4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
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.
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).
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
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
27 * Copyright (c) 2001, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Whamcloud, Inc.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/ost/ost_handler.c
38 * Author: Peter J. Braam <braam@clusterfs.com>
39 * Author: Phil Schwan <phil@clusterfs.com>
42 #define DEBUG_SUBSYSTEM S_OST
44 #include <linux/module.h>
45 #include <obd_cksum.h>
47 #include <lustre_net.h>
48 #include <lustre_dlm.h>
49 #include <lustre_export.h>
50 #include <lustre_debug.h>
51 #include <linux/init.h>
52 #include <lprocfs_status.h>
53 #include <libcfs/list.h>
54 #include <lustre_quota.h>
55 #include "ost_internal.h"
57 static int oss_num_threads;
58 CFS_MODULE_PARM(oss_num_threads, "i", int, 0444,
59 "number of OSS service threads to start");
61 static int ost_num_threads;
62 CFS_MODULE_PARM(ost_num_threads, "i", int, 0444,
63 "number of OST service threads to start (deprecated)");
65 static int oss_num_create_threads;
66 CFS_MODULE_PARM(oss_num_create_threads, "i", int, 0444,
67 "number of OSS create threads to start");
69 static char *oss_cpts;
70 CFS_MODULE_PARM(oss_cpts, "s", charp, 0444,
71 "CPU partitions OSS threads should run on");
73 static char *oss_io_cpts;
74 CFS_MODULE_PARM(oss_io_cpts, "s", charp, 0444,
75 "CPU partitions OSS IO threads should run on");
78 * Do not return server-side uid/gid to remote client
80 static void ost_drop_id(struct obd_export *exp, struct obdo *oa)
82 if (exp_connect_rmtclient(exp)) {
85 oa->o_valid &= ~(OBD_MD_FLUID | OBD_MD_FLGID);
90 * Validate oa from client.
91 * If the request comes from 2.0 clients, currently only RSVD seq and IDIF
93 * a. for single MDS seq = FID_SEQ_OST_MDT0,
94 * b. for CMD, seq = FID_SEQ_OST_MDT0, FID_SEQ_OST_MDT1 - FID_SEQ_OST_MAX
96 static int ost_validate_obdo(struct obd_export *exp, struct obdo *oa,
97 struct obd_ioobj *ioobj)
99 if (oa != NULL && !(oa->o_valid & OBD_MD_FLGROUP)) {
100 oa->o_seq = FID_SEQ_OST_MDT0;
102 ioobj->ioo_seq = FID_SEQ_OST_MDT0;
103 /* remove fid_seq_is_rsvd() after FID-on-OST allows SEQ > 9 */
104 } else if (oa == NULL || !(fid_seq_is_rsvd(oa->o_seq) ||
105 fid_seq_is_mdt0(oa->o_seq))) {
106 CERROR("%s: client %s sent invalid object "POSTID"\n",
107 exp->exp_obd->obd_name, obd_export_nid2str(exp),
108 oa ? oa->o_id : -1, oa ? oa->o_seq : -1);
111 obdo_from_ostid(oa, &oa->o_oi);
113 ioobj_from_obdo(ioobj, oa);
117 void oti_to_request(struct obd_trans_info *oti, struct ptlrpc_request *req)
119 struct oti_req_ack_lock *ack_lock;
125 if (req->rq_repmsg) {
126 __u64 versions[PTLRPC_NUM_VERSIONS] = { 0 };
127 lustre_msg_set_transno(req->rq_repmsg, oti->oti_transno);
128 versions[0] = oti->oti_pre_version;
129 lustre_msg_set_versions(req->rq_repmsg, versions);
131 req->rq_transno = oti->oti_transno;
133 /* XXX 4 == entries in oti_ack_locks??? */
134 for (ack_lock = oti->oti_ack_locks, i = 0; i < 4; i++, ack_lock++) {
137 /* XXX not even calling target_send_reply in some cases... */
138 ptlrpc_save_lock (req, &ack_lock->lock, ack_lock->mode, 0);
142 static int ost_destroy(struct obd_export *exp, struct ptlrpc_request *req,
143 struct obd_trans_info *oti)
145 struct ost_body *body, *repbody;
146 struct lustre_capa *capa = NULL;
150 /* Get the request body */
151 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
155 if (body->oa.o_id == 0)
158 rc = ost_validate_obdo(exp, &body->oa, NULL);
162 /* If there's a DLM request, cancel the locks mentioned in it*/
163 if (req_capsule_field_present(&req->rq_pill, &RMF_DLM_REQ, RCL_CLIENT)) {
164 struct ldlm_request *dlm;
166 dlm = req_capsule_client_get(&req->rq_pill, &RMF_DLM_REQ);
169 ldlm_request_cancel(req, dlm, 0);
172 /* If there's a capability, get it */
173 if (body->oa.o_valid & OBD_MD_FLOSSCAPA) {
174 capa = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
176 CERROR("Missing capability for OST DESTROY");
181 /* Prepare the reply */
182 rc = req_capsule_server_pack(&req->rq_pill);
186 /* Get the log cancellation cookie */
187 if (body->oa.o_valid & OBD_MD_FLCOOKIE)
188 oti->oti_logcookies = &body->oa.o_lcookie;
190 /* Finish the reply */
191 repbody = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
192 memcpy(&repbody->oa, &body->oa, sizeof(body->oa));
194 /* Do the destroy and set the reply status accordingly */
195 req->rq_status = obd_destroy(req->rq_svc_thread->t_env, exp,
196 &repbody->oa, NULL, oti, NULL, capa);
201 * Helper function for getting server side [start, start+count] DLM lock
202 * if asked by client.
204 static int ost_lock_get(struct obd_export *exp, struct obdo *oa,
205 __u64 start, __u64 count, struct lustre_handle *lh,
208 struct ldlm_res_id res_id;
209 ldlm_policy_data_t policy;
210 __u64 end = start + count;
214 LASSERT(!lustre_handle_is_used(lh));
215 /* o_id and o_gr are used for localizing resource, if client miss to set
216 * them, do not trigger ASSERTION. */
217 if (unlikely((oa->o_valid & (OBD_MD_FLID | OBD_MD_FLGROUP)) !=
218 (OBD_MD_FLID | OBD_MD_FLGROUP)))
221 if (!(oa->o_valid & OBD_MD_FLFLAGS) ||
222 !(oa->o_flags & OBD_FL_SRVLOCK))
225 osc_build_res_name(oa->o_id, oa->o_seq, &res_id);
226 CDEBUG(D_INODE, "OST-side extent lock.\n");
228 policy.l_extent.start = start & CFS_PAGE_MASK;
230 /* If ->o_blocks is EOF it means "lock till the end of the
231 * file". Otherwise, it's size of a hole being punched (in bytes) */
232 if (count == OBD_OBJECT_EOF || end < start)
233 policy.l_extent.end = OBD_OBJECT_EOF;
235 policy.l_extent.end = end | ~CFS_PAGE_MASK;
237 RETURN(ldlm_cli_enqueue_local(exp->exp_obd->obd_namespace, &res_id,
238 LDLM_EXTENT, &policy, mode, &flags,
239 ldlm_blocking_ast, ldlm_completion_ast,
240 ldlm_glimpse_ast, NULL, 0, NULL, lh));
243 /* Helper function: release lock, if any. */
244 static void ost_lock_put(struct obd_export *exp,
245 struct lustre_handle *lh, int mode)
248 if (lustre_handle_is_used(lh))
249 ldlm_lock_decref(lh, mode);
253 static int ost_getattr(struct obd_export *exp, struct ptlrpc_request *req)
255 struct ost_body *body, *repbody;
256 struct obd_info *oinfo;
257 struct lustre_handle lh = { 0 };
258 struct lustre_capa *capa = NULL;
262 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
266 rc = ost_validate_obdo(exp, &body->oa, NULL);
270 if (body->oa.o_valid & OBD_MD_FLOSSCAPA) {
271 capa = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
273 CERROR("Missing capability for OST GETATTR");
278 rc = req_capsule_server_pack(&req->rq_pill);
282 repbody = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
283 repbody->oa = body->oa;
285 rc = ost_lock_get(exp, &repbody->oa, 0, OBD_OBJECT_EOF, &lh, LCK_PR, 0);
289 OBD_ALLOC_PTR(oinfo);
291 GOTO(unlock, rc = -ENOMEM);
292 oinfo->oi_oa = &repbody->oa;
293 oinfo->oi_capa = capa;
295 req->rq_status = obd_getattr(req->rq_svc_thread->t_env, exp, oinfo);
299 ost_drop_id(exp, &repbody->oa);
302 ost_lock_put(exp, &lh, LCK_PR);
306 static int ost_statfs(struct ptlrpc_request *req)
308 struct obd_statfs *osfs;
312 rc = req_capsule_server_pack(&req->rq_pill);
316 osfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
318 req->rq_status = obd_statfs(req->rq_svc_thread->t_env, req->rq_export,
320 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
322 if (req->rq_status != 0)
323 CERROR("ost: statfs failed: rc %d\n", req->rq_status);
328 static int ost_create(struct obd_export *exp, struct ptlrpc_request *req,
329 struct obd_trans_info *oti)
331 struct ost_body *body, *repbody;
335 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
339 rc = ost_validate_obdo(req->rq_export, &body->oa, NULL);
343 rc = req_capsule_server_pack(&req->rq_pill);
347 repbody = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
348 repbody->oa = body->oa;
349 oti->oti_logcookies = &body->oa.o_lcookie;
351 req->rq_status = obd_create(req->rq_svc_thread->t_env, exp,
352 &repbody->oa, NULL, oti);
353 //obd_log_cancel(conn, NULL, 1, oti->oti_logcookies, 0);
357 static int ost_punch(struct obd_export *exp, struct ptlrpc_request *req,
358 struct obd_trans_info *oti)
360 struct ost_body *body, *repbody;
362 struct lustre_handle lh = {0,};
365 /* check that we do support OBD_CONNECT_TRUNCLOCK. */
366 CLASSERT(OST_CONNECT_SUPPORTED & OBD_CONNECT_TRUNCLOCK);
368 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
372 rc = ost_validate_obdo(exp, &body->oa, NULL);
376 if ((body->oa.o_valid & (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS)) !=
377 (OBD_MD_FLSIZE | OBD_MD_FLBLOCKS))
380 rc = req_capsule_server_pack(&req->rq_pill);
384 /* standard truncate optimization: if file body is completely
385 * destroyed, don't send data back to the server. */
386 if (body->oa.o_size == 0)
387 flags |= LDLM_AST_DISCARD_DATA;
389 repbody = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
390 repbody->oa = body->oa;
392 rc = ost_lock_get(exp, &repbody->oa, repbody->oa.o_size,
393 repbody->oa.o_blocks, &lh, LCK_PW, flags);
395 struct obd_info *oinfo;
396 struct lustre_capa *capa = NULL;
398 if (repbody->oa.o_valid & OBD_MD_FLFLAGS &&
399 repbody->oa.o_flags == OBD_FL_SRVLOCK)
401 * If OBD_FL_SRVLOCK is the only bit set in
402 * ->o_flags, clear OBD_MD_FLFLAGS to avoid falling
403 * through filter_setattr() to filter_iocontrol().
