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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2015, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
33 #define DEBUG_SUBSYSTEM S_ECHO
34 #include <linux/libcfs/libcfs.h>
37 #include <obd_support.h>
38 #include <obd_class.h>
39 #include <lustre_debug.h>
40 #include <lprocfs_status.h>
41 #include <cl_object.h>
42 #include <lustre_fid.h>
43 #include <lustre_acl.h>
44 #include <uapi/linux/lustre/lustre_ioctl.h>
45 #include <lustre_net.h>
47 #include "echo_internal.h"
49 /** \defgroup echo_client Echo Client
54 struct cl_device ed_cl;
55 struct echo_client_obd *ed_ec;
57 struct cl_site ed_site_myself;
58 struct lu_site *ed_site;
59 struct lu_device *ed_next;
63 struct cl_object eo_cl;
64 struct cl_object_header eo_hdr;
66 struct echo_device *eo_dev;
67 struct list_head eo_obj_chain;
68 struct lov_oinfo *eo_oinfo;
73 struct echo_object_conf {
74 struct cl_object_conf eoc_cl;
75 struct lov_oinfo **eoc_oinfo;
79 struct cl_page_slice ep_cl;
84 struct cl_lock_slice el_cl;
85 struct list_head el_chain;
86 struct echo_object *el_object;
91 static int echo_client_setup(const struct lu_env *env,
92 struct obd_device *obddev,
93 struct lustre_cfg *lcfg);
94 static int echo_client_cleanup(struct obd_device *obddev);
96 /** \defgroup echo_helpers Helper functions
99 static inline struct echo_device *cl2echo_dev(const struct cl_device *dev)
101 return container_of0(dev, struct echo_device, ed_cl);
104 static inline struct cl_device *echo_dev2cl(struct echo_device *d)
109 static inline struct echo_device *obd2echo_dev(const struct obd_device *obd)
111 return cl2echo_dev(lu2cl_dev(obd->obd_lu_dev));
114 static inline struct cl_object *echo_obj2cl(struct echo_object *eco)
119 static inline struct echo_object *cl2echo_obj(const struct cl_object *o)
121 return container_of(o, struct echo_object, eo_cl);
124 static inline struct echo_page *cl2echo_page(const struct cl_page_slice *s)
126 return container_of(s, struct echo_page, ep_cl);
129 static inline struct echo_lock *cl2echo_lock(const struct cl_lock_slice *s)
131 return container_of(s, struct echo_lock, el_cl);
134 static inline struct cl_lock *echo_lock2cl(const struct echo_lock *ecl)
136 return ecl->el_cl.cls_lock;
139 static struct lu_context_key echo_thread_key;
140 static inline struct echo_thread_info *echo_env_info(const struct lu_env *env)
142 struct echo_thread_info *info;
144 info = lu_context_key_get(&env->le_ctx, &echo_thread_key);
150 struct echo_object_conf *cl2echo_conf(const struct cl_object_conf *c)
152 return container_of(c, struct echo_object_conf, eoc_cl);
155 /** @} echo_helpers */
156 static int cl_echo_object_put(struct echo_object *eco);
157 static int cl_echo_object_brw(struct echo_object *eco, int rw, u64 offset,
158 struct page **pages, int npages, int async);
160 struct echo_thread_info {
161 struct echo_object_conf eti_conf;
162 struct lustre_md eti_md;
164 struct cl_2queue eti_queue;
166 struct cl_lock eti_lock;
167 struct lu_fid eti_fid;
168 struct lu_fid eti_fid2;
171 /* No session used right now */
172 struct echo_session_info {
176 static struct kmem_cache *echo_lock_kmem;
177 static struct kmem_cache *echo_object_kmem;
178 static struct kmem_cache *echo_thread_kmem;
179 static struct kmem_cache *echo_session_kmem;
181 static struct lu_kmem_descr echo_caches[] = {
183 .ckd_cache = &echo_lock_kmem,
184 .ckd_name = "echo_lock_kmem",
185 .ckd_size = sizeof(struct echo_lock)
188 .ckd_cache = &echo_object_kmem,
189 .ckd_name = "echo_object_kmem",
190 .ckd_size = sizeof(struct echo_object)
193 .ckd_cache = &echo_thread_kmem,
194 .ckd_name = "echo_thread_kmem",
195 .ckd_size = sizeof(struct echo_thread_info)
198 .ckd_cache = &echo_session_kmem,
199 .ckd_name = "echo_session_kmem",
200 .ckd_size = sizeof(struct echo_session_info)
207 /** \defgroup echo_page Page operations
209 * Echo page operations.
