Whamcloud - gitweb
Revert "b=22176 Add .sync_fs super block handler"
[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 (c) 2002, 2010, Oracle and/or its affiliates. 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_SEQ_IS_MDT(lsm->lsm_object_seq);
100                 (*lmmp)->lmm_object_id = cpu_to_le64(lsm->lsm_object_id);
101                 (*lmmp)->lmm_object_seq = cpu_to_le64(lsm->lsm_object_seq);
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_seq = le64_to_cpu (lmm->lmm_object_seq);
155                 LASSERT((*lsmp)->lsm_object_id);
156                 LASSERT_SEQ_IS_MDT((*lsmp)->lsm_object_seq);
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         lustre_set_wire_obdo(&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 = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
214         if (body) {
215                 CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
216                 lustre_get_wire_obdo(aa->aa_oi->oi_oa, &body->oa);
217
218                 /* This should really be sent by the OST */
219                 aa->aa_oi->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
220                 aa->aa_oi->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
221         } else {
222                 CDEBUG(D_INFO, "can't unpack ost_body\n");
223                 rc = -EPROTO;
224                 aa->aa_oi->oi_oa->o_valid = 0;
225         }
226 out:
227         rc = aa->aa_oi->oi_cb_up(aa->aa_oi, rc);
228         RETURN(rc);
229 }
230
231 static int osc_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
232                              struct ptlrpc_request_set *set)
233 {
234         struct ptlrpc_request *req;
235         struct osc_async_args *aa;
236         int                    rc;
237         ENTRY;
238
239         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
240         if (req == NULL)
241                 RETURN(-ENOMEM);
242
243         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
244         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
245         if (rc) {
246                 ptlrpc_request_free(req);
247                 RETURN(rc);
248         }
249
250         osc_pack_req_body(req, oinfo);
251
252         ptlrpc_request_set_replen(req);
253         req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_getattr_interpret;
254
255         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
256         aa = ptlrpc_req_async_args(req);
257         aa->aa_oi = oinfo;
258
259         ptlrpc_set_add_req(set, req);
260         RETURN(0);
261 }
262
263 static int osc_getattr(struct obd_export *exp, struct obd_info *oinfo)
264 {
265         struct ptlrpc_request *req;
266         struct ost_body       *body;
267         int                    rc;
268         ENTRY;
269
270         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_GETATTR);
271         if (req == NULL)
272                 RETURN(-ENOMEM);
273
274         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
275         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GETATTR);
276         if (rc) {
277                 ptlrpc_request_free(req);
278                 RETURN(rc);
279         }
280
281         osc_pack_req_body(req, oinfo);
282
283         ptlrpc_request_set_replen(req);
284
285         rc = ptlrpc_queue_wait(req);
286         if (rc)
287                 GOTO(out, rc);
288
289         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
290         if (body == NULL)
291                 GOTO(out, rc = -EPROTO);
292
293         CDEBUG(D_INODE, "mode: %o\n", body->oa.o_mode);
294         lustre_get_wire_obdo(oinfo->oi_oa, &body->oa);
295
296         /* This should really be sent by the OST */
297         oinfo->oi_oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
298         oinfo->oi_oa->o_valid |= OBD_MD_FLBLKSZ;
299
300         EXIT;
301  out:
302         ptlrpc_req_finished(req);
303         return rc;
304 }
305
306 static int osc_setattr(struct obd_export *exp, struct obd_info *oinfo,
307                        struct obd_trans_info *oti)
308 {
309         struct ptlrpc_request *req;
310         struct ost_body       *body;
311         int                    rc;
312         ENTRY;
313
314         LASSERT(oinfo->oi_oa->o_valid & OBD_MD_FLGROUP);
315
316         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
317         if (req == NULL)
318                 RETURN(-ENOMEM);
319
320         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
321         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
322         if (rc) {
323                 ptlrpc_request_free(req);
324                 RETURN(rc);
325         }
326
327         osc_pack_req_body(req, oinfo);
328
329         ptlrpc_request_set_replen(req);
330
331         rc = ptlrpc_queue_wait(req);
332         if (rc)
333                 GOTO(out, rc);
334
335         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
336         if (body == NULL)
337                 GOTO(out, rc = -EPROTO);
338
339         lustre_get_wire_obdo(oinfo->oi_oa, &body->oa);
340
341         EXIT;
342 out:
343         ptlrpc_req_finished(req);
344         RETURN(rc);
345 }
346
347 static int osc_setattr_interpret(const struct lu_env *env,
348                                  struct ptlrpc_request *req,
349                                  struct osc_setattr_args *sa, int rc)
350 {
351         struct ost_body *body;
352         ENTRY;
353
354         if (rc != 0)
355                 GOTO(out, rc);
356
357         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
358         if (body == NULL)
359                 GOTO(out, rc = -EPROTO);
360
361         lustre_get_wire_obdo(sa->sa_oa, &body->oa);
362 out:
363         rc = sa->sa_upcall(sa->sa_cookie, rc);
364         RETURN(rc);
365 }
366
367 int osc_setattr_async_base(struct obd_export *exp, struct obd_info *oinfo,
368                            struct obd_trans_info *oti,
369                            obd_enqueue_update_f upcall, void *cookie,
370                            struct ptlrpc_request_set *rqset)
371 {
372         struct ptlrpc_request   *req;
373         struct osc_setattr_args *sa;
374         int                      rc;
375         ENTRY;
376
377         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SETATTR);
378         if (req == NULL)
379                 RETURN(-ENOMEM);
380
381         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
382         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SETATTR);
383         if (rc) {
384                 ptlrpc_request_free(req);
385                 RETURN(rc);
386         }
387
388         if (oti && oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE)
389                 oinfo->oi_oa->o_lcookie = *oti->oti_logcookies;
390
391         osc_pack_req_body(req, oinfo);
392
393         ptlrpc_request_set_replen(req);
394
395         /* do mds to ost setattr asynchronously */
396         if (!rqset) {
397                 /* Do not wait for response. */
398                 ptlrpcd_add_req(req, PSCOPE_OTHER);
399         } else {
400                 req->rq_interpret_reply =
401                         (ptlrpc_interpterer_t)osc_setattr_interpret;
402
403                 CLASSERT (sizeof(*sa) <= sizeof(req->rq_async_args));
404                 sa = ptlrpc_req_async_args(req);
405                 sa->sa_oa = oinfo->oi_oa;
406                 sa->sa_upcall = upcall;
407                 sa->sa_cookie = cookie;
408
409                 if (rqset == PTLRPCD_SET)
410                         ptlrpcd_add_req(req, PSCOPE_OTHER);
411                 else
412                         ptlrpc_set_add_req(rqset, req);
413         }
414
415         RETURN(0);
416 }
417
418 static int osc_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
419                              struct obd_trans_info *oti,
420                              struct ptlrpc_request_set *rqset)
421 {
422         return osc_setattr_async_base(exp, oinfo, oti,
423                                       oinfo->oi_cb_up, oinfo, rqset);
424 }
425
426 int osc_real_create(struct obd_export *exp, struct obdo *oa,
427                     struct lov_stripe_md **ea, struct obd_trans_info *oti)
428 {
429         struct ptlrpc_request *req;
430         struct ost_body       *body;
431         struct lov_stripe_md  *lsm;
432         int                    rc;
433         ENTRY;
434
435         LASSERT(oa);
436         LASSERT(ea);
437
438         lsm = *ea;
439         if (!lsm) {
440                 rc = obd_alloc_memmd(exp, &lsm);
441                 if (rc < 0)
442                         RETURN(rc);
443         }
444
445         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_CREATE);
446         if (req == NULL)
447                 GOTO(out, rc = -ENOMEM);
448
449         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
450         if (rc) {
451                 ptlrpc_request_free(req);
452                 GOTO(out, rc);
453         }
454
455         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
456         LASSERT(body);
457         lustre_set_wire_obdo(&body->oa, oa);
458
459         ptlrpc_request_set_replen(req);
460
461         if ((oa->o_valid & OBD_MD_FLFLAGS) &&
462             oa->o_flags == OBD_FL_DELORPHAN) {
463                 DEBUG_REQ(D_HA, req,
464                           "delorphan from OST integration");
465                 /* Don't resend the delorphan req */
466                 req->rq_no_resend = req->rq_no_delay = 1;
467         }
468
469         rc = ptlrpc_queue_wait(req);
470         if (rc)
471                 GOTO(out_req, rc);
472
473         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
474         if (body == NULL)
475                 GOTO(out_req, rc = -EPROTO);
476
477         lustre_get_wire_obdo(oa, &body->oa);
478
479         /* This should really be sent by the OST */
480         oa->o_blksize = PTLRPC_MAX_BRW_SIZE;
481         oa->o_valid |= OBD_MD_FLBLKSZ;
482
483         /* XXX LOV STACKING: the lsm that is passed to us from LOV does not
484          * have valid lsm_oinfo data structs, so don't go touching that.
485          * This needs to be fixed in a big way.
486          */
487         lsm->lsm_object_id = oa->o_id;
488         lsm->lsm_object_seq = oa->o_seq;
489         *ea = lsm;
490
491         if (oti != NULL) {
492                 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
493
494                 if (oa->o_valid & OBD_MD_FLCOOKIE) {
495                         if (!oti->oti_logcookies)
496                                 oti_alloc_cookies(oti, 1);
497                         *oti->oti_logcookies = oa->o_lcookie;
498                 }
499         }
500
501         CDEBUG(D_HA, "transno: "LPD64"\n",
502                lustre_msg_get_transno(req->rq_repmsg));
503 out_req:
504         ptlrpc_req_finished(req);
505 out:
506         if (rc && !*ea)
507                 obd_free_memmd(exp, &lsm);
508         RETURN(rc);
509 }
510
511 int osc_punch_base(struct obd_export *exp, struct obd_info *oinfo,
512                    obd_enqueue_update_f upcall, void *cookie,
513                    struct ptlrpc_request_set *rqset)
514 {
515         struct ptlrpc_request   *req;
516         struct osc_setattr_args *sa;
517         struct ost_body         *body;
518         int                      rc;
519         ENTRY;
520
521         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_PUNCH);
522         if (req == NULL)
523                 RETURN(-ENOMEM);
524
525         osc_set_capa_size(req, &RMF_CAPA1, oinfo->oi_capa);
526         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
527         if (rc) {
528                 ptlrpc_request_free(req);
529                 RETURN(rc);
530         }
531         req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
532         ptlrpc_at_set_req_timeout(req);
533
534         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
535         LASSERT(body);
536         lustre_set_wire_obdo(&body->oa, oinfo->oi_oa);
537         osc_pack_capa(req, body, oinfo->oi_capa);
538
539         ptlrpc_request_set_replen(req);
540
541
542         req->rq_interpret_reply = (ptlrpc_interpterer_t)osc_setattr_interpret;
543         CLASSERT (sizeof(*sa) <= sizeof(req->rq_async_args));
544         sa = ptlrpc_req_async_args(req);
545         sa->sa_oa     = oinfo->oi_oa;
546         sa->sa_upcall = upcall;
547         sa->sa_cookie = cookie;
548         if (rqset == PTLRPCD_SET)
549                 ptlrpcd_add_req(req, PSCOPE_OTHER);
550         else
551                 ptlrpc_set_add_req(rqset, req);
552
553         RETURN(0);
554 }
555
556 static int osc_punch(struct obd_export *exp, struct obd_info *oinfo,
557                      struct obd_trans_info *oti,
558                      struct ptlrpc_request_set *rqset)
559 {
560         oinfo->oi_oa->o_size   = oinfo->oi_policy.l_extent.start;
561         oinfo->oi_oa->o_blocks = oinfo->oi_policy.l_extent.end;
562         oinfo->oi_oa->o_valid |= OBD_MD_FLSIZE | OBD_MD_FLBLOCKS;
563         return osc_punch_base(exp, oinfo,
564                               oinfo->oi_cb_up, oinfo, rqset);
565 }
566
567 static int osc_sync(struct obd_export *exp, struct obdo *oa,
568                     struct lov_stripe_md *md, obd_size start, obd_size end,
569                     void *capa)
570 {
571         struct ptlrpc_request *req;
572         struct ost_body       *body;
573         int                    rc;
574         ENTRY;
575
576         if (!oa) {
577                 CDEBUG(D_INFO, "oa NULL\n");
578                 RETURN(-EINVAL);
579         }
580
581         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_SYNC);
582         if (req == NULL)
583                 RETURN(-ENOMEM);
584
585         osc_set_capa_size(req, &RMF_CAPA1, capa);
586         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_SYNC);
587         if (rc) {
588                 ptlrpc_request_free(req);
589                 RETURN(rc);
590         }
591
592         /* overload the size and blocks fields in the oa with start/end */
593         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
594         LASSERT(body);
595         lustre_set_wire_obdo(&body->oa, oa);
596         body->oa.o_size = start;
597         body->oa.o_blocks = end;
598         body->oa.o_valid |= (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS);
599         osc_pack_capa(req, body, capa);
600
601         ptlrpc_request_set_replen(req);
602
603         rc = ptlrpc_queue_wait(req);
604         if (rc)
605                 GOTO(out, rc);
606
607         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
608         if (body == NULL)
609                 GOTO(out, rc = -EPROTO);
610
611         lustre_get_wire_obdo(oa, &body->oa);
612
613         EXIT;
614  out:
615         ptlrpc_req_finished(req);
616         return rc;
617 }
618
619 /* Find and cancel locally locks matched by @mode in the resource found by
620  * @objid. Found locks are added into @cancel list. Returns the amount of
621  * locks added to @cancels list. */
622 static int osc_resource_get_unused(struct obd_export *exp, struct obdo *oa,
623                                    cfs_list_t *cancels,
624                                    ldlm_mode_t mode, int lock_flags)
625 {
626         struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
627         struct ldlm_res_id res_id;
628         struct ldlm_resource *res;
629         int count;
630         ENTRY;
631
632         osc_build_res_name(oa->o_id, oa->o_seq, &res_id);
633         res = ldlm_resource_get(ns, NULL, &res_id, 0, 0);
634         if (res == NULL)
635                 RETURN(0);
636
637         LDLM_RESOURCE_ADDREF(res);
638         count = ldlm_cancel_resource_local(res, cancels, NULL, mode,
639                                            lock_flags, 0, NULL);
640         LDLM_RESOURCE_DELREF(res);
641         ldlm_resource_putref(res);
642         RETURN(count);
643 }
644
645 static int osc_destroy_interpret(const struct lu_env *env,
646                                  struct ptlrpc_request *req, void *data,
647                                  int rc)
648 {
649         struct client_obd *cli = &req->rq_import->imp_obd->u.cli;
650
651         cfs_atomic_dec(&cli->cl_destroy_in_flight);
652         cfs_waitq_signal(&cli->cl_destroy_waitq);
653         return 0;
654 }
655
656 static int osc_can_send_destroy(struct client_obd *cli)
657 {
658         if (cfs_atomic_inc_return(&cli->cl_destroy_in_flight) <=
659             cli->cl_max_rpcs_in_flight) {
660                 /* The destroy request can be sent */
661                 return 1;
662         }
663         if (cfs_atomic_dec_return(&cli->cl_destroy_in_flight) <
664             cli->cl_max_rpcs_in_flight) {
665                 /*
666                  * The counter has been modified between the two atomic
667                  * operations.
668                  */
669                 cfs_waitq_signal(&cli->cl_destroy_waitq);
670         }
671         return 0;
672 }
673
674 /* Destroy requests can be async always on the client, and we don't even really
675  * care about the return code since the client cannot do anything at all about
676  * a destroy failure.
677  * When the MDS is unlinking a filename, it saves the file objects into a
678  * recovery llog, and these object records are cancelled when the OST reports
679  * they were destroyed and sync'd to disk (i.e. transaction committed).