405 repbody->oa.o_valid &= ~OBD_MD_FLFLAGS;
407 if (repbody->oa.o_valid & OBD_MD_FLOSSCAPA) {
408 capa = req_capsule_client_get(&req->rq_pill,
411 CERROR("Missing capability for OST PUNCH");
412 GOTO(unlock, rc = -EFAULT);
416 OBD_ALLOC_PTR(oinfo);
418 GOTO(unlock, rc = -ENOMEM);
419 oinfo->oi_oa = &repbody->oa;
420 oinfo->oi_policy.l_extent.start = oinfo->oi_oa->o_size;
421 oinfo->oi_policy.l_extent.end = oinfo->oi_oa->o_blocks;
422 oinfo->oi_capa = capa;
423 oinfo->oi_flags = OBD_FL_PUNCH;
425 req->rq_status = obd_punch(req->rq_svc_thread->t_env, exp,
429 ost_lock_put(exp, &lh, LCK_PW);
432 ost_drop_id(exp, &repbody->oa);
436 static int ost_sync(struct obd_export *exp, struct ptlrpc_request *req,
437 struct obd_trans_info *oti)
439 struct ost_body *body, *repbody;
440 struct obd_info *oinfo;
441 struct lustre_capa *capa = NULL;
445 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
449 rc = ost_validate_obdo(exp, &body->oa, NULL);
453 if (body->oa.o_valid & OBD_MD_FLOSSCAPA) {
454 capa = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
456 CERROR("Missing capability for OST SYNC");
461 rc = req_capsule_server_pack(&req->rq_pill);
465 repbody = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
466 repbody->oa = body->oa;
468 OBD_ALLOC_PTR(oinfo);
472 oinfo->oi_oa = &repbody->oa;
473 oinfo->oi_capa = capa;
474 oinfo->oi_jobid = oti->oti_jobid;
475 req->rq_status = obd_sync(req->rq_svc_thread->t_env, exp, oinfo,
476 repbody->oa.o_size, repbody->oa.o_blocks,
480 ost_drop_id(exp, &repbody->oa);
484 static int ost_setattr(struct obd_export *exp, struct ptlrpc_request *req,
485 struct obd_trans_info *oti)
487 struct ost_body *body, *repbody;
488 struct obd_info *oinfo;
489 struct lustre_capa *capa = NULL;
493 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
497 rc = ost_validate_obdo(req->rq_export, &body->oa, NULL);
501 rc = req_capsule_server_pack(&req->rq_pill);
505 if (body->oa.o_valid & OBD_MD_FLOSSCAPA) {
506 capa = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
508 CERROR("Missing capability for OST SETATTR");
513 repbody = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
514 repbody->oa = body->oa;
516 OBD_ALLOC_PTR(oinfo);
519 oinfo->oi_oa = &repbody->oa;
520 oinfo->oi_capa = capa;
522 req->rq_status = obd_setattr(req->rq_svc_thread->t_env, exp, oinfo,
527 ost_drop_id(exp, &repbody->oa);
531 static __u32 ost_checksum_bulk(struct ptlrpc_bulk_desc *desc, int opc,
532 cksum_type_t cksum_type)
534 struct cfs_crypto_hash_desc *hdesc;
535 unsigned int bufsize;
537 unsigned char cfs_alg = cksum_obd2cfs(cksum_type);
540 hdesc = cfs_crypto_hash_init(cfs_alg, NULL, 0);
542 CERROR("Unable to initialize checksum hash %s\n",
543 cfs_crypto_hash_name(cfs_alg));
544 return PTR_ERR(hdesc);
546 CDEBUG(D_INFO, "Checksum for algo %s\n", cfs_crypto_hash_name(cfs_alg));
547 for (i = 0; i < desc->bd_iov_count; i++) {
549 /* corrupt the data before we compute the checksum, to
550 * simulate a client->OST data error */
551 if (i == 0 && opc == OST_WRITE &&
552 OBD_FAIL_CHECK(OBD_FAIL_OST_CHECKSUM_RECEIVE)) {
553 int off = desc->bd_iov[i].kiov_offset & ~CFS_PAGE_MASK;
554 int len = desc->bd_iov[i].kiov_len;
555 char *ptr = kmap(desc->bd_iov[i].kiov_page) + off;
556 memcpy(ptr, "bad3", min(4, len));
557 kunmap(desc->bd_iov[i].kiov_page);
559 cfs_crypto_hash_update_page(hdesc, desc->bd_iov[i].kiov_page,
560 desc->bd_iov[i].kiov_offset & ~CFS_PAGE_MASK,
561 desc->bd_iov[i].kiov_len);
563 /* corrupt the data after we compute the checksum, to
564 * simulate an OST->client data error */
565 if (i == 0 && opc == OST_READ &&
566 OBD_FAIL_CHECK(OBD_FAIL_OST_CHECKSUM_SEND)) {
567 int off = desc->bd_iov[i].kiov_offset & ~CFS_PAGE_MASK;
568 int len = desc->bd_iov[i].kiov_len;
569 char *ptr = kmap(desc->bd_iov[i].kiov_page) + off;
570 memcpy(ptr, "bad4", min(4, len));
571 kunmap(desc->bd_iov[i].kiov_page);
572 /* nobody should use corrupted page again */
573 ClearPageUptodate(desc->bd_iov[i].kiov_page);
578 err = cfs_crypto_hash_final(hdesc, (unsigned char *)&cksum, &bufsize);
580 cfs_crypto_hash_final(hdesc, NULL, NULL);
585 static int ost_brw_lock_get(int mode, struct obd_export *exp,
586 struct obd_ioobj *obj, struct niobuf_remote *nb,
587 struct lustre_handle *lh)
590 int nrbufs = obj->ioo_bufcnt;
591 struct ldlm_res_id res_id;
592 ldlm_policy_data_t policy;
596 osc_build_res_name(obj->ioo_id, obj->ioo_seq, &res_id);
597 LASSERT(mode == LCK_PR || mode == LCK_PW);
598 LASSERT(!lustre_handle_is_used(lh));
600 if (nrbufs == 0 || !(nb[0].flags & OBD_BRW_SRVLOCK))
603 for (i = 1; i < nrbufs; i ++)
604 if ((nb[0].flags & OBD_BRW_SRVLOCK) !=
605 (nb[i].flags & OBD_BRW_SRVLOCK))
608 policy.l_extent.start = nb[0].offset & CFS_PAGE_MASK;
609 policy.l_extent.end = (nb[nrbufs - 1].offset +
610 nb[nrbufs - 1].len - 1) | ~CFS_PAGE_MASK;
612 RETURN(ldlm_cli_enqueue_local(exp->exp_obd->obd_namespace, &res_id,
613 LDLM_EXTENT, &policy, mode, &flags,
614 ldlm_blocking_ast, ldlm_completion_ast,
615 ldlm_glimpse_ast, NULL, 0, NULL, lh));
618 static void ost_brw_lock_put(int mode,
619 struct obd_ioobj *obj, struct niobuf_remote *niob,
620 struct lustre_handle *lh)
623 LASSERT(mode == LCK_PR || mode == LCK_PW);
624 LASSERT((obj->ioo_bufcnt > 0 && (niob[0].flags & OBD_BRW_SRVLOCK)) ==
625 lustre_handle_is_used(lh));
626 if (lustre_handle_is_used(lh))
627 ldlm_lock_decref(lh, mode);
631 /* Allocate thread local buffers if needed */
632 static struct ost_thread_local_cache *ost_tls_get(struct ptlrpc_request *r)
634 struct ost_thread_local_cache *tls =
635 (struct ost_thread_local_cache *)(r->rq_svc_thread->t_data);
637 /* In normal mode of operation an I/O request is serviced only
638 * by ll_ost_io threads each of them has own tls buffers allocated by
640 * During recovery, an I/O request may be queued until any of the ost
641 * service threads process it. Not necessary it should be one of
642 * ll_ost_io threads. In that case we dynamically allocating tls
643 * buffers for the request service time. */
644 if (unlikely(tls == NULL)) {
645 LASSERT(r->rq_export->exp_in_recovery);
649 r->rq_svc_thread->t_data = tls;
655 /* Free thread local buffers if they were allocated only for servicing
656 * this one request */
657 static void ost_tls_put(struct ptlrpc_request *r)
659 struct ost_thread_local_cache *tls =
660 (struct ost_thread_local_cache *)(r->rq_svc_thread->t_data);
662 if (unlikely(tls->temporary)) {
664 r->rq_svc_thread->t_data = NULL;
668 static int ost_brw_read(struct ptlrpc_request *req, struct obd_trans_info *oti)
670 struct ptlrpc_bulk_desc *desc = NULL;
671 struct obd_export *exp = req->rq_export;
672 struct niobuf_remote *remote_nb;
673 struct niobuf_local *local_nb;
674 struct obd_ioobj *ioo;
675 struct ost_body *body, *repbody;
676 struct lustre_capa *capa = NULL;
677 struct l_wait_info lwi;
678 struct lustre_handle lockh = { 0 };
679 int niocount, npages, nob = 0, rc, i;
681 struct ost_thread_local_cache *tls;
684 req->rq_bulk_read = 1;
686 if (OBD_FAIL_CHECK(OBD_FAIL_OST_BRW_READ_BULK))
687 GOTO(out, rc = -EIO);
689 OBD_FAIL_TIMEOUT(OBD_FAIL_OST_BRW_PAUSE_BULK, (obd_timeout + 1) / 4);
691 /* Check if there is eviction in progress, and if so, wait for it to
693 if (unlikely(cfs_atomic_read(&exp->exp_obd->obd_evict_inprogress))) {
694 lwi = LWI_INTR(NULL, NULL); // We do not care how long it takes
695 rc = l_wait_event(exp->exp_obd->obd_evict_inprogress_waitq,
696 !cfs_atomic_read(&exp->exp_obd->obd_evict_inprogress),
700 GOTO(out, rc = -ENOTCONN);
702 /* ost_body, ioobj & noibuf_remote are verified and swabbed in
703 * ost_rw_hpreq_check(). */
704 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
706 GOTO(out, rc = -EFAULT);
709 * A req_capsule_X_get_array(pill, field, ptr_to_element_count) function
710 * would be useful here and wherever we get &RMF_OBD_IOOBJ and
711 * &RMF_NIOBUF_REMOTE.
