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) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #define DEBUG_SUBSYSTEM S_MDC
38 #include "../include/lustre_net.h"
39 #include "../include/lustre/lustre_idl.h"
40 #include "mdc_internal.h"
42 static void __mdc_pack_body(struct mdt_body *b, __u32 suppgid)
47 b->uid = from_kuid(&init_user_ns, current_uid());
48 b->gid = from_kgid(&init_user_ns, current_gid());
49 b->fsuid = from_kuid(&init_user_ns, current_fsuid());
50 b->fsgid = from_kgid(&init_user_ns, current_fsgid());
51 b->capability = cfs_curproc_cap_pack();
54 void mdc_is_subdir_pack(struct ptlrpc_request *req, const struct lu_fid *pfid,
55 const struct lu_fid *cfid, int flags)
57 struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
62 b->valid = OBD_MD_FLID;
69 void mdc_swap_layouts_pack(struct ptlrpc_request *req,
70 struct md_op_data *op_data)
72 struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
75 __mdc_pack_body(b, op_data->op_suppgids[0]);
76 b->fid1 = op_data->op_fid1;
77 b->fid2 = op_data->op_fid2;
78 b->valid |= OBD_MD_FLID;
81 void mdc_pack_body(struct ptlrpc_request *req, const struct lu_fid *fid,
82 __u64 valid, int ea_size, __u32 suppgid, int flags)
84 struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
88 b->eadatasize = ea_size;
90 __mdc_pack_body(b, suppgid);
93 b->valid |= OBD_MD_FLID;
97 void mdc_readdir_pack(struct ptlrpc_request *req, __u64 pgoff,
98 __u32 size, const struct lu_fid *fid)
100 struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
103 b->valid |= OBD_MD_FLID;
104 b->size = pgoff; /* !! */
105 b->nlink = size; /* !! */
106 __mdc_pack_body(b, -1);
107 b->mode = LUDA_FID | LUDA_TYPE;
110 /* packing of MDS records */
111 void mdc_create_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
112 const void *data, int datalen, __u32 mode,
113 __u32 uid, __u32 gid, cfs_cap_t cap_effective, __u64 rdev)
115 struct mdt_rec_create *rec;
119 CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create));
120 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
122 rec->cr_opcode = REINT_CREATE;
125 rec->cr_cap = cap_effective;
126 rec->cr_fid1 = op_data->op_fid1;
127 rec->cr_fid2 = op_data->op_fid2;
130 rec->cr_time = op_data->op_mod_time;
131 rec->cr_suppgid1 = op_data->op_suppgids[0];
132 rec->cr_suppgid2 = op_data->op_suppgids[1];
133 flags = op_data->op_flags & MF_SOM_LOCAL_FLAGS;
134 if (op_data->op_bias & MDS_CREATE_VOLATILE)
135 flags |= MDS_OPEN_VOLATILE;
136 set_mrc_cr_flags(rec, flags);
137 rec->cr_bias = op_data->op_bias;
138 rec->cr_umask = current_umask();
140 tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
141 LOGL0(op_data->op_name, op_data->op_namelen, tmp);
144 tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
145 memcpy(tmp, data, datalen);
149 static __u64 mds_pack_open_flags(__u64 flags, __u32 mode)
151 __u64 cr_flags = (flags & (FMODE_READ | FMODE_WRITE |
152 MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS |
153 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK |
154 MDS_OPEN_BY_FID | MDS_OPEN_LEASE |
157 cr_flags |= MDS_OPEN_CREAT;
159 cr_flags |= MDS_OPEN_EXCL;
161 cr_flags |= MDS_OPEN_TRUNC;
162 if (flags & O_APPEND)
163 cr_flags |= MDS_OPEN_APPEND;
165 cr_flags |= MDS_OPEN_SYNC;
166 if (flags & O_DIRECTORY)
167 cr_flags |= MDS_OPEN_DIRECTORY;
168 if (flags & __FMODE_EXEC)
169 cr_flags |= MDS_FMODE_EXEC;
170 if (cl_is_lov_delay_create(flags))
171 cr_flags |= MDS_OPEN_DELAY_CREATE;
173 if (flags & O_NONBLOCK)
174 cr_flags |= MDS_OPEN_NORESTORE;
179 /* packing of MDS records */
180 void mdc_open_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
