
591                 RETURN(-EINVAL);
592         }
593
594         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
595         if (req == NULL)
596                 RETURN(-ENOMEM);
597
598         osc_set_capa_size(req, &RMF_CAPA1, capa);
599         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
600         if (rc) {
601                 ptlrpc_request_free(req);
602                 RETURN(rc);
603         }
604
605         /* overload the size and blocks fields in the oa with start/end */
606         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
607         LASSERT(body);
608         body->oa = *oa;
609         body->oa.o_size = start;
610         body->oa.o_blocks = end;
611         body->oa.o_valid |= (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS);
612         osc_pack_capa(req, body, capa);
613
614         ptlrpc_request_set_replen(req);
615
616         rc = ptlrpc_queue_wait(req);
617         if (rc)
618                 GOTO(out, rc);
619
620         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
621         if (body == NULL)
622                 GOTO(out, rc = -EPROTO);
623
624         *oa = body->oa;
625
626         EXIT;
627  out:
628         ptlrpc_req_finished(req);
629         return rc;
630 }
631
632 /* Find and cancel locally locks matched by @mode in the resource found by
633  * @objid. Found locks are added into @cancel list. Returns the amount of
634  * locks added to @cancels list. */
635 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
636                                    struct list_head *cancels, ldlm_mode_t mode,
637                                    int lock_flags)
638 {
639         struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
640         struct ldlm_res_id res_id;
641         struct ldlm_resource *res;
642         int count;
643         ENTRY;
644
645         osc_build_res_name(oa->o_id, oa->o_gr, &res_id);
646         res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
647         if (res == NULL)
648                 RETURN(0);
649
650         LDLM_RESOURCE_ADDREF(res);
651         count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
652                                            lock_flags, 0, NULL);
653         LDLM_RESOURCE_DELREF(res);
654         ldlm_resource_putref(res);
655         RETURN(count);
656 }
657
658 static int osc_destroy_interpret(const struct lu_env *env,
659                                  struct ptlrpc_request *req, void *data,
660                                  int rc)
661 {
662         struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
663
664         atomic_dec(&cli->cl_destroy_in_flight);
665         cfs_waitq_signal(&cli->cl_destroy_waitq);
666         return 0;
667 }
668
669 static int osc_can_send_destroy(struct client_obd *cli)
670 {
671         if (atomic_inc_return(&cli->cl_destroy_in_flight) <=
672             cli->cl_max_rpcs_in_flight) {
673                 /* The destroy request can be sent */
674                 return 1;
675         }
676         if (atomic_dec_return(&cli->cl_destroy_in_flight) <
677             cli->cl_max_rpcs_in_flight) {
678                 /*
679                  * The counter has been modified between the two atomic
680                  * operations.
681                  */
682                 cfs_waitq_signal(&cli->cl_destroy_waitq);
683         }
684         return 0;
685 }
686
687 /* Destroy requests can be async always on the client, and we don't even really
688  * care about the return code since the client cannot do anything at all about
689  * a destroy failure.
690  * When the MDS is unlinking a filename, it saves the file objects into a
691  * recovery llog, and these object records are cancelled when the OST reports
692  * they were destroyed and sync'd to disk (i.e. transaction committed).
693  * If the client dies, or the OST is down when the object should be destroyed,
694  * the records are not cancelled, and when the OST reconnects to the MDS next,
695  * it will retrieve the llog unlink logs and then sends the log cancellation
696  * cookies to the MDS after committing destroy transactions. */
697 static int osc_destroy(struct obd_export *exp, struct obdo *oa,
698                        struct lov_stripe_md *ea, struct obd_trans_info *oti,
699                        struct obd_export *md_export, void *capa)
700 {
701         struct client_obd     *cli = &exp->exp_obd->u.cli;
702         struct ptlrpc_request *req;
703         struct ost_body       *body;
704         CFS_LIST_HEAD(cancels);
705         int rc, count;
706         ENTRY;
707
708         if (!oa) {
709                 CDEBUG(D_INFO, "oa NULL\n");
710                 RETURN(-EINVAL);
711         }
712
713         count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
714                                         LDLM_FL_DISCARD_DATA);
715
716         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
717         if (req == NULL) {
718                 ldlm_lock_list_put(&cancels, l_bl_ast, count);
719                 RETURN(-ENOMEM);
720         }
721
722         osc_set_capa_size(req, &RMF_CAPA1, (struct obd_capa *)capa);
723         rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
724                                0, &cancels, count);
725         if (rc) {
726                 ptlrpc_request_free(req);
727                 RETURN(rc);
728         }
729
730         req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
731         ptlrpc_at_set_req_timeout(req);
732
733         if (oti != NULL && oa->o_valid & OBD_MD_FLCOOKIE)
734                 oa->o_lcookie = *oti->oti_logcookies;
735         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
736         LASSERT(body);
737         body->oa = *oa;
738
739         osc_pack_capa(req, body, (struct obd_capa *)capa);
740         ptlrpc_request_set_replen(req);
741
742         /* don't throttle destroy RPCs for the MDT */
743         if (!(cli->cl_import->imp_connect_flags_orig & OBD_CONNECT_MDS)) {
744                 req->rq_interpret_reply = osc_destroy_interpret;
745                 if (!osc_can_send_destroy(cli)) {
746                         struct l_wait_info lwi = { 0 };
747
748                         /*
749                          * Wait until the number of on-going destroy RPCs drops
750                          * under max_rpc_in_flight
751                          */
752                         l_wait_event_exclusive(cli->cl_destroy_waitq,
753                                                osc_can_send_destroy(cli), &lwi);
754                 }
755         }
756
757         /* Do not wait for response */
758         ptlrpcd_add_req(req, PSCOPE_OTHER);
759         RETURN(0);
760 }
761
762 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
763                                 long writing_bytes)
764 {
765         obd_flag bits = OBD_MD_FLBLOCKS|OBD_MD_FLGRANT;
766
767         LASSERT(!(oa->o_valid & bits));
768
769         oa->o_valid |= bits;
770         client_obd_list_lock(&cli->cl_loi_list_lock);
771         oa->o_dirty = cli->cl_dirty;
772         if (cli->cl_dirty - cli->cl_dirty_transit > cli->cl_dirty_max) {
773                 CERROR("dirty %lu - %lu > dirty_max %lu\n",
774                        cli->cl_dirty, cli->cl_dirty_transit, cli->cl_dirty_max);
775                 oa->o_undirty = 0;
776         } else if (atomic_read(&obd_dirty_pages) -
777                    atomic_read(&obd_dirty_transit_pages) > obd_max_dirty_pages){
778                 CERROR("dirty %d - %d > system dirty_max %d\n",
779                        atomic_read(&obd_dirty_pages),
780                        atomic_read(&obd_dirty_transit_pages),
781                        obd_max_dirty_pages);
782                 oa->o_undirty = 0;
783         } else if (cli->cl_dirty_max - cli->cl_dirty > 0x7fffffff) {
784                 CERROR("dirty %lu - dirty_max %lu too big???\n",
785                        cli->cl_dirty, cli->cl_dirty_max);
786                 oa->o_undirty = 0;
787         } else {
788                 long max_in_flight = (cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT)*
789                                 (cli->cl_max_rpcs_in_flight + 1);
790                 oa->o_undirty = max(cli->cl_dirty_max, max_in_flight);
791         }
792         oa->o_grant = cli->cl_avail_grant;
793         oa->o_dropped = cli->cl_lost_grant;
794         cli->cl_lost_grant = 0;
795         client_obd_list_unlock(&cli->cl_loi_list_lock);
796         CDEBUG(D_CACHE,"dirty: "LPU64" undirty: %u dropped %u grant: "LPU64"\n",
797                oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
798
799 }
800
801 static void osc_update_next_shrink(struct client_obd *cli)
802 {
803         int time = GRANT_SHRINK_INTERVAL;
804         cli->cl_next_shrink_grant = cfs_time_shift(time);
805         CDEBUG(D_CACHE, "next time %ld to shrink grant \n",
806                cli->cl_next_shrink_grant);
807 }
808
809 /* caller must hold loi_list_lock */
810 static void osc_consume_write_grant(struct client_obd *cli,
811                                     struct brw_page *pga)
812 {
813         LASSERT(!(pga->flag & OBD_BRW_FROM_GRANT));
814         atomic_inc(&obd_dirty_pages);
815         cli->cl_dirty += CFS_PAGE_SIZE;
816         cli->cl_avail_grant -= CFS_PAGE_SIZE;
817         pga->flag |= OBD_BRW_FROM_GRANT;
818         CDEBUG(D_CACHE, "using %lu grant credits for brw %p page %p\n",
819                CFS_PAGE_SIZE, pga, pga->pg);
820         LASSERT(cli->cl_avail_grant >= 0);
821         osc_update_next_shrink(cli);
822 }
823
824 /* the companion to osc_consume_write_grant, called when a brw has completed.
825  * must be called with the loi lock held. */
826 static void osc_release_write_grant(struct client_obd *cli,
827                                     struct brw_page *pga, int sent)
828 {
829         int blocksize = cli->cl_import->imp_obd->obd_osfs.os_bsize ? : 4096;
830         ENTRY;
831
832         if (!(pga->flag & OBD_BRW_FROM_GRANT)) {
833                 EXIT;
834                 return;
835         }
836
837         pga->flag &= ~OBD_BRW_FROM_GRANT;
838         atomic_dec(&obd_dirty_pages);
839         cli->cl_dirty -= CFS_PAGE_SIZE;
840         if (pga->flag & OBD_BRW_NOCACHE) {
841                 pga->flag &= ~OBD_BRW_NOCACHE;
842                 atomic_dec(&obd_dirty_transit_pages);
843                 cli->cl_dirty_transit -= CFS_PAGE_SIZE;
844         }
845         if (!sent) {
846                 cli->cl_lost_grant += CFS_PAGE_SIZE;
847                 CDEBUG(D_CACHE, "lost grant: %lu avail grant: %lu dirty: %lu\n",
848                        cli->cl_lost_grant, cli->cl_avail_grant, cli->cl_dirty);
849         } else if (CFS_PAGE_SIZE != blocksize && pga->count != CFS_PAGE_SIZE) {
850                 /* For short writes we shouldn't count parts of pages that
851                  * span a whole block on the OST side, or our accounting goes
852                  * wrong.  Should match the code in filter_grant_check. */
853                 int offset = pga->off & ~CFS_PAGE_MASK;
854                 int count = pga->count + (offset & (blocksize - 1));
855                 int end = (offset + pga->count) & (blocksize - 1);
856                 if (end)
857                         count += blocksize - end;
858
859                 cli->cl_lost_grant += CFS_PAGE_SIZE - count;
860                 CDEBUG(D_CACHE, "lost %lu grant: %lu avail: %lu dirty: %lu\n",
861                        CFS_PAGE_SIZE - count, cli->cl_lost_grant,
862                        cli->cl_avail_grant, cli->cl_dirty);
863         }
864
865         EXIT;
866 }
867
868 static unsigned long rpcs_in_flight(struct client_obd *cli)
869 {
870         return cli->cl_r_in_flight + cli->cl_w_in_flight;
871 }
872
873 /* caller must hold loi_list_lock */
874 void osc_wake_cache_waiters(struct client_obd *cli)
875 {
876         struct list_head *l, *tmp;
877         struct osc_cache_waiter *ocw;
878
879         ENTRY;
880         list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
881                 /* if we can't dirty more, we must wait until some is written */
882                 if ((cli->cl_dirty + CFS_PAGE_SIZE > cli->cl_dirty_max) ||
883                    (atomic_read(&obd_dirty_pages) + 1 > obd_max_dirty_pages)) {
884                         CDEBUG(D_CACHE, "no dirty room: dirty: %ld "
885                                "osc max %ld, sys max %d\n", cli->cl_dirty,
886                                cli->cl_dirty_max, obd_max_dirty_pages);
887                         return;
888                 }
889
890                 /* if still dirty cache but no grant wait for pending RPCs that
891                  * may yet return us some grant before doing sync writes */
892                 if (cli->cl_w_in_flight && cli->cl_avail_grant < CFS_PAGE_SIZE) {
893                         CDEBUG(D_CACHE, "%u BRW writes in flight, no grant\n",
894                                cli->cl_w_in_flight);
895                         return;
896                 }
897
898                 ocw = list_entry(l, struct osc_cache_waiter, ocw_entry);
899                 list_del_init(&ocw->ocw_entry);
900                 if (cli->cl_avail_grant < CFS_PAGE_SIZE) {
901                         /* no more RPCs in flight to return grant, do sync IO */
902                         ocw->ocw_rc = -EDQUOT;
903                         CDEBUG(D_INODE, "wake oap %p for sync\n", ocw->ocw_oap);
904                 } else {
905                         osc_consume_write_grant(cli,
906                                                 &ocw->ocw_oap->oap_brw_page);
907                 }
908
909                 cfs_waitq_signal(&ocw->ocw_waitq);
910         }
911
912         EXIT;
913 }
914
915 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
916 {
917         client_obd_list_lock(&cli->cl_loi_list_lock);
918         CDEBUG(D_CACHE, "got "LPU64" extra grant\n", body->oa.o_grant);
919         if (body->oa.o_valid & OBD_MD_FLGRANT)
920                 cli->cl_avail_grant += body->oa.o_grant;
921         /* waiters are woken in brw_interpret */
922         client_obd_list_unlock(&cli->cl_loi_list_lock);
923 }
924
925 static int osc_set_info_async(struct obd_export *exp, obd_count keylen,
926                               void *key, obd_count vallen, void *val,
927                               struct ptlrpc_request_set *set);
928
929 static int osc_shrink_grant_interpret(const struct lu_env *env,
930                                       struct ptlrpc_request *req,
931                                       void *aa, int rc)
932 {
933         struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
934         struct obdo *oa = ((struct osc_grant_args *)aa)->aa_oa;
935         struct ost_body *body;
936         
937         if (rc != 0) {
938                 client_obd_list_lock(&cli->cl_loi_list_lock);
939                 cli->cl_avail_grant += oa->o_grant;
940                 client_obd_list_unlock(&cli->cl_loi_list_lock);
941                 GOTO(out, rc);
942         }
943
944         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
945         LASSERT(body);
946         osc_update_grant(cli, body);
947 out:
948         OBD_FREE_PTR(oa);
949         return rc;        
950 }
951
952 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
953 {
954         client_obd_list_lock(&cli->cl_loi_list_lock);
955         oa->o_grant = cli->cl_avail_grant / 4;
956         cli->cl_avail_grant -= oa->o_grant; 
957         client_obd_list_unlock(&cli->cl_loi_list_lock);
958         oa->o_flags |= OBD_FL_SHRINK_GRANT;
959         osc_update_next_shrink(cli);
960 }
961
962 static int osc_shrink_grant(struct client_obd *cli)
963 {
964         int    rc = 0;
965         struct ost_body     *body;
966         ENTRY;
967
