projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
staging: lustre: remove RETURN macro
[deliverable/linux.git] / drivers / staging / lustre / lustre / lmv / lmv_obd.c
1 /*
2 * GPL HEADER START
3 *
4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
9 *
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.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) 2004, 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
37 #define DEBUG_SUBSYSTEM S_LMV
38 #include <linux/slab.h>
39 #include <linux/module.h>
40 #include <linux/init.h>
41 #include <linux/pagemap.h>
42 #include <linux/mm.h>
43 #include <asm/div64.h>
44 #include <linux/seq_file.h>
45 #include <linux/namei.h>
46 #include <asm/uaccess.h>
47
48 #include <lustre/lustre_idl.h>
49 #include <obd_support.h>
50 #include <lustre_lib.h>
51 #include <lustre_net.h>
52 #include <obd_class.h>
53 #include <lprocfs_status.h>
54 #include <lustre_lite.h>
55 #include <lustre_fid.h>
56 #include "lmv_internal.h"
57
58 static void lmv_activate_target(struct lmv_obd *lmv,
59 struct lmv_tgt_desc *tgt,
60 int activate)
61 {
62 if (tgt->ltd_active == activate)
63 return;
64
65 tgt->ltd_active = activate;
66 lmv->desc.ld_active_tgt_count += (activate ? 1 : -1);
67 }
68
69 /**
70 * Error codes:
71 *
72 * -EINVAL : UUID can't be found in the LMV's target list
73 * -ENOTCONN: The UUID is found, but the target connection is bad (!)
74 * -EBADF : The UUID is found, but the OBD of the wrong type (!)
75 */
76 static int lmv_set_mdc_active(struct lmv_obd *lmv, struct obd_uuid *uuid,
77 int activate)
78 {
79 struct lmv_tgt_desc *uninitialized_var(tgt);
80 struct obd_device *obd;
81 int i;
82 int rc = 0;
83
84 CDEBUG(D_INFO, "Searching in lmv %p for uuid %s (activate=%d)\n",
85 lmv, uuid->uuid, activate);
86
87 spin_lock(&lmv->lmv_lock);
88 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
89 tgt = lmv->tgts[i];
90 if (tgt == NULL || tgt->ltd_exp == NULL)
91 continue;
92
93 CDEBUG(D_INFO, "Target idx %d is %s conn "LPX64"\n", i,
94 tgt->ltd_uuid.uuid, tgt->ltd_exp->exp_handle.h_cookie);
95
96 if (obd_uuid_equals(uuid, &tgt->ltd_uuid))
97 break;
98 }
99
100 if (i == lmv->desc.ld_tgt_count)
101 GOTO(out_lmv_lock, rc = -EINVAL);
102
103 obd = class_exp2obd(tgt->ltd_exp);
104 if (obd == NULL)
105 GOTO(out_lmv_lock, rc = -ENOTCONN);
106
107 CDEBUG(D_INFO, "Found OBD %s=%s device %d (%p) type %s at LMV idx %d\n",
108 obd->obd_name, obd->obd_uuid.uuid, obd->obd_minor, obd,
109 obd->obd_type->typ_name, i);
110 LASSERT(strcmp(obd->obd_type->typ_name, LUSTRE_MDC_NAME) == 0);
111
112 if (tgt->ltd_active == activate) {
113 CDEBUG(D_INFO, "OBD %p already %sactive!\n", obd,
114 activate ? "" : "in");
115 GOTO(out_lmv_lock, rc);
116 }
117
118 CDEBUG(D_INFO, "Marking OBD %p %sactive\n", obd,
119 activate ? "" : "in");
120 lmv_activate_target(lmv, tgt, activate);
121
122 out_lmv_lock:
123 spin_unlock(&lmv->lmv_lock);
124 return rc;
125 }
126
127 struct obd_uuid *lmv_get_uuid(struct obd_export *exp)
128 {
129 struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
130
131 return obd_get_uuid(lmv->tgts[0]->ltd_exp);
132 }
133
134 static int lmv_notify(struct obd_device *obd, struct obd_device *watched,
135 enum obd_notify_event ev, void *data)
136 {
137 struct obd_connect_data *conn_data;
138 struct lmv_obd *lmv = &obd->u.lmv;
139 struct obd_uuid *uuid;
140 int rc = 0;
141
142 if (strcmp(watched->obd_type->typ_name, LUSTRE_MDC_NAME)) {
143 CERROR("unexpected notification of %s %s!\n",
144 watched->obd_type->typ_name,
145 watched->obd_name);
146 return -EINVAL;
147 }
148
149 uuid = &watched->u.cli.cl_target_uuid;
150 if (ev == OBD_NOTIFY_ACTIVE || ev == OBD_NOTIFY_INACTIVE) {
151 /*
152 * Set MDC as active before notifying the observer, so the
153 * observer can use the MDC normally.
154 */
155 rc = lmv_set_mdc_active(lmv, uuid,
156 ev == OBD_NOTIFY_ACTIVE);
157 if (rc) {
158 CERROR("%sactivation of %s failed: %d\n",
159 ev == OBD_NOTIFY_ACTIVE ? "" : "de",
160 uuid->uuid, rc);
161 return rc;
162 }
163 } else if (ev == OBD_NOTIFY_OCD) {
164 conn_data = &watched->u.cli.cl_import->imp_connect_data;
165 /*
166 * XXX: Make sure that ocd_connect_flags from all targets are
167 * the same. Otherwise one of MDTs runs wrong version or
168 * something like this. --umka
169 */
170 obd->obd_self_export->exp_connect_data = *conn_data;
171 }
172 #if 0
173 else if (ev == OBD_NOTIFY_DISCON) {
174 /*
175 * For disconnect event, flush fld cache for failout MDS case.
176 */
177 fld_client_flush(&lmv->lmv_fld);
178 }
179 #endif
180 /*
181 * Pass the notification up the chain.
182 */
183 if (obd->obd_observer)
184 rc = obd_notify(obd->obd_observer, watched, ev, data);
185
186 return rc;
187 }
188
189 /**
190 * This is fake connect function. Its purpose is to initialize lmv and say
191 * caller that everything is okay. Real connection will be performed later.
192 */
193 static int lmv_connect(const struct lu_env *env,
194 struct obd_export **exp, struct obd_device *obd,
195 struct obd_uuid *cluuid, struct obd_connect_data *data,
196 void *localdata)
197 {
198 struct proc_dir_entry *lmv_proc_dir;
199 struct lmv_obd *lmv = &obd->u.lmv;
200 struct lustre_handle conn = { 0 };
201 int rc = 0;
202
203 /*
204 * We don't want to actually do the underlying connections more than
205 * once, so keep track.
206 */
207 lmv->refcount++;
208 if (lmv->refcount > 1) {
209 *exp = NULL;
210 return 0;
211 }
212
213 rc = class_connect(&conn, obd, cluuid);
214 if (rc) {
215 CERROR("class_connection() returned %d\n", rc);
216 return rc;
217 }
218
219 *exp = class_conn2export(&conn);
220 class_export_get(*exp);
221
222 lmv->exp = *exp;
223 lmv->connected = 0;
224 lmv->cluuid = *cluuid;
225
226 if (data)
227 lmv->conn_data = *data;
228
229 if (obd->obd_proc_private != NULL) {
230 lmv_proc_dir = obd->obd_proc_private;
231 } else {
232 lmv_proc_dir = lprocfs_register("target_obds", obd->obd_proc_entry,
233 NULL, NULL);
234 if (IS_ERR(lmv_proc_dir)) {
235 CERROR("could not register /proc/fs/lustre/%s/%s/target_obds.",
236 obd->obd_type->typ_name, obd->obd_name);
237 lmv_proc_dir = NULL;
238 }
239 obd->obd_proc_private = lmv_proc_dir;
240 }
241
242 /*
243 * All real clients should perform actual connection right away, because
244 * it is possible, that LMV will not have opportunity to connect targets
245 * and MDC stuff will be called directly, for instance while reading
246 * ../mdc/../kbytesfree procfs file, etc.
247 */
248 if (data->ocd_connect_flags & OBD_CONNECT_REAL)
249 rc = lmv_check_connect(obd);
250
251 if (rc && lmv_proc_dir) {
252 lprocfs_remove(&lmv_proc_dir);
253 obd->obd_proc_private = NULL;
254 }
255
256 return rc;
257 }
258
259 static void lmv_set_timeouts(struct obd_device *obd)
260 {
261 struct lmv_tgt_desc *tgt;
262 struct lmv_obd *lmv;
263 int i;
264
265 lmv = &obd->u.lmv;
266 if (lmv->server_timeout == 0)
267 return;
268
269 if (lmv->connected == 0)
270 return;
271
272 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
273 tgt = lmv->tgts[i];
274 if (tgt == NULL || tgt->ltd_exp == NULL || tgt->ltd_active == 0)
275 continue;
276
277 obd_set_info_async(NULL, tgt->ltd_exp, sizeof(KEY_INTERMDS),
278 KEY_INTERMDS, 0, NULL, NULL);
279 }
280 }
281
282 static int lmv_init_ea_size(struct obd_export *exp, int easize,
283 int def_easize, int cookiesize)
284 {
285 struct obd_device *obd = exp->exp_obd;
286 struct lmv_obd *lmv = &obd->u.lmv;
287 int i;
288 int rc = 0;
289 int change = 0;
290
291 if (lmv->max_easize < easize) {
292 lmv->max_easize = easize;
293 change = 1;
294 }
295 if (lmv->max_def_easize < def_easize) {
296 lmv->max_def_easize = def_easize;
297 change = 1;
298 }
299 if (lmv->max_cookiesize < cookiesize) {
300 lmv->max_cookiesize = cookiesize;
301 change = 1;
302 }
303 if (change == 0)
304 return 0;
305
306 if (lmv->connected == 0)
307 return 0;
308
309 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
310 if (lmv->tgts[i] == NULL ||
311 lmv->tgts[i]->ltd_exp == NULL ||
312 lmv->tgts[i]->ltd_active == 0) {
313 CWARN("%s: NULL export for %d\n", obd->obd_name, i);
314 continue;
315 }
316
317 rc = md_init_ea_size(lmv->tgts[i]->ltd_exp, easize, def_easize,
318 cookiesize);
319 if (rc) {
320 CERROR("%s: obd_init_ea_size() failed on MDT target %d:"
321 " rc = %d.\n", obd->obd_name, i, rc);
322 break;
323 }
324 }
325 return rc;
326 }
327
328 #define MAX_STRING_SIZE 128
329
330 int lmv_connect_mdc(struct obd_device *obd, struct lmv_tgt_desc *tgt)
331 {
332 struct proc_dir_entry *lmv_proc_dir;
333 struct lmv_obd *lmv = &obd->u.lmv;
334 struct obd_uuid *cluuid = &lmv->cluuid;
335 struct obd_uuid lmv_mdc_uuid = { "LMV_MDC_UUID" };
336 struct obd_device *mdc_obd;
337 struct obd_export *mdc_exp;
338 struct lu_fld_target target;
339 int rc;
340
341 mdc_obd = class_find_client_obd(&tgt->ltd_uuid, LUSTRE_MDC_NAME,
342 &obd->obd_uuid);
343 if (!mdc_obd) {
344 CERROR("target %s not attached\n", tgt->ltd_uuid.uuid);
345 return -EINVAL;
346 }
347
348 CDEBUG(D_CONFIG, "connect to %s(%s) - %s, %s FOR %s\n",
349 mdc_obd->obd_name, mdc_obd->obd_uuid.uuid,
350 tgt->ltd_uuid.uuid, obd->obd_uuid.uuid,
351 cluuid->uuid);
352
353 if (!mdc_obd->obd_set_up) {
354 CERROR("target %s is not set up\n", tgt->ltd_uuid.uuid);
355 return -EINVAL;
356 }
357
358 rc = obd_connect(NULL, &mdc_exp, mdc_obd, &lmv_mdc_uuid,
359 &lmv->conn_data, NULL);
360 if (rc) {
361 CERROR("target %s connect error %d\n", tgt->ltd_uuid.uuid, rc);
362 return rc;
363 }
364
365 /*
366 * Init fid sequence client for this mdc and add new fld target.
367 */
368 rc = obd_fid_init(mdc_obd, mdc_exp, LUSTRE_SEQ_METADATA);
369 if (rc)
370 return rc;
371
372 target.ft_srv = NULL;
373 target.ft_exp = mdc_exp;
374 target.ft_idx = tgt->ltd_idx;
375
376 fld_client_add_target(&lmv->lmv_fld, &target);
377
378 rc = obd_register_observer(mdc_obd, obd);
379 if (rc) {
380 obd_disconnect(mdc_exp);
381 CERROR("target %s register_observer error %d\n",
382 tgt->ltd_uuid.uuid, rc);
383 return rc;
384 }
385
386 if (obd->obd_observer) {
387 /*
388 * Tell the observer about the new target.