213 static int echo_page_own(const struct lu_env *env,
214 const struct cl_page_slice *slice,
215 struct cl_io *io, int nonblock)
217 struct echo_page *ep = cl2echo_page(slice);
220 mutex_lock(&ep->ep_lock);
221 else if (!mutex_trylock(&ep->ep_lock))
226 static void echo_page_disown(const struct lu_env *env,
227 const struct cl_page_slice *slice,
230 struct echo_page *ep = cl2echo_page(slice);
232 LASSERT(mutex_is_locked(&ep->ep_lock));
233 mutex_unlock(&ep->ep_lock);
236 static void echo_page_discard(const struct lu_env *env,
237 const struct cl_page_slice *slice,
238 struct cl_io *unused)
240 cl_page_delete(env, slice->cpl_page);
243 static int echo_page_is_vmlocked(const struct lu_env *env,
244 const struct cl_page_slice *slice)
246 if (mutex_is_locked(&cl2echo_page(slice)->ep_lock))
251 static void echo_page_completion(const struct lu_env *env,
252 const struct cl_page_slice *slice,
255 LASSERT(slice->cpl_page->cp_sync_io);
258 static void echo_page_fini(const struct lu_env *env,
259 struct cl_page_slice *slice)
261 struct echo_object *eco = cl2echo_obj(slice->cpl_obj);
263 atomic_dec(&eco->eo_npages);
264 put_page(slice->cpl_page->cp_vmpage);
267 static int echo_page_prep(const struct lu_env *env,
268 const struct cl_page_slice *slice,
269 struct cl_io *unused)
274 static int echo_page_print(const struct lu_env *env,
275 const struct cl_page_slice *slice,
276 void *cookie, lu_printer_t printer)
278 struct echo_page *ep = cl2echo_page(slice);
280 (*printer)(env, cookie, LUSTRE_ECHO_CLIENT_NAME "-page@%p %d vm@%p\n",
281 ep, mutex_is_locked(&ep->ep_lock),
282 slice->cpl_page->cp_vmpage);
286 static const struct cl_page_operations echo_page_ops = {
287 .cpo_own = echo_page_own,
288 .cpo_disown = echo_page_disown,
289 .cpo_discard = echo_page_discard,
290 .cpo_fini = echo_page_fini,
291 .cpo_print = echo_page_print,
292 .cpo_is_vmlocked = echo_page_is_vmlocked,
295 .cpo_prep = echo_page_prep,
296 .cpo_completion = echo_page_completion,
299 .cpo_prep = echo_page_prep,
300 .cpo_completion = echo_page_completion,
307 /** \defgroup echo_lock Locking
309 * echo lock operations
313 static void echo_lock_fini(const struct lu_env *env,
314 struct cl_lock_slice *slice)
316 struct echo_lock *ecl = cl2echo_lock(slice);
318 LASSERT(list_empty(&ecl->el_chain));
319 kmem_cache_free(echo_lock_kmem, ecl);
322 static const struct cl_lock_operations echo_lock_ops = {
323 .clo_fini = echo_lock_fini,
328 /** \defgroup echo_cl_ops cl_object operations
330 * operations for cl_object
334 static int echo_page_init(const struct lu_env *env, struct cl_object *obj,
335 struct cl_page *page, pgoff_t index)
337 struct echo_page *ep = cl_object_page_slice(obj, page);
338 struct echo_object *eco = cl2echo_obj(obj);
340 get_page(page->cp_vmpage);
341 mutex_init(&ep->ep_lock);
342 cl_page_slice_add(page, &ep->ep_cl, obj, index, &echo_page_ops);
343 atomic_inc(&eco->eo_npages);
347 static int echo_io_init(const struct lu_env *env, struct cl_object *obj,
353 static int echo_lock_init(const struct lu_env *env,
354 struct cl_object *obj, struct cl_lock *lock,
355 const struct cl_io *unused)
357 struct echo_lock *el;
359 el = kmem_cache_zalloc(echo_lock_kmem, GFP_NOFS);
361 cl_lock_slice_add(lock, &el->el_cl, obj, &echo_lock_ops);
362 el->el_object = cl2echo_obj(obj);
363 INIT_LIST_HEAD(&el->el_chain);
364 atomic_set(&el->el_refcount, 0);
366 return !el ? -ENOMEM : 0;
369 static int echo_conf_set(const struct lu_env *env, struct cl_object *obj,
370 const struct cl_object_conf *conf)
375 static const struct cl_object_operations echo_cl_obj_ops = {
376 .coo_page_init = echo_page_init,
377 .coo_lock_init = echo_lock_init,
378 .coo_io_init = echo_io_init,
379 .coo_conf_set = echo_conf_set
382 /** @} echo_cl_ops */
384 /** \defgroup echo_lu_ops lu_object operations
386 * operations for echo lu object.
390 static int echo_object_init(const struct lu_env *env, struct lu_object *obj,
391 const struct lu_object_conf *conf)
393 struct echo_device *ed = cl2echo_dev(lu2cl_dev(obj->lo_dev));
394 struct echo_client_obd *ec = ed->ed_ec;
395 struct echo_object *eco = cl2echo_obj(lu2cl(obj));
396 const struct cl_object_conf *cconf;
397 struct echo_object_conf *econf;
400 struct lu_object *below;
401 struct lu_device *under;
404 below = under->ld_ops->ldo_object_alloc(env, obj->lo_header,
408 lu_object_add(obj, below);
411 cconf = lu2cl_conf(conf);
412 econf = cl2echo_conf(cconf);
414 LASSERT(econf->eoc_oinfo);
416 * Transfer the oinfo pointer to eco that it won't be
419 eco->eo_oinfo = *econf->eoc_oinfo;
420 *econf->eoc_oinfo = NULL;
423 atomic_set(&eco->eo_npages, 0);
424 cl_object_page_init(lu2cl(obj), sizeof(struct echo_page));
426 spin_lock(&ec->ec_lock);
427 list_add_tail(&eco->eo_obj_chain, &ec->ec_objects);
428 spin_unlock(&ec->ec_lock);
433 static void echo_object_free(const struct lu_env *env, struct lu_object *obj)
435 struct echo_object *eco = cl2echo_obj(lu2cl(obj));
436 struct echo_client_obd *ec = eco->eo_dev->ed_ec;
438 LASSERT(atomic_read(&eco->eo_npages) == 0);
440 spin_lock(&ec->ec_lock);
441 list_del_init(&eco->eo_obj_chain);
442 spin_unlock(&ec->ec_lock);
445 lu_object_header_fini(obj->lo_header);
447 kfree(eco->eo_oinfo);
448 kmem_cache_free(echo_object_kmem, eco);
451 static int echo_object_print(const struct lu_env *env, void *cookie,
452 lu_printer_t p, const struct lu_object *o)
454 struct echo_object *obj = cl2echo_obj(lu2cl(o));
456 return (*p)(env, cookie, "echoclient-object@%p", obj);
459 static const struct lu_object_operations echo_lu_obj_ops = {
460 .loo_object_init = echo_object_init,
461 .loo_object_delete = NULL,
462 .loo_object_release = NULL,
463 .loo_object_free = echo_object_free,
464 .loo_object_print = echo_object_print,
465 .loo_object_invariant = NULL
468 /** @} echo_lu_ops */
470 /** \defgroup echo_lu_dev_ops lu_device operations
472 * Operations for echo lu device.