968         OBD_ALLOC_PTR(body);
969         if (!body)
970                 RETURN(-ENOMEM);
971
972         osc_announce_cached(cli, &body->oa, 0);
973         osc_shrink_grant_local(cli, &body->oa);
974         rc = osc_set_info_async(cli->cl_import->imp_obd->obd_self_export,
975                                 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
976                                 sizeof(*body), body, NULL);
977         if (rc) {
978                 client_obd_list_lock(&cli->cl_loi_list_lock);
979                 cli->cl_avail_grant += body->oa.o_grant;
980                 client_obd_list_unlock(&cli->cl_loi_list_lock);
981         }
982         if (body)
983                OBD_FREE_PTR(body);
984         RETURN(rc);
985 }
986
987 #define GRANT_SHRINK_LIMIT PTLRPC_MAX_BRW_SIZE
988 static int osc_should_shrink_grant(struct client_obd *client)
989 {
990         cfs_time_t time = cfs_time_current();
991         cfs_time_t next_shrink = client->cl_next_shrink_grant;
992         if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
993                 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
994                     client->cl_avail_grant > GRANT_SHRINK_LIMIT)
995                         return 1;
996                 else
997                         osc_update_next_shrink(client);
998         }
999         return 0;
1000 }
1001
1002 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
1003 {
1004         struct client_obd *client;
1005
1006         list_for_each_entry(client, &item->ti_obd_list, cl_grant_shrink_list) {
1007                 if (osc_should_shrink_grant(client))
1008                         osc_shrink_grant(client);
1009         }
1010         return 0;
1011 }
1012
1013 static int osc_add_shrink_grant(struct client_obd *client)
1014 {
1015         int rc;
1016
1017         rc = ptlrpc_add_timeout_client(GRANT_SHRINK_INTERVAL, 
1018                                          TIMEOUT_GRANT,
1019                                          osc_grant_shrink_grant_cb, NULL,
1020                                          &client->cl_grant_shrink_list);
1021         if (rc) {
1022                 CERROR("add grant client %s error %d\n", 
1023                         client->cl_import->imp_obd->obd_name, rc);
1024                 return rc;
1025         }
1026         CDEBUG(D_CACHE, "add grant client %s \n", 
1027                client->cl_import->imp_obd->obd_name);
1028         osc_update_next_shrink(client);
1029         return 0; 
1030 }
1031
1032 static int osc_del_shrink_grant(struct client_obd *client)
1033 {
1034         CDEBUG(D_CACHE, "del grant client %s \n", 
1035                client->cl_import->imp_obd->obd_name);
1036         return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list);
1037 }
1038
1039 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1040 {
1041         client_obd_list_lock(&cli->cl_loi_list_lock);
1042         cli->cl_avail_grant = ocd->ocd_grant;
1043         client_obd_list_unlock(&cli->cl_loi_list_lock);
1044
1045         if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
1046             list_empty(&cli->cl_grant_shrink_list))
1047                 osc_add_shrink_grant(cli);
1048
1049         CDEBUG(D_CACHE, "setting cl_avail_grant: %ld cl_lost_grant: %ld \n",
1050                cli->cl_avail_grant, cli->cl_lost_grant);
1051         LASSERT(cli->cl_avail_grant >= 0);
1052 }
1053
1054 /* We assume that the reason this OSC got a short read is because it read
1055  * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1056  * via the LOV, and it _knows_ it's reading inside the file, it's just that
1057  * this stripe never got written at or beyond this stripe offset yet. */
1058 static void handle_short_read(int nob_read, obd_count page_count,
1059                               struct brw_page **pga)
1060 {
1061         char *ptr;
1062         int i = 0;
1063
1064         /* skip bytes read OK */
1065         while (nob_read > 0) {
1066                 LASSERT (page_count > 0);
1067
1068                 if (pga[i]->count > nob_read) {
1069                         /* EOF inside this page */
1070                         ptr = cfs_kmap(pga[i]->pg) +
1071                                 (pga[i]->off & ~CFS_PAGE_MASK);
1072                         memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1073                         cfs_kunmap(pga[i]->pg);
1074                         page_count--;
1075                         i++;
1076                         break;
1077                 }
1078
1079                 nob_read -= pga[i]->count;
1080                 page_count--;
1081                 i++;
1082         }
1083
1084         /* zero remaining pages */
1085         while (page_count-- > 0) {
1086                 ptr = cfs_kmap(pga[i]->pg) + (pga[i]->off & ~CFS_PAGE_MASK);
1087                 memset(ptr, 0, pga[i]->count);
1088                 cfs_kunmap(pga[i]->pg);
1089                 i++;
1090         }
1091 }
1092
1093 static int check_write_rcs(struct ptlrpc_request *req,
1094                            int requested_nob, int niocount,
1095                            obd_count page_count, struct brw_page **pga)
1096 {
1097         int    *remote_rcs, i;
1098
1099         /* return error if any niobuf was in error */
1100         remote_rcs = lustre_swab_repbuf(req, REQ_REC_OFF + 1,
1101                                         sizeof(*remote_rcs) * niocount, NULL);
1102         if (remote_rcs == NULL) {
1103                 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1104                 return(-EPROTO);
1105         }
1106         if (lustre_msg_swabbed(req->rq_repmsg))
1107                 for (i = 0; i < niocount; i++)
1108                         __swab32s(&remote_rcs[i]);
1109
1110         for (i = 0; i < niocount; i++) {
1111                 if (remote_rcs[i] < 0)
1112                         return(remote_rcs[i]);
1113
1114                 if (remote_rcs[i] != 0) {
1115                         CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1116                                 i, remote_rcs[i], req);
1117                         return(-EPROTO);
1118                 }
1119         }
1120
1121         if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1122                 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1123                        req->rq_bulk->bd_nob_transferred, requested_nob);
1124                 return(-EPROTO);
1125         }
1126
1127         return (0);
1128 }
1129
1130 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1131 {
1132         if (p1->flag != p2->flag) {
1133                 unsigned mask = ~(OBD_BRW_FROM_GRANT|
1134                                   OBD_BRW_NOCACHE|OBD_BRW_SYNC);
1135
1136                 /* warn if we try to combine flags that we don't know to be
1137                  * safe to combine */
1138                 if ((p1->flag & mask) != (p2->flag & mask))
1139                         CERROR("is it ok to have flags 0x%x and 0x%x in the "
1140                                "same brw?\n", p1->flag, p2->flag);
1141                 return 0;
1142         }
1143
1144         return (p1->off + p1->count == p2->off);
1145 }
1146
1147 static obd_count osc_checksum_bulk(int nob, obd_count pg_count,
1148                                    struct brw_page **pga, int opc,
1149                                    cksum_type_t cksum_type)
1150 {
1151         __u32 cksum;
1152         int i = 0;
1153
1154         LASSERT (pg_count > 0);
1155         cksum = init_checksum(cksum_type);
1156         while (nob > 0 && pg_count > 0) {
1157                 unsigned char *ptr = cfs_kmap(pga[i]->pg);
1158                 int off = pga[i]->off & ~CFS_PAGE_MASK;
1159                 int count = pga[i]->count > nob ? nob : pga[i]->count;
1160
1161                 /* corrupt the data before we compute the checksum, to
1162                  * simulate an OST->client data error */
1163                 if (i == 0 && opc == OST_READ &&
1164                     OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE))
1165                         memcpy(ptr + off, "bad1", min(4, nob));
1166                 cksum = compute_checksum(cksum, ptr + off, count, cksum_type);
1167                 cfs_kunmap(pga[i]->pg);
1168                 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d checksum %x\n",
1169                                off, cksum);
1170
1171                 nob -= pga[i]->count;
1172                 pg_count--;
1173                 i++;
1174         }
1175         /* For sending we only compute the wrong checksum instead
1176          * of corrupting the data so it is still correct on a redo */
1177         if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1178                 cksum++;
1179
1180         return cksum;
1181 }
1182
1183 static int osc_brw_prep_request(int cmd, struct client_obd *cli,struct obdo *oa,
1184                                 struct lov_stripe_md *lsm, obd_count page_count,
1185                                 struct brw_page **pga,
1186                                 struct ptlrpc_request **reqp,
1187                                 struct obd_capa *ocapa, int reserve)
1188 {
1189         struct ptlrpc_request   *req;
1190         struct ptlrpc_bulk_desc *desc;
1191         struct ost_body         *body;
1192         struct obd_ioobj        *ioobj;
1193         struct niobuf_remote    *niobuf;
1194         int niocount, i, requested_nob, opc, rc;
1195         struct osc_brw_async_args *aa;
1196         struct req_capsule      *pill;
1197         struct brw_page *pg_prev;
1198
1199         ENTRY;
1200         if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1201                 RETURN(-ENOMEM); /* Recoverable */
1202         if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1203                 RETURN(-EINVAL); /* Fatal */
1204
1205         if ((cmd & OBD_BRW_WRITE) != 0) {
1206                 opc = OST_WRITE;
1207                 req = ptlrpc_request_alloc_pool(cli->cl_import,
1208                                                 cli->cl_import->imp_rq_pool,
1209                                                 &RQF_OST_BRW);
1210         } else {
1211                 opc = OST_READ;
1212                 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW);
1213         }
1214         if (req == NULL)
1215                 RETURN(-ENOMEM);
1216
1217         for (niocount = i = 1; i < page_count; i++) {
1218                 if (!can_merge_pages(pga[i - 1], pga[i]))
1219                         niocount++;
1220         }
1221
1222         pill = &req->rq_pill;
1223         req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1224                              niocount * sizeof(*niobuf));
1225         osc_set_capa_size(req, &RMF_CAPA1, ocapa);
1226
1227         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1228         if (rc) {
1229                 ptlrpc_request_free(req);
1230                 RETURN(rc);
1231         }
1232         req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1233         ptlrpc_at_set_req_timeout(req);
1234
1235         if (opc == OST_WRITE)
1236                 desc = ptlrpc_prep_bulk_imp(req, page_count,
1237                                             BULK_GET_SOURCE, OST_BULK_PORTAL);
1238         else
1239                 desc = ptlrpc_prep_bulk_imp(req, page_count,
1240                                             BULK_PUT_SINK, OST_BULK_PORTAL);
1241
1242         if (desc == NULL)
1243                 GOTO(out, rc = -ENOMEM);
1244         /* NB request now owns desc and will free it when it gets freed */
1245
1246         body = req_capsule_client_get(pill, &RMF_OST_BODY);
1247         ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1248         niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1249         LASSERT(body && ioobj && niobuf);
1250
1251         body->oa = *oa;
1252
1253         obdo_to_ioobj(oa, ioobj);
1254         ioobj->ioo_bufcnt = niocount;
1255         osc_pack_capa(req, body, ocapa);
1256         LASSERT (page_count > 0);
1257         pg_prev = pga[0];
1258         for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1259                 struct brw_page *pg = pga[i];
1260
1261                 LASSERT(pg->count > 0);
1262                 LASSERTF((pg->off & ~CFS_PAGE_MASK) + pg->count <= CFS_PAGE_SIZE,
1263                          "i: %d pg: %p off: "LPU64", count: %u\n", i, pg,
1264                          pg->off, pg->count);
1265 #ifdef __linux__
1266                 LASSERTF(i == 0 || pg->off > pg_prev->off,
1267                          "i %d p_c %u pg %p [pri %lu ind %lu] off "LPU64
1268                          " prev_pg %p [pri %lu ind %lu] off "LPU64"\n",
1269                          i, page_count,
1270                          pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1271                          pg_prev->pg, page_private(pg_prev->pg),
1272                          pg_prev->pg->index, pg_prev->off);
1273 #else
1274                 LASSERTF(i == 0 || pg->off > pg_prev->off,
1275                          "i %d p_c %u\n", i, page_count);
1276 #endif
1277                 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1278                         (pg->flag & OBD_BRW_SRVLOCK));
1279
1280                 ptlrpc_prep_bulk_page(desc, pg->pg, pg->off & ~CFS_PAGE_MASK,
1281                                       pg->count);
1282                 requested_nob += pg->count;
1283
1284                 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1285                         niobuf--;
1286                         niobuf->len += pg->count;
1287                 } else {
1288                         niobuf->offset = pg->off;
1289                         niobuf->len    = pg->count;
1290                         niobuf->flags  = pg->flag;
1291                 }
1292                 pg_prev = pg;
1293         }
1294
1295         LASSERTF((void *)(niobuf - niocount) ==
1296                 lustre_msg_buf(req->rq_reqmsg, REQ_REC_OFF + 2,
1297                                niocount * sizeof(*niobuf)),
1298                 "want %p - real %p\n", lustre_msg_buf(req->rq_reqmsg,
1299                 REQ_REC_OFF + 2, niocount * sizeof(*niobuf)),
1300                 (void *)(niobuf - niocount));
1301
1302         osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1303         if (osc_should_shrink_grant(cli))
1304                 osc_shrink_grant_local(cli, &body->oa); 
1305
1306         /* size[REQ_REC_OFF] still sizeof (*body) */
1307         if (opc == OST_WRITE) {
1308                 if (unlikely(cli->cl_checksum) &&
1309                     !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1310                         /* store cl_cksum_type in a local variable since
1311                          * it can be changed via lprocfs */
1312                         cksum_type_t cksum_type = cli->cl_cksum_type;
1313
1314                         if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1315                                 oa->o_flags = body->oa.o_flags = 0;
1316                         body->oa.o_flags |= cksum_type_pack(cksum_type);
1317                         body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1318                         body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1319                                                              page_count, pga,
1320                                                              OST_WRITE,
1321                                                              cksum_type);
1322                         CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1323                                body->oa.o_cksum);
1324                         /* save this in 'oa', too, for later checking */