389 */
390 rc = obd_notify(obd->obd_observer, mdc_exp->exp_obd,
391 OBD_NOTIFY_ACTIVE,
392 (void *)(tgt - lmv->tgts[0]));
393 if (rc) {
394 obd_disconnect(mdc_exp);
395 return rc;
396 }
397 }
398
399 tgt->ltd_active = 1;
400 tgt->ltd_exp = mdc_exp;
401 lmv->desc.ld_active_tgt_count++;
402
403 md_init_ea_size(tgt->ltd_exp, lmv->max_easize,
404 lmv->max_def_easize, lmv->max_cookiesize);
405
406 CDEBUG(D_CONFIG, "Connected to %s(%s) successfully (%d)\n",
407 mdc_obd->obd_name, mdc_obd->obd_uuid.uuid,
408 atomic_read(&obd->obd_refcount));
409
410 lmv_proc_dir = obd->obd_proc_private;
411 if (lmv_proc_dir) {
412 struct proc_dir_entry *mdc_symlink;
413
414 LASSERT(mdc_obd->obd_type != NULL);
415 LASSERT(mdc_obd->obd_type->typ_name != NULL);
416 mdc_symlink = lprocfs_add_symlink(mdc_obd->obd_name,
417 lmv_proc_dir,
418 "../../../%s/%s",
419 mdc_obd->obd_type->typ_name,
420 mdc_obd->obd_name);
421 if (mdc_symlink == NULL) {
422 CERROR("Could not register LMV target "
423 "/proc/fs/lustre/%s/%s/target_obds/%s.",
424 obd->obd_type->typ_name, obd->obd_name,
425 mdc_obd->obd_name);
426 lprocfs_remove(&lmv_proc_dir);
427 obd->obd_proc_private = NULL;
428 }
429 }
430 return 0;
431 }
432
433 static void lmv_del_target(struct lmv_obd *lmv, int index)
434 {
435 if (lmv->tgts[index] == NULL)
436 return;
437
438 OBD_FREE_PTR(lmv->tgts[index]);
439 lmv->tgts[index] = NULL;
440 return;
441 }
442
443 static int lmv_add_target(struct obd_device *obd, struct obd_uuid *uuidp,
444 __u32 index, int gen)
445 {
446 struct lmv_obd *lmv = &obd->u.lmv;
447 struct lmv_tgt_desc *tgt;
448 int rc = 0;
449
450 CDEBUG(D_CONFIG, "Target uuid: %s. index %d\n", uuidp->uuid, index);
451
452 lmv_init_lock(lmv);
453
454 if (lmv->desc.ld_tgt_count == 0) {
455 struct obd_device *mdc_obd;
456
457 mdc_obd = class_find_client_obd(uuidp, LUSTRE_MDC_NAME,
458 &obd->obd_uuid);
459 if (!mdc_obd) {
460 lmv_init_unlock(lmv);
461 CERROR("%s: Target %s not attached: rc = %d\n",
462 obd->obd_name, uuidp->uuid, -EINVAL);
463 return -EINVAL;
464 }
465 }
466
467 if ((index < lmv->tgts_size) && (lmv->tgts[index] != NULL)) {
468 tgt = lmv->tgts[index];
469 CERROR("%s: UUID %s already assigned at LOV target index %d:"
470 " rc = %d\n", obd->obd_name,
471 obd_uuid2str(&tgt->ltd_uuid), index, -EEXIST);
472 lmv_init_unlock(lmv);
473 return -EEXIST;
474 }
475
476 if (index >= lmv->tgts_size) {
477 /* We need to reallocate the lmv target array. */
478 struct lmv_tgt_desc **newtgts, **old = NULL;
479 __u32 newsize = 1;
480 __u32 oldsize = 0;
481
482 while (newsize < index + 1)
483 newsize = newsize << 1;
484 OBD_ALLOC(newtgts, sizeof(*newtgts) * newsize);
485 if (newtgts == NULL) {
486 lmv_init_unlock(lmv);
487 return -ENOMEM;
488 }
489
490 if (lmv->tgts_size) {
491 memcpy(newtgts, lmv->tgts,
492 sizeof(*newtgts) * lmv->tgts_size);
493 old = lmv->tgts;
494 oldsize = lmv->tgts_size;
495 }
496
497 lmv->tgts = newtgts;
498 lmv->tgts_size = newsize;
499 smp_rmb();
500 if (old)
501 OBD_FREE(old, sizeof(*old) * oldsize);
502
503 CDEBUG(D_CONFIG, "tgts: %p size: %d\n", lmv->tgts,
504 lmv->tgts_size);
505 }
506
507 OBD_ALLOC_PTR(tgt);
508 if (!tgt) {
509 lmv_init_unlock(lmv);
510 return -ENOMEM;
511 }
512
513 mutex_init(&tgt->ltd_fid_mutex);
514 tgt->ltd_idx = index;
515 tgt->ltd_uuid = *uuidp;
516 tgt->ltd_active = 0;
517 lmv->tgts[index] = tgt;
518 if (index >= lmv->desc.ld_tgt_count)
519 lmv->desc.ld_tgt_count = index + 1;
520
521 if (lmv->connected) {
522 rc = lmv_connect_mdc(obd, tgt);
523 if (rc) {
524 spin_lock(&lmv->lmv_lock);
525 lmv->desc.ld_tgt_count--;
526 memset(tgt, 0, sizeof(*tgt));
527 spin_unlock(&lmv->lmv_lock);
528 } else {
529 int easize = sizeof(struct lmv_stripe_md) +
530 lmv->desc.ld_tgt_count *
531 sizeof(struct lu_fid);
532 lmv_init_ea_size(obd->obd_self_export, easize, 0, 0);
533 }
534 }
535
536 lmv_init_unlock(lmv);
537 return rc;
538 }
539
540 int lmv_check_connect(struct obd_device *obd)
541 {
542 struct lmv_obd *lmv = &obd->u.lmv;
543 struct lmv_tgt_desc *tgt;
544 int i;
545 int rc;
546 int easize;
547
548 if (lmv->connected)
549 return 0;
550
551 lmv_init_lock(lmv);
552 if (lmv->connected) {
553 lmv_init_unlock(lmv);
554 return 0;
555 }
556
557 if (lmv->desc.ld_tgt_count == 0) {
558 lmv_init_unlock(lmv);
559 CERROR("%s: no targets configured.\n", obd->obd_name);
560 return -EINVAL;
561 }
562
563 CDEBUG(D_CONFIG, "Time to connect %s to %s\n",
564 lmv->cluuid.uuid, obd->obd_name);
565
566 LASSERT(lmv->tgts != NULL);
567
568 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
569 tgt = lmv->tgts[i];
570 if (tgt == NULL)
571 continue;
572 rc = lmv_connect_mdc(obd, tgt);
573 if (rc)
574 GOTO(out_disc, rc);
575 }
576
577 lmv_set_timeouts(obd);
578 class_export_put(lmv->exp);
579 lmv->connected = 1;
580 easize = lmv_get_easize(lmv);
581 lmv_init_ea_size(obd->obd_self_export, easize, 0, 0);
582 lmv_init_unlock(lmv);
583 return 0;
584
585 out_disc:
586 while (i-- > 0) {
587 int rc2;
588 tgt = lmv->tgts[i];
589 if (tgt == NULL)
590 continue;
591 tgt->ltd_active = 0;
592 if (tgt->ltd_exp) {
593 --lmv->desc.ld_active_tgt_count;
594 rc2 = obd_disconnect(tgt->ltd_exp);
595 if (rc2) {
596 CERROR("LMV target %s disconnect on "
597 "MDC idx %d: error %d\n",
598 tgt->ltd_uuid.uuid, i, rc2);
599 }
600 }
601 }
602 class_disconnect(lmv->exp);
603 lmv_init_unlock(lmv);
604 return rc;
605 }
606
607 static int lmv_disconnect_mdc(struct obd_device *obd, struct lmv_tgt_desc *tgt)
608 {
609 struct proc_dir_entry *lmv_proc_dir;
610 struct lmv_obd *lmv = &obd->u.lmv;
611 struct obd_device *mdc_obd;
612 int rc;
613
614 LASSERT(tgt != NULL);
615 LASSERT(obd != NULL);
616
617 mdc_obd = class_exp2obd(tgt->ltd_exp);
618
619 if (mdc_obd) {
620 mdc_obd->obd_force = obd->obd_force;
621 mdc_obd->obd_fail = obd->obd_fail;
622 mdc_obd->obd_no_recov = obd->obd_no_recov;
623 }
624
625 lmv_proc_dir = obd->obd_proc_private;
626 if (lmv_proc_dir)
627 lprocfs_remove_proc_entry(mdc_obd->obd_name, lmv_proc_dir);
628
629 rc = obd_fid_fini(tgt->ltd_exp->exp_obd);
630 if (rc)
631 CERROR("Can't finanize fids factory\n");
632
633 CDEBUG(D_INFO, "Disconnected from %s(%s) successfully\n",
634 tgt->ltd_exp->exp_obd->obd_name,
635 tgt->ltd_exp->exp_obd->obd_uuid.uuid);
636
637 obd_register_observer(tgt->ltd_exp->exp_obd, NULL);
638 rc = obd_disconnect(tgt->ltd_exp);
639 if (rc) {
640 if (tgt->ltd_active) {
641 CERROR("Target %s disconnect error %d\n",
642 tgt->ltd_uuid.uuid, rc);
643 }
644 }
645
646 lmv_activate_target(lmv, tgt, 0);
647 tgt->ltd_exp = NULL;
648 return 0;
649 }
650
651 static int lmv_disconnect(struct obd_export *exp)
652 {
653 struct obd_device *obd = class_exp2obd(exp);
654 struct lmv_obd *lmv = &obd->u.lmv;
655 int rc;
656 int i;
657
658 if (!lmv->tgts)
659 goto out_local;
660
661 /*
662 * Only disconnect the underlying layers on the final disconnect.
663 */
664 lmv->refcount--;
665 if (lmv->refcount != 0)
666 goto out_local;
667
668 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
669 if (lmv->tgts[i] == NULL || lmv->tgts[i]->ltd_exp == NULL)
670 continue;
671
672 lmv_disconnect_mdc(obd, lmv->tgts[i]);
673 }
674
675 if (obd->obd_proc_private)
676 lprocfs_remove((proc_dir_entry_t **)&obd->obd_proc_private);
677 else
678 CERROR("/proc/fs/lustre/%s/%s/target_obds missing\n",
679 obd->obd_type->typ_name, obd->obd_name);
680
681 out_local:
682 /*
683 * This is the case when no real connection is established by
684 * lmv_check_connect().
685 */
686 if (!lmv->connected)
687 class_export_put(exp);
688 rc = class_disconnect(exp);
689 if (lmv->refcount == 0)
690 lmv->connected = 0;
691 return rc;
692 }
693
694 static int lmv_fid2path(struct obd_export *exp, int len, void *karg, void *uarg)
695 {
696 struct obd_device *obddev = class_exp2obd(exp);
697 struct lmv_obd *lmv = &obddev->u.lmv;
698 struct getinfo_fid2path *gf;
699 struct lmv_tgt_desc *tgt;
700 struct getinfo_fid2path *remote_gf = NULL;
701 int remote_gf_size = 0;
702 int rc;
703
704 gf = (struct getinfo_fid2path *)karg;
705 tgt = lmv_find_target(lmv, &gf->gf_fid);
706 if (IS_ERR(tgt))
707 return PTR_ERR(tgt);
708
709 repeat_fid2path:
710 rc = obd_iocontrol(OBD_IOC_FID2PATH, tgt->ltd_exp, len, gf, uarg);
711 if (rc != 0 && rc != -EREMOTE)
712 GOTO(out_fid2path, rc);
713
714 /* If remote_gf != NULL, it means just building the
715 * path on the remote MDT, copy this path segement to gf */
716 if (remote_gf != NULL) {
717 struct getinfo_fid2path *ori_gf;
718 char *ptr;
719
720 ori_gf = (struct getinfo_fid2path *)karg;
721 if (strlen(ori_gf->gf_path) +
722 strlen(gf->gf_path) > ori_gf->gf_pathlen)
723 GOTO(out_fid2path, rc = -EOVERFLOW);
724
725 ptr = ori_gf->gf_path;
726
727 memmove(ptr + strlen(gf->gf_path) + 1, ptr,
728 strlen(ori_gf->gf_path));
729
730 strncpy(ptr, gf->gf_path, strlen(gf->gf_path));
731 ptr += strlen(gf->gf_path);
732 *ptr = '/';
733 }
734
735 CDEBUG(D_INFO, "%s: get path %s "DFID" rec: "LPU64" ln: %u\n",
736 tgt->ltd_exp->exp_obd->obd_name,
737 gf->gf_path, PFID(&gf->gf_fid), gf->gf_recno,
738 gf->gf_linkno);
739
740 if (rc == 0)
741 GOTO(out_fid2path, rc);
742
743 /* sigh, has to go to another MDT to do path building further */
744 if (remote_gf == NULL) {
745 remote_gf_size = sizeof(*remote_gf) + PATH_MAX;
746 OBD_ALLOC(remote_gf, remote_gf_size);
747 if (remote_gf == NULL)
748 GOTO(out_fid2path, rc = -ENOMEM);
749 remote_gf->gf_pathlen = PATH_MAX;
750 }
751
752 if (!fid_is_sane(&gf->gf_fid)) {
753 CERROR("%s: invalid FID "DFID": rc = %d\n",
754 tgt->ltd_exp->exp_obd->obd_name,
755 PFID(&gf->gf_fid), -EINVAL);
756 GOTO(out_fid2path, rc = -EINVAL);
757 }
758
759 tgt = lmv_find_target(lmv, &gf->gf_fid);
760 if (IS_ERR(tgt))
761 GOTO(out_fid2path, rc = -EINVAL);
762
763 remote_gf->gf_fid = gf->gf_fid;
764 remote_gf->gf_recno = -1;
765 remote_gf->gf_linkno = -1;
766 memset(remote_gf->gf_path, 0, remote_gf->gf_pathlen);
767 gf = remote_gf;
768 goto repeat_fid2path;
769
770 out_fid2path:
771 if (remote_gf != NULL)
772 OBD_FREE(remote_gf, remote_gf_size);
773 return rc;
774 }
775
776 static int lmv_hsm_req_count(struct lmv_obd *lmv,
777 const struct hsm_user_request *hur,
778 const struct lmv_tgt_desc *tgt_mds)
779 {
780 int i, nr = 0;
781 struct lmv_tgt_desc *curr_tgt;
782
783 /* count how many requests must be sent to the given target */
784 for (i = 0; i < hur->hur_request.hr_itemcount; i++) {
785 curr_tgt = lmv_find_target(lmv, &hur->hur_user_item[i].hui_fid);
786 if (obd_uuid_equals(&curr_tgt->ltd_uuid, &tgt_mds->ltd_uuid))
787 nr++;
788 }
789 return nr;
790 }
791
792 static void lmv_hsm_req_build(struct lmv_obd *lmv,
793 struct hsm_user_request *hur_in,
794 const struct lmv_tgt_desc *tgt_mds,
795 struct hsm_user_request *hur_out)
796 {
797 int i, nr_out;
798 struct lmv_tgt_desc *curr_tgt;
799
800 /* build the hsm_user_request for the given target */
801 hur_out->hur_request = hur_in->hur_request;
802 nr_out = 0;
803 for (i = 0; i < hur_in->hur_request.hr_itemcount; i++) {
804 curr_tgt = lmv_find_target(lmv,
805 &hur_in->hur_user_item[i].hui_fid);
806 if (obd_uuid_equals(&curr_tgt->ltd_uuid, &tgt_mds->ltd_uuid)) {
807 hur_out->hur_user_item[nr_out] =
808 hur_in->hur_user_item[i];
809 nr_out++;
810 }
811 }
812 hur_out->hur_request.hr_itemcount = nr_out;
813 memcpy(hur_data(hur_out), hur_data(hur_in),
814 hur_in->hur_request.hr_data_len);
815 }
816
817 static int lmv_hsm_ct_unregister(struct lmv_obd *lmv, unsigned int cmd, int len,
818 struct lustre_kernelcomm *lk, void *uarg)
819 {
820 int i, rc = 0;
821
822 /* unregister request (call from llapi_hsm_copytool_fini) */
823 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
824 /* best effort: try to clean as much as possible
825 * (continue on error) */
826 obd_iocontrol(cmd, lmv->tgts[i]->ltd_exp, len, lk, uarg);
827 }
828
829 /* Whatever the result, remove copytool from kuc groups.