680  * If the client dies, or the OST is down when the object should be destroyed,
681  * the records are not cancelled, and when the OST reconnects to the MDS next,
682  * it will retrieve the llog unlink logs and then sends the log cancellation
683  * cookies to the MDS after committing destroy transactions. */
684 static int osc_destroy(struct obd_export *exp, struct obdo *oa,
685                        struct lov_stripe_md *ea, struct obd_trans_info *oti,
686                        struct obd_export *md_export, void *capa)
687 {
688         struct client_obd     *cli = &exp->exp_obd->u.cli;
689         struct ptlrpc_request *req;
690         struct ost_body       *body;
691         CFS_LIST_HEAD(cancels);
692         int rc, count;
693         ENTRY;
694
695         if (!oa) {
696                 CDEBUG(D_INFO, "oa NULL\n");
697                 RETURN(-EINVAL);
698         }
699
700         count = osc_resource_get_unused(exp, oa, &cancels, LCK_PW,
701                                         LDLM_FL_DISCARD_DATA);
702
703         req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_OST_DESTROY);
704         if (req == NULL) {
705                 ldlm_lock_list_put(&cancels, l_bl_ast, count);
706                 RETURN(-ENOMEM);
707         }
708
709         osc_set_capa_size(req, &RMF_CAPA1, (struct obd_capa *)capa);
710         rc = ldlm_prep_elc_req(exp, req, LUSTRE_OST_VERSION, OST_DESTROY,
711                                0, &cancels, count);
712         if (rc) {
713                 ptlrpc_request_free(req);
714                 RETURN(rc);
715         }
716
717         req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
718         ptlrpc_at_set_req_timeout(req);
719
720         if (oti != NULL && oa->o_valid & OBD_MD_FLCOOKIE)
721                 oa->o_lcookie = *oti->oti_logcookies;
722         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
723         LASSERT(body);
724         lustre_set_wire_obdo(&body->oa, oa);
725
726         osc_pack_capa(req, body, (struct obd_capa *)capa);
727         ptlrpc_request_set_replen(req);
728
729         /* don't throttle destroy RPCs for the MDT */
730         if (!(cli->cl_import->imp_connect_flags_orig & OBD_CONNECT_MDS)) {
731                 req->rq_interpret_reply = osc_destroy_interpret;
732                 if (!osc_can_send_destroy(cli)) {
733                         struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP,
734                                                           NULL);
735
736                         /*
737                          * Wait until the number of on-going destroy RPCs drops
738                          * under max_rpc_in_flight
739                          */
740                         l_wait_event_exclusive(cli->cl_destroy_waitq,
741                                                osc_can_send_destroy(cli), &lwi);
742                 }
743         }
744
745         /* Do not wait for response */
746         ptlrpcd_add_req(req, PSCOPE_OTHER);
747         RETURN(0);
748 }
749
750 static void osc_announce_cached(struct client_obd *cli, struct obdo *oa,
751                                 long writing_bytes)
752 {
753         obd_flag bits = OBD_MD_FLBLOCKS|OBD_MD_FLGRANT;
754
755         LASSERT(!(oa->o_valid & bits));
756
757         oa->o_valid |= bits;
758         client_obd_list_lock(&cli->cl_loi_list_lock);
759         oa->o_dirty = cli->cl_dirty;
760         if (cli->cl_dirty - cli->cl_dirty_transit > cli->cl_dirty_max) {
761                 CERROR("dirty %lu - %lu > dirty_max %lu\n",
762                        cli->cl_dirty, cli->cl_dirty_transit, cli->cl_dirty_max);
763                 oa->o_undirty = 0;
764         } else if (cfs_atomic_read(&obd_dirty_pages) -
765                    cfs_atomic_read(&obd_dirty_transit_pages) >
766                    obd_max_dirty_pages + 1){
767                 /* The cfs_atomic_read() allowing the cfs_atomic_inc() are
768                  * not covered by a lock thus they may safely race and trip
769                  * this CERROR() unless we add in a small fudge factor (+1). */
770                 CERROR("dirty %d - %d > system dirty_max %d\n",
771                        cfs_atomic_read(&obd_dirty_pages),
772                        cfs_atomic_read(&obd_dirty_transit_pages),
773                        obd_max_dirty_pages);
774                 oa->o_undirty = 0;
775         } else if (cli->cl_dirty_max - cli->cl_dirty > 0x7fffffff) {
776                 CERROR("dirty %lu - dirty_max %lu too big???\n",
777                        cli->cl_dirty, cli->cl_dirty_max);
778                 oa->o_undirty = 0;
779         } else {
780                 long max_in_flight = (cli->cl_max_pages_per_rpc << CFS_PAGE_SHIFT)*
781                                 (cli->cl_max_rpcs_in_flight + 1);
782                 oa->o_undirty = max(cli->cl_dirty_max, max_in_flight);
783         }
784         oa->o_grant = cli->cl_avail_grant;
785         oa->o_dropped = cli->cl_lost_grant;
786         cli->cl_lost_grant = 0;
787         client_obd_list_unlock(&cli->cl_loi_list_lock);
788         CDEBUG(D_CACHE,"dirty: "LPU64" undirty: %u dropped %u grant: "LPU64"\n",
789                oa->o_dirty, oa->o_undirty, oa->o_dropped, oa->o_grant);
790
791 }
792
793 static void osc_update_next_shrink(struct client_obd *cli)
794 {
795         cli->cl_next_shrink_grant =
796                 cfs_time_shift(cli->cl_grant_shrink_interval);
797         CDEBUG(D_CACHE, "next time %ld to shrink grant \n",
798                cli->cl_next_shrink_grant);
799 }
800
801 /* caller must hold loi_list_lock */
802 static void osc_consume_write_grant(struct client_obd *cli,
803                                     struct brw_page *pga)
804 {
805         LASSERT_SPIN_LOCKED(&cli->cl_loi_list_lock.lock);
806         LASSERT(!(pga->flag & OBD_BRW_FROM_GRANT));
807         cfs_atomic_inc(&obd_dirty_pages);
808         cli->cl_dirty += CFS_PAGE_SIZE;
809         cli->cl_avail_grant -= CFS_PAGE_SIZE;
810         pga->flag |= OBD_BRW_FROM_GRANT;
811         CDEBUG(D_CACHE, "using %lu grant credits for brw %p page %p\n",
812                CFS_PAGE_SIZE, pga, pga->pg);
813         LASSERT(cli->cl_avail_grant >= 0);
814         osc_update_next_shrink(cli);
815 }
816
817 /* the companion to osc_consume_write_grant, called when a brw has completed.
818  * must be called with the loi lock held. */
819 static void osc_release_write_grant(struct client_obd *cli,
820                                     struct brw_page *pga, int sent)
821 {
822         int blocksize = cli->cl_import->imp_obd->obd_osfs.os_bsize ? : 4096;
823         ENTRY;
824
825         LASSERT_SPIN_LOCKED(&cli->cl_loi_list_lock.lock);
826         if (!(pga->flag & OBD_BRW_FROM_GRANT)) {
827                 EXIT;
828                 return;
829         }
830
831         pga->flag &= ~OBD_BRW_FROM_GRANT;
832         cfs_atomic_dec(&obd_dirty_pages);
833         cli->cl_dirty -= CFS_PAGE_SIZE;
834         if (pga->flag & OBD_BRW_NOCACHE) {
835                 pga->flag &= ~OBD_BRW_NOCACHE;
836                 cfs_atomic_dec(&obd_dirty_transit_pages);
837                 cli->cl_dirty_transit -= CFS_PAGE_SIZE;
838         }
839         if (!sent) {
840                 cli->cl_lost_grant += CFS_PAGE_SIZE;
841                 CDEBUG(D_CACHE, "lost grant: %lu avail grant: %lu dirty: %lu\n",
842                        cli->cl_lost_grant, cli->cl_avail_grant, cli->cl_dirty);
843         } else if (CFS_PAGE_SIZE != blocksize && pga->count != CFS_PAGE_SIZE) {
844                 /* For short writes we shouldn't count parts of pages that
845                  * span a whole block on the OST side, or our accounting goes
846                  * wrong.  Should match the code in filter_grant_check. */
847                 int offset = pga->off & ~CFS_PAGE_MASK;
848                 int count = pga->count + (offset & (blocksize - 1));
849                 int end = (offset + pga->count) & (blocksize - 1);
850                 if (end)
851                         count += blocksize - end;
852
853                 cli->cl_lost_grant += CFS_PAGE_SIZE - count;
854                 CDEBUG(D_CACHE, "lost %lu grant: %lu avail: %lu dirty: %lu\n",
855                        CFS_PAGE_SIZE - count, cli->cl_lost_grant,
856                        cli->cl_avail_grant, cli->cl_dirty);
857         }
858
859         EXIT;
860 }
861
862 static unsigned long rpcs_in_flight(struct client_obd *cli)
863 {
864         return cli->cl_r_in_flight + cli->cl_w_in_flight;
865 }
866
867 /* caller must hold loi_list_lock */
868 void osc_wake_cache_waiters(struct client_obd *cli)
869 {
870         cfs_list_t *l, *tmp;
871         struct osc_cache_waiter *ocw;
872
873         ENTRY;
874         cfs_list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
875                 /* if we can't dirty more, we must wait until some is written */
876                 if ((cli->cl_dirty + CFS_PAGE_SIZE > cli->cl_dirty_max) ||
877                    (cfs_atomic_read(&obd_dirty_pages) + 1 >
878                     obd_max_dirty_pages)) {
879                         CDEBUG(D_CACHE, "no dirty room: dirty: %ld "
880                                "osc max %ld, sys max %d\n", cli->cl_dirty,
881                                cli->cl_dirty_max, obd_max_dirty_pages);
882                         return;
883                 }
884
885                 /* if still dirty cache but no grant wait for pending RPCs that
886                  * may yet return us some grant before doing sync writes */
887                 if (cli->cl_w_in_flight && cli->cl_avail_grant < CFS_PAGE_SIZE) {
888                         CDEBUG(D_CACHE, "%u BRW writes in flight, no grant\n",
889                                cli->cl_w_in_flight);
890                         return;
891                 }
892
893                 ocw = cfs_list_entry(l, struct osc_cache_waiter, ocw_entry);
894                 cfs_list_del_init(&ocw->ocw_entry);
895                 if (cli->cl_avail_grant < CFS_PAGE_SIZE) {
896                         /* no more RPCs in flight to return grant, do sync IO */
897                         ocw->ocw_rc = -EDQUOT;
898                         CDEBUG(D_INODE, "wake oap %p for sync\n", ocw->ocw_oap);
899                 } else {
900                         osc_consume_write_grant(cli,
901                                                 &ocw->ocw_oap->oap_brw_page);
902                 }
903
904                 cfs_waitq_signal(&ocw->ocw_waitq);
905         }
906
907         EXIT;
908 }
909
910 static void __osc_update_grant(struct client_obd *cli, obd_size grant)
911 {
912         client_obd_list_lock(&cli->cl_loi_list_lock);
913         cli->cl_avail_grant += grant;
914         client_obd_list_unlock(&cli->cl_loi_list_lock);
915 }
916
917 static void osc_update_grant(struct client_obd *cli, struct ost_body *body)
918 {
919         if (body->oa.o_valid & OBD_MD_FLGRANT) {
920                 CDEBUG(D_CACHE, "got "LPU64" extra grant\n", body->oa.o_grant);
921                 __osc_update_grant(cli, body->oa.o_grant);
922         }
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                 __osc_update_grant(cli, oa->o_grant);
939                 GOTO(out, rc);
940         }
941
942         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
943         LASSERT(body);
944         osc_update_grant(cli, body);
945 out:
946         OBDO_FREE(oa);
947         return rc;
948 }
949
950 static void osc_shrink_grant_local(struct client_obd *cli, struct obdo *oa)
951 {
952         client_obd_list_lock(&cli->cl_loi_list_lock);
953         oa->o_grant = cli->cl_avail_grant / 4;
954         cli->cl_avail_grant -= oa->o_grant;
955         client_obd_list_unlock(&cli->cl_loi_list_lock);
956         if (!(oa->o_valid & OBD_MD_FLFLAGS)) {
957                 oa->o_valid |= OBD_MD_FLFLAGS;
958                 oa->o_flags = 0;
959         }
960         oa->o_flags |= OBD_FL_SHRINK_GRANT;
961         osc_update_next_shrink(cli);
962 }
963
964 /* Shrink the current grant, either from some large amount to enough for a
965  * full set of in-flight RPCs, or if we have already shrunk to that limit
966  * then to enough for a single RPC.  This avoids keeping more grant than
967  * needed, and avoids shrinking the grant piecemeal. */
968 static int osc_shrink_grant(struct client_obd *cli)
969 {
970         long target = (cli->cl_max_rpcs_in_flight + 1) *
971                       cli->cl_max_pages_per_rpc;
972
973         client_obd_list_lock(&cli->cl_loi_list_lock);
974         if (cli->cl_avail_grant <= target)
975                 target = cli->cl_max_pages_per_rpc;
976         client_obd_list_unlock(&cli->cl_loi_list_lock);
977
978         return osc_shrink_grant_to_target(cli, target);
979 }
980
981 int osc_shrink_grant_to_target(struct client_obd *cli, long target)
982 {
983         int    rc = 0;
984         struct ost_body     *body;
985         ENTRY;
986
987         client_obd_list_lock(&cli->cl_loi_list_lock);
988         /* Don't shrink if we are already above or below the desired limit
989          * We don't want to shrink below a single RPC, as that will negatively
990          * impact block allocation and long-term performance. */
991         if (target < cli->cl_max_pages_per_rpc)
992                 target = cli->cl_max_pages_per_rpc;
993
994         if (target >= cli->cl_avail_grant) {
995                 client_obd_list_unlock(&cli->cl_loi_list_lock);
996                 RETURN(0);
997         }
998         client_obd_list_unlock(&cli->cl_loi_list_lock);
999
1000         OBD_ALLOC_PTR(body);
1001         if (!body)
1002                 RETURN(-ENOMEM);
1003
1004         osc_announce_cached(cli, &body->oa, 0);
1005
1006         client_obd_list_lock(&cli->cl_loi_list_lock);
1007         body->oa.o_grant = cli->cl_avail_grant - target;
1008         cli->cl_avail_grant = target;
1009         client_obd_list_unlock(&cli->cl_loi_list_lock);
1010         if (!(body->oa.o_valid & OBD_MD_FLFLAGS)) {
1011                 body->oa.o_valid |= OBD_MD_FLFLAGS;
1012                 body->oa.o_flags = 0;
1013         }
1014         body->oa.o_flags |= OBD_FL_SHRINK_GRANT;
1015         osc_update_next_shrink(cli);
1016
1017         rc = osc_set_info_async(cli->cl_import->imp_obd->obd_self_export,
1018                                 sizeof(KEY_GRANT_SHRINK), KEY_GRANT_SHRINK,
1019                                 sizeof(*body), body, NULL);
1020         if (rc != 0)
1021                 __osc_update_grant(cli, body->oa.o_grant);
1022         OBD_FREE_PTR(body);
1023         RETURN(rc);
1024 }
1025
1026 #define GRANT_SHRINK_LIMIT PTLRPC_MAX_BRW_SIZE
1027 static int osc_should_shrink_grant(struct client_obd *client)
1028 {
1029         cfs_time_t time = cfs_time_current();
1030         cfs_time_t next_shrink = client->cl_next_shrink_grant;
1031
1032         if ((client->cl_import->imp_connect_data.ocd_connect_flags &
1033              OBD_CONNECT_GRANT_SHRINK) == 0)
1034                 return 0;
1035
1036         if (cfs_time_aftereq(time, next_shrink - 5 * CFS_TICK)) {
1037                 if (client->cl_import->imp_state == LUSTRE_IMP_FULL &&
1038                     client->cl_avail_grant > GRANT_SHRINK_LIMIT)
1039                         return 1;
1040                 else
1041                         osc_update_next_shrink(client);
1042         }
1043         return 0;
1044 }
1045
1046 static int osc_grant_shrink_grant_cb(struct timeout_item *item, void *data)
1047 {
1048         struct client_obd *client;
1049
1050         cfs_list_for_each_entry(client, &item->ti_obd_list,
1051                                 cl_grant_shrink_list) {
1052                 if (osc_should_shrink_grant(client))
1053                         osc_shrink_grant(client);
1054         }
1055         return 0;
1056 }
1057
1058 static int osc_add_shrink_grant(struct client_obd *client)
1059 {
1060         int rc;
1061
1062         rc = ptlrpc_add_timeout_client(client->cl_grant_shrink_interval,
1063                                        TIMEOUT_GRANT,
1064                                        osc_grant_shrink_grant_cb, NULL,
1065                                        &client->cl_grant_shrink_list);
1066         if (rc) {
1067                 CERROR("add grant client %s error %d\n",
1068                         client->cl_import->imp_obd->obd_name, rc);
1069                 return rc;
1070         }
1071         CDEBUG(D_CACHE, "add grant client %s \n",
1072                client->cl_import->imp_obd->obd_name);
1073         osc_update_next_shrink(client);
1074         return 0;
1075 }
1076
1077 static int osc_del_shrink_grant(struct client_obd *client)
1078 {
1079         return ptlrpc_del_timeout_client(&client->cl_grant_shrink_list,
1080                                          TIMEOUT_GRANT);
1081 }
1082
1083 static void osc_init_grant(struct client_obd *cli, struct obd_connect_data *ocd)
1084 {
1085         /*
1086          * ocd_grant is the total grant amount we're expect to hold: if we've
1087          * been evicted, it's the new avail_grant amount, cl_dirty will drop
1088          * to 0 as inflight RPCs fail out; otherwise, it's avail_grant + dirty.