713 ioo = req_capsule_client_get(&req->rq_pill, &RMF_OBD_IOOBJ);
715 GOTO(out, rc = -EFAULT);
717 rc = ost_validate_obdo(exp, &body->oa, ioo);
721 niocount = ioo->ioo_bufcnt;
722 remote_nb = req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE);
723 if (remote_nb == NULL)
724 GOTO(out, rc = -EFAULT);
726 if (body->oa.o_valid & OBD_MD_FLOSSCAPA) {
727 capa = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
729 CERROR("Missing capability for OST BRW READ");
730 GOTO(out, rc = -EFAULT);
734 rc = req_capsule_server_pack(&req->rq_pill);
738 tls = ost_tls_get(req);
740 GOTO(out_bulk, rc = -ENOMEM);
741 local_nb = tls->local;
743 rc = ost_brw_lock_get(LCK_PR, exp, ioo, remote_nb, &lockh);
748 * If getting the lock took more time than
749 * client was willing to wait, drop it. b=11330
751 if (cfs_time_current_sec() > req->rq_deadline ||
752 OBD_FAIL_CHECK(OBD_FAIL_OST_DROP_REQ)) {
754 CERROR("Dropping timed-out read from %s because locking"
755 "object "LPX64" took %ld seconds (limit was %ld).\n",
756 libcfs_id2str(req->rq_peer), ioo->ioo_id,
757 cfs_time_current_sec() - req->rq_arrival_time.tv_sec,
758 req->rq_deadline - req->rq_arrival_time.tv_sec);
759 GOTO(out_lock, rc = -ETIMEDOUT);
762 repbody = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
763 memcpy(&repbody->oa, &body->oa, sizeof(repbody->oa));
765 npages = OST_THREAD_POOL_SIZE;
766 rc = obd_preprw(req->rq_svc_thread->t_env, OBD_BRW_READ, exp,
767 &repbody->oa, 1, ioo, remote_nb, &npages, local_nb,
772 desc = ptlrpc_prep_bulk_exp(req, npages,
773 BULK_PUT_SOURCE, OST_BULK_PORTAL);
775 GOTO(out_commitrw, rc = -ENOMEM);
778 for (i = 0; i < npages; i++) {
779 int page_rc = local_nb[i].rc;
781 if (page_rc < 0) { /* error */
787 if (page_rc != 0) { /* some data! */
788 LASSERT (local_nb[i].page != NULL);
789 ptlrpc_prep_bulk_page(desc, local_nb[i].page,
790 local_nb[i].offset & ~CFS_PAGE_MASK,
794 if (page_rc != local_nb[i].len) { /* short read */
795 /* All subsequent pages should be 0 */
797 LASSERT(local_nb[i].rc == 0);
802 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
803 cksum_type_t cksum_type =
804 cksum_type_unpack(repbody->oa.o_valid & OBD_MD_FLFLAGS ?
805 repbody->oa.o_flags : 0);
806 repbody->oa.o_flags = cksum_type_pack(cksum_type);
807 repbody->oa.o_valid = OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
808 repbody->oa.o_cksum = ost_checksum_bulk(desc, OST_READ,cksum_type);
809 CDEBUG(D_PAGE, "checksum at read origin: %x\n",
810 repbody->oa.o_cksum);
812 repbody->oa.o_valid = 0;
814 /* We're finishing using body->oa as an input variable */
816 /* Check if client was evicted while we were doing i/o before touching
819 if (likely(!CFS_FAIL_PRECHECK(OBD_FAIL_PTLRPC_CLIENT_BULK_CB2)))
820 rc = target_bulk_io(exp, desc, &lwi);
825 /* Must commit after prep above in all cases */
826 rc = obd_commitrw(req->rq_svc_thread->t_env, OBD_BRW_READ, exp,
827 &repbody->oa, 1, ioo, remote_nb, npages, local_nb,
831 ost_drop_id(exp, &repbody->oa);
834 ost_brw_lock_put(LCK_PR, ioo, remote_nb, &lockh);
838 if (desc && !CFS_FAIL_PRECHECK(OBD_FAIL_PTLRPC_CLIENT_BULK_CB2))
839 ptlrpc_free_bulk(desc);
843 req->rq_status = nob;
844 ptlrpc_lprocfs_brw(req, nob);
845 target_committed_to_req(req);
847 } else if (!no_reply) {
848 /* Only reply if there was no comms problem with bulk */
849 target_committed_to_req(req);
853 /* reply out callback would free */
854 ptlrpc_req_drop_rs(req);
855 LCONSOLE_WARN("%s: Bulk IO read error with %s (at %s), "
856 "client will retry: rc %d\n",
857 exp->exp_obd->obd_name,
858 obd_uuid2str(&exp->exp_client_uuid),
859 obd_export_nid2str(exp), rc);
861 /* send a bulk after reply to simulate a network delay or reordering
863 if (unlikely(CFS_FAIL_PRECHECK(OBD_FAIL_PTLRPC_CLIENT_BULK_CB2))) {
865 struct l_wait_info lwi1;
867 CDEBUG(D_INFO, "reorder BULK\n");
868 cfs_waitq_init(&waitq);
870 lwi1 = LWI_TIMEOUT_INTR(cfs_time_seconds(3), NULL, NULL, NULL);
871 l_wait_event(waitq, 0, &lwi1);
872 rc = target_bulk_io(exp, desc, &lwi);
873 ptlrpc_free_bulk(desc);
879 static int ost_brw_write(struct ptlrpc_request *req, struct obd_trans_info *oti)
881 struct ptlrpc_bulk_desc *desc = NULL;
882 struct obd_export *exp = req->rq_export;
883 struct niobuf_remote *remote_nb;
884 struct niobuf_local *local_nb;
885 struct obd_ioobj *ioo;
886 struct ost_body *body, *repbody;
887 struct l_wait_info lwi;
888 struct lustre_handle lockh = {0};
889 struct lustre_capa *capa = NULL;
891 int objcount, niocount, npages;
893 obd_count client_cksum = 0, server_cksum = 0;
894 cksum_type_t cksum_type = OBD_CKSUM_CRC32;
895 int no_reply = 0, mmap = 0;
896 __u32 o_uid = 0, o_gid = 0;
897 struct ost_thread_local_cache *tls;
900 req->rq_bulk_write = 1;
902 if (OBD_FAIL_CHECK(OBD_FAIL_OST_BRW_WRITE_BULK))
903 GOTO(out, rc = -EIO);
904 if (OBD_FAIL_CHECK(OBD_FAIL_OST_BRW_WRITE_BULK2))
905 GOTO(out, rc = -EFAULT);
907 /* pause before transaction has been started */
908 OBD_FAIL_TIMEOUT(OBD_FAIL_OST_BRW_PAUSE_BULK, (obd_timeout + 1) / 4);
910 /* ost_body, ioobj & noibuf_remote are verified and swabbed in
911 * ost_rw_hpreq_check(). */
912 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
914 GOTO(out, rc = -EFAULT);
916 objcount = req_capsule_get_size(&req->rq_pill, &RMF_OBD_IOOBJ,
917 RCL_CLIENT) / sizeof(*ioo);
918 ioo = req_capsule_client_get(&req->rq_pill, &RMF_OBD_IOOBJ);
920 GOTO(out, rc = -EFAULT);
922 rc = ost_validate_obdo(exp, &body->oa, ioo);
926 for (niocount = i = 0; i < objcount; i++)
927 niocount += ioo[i].ioo_bufcnt;
930 * It'd be nice to have a capsule function to indicate how many elements
931 * there were in a buffer for an RMF that's declared to be an array.
932 * It's easy enough to compute the number of elements here though.
934 remote_nb = req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE);
935 if (remote_nb == NULL || niocount != (req_capsule_get_size(&req->rq_pill,
936 &RMF_NIOBUF_REMOTE, RCL_CLIENT) / sizeof(*remote_nb)))
937 GOTO(out, rc = -EFAULT);
939 if ((remote_nb[0].flags & OBD_BRW_MEMALLOC) &&
940 (exp->exp_connection->c_peer.nid == exp->exp_connection->c_self))
941 cfs_memory_pressure_set();
943 if (body->oa.o_valid & OBD_MD_FLOSSCAPA) {
944 capa = req_capsule_client_get(&req->rq_pill, &RMF_CAPA1);
946 CERROR("Missing capability for OST BRW WRITE");
947 GOTO(out, rc = -EFAULT);
951 req_capsule_set_size(&req->rq_pill, &RMF_RCS, RCL_SERVER,
952 niocount * sizeof(*rcs));
953 rc = req_capsule_server_pack(&req->rq_pill);
956 CFS_FAIL_TIMEOUT(OBD_FAIL_OST_BRW_PAUSE_PACK, cfs_fail_val);
957 rcs = req_capsule_server_get(&req->rq_pill, &RMF_RCS);
959 tls = ost_tls_get(req);
961 GOTO(out_bulk, rc = -ENOMEM);
962 local_nb = tls->local;
964 rc = ost_brw_lock_get(LCK_PW, exp, ioo, remote_nb, &lockh);
969 * If getting the lock took more time than
970 * client was willing to wait, drop it. b=11330
972 if (cfs_time_current_sec() > req->rq_deadline ||
973 OBD_FAIL_CHECK(OBD_FAIL_OST_DROP_REQ)) {
975 CERROR("Dropping timed-out write from %s because locking "
976 "object "LPX64" took %ld seconds (limit was %ld).\n",
977 libcfs_id2str(req->rq_peer), ioo->ioo_id,
978 cfs_time_current_sec() - req->rq_arrival_time.tv_sec,
979 req->rq_deadline - req->rq_arrival_time.tv_sec);
980 GOTO(out_lock, rc = -ETIMEDOUT);
983 /* obd_preprw clobbers oa->valid, so save what we need */
984 if (body->oa.o_valid & OBD_MD_FLCKSUM) {
985 client_cksum = body->oa.o_cksum;
986 if (body->oa.o_valid & OBD_MD_FLFLAGS)
987 cksum_type = cksum_type_unpack(body->oa.o_flags);
989 if (body->oa.o_valid & OBD_MD_FLFLAGS && body->oa.o_flags & OBD_FL_MMAP)
992 /* Because we already sync grant info with client when reconnect,
993 * grant info will be cleared for resent req, then fed_grant and
994 * total_grant will not be modified in following preprw_write */
995 if (lustre_msg_get_flags(req->rq_reqmsg) & (MSG_RESENT | MSG_REPLAY)) {
996 DEBUG_REQ(D_CACHE, req, "clear resent/replay req grant info");
997 body->oa.o_valid &= ~OBD_MD_FLGRANT;
1000 if (exp_connect_rmtclient(exp)) {
1001 o_uid = body->oa.o_uid;
1002 o_gid = body->oa.o_gid;
1005 repbody = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
1006 memcpy(&repbody->oa, &body->oa, sizeof(repbody->oa));
1008 npages = OST_THREAD_POOL_SIZE;
1009 rc = obd_preprw(req->rq_svc_thread->t_env, OBD_BRW_WRITE, exp,
1010 &repbody->oa, objcount, ioo, remote_nb, &npages,
1011 local_nb, oti, capa);
1015 desc = ptlrpc_prep_bulk_exp(req, npages,
1016 BULK_GET_SINK, OST_BULK_PORTAL);
1018 GOTO(skip_transfer, rc = -ENOMEM);
1020 /* NB Having prepped, we must commit... */
1022 for (i = 0; i < npages; i++)
1023 ptlrpc_prep_bulk_page(desc, local_nb[i].page,
1024 local_nb[i].offset & ~CFS_PAGE_MASK,
1027 rc = sptlrpc_svc_prep_bulk(req, desc);
1031 rc = target_bulk_io(exp, desc, &lwi);
1035 if (client_cksum != 0 && rc == 0) {
1036 static int cksum_counter;
1037 repbody->oa.o_valid |= OBD_MD_FLCKSUM | OBD_MD_FLFLAGS;
1038 repbody->oa.o_flags &= ~OBD_FL_CKSUM_ALL;
1039 repbody->oa.o_flags |= cksum_type_pack(cksum_type);
1040 server_cksum = ost_checksum_bulk(desc, OST_WRITE, cksum_type);
1041 repbody->oa.o_cksum = server_cksum;
1043 if (unlikely(client_cksum != server_cksum)) {
1044 CDEBUG_LIMIT(mmap ? D_INFO : D_ERROR,
1045 "client csum %x, server csum %x\n",
1046 client_cksum, server_cksum);
1048 } else if ((cksum_counter & (-cksum_counter)) == cksum_counter){
1049 CDEBUG(D_INFO, "Checksum %u from %s OK: %x\n",
1050 cksum_counter, libcfs_id2str(req->rq_peer),
1055 /* Must commit after prep above in all cases */
1056 rc = obd_commitrw(req->rq_svc_thread->t_env, OBD_BRW_WRITE, exp,
1057 &repbody->oa, objcount, ioo, remote_nb, npages,
1059 if (rc == -ENOTCONN)
1060 /* quota acquire process has been given up because
1061 * either the client has been evicted or the client
1062 * has timed out the request already */
1065 if (exp_connect_rmtclient(exp)) {
1066 repbody->oa.o_uid = o_uid;
1067 repbody->oa.o_gid = o_gid;
1071 * Disable sending mtime back to the client. If the client locked the
1072 * whole object, then it has already updated the mtime on its side,
1073 * otherwise it will have to glimpse anyway (see bug 21489, comment 32)
1075 repbody->oa.o_valid &= ~(OBD_MD_FLMTIME | OBD_MD_FLATIME);
1077 if (unlikely(client_cksum != server_cksum && rc == 0 && !mmap)) {
1078 int new_cksum = ost_checksum_bulk(desc, OST_WRITE, cksum_type);
1083 if (new_cksum == server_cksum)
1084 msg = "changed in transit before arrival at OST";
1085 else if (new_cksum == client_cksum)
1086 msg = "initial checksum before message complete";
1088 msg = "changed in transit AND after initial checksum";
1090 if (req->rq_peer.nid == desc->bd_sender) {
1094 router = libcfs_nid2str(desc->bd_sender);
1097 LCONSOLE_ERROR_MSG(0x168, "%s: BAD WRITE CHECKSUM: %s from "
1098 "%s%s%s inode "DFID" object "
1099 LPU64"/"LPU64" extent ["LPU64"-"LPU64"]\n",
1100 exp->exp_obd->obd_name, msg,
1101 libcfs_id2str(req->rq_peer),
1103 body->oa.o_valid & OBD_MD_FLFID ?