830 * Unreached coordinators will get EPIPE on next requests
831 * and will unregister automatically.
832 */
833 rc = libcfs_kkuc_group_rem(lk->lk_uid, lk->lk_group);
834 return rc;
835 }
836
837 static int lmv_hsm_ct_register(struct lmv_obd *lmv, unsigned int cmd, int len,
838 struct lustre_kernelcomm *lk, void *uarg)
839 {
840 struct file *filp;
841 int i, j, err;
842 int rc = 0;
843 bool any_set = false;
844
845 /* All or nothing: try to register to all MDS.
846 * In case of failure, unregister from previous MDS,
847 * except if it because of inactive target. */
848 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
849 err = obd_iocontrol(cmd, lmv->tgts[i]->ltd_exp,
850 len, lk, uarg);
851 if (err) {
852 if (lmv->tgts[i]->ltd_active) {
853 /* permanent error */
854 CERROR("error: iocontrol MDC %s on MDT"
855 "idx %d cmd %x: err = %d\n",
856 lmv->tgts[i]->ltd_uuid.uuid,
857 i, cmd, err);
858 rc = err;
859 lk->lk_flags |= LK_FLG_STOP;
860 /* unregister from previous MDS */
861 for (j = 0; j < i; j++)
862 obd_iocontrol(cmd,
863 lmv->tgts[j]->ltd_exp,
864 len, lk, uarg);
865 return rc;
866 }
867 /* else: transient error.
868 * kuc will register to the missing MDT
869 * when it is back */
870 } else {
871 any_set = true;
872 }
873 }
874
875 if (!any_set)
876 /* no registration done: return error */
877 return -ENOTCONN;
878
879 /* at least one registration done, with no failure */
880 filp = fget(lk->lk_wfd);
881 if (filp == NULL) {
882 return -EBADF;
883 }
884 rc = libcfs_kkuc_group_add(filp, lk->lk_uid, lk->lk_group, lk->lk_data);
885 if (rc != 0 && filp != NULL)
886 fput(filp);
887 return rc;
888 }
889
890
891
892
893 static int lmv_iocontrol(unsigned int cmd, struct obd_export *exp,
894 int len, void *karg, void *uarg)
895 {
896 struct obd_device *obddev = class_exp2obd(exp);
897 struct lmv_obd *lmv = &obddev->u.lmv;
898 int i = 0;
899 int rc = 0;
900 int set = 0;
901 int count = lmv->desc.ld_tgt_count;
902
903 if (count == 0)
904 return -ENOTTY;
905
906 switch (cmd) {
907 case IOC_OBD_STATFS: {
908 struct obd_ioctl_data *data = karg;
909 struct obd_device *mdc_obd;
910 struct obd_statfs stat_buf = {0};
911 __u32 index;
912
913 memcpy(&index, data->ioc_inlbuf2, sizeof(__u32));
914 if ((index >= count))
915 return -ENODEV;
916
917 if (lmv->tgts[index] == NULL ||
918 lmv->tgts[index]->ltd_active == 0)
919 return -ENODATA;
920
921 mdc_obd = class_exp2obd(lmv->tgts[index]->ltd_exp);
922 if (!mdc_obd)
923 return -EINVAL;
924
925 /* copy UUID */
926 if (copy_to_user(data->ioc_pbuf2, obd2cli_tgt(mdc_obd),
927 min((int) data->ioc_plen2,
928 (int) sizeof(struct obd_uuid))))
929 return -EFAULT;
930
931 rc = obd_statfs(NULL, lmv->tgts[index]->ltd_exp, &stat_buf,
932 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
933 0);
934 if (rc)
935 return rc;
936 if (copy_to_user(data->ioc_pbuf1, &stat_buf,
937 min((int) data->ioc_plen1,
938 (int) sizeof(stat_buf))))
939 return -EFAULT;
940 break;
941 }
942 case OBD_IOC_QUOTACTL: {
943 struct if_quotactl *qctl = karg;
944 struct lmv_tgt_desc *tgt = NULL;
945 struct obd_quotactl *oqctl;
946
947 if (qctl->qc_valid == QC_MDTIDX) {
948 if (qctl->qc_idx < 0 || count <= qctl->qc_idx)
949 return -EINVAL;
950
951 tgt = lmv->tgts[qctl->qc_idx];
952 if (tgt == NULL || tgt->ltd_exp == NULL)
953 return -EINVAL;
954 } else if (qctl->qc_valid == QC_UUID) {
955 for (i = 0; i < count; i++) {
956 tgt = lmv->tgts[i];
957 if (tgt == NULL)
958 continue;
959 if (!obd_uuid_equals(&tgt->ltd_uuid,
960 &qctl->obd_uuid))
961 continue;
962
963 if (tgt->ltd_exp == NULL)
964 return -EINVAL;
965
966 break;
967 }
968 } else {
969 return -EINVAL;
970 }
971
972 if (i >= count)
973 return -EAGAIN;
974
975 LASSERT(tgt && tgt->ltd_exp);
976 OBD_ALLOC_PTR(oqctl);
977 if (!oqctl)
978 return -ENOMEM;
979
980 QCTL_COPY(oqctl, qctl);
981 rc = obd_quotactl(tgt->ltd_exp, oqctl);
982 if (rc == 0) {
983 QCTL_COPY(qctl, oqctl);
984 qctl->qc_valid = QC_MDTIDX;
985 qctl->obd_uuid = tgt->ltd_uuid;
986 }
987 OBD_FREE_PTR(oqctl);
988 break;
989 }
990 case OBD_IOC_CHANGELOG_SEND:
991 case OBD_IOC_CHANGELOG_CLEAR: {
992 struct ioc_changelog *icc = karg;
993
994 if (icc->icc_mdtindex >= count)
995 return -ENODEV;
996
997 if (lmv->tgts[icc->icc_mdtindex] == NULL ||
998 lmv->tgts[icc->icc_mdtindex]->ltd_exp == NULL ||
999 lmv->tgts[icc->icc_mdtindex]->ltd_active == 0)
1000 return -ENODEV;
1001 rc = obd_iocontrol(cmd, lmv->tgts[icc->icc_mdtindex]->ltd_exp,
1002 sizeof(*icc), icc, NULL);
1003 break;
1004 }
1005 case LL_IOC_GET_CONNECT_FLAGS: {
1006 if (lmv->tgts[0] == NULL)
1007 return -ENODATA;
1008 rc = obd_iocontrol(cmd, lmv->tgts[0]->ltd_exp, len, karg, uarg);
1009 break;
1010 }
1011 case OBD_IOC_FID2PATH: {
1012 rc = lmv_fid2path(exp, len, karg, uarg);
1013 break;
1014 }
1015 case LL_IOC_HSM_STATE_GET:
1016 case LL_IOC_HSM_STATE_SET:
1017 case LL_IOC_HSM_ACTION: {
1018 struct md_op_data *op_data = karg;
1019 struct lmv_tgt_desc *tgt;
1020
1021 tgt = lmv_find_target(lmv, &op_data->op_fid1);
1022 if (IS_ERR(tgt))
1023 return PTR_ERR(tgt);
1024
1025 if (tgt->ltd_exp == NULL)
1026 return -EINVAL;
1027
1028 rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
1029 break;
1030 }
1031 case LL_IOC_HSM_PROGRESS: {
1032 const struct hsm_progress_kernel *hpk = karg;
1033 struct lmv_tgt_desc *tgt;
1034
1035 tgt = lmv_find_target(lmv, &hpk->hpk_fid);
1036 if (IS_ERR(tgt))
1037 return PTR_ERR(tgt);
1038 rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
1039 break;
1040 }
1041 case LL_IOC_HSM_REQUEST: {
1042 struct hsm_user_request *hur = karg;
1043 struct lmv_tgt_desc *tgt;
1044 unsigned int reqcount = hur->hur_request.hr_itemcount;
1045
1046 if (reqcount == 0)
1047 return 0;
1048
1049 /* if the request is about a single fid
1050 * or if there is a single MDS, no need to split
1051 * the request. */
1052 if (reqcount == 1 || count == 1) {
1053 tgt = lmv_find_target(lmv,
1054 &hur->hur_user_item[0].hui_fid);
1055 if (IS_ERR(tgt))
1056 return PTR_ERR(tgt);
1057 rc = obd_iocontrol(cmd, tgt->ltd_exp, len, karg, uarg);
1058 } else {
1059 /* split fid list to their respective MDS */
1060 for (i = 0; i < count; i++) {
1061 unsigned int nr, reqlen;
1062 int rc1;
1063 struct hsm_user_request *req;
1064
1065 nr = lmv_hsm_req_count(lmv, hur, lmv->tgts[i]);
1066 if (nr == 0) /* nothing for this MDS */
1067 continue;
1068
1069 /* build a request with fids for this MDS */
1070 reqlen = offsetof(typeof(*hur),
1071 hur_user_item[nr])
1072 + hur->hur_request.hr_data_len;
1073 OBD_ALLOC_LARGE(req, reqlen);
1074 if (req == NULL)
1075 return -ENOMEM;
1076
1077 lmv_hsm_req_build(lmv, hur, lmv->tgts[i], req);
1078
1079 rc1 = obd_iocontrol(cmd, lmv->tgts[i]->ltd_exp,
1080 reqlen, req, uarg);
1081 if (rc1 != 0 && rc == 0)
1082 rc = rc1;
1083 OBD_FREE_LARGE(req, reqlen);
1084 }
1085 }
1086 break;
1087 }
1088 case LL_IOC_LOV_SWAP_LAYOUTS: {
1089 struct md_op_data *op_data = karg;
1090 struct lmv_tgt_desc *tgt1, *tgt2;
1091
1092 tgt1 = lmv_find_target(lmv, &op_data->op_fid1);
1093 if (IS_ERR(tgt1))
1094 return PTR_ERR(tgt1);
1095
1096 tgt2 = lmv_find_target(lmv, &op_data->op_fid2);
1097 if (IS_ERR(tgt2))
1098 return PTR_ERR(tgt2);
1099
1100 if ((tgt1->ltd_exp == NULL) || (tgt2->ltd_exp == NULL))
1101 return -EINVAL;
1102
1103 /* only files on same MDT can have their layouts swapped */
1104 if (tgt1->ltd_idx != tgt2->ltd_idx)
1105 return -EPERM;
1106
1107 rc = obd_iocontrol(cmd, tgt1->ltd_exp, len, karg, uarg);
1108 break;
1109 }
1110 case LL_IOC_HSM_CT_START: {
1111 struct lustre_kernelcomm *lk = karg;
1112 if (lk->lk_flags & LK_FLG_STOP)
1113 rc = lmv_hsm_ct_unregister(lmv, cmd, len, lk, uarg);
1114 else
1115 rc = lmv_hsm_ct_register(lmv, cmd, len, lk, uarg);
1116 break;
1117 }
1118 default:
1119 for (i = 0; i < count; i++) {
1120 struct obd_device *mdc_obd;
1121 int err;
1122
1123 if (lmv->tgts[i] == NULL ||
1124 lmv->tgts[i]->ltd_exp == NULL)
1125 continue;
1126 /* ll_umount_begin() sets force flag but for lmv, not
1127 * mdc. Let's pass it through */
1128 mdc_obd = class_exp2obd(lmv->tgts[i]->ltd_exp);
1129 mdc_obd->obd_force = obddev->obd_force;
1130 err = obd_iocontrol(cmd, lmv->tgts[i]->ltd_exp, len,
1131 karg, uarg);
1132 if (err == -ENODATA && cmd == OBD_IOC_POLL_QUOTACHECK) {
1133 return err;
1134 } else if (err) {
1135 if (lmv->tgts[i]->ltd_active) {
1136 CERROR("error: iocontrol MDC %s on MDT"
1137 "idx %d cmd %x: err = %d\n",
1138 lmv->tgts[i]->ltd_uuid.uuid,
1139 i, cmd, err);
1140 if (!rc)
1141 rc = err;
1142 }
1143 } else
1144 set = 1;
1145 }
1146 if (!set && !rc)
1147 rc = -EIO;
1148 }
1149 return rc;
1150 }
1151
1152 #if 0
1153 static int lmv_all_chars_policy(int count, const char *name,
1154 int len)
1155 {
1156 unsigned int c = 0;
1157
1158 while (len > 0)
1159 c += name[--len];
1160 c = c % count;
1161 return c;
1162 }
1163
1164 static int lmv_nid_policy(struct lmv_obd *lmv)
1165 {
1166 struct obd_import *imp;
1167 __u32 id;
1168
1169 /*
1170 * XXX: To get nid we assume that underlying obd device is mdc.
