projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
staging: add Lustre file system client support
[deliverable/linux.git] / drivers / staging / lustre / lustre / lov / lov_obd.c
CommitLineData
d7e09d03
PT		 1/*
2 * GPL HEADER START
3 *
4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
19 *
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
22 * have any questions.
23 *
24 * GPL HEADER END
25 */
26/*
27 * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
29 *
30 * Copyright (c) 2011, 2012, Intel Corporation.
31 */
32/*
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
35 *
36 * lustre/lov/lov_obd.c
37 *
38 * Author: Phil Schwan <phil@clusterfs.com>
39 * Author: Peter Braam <braam@clusterfs.com>
40 * Author: Mike Shaver <shaver@clusterfs.com>
41 * Author: Nathan Rutman <nathan@clusterfs.com>
42 */
43
44#define DEBUG_SUBSYSTEM S_LOV
45#include <linux/libcfs/libcfs.h>
46
47#include <obd_support.h>
48#include <lustre_lib.h>
49#include <lustre_net.h>
50#include <lustre/lustre_idl.h>
51#include <lustre_dlm.h>
52#include <lustre_mds.h>
53#include <lustre_debug.h>
54#include <obd_class.h>
55#include <obd_lov.h>
56#include <obd_ost.h>
57#include <lprocfs_status.h>
58#include <lustre_param.h>
59#include <cl_object.h>
60#include <lclient.h>
61#include <lustre/ll_fiemap.h>
62#include <lustre_log.h>
63#include <lustre_fid.h>
64
65#include "lov_internal.h"
66
67/* Keep a refcount of lov->tgt usage to prevent racing with addition/deletion.
68 Any function that expects lov_tgts to remain stationary must take a ref. */
69static void lov_getref(struct obd_device *obd)
70{
71 struct lov_obd *lov = &obd->u.lov;
72
73 /* nobody gets through here until lov_putref is done */
74 mutex_lock(&lov->lov_lock);
75 atomic_inc(&lov->lov_refcount);
76 mutex_unlock(&lov->lov_lock);
77 return;
78}
79
80static void __lov_del_obd(struct obd_device *obd, struct lov_tgt_desc *tgt);
81
82static void lov_putref(struct obd_device *obd)
83{
84 struct lov_obd *lov = &obd->u.lov;
85
86 mutex_lock(&lov->lov_lock);
87 /* ok to dec to 0 more than once -- ltd_exp's will be null */
88 if (atomic_dec_and_test(&lov->lov_refcount) && lov->lov_death_row) {
89 LIST_HEAD(kill);
90 int i;
91 struct lov_tgt_desc *tgt, *n;
92 CDEBUG(D_CONFIG, "destroying %d lov targets\n",
93 lov->lov_death_row);
94 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
95 tgt = lov->lov_tgts[i];
96
97 if (!tgt || !tgt->ltd_reap)
98 continue;
99 list_add(&tgt->ltd_kill, &kill);
100 /* XXX - right now there is a dependency on ld_tgt_count
101 * being the maximum tgt index for computing the
102 * mds_max_easize. So we can't shrink it. */
103 lov_ost_pool_remove(&lov->lov_packed, i);
104 lov->lov_tgts[i] = NULL;
105 lov->lov_death_row--;
106 }
107 mutex_unlock(&lov->lov_lock);
108
109 list_for_each_entry_safe(tgt, n, &kill, ltd_kill) {
110 list_del(&tgt->ltd_kill);
111 /* Disconnect */
112 __lov_del_obd(obd, tgt);
113 }
114 } else {
115 mutex_unlock(&lov->lov_lock);
116 }
117}
118
119static int lov_set_osc_active(struct obd_device *obd, struct obd_uuid *uuid,
120 enum obd_notify_event ev);
121static int lov_notify(struct obd_device *obd, struct obd_device *watched,
122 enum obd_notify_event ev, void *data);
123
124
125#define MAX_STRING_SIZE 128
126int lov_connect_obd(struct obd_device *obd, __u32 index, int activate,
127 struct obd_connect_data *data)
128{
129 struct lov_obd *lov = &obd->u.lov;
130 struct obd_uuid *tgt_uuid;
131 struct obd_device *tgt_obd;
132 static struct obd_uuid lov_osc_uuid = { "LOV_OSC_UUID" };
133 struct obd_import *imp;
134 proc_dir_entry_t *lov_proc_dir;
135 int rc;
136 ENTRY;
137
138 if (!lov->lov_tgts[index])
139 RETURN(-EINVAL);
140
141 tgt_uuid = &lov->lov_tgts[index]->ltd_uuid;
142 tgt_obd = lov->lov_tgts[index]->ltd_obd;
143
144 if (!tgt_obd->obd_set_up) {
145 CERROR("Target %s not set up\n", obd_uuid2str(tgt_uuid));
146 RETURN(-EINVAL);
147 }
148
149 /* override the sp_me from lov */
150 tgt_obd->u.cli.cl_sp_me = lov->lov_sp_me;
151
152 if (data && (data->ocd_connect_flags & OBD_CONNECT_INDEX))
153 data->ocd_index = index;
154
155 /*
156 * Divine LOV knows that OBDs under it are OSCs.
157 */
158 imp = tgt_obd->u.cli.cl_import;
159
160 if (activate) {
161 tgt_obd->obd_no_recov = 0;
162 /* FIXME this is probably supposed to be
163 ptlrpc_set_import_active. Horrible naming. */
164 ptlrpc_activate_import(imp);
165 }
166
167 rc = obd_register_observer(tgt_obd, obd);
168 if (rc) {
169 CERROR("Target %s register_observer error %d\n",
170 obd_uuid2str(tgt_uuid), rc);
171 RETURN(rc);
172 }
173
174
175 if (imp->imp_invalid) {
176 CDEBUG(D_CONFIG, "not connecting OSC %s; administratively "
177 "disabled\n", obd_uuid2str(tgt_uuid));
178 RETURN(0);
179 }
180
181 rc = obd_connect(NULL, &lov->lov_tgts[index]->ltd_exp, tgt_obd,
182 &lov_osc_uuid, data, NULL);
183 if (rc || !lov->lov_tgts[index]->ltd_exp) {
184 CERROR("Target %s connect error %d\n",
185 obd_uuid2str(tgt_uuid), rc);
186 RETURN(-ENODEV);
187 }
188
189 lov->lov_tgts[index]->ltd_reap = 0;
190
191 CDEBUG(D_CONFIG, "Connected tgt idx %d %s (%s) %sactive\n", index,
192 obd_uuid2str(tgt_uuid), tgt_obd->obd_name, activate ? "":"in");
193
194 lov_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds");
195 if (lov_proc_dir) {
196 struct obd_device *osc_obd = lov->lov_tgts[index]->ltd_exp->exp_obd;
197 proc_dir_entry_t *osc_symlink;
198
199 LASSERT(osc_obd != NULL);
200 LASSERT(osc_obd->obd_magic == OBD_DEVICE_MAGIC);
201 LASSERT(osc_obd->obd_type->typ_name != NULL);
202
203 osc_symlink = lprocfs_add_symlink(osc_obd->obd_name,
204 lov_proc_dir,
205 "../../../%s/%s",
206 osc_obd->obd_type->typ_name,
207 osc_obd->obd_name);
208 if (osc_symlink == NULL) {
209 CERROR("could not register LOV target "
210 "/proc/fs/lustre/%s/%s/target_obds/%s.",
211 obd->obd_type->typ_name, obd->obd_name,
212 osc_obd->obd_name);
213 lprocfs_remove(&lov_proc_dir);
214 }
215 }
216
217 RETURN(0);
218}
219
220static int lov_connect(const struct lu_env *env,
221 struct obd_export **exp, struct obd_device *obd,
222 struct obd_uuid *cluuid, struct obd_connect_data *data,
223 void *localdata)
224{
225 struct lov_obd *lov = &obd->u.lov;
226 struct lov_tgt_desc *tgt;
227 struct lustre_handle conn;
228 int i, rc;
229 ENTRY;
230
231 CDEBUG(D_CONFIG, "connect #%d\n", lov->lov_connects);
232
233 rc = class_connect(&conn, obd, cluuid);
234 if (rc)
235 RETURN(rc);
236
237 *exp = class_conn2export(&conn);
238
239 /* Why should there ever be more than 1 connect? */
240 lov->lov_connects++;
241 LASSERT(lov->lov_connects == 1);
242
243 memset(&lov->lov_ocd, 0, sizeof(lov->lov_ocd));
244 if (data)
245 lov->lov_ocd = *data;
246
247 obd_getref(obd);
248 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
249 tgt = lov->lov_tgts[i];
250 if (!tgt || obd_uuid_empty(&tgt->ltd_uuid))
251 continue;
252 /* Flags will be lowest common denominator */
253 rc = lov_connect_obd(obd, i, tgt->ltd_activate, &lov->lov_ocd);
254 if (rc) {
255 CERROR("%s: lov connect tgt %d failed: %d\n",
256 obd->obd_name, i, rc);
257 continue;
258 }
259 /* connect to administrative disabled ost */
260 if (!lov->lov_tgts[i]->ltd_exp)
261 continue;
262
263 rc = lov_notify(obd, lov->lov_tgts[i]->ltd_exp->exp_obd,
264 OBD_NOTIFY_CONNECT, (void *)&i);
265 if (rc) {
266 CERROR("%s error sending notify %d\n",
267 obd->obd_name, rc);
268 }
269 }
270 obd_putref(obd);
271
272 RETURN(0);
273}
274
275static int lov_disconnect_obd(struct obd_device *obd, struct lov_tgt_desc *tgt)
276{
277 proc_dir_entry_t *lov_proc_dir;
278 struct lov_obd *lov = &obd->u.lov;
279 struct obd_device *osc_obd;
280 int rc;
281 ENTRY;
282
283 osc_obd = class_exp2obd(tgt->ltd_exp);
284 CDEBUG(D_CONFIG, "%s: disconnecting target %s\n",
285 obd->obd_name, osc_obd->obd_name);
286
287 if (tgt->ltd_active) {
288 tgt->ltd_active = 0;
289 lov->desc.ld_active_tgt_count--;
290 tgt->ltd_exp->exp_obd->obd_inactive = 1;
291 }
292
293 lov_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds");
294 if (lov_proc_dir) {
295 proc_dir_entry_t *osc_symlink;
296
297 osc_symlink = lprocfs_srch(lov_proc_dir, osc_obd->obd_name);
298 if (osc_symlink) {
299 lprocfs_remove(&osc_symlink);
300 } else {
301 CERROR("/proc/fs/lustre/%s/%s/target_obds/%s missing.",
302 obd->obd_type->typ_name, obd->obd_name,
303 osc_obd->obd_name);
304 }
305 }
306
307 if (osc_obd) {
308 /* Pass it on to our clients.
309 * XXX This should be an argument to disconnect,
310 * XXX not a back-door flag on the OBD. Ah well.
311 */
312 osc_obd->obd_force = obd->obd_force;
313 osc_obd->obd_fail = obd->obd_fail;
314 osc_obd->obd_no_recov = obd->obd_no_recov;
315 }
316
317 obd_register_observer(osc_obd, NULL);
318
319 rc = obd_disconnect(tgt->ltd_exp);
320 if (rc) {
321 CERROR("Target %s disconnect error %d\n",
322 tgt->ltd_uuid.uuid, rc);
323 rc = 0;
324 }
325
326 tgt->ltd_exp = NULL;
327 RETURN(0);
328}
329
330static int lov_disconnect(struct obd_export *exp)
331{
332 struct obd_device *obd = class_exp2obd(exp);
333 struct lov_obd *lov = &obd->u.lov;
334 int i, rc;
335 ENTRY;
336
337 if (!lov->lov_tgts)
338 goto out;
339
340 /* Only disconnect the underlying layers on the final disconnect. */
341 lov->lov_connects--;
342 if (lov->lov_connects != 0) {
343 /* why should there be more than 1 connect? */
344 CERROR("disconnect #%d\n", lov->lov_connects);
345 goto out;
346 }
347
348 /* Let's hold another reference so lov_del_obd doesn't spin through
349 putref every time */
350 obd_getref(obd);
351
352 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
353 if (lov->lov_tgts[i] && lov->lov_tgts[i]->ltd_exp) {
354 /* Disconnection is the last we know about an obd */
355 lov_del_target(obd, i, 0, lov->lov_tgts[i]->ltd_gen);
356 }
357 }
358 obd_putref(obd);
359
360out:
361 rc = class_disconnect(exp); /* bz 9811 */
362 RETURN(rc);
363}
364
365/* Error codes:
366 *
367 * -EINVAL : UUID can't be found in the LOV's target list
368 * -ENOTCONN: The UUID is found, but the target connection is bad (!)
369 * -EBADF : The UUID is found, but the OBD is the wrong type (!)
370 * any >= 0 : is log target index
371 */
372static int lov_set_osc_active(struct obd_device *obd, struct obd_uuid *uuid,
373 enum obd_notify_event ev)
374{
375 struct lov_obd *lov = &obd->u.lov;
376 struct lov_tgt_desc *tgt;
377 int index, activate, active;
378 ENTRY;
379
380 CDEBUG(D_INFO, "Searching in lov %p for uuid %s event(%d)\n",
381 lov, uuid->uuid, ev);
382
383 obd_getref(obd);
384 for (index = 0; index < lov->desc.ld_tgt_count; index++) {
385 tgt = lov->lov_tgts[index];
386 if (!tgt)
387 continue;
388 /*
389 * LU-642, initially inactive OSC could miss the obd_connect,
390 * we make up for it here.