1104 body->oa.o_parent_seq : (__u64)0,
1105 body->oa.o_valid & OBD_MD_FLFID ?
1106 body->oa.o_parent_oid : 0,
1107 body->oa.o_valid & OBD_MD_FLFID ?
1108 body->oa.o_parent_ver : 0,
1110 body->oa.o_valid & OBD_MD_FLGROUP ?
1111 body->oa.o_seq : (__u64)0,
1113 local_nb[npages-1].offset +
1114 local_nb[npages-1].len - 1 );
1115 CERROR("client csum %x, original server csum %x, "
1116 "server csum now %x\n",
1117 client_cksum, server_cksum, new_cksum);
1123 /* set per-requested niobuf return codes */
1124 for (i = j = 0; i < niocount; i++) {
1125 int len = remote_nb[i].len;
1130 LASSERT(j < npages);
1131 if (local_nb[j].rc < 0)
1132 rcs[i] = local_nb[j].rc;
1133 len -= local_nb[j].len;
1138 LASSERT(j == npages);
1139 ptlrpc_lprocfs_brw(req, nob);
1143 ost_brw_lock_put(LCK_PW, ioo, remote_nb, &lockh);
1148 ptlrpc_free_bulk(desc);
1151 oti_to_request(oti, req);
1152 target_committed_to_req(req);
1153 rc = ptlrpc_reply(req);
1154 } else if (!no_reply) {
1155 /* Only reply if there was no comms problem with bulk */
1156 target_committed_to_req(req);
1157 req->rq_status = rc;
1160 /* reply out callback would free */
1161 ptlrpc_req_drop_rs(req);
1162 LCONSOLE_WARN("%s: Bulk IO write error with %s (at %s), "
1163 "client will retry: rc %d\n",
1164 exp->exp_obd->obd_name,
1165 obd_uuid2str(&exp->exp_client_uuid),
1166 obd_export_nid2str(exp), rc);
1168 cfs_memory_pressure_clr();
1173 * Implementation of OST_SET_INFO.
1175 * OST_SET_INFO is like ioctl(): heavily overloaded. Specifically, it takes a
1176 * "key" and a value RPC buffers as arguments, with the value's contents
1177 * interpreted according to the key.
1179 * Value types that need swabbing have swabbing done explicitly, either here or
1180 * in functions called from here. This should be corrected: all swabbing should
1181 * be done in the capsule abstraction, as that will then allow us to move
1182 * swabbing exclusively to the client without having to modify server code
1183 * outside the capsule abstraction's implementation itself. To correct this
1184 * will require minor changes to the capsule abstraction; see the comments for
1185 * req_capsule_extend() in layout.c.
1187 static int ost_set_info(struct obd_export *exp, struct ptlrpc_request *req)
1189 struct ost_body *body = NULL, *repbody;
1190 char *key, *val = NULL;
1191 int keylen, vallen, rc = 0;
1192 int is_grant_shrink = 0;
1195 key = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
1197 DEBUG_REQ(D_HA, req, "no set_info key");
1200 keylen = req_capsule_get_size(&req->rq_pill, &RMF_SETINFO_KEY,
1203 vallen = req_capsule_get_size(&req->rq_pill, &RMF_SETINFO_VAL,
1206 if ((is_grant_shrink = KEY_IS(KEY_GRANT_SHRINK)))
1207 /* In this case the value is actually an RMF_OST_BODY, so we
1208 * transmutate the type of this PTLRPC */
1209 req_capsule_extend(&req->rq_pill, &RQF_OST_SET_GRANT_INFO);
1211 rc = req_capsule_server_pack(&req->rq_pill);
1216 if (is_grant_shrink) {
1217 body = req_capsule_client_get(&req->rq_pill,
1222 repbody = req_capsule_server_get(&req->rq_pill,
1224 memcpy(repbody, body, sizeof(*body));
1225 val = (char*)repbody;
1227 val = req_capsule_client_get(&req->rq_pill,
1232 if (KEY_IS(KEY_EVICT_BY_NID)) {
1234 obd_export_evict_by_nid(exp->exp_obd, val);
1236 } else if (KEY_IS(KEY_MDS_CONN) && ptlrpc_req_need_swab(req)) {
1237 if (vallen < sizeof(__u32))
1239 __swab32s((__u32 *)val);
1242 /* OBD will also check if KEY_IS(KEY_GRANT_SHRINK), and will cast val to
1243 * a struct ost_body * value */
1244 rc = obd_set_info_async(req->rq_svc_thread->t_env, exp, keylen,
1245 key, vallen, val, NULL);
1247 lustre_msg_set_status(req->rq_repmsg, 0);
1251 static int ost_get_info(struct obd_export *exp, struct ptlrpc_request *req)
1254 int keylen, replylen, rc = 0;
1255 struct req_capsule *pill = &req->rq_pill;
1258 /* this common part for get_info rpc */
1259 key = req_capsule_client_get(pill, &RMF_SETINFO_KEY);
1261 DEBUG_REQ(D_HA, req, "no get_info key");
1264 keylen = req_capsule_get_size(pill, &RMF_SETINFO_KEY, RCL_CLIENT);
1266 if (KEY_IS(KEY_FIEMAP)) {
1267 struct ll_fiemap_info_key *fm_key = key;
1270 rc = ost_validate_obdo(exp, &fm_key->oa, NULL);
1275 rc = obd_get_info(req->rq_svc_thread->t_env, exp, keylen, key,
1276 &replylen, NULL, NULL);
1280 req_capsule_set_size(pill, &RMF_GENERIC_DATA,
1281 RCL_SERVER, replylen);
1283 rc = req_capsule_server_pack(pill);
1287 reply = req_capsule_server_get(pill, &RMF_GENERIC_DATA);
1291 /* call again to fill in the reply buffer */
1292 rc = obd_get_info(req->rq_svc_thread->t_env, exp, keylen, key,
1293 &replylen, reply, NULL);
1295 lustre_msg_set_status(req->rq_repmsg, 0);
1299 #ifdef HAVE_QUOTA_SUPPORT
1300 static int ost_handle_quotactl(struct ptlrpc_request *req)
1302 struct obd_quotactl *oqctl, *repoqc;
1306 oqctl = req_capsule_client_get(&req->rq_pill, &RMF_OBD_QUOTACTL);
1308 GOTO(out, rc = -EPROTO);
1310 rc = req_capsule_server_pack(&req->rq_pill);
1314 repoqc = req_capsule_server_get(&req->rq_pill, &RMF_OBD_QUOTACTL);
1315 req->rq_status = obd_quotactl(req->rq_export, oqctl);
1322 static int ost_handle_quotacheck(struct ptlrpc_request *req)
1324 struct obd_quotactl *oqctl;
1328 oqctl = req_capsule_client_get(&req->rq_pill, &RMF_OBD_QUOTACTL);
1332 rc = req_capsule_server_pack(&req->rq_pill);
1336 req->rq_status = obd_quotacheck(req->rq_export, oqctl);
1340 static int ost_handle_quota_adjust_qunit(struct ptlrpc_request *req)
1342 struct quota_adjust_qunit *oqaq, *repoqa;
1343 struct lustre_quota_ctxt *qctxt;
1347 qctxt = &req->rq_export->exp_obd->u.obt.obt_qctxt;
1348 oqaq = req_capsule_client_get(&req->rq_pill, &RMF_QUOTA_ADJUST_QUNIT);
1350 GOTO(out, rc = -EPROTO);
1352 rc = req_capsule_server_pack(&req->rq_pill);
1356 repoqa = req_capsule_server_get(&req->rq_pill, &RMF_QUOTA_ADJUST_QUNIT);
1357 req->rq_status = obd_quota_adjust_qunit(req->rq_export, oqaq, qctxt, NULL);
1365 static int ost_llog_handle_connect(struct obd_export *exp,
1366 struct ptlrpc_request *req)
1368 struct llogd_conn_body *body;
1372 body = req_capsule_client_get(&req->rq_pill, &RMF_LLOGD_CONN_BODY);
1373 rc = obd_llog_connect(exp, body);
1377 #define ost_init_sec_none(reply, exp) \
1379 reply->ocd_connect_flags &= ~(OBD_CONNECT_RMT_CLIENT | \
1380 OBD_CONNECT_RMT_CLIENT_FORCE | \
1381 OBD_CONNECT_OSS_CAPA); \
1382 cfs_spin_lock(&exp->exp_lock); \
1383 exp->exp_connect_flags = reply->ocd_connect_flags; \
1384 cfs_spin_unlock(&exp->exp_lock); \
1387 static int ost_init_sec_level(struct ptlrpc_request *req)
1389 struct obd_export *exp = req->rq_export;
1390 struct req_capsule *pill = &req->rq_pill;
1391 struct obd_device *obd = exp->exp_obd;
1392 struct filter_obd *filter = &obd->u.filter;
1393 char *client = libcfs_nid2str(req->rq_peer.nid);
1394 struct obd_connect_data *data, *reply;
1398 data = req_capsule_client_get(pill, &RMF_CONNECT_DATA);
1399 reply = req_capsule_server_get(pill, &RMF_CONNECT_DATA);
1400 if (data == NULL || reply == NULL)
1403 /* connection from MDT is always trusted */
1404 if (req->rq_auth_usr_mdt) {
1405 ost_init_sec_none(reply, exp);
1409 /* no GSS support case */
1410 if (!req->rq_auth_gss) {
1411 if (filter->fo_sec_level > LUSTRE_SEC_NONE) {
1412 CWARN("client %s -> target %s does not user GSS, "
1413 "can not run under security level %d.\n",
1414 client, obd->obd_name, filter->fo_sec_level);
1417 ost_init_sec_none(reply, exp);
1422 /* old version case */
1423 if (unlikely(!(data->ocd_connect_flags & OBD_CONNECT_RMT_CLIENT) ||
1424 !(data->ocd_connect_flags & OBD_CONNECT_OSS_CAPA))) {
1425 if (filter->fo_sec_level > LUSTRE_SEC_NONE) {
1426 CWARN("client %s -> target %s uses old version, "
1427 "can not run under security level %d.\n",
1428 client, obd->obd_name, filter->fo_sec_level);
1431 CWARN("client %s -> target %s uses old version, "
1432 "run under security level %d.\n",
1433 client, obd->obd_name, filter->fo_sec_level);
1434 ost_init_sec_none(reply, exp);
1439 remote = data->ocd_connect_flags & OBD_CONNECT_RMT_CLIENT_FORCE;
1441 if (!req->rq_auth_remote)
1442 CDEBUG(D_SEC, "client (local realm) %s -> target %s "
1443 "asked to be remote.\n", client, obd->obd_name);
1444 } else if (req->rq_auth_remote) {
1446 CDEBUG(D_SEC, "client (remote realm) %s -> target %s is set "
1447 "as remote by default.\n", client, obd->obd_name);
1451 if (!filter->fo_fl_oss_capa) {
1452 CDEBUG(D_SEC, "client %s -> target %s is set as remote,"
1453 " but OSS capabilities are not enabled: %d.\n",
1454 client, obd->obd_name, filter->fo_fl_oss_capa);
1459 switch (filter->fo_sec_level) {
1460 case LUSTRE_SEC_NONE:
1462 ost_init_sec_none(reply, exp);
1465 CDEBUG(D_SEC, "client %s -> target %s is set as remote, "
1466 "can not run under security level %d.\n",
1467 client, obd->obd_name, filter->fo_sec_level);
1470 case LUSTRE_SEC_REMOTE:
1472 ost_init_sec_none(reply, exp);
1474 case LUSTRE_SEC_ALL:
1476 reply->ocd_connect_flags &= ~(OBD_CONNECT_RMT_CLIENT |
1477 OBD_CONNECT_RMT_CLIENT_FORCE);
1478 if (!filter->fo_fl_oss_capa)
1479 reply->ocd_connect_flags &= ~OBD_CONNECT_OSS_CAPA;
1481 cfs_spin_lock(&exp->exp_lock);
1482 exp->exp_connect_flags = reply->ocd_connect_flags;
1483 cfs_spin_unlock(&exp->exp_lock);
1495 * this should be done in filter_connect()/filter_reconnect(), but
1496 * we can't obtain information like NID, which stored in incoming
1497 * request, thus can't decide what flavor to use. so we do it here.