181 __u32 mode, __u64 rdev, __u64 flags, const void *lmm,
184 struct mdt_rec_create *rec;
188 CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_create));
189 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
191 /* XXX do something about time, uid, gid */
192 rec->cr_opcode = REINT_OPEN;
193 rec->cr_fsuid = from_kuid(&init_user_ns, current_fsuid());
194 rec->cr_fsgid = from_kgid(&init_user_ns, current_fsgid());
195 rec->cr_cap = cfs_curproc_cap_pack();
196 rec->cr_fid1 = op_data->op_fid1;
197 rec->cr_fid2 = op_data->op_fid2;
200 cr_flags = mds_pack_open_flags(flags, mode);
202 rec->cr_time = op_data->op_mod_time;
203 rec->cr_suppgid1 = op_data->op_suppgids[0];
204 rec->cr_suppgid2 = op_data->op_suppgids[1];
205 rec->cr_bias = op_data->op_bias;
206 rec->cr_umask = current_umask();
207 rec->cr_old_handle = op_data->op_handle;
209 if (op_data->op_name) {
210 tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
211 LOGL0(op_data->op_name, op_data->op_namelen, tmp);
212 if (op_data->op_bias & MDS_CREATE_VOLATILE)
213 cr_flags |= MDS_OPEN_VOLATILE;
217 cr_flags |= MDS_OPEN_HAS_EA;
218 tmp = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
219 memcpy(tmp, lmm, lmmlen);
221 set_mrc_cr_flags(rec, cr_flags);
224 static inline __u64 attr_pack(unsigned int ia_valid)
228 if (ia_valid & ATTR_MODE)
229 sa_valid |= MDS_ATTR_MODE;
230 if (ia_valid & ATTR_UID)
231 sa_valid |= MDS_ATTR_UID;
232 if (ia_valid & ATTR_GID)
233 sa_valid |= MDS_ATTR_GID;
234 if (ia_valid & ATTR_SIZE)
235 sa_valid |= MDS_ATTR_SIZE;
236 if (ia_valid & ATTR_ATIME)
237 sa_valid |= MDS_ATTR_ATIME;
238 if (ia_valid & ATTR_MTIME)
239 sa_valid |= MDS_ATTR_MTIME;
240 if (ia_valid & ATTR_CTIME)
241 sa_valid |= MDS_ATTR_CTIME;
242 if (ia_valid & ATTR_ATIME_SET)
243 sa_valid |= MDS_ATTR_ATIME_SET;
244 if (ia_valid & ATTR_MTIME_SET)
245 sa_valid |= MDS_ATTR_MTIME_SET;
246 if (ia_valid & ATTR_FORCE)
247 sa_valid |= MDS_ATTR_FORCE;
248 if (ia_valid & ATTR_ATTR_FLAG)
249 sa_valid |= MDS_ATTR_ATTR_FLAG;
250 if (ia_valid & ATTR_KILL_SUID)
251 sa_valid |= MDS_ATTR_KILL_SUID;
252 if (ia_valid & ATTR_KILL_SGID)
253 sa_valid |= MDS_ATTR_KILL_SGID;
254 if (ia_valid & ATTR_CTIME_SET)
255 sa_valid |= MDS_ATTR_CTIME_SET;
256 if (ia_valid & ATTR_OPEN)
257 sa_valid |= MDS_ATTR_FROM_OPEN;
258 if (ia_valid & ATTR_BLOCKS)
259 sa_valid |= MDS_ATTR_BLOCKS;
260 if (ia_valid & MDS_OPEN_OWNEROVERRIDE)
261 /* NFSD hack (see bug 5781) */
262 sa_valid |= MDS_OPEN_OWNEROVERRIDE;
266 static void mdc_setattr_pack_rec(struct mdt_rec_setattr *rec,
267 struct md_op_data *op_data)
269 rec->sa_opcode = REINT_SETATTR;
270 rec->sa_fsuid = from_kuid(&init_user_ns, current_fsuid());
271 rec->sa_fsgid = from_kgid(&init_user_ns, current_fsgid());
272 rec->sa_cap = cfs_curproc_cap_pack();
273 rec->sa_suppgid = -1;
275 rec->sa_fid = op_data->op_fid1;
276 rec->sa_valid = attr_pack(op_data->op_attr.ia_valid);
277 rec->sa_mode = op_data->op_attr.ia_mode;
278 rec->sa_uid = from_kuid(&init_user_ns, op_data->op_attr.ia_uid);
279 rec->sa_gid = from_kgid(&init_user_ns, op_data->op_attr.ia_gid);
280 rec->sa_size = op_data->op_attr.ia_size;
281 rec->sa_blocks = op_data->op_attr_blocks;
282 rec->sa_atime = LTIME_S(op_data->op_attr.ia_atime);
283 rec->sa_mtime = LTIME_S(op_data->op_attr.ia_mtime);
284 rec->sa_ctime = LTIME_S(op_data->op_attr.ia_ctime);
286 ((struct ll_iattr *)&op_data->op_attr)->ia_attr_flags;
287 if ((op_data->op_attr.ia_valid & ATTR_GID) &&
288 in_group_p(op_data->op_attr.ia_gid))
290 from_kgid(&init_user_ns, op_data->op_attr.ia_gid);
292 rec->sa_suppgid = op_data->op_suppgids[0];
294 rec->sa_bias = op_data->op_bias;
297 static void mdc_ioepoch_pack(struct mdt_ioepoch *epoch,