476 static struct lu_object *echo_object_alloc(const struct lu_env *env,
477 const struct lu_object_header *hdr,
478 struct lu_device *dev)
480 struct echo_object *eco;
481 struct lu_object *obj = NULL;
483 /* we're the top dev. */
485 eco = kmem_cache_zalloc(echo_object_kmem, GFP_NOFS);
487 struct cl_object_header *hdr = &eco->eo_hdr;
489 obj = &echo_obj2cl(eco)->co_lu;
490 cl_object_header_init(hdr);
491 hdr->coh_page_bufsize = cfs_size_round(sizeof(struct cl_page));
493 lu_object_init(obj, &hdr->coh_lu, dev);
494 lu_object_add_top(&hdr->coh_lu, obj);
496 eco->eo_cl.co_ops = &echo_cl_obj_ops;
497 obj->lo_ops = &echo_lu_obj_ops;
502 static const struct lu_device_operations echo_device_lu_ops = {
503 .ldo_object_alloc = echo_object_alloc,
506 /** @} echo_lu_dev_ops */
508 /** \defgroup echo_init Setup and teardown
510 * Init and fini functions for echo client.
514 static int echo_site_init(const struct lu_env *env, struct echo_device *ed)
516 struct cl_site *site = &ed->ed_site_myself;
519 /* initialize site */
520 rc = cl_site_init(site, &ed->ed_cl);
522 CERROR("Cannot initialize site for echo client(%d)\n", rc);
526 rc = lu_site_init_finish(&site->cs_lu);
532 ed->ed_site = &site->cs_lu;
536 static void echo_site_fini(const struct lu_env *env, struct echo_device *ed)
539 lu_site_fini(ed->ed_site);
544 static void *echo_thread_key_init(const struct lu_context *ctx,
545 struct lu_context_key *key)
547 struct echo_thread_info *info;
549 info = kmem_cache_zalloc(echo_thread_kmem, GFP_NOFS);
551 info = ERR_PTR(-ENOMEM);
555 static void echo_thread_key_fini(const struct lu_context *ctx,
556 struct lu_context_key *key, void *data)
558 struct echo_thread_info *info = data;
560 kmem_cache_free(echo_thread_kmem, info);
563 static struct lu_context_key echo_thread_key = {
564 .lct_tags = LCT_CL_THREAD,
565 .lct_init = echo_thread_key_init,
566 .lct_fini = echo_thread_key_fini,
569 static void *echo_session_key_init(const struct lu_context *ctx,
570 struct lu_context_key *key)
572 struct echo_session_info *session;
574 session = kmem_cache_zalloc(echo_session_kmem, GFP_NOFS);
576 session = ERR_PTR(-ENOMEM);
580 static void echo_session_key_fini(const struct lu_context *ctx,
581 struct lu_context_key *key, void *data)
583 struct echo_session_info *session = data;
585 kmem_cache_free(echo_session_kmem, session);
588 static struct lu_context_key echo_session_key = {
589 .lct_tags = LCT_SESSION,
590 .lct_init = echo_session_key_init,
591 .lct_fini = echo_session_key_fini,
594 LU_TYPE_INIT_FINI(echo, &echo_thread_key, &echo_session_key);
596 static struct lu_device *echo_device_alloc(const struct lu_env *env,
597 struct lu_device_type *t,
598 struct lustre_cfg *cfg)
600 struct lu_device *next;
601 struct echo_device *ed;
602 struct cl_device *cd;
603 struct obd_device *obd = NULL; /* to keep compiler happy */
604 struct obd_device *tgt;
605 const char *tgt_type_name;
608 ed = kzalloc(sizeof(*ed), GFP_NOFS);
615 rc = cl_device_init(cd, t);
619 cd->cd_lu_dev.ld_ops = &echo_device_lu_ops;
621 obd = class_name2obd(lustre_cfg_string(cfg, 0));
625 tgt = class_name2obd(lustre_cfg_string(cfg, 1));
627 CERROR("Can not find tgt device %s\n",
628 lustre_cfg_string(cfg, 1));
630 goto out_device_fini;
633 next = tgt->obd_lu_dev;
634 if (!strcmp(tgt->obd_type->typ_name, LUSTRE_MDT_NAME)) {
635 CERROR("echo MDT client must be run on server\n");
637 goto out_device_fini;
640 rc = echo_site_init(env, ed);
642 goto out_device_fini;
644 rc = echo_client_setup(env, obd, cfg);
648 ed->ed_ec = &obd->u.echo_client;
650 /* if echo client is to be stacked upon ost device, the next is
651 * NULL since ost is not a clio device so far
653 if (next && !lu_device_is_cl(next))
656 tgt_type_name = tgt->obd_type->typ_name;
663 next->ld_site = ed->ed_site;
664 rc = next->ld_type->ldt_ops->ldto_device_init(env, next,
665 next->ld_type->ldt_name,
671 LASSERT(strcmp(tgt_type_name, LUSTRE_OST_NAME) == 0);
675 return &cd->cd_lu_dev;
678 err = echo_client_cleanup(obd);
680 CERROR("Cleanup obd device %s error(%d)\n",
683 echo_site_fini(env, ed);
685 cl_device_fini(&ed->ed_cl);
692 static int echo_device_init(const struct lu_env *env, struct lu_device *d,
693 const char *name, struct lu_device *next)
699 static struct lu_device *echo_device_fini(const struct lu_env *env,
702 struct echo_device *ed = cl2echo_dev(lu2cl_dev(d));
703 struct lu_device *next = ed->ed_next;
706 next = next->ld_type->ldt_ops->ldto_device_fini(env, next);
710 static void echo_lock_release(const struct lu_env *env,
711 struct echo_lock *ecl,
714 struct cl_lock *clk = echo_lock2cl(ecl);
716 cl_lock_release(env, clk);
719 static struct lu_device *echo_device_free(const struct lu_env *env,
722 struct echo_device *ed = cl2echo_dev(lu2cl_dev(d));
723 struct echo_client_obd *ec = ed->ed_ec;
724 struct echo_object *eco;
725 struct lu_device *next = ed->ed_next;
727 CDEBUG(D_INFO, "echo device:%p is going to be freed, next = %p\n",
730 lu_site_purge(env, ed->ed_site, -1);
732 /* check if there are objects still alive.