1171 */
1172 imp = class_exp2cliimp(lmv->tgts[0].ltd_exp);
1173 id = imp->imp_connection->c_self ^ (imp->imp_connection->c_self >> 32);
1174 return id % lmv->desc.ld_tgt_count;
1175 }
1176
1177 static int lmv_choose_mds(struct lmv_obd *lmv, struct md_op_data *op_data,
1178 placement_policy_t placement)
1179 {
1180 switch (placement) {
1181 case PLACEMENT_CHAR_POLICY:
1182 return lmv_all_chars_policy(lmv->desc.ld_tgt_count,
1183 op_data->op_name,
1184 op_data->op_namelen);
1185 case PLACEMENT_NID_POLICY:
1186 return lmv_nid_policy(lmv);
1187
1188 default:
1189 break;
1190 }
1191
1192 CERROR("Unsupported placement policy %x\n", placement);
1193 return -EINVAL;
1194 }
1195 #endif
1196
1197 /**
1198 * This is _inode_ placement policy function (not name).
1199 */
1200 static int lmv_placement_policy(struct obd_device *obd,
1201 struct md_op_data *op_data,
1202 mdsno_t *mds)
1203 {
1204 struct lmv_obd *lmv = &obd->u.lmv;
1205
1206 LASSERT(mds != NULL);
1207
1208 if (lmv->desc.ld_tgt_count == 1) {
1209 *mds = 0;
1210 return 0;
1211 }
1212
1213 /**
1214 * If stripe_offset is provided during setdirstripe
1215 * (setdirstripe -i xx), xx MDS will be choosen.
1216 */
1217 if (op_data->op_cli_flags & CLI_SET_MEA) {
1218 struct lmv_user_md *lum;
1219
1220 lum = (struct lmv_user_md *)op_data->op_data;
1221 if (lum->lum_type == LMV_STRIPE_TYPE &&
1222 lum->lum_stripe_offset != -1) {
1223 if (lum->lum_stripe_offset >= lmv->desc.ld_tgt_count) {
1224 CERROR("%s: Stripe_offset %d > MDT count %d:"
1225 " rc = %d\n", obd->obd_name,
1226 lum->lum_stripe_offset,
1227 lmv->desc.ld_tgt_count, -ERANGE);
1228 return -ERANGE;
1229 }
1230 *mds = lum->lum_stripe_offset;
1231 return 0;
1232 }
1233 }
1234
1235 /* Allocate new fid on target according to operation type and parent
1236 * home mds. */
1237 *mds = op_data->op_mds;
1238 return 0;
1239 }
1240
1241 int __lmv_fid_alloc(struct lmv_obd *lmv, struct lu_fid *fid,
1242 mdsno_t mds)
1243 {
1244 struct lmv_tgt_desc *tgt;
1245 int rc;
1246
1247 tgt = lmv_get_target(lmv, mds);
1248 if (IS_ERR(tgt))
1249 return PTR_ERR(tgt);
1250
1251 /*
1252 * New seq alloc and FLD setup should be atomic. Otherwise we may find
1253 * on server that seq in new allocated fid is not yet known.
1254 */
1255 mutex_lock(&tgt->ltd_fid_mutex);
1256
1257 if (tgt->ltd_active == 0 || tgt->ltd_exp == NULL)
1258 GOTO(out, rc = -ENODEV);
1259
1260 /*
1261 * Asking underlaying tgt layer to allocate new fid.
1262 */
1263 rc = obd_fid_alloc(tgt->ltd_exp, fid, NULL);
1264 if (rc > 0) {
1265 LASSERT(fid_is_sane(fid));
1266 rc = 0;
1267 }
1268
1269 out:
1270 mutex_unlock(&tgt->ltd_fid_mutex);
1271 return rc;
1272 }
1273
1274 int lmv_fid_alloc(struct obd_export *exp, struct lu_fid *fid,
1275 struct md_op_data *op_data)
1276 {
1277 struct obd_device *obd = class_exp2obd(exp);
1278 struct lmv_obd *lmv = &obd->u.lmv;
1279 mdsno_t mds = 0;
1280 int rc;
1281
1282 LASSERT(op_data != NULL);
1283 LASSERT(fid != NULL);
1284
1285 rc = lmv_placement_policy(obd, op_data, &mds);
1286 if (rc) {
1287 CERROR("Can't get target for allocating fid, "
1288 "rc %d\n", rc);
1289 return rc;
1290 }
1291
1292 rc = __lmv_fid_alloc(lmv, fid, mds);
1293 if (rc) {
1294 CERROR("Can't alloc new fid, rc %d\n", rc);
1295 return rc;
1296 }
1297
1298 return rc;
1299 }
1300
1301 static int lmv_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
1302 {
1303 struct lmv_obd *lmv = &obd->u.lmv;
1304 struct lprocfs_static_vars lvars;
1305 struct lmv_desc *desc;
1306 int rc;
1307
1308 if (LUSTRE_CFG_BUFLEN(lcfg, 1) < 1) {
1309 CERROR("LMV setup requires a descriptor\n");
1310 return -EINVAL;
1311 }
1312
1313 desc = (struct lmv_desc *)lustre_cfg_buf(lcfg, 1);
1314 if (sizeof(*desc) > LUSTRE_CFG_BUFLEN(lcfg, 1)) {
1315 CERROR("Lmv descriptor size wrong: %d > %d\n",
1316 (int)sizeof(*desc), LUSTRE_CFG_BUFLEN(lcfg, 1));
1317 return -EINVAL;
1318 }
1319
1320 OBD_ALLOC(lmv->tgts, sizeof(*lmv->tgts) * 32);
1321 if (lmv->tgts == NULL)
1322 return -ENOMEM;
1323 lmv->tgts_size = 32;
1324
1325 obd_str2uuid(&lmv->desc.ld_uuid, desc->ld_uuid.uuid);
1326 lmv->desc.ld_tgt_count = 0;
1327 lmv->desc.ld_active_tgt_count = 0;
1328 lmv->max_cookiesize = 0;
1329 lmv->max_def_easize = 0;
1330 lmv->max_easize = 0;
1331 lmv->lmv_placement = PLACEMENT_CHAR_POLICY;
1332
1333 spin_lock_init(&lmv->lmv_lock);
1334 mutex_init(&lmv->init_mutex);
1335
1336 lprocfs_lmv_init_vars(&lvars);
1337
1338 lprocfs_obd_setup(obd, lvars.obd_vars);
1339 #ifdef LPROCFS
1340 {
1341 rc = lprocfs_seq_create(obd->obd_proc_entry, "target_obd",
1342 0444, &lmv_proc_target_fops, obd);
1343 if (rc)
1344 CWARN("%s: error adding LMV target_obd file: rc = %d\n",
1345 obd->obd_name, rc);
1346 }
1347 #endif
1348 rc = fld_client_init(&lmv->lmv_fld, obd->obd_name,
1349 LUSTRE_CLI_FLD_HASH_DHT);
1350 if (rc) {
1351 CERROR("Can't init FLD, err %d\n", rc);
1352 GOTO(out, rc);
1353 }
1354
1355 return 0;
1356
1357 out:
1358 return rc;
1359 }
1360
1361 static int lmv_cleanup(struct obd_device *obd)
1362 {
1363 struct lmv_obd *lmv = &obd->u.lmv;
1364
1365 fld_client_fini(&lmv->lmv_fld);
1366 if (lmv->tgts != NULL) {
1367 int i;
1368 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
1369 if (lmv->tgts[i] == NULL)
1370 continue;
1371 lmv_del_target(lmv, i);
1372 }
1373 OBD_FREE(lmv->tgts, sizeof(*lmv->tgts) * lmv->tgts_size);
1374 lmv->tgts_size = 0;
1375 }
1376 return 0;
1377 }
1378
1379 static int lmv_process_config(struct obd_device *obd, obd_count len, void *buf)
1380 {
1381 struct lustre_cfg *lcfg = buf;
1382 struct obd_uuid obd_uuid;
1383 int gen;
1384 __u32 index;
1385 int rc;
1386
1387 switch (lcfg->lcfg_command) {
1388 case LCFG_ADD_MDC:
1389 /* modify_mdc_tgts add 0:lustre-clilmv 1:lustre-MDT0000_UUID
1390 * 2:0 3:1 4:lustre-MDT0000-mdc_UUID */
1391 if (LUSTRE_CFG_BUFLEN(lcfg, 1) > sizeof(obd_uuid.uuid))
1392 GOTO(out, rc = -EINVAL);
1393
1394 obd_str2uuid(&obd_uuid, lustre_cfg_buf(lcfg, 1));
1395
1396 if (sscanf(lustre_cfg_buf(lcfg, 2), "%d", &index) != 1)
1397 GOTO(out, rc = -EINVAL);
1398 if (sscanf(lustre_cfg_buf(lcfg, 3), "%d", &gen) != 1)
1399 GOTO(out, rc = -EINVAL);
1400 rc = lmv_add_target(obd, &obd_uuid, index, gen);
1401 GOTO(out, rc);
1402 default:
1403 CERROR("Unknown command: %d\n", lcfg->lcfg_command);
1404 GOTO(out, rc = -EINVAL);
1405 }
1406 out:
1407 return rc;
1408 }
1409
1410 static int lmv_statfs(const struct lu_env *env, struct obd_export *exp,
1411 struct obd_statfs *osfs, __u64 max_age, __u32 flags)
1412 {
1413 struct obd_device *obd = class_exp2obd(exp);
1414 struct lmv_obd *lmv = &obd->u.lmv;
1415 struct obd_statfs *temp;
1416 int rc = 0;
1417 int i;
1418
1419 rc = lmv_check_connect(obd);
1420 if (rc)
1421 return rc;
1422
1423 OBD_ALLOC(temp, sizeof(*temp));
1424 if (temp == NULL)
1425 return -ENOMEM;
1426
1427 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
1428 if (lmv->tgts[i] == NULL || lmv->tgts[i]->ltd_exp == NULL)
1429 continue;
1430
1431 rc = obd_statfs(env, lmv->tgts[i]->ltd_exp, temp,
1432 max_age, flags);
1433 if (rc) {
1434 CERROR("can't stat MDS #%d (%s), error %d\n", i,
1435 lmv->tgts[i]->ltd_exp->exp_obd->obd_name,
1436 rc);
1437 GOTO(out_free_temp, rc);
1438 }
1439
1440 if (i == 0) {
1441 *osfs = *temp;
1442 /* If the statfs is from mount, it will needs
1443 * retrieve necessary information from MDT0.
1444 * i.e. mount does not need the merged osfs
1445 * from all of MDT.