298 struct md_op_data *op_data)
300 memcpy(&epoch->handle, &op_data->op_handle, sizeof(epoch->handle));
301 epoch->ioepoch = op_data->op_ioepoch;
302 epoch->flags = op_data->op_flags & MF_SOM_LOCAL_FLAGS;
305 void mdc_setattr_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
306 void *ea, int ealen, void *ea2, int ea2len)
308 struct mdt_rec_setattr *rec;
309 struct mdt_ioepoch *epoch;
310 struct lov_user_md *lum = NULL;
312 CLASSERT(sizeof(struct mdt_rec_reint) ==
313 sizeof(struct mdt_rec_setattr));
314 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
315 mdc_setattr_pack_rec(rec, op_data);
317 if (op_data->op_flags & (MF_SOM_CHANGE | MF_EPOCH_OPEN)) {
318 epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH);
319 mdc_ioepoch_pack(epoch, op_data);
325 lum = req_capsule_client_get(&req->rq_pill, &RMF_EADATA);
326 if (ea == NULL) { /* Remove LOV EA */
327 lum->lmm_magic = LOV_USER_MAGIC_V1;
328 lum->lmm_stripe_size = 0;
329 lum->lmm_stripe_count = 0;
330 lum->lmm_stripe_offset = (typeof(lum->lmm_stripe_offset))(-1);
332 memcpy(lum, ea, ealen);
338 memcpy(req_capsule_client_get(&req->rq_pill, &RMF_LOGCOOKIES), ea2,
342 void mdc_unlink_pack(struct ptlrpc_request *req, struct md_op_data *op_data)
344 struct mdt_rec_unlink *rec;
347 CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_unlink));
348 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
349 LASSERT(rec != NULL);
351 rec->ul_opcode = op_data->op_cli_flags & CLI_RM_ENTRY ?
352 REINT_RMENTRY : REINT_UNLINK;
353 rec->ul_fsuid = op_data->op_fsuid;
354 rec->ul_fsgid = op_data->op_fsgid;
355 rec->ul_cap = op_data->op_cap;
356 rec->ul_mode = op_data->op_mode;
357 rec->ul_suppgid1 = op_data->op_suppgids[0];
358 rec->ul_suppgid2 = -1;
359 rec->ul_fid1 = op_data->op_fid1;
360 rec->ul_fid2 = op_data->op_fid2;
361 rec->ul_time = op_data->op_mod_time;
362 rec->ul_bias = op_data->op_bias;
364 tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
365 LASSERT(tmp != NULL);
366 LOGL0(op_data->op_name, op_data->op_namelen, tmp);
369 void mdc_link_pack(struct ptlrpc_request *req, struct md_op_data *op_data)
371 struct mdt_rec_link *rec;
374 CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_link));
375 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
376 LASSERT(rec != NULL);
378 rec->lk_opcode = REINT_LINK;
379 rec->lk_fsuid = op_data->op_fsuid; /* current->fsuid; */
380 rec->lk_fsgid = op_data->op_fsgid; /* current->fsgid; */
381 rec->lk_cap = op_data->op_cap; /* current->cap_effective; */
382 rec->lk_suppgid1 = op_data->op_suppgids[0];
383 rec->lk_suppgid2 = op_data->op_suppgids[1];
384 rec->lk_fid1 = op_data->op_fid1;
385 rec->lk_fid2 = op_data->op_fid2;
386 rec->lk_time = op_data->op_mod_time;
387 rec->lk_bias = op_data->op_bias;
389 tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
390 LOGL0(op_data->op_name, op_data->op_namelen, tmp);
393 void mdc_rename_pack(struct ptlrpc_request *req, struct md_op_data *op_data,
394 const char *old, int oldlen, const char *new, int newlen)
396 struct mdt_rec_rename *rec;
399 CLASSERT(sizeof(struct mdt_rec_reint) == sizeof(struct mdt_rec_rename));
400 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
402 /* XXX do something about time, uid, gid */
403 rec->rn_opcode = REINT_RENAME;
404 rec->rn_fsuid = op_data->op_fsuid;
405 rec->rn_fsgid = op_data->op_fsgid;
406 rec->rn_cap = op_data->op_cap;
407 rec->rn_suppgid1 = op_data->op_suppgids[0];
408 rec->rn_suppgid2 = op_data->op_suppgids[1];
409 rec->rn_fid1 = op_data->op_fid1;
410 rec->rn_fid2 = op_data->op_fid2;
411 rec->rn_time = op_data->op_mod_time;
412 rec->rn_mode = op_data->op_mode;