391 */
392 if (ev == OBD_NOTIFY_ACTIVATE && tgt->ltd_exp == NULL &&
393 obd_uuid_equals(uuid, &tgt->ltd_uuid)) {
394 struct obd_uuid lov_osc_uuid = {"LOV_OSC_UUID"};
395
396 obd_connect(NULL, &tgt->ltd_exp, tgt->ltd_obd,
397 &lov_osc_uuid, &lov->lov_ocd, NULL);
398 }
399 if (!tgt->ltd_exp)
400 continue;
401
402 CDEBUG(D_INFO, "lov idx %d is %s conn "LPX64"\n",
403 index, obd_uuid2str(&tgt->ltd_uuid),
404 tgt->ltd_exp->exp_handle.h_cookie);
405 if (obd_uuid_equals(uuid, &tgt->ltd_uuid))
406 break;
407 }
408
409 if (index == lov->desc.ld_tgt_count)
410 GOTO(out, index = -EINVAL);
411
412 if (ev == OBD_NOTIFY_DEACTIVATE || ev == OBD_NOTIFY_ACTIVATE) {
413 activate = (ev == OBD_NOTIFY_ACTIVATE) ? 1 : 0;
414
415 if (lov->lov_tgts[index]->ltd_activate == activate) {
416 CDEBUG(D_INFO, "OSC %s already %sactivate!\n",
417 uuid->uuid, activate ? "" : "de");
418 } else {
419 lov->lov_tgts[index]->ltd_activate = activate;
420 CDEBUG(D_CONFIG, "%sactivate OSC %s\n",
421 activate ? "" : "de", obd_uuid2str(uuid));
422 }
423
424 } else if (ev == OBD_NOTIFY_INACTIVE || ev == OBD_NOTIFY_ACTIVE) {
425 active = (ev == OBD_NOTIFY_ACTIVE) ? 1 : 0;
426
427 if (lov->lov_tgts[index]->ltd_active == active) {
428 CDEBUG(D_INFO, "OSC %s already %sactive!\n",
429 uuid->uuid, active ? "" : "in");
430 GOTO(out, index);
431 } else {
432 CDEBUG(D_CONFIG, "Marking OSC %s %sactive\n",
433 obd_uuid2str(uuid), active ? "" : "in");
434 }
435
436 lov->lov_tgts[index]->ltd_active = active;
437 if (active) {
438 lov->desc.ld_active_tgt_count++;
439 lov->lov_tgts[index]->ltd_exp->exp_obd->obd_inactive = 0;
440 } else {
441 lov->desc.ld_active_tgt_count--;
442 lov->lov_tgts[index]->ltd_exp->exp_obd->obd_inactive = 1;
443 }
444 } else {
445 CERROR("Unknown event(%d) for uuid %s", ev, uuid->uuid);
446 }
447
448 out:
449 obd_putref(obd);
450 RETURN(index);
451}
452
453static int lov_notify(struct obd_device *obd, struct obd_device *watched,
454 enum obd_notify_event ev, void *data)
455{
456 int rc = 0;
457 struct lov_obd *lov = &obd->u.lov;
458 ENTRY;
459
460 down_read(&lov->lov_notify_lock);
461 if (!lov->lov_connects) {
462 up_read(&lov->lov_notify_lock);
463 RETURN(rc);
464 }
465
466 if (ev == OBD_NOTIFY_ACTIVE || ev == OBD_NOTIFY_INACTIVE ||
467 ev == OBD_NOTIFY_ACTIVATE || ev == OBD_NOTIFY_DEACTIVATE) {
468 struct obd_uuid *uuid;
469
470 LASSERT(watched);
471
472 if (strcmp(watched->obd_type->typ_name, LUSTRE_OSC_NAME)) {
473 up_read(&lov->lov_notify_lock);
474 CERROR("unexpected notification of %s %s!\n",
475 watched->obd_type->typ_name,
476 watched->obd_name);
477 RETURN(-EINVAL);
478 }
479 uuid = &watched->u.cli.cl_target_uuid;
480
481 /* Set OSC as active before notifying the observer, so the
482 * observer can use the OSC normally.
483 */
484 rc = lov_set_osc_active(obd, uuid, ev);
485 if (rc < 0) {
486 up_read(&lov->lov_notify_lock);
487 CERROR("event(%d) of %s failed: %d\n", ev,
488 obd_uuid2str(uuid), rc);
489 RETURN(rc);
490 }
491 /* active event should be pass lov target index as data */
492 data = &rc;
493 }
494
495 /* Pass the notification up the chain. */
496 if (watched) {
497 rc = obd_notify_observer(obd, watched, ev, data);
498 } else {
499 /* NULL watched means all osc's in the lov (only for syncs) */
500 /* sync event should be send lov idx as data */
501 struct lov_obd *lov = &obd->u.lov;
502 int i, is_sync;
503
504 data = &i;
505 is_sync = (ev == OBD_NOTIFY_SYNC) ||
506 (ev == OBD_NOTIFY_SYNC_NONBLOCK);
507
508 obd_getref(obd);
509 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
510 if (!lov->lov_tgts[i])
511 continue;
512
513 /* don't send sync event if target not
514 * connected/activated */
515 if (is_sync && !lov->lov_tgts[i]->ltd_active)
516 continue;
517
518 rc = obd_notify_observer(obd, lov->lov_tgts[i]->ltd_obd,
519 ev, data);
520 if (rc) {
521 CERROR("%s: notify %s of %s failed %d\n",
522 obd->obd_name,
523 obd->obd_observer->obd_name,
524 lov->lov_tgts[i]->ltd_obd->obd_name,
525 rc);
526 }
527 }
528 obd_putref(obd);
529 }
530
531 up_read(&lov->lov_notify_lock);
532 RETURN(rc);
533}
534
535static int lov_add_target(struct obd_device *obd, struct obd_uuid *uuidp,
536 __u32 index, int gen, int active)
537{
538 struct lov_obd *lov = &obd->u.lov;
539 struct lov_tgt_desc *tgt;
540 struct obd_device *tgt_obd;
541 int rc;
542 ENTRY;
543
544 CDEBUG(D_CONFIG, "uuid:%s idx:%d gen:%d active:%d\n",
545 uuidp->uuid, index, gen, active);
546
547 if (gen <= 0) {
548 CERROR("request to add OBD %s with invalid generation: %d\n",
549 uuidp->uuid, gen);
550 RETURN(-EINVAL);
551 }
552
553 tgt_obd = class_find_client_obd(uuidp, LUSTRE_OSC_NAME,
554 &obd->obd_uuid);
555 if (tgt_obd == NULL)
556 RETURN(-EINVAL);
557
558 mutex_lock(&lov->lov_lock);
559
560 if ((index < lov->lov_tgt_size) && (lov->lov_tgts[index] != NULL)) {
561 tgt = lov->lov_tgts[index];
562 CERROR("UUID %s already assigned at LOV target index %d\n",
563 obd_uuid2str(&tgt->ltd_uuid), index);
564 mutex_unlock(&lov->lov_lock);
565 RETURN(-EEXIST);
566 }
567
568 if (index >= lov->lov_tgt_size) {
569 /* We need to reallocate the lov target array. */
570 struct lov_tgt_desc **newtgts, **old = NULL;
571 __u32 newsize, oldsize = 0;
572
573 newsize = max(lov->lov_tgt_size, (__u32)2);
574 while (newsize < index + 1)
575 newsize = newsize << 1;
576 OBD_ALLOC(newtgts, sizeof(*newtgts) * newsize);
577 if (newtgts == NULL) {
578 mutex_unlock(&lov->lov_lock);
579 RETURN(-ENOMEM);
580 }
581
582 if (lov->lov_tgt_size) {
583 memcpy(newtgts, lov->lov_tgts, sizeof(*newtgts) *
584 lov->lov_tgt_size);
585 old = lov->lov_tgts;
586 oldsize = lov->lov_tgt_size;
587 }
588
589 lov->lov_tgts = newtgts;
590 lov->lov_tgt_size = newsize;
591 smp_rmb();
592 if (old)
593 OBD_FREE(old, sizeof(*old) * oldsize);
594
595 CDEBUG(D_CONFIG, "tgts: %p size: %d\n",
596 lov->lov_tgts, lov->lov_tgt_size);
597 }
598
599 OBD_ALLOC_PTR(tgt);
600 if (!tgt) {
601 mutex_unlock(&lov->lov_lock);
602 RETURN(-ENOMEM);
603 }
604
605 rc = lov_ost_pool_add(&lov->lov_packed, index, lov->lov_tgt_size);
606 if (rc) {
607 mutex_unlock(&lov->lov_lock);
608 OBD_FREE_PTR(tgt);
609 RETURN(rc);
610 }
611
612 tgt->ltd_uuid = *uuidp;
613 tgt->ltd_obd = tgt_obd;
614 /* XXX - add a sanity check on the generation number. */
615 tgt->ltd_gen = gen;
616 tgt->ltd_index = index;
617 tgt->ltd_activate = active;
618 lov->lov_tgts[index] = tgt;
619 if (index >= lov->desc.ld_tgt_count)
620 lov->desc.ld_tgt_count = index + 1;
621
622 mutex_unlock(&lov->lov_lock);
623
624 CDEBUG(D_CONFIG, "idx=%d ltd_gen=%d ld_tgt_count=%d\n",
625 index, tgt->ltd_gen, lov->desc.ld_tgt_count);
626
627 rc = obd_notify(obd, tgt_obd, OBD_NOTIFY_CREATE, &index);
628
629 if (lov->lov_connects == 0) {
630 /* lov_connect hasn't been called yet. We'll do the
631 lov_connect_obd on this target when that fn first runs,
632 because we don't know the connect flags yet. */
633 RETURN(0);
634 }
635
636 obd_getref(obd);
637
638 rc = lov_connect_obd(obd, index, active, &lov->lov_ocd);
639 if (rc)
640 GOTO(out, rc);
641
642 /* connect to administrative disabled ost */
643 if (!tgt->ltd_exp)
644 GOTO(out, rc = 0);
645
646 if (lov->lov_cache != NULL) {
647 rc = obd_set_info_async(NULL, tgt->ltd_exp,
648 sizeof(KEY_CACHE_SET), KEY_CACHE_SET,
649 sizeof(struct cl_client_cache), lov->lov_cache,
650 NULL);
651 if (rc < 0)
652 GOTO(out, rc);
653 }
654
655 rc = lov_notify(obd, tgt->ltd_exp->exp_obd,
656 active ? OBD_NOTIFY_CONNECT : OBD_NOTIFY_INACTIVE,
657 (void *)&index);
658
659out:
660 if (rc) {
661 CERROR("add failed (%d), deleting %s\n", rc,
662 obd_uuid2str(&tgt->ltd_uuid));
663 lov_del_target(obd, index, 0, 0);
664 }
665 obd_putref(obd);
666 RETURN(rc);
667}
668
669/* Schedule a target for deletion */
670int lov_del_target(struct obd_device *obd, __u32 index,
671 struct obd_uuid *uuidp, int gen)
672{
673 struct lov_obd *lov = &obd->u.lov;
674 int count = lov->desc.ld_tgt_count;
675 int rc = 0;
676 ENTRY;
677
678 if (index >= count) {
679 CERROR("LOV target index %d >= number of LOV OBDs %d.\n",
680 index, count);
681 RETURN(-EINVAL);
682 }
683
684 /* to make sure there's no ongoing lov_notify() now */
685 down_write(&lov->lov_notify_lock);
686 obd_getref(obd);
687
688 if (!lov->lov_tgts[index]) {
689 CERROR("LOV target at index %d is not setup.\n", index);
690 GOTO(out, rc = -EINVAL);
691 }
692
693 if (uuidp && !obd_uuid_equals(uuidp, &lov->lov_tgts[index]->ltd_uuid)) {
694 CERROR("LOV target UUID %s at index %d doesn't match %s.\n",
695 lov_uuid2str(lov, index), index,
696 obd_uuid2str(uuidp));
697 GOTO(out, rc = -EINVAL);
698 }
699
700 CDEBUG(D_CONFIG, "uuid: %s idx: %d gen: %d exp: %p active: %d\n",
701 lov_uuid2str(lov, index), index,
702 lov->lov_tgts[index]->ltd_gen, lov->lov_tgts[index]->ltd_exp,
703 lov->lov_tgts[index]->ltd_active);
704
705 lov->lov_tgts[index]->ltd_reap = 1;
706 lov->lov_death_row++;
707 /* we really delete it from obd_putref */
708out:
709 obd_putref(obd);
710 up_write(&lov->lov_notify_lock);
711
712 RETURN(rc);
713}
714
715static void __lov_del_obd(struct obd_device *obd, struct lov_tgt_desc *tgt)
716{
717 struct obd_device *osc_obd;
718
719 LASSERT(tgt);
720 LASSERT(tgt->ltd_reap);
721
722 osc_obd = class_exp2obd(tgt->ltd_exp);
723
724 CDEBUG(D_CONFIG, "Removing tgt %s : %s\n",
725 tgt->ltd_uuid.uuid,
726 osc_obd ? osc_obd->obd_name : "<no obd>");
727
728 if (tgt->ltd_exp)
729 lov_disconnect_obd(obd, tgt);
730
731 OBD_FREE_PTR(tgt);
732
733 /* Manual cleanup - no cleanup logs to clean up the osc's. We must
734 do it ourselves. And we can't do it from lov_cleanup,
735 because we just lost our only reference to it. */
736 if (osc_obd)
737 class_manual_cleanup(osc_obd);
738}
739
740void lov_fix_desc_stripe_size(__u64 *val)
741{
742 if (*val < LOV_DEFAULT_STRIPE_SIZE) {
743 LCONSOLE_WARN("Increasing default stripe size to min %u\n",
744 LOV_DEFAULT_STRIPE_SIZE);
745 *val = LOV_DEFAULT_STRIPE_SIZE;
746 } else if (*val & (LOV_MIN_STRIPE_SIZE - 1)) {
747 *val &= ~(LOV_MIN_STRIPE_SIZE - 1);
748 LCONSOLE_WARN("Changing default stripe size to "LPU64" (a "
749 "multiple of %u)\n",
750 *val, LOV_MIN_STRIPE_SIZE);
751 }
752}
753
754void lov_fix_desc_stripe_count(__u32 *val)
755{
756 if (*val == 0)
757 *val = 1;
758}
759
760void lov_fix_desc_pattern(__u32 *val)