1446 * And also clients can be mounted as long as
1447 * MDT0 is in service*/
1448 if (flags & OBD_STATFS_FOR_MDT0)
1449 GOTO(out_free_temp, rc);
1450 } else {
1451 osfs->os_bavail += temp->os_bavail;
1452 osfs->os_blocks += temp->os_blocks;
1453 osfs->os_ffree += temp->os_ffree;
1454 osfs->os_files += temp->os_files;
1455 }
1456 }
1457
1458 out_free_temp:
1459 OBD_FREE(temp, sizeof(*temp));
1460 return rc;
1461 }
1462
1463 static int lmv_getstatus(struct obd_export *exp,
1464 struct lu_fid *fid,
1465 struct obd_capa **pc)
1466 {
1467 struct obd_device *obd = exp->exp_obd;
1468 struct lmv_obd *lmv = &obd->u.lmv;
1469 int rc;
1470
1471 rc = lmv_check_connect(obd);
1472 if (rc)
1473 return rc;
1474
1475 rc = md_getstatus(lmv->tgts[0]->ltd_exp, fid, pc);
1476 return rc;
1477 }
1478
1479 static int lmv_getxattr(struct obd_export *exp, const struct lu_fid *fid,
1480 struct obd_capa *oc, obd_valid valid, const char *name,
1481 const char *input, int input_size, int output_size,
1482 int flags, struct ptlrpc_request **request)
1483 {
1484 struct obd_device *obd = exp->exp_obd;
1485 struct lmv_obd *lmv = &obd->u.lmv;
1486 struct lmv_tgt_desc *tgt;
1487 int rc;
1488
1489 rc = lmv_check_connect(obd);
1490 if (rc)
1491 return rc;
1492
1493 tgt = lmv_find_target(lmv, fid);
1494 if (IS_ERR(tgt))
1495 return PTR_ERR(tgt);
1496
1497 rc = md_getxattr(tgt->ltd_exp, fid, oc, valid, name, input,
1498 input_size, output_size, flags, request);
1499
1500 return rc;
1501 }
1502
1503 static int lmv_setxattr(struct obd_export *exp, const struct lu_fid *fid,
1504 struct obd_capa *oc, obd_valid valid, const char *name,
1505 const char *input, int input_size, int output_size,
1506 int flags, __u32 suppgid,
1507 struct ptlrpc_request **request)
1508 {
1509 struct obd_device *obd = exp->exp_obd;
1510 struct lmv_obd *lmv = &obd->u.lmv;
1511 struct lmv_tgt_desc *tgt;
1512 int rc;
1513
1514 rc = lmv_check_connect(obd);
1515 if (rc)
1516 return rc;
1517
1518 tgt = lmv_find_target(lmv, fid);
1519 if (IS_ERR(tgt))
1520 return PTR_ERR(tgt);
1521
1522 rc = md_setxattr(tgt->ltd_exp, fid, oc, valid, name, input,
1523 input_size, output_size, flags, suppgid,
1524 request);
1525
1526 return rc;
1527 }
1528
1529 static int lmv_getattr(struct obd_export *exp, struct md_op_data *op_data,
1530 struct ptlrpc_request **request)
1531 {
1532 struct obd_device *obd = exp->exp_obd;
1533 struct lmv_obd *lmv = &obd->u.lmv;
1534 struct lmv_tgt_desc *tgt;
1535 int rc;
1536
1537 rc = lmv_check_connect(obd);
1538 if (rc)
1539 return rc;
1540
1541 tgt = lmv_find_target(lmv, &op_data->op_fid1);
1542 if (IS_ERR(tgt))
1543 return PTR_ERR(tgt);
1544
1545 if (op_data->op_flags & MF_GET_MDT_IDX) {
1546 op_data->op_mds = tgt->ltd_idx;
1547 return 0;
1548 }
1549
1550 rc = md_getattr(tgt->ltd_exp, op_data, request);
1551
1552 return rc;
1553 }
1554
1555 static int lmv_null_inode(struct obd_export *exp, const struct lu_fid *fid)
1556 {
1557 struct obd_device *obd = exp->exp_obd;
1558 struct lmv_obd *lmv = &obd->u.lmv;
1559 int i;
1560 int rc;
1561
1562 rc = lmv_check_connect(obd);
1563 if (rc)
1564 return rc;
1565
1566 CDEBUG(D_INODE, "CBDATA for "DFID"\n", PFID(fid));
1567
1568 /*
1569 * With DNE every object can have two locks in different namespaces:
1570 * lookup lock in space of MDT storing direntry and update/open lock in
1571 * space of MDT storing inode.
1572 */
1573 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
1574 if (lmv->tgts[i] == NULL || lmv->tgts[i]->ltd_exp == NULL)
1575 continue;
1576 md_null_inode(lmv->tgts[i]->ltd_exp, fid);
1577 }
1578
1579 return 0;
1580 }
1581
1582 static int lmv_find_cbdata(struct obd_export *exp, const struct lu_fid *fid,
1583 ldlm_iterator_t it, void *data)
1584 {
1585 struct obd_device *obd = exp->exp_obd;
1586 struct lmv_obd *lmv = &obd->u.lmv;
1587 int i;
1588 int rc;
1589
1590 rc = lmv_check_connect(obd);
1591 if (rc)
1592 return rc;
1593
1594 CDEBUG(D_INODE, "CBDATA for "DFID"\n", PFID(fid));
1595
1596 /*
1597 * With DNE every object can have two locks in different namespaces:
1598 * lookup lock in space of MDT storing direntry and update/open lock in
1599 * space of MDT storing inode.
1600 */
1601 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
1602 if (lmv->tgts[i] == NULL || lmv->tgts[i]->ltd_exp == NULL)
1603 continue;
1604 rc = md_find_cbdata(lmv->tgts[i]->ltd_exp, fid, it, data);
1605 if (rc)
1606 return rc;
1607 }
1608
1609 return rc;
1610 }
1611
1612
1613 static int lmv_close(struct obd_export *exp, struct md_op_data *op_data,
1614 struct md_open_data *mod, struct ptlrpc_request **request)
1615 {
1616 struct obd_device *obd = exp->exp_obd;
1617 struct lmv_obd *lmv = &obd->u.lmv;
1618 struct lmv_tgt_desc *tgt;
1619 int rc;
1620
1621 rc = lmv_check_connect(obd);
1622 if (rc)
1623 return rc;
1624
1625 tgt = lmv_find_target(lmv, &op_data->op_fid1);
1626 if (IS_ERR(tgt))
1627 return PTR_ERR(tgt);
1628
1629 CDEBUG(D_INODE, "CLOSE "DFID"\n", PFID(&op_data->op_fid1));
1630 rc = md_close(tgt->ltd_exp, op_data, mod, request);
1631 return rc;
1632 }
1633
1634 struct lmv_tgt_desc
1635 *lmv_locate_mds(struct lmv_obd *lmv, struct md_op_data *op_data,
1636 struct lu_fid *fid)
1637 {
1638 struct lmv_tgt_desc *tgt;
1639
1640 tgt = lmv_find_target(lmv, fid);
1641 if (IS_ERR(tgt))
1642 return tgt;
1643
1644 op_data->op_mds = tgt->ltd_idx;
1645
1646 return tgt;
1647 }
1648
1649 int lmv_create(struct obd_export *exp, struct md_op_data *op_data,
1650 const void *data, int datalen, int mode, __u32 uid,
1651 __u32 gid, cfs_cap_t cap_effective, __u64 rdev,
1652 struct ptlrpc_request **request)
1653 {
1654 struct obd_device *obd = exp->exp_obd;
1655 struct lmv_obd *lmv = &obd->u.lmv;
1656 struct lmv_tgt_desc *tgt;
1657 int rc;
1658
1659 rc = lmv_check_connect(obd);
1660 if (rc)
1661 return rc;
1662
1663 if (!lmv->desc.ld_active_tgt_count)
1664 return -EIO;
1665
1666 tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
1667 if (IS_ERR(tgt))
1668 return PTR_ERR(tgt);
1669
1670 rc = lmv_fid_alloc(exp, &op_data->op_fid2, op_data);
1671 if (rc)
1672 return rc;
1673
1674 CDEBUG(D_INODE, "CREATE '%*s' on "DFID" -> mds #%x\n",
1675 op_data->op_namelen, op_data->op_name, PFID(&op_data->op_fid1),
1676 op_data->op_mds);
1677
1678 op_data->op_flags |= MF_MDC_CANCEL_FID1;
1679 rc = md_create(tgt->ltd_exp, op_data, data, datalen, mode, uid, gid,
1680 cap_effective, rdev, request);
1681
1682 if (rc == 0) {
1683 if (*request == NULL)
1684 return rc;
1685 CDEBUG(D_INODE, "Created - "DFID"\n", PFID(&op_data->op_fid2));
1686 }
1687 return rc;
1688 }
1689
1690 static int lmv_done_writing(struct obd_export *exp,
1691 struct md_op_data *op_data,
1692 struct md_open_data *mod)
1693 {
1694 struct obd_device *obd = exp->exp_obd;
1695 struct lmv_obd *lmv = &obd->u.lmv;
1696 struct lmv_tgt_desc *tgt;
1697 int rc;
1698
1699 rc = lmv_check_connect(obd);
1700 if (rc)
1701 return rc;
1702
1703 tgt = lmv_find_target(lmv, &op_data->op_fid1);
1704 if (IS_ERR(tgt))
1705 return PTR_ERR(tgt);
1706
1707 rc = md_done_writing(tgt->ltd_exp, op_data, mod);
1708 return rc;
1709 }
1710
1711 static int
1712 lmv_enqueue_remote(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
1713 struct lookup_intent *it, struct md_op_data *op_data,
1714 struct lustre_handle *lockh, void *lmm, int lmmsize,
1715 int extra_lock_flags)
1716 {
1717 struct ptlrpc_request *req = it->d.lustre.it_data;
1718 struct obd_device *obd = exp->exp_obd;
1719 struct lmv_obd *lmv = &obd->u.lmv;
1720 struct lustre_handle plock;
1721 struct lmv_tgt_desc *tgt;
1722 struct md_op_data *rdata;
1723 struct lu_fid fid1;
1724 struct mdt_body *body;
1725 int rc = 0;
1726 int pmode;
1727
1728 body = req_capsule_server_get(&req->rq_pill, &RMF_MDT_BODY);
1729 LASSERT(body != NULL);
1730
1731 if (!(body->valid & OBD_MD_MDS))
1732 return 0;
1733
1734 CDEBUG(D_INODE, "REMOTE_ENQUEUE '%s' on "DFID" -> "DFID"\n",
1735 LL_IT2STR(it), PFID(&op_data->op_fid1), PFID(&body->fid1));
1736
1737 /*
1738 * We got LOOKUP lock, but we really need attrs.
1739 */
1740 pmode = it->d.lustre.it_lock_mode;
1741 LASSERT(pmode != 0);
1742 memcpy(&plock, lockh, sizeof(plock));
1743 it->d.lustre.it_lock_mode = 0;
1744 it->d.lustre.it_data = NULL;
1745 fid1 = body->fid1;
1746
1747 it->d.lustre.it_disposition &= ~DISP_ENQ_COMPLETE;
1748 ptlrpc_req_finished(req);
1749
1750 tgt = lmv_find_target(lmv, &fid1);
1751 if (IS_ERR(tgt))
1752 GOTO(out, rc = PTR_ERR(tgt));
1753
1754 OBD_ALLOC_PTR(rdata);
1755 if (rdata == NULL)
1756 GOTO(out, rc = -ENOMEM);
1757
1758 rdata->op_fid1 = fid1;
1759 rdata->op_bias = MDS_CROSS_REF;
1760
1761 rc = md_enqueue(tgt->ltd_exp, einfo, it, rdata, lockh,
1762 lmm, lmmsize, NULL, extra_lock_flags);
1763 OBD_FREE_PTR(rdata);
1764 out:
1765 ldlm_lock_decref(&plock, pmode);
1766 return rc;
1767 }
1768
1769 static int
1770 lmv_enqueue(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
1771 struct lookup_intent *it, struct md_op_data *op_data,
1772 struct lustre_handle *lockh, void *lmm, int lmmsize,
1773 struct ptlrpc_request **req, __u64 extra_lock_flags)
1774 {
1775 struct obd_device *obd = exp->exp_obd;
1776 struct lmv_obd *lmv = &obd->u.lmv;
1777 struct lmv_tgt_desc *tgt;
1778 int rc;
1779
1780 rc = lmv_check_connect(obd);
1781 if (rc)
1782 return rc;
1783
1784 CDEBUG(D_INODE, "ENQUEUE '%s' on "DFID"\n",
1785 LL_IT2STR(it), PFID(&op_data->op_fid1));
1786
1787 tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
1788 if (IS_ERR(tgt))
1789 return PTR_ERR(tgt);
1790
1791 CDEBUG(D_INODE, "ENQUEUE '%s' on "DFID" -> mds #%d\n",
1792 LL_IT2STR(it), PFID(&op_data->op_fid1), tgt->ltd_idx);
1793
1794 rc = md_enqueue(tgt->ltd_exp, einfo, it, op_data, lockh,
1795 lmm, lmmsize, req, extra_lock_flags);
1796
1797 if (rc == 0 && it && it->it_op == IT_OPEN) {
1798 rc = lmv_enqueue_remote(exp, einfo, it, op_data, lockh,
1799 lmm, lmmsize, extra_lock_flags);
1800 }
1801 return rc;
1802 }
1803
1804 static int
1805 lmv_getattr_name(struct obd_export *exp,struct md_op_data *op_data,
1806 struct ptlrpc_request **request)
1807 {
1808 struct ptlrpc_request *req = NULL;
1809 struct obd_device *obd = exp->exp_obd;
1810 struct lmv_obd *lmv = &obd->u.lmv;
1811 struct lmv_tgt_desc *tgt;
1812 struct mdt_body *body;
1813 int rc;
1814
1815 rc = lmv_check_connect(obd);
1816 if (rc)
1817 return rc;
1818
1819 tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
1820 if (IS_ERR(tgt))
1821 return PTR_ERR(tgt);
1822
1823 CDEBUG(D_INODE, "GETATTR_NAME for %*s on "DFID" -> mds #%d\n",
1824 op_data->op_namelen, op_data->op_name, PFID(&op_data->op_fid1),
1825 tgt->ltd_idx);
1826
1827 rc = md_getattr_name(tgt->ltd_exp, op_data, request);
1828 if (rc != 0)
1829 return rc;
1830
1831 body = req_capsule_server_get(&(*request)->rq_pill,
1832 &RMF_MDT_BODY);
1833 LASSERT(body != NULL);
1834
1835 if (body->valid & OBD_MD_MDS) {
1836 struct lu_fid rid = body->fid1;
1837 CDEBUG(D_INODE, "Request attrs for "DFID"\n",
1838 PFID(&rid));
1839
1840 tgt = lmv_find_target(lmv, &rid);
1841 if (IS_ERR(tgt)) {
1842 ptlrpc_req_finished(*request);
1843 return PTR_ERR(tgt);
1844 }
1845
1846 op_data->op_fid1 = rid;
1847 op_data->op_valid |= OBD_MD_FLCROSSREF;
1848 op_data->op_namelen = 0;
1849 op_data->op_name = NULL;
1850 rc = md_getattr_name(tgt->ltd_exp, op_data, &req);
1851 ptlrpc_req_finished(*request);
1852 *request = req;
1853 }
1854
1855 return rc;
1856 }
1857
1858 #define md_op_data_fid(op_data, fl) \
1859 (fl == MF_MDC_CANCEL_FID1 ? &op_data->op_fid1 : \
1860 fl == MF_MDC_CANCEL_FID2 ? &op_data->op_fid2 : \
1861 fl == MF_MDC_CANCEL_FID3 ? &op_data->op_fid3 : \
1862 fl == MF_MDC_CANCEL_FID4 ? &op_data->op_fid4 : \
1863 NULL)
1864
1865 static int lmv_early_cancel(struct obd_export *exp, struct md_op_data *op_data,
1866 int op_tgt, ldlm_mode_t mode, int bits, int flag)
1867 {
1868 struct lu_fid *fid = md_op_data_fid(op_data, flag);
1869 struct obd_device *obd = exp->exp_obd;
1870 struct lmv_obd *lmv = &obd->u.lmv;
1871 struct lmv_tgt_desc *tgt;
1872 ldlm_policy_data_t policy = {{0}};
1873 int rc = 0;
1874
1875 if (!fid_is_sane(fid))
1876 return 0;
1877
1878 tgt = lmv_find_target(lmv, fid);
1879 if (IS_ERR(tgt))
1880 return PTR_ERR(tgt);
1881
1882 if (tgt->ltd_idx != op_tgt) {
1883 CDEBUG(D_INODE, "EARLY_CANCEL on "DFID"\n", PFID(fid));
1884 policy.l_inodebits.bits = bits;
1885 rc = md_cancel_unused(tgt->ltd_exp, fid, &policy,
1886 mode, LCF_ASYNC, NULL);
1887 } else {
1888 CDEBUG(D_INODE,
1889 "EARLY_CANCEL skip operation target %d on "DFID"\n",
1890 op_tgt, PFID(fid));
1891 op_data->op_flags |= flag;
1892 rc = 0;
1893 }
1894
1895 return rc;
1896 }
1897
1898 /*
1899 * llite passes fid of an target inode in op_data->op_fid1 and id of directory in
1900 * op_data->op_fid2
1901 */
1902 static int lmv_link(struct obd_export *exp, struct md_op_data *op_data,
1903 struct ptlrpc_request **request)
1904 {
1905 struct obd_device *obd = exp->exp_obd;
1906 struct lmv_obd *lmv = &obd->u.lmv;
1907 struct lmv_tgt_desc *tgt;
1908 int rc;
1909
1910 rc = lmv_check_connect(obd);
1911 if (rc)
1912 return rc;
1913
1914 LASSERT(op_data->op_namelen != 0);
1915
1916 CDEBUG(D_INODE, "LINK "DFID":%*s to "DFID"\n",
1917 PFID(&op_data->op_fid2), op_data->op_namelen,
1918 op_data->op_name, PFID(&op_data->op_fid1));
1919
1920 op_data->op_fsuid = from_kuid(&init_user_ns, current_fsuid());
1921 op_data->op_fsgid = from_kgid(&init_user_ns, current_fsgid());
1922 op_data->op_cap = cfs_curproc_cap_pack();
1923 tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid2);
1924 if (IS_ERR(tgt))
1925 return PTR_ERR(tgt);
1926
1927 /*
1928 * Cancel UPDATE lock on child (fid1).