413 rec->rn_bias = op_data->op_bias;
415 tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
416 LOGL0(old, oldlen, tmp);
419 tmp = req_capsule_client_get(&req->rq_pill, &RMF_SYMTGT);
420 LOGL0(new, newlen, tmp);
424 void mdc_getattr_pack(struct ptlrpc_request *req, __u64 valid, int flags,
425 struct md_op_data *op_data, int ea_size)
427 struct mdt_body *b = req_capsule_client_get(&req->rq_pill,
431 if (op_data->op_bias & MDS_CHECK_SPLIT)
432 b->valid |= OBD_MD_FLCKSPLIT;
433 if (op_data->op_bias & MDS_CROSS_REF)
434 b->valid |= OBD_MD_FLCROSSREF;
435 b->eadatasize = ea_size;
437 __mdc_pack_body(b, op_data->op_suppgids[0]);
439 b->fid1 = op_data->op_fid1;
440 b->fid2 = op_data->op_fid2;
441 b->valid |= OBD_MD_FLID;
443 if (op_data->op_name) {
444 char *tmp = req_capsule_client_get(&req->rq_pill, &RMF_NAME);
446 LOGL0(op_data->op_name, op_data->op_namelen, tmp);
451 static void mdc_hsm_release_pack(struct ptlrpc_request *req,
452 struct md_op_data *op_data)
454 if (op_data->op_bias & MDS_HSM_RELEASE) {
455 struct close_data *data;
456 struct ldlm_lock *lock;
458 data = req_capsule_client_get(&req->rq_pill, &RMF_CLOSE_DATA);
459 LASSERT(data != NULL);
461 lock = ldlm_handle2lock(&op_data->op_lease_handle);
463 data->cd_handle = lock->l_remote_handle;
466 ldlm_cli_cancel(&op_data->op_lease_handle, LCF_LOCAL);
468 data->cd_data_version = op_data->op_data_version;
469 data->cd_fid = op_data->op_fid2;
473 void mdc_close_pack(struct ptlrpc_request *req, struct md_op_data *op_data)
475 struct mdt_ioepoch *epoch;
476 struct mdt_rec_setattr *rec;
478 epoch = req_capsule_client_get(&req->rq_pill, &RMF_MDT_EPOCH);
479 rec = req_capsule_client_get(&req->rq_pill, &RMF_REC_REINT);
481 mdc_setattr_pack_rec(rec, op_data);
482 mdc_ioepoch_pack(epoch, op_data);
483 mdc_hsm_release_pack(req, op_data);
486 static int mdc_req_avail(struct client_obd *cli, struct mdc_cache_waiter *mcw)
490 client_obd_list_lock(&cli->cl_loi_list_lock);
491 rc = list_empty(&mcw->mcw_entry);
492 client_obd_list_unlock(&cli->cl_loi_list_lock);
496 /* We record requests in flight in cli->cl_r_in_flight here.
497 * There is only one write rpc possible in mdc anyway. If this to change
498 * in the future - the code may need to be revisited. */
499 int mdc_enter_request(struct client_obd *cli)
502 struct mdc_cache_waiter mcw;
503 struct l_wait_info lwi = LWI_INTR(LWI_ON_SIGNAL_NOOP, NULL);
505 client_obd_list_lock(&cli->cl_loi_list_lock);
506 if (cli->cl_r_in_flight >= cli->cl_max_rpcs_in_flight) {
507 list_add_tail(&mcw.mcw_entry, &cli->cl_cache_waiters);
508 init_waitqueue_head(&mcw.mcw_waitq);
509 client_obd_list_unlock(&cli->cl_loi_list_lock);
510 rc = l_wait_event(mcw.mcw_waitq, mdc_req_avail(cli, &mcw),
513 client_obd_list_lock(&cli->cl_loi_list_lock);
514 if (list_empty(&mcw.mcw_entry))
515 cli->cl_r_in_flight--;
516 list_del_init(&mcw.mcw_entry);
517 client_obd_list_unlock(&cli->cl_loi_list_lock);
520 cli->cl_r_in_flight++;
521 client_obd_list_unlock(&cli->cl_loi_list_lock);
526 void mdc_exit_request(struct client_obd *cli)
528 struct list_head *l, *tmp;
529 struct mdc_cache_waiter *mcw;
531 client_obd_list_lock(&cli->cl_loi_list_lock);
532 cli->cl_r_in_flight--;
533 list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
534 if (cli->cl_r_in_flight >= cli->cl_max_rpcs_in_flight) {
535 /* No free request slots anymore */
539 mcw = list_entry(l, struct mdc_cache_waiter, mcw_entry);
540 list_del_init(&mcw->mcw_entry);
541 cli->cl_r_in_flight++;
542 wake_up(&mcw->mcw_waitq);
544 /* Empty waiting list? Decrease reqs in-flight number */
546 client_obd_list_unlock(&cli->cl_loi_list_lock);