733 * It shouldn't have any object because lu_site_purge would cleanup
734 * all of cached objects. Anyway, probably the echo device is being
735 * parallelly accessed.
737 spin_lock(&ec->ec_lock);
738 list_for_each_entry(eco, &ec->ec_objects, eo_obj_chain)
740 spin_unlock(&ec->ec_lock);
743 lu_site_purge(env, ed->ed_site, -1);
746 "Waiting for the reference of echo object to be dropped\n");
748 /* Wait for the last reference to be dropped. */
749 spin_lock(&ec->ec_lock);
750 while (!list_empty(&ec->ec_objects)) {
751 spin_unlock(&ec->ec_lock);
752 CERROR("echo_client still has objects at cleanup time, wait for 1 second\n");
753 set_current_state(TASK_UNINTERRUPTIBLE);
754 schedule_timeout(cfs_time_seconds(1));
755 lu_site_purge(env, ed->ed_site, -1);
756 spin_lock(&ec->ec_lock);
758 spin_unlock(&ec->ec_lock);
760 LASSERT(list_empty(&ec->ec_locks));
762 CDEBUG(D_INFO, "No object exists, exiting...\n");
764 echo_client_cleanup(d->ld_obd);
767 next = next->ld_type->ldt_ops->ldto_device_free(env, next);
769 LASSERT(ed->ed_site == d->ld_site);
770 echo_site_fini(env, ed);
771 cl_device_fini(&ed->ed_cl);
774 cl_env_cache_purge(~0);
779 static const struct lu_device_type_operations echo_device_type_ops = {
780 .ldto_init = echo_type_init,
781 .ldto_fini = echo_type_fini,
783 .ldto_start = echo_type_start,
784 .ldto_stop = echo_type_stop,
786 .ldto_device_alloc = echo_device_alloc,
787 .ldto_device_free = echo_device_free,
788 .ldto_device_init = echo_device_init,
789 .ldto_device_fini = echo_device_fini
792 static struct lu_device_type echo_device_type = {
793 .ldt_tags = LU_DEVICE_CL,
794 .ldt_name = LUSTRE_ECHO_CLIENT_NAME,
795 .ldt_ops = &echo_device_type_ops,
796 .ldt_ctx_tags = LCT_CL_THREAD,
801 /** \defgroup echo_exports Exported operations
803 * exporting functions to echo client
808 /* Interfaces to echo client obd device */
809 static struct echo_object *
810 cl_echo_object_find(struct echo_device *d, const struct ost_id *oi)
813 struct echo_thread_info *info;
814 struct echo_object_conf *conf;
815 struct lov_oinfo *oinfo = NULL;
816 struct echo_object *eco;
817 struct cl_object *obj;
822 LASSERTF(ostid_id(oi), DOSTID "\n", POSTID(oi));
823 LASSERTF(ostid_seq(oi) == FID_SEQ_ECHO, DOSTID "\n", POSTID(oi));
825 /* Never return an object if the obd is to be freed. */
826 if (echo_dev2cl(d)->cd_lu_dev.ld_obd->obd_stopping)
827 return ERR_PTR(-ENODEV);
829 env = cl_env_get(&refcheck);
833 info = echo_env_info(env);
834 conf = &info->eti_conf;
836 oinfo = kzalloc(sizeof(*oinfo), GFP_NOFS);
838 eco = ERR_PTR(-ENOMEM);
843 conf->eoc_cl.u.coc_oinfo = oinfo;
847 * If echo_object_init() is successful then ownership of oinfo
848 * is transferred to the object.
