1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
5 * Lustre Metadata Target (mdt) open/close file handling
7 * Copyright (C) 2002-2006 Cluster File Systems, Inc.
8 * Author: Huang Hua <huanghua@clusterfs.com>
10 * This file is part of the Lustre file system, http://www.lustre.org
11 * Lustre is a trademark of Cluster File Systems, Inc.
13 * You may have signed or agreed to another license before downloading
14 * this software. If so, you are bound by the terms and conditions
15 * of that agreement, and the following does not apply to you. See the
16 * LICENSE file included with this distribution for more information.
18 * If you did not agree to a different license, then this copy of Lustre
19 * is open source software; you can redistribute it and/or modify it
20 * under the terms of version 2 of the GNU General Public License as
21 * published by the Free Software Foundation.
23 * In either case, Lustre is distributed in the hope that it will be
24 * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
25 * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
26 * license text for more details.
30 # define EXPORT_SYMTAB
32 #define DEBUG_SUBSYSTEM S_MDS
34 #include <lustre_mds.h>
35 #include "mdt_internal.h"
37 /* we do nothing because we do not have refcount now */
38 static void mdt_mfd_get(void *mfdp)
42 /* Create a new mdt_file_data struct, initialize it,
43 * and insert it to global hash table */
44 static struct mdt_file_data *mdt_mfd_new(void)
46 struct mdt_file_data *mfd;
51 INIT_LIST_HEAD(&mfd->mfd_handle.h_link);
52 INIT_LIST_HEAD(&mfd->mfd_list);
53 class_handle_hash(&mfd->mfd_handle, mdt_mfd_get);
58 /* Find the mfd pointed to by handle in global hash table. */
59 static struct mdt_file_data *mdt_handle2mfd(const struct lustre_handle *handle)
62 LASSERT(handle != NULL);
63 RETURN(class_handle2object(handle->cookie));
67 static void mdt_mfd_free(struct mdt_file_data *mfd)
69 LASSERT(list_empty(&mfd->mfd_handle.h_link));
70 LASSERT(list_empty(&mfd->mfd_list));
74 static int mdt_create_data_obj(struct mdt_thread_info *info,
75 struct mdt_object *p, struct mdt_object *o)
77 struct md_attr *ma = &info->mti_attr;
78 struct mdt_reint_record *mrr = &info->mti_rr;
80 return mdo_create_data_object(info->mti_ctxt, mdt_object_child(p),
81 mdt_object_child(o), mrr->rr_eadata,
82 mrr->rr_eadatalen, ma);
85 static int mdt_mfd_open(struct mdt_thread_info *info,
88 int flags, int created)
90 struct mdt_export_data *med;
91 struct mdt_file_data *mfd;
92 struct mdt_body *repbody;
93 struct md_attr *ma = &info->mti_attr;
94 struct lu_attr *la = &ma->ma_attr;
95 struct ptlrpc_request *req = mdt_info_req(info);
99 med = &req->rq_export->exp_mdt_data;
100 repbody = req_capsule_server_get(&info->mti_pill, &RMF_MDT_BODY);
103 /* we have to get attr & lov ea for this object*/
104 rc = mo_attr_get(info->mti_ctxt, mdt_object_child(o), ma);
107 if (ma->ma_valid & MA_INODE)
108 mdt_pack_attr2body(repbody, la, mdt_object_fid(o));
110 if (!S_ISREG(la->la_mode) &&
111 !S_ISDIR(la->la_mode) &&
112 (req->rq_export->exp_connect_flags & OBD_CONNECT_NODEVOH))
113 /* If client supports this, do not return open handle
114 * for special nodes */
117 if ((S_ISREG(la->la_mode) || S_ISDIR(la->la_mode))
118 && !created && !(ma->ma_valid & MA_LOV)) {
119 /*No EA, check whether it is will set regEA and dirEA
120 *since in above attr get, these size might be zero,
121 *so reset it, to retrieve the MD after create obj*/
124 rc = mdt_create_data_obj(info, p, o);
128 /* FIXME:maybe this can be done earlier? */
129 if (S_ISDIR(la->la_mode)) {
130 if (flags & (MDS_OPEN_CREAT | FMODE_WRITE)) {
131 /* we are trying to create or
132 * write an existing dir. */
135 } else if (flags & MDS_OPEN_DIRECTORY)
141 CDEBUG(D_INODE, "after open, ma_valid bit = "LPX64"\n", ma->ma_valid);
142 CDEBUG(D_INODE, "after open, lmm_size = %d\n", ma->ma_lmm_size);
143 repbody->eadatasize = 0;
144 repbody->aclsize = 0;
146 if (ma->ma_lmm_size && ma->ma_valid & MA_LOV) {
147 repbody->eadatasize = ma->ma_lmm_size;
148 if (S_ISDIR(la->la_mode))
149 repbody->valid |= OBD_MD_FLDIREA;
151 repbody->valid |= OBD_MD_FLEASIZE;
154 /*FIXME: should determine the offset dynamicly */
155 lustre_shrink_reply(req, 2, repbody->eadatasize, 1);
156 lustre_shrink_reply(req, repbody->eadatasize ? 3 : 2, repbody->aclsize, 0);
158 if (flags & FMODE_WRITE) {
159 /*mds_get_write_access*/
160 } else if (flags & MDS_FMODE_EXEC) {
161 /*mds_deny_write_access*/
164 /* (1) client wants transno when open to keep a ref count for replay;
165 * see after_reply() and mdc_close_commit();
166 * (2) we need to record the transaction related stuff onto disk;
167 * But, question is: when do a rean only open, do we still need transno?