1499 * This hack should be removed after the OST stack be rewritten, just
1500 * like what we are doing in mdt_obd_connect()/mdt_obd_reconnect().
1502 static int ost_connect_check_sptlrpc(struct ptlrpc_request *req)
1504 struct obd_export *exp = req->rq_export;
1505 struct filter_obd *filter = &exp->exp_obd->u.filter;
1506 struct sptlrpc_flavor flvr;
1509 if (unlikely(strcmp(exp->exp_obd->obd_type->typ_name,
1510 LUSTRE_ECHO_NAME) == 0)) {
1511 exp->exp_flvr.sf_rpc = SPTLRPC_FLVR_ANY;
1515 if (exp->exp_flvr.sf_rpc == SPTLRPC_FLVR_INVALID) {
1516 cfs_read_lock(&filter->fo_sptlrpc_lock);
1517 sptlrpc_target_choose_flavor(&filter->fo_sptlrpc_rset,
1521 cfs_read_unlock(&filter->fo_sptlrpc_lock);
1523 cfs_spin_lock(&exp->exp_lock);
1525 exp->exp_sp_peer = req->rq_sp_from;
1526 exp->exp_flvr = flvr;
1528 if (exp->exp_flvr.sf_rpc != SPTLRPC_FLVR_ANY &&
1529 exp->exp_flvr.sf_rpc != req->rq_flvr.sf_rpc) {
1530 CERROR("unauthorized rpc flavor %x from %s, "
1531 "expect %x\n", req->rq_flvr.sf_rpc,
1532 libcfs_nid2str(req->rq_peer.nid),
1533 exp->exp_flvr.sf_rpc);
1537 cfs_spin_unlock(&exp->exp_lock);
1539 if (exp->exp_sp_peer != req->rq_sp_from) {
1540 CERROR("RPC source %s doesn't match %s\n",
1541 sptlrpc_part2name(req->rq_sp_from),
1542 sptlrpc_part2name(exp->exp_sp_peer));
1545 rc = sptlrpc_target_export_check(exp, req);
1552 /* Ensure that data and metadata are synced to the disk when lock is cancelled
1554 int ost_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
1555 void *data, int flag)
1558 __u32 sync_lock_cancel = 0;
1559 __u32 len = sizeof(sync_lock_cancel);
1564 rc = lu_env_init(&env, LCT_DT_THREAD);
1565 if (unlikely(rc != 0))
1568 rc = obd_get_info(&env, lock->l_export, sizeof(KEY_SYNC_LOCK_CANCEL),
1569 KEY_SYNC_LOCK_CANCEL, &len, &sync_lock_cancel, NULL);
1570 if (rc == 0 && flag == LDLM_CB_CANCELING &&
1571 (lock->l_granted_mode & (LCK_PW|LCK_GROUP)) &&
1572 (sync_lock_cancel == ALWAYS_SYNC_ON_CANCEL ||
1573 (sync_lock_cancel == BLOCKING_SYNC_ON_CANCEL &&
1574 lock->l_flags & LDLM_FL_CBPENDING))) {
1575 struct obd_info *oinfo;
1579 OBD_ALLOC_PTR(oinfo);
1581 GOTO(out_env, rc = -ENOMEM);
1584 OBD_FREE_PTR(oinfo);
1585 GOTO(out_env, rc = -ENOMEM);
1587 oa->o_id = lock->l_resource->lr_name.name[0];
1588 oa->o_seq = lock->l_resource->lr_name.name[1];
1589 oa->o_valid = OBD_MD_FLID|OBD_MD_FLGROUP;
1592 rc = obd_sync(&env, lock->l_export, oinfo,
1593 lock->l_policy_data.l_extent.start,
1594 lock->l_policy_data.l_extent.end, NULL);
1596 CERROR("Error %d syncing data on lock cancel\n", rc);
1599 OBD_FREE_PTR(oinfo);
1602 rc = ldlm_server_blocking_ast(lock, desc, data, flag);
1608 static int ost_filter_recovery_request(struct ptlrpc_request *req,
1609 struct obd_device *obd, int *process)
1611 switch (lustre_msg_get_opc(req->rq_reqmsg)) {
1612 case OST_CONNECT: /* This will never get here, but for completeness. */
1613 case OST_DISCONNECT:
1624 case OBD_LOG_CANCEL:
1626 *process = target_queue_recovery_request(req, obd);
1630 DEBUG_REQ(D_WARNING, req, "not permitted during recovery");
1636 int ost_msg_check_version(struct lustre_msg *msg)
1640 switch(lustre_msg_get_opc(msg)) {
1642 case OST_DISCONNECT:
1645 case SEC_CTX_INIT_CONT:
1647 rc = lustre_msg_check_version(msg, LUSTRE_OBD_VERSION);
1649 CERROR("bad opc %u version %08x, expecting %08x\n",
1650 lustre_msg_get_opc(msg),
1651 lustre_msg_get_version(msg),
1652 LUSTRE_OBD_VERSION);
1665 #ifdef HAVE_QUOTA_SUPPORT
1666 case OST_QUOTACHECK:
1668 case OST_QUOTA_ADJUST_QUNIT:
1670 rc = lustre_msg_check_version(msg, LUSTRE_OST_VERSION);
1672 CERROR("bad opc %u version %08x, expecting %08x\n",
1673 lustre_msg_get_opc(msg),
1674 lustre_msg_get_version(msg),
1675 LUSTRE_OST_VERSION);
1680 case LDLM_BL_CALLBACK:
1681 case LDLM_CP_CALLBACK:
1682 rc = lustre_msg_check_version(msg, LUSTRE_DLM_VERSION);
1684 CERROR("bad opc %u version %08x, expecting %08x\n",
1685 lustre_msg_get_opc(msg),
1686 lustre_msg_get_version(msg),
1687 LUSTRE_DLM_VERSION);
1689 case LLOG_ORIGIN_CONNECT:
1690 case OBD_LOG_CANCEL:
1691 rc = lustre_msg_check_version(msg, LUSTRE_LOG_VERSION);
1693 CERROR("bad opc %u version %08x, expecting %08x\n",
1694 lustre_msg_get_opc(msg),
1695 lustre_msg_get_version(msg),
1696 LUSTRE_LOG_VERSION);
1699 CERROR("Unexpected opcode %d\n", lustre_msg_get_opc(msg));
1705 struct ost_prolong_data {
1706 struct ptlrpc_request *opd_req;
1707 struct obd_export *opd_exp;
1708 struct obdo *opd_oa;
1709 struct ldlm_res_id opd_resid;
1710 struct ldlm_extent opd_extent;
1711 ldlm_mode_t opd_mode;
1712 unsigned int opd_locks;
1716 /* prolong locks for the current service time of the corresponding
1717 * portal (= OST_IO_PORTAL)
1719 static inline int prolong_timeout(struct ptlrpc_request *req)
1721 struct ptlrpc_service_part *svcpt = req->rq_rqbd->rqbd_svcpt;
1724 return obd_timeout / 2;
1726 return max(at_est2timeout(at_get(&svcpt->scp_at_estimate)),
1730 static void ost_prolong_lock_one(struct ost_prolong_data *opd,
1731 struct ldlm_lock *lock)
1733 LASSERT(lock->l_export == opd->opd_exp);
1735 if (lock->l_destroyed) /* lock already cancelled */
1738 /* XXX: never try to grab resource lock here because we're inside
1739 * exp_bl_list_lock; in ldlm_lockd.c to handle waiting list we take
1740 * res lock and then exp_bl_list_lock. */
1742 if (!(lock->l_flags & LDLM_FL_AST_SENT))
1743 /* ignore locks not being cancelled */
1747 "refreshed for req x"LPU64" ext("LPU64"->"LPU64") to %ds.\n",
1748 opd->opd_req->rq_xid, opd->opd_extent.start,
1749 opd->opd_extent.end, opd->opd_timeout);
1751 /* OK. this is a possible lock the user holds doing I/O
1752 * let's refresh eviction timer for it */
1753 ldlm_refresh_waiting_lock(lock, opd->opd_timeout);
1757 static void ost_prolong_locks(struct ost_prolong_data *data)
1759 struct obd_export *exp = data->opd_exp;
1760 struct obdo *oa = data->opd_oa;
1761 struct ldlm_lock *lock;
1764 if (oa->o_valid & OBD_MD_FLHANDLE) {
1765 /* mostly a request should be covered by only one lock, try
1767 lock = ldlm_handle2lock(&oa->o_handle);
1769 /* Fast path to check if the lock covers the whole IO
1770 * region exclusively. */
1771 if (lock->l_granted_mode == LCK_PW &&
1772 ldlm_extent_contain(&lock->l_policy_data.l_extent,
1773 &data->opd_extent)) {
1775 ost_prolong_lock_one(data, lock);
1776 LDLM_LOCK_PUT(lock);
1779 LDLM_LOCK_PUT(lock);
1784 cfs_spin_lock_bh(&exp->exp_bl_list_lock);
1785 cfs_list_for_each_entry(lock, &exp->exp_bl_list, l_exp_list) {
1786 LASSERT(lock->l_flags & LDLM_FL_AST_SENT);
1787 LASSERT(lock->l_resource->lr_type == LDLM_EXTENT);
1789 if (!ldlm_res_eq(&data->opd_resid, &lock->l_resource->lr_name))
1792 if (!ldlm_extent_overlap(&lock->l_policy_data.l_extent,
1796 ost_prolong_lock_one(data, lock);
1798 cfs_spin_unlock_bh(&exp->exp_bl_list_lock);
1804 * Returns 1 if the given PTLRPC matches the given LDLM locks, or 0 if it does
1807 static int ost_rw_hpreq_lock_match(struct ptlrpc_request *req,
1808 struct ldlm_lock *lock)
1810 struct niobuf_remote *nb;
1811 struct obd_ioobj *ioo;
1813 struct ldlm_extent ext;
1816 opc = lustre_msg_get_opc(req->rq_reqmsg);
1817 LASSERT(opc == OST_READ || opc == OST_WRITE);
1819 ioo = req_capsule_client_get(&req->rq_pill, &RMF_OBD_IOOBJ);
1820 LASSERT(ioo != NULL);
1822 nb = req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE);
1823 LASSERT(nb != NULL);
1825 ext.start = nb->offset;
1826 nb += ioo->ioo_bufcnt - 1;
1827 ext.end = nb->offset + nb->len - 1;
1829 LASSERT(lock->l_resource != NULL);
1830 if (!osc_res_name_eq(ioo->ioo_id, ioo->ioo_seq,
1831 &lock->l_resource->lr_name))
1835 if (opc == OST_READ)
1837 if (!(lock->l_granted_mode & mode))
1840 RETURN(ldlm_extent_overlap(&lock->l_policy_data.l_extent, &ext));
1844 * High-priority queue request check for whether the given PTLRPC request (\a
1845 * req) is blocking an LDLM lock cancel.