850 conf->eoc_oinfo = &oinfo;
852 fid = &info->eti_fid;
853 rc = ostid_to_fid(fid, (struct ost_id *)oi, 0);
859 /* In the function below, .hs_keycmp resolves to
860 * lu_obj_hop_keycmp()
862 /* coverity[overrun-buffer-val] */
863 obj = cl_object_find(env, echo_dev2cl(d), fid, &conf->eoc_cl);
869 eco = cl2echo_obj(obj);
870 if (eco->eo_deleted) {
871 cl_object_put(env, obj);
872 eco = ERR_PTR(-EAGAIN);
877 cl_env_put(env, &refcheck);
881 static int cl_echo_object_put(struct echo_object *eco)
884 struct cl_object *obj = echo_obj2cl(eco);
887 env = cl_env_get(&refcheck);
891 /* an external function to kill an object? */
892 if (eco->eo_deleted) {
893 struct lu_object_header *loh = obj->co_lu.lo_header;
895 LASSERT(&eco->eo_hdr == luh2coh(loh));
896 set_bit(LU_OBJECT_HEARD_BANSHEE, &loh->loh_flags);
899 cl_object_put(env, obj);
900 cl_env_put(env, &refcheck);
904 static int cl_echo_enqueue0(struct lu_env *env, struct echo_object *eco,
905 u64 start, u64 end, int mode,
906 __u64 *cookie, __u32 enqflags)
910 struct cl_object *obj;
911 struct cl_lock_descr *descr;
912 struct echo_thread_info *info;
915 info = echo_env_info(env);
917 lck = &info->eti_lock;
918 obj = echo_obj2cl(eco);
920 memset(lck, 0, sizeof(*lck));
921 descr = &lck->cll_descr;
922 descr->cld_obj = obj;
923 descr->cld_start = cl_index(obj, start);
924 descr->cld_end = cl_index(obj, end);
925 descr->cld_mode = mode == LCK_PW ? CLM_WRITE : CLM_READ;
926 descr->cld_enq_flags = enqflags;
929 rc = cl_lock_request(env, io, lck);
931 struct echo_client_obd *ec = eco->eo_dev->ed_ec;
932 struct echo_lock *el;
934 el = cl2echo_lock(cl_lock_at(lck, &echo_device_type));
935 spin_lock(&ec->ec_lock);
936 if (list_empty(&el->el_chain)) {
937 list_add(&el->el_chain, &ec->ec_locks);
938 el->el_cookie = ++ec->ec_unique;
940 atomic_inc(&el->el_refcount);
941 *cookie = el->el_cookie;
942 spin_unlock(&ec->ec_lock);
947 static int cl_echo_cancel0(struct lu_env *env, struct echo_device *ed,
950 struct echo_client_obd *ec = ed->ed_ec;
951 struct echo_lock *ecl = NULL;
952 struct list_head *el;
953 int found = 0, still_used = 0;
955 spin_lock(&ec->ec_lock);
956 list_for_each(el, &ec->ec_locks) {
957 ecl = list_entry(el, struct echo_lock, el_chain);
958 CDEBUG(D_INFO, "ecl: %p, cookie: %#llx\n", ecl, ecl->el_cookie);
959 found = (ecl->el_cookie == cookie);
961 if (atomic_dec_and_test(&ecl->el_refcount))
962 list_del_init(&ecl->el_chain);
968 spin_unlock(&ec->ec_lock);
973 echo_lock_release(env, ecl, still_used);
977 static void echo_commit_callback(const struct lu_env *env, struct cl_io *io,
978 struct cl_page *page)
980 struct echo_thread_info *info;
981 struct cl_2queue *queue;
983 info = echo_env_info(env);
984 LASSERT(io == &info->eti_io);
986 queue = &info->eti_queue;
987 cl_page_list_add(&queue->c2_qout, page);
990 static int cl_echo_object_brw(struct echo_object *eco, int rw, u64 offset,
991 struct page **pages, int npages, int async)
994 struct echo_thread_info *info;
995 struct cl_object *obj = echo_obj2cl(eco);
996 struct echo_device *ed = eco->eo_dev;
997 struct cl_2queue *queue;
1000 struct lustre_handle lh = { 0 };
1001 size_t page_size = cl_page_size(obj);
1006 LASSERT((offset & ~PAGE_MASK) == 0);
1007 LASSERT(ed->ed_next);
1008 env = cl_env_get(&refcheck);
1010 return PTR_ERR(env);
1012 info = echo_env_info(env);
1014 queue = &info->eti_queue;
1016 cl_2queue_init(queue);
1018 io->ci_ignore_layout = 1;
1019 rc = cl_io_init(env, io, CIT_MISC, obj);
1024 rc = cl_echo_enqueue0(env, eco, offset,
1025 offset + npages * PAGE_SIZE - 1,
1026 rw == READ ? LCK_PR : LCK_PW, &lh.cookie,
1031 for (i = 0; i < npages; i++) {
1033 clp = cl_page_find(env, obj, cl_index(obj, offset),
1034 pages[i], CPT_TRANSIENT);
1039 LASSERT(clp->cp_type == CPT_TRANSIENT);
1041 rc = cl_page_own(env, io, clp);
1043 LASSERT(clp->cp_state == CPS_FREEING);
1044 cl_page_put(env, clp);
1048 * Add a page to the incoming page list of 2-queue.
1050 cl_page_list_add(&queue->c2_qin, clp);
1052 /* drop the reference count for cl_page_find, so that the page
1053 * will be freed in cl_2queue_fini.