1325                         oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1326                         oa->o_flags |= cksum_type_pack(cksum_type);
1327                 } else {
1328                         /* clear out the checksum flag, in case this is a
1329                          * resend but cl_checksum is no longer set. b=11238 */
1330                         oa->o_valid &= ~OBD_MD_FLCKSUM;
1331                 }
1332                 oa->o_cksum = body->oa.o_cksum;
1333                 /* 1 RC per niobuf */
1334                 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_SERVER,
1335                                      sizeof(__u32) * niocount);
1336         } else {
1337                 if (unlikely(cli->cl_checksum) &&
1338                     !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1339                         if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1340                                 body->oa.o_flags = 0;
1341                         body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1342                         body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1343                 }
1344                 req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_SERVER, 0);
1345                 /* 1 RC for the whole I/O */
1346         }
1347         ptlrpc_request_set_replen(req);
1348
1349         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1350         aa = ptlrpc_req_async_args(req);
1351         aa->aa_oa = oa;
1352         aa->aa_requested_nob = requested_nob;
1353         aa->aa_nio_count = niocount;
1354         aa->aa_page_count = page_count;
1355         aa->aa_resends = 0;
1356         aa->aa_ppga = pga;
1357         aa->aa_cli = cli;
1358         CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1359         if (ocapa && reserve)
1360                 aa->aa_ocapa = capa_get(ocapa);
1361
1362         *reqp = req;
1363         RETURN(0);
1364
1365  out:
1366         ptlrpc_req_finished(req);
1367         RETURN(rc);
1368 }
1369
1370 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1371                                 __u32 client_cksum, __u32 server_cksum, int nob,
1372                                 obd_count page_count, struct brw_page **pga,
1373                                 cksum_type_t client_cksum_type)
1374 {
1375         __u32 new_cksum;
1376         char *msg;
1377         cksum_type_t cksum_type;
1378
1379         if (server_cksum == client_cksum) {
1380                 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1381                 return 0;
1382         }
1383
1384         if (oa->o_valid & OBD_MD_FLFLAGS)
1385                 cksum_type = cksum_type_unpack(oa->o_flags);
1386         else
1387                 cksum_type = OBD_CKSUM_CRC32;
1388
1389         new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1390                                       cksum_type);
1391
1392         if (cksum_type != client_cksum_type)
1393                 msg = "the server did not use the checksum type specified in "
1394                       "the original request - likely a protocol problem";
1395         else if (new_cksum == server_cksum)
1396                 msg = "changed on the client after we checksummed it - "
1397                       "likely false positive due to mmap IO (bug 11742)";
1398         else if (new_cksum == client_cksum)
1399                 msg = "changed in transit before arrival at OST";
1400         else
1401                 msg = "changed in transit AND doesn't match the original - "
1402                       "likely false positive due to mmap IO (bug 11742)";
1403
1404         LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inum "
1405                            LPU64"/"LPU64" object "LPU64"/"LPU64" extent "
1406                            "["LPU64"-"LPU64"]\n",
1407                            msg, libcfs_nid2str(peer->nid),
1408                            oa->o_valid & OBD_MD_FLFID ? oa->o_fid : (__u64)0,
1409                            oa->o_valid & OBD_MD_FLFID ? oa->o_generation :
1410                                                         (__u64)0,
1411                            oa->o_id,
1412                            oa->o_valid & OBD_MD_FLGROUP ? oa->o_gr : (__u64)0,
1413                            pga[0]->off,
1414                            pga[page_count-1]->off + pga[page_count-1]->count - 1);
1415         CERROR("original client csum %x (type %x), server csum %x (type %x), "
1416                "client csum now %x\n", client_cksum, client_cksum_type,
1417                server_cksum, cksum_type, new_cksum);
1418         return 1;
1419 }
1420
1421 /* Note rc enters this function as number of bytes transferred */
1422 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1423 {
1424         struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1425         const lnet_process_id_t *peer =
1426                         &req->rq_import->imp_connection->c_peer;
1427         struct client_obd *cli = aa->aa_cli;
1428         struct ost_body *body;
1429         __u32 client_cksum = 0;
1430         ENTRY;
1431
1432         if (rc < 0 && rc != -EDQUOT)
1433                 RETURN(rc);
1434
1435         LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1436         body = lustre_swab_repbuf(req, REPLY_REC_OFF, sizeof(*body),
1437                                   lustre_swab_ost_body);
1438         if (body == NULL) {
1439                 CDEBUG(D_INFO, "Can't unpack body\n");
1440                 RETURN(-EPROTO);
1441         }
1442
1443         /* set/clear over quota flag for a uid/gid */
1444         if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1445             body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA))
1446                 lquota_setdq(quota_interface, cli, body->oa.o_uid,
1447                              body->oa.o_gid, body->oa.o_valid,
1448                              body->oa.o_flags);
1449
1450         if (rc < 0)
1451                 RETURN(rc);
1452
1453         if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1454                 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1455
1456         osc_update_grant(cli, body);
1457
1458         if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1459                 if (rc > 0) {
1460                         CERROR("Unexpected +ve rc %d\n", rc);
1461                         RETURN(-EPROTO);
1462                 }
1463                 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1464
1465                 if (sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1466                         RETURN(-EAGAIN);
1467
1468                 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1469                     check_write_checksum(&body->oa, peer, client_cksum,
1470                                          body->oa.o_cksum, aa->aa_requested_nob,
1471                                          aa->aa_page_count, aa->aa_ppga,
1472                                          cksum_type_unpack(aa->aa_oa->o_flags)))
1473                         RETURN(-EAGAIN);
1474
1475                 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1476                                      aa->aa_page_count, aa->aa_ppga);
1477                 GOTO(out, rc);
1478         }
1479
1480         /* The rest of this function executes only for OST_READs */
1481
1482         rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1483         if (rc < 0)
1484                 GOTO(out, rc);
1485
1486         if (rc > aa->aa_requested_nob) {
1487                 CERROR("Unexpected rc %d (%d requested)\n", rc,
1488                        aa->aa_requested_nob);
1489                 RETURN(-EPROTO);
1490         }
1491
1492         if (rc != req->rq_bulk->bd_nob_transferred) {
1493                 CERROR ("Unexpected rc %d (%d transferred)\n",
1494                         rc, req->rq_bulk->bd_nob_transferred);
1495                 return (-EPROTO);
1496         }
1497
1498         if (rc < aa->aa_requested_nob)
1499                 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1500
1501         if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1502                 static int cksum_counter;
1503                 __u32      server_cksum = body->oa.o_cksum;
1504                 char      *via;
1505                 char      *router;
1506                 cksum_type_t cksum_type;
1507
1508                 if (body->oa.o_valid & OBD_MD_FLFLAGS)
1509                         cksum_type = cksum_type_unpack(body->oa.o_flags);
1510                 else
1511                         cksum_type = OBD_CKSUM_CRC32;
1512                 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1513                                                  aa->aa_ppga, OST_READ,
1514                                                  cksum_type);
1515
1516                 if (peer->nid == req->rq_bulk->bd_sender) {
1517                         via = router = "";
1518                 } else {
1519                         via = " via ";
1520                         router = libcfs_nid2str(req->rq_bulk->bd_sender);
1521                 }
1522
1523                 if (server_cksum == ~0 && rc > 0) {
1524                         CERROR("Protocol error: server %s set the 'checksum' "
1525                                "bit, but didn't send a checksum.  Not fatal, "
1526                                "but please notify on http://bugzilla.lustre.org/\n",
1527                                libcfs_nid2str(peer->nid));
1528                 } else if (server_cksum != client_cksum) {
1529                         LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1530                                            "%s%s%s inum "LPU64"/"LPU64" object "
1531                                            LPU64"/"LPU64" extent "
1532                                            "["LPU64"-"LPU64"]\n",
1533                                            req->rq_import->imp_obd->obd_name,
1534                                            libcfs_nid2str(peer->nid),
1535                                            via, router,
1536                                            body->oa.o_valid & OBD_MD_FLFID ?
1537                                                 body->oa.o_fid : (__u64)0,
1538                                            body->oa.o_valid & OBD_MD_FLFID ?
1539                                                 body->oa.o_generation :(__u64)0,
1540                                            body->oa.o_id,
1541                                            body->oa.o_valid & OBD_MD_FLGROUP ?
1542                                                 body->oa.o_gr : (__u64)0,
1543                                            aa->aa_ppga[0]->off,
1544                                            aa->aa_ppga[aa->aa_page_count-1]->off +
1545                                            aa->aa_ppga[aa->aa_page_count-1]->count -
1546                                                                         1);
1547                         CERROR("client %x, server %x, cksum_type %x\n",
1548                                client_cksum, server_cksum, cksum_type);
1549                         cksum_counter = 0;
1550                         aa->aa_oa->o_cksum = client_cksum;
1551                         rc = -EAGAIN;
1552                 } else {
1553                         cksum_counter++;
1554                         CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1555                         rc = 0;
1556                 }
1557         } else if (unlikely(client_cksum)) {
1558                 static int cksum_missed;
1559
1560                 cksum_missed++;
1561                 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1562                         CERROR("Checksum %u requested from %s but not sent\n",
1563                                cksum_missed, libcfs_nid2str(peer->nid));
1564         } else {
1565                 rc = 0;
1566         }
1567 out:
1568         if (rc >= 0)
1569                 *aa->aa_oa = body->oa;
1570
1571         RETURN(rc);
1572 }
1573
1574 static int osc_brw_internal(int cmd, struct obd_export *exp, struct obdo *oa,
1575                             struct lov_stripe_md *lsm,
1576                             obd_count page_count, struct brw_page **pga,
1577                             struct obd_capa *ocapa)
1578 {
1579         struct ptlrpc_request *req;
1580         int                    rc;
1581         cfs_waitq_t            waitq;
1582         int                    resends = 0;
1583         struct l_wait_info     lwi;
1584
1585         ENTRY;
1586
1587         cfs_waitq_init(&waitq);
1588
1589 restart_bulk:
1590         rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1591                                   page_count, pga, &req, ocapa, 0);
1592         if (rc != 0)
1593                 return (rc);
1594
1595         rc = ptlrpc_queue_wait(req);
1596
1597         if (rc == -ETIMEDOUT && req->rq_resend) {
1598                 DEBUG_REQ(D_HA, req,  "BULK TIMEOUT");
1599                 ptlrpc_req_finished(req);
1600                 goto restart_bulk;
1601         }
1602
1603         rc = osc_brw_fini_request(req, rc);
1604
1605         ptlrpc_req_finished(req);
1606         if (osc_recoverable_error(rc)) {
1607                 resends++;
1608                 if (!osc_should_resend(resends, &exp->exp_obd->u.cli)) {
1609                         CERROR("too many resend retries, returning error\n");
1610                         RETURN(-EIO);
1611                 }
1612
1613                 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends), NULL, NULL, NULL);
1614                 l_wait_event(waitq, 0, &lwi);
1615
1616                 goto restart_bulk;
1617         }
1618
1619         RETURN (rc);
1620 }
1621
1622 int osc_brw_redo_request(struct ptlrpc_request *request,
1623                          struct osc_brw_async_args *aa)
1624 {
1625         struct ptlrpc_request *new_req;
1626         struct ptlrpc_request_set *set = request->rq_set;
1627         struct osc_brw_async_args *new_aa;
1628         struct osc_async_page *oap;
1629         int rc = 0;
1630         ENTRY;
1631
1632         if (!osc_should_resend(aa->aa_resends, aa->aa_cli)) {
1633                 CERROR("too many resend retries, returning error\n");
1634                 RETURN(-EIO);
1635         }
1636
1637         DEBUG_REQ(D_ERROR, request, "redo for recoverable error");
1638
1639         rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1640                                         OST_WRITE ? OBD_BRW_WRITE :OBD_BRW_READ,
1641                                   aa->aa_cli, aa->aa_oa,
1642                                   NULL /* lsm unused by osc currently */,
1643                                   aa->aa_page_count, aa->aa_ppga,
1644                                   &new_req, aa->aa_ocapa, 0);
1645         if (rc)
1646                 RETURN(rc);
1647
1648         client_obd_list_lock(&aa->aa_cli->cl_loi_list_lock);
1649
1650         list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1651                 if (oap->oap_request != NULL) {
1652                         LASSERTF(request == oap->oap_request,
1653                                  "request %p != oap_request %p\n",
1654                                  request, oap->oap_request);
1655                         if (oap->oap_interrupted) {
1656                                 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1657                                 ptlrpc_req_finished(new_req);
1658                                 RETURN(-EINTR);
1659                         }
1660                 }
1661         }
1662         /* New request takes over pga and oaps from old request.