761{
762 /* from lov_setstripe */
763 if ((*val != 0) && (*val != LOV_PATTERN_RAID0)) {
764 LCONSOLE_WARN("Unknown stripe pattern: %#x\n", *val);
765 *val = 0;
766 }
767}
768
769void lov_fix_desc_qos_maxage(__u32 *val)
770{
771 /* fix qos_maxage */
772 if (*val == 0)
773 *val = QOS_DEFAULT_MAXAGE;
774}
775
776void lov_fix_desc(struct lov_desc *desc)
777{
778 lov_fix_desc_stripe_size(&desc->ld_default_stripe_size);
779 lov_fix_desc_stripe_count(&desc->ld_default_stripe_count);
780 lov_fix_desc_pattern(&desc->ld_pattern);
781 lov_fix_desc_qos_maxage(&desc->ld_qos_maxage);
782}
783
784int lov_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
785{
786 struct lprocfs_static_vars lvars = { 0 };
787 struct lov_desc *desc;
788 struct lov_obd *lov = &obd->u.lov;
789 int rc;
790 ENTRY;
791
792 if (LUSTRE_CFG_BUFLEN(lcfg, 1) < 1) {
793 CERROR("LOV setup requires a descriptor\n");
794 RETURN(-EINVAL);
795 }
796
797 desc = (struct lov_desc *)lustre_cfg_buf(lcfg, 1);
798
799 if (sizeof(*desc) > LUSTRE_CFG_BUFLEN(lcfg, 1)) {
800 CERROR("descriptor size wrong: %d > %d\n",
801 (int)sizeof(*desc), LUSTRE_CFG_BUFLEN(lcfg, 1));
802 RETURN(-EINVAL);
803 }
804
805 if (desc->ld_magic != LOV_DESC_MAGIC) {
806 if (desc->ld_magic == __swab32(LOV_DESC_MAGIC)) {
807 CDEBUG(D_OTHER, "%s: Swabbing lov desc %p\n",
808 obd->obd_name, desc);
809 lustre_swab_lov_desc(desc);
810 } else {
811 CERROR("%s: Bad lov desc magic: %#x\n",
812 obd->obd_name, desc->ld_magic);
813 RETURN(-EINVAL);
814 }
815 }
816
817 lov_fix_desc(desc);
818
819 desc->ld_active_tgt_count = 0;
820 lov->desc = *desc;
821 lov->lov_tgt_size = 0;
822
823 mutex_init(&lov->lov_lock);
824 atomic_set(&lov->lov_refcount, 0);
825 lov->lov_sp_me = LUSTRE_SP_CLI;
826
827 init_rwsem(&lov->lov_notify_lock);
828
829 lov->lov_pools_hash_body = cfs_hash_create("POOLS", HASH_POOLS_CUR_BITS,
830 HASH_POOLS_MAX_BITS,
831 HASH_POOLS_BKT_BITS, 0,
832 CFS_HASH_MIN_THETA,
833 CFS_HASH_MAX_THETA,
834 &pool_hash_operations,
835 CFS_HASH_DEFAULT);
836 INIT_LIST_HEAD(&lov->lov_pool_list);
837 lov->lov_pool_count = 0;
838 rc = lov_ost_pool_init(&lov->lov_packed, 0);
839 if (rc)
840 GOTO(out, rc);
841
842 lprocfs_lov_init_vars(&lvars);
843 lprocfs_obd_setup(obd, lvars.obd_vars);
844#ifdef LPROCFS
845 {
846 int rc;
847
848 rc = lprocfs_seq_create(obd->obd_proc_entry, "target_obd",
849 0444, &lov_proc_target_fops, obd);
850 if (rc)
851 CWARN("Error adding the target_obd file\n");
852 }
853#endif
854 lov->lov_pool_proc_entry = lprocfs_register("pools",
855 obd->obd_proc_entry,
856 NULL, NULL);
857
858 RETURN(0);
859
860out:
861 return rc;
862}
863
864static int lov_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
865{
866 int rc = 0;
867 struct lov_obd *lov = &obd->u.lov;
868
869 ENTRY;
870
871 switch (stage) {
872 case OBD_CLEANUP_EARLY: {
873 int i;
874 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
875 if (!lov->lov_tgts[i] || !lov->lov_tgts[i]->ltd_active)
876 continue;
877 obd_precleanup(class_exp2obd(lov->lov_tgts[i]->ltd_exp),
878 OBD_CLEANUP_EARLY);
879 }
880 break;
881 }
882 case OBD_CLEANUP_EXPORTS:
883 rc = obd_llog_finish(obd, 0);
884 if (rc != 0)
885 CERROR("failed to cleanup llogging subsystems\n");
886 break;
887 }
888 RETURN(rc);
889}
890
891static int lov_cleanup(struct obd_device *obd)
892{
893 struct lov_obd *lov = &obd->u.lov;
894 struct list_head *pos, *tmp;
895 struct pool_desc *pool;
896 ENTRY;
897
898 list_for_each_safe(pos, tmp, &lov->lov_pool_list) {
899 pool = list_entry(pos, struct pool_desc, pool_list);
900 /* free pool structs */
901 CDEBUG(D_INFO, "delete pool %p\n", pool);
902 /* In the function below, .hs_keycmp resolves to
903 * pool_hashkey_keycmp() */
904 /* coverity[overrun-buffer-val] */
905 lov_pool_del(obd, pool->pool_name);
906 }
907 cfs_hash_putref(lov->lov_pools_hash_body);
908 lov_ost_pool_free(&lov->lov_packed);
909
910 lprocfs_obd_cleanup(obd);
911 if (lov->lov_tgts) {
912 int i;
913 obd_getref(obd);
914 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
915 if (!lov->lov_tgts[i])
916 continue;
917
918 /* Inactive targets may never have connected */
919 if (lov->lov_tgts[i]->ltd_active ||
920 atomic_read(&lov->lov_refcount))
921 /* We should never get here - these
922 should have been removed in the
923 disconnect. */
924 CERROR("lov tgt %d not cleaned!"
925 " deathrow=%d, lovrc=%d\n",
926 i, lov->lov_death_row,
927 atomic_read(&lov->lov_refcount));
928 lov_del_target(obd, i, 0, 0);
929 }
930 obd_putref(obd);
931 OBD_FREE(lov->lov_tgts, sizeof(*lov->lov_tgts) *
932 lov->lov_tgt_size);
933 lov->lov_tgt_size = 0;
934 }
935 RETURN(0);
936}
937
938int lov_process_config_base(struct obd_device *obd, struct lustre_cfg *lcfg,
939 __u32 *indexp, int *genp)
940{
941 struct obd_uuid obd_uuid;
942 int cmd;
943 int rc = 0;
944 ENTRY;
945
946 switch(cmd = lcfg->lcfg_command) {
947 case LCFG_LOV_ADD_OBD:
948 case LCFG_LOV_ADD_INA:
949 case LCFG_LOV_DEL_OBD: {
950 __u32 index;
951 int gen;
952 /* lov_modify_tgts add 0:lov_mdsA 1:ost1_UUID 2:0 3:1 */
953 if (LUSTRE_CFG_BUFLEN(lcfg, 1) > sizeof(obd_uuid.uuid))
954 GOTO(out, rc = -EINVAL);
955
956 obd_str2uuid(&obd_uuid, lustre_cfg_buf(lcfg, 1));
957
958 if (sscanf(lustre_cfg_buf(lcfg, 2), "%d", indexp) != 1)
959 GOTO(out, rc = -EINVAL);
960 if (sscanf(lustre_cfg_buf(lcfg, 3), "%d", genp) != 1)
961 GOTO(out, rc = -EINVAL);
962 index = *indexp;
963 gen = *genp;
964 if (cmd == LCFG_LOV_ADD_OBD)
965 rc = lov_add_target(obd, &obd_uuid, index, gen, 1);
966 else if (cmd == LCFG_LOV_ADD_INA)
967 rc = lov_add_target(obd, &obd_uuid, index, gen, 0);
968 else
969 rc = lov_del_target(obd, index, &obd_uuid, gen);
970 GOTO(out, rc);
971 }
972 case LCFG_PARAM: {
973 struct lprocfs_static_vars lvars = { 0 };
974 struct lov_desc *desc = &(obd->u.lov.desc);
975
976 if (!desc)
977 GOTO(out, rc = -EINVAL);
978
979 lprocfs_lov_init_vars(&lvars);
980
981 rc = class_process_proc_param(PARAM_LOV, lvars.obd_vars,
982 lcfg, obd);
983 if (rc > 0)
984 rc = 0;
985 GOTO(out, rc);
986 }
987 case LCFG_POOL_NEW:
988 case LCFG_POOL_ADD:
989 case LCFG_POOL_DEL:
990 case LCFG_POOL_REM:
991 GOTO(out, rc);
992
993 default: {
994 CERROR("Unknown command: %d\n", lcfg->lcfg_command);
995 GOTO(out, rc = -EINVAL);
996
997 }
998 }
999out:
1000 RETURN(rc);
1001}
1002
1003static int lov_recreate(struct obd_export *exp, struct obdo *src_oa,
1004 struct lov_stripe_md **ea, struct obd_trans_info *oti)
1005{
1006 struct lov_stripe_md *obj_mdp, *lsm;
1007 struct lov_obd *lov = &exp->exp_obd->u.lov;
1008 unsigned ost_idx;
1009 int rc, i;
1010 ENTRY;
1011
1012 LASSERT(src_oa->o_valid & OBD_MD_FLFLAGS &&
1013 src_oa->o_flags & OBD_FL_RECREATE_OBJS);
1014
1015 OBD_ALLOC(obj_mdp, sizeof(*obj_mdp));
1016 if (obj_mdp == NULL)
1017 RETURN(-ENOMEM);
1018
1019 ost_idx = src_oa->o_nlink;
1020 lsm = *ea;
1021 if (lsm == NULL)
1022 GOTO(out, rc = -EINVAL);
1023 if (ost_idx >= lov->desc.ld_tgt_count ||
1024 !lov->lov_tgts[ost_idx])
1025 GOTO(out, rc = -EINVAL);
1026
1027 for (i = 0; i < lsm->lsm_stripe_count; i++) {
1028 if (lsm->lsm_oinfo[i]->loi_ost_idx == ost_idx) {
1029 if (ostid_id(&lsm->lsm_oinfo[i]->loi_oi) !=
1030 ostid_id(&src_oa->o_oi))
1031 GOTO(out, rc = -EINVAL);
1032 break;
1033 }
1034 }
1035 if (i == lsm->lsm_stripe_count)
1036 GOTO(out, rc = -EINVAL);
1037
1038 rc = obd_create(NULL, lov->lov_tgts[ost_idx]->ltd_exp,
1039 src_oa, &obj_mdp, oti);
1040out:
1041 OBD_FREE(obj_mdp, sizeof(*obj_mdp));
1042 RETURN(rc);
1043}
1044
1045/* the LOV expects oa->o_id to be set to the LOV object id */
1046static int lov_create(const struct lu_env *env, struct obd_export *exp,
1047 struct obdo *src_oa, struct lov_stripe_md **ea,
1048 struct obd_trans_info *oti)
1049{
1050 struct lov_obd *lov;
1051 int rc = 0;
1052 ENTRY;
1053
1054 LASSERT(ea != NULL);
1055 if (exp == NULL)
1056 RETURN(-EINVAL);
1057
1058 if ((src_oa->o_valid & OBD_MD_FLFLAGS) &&
1059 src_oa->o_flags == OBD_FL_DELORPHAN) {
1060 /* should be used with LOV anymore */
1061 LBUG();
1062 }
1063
1064 lov = &exp->exp_obd->u.lov;
1065 if (!lov->desc.ld_active_tgt_count)
1066 RETURN(-EIO);
1067
1068 obd_getref(exp->exp_obd);
1069 /* Recreate a specific object id at the given OST index */
1070 if ((src_oa->o_valid & OBD_MD_FLFLAGS) &&
1071 (src_oa->o_flags & OBD_FL_RECREATE_OBJS)) {
1072 rc = lov_recreate(exp, src_oa, ea, oti);
1073 }
1074
1075 obd_putref(exp->exp_obd);
1076 RETURN(rc);
1077}
1078
1079#define ASSERT_LSM_MAGIC(lsmp) \
1080do { \
1081 LASSERT((lsmp) != NULL); \
1082 LASSERTF(((lsmp)->lsm_magic == LOV_MAGIC_V1 || \
1083 (lsmp)->lsm_magic == LOV_MAGIC_V3), \
1084 "%p->lsm_magic=%x\n", (lsmp), (lsmp)->lsm_magic); \
1085} while (0)
1086
1087static int lov_destroy(const struct lu_env *env, struct obd_export *exp,
1088 struct obdo *oa, struct lov_stripe_md *lsm,
1089 struct obd_trans_info *oti, struct obd_export *md_exp,
1090 void *capa)
1091{
1092 struct lov_request_set *set;
1093 struct obd_info oinfo;
1094 struct lov_request *req;
1095 struct list_head *pos;
1096 struct lov_obd *lov;
1097 int rc = 0, err = 0;
1098 ENTRY;
1099
1100 ASSERT_LSM_MAGIC(lsm);
1101
1102 if (!exp || !exp->exp_obd)
1103 RETURN(-ENODEV);
1104
1105 if (oa->o_valid & OBD_MD_FLCOOKIE) {
1106 LASSERT(oti);
1107 LASSERT(oti->oti_logcookies);
1108 }
1109
1110 lov = &exp->exp_obd->u.lov;
1111 obd_getref(exp->exp_obd);
1112 rc = lov_prep_destroy_set(exp, &oinfo, oa, lsm, oti, &set);
1113 if (rc)
1114 GOTO(out, rc);
1115
1116 list_for_each (pos, &set->set_list) {
1117 req = list_entry(pos, struct lov_request, rq_link);
1118
1119 if (oa->o_valid & OBD_MD_FLCOOKIE)
1120 oti->oti_logcookies = set->set_cookies + req->rq_stripe;
1121
1122 err = obd_destroy(env, lov->lov_tgts[req->rq_idx]->ltd_exp,
1123 req->rq_oi.oi_oa, NULL, oti, NULL, capa);
1124 err = lov_update_common_set(set, req, err);
1125 if (err) {
1126 CERROR("%s: destroying objid "DOSTID" subobj "
1127 DOSTID" on OST idx %d: rc = %d\n",
1128 exp->exp_obd->obd_name, POSTID(&oa->o_oi),
1129 POSTID(&req->rq_oi.oi_oa->o_oi),
1130 req->rq_idx, err);
1131 if (!rc)
1132 rc = err;
1133 }
1134 }
1135
1136 if (rc == 0) {
1137 LASSERT(lsm_op_find(lsm->lsm_magic) != NULL);
1138 rc = lsm_op_find(lsm->lsm_magic)->lsm_destroy(lsm, oa, md_exp);
1139 }
1140 err = lov_fini_destroy_set(set);