1089          *
1090          * race is tolerable here: if we're evicted, but imp_state already
1091          * left EVICTED state, then cl_dirty must be 0 already.
1092          */
1093         client_obd_list_lock(&cli->cl_loi_list_lock);
1094         if (cli->cl_import->imp_state == LUSTRE_IMP_EVICTED)
1095                 cli->cl_avail_grant = ocd->ocd_grant;
1096         else
1097                 cli->cl_avail_grant = ocd->ocd_grant - cli->cl_dirty;
1098
1099         if (cli->cl_avail_grant < 0) {
1100                 CWARN("%s: available grant < 0, the OSS is probably not running"
1101                       " with patch from bug20278 (%ld) \n",
1102                       cli->cl_import->imp_obd->obd_name, cli->cl_avail_grant);
1103                 /* workaround for 1.6 servers which do not have 
1104                  * the patch from bug20278 */
1105                 cli->cl_avail_grant = ocd->ocd_grant;
1106         }
1107
1108         client_obd_list_unlock(&cli->cl_loi_list_lock);
1109
1110         CDEBUG(D_CACHE, "%s, setting cl_avail_grant: %ld cl_lost_grant: %ld \n",
1111                cli->cl_import->imp_obd->obd_name,
1112                cli->cl_avail_grant, cli->cl_lost_grant);
1113
1114         if (ocd->ocd_connect_flags & OBD_CONNECT_GRANT_SHRINK &&
1115             cfs_list_empty(&cli->cl_grant_shrink_list))
1116                 osc_add_shrink_grant(cli);
1117 }
1118
1119 /* We assume that the reason this OSC got a short read is because it read
1120  * beyond the end of a stripe file; i.e. lustre is reading a sparse file
1121  * via the LOV, and it _knows_ it's reading inside the file, it's just that
1122  * this stripe never got written at or beyond this stripe offset yet. */
1123 static void handle_short_read(int nob_read, obd_count page_count,
1124                               struct brw_page **pga)
1125 {
1126         char *ptr;
1127         int i = 0;
1128
1129         /* skip bytes read OK */
1130         while (nob_read > 0) {
1131                 LASSERT (page_count > 0);
1132
1133                 if (pga[i]->count > nob_read) {
1134                         /* EOF inside this page */
1135                         ptr = cfs_kmap(pga[i]->pg) +
1136                                 (pga[i]->off & ~CFS_PAGE_MASK);
1137                         memset(ptr + nob_read, 0, pga[i]->count - nob_read);
1138                         cfs_kunmap(pga[i]->pg);
1139                         page_count--;
1140                         i++;
1141                         break;
1142                 }
1143
1144                 nob_read -= pga[i]->count;
1145                 page_count--;
1146                 i++;
1147         }
1148
1149         /* zero remaining pages */
1150         while (page_count-- > 0) {
1151                 ptr = cfs_kmap(pga[i]->pg) + (pga[i]->off & ~CFS_PAGE_MASK);
1152                 memset(ptr, 0, pga[i]->count);
1153                 cfs_kunmap(pga[i]->pg);
1154                 i++;
1155         }
1156 }
1157
1158 static int check_write_rcs(struct ptlrpc_request *req,
1159                            int requested_nob, int niocount,
1160                            obd_count page_count, struct brw_page **pga)
1161 {
1162         int     i;
1163         __u32   *remote_rcs;
1164
1165         remote_rcs = req_capsule_server_sized_get(&req->rq_pill, &RMF_RCS,
1166                                                   sizeof(*remote_rcs) *
1167                                                   niocount);
1168         if (remote_rcs == NULL) {
1169                 CDEBUG(D_INFO, "Missing/short RC vector on BRW_WRITE reply\n");
1170                 return(-EPROTO);
1171         }
1172
1173         /* return error if any niobuf was in error */
1174         for (i = 0; i < niocount; i++) {
1175                 if (remote_rcs[i] < 0)
1176                         return(remote_rcs[i]);
1177
1178                 if (remote_rcs[i] != 0) {
1179                         CDEBUG(D_INFO, "rc[%d] invalid (%d) req %p\n",
1180                                 i, remote_rcs[i], req);
1181                         return(-EPROTO);
1182                 }
1183         }
1184
1185         if (req->rq_bulk->bd_nob_transferred != requested_nob) {
1186                 CERROR("Unexpected # bytes transferred: %d (requested %d)\n",
1187                        req->rq_bulk->bd_nob_transferred, requested_nob);
1188                 return(-EPROTO);
1189         }
1190
1191         return (0);
1192 }
1193
1194 static inline int can_merge_pages(struct brw_page *p1, struct brw_page *p2)
1195 {
1196         if (p1->flag != p2->flag) {
1197                 unsigned mask = ~(OBD_BRW_FROM_GRANT|
1198                                   OBD_BRW_NOCACHE|OBD_BRW_SYNC|OBD_BRW_ASYNC);
1199
1200                 /* warn if we try to combine flags that we don't know to be
1201                  * safe to combine */
1202                 if ((p1->flag & mask) != (p2->flag & mask))
1203                         CERROR("is it ok to have flags 0x%x and 0x%x in the "
1204                                "same brw?\n", p1->flag, p2->flag);
1205                 return 0;
1206         }
1207
1208         return (p1->off + p1->count == p2->off);
1209 }
1210
1211 static obd_count osc_checksum_bulk(int nob, obd_count pg_count,
1212                                    struct brw_page **pga, int opc,
1213                                    cksum_type_t cksum_type)
1214 {
1215         __u32 cksum;
1216         int i = 0;
1217
1218         LASSERT (pg_count > 0);
1219         cksum = init_checksum(cksum_type);
1220         while (nob > 0 && pg_count > 0) {
1221                 unsigned char *ptr = cfs_kmap(pga[i]->pg);
1222                 int off = pga[i]->off & ~CFS_PAGE_MASK;
1223                 int count = pga[i]->count > nob ? nob : pga[i]->count;
1224
1225                 /* corrupt the data before we compute the checksum, to
1226                  * simulate an OST->client data error */
1227                 if (i == 0 && opc == OST_READ &&
1228                     OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_RECEIVE))
1229                         memcpy(ptr + off, "bad1", min(4, nob));
1230                 cksum = compute_checksum(cksum, ptr + off, count, cksum_type);
1231                 cfs_kunmap(pga[i]->pg);
1232                 LL_CDEBUG_PAGE(D_PAGE, pga[i]->pg, "off %d checksum %x\n",
1233                                off, cksum);
1234
1235                 nob -= pga[i]->count;
1236                 pg_count--;
1237                 i++;
1238         }
1239         /* For sending we only compute the wrong checksum instead
1240          * of corrupting the data so it is still correct on a redo */
1241         if (opc == OST_WRITE && OBD_FAIL_CHECK(OBD_FAIL_OSC_CHECKSUM_SEND))
1242                 cksum++;
1243
1244         return cksum;
1245 }
1246
1247 static int osc_brw_prep_request(int cmd, struct client_obd *cli,struct obdo *oa,
1248                                 struct lov_stripe_md *lsm, obd_count page_count,
1249                                 struct brw_page **pga,
1250                                 struct ptlrpc_request **reqp,
1251                                 struct obd_capa *ocapa, int reserve)
1252 {
1253         struct ptlrpc_request   *req;
1254         struct ptlrpc_bulk_desc *desc;
1255         struct ost_body         *body;
1256         struct obd_ioobj        *ioobj;
1257         struct niobuf_remote    *niobuf;
1258         int niocount, i, requested_nob, opc, rc;
1259         struct osc_brw_async_args *aa;
1260         struct req_capsule      *pill;
1261         struct brw_page *pg_prev;
1262
1263         ENTRY;
1264         if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ))
1265                 RETURN(-ENOMEM); /* Recoverable */
1266         if (OBD_FAIL_CHECK(OBD_FAIL_OSC_BRW_PREP_REQ2))
1267                 RETURN(-EINVAL); /* Fatal */
1268
1269         if ((cmd & OBD_BRW_WRITE) != 0) {
1270                 opc = OST_WRITE;
1271                 req = ptlrpc_request_alloc_pool(cli->cl_import,
1272                                                 cli->cl_import->imp_rq_pool,
1273                                                 &RQF_OST_BRW_WRITE);
1274         } else {
1275                 opc = OST_READ;
1276                 req = ptlrpc_request_alloc(cli->cl_import, &RQF_OST_BRW_READ);
1277         }
1278         if (req == NULL)
1279                 RETURN(-ENOMEM);
1280
1281         for (niocount = i = 1; i < page_count; i++) {
1282                 if (!can_merge_pages(pga[i - 1], pga[i]))
1283                         niocount++;
1284         }
1285
1286         pill = &req->rq_pill;
1287         req_capsule_set_size(pill, &RMF_OBD_IOOBJ, RCL_CLIENT,
1288                              sizeof(*ioobj));
1289         req_capsule_set_size(pill, &RMF_NIOBUF_REMOTE, RCL_CLIENT,
1290                              niocount * sizeof(*niobuf));
1291         osc_set_capa_size(req, &RMF_CAPA1, ocapa);
1292
1293         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, opc);
1294         if (rc) {
1295                 ptlrpc_request_free(req);
1296                 RETURN(rc);
1297         }
1298         req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1299         ptlrpc_at_set_req_timeout(req);
1300
1301         if (opc == OST_WRITE)
1302                 desc = ptlrpc_prep_bulk_imp(req, page_count,
1303                                             BULK_GET_SOURCE, OST_BULK_PORTAL);
1304         else
1305                 desc = ptlrpc_prep_bulk_imp(req, page_count,
1306                                             BULK_PUT_SINK, OST_BULK_PORTAL);
1307
1308         if (desc == NULL)
1309                 GOTO(out, rc = -ENOMEM);
1310         /* NB request now owns desc and will free it when it gets freed */
1311
1312         body = req_capsule_client_get(pill, &RMF_OST_BODY);
1313         ioobj = req_capsule_client_get(pill, &RMF_OBD_IOOBJ);
1314         niobuf = req_capsule_client_get(pill, &RMF_NIOBUF_REMOTE);
1315         LASSERT(body != NULL && ioobj != NULL && niobuf != NULL);
1316
1317         lustre_set_wire_obdo(&body->oa, oa);
1318
1319         obdo_to_ioobj(oa, ioobj);
1320         ioobj->ioo_bufcnt = niocount;
1321         osc_pack_capa(req, body, ocapa);
1322         LASSERT (page_count > 0);
1323         pg_prev = pga[0];
1324         for (requested_nob = i = 0; i < page_count; i++, niobuf++) {
1325                 struct brw_page *pg = pga[i];
1326
1327                 LASSERT(pg->count > 0);
1328                 LASSERTF((pg->off & ~CFS_PAGE_MASK) + pg->count <= CFS_PAGE_SIZE,
1329                          "i: %d pg: %p off: "LPU64", count: %u\n", i, pg,
1330                          pg->off, pg->count);
1331 #ifdef __linux__
1332                 LASSERTF(i == 0 || pg->off > pg_prev->off,
1333                          "i %d p_c %u pg %p [pri %lu ind %lu] off "LPU64
1334                          " prev_pg %p [pri %lu ind %lu] off "LPU64"\n",
1335                          i, page_count,
1336                          pg->pg, page_private(pg->pg), pg->pg->index, pg->off,
1337                          pg_prev->pg, page_private(pg_prev->pg),
1338                          pg_prev->pg->index, pg_prev->off);
1339 #else
1340                 LASSERTF(i == 0 || pg->off > pg_prev->off,
1341                          "i %d p_c %u\n", i, page_count);
1342 #endif
1343                 LASSERT((pga[0]->flag & OBD_BRW_SRVLOCK) ==
1344                         (pg->flag & OBD_BRW_SRVLOCK));
1345
1346                 ptlrpc_prep_bulk_page(desc, pg->pg, pg->off & ~CFS_PAGE_MASK,
1347                                       pg->count);
1348                 requested_nob += pg->count;
1349
1350                 if (i > 0 && can_merge_pages(pg_prev, pg)) {
1351                         niobuf--;
1352                         niobuf->len += pg->count;
1353                 } else {
1354                         niobuf->offset = pg->off;
1355                         niobuf->len    = pg->count;
1356                         niobuf->flags  = pg->flag;
1357                 }
1358                 pg_prev = pg;
1359         }
1360
1361         LASSERTF((void *)(niobuf - niocount) ==
1362                 req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE),
1363                 "want %p - real %p\n", req_capsule_client_get(&req->rq_pill,
1364                 &RMF_NIOBUF_REMOTE), (void *)(niobuf - niocount));
1365
1366         osc_announce_cached(cli, &body->oa, opc == OST_WRITE ? requested_nob:0);
1367         if (osc_should_shrink_grant(cli))
1368                 osc_shrink_grant_local(cli, &body->oa);
1369
1370         /* size[REQ_REC_OFF] still sizeof (*body) */
1371         if (opc == OST_WRITE) {
1372                 if (unlikely(cli->cl_checksum) &&
1373                     !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1374                         /* store cl_cksum_type in a local variable since
1375                          * it can be changed via lprocfs */
1376                         cksum_type_t cksum_type = cli->cl_cksum_type;
1377
1378                         if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0) {
1379                                 oa->o_flags &= OBD_FL_LOCAL_MASK;
1380                                 body->oa.o_flags = 0;
1381                         }
1382                         body->oa.o_flags |= cksum_type_pack(cksum_type);
1383                         body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1384                         body->oa.o_cksum = osc_checksum_bulk(requested_nob,
1385                                                              page_count, pga,
1386                                                              OST_WRITE,
1387                                                              cksum_type);
1388                         CDEBUG(D_PAGE, "checksum at write origin: %x\n",
1389                                body->oa.o_cksum);
1390                         /* save this in 'oa', too, for later checking */
1391                         oa->o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1392                         oa->o_flags |= cksum_type_pack(cksum_type);
1393                 } else {
1394                         /* clear out the checksum flag, in case this is a
1395                          * resend but cl_checksum is no longer set. b=11238 */
1396                         oa->o_valid &= ~OBD_MD_FLCKSUM;
1397                 }
1398                 oa->o_cksum = body->oa.o_cksum;
1399                 /* 1 RC per niobuf */
1400                 req_capsule_set_size(pill, &RMF_RCS, RCL_SERVER,
1401                                      sizeof(__u32) * niocount);
1402         } else {
1403                 if (unlikely(cli->cl_checksum) &&
1404                     !sptlrpc_flavor_has_bulk(&req->rq_flvr)) {
1405                         if ((body->oa.o_valid & OBD_MD_FLFLAGS) == 0)
1406                                 body->oa.o_flags = 0;
1407                         body->oa.o_flags |= cksum_type_pack(cli->cl_cksum_type);
1408                         body->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1409                 }
1410         }
1411         ptlrpc_request_set_replen(req);
1412
1413         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