1663          * Note that copying a list_head doesn't work, need to move it... */
1664         aa->aa_resends++;
1665         new_req->rq_interpret_reply = request->rq_interpret_reply;
1666         new_req->rq_async_args = request->rq_async_args;
1667         new_req->rq_sent = cfs_time_current_sec() + aa->aa_resends;
1668
1669         new_aa = ptlrpc_req_async_args(new_req);
1670
1671         CFS_INIT_LIST_HEAD(&new_aa->aa_oaps);
1672         list_splice(&aa->aa_oaps, &new_aa->aa_oaps);
1673         CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1674
1675         list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1676                 if (oap->oap_request) {
1677                         ptlrpc_req_finished(oap->oap_request);
1678                         oap->oap_request = ptlrpc_request_addref(new_req);
1679                 }
1680         }
1681
1682         new_aa->aa_ocapa = aa->aa_ocapa;
1683         aa->aa_ocapa = NULL;
1684
1685         /* use ptlrpc_set_add_req is safe because interpret functions work
1686          * in check_set context. only one way exist with access to request
1687          * from different thread got -EINTR - this way protected with
1688          * cl_loi_list_lock */
1689         ptlrpc_set_add_req(set, new_req);
1690
1691         client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1692
1693         DEBUG_REQ(D_INFO, new_req, "new request");
1694         RETURN(0);
1695 }
1696
1697 /*
1698  * ugh, we want disk allocation on the target to happen in offset order.  we'll
1699  * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1700  * fine for our small page arrays and doesn't require allocation.  its an
1701  * insertion sort that swaps elements that are strides apart, shrinking the
1702  * stride down until its '1' and the array is sorted.
1703  */
1704 static void sort_brw_pages(struct brw_page **array, int num)
1705 {
1706         int stride, i, j;
1707         struct brw_page *tmp;
1708
1709         if (num == 1)
1710                 return;
1711         for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1712                 ;
1713
1714         do {
1715                 stride /= 3;
1716                 for (i = stride ; i < num ; i++) {
1717                         tmp = array[i];
1718                         j = i;
1719                         while (j >= stride && array[j - stride]->off > tmp->off) {
1720                                 array[j] = array[j - stride];
1721                                 j -= stride;
1722                         }
1723                         array[j] = tmp;
1724                 }
1725         } while (stride > 1);
1726 }
1727
1728 static obd_count max_unfragmented_pages(struct brw_page **pg, obd_count pages)
1729 {
1730         int count = 1;
1731         int offset;
1732         int i = 0;
1733
1734         LASSERT (pages > 0);
1735         offset = pg[i]->off & ~CFS_PAGE_MASK;
1736
1737         for (;;) {
1738                 pages--;
1739                 if (pages == 0)         /* that's all */
1740                         return count;
1741
1742                 if (offset + pg[i]->count < CFS_PAGE_SIZE)
1743                         return count;   /* doesn't end on page boundary */
1744
1745                 i++;
1746                 offset = pg[i]->off & ~CFS_PAGE_MASK;
1747                 if (offset != 0)        /* doesn't start on page boundary */
1748                         return count;
1749
1750                 count++;
1751         }
1752 }
1753
1754 static struct brw_page **osc_build_ppga(struct brw_page *pga, obd_count count)
1755 {
1756         struct brw_page **ppga;
1757         int i;
1758
1759         OBD_ALLOC(ppga, sizeof(*ppga) * count);
1760         if (ppga == NULL)
1761                 return NULL;
1762
1763         for (i = 0; i < count; i++)
1764                 ppga[i] = pga + i;
1765         return ppga;
1766 }
1767
1768 static void osc_release_ppga(struct brw_page **ppga, obd_count count)
1769 {
1770         LASSERT(ppga != NULL);
1771         OBD_FREE(ppga, sizeof(*ppga) * count);
1772 }
1773
1774 static int osc_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1775                    obd_count page_count, struct brw_page *pga,
1776                    struct obd_trans_info *oti)
1777 {
1778         struct obdo *saved_oa = NULL;
1779         struct brw_page **ppga, **orig;
1780         struct obd_import *imp = class_exp2cliimp(exp);
1781         struct client_obd *cli = &imp->imp_obd->u.cli;
1782         int rc, page_count_orig;
1783         ENTRY;
1784
1785         if (cmd & OBD_BRW_CHECK) {
1786                 /* The caller just wants to know if there's a chance that this
1787                  * I/O can succeed */
1788
1789                 if (imp == NULL || imp->imp_invalid)
1790                         RETURN(-EIO);
1791                 RETURN(0);
1792         }
1793
1794         /* test_brw with a failed create can trip this, maybe others. */
1795         LASSERT(cli->cl_max_pages_per_rpc);
1796
1797         rc = 0;
1798
1799         orig = ppga = osc_build_ppga(pga, page_count);
1800         if (ppga == NULL)
1801                 RETURN(-ENOMEM);
1802         page_count_orig = page_count;
1803
1804         sort_brw_pages(ppga, page_count);
1805         while (page_count) {
1806                 obd_count pages_per_brw;
1807
1808                 if (page_count > cli->cl_max_pages_per_rpc)
1809                         pages_per_brw = cli->cl_max_pages_per_rpc;
1810                 else
1811                         pages_per_brw = page_count;
1812
1813                 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1814
1815                 if (saved_oa != NULL) {
1816                         /* restore previously saved oa */
1817                         *oinfo->oi_oa = *saved_oa;
1818                 } else if (page_count > pages_per_brw) {
1819                         /* save a copy of oa (brw will clobber it) */
1820                         OBDO_ALLOC(saved_oa);
1821                         if (saved_oa == NULL)
1822                                 GOTO(out, rc = -ENOMEM);
1823                         *saved_oa = *oinfo->oi_oa;
1824                 }
1825
1826                 rc = osc_brw_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1827                                       pages_per_brw, ppga, oinfo->oi_capa);
1828
1829                 if (rc != 0)
1830                         break;
1831
1832                 page_count -= pages_per_brw;
1833                 ppga += pages_per_brw;
1834         }
1835
1836 out:
1837         osc_release_ppga(orig, page_count_orig);
1838
1839         if (saved_oa != NULL)
1840                 OBDO_FREE(saved_oa);
1841
1842         RETURN(rc);
1843 }
1844
1845 /* The companion to osc_enter_cache(), called when @oap is no longer part of
1846  * the dirty accounting.  Writeback completes or truncate happens before
1847  * writing starts.  Must be called with the loi lock held. */
1848 static void osc_exit_cache(struct client_obd *cli, struct osc_async_page *oap,
1849                            int sent)
1850 {
1851         osc_release_write_grant(cli, &oap->oap_brw_page, sent);
1852 }
1853
1854
1855 /* This maintains the lists of pending pages to read/write for a given object
1856  * (lop).  This is used by osc_check_rpcs->osc_next_loi() and loi_list_maint()
1857  * to quickly find objects that are ready to send an RPC. */
1858 static int lop_makes_rpc(struct client_obd *cli, struct loi_oap_pages *lop,
1859                          int cmd)
1860 {
1861         int optimal;
1862         ENTRY;
1863
1864         if (lop->lop_num_pending == 0)
1865                 RETURN(0);
1866
1867         /* if we have an invalid import we want to drain the queued pages
1868          * by forcing them through rpcs that immediately fail and complete
1869          * the pages.  recovery relies on this to empty the queued pages
1870          * before canceling the locks and evicting down the llite pages */
1871         if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
1872                 RETURN(1);
1873
1874         /* stream rpcs in queue order as long as as there is an urgent page
1875          * queued.  this is our cheap solution for good batching in the case
1876          * where writepage marks some random page in the middle of the file
1877          * as urgent because of, say, memory pressure */
1878         if (!list_empty(&lop->lop_urgent)) {
1879                 CDEBUG(D_CACHE, "urgent request forcing RPC\n");
1880                 RETURN(1);
1881         }
1882         /* fire off rpcs when we have 'optimal' rpcs as tuned for the wire. */
1883         optimal = cli->cl_max_pages_per_rpc;
1884         if (cmd & OBD_BRW_WRITE) {
1885                 /* trigger a write rpc stream as long as there are dirtiers
1886                  * waiting for space.  as they're waiting, they're not going to
1887                  * create more pages to coallesce with what's waiting.. */
1888                 if (!list_empty(&cli->cl_cache_waiters)) {
1889                         CDEBUG(D_CACHE, "cache waiters forcing RPC\n");
1890                         RETURN(1);
1891                 }
1892                 /* +16 to avoid triggering rpcs that would want to include pages
1893                  * that are being queued but which can't be made ready until
1894                  * the queuer finishes with the page. this is a wart for
1895                  * llite::commit_write() */
1896                 optimal += 16;
1897         }
1898         if (lop->lop_num_pending >= optimal)
1899                 RETURN(1);
1900
1901         RETURN(0);
1902 }
1903
1904 static void on_list(struct list_head *item, struct list_head *list,
1905                     int should_be_on)
1906 {
1907         if (list_empty(item) && should_be_on)
1908                 list_add_tail(item, list);
1909         else if (!list_empty(item) && !should_be_on)
1910                 list_del_init(item);
1911 }
1912
1913 /* maintain the loi's cli list membership invariants so that osc_send_oap_rpc
1914  * can find pages to build into rpcs quickly */
1915 void loi_list_maint(struct client_obd *cli, struct lov_oinfo *loi)
1916 {
1917         on_list(&loi->loi_cli_item, &cli->cl_loi_ready_list,
1918                 lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE) ||
1919                 lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ));
1920
1921         on_list(&loi->loi_write_item, &cli->cl_loi_write_list,
1922                 loi->loi_write_lop.lop_num_pending);
1923
1924         on_list(&loi->loi_read_item, &cli->cl_loi_read_list,
1925                 loi->loi_read_lop.lop_num_pending);
1926 }
1927
1928 static void lop_update_pending(struct client_obd *cli,
1929                                struct loi_oap_pages *lop, int cmd, int delta)
1930 {
1931         lop->lop_num_pending += delta;
1932         if (cmd & OBD_BRW_WRITE)
1933                 cli->cl_pending_w_pages += delta;
1934         else
1935                 cli->cl_pending_r_pages += delta;
1936 }
1937
1938 /**
1939  * this is called when a sync waiter receives an interruption.  Its job is to
1940  * get the caller woken as soon as possible.  If its page hasn't been put in an
1941  * rpc yet it can dequeue immediately.  Otherwise it has to mark the rpc as
1942  * desiring interruption which will forcefully complete the rpc once the rpc
1943  * has timed out.
1944  */
1945 int osc_oap_interrupted(const struct lu_env *env, struct osc_async_page *oap)
1946 {
1947         struct loi_oap_pages *lop;
1948         struct lov_oinfo *loi;
1949         int rc = -EBUSY;
1950         ENTRY;
1951
1952         LASSERT(!oap->oap_interrupted);
1953         oap->oap_interrupted = 1;
1954
1955         /* ok, it's been put in an rpc. only one oap gets a request reference */
1956         if (oap->oap_request != NULL) {
1957                 ptlrpc_mark_interrupted(oap->oap_request);
1958                 ptlrpcd_wake(oap->oap_request);
1959                 ptlrpc_req_finished(oap->oap_request);
1960                 oap->oap_request = NULL;
1961         }
1962
1963         /*
1964          * page completion may be called only if ->cpo_prep() method was
1965          * executed by osc_io_submit(), that also adds page the to pending list
1966          */
1967         if (!list_empty(&oap->oap_pending_item)) {
1968                 list_del_init(&oap->oap_pending_item);
1969                 list_del_init(&oap->oap_urgent_item);
1970
1971                 loi = oap->oap_loi;
1972                 lop = (oap->oap_cmd & OBD_BRW_WRITE) ?