1055 cl_page_put(env, clp);
1056 cl_page_clip(env, clp, 0, page_size);
1058 offset += page_size;
1062 enum cl_req_type typ = rw == READ ? CRT_READ : CRT_WRITE;
1064 async = async && (typ == CRT_WRITE);
1066 rc = cl_io_commit_async(env, io, &queue->c2_qin,
1068 echo_commit_callback);
1070 rc = cl_io_submit_sync(env, io, typ, queue, 0);
1071 CDEBUG(D_INFO, "echo_client %s write returns %d\n",
1072 async ? "async" : "sync", rc);
1075 cl_echo_cancel0(env, ed, lh.cookie);
1077 cl_2queue_discard(env, io, queue);
1078 cl_2queue_disown(env, io, queue);
1079 cl_2queue_fini(env, queue);
1080 cl_io_fini(env, io);
1082 cl_env_put(env, &refcheck);
1086 /** @} echo_exports */
1088 static u64 last_object_id;
1090 static int echo_create_object(const struct lu_env *env, struct echo_device *ed,
1093 struct echo_object *eco;
1094 struct echo_client_obd *ec = ed->ed_ec;
1098 if (!(oa->o_valid & OBD_MD_FLID) ||
1099 !(oa->o_valid & OBD_MD_FLGROUP) ||
1100 !fid_seq_is_echo(ostid_seq(&oa->o_oi))) {
1101 CERROR("invalid oid " DOSTID "\n", POSTID(&oa->o_oi));
1105 if (!ostid_id(&oa->o_oi)) {
1106 rc = ostid_set_id(&oa->o_oi, ++last_object_id);
1111 rc = obd_create(env, ec->ec_exp, oa);
1113 CERROR("Cannot create objects: rc = %d\n", rc);
1118 oa->o_valid |= OBD_MD_FLID;
1120 eco = cl_echo_object_find(ed, &oa->o_oi);
1125 cl_echo_object_put(eco);
1127 CDEBUG(D_INFO, "oa oid " DOSTID "\n", POSTID(&oa->o_oi));
1131 obd_destroy(env, ec->ec_exp, oa);
1133 CERROR("create object failed with: rc = %d\n", rc);
1137 static int echo_get_object(struct echo_object **ecop, struct echo_device *ed,
1140 struct echo_object *eco;
1143 if (!(oa->o_valid & OBD_MD_FLID) || !(oa->o_valid & OBD_MD_FLGROUP) ||
1144 !ostid_id(&oa->o_oi)) {
1145 CERROR("invalid oid " DOSTID "\n", POSTID(&oa->o_oi));
1150 eco = cl_echo_object_find(ed, &oa->o_oi);
1158 static void echo_put_object(struct echo_object *eco)
1162 rc = cl_echo_object_put(eco);
1164 CERROR("%s: echo client drop an object failed: rc = %d\n",
1165 eco->eo_dev->ed_ec->ec_exp->exp_obd->obd_name, rc);
1169 echo_client_page_debug_setup(struct page *page, int rw, u64 id,
1170 u64 offset, u64 count)
1177 /* no partial pages on the client */
1178 LASSERT(count == PAGE_SIZE);
1182 for (delta = 0; delta < PAGE_SIZE; delta += OBD_ECHO_BLOCK_SIZE) {
1183 if (rw == OBD_BRW_WRITE) {
1184 stripe_off = offset + delta;
1187 stripe_off = 0xdeadbeef00c0ffeeULL;
1188 stripe_id = 0xdeadbeef00c0ffeeULL;
1190 block_debug_setup(addr + delta, OBD_ECHO_BLOCK_SIZE,
1191 stripe_off, stripe_id);
1197 static int echo_client_page_debug_check(struct page *page, u64 id,
1198 u64 offset, u64 count)
1207 /* no partial pages on the client */
1208 LASSERT(count == PAGE_SIZE);
1212 for (rc = delta = 0; delta < PAGE_SIZE; delta += OBD_ECHO_BLOCK_SIZE) {
1213 stripe_off = offset + delta;
1216 rc2 = block_debug_check("test_brw",
1217 addr + delta, OBD_ECHO_BLOCK_SIZE,
1218 stripe_off, stripe_id);
1220 CERROR("Error in echo object %#llx\n", id);
1229 static int echo_client_kbrw(struct echo_device *ed, int rw, struct obdo *oa,
1230 struct echo_object *eco, u64 offset,
1231 u64 count, int async)
1234 struct brw_page *pga;
1235 struct brw_page *pgp;
1236 struct page **pages;
1244 verify = (ostid_id(&oa->o_oi) != ECHO_PERSISTENT_OBJID &&
1245 (oa->o_valid & OBD_MD_FLFLAGS) != 0 &&
1246 (oa->o_flags & OBD_FL_DEBUG_CHECK) != 0);
1248 gfp_mask = ((ostid_id(&oa->o_oi) & 2) == 0) ? GFP_KERNEL : GFP_HIGHUSER;
1250 LASSERT(rw == OBD_BRW_WRITE || rw == OBD_BRW_READ);
1253 (count & (~PAGE_MASK)) != 0)
1256 /* XXX think again with misaligned I/O */
1257 npages = count >> PAGE_SHIFT;
1259 if (rw == OBD_BRW_WRITE)
1260 brw_flags = OBD_BRW_ASYNC;
1262 pga = kcalloc(npages, sizeof(*pga), GFP_NOFS);
1266 pages = kcalloc(npages, sizeof(*pages), GFP_NOFS);
1272 for (i = 0, pgp = pga, off = offset;
1274 i++, pgp++, off += PAGE_SIZE) {
1275 LASSERT(!pgp->pg); /* for cleanup */
1278 pgp->pg = alloc_page(gfp_mask);