1414         aa = ptlrpc_req_async_args(req);
1415         aa->aa_oa = oa;
1416         aa->aa_requested_nob = requested_nob;
1417         aa->aa_nio_count = niocount;
1418         aa->aa_page_count = page_count;
1419         aa->aa_resends = 0;
1420         aa->aa_ppga = pga;
1421         aa->aa_cli = cli;
1422         CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1423         if (ocapa && reserve)
1424                 aa->aa_ocapa = capa_get(ocapa);
1425
1426         *reqp = req;
1427         RETURN(0);
1428
1429  out:
1430         ptlrpc_req_finished(req);
1431         RETURN(rc);
1432 }
1433
1434 static int check_write_checksum(struct obdo *oa, const lnet_process_id_t *peer,
1435                                 __u32 client_cksum, __u32 server_cksum, int nob,
1436                                 obd_count page_count, struct brw_page **pga,
1437                                 cksum_type_t client_cksum_type)
1438 {
1439         __u32 new_cksum;
1440         char *msg;
1441         cksum_type_t cksum_type;
1442
1443         if (server_cksum == client_cksum) {
1444                 CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1445                 return 0;
1446         }
1447
1448         /* If this is mmaped file - it can be changed at any time */
1449         if (oa->o_valid & OBD_MD_FLFLAGS && oa->o_flags & OBD_FL_MMAP)
1450                 return 1;
1451
1452         if (oa->o_valid & OBD_MD_FLFLAGS)
1453                 cksum_type = cksum_type_unpack(oa->o_flags);
1454         else
1455                 cksum_type = OBD_CKSUM_CRC32;
1456
1457         new_cksum = osc_checksum_bulk(nob, page_count, pga, OST_WRITE,
1458                                       cksum_type);
1459
1460         if (cksum_type != client_cksum_type)
1461                 msg = "the server did not use the checksum type specified in "
1462                       "the original request - likely a protocol problem";
1463         else if (new_cksum == server_cksum)
1464                 msg = "changed on the client after we checksummed it - "
1465                       "likely false positive due to mmap IO (bug 11742)";
1466         else if (new_cksum == client_cksum)
1467                 msg = "changed in transit before arrival at OST";
1468         else
1469                 msg = "changed in transit AND doesn't match the original - "
1470                       "likely false positive due to mmap IO (bug 11742)";
1471
1472         LCONSOLE_ERROR_MSG(0x132, "BAD WRITE CHECKSUM: %s: from %s inode "DFID
1473                            " object "LPU64"/"LPU64" extent ["LPU64"-"LPU64"]\n",
1474                            msg, libcfs_nid2str(peer->nid),
1475                            oa->o_valid & OBD_MD_FLFID ? oa->o_parent_seq : (__u64)0,
1476                            oa->o_valid & OBD_MD_FLFID ? oa->o_parent_oid : 0,
1477                            oa->o_valid & OBD_MD_FLFID ? oa->o_parent_ver : 0,
1478                            oa->o_id,
1479                            oa->o_valid & OBD_MD_FLGROUP ? oa->o_seq : (__u64)0,
1480                            pga[0]->off,
1481                            pga[page_count-1]->off + pga[page_count-1]->count - 1);
1482         CERROR("original client csum %x (type %x), server csum %x (type %x), "
1483                "client csum now %x\n", client_cksum, client_cksum_type,
1484                server_cksum, cksum_type, new_cksum);
1485         return 1;
1486 }
1487
1488 /* Note rc enters this function as number of bytes transferred */
1489 static int osc_brw_fini_request(struct ptlrpc_request *req, int rc)
1490 {
1491         struct osc_brw_async_args *aa = (void *)&req->rq_async_args;
1492         const lnet_process_id_t *peer =
1493                         &req->rq_import->imp_connection->c_peer;
1494         struct client_obd *cli = aa->aa_cli;
1495         struct ost_body *body;
1496         __u32 client_cksum = 0;
1497         ENTRY;
1498
1499         if (rc < 0 && rc != -EDQUOT) {
1500                 DEBUG_REQ(D_INFO, req, "Failed request with rc = %d\n", rc);
1501                 RETURN(rc);
1502         }
1503
1504         LASSERTF(req->rq_repmsg != NULL, "rc = %d\n", rc);
1505         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
1506         if (body == NULL) {
1507                 DEBUG_REQ(D_INFO, req, "Can't unpack body\n");
1508                 RETURN(-EPROTO);
1509         }
1510
1511 #ifdef HAVE_QUOTA_SUPPORT
1512         /* set/clear over quota flag for a uid/gid */
1513         if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE &&
1514             body->oa.o_valid & (OBD_MD_FLUSRQUOTA | OBD_MD_FLGRPQUOTA)) {
1515                 unsigned int qid[MAXQUOTAS] = { body->oa.o_uid, body->oa.o_gid };
1516
1517                 CDEBUG(D_QUOTA, "setdq for [%u %u] with valid "LPX64", flags %x\n",
1518                        body->oa.o_uid, body->oa.o_gid, body->oa.o_valid,
1519                        body->oa.o_flags);
1520                 lquota_setdq(quota_interface, cli, qid, body->oa.o_valid,
1521                              body->oa.o_flags);
1522         }
1523 #endif
1524
1525         osc_update_grant(cli, body);
1526
1527         if (rc < 0)
1528                 RETURN(rc);
1529
1530         if (aa->aa_oa->o_valid & OBD_MD_FLCKSUM)
1531                 client_cksum = aa->aa_oa->o_cksum; /* save for later */
1532
1533         if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE) {
1534                 if (rc > 0) {
1535                         CERROR("Unexpected +ve rc %d\n", rc);
1536                         RETURN(-EPROTO);
1537                 }
1538                 LASSERT(req->rq_bulk->bd_nob == aa->aa_requested_nob);
1539
1540                 if (sptlrpc_cli_unwrap_bulk_write(req, req->rq_bulk))
1541                         RETURN(-EAGAIN);
1542
1543                 if ((aa->aa_oa->o_valid & OBD_MD_FLCKSUM) && client_cksum &&
1544                     check_write_checksum(&body->oa, peer, client_cksum,
1545                                          body->oa.o_cksum, aa->aa_requested_nob,
1546                                          aa->aa_page_count, aa->aa_ppga,
1547                                          cksum_type_unpack(aa->aa_oa->o_flags)))
1548                         RETURN(-EAGAIN);
1549
1550                 rc = check_write_rcs(req, aa->aa_requested_nob,aa->aa_nio_count,
1551                                      aa->aa_page_count, aa->aa_ppga);
1552                 GOTO(out, rc);
1553         }
1554
1555         /* The rest of this function executes only for OST_READs */
1556
1557         /* if unwrap_bulk failed, return -EAGAIN to retry */
1558         rc = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, rc);
1559         if (rc < 0)
1560                 GOTO(out, rc = -EAGAIN);
1561
1562         if (rc > aa->aa_requested_nob) {
1563                 CERROR("Unexpected rc %d (%d requested)\n", rc,
1564                        aa->aa_requested_nob);
1565                 RETURN(-EPROTO);
1566         }
1567
1568         if (rc != req->rq_bulk->bd_nob_transferred) {
1569                 CERROR ("Unexpected rc %d (%d transferred)\n",
1570                         rc, req->rq_bulk->bd_nob_transferred);
1571                 return (-EPROTO);
1572         }
1573
1574         if (rc < aa->aa_requested_nob)
1575                 handle_short_read(rc, aa->aa_page_count, aa->aa_ppga);
1576
1577         if (body->oa.o_valid & OBD_MD_FLCKSUM) {
1578                 static int cksum_counter;
1579                 __u32      server_cksum = body->oa.o_cksum;
1580                 char      *via;
1581                 char      *router;
1582                 cksum_type_t cksum_type;
1583
1584                 if (body->oa.o_valid & OBD_MD_FLFLAGS)
1585                         cksum_type = cksum_type_unpack(body->oa.o_flags);
1586                 else
1587                         cksum_type = OBD_CKSUM_CRC32;
1588                 client_cksum = osc_checksum_bulk(rc, aa->aa_page_count,
1589                                                  aa->aa_ppga, OST_READ,
1590                                                  cksum_type);
1591
1592                 if (peer->nid == req->rq_bulk->bd_sender) {
1593                         via = router = "";
1594                 } else {
1595                         via = " via ";
1596                         router = libcfs_nid2str(req->rq_bulk->bd_sender);
1597                 }
1598
1599                 if (server_cksum == ~0 && rc > 0) {
1600                         CERROR("Protocol error: server %s set the 'checksum' "
1601                                "bit, but didn't send a checksum.  Not fatal, "
1602                                "but please notify on http://bugzilla.lustre.org/\n",
1603                                libcfs_nid2str(peer->nid));
1604                 } else if (server_cksum != client_cksum) {
1605                         LCONSOLE_ERROR_MSG(0x133, "%s: BAD READ CHECKSUM: from "
1606                                            "%s%s%s inode "DFID" object "
1607                                            LPU64"/"LPU64" extent "
1608                                            "["LPU64"-"LPU64"]\n",
1609                                            req->rq_import->imp_obd->obd_name,
1610                                            libcfs_nid2str(peer->nid),
1611                                            via, router,
1612                                            body->oa.o_valid & OBD_MD_FLFID ?
1613                                                 body->oa.o_parent_seq : (__u64)0,
1614                                            body->oa.o_valid & OBD_MD_FLFID ?
1615                                                 body->oa.o_parent_oid : 0,
1616                                            body->oa.o_valid & OBD_MD_FLFID ?
1617                                                 body->oa.o_parent_ver : 0,
1618                                            body->oa.o_id,
1619                                            body->oa.o_valid & OBD_MD_FLGROUP ?
1620                                                 body->oa.o_seq : (__u64)0,
1621                                            aa->aa_ppga[0]->off,
1622                                            aa->aa_ppga[aa->aa_page_count-1]->off +
1623                                            aa->aa_ppga[aa->aa_page_count-1]->count -
1624                                                                         1);
1625                         CERROR("client %x, server %x, cksum_type %x\n",
1626                                client_cksum, server_cksum, cksum_type);
1627                         cksum_counter = 0;
1628                         aa->aa_oa->o_cksum = client_cksum;
1629                         rc = -EAGAIN;
1630                 } else {
1631                         cksum_counter++;
1632                         CDEBUG(D_PAGE, "checksum %x confirmed\n", client_cksum);
1633                         rc = 0;
1634                 }
1635         } else if (unlikely(client_cksum)) {
1636                 static int cksum_missed;
1637
1638                 cksum_missed++;
1639                 if ((cksum_missed & (-cksum_missed)) == cksum_missed)
1640                         CERROR("Checksum %u requested from %s but not sent\n",
1641                                cksum_missed, libcfs_nid2str(peer->nid));
1642         } else {
1643                 rc = 0;
1644         }
1645 out:
1646         if (rc >= 0)
1647                 lustre_get_wire_obdo(aa->aa_oa, &body->oa);
1648
1649         RETURN(rc);
1650 }
1651
1652 static int osc_brw_internal(int cmd, struct obd_export *exp, struct obdo *oa,
1653                             struct lov_stripe_md *lsm,
1654                             obd_count page_count, struct brw_page **pga,
1655                             struct obd_capa *ocapa)
1656 {
1657         struct ptlrpc_request *req;
1658         int                    rc;
1659         cfs_waitq_t            waitq;
1660         int                    resends = 0;
1661         struct l_wait_info     lwi;
1662
1663         ENTRY;
1664
1665         cfs_waitq_init(&waitq);
1666
1667 restart_bulk:
1668         rc = osc_brw_prep_request(cmd, &exp->exp_obd->u.cli, oa, lsm,
1669                                   page_count, pga, &req, ocapa, 0);
1670         if (rc != 0)
1671                 return (rc);
1672
1673         rc = ptlrpc_queue_wait(req);
1674
1675         if (rc == -ETIMEDOUT && req->rq_resend) {
1676                 DEBUG_REQ(D_HA, req,  "BULK TIMEOUT");
1677                 ptlrpc_req_finished(req);
1678                 goto restart_bulk;
1679         }
1680
1681         rc = osc_brw_fini_request(req, rc);
1682
1683         ptlrpc_req_finished(req);
1684         if (osc_recoverable_error(rc)) {
1685                 resends++;
1686                 if (!osc_should_resend(resends, &exp->exp_obd->u.cli)) {
1687                         CERROR("too many resend retries, returning error\n");
1688                         RETURN(-EIO);
1689                 }
1690
1691                 lwi = LWI_TIMEOUT_INTR(cfs_time_seconds(resends), NULL, NULL, NULL);
1692                 l_wait_event(waitq, 0, &lwi);
1693
1694                 goto restart_bulk;
1695         }
1696
1697         RETURN (rc);
1698 }
1699
1700 int osc_brw_redo_request(struct ptlrpc_request *request,
1701                          struct osc_brw_async_args *aa)
1702 {
1703         struct ptlrpc_request *new_req;
1704         struct ptlrpc_request_set *set = request->rq_set;
1705         struct osc_brw_async_args *new_aa;
1706         struct osc_async_page *oap;
1707         int rc = 0;
1708         ENTRY;
1709
1710         if (!osc_should_resend(aa->aa_resends, aa->aa_cli)) {
1711                 CERROR("too many resent retries, returning error\n");
1712                 RETURN(-EIO);
1713         }
1714
1715         DEBUG_REQ(D_ERROR, request, "redo for recoverable error");
1716
1717         rc = osc_brw_prep_request(lustre_msg_get_opc(request->rq_reqmsg) ==
1718                                         OST_WRITE ? OBD_BRW_WRITE :OBD_BRW_READ,
1719                                   aa->aa_cli, aa->aa_oa,
1720                                   NULL /* lsm unused by osc currently */,
1721                                   aa->aa_page_count, aa->aa_ppga,
1722                                   &new_req, aa->aa_ocapa, 0);
1723         if (rc)
1724                 RETURN(rc);
1725
1726         client_obd_list_lock(&aa->aa_cli->cl_loi_list_lock);
1727
1728         cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
1729                 if (oap->oap_request != NULL) {
1730                         LASSERTF(request == oap->oap_request,
1731                                  "request %p != oap_request %p\n",
1732                                  request, oap->oap_request);
1733                         if (oap->oap_interrupted) {
1734                                 client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1735                                 ptlrpc_req_finished(new_req);
1736                                 RETURN(-EINTR);
1737                         }
1738                 }
1739         }
1740         /* New request takes over pga and oaps from old request.