1847 * Returns 1 if the given given PTLRPC request (\a req) is blocking an LDLM lock
1848 * cancel, 0 if it is not, and -EFAULT if the request is malformed.
1850 * Only OST_READs, OST_WRITEs and OST_PUNCHes go on the h-p RPC queue. This
1851 * function looks only at OST_READs and OST_WRITEs.
1853 static int ost_rw_hpreq_check(struct ptlrpc_request *req)
1855 struct obd_device *obd = req->rq_export->exp_obd;
1856 struct ost_body *body;
1857 struct obd_ioobj *ioo;
1858 struct niobuf_remote *nb;
1859 struct ost_prolong_data opd = { 0 };
1864 * Use LASSERT to do sanity check because malformed RPCs should have
1865 * been filtered out in ost_hpreq_handler().
1867 opc = lustre_msg_get_opc(req->rq_reqmsg);
1868 LASSERT(opc == OST_READ || opc == OST_WRITE);
1870 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
1871 LASSERT(body != NULL);
1873 ioo = req_capsule_client_get(&req->rq_pill, &RMF_OBD_IOOBJ);
1874 LASSERT(ioo != NULL);
1876 nb = req_capsule_client_get(&req->rq_pill, &RMF_NIOBUF_REMOTE);
1877 LASSERT(nb != NULL);
1878 LASSERT(!(nb->flags & OBD_BRW_SRVLOCK));
1880 osc_build_res_name(ioo->ioo_id, ioo->ioo_seq, &opd.opd_resid);
1884 if (opc == OST_READ)
1886 opd.opd_mode = mode;
1887 opd.opd_exp = req->rq_export;
1888 opd.opd_oa = &body->oa;
1889 opd.opd_extent.start = nb->offset;
1890 nb += ioo->ioo_bufcnt - 1;
1891 opd.opd_extent.end = nb->offset + nb->len - 1;
1892 opd.opd_timeout = prolong_timeout(req);
1894 DEBUG_REQ(D_RPCTRACE, req,
1895 "%s %s: refresh rw locks: " LPU64"/"LPU64" ("LPU64"->"LPU64")\n",
1896 obd->obd_name, cfs_current()->comm,
1897 opd.opd_resid.name[0], opd.opd_resid.name[1],
1898 opd.opd_extent.start, opd.opd_extent.end);
1900 ost_prolong_locks(&opd);
1902 CDEBUG(D_DLMTRACE, "%s: refreshed %u locks timeout for req %p.\n",
1903 obd->obd_name, opd.opd_locks, req);
1905 RETURN(opd.opd_locks);
1908 static void ost_rw_hpreq_fini(struct ptlrpc_request *req)
1910 (void)ost_rw_hpreq_check(req);
1914 * Like ost_rw_hpreq_lock_match(), but for OST_PUNCH RPCs.
1916 static int ost_punch_hpreq_lock_match(struct ptlrpc_request *req,
1917 struct ldlm_lock *lock)
1919 struct ost_body *body;
1922 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
1923 LASSERT(body != NULL);
1925 if (body->oa.o_valid & OBD_MD_FLHANDLE &&
1926 body->oa.o_handle.cookie == lock->l_handle.h_cookie)
1933 * Like ost_rw_hpreq_check(), but for OST_PUNCH RPCs.
1935 static int ost_punch_hpreq_check(struct ptlrpc_request *req)
1937 struct obd_device *obd = req->rq_export->exp_obd;
1938 struct ost_body *body;
1940 struct ost_prolong_data opd = { 0 };
1944 body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
1945 LASSERT(body != NULL);
1948 LASSERT(!(oa->o_valid & OBD_MD_FLFLAGS) ||
1949 !(oa->o_flags & OBD_FL_SRVLOCK));
1952 end = start + oa->o_blocks;
1955 opd.opd_mode = LCK_PW;
1956 opd.opd_exp = req->rq_export;
1958 opd.opd_extent.start = start;
1959 opd.opd_extent.end = end;
1960 if (oa->o_blocks == OBD_OBJECT_EOF)
1961 opd.opd_extent.end = OBD_OBJECT_EOF;
1962 opd.opd_timeout = prolong_timeout(req);
1964 osc_build_res_name(oa->o_id, oa->o_seq, &opd.opd_resid);
1967 "%s: refresh locks: "LPU64"/"LPU64" ("LPU64"->"LPU64")\n",
1969 opd.opd_resid.name[0], opd.opd_resid.name[1],
1970 opd.opd_extent.start, opd.opd_extent.end);
1972 ost_prolong_locks(&opd);
1974 CDEBUG(D_DLMTRACE, "%s: refreshed %u locks timeout for req %p.\n",
1975 obd->obd_name, opd.opd_locks, req);
1977 RETURN(opd.opd_locks > 0);
1980 static void ost_punch_hpreq_fini(struct ptlrpc_request *req)
1982 (void)ost_punch_hpreq_check(req);
1985 struct ptlrpc_hpreq_ops ost_hpreq_rw = {
1986 .hpreq_lock_match = ost_rw_hpreq_lock_match,
1987 .hpreq_check = ost_rw_hpreq_check,
1988 .hpreq_fini = ost_rw_hpreq_fini
1991 struct ptlrpc_hpreq_ops ost_hpreq_punch = {
1992 .hpreq_lock_match = ost_punch_hpreq_lock_match,
1993 .hpreq_check = ost_punch_hpreq_check,
1994 .hpreq_fini = ost_punch_hpreq_fini
1997 /** Assign high priority operations to the request if needed. */
1998 static int ost_hpreq_handler(struct ptlrpc_request *req)
2001 if (req->rq_export) {
2002 int opc = lustre_msg_get_opc(req->rq_reqmsg);
2003 struct ost_body *body;
2005 if (opc == OST_READ || opc == OST_WRITE) {
2006 struct niobuf_remote *nb;
2007 struct obd_ioobj *ioo;
2008 int objcount, niocount;
2012 /* RPCs on the H-P queue can be inspected before
2013 * ost_handler() initializes their pills, so we
2014 * initialize that here. Capsule initialization is
2015 * idempotent, as is setting the pill's format (provided
2016 * it doesn't change).