1283 pgp->count = PAGE_SIZE;
1285 pgp->flag = brw_flags;
1288 echo_client_page_debug_setup(pgp->pg, rw,
1289 ostid_id(&oa->o_oi), off,
1293 /* brw mode can only be used at client */
1294 LASSERT(ed->ed_next);
1295 rc = cl_echo_object_brw(eco, rw, offset, pages, npages, async);
1298 if (rc != 0 || rw != OBD_BRW_READ)
1301 for (i = 0, pgp = pga; i < npages; i++, pgp++) {
1308 vrc = echo_client_page_debug_check(pgp->pg,
1309 ostid_id(&oa->o_oi),
1310 pgp->off, pgp->count);
1311 if (vrc != 0 && rc == 0)
1314 __free_page(pgp->pg);
1321 static int echo_client_prep_commit(const struct lu_env *env,
1322 struct obd_export *exp, int rw,
1323 struct obdo *oa, struct echo_object *eco,
1324 u64 offset, u64 count,
1325 u64 batch, int async)
1327 struct obd_ioobj ioo;
1328 struct niobuf_local *lnb;
1329 struct niobuf_remote rnb;
1331 u64 npages, tot_pages;
1332 int i, ret = 0, brw_flags = 0;
1334 if (count <= 0 || (count & (~PAGE_MASK)) != 0)
1337 npages = batch >> PAGE_SHIFT;
1338 tot_pages = count >> PAGE_SHIFT;
1340 lnb = kcalloc(npages, sizeof(struct niobuf_local), GFP_NOFS);
1346 if (rw == OBD_BRW_WRITE && async)
1347 brw_flags |= OBD_BRW_ASYNC;
1349 obdo_to_ioobj(oa, &ioo);
1353 for (; tot_pages > 0; tot_pages -= npages) {
1356 if (tot_pages < npages)
1359 rnb.rnb_offset = off;
1360 rnb.rnb_len = npages * PAGE_SIZE;
1361 rnb.rnb_flags = brw_flags;
1363 off += npages * PAGE_SIZE;
1366 ret = obd_preprw(env, rw, exp, oa, 1, &ioo, &rnb, &lpages, lnb);
1370 for (i = 0; i < lpages; i++) {
1371 struct page *page = lnb[i].lnb_page;
1373 /* read past eof? */
1374 if (!page && lnb[i].lnb_rc == 0)
1378 lnb[i].lnb_flags |= OBD_BRW_ASYNC;
1380 if (ostid_id(&oa->o_oi) == ECHO_PERSISTENT_OBJID ||
1381 (oa->o_valid & OBD_MD_FLFLAGS) == 0 ||
1382 (oa->o_flags & OBD_FL_DEBUG_CHECK) == 0)
1385 if (rw == OBD_BRW_WRITE)
1386 echo_client_page_debug_setup(page, rw,
1387 ostid_id(&oa->o_oi),
1388 lnb[i].lnb_file_offset,
1391 echo_client_page_debug_check(page,
1392 ostid_id(&oa->o_oi),
1393 lnb[i].lnb_file_offset,
1397 ret = obd_commitrw(env, rw, exp, oa, 1, &ioo, &rnb, npages, lnb,
1402 /* Reuse env context. */
1403 lu_context_exit((struct lu_context *)&env->le_ctx);
1404 lu_context_enter((struct lu_context *)&env->le_ctx);
1412 static int echo_client_brw_ioctl(const struct lu_env *env, int rw,
1413 struct obd_export *exp,
1414 struct obd_ioctl_data *data)
1416 struct obd_device *obd = class_exp2obd(exp);
1417 struct echo_device *ed = obd2echo_dev(obd);
1418 struct echo_client_obd *ec = ed->ed_ec;
1419 struct obdo *oa = &data->ioc_obdo1;
1420 struct echo_object *eco;
1425 LASSERT(oa->o_valid & OBD_MD_FLGROUP);
1427 rc = echo_get_object(&eco, ed, oa);
1431 oa->o_valid &= ~OBD_MD_FLHANDLE;
1433 /* OFD/obdfilter works only via prep/commit */
1434 test_mode = (long)data->ioc_pbuf1;
1438 if (!ed->ed_next && test_mode != 3) {
1440 data->ioc_plen1 = data->ioc_count;
1443 /* Truncate batch size to maximum */
1444 if (data->ioc_plen1 > PTLRPC_MAX_BRW_SIZE)
1445 data->ioc_plen1 = PTLRPC_MAX_BRW_SIZE;
1447 switch (test_mode) {
1451 rc = echo_client_kbrw(ed, rw, oa, eco, data->ioc_offset,
1452 data->ioc_count, async);
1455 rc = echo_client_prep_commit(env, ec->ec_exp, rw, oa, eco,
1456 data->ioc_offset, data->ioc_count,
1457 data->ioc_plen1, async);
1462 echo_put_object(eco);
1467 echo_client_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
1468 void *karg, void __user *uarg)
1470 struct obd_device *obd = exp->exp_obd;
1471 struct echo_device *ed = obd2echo_dev(obd);
1472 struct echo_client_obd *ec = ed->ed_ec;
1473 struct echo_object *eco;
1474 struct obd_ioctl_data *data = karg;
1478 int rw = OBD_BRW_READ;
1481 oa = &data->ioc_obdo1;
1482 if (!(oa->o_valid & OBD_MD_FLGROUP)) {
1483 oa->o_valid |= OBD_MD_FLGROUP;
1484 ostid_set_seq_echo(&oa->o_oi);
1487 /* This FID is unpacked just for validation at this point */
1488 rc = ostid_to_fid(&fid, &oa->o_oi, 0);
1492 env = kzalloc(sizeof(*env), GFP_NOFS);