1141out:
1142 obd_putref(exp->exp_obd);
1143 RETURN(rc ? rc : err);
1144}
1145
1146static int lov_getattr(const struct lu_env *env, struct obd_export *exp,
1147 struct obd_info *oinfo)
1148{
1149 struct lov_request_set *set;
1150 struct lov_request *req;
1151 struct list_head *pos;
1152 struct lov_obd *lov;
1153 int err = 0, rc = 0;
1154 ENTRY;
1155
1156 LASSERT(oinfo);
1157 ASSERT_LSM_MAGIC(oinfo->oi_md);
1158
1159 if (!exp || !exp->exp_obd)
1160 RETURN(-ENODEV);
1161
1162 lov = &exp->exp_obd->u.lov;
1163
1164 rc = lov_prep_getattr_set(exp, oinfo, &set);
1165 if (rc)
1166 RETURN(rc);
1167
1168 list_for_each (pos, &set->set_list) {
1169 req = list_entry(pos, struct lov_request, rq_link);
1170
1171 CDEBUG(D_INFO, "objid "DOSTID"[%d] has subobj "DOSTID" at idx"
1172 " %u\n", POSTID(&oinfo->oi_oa->o_oi), req->rq_stripe,
1173 POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx);
1174
1175 rc = obd_getattr(env, lov->lov_tgts[req->rq_idx]->ltd_exp,
1176 &req->rq_oi);
1177 err = lov_update_common_set(set, req, rc);
1178 if (err) {
1179 CERROR("%s: getattr objid "DOSTID" subobj "
1180 DOSTID" on OST idx %d: rc = %d\n",
1181 exp->exp_obd->obd_name,
1182 POSTID(&oinfo->oi_oa->o_oi),
1183 POSTID(&req->rq_oi.oi_oa->o_oi),
1184 req->rq_idx, err);
1185 break;
1186 }
1187 }
1188
1189 rc = lov_fini_getattr_set(set);
1190 if (err)
1191 rc = err;
1192 RETURN(rc);
1193}
1194
1195static int lov_getattr_interpret(struct ptlrpc_request_set *rqset,
1196 void *data, int rc)
1197{
1198 struct lov_request_set *lovset = (struct lov_request_set *)data;
1199 int err;
1200 ENTRY;
1201
1202 /* don't do attribute merge if this aysnc op failed */
1203 if (rc)
1204 atomic_set(&lovset->set_completes, 0);
1205 err = lov_fini_getattr_set(lovset);
1206 RETURN(rc ? rc : err);
1207}
1208
1209static int lov_getattr_async(struct obd_export *exp, struct obd_info *oinfo,
1210 struct ptlrpc_request_set *rqset)
1211{
1212 struct lov_request_set *lovset;
1213 struct lov_obd *lov;
1214 struct list_head *pos;
1215 struct lov_request *req;
1216 int rc = 0, err;
1217 ENTRY;
1218
1219 LASSERT(oinfo);
1220 ASSERT_LSM_MAGIC(oinfo->oi_md);
1221
1222 if (!exp || !exp->exp_obd)
1223 RETURN(-ENODEV);
1224
1225 lov = &exp->exp_obd->u.lov;
1226
1227 rc = lov_prep_getattr_set(exp, oinfo, &lovset);
1228 if (rc)
1229 RETURN(rc);
1230
1231 CDEBUG(D_INFO, "objid "DOSTID": %ux%u byte stripes\n",
1232 POSTID(&oinfo->oi_md->lsm_oi), oinfo->oi_md->lsm_stripe_count,
1233 oinfo->oi_md->lsm_stripe_size);
1234
1235 list_for_each(pos, &lovset->set_list) {
1236 req = list_entry(pos, struct lov_request, rq_link);
1237
1238 CDEBUG(D_INFO, "objid "DOSTID"[%d] has subobj "DOSTID" at idx"
1239 "%u\n", POSTID(&oinfo->oi_oa->o_oi), req->rq_stripe,
1240 POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx);
1241 rc = obd_getattr_async(lov->lov_tgts[req->rq_idx]->ltd_exp,
1242 &req->rq_oi, rqset);
1243 if (rc) {
1244 CERROR("%s: getattr objid "DOSTID" subobj"
1245 DOSTID" on OST idx %d: rc = %d\n",
1246 exp->exp_obd->obd_name,
1247 POSTID(&oinfo->oi_oa->o_oi),
1248 POSTID(&req->rq_oi.oi_oa->o_oi),
1249 req->rq_idx, rc);
1250 GOTO(out, rc);
1251 }
1252 }
1253
1254 if (!list_empty(&rqset->set_requests)) {
1255 LASSERT(rc == 0);
1256 LASSERT (rqset->set_interpret == NULL);
1257 rqset->set_interpret = lov_getattr_interpret;
1258 rqset->set_arg = (void *)lovset;
1259 RETURN(rc);
1260 }
1261out:
1262 if (rc)
1263 atomic_set(&lovset->set_completes, 0);
1264 err = lov_fini_getattr_set(lovset);
1265 RETURN(rc ? rc : err);
1266}
1267
1268static int lov_setattr(const struct lu_env *env, struct obd_export *exp,
1269 struct obd_info *oinfo, struct obd_trans_info *oti)
1270{
1271 struct lov_request_set *set;
1272 struct lov_obd *lov;
1273 struct list_head *pos;
1274 struct lov_request *req;
1275 int err = 0, rc = 0;
1276 ENTRY;
1277
1278 LASSERT(oinfo);
1279 ASSERT_LSM_MAGIC(oinfo->oi_md);
1280
1281 if (!exp || !exp->exp_obd)
1282 RETURN(-ENODEV);
1283
1284 /* for now, we only expect the following updates here */
1285 LASSERT(!(oinfo->oi_oa->o_valid & ~(OBD_MD_FLID | OBD_MD_FLTYPE |
1286 OBD_MD_FLMODE | OBD_MD_FLATIME |
1287 OBD_MD_FLMTIME | OBD_MD_FLCTIME |
1288 OBD_MD_FLFLAGS | OBD_MD_FLSIZE |
1289 OBD_MD_FLGROUP | OBD_MD_FLUID |
1290 OBD_MD_FLGID | OBD_MD_FLFID |
1291 OBD_MD_FLGENER)));
1292 lov = &exp->exp_obd->u.lov;
1293 rc = lov_prep_setattr_set(exp, oinfo, oti, &set);
1294 if (rc)
1295 RETURN(rc);
1296
1297 list_for_each (pos, &set->set_list) {
1298 req = list_entry(pos, struct lov_request, rq_link);
1299
1300 rc = obd_setattr(env, lov->lov_tgts[req->rq_idx]->ltd_exp,
1301 &req->rq_oi, NULL);
1302 err = lov_update_setattr_set(set, req, rc);
1303 if (err) {
1304 CERROR("%s: setattr objid "DOSTID" subobj "
1305 DOSTID" on OST idx %d: rc = %d\n",
1306 exp->exp_obd->obd_name,
1307 POSTID(&set->set_oi->oi_oa->o_oi),
1308 POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx,
1309 err);
1310 if (!rc)
1311 rc = err;
1312 }
1313 }
1314 err = lov_fini_setattr_set(set);
1315 if (!rc)
1316 rc = err;
1317 RETURN(rc);
1318}
1319
1320static int lov_setattr_interpret(struct ptlrpc_request_set *rqset,
1321 void *data, int rc)
1322{
1323 struct lov_request_set *lovset = (struct lov_request_set *)data;
1324 int err;
1325 ENTRY;
1326
1327 if (rc)
1328 atomic_set(&lovset->set_completes, 0);
1329 err = lov_fini_setattr_set(lovset);
1330 RETURN(rc ? rc : err);
1331}
1332
1333/* If @oti is given, the request goes from MDS and responses from OSTs are not
1334 needed. Otherwise, a client is waiting for responses. */
1335static int lov_setattr_async(struct obd_export *exp, struct obd_info *oinfo,
1336 struct obd_trans_info *oti,
1337 struct ptlrpc_request_set *rqset)
1338{
1339 struct lov_request_set *set;
1340 struct lov_request *req;
1341 struct list_head *pos;
1342 struct lov_obd *lov;
1343 int rc = 0;
1344 ENTRY;
1345
1346 LASSERT(oinfo);
1347 ASSERT_LSM_MAGIC(oinfo->oi_md);
1348 if (oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE) {
1349 LASSERT(oti);
1350 LASSERT(oti->oti_logcookies);
1351 }
1352
1353 if (!exp || !exp->exp_obd)
1354 RETURN(-ENODEV);
1355
1356 lov = &exp->exp_obd->u.lov;
1357 rc = lov_prep_setattr_set(exp, oinfo, oti, &set);
1358 if (rc)
1359 RETURN(rc);
1360
1361 CDEBUG(D_INFO, "objid "DOSTID": %ux%u byte stripes\n",
1362 POSTID(&oinfo->oi_md->lsm_oi),
1363 oinfo->oi_md->lsm_stripe_count,
1364 oinfo->oi_md->lsm_stripe_size);
1365
1366 list_for_each(pos, &set->set_list) {
1367 req = list_entry(pos, struct lov_request, rq_link);
1368
1369 if (oinfo->oi_oa->o_valid & OBD_MD_FLCOOKIE)
1370 oti->oti_logcookies = set->set_cookies + req->rq_stripe;
1371
1372 CDEBUG(D_INFO, "objid "DOSTID"[%d] has subobj "DOSTID" at idx"
1373 "%u\n", POSTID(&oinfo->oi_oa->o_oi), req->rq_stripe,
1374 POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx);
1375
1376 rc = obd_setattr_async(lov->lov_tgts[req->rq_idx]->ltd_exp,
1377 &req->rq_oi, oti, rqset);
1378 if (rc) {
1379 CERROR("error: setattr objid "DOSTID" subobj"
1380 DOSTID" on OST idx %d: rc = %d\n",
1381 POSTID(&set->set_oi->oi_oa->o_oi),
1382 POSTID(&req->rq_oi.oi_oa->o_oi),
1383 req->rq_idx, rc);
1384 break;
1385 }
1386 }
1387
1388 /* If we are not waiting for responses on async requests, return. */
1389 if (rc || !rqset || list_empty(&rqset->set_requests)) {
1390 int err;
1391 if (rc)
1392 atomic_set(&set->set_completes, 0);
1393 err = lov_fini_setattr_set(set);
1394 RETURN(rc ? rc : err);
1395 }
1396
1397 LASSERT(rqset->set_interpret == NULL);
1398 rqset->set_interpret = lov_setattr_interpret;
1399 rqset->set_arg = (void *)set;
1400
1401 RETURN(0);
1402}
1403
1404static int lov_punch_interpret(struct ptlrpc_request_set *rqset,
1405 void *data, int rc)
1406{
1407 struct lov_request_set *lovset = (struct lov_request_set *)data;
1408 int err;
1409 ENTRY;
1410
1411 if (rc)
1412 atomic_set(&lovset->set_completes, 0);
1413 err = lov_fini_punch_set(lovset);
1414 RETURN(rc ? rc : err);
1415}
1416
1417/* FIXME: maybe we'll just make one node the authoritative attribute node, then
1418 * we can send this 'punch' to just the authoritative node and the nodes
1419 * that the punch will affect. */
1420static int lov_punch(const struct lu_env *env, struct obd_export *exp,
1421 struct obd_info *oinfo, struct obd_trans_info *oti,
1422 struct ptlrpc_request_set *rqset)
1423{
1424 struct lov_request_set *set;
1425 struct lov_obd *lov;
1426 struct list_head *pos;
1427 struct lov_request *req;
1428 int rc = 0;
1429 ENTRY;
1430
1431 LASSERT(oinfo);
1432 ASSERT_LSM_MAGIC(oinfo->oi_md);
1433
1434 if (!exp || !exp->exp_obd)
1435 RETURN(-ENODEV);
1436
1437 lov = &exp->exp_obd->u.lov;
1438 rc = lov_prep_punch_set(exp, oinfo, oti, &set);
1439 if (rc)
1440 RETURN(rc);
1441
1442 list_for_each (pos, &set->set_list) {
1443 req = list_entry(pos, struct lov_request, rq_link);
1444
1445 rc = obd_punch(env, lov->lov_tgts[req->rq_idx]->ltd_exp,
1446 &req->rq_oi, NULL, rqset);
1447 if (rc) {
1448 CERROR("%s: punch objid "DOSTID" subobj "DOSTID
1449 " on OST idx %d: rc = %d\n",
1450 exp->exp_obd->obd_name,
1451 POSTID(&set->set_oi->oi_oa->o_oi),
1452 POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx, rc);
1453 break;
1454 }
1455 }
1456
1457 if (rc || list_empty(&rqset->set_requests)) {
1458 int err;
1459 err = lov_fini_punch_set(set);
1460 RETURN(rc ? rc : err);
1461 }
1462
1463 LASSERT(rqset->set_interpret == NULL);
1464 rqset->set_interpret = lov_punch_interpret;
1465 rqset->set_arg = (void *)set;
1466
1467 RETURN(0);
1468}
1469
1470static int lov_sync_interpret(struct ptlrpc_request_set *rqset,
1471 void *data, int rc)
1472{
1473 struct lov_request_set *lovset = data;
1474 int err;
1475 ENTRY;
1476
1477 if (rc)
1478 atomic_set(&lovset->set_completes, 0);
1479 err = lov_fini_sync_set(lovset);
1480 RETURN(rc ?: err);
1481}
1482
1483static int lov_sync(const struct lu_env *env, struct obd_export *exp,
1484 struct obd_info *oinfo, obd_off start, obd_off end,
1485 struct ptlrpc_request_set *rqset)
1486{
1487 struct lov_request_set *set = NULL;
1488 struct lov_obd *lov;
1489 struct list_head *pos;
1490 struct lov_request *req;
1491 int rc = 0;
1492 ENTRY;
1493
1494 ASSERT_LSM_MAGIC(oinfo->oi_md);
1495 LASSERT(rqset != NULL);
1496