1973                         &loi->loi_write_lop : &loi->loi_read_lop;
1974                 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, -1);
1975                 loi_list_maint(oap->oap_cli, oap->oap_loi);
1976                 rc = oap->oap_caller_ops->ap_completion(env,
1977                                           oap->oap_caller_data,
1978                                           oap->oap_cmd, NULL, -EINTR);
1979         }
1980
1981         RETURN(rc);
1982 }
1983
1984 /* this is trying to propogate async writeback errors back up to the
1985  * application.  As an async write fails we record the error code for later if
1986  * the app does an fsync.  As long as errors persist we force future rpcs to be
1987  * sync so that the app can get a sync error and break the cycle of queueing
1988  * pages for which writeback will fail. */
1989 static void osc_process_ar(struct osc_async_rc *ar, __u64 xid,
1990                            int rc)
1991 {
1992         if (rc) {
1993                 if (!ar->ar_rc)
1994                         ar->ar_rc = rc;
1995
1996                 ar->ar_force_sync = 1;
1997                 ar->ar_min_xid = ptlrpc_sample_next_xid();
1998                 return;
1999
2000         }
2001
2002         if (ar->ar_force_sync && (xid >= ar->ar_min_xid))
2003                 ar->ar_force_sync = 0;
2004 }
2005
2006 void osc_oap_to_pending(struct osc_async_page *oap)
2007 {
2008         struct loi_oap_pages *lop;
2009
2010         if (oap->oap_cmd & OBD_BRW_WRITE)
2011                 lop = &oap->oap_loi->loi_write_lop;
2012         else
2013                 lop = &oap->oap_loi->loi_read_lop;
2014
2015         if (oap->oap_async_flags & ASYNC_URGENT)
2016                 list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2017         list_add_tail(&oap->oap_pending_item, &lop->lop_pending);
2018         lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, 1);
2019 }
2020
2021 /* this must be called holding the loi list lock to give coverage to exit_cache,
2022  * async_flag maintenance, and oap_request */
2023 static void osc_ap_completion(const struct lu_env *env,
2024                               struct client_obd *cli, struct obdo *oa,
2025                               struct osc_async_page *oap, int sent, int rc)
2026 {
2027         __u64 xid = 0;
2028
2029         ENTRY;
2030         if (oap->oap_request != NULL) {
2031                 xid = ptlrpc_req_xid(oap->oap_request);
2032                 ptlrpc_req_finished(oap->oap_request);
2033                 oap->oap_request = NULL;
2034         }
2035
2036         oap->oap_async_flags = 0;
2037         oap->oap_interrupted = 0;
2038
2039         if (oap->oap_cmd & OBD_BRW_WRITE) {
2040                 osc_process_ar(&cli->cl_ar, xid, rc);
2041                 osc_process_ar(&oap->oap_loi->loi_ar, xid, rc);
2042         }
2043
2044         if (rc == 0 && oa != NULL) {
2045                 if (oa->o_valid & OBD_MD_FLBLOCKS)
2046                         oap->oap_loi->loi_lvb.lvb_blocks = oa->o_blocks;
2047                 if (oa->o_valid & OBD_MD_FLMTIME)
2048                         oap->oap_loi->loi_lvb.lvb_mtime = oa->o_mtime;
2049                 if (oa->o_valid & OBD_MD_FLATIME)
2050                         oap->oap_loi->loi_lvb.lvb_atime = oa->o_atime;
2051                 if (oa->o_valid & OBD_MD_FLCTIME)
2052                         oap->oap_loi->loi_lvb.lvb_ctime = oa->o_ctime;
2053         }
2054
2055         rc = oap->oap_caller_ops->ap_completion(env, oap->oap_caller_data,
2056                                                 oap->oap_cmd, oa, rc);
2057
2058         /* ll_ap_completion (from llite) drops PG_locked. so, a new
2059          * I/O on the page could start, but OSC calls it under lock
2060          * and thus we can add oap back to pending safely */
2061         if (rc)
2062                 /* upper layer wants to leave the page on pending queue */
2063                 osc_oap_to_pending(oap);
2064         else
2065                 osc_exit_cache(cli, oap, sent);
2066         EXIT;
2067 }
2068
2069 static int brw_interpret(const struct lu_env *env,
2070                          struct ptlrpc_request *req, void *data, int rc)
2071 {
2072         struct osc_brw_async_args *aa = data;
2073         struct client_obd *cli;
2074         int async;
2075         ENTRY;
2076
2077         rc = osc_brw_fini_request(req, rc);
2078         CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
2079         if (osc_recoverable_error(rc)) {
2080                 rc = osc_brw_redo_request(req, aa);
2081                 if (rc == 0)
2082                         RETURN(0);
2083         }
2084
2085         if (aa->aa_ocapa) {
2086                 capa_put(aa->aa_ocapa);
2087                 aa->aa_ocapa = NULL;
2088         }
2089
2090         cli = aa->aa_cli;
2091
2092         client_obd_list_lock(&cli->cl_loi_list_lock);
2093
2094         /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2095          * is called so we know whether to go to sync BRWs or wait for more
2096          * RPCs to complete */
2097         if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
2098                 cli->cl_w_in_flight--;
2099         else
2100                 cli->cl_r_in_flight--;
2101
2102         async = list_empty(&aa->aa_oaps);
2103         if (!async) { /* from osc_send_oap_rpc() */
2104                 struct osc_async_page *oap, *tmp;
2105                 /* the caller may re-use the oap after the completion call so
2106                  * we need to clean it up a little */
2107                 list_for_each_entry_safe(oap, tmp, &aa->aa_oaps, oap_rpc_item) {
2108                         list_del_init(&oap->oap_rpc_item);
2109                         osc_ap_completion(env, cli, aa->aa_oa, oap, 1, rc);
2110                 }
2111                 OBDO_FREE(aa->aa_oa);
2112         } else { /* from async_internal() */
2113                 int i;
2114                 for (i = 0; i < aa->aa_page_count; i++)
2115                         osc_release_write_grant(aa->aa_cli, aa->aa_ppga[i], 1);
2116         }
2117         osc_wake_cache_waiters(cli);
2118         osc_check_rpcs(env, cli);
2119         client_obd_list_unlock(&cli->cl_loi_list_lock);
2120         if (!async)
2121                 cl_req_completion(env, aa->aa_clerq, rc);
2122         osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2123         RETURN(rc);
2124 }
2125
2126 static struct ptlrpc_request *osc_build_req(const struct lu_env *env,
2127                                             struct client_obd *cli,
2128                                             struct list_head *rpc_list,
2129                                             int page_count, int cmd)
2130 {
2131         struct ptlrpc_request *req;
2132         struct brw_page **pga = NULL;
2133         struct osc_brw_async_args *aa;
2134         struct obdo *oa = NULL;
2135         const struct obd_async_page_ops *ops = NULL;
2136         void *caller_data = NULL;
2137         struct osc_async_page *oap;
2138         struct osc_async_page *tmp;
2139         struct ost_body *body;
2140         struct cl_req *clerq = NULL;
2141         enum cl_req_type crt = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
2142         struct ldlm_lock *lock = NULL;
2143         struct cl_req_attr crattr;
2144         int i, rc;
2145
2146         ENTRY;
2147         LASSERT(!list_empty(rpc_list));
2148
2149         memset(&crattr, 0, sizeof crattr);
2150         OBD_ALLOC(pga, sizeof(*pga) * page_count);
2151         if (pga == NULL)
2152                 GOTO(out, req = ERR_PTR(-ENOMEM));
2153
2154         OBDO_ALLOC(oa);
2155         if (oa == NULL)
2156                 GOTO(out, req = ERR_PTR(-ENOMEM));
2157
2158         i = 0;
2159         list_for_each_entry(oap, rpc_list, oap_rpc_item) {
2160                 struct cl_page *page = osc_oap2cl_page(oap);
2161                 if (ops == NULL) {
2162                         ops = oap->oap_caller_ops;
2163                         caller_data = oap->oap_caller_data;
2164
2165                         clerq = cl_req_alloc(env, page, crt,
2166                                              1 /* only 1-object rpcs for
2167                                                 * now */);
2168                         if (IS_ERR(clerq))
2169                                 GOTO(out, req = (void *)clerq);
2170                         lock = oap->oap_ldlm_lock;
2171                 }
2172                 pga[i] = &oap->oap_brw_page;
2173                 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2174                 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
2175                        pga[i]->pg, cfs_page_index(oap->oap_page), oap, pga[i]->flag);
2176                 i++;
2177                 cl_req_page_add(env, clerq, page);
2178         }
2179
2180         /* always get the data for the obdo for the rpc */
2181         LASSERT(ops != NULL);
2182         crattr.cra_oa = oa;
2183         crattr.cra_capa = NULL;
2184         cl_req_attr_set(env, clerq, &crattr, ~0ULL);
2185         if (lock) {
2186                 oa->o_handle = lock->l_remote_handle;
2187                 oa->o_valid |= OBD_MD_FLHANDLE;
2188         }
2189
2190         rc = cl_req_prep(env, clerq);
2191         if (rc != 0) {
2192                 CERROR("cl_req_prep failed: %d\n", rc);
2193                 GOTO(out, req = ERR_PTR(rc));
2194         }
2195
2196         sort_brw_pages(pga, page_count);
2197         rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count,
2198                                   pga, &req, crattr.cra_capa, 1);
2199         if (rc != 0) {
2200                 CERROR("prep_req failed: %d\n", rc);
2201                 GOTO(out, req = ERR_PTR(rc));
2202         }
2203
2204         /* Need to update the timestamps after the request is built in case
2205          * we race with setattr (locally or in queue at OST).  If OST gets
2206          * later setattr before earlier BRW (as determined by the request xid),
2207          * the OST will not use BRW timestamps.  Sadly, there is no obvious
2208          * way to do this in a single call.  bug 10150 */
2209         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
2210         cl_req_attr_set(env, clerq, &crattr,
2211                         OBD_MD_FLMTIME|OBD_MD_FLCTIME|OBD_MD_FLATIME);
2212
2213         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2214         aa = ptlrpc_req_async_args(req);
2215         CFS_INIT_LIST_HEAD(&aa->aa_oaps);
2216         list_splice(rpc_list, &aa->aa_oaps);
2217         CFS_INIT_LIST_HEAD(rpc_list);
2218         aa->aa_clerq = clerq;
2219 out:
2220         capa_put(crattr.cra_capa);
2221         if (IS_ERR(req)) {
2222                 if (oa)
2223                         OBDO_FREE(oa);
2224                 if (pga)
2225                         OBD_FREE(pga, sizeof(*pga) * page_count);
2226                 /* this should happen rarely and is pretty bad, it makes the
2227                  * pending list not follow the dirty order */
2228                 client_obd_list_lock(&cli->cl_loi_list_lock);
2229                 list_for_each_entry_safe(oap, tmp, rpc_list, oap_rpc_item) {
2230                         list_del_init(&oap->oap_rpc_item);
2231
2232                         /* queued sync pages can be torn down while the pages
2233                          * were between the pending list and the rpc */
2234                         if (oap->oap_interrupted) {
2235                                 CDEBUG(D_INODE, "oap %p interrupted\n", oap);
2236                                 osc_ap_completion(env, cli, NULL, oap, 0,
2237                                                   oap->oap_count);
2238                                 continue;
2239                         }
2240                         osc_ap_completion(env, cli, NULL, oap, 0, PTR_ERR(req));
2241                 }
2242                 if (clerq && !IS_ERR(clerq))
2243                         cl_req_completion(env, clerq, PTR_ERR(req));
2244         }
2245         RETURN(req);
2246 }
2247
2248 /**
2249  * prepare pages for ASYNC io and put pages in send queue.
2250  *
2251  * \param cli -
2252  * \param loi -
2253  * \param cmd - OBD_BRW_* macroses
2254  * \param lop - pending pages
2255  *
2256  * \return zero if pages successfully add to send queue.
2257  * \return not zere if error occurring.