2018 req_capsule_init(&req->rq_pill, req, RCL_SERVER);
2019 if (opc == OST_READ)
2020 req_capsule_set(&req->rq_pill,
2023 req_capsule_set(&req->rq_pill,
2024 &RQF_OST_BRW_WRITE);
2026 body = req_capsule_client_get(&req->rq_pill,
2029 CERROR("Missing/short ost_body\n");
2033 objcount = req_capsule_get_size(&req->rq_pill,
2037 if (objcount == 0) {
2038 CERROR("Missing/short ioobj\n");
2042 CERROR("too many ioobjs (%d)\n", objcount);
2046 ioo = req_capsule_client_get(&req->rq_pill,
2049 CERROR("Missing/short ioobj\n");
2053 rc = ost_validate_obdo(req->rq_export, &body->oa, ioo);
2055 CERROR("invalid object ids\n");
2059 for (niocount = i = 0; i < objcount; i++) {
2060 if (ioo[i].ioo_bufcnt == 0) {
2061 CERROR("ioo[%d] has zero bufcnt\n", i);
2064 niocount += ioo[i].ioo_bufcnt;
2066 if (niocount > PTLRPC_MAX_BRW_PAGES) {
2067 DEBUG_REQ(D_RPCTRACE, req,
2068 "bulk has too many pages (%d)",
2073 nb = req_capsule_client_get(&req->rq_pill,
2074 &RMF_NIOBUF_REMOTE);
2076 CERROR("Missing/short niobuf\n");
2080 if (niocount == 0 || !(nb[0].flags & OBD_BRW_SRVLOCK))
2081 req->rq_ops = &ost_hpreq_rw;
2082 } else if (opc == OST_PUNCH) {
2083 req_capsule_init(&req->rq_pill, req, RCL_SERVER);
2084 req_capsule_set(&req->rq_pill, &RQF_OST_PUNCH);
2086 body = req_capsule_client_get(&req->rq_pill,
2089 CERROR("Missing/short ost_body\n");
2093 if (!(body->oa.o_valid & OBD_MD_FLFLAGS) ||
2094 !(body->oa.o_flags & OBD_FL_SRVLOCK))
2095 req->rq_ops = &ost_hpreq_punch;
2101 /* TODO: handle requests in a similar way as MDT: see mdt_handle_common() */
2102 int ost_handle(struct ptlrpc_request *req)
2104 struct obd_trans_info trans_info = { 0, };
2105 struct obd_trans_info *oti = &trans_info;
2106 int should_process, fail = OBD_FAIL_OST_ALL_REPLY_NET, rc = 0;
2107 struct obd_device *obd = NULL;
2110 /* OST module is kept between remounts, but the last reference
2111 * to specific module (say, osd or ofd) kills all related keys
2112 * from the environment. so we have to refill it until the root
2113 * cause is fixed properly */
2114 lu_env_refill(req->rq_svc_thread->t_env);
2116 LASSERT(current->journal_info == NULL);
2118 /* primordial rpcs don't affect server recovery */
2119 switch (lustre_msg_get_opc(req->rq_reqmsg)) {
2121 case SEC_CTX_INIT_CONT:
2126 req_capsule_init(&req->rq_pill, req, RCL_SERVER);
2128 if (lustre_msg_get_opc(req->rq_reqmsg) != OST_CONNECT) {
2129 if (!class_connected_export(req->rq_export)) {
2130 CDEBUG(D_HA,"operation %d on unconnected OST from %s\n",
2131 lustre_msg_get_opc(req->rq_reqmsg),
2132 libcfs_id2str(req->rq_peer));
2133 req->rq_status = -ENOTCONN;
2134 GOTO(out, rc = -ENOTCONN);
2137 obd = req->rq_export->exp_obd;
2139 /* Check for aborted recovery. */
2140 if (obd->obd_recovering) {
2141 rc = ost_filter_recovery_request(req, obd,
2143 if (rc || !should_process)
2145 else if (should_process < 0) {
2146 req->rq_status = should_process;
2147 rc = ptlrpc_error(req);
2155 rc = ost_msg_check_version(req->rq_reqmsg);
2159 if (req && req->rq_reqmsg && req->rq_export &&
2160 (req->rq_export->exp_connect_flags & OBD_CONNECT_JOBSTATS))
2161 oti->oti_jobid = lustre_msg_get_jobid(req->rq_reqmsg);
2163 switch (lustre_msg_get_opc(req->rq_reqmsg)) {
2165 CDEBUG(D_INODE, "connect\n");
2166 req_capsule_set(&req->rq_pill, &RQF_OST_CONNECT);
2167 if (OBD_FAIL_CHECK(OBD_FAIL_OST_CONNECT_NET))
2169 rc = target_handle_connect(req);
2170 if (OBD_FAIL_CHECK(OBD_FAIL_OST_CONNECT_NET2))
2173 rc = ost_init_sec_level(req);
2175 rc = ost_connect_check_sptlrpc(req);
2179 case OST_DISCONNECT:
2180 CDEBUG(D_INODE, "disconnect\n");
2181 req_capsule_set(&req->rq_pill, &RQF_OST_DISCONNECT);
2182 if (OBD_FAIL_CHECK(OBD_FAIL_OST_DISCONNECT_NET))
2184 rc = target_handle_disconnect(req);
2187 CDEBUG(D_INODE, "create\n");
2188 req_capsule_set(&req->rq_pill, &RQF_OST_CREATE);
2189 if (OBD_FAIL_CHECK(OBD_FAIL_OST_CREATE_NET))
2191 if (OBD_FAIL_CHECK(OBD_FAIL_OST_EROFS))
2192 GOTO(out, rc = -EROFS);
2193 rc = ost_create(req->rq_export, req, oti);
2196 CDEBUG(D_INODE, "destroy\n");
2197 req_capsule_set(&req->rq_pill, &RQF_OST_DESTROY);
2198 if (OBD_FAIL_CHECK(OBD_FAIL_OST_DESTROY_NET))
2200 if (OBD_FAIL_CHECK(OBD_FAIL_OST_EROFS))
2201 GOTO(out, rc = -EROFS);
2202 rc = ost_destroy(req->rq_export, req, oti);
2205 CDEBUG(D_INODE, "getattr\n");
2206 req_capsule_set(&req->rq_pill, &RQF_OST_GETATTR);
2207 if (OBD_FAIL_CHECK(OBD_FAIL_OST_GETATTR_NET))
2209 rc = ost_getattr(req->rq_export, req);
2212 CDEBUG(D_INODE, "setattr\n");
2213 req_capsule_set(&req->rq_pill, &RQF_OST_SETATTR);
2214 if (OBD_FAIL_CHECK(OBD_FAIL_OST_SETATTR_NET))
2216 rc = ost_setattr(req->rq_export, req, oti);
2219 req_capsule_set(&req->rq_pill, &RQF_OST_BRW_WRITE);
2220 CDEBUG(D_INODE, "write\n");
2221 /* req->rq_request_portal would be nice, if it was set */
2222 if (ptlrpc_req2svc(req)->srv_req_portal != OST_IO_PORTAL) {
2223 CERROR("%s: deny write request from %s to portal %u\n",
2224 req->rq_export->exp_obd->obd_name,
2225 obd_export_nid2str(req->rq_export),
2226 ptlrpc_req2svc(req)->srv_req_portal);
2227 GOTO(out, rc = -EPROTO);
2229 if (OBD_FAIL_CHECK(OBD_FAIL_OST_BRW_NET))
2231 if (OBD_FAIL_CHECK(OBD_FAIL_OST_ENOSPC))
2232 GOTO(out, rc = -ENOSPC);
2233 if (OBD_FAIL_CHECK(OBD_FAIL_OST_EROFS))
2234 GOTO(out, rc = -EROFS);
2235 rc = ost_brw_write(req, oti);
2236 LASSERT(current->journal_info == NULL);
2237 /* ost_brw_write sends its own replies */
2240 req_capsule_set(&req->rq_pill, &RQF_OST_BRW_READ);
2241 CDEBUG(D_INODE, "read\n");
2242 /* req->rq_request_portal would be nice, if it was set */
2243 if (ptlrpc_req2svc(req)->srv_req_portal != OST_IO_PORTAL) {
2244 CERROR("%s: deny read request from %s to portal %u\n",
2245 req->rq_export->exp_obd->obd_name,
2246 obd_export_nid2str(req->rq_export),
2247 ptlrpc_req2svc(req)->srv_req_portal);
2248 GOTO(out, rc = -EPROTO);
2250 if (OBD_FAIL_CHECK(OBD_FAIL_OST_BRW_NET))
2252 rc = ost_brw_read(req, oti);
2253 LASSERT(current->journal_info == NULL);
2254 /* ost_brw_read sends its own replies */
2257 CDEBUG(D_INODE, "punch\n");
2258 req_capsule_set(&req->rq_pill, &RQF_OST_PUNCH);
2259 if (OBD_FAIL_CHECK(OBD_FAIL_OST_PUNCH_NET))
2261 if (OBD_FAIL_CHECK(OBD_FAIL_OST_EROFS))
2262 GOTO(out, rc = -EROFS);
2263 rc = ost_punch(req->rq_export, req, oti);
2266 CDEBUG(D_INODE, "statfs\n");
2267 req_capsule_set(&req->rq_pill, &RQF_OST_STATFS);
2268 if (OBD_FAIL_CHECK(OBD_FAIL_OST_STATFS_NET))
2270 rc = ost_statfs(req);
2273 CDEBUG(D_INODE, "sync\n");
2274 req_capsule_set(&req->rq_pill, &RQF_OST_SYNC);
2275 if (OBD_FAIL_CHECK(OBD_FAIL_OST_SYNC_NET))
2277 rc = ost_sync(req->rq_export, req, oti);
2280 DEBUG_REQ(D_INODE, req, "set_info");
2281 req_capsule_set(&req->rq_pill, &RQF_OBD_SET_INFO);
2282 rc = ost_set_info(req->rq_export, req);
2285 DEBUG_REQ(D_INODE, req, "get_info");
2286 req_capsule_set(&req->rq_pill, &RQF_OST_GET_INFO_GENERIC);
2287 rc = ost_get_info(req->rq_export, req);
2289 #ifdef HAVE_QUOTA_SUPPORT
2290 case OST_QUOTACHECK:
2291 CDEBUG(D_INODE, "quotacheck\n");
2292 req_capsule_set(&req->rq_pill, &RQF_OST_QUOTACHECK);
2293 if (OBD_FAIL_CHECK(OBD_FAIL_OST_QUOTACHECK_NET))
2295 rc = ost_handle_quotacheck(req);
2298 CDEBUG(D_INODE, "quotactl\n");
2299 req_capsule_set(&req->rq_pill, &RQF_OST_QUOTACTL);
2300 if (OBD_FAIL_CHECK(OBD_FAIL_OST_QUOTACTL_NET))
2302 rc = ost_handle_quotactl(req);
2304 case OST_QUOTA_ADJUST_QUNIT:
2305 CDEBUG(D_INODE, "quota_adjust_qunit\n");
2306 req_capsule_set(&req->rq_pill, &RQF_OST_QUOTA_ADJUST_QUNIT);
2307 rc = ost_handle_quota_adjust_qunit(req);
2311 DEBUG_REQ(D_INODE, req, "ping");
2312 req_capsule_set(&req->rq_pill, &RQF_OBD_PING);
2313 rc = target_handle_ping(req);
2315 /* FIXME - just reply status */
2316 case LLOG_ORIGIN_CONNECT:
2317 DEBUG_REQ(D_INODE, req, "log connect");
2318 req_capsule_set(&req->rq_pill, &RQF_LLOG_ORIGIN_CONNECT);
2319 rc = ost_llog_handle_connect(req->rq_export, req);
2320 req->rq_status = rc;
2321 rc = req_capsule_server_pack(&req->rq_pill);
2324 RETURN(ptlrpc_reply(req));
2325 case OBD_LOG_CANCEL:
2326 CDEBUG(D_INODE, "log cancel\n");
2327 req_capsule_set(&req->rq_pill, &RQF_LOG_CANCEL);
2328 if (OBD_FAIL_CHECK(OBD_FAIL_OBD_LOG_CANCEL_NET))
2330 rc = llog_origin_handle_cancel(req);
2331 if (OBD_FAIL_CHECK(OBD_FAIL_OBD_LOG_CANCEL_REP))
2333 req->rq_status = rc;
2334 rc = req_capsule_server_pack(&req->rq_pill);
2337 RETURN(ptlrpc_reply(req));
2339 CDEBUG(D_INODE, "enqueue\n");
2340 req_capsule_set(&req->rq_pill, &RQF_LDLM_ENQUEUE);
2341 if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_ENQUEUE))
2343 rc = ldlm_handle_enqueue(req, ldlm_server_completion_ast,
2345 ldlm_server_glimpse_ast);
2346 fail = OBD_FAIL_OST_LDLM_REPLY_NET;
2349 CDEBUG(D_INODE, "convert\n");
2350 req_capsule_set(&req->rq_pill, &RQF_LDLM_CONVERT);
2351 if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_CONVERT))
2353 rc = ldlm_handle_convert(req);
2356 CDEBUG(D_INODE, "cancel\n");
2357 req_capsule_set(&req->rq_pill, &RQF_LDLM_CANCEL);
2358 if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_CANCEL))
2360 rc = ldlm_handle_cancel(req);
2362 case LDLM_BL_CALLBACK:
2363 case LDLM_CP_CALLBACK:
2364 CDEBUG(D_INODE, "callback\n");
2365 CERROR("callbacks should not happen on OST\n");
2368 CERROR("Unexpected opcode %d\n",
2369 lustre_msg_get_opc(req->rq_reqmsg));
2370 req->rq_status = -ENOTSUPP;
2371 rc = ptlrpc_error(req);
2375 LASSERT(current->journal_info == NULL);
2378 /* If we're DISCONNECTing, the export_data is already freed */
2379 if (!rc && lustre_msg_get_opc(req->rq_reqmsg) != OST_DISCONNECT)
2380 target_committed_to_req(req);
2384 oti_to_request(oti, req);
2386 target_send_reply(req, rc, fail);
2389 EXPORT_SYMBOL(ost_handle);
2391 * free per-thread pool created by ost_thread_init().