1497 if (!exp->exp_obd)
1498 RETURN(-ENODEV);
1499
1500 lov = &exp->exp_obd->u.lov;
1501 rc = lov_prep_sync_set(exp, oinfo, start, end, &set);
1502 if (rc)
1503 RETURN(rc);
1504
1505 CDEBUG(D_INFO, "fsync objid "DOSTID" ["LPX64", "LPX64"]\n",
1506 POSTID(&set->set_oi->oi_oa->o_oi), start, end);
1507
1508 list_for_each (pos, &set->set_list) {
1509 req = list_entry(pos, struct lov_request, rq_link);
1510
1511 rc = obd_sync(env, lov->lov_tgts[req->rq_idx]->ltd_exp,
1512 &req->rq_oi, req->rq_oi.oi_policy.l_extent.start,
1513 req->rq_oi.oi_policy.l_extent.end, rqset);
1514 if (rc) {
1515 CERROR("%s: fsync objid "DOSTID" subobj "DOSTID
1516 " on OST idx %d: rc = %d\n",
1517 exp->exp_obd->obd_name,
1518 POSTID(&set->set_oi->oi_oa->o_oi),
1519 POSTID(&req->rq_oi.oi_oa->o_oi), req->rq_idx,
1520 rc);
1521 break;
1522 }
1523 }
1524
1525 /* If we are not waiting for responses on async requests, return. */
1526 if (rc || list_empty(&rqset->set_requests)) {
1527 int err = lov_fini_sync_set(set);
1528
1529 RETURN(rc ?: err);
1530 }
1531
1532 LASSERT(rqset->set_interpret == NULL);
1533 rqset->set_interpret = lov_sync_interpret;
1534 rqset->set_arg = (void *)set;
1535
1536 RETURN(0);
1537}
1538
1539static int lov_brw_check(struct lov_obd *lov, struct obd_info *lov_oinfo,
1540 obd_count oa_bufs, struct brw_page *pga)
1541{
1542 struct obd_info oinfo = { { { 0 } } };
1543 int i, rc = 0;
1544
1545 oinfo.oi_oa = lov_oinfo->oi_oa;
1546
1547 /* The caller just wants to know if there's a chance that this
1548 * I/O can succeed */
1549 for (i = 0; i < oa_bufs; i++) {
1550 int stripe = lov_stripe_number(lov_oinfo->oi_md, pga[i].off);
1551 int ost = lov_oinfo->oi_md->lsm_oinfo[stripe]->loi_ost_idx;
1552 obd_off start, end;
1553
1554 if (!lov_stripe_intersects(lov_oinfo->oi_md, i, pga[i].off,
1555 pga[i].off + pga[i].count - 1,
1556 &start, &end))
1557 continue;
1558
1559 if (!lov->lov_tgts[ost] || !lov->lov_tgts[ost]->ltd_active) {
1560 CDEBUG(D_HA, "lov idx %d inactive\n", ost);
1561 return -EIO;
1562 }
1563
1564 rc = obd_brw(OBD_BRW_CHECK, lov->lov_tgts[ost]->ltd_exp, &oinfo,
1565 1, &pga[i], NULL);
1566 if (rc)
1567 break;
1568 }
1569 return rc;
1570}
1571
1572static int lov_brw(int cmd, struct obd_export *exp, struct obd_info *oinfo,
1573 obd_count oa_bufs, struct brw_page *pga,
1574 struct obd_trans_info *oti)
1575{
1576 struct lov_request_set *set;
1577 struct lov_request *req;
1578 struct list_head *pos;
1579 struct lov_obd *lov = &exp->exp_obd->u.lov;
1580 int err, rc = 0;
1581 ENTRY;
1582
1583 ASSERT_LSM_MAGIC(oinfo->oi_md);
1584
1585 if (cmd == OBD_BRW_CHECK) {
1586 rc = lov_brw_check(lov, oinfo, oa_bufs, pga);
1587 RETURN(rc);
1588 }
1589
1590 rc = lov_prep_brw_set(exp, oinfo, oa_bufs, pga, oti, &set);
1591 if (rc)
1592 RETURN(rc);
1593
1594 list_for_each (pos, &set->set_list) {
1595 struct obd_export *sub_exp;
1596 struct brw_page *sub_pga;
1597 req = list_entry(pos, struct lov_request, rq_link);
1598
1599 sub_exp = lov->lov_tgts[req->rq_idx]->ltd_exp;
1600 sub_pga = set->set_pga + req->rq_pgaidx;
1601 rc = obd_brw(cmd, sub_exp, &req->rq_oi, req->rq_oabufs,
1602 sub_pga, oti);
1603 if (rc)
1604 break;
1605 lov_update_common_set(set, req, rc);
1606 }
1607
1608 err = lov_fini_brw_set(set);
1609 if (!rc)
1610 rc = err;
1611 RETURN(rc);
1612}
1613
1614static int lov_enqueue_interpret(struct ptlrpc_request_set *rqset,
1615 void *data, int rc)
1616{
1617 struct lov_request_set *lovset = (struct lov_request_set *)data;
1618 ENTRY;
1619 rc = lov_fini_enqueue_set(lovset, lovset->set_ei->ei_mode, rc, rqset);
1620 RETURN(rc);
1621}
1622
1623static int lov_enqueue(struct obd_export *exp, struct obd_info *oinfo,
1624 struct ldlm_enqueue_info *einfo,
1625 struct ptlrpc_request_set *rqset)
1626{
1627 ldlm_mode_t mode = einfo->ei_mode;
1628 struct lov_request_set *set;
1629 struct lov_request *req;
1630 struct list_head *pos;
1631 struct lov_obd *lov;
1632 ldlm_error_t rc;
1633 ENTRY;
1634
1635 LASSERT(oinfo);
1636 ASSERT_LSM_MAGIC(oinfo->oi_md);
1637 LASSERT(mode == (mode & -mode));
1638
1639 /* we should never be asked to replay a lock this way. */
1640 LASSERT((oinfo->oi_flags & LDLM_FL_REPLAY) == 0);
1641
1642 if (!exp || !exp->exp_obd)
1643 RETURN(-ENODEV);
1644
1645 lov = &exp->exp_obd->u.lov;
1646 rc = lov_prep_enqueue_set(exp, oinfo, einfo, &set);
1647 if (rc)
1648 RETURN(rc);
1649
1650 list_for_each (pos, &set->set_list) {
1651 req = list_entry(pos, struct lov_request, rq_link);
1652
1653 rc = obd_enqueue(lov->lov_tgts[req->rq_idx]->ltd_exp,
1654 &req->rq_oi, einfo, rqset);
1655 if (rc != ELDLM_OK)
1656 GOTO(out, rc);
1657 }
1658
1659 if (rqset && !list_empty(&rqset->set_requests)) {
1660 LASSERT(rc == 0);
1661 LASSERT(rqset->set_interpret == NULL);
1662 rqset->set_interpret = lov_enqueue_interpret;
1663 rqset->set_arg = (void *)set;
1664 RETURN(rc);
1665 }
1666out:
1667 rc = lov_fini_enqueue_set(set, mode, rc, rqset);
1668 RETURN(rc);
1669}
1670
1671static int lov_change_cbdata(struct obd_export *exp,
1672 struct lov_stripe_md *lsm, ldlm_iterator_t it,
1673 void *data)
1674{
1675 struct lov_obd *lov;
1676 int rc = 0, i;
1677 ENTRY;
1678
1679 ASSERT_LSM_MAGIC(lsm);
1680
1681 if (!exp || !exp->exp_obd)
1682 RETURN(-ENODEV);
1683
1684 lov = &exp->exp_obd->u.lov;
1685 for (i = 0; i < lsm->lsm_stripe_count; i++) {
1686 struct lov_stripe_md submd;
1687 struct lov_oinfo *loi = lsm->lsm_oinfo[i];
1688
1689 if (!lov->lov_tgts[loi->loi_ost_idx]) {
1690 CDEBUG(D_HA, "lov idx %d NULL \n", loi->loi_ost_idx);
1691 continue;
1692 }
1693
1694 submd.lsm_oi = loi->loi_oi;
1695 submd.lsm_stripe_count = 0;
1696 rc = obd_change_cbdata(lov->lov_tgts[loi->loi_ost_idx]->ltd_exp,
1697 &submd, it, data);
1698 }
1699 RETURN(rc);
1700}
1701
1702/* find any ldlm lock of the inode in lov
1703 * return 0 not find
1704 * 1 find one
1705 * < 0 error */
1706static int lov_find_cbdata(struct obd_export *exp,
1707 struct lov_stripe_md *lsm, ldlm_iterator_t it,
1708 void *data)
1709{
1710 struct lov_obd *lov;
1711 int rc = 0, i;
1712 ENTRY;
1713
1714 ASSERT_LSM_MAGIC(lsm);
1715
1716 if (!exp || !exp->exp_obd)
1717 RETURN(-ENODEV);
1718
1719 lov = &exp->exp_obd->u.lov;
1720 for (i = 0; i < lsm->lsm_stripe_count; i++) {
1721 struct lov_stripe_md submd;
1722 struct lov_oinfo *loi = lsm->lsm_oinfo[i];
1723
1724 if (!lov->lov_tgts[loi->loi_ost_idx]) {
1725 CDEBUG(D_HA, "lov idx %d NULL \n", loi->loi_ost_idx);
1726 continue;
1727 }
1728 submd.lsm_oi = loi->loi_oi;
1729 submd.lsm_stripe_count = 0;
1730 rc = obd_find_cbdata(lov->lov_tgts[loi->loi_ost_idx]->ltd_exp,
1731 &submd, it, data);
1732 if (rc != 0)
1733 RETURN(rc);
1734 }
1735 RETURN(rc);
1736}
1737
1738static int lov_cancel(struct obd_export *exp, struct lov_stripe_md *lsm,
1739 __u32 mode, struct lustre_handle *lockh)
1740{
1741 struct lov_request_set *set;
1742 struct obd_info oinfo;
1743 struct lov_request *req;
1744 struct list_head *pos;
1745 struct lov_obd *lov;
1746 struct lustre_handle *lov_lockhp;
1747 int err = 0, rc = 0;
1748 ENTRY;
1749
1750 ASSERT_LSM_MAGIC(lsm);
1751
1752 if (!exp || !exp->exp_obd)
1753 RETURN(-ENODEV);
1754
1755 LASSERT(lockh);
1756 lov = &exp->exp_obd->u.lov;
1757 rc = lov_prep_cancel_set(exp, &oinfo, lsm, mode, lockh, &set);
1758 if (rc)
1759 RETURN(rc);
1760
1761 list_for_each(pos, &set->set_list) {
1762 req = list_entry(pos, struct lov_request, rq_link);
1763 lov_lockhp = set->set_lockh->llh_handles + req->rq_stripe;
1764
1765 rc = obd_cancel(lov->lov_tgts[req->rq_idx]->ltd_exp,
1766 req->rq_oi.oi_md, mode, lov_lockhp);
1767 rc = lov_update_common_set(set, req, rc);
1768 if (rc) {
1769 CERROR("%s: cancel objid "DOSTID" subobj "
1770 DOSTID" on OST idx %d: rc = %d\n",
1771 exp->exp_obd->obd_name, POSTID(&lsm->lsm_oi),
1772 POSTID(&req->rq_oi.oi_md->lsm_oi),
1773 req->rq_idx, rc);
1774 err = rc;
1775 }
1776
1777 }
1778 lov_fini_cancel_set(set);
1779 RETURN(err);
1780}
1781
1782static int lov_cancel_unused(struct obd_export *exp,
1783 struct lov_stripe_md *lsm,
1784 ldlm_cancel_flags_t flags, void *opaque)
1785{
1786 struct lov_obd *lov;
1787 int rc = 0, i;
1788 ENTRY;
1789
1790 if (!exp || !exp->exp_obd)
1791 RETURN(-ENODEV);
1792
1793 lov = &exp->exp_obd->u.lov;
1794 if (lsm == NULL) {
1795 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
1796 int err;
1797 if (!lov->lov_tgts[i] || !lov->lov_tgts[i]->ltd_exp)
1798 continue;
1799
1800 err = obd_cancel_unused(lov->lov_tgts[i]->ltd_exp, NULL,
1801 flags, opaque);
1802 if (!rc)
1803 rc = err;
1804 }
1805 RETURN(rc);
1806 }
1807
1808 ASSERT_LSM_MAGIC(lsm);
1809
1810 for (i = 0; i < lsm->lsm_stripe_count; i++) {
1811 struct lov_stripe_md submd;
1812 struct lov_oinfo *loi = lsm->lsm_oinfo[i];
1813 int idx = loi->loi_ost_idx;
1814 int err;
1815
1816 if (!lov->lov_tgts[idx]) {
1817 CDEBUG(D_HA, "lov idx %d NULL\n", idx);
1818 continue;
1819 }
1820
1821 if (!lov->lov_tgts[idx]->ltd_active)
1822 CDEBUG(D_HA, "lov idx %d inactive\n", idx);
1823
1824 submd.lsm_oi = loi->loi_oi;
1825 submd.lsm_stripe_count = 0;
1826 err = obd_cancel_unused(lov->lov_tgts[idx]->ltd_exp,
1827 &submd, flags, opaque);
1828 if (err && lov->lov_tgts[idx]->ltd_active) {
1829 CERROR("%s: cancel unused objid "DOSTID
1830 " subobj "DOSTID" on OST idx %d: rc = %d\n",
1831 exp->exp_obd->obd_name, POSTID(&lsm->lsm_oi),
1832 POSTID(&loi->loi_oi), idx, err);
1833 if (!rc)
1834 rc = err;
1835 }
1836 }
1837 RETURN(rc);
1838}
1839
1840int lov_statfs_interpret(struct ptlrpc_request_set *rqset, void *data, int rc)
1841{
1842 struct lov_request_set *lovset = (struct lov_request_set *)data;
1843 int err;
1844 ENTRY;
1845
1846 if (rc)
1847 atomic_set(&lovset->set_completes, 0);
1848
1849 err = lov_fini_statfs_set(lovset);
1850 RETURN(rc ? rc : err);
1851}
1852
1853static int lov_statfs_async(struct obd_export *exp, struct obd_info *oinfo,
1854 __u64 max_age, struct ptlrpc_request_set *rqset)
1855{
1856 struct obd_device *obd = class_exp2obd(exp);
1857 struct lov_request_set *set;
1858 struct lov_request *req;
1859 struct list_head *pos;
1860 struct lov_obd *lov;
1861 int rc = 0;
1862 ENTRY;
1863
1864 LASSERT(oinfo != NULL);
1865 LASSERT(oinfo->oi_osfs != NULL);
1866
1867 lov = &obd->u.lov;
1868 rc = lov_prep_statfs_set(obd, oinfo, &set);
1869 if (rc)
1870 RETURN(rc);
1871