170 struct txn_param txn;
172 struct dt_device *dt = info->mti_mdt->mdt_bottom;
176 th = dt->dd_ops->dt_trans_start(info->mti_ctxt, dt, &txn);
178 dt->dd_ops->dt_trans_stop(info->mti_ctxt, th);
185 /* keep a reference on this object for this open,
186 * and is released by mdt_mfd_close() */
187 mdt_object_get(info->mti_ctxt, o);
189 mfd->mfd_mode = flags;
191 mfd->mfd_xid = mdt_info_req(info)->rq_xid;
193 spin_lock(&med->med_open_lock);
194 list_add(&mfd->mfd_list, &med->med_open_head);
195 spin_unlock(&med->med_open_lock);
197 repbody->handle.cookie = mfd->mfd_handle.h_cookie;
204 int mdt_open_by_fid(struct mdt_thread_info* info, const struct lu_fid *fid,
207 struct mdt_object *o;
208 struct lu_attr *la = &info->mti_attr.ma_attr;
212 o = mdt_object_find(info->mti_ctxt, info->mti_mdt, fid);
214 if (mdt_object_exists(info->mti_ctxt, &o->mot_obj.mo_lu) > 0) {
215 if (la->la_flags & MDS_OPEN_EXCL &&
216 la->la_flags & MDS_OPEN_CREAT)
219 rc = mdt_mfd_open(info, NULL, o, flags, 0);
222 if (la->la_flags & MDS_OPEN_CREAT) {
223 rc = mo_object_create(info->mti_ctxt,
227 rc = mdt_mfd_open(info, NULL, o, flags, 1);
230 mdt_object_put(info->mti_ctxt, o);
237 int mdt_pin(struct mdt_thread_info* info)
239 struct mdt_body *body;
243 rc = req_capsule_pack(&info->mti_pill);
245 body = req_capsule_client_get(&info->mti_pill, &RMF_MDT_BODY);
246 rc = mdt_open_by_fid(info, &body->fid1, body->flags);
251 int mdt_reint_open(struct mdt_thread_info *info)
253 struct mdt_device *mdt = info->mti_mdt;
254 struct mdt_object *parent;
255 struct mdt_object *child;
256 struct mdt_lock_handle *lh;
257 struct ldlm_reply *ldlm_rep;
258 struct lu_fid *child_fid = &info->mti_tmp_fid1;
259 struct md_attr *ma = &info->mti_attr;
260 struct lu_attr *la = &ma->ma_attr;
263 struct mdt_reint_record *rr = &info->mti_rr;
266 req_capsule_set_size(&info->mti_pill, &RMF_MDT_MD, RCL_SERVER,
267 mdt->mdt_max_mdsize);
269 req_capsule_set_size(&info->mti_pill, &RMF_EADATA, RCL_SERVER,
270 LUSTRE_POSIX_ACL_MAX_SIZE);
272 result = req_capsule_pack(&info->mti_pill);
276 ma->ma_lmm = req_capsule_server_get(&info->mti_pill,
278 ma->ma_lmm_size = req_capsule_get_size(&info->mti_pill,
282 if (rr->rr_name[0] == 0) {
283 /* reint partial remote open */
284 RETURN(mdt_open_by_fid(info, rr->rr_fid1, la->la_flags));
287 /* we now have no resent message, so it must be an intent */
288 /*TODO: remove this and add MDS_CHECK_RESENT if resent enabled*/
289 LASSERT(info->mti_pill.rc_fmt == &RQF_LDLM_INTENT_OPEN);
291 ldlm_rep = req_capsule_server_get(&info->mti_pill, &RMF_DLM_REP);
293 intent_set_disposition(ldlm_rep, DISP_LOOKUP_EXECD);
294 lh = &info->mti_lh[MDT_LH_PARENT];
295 if (!(la->la_flags & MDS_OPEN_CREAT))
296 lh->mlh_mode = LCK_CR;
298 lh->mlh_mode = LCK_EX;
299 parent = mdt_object_find_lock(info, rr->rr_fid1, lh,
300 MDS_INODELOCK_UPDATE);
302 GOTO(out, result = PTR_ERR(parent));
304 result = mdo_lookup(info->mti_ctxt, mdt_object_child(parent),