2258  */
2259 static int
2260 osc_send_oap_rpc(const struct lu_env *env, struct client_obd *cli,
2261                  struct lov_oinfo *loi,
2262                  int cmd, struct loi_oap_pages *lop)
2263 {
2264         struct ptlrpc_request *req;
2265         obd_count page_count = 0;
2266         struct osc_async_page *oap = NULL, *tmp;
2267         struct osc_brw_async_args *aa;
2268         const struct obd_async_page_ops *ops;
2269         CFS_LIST_HEAD(rpc_list);
2270         unsigned int ending_offset;
2271         unsigned  starting_offset = 0;
2272         int srvlock = 0;
2273         struct cl_object *clob = NULL;
2274         ENTRY;
2275
2276         /* first we find the pages we're allowed to work with */
2277         list_for_each_entry_safe(oap, tmp, &lop->lop_pending,
2278                                  oap_pending_item) {
2279                 ops = oap->oap_caller_ops;
2280
2281                 LASSERTF(oap->oap_magic == OAP_MAGIC, "Bad oap magic: oap %p, "
2282                          "magic 0x%x\n", oap, oap->oap_magic);
2283
2284                 if (clob == NULL) {
2285                         /* pin object in memory, so that completion call-backs
2286                          * can be safely called under client_obd_list lock. */
2287                         clob = osc_oap2cl_page(oap)->cp_obj;
2288                         cl_object_get(clob);
2289                 }
2290
2291                 if (page_count != 0 &&
2292                     srvlock != !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK)) {
2293                         CDEBUG(D_PAGE, "SRVLOCK flag mismatch,"
2294                                " oap %p, page %p, srvlock %u\n",
2295                                oap, oap->oap_brw_page.pg, (unsigned)!srvlock);
2296                         break;
2297                 }
2298                 /* in llite being 'ready' equates to the page being locked
2299                  * until completion unlocks it.  commit_write submits a page
2300                  * as not ready because its unlock will happen unconditionally
2301                  * as the call returns.  if we race with commit_write giving
2302                  * us that page we dont' want to create a hole in the page
2303                  * stream, so we stop and leave the rpc to be fired by
2304                  * another dirtier or kupdated interval (the not ready page
2305                  * will still be on the dirty list).  we could call in
2306                  * at the end of ll_file_write to process the queue again. */
2307                 if (!(oap->oap_async_flags & ASYNC_READY)) {
2308                         int rc = ops->ap_make_ready(env, oap->oap_caller_data,
2309                                                     cmd);
2310                         if (rc < 0)
2311                                 CDEBUG(D_INODE, "oap %p page %p returned %d "
2312                                                 "instead of ready\n", oap,
2313                                                 oap->oap_page, rc);
2314                         switch (rc) {
2315                         case -EAGAIN:
2316                                 /* llite is telling us that the page is still
2317                                  * in commit_write and that we should try
2318                                  * and put it in an rpc again later.  we
2319                                  * break out of the loop so we don't create
2320                                  * a hole in the sequence of pages in the rpc
2321                                  * stream.*/
2322                                 oap = NULL;
2323                                 break;
2324                         case -EINTR:
2325                                 /* the io isn't needed.. tell the checks
2326                                  * below to complete the rpc with EINTR */
2327                                 oap->oap_async_flags |= ASYNC_COUNT_STABLE;
2328                                 oap->oap_count = -EINTR;
2329                                 break;
2330                         case 0:
2331                                 oap->oap_async_flags |= ASYNC_READY;
2332                                 break;
2333                         default:
2334                                 LASSERTF(0, "oap %p page %p returned %d "
2335                                             "from make_ready\n", oap,
2336                                             oap->oap_page, rc);
2337                                 break;
2338                         }
2339                 }
2340                 if (oap == NULL)
2341                         break;
2342                 /*
2343                  * Page submitted for IO has to be locked. Either by
2344                  * ->ap_make_ready() or by higher layers.
2345                  */
2346 #if defined(__KERNEL__) && defined(__linux__)
2347                 {
2348                         struct cl_page *page;
2349
2350                         page = osc_oap2cl_page(oap);
2351
2352                         if (page->cp_type == CPT_CACHEABLE &&
2353                             !(PageLocked(oap->oap_page) &&
2354                               (CheckWriteback(oap->oap_page, cmd)))) {
2355                                 CDEBUG(D_PAGE, "page %p lost wb %lx/%x\n",
2356                                        oap->oap_page,
2357                                        (long)oap->oap_page->flags,
2358                                        oap->oap_async_flags);
2359                                 LBUG();
2360                         }
2361                 }
2362 #endif
2363                 /* If there is a gap at the start of this page, it can't merge
2364                  * with any previous page, so we'll hand the network a
2365                  * "fragmented" page array that it can't transfer in 1 RDMA */
2366                 if (page_count != 0 && oap->oap_page_off != 0)
2367                         break;
2368
2369                 /* take the page out of our book-keeping */
2370                 list_del_init(&oap->oap_pending_item);
2371                 lop_update_pending(cli, lop, cmd, -1);
2372                 list_del_init(&oap->oap_urgent_item);
2373
2374                 if (page_count == 0)
2375                         starting_offset = (oap->oap_obj_off+oap->oap_page_off) &
2376                                           (PTLRPC_MAX_BRW_SIZE - 1);
2377
2378                 /* ask the caller for the size of the io as the rpc leaves. */
2379                 if (!(oap->oap_async_flags & ASYNC_COUNT_STABLE)) {
2380                         oap->oap_count =
2381                                 ops->ap_refresh_count(env, oap->oap_caller_data,
2382                                                       cmd);
2383                         LASSERT(oap->oap_page_off + oap->oap_count <= CFS_PAGE_SIZE);
2384                 }
2385                 if (oap->oap_count <= 0) {
2386                         CDEBUG(D_CACHE, "oap %p count %d, completing\n", oap,
2387                                oap->oap_count);
2388                         osc_ap_completion(env, cli, NULL,
2389                                           oap, 0, oap->oap_count);
2390                         continue;
2391                 }
2392
2393                 /* now put the page back in our accounting */
2394                 list_add_tail(&oap->oap_rpc_item, &rpc_list);
2395                 if (page_count == 0)
2396                         srvlock = !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK);
2397                 if (++page_count >= cli->cl_max_pages_per_rpc)
2398                         break;
2399
2400                 /* End on a PTLRPC_MAX_BRW_SIZE boundary.  We want full-sized
2401                  * RPCs aligned on PTLRPC_MAX_BRW_SIZE boundaries to help reads
2402                  * have the same alignment as the initial writes that allocated
2403                  * extents on the server. */
2404                 ending_offset = (oap->oap_obj_off + oap->oap_page_off +
2405                                  oap->oap_count) & (PTLRPC_MAX_BRW_SIZE - 1);
2406                 if (ending_offset == 0)
2407                         break;
2408
2409                 /* If there is a gap at the end of this page, it can't merge
2410                  * with any subsequent pages, so we'll hand the network a
2411                  * "fragmented" page array that it can't transfer in 1 RDMA */
2412                 if (oap->oap_page_off + oap->oap_count < CFS_PAGE_SIZE)
2413                         break;
2414         }
2415
2416         osc_wake_cache_waiters(cli);
2417
2418         loi_list_maint(cli, loi);
2419
2420         client_obd_list_unlock(&cli->cl_loi_list_lock);
2421
2422         if (clob != NULL)
2423                 cl_object_put(env, clob);
2424
2425         if (page_count == 0) {
2426                 client_obd_list_lock(&cli->cl_loi_list_lock);
2427                 RETURN(0);
2428         }
2429
2430         req = osc_build_req(env, cli, &rpc_list, page_count, cmd);
2431         if (IS_ERR(req)) {
2432                 LASSERT(list_empty(&rpc_list));
2433                 loi_list_maint(cli, loi);
2434                 RETURN(PTR_ERR(req));
2435         }
2436
2437         aa = ptlrpc_req_async_args(req);
2438
2439         if (cmd == OBD_BRW_READ) {
2440                 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2441                 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2442                 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2443                                       (starting_offset >> CFS_PAGE_SHIFT) + 1);
2444         } else {
2445                 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2446                 lprocfs_oh_tally(&cli->cl_write_rpc_hist,
2447                                  cli->cl_w_in_flight);
2448                 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2449                                       (starting_offset >> CFS_PAGE_SHIFT) + 1);
2450         }
2451         ptlrpc_lprocfs_brw(req, aa->aa_requested_nob);
2452
2453         client_obd_list_lock(&cli->cl_loi_list_lock);
2454
2455         if (cmd == OBD_BRW_READ)
2456                 cli->cl_r_in_flight++;
2457         else
2458                 cli->cl_w_in_flight++;
2459
2460         /* queued sync pages can be torn down while the pages
2461          * were between the pending list and the rpc */
2462         tmp = NULL;
2463         list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2464                 /* only one oap gets a request reference */
2465                 if (tmp == NULL)
2466                         tmp = oap;
2467                 if (oap->oap_interrupted && !req->rq_intr) {
2468                         CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2469                                oap, req);
2470                         ptlrpc_mark_interrupted(req);
2471                 }
2472         }
2473         if (tmp != NULL)
2474                 tmp->oap_request = ptlrpc_request_addref(req);
2475
2476         DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2477                   page_count, aa, cli->cl_r_in_flight, cli->cl_w_in_flight);
2478
2479         req->rq_interpret_reply = brw_interpret;
2480         ptlrpcd_add_req(req, PSCOPE_BRW);
2481         RETURN(1);
2482 }
2483
2484 #define LOI_DEBUG(LOI, STR, args...)                                     \
2485         CDEBUG(D_INODE, "loi ready %d wr %d:%d rd %d:%d " STR,           \
2486                !list_empty(&(LOI)->loi_cli_item),                        \
2487                (LOI)->loi_write_lop.lop_num_pending,                     \
2488                !list_empty(&(LOI)->loi_write_lop.lop_urgent),            \
2489                (LOI)->loi_read_lop.lop_num_pending,                      \
2490                !list_empty(&(LOI)->loi_read_lop.lop_urgent),             \
2491                args)                                                     \
2492
2493 /* This is called by osc_check_rpcs() to find which objects have pages that
2494  * we could be sending.  These lists are maintained by lop_makes_rpc(). */
2495 struct lov_oinfo *osc_next_loi(struct client_obd *cli)
2496 {
2497         ENTRY;
2498         /* first return all objects which we already know to have
2499          * pages ready to be stuffed into rpcs */
2500         if (!list_empty(&cli->cl_loi_ready_list))
2501                 RETURN(list_entry(cli->cl_loi_ready_list.next,
2502                                   struct lov_oinfo, loi_cli_item));
2503
2504         /* then if we have cache waiters, return all objects with queued
2505          * writes.  This is especially important when many small files
2506          * have filled up the cache and not been fired into rpcs because
2507          * they don't pass the nr_pending/object threshhold */
2508         if (!list_empty(&cli->cl_cache_waiters) &&
2509             !list_empty(&cli->cl_loi_write_list))
2510                 RETURN(list_entry(cli->cl_loi_write_list.next,
2511                                   struct lov_oinfo, loi_write_item));
2512
2513         /* then return all queued objects when we have an invalid import
2514          * so that they get flushed */
2515         if (cli->cl_import == NULL || cli->cl_import->imp_invalid) {
2516                 if (!list_empty(&cli->cl_loi_write_list))
2517                         RETURN(list_entry(cli->cl_loi_write_list.next,
2518                                           struct lov_oinfo, loi_write_item));
2519                 if (!list_empty(&cli->cl_loi_read_list))
2520                         RETURN(list_entry(cli->cl_loi_read_list.next,
2521                                           struct lov_oinfo, loi_read_item));
2522         }
2523         RETURN(NULL);
2524 }
2525
2526 /* called with the loi list lock held */
2527 void osc_check_rpcs(const struct lu_env *env, struct client_obd *cli)
2528 {
2529         struct lov_oinfo *loi;
2530         int rc = 0, race_counter = 0;
2531         ENTRY;
2532
2533         while ((loi = osc_next_loi(cli)) != NULL) {
2534                 LOI_DEBUG(loi, "%lu in flight\n", rpcs_in_flight(cli));
2535
2536                 if (rpcs_in_flight(cli) >= cli->cl_max_rpcs_in_flight)
2537                         break;
2538
2539                 /* attempt some read/write balancing by alternating between
2540                  * reads and writes in an object.  The makes_rpc checks here
2541                  * would be redundant if we were getting read/write work items
2542                  * instead of objects.  we don't want send_oap_rpc to drain a
2543                  * partial read pending queue when we're given this object to
2544                  * do io on writes while there are cache waiters */
2545                 if (lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)) {
2546                         rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_WRITE,
2547                                               &loi->loi_write_lop);
2548                         if (rc < 0)
2549                                 break;
2550                         if (rc > 0)
2551                                 race_counter = 0;
2552                         else
2553                                 race_counter++;
2554                 }
2555                 if (lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ)) {
2556                         rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_READ,
2557                                               &loi->loi_read_lop);
2558                         if (rc < 0)
2559                                 break;
2560                         if (rc > 0)
2561                                 race_counter = 0;
2562                         else
2563                                 race_counter++;
2564                 }
2565
2566                 /* attempt some inter-object balancing by issueing rpcs
2567                  * for each object in turn */
2568                 if (!list_empty(&loi->loi_cli_item))
2569                         list_del_init(&loi->loi_cli_item);
2570                 if (!list_empty(&loi->loi_write_item))
2571                         list_del_init(&loi->loi_write_item);
2572                 if (!list_empty(&loi->loi_read_item))
2573                         list_del_init(&loi->loi_read_item);
2574
2575                 loi_list_maint(cli, loi);
2576
2577                 /* send_oap_rpc fails with 0 when make_ready tells it to
2578                  * back off.  llite's make_ready does this when it tries
2579                  * to lock a page queued for write that is already locked.