1496 rc = lu_env_init(env, LCT_DT_THREAD);
1503 case OBD_IOC_CREATE: /* may create echo object */
1504 if (!capable(CFS_CAP_SYS_ADMIN)) {
1509 rc = echo_create_object(env, ed, oa);
1512 case OBD_IOC_DESTROY:
1513 if (!capable(CFS_CAP_SYS_ADMIN)) {
1518 rc = echo_get_object(&eco, ed, oa);
1520 rc = obd_destroy(env, ec->ec_exp, oa);
1522 eco->eo_deleted = 1;
1523 echo_put_object(eco);
1527 case OBD_IOC_GETATTR:
1528 rc = echo_get_object(&eco, ed, oa);
1530 rc = obd_getattr(env, ec->ec_exp, oa);
1531 echo_put_object(eco);
1535 case OBD_IOC_SETATTR:
1536 if (!capable(CFS_CAP_SYS_ADMIN)) {
1541 rc = echo_get_object(&eco, ed, oa);
1543 rc = obd_setattr(env, ec->ec_exp, oa);
1544 echo_put_object(eco);
1548 case OBD_IOC_BRW_WRITE:
1549 if (!capable(CFS_CAP_SYS_ADMIN)) {
1556 case OBD_IOC_BRW_READ:
1557 rc = echo_client_brw_ioctl(env, rw, exp, data);
1561 CERROR("echo_ioctl(): unrecognised ioctl %#x\n", cmd);
1573 static int echo_client_setup(const struct lu_env *env,
1574 struct obd_device *obddev, struct lustre_cfg *lcfg)
1576 struct echo_client_obd *ec = &obddev->u.echo_client;
1577 struct obd_device *tgt;
1578 struct obd_uuid echo_uuid = { "ECHO_UUID" };
1579 struct obd_connect_data *ocd = NULL;
1582 if (lcfg->lcfg_bufcount < 2 || LUSTRE_CFG_BUFLEN(lcfg, 1) < 1) {
1583 CERROR("requires a TARGET OBD name\n");
1587 tgt = class_name2obd(lustre_cfg_string(lcfg, 1));
1588 if (!tgt || !tgt->obd_attached || !tgt->obd_set_up) {
1589 CERROR("device not attached or not set up (%s)\n",
1590 lustre_cfg_string(lcfg, 1));
1594 spin_lock_init(&ec->ec_lock);
1595 INIT_LIST_HEAD(&ec->ec_objects);
1596 INIT_LIST_HEAD(&ec->ec_locks);
1599 ocd = kzalloc(sizeof(*ocd), GFP_NOFS);
1603 ocd->ocd_connect_flags = OBD_CONNECT_VERSION | OBD_CONNECT_REQPORTAL |
1604 OBD_CONNECT_BRW_SIZE |
1605 OBD_CONNECT_GRANT | OBD_CONNECT_FULL20 |
1606 OBD_CONNECT_64BITHASH | OBD_CONNECT_LVB_TYPE |
1608 ocd->ocd_brw_size = DT_MAX_BRW_SIZE;
1609 ocd->ocd_version = LUSTRE_VERSION_CODE;
1610 ocd->ocd_group = FID_SEQ_ECHO;
1612 rc = obd_connect(env, &ec->ec_exp, tgt, &echo_uuid, ocd, NULL);
1617 CERROR("fail to connect to device %s\n",
1618 lustre_cfg_string(lcfg, 1));
1625 static int echo_client_cleanup(struct obd_device *obddev)
1627 struct echo_client_obd *ec = &obddev->u.echo_client;
1630 if (!list_empty(&obddev->obd_exports)) {
1631 CERROR("still has clients!\n");
1635 LASSERT(atomic_read(&ec->ec_exp->exp_refcount) > 0);
1636 rc = obd_disconnect(ec->ec_exp);
1638 CERROR("fail to disconnect device: %d\n", rc);
1643 static int echo_client_connect(const struct lu_env *env,
1644 struct obd_export **exp,
1645 struct obd_device *src, struct obd_uuid *cluuid,
1646 struct obd_connect_data *data, void *localdata)
1649 struct lustre_handle conn = { 0 };
1651 rc = class_connect(&conn, src, cluuid);
1653 *exp = class_conn2export(&conn);
1658 static int echo_client_disconnect(struct obd_export *exp)
1667 rc = class_disconnect(exp);
1673 static struct obd_ops echo_client_obd_ops = {
1674 .owner = THIS_MODULE,
1675 .iocontrol = echo_client_iocontrol,
1676 .connect = echo_client_connect,
1677 .disconnect = echo_client_disconnect
1680 static int echo_client_init(void)
1684 rc = lu_kmem_init(echo_caches);
1686 rc = class_register_type(&echo_client_obd_ops, NULL,
1687 LUSTRE_ECHO_CLIENT_NAME,
1690 lu_kmem_fini(echo_caches);
1695 static void echo_client_exit(void)
1697 class_unregister_type(LUSTRE_ECHO_CLIENT_NAME);
1698 lu_kmem_fini(echo_caches);
1701 static int __init obdecho_init(void)
1703 LCONSOLE_INFO("Echo OBD driver; http://www.lustre.org/\n");
1705 LASSERT(PAGE_SIZE % OBD_ECHO_BLOCK_SIZE == 0);
1707 return echo_client_init();
1710 static void /*__exit*/ obdecho_exit(void)
1715 MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
1716 MODULE_DESCRIPTION("Lustre Echo Client test driver");
1717 MODULE_VERSION(LUSTRE_VERSION_STRING);
1718 MODULE_LICENSE("GPL");
1720 module_init(obdecho_init);
1721 module_exit(obdecho_exit);
1723 /** @} echo_client */