1741          * Note that copying a list_head doesn't work, need to move it... */
1742         aa->aa_resends++;
1743         new_req->rq_interpret_reply = request->rq_interpret_reply;
1744         new_req->rq_async_args = request->rq_async_args;
1745         new_req->rq_sent = cfs_time_current_sec() + aa->aa_resends;
1746
1747         new_aa = ptlrpc_req_async_args(new_req);
1748
1749         CFS_INIT_LIST_HEAD(&new_aa->aa_oaps);
1750         cfs_list_splice(&aa->aa_oaps, &new_aa->aa_oaps);
1751         CFS_INIT_LIST_HEAD(&aa->aa_oaps);
1752
1753         cfs_list_for_each_entry(oap, &new_aa->aa_oaps, oap_rpc_item) {
1754                 if (oap->oap_request) {
1755                         ptlrpc_req_finished(oap->oap_request);
1756                         oap->oap_request = ptlrpc_request_addref(new_req);
1757                 }
1758         }
1759
1760         new_aa->aa_ocapa = aa->aa_ocapa;
1761         aa->aa_ocapa = NULL;
1762
1763         /* use ptlrpc_set_add_req is safe because interpret functions work
1764          * in check_set context. only one way exist with access to request
1765          * from different thread got -EINTR - this way protected with
1766          * cl_loi_list_lock */
1767         ptlrpc_set_add_req(set, new_req);
1768
1769         client_obd_list_unlock(&aa->aa_cli->cl_loi_list_lock);
1770
1771         DEBUG_REQ(D_INFO, new_req, "new request");
1772         RETURN(0);
1773 }
1774
1775 /*
1776  * ugh, we want disk allocation on the target to happen in offset order.  we'll
1777  * follow sedgewicks advice and stick to the dead simple shellsort -- it'll do
1778  * fine for our small page arrays and doesn't require allocation.  its an
1779  * insertion sort that swaps elements that are strides apart, shrinking the
1780  * stride down until its '1' and the array is sorted.
1781  */
1782 static void sort_brw_pages(struct brw_page **array, int num)
1783 {
1784         int stride, i, j;
1785         struct brw_page *tmp;
1786
1787         if (num == 1)
1788                 return;
1789         for (stride = 1; stride < num ; stride = (stride * 3) + 1)
1790                 ;
1791
1792         do {
1793                 stride /= 3;
1794                 for (i = stride ; i < num ; i++) {
1795                         tmp = array[i];
1796                         j = i;
1797                         while (j >= stride && array[j - stride]->off > tmp->off) {
1798                                 array[j] = array[j - stride];
1799                                 j -= stride;
1800                         }
1801                         array[j] = tmp;
1802                 }
1803         } while (stride > 1);
1804 }
1805
1806 static obd_count max_unfragmented_pages(struct brw_page **pg, obd_count pages)
1807 {
1808         int count = 1;
1809         int offset;
1810         int i = 0;
1811
1812         LASSERT (pages > 0);
1813         offset = pg[i]->off & ~CFS_PAGE_MASK;
1814
1815         for (;;) {
1816                 pages--;
1817                 if (pages == 0)         /* that's all */
1818                         return count;
1819
1820                 if (offset + pg[i]->count < CFS_PAGE_SIZE)
1821                         return count;   /* doesn't end on page boundary */
1822
1823                 i++;
1824                 offset = pg[i]->off & ~CFS_PAGE_MASK;
1825                 if (offset != 0)        /* doesn't start on page boundary */
1826                         return count;
1827
1828                 count++;
1829         }
1830 }
1831
1832 static struct brw_page **osc_build_ppga(struct brw_page *pga, obd_count count)
1833 {
1834         struct brw_page **ppga;
1835         int i;
1836
1837         OBD_ALLOC(ppga, sizeof(*ppga) * count);
1838         if (ppga == NULL)
1839                 return NULL;
1840
1841         for (i = 0; i < count; i++)
1842                 ppga[i] = pga + i;
1843         return ppga;
1844 }
1845
1846 static void osc_release_ppga(struct brw_page **ppga, obd_count count)
1847 {
1848         LASSERT(ppga != NULL);
1849         OBD_FREE(ppga, sizeof(*ppga) * count);
1850 }
1851
1852 static int osc_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1853                    obd_count page_count, struct brw_page *pga,
1854                    struct obd_trans_info *oti)
1855 {
1856         struct obdo *saved_oa = NULL;
1857         struct brw_page **ppga, **orig;
1858         struct obd_import *imp = class_exp2cliimp(exp);
1859         struct client_obd *cli;
1860         int rc, page_count_orig;
1861         ENTRY;
1862
1863         LASSERT((imp != NULL) && (imp->imp_obd != NULL));
1864         cli = &imp->imp_obd->u.cli;
1865
1866         if (cmd & OBD_BRW_CHECK) {
1867                 /* The caller just wants to know if there's a chance that this
1868                  * I/O can succeed */
1869
1870                 if (imp->imp_invalid)
1871                         RETURN(-EIO);
1872                 RETURN(0);
1873         }
1874
1875         /* test_brw with a failed create can trip this, maybe others. */
1876         LASSERT(cli->cl_max_pages_per_rpc);
1877
1878         rc = 0;
1879
1880         orig = ppga = osc_build_ppga(pga, page_count);
1881         if (ppga == NULL)
1882                 RETURN(-ENOMEM);
1883         page_count_orig = page_count;
1884
1885         sort_brw_pages(ppga, page_count);
1886         while (page_count) {
1887                 obd_count pages_per_brw;
1888
1889                 if (page_count > cli->cl_max_pages_per_rpc)
1890                         pages_per_brw = cli->cl_max_pages_per_rpc;
1891                 else
1892                         pages_per_brw = page_count;
1893
1894                 pages_per_brw = max_unfragmented_pages(ppga, pages_per_brw);
1895
1896                 if (saved_oa != NULL) {
1897                         /* restore previously saved oa */
1898                         *oinfo->oi_oa = *saved_oa;
1899                 } else if (page_count > pages_per_brw) {
1900                         /* save a copy of oa (brw will clobber it) */
1901                         OBDO_ALLOC(saved_oa);
1902                         if (saved_oa == NULL)
1903                                 GOTO(out, rc = -ENOMEM);
1904                         *saved_oa = *oinfo->oi_oa;
1905                 }
1906
1907                 rc = osc_brw_internal(cmd, exp, oinfo->oi_oa, oinfo->oi_md,
1908                                       pages_per_brw, ppga, oinfo->oi_capa);
1909
1910                 if (rc != 0)
1911                         break;
1912
1913                 page_count -= pages_per_brw;
1914                 ppga += pages_per_brw;
1915         }
1916
1917 out:
1918         osc_release_ppga(orig, page_count_orig);
1919
1920         if (saved_oa != NULL)
1921                 OBDO_FREE(saved_oa);
1922
1923         RETURN(rc);
1924 }
1925
1926 /* The companion to osc_enter_cache(), called when @oap is no longer part of
1927  * the dirty accounting.  Writeback completes or truncate happens before
1928  * writing starts.  Must be called with the loi lock held. */
1929 static void osc_exit_cache(struct client_obd *cli, struct osc_async_page *oap,
1930                            int sent)
1931 {
1932         osc_release_write_grant(cli, &oap->oap_brw_page, sent);
1933 }
1934
1935
1936 /* This maintains the lists of pending pages to read/write for a given object
1937  * (lop).  This is used by osc_check_rpcs->osc_next_loi() and loi_list_maint()
1938  * to quickly find objects that are ready to send an RPC. */
1939 static int lop_makes_rpc(struct client_obd *cli, struct loi_oap_pages *lop,
1940                          int cmd)
1941 {
1942         int optimal;
1943         ENTRY;
1944
1945         if (lop->lop_num_pending == 0)
1946                 RETURN(0);
1947
1948         /* if we have an invalid import we want to drain the queued pages
1949          * by forcing them through rpcs that immediately fail and complete
1950          * the pages.  recovery relies on this to empty the queued pages
1951          * before canceling the locks and evicting down the llite pages */
1952         if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
1953                 RETURN(1);
1954
1955         /* stream rpcs in queue order as long as as there is an urgent page
1956          * queued.  this is our cheap solution for good batching in the case
1957          * where writepage marks some random page in the middle of the file
1958          * as urgent because of, say, memory pressure */
1959         if (!cfs_list_empty(&lop->lop_urgent)) {
1960                 CDEBUG(D_CACHE, "urgent request forcing RPC\n");
1961                 RETURN(1);
1962         }
1963         /* fire off rpcs when we have 'optimal' rpcs as tuned for the wire. */
1964         optimal = cli->cl_max_pages_per_rpc;
1965         if (cmd & OBD_BRW_WRITE) {
1966                 /* trigger a write rpc stream as long as there are dirtiers
1967                  * waiting for space.  as they're waiting, they're not going to
1968                  * create more pages to coalesce with what's waiting.. */
1969                 if (!cfs_list_empty(&cli->cl_cache_waiters)) {
1970                         CDEBUG(D_CACHE, "cache waiters forcing RPC\n");
1971                         RETURN(1);
1972                 }
1973                 /* +16 to avoid triggering rpcs that would want to include pages
1974                  * that are being queued but which can't be made ready until
1975                  * the queuer finishes with the page. this is a wart for
1976                  * llite::commit_write() */
1977                 optimal += 16;
1978         }
1979         if (lop->lop_num_pending >= optimal)
1980                 RETURN(1);
1981
1982         RETURN(0);
1983 }
1984
1985 static int lop_makes_hprpc(struct loi_oap_pages *lop)
1986 {
1987         struct osc_async_page *oap;
1988         ENTRY;
1989
1990         if (cfs_list_empty(&lop->lop_urgent))
1991                 RETURN(0);
1992
1993         oap = cfs_list_entry(lop->lop_urgent.next,
1994                          struct osc_async_page, oap_urgent_item);
1995
1996         if (oap->oap_async_flags & ASYNC_HP) {
1997                 CDEBUG(D_CACHE, "hp request forcing RPC\n");
1998                 RETURN(1);
1999         }
2000
2001         RETURN(0);
2002 }
2003
2004 static void on_list(cfs_list_t *item, cfs_list_t *list,
2005                     int should_be_on)
2006 {
2007         if (cfs_list_empty(item) && should_be_on)
2008                 cfs_list_add_tail(item, list);
2009         else if (!cfs_list_empty(item) && !should_be_on)
2010                 cfs_list_del_init(item);
2011 }
2012
2013 /* maintain the loi's cli list membership invariants so that osc_send_oap_rpc
2014  * can find pages to build into rpcs quickly */
2015 void loi_list_maint(struct client_obd *cli, struct lov_oinfo *loi)
2016 {
2017         if (lop_makes_hprpc(&loi->loi_write_lop) ||
2018             lop_makes_hprpc(&loi->loi_read_lop)) {
2019                 /* HP rpc */
2020                 on_list(&loi->loi_ready_item, &cli->cl_loi_ready_list, 0);
2021                 on_list(&loi->loi_hp_ready_item, &cli->cl_loi_hp_ready_list, 1);
2022         } else {
2023                 on_list(&loi->loi_hp_ready_item, &cli->cl_loi_hp_ready_list, 0);
2024                 on_list(&loi->loi_ready_item, &cli->cl_loi_ready_list,
2025                         lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)||
2026                         lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ));
2027         }
2028
2029         on_list(&loi->loi_write_item, &cli->cl_loi_write_list,
2030                 loi->loi_write_lop.lop_num_pending);
2031
2032         on_list(&loi->loi_read_item, &cli->cl_loi_read_list,
2033                 loi->loi_read_lop.lop_num_pending);
2034 }
2035
2036 static void lop_update_pending(struct client_obd *cli,
2037                                struct loi_oap_pages *lop, int cmd, int delta)
2038 {
2039         lop->lop_num_pending += delta;
2040         if (cmd & OBD_BRW_WRITE)
2041                 cli->cl_pending_w_pages += delta;
2042         else
2043                 cli->cl_pending_r_pages += delta;
2044 }
2045
2046 /**
2047  * this is called when a sync waiter receives an interruption.  Its job is to
2048  * get the caller woken as soon as possible.  If its page hasn't been put in an
2049  * rpc yet it can dequeue immediately.  Otherwise it has to mark the rpc as
2050  * desiring interruption which will forcefully complete the rpc once the rpc
2051  * has timed out.
2052  */
2053 int osc_oap_interrupted(const struct lu_env *env, struct osc_async_page *oap)
2054 {
2055         struct loi_oap_pages *lop;
2056         struct lov_oinfo *loi;
2057         int rc = -EBUSY;
2058         ENTRY;
2059
2060         LASSERT(!oap->oap_interrupted);
2061         oap->oap_interrupted = 1;
2062
2063         /* ok, it's been put in an rpc. only one oap gets a request reference */
2064         if (oap->oap_request != NULL) {
2065                 ptlrpc_mark_interrupted(oap->oap_request);
2066                 ptlrpcd_wake(oap->oap_request);
2067                 ptlrpc_req_finished(oap->oap_request);
2068                 oap->oap_request = NULL;
2069         }
2070
2071         /*
2072          * page completion may be called only if ->cpo_prep() method was
2073          * executed by osc_io_submit(), that also adds page the to pending list
2074          */
2075         if (!cfs_list_empty(&oap->oap_pending_item)) {
2076                 cfs_list_del_init(&oap->oap_pending_item);
2077                 cfs_list_del_init(&oap->oap_urgent_item);
2078
2079                 loi = oap->oap_loi;
2080                 lop = (oap->oap_cmd & OBD_BRW_WRITE) ?