1929 */
1930 op_data->op_flags |= MF_MDC_CANCEL_FID2;
1931 rc = lmv_early_cancel(exp, op_data, tgt->ltd_idx, LCK_EX,
1932 MDS_INODELOCK_UPDATE, MF_MDC_CANCEL_FID1);
1933 if (rc != 0)
1934 return rc;
1935
1936 rc = md_link(tgt->ltd_exp, op_data, request);
1937
1938 return rc;
1939 }
1940
1941 static int lmv_rename(struct obd_export *exp, struct md_op_data *op_data,
1942 const char *old, int oldlen, const char *new, int newlen,
1943 struct ptlrpc_request **request)
1944 {
1945 struct obd_device *obd = exp->exp_obd;
1946 struct lmv_obd *lmv = &obd->u.lmv;
1947 struct lmv_tgt_desc *src_tgt;
1948 struct lmv_tgt_desc *tgt_tgt;
1949 int rc;
1950
1951 LASSERT(oldlen != 0);
1952
1953 CDEBUG(D_INODE, "RENAME %*s in "DFID" to %*s in "DFID"\n",
1954 oldlen, old, PFID(&op_data->op_fid1),
1955 newlen, new, PFID(&op_data->op_fid2));
1956
1957 rc = lmv_check_connect(obd);
1958 if (rc)
1959 return rc;
1960
1961 op_data->op_fsuid = from_kuid(&init_user_ns, current_fsuid());
1962 op_data->op_fsgid = from_kgid(&init_user_ns, current_fsgid());
1963 op_data->op_cap = cfs_curproc_cap_pack();
1964 src_tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
1965 if (IS_ERR(src_tgt))
1966 return PTR_ERR(src_tgt);
1967
1968 tgt_tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid2);
1969 if (IS_ERR(tgt_tgt))
1970 return PTR_ERR(tgt_tgt);
1971 /*
1972 * LOOKUP lock on src child (fid3) should also be cancelled for
1973 * src_tgt in mdc_rename.
1974 */
1975 op_data->op_flags |= MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID3;
1976
1977 /*
1978 * Cancel UPDATE locks on tgt parent (fid2), tgt_tgt is its
1979 * own target.
1980 */
1981 rc = lmv_early_cancel(exp, op_data, src_tgt->ltd_idx,
1982 LCK_EX, MDS_INODELOCK_UPDATE,
1983 MF_MDC_CANCEL_FID2);
1984
1985 /*
1986 * Cancel LOOKUP locks on tgt child (fid4) for parent tgt_tgt.
1987 */
1988 if (rc == 0) {
1989 rc = lmv_early_cancel(exp, op_data, src_tgt->ltd_idx,
1990 LCK_EX, MDS_INODELOCK_LOOKUP,
1991 MF_MDC_CANCEL_FID4);
1992 }
1993
1994 /*
1995 * Cancel all the locks on tgt child (fid4).
1996 */
1997 if (rc == 0)
1998 rc = lmv_early_cancel(exp, op_data, src_tgt->ltd_idx,
1999 LCK_EX, MDS_INODELOCK_FULL,
2000 MF_MDC_CANCEL_FID4);
2001
2002 if (rc == 0)
2003 rc = md_rename(src_tgt->ltd_exp, op_data, old, oldlen,
2004 new, newlen, request);
2005 return rc;
2006 }
2007
2008 static int lmv_setattr(struct obd_export *exp, struct md_op_data *op_data,
2009 void *ea, int ealen, void *ea2, int ea2len,
2010 struct ptlrpc_request **request,
2011 struct md_open_data **mod)
2012 {
2013 struct obd_device *obd = exp->exp_obd;
2014 struct lmv_obd *lmv = &obd->u.lmv;
2015 struct lmv_tgt_desc *tgt;
2016 int rc = 0;
2017
2018 rc = lmv_check_connect(obd);
2019 if (rc)
2020 return rc;
2021
2022 CDEBUG(D_INODE, "SETATTR for "DFID", valid 0x%x\n",
2023 PFID(&op_data->op_fid1), op_data->op_attr.ia_valid);
2024
2025 op_data->op_flags |= MF_MDC_CANCEL_FID1;
2026 tgt = lmv_find_target(lmv, &op_data->op_fid1);
2027 if (IS_ERR(tgt))
2028 return PTR_ERR(tgt);
2029
2030 rc = md_setattr(tgt->ltd_exp, op_data, ea, ealen, ea2,
2031 ea2len, request, mod);
2032
2033 return rc;
2034 }
2035
2036 static int lmv_sync(struct obd_export *exp, const struct lu_fid *fid,
2037 struct obd_capa *oc, struct ptlrpc_request **request)
2038 {
2039 struct obd_device *obd = exp->exp_obd;
2040 struct lmv_obd *lmv = &obd->u.lmv;
2041 struct lmv_tgt_desc *tgt;
2042 int rc;
2043
2044 rc = lmv_check_connect(obd);
2045 if (rc)
2046 return rc;
2047
2048 tgt = lmv_find_target(lmv, fid);
2049 if (IS_ERR(tgt))
2050 return PTR_ERR(tgt);
2051
2052 rc = md_sync(tgt->ltd_exp, fid, oc, request);
2053 return rc;
2054 }
2055
2056 /*
2057 * Adjust a set of pages, each page containing an array of lu_dirpages,
2058 * so that each page can be used as a single logical lu_dirpage.
2059 *
2060 * A lu_dirpage is laid out as follows, where s = ldp_hash_start,
2061 * e = ldp_hash_end, f = ldp_flags, p = padding, and each "ent" is a
2062 * struct lu_dirent. It has size up to LU_PAGE_SIZE. The ldp_hash_end
2063 * value is used as a cookie to request the next lu_dirpage in a
2064 * directory listing that spans multiple pages (two in this example):
2065 * ________
2066 * | |
2067 * .|--------v------- -----.
2068 * |s|e|f|p|ent|ent| ... |ent|
2069 * '--|-------------- -----' Each CFS_PAGE contains a single
2070 * '------. lu_dirpage.
2071 * .---------v------- -----.
2072 * |s|e|f|p|ent| 0 | ... | 0 |
2073 * '----------------- -----'
2074 *
2075 * However, on hosts where the native VM page size (PAGE_CACHE_SIZE) is
2076 * larger than LU_PAGE_SIZE, a single host page may contain multiple
2077 * lu_dirpages. After reading the lu_dirpages from the MDS, the
2078 * ldp_hash_end of the first lu_dirpage refers to the one immediately
2079 * after it in the same CFS_PAGE (arrows simplified for brevity, but
2080 * in general e0==s1, e1==s2, etc.):
2081 *
2082 * .-------------------- -----.
2083 * |s0|e0|f0|p|ent|ent| ... |ent|
2084 * |---v---------------- -----|
2085 * |s1|e1|f1|p|ent|ent| ... |ent|
2086 * |---v---------------- -----| Here, each CFS_PAGE contains
2087 * ... multiple lu_dirpages.
2088 * |---v---------------- -----|
2089 * |s'|e'|f'|p|ent|ent| ... |ent|
2090 * '---|---------------- -----'
2091 * v
2092 * .----------------------------.
2093 * | next CFS_PAGE |
2094 *
2095 * This structure is transformed into a single logical lu_dirpage as follows:
2096 *
2097 * - Replace e0 with e' so the request for the next lu_dirpage gets the page
2098 * labeled 'next CFS_PAGE'.
2099 *
2100 * - Copy the LDF_COLLIDE flag from f' to f0 to correctly reflect whether
2101 * a hash collision with the next page exists.
2102 *
2103 * - Adjust the lde_reclen of the ending entry of each lu_dirpage to span
2104 * to the first entry of the next lu_dirpage.
2105 */
2106 #if PAGE_CACHE_SIZE > LU_PAGE_SIZE
2107 static void lmv_adjust_dirpages(struct page **pages, int ncfspgs, int nlupgs)
2108 {
2109 int i;
2110
2111 for (i = 0; i < ncfspgs; i++) {
2112 struct lu_dirpage *dp = kmap(pages[i]);
2113 struct lu_dirpage *first = dp;
2114 struct lu_dirent *end_dirent = NULL;
2115 struct lu_dirent *ent;
2116 __u64 hash_end = dp->ldp_hash_end;
2117 __u32 flags = dp->ldp_flags;
2118
2119 while (--nlupgs > 0) {
2120 ent = lu_dirent_start(dp);
2121 for (end_dirent = ent; ent != NULL;
2122 end_dirent = ent, ent = lu_dirent_next(ent));
2123
2124 /* Advance dp to next lu_dirpage. */
2125 dp = (struct lu_dirpage *)((char *)dp + LU_PAGE_SIZE);
2126
2127 /* Check if we've reached the end of the CFS_PAGE. */
2128 if (!((unsigned long)dp & ~CFS_PAGE_MASK))
2129 break;
2130
2131 /* Save the hash and flags of this lu_dirpage. */
2132 hash_end = dp->ldp_hash_end;
2133 flags = dp->ldp_flags;
2134
2135 /* Check if lu_dirpage contains no entries. */
2136 if (!end_dirent)
2137 break;
2138
2139 /* Enlarge the end entry lde_reclen from 0 to
2140 * first entry of next lu_dirpage. */
2141 LASSERT(le16_to_cpu(end_dirent->lde_reclen) == 0);
2142 end_dirent->lde_reclen =
2143 cpu_to_le16((char *)(dp->ldp_entries) -
2144 (char *)end_dirent);
2145 }
2146
2147 first->ldp_hash_end = hash_end;
2148 first->ldp_flags &= ~cpu_to_le32(LDF_COLLIDE);
2149 first->ldp_flags |= flags & cpu_to_le32(LDF_COLLIDE);
2150
2151 kunmap(pages[i]);
2152 }
2153 LASSERTF(nlupgs == 0, "left = %d", nlupgs);
2154 }
2155 #else
2156 #define lmv_adjust_dirpages(pages, ncfspgs, nlupgs) do {} while (0)
2157 #endif /* PAGE_CACHE_SIZE > LU_PAGE_SIZE */
2158
2159 static int lmv_readpage(struct obd_export *exp, struct md_op_data *op_data,
2160 struct page **pages, struct ptlrpc_request **request)
2161 {
2162 struct obd_device *obd = exp->exp_obd;
2163 struct lmv_obd *lmv = &obd->u.lmv;
2164 __u64 offset = op_data->op_offset;
2165 int rc;
2166 int ncfspgs; /* pages read in PAGE_CACHE_SIZE */
2167 int nlupgs; /* pages read in LU_PAGE_SIZE */
2168 struct lmv_tgt_desc *tgt;
2169
2170 rc = lmv_check_connect(obd);
2171 if (rc)
2172 return rc;
2173
2174 CDEBUG(D_INODE, "READPAGE at "LPX64" from "DFID"\n",
2175 offset, PFID(&op_data->op_fid1));
2176
2177 tgt = lmv_find_target(lmv, &op_data->op_fid1);
2178 if (IS_ERR(tgt))
2179 return PTR_ERR(tgt);
2180
2181 rc = md_readpage(tgt->ltd_exp, op_data, pages, request);
2182 if (rc != 0)
2183 return rc;
2184
2185 ncfspgs = ((*request)->rq_bulk->bd_nob_transferred + PAGE_CACHE_SIZE - 1)
2186 >> PAGE_CACHE_SHIFT;
2187 nlupgs = (*request)->rq_bulk->bd_nob_transferred >> LU_PAGE_SHIFT;
2188 LASSERT(!((*request)->rq_bulk->bd_nob_transferred & ~LU_PAGE_MASK));
2189 LASSERT(ncfspgs > 0 && ncfspgs <= op_data->op_npages);
2190
2191 CDEBUG(D_INODE, "read %d(%d)/%d pages\n", ncfspgs, nlupgs,
2192 op_data->op_npages);
2193
2194 lmv_adjust_dirpages(pages, ncfspgs, nlupgs);
2195
2196 return rc;
2197 }
2198
2199 static int lmv_unlink(struct obd_export *exp, struct md_op_data *op_data,
2200 struct ptlrpc_request **request)
2201 {
2202 struct obd_device *obd = exp->exp_obd;
2203 struct lmv_obd *lmv = &obd->u.lmv;
2204 struct lmv_tgt_desc *tgt = NULL;
2205 struct mdt_body *body;
2206 int rc;
2207
2208 rc = lmv_check_connect(obd);
2209 if (rc)
2210 return rc;
2211 retry:
2212 /* Send unlink requests to the MDT where the child is located */
2213 if (likely(!fid_is_zero(&op_data->op_fid2)))
2214 tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid2);
2215 else
2216 tgt = lmv_locate_mds(lmv, op_data, &op_data->op_fid1);
2217 if (IS_ERR(tgt))
2218 return PTR_ERR(tgt);
2219
2220 op_data->op_fsuid = from_kuid(&init_user_ns, current_fsuid());
2221 op_data->op_fsgid = from_kgid(&init_user_ns, current_fsgid());
2222 op_data->op_cap = cfs_curproc_cap_pack();
2223
2224 /*
2225 * If child's fid is given, cancel unused locks for it if it is from
2226 * another export than parent.