1872 list_for_each (pos, &set->set_list) {
1873 req = list_entry(pos, struct lov_request, rq_link);
1874 rc = obd_statfs_async(lov->lov_tgts[req->rq_idx]->ltd_exp,
1875 &req->rq_oi, max_age, rqset);
1876 if (rc)
1877 break;
1878 }
1879
1880 if (rc || list_empty(&rqset->set_requests)) {
1881 int err;
1882 if (rc)
1883 atomic_set(&set->set_completes, 0);
1884 err = lov_fini_statfs_set(set);
1885 RETURN(rc ? rc : err);
1886 }
1887
1888 LASSERT(rqset->set_interpret == NULL);
1889 rqset->set_interpret = lov_statfs_interpret;
1890 rqset->set_arg = (void *)set;
1891 RETURN(0);
1892}
1893
1894static int lov_statfs(const struct lu_env *env, struct obd_export *exp,
1895 struct obd_statfs *osfs, __u64 max_age, __u32 flags)
1896{
1897 struct ptlrpc_request_set *set = NULL;
1898 struct obd_info oinfo = { { { 0 } } };
1899 int rc = 0;
1900 ENTRY;
1901
1902
1903 /* for obdclass we forbid using obd_statfs_rqset, but prefer using async
1904 * statfs requests */
1905 set = ptlrpc_prep_set();
1906 if (set == NULL)
1907 RETURN(-ENOMEM);
1908
1909 oinfo.oi_osfs = osfs;
1910 oinfo.oi_flags = flags;
1911 rc = lov_statfs_async(exp, &oinfo, max_age, set);
1912 if (rc == 0)
1913 rc = ptlrpc_set_wait(set);
1914 ptlrpc_set_destroy(set);
1915
1916 RETURN(rc);
1917}
1918
1919static int lov_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
1920 void *karg, void *uarg)
1921{
1922 struct obd_device *obddev = class_exp2obd(exp);
1923 struct lov_obd *lov = &obddev->u.lov;
1924 int i = 0, rc = 0, count = lov->desc.ld_tgt_count;
1925 struct obd_uuid *uuidp;
1926 ENTRY;
1927
1928 switch (cmd) {
1929 case IOC_OBD_STATFS: {
1930 struct obd_ioctl_data *data = karg;
1931 struct obd_device *osc_obd;
1932 struct obd_statfs stat_buf = {0};
1933 __u32 index;
1934 __u32 flags;
1935
1936 memcpy(&index, data->ioc_inlbuf2, sizeof(__u32));
1937 if ((index >= count))
1938 RETURN(-ENODEV);
1939
1940 if (!lov->lov_tgts[index])
1941 /* Try again with the next index */
1942 RETURN(-EAGAIN);
1943 if (!lov->lov_tgts[index]->ltd_active)
1944 RETURN(-ENODATA);
1945
1946 osc_obd = class_exp2obd(lov->lov_tgts[index]->ltd_exp);
1947 if (!osc_obd)
1948 RETURN(-EINVAL);
1949
1950 /* copy UUID */
1951 if (copy_to_user(data->ioc_pbuf2, obd2cli_tgt(osc_obd),
1952 min((int) data->ioc_plen2,
1953 (int) sizeof(struct obd_uuid))))
1954 RETURN(-EFAULT);
1955
1956 flags = uarg ? *(__u32*)uarg : 0;
1957 /* got statfs data */
1958 rc = obd_statfs(NULL, lov->lov_tgts[index]->ltd_exp, &stat_buf,
1959 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
1960 flags);
1961 if (rc)
1962 RETURN(rc);
1963 if (copy_to_user(data->ioc_pbuf1, &stat_buf,
1964 min((int) data->ioc_plen1,
1965 (int) sizeof(stat_buf))))
1966 RETURN(-EFAULT);
1967 break;
1968 }
1969 case OBD_IOC_LOV_GET_CONFIG: {
1970 struct obd_ioctl_data *data;
1971 struct lov_desc *desc;
1972 char *buf = NULL;
1973 __u32 *genp;
1974
1975 len = 0;
1976 if (obd_ioctl_getdata(&buf, &len, (void *)uarg))
1977 RETURN(-EINVAL);
1978
1979 data = (struct obd_ioctl_data *)buf;
1980
1981 if (sizeof(*desc) > data->ioc_inllen1) {
1982 obd_ioctl_freedata(buf, len);
1983 RETURN(-EINVAL);
1984 }
1985
1986 if (sizeof(uuidp->uuid) * count > data->ioc_inllen2) {
1987 obd_ioctl_freedata(buf, len);
1988 RETURN(-EINVAL);
1989 }
1990
1991 if (sizeof(__u32) * count > data->ioc_inllen3) {
1992 obd_ioctl_freedata(buf, len);
1993 RETURN(-EINVAL);
1994 }
1995
1996 desc = (struct lov_desc *)data->ioc_inlbuf1;
1997 memcpy(desc, &(lov->desc), sizeof(*desc));
1998
1999 uuidp = (struct obd_uuid *)data->ioc_inlbuf2;
2000 genp = (__u32 *)data->ioc_inlbuf3;
2001 /* the uuid will be empty for deleted OSTs */
2002 for (i = 0; i < count; i++, uuidp++, genp++) {
2003 if (!lov->lov_tgts[i])
2004 continue;
2005 *uuidp = lov->lov_tgts[i]->ltd_uuid;
2006 *genp = lov->lov_tgts[i]->ltd_gen;
2007 }
2008
2009 if (copy_to_user((void *)uarg, buf, len))
2010 rc = -EFAULT;
2011 obd_ioctl_freedata(buf, len);
2012 break;
2013 }
2014 case LL_IOC_LOV_SETSTRIPE:
2015 rc = lov_setstripe(exp, len, karg, uarg);
2016 break;
2017 case LL_IOC_LOV_GETSTRIPE:
2018 rc = lov_getstripe(exp, karg, uarg);
2019 break;
2020 case LL_IOC_LOV_SETEA:
2021 rc = lov_setea(exp, karg, uarg);
2022 break;
2023 case OBD_IOC_QUOTACTL: {
2024 struct if_quotactl *qctl = karg;
2025 struct lov_tgt_desc *tgt = NULL;
2026 struct obd_quotactl *oqctl;
2027
2028 if (qctl->qc_valid == QC_OSTIDX) {
2029 if (qctl->qc_idx < 0 || count <= qctl->qc_idx)
2030 RETURN(-EINVAL);
2031
2032 tgt = lov->lov_tgts[qctl->qc_idx];
2033 if (!tgt || !tgt->ltd_exp)
2034 RETURN(-EINVAL);
2035 } else if (qctl->qc_valid == QC_UUID) {
2036 for (i = 0; i < count; i++) {
2037 tgt = lov->lov_tgts[i];
2038 if (!tgt ||
2039 !obd_uuid_equals(&tgt->ltd_uuid,
2040 &qctl->obd_uuid))
2041 continue;
2042
2043 if (tgt->ltd_exp == NULL)
2044 RETURN(-EINVAL);
2045
2046 break;
2047 }
2048 } else {
2049 RETURN(-EINVAL);
2050 }
2051
2052 if (i >= count)
2053 RETURN(-EAGAIN);
2054
2055 LASSERT(tgt && tgt->ltd_exp);
2056 OBD_ALLOC_PTR(oqctl);
2057 if (!oqctl)
2058 RETURN(-ENOMEM);
2059
2060 QCTL_COPY(oqctl, qctl);
2061 rc = obd_quotactl(tgt->ltd_exp, oqctl);
2062 if (rc == 0) {
2063 QCTL_COPY(qctl, oqctl);
2064 qctl->qc_valid = QC_OSTIDX;
2065 qctl->obd_uuid = tgt->ltd_uuid;
2066 }
2067 OBD_FREE_PTR(oqctl);
2068 break;
2069 }
2070 default: {
2071 int set = 0;
2072
2073 if (count == 0)
2074 RETURN(-ENOTTY);
2075
2076 for (i = 0; i < count; i++) {
2077 int err;
2078 struct obd_device *osc_obd;
2079
2080 /* OST was disconnected */
2081 if (!lov->lov_tgts[i] || !lov->lov_tgts[i]->ltd_exp)
2082 continue;
2083
2084 /* ll_umount_begin() sets force flag but for lov, not
2085 * osc. Let's pass it through */
2086 osc_obd = class_exp2obd(lov->lov_tgts[i]->ltd_exp);
2087 osc_obd->obd_force = obddev->obd_force;
2088 err = obd_iocontrol(cmd, lov->lov_tgts[i]->ltd_exp,
2089 len, karg, uarg);
2090 if (err == -ENODATA && cmd == OBD_IOC_POLL_QUOTACHECK) {
2091 RETURN(err);
2092 } else if (err) {
2093 if (lov->lov_tgts[i]->ltd_active) {
2094 CDEBUG(err == -ENOTTY ?
2095 D_IOCTL : D_WARNING,
2096 "iocontrol OSC %s on OST "
2097 "idx %d cmd %x: err = %d\n",
2098 lov_uuid2str(lov, i),
2099 i, cmd, err);
2100 if (!rc)
2101 rc = err;
2102 }
2103 } else {
2104 set = 1;
2105 }
2106 }
2107 if (!set && !rc)
2108 rc = -EIO;
2109 }
2110 }
2111
2112 RETURN(rc);
2113}
2114
2115#define FIEMAP_BUFFER_SIZE 4096
2116
2117/**
2118 * Non-zero fe_logical indicates that this is a continuation FIEMAP
2119 * call. The local end offset and the device are sent in the first
2120 * fm_extent. This function calculates the stripe number from the index.
2121 * This function returns a stripe_no on which mapping is to be restarted.
2122 *
2123 * This function returns fm_end_offset which is the in-OST offset at which
2124 * mapping should be restarted. If fm_end_offset=0 is returned then caller
2125 * will re-calculate proper offset in next stripe.
2126 * Note that the first extent is passed to lov_get_info via the value field.
2127 *
2128 * \param fiemap fiemap request header
2129 * \param lsm striping information for the file
2130 * \param fm_start logical start of mapping
2131 * \param fm_end logical end of mapping
2132 * \param start_stripe starting stripe will be returned in this
2133 */
2134obd_size fiemap_calc_fm_end_offset(struct ll_user_fiemap *fiemap,
2135 struct lov_stripe_md *lsm, obd_size fm_start,
2136 obd_size fm_end, int *start_stripe)
2137{
2138 obd_size local_end = fiemap->fm_extents[0].fe_logical;
2139 obd_off lun_start, lun_end;
2140 obd_size fm_end_offset;
2141 int stripe_no = -1, i;
2142
2143 if (fiemap->fm_extent_count == 0 ||
2144 fiemap->fm_extents[0].fe_logical == 0)
2145 return 0;
2146
2147 /* Find out stripe_no from ost_index saved in the fe_device */
2148 for (i = 0; i < lsm->lsm_stripe_count; i++) {
2149 if (lsm->lsm_oinfo[i]->loi_ost_idx ==
2150 fiemap->fm_extents[0].fe_device) {
2151 stripe_no = i;
2152 break;
2153 }
2154 }
2155 if (stripe_no == -1)
2156 return -EINVAL;
2157
2158 /* If we have finished mapping on previous device, shift logical
2159 * offset to start of next device */
2160 if ((lov_stripe_intersects(lsm, stripe_no, fm_start, fm_end,
2161 &lun_start, &lun_end)) != 0 &&
2162 local_end < lun_end) {
2163 fm_end_offset = local_end;
2164 *start_stripe = stripe_no;
2165 } else {
2166 /* This is a special value to indicate that caller should
2167 * calculate offset in next stripe. */
2168 fm_end_offset = 0;
2169 *start_stripe = (stripe_no + 1) % lsm->lsm_stripe_count;
2170 }
2171
2172 return fm_end_offset;
2173}
2174
2175/**
2176 * We calculate on which OST the mapping will end. If the length of mapping
2177 * is greater than (stripe_size * stripe_count) then the last_stripe will
2178 * will be one just before start_stripe. Else we check if the mapping
2179 * intersects each OST and find last_stripe.
2180 * This function returns the last_stripe and also sets the stripe_count
2181 * over which the mapping is spread
2182 *
2183 * \param lsm striping information for the file
2184 * \param fm_start logical start of mapping
2185 * \param fm_end logical end of mapping
2186 * \param start_stripe starting stripe of the mapping
2187 * \param stripe_count the number of stripes across which to map is returned
2188 *
2189 * \retval last_stripe return the last stripe of the mapping
2190 */
2191int fiemap_calc_last_stripe(struct lov_stripe_md *lsm, obd_size fm_start,
2192 obd_size fm_end, int start_stripe,
2193 int *stripe_count)
2194{
2195 int last_stripe;
2196 obd_off obd_start, obd_end;
2197 int i, j;
2198
2199 if (fm_end - fm_start > lsm->lsm_stripe_size * lsm->lsm_stripe_count) {
2200 last_stripe = (start_stripe < 1 ? lsm->lsm_stripe_count - 1 :
2201 start_stripe - 1);
2202 *stripe_count = lsm->lsm_stripe_count;
2203 } else {
2204 for (j = 0, i = start_stripe; j < lsm->lsm_stripe_count;
2205 i = (i + 1) % lsm->lsm_stripe_count, j++) {
2206 if ((lov_stripe_intersects(lsm, i, fm_start, fm_end,
2207 &obd_start, &obd_end)) == 0)
2208 break;
2209 }
2210 *stripe_count = j;
2211 last_stripe = (start_stripe + j - 1) %lsm->lsm_stripe_count;
2212 }
2213
2214 return last_stripe;
2215}
2216
2217/**
2218 * Set fe_device and copy extents from local buffer into main return buffer.