2081                         &loi->loi_write_lop : &loi->loi_read_lop;
2082                 lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, -1);
2083                 loi_list_maint(oap->oap_cli, oap->oap_loi);
2084                 rc = oap->oap_caller_ops->ap_completion(env,
2085                                           oap->oap_caller_data,
2086                                           oap->oap_cmd, NULL, -EINTR);
2087         }
2088
2089         RETURN(rc);
2090 }
2091
2092 /* this is trying to propogate async writeback errors back up to the
2093  * application.  As an async write fails we record the error code for later if
2094  * the app does an fsync.  As long as errors persist we force future rpcs to be
2095  * sync so that the app can get a sync error and break the cycle of queueing
2096  * pages for which writeback will fail. */
2097 static void osc_process_ar(struct osc_async_rc *ar, __u64 xid,
2098                            int rc)
2099 {
2100         if (rc) {
2101                 if (!ar->ar_rc)
2102                         ar->ar_rc = rc;
2103
2104                 ar->ar_force_sync = 1;
2105                 ar->ar_min_xid = ptlrpc_sample_next_xid();
2106                 return;
2107
2108         }
2109
2110         if (ar->ar_force_sync && (xid >= ar->ar_min_xid))
2111                 ar->ar_force_sync = 0;
2112 }
2113
2114 void osc_oap_to_pending(struct osc_async_page *oap)
2115 {
2116         struct loi_oap_pages *lop;
2117
2118         if (oap->oap_cmd & OBD_BRW_WRITE)
2119                 lop = &oap->oap_loi->loi_write_lop;
2120         else
2121                 lop = &oap->oap_loi->loi_read_lop;
2122
2123         if (oap->oap_async_flags & ASYNC_HP)
2124                 cfs_list_add(&oap->oap_urgent_item, &lop->lop_urgent);
2125         else if (oap->oap_async_flags & ASYNC_URGENT)
2126                 cfs_list_add_tail(&oap->oap_urgent_item, &lop->lop_urgent);
2127         cfs_list_add_tail(&oap->oap_pending_item, &lop->lop_pending);
2128         lop_update_pending(oap->oap_cli, lop, oap->oap_cmd, 1);
2129 }
2130
2131 /* this must be called holding the loi list lock to give coverage to exit_cache,
2132  * async_flag maintenance, and oap_request */
2133 static void osc_ap_completion(const struct lu_env *env,
2134                               struct client_obd *cli, struct obdo *oa,
2135                               struct osc_async_page *oap, int sent, int rc)
2136 {
2137         __u64 xid = 0;
2138
2139         ENTRY;
2140         if (oap->oap_request != NULL) {
2141                 xid = ptlrpc_req_xid(oap->oap_request);
2142                 ptlrpc_req_finished(oap->oap_request);
2143                 oap->oap_request = NULL;
2144         }
2145
2146         cfs_spin_lock(&oap->oap_lock);
2147         oap->oap_async_flags = 0;
2148         cfs_spin_unlock(&oap->oap_lock);
2149         oap->oap_interrupted = 0;
2150
2151         if (oap->oap_cmd & OBD_BRW_WRITE) {
2152                 osc_process_ar(&cli->cl_ar, xid, rc);
2153                 osc_process_ar(&oap->oap_loi->loi_ar, xid, rc);
2154         }
2155
2156         if (rc == 0 && oa != NULL) {
2157                 if (oa->o_valid & OBD_MD_FLBLOCKS)
2158                         oap->oap_loi->loi_lvb.lvb_blocks = oa->o_blocks;
2159                 if (oa->o_valid & OBD_MD_FLMTIME)
2160                         oap->oap_loi->loi_lvb.lvb_mtime = oa->o_mtime;
2161                 if (oa->o_valid & OBD_MD_FLATIME)
2162                         oap->oap_loi->loi_lvb.lvb_atime = oa->o_atime;
2163                 if (oa->o_valid & OBD_MD_FLCTIME)
2164                         oap->oap_loi->loi_lvb.lvb_ctime = oa->o_ctime;
2165         }
2166
2167         rc = oap->oap_caller_ops->ap_completion(env, oap->oap_caller_data,
2168                                                 oap->oap_cmd, oa, rc);
2169
2170         /* ll_ap_completion (from llite) drops PG_locked. so, a new
2171          * I/O on the page could start, but OSC calls it under lock
2172          * and thus we can add oap back to pending safely */
2173         if (rc)
2174                 /* upper layer wants to leave the page on pending queue */
2175                 osc_oap_to_pending(oap);
2176         else
2177                 osc_exit_cache(cli, oap, sent);
2178         EXIT;
2179 }
2180
2181 static int brw_interpret(const struct lu_env *env,
2182                          struct ptlrpc_request *req, void *data, int rc)
2183 {
2184         struct osc_brw_async_args *aa = data;
2185         struct client_obd *cli;
2186         int async;
2187         ENTRY;
2188
2189         rc = osc_brw_fini_request(req, rc);
2190         CDEBUG(D_INODE, "request %p aa %p rc %d\n", req, aa, rc);
2191         if (osc_recoverable_error(rc)) {
2192                 /* Only retry once for mmaped files since the mmaped page
2193                  * might be modified at anytime. We have to retry at least
2194                  * once in case there WAS really a corruption of the page
2195                  * on the network, that was not caused by mmap() modifying
2196                  * the page. Bug11742 */
2197                 if ((rc == -EAGAIN) && (aa->aa_resends > 0) &&
2198                     aa->aa_oa->o_valid & OBD_MD_FLFLAGS &&
2199                     aa->aa_oa->o_flags & OBD_FL_MMAP) {
2200                         rc = 0;
2201                 } else {
2202                         rc = osc_brw_redo_request(req, aa);
2203                         if (rc == 0)
2204                                 RETURN(0);
2205                 }
2206         }
2207
2208         if (aa->aa_ocapa) {
2209                 capa_put(aa->aa_ocapa);
2210                 aa->aa_ocapa = NULL;
2211         }
2212
2213         cli = aa->aa_cli;
2214
2215         client_obd_list_lock(&cli->cl_loi_list_lock);
2216
2217         /* We need to decrement before osc_ap_completion->osc_wake_cache_waiters
2218          * is called so we know whether to go to sync BRWs or wait for more
2219          * RPCs to complete */
2220         if (lustre_msg_get_opc(req->rq_reqmsg) == OST_WRITE)
2221                 cli->cl_w_in_flight--;
2222         else
2223                 cli->cl_r_in_flight--;
2224
2225         async = cfs_list_empty(&aa->aa_oaps);
2226         if (!async) { /* from osc_send_oap_rpc() */
2227                 struct osc_async_page *oap, *tmp;
2228                 /* the caller may re-use the oap after the completion call so
2229                  * we need to clean it up a little */
2230                 cfs_list_for_each_entry_safe(oap, tmp, &aa->aa_oaps,
2231                                              oap_rpc_item) {
2232                         cfs_list_del_init(&oap->oap_rpc_item);
2233                         osc_ap_completion(env, cli, aa->aa_oa, oap, 1, rc);
2234                 }
2235                 OBDO_FREE(aa->aa_oa);
2236         } else { /* from async_internal() */
2237                 obd_count i;
2238                 for (i = 0; i < aa->aa_page_count; i++)
2239                         osc_release_write_grant(aa->aa_cli, aa->aa_ppga[i], 1);
2240         }
2241         osc_wake_cache_waiters(cli);
2242         osc_check_rpcs(env, cli);
2243         client_obd_list_unlock(&cli->cl_loi_list_lock);
2244         if (!async)
2245                 cl_req_completion(env, aa->aa_clerq, rc);
2246         osc_release_ppga(aa->aa_ppga, aa->aa_page_count);
2247
2248         RETURN(rc);
2249 }
2250
2251 static struct ptlrpc_request *osc_build_req(const struct lu_env *env,
2252                                             struct client_obd *cli,
2253                                             cfs_list_t *rpc_list,
2254                                             int page_count, int cmd)
2255 {
2256         struct ptlrpc_request *req;
2257         struct brw_page **pga = NULL;
2258         struct osc_brw_async_args *aa;
2259         struct obdo *oa = NULL;
2260         const struct obd_async_page_ops *ops = NULL;
2261         void *caller_data = NULL;
2262         struct osc_async_page *oap;
2263         struct osc_async_page *tmp;
2264         struct ost_body *body;
2265         struct cl_req *clerq = NULL;
2266         enum cl_req_type crt = (cmd & OBD_BRW_WRITE) ? CRT_WRITE : CRT_READ;
2267         struct ldlm_lock *lock = NULL;
2268         struct cl_req_attr crattr;
2269         int i, rc, mpflag = 0;
2270
2271         ENTRY;
2272         LASSERT(!cfs_list_empty(rpc_list));
2273
2274         if (cmd & OBD_BRW_MEMALLOC)
2275                 mpflag = cfs_memory_pressure_get_and_set();
2276
2277         memset(&crattr, 0, sizeof crattr);
2278         OBD_ALLOC(pga, sizeof(*pga) * page_count);
2279         if (pga == NULL)
2280                 GOTO(out, req = ERR_PTR(-ENOMEM));
2281
2282         OBDO_ALLOC(oa);
2283         if (oa == NULL)
2284                 GOTO(out, req = ERR_PTR(-ENOMEM));
2285
2286         i = 0;
2287         cfs_list_for_each_entry(oap, rpc_list, oap_rpc_item) {
2288                 struct cl_page *page = osc_oap2cl_page(oap);
2289                 if (ops == NULL) {
2290                         ops = oap->oap_caller_ops;
2291                         caller_data = oap->oap_caller_data;
2292
2293                         clerq = cl_req_alloc(env, page, crt,
2294                                              1 /* only 1-object rpcs for
2295                                                 * now */);
2296                         if (IS_ERR(clerq))
2297                                 GOTO(out, req = (void *)clerq);
2298                         lock = oap->oap_ldlm_lock;
2299                 }
2300                 pga[i] = &oap->oap_brw_page;
2301                 pga[i]->off = oap->oap_obj_off + oap->oap_page_off;
2302                 CDEBUG(0, "put page %p index %lu oap %p flg %x to pga\n",
2303                        pga[i]->pg, cfs_page_index(oap->oap_page), oap, pga[i]->flag);
2304                 i++;
2305                 cl_req_page_add(env, clerq, page);
2306         }
2307
2308         /* always get the data for the obdo for the rpc */
2309         LASSERT(ops != NULL);
2310         crattr.cra_oa = oa;
2311         crattr.cra_capa = NULL;
2312         cl_req_attr_set(env, clerq, &crattr, ~0ULL);
2313         if (lock) {
2314                 oa->o_handle = lock->l_remote_handle;
2315                 oa->o_valid |= OBD_MD_FLHANDLE;
2316         }
2317
2318         rc = cl_req_prep(env, clerq);
2319         if (rc != 0) {
2320                 CERROR("cl_req_prep failed: %d\n", rc);
2321                 GOTO(out, req = ERR_PTR(rc));
2322         }
2323
2324         sort_brw_pages(pga, page_count);
2325         rc = osc_brw_prep_request(cmd, cli, oa, NULL, page_count,
2326                                   pga, &req, crattr.cra_capa, 1);
2327         if (rc != 0) {
2328                 CERROR("prep_req failed: %d\n", rc);
2329                 GOTO(out, req = ERR_PTR(rc));
2330         }
2331
2332         if (cmd & OBD_BRW_MEMALLOC)
2333                 req->rq_memalloc = 1;
2334
2335         /* Need to update the timestamps after the request is built in case
2336          * we race with setattr (locally or in queue at OST).  If OST gets
2337          * later setattr before earlier BRW (as determined by the request xid),
2338          * the OST will not use BRW timestamps.  Sadly, there is no obvious
2339          * way to do this in a single call.  bug 10150 */
2340         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
2341         cl_req_attr_set(env, clerq, &crattr,
2342                         OBD_MD_FLMTIME|OBD_MD_FLCTIME|OBD_MD_FLATIME);
2343
2344         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
2345         aa = ptlrpc_req_async_args(req);
2346         CFS_INIT_LIST_HEAD(&aa->aa_oaps);
2347         cfs_list_splice(rpc_list, &aa->aa_oaps);
2348         CFS_INIT_LIST_HEAD(rpc_list);
2349         aa->aa_clerq = clerq;
2350 out:
2351         if (cmd & OBD_BRW_MEMALLOC)
2352                 cfs_memory_pressure_restore(mpflag);
2353
2354         capa_put(crattr.cra_capa);
2355         if (IS_ERR(req)) {
2356                 if (oa)
2357                         OBDO_FREE(oa);
2358                 if (pga)
2359                         OBD_FREE(pga, sizeof(*pga) * page_count);
2360                 /* this should happen rarely and is pretty bad, it makes the
2361                  * pending list not follow the dirty order */
2362                 client_obd_list_lock(&cli->cl_loi_list_lock);
2363                 cfs_list_for_each_entry_safe(oap, tmp, rpc_list, oap_rpc_item) {
2364                         cfs_list_del_init(&oap->oap_rpc_item);
2365
2366                         /* queued sync pages can be torn down while the pages
2367                          * were between the pending list and the rpc */
2368                         if (oap->oap_interrupted) {
2369                                 CDEBUG(D_INODE, "oap %p interrupted\n", oap);
2370                                 osc_ap_completion(env, cli, NULL, oap, 0,
2371                                                   oap->oap_count);
2372                                 continue;
2373                         }
2374                         osc_ap_completion(env, cli, NULL, oap, 0, PTR_ERR(req));
2375                 }
2376                 if (clerq && !IS_ERR(clerq))
2377                         cl_req_completion(env, clerq, PTR_ERR(req));
2378         }
2379         RETURN(req);
2380 }
2381
2382 /**
2383  * prepare pages for ASYNC io and put pages in send queue.
2384  *
2385  * \param cmd OBD_BRW_* macroses
2386  * \param lop pending pages
2387  *
2388  * \return zero if no page added to send queue.
2389  * \return 1 if pages successfully added to send queue.
2390  * \return negative on errors.
2391  */
2392 static int
2393 osc_send_oap_rpc(const struct lu_env *env, struct client_obd *cli,
2394                  struct lov_oinfo *loi,
2395                  int cmd, struct loi_oap_pages *lop)
2396 {
2397         struct ptlrpc_request *req;
2398         obd_count page_count = 0;
2399         struct osc_async_page *oap = NULL, *tmp;
2400         struct osc_brw_async_args *aa;
2401         const struct obd_async_page_ops *ops;
2402         CFS_LIST_HEAD(rpc_list);
2403         CFS_LIST_HEAD(tmp_list);
2404         unsigned int ending_offset;
2405         unsigned  starting_offset = 0;
2406         int srvlock = 0, mem_tight = 0;
2407         struct cl_object *clob = NULL;
2408         ENTRY;
2409
2410         /* ASYNC_HP pages first. At present, when the lock the pages is
2411          * to be canceled, the pages covered by the lock will be sent out
2412          * with ASYNC_HP. We have to send out them as soon as possible. */
2413         cfs_list_for_each_entry_safe(oap, tmp, &lop->lop_urgent, oap_urgent_item) {
2414                 if (oap->oap_async_flags & ASYNC_HP) 
2415                         cfs_list_move(&oap->oap_pending_item, &tmp_list);
2416                 else
2417                         cfs_list_move_tail(&oap->oap_pending_item, &tmp_list);
2418                 if (++page_count >= cli->cl_max_pages_per_rpc)
2419                         break;
2420         }
2421
2422         cfs_list_splice(&tmp_list, &lop->lop_pending);
2423         page_count = 0;
2424
2425         /* first we find the pages we're allowed to work with */
2426         cfs_list_for_each_entry_safe(oap, tmp, &lop->lop_pending,
2427                                      oap_pending_item) {
2428                 ops = oap->oap_caller_ops;
2429
2430                 LASSERTF(oap->oap_magic == OAP_MAGIC, "Bad oap magic: oap %p, "
2431                          "magic 0x%x\n", oap, oap->oap_magic);
2432
2433                 if (clob == NULL) {
2434                         /* pin object in memory, so that completion call-backs
2435                          * can be safely called under client_obd_list lock. */
2436                         clob = osc_oap2cl_page(oap)->cp_obj;
2437                         cl_object_get(clob);
2438                 }
2439
2440                 if (page_count != 0 &&
2441                     srvlock != !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK)) {
2442                         CDEBUG(D_PAGE, "SRVLOCK flag mismatch,"
2443                                " oap %p, page %p, srvlock %u\n",
2444                                oap, oap->oap_brw_page.pg, (unsigned)!srvlock);
2445                         break;
2446                 }
2447
2448                 /* If there is a gap at the start of this page, it can't merge
2449                  * with any previous page, so we'll hand the network a
2450                  * "fragmented" page array that it can't transfer in 1 RDMA */
2451                 if (page_count != 0 && oap->oap_page_off != 0)
2452                         break;
2453
2454                 /* in llite being 'ready' equates to the page being locked
2455                  * until completion unlocks it.  commit_write submits a page
2456                  * as not ready because its unlock will happen unconditionally
2457                  * as the call returns.  if we race with commit_write giving
2458                  * us that page we don't want to create a hole in the page
2459                  * stream, so we stop and leave the rpc to be fired by
2460                  * another dirtier or kupdated interval (the not ready page
2461                  * will still be on the dirty list).  we could call in
2462                  * at the end of ll_file_write to process the queue again. */
2463                 if (!(oap->oap_async_flags & ASYNC_READY)) {
2464                         int rc = ops->ap_make_ready(env, oap->oap_caller_data,
2465                                                     cmd);
2466                         if (rc < 0)
2467                                 CDEBUG(D_INODE, "oap %p page %p returned %d "
2468                                                 "instead of ready\n", oap,
2469                                                 oap->oap_page, rc);
2470                         switch (rc) {
2471                         case -EAGAIN:
2472                                 /* llite is telling us that the page is still
2473                                  * in commit_write and that we should try
2474                                  * and put it in an rpc again later.  we
2475                                  * break out of the loop so we don't create
2476                                  * a hole in the sequence of pages in the rpc
2477                                  * stream.*/
2478                                 oap = NULL;
2479                                 break;
2480                         case -EINTR:
2481                                 /* the io isn't needed.. tell the checks
2482                                  * below to complete the rpc with EINTR */
2483                                 cfs_spin_lock(&oap->oap_lock);
2484                                 oap->oap_async_flags |= ASYNC_COUNT_STABLE;
2485                                 cfs_spin_unlock(&oap->oap_lock);
2486                                 oap->oap_count = -EINTR;
2487                                 break;
2488                         case 0:
2489                                 cfs_spin_lock(&oap->oap_lock);
2490                                 oap->oap_async_flags |= ASYNC_READY;
2491                                 cfs_spin_unlock(&oap->oap_lock);
2492                                 break;
2493                         default:
2494                                 LASSERTF(0, "oap %p page %p returned %d "
2495                                             "from make_ready\n", oap,
2496                                             oap->oap_page, rc);
2497                                 break;
2498                         }
2499                 }
2500                 if (oap == NULL)
2501                         break;
2502                 /*
2503                  * Page submitted for IO has to be locked. Either by
2504                  * ->ap_make_ready() or by higher layers.