2393 static void ost_thread_done(struct ptlrpc_thread *thread)
2395 struct ost_thread_local_cache *tls; /* TLS stands for Thread-Local
2400 LASSERT(thread != NULL);
2403 * be prepared to handle partially-initialized pools (because this is
2404 * called from ost_thread_init() for cleanup.
2406 tls = thread->t_data;
2409 thread->t_data = NULL;
2415 * initialize per-thread page pool (bug 5137).
2417 static int ost_thread_init(struct ptlrpc_thread *thread)
2419 struct ost_thread_local_cache *tls;
2423 LASSERT(thread != NULL);
2424 LASSERT(thread->t_data == NULL);
2429 thread->t_data = tls;
2433 #define OST_WATCHDOG_TIMEOUT (obd_timeout * 1000)
2435 static struct cfs_cpt_table *ost_io_cptable;
2437 /* Sigh - really, this is an OSS, the _server_, not the _target_ */
2438 static int ost_setup(struct obd_device *obd, struct lustre_cfg* lcfg)
2440 static struct ptlrpc_service_conf svc_conf;
2441 struct ost_obd *ost = &obd->u.ost;
2442 struct lprocfs_static_vars lvars;
2447 rc = cfs_cleanup_group_info();
2451 lprocfs_ost_init_vars(&lvars);
2452 lprocfs_obd_setup(obd, lvars.obd_vars);
2454 cfs_mutex_init(&ost->ost_health_mutex);
2456 svc_conf = (typeof(svc_conf)) {
2457 .psc_name = LUSTRE_OSS_NAME,
2458 .psc_watchdog_factor = OSS_SERVICE_WATCHDOG_FACTOR,
2460 .bc_nbufs = OST_NBUFS,
2461 .bc_buf_size = OST_BUFSIZE,
2462 .bc_req_max_size = OST_MAXREQSIZE,
2463 .bc_rep_max_size = OST_MAXREPSIZE,
2464 .bc_req_portal = OST_REQUEST_PORTAL,
2465 .bc_rep_portal = OSC_REPLY_PORTAL,
2468 .tc_thr_name = "ll_ost",
2469 .tc_thr_factor = OSS_THR_FACTOR,
2470 .tc_nthrs_init = OSS_NTHRS_INIT,
2471 .tc_nthrs_base = OSS_NTHRS_BASE,
2472 .tc_nthrs_max = OSS_NTHRS_MAX,
2473 .tc_nthrs_user = oss_num_threads,
2474 .tc_cpu_affinity = 1,
2475 .tc_ctx_tags = LCT_DT_THREAD,
2478 .cc_pattern = oss_cpts,
2481 .so_req_handler = ost_handle,
2482 .so_req_printer = target_print_req,
2483 .so_hpreq_handler = ptlrpc_hpreq_handler,
2486 ost->ost_service = ptlrpc_register_service(&svc_conf,
2487 obd->obd_proc_entry);
2488 if (IS_ERR(ost->ost_service)) {
2489 rc = PTR_ERR(ost->ost_service);
2490 CERROR("failed to start service: %d\n", rc);
2491 GOTO(out_lprocfs, rc);
2494 memset(&svc_conf, 0, sizeof(svc_conf));
2495 svc_conf = (typeof(svc_conf)) {
2496 .psc_name = "ost_create",
2497 .psc_watchdog_factor = OSS_SERVICE_WATCHDOG_FACTOR,
2499 .bc_nbufs = OST_NBUFS,
2500 .bc_buf_size = OST_BUFSIZE,
2501 .bc_req_max_size = OST_MAXREQSIZE,
2502 .bc_rep_max_size = OST_MAXREPSIZE,
2503 .bc_req_portal = OST_CREATE_PORTAL,
2504 .bc_rep_portal = OSC_REPLY_PORTAL,
2507 .tc_thr_name = "ll_ost_create",
2508 .tc_thr_factor = OSS_CR_THR_FACTOR,
2509 .tc_nthrs_init = OSS_CR_NTHRS_INIT,
2510 .tc_nthrs_base = OSS_CR_NTHRS_BASE,
2511 .tc_nthrs_max = OSS_CR_NTHRS_MAX,
2512 .tc_nthrs_user = oss_num_create_threads,
2513 .tc_cpu_affinity = 1,
2514 .tc_ctx_tags = LCT_DT_THREAD,
2517 .cc_pattern = oss_cpts,
2520 .so_req_handler = ost_handle,
2521 .so_req_printer = target_print_req,
2522 .so_hpreq_handler = NULL,
2525 ost->ost_create_service = ptlrpc_register_service(&svc_conf,
2526 obd->obd_proc_entry);
2527 if (IS_ERR(ost->ost_create_service)) {
2528 rc = PTR_ERR(ost->ost_create_service);
2529 CERROR("failed to start OST create service: %d\n", rc);
2530 GOTO(out_service, rc);
2533 mask = cfs_cpt_table->ctb_nodemask;
2534 /* event CPT feature is disabled in libcfs level by set partition
2535 * number to 1, we still want to set node affinity for io service */
2536 if (cfs_cpt_number(cfs_cpt_table) == 1 && nodes_weight(*mask) > 1) {
2540 ost_io_cptable = cfs_cpt_table_alloc(nodes_weight(*mask));
2541 for_each_node_mask(i, *mask) {
2542 if (ost_io_cptable == NULL) {
2543 CWARN("OSS failed to create CPT table\n");
2547 rc = cfs_cpt_set_node(ost_io_cptable, cpt++, i);
2549 CWARN("OSS Failed to set node %d for"
2550 "IO CPT table\n", i);
2551 cfs_cpt_table_free(ost_io_cptable);
2552 ost_io_cptable = NULL;
2558 memset(&svc_conf, 0, sizeof(svc_conf));
2559 svc_conf = (typeof(svc_conf)) {
2560 .psc_name = "ost_io",
2561 .psc_watchdog_factor = OSS_SERVICE_WATCHDOG_FACTOR,
2563 .bc_nbufs = OST_NBUFS,
2564 .bc_buf_size = OST_BUFSIZE,
2565 .bc_req_max_size = OST_MAXREQSIZE,
2566 .bc_rep_max_size = OST_MAXREPSIZE,
2567 .bc_req_portal = OST_IO_PORTAL,
2568 .bc_rep_portal = OSC_REPLY_PORTAL,
2571 .tc_thr_name = "ll_ost_io",
2572 .tc_thr_factor = OSS_THR_FACTOR,
2573 .tc_nthrs_init = OSS_NTHRS_INIT,
2574 .tc_nthrs_base = OSS_NTHRS_BASE,
2575 .tc_nthrs_max = OSS_NTHRS_MAX,
2576 .tc_nthrs_user = oss_num_threads,
2577 .tc_cpu_affinity = 1,
2578 .tc_ctx_tags = LCT_DT_THREAD,
2581 .cc_cptable = ost_io_cptable,
2582 .cc_pattern = ost_io_cptable == NULL ?
2586 .so_thr_init = ost_thread_init,
2587 .so_thr_done = ost_thread_done,
2588 .so_req_handler = ost_handle,
2589 .so_hpreq_handler = ost_hpreq_handler,
2590 .so_req_printer = target_print_req,
2591 .so_hpreq_handler = NULL,
2594 ost->ost_io_service = ptlrpc_register_service(&svc_conf,
2595 obd->obd_proc_entry);
2596 if (IS_ERR(ost->ost_io_service)) {
2597 rc = PTR_ERR(ost->ost_io_service);
2598 CERROR("failed to start OST I/O service: %d\n", rc);
2599 ost->ost_io_service = NULL;
2600 GOTO(out_create, rc);
2603 ping_evictor_start();
2608 ptlrpc_unregister_service(ost->ost_create_service);
2609 ost->ost_create_service = NULL;
2611 ptlrpc_unregister_service(ost->ost_service);
2612 ost->ost_service = NULL;
2614 lprocfs_obd_cleanup(obd);
2618 static int ost_cleanup(struct obd_device *obd)
2620 struct ost_obd *ost = &obd->u.ost;
2624 ping_evictor_stop();
2626 /* there is no recovery for OST OBD, all recovery is controlled by
2628 LASSERT(obd->obd_recovering == 0);
2629 cfs_mutex_lock(&ost->ost_health_mutex);
2630 ptlrpc_unregister_service(ost->ost_service);
2631 ptlrpc_unregister_service(ost->ost_create_service);
2632 ptlrpc_unregister_service(ost->ost_io_service);
2633 ost->ost_service = NULL;
2634 ost->ost_create_service = NULL;
2635 ost->ost_io_service = NULL;
2637 cfs_mutex_unlock(&ost->ost_health_mutex);
2639 lprocfs_obd_cleanup(obd);
2641 if (ost_io_cptable != NULL) {
2642 cfs_cpt_table_free(ost_io_cptable);
2643 ost_io_cptable = NULL;
2649 static int ost_health_check(const struct lu_env *env, struct obd_device *obd)
2651 struct ost_obd *ost = &obd->u.ost;
2654 cfs_mutex_lock(&ost->ost_health_mutex);
2655 rc |= ptlrpc_service_health_check(ost->ost_service);
2656 rc |= ptlrpc_service_health_check(ost->ost_create_service);
2657 rc |= ptlrpc_service_health_check(ost->ost_io_service);
2658 cfs_mutex_unlock(&ost->ost_health_mutex);
2661 * health_check to return 0 on healthy
2662 * and 1 on unhealthy.
2670 struct ost_thread_local_cache *ost_tls(struct ptlrpc_request *r)
2672 return (struct ost_thread_local_cache *)(r->rq_svc_thread->t_data);
2675 /* use obd ops to offer management infrastructure */
2676 static struct obd_ops ost_obd_ops = {
2677 .o_owner = THIS_MODULE,
2678 .o_setup = ost_setup,
2679 .o_cleanup = ost_cleanup,
2680 .o_health_check = ost_health_check,
2684 static int __init ost_init(void)
2686 struct lprocfs_static_vars lvars;
2690 lprocfs_ost_init_vars(&lvars);
2691 rc = class_register_type(&ost_obd_ops, NULL, lvars.module_vars,
2692 LUSTRE_OSS_NAME, NULL);
2694 if (ost_num_threads != 0 && oss_num_threads == 0) {
2695 LCONSOLE_INFO("ost_num_threads module parameter is deprecated, "
2696 "use oss_num_threads instead or unset both for "
2697 "dynamic thread startup\n");
2698 oss_num_threads = ost_num_threads;
2704 static void /*__exit*/ ost_exit(void)
2706 class_unregister_type(LUSTRE_OSS_NAME);
2709 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
2710 MODULE_DESCRIPTION("Lustre Object Storage Target (OST) v0.01");
2711 MODULE_LICENSE("GPL");
2713 module_init(ost_init);
2714 module_exit(ost_exit);