2219 *
2220 * \param fiemap fiemap request header
2221 * \param lcl_fm_ext array of local fiemap extents to be copied
2222 * \param ost_index OST index to be written into the fm_device field for each
2223 extent
2224 * \param ext_count number of extents to be copied
2225 * \param current_extent where to start copying in main extent array
2226 */
2227void fiemap_prepare_and_copy_exts(struct ll_user_fiemap *fiemap,
2228 struct ll_fiemap_extent *lcl_fm_ext,
2229 int ost_index, unsigned int ext_count,
2230 int current_extent)
2231{
2232 char *to;
2233 int ext;
2234
2235 for (ext = 0; ext < ext_count; ext++) {
2236 lcl_fm_ext[ext].fe_device = ost_index;
2237 lcl_fm_ext[ext].fe_flags |= FIEMAP_EXTENT_NET;
2238 }
2239
2240 /* Copy fm_extent's from fm_local to return buffer */
2241 to = (char *)fiemap + fiemap_count_to_size(current_extent);
2242 memcpy(to, lcl_fm_ext, ext_count * sizeof(struct ll_fiemap_extent));
2243}
2244
2245/**
2246 * Break down the FIEMAP request and send appropriate calls to individual OSTs.
2247 * This also handles the restarting of FIEMAP calls in case mapping overflows
2248 * the available number of extents in single call.
2249 */
2250static int lov_fiemap(struct lov_obd *lov, __u32 keylen, void *key,
2251 __u32 *vallen, void *val, struct lov_stripe_md *lsm)
2252{
2253 struct ll_fiemap_info_key *fm_key = key;
2254 struct ll_user_fiemap *fiemap = val;
2255 struct ll_user_fiemap *fm_local = NULL;
2256 struct ll_fiemap_extent *lcl_fm_ext;
2257 int count_local;
2258 unsigned int get_num_extents = 0;
2259 int ost_index = 0, actual_start_stripe, start_stripe;
2260 obd_size fm_start, fm_end, fm_length, fm_end_offset;
2261 obd_size curr_loc;
2262 int current_extent = 0, rc = 0, i;
2263 int ost_eof = 0; /* EOF for object */
2264 int ost_done = 0; /* done with required mapping for this OST? */
2265 int last_stripe;
2266 int cur_stripe = 0, cur_stripe_wrap = 0, stripe_count;
2267 unsigned int buffer_size = FIEMAP_BUFFER_SIZE;
2268
2269 if (lsm == NULL)
2270 GOTO(out, rc = 0);
2271
2272 if (fiemap_count_to_size(fm_key->fiemap.fm_extent_count) < buffer_size)
2273 buffer_size = fiemap_count_to_size(fm_key->fiemap.fm_extent_count);
2274
2275 OBD_ALLOC_LARGE(fm_local, buffer_size);
2276 if (fm_local == NULL)
2277 GOTO(out, rc = -ENOMEM);
2278 lcl_fm_ext = &fm_local->fm_extents[0];
2279
2280 count_local = fiemap_size_to_count(buffer_size);
2281
2282 memcpy(fiemap, &fm_key->fiemap, sizeof(*fiemap));
2283 fm_start = fiemap->fm_start;
2284 fm_length = fiemap->fm_length;
2285 /* Calculate start stripe, last stripe and length of mapping */
2286 actual_start_stripe = start_stripe = lov_stripe_number(lsm, fm_start);
2287 fm_end = (fm_length == ~0ULL ? fm_key->oa.o_size :
2288 fm_start + fm_length - 1);
2289 /* If fm_length != ~0ULL but fm_start+fm_length-1 exceeds file size */
2290 if (fm_end > fm_key->oa.o_size)
2291 fm_end = fm_key->oa.o_size;
2292
2293 last_stripe = fiemap_calc_last_stripe(lsm, fm_start, fm_end,
2294 actual_start_stripe, &stripe_count);
2295
2296 fm_end_offset = fiemap_calc_fm_end_offset(fiemap, lsm, fm_start,
2297 fm_end, &start_stripe);
2298 if (fm_end_offset == -EINVAL)
2299 GOTO(out, rc = -EINVAL);
2300
2301 if (fiemap->fm_extent_count == 0) {
2302 get_num_extents = 1;
2303 count_local = 0;
2304 }
2305
2306 /* Check each stripe */
2307 for (cur_stripe = start_stripe, i = 0; i < stripe_count;
2308 i++, cur_stripe = (cur_stripe + 1) % lsm->lsm_stripe_count) {
2309 obd_size req_fm_len; /* Stores length of required mapping */
2310 obd_size len_mapped_single_call;
2311 obd_off lun_start, lun_end, obd_object_end;
2312 unsigned int ext_count;
2313
2314 cur_stripe_wrap = cur_stripe;
2315
2316 /* Find out range of mapping on this stripe */
2317 if ((lov_stripe_intersects(lsm, cur_stripe, fm_start, fm_end,
2318 &lun_start, &obd_object_end)) == 0)
2319 continue;
2320
2321 /* If this is a continuation FIEMAP call and we are on
2322 * starting stripe then lun_start needs to be set to
2323 * fm_end_offset */
2324 if (fm_end_offset != 0 && cur_stripe == start_stripe)
2325 lun_start = fm_end_offset;
2326
2327 if (fm_length != ~0ULL) {
2328 /* Handle fm_start + fm_length overflow */
2329 if (fm_start + fm_length < fm_start)
2330 fm_length = ~0ULL - fm_start;
2331 lun_end = lov_size_to_stripe(lsm, fm_start + fm_length,
2332 cur_stripe);
2333 } else {
2334 lun_end = ~0ULL;
2335 }
2336
2337 if (lun_start == lun_end)
2338 continue;
2339
2340 req_fm_len = obd_object_end - lun_start;
2341 fm_local->fm_length = 0;
2342 len_mapped_single_call = 0;
2343
2344 /* If the output buffer is very large and the objects have many
2345 * extents we may need to loop on a single OST repeatedly */
2346 ost_eof = 0;
2347 ost_done = 0;
2348 do {
2349 if (get_num_extents == 0) {
2350 /* Don't get too many extents. */
2351 if (current_extent + count_local >
2352 fiemap->fm_extent_count)
2353 count_local = fiemap->fm_extent_count -
2354 current_extent;
2355 }
2356
2357 lun_start += len_mapped_single_call;
2358 fm_local->fm_length = req_fm_len - len_mapped_single_call;
2359 req_fm_len = fm_local->fm_length;
2360 fm_local->fm_extent_count = count_local;
2361 fm_local->fm_mapped_extents = 0;
2362 fm_local->fm_flags = fiemap->fm_flags;
2363
2364 fm_key->oa.o_oi = lsm->lsm_oinfo[cur_stripe]->loi_oi;
2365 ost_index = lsm->lsm_oinfo[cur_stripe]->loi_ost_idx;
2366
2367 if (ost_index < 0 || ost_index >=lov->desc.ld_tgt_count)
2368 GOTO(out, rc = -EINVAL);
2369
2370 /* If OST is inactive, return extent with UNKNOWN flag */
2371 if (!lov->lov_tgts[ost_index]->ltd_active) {
2372 fm_local->fm_flags |= FIEMAP_EXTENT_LAST;
2373 fm_local->fm_mapped_extents = 1;
2374
2375 lcl_fm_ext[0].fe_logical = lun_start;
2376 lcl_fm_ext[0].fe_length = obd_object_end -
2377 lun_start;
2378 lcl_fm_ext[0].fe_flags |= FIEMAP_EXTENT_UNKNOWN;
2379
2380 goto inactive_tgt;
2381 }
2382
2383 fm_local->fm_start = lun_start;
2384 fm_local->fm_flags &= ~FIEMAP_FLAG_DEVICE_ORDER;
2385 memcpy(&fm_key->fiemap, fm_local, sizeof(*fm_local));
2386 *vallen=fiemap_count_to_size(fm_local->fm_extent_count);
2387 rc = obd_get_info(NULL,
2388 lov->lov_tgts[ost_index]->ltd_exp,
2389 keylen, key, vallen, fm_local, lsm);
2390 if (rc != 0)
2391 GOTO(out, rc);
2392
2393inactive_tgt:
2394 ext_count = fm_local->fm_mapped_extents;
2395 if (ext_count == 0) {
2396 ost_done = 1;
2397 /* If last stripe has hole at the end,
2398 * then we need to return */
2399 if (cur_stripe_wrap == last_stripe) {
2400 fiemap->fm_mapped_extents = 0;
2401 goto finish;
2402 }
2403 break;
2404 }
2405
2406 /* If we just need num of extents then go to next device */
2407 if (get_num_extents) {
2408 current_extent += ext_count;
2409 break;
2410 }
2411
2412 len_mapped_single_call = lcl_fm_ext[ext_count-1].fe_logical -
2413 lun_start + lcl_fm_ext[ext_count - 1].fe_length;
2414
2415 /* Have we finished mapping on this device? */
2416 if (req_fm_len <= len_mapped_single_call)
2417 ost_done = 1;
2418
2419 /* Clear the EXTENT_LAST flag which can be present on
2420 * last extent */
2421 if (lcl_fm_ext[ext_count-1].fe_flags & FIEMAP_EXTENT_LAST)
2422 lcl_fm_ext[ext_count - 1].fe_flags &=
2423 ~FIEMAP_EXTENT_LAST;
2424
2425 curr_loc = lov_stripe_size(lsm,
2426 lcl_fm_ext[ext_count - 1].fe_logical+
2427 lcl_fm_ext[ext_count - 1].fe_length,
2428 cur_stripe);
2429 if (curr_loc >= fm_key->oa.o_size)
2430 ost_eof = 1;
2431
2432 fiemap_prepare_and_copy_exts(fiemap, lcl_fm_ext,
2433 ost_index, ext_count,
2434 current_extent);
2435
2436 current_extent += ext_count;
2437
2438 /* Ran out of available extents? */
2439 if (current_extent >= fiemap->fm_extent_count)
2440 goto finish;
2441 } while (ost_done == 0 && ost_eof == 0);
2442
2443 if (cur_stripe_wrap == last_stripe)
2444 goto finish;
2445 }
2446
2447finish:
2448 /* Indicate that we are returning device offsets unless file just has
2449 * single stripe */
2450 if (lsm->lsm_stripe_count > 1)
2451 fiemap->fm_flags |= FIEMAP_FLAG_DEVICE_ORDER;
2452
2453 if (get_num_extents)
2454 goto skip_last_device_calc;
2455
2456 /* Check if we have reached the last stripe and whether mapping for that
2457 * stripe is done. */
2458 if (cur_stripe_wrap == last_stripe) {
2459 if (ost_done || ost_eof)
2460 fiemap->fm_extents[current_extent - 1].fe_flags |=
2461 FIEMAP_EXTENT_LAST;
2462 }
2463
2464skip_last_device_calc:
2465 fiemap->fm_mapped_extents = current_extent;
2466
2467out:
2468 OBD_FREE_LARGE(fm_local, buffer_size);
2469 return rc;
2470}
2471
2472static int lov_get_info(const struct lu_env *env, struct obd_export *exp,
2473 __u32 keylen, void *key, __u32 *vallen, void *val,
2474 struct lov_stripe_md *lsm)
2475{
2476 struct obd_device *obddev = class_exp2obd(exp);
2477 struct lov_obd *lov = &obddev->u.lov;
2478 int i, rc;
2479 ENTRY;
2480
2481 if (!vallen || !val)
2482 RETURN(-EFAULT);
2483
2484 obd_getref(obddev);
2485
2486 if (KEY_IS(KEY_LOCK_TO_STRIPE)) {
2487 struct {
2488 char name[16];
2489 struct ldlm_lock *lock;
2490 } *data = key;
2491 struct ldlm_res_id *res_id = &data->lock->l_resource->lr_name;
2492 struct lov_oinfo *loi;
2493 __u32 *stripe = val;
2494
2495 if (*vallen < sizeof(*stripe))
2496 GOTO(out, rc = -EFAULT);
2497 *vallen = sizeof(*stripe);
2498
2499 /* XXX This is another one of those bits that will need to
2500 * change if we ever actually support nested LOVs. It uses
2501 * the lock's export to find out which stripe it is. */
2502 /* XXX - it's assumed all the locks for deleted OSTs have
2503 * been cancelled. Also, the export for deleted OSTs will
2504 * be NULL and won't match the lock's export. */
2505 for (i = 0; i < lsm->lsm_stripe_count; i++) {
2506 loi = lsm->lsm_oinfo[i];
2507 if (!lov->lov_tgts[loi->loi_ost_idx])
2508 continue;
2509 if (lov->lov_tgts[loi->loi_ost_idx]->ltd_exp ==
2510 data->lock->l_conn_export &&
2511 ostid_res_name_eq(&loi->loi_oi, res_id)) {
2512 *stripe = i;
2513 GOTO(out, rc = 0);
2514 }
2515 }
2516 LDLM_ERROR(data->lock, "lock on inode without such object");
2517 dump_lsm(D_ERROR, lsm);
2518 GOTO(out, rc = -ENXIO);
2519 } else if (KEY_IS(KEY_LAST_ID)) {
2520 struct obd_id_info *info = val;
2521 __u32 size = sizeof(obd_id);
2522 struct lov_tgt_desc *tgt;
2523
2524 LASSERT(*vallen == sizeof(struct obd_id_info));
2525 tgt = lov->lov_tgts[info->idx];
2526
2527 if (!tgt || !tgt->ltd_active)
2528 GOTO(out, rc = -ESRCH);
2529
2530 rc = obd_get_info(env, tgt->ltd_exp, keylen, key,
2531 &size, info->data, NULL);
2532 GOTO(out, rc = 0);
2533 } else if (KEY_IS(KEY_LOVDESC)) {
2534 struct lov_desc *desc_ret = val;
2535 *desc_ret = lov->desc;