2505                  */
2506 #if defined(__KERNEL__) && defined(__linux__)
2507                 {
2508                         struct cl_page *page;
2509
2510                         page = osc_oap2cl_page(oap);
2511
2512                         if (page->cp_type == CPT_CACHEABLE &&
2513                             !(PageLocked(oap->oap_page) &&
2514                               (CheckWriteback(oap->oap_page, cmd)))) {
2515                                 CDEBUG(D_PAGE, "page %p lost wb %lx/%x\n",
2516                                        oap->oap_page,
2517                                        (long)oap->oap_page->flags,
2518                                        oap->oap_async_flags);
2519                                 LBUG();
2520                         }
2521                 }
2522 #endif
2523
2524                 /* take the page out of our book-keeping */
2525                 cfs_list_del_init(&oap->oap_pending_item);
2526                 lop_update_pending(cli, lop, cmd, -1);
2527                 cfs_list_del_init(&oap->oap_urgent_item);
2528
2529                 if (page_count == 0)
2530                         starting_offset = (oap->oap_obj_off+oap->oap_page_off) &
2531                                           (PTLRPC_MAX_BRW_SIZE - 1);
2532
2533                 /* ask the caller for the size of the io as the rpc leaves. */
2534                 if (!(oap->oap_async_flags & ASYNC_COUNT_STABLE)) {
2535                         oap->oap_count =
2536                                 ops->ap_refresh_count(env, oap->oap_caller_data,
2537                                                       cmd);
2538                         LASSERT(oap->oap_page_off + oap->oap_count <= CFS_PAGE_SIZE);
2539                 }
2540                 if (oap->oap_count <= 0) {
2541                         CDEBUG(D_CACHE, "oap %p count %d, completing\n", oap,
2542                                oap->oap_count);
2543                         osc_ap_completion(env, cli, NULL,
2544                                           oap, 0, oap->oap_count);
2545                         continue;
2546                 }
2547
2548                 /* now put the page back in our accounting */
2549                 cfs_list_add_tail(&oap->oap_rpc_item, &rpc_list);
2550                 if (oap->oap_brw_flags & OBD_BRW_MEMALLOC)
2551                         mem_tight = 1;
2552                 if (page_count == 0)
2553                         srvlock = !!(oap->oap_brw_flags & OBD_BRW_SRVLOCK);
2554                 if (++page_count >= cli->cl_max_pages_per_rpc)
2555                         break;
2556
2557                 /* End on a PTLRPC_MAX_BRW_SIZE boundary.  We want full-sized
2558                  * RPCs aligned on PTLRPC_MAX_BRW_SIZE boundaries to help reads
2559                  * have the same alignment as the initial writes that allocated
2560                  * extents on the server. */
2561                 ending_offset = (oap->oap_obj_off + oap->oap_page_off +
2562                                  oap->oap_count) & (PTLRPC_MAX_BRW_SIZE - 1);
2563                 if (ending_offset == 0)
2564                         break;
2565
2566                 /* If there is a gap at the end of this page, it can't merge
2567                  * with any subsequent pages, so we'll hand the network a
2568                  * "fragmented" page array that it can't transfer in 1 RDMA */
2569                 if (oap->oap_page_off + oap->oap_count < CFS_PAGE_SIZE)
2570                         break;
2571         }
2572
2573         osc_wake_cache_waiters(cli);
2574
2575         loi_list_maint(cli, loi);
2576
2577         client_obd_list_unlock(&cli->cl_loi_list_lock);
2578
2579         if (clob != NULL)
2580                 cl_object_put(env, clob);
2581
2582         if (page_count == 0) {
2583                 client_obd_list_lock(&cli->cl_loi_list_lock);
2584                 RETURN(0);
2585         }
2586
2587         req = osc_build_req(env, cli, &rpc_list, page_count,
2588                             mem_tight ? (cmd | OBD_BRW_MEMALLOC) : cmd);
2589         if (IS_ERR(req)) {
2590                 LASSERT(cfs_list_empty(&rpc_list));
2591                 loi_list_maint(cli, loi);
2592                 RETURN(PTR_ERR(req));
2593         }
2594
2595         aa = ptlrpc_req_async_args(req);
2596
2597         if (cmd == OBD_BRW_READ) {
2598                 lprocfs_oh_tally_log2(&cli->cl_read_page_hist, page_count);
2599                 lprocfs_oh_tally(&cli->cl_read_rpc_hist, cli->cl_r_in_flight);
2600                 lprocfs_oh_tally_log2(&cli->cl_read_offset_hist,
2601                                       (starting_offset >> CFS_PAGE_SHIFT) + 1);
2602         } else {
2603                 lprocfs_oh_tally_log2(&cli->cl_write_page_hist, page_count);
2604                 lprocfs_oh_tally(&cli->cl_write_rpc_hist,
2605                                  cli->cl_w_in_flight);
2606                 lprocfs_oh_tally_log2(&cli->cl_write_offset_hist,
2607                                       (starting_offset >> CFS_PAGE_SHIFT) + 1);
2608         }
2609         ptlrpc_lprocfs_brw(req, aa->aa_requested_nob);
2610
2611         client_obd_list_lock(&cli->cl_loi_list_lock);
2612
2613         if (cmd == OBD_BRW_READ)
2614                 cli->cl_r_in_flight++;
2615         else
2616                 cli->cl_w_in_flight++;
2617
2618         /* queued sync pages can be torn down while the pages
2619          * were between the pending list and the rpc */
2620         tmp = NULL;
2621         cfs_list_for_each_entry(oap, &aa->aa_oaps, oap_rpc_item) {
2622                 /* only one oap gets a request reference */
2623                 if (tmp == NULL)
2624                         tmp = oap;
2625                 if (oap->oap_interrupted && !req->rq_intr) {
2626                         CDEBUG(D_INODE, "oap %p in req %p interrupted\n",
2627                                oap, req);
2628                         ptlrpc_mark_interrupted(req);
2629                 }
2630         }
2631         if (tmp != NULL)
2632                 tmp->oap_request = ptlrpc_request_addref(req);
2633
2634         DEBUG_REQ(D_INODE, req, "%d pages, aa %p. now %dr/%dw in flight",
2635                   page_count, aa, cli->cl_r_in_flight, cli->cl_w_in_flight);
2636
2637         req->rq_interpret_reply = brw_interpret;
2638         ptlrpcd_add_req(req, PSCOPE_BRW);
2639         RETURN(1);
2640 }
2641
2642 #define LOI_DEBUG(LOI, STR, args...)                                     \
2643         CDEBUG(D_INODE, "loi ready %d wr %d:%d rd %d:%d " STR,           \
2644                !cfs_list_empty(&(LOI)->loi_ready_item) ||                \
2645                !cfs_list_empty(&(LOI)->loi_hp_ready_item),               \
2646                (LOI)->loi_write_lop.lop_num_pending,                     \
2647                !cfs_list_empty(&(LOI)->loi_write_lop.lop_urgent),        \
2648                (LOI)->loi_read_lop.lop_num_pending,                      \
2649                !cfs_list_empty(&(LOI)->loi_read_lop.lop_urgent),         \
2650                args)                                                     \
2651
2652 /* This is called by osc_check_rpcs() to find which objects have pages that
2653  * we could be sending.  These lists are maintained by lop_makes_rpc(). */
2654 struct lov_oinfo *osc_next_loi(struct client_obd *cli)
2655 {
2656         ENTRY;
2657
2658         /* First return objects that have blocked locks so that they
2659          * will be flushed quickly and other clients can get the lock,
2660          * then objects which have pages ready to be stuffed into RPCs */
2661         if (!cfs_list_empty(&cli->cl_loi_hp_ready_list))
2662                 RETURN(cfs_list_entry(cli->cl_loi_hp_ready_list.next,
2663                                       struct lov_oinfo, loi_hp_ready_item));
2664         if (!cfs_list_empty(&cli->cl_loi_ready_list))
2665                 RETURN(cfs_list_entry(cli->cl_loi_ready_list.next,
2666                                       struct lov_oinfo, loi_ready_item));
2667
2668         /* then if we have cache waiters, return all objects with queued
2669          * writes.  This is especially important when many small files
2670          * have filled up the cache and not been fired into rpcs because
2671          * they don't pass the nr_pending/object threshhold */
2672         if (!cfs_list_empty(&cli->cl_cache_waiters) &&
2673             !cfs_list_empty(&cli->cl_loi_write_list))
2674                 RETURN(cfs_list_entry(cli->cl_loi_write_list.next,
2675                                       struct lov_oinfo, loi_write_item));
2676
2677         /* then return all queued objects when we have an invalid import
2678          * so that they get flushed */
2679         if (cli->cl_import == NULL || cli->cl_import->imp_invalid) {
2680                 if (!cfs_list_empty(&cli->cl_loi_write_list))
2681                         RETURN(cfs_list_entry(cli->cl_loi_write_list.next,
2682                                               struct lov_oinfo,
2683                                               loi_write_item));
2684                 if (!cfs_list_empty(&cli->cl_loi_read_list))
2685                         RETURN(cfs_list_entry(cli->cl_loi_read_list.next,
2686                                               struct lov_oinfo, loi_read_item));
2687         }
2688         RETURN(NULL);
2689 }
2690
2691 static int osc_max_rpc_in_flight(struct client_obd *cli, struct lov_oinfo *loi)
2692 {
2693         struct osc_async_page *oap;
2694         int hprpc = 0;
2695
2696         if (!cfs_list_empty(&loi->loi_write_lop.lop_urgent)) {
2697                 oap = cfs_list_entry(loi->loi_write_lop.lop_urgent.next,
2698                                      struct osc_async_page, oap_urgent_item);
2699                 hprpc = !!(oap->oap_async_flags & ASYNC_HP);
2700         }
2701
2702         if (!hprpc && !cfs_list_empty(&loi->loi_read_lop.lop_urgent)) {
2703                 oap = cfs_list_entry(loi->loi_read_lop.lop_urgent.next,
2704                                      struct osc_async_page, oap_urgent_item);
2705                 hprpc = !!(oap->oap_async_flags & ASYNC_HP);
2706         }
2707
2708         return rpcs_in_flight(cli) >= cli->cl_max_rpcs_in_flight + hprpc;
2709 }
2710
2711 /* called with the loi list lock held */
2712 void osc_check_rpcs(const struct lu_env *env, struct client_obd *cli)
2713 {
2714         struct lov_oinfo *loi;
2715         int rc = 0, race_counter = 0;
2716         ENTRY;
2717
2718         while ((loi = osc_next_loi(cli)) != NULL) {
2719                 LOI_DEBUG(loi, "%lu in flight\n", rpcs_in_flight(cli));
2720
2721                 if (osc_max_rpc_in_flight(cli, loi))
2722                         break;
2723
2724                 /* attempt some read/write balancing by alternating between
2725                  * reads and writes in an object.  The makes_rpc checks here
2726                  * would be redundant if we were getting read/write work items
2727                  * instead of objects.  we don't want send_oap_rpc to drain a
2728                  * partial read pending queue when we're given this object to
2729                  * do io on writes while there are cache waiters */
2730                 if (lop_makes_rpc(cli, &loi->loi_write_lop, OBD_BRW_WRITE)) {
2731                         rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_WRITE,
2732                                               &loi->loi_write_lop);
2733                         if (rc < 0) {
2734                                 CERROR("Write request failed with %d\n", rc);
2735
2736                                 /* osc_send_oap_rpc failed, mostly because of
2737                                  * memory pressure.
2738                                  *
2739                                  * It can't break here, because if:
2740                                  *  - a page was submitted by osc_io_submit, so
2741                                  *    page locked;
2742                                  *  - no request in flight
2743                                  *  - no subsequent request
2744                                  * The system will be in live-lock state,
2745                                  * because there is no chance to call
2746                                  * osc_io_unplug() and osc_check_rpcs() any
2747                                  * more. pdflush can't help in this case,
2748                                  * because it might be blocked at grabbing
2749                                  * the page lock as we mentioned.
2750                                  *
2751                                  * Anyway, continue to drain pages. */
2752                                 /* break; */
2753                         }
2754
2755                         if (rc > 0)
2756                                 race_counter = 0;
2757                         else
2758                                 race_counter++;
2759                 }
2760                 if (lop_makes_rpc(cli, &loi->loi_read_lop, OBD_BRW_READ)) {
2761                         rc = osc_send_oap_rpc(env, cli, loi, OBD_BRW_READ,
2762                                               &loi->loi_read_lop);
2763                         if (rc < 0)
2764                                 CERROR("Read request failed with %d\n", rc);
2765
2766                         if (rc > 0)
2767                                 race_counter = 0;
2768                         else
2769                                 race_counter++;
2770                 }
2771
2772                 /* attempt some inter-object balancing by issuing rpcs
2773                  * for each object in turn */
2774                 if (!cfs_list_empty(&loi->loi_hp_ready_item))
2775                         cfs_list_del_init(&loi->loi_hp_ready_item);
2776                 if (!cfs_list_empty(&loi->loi_ready_item))
2777                         cfs_list_del_init(&loi->loi_ready_item);
2778                 if (!cfs_list_empty(&loi->loi_write_item))
2779                         cfs_list_del_init(&loi->loi_write_item);
2780                 if (!cfs_list_empty(&loi->loi_read_item))
2781                         cfs_list_del_init(&loi->loi_read_item);
2782
2783                 loi_list_maint(cli, loi);
2784
2785                 /* send_oap_rpc fails with 0 when make_ready tells it to
2786                  * back off.  llite's make_ready does this when it tries
2787                  * to lock a page queued for write that is already locked.
2788                  * we want to try sending rpcs from many objects, but we
2789                  * don't want to spin failing with 0.  */
2790                 if (race_counter == 10)
2791                         break;
2792         }
2793         EXIT;
2794 }
2795
2796 /* we're trying to queue a page in the osc so we're subject to the
2797  * 'cl_dirty_max' limit on the number of pages that can be queued in the osc.
2798  * If the osc's queued pages are already at that limit, then we want to sleep
2799  * until there is space in the osc's queue for us.  We also may be waiting for
2800  * write credits from the OST if there are RPCs in flight that may return some
2801  * before we fall back to sync writes.
2802  *
2803  * We need this know our allocation was granted in the presence of signals */
2804 static int ocw_granted(struct client_obd *cli, struct osc_cache_waiter *ocw)
2805 {
2806         int rc;
2807         ENTRY;
2808         client_obd_list_lock(&cli->cl_loi_list_lock);
2809         rc = cfs_list_empty(&ocw->ocw_entry) || rpcs_in_flight(cli) == 0;
2810         client_obd_list_unlock(&cli->cl_loi_list_lock);
2811         RETURN(rc);
2812 };
2813
2814 /**
2815  * Non-blocking version of osc_enter_cache() that consumes grant only when it
2816  * is available.