2580                  * we want to try sending rpcs from many objects, but we
2581                  * don't want to spin failing with 0.  */
2582                 if (race_counter == 10)
2583                         break;
2584         }
2585         EXIT;
2586 }
2587
2588 /* we're trying to queue a page in the osc so we're subject to the
2589  * 'cl_dirty_max' limit on the number of pages that can be queued in the osc.
2590  * If the osc's queued pages are already at that limit, then we want to sleep
2591  * until there is space in the osc's queue for us.  We also may be waiting for
2592  * write credits from the OST if there are RPCs in flight that may return some
2593  * before we fall back to sync writes.
2594  *
2595  * We need this know our allocation was granted in the presence of signals */
2596 static int ocw_granted(struct client_obd *cli, struct osc_cache_waiter *ocw)
2597 {
2598         int rc;
2599         ENTRY;
2600         client_obd_list_lock(&cli->cl_loi_list_lock);
2601         rc = list_empty(&ocw->ocw_entry) || rpcs_in_flight(cli) == 0;
2602         client_obd_list_unlock(&cli->cl_loi_list_lock);
2603         RETURN(rc);
2604 };
2605
2606 /**
2607  * Non-blocking version of osc_enter_cache() that consumes grant only when it
2608  * is available.
2609  */
2610 int osc_enter_cache_try(const struct lu_env *env,
2611                         struct client_obd *cli, struct lov_oinfo *loi,
2612                         struct osc_async_page *oap, int transient)
2613 {
2614         int has_grant;
2615
2616         has_grant = cli->cl_avail_grant >= CFS_PAGE_SIZE;
2617         if (has_grant) {
2618                 osc_consume_write_grant(cli, &oap->oap_brw_page);
2619                 if (transient) {
2620                         cli->cl_dirty_transit += CFS_PAGE_SIZE;
2621                         atomic_inc(&obd_dirty_transit_pages);
2622                         oap->oap_brw_flags |= OBD_BRW_NOCACHE;
2623                 }
2624         }
2625         return has_grant;
2626 }
2627
2628 /* Caller must hold loi_list_lock - we drop/regain it if we need to wait for
2629  * grant or cache space. */
2630 static int osc_enter_cache(const struct lu_env *env,
2631                            struct client_obd *cli, struct lov_oinfo *loi,
2632                            struct osc_async_page *oap)
2633 {
2634         struct osc_cache_waiter ocw;
2635         struct l_wait_info lwi = { 0 };
2636
2637         ENTRY;
2638
2639         CDEBUG(D_CACHE, "dirty: %ld/%d dirty_max: %ld/%d dropped: %lu "
2640                "grant: %lu\n", cli->cl_dirty, atomic_read(&obd_dirty_pages),
2641                cli->cl_dirty_max, obd_max_dirty_pages,
2642                cli->cl_lost_grant, cli->cl_avail_grant);
2643
2644         /* force the caller to try sync io.  this can jump the list
2645          * of queued writes and create a discontiguous rpc stream */
2646         if (cli->cl_dirty_max < CFS_PAGE_SIZE || cli->cl_ar.ar_force_sync ||
2647             loi->loi_ar.ar_force_sync)
2648                 RETURN(-EDQUOT);
2649
2650         /* Hopefully normal case - cache space and write credits available */
2651         if (cli->cl_dirty + CFS_PAGE_SIZE <= cli->cl_dirty_max &&
2652             atomic_read(&obd_dirty_pages) + 1 <= obd_max_dirty_pages &&
2653             osc_enter_cache_try(env, cli, loi, oap, 0))
2654                 RETURN(0);
2655
2656         /* Make sure that there are write rpcs in flight to wait for.  This
2657          * is a little silly as this object may not have any pending but
2658          * other objects sure might. */
2659         if (cli->cl_w_in_flight) {
2660                 list_add_tail(&ocw.ocw_entry, &cli->cl_cache_waiters);
2661                 cfs_waitq_init(&ocw.ocw_waitq);
2662                 ocw.ocw_oap = oap;
2663                 ocw.ocw_rc = 0;
2664
2665                 loi_list_maint(cli, loi);
2666                 osc_check_rpcs(env, cli);
2667                 client_obd_list_unlock(&cli->cl_loi_list_lock);
2668
2669                 CDEBUG(D_CACHE, "sleeping for cache space\n");
2670                 l_wait_event(ocw.ocw_waitq, ocw_granted(cli, &ocw), &lwi);
2671
2672                 client_obd_list_lock(&cli->cl_loi_list_lock);
2673                 if (!list_empty(&ocw.ocw_entry)) {
2674                         list_del(&ocw.ocw_entry);
2675                         RETURN(-EINTR);
2676                 }
2677                 RETURN(ocw.ocw_rc);
2678         }
2679
2680         RETURN(-EDQUOT);
2681 }
2682
2683
2684 int osc_prep_async_page(struct obd_export *exp, struct lov_stripe_md *lsm,
2685                         struct lov_oinfo *loi, cfs_page_t *page,
2686                         obd_off offset, const struct obd_async_page_ops *ops,
2687                         void *data, void **res, int nocache,
2688                         struct lustre_handle *lockh)
2689 {
2690         struct osc_async_page *oap;
2691
2692         ENTRY;
2693
2694         if (!page)
2695                 return size_round(sizeof(*oap));
2696
2697         oap = *res;
2698         oap->oap_magic = OAP_MAGIC;
2699         oap->oap_cli = &exp->exp_obd->u.cli;
2700         oap->oap_loi = loi;
2701
2702         oap->oap_caller_ops = ops;
2703         oap->oap_caller_data = data;
2704
2705         oap->oap_page = page;
2706         oap->oap_obj_off = offset;
2707         if (!client_is_remote(exp) &&
2708             cfs_capable(CFS_CAP_SYS_RESOURCE))
2709                 oap->oap_brw_flags = OBD_BRW_NOQUOTA;
2710
2711         LASSERT(!(offset & ~CFS_PAGE_MASK));
2712
2713         CFS_INIT_LIST_HEAD(&oap->oap_pending_item);
2714         CFS_INIT_LIST_HEAD(&oap->oap_urgent_item);
2715         CFS_INIT_LIST_HEAD(&oap->oap_rpc_item);
2716         CFS_INIT_LIST_HEAD(&oap->oap_page_list);
2717
2718         spin_lock_init(&oap->oap_lock);
2719         CDEBUG(D_CACHE, "oap %p page %p obj off "LPU64"\n", oap, page, offset);
2720         RETURN(0);
2721 }
2722
2723 struct osc_async_page *oap_from_cookie(void *cookie)
2724 {
2725         struct osc_async_page *oap = cookie;
2726         if (oap->oap_magic != OAP_MAGIC)
2727                 return ERR_PTR(-EINVAL);
2728         return oap;
2729 };
2730
2731 int osc_queue_async_io(const struct lu_env *env,
2732                        struct obd_export *exp, struct lov_stripe_md *lsm,
2733                        struct lov_oinfo *loi, void *cookie,
2734                        int cmd, obd_off off, int count,
2735                        obd_flag brw_flags, enum async_flags async_flags)
2736 {
2737         struct client_obd *cli = &exp->exp_obd->u.cli;
2738         struct osc_async_page *oap;
2739         int rc = 0;
2740         ENTRY;
2741
2742         oap = oap_from_cookie(cookie);
2743         if (IS_ERR(oap))
2744                 RETURN(PTR_ERR(oap));
2745
2746         if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
2747                 RETURN(-EIO);
2748
2749         if (!list_empty(&oap->oap_pending_item) ||
2750             !list_empty(&oap->oap_urgent_item) ||
2751             !list_empty(&oap->oap_rpc_item))
2752                 RETURN(-EBUSY);
2753
2754         /* check if the file's owner/group is over quota */
2755         if ((cmd & OBD_BRW_WRITE) && !(cmd & OBD_BRW_NOQUOTA)) {
2756                 struct cl_object *obj;
2757                 struct cl_attr    attr; /* XXX put attr into thread info */
2758
2759                 obj = cl_object_top(osc_oap2cl_page(oap)->cp_obj);
2760
2761                 cl_object_attr_lock(obj);
2762                 rc = cl_object_attr_get(env, obj, &attr);
2763                 cl_object_attr_unlock(obj);
2764
2765                 if (rc == 0 && lquota_chkdq(quota_interface, cli, attr.cat_uid,
2766                                             attr.cat_gid) == NO_QUOTA)
2767                         rc = -EDQUOT;
2768                 if (rc)
2769                         RETURN(rc);
2770         }
2771
2772         if (loi == NULL)
2773                 loi = lsm->lsm_oinfo[0];
2774
2775         client_obd_list_lock(&cli->cl_loi_list_lock);
2776
2777         LASSERT(off + count <= CFS_PAGE_SIZE);
2778         oap->oap_cmd = cmd;
2779         oap->oap_page_off = off;
2780         oap->oap_count = count;
2781         oap->oap_brw_flags = brw_flags;
2782         oap->oap_async_flags = async_flags;
2783
2784         if (cmd & OBD_BRW_WRITE) {
2785                 rc = osc_enter_cache(env, cli, loi, oap);
2786                 if (rc) {
2787                         client_obd_list_unlock(&cli->cl_loi_list_lock);
2788                         RETURN(rc);
2789                 }
2790         }
2791
2792         osc_oap_to_pending(oap);
2793         loi_list_maint(cli, loi);
2794
2795         LOI_DEBUG(loi, "oap %p page %p added for cmd %d\n", oap, oap->oap_page,
2796                   cmd);
2797
2798         osc_check_rpcs(env, cli);
2799         client_obd_list_unlock(&cli->cl_loi_list_lock);
2800
2801         RETURN(0);
2802 }
2803
2804 /* aka (~was & now & flag), but this is more clear :) */
2805 #define SETTING(was, now, flag) (!(was & flag) && (now & flag))
2806
2807 int osc_set_async_flags_base(struct client_obd *cli,
2808                              struct lov_oinfo *loi, struct osc_async_page *oap,
2809                              obd_flag async_flags)
2810 {
2811         struct loi_oap_pages *lop;
2812         ENTRY;
2813
2814         if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
2815                 RETURN(-EIO);
2816
2817         if (oap->oap_cmd & OBD_BRW_WRITE) {
2818                 lop = &loi->loi_write_lop;
2819         } else {
2820                 lop = &loi->loi_read_lop;
2821         }
2822
2823         if (list_empty(&oap->oap_pending_item))
2824                 RETURN(-EINVAL);
2825
2826         if ((oap->oap_async_flags & async_flags) == async_flags)
2827                 RETURN(0);
2828
2829         if (SETTING(oap->oap_async_flags, async_flags, ASYNC_READY))
2830                 oap->oap_async_flags |= ASYNC_READY;
2831
2832         if (SETTING(oap->oap_async_flags, async_flags, ASYNC_URGENT)) {
2833                 if (list_empty(&oap->oap_rpc_item)) {
2834                         list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2835                         loi_list_maint(cli, loi);
2836                 }
2837         }
2838
2839         LOI_DEBUG(loi, "oap %p page %p has flags %x\n", oap, oap->oap_page,
2840                         oap->oap_async_flags);
2841         RETURN(0);
2842 }
2843
2844 int osc_teardown_async_page(struct obd_export *exp,
2845                             struct lov_stripe_md *lsm,
2846                             struct lov_oinfo *loi, void *cookie)
2847 {
2848         struct client_obd *cli = &exp->exp_obd->u.cli;
2849         struct loi_oap_pages *lop;
2850         struct osc_async_page *oap;
2851         int rc = 0;
2852         ENTRY;
2853
2854         oap = oap_from_cookie(cookie);
2855         if (IS_ERR(oap))
2856                 RETURN(PTR_ERR(oap));
2857
2858         if (loi == NULL)
2859                 loi = lsm->lsm_oinfo[0];
2860
2861         if (oap->oap_cmd & OBD_BRW_WRITE) {
2862                 lop = &loi->loi_write_lop;
2863         } else {
2864                 lop = &loi->loi_read_lop;
2865         }
2866
2867         client_obd_list_lock(&cli->cl_loi_list_lock);
2868
2869         if (!list_empty(&oap->oap_rpc_item))
2870                 GOTO(out, rc = -EBUSY);
2871
2872         osc_exit_cache(cli, oap, 0);
2873         osc_wake_cache_waiters(cli);
2874
2875         if (!list_empty(&oap->oap_urgent_item)) {
2876                 list_del_init(&oap->oap_urgent_item);
2877                 oap->oap_async_flags &= ~ASYNC_URGENT;
2878         }
2879         if (!list_empty(&oap->oap_pending_item)) {
2880                 list_del_init(&oap->oap_pending_item);
2881                 lop_update_pending(cli, lop, oap->oap_cmd, -1);
2882         }
2883         loi_list_maint(cli, loi);
2884         LOI_DEBUG(loi, "oap %p page %p torn down\n", oap, oap->oap_page);
2885 out:
2886         client_obd_list_unlock(&cli->cl_loi_list_lock);
2887         RETURN(rc);
2888 }
2889
2890 static void osc_set_lock_data_with_check(struct ldlm_lock *lock,
2891                                          struct ldlm_enqueue_info *einfo,
2892                                          int flags)
2893 {
2894         void *data = einfo->ei_cbdata;
2895
2896         LASSERT(lock != NULL);
2897         LASSERT(lock->l_blocking_ast == einfo->ei_cb_bl);
2898         LASSERT(lock->l_resource->lr_type == einfo->ei_type);
2899         LASSERT(lock->l_completion_ast == einfo->ei_cb_cp);
2900         LASSERT(lock->l_glimpse_ast == einfo->ei_cb_gl);
2901
2902         lock_res_and_lock(lock);
2903         spin_lock(&osc_ast_guard);
2904         LASSERT(lock->l_ast_data == NULL || lock->l_ast_data == data);