2536
2537 GOTO(out, rc = 0);
2538 } else if (KEY_IS(KEY_FIEMAP)) {
2539 rc = lov_fiemap(lov, keylen, key, vallen, val, lsm);
2540 GOTO(out, rc);
2541 } else if (KEY_IS(KEY_CONNECT_FLAG)) {
2542 struct lov_tgt_desc *tgt;
2543 __u64 ost_idx = *((__u64*)val);
2544
2545 LASSERT(*vallen == sizeof(__u64));
2546 LASSERT(ost_idx < lov->desc.ld_tgt_count);
2547 tgt = lov->lov_tgts[ost_idx];
2548
2549 if (!tgt || !tgt->ltd_exp)
2550 GOTO(out, rc = -ESRCH);
2551
2552 *((__u64 *)val) = exp_connect_flags(tgt->ltd_exp);
2553 GOTO(out, rc = 0);
2554 } else if (KEY_IS(KEY_TGT_COUNT)) {
2555 *((int *)val) = lov->desc.ld_tgt_count;
2556 GOTO(out, rc = 0);
2557 }
2558
2559 rc = -EINVAL;
2560
2561out:
2562 obd_putref(obddev);
2563 RETURN(rc);
2564}
2565
2566static int lov_set_info_async(const struct lu_env *env, struct obd_export *exp,
2567 obd_count keylen, void *key, obd_count vallen,
2568 void *val, struct ptlrpc_request_set *set)
2569{
2570 struct obd_device *obddev = class_exp2obd(exp);
2571 struct lov_obd *lov = &obddev->u.lov;
2572 obd_count count;
2573 int i, rc = 0, err;
2574 struct lov_tgt_desc *tgt;
2575 unsigned incr, check_uuid,
2576 do_inactive, no_set;
2577 unsigned next_id = 0, mds_con = 0, capa = 0;
2578 ENTRY;
2579
2580 incr = check_uuid = do_inactive = no_set = 0;
2581 if (set == NULL) {
2582 no_set = 1;
2583 set = ptlrpc_prep_set();
2584 if (!set)
2585 RETURN(-ENOMEM);
2586 }
2587
2588 obd_getref(obddev);
2589 count = lov->desc.ld_tgt_count;
2590
2591 if (KEY_IS(KEY_NEXT_ID)) {
2592 count = vallen / sizeof(struct obd_id_info);
2593 vallen = sizeof(obd_id);
2594 incr = sizeof(struct obd_id_info);
2595 do_inactive = 1;
2596 next_id = 1;
2597 } else if (KEY_IS(KEY_CHECKSUM)) {
2598 do_inactive = 1;
2599 } else if (KEY_IS(KEY_EVICT_BY_NID)) {
2600 /* use defaults: do_inactive = incr = 0; */
2601 } else if (KEY_IS(KEY_MDS_CONN)) {
2602 mds_con = 1;
2603 } else if (KEY_IS(KEY_CAPA_KEY)) {
2604 capa = 1;
2605 } else if (KEY_IS(KEY_CACHE_SET)) {
2606 LASSERT(lov->lov_cache == NULL);
2607 lov->lov_cache = val;
2608 do_inactive = 1;
2609 }
2610
2611 for (i = 0; i < count; i++, val = (char *)val + incr) {
2612 if (next_id) {
2613 tgt = lov->lov_tgts[((struct obd_id_info*)val)->idx];
2614 } else {
2615 tgt = lov->lov_tgts[i];
2616 }
2617 /* OST was disconnected */
2618 if (!tgt || !tgt->ltd_exp)
2619 continue;
2620
2621 /* OST is inactive and we don't want inactive OSCs */
2622 if (!tgt->ltd_active && !do_inactive)
2623 continue;
2624
2625 if (mds_con) {
2626 struct mds_group_info *mgi;
2627
2628 LASSERT(vallen == sizeof(*mgi));
2629 mgi = (struct mds_group_info *)val;
2630
2631 /* Only want a specific OSC */
2632 if (mgi->uuid && !obd_uuid_equals(mgi->uuid,
2633 &tgt->ltd_uuid))
2634 continue;
2635
2636 err = obd_set_info_async(env, tgt->ltd_exp,
2637 keylen, key, sizeof(int),
2638 &mgi->group, set);
2639 } else if (next_id) {
2640 err = obd_set_info_async(env, tgt->ltd_exp,
2641 keylen, key, vallen,
2642 ((struct obd_id_info*)val)->data, set);
2643 } else if (capa) {
2644 struct mds_capa_info *info = (struct mds_capa_info*)val;
2645
2646 LASSERT(vallen == sizeof(*info));
2647
2648 /* Only want a specific OSC */
2649 if (info->uuid &&
2650 !obd_uuid_equals(info->uuid, &tgt->ltd_uuid))
2651 continue;
2652
2653 err = obd_set_info_async(env, tgt->ltd_exp, keylen,
2654 key, sizeof(*info->capa),
2655 info->capa, set);
2656 } else {
2657 /* Only want a specific OSC */
2658 if (check_uuid &&
2659 !obd_uuid_equals(val, &tgt->ltd_uuid))
2660 continue;
2661
2662 err = obd_set_info_async(env, tgt->ltd_exp,
2663 keylen, key, vallen, val, set);
2664 }
2665
2666 if (!rc)
2667 rc = err;
2668 }
2669
2670 obd_putref(obddev);
2671 if (no_set) {
2672 err = ptlrpc_set_wait(set);
2673 if (!rc)
2674 rc = err;
2675 ptlrpc_set_destroy(set);
2676 }
2677 RETURN(rc);
2678}
2679
2680static int lov_extent_calc(struct obd_export *exp, struct lov_stripe_md *lsm,
2681 int cmd, __u64 *offset)
2682{
2683 __u32 ssize = lsm->lsm_stripe_size;
2684 __u64 start;
2685
2686 start = *offset;
2687 lov_do_div64(start, ssize);
2688 start = start * ssize;
2689
2690 CDEBUG(D_DLMTRACE, "offset "LPU64", stripe %u, start "LPU64
2691 ", end "LPU64"\n", *offset, ssize, start,
2692 start + ssize - 1);
2693 if (cmd == OBD_CALC_STRIPE_END) {
2694 *offset = start + ssize - 1;
2695 } else if (cmd == OBD_CALC_STRIPE_START) {
2696 *offset = start;
2697 } else {
2698 LBUG();
2699 }
2700
2701 RETURN(0);
2702}
2703
2704void lov_stripe_lock(struct lov_stripe_md *md)
2705{
2706 LASSERT(md->lsm_lock_owner != current_pid());
2707 spin_lock(&md->lsm_lock);
2708 LASSERT(md->lsm_lock_owner == 0);
2709 md->lsm_lock_owner = current_pid();
2710}
2711EXPORT_SYMBOL(lov_stripe_lock);
2712
2713void lov_stripe_unlock(struct lov_stripe_md *md)
2714{
2715 LASSERT(md->lsm_lock_owner == current_pid());
2716 md->lsm_lock_owner = 0;
2717 spin_unlock(&md->lsm_lock);
2718}
2719EXPORT_SYMBOL(lov_stripe_unlock);
2720
2721static int lov_quotactl(struct obd_device *obd, struct obd_export *exp,
2722 struct obd_quotactl *oqctl)
2723{
2724 struct lov_obd *lov = &obd->u.lov;
2725 struct lov_tgt_desc *tgt;
2726 __u64 curspace = 0;
2727 __u64 bhardlimit = 0;
2728 int i, rc = 0;
2729 ENTRY;
2730
2731 if (oqctl->qc_cmd != LUSTRE_Q_QUOTAON &&
2732 oqctl->qc_cmd != LUSTRE_Q_QUOTAOFF &&
2733 oqctl->qc_cmd != Q_GETOQUOTA &&
2734 oqctl->qc_cmd != Q_INITQUOTA &&
2735 oqctl->qc_cmd != LUSTRE_Q_SETQUOTA &&
2736 oqctl->qc_cmd != Q_FINVALIDATE) {
2737 CERROR("bad quota opc %x for lov obd", oqctl->qc_cmd);
2738 RETURN(-EFAULT);
2739 }
2740
2741 /* for lov tgt */
2742 obd_getref(obd);
2743 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
2744 int err;
2745
2746 tgt = lov->lov_tgts[i];
2747
2748 if (!tgt)
2749 continue;
2750
2751 if (!tgt->ltd_active || tgt->ltd_reap) {
2752 if (oqctl->qc_cmd == Q_GETOQUOTA &&
2753 lov->lov_tgts[i]->ltd_activate) {
2754 rc = -EREMOTEIO;
2755 CERROR("ost %d is inactive\n", i);
2756 } else {
2757 CDEBUG(D_HA, "ost %d is inactive\n", i);
2758 }
2759 continue;
2760 }
2761
2762 err = obd_quotactl(tgt->ltd_exp, oqctl);
2763 if (err) {
2764 if (tgt->ltd_active && !rc)
2765 rc = err;
2766 continue;
2767 }
2768
2769 if (oqctl->qc_cmd == Q_GETOQUOTA) {
2770 curspace += oqctl->qc_dqblk.dqb_curspace;
2771 bhardlimit += oqctl->qc_dqblk.dqb_bhardlimit;
2772 }
2773 }
2774 obd_putref(obd);
2775
2776 if (oqctl->qc_cmd == Q_GETOQUOTA) {
2777 oqctl->qc_dqblk.dqb_curspace = curspace;
2778 oqctl->qc_dqblk.dqb_bhardlimit = bhardlimit;
2779 }
2780 RETURN(rc);
2781}
2782
2783static int lov_quotacheck(struct obd_device *obd, struct obd_export *exp,
2784 struct obd_quotactl *oqctl)
2785{
2786 struct lov_obd *lov = &obd->u.lov;
2787 int i, rc = 0;
2788 ENTRY;
2789
2790 obd_getref(obd);
2791
2792 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
2793 if (!lov->lov_tgts[i])
2794 continue;
2795
2796 /* Skip quota check on the administratively disabled OSTs. */
2797 if (!lov->lov_tgts[i]->ltd_activate) {
2798 CWARN("lov idx %d was administratively disabled, "
2799 "skip quotacheck on it.\n", i);
2800 continue;
2801 }
2802
2803 if (!lov->lov_tgts[i]->ltd_active) {
2804 CERROR("lov idx %d inactive\n", i);
2805 rc = -EIO;
2806 goto out;
2807 }
2808 }
2809
2810 for (i = 0; i < lov->desc.ld_tgt_count; i++) {
2811 int err;
2812
2813 if (!lov->lov_tgts[i] || !lov->lov_tgts[i]->ltd_activate)
2814 continue;
2815
2816 err = obd_quotacheck(lov->lov_tgts[i]->ltd_exp, oqctl);
2817 if (err && !rc)
2818 rc = err;
2819 }
2820
2821out:
2822 obd_putref(obd);
2823
2824 RETURN(rc);
2825}
2826
2827struct obd_ops lov_obd_ops = {
2828 .o_owner = THIS_MODULE,
2829 .o_setup = lov_setup,
2830 .o_precleanup = lov_precleanup,
2831 .o_cleanup = lov_cleanup,
2832 //.o_process_config = lov_process_config,
2833 .o_connect = lov_connect,
2834 .o_disconnect = lov_disconnect,
2835 .o_statfs = lov_statfs,
2836 .o_statfs_async = lov_statfs_async,
2837 .o_packmd = lov_packmd,
2838 .o_unpackmd = lov_unpackmd,
2839 .o_create = lov_create,
2840 .o_destroy = lov_destroy,
2841 .o_getattr = lov_getattr,
2842 .o_getattr_async = lov_getattr_async,
2843 .o_setattr = lov_setattr,
2844 .o_setattr_async = lov_setattr_async,
2845 .o_brw = lov_brw,
2846 .o_merge_lvb = lov_merge_lvb,
2847 .o_adjust_kms = lov_adjust_kms,
2848 .o_punch = lov_punch,
2849 .o_sync = lov_sync,
2850 .o_enqueue = lov_enqueue,
2851 .o_change_cbdata = lov_change_cbdata,
2852 .o_find_cbdata = lov_find_cbdata,
2853 .o_cancel = lov_cancel,
2854 .o_cancel_unused = lov_cancel_unused,
2855 .o_iocontrol = lov_iocontrol,
2856 .o_get_info = lov_get_info,
2857 .o_set_info_async = lov_set_info_async,
2858 .o_extent_calc = lov_extent_calc,
2859 .o_llog_init = lov_llog_init,
2860 .o_llog_finish = lov_llog_finish,
2861 .o_notify = lov_notify,
2862 .o_pool_new = lov_pool_new,
2863 .o_pool_rem = lov_pool_remove,
2864 .o_pool_add = lov_pool_add,
2865 .o_pool_del = lov_pool_del,
2866 .o_getref = lov_getref,
2867 .o_putref = lov_putref,
2868 .o_quotactl = lov_quotactl,
2869 .o_quotacheck = lov_quotacheck,
2870};
2871
2872struct kmem_cache *lov_oinfo_slab;
2873
2874extern struct lu_kmem_descr lov_caches[];
2875
2876int __init lov_init(void)
2877{
2878 struct lprocfs_static_vars lvars = { 0 };
2879 int rc;
2880 ENTRY;
2881
2882 /* print an address of _any_ initialized kernel symbol from this
2883 * module, to allow debugging with gdb that doesn't support data
2884 * symbols from modules.*/
2885 CDEBUG(D_INFO, "Lustre LOV module (%p).\n", &lov_caches);
2886
2887 rc = lu_kmem_init(lov_caches);
2888 if (rc)
2889 return rc;
2890
2891 lov_oinfo_slab = kmem_cache_create("lov_oinfo",
2892 sizeof(struct lov_oinfo),
2893 0, SLAB_HWCACHE_ALIGN, NULL);
2894 if (lov_oinfo_slab == NULL) {
2895 lu_kmem_fini(lov_caches);
2896 return -ENOMEM;
2897 }
2898 lprocfs_lov_init_vars(&lvars);
2899
2900 rc = class_register_type(&lov_obd_ops, NULL, lvars.module_vars,
2901 LUSTRE_LOV_NAME, &lov_device_type);
2902
2903 if (rc) {
2904 kmem_cache_destroy(lov_oinfo_slab);
2905 lu_kmem_fini(lov_caches);
2906 }
2907
2908 RETURN(rc);
2909}
2910
2911static void /*__exit*/ lov_exit(void)
2912{
2913 class_unregister_type(LUSTRE_LOV_NAME);
2914 kmem_cache_destroy(lov_oinfo_slab);
2915
2916 lu_kmem_fini(lov_caches);
2917}
2918
2919MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
2920MODULE_DESCRIPTION("Lustre Logical Object Volume OBD driver");
2921MODULE_LICENSE("GPL");
2922
2923cfs_module(lov, LUSTRE_VERSION_STRING, lov_init, lov_exit);
Linux kernel - Deliverables
This page took 0.142197 seconds and 5 git commands to generate.