2817  */
2818 int osc_enter_cache_try(const struct lu_env *env,
2819                         struct client_obd *cli, struct lov_oinfo *loi,
2820                         struct osc_async_page *oap, int transient)
2821 {
2822         int has_grant;
2823
2824         has_grant = cli->cl_avail_grant >= CFS_PAGE_SIZE;
2825         if (has_grant) {
2826                 osc_consume_write_grant(cli, &oap->oap_brw_page);
2827                 if (transient) {
2828                         cli->cl_dirty_transit += CFS_PAGE_SIZE;
2829                         cfs_atomic_inc(&obd_dirty_transit_pages);
2830                         oap->oap_brw_flags |= OBD_BRW_NOCACHE;
2831                 }
2832         }
2833         return has_grant;
2834 }
2835
2836 /* Caller must hold loi_list_lock - we drop/regain it if we need to wait for
2837  * grant or cache space. */
2838 static int osc_enter_cache(const struct lu_env *env,
2839                            struct client_obd *cli, struct lov_oinfo *loi,
2840                            struct osc_async_page *oap)
2841 {
2842         struct osc_cache_waiter ocw;
2843         struct l_wait_info lwi = { 0 };
2844
2845         ENTRY;
2846
2847         CDEBUG(D_CACHE, "dirty: %ld/%d dirty_max: %ld/%d dropped: %lu "
2848                "grant: %lu\n", cli->cl_dirty, cfs_atomic_read(&obd_dirty_pages),
2849                cli->cl_dirty_max, obd_max_dirty_pages,
2850                cli->cl_lost_grant, cli->cl_avail_grant);
2851
2852         /* force the caller to try sync io.  this can jump the list
2853          * of queued writes and create a discontiguous rpc stream */
2854         if (cli->cl_dirty_max < CFS_PAGE_SIZE || cli->cl_ar.ar_force_sync ||
2855             loi->loi_ar.ar_force_sync)
2856                 RETURN(-EDQUOT);
2857
2858         /* Hopefully normal case - cache space and write credits available */
2859         if (cli->cl_dirty + CFS_PAGE_SIZE <= cli->cl_dirty_max &&
2860             cfs_atomic_read(&obd_dirty_pages) + 1 <= obd_max_dirty_pages &&
2861             osc_enter_cache_try(env, cli, loi, oap, 0))
2862                 RETURN(0);
2863
2864         /* It is safe to block as a cache waiter as long as there is grant
2865          * space available or the hope of additional grant being returned
2866          * when an in flight write completes.  Using the write back cache
2867          * if possible is preferable to sending the data synchronously
2868          * because write pages can then be merged in to large requests.
2869          * The addition of this cache waiter will causing pending write
2870          * pages to be sent immediately. */
2871         if (cli->cl_w_in_flight || cli->cl_avail_grant >= CFS_PAGE_SIZE) {
2872                 cfs_list_add_tail(&ocw.ocw_entry, &cli->cl_cache_waiters);
2873                 cfs_waitq_init(&ocw.ocw_waitq);
2874                 ocw.ocw_oap = oap;
2875                 ocw.ocw_rc = 0;
2876
2877                 loi_list_maint(cli, loi);
2878                 osc_check_rpcs(env, cli);
2879                 client_obd_list_unlock(&cli->cl_loi_list_lock);
2880
2881                 CDEBUG(D_CACHE, "sleeping for cache space\n");
2882                 l_wait_event(ocw.ocw_waitq, ocw_granted(cli, &ocw), &lwi);
2883
2884                 client_obd_list_lock(&cli->cl_loi_list_lock);
2885                 if (!cfs_list_empty(&ocw.ocw_entry)) {
2886                         cfs_list_del(&ocw.ocw_entry);
2887                         RETURN(-EINTR);
2888                 }
2889                 RETURN(ocw.ocw_rc);
2890         }
2891
2892         RETURN(-EDQUOT);
2893 }
2894
2895
2896 int osc_prep_async_page(struct obd_export *exp, struct lov_stripe_md *lsm,
2897                         struct lov_oinfo *loi, cfs_page_t *page,
2898                         obd_off offset, const struct obd_async_page_ops *ops,
2899                         void *data, void **res, int nocache,
2900                         struct lustre_handle *lockh)
2901 {
2902         struct osc_async_page *oap;
2903
2904         ENTRY;
2905
2906         if (!page)
2907                 return cfs_size_round(sizeof(*oap));
2908
2909         oap = *res;
2910         oap->oap_magic = OAP_MAGIC;
2911         oap->oap_cli = &exp->exp_obd->u.cli;
2912         oap->oap_loi = loi;
2913
2914         oap->oap_caller_ops = ops;
2915         oap->oap_caller_data = data;
2916
2917         oap->oap_page = page;
2918         oap->oap_obj_off = offset;
2919         if (!client_is_remote(exp) &&
2920             cfs_capable(CFS_CAP_SYS_RESOURCE))
2921                 oap->oap_brw_flags = OBD_BRW_NOQUOTA;
2922
2923         LASSERT(!(offset & ~CFS_PAGE_MASK));
2924
2925         CFS_INIT_LIST_HEAD(&oap->oap_pending_item);
2926         CFS_INIT_LIST_HEAD(&oap->oap_urgent_item);
2927         CFS_INIT_LIST_HEAD(&oap->oap_rpc_item);
2928         CFS_INIT_LIST_HEAD(&oap->oap_page_list);
2929
2930         cfs_spin_lock_init(&oap->oap_lock);
2931         CDEBUG(D_CACHE, "oap %p page %p obj off "LPU64"\n", oap, page, offset);
2932         RETURN(0);
2933 }
2934
2935 struct osc_async_page *oap_from_cookie(void *cookie)
2936 {
2937         struct osc_async_page *oap = cookie;
2938         if (oap->oap_magic != OAP_MAGIC)
2939                 return ERR_PTR(-EINVAL);
2940         return oap;
2941 };
2942
2943 int osc_queue_async_io(const struct lu_env *env,
2944                        struct obd_export *exp, struct lov_stripe_md *lsm,
2945                        struct lov_oinfo *loi, void *cookie,
2946                        int cmd, obd_off off, int count,
2947                        obd_flag brw_flags, enum async_flags async_flags)
2948 {
2949         struct client_obd *cli = &exp->exp_obd->u.cli;
2950         struct osc_async_page *oap;
2951         int rc = 0;
2952         ENTRY;
2953
2954         oap = oap_from_cookie(cookie);
2955         if (IS_ERR(oap))
2956                 RETURN(PTR_ERR(oap));
2957
2958         if (cli->cl_import == NULL || cli->cl_import->imp_invalid)
2959                 RETURN(-EIO);
2960
2961         if (!cfs_list_empty(&oap->oap_pending_item) ||
2962             !cfs_list_empty(&oap->oap_urgent_item) ||
2963             !cfs_list_empty(&oap->oap_rpc_item))
2964                 RETURN(-EBUSY);
2965
2966         /* check if the file's owner/group is over quota */
2967         if ((cmd & OBD_BRW_WRITE) && !(cmd & OBD_BRW_NOQUOTA)) {
2968                 struct cl_object *obj;
2969                 struct cl_attr    attr; /* XXX put attr into thread info */
2970                 unsigned int qid[MAXQUOTAS];
2971
2972                 obj = cl_object_top(osc_oap2cl_page(oap)->cp_obj);
2973
2974                 cl_object_attr_lock(obj);
2975                 rc = cl_object_attr_get(env, obj, &attr);
2976                 cl_object_attr_unlock(obj);
2977
2978                 qid[USRQUOTA] = attr.cat_uid;
2979                 qid[GRPQUOTA] = attr.cat_gid;
2980                 if (rc == 0 &&
2981                     lquota_chkdq(quota_interface, cli, qid) == NO_QUOTA)
2982                         rc = -EDQUOT;
2983                 if (rc)
2984                         RETURN(rc);
2985         }
2986
2987         if (loi == NULL)
2988                 loi = lsm->lsm_oinfo[0];
2989
2990         client_obd_list_lock(&cli->cl_loi_list_lock);
2991
2992         LASSERT(off + count <= CFS_PAGE_SIZE);
2993         oap->oap_cmd = cmd;
2994         oap->oap_page_off = off;
2995         oap->oap_count = count;
2996         oap->oap_brw_flags = brw_flags;
2997         /* Give a hint to OST that requests are coming from kswapd - bug19529 */
2998         if (cfs_memory_pressure_get())
2999                 oap->oap_brw_flags |= OBD_BRW_MEMALLOC;
3000         cfs_spin_lock(&oap->oap_lock);
3001         oap->oap_async_flags = async_flags;
3002         cfs_spin_unlock(&oap->oap_lock);
3003
3004         if (cmd & OBD_BRW_WRITE) {
3005                 rc = osc_enter_cache(env, cli, loi, oap);
3006                 if (rc) {
3007                         client_obd_list_unlock(&cli->cl_loi_list_lock);
3008                         RETURN(rc);
3009                 }
3010         }
3011
3012         osc_oap_to_pending(oap);
3013         loi_list_maint(cli, loi);
3014
3015         LOI_DEBUG(loi, "oap %p page %p added for cmd %d\n", oap, oap->oap_page,
3016                   cmd);
3017
3018         osc_check_rpcs(env, cli);
3019         client_obd_list_unlock(&cli->cl_loi_list_lock);
3020
3021         RETURN(0);
3022 }
3023
3024 /* aka (~was & now & flag), but this is more clear :) */
3025 #define SETTING(was, now, flag) (!(was & flag) && (now & flag))
3026
3027 int osc_set_async_flags_base(struct client_obd *cli,
3028                              struct lov_oinfo *loi, struct osc_async_page *oap,
3029                              obd_flag async_flags)
3030 {
3031         struct loi_oap_pages *lop;
3032         int flags = 0;
3033         ENTRY;
3034
3035         LASSERT(!cfs_list_empty(&oap->oap_pending_item));
3036
3037         if (oap->oap_cmd & OBD_BRW_WRITE) {
3038                 lop = &loi->loi_write_lop;
3039         } else {
3040                 lop = &loi->loi_read_lop;
3041         }
3042
3043         if ((oap->oap_async_flags & async_flags) == async_flags)
3044                 RETURN(0);
3045
3046         if (SETTING(oap->oap_async_flags, async_flags, ASYNC_READY))
3047                 flags |= ASYNC_READY;
3048
3049         if (SETTING(oap->oap_async_flags, async_flags, ASYNC_URGENT) &&
3050             cfs_list_empty(&oap->oap_rpc_item)) {
3051                 if (oap->oap_async_flags & ASYNC_HP)
3052                         cfs_list_add(&oap->oap_urgent_item, &lop->lop_urgent);
3053                 else
3054                         cfs_list_add_tail(&oap->oap_urgent_item,
3055                                           &lop->lop_urgent);
3056                 flags |= ASYNC_URGENT;
3057                 loi_list_maint(cli, loi);
3058         }
3059         cfs_spin_lock(&oap->oap_lock);
3060         oap->oap_async_flags |= flags;
3061         cfs_spin_unlock(&oap->oap_lock);
3062
3063         LOI_DEBUG(loi, "oap %p page %p has flags %x\n", oap, oap->oap_page,
3064                         oap->oap_async_flags);
3065         RETURN(0);
3066 }
3067
3068 int osc_teardown_async_page(struct obd_export *exp,
3069                             struct lov_stripe_md *lsm,
3070                             struct lov_oinfo *loi, void *cookie)
3071 {
3072         struct client_obd *cli = &exp->exp_obd->u.cli;
3073         struct loi_oap_pages *lop;
3074         struct osc_async_page *oap;
3075         int rc = 0;
3076         ENTRY;
3077
3078         oap = oap_from_cookie(cookie);
3079         if (IS_ERR(oap))
3080                 RETURN(PTR_ERR(oap));
3081
3082         if (loi == NULL)
3083                 loi = lsm->lsm_oinfo[0];
3084
3085         if (oap->oap_cmd & OBD_BRW_WRITE) {
3086                 lop = &loi->loi_write_lop;
3087         } else {
3088                 lop = &loi->loi_read_lop;
3089         }
3090
3091         client_obd_list_lock(&cli->cl_loi_list_lock);
3092
3093         if (!cfs_list_empty(&oap->oap_rpc_item))
3094                 GOTO(out, rc = -EBUSY);
3095
3096         osc_exit_cache(cli, oap, 0);
3097         osc_wake_cache_waiters(cli);
3098
3099         if (!cfs_list_empty(&oap->oap_urgent_item)) {
3100                 cfs_list_del_init(&oap->oap_urgent_item);
3101                 cfs_spin_lock(&oap->oap_lock);
3102                 oap->oap_async_flags &= ~(ASYNC_URGENT | ASYNC_HP);
3103                 cfs_spin_unlock(&oap->oap_lock);
3104         }
3105         if (!cfs_list_empty(&oap->oap_pending_item)) {
3106                 cfs_list_del_init(&oap->oap_pending_item);
3107                 lop_update_pending(cli, lop, oap->oap_cmd, -1);
3108         }
3109         loi_list_maint(cli, loi);
3110         LOI_DEBUG(loi, "oap %p page %p torn down\n", oap, oap->oap_page);
3111 out:
3112         client_obd_list_unlock(&cli->cl_loi_list_lock);
3113         RETURN(rc);
3114 }
3115
3116 static void osc_set_lock_data_with_check(struct ldlm_lock *lock,
3117                                          struct ldlm_enqueue_info *einfo,
3118                                          int flags)
3119 {
3120         void *data = einfo->ei_cbdata;
3121
3122         LASSERT(lock != NULL);
3123         LASSERT(lock->l_blocking_ast == einfo->ei_cb_bl);
3124         LASSERT(lock->l_resource->lr_type == einfo->ei_type);
3125         LASSERT(lock->l_completion_ast == einfo->ei_cb_cp);
3126         LASSERT(lock->l_glimpse_ast == einfo->ei_cb_gl);
3127
3128         lock_res_and_lock(lock);
3129         cfs_spin_lock(&osc_ast_guard);
3130         LASSERT(lock->l_ast_data == NULL || lock->l_ast_data == data);
3131         lock->l_ast_data = data;
3132         cfs_spin_unlock(&osc_ast_guard);
3133         unlock_res_and_lock(lock);
3134 }
3135
3136 static void osc_set_data_with_check(struct lustre_handle *lockh,
3137                                     struct ldlm_enqueue_info *einfo,
3138                                     int flags)
3139 {
3140         struct ldlm_lock *lock = ldlm_handle2lock(lockh);
3141
3142         if (lock != NULL) {
3143                 osc_set_lock_data_with_check(lock, einfo, flags);
3144                 LDLM_LOCK_PUT(lock);
3145         } else
3146                 CERROR("lockh %p, data %p - client evicted?\n",
3147                        lockh, einfo->ei_cbdata);
3148 }
3149
3150 static int osc_change_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
3151                              ldlm_iterator_t replace, void *data)
3152 {
3153         struct ldlm_res_id res_id;
3154         struct obd_device *obd = class_exp2obd(exp);
3155
3156         osc_build_res_name(lsm->lsm_object_id, lsm->lsm_object_seq, &res_id);
3157         ldlm_resource_iterate(obd->obd_namespace, &res_id, replace, data);
3158         return 0;
3159 }
3160
3161 /* find any ldlm lock of the inode in osc
3162  * return 0    not find
3163  *        1    find one
3164  *      < 0    error */
3165 static int osc_find_cbdata(struct obd_export *exp, struct lov_stripe_md *lsm,
3166                            ldlm_iterator_t replace, void *data)
3167 {