2905         lock->l_ast_data = data;
2906         spin_unlock(&osc_ast_guard);
2907         unlock_res_and_lock(lock);
2908 }
2909
2910 static void osc_set_data_with_check(struct lustre_handle *lockh,
2911                                     struct ldlm_enqueue_info *einfo,
2912                                     int flags)
2913 {
2914         struct ldlm_lock *lock = ldlm_handle2lock(lockh);
2915
2916         if (lock != NULL) {
2917                 osc_set_lock_data_with_check(lock, einfo, flags);
2918                 LDLM_LOCK_PUT(lock);
2919         } else
2920                 CERROR("lockh %p, data %p - client evicted?\n",
2921                        lockh, einfo->ei_cbdata);
2922 }
2923
2924 static int osc_change_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
2925                              ldlm_iterator_t replace, void *data)
2926 {
2927         struct ldlm_res_id res_id;
2928         struct obd_device *obd = class_exp2obd(exp);
2929
2930         osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_gr, &res_id);
2931         ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
2932         return 0;
2933 }
2934
2935 static int osc_enqueue_fini(struct ptlrpc_request *req, struct ost_lvb *lvb,
2936                             obd_enqueue_update_f upcall, void *cookie,
2937                             int *flags, int rc)
2938 {
2939         int intent = *flags & LDLM_FL_HAS_INTENT;
2940         ENTRY;
2941
2942         if (intent) {
2943                 /* The request was created before ldlm_cli_enqueue call. */
2944                 if (rc == ELDLM_LOCK_ABORTED) {
2945                         struct ldlm_reply *rep;
2946                         rep = req_capsule_server_get(&req->rq_pill,
2947                                                      &RMF_DLM_REP);
2948
2949                         LASSERT(rep != NULL);
2950                         if (rep->lock_policy_res1)
2951                                 rc = rep->lock_policy_res1;
2952                 }
2953         }
2954
2955         if ((intent && rc == ELDLM_LOCK_ABORTED) || !rc) {
2956                 *flags |= LDLM_FL_LVB_READY;
2957                 CDEBUG(D_INODE,"got kms "LPU64" blocks "LPU64" mtime "LPU64"\n",
2958                        lvb->lvb_size, lvb->lvb_blocks, lvb->lvb_mtime);
2959         }
2960
2961         /* Call the update callback. */
2962         rc = (*upcall)(cookie, rc);
2963         RETURN(rc);
2964 }
2965
2966 static int osc_enqueue_interpret(const struct lu_env *env,
2967                                  struct ptlrpc_request *req,
2968                                  struct osc_enqueue_args *aa, int rc)
2969 {
2970         struct ldlm_lock *lock;
2971         struct lustre_handle handle;
2972         __u32 mode;
2973
2974         /* Make a local copy of a lock handle and a mode, because aa->oa_*
2975          * might be freed anytime after lock upcall has been called. */
2976         lustre_handle_copy(&handle, aa->oa_lockh);
2977         mode = aa->oa_ei->ei_mode;
2978
2979         /* ldlm_cli_enqueue is holding a reference on the lock, so it must
2980          * be valid. */
2981         lock = ldlm_handle2lock(&handle);
2982
2983         /* Take an additional reference so that a blocking AST that
2984          * ldlm_cli_enqueue_fini() might post for a failed lock, is guaranteed
2985          * to arrive after an upcall has been executed by
2986          * osc_enqueue_fini(). */
2987         ldlm_lock_addref(&handle, mode);
2988
2989         /* Complete obtaining the lock procedure. */
2990         rc = ldlm_cli_enqueue_fini(aa->oa_exp, req, aa->oa_ei->ei_type, 1,
2991                                    mode, aa->oa_flags, aa->oa_lvb,
2992                                    sizeof(*aa->oa_lvb), lustre_swab_ost_lvb,
2993                                    &handle, rc);
2994         /* Complete osc stuff. */
2995         rc = osc_enqueue_fini(req, aa->oa_lvb,
2996                               aa->oa_upcall, aa->oa_cookie, aa->oa_flags, rc);
2997         /* Release the lock for async request. */
2998         if (lustre_handle_is_used(&handle) && rc == ELDLM_OK)
2999                 /*
3000                  * Releases a reference taken by ldlm_cli_enqueue(), if it is
3001                  * not already released by
3002                  * ldlm_cli_enqueue_fini()->failed_lock_cleanup()
3003                  */
3004                 ldlm_lock_decref(&handle, mode);
3005
3006         LASSERTF(lock != NULL, "lockh %p, req %p, aa %p - client evicted?\n",
3007                  aa->oa_lockh, req, aa);
3008         ldlm_lock_decref(&handle, mode);
3009         LDLM_LOCK_PUT(lock);
3010         return rc;
3011 }
3012
3013 void osc_update_enqueue(struct lustre_handle *lov_lockhp,
3014                         struct lov_oinfo *loi, int flags,
3015                         struct ost_lvb *lvb, __u32 mode, int rc)
3016 {
3017         if (rc == ELDLM_OK) {
3018                 struct ldlm_lock *lock = ldlm_handle2lock(lov_lockhp);
3019                 __u64 tmp;
3020
3021                 LASSERT(lock != NULL);
3022                 loi->loi_lvb = *lvb;
3023                 tmp = loi->loi_lvb.lvb_size;
3024                 /* Extend KMS up to the end of this lock and no further
3025                  * A lock on [x,y] means a KMS of up to y + 1 bytes! */
3026                 if (tmp > lock->l_policy_data.l_extent.end)
3027                         tmp = lock->l_policy_data.l_extent.end + 1;
3028                 if (tmp >= loi->loi_kms) {
3029                         LDLM_DEBUG(lock, "lock acquired, setting rss="LPU64
3030                                    ", kms="LPU64, loi->loi_lvb.lvb_size, tmp);
3031                         loi_kms_set(loi, tmp);
3032                 } else {
3033                         LDLM_DEBUG(lock, "lock acquired, setting rss="
3034                                    LPU64"; leaving kms="LPU64", end="LPU64,
3035                                    loi->loi_lvb.lvb_size, loi->loi_kms,
3036                                    lock->l_policy_data.l_extent.end);
3037                 }
3038                 ldlm_lock_allow_match(lock);
3039                 LDLM_LOCK_PUT(lock);
3040         } else if (rc == ELDLM_LOCK_ABORTED && (flags & LDLM_FL_HAS_INTENT)) {
3041                 loi->loi_lvb = *lvb;
3042                 CDEBUG(D_INODE, "glimpsed, setting rss="LPU64"; leaving"
3043                        " kms="LPU64"\n", loi->loi_lvb.lvb_size, loi->loi_kms);
3044                 rc = ELDLM_OK;
3045         }
3046 }
3047 EXPORT_SYMBOL(osc_update_enqueue);
3048
3049 struct ptlrpc_request_set *PTLRPCD_SET = (void *)1;
3050
3051 /* When enqueuing asynchronously, locks are not ordered, we can obtain a lock
3052  * from the 2nd OSC before a lock from the 1st one. This does not deadlock with
3053  * other synchronous requests, however keeping some locks and trying to obtain
3054  * others may take a considerable amount of time in a case of ost failure; and
3055  * when other sync requests do not get released lock from a client, the client
3056  * is excluded from the cluster -- such scenarious make the life difficult, so
3057  * release locks just after they are obtained. */
3058 int osc_enqueue_base(struct obd_export *exp, struct ldlm_res_id *res_id,
3059                      int *flags, ldlm_policy_data_t *policy,
3060                      struct ost_lvb *lvb, int kms_valid,
3061                      obd_enqueue_update_f upcall, void *cookie,
3062                      struct ldlm_enqueue_info *einfo,
3063                      struct lustre_handle *lockh,
3064                      struct ptlrpc_request_set *rqset, int async)
3065 {
3066         struct obd_device *obd = exp->exp_obd;
3067         struct ptlrpc_request *req = NULL;
3068         int intent = *flags & LDLM_FL_HAS_INTENT;
3069         ldlm_mode_t mode;
3070         int rc;
3071         ENTRY;
3072
3073         /* Filesystem lock extents are extended to page boundaries so that
3074          * dealing with the page cache is a little smoother.  */
3075         policy->l_extent.start -= policy->l_extent.start & ~CFS_PAGE_MASK;
3076         policy->l_extent.end |= ~CFS_PAGE_MASK;
3077
3078         /*
3079          * kms is not valid when either object is completely fresh (so that no
3080          * locks are cached), or object was evicted. In the latter case cached
3081          * lock cannot be used, because it would prime inode state with
3082          * potentially stale LVB.
3083          */
3084         if (!kms_valid)
3085                 goto no_match;
3086
3087         /* Next, search for already existing extent locks that will cover us */
3088         /* If we're trying to read, we also search for an existing PW lock.  The
3089          * VFS and page cache already protect us locally, so lots of readers/
3090          * writers can share a single PW lock.
3091          *
3092          * There are problems with conversion deadlocks, so instead of
3093          * converting a read lock to a write lock, we'll just enqueue a new
3094          * one.
3095          *
3096          * At some point we should cancel the read lock instead of making them
3097          * send us a blocking callback, but there are problems with canceling
3098          * locks out from other users right now, too. */
3099         mode = einfo->ei_mode;
3100         if (einfo->ei_mode == LCK_PR)
3101                 mode |= LCK_PW;
3102         mode = ldlm_lock_match(obd->obd_namespace,
3103                                *flags | LDLM_FL_LVB_READY, res_id,
3104                                einfo->ei_type, policy, mode, lockh, 0);
3105         if (mode) {
3106                 struct ldlm_lock *matched = ldlm_handle2lock(lockh);
3107
3108                 if (matched->l_ast_data == NULL ||
3109                     matched->l_ast_data == einfo->ei_cbdata) {
3110                         /* addref the lock only if not async requests and PW
3111                          * lock is matched whereas we asked for PR. */
3112                         if (!rqset && einfo->ei_mode != mode)
3113                                 ldlm_lock_addref(lockh, LCK_PR);
3114                         osc_set_lock_data_with_check(matched, einfo, *flags);
3115                         if (intent) {
3116                                 /* I would like to be able to ASSERT here that
3117                                  * rss <= kms, but I can't, for reasons which
3118                                  * are explained in lov_enqueue() */
3119                         }
3120
3121                         /* We already have a lock, and it's referenced */
3122                         (*upcall)(cookie, ELDLM_OK);
3123
3124                         /* For async requests, decref the lock. */
3125                         if (einfo->ei_mode != mode)
3126                                 ldlm_lock_decref(lockh, LCK_PW);
3127                         else if (rqset)
3128                                 ldlm_lock_decref(lockh, einfo->ei_mode);
3129                         LDLM_LOCK_PUT(matched);
3130                         RETURN(ELDLM_OK);
3131                 } else
3132                         ldlm_lock_decref(lockh, mode);
3133                 LDLM_LOCK_PUT(matched);
3134         }
3135
3136  no_match:
3137         if (intent) {
3138                 CFS_LIST_HEAD(cancels);
3139                 req = ptlrpc_request_alloc(class_exp2cliimp(exp),
3140                                            &RQF_LDLM_ENQUEUE_LVB);
3141                 if (req == NULL)
3142                         RETURN(-ENOMEM);
3143
3144                 rc = ldlm_prep_enqueue_req(exp, req, &cancels, 0);
3145                 if (rc)
3146                         RETURN(rc);
3147
3148                 req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER,
3149                                      sizeof *lvb);
3150                 ptlrpc_request_set_replen(req);
3151         }
3152
3153         /* users of osc_enqueue() can pass this flag for ldlm_lock_match() */
3154         *flags &= ~LDLM_FL_BLOCK_GRANTED;
3155
3156         rc = ldlm_cli_enqueue(exp, &req, einfo, res_id, policy, flags, lvb,
3157                               sizeof(*lvb), lustre_swab_ost_lvb, lockh, async);
3158         if (rqset) {
3159                 if (!rc) {
3160                         struct osc_enqueue_args *aa;
3161                         CLASSERT (sizeof(*aa) <= sizeof(req->rq_async_args));
3162                         aa = ptlrpc_req_async_args(req);
3163                         aa->oa_ei = einfo;
3164                         aa->oa_exp = exp;
3165                         aa->oa_flags  = flags;
3166                         aa->oa_upcall = upcall;
3167                         aa->oa_cookie = cookie;
3168                         aa->oa_lvb    =&nb