2227 *
2228 * LOOKUP lock for child (fid3) should also be cancelled on parent
2229 * tgt_tgt in mdc_unlink().
2230 */
2231 op_data->op_flags |= MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID3;
2232
2233 /*
2234 * Cancel FULL locks on child (fid3).
2235 */
2236 rc = lmv_early_cancel(exp, op_data, tgt->ltd_idx, LCK_EX,
2237 MDS_INODELOCK_FULL, MF_MDC_CANCEL_FID3);
2238
2239 if (rc != 0)
2240 return rc;
2241
2242 CDEBUG(D_INODE, "unlink with fid="DFID"/"DFID" -> mds #%d\n",
2243 PFID(&op_data->op_fid1), PFID(&op_data->op_fid2), tgt->ltd_idx);
2244
2245 rc = md_unlink(tgt->ltd_exp, op_data, request);
2246 if (rc != 0 && rc != -EREMOTE)
2247 return rc;
2248
2249 body = req_capsule_server_get(&(*request)->rq_pill, &RMF_MDT_BODY);
2250 if (body == NULL)
2251 return -EPROTO;
2252
2253 /* Not cross-ref case, just get out of here. */
2254 if (likely(!(body->valid & OBD_MD_MDS)))
2255 return 0;
2256
2257 CDEBUG(D_INODE, "%s: try unlink to another MDT for "DFID"\n",
2258 exp->exp_obd->obd_name, PFID(&body->fid1));
2259
2260 /* This is a remote object, try remote MDT, Note: it may
2261 * try more than 1 time here, Considering following case
2262 * /mnt/lustre is root on MDT0, remote1 is on MDT1
2263 * 1. Initially A does not know where remote1 is, it send
2264 * unlink RPC to MDT0, MDT0 return -EREMOTE, it will
2265 * resend unlink RPC to MDT1 (retry 1st time).
2266 *
2267 * 2. During the unlink RPC in flight,
2268 * client B mv /mnt/lustre/remote1 /mnt/lustre/remote2
2269 * and create new remote1, but on MDT0
2270 *
2271 * 3. MDT1 get unlink RPC(from A), then do remote lock on
2272 * /mnt/lustre, then lookup get fid of remote1, and find
2273 * it is remote dir again, and replay -EREMOTE again.
2274 *
2275 * 4. Then A will resend unlink RPC to MDT0. (retry 2nd times).
2276 *
2277 * In theory, it might try unlimited time here, but it should
2278 * be very rare case. */
2279 op_data->op_fid2 = body->fid1;
2280 ptlrpc_req_finished(*request);
2281 *request = NULL;
2282
2283 goto retry;
2284 }
2285
2286 static int lmv_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
2287 {
2288 struct lmv_obd *lmv = &obd->u.lmv;
2289 int rc = 0;
2290
2291 switch (stage) {
2292 case OBD_CLEANUP_EARLY:
2293 /* XXX: here should be calling obd_precleanup() down to
2294 * stack. */
2295 break;
2296 case OBD_CLEANUP_EXPORTS:
2297 fld_client_proc_fini(&lmv->lmv_fld);
2298 lprocfs_obd_cleanup(obd);
2299 break;
2300 default:
2301 break;
2302 }
2303 return rc;
2304 }
2305
2306 static int lmv_get_info(const struct lu_env *env, struct obd_export *exp,
2307 __u32 keylen, void *key, __u32 *vallen, void *val,
2308 struct lov_stripe_md *lsm)
2309 {
2310 struct obd_device *obd;
2311 struct lmv_obd *lmv;
2312 int rc = 0;
2313
2314 obd = class_exp2obd(exp);
2315 if (obd == NULL) {
2316 CDEBUG(D_IOCTL, "Invalid client cookie "LPX64"\n",
2317 exp->exp_handle.h_cookie);
2318 return -EINVAL;
2319 }
2320
2321 lmv = &obd->u.lmv;
2322 if (keylen >= strlen("remote_flag") && !strcmp(key, "remote_flag")) {
2323 struct lmv_tgt_desc *tgt;
2324 int i;
2325
2326 rc = lmv_check_connect(obd);
2327 if (rc)
2328 return rc;
2329
2330 LASSERT(*vallen == sizeof(__u32));
2331 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2332 tgt = lmv->tgts[i];
2333 /*
2334 * All tgts should be connected when this gets called.
2335 */
2336 if (tgt == NULL || tgt->ltd_exp == NULL)
2337 continue;
2338
2339 if (!obd_get_info(env, tgt->ltd_exp, keylen, key,
2340 vallen, val, NULL))
2341 return 0;
2342 }
2343 return -EINVAL;
2344 } else if (KEY_IS(KEY_MAX_EASIZE) || KEY_IS(KEY_CONN_DATA)) {
2345 rc = lmv_check_connect(obd);
2346 if (rc)
2347 return rc;
2348
2349 /*
2350 * Forwarding this request to first MDS, it should know LOV
2351 * desc.
2352 */
2353 rc = obd_get_info(env, lmv->tgts[0]->ltd_exp, keylen, key,
2354 vallen, val, NULL);
2355 if (!rc && KEY_IS(KEY_CONN_DATA))
2356 exp->exp_connect_data = *(struct obd_connect_data *)val;
2357 return rc;
2358 } else if (KEY_IS(KEY_TGT_COUNT)) {
2359 *((int *)val) = lmv->desc.ld_tgt_count;
2360 return 0;
2361 }
2362
2363 CDEBUG(D_IOCTL, "Invalid key\n");
2364 return -EINVAL;
2365 }
2366
2367 int lmv_set_info_async(const struct lu_env *env, struct obd_export *exp,
2368 obd_count keylen, void *key, obd_count vallen,
2369 void *val, struct ptlrpc_request_set *set)
2370 {
2371 struct lmv_tgt_desc *tgt;
2372 struct obd_device *obd;
2373 struct lmv_obd *lmv;
2374 int rc = 0;
2375
2376 obd = class_exp2obd(exp);
2377 if (obd == NULL) {
2378 CDEBUG(D_IOCTL, "Invalid client cookie "LPX64"\n",
2379 exp->exp_handle.h_cookie);
2380 return -EINVAL;
2381 }
2382 lmv = &obd->u.lmv;
2383
2384 if (KEY_IS(KEY_READ_ONLY) || KEY_IS(KEY_FLUSH_CTX)) {
2385 int i, err = 0;
2386
2387 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2388 tgt = lmv->tgts[i];
2389
2390 if (tgt == NULL || tgt->ltd_exp == NULL)
2391 continue;
2392
2393 err = obd_set_info_async(env, tgt->ltd_exp,
2394 keylen, key, vallen, val, set);
2395 if (err && rc == 0)
2396 rc = err;
2397 }
2398
2399 return rc;
2400 }
2401
2402 return -EINVAL;
2403 }
2404
2405 int lmv_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
2406 struct lov_stripe_md *lsm)
2407 {
2408 struct obd_device *obd = class_exp2obd(exp);
2409 struct lmv_obd *lmv = &obd->u.lmv;
2410 struct lmv_stripe_md *meap;
2411 struct lmv_stripe_md *lsmp;
2412 int mea_size;
2413 int i;
2414
2415 mea_size = lmv_get_easize(lmv);
2416 if (!lmmp)
2417 return mea_size;
2418
2419 if (*lmmp && !lsm) {
2420 OBD_FREE_LARGE(*lmmp, mea_size);
2421 *lmmp = NULL;
2422 return 0;
2423 }
2424
2425 if (*lmmp == NULL) {
2426 OBD_ALLOC_LARGE(*lmmp, mea_size);
2427 if (*lmmp == NULL)
2428 return -ENOMEM;
2429 }
2430
2431 if (!lsm)
2432 return mea_size;
2433
2434 lsmp = (struct lmv_stripe_md *)lsm;
2435 meap = (struct lmv_stripe_md *)*lmmp;
2436
2437 if (lsmp->mea_magic != MEA_MAGIC_LAST_CHAR &&
2438 lsmp->mea_magic != MEA_MAGIC_ALL_CHARS)
2439 return -EINVAL;
2440
2441 meap->mea_magic = cpu_to_le32(lsmp->mea_magic);
2442 meap->mea_count = cpu_to_le32(lsmp->mea_count);
2443 meap->mea_master = cpu_to_le32(lsmp->mea_master);
2444
2445 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2446 meap->mea_ids[i] = lsmp->mea_ids[i];
2447 fid_cpu_to_le(&meap->mea_ids[i], &lsmp->mea_ids[i]);
2448 }
2449
2450 return mea_size;
2451 }
2452
2453 int lmv_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
2454 struct lov_mds_md *lmm, int lmm_size)
2455 {
2456 struct obd_device *obd = class_exp2obd(exp);
2457 struct lmv_stripe_md **tmea = (struct lmv_stripe_md **)lsmp;
2458 struct lmv_stripe_md *mea = (struct lmv_stripe_md *)lmm;
2459 struct lmv_obd *lmv = &obd->u.lmv;
2460 int mea_size;
2461 int i;
2462 __u32 magic;
2463
2464 mea_size = lmv_get_easize(lmv);
2465 if (lsmp == NULL)
2466 return mea_size;
2467
2468 if (*lsmp != NULL && lmm == NULL) {
2469 OBD_FREE_LARGE(*tmea, mea_size);
2470 *lsmp = NULL;
2471 return 0;
2472 }
2473
2474 LASSERT(mea_size == lmm_size);
2475
2476 OBD_ALLOC_LARGE(*tmea, mea_size);
2477 if (*tmea == NULL)
2478 return -ENOMEM;
2479
2480 if (!lmm)
2481 return mea_size;
2482
2483 if (mea->mea_magic == MEA_MAGIC_LAST_CHAR ||
2484 mea->mea_magic == MEA_MAGIC_ALL_CHARS ||
2485 mea->mea_magic == MEA_MAGIC_HASH_SEGMENT)
2486 {
2487 magic = le32_to_cpu(mea->mea_magic);
2488 } else {
2489 /*
2490 * Old mea is not handled here.