305 rr->rr_name, child_fid);
306 if (result != 0 && result != -ENOENT)
307 GOTO(out_parent, result);
309 if (result == -ENOENT) {
310 intent_set_disposition(ldlm_rep, DISP_LOOKUP_NEG);
311 if (!(la->la_flags & MDS_OPEN_CREAT))
312 GOTO(out_parent, result);
313 *child_fid = *info->mti_rr.rr_fid2;
314 /* new object will be created. see the following */
316 intent_set_disposition(ldlm_rep, DISP_LOOKUP_POS);
317 if ((la->la_flags & MDS_OPEN_EXCL &&
318 la->la_flags & MDS_OPEN_CREAT))
319 GOTO(out_parent, result = -EEXIST);
322 child = mdt_object_find(info->mti_ctxt, mdt, child_fid);
324 GOTO(out_parent, result = PTR_ERR(child));
326 if (result == -ENOENT) {
327 /* not found and with MDS_OPEN_CREAT: let's create it */
328 result = mdo_create(info->mti_ctxt,
329 mdt_object_child(parent),
331 mdt_object_child(child),
332 rr->rr_tgt, rr->rr_eadata,
333 rr->rr_eadatalen, &info->mti_attr);
334 intent_set_disposition(ldlm_rep, DISP_OPEN_CREATE);
336 GOTO(out_child, result);
341 result = mdt_mfd_open(info, parent, child, la->la_flags, created);
342 intent_set_disposition(ldlm_rep, DISP_OPEN_OPEN);
343 GOTO(finish_open, result);
346 if (result != 0 && created) {
347 int rc2 = mdo_unlink(info->mti_ctxt, mdt_object_child(parent),
348 mdt_object_child(child), rr->rr_name,
351 CERROR("error in cleanup of open");
354 mdt_object_put(info->mti_ctxt, child);
356 mdt_object_unlock_put(info, parent, lh, result);
361 void mdt_mfd_close(const struct lu_context *ctxt,
362 struct mdt_file_data *mfd)
366 if (mfd->mfd_mode & FMODE_WRITE) {
367 /*mdt_put_write_access*/
368 } else if (mfd->mfd_mode & MDS_FMODE_EXEC) {
369 /*mdt_allow_write_access*/
372 /* release reference on this object.
373 * it will be destroyed by lower layer if necessary.
375 mdt_object_put(ctxt, mfd->mfd_object);
381 int mdt_close(struct mdt_thread_info *info)
383 struct md_attr *ma = &info->mti_attr;
384 struct mdt_export_data *med;
385 struct mdt_file_data *mfd;
386 struct mdt_object *o;
390 med = &mdt_info_req(info)->rq_export->exp_mdt_data;
392 spin_lock(&med->med_open_lock);
393 mfd = mdt_handle2mfd(&(info->mti_body->handle));
395 spin_unlock(&med->med_open_lock);
396 CDEBUG(D_INODE, "no handle for file close: fid = "DFID3
397 ": cookie = "LPX64"\n", PFID3(&info->mti_body->fid1),
398 info->mti_body->handle.cookie);
401 class_handle_unhash(&mfd->mfd_handle);
402 list_del_init(&mfd->mfd_list);
403 spin_unlock(&med->med_open_lock);
406 ma->ma_lmm = req_capsule_server_get(&info->mti_pill,
408 ma->ma_lmm_size = req_capsule_get_size(&info->mti_pill,
409 &RMF_MDT_MD, RCL_SERVER);
410 rc = mo_attr_get(info->mti_ctxt, mdt_object_child(o), ma);
412 rc = mdt_handle_last_unlink(info, o);
414 mdt_mfd_close(info->mti_ctxt, mfd);
416 mdt_shrink_reply(info);
420 int mdt_done_writing(struct mdt_thread_info *info)
425 req_capsule_set(&info->mti_pill, &RQF_MDS_DONE_WRITING);
426 rc = req_capsule_pack(&info->mti_pill);