2491 */
2492 CERROR("Old not supportable EA is found\n");
2493 LBUG();
2494 }
2495
2496 (*tmea)->mea_magic = magic;
2497 (*tmea)->mea_count = le32_to_cpu(mea->mea_count);
2498 (*tmea)->mea_master = le32_to_cpu(mea->mea_master);
2499
2500 for (i = 0; i < (*tmea)->mea_count; i++) {
2501 (*tmea)->mea_ids[i] = mea->mea_ids[i];
2502 fid_le_to_cpu(&(*tmea)->mea_ids[i], &(*tmea)->mea_ids[i]);
2503 }
2504 return mea_size;
2505 }
2506
2507 static int lmv_cancel_unused(struct obd_export *exp, const struct lu_fid *fid,
2508 ldlm_policy_data_t *policy, ldlm_mode_t mode,
2509 ldlm_cancel_flags_t flags, void *opaque)
2510 {
2511 struct obd_device *obd = exp->exp_obd;
2512 struct lmv_obd *lmv = &obd->u.lmv;
2513 int rc = 0;
2514 int err;
2515 int i;
2516
2517 LASSERT(fid != NULL);
2518
2519 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2520 if (lmv->tgts[i] == NULL || lmv->tgts[i]->ltd_exp == NULL ||
2521 lmv->tgts[i]->ltd_active == 0)
2522 continue;
2523
2524 err = md_cancel_unused(lmv->tgts[i]->ltd_exp, fid,
2525 policy, mode, flags, opaque);
2526 if (!rc)
2527 rc = err;
2528 }
2529 return rc;
2530 }
2531
2532 int lmv_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data,
2533 __u64 *bits)
2534 {
2535 struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
2536 int rc;
2537
2538 rc = md_set_lock_data(lmv->tgts[0]->ltd_exp, lockh, data, bits);
2539 return rc;
2540 }
2541
2542 ldlm_mode_t lmv_lock_match(struct obd_export *exp, __u64 flags,
2543 const struct lu_fid *fid, ldlm_type_t type,
2544 ldlm_policy_data_t *policy, ldlm_mode_t mode,
2545 struct lustre_handle *lockh)
2546 {
2547 struct obd_device *obd = exp->exp_obd;
2548 struct lmv_obd *lmv = &obd->u.lmv;
2549 ldlm_mode_t rc;
2550 int i;
2551
2552 CDEBUG(D_INODE, "Lock match for "DFID"\n", PFID(fid));
2553
2554 /*
2555 * With CMD every object can have two locks in different namespaces:
2556 * lookup lock in space of mds storing direntry and update/open lock in
2557 * space of mds storing inode. Thus we check all targets, not only that
2558 * one fid was created in.
2559 */
2560 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2561 if (lmv->tgts[i] == NULL ||
2562 lmv->tgts[i]->ltd_exp == NULL ||
2563 lmv->tgts[i]->ltd_active == 0)
2564 continue;
2565
2566 rc = md_lock_match(lmv->tgts[i]->ltd_exp, flags, fid,
2567 type, policy, mode, lockh);
2568 if (rc)
2569 return rc;
2570 }
2571
2572 return 0;
2573 }
2574
2575 int lmv_get_lustre_md(struct obd_export *exp, struct ptlrpc_request *req,
2576 struct obd_export *dt_exp, struct obd_export *md_exp,
2577 struct lustre_md *md)
2578 {
2579 struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
2580
2581 return md_get_lustre_md(lmv->tgts[0]->ltd_exp, req, dt_exp, md_exp, md);
2582 }
2583
2584 int lmv_free_lustre_md(struct obd_export *exp, struct lustre_md *md)
2585 {
2586 struct obd_device *obd = exp->exp_obd;
2587 struct lmv_obd *lmv = &obd->u.lmv;
2588
2589 if (md->mea)
2590 obd_free_memmd(exp, (void *)&md->mea);
2591 return md_free_lustre_md(lmv->tgts[0]->ltd_exp, md);
2592 }
2593
2594 int lmv_set_open_replay_data(struct obd_export *exp,
2595 struct obd_client_handle *och,
2596 struct ptlrpc_request *open_req)
2597 {
2598 struct obd_device *obd = exp->exp_obd;
2599 struct lmv_obd *lmv = &obd->u.lmv;
2600 struct lmv_tgt_desc *tgt;
2601
2602 tgt = lmv_find_target(lmv, &och->och_fid);
2603 if (IS_ERR(tgt))
2604 return PTR_ERR(tgt);
2605
2606 return md_set_open_replay_data(tgt->ltd_exp, och, open_req);
2607 }
2608
2609 int lmv_clear_open_replay_data(struct obd_export *exp,
2610 struct obd_client_handle *och)
2611 {
2612 struct obd_device *obd = exp->exp_obd;
2613 struct lmv_obd *lmv = &obd->u.lmv;
2614 struct lmv_tgt_desc *tgt;
2615
2616 tgt = lmv_find_target(lmv, &och->och_fid);
2617 if (IS_ERR(tgt))
2618 return PTR_ERR(tgt);
2619
2620 return md_clear_open_replay_data(tgt->ltd_exp, och);
2621 }
2622
2623 static int lmv_get_remote_perm(struct obd_export *exp,
2624 const struct lu_fid *fid,
2625 struct obd_capa *oc, __u32 suppgid,
2626 struct ptlrpc_request **request)
2627 {
2628 struct obd_device *obd = exp->exp_obd;
2629 struct lmv_obd *lmv = &obd->u.lmv;
2630 struct lmv_tgt_desc *tgt;
2631 int rc;
2632
2633 rc = lmv_check_connect(obd);
2634 if (rc)
2635 return rc;
2636
2637 tgt = lmv_find_target(lmv, fid);
2638 if (IS_ERR(tgt))
2639 return PTR_ERR(tgt);
2640
2641 rc = md_get_remote_perm(tgt->ltd_exp, fid, oc, suppgid, request);
2642 return rc;
2643 }
2644
2645 static int lmv_renew_capa(struct obd_export *exp, struct obd_capa *oc,
2646 renew_capa_cb_t cb)
2647 {
2648 struct obd_device *obd = exp->exp_obd;
2649 struct lmv_obd *lmv = &obd->u.lmv;
2650 struct lmv_tgt_desc *tgt;
2651 int rc;
2652
2653 rc = lmv_check_connect(obd);
2654 if (rc)
2655 return rc;
2656
2657 tgt = lmv_find_target(lmv, &oc->c_capa.lc_fid);
2658 if (IS_ERR(tgt))
2659 return PTR_ERR(tgt);
2660
2661 rc = md_renew_capa(tgt->ltd_exp, oc, cb);
2662 return rc;
2663 }
2664
2665 int lmv_unpack_capa(struct obd_export *exp, struct ptlrpc_request *req,
2666 const struct req_msg_field *field, struct obd_capa **oc)
2667 {
2668 struct lmv_obd *lmv = &exp->exp_obd->u.lmv;
2669
2670 return md_unpack_capa(lmv->tgts[0]->ltd_exp, req, field, oc);
2671 }
2672
2673 int lmv_intent_getattr_async(struct obd_export *exp,
2674 struct md_enqueue_info *minfo,
2675 struct ldlm_enqueue_info *einfo)
2676 {
2677 struct md_op_data *op_data = &minfo->mi_data;
2678 struct obd_device *obd = exp->exp_obd;
2679 struct lmv_obd *lmv = &obd->u.lmv;
2680 struct lmv_tgt_desc *tgt = NULL;
2681 int rc;
2682
2683 rc = lmv_check_connect(obd);
2684 if (rc)
2685 return rc;
2686
2687 tgt = lmv_find_target(lmv, &op_data->op_fid1);
2688 if (IS_ERR(tgt))
2689 return PTR_ERR(tgt);
2690
2691 rc = md_intent_getattr_async(tgt->ltd_exp, minfo, einfo);
2692 return rc;
2693 }
2694
2695 int lmv_revalidate_lock(struct obd_export *exp, struct lookup_intent *it,
2696 struct lu_fid *fid, __u64 *bits)
2697 {
2698 struct obd_device *obd = exp->exp_obd;
2699 struct lmv_obd *lmv = &obd->u.lmv;
2700 struct lmv_tgt_desc *tgt;
2701 int rc;
2702
2703 rc = lmv_check_connect(obd);
2704 if (rc)
2705 return rc;
2706
2707 tgt = lmv_find_target(lmv, fid);
2708 if (IS_ERR(tgt))
2709 return PTR_ERR(tgt);
2710
2711 rc = md_revalidate_lock(tgt->ltd_exp, it, fid, bits);
2712 return rc;
2713 }
2714
2715 /**
2716 * For lmv, only need to send request to master MDT, and the master MDT will
2717 * process with other slave MDTs. The only exception is Q_GETOQUOTA for which
2718 * we directly fetch data from the slave MDTs.
2719 */
2720 int lmv_quotactl(struct obd_device *unused, struct obd_export *exp,
2721 struct obd_quotactl *oqctl)
2722 {
2723 struct obd_device *obd = class_exp2obd(exp);
2724 struct lmv_obd *lmv = &obd->u.lmv;
2725 struct lmv_tgt_desc *tgt = lmv->tgts[0];
2726 int rc = 0, i;
2727 __u64 curspace, curinodes;
2728
2729 if (!lmv->desc.ld_tgt_count || !tgt->ltd_active) {
2730 CERROR("master lmv inactive\n");
2731 return -EIO;
2732 }
2733
2734 if (oqctl->qc_cmd != Q_GETOQUOTA) {
2735 rc = obd_quotactl(tgt->ltd_exp, oqctl);
2736 return rc;
2737 }
2738
2739 curspace = curinodes = 0;
2740 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2741 int err;
2742 tgt = lmv->tgts[i];
2743
2744 if (tgt == NULL || tgt->ltd_exp == NULL || tgt->ltd_active == 0)
2745 continue;
2746 if (!tgt->ltd_active) {
2747 CDEBUG(D_HA, "mdt %d is inactive.\n", i);
2748 continue;
2749 }
2750
2751 err = obd_quotactl(tgt->ltd_exp, oqctl);
2752 if (err) {
2753 CERROR("getquota on mdt %d failed. %d\n", i, err);
2754 if (!rc)
2755 rc = err;
2756 } else {
2757 curspace += oqctl->qc_dqblk.dqb_curspace;
2758 curinodes += oqctl->qc_dqblk.dqb_curinodes;
2759 }
2760 }
2761 oqctl->qc_dqblk.dqb_curspace = curspace;
2762 oqctl->qc_dqblk.dqb_curinodes = curinodes;
2763
2764 return rc;
2765 }
2766
2767 int lmv_quotacheck(struct obd_device *unused, struct obd_export *exp,
2768 struct obd_quotactl *oqctl)
2769 {
2770 struct obd_device *obd = class_exp2obd(exp);
2771 struct lmv_obd *lmv = &obd->u.lmv;
2772 struct lmv_tgt_desc *tgt;
2773 int i, rc = 0;
2774
2775 for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
2776 int err;
2777 tgt = lmv->tgts[i];
2778 if (tgt == NULL || tgt->ltd_exp == NULL || !tgt->ltd_active) {
2779 CERROR("lmv idx %d inactive\n", i);
2780 return -EIO;
2781 }
2782
2783 err = obd_quotacheck(tgt->ltd_exp, oqctl);
2784 if (err && !rc)
2785 rc = err;
2786 }
2787
2788 return rc;
2789 }
2790
2791 struct obd_ops lmv_obd_ops = {
2792 .o_owner = THIS_MODULE,
2793 .o_setup = lmv_setup,
2794 .o_cleanup = lmv_cleanup,
2795 .o_precleanup = lmv_precleanup,
2796 .o_process_config = lmv_process_config,
2797 .o_connect = lmv_connect,
2798 .o_disconnect = lmv_disconnect,
2799 .o_statfs = lmv_statfs,
2800 .o_get_info = lmv_get_info,
2801 .o_set_info_async = lmv_set_info_async,
2802 .o_packmd = lmv_packmd,
2803 .o_unpackmd = lmv_unpackmd,
2804 .o_notify = lmv_notify,
2805 .o_get_uuid = lmv_get_uuid,
2806 .o_iocontrol = lmv_iocontrol,
2807 .o_quotacheck = lmv_quotacheck,
2808 .o_quotactl = lmv_quotactl
2809 };
2810
2811 struct md_ops lmv_md_ops = {
2812 .m_getstatus = lmv_getstatus,
2813 .m_null_inode = lmv_null_inode,
2814 .m_find_cbdata = lmv_find_cbdata,
2815 .m_close = lmv_close,
2816 .m_create = lmv_create,
2817 .m_done_writing = lmv_done_writing,
2818 .m_enqueue = lmv_enqueue,
2819 .m_getattr = lmv_getattr,
2820 .m_getxattr = lmv_getxattr,
2821 .m_getattr_name = lmv_getattr_name,
2822 .m_intent_lock = lmv_intent_lock,
2823 .m_link = lmv_link,
2824 .m_rename = lmv_rename,
2825 .m_setattr = lmv_setattr,
2826 .m_setxattr = lmv_setxattr,
2827 .m_sync = lmv_sync,
2828 .m_readpage = lmv_readpage,
2829 .m_unlink = lmv_unlink,
2830 .m_init_ea_size = lmv_init_ea_size,
2831 .m_cancel_unused = lmv_cancel_unused,
2832 .m_set_lock_data = lmv_set_lock_data,
2833 .m_lock_match = lmv_lock_match,
2834 .m_get_lustre_md = lmv_get_lustre_md,
2835 .m_free_lustre_md = lmv_free_lustre_md,
2836 .m_set_open_replay_data = lmv_set_open_replay_data,
2837 .m_clear_open_replay_data = lmv_clear_open_replay_data,
2838 .m_renew_capa = lmv_renew_capa,
2839 .m_unpack_capa = lmv_unpack_capa,
2840 .m_get_remote_perm = lmv_get_remote_perm,
2841 .m_intent_getattr_async = lmv_intent_getattr_async,
2842 .m_revalidate_lock = lmv_revalidate_lock
2843 };
2844
2845 int __init lmv_init(void)
2846 {
2847 struct lprocfs_static_vars lvars;
2848 int rc;
2849
2850 lprocfs_lmv_init_vars(&lvars);
2851
2852 rc = class_register_type(&lmv_obd_ops, &lmv_md_ops,
2853 lvars.module_vars, LUSTRE_LMV_NAME, NULL);
2854 return rc;
2855 }
2856
2857 static void lmv_exit(void)
2858 {
2859 class_unregister_type(LUSTRE_LMV_NAME);
2860 }
2861
2862 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
2863 MODULE_DESCRIPTION("Lustre Logical Metadata Volume OBD driver");
2864 MODULE_LICENSE("GPL");
2865
2866 module_init(lmv_init);
2867 module_exit(lmv_exit);
Linux kernel - Deliverables
This page took 0.134525 seconds and 5 